Association of CTNNB1 (β-Catenin) Alterations, Body Mass Index, and Physical Activity With Survival in Patients With Colorectal Cancer FREE

Activation of the WNT signaling pathway and its major mediator cadherin-associated protein β 1 (CTNNB1; the HUGO-approved official symbol for β-catenin), most commonly by loss of adenomatous polyposis coli (APC), plays a critical role in colorectal carcinogenesis.¹ The WNT signaling pathway is a potential therapeutic target.^2- 3 Inactivation of kinases in the APC complex leads to accumulation of cytoplasmic CTNNB1 and its translocation to the nucleus, in which it acts as a coactivator with the transcription factor (TCF) family (Figure 1).⁴ Proliferative genes regulated by WNT-CTNNB1 signaling such as the oncogene MYC and cell cycle regulator cyclin D1 (CCND1) contribute to tumor progression.¹

Accumulating evidence indicates a role of WNT-CTNNB1 signaling in adipogenesis, obesity, and metabolic diseases.^5- 7 Signaling of WNT has been shown to be activated by PRKA ([adenosine monophosphate] AMP-activated protein kinase),⁸ which is a key regulator of energy metabolism.⁹ Considering the dual roles of CTNNB1 in carcinogenesis and energy metabolism, we hypothesized that activation of WNT-CTNNB1 signaling (ie, accumulation of nuclear CTNNB1) might confer proliferative ability to cancer cells, even when not under excess energy balance status. In addition, epidemiological evidence suggests causal effects of obesity or excess energy balance on colon cancer incidence^10- 11 and mortality.^12- 13 Notably, physical activity (exercise) has emerged as a modifiable lifestyle factor that may improve cancer survival.^14- 15

Data from patients with colorectal cancer (N = 955) in 2 nationwide prospective cohort studies were used to test the hypothesis that tumor alterations of CTNNB1 modify prognostic associations of body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) and postdiagnosis physical activity. Because patient characteristics, lifestyle factors, and major tumor molecular features had been collected, we were able to analyze an interactive effect of CTNNB1 tumor status and BMI and level of postdiagnosis physical activity.

The data were collected from participants in 2 nationwide prospective cohort studies: the Nurses' Health Study (N = 121 701 women followed up since 1976) and the Health Professionals Follow-up Study (N = 51 529 men followed up since 1986).¹⁶ Cohort participants were sent follow-up questionnaires every 2 years to update information on dietary and lifestyle factors (including BMI and level of physical activity), and to identify newly diagnosed cancers in themselves and in their first-degree relatives. We collected paraffin-embedded tissue blocks from hospitals throughout the United States where patients underwent colorectal cancer resections.¹⁶

Hematoxylin and eosin−stained tissue sections from all colorectal cancer cases were reviewed by a pathologist (S.O.) who was unaware of the other study data. The tumor differentiation was categorized as well to moderate vs poor (>50% vs ≤50% gland formation). We excluded individuals who had been preoperatively treated. Participants were observed until death or June 30, 2009, whichever came first. Death of a participant was confirmed using the National Death Index. Informed consent from study participants was indicated if the follow-up questionnaires were returned. The study was approved by the human subjects committees at Harvard School of Public Health and Brigham and Women's Hospital (both in Boston, Massachusetts).

Leisure-time physical activity was assessed every 2 years in both study cohorts and was described and validated against physical activity diaries.^17- 18 Participants reported duration of participation (range: 0-≥11 hours/week) spent walking (along with usual pace), jogging, running, bicycling, swimming laps, playing racket sports, participating in other aerobic exercises or lower-intensity exercise (yoga, toning, stretching), and other vigorous activities. Each activity on the questionnaire was assigned a metabolic equivalent task (MET) score.¹⁹ One MET is the energy expenditure for sitting quietly; MET scores are defined as the ratio of the metabolic rate associated with specific activities divided by the resting metabolic rate. The values from the individual activities were summed for a total score of MET hours per week.

Based on prior studies of physical activity and survival after colorectal cancer,^14- 15 individuals who engaged in at least 18 MET-h/wk had significantly better colorectal cancer–specific mortality. To avoid assessment during the period of active treatment, the first assessment of physical activity was collected at least 1 year but no more than 4 years postdiagnosis (median, 17 months). To avoid bias due to declining physical activity in the period around cancer recurrence or death, physical activity was not updated but was assessed at a single point postdiagnosis.^14- 15

DNA was extracted from paraffin-embedded tissue. Polymerase chain reaction and pyrosequencing were used for v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) (HGNC 6407; codons 12 and 13),²⁰ v-raf murine sarcoma viral oncogene homolog B1 (BRAF) (HGNC 1097; codon 600),²¹ and phosphoinositide-3-kinase catalytic alpha polypeptide (PIK3CA) (HGNC 8975; exons 9 and 20).²² Microsatellite instability analysis was performed using 10 microsatellite markers (D2S123, D5S346, D17S250, BAT25, BAT26, BAT40, D18S55, D18S56, D18S67, and D18S487).²³ High microsatellite instability was defined as instability in 30% or more of the markers; low microsatellite instability and microsatellite stability were defined as instability in less than 30% of the markers.

Using validated bisulfite DNA treatment and real-time polymerase chain reaction (MethyLight; Qiagen, Valencia, California),^24- 25 we quantified DNA methylation in these 8 promoters to determine CpG island methylator phenotype (CIMP): calcium channel voltage-dependent T-type α 1G subunit (CACNA1G) (HGNC 1394), cyclin-dependent kinase inhibitor 2A (CDKN2A) (p16; HGNC 1787), cellular retinoic acid binding protein 1 (CRABP1) (HGNC 2338), insulin-like growth factor 2 (IGF2) (HGNC 5466), mutL homolog 1 (MLH1) (HGNC 7127), neurogenin 1 (NEUROG1) (HGNC 7764), runt-related transcription factor 3 (RUNX3) (HGNC 10473), and suppressor of cytokine signaling 1 (SOCS1) (HGNC 19383).^23,26 According to established criteria,²⁷ high CIMP was defined as the presence of 6 or more of the 8 methylated promoters; low CIMP was defined as the presence of 1 of the 8 methylated promoters to presence of 5 of the 8 methylated promoters; and CIMP-0 was defined as presence of 0 of the 8 methylated promoters. To accurately quantify relatively high levels of long interspersed nucleotide element 1 (LINE-1) methylation, we used pyrosequencing methods as described by Irahara et al.²⁸

The methods of immunohistochemistry were previously described for tumor protein p53 (TP53) (HGNC 11998)²⁹ and CTNNB1 (HGNC 2514).³⁰ Antigen retrieval was performed for CTNNB1 using deparaffinized tissue sections in citrate buffer (BioGenex, San Ramon, California) that were treated in a microwave and then in a pressure cooker for 15 minutes. Tissue sections were incubated for 15 minutes with 3% H₂O₂ (to block endogenous peroxidase) with 10% normal goat serum (Vector Laboratories, Burlingame, California) in phosphate-buffered saline for 10 minutes and with serum-free protein block for 10 minutes (DAKO, Carpinteria, California). The primary antibody against CTNNB1 (β-catenin, clone 14, 1:400 dilution; BD Transduction Laboratories, Franklin Lakes, New Jersey) was applied for 1 hour at room temperature. Later, the secondary antibody was applied for 20 minutes (BioGenex) and then streptavidin peroxidase conjugate was applied for 20 minutes (BioGenex). Sections were visualized with diaminobenzidine for 2 minutes and by methyl green counterstain. Appropriate positive and negative controls were included in each immunohistochemistry run. For CTNNB1, normal colonic epithelial cells served as internal positive controls with membrane staining (Figure 2). The methods of interpretation and validation of cutoffs are described in detail in the eAppendix.

All statistical analyses were performed using SAS software version 9.1 (SAS Institute Inc, Cary, North Carolina). All P values were 2-sided and statistical significance was set at a P value of .05. The P value of .003 for significance level was used when multiple hypothesis testing was performed and was adjusted by Bonferroni correction (ie, P = .05/15). The χ² test was used for categorical data. The unpaired t test (under the assumption of equal variances) was used to compare mean age and mean level of LINE-1 methylation. The Kaplan-Meier method and log-rank tests were used for survival analyses. Deaths by other causes were censored for the analyses of colorectal cancer–specific mortality. With a P value of .05 for significance and total sample size of 955 with 266 events, there was 80% power to detect colorectal cancer–specific mortality with a hazard ratio (HR) of 1.57; with 440 events, there was 80% power to detect overall mortality with an HR of 1.50. We calculated age-adjusted incidence rates of death from colorectal cancer or from all causes in a specific subgroup by dividing the number of deaths by the sum of person-years of follow-up, which were adjusted for age.

To control for confounding, we used multivariate Cox proportional hazards regression models. A multivariate model initially included sex, age at diagnosis (continuous), BMI (<30 vs ≥30), family history of colorectal cancer in first-degree relatives (present vs absent), tumor location (proximal vs distal), tumor differentiation (well to moderate vs poor), microsatellite instability (high microsatellite instability, low microsatellite instability, or microsatellite stability), CIMP (high, low, or 0), LINE-1 methylation (continuous), TP53, KRAS, BRAF, and PIK3CA. Disease stage (I, IIA, IIB, IIIA, IIIB, IIIC, IV, or unknown) was used as a stratifying variable using the strata option in the SAS proc phreg command to avoid overfitting and residual confounding. A backward stepwise elimination was performed with P value threshold of .20 to avoid overfitting. Excluding cases with missing information in any of the covariates did not substantially alter results (detail about the handling of missing data is described in the eAppendix). The proportionality of hazards assumption was satisfied by evaluating time-dependent variables, which were the cross-product of the CTNNB1 variable and survival time (P = .10). We tested the models by removing individual variables and did not detect problems with colinearity.

An interaction was assessed using the Wald test on the cross-product of CTNNB1 and another variable of interest (excluding cases missing data) in a multivariate Cox model. Based on our a priori hypothesis on an interaction between CTNNB1 and energy balance status, we examined an interaction between nuclear CTNNB1 and prediagnosis BMI (or postdiagnosis level of physical activity). In exploratory analyses, we examined a potential interaction between nuclear CTNNB1 and any of the other covariates (including age, sex, family history of colorectal cancer, tumor location, cancer stage, microsatellite instability, CIMP, LINE-1 methylation, KRAS, BRAF, PIK3CA, and TP53), and found no significant interaction (all P >.01 for interaction; given multiple hypothesis testing, a statistical significance level was adjusted to P = .003 for interaction). The effect of nuclear CTNNB1 did not significantly differ between the 2 independent cohort studies (P = .86 for interaction for cancer-specific survival and P = .78 for interaction for overall survival).

Based on the availability of tumor tissue and data up to 2004, there were 955 patients diagnosed with colorectal cancer in stage I, II, III, or IV and included in this study (Table 1). Among the 955 patients, 455 had positive status for cytoplasmic CTNNB1 (48%; defined as moderate to strong expression), 439 had positive status for nuclear CTNNB1 (46%; defined as moderate to strong expression), and 488 had loss of CTNNB1 membrane (51%; defined as no expression or weak membrane expression). Patients with positive status for cytoplasmic and nuclear CTNNB1 were correlated with each other (P <.001) and with loss of CTNNB1 membrane expression (P <.001). As shown in Table 1, proximal tumor location, high level of microsatellite instability, high CIMP, and presence of the BRAF mutation were associated inversely with positive status for cytoplasmic and nuclear CTNNB1 and loss of CTNNB1 membrane (P <.001 for all comparisons).

During follow-up (median, 141 [interquartile range, 105-192] months for all censored cases), there were 440 deaths, which included 266 colorectal cancer–specific deaths. For either colorectal cancer–specific mortality or overall mortality, CTNNB1 status (positive or negative for cytoplasmic or nuclear or having loss or not having loss of membrane) was not significantly associated with patient survival in analyses using the log-rank test or in univariate or multivariate Cox regression analysis (Table 2).

Because WNT-CTNNB1 signaling has been implicated in obesity and metabolic diseases,^5- 7 we examined a potential modifying effect of prediagnosis BMI on the relationship between CTNNB1 alterations and patient survival (Table 3). There was a significant modifying effect of BMI. In obese patients (BMI ≥30), positive status for nuclear CTNNB1 was associated with significantly better cancer-specific mortality (adjusted HR, 0.24 [95% confidence interval {CI}, 0.12-0.49]) and overall mortality (adjusted HR, 0.56 [95% CI, 0.35-0.90]). In contrast, among nonobese patients (BMI <30), positive status for nuclear CTNNB1 was not significantly associated with cancer-specific survival (P <.001 for interaction) or overall survival (P = .03 for interaction). A similar but weaker modifying effect of BMI was found in survival analyses using cytoplasmic CTNNB1 or membrane status (eTable 1 and eTable 2).

We examined the effect of level of postdiagnosis physical activity on mortality (in patients with stage I, II, or III cancer, n = 497) in strata of nuclear CTNNB1 status (Figure 3 and Table 4). We excluded patients with stage IV cancer as in our previous analysis³¹ because of the likely influence of disease severity on a patient's ability to exercise. For patients with negative status for nuclear CTNNB1, high level of postdiagnosis physical activity (≥18 vs <18 MET-h/wk) was associated with significantly better colorectal cancer–specific mortality (adjusted HR, 0.33 [95% CI, 0.13-0.81]). However, for patients with positive status for CTNNB1, physical activity was not associated with colorectal cancer–specific survival (adjusted HR, 1.07 [95% CI, 0.50-2.30], P = .05 for interaction) or with overall survival (adjusted HR, 0.86 [95% CI, 0.55-1.34]). The differential effect of postdiagnosis physical activity on patient survival according to status of nuclear CTNNB1 also was evident using Kaplan-Meier analyses (Figure 3).

We performed this study to test the hypothesis that the CTNNB1 status in colorectal cancers interacts with patients' BMI or postdiagnosis level of physical activity and modifies tumor cell behavior. We found substantial interactive prognostic associations of tumor CTNNB1 and self-reported BMI or postdiagnosis level of physical activity. Specifically, positive status for nuclear CTNNB1 was associated with significantly better colorectal cancer–specific survival and overall survival in obese patients, while positive or negative status for nuclear CTNNB1 was not associated with survival among nonobese patients. Furthermore, in patients with stage I, II, or III cancer and negative status for nuclear CTNNB1, high level of postdiagnosis physical activity (≥18 MET-h/wk) was associated with significantly better colorectal cancer–specific survival, while physical activity was not associated with survival among patients with stage I, II, or III cancer and positive status for nuclear CTNNB1. These results provide evidence for a possible interactive effect of tumor CTNNB1 signaling and patients' energy balance status in determining tumor cell behavior. Our data support the hypothesis that progression of a tumor with an inactive WNT-CTNNB1 signaling pathway might be influenced by energy intake and expenditure, whereas a tumor with an active WNT-CTNNB1 signaling pathway might progress independent of energy balance status. Although our data need to be confirmed by independent data sets, tumor CTNNB1 status may serve as a predictive biomarker for response to a prescription for physical activity (exercise) in clinical practice. Because physical activity is a modifiable lifestyle factor, our data may have considerable clinical implications.

Examining molecular changes or prognostic factors is important in cancer research.^32- 33 Previous prognostic data on CTNNB1 alterations in colorectal cancer were inconclusive.^34- 52 Two large studies^42- 43 (N>540) demonstrated no prognostic role of CTNNB1 alterations. However, none of the previous studies^34- 52 had examined interactive associations of tumor CTNNB1 signaling and patients' BMI or physical activity.

An interaction between the host patient factor and the tumor modifying cell behavior was first described between BMI and FASN expression in colon cancer.⁵³ Examining interactions between host patient factors and tumors has been a novel research paradigm in the evolving interdisciplinary field of molecular pathological epidemiology.^54- 55 Investigating whether lifestyle interventions are more or less beneficial based on tumor molecular subtypes provides important information regarding mechanisms of action that can, in turn, inform mechanistically driven clinical trials to optimize the efficacy of such lifestyle interventions.

We have previously examined interactions of several molecular markers with physical activity, and found that the benefit of physical activity may be influenced by tumor CDKN1B (p27) status.³¹ Specifically, physical activity after colon cancer diagnosis was associated with better cancer-specific survival in CDKN1B-expressing tumors but not in CDKN1B-lost tumors.³¹ It would be of particular interest to examine interactive effects of CDKN1B, CTNNB1, and physical activity in well-powered trial settings in the future. In addition, further studies are warranted to examine the exact mechanism of how physical activity modulates the CTNNB1 or CDKN1B signaling pathway to influence tumor cell behavior.

Prospective observational data suggest that patients who survive colorectal cancer and are physically active have lower rates of cancer recurrence and better survival compared with those who are physically inactive.^14- 15,56 Physical activity is a modifiable lifestyle factor; therefore, its beneficial effect on cancer survivors has considerable clinical implications. However, as with any other oncological intervention, it is unlikely that all patients will universally gain a benefit from exercise. As evidence grows that nondrug interventions such as physical activity can influence outcome of patients with established cancer, there is a need to better delineate subpopulations of cancers that may or may not be more likely to be affected by such an intervention. Thus, it is of particular interest to identify patient characteristics or tumor biomarkers that can predict response to an exercise intervention. In the current study, we have found CTNNB1 status as such a candidate predictor. Although our data need to be confirmed by additional studies, determining status for nuclear CTNNB1 may improve the identification of patients who will benefit most from physical activity. Future prospective intervention studies are warranted.

There are limitations to this study. For example, data on cancer treatment were limited. Nonetheless, it is not likely that chemotherapy use substantially differed according to CTNNB1 tumor alterations because such data were not available for treatment decision making. In addition, our multivariate survival analysis finely adjusted for disease stage (I, IIA, IIB, IIIA, IIIB, IIIC, IV, or unknown) on which treatment decision making was mostly based. We recognize that we cannot exclude the possibility that dosing, completion, or dose-modification rates of adjuvant therapy might vary according to BMI or physical activity. However, approximately 60% of patients included in our analysis had stage I or stage II disease and surgery alone would generally be the standard care; no interaction between physical activity and disease stage was observed in our previous analyses of this cohort.⁵⁶ Thus, we consider that the confounding effects of adjuvant therapy would not have been substantial. Another limitation is that data on cancer recurrences beyond cause of mortality were not available in these cohorts. Nonetheless, colorectal cancer–specific survival is a reasonable surrogate of colorectal cancer–specific outcome.

There are advantages in using the database of the 2 prospective cohort studies. Data on anthropometric measurements, physical activity, cancer staging, and other clinical, pathological, and tumoral molecular variables were prospectively collected and investigators were blinded to patient outcome. Cohort participants who developed cancer were treated at hospitals throughout the United States and were more representative of colorectal cancer cases in the general US population than patients in only a few academic hospitals. In addition, we assessed the effect of CTNNB1 alterations independent of other critical molecular events such as BRAF and PIK3CA mutations, LINE-1 hypomethylation, CIMP, and microsatellite instability, all of which have been associated with colorectal cancer prognosis.^23,57- 58

In conclusion, our data provide evidence for an interaction between CTNNB1 alterations in colorectal cancer and patients' energy balance status, which influences tumor cell behavior. Notably, there appear to be substantial modifying effects of tumor CTNNB1 status on the beneficial prognostic role of postdiagnosis physical activity. Physical activity was associated with better cancer-specific survival only in patients with negative status for nuclear CTNNB1 colorectal cancers, whereas physical activity was not associated with survival in patients with positive status for nuclear CTNNB1. Our findings may have considerable clinical implications because physical activity is a modifiable lifestyle factor.