Advanced Glycation End Products, Soluble Receptor for Advanced Glycation End Products, and Risk of Colorectal Cancer Free

This article is featured in Highlights of This Issue, p. 1261

Energy imbalance, insulin resistance, and chronic inflammation are underlying mechanisms in colorectal cancer development. Further understanding of interrelations among environmental exposure (such as dietary intake and smoking) and these etiologic mechanisms may provide novel opportunities for prevention and clinical management of colorectal cancer. We proposed that the axis of advanced glycation end products (AGE) and receptor for AGEs (RAGE) contributes to the development of colorectal cancer.

AGEs are a group of irreversible adducts or cross-linking created through a nonenzymatic glycosylation between reducing sugars and free amino groups of proteins, lipids, or nucleic acids (1). AGEs form endogenously during normal metabolism, and exogenously from foods processed at high temperatures as well as from tobacco smoking (2, 3). AGEs accumulate slowly in human tissue proteins during aging and more rapidly in the states associated with hyperglycemia and enhanced oxidative and carbonyl stress. Circulating indicators of AGEs and oxidative stress are directly influenced by the intake of dietary AGEs (4).

The principal mechanism by which AGEs elicit biological function is through their receptors. The ligation of AGEs and membrane-bound full-length RAGE can trigger an array of signaling pathways that are involved in inflammation and tumorigenesis (5). Soluble form of RAGE (sRAGE) is found in the circulation in humans (6). By binding AGEs or other ligands and acting as a “receptor decoy,” sRAGE represents a naturally occurring competitive inhibitor of RAGE-mediated signaling pathways (7).

The AGE–RAGE axis plays a critical role in the pathologic interplay between hyperglycemia and vascular homeostasis. However, the role of AGEs in cancer development is largely unknown. Several hospital-based studies found decreased sRAGE levels in patients with breast cancer, lung cancer, and pancreatic cancer, compared with healthy controls (8–10). In this study, we prospectively investigated the associations of N^ϵ-(carboxymethyl)lysine (CML), a prominent type of AGE, and sRAGE with risk of colorectal cancer in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study (11). We hypothesized that higher levels of CML-AGE and lower levels of sRAGE are associated with a greater risk of colorectal cancer.

The ATBC Study is a primary prevention trial conducted in southwest Finland. Between 1985 and 1988, 29,133 men ages 50 to 69 years who smoked at least 5 cigarettes per day were randomized to receive α-tocopherol, β-carotene, α-tocopherol and β-carotene, or placebo. Exclusion criteria for the participation in the trial included the following: malignancy other than nonmelanoma skin cancer or carcinoma in situ; severe angina on exertion; chronic renal insufficiency; cirrhosis of liver; chronic alcoholism; receiving anticoagulant therapy; other medical problems that might limit participation for 6 years; and current use of supplements containing vitamin E, vitamin A, or β-carotene (11). The trial ended in April 1993 and the follow-up for health outcomes continues through national registries. The ATBC Study was approved by the institutional review groups of both U.S. National Cancer Institute and the National Public Health Institute of Finland (now National Institute for Health and Welfare, Helsinki, Finland). All participants provided the written informed consent before randomization.

This study is a case–cohort study within the ATBC Study. Eligible study subjects were alive without cancer within the first 5 years of follow-up after baseline (N = 24,708). The follow-up ended at death, at diagnosis of colorectal cancer, or on April 30, 2006. A total of 508 incident cases of colorectal cancer were identified from the Finnish Cancer Registry (12), including proximal tumors [International Classification of Diseases (ICD)-9 codes 153.0, 153.1, 153.4, 153.6, 153.7] and distal tumors (ICD-9 codes 153.2, 153.3, 154.0, 154.1). The reference group comprised 500 subcohort participants who were randomly selected from all eligible study subjects. After excluding 25 cases and 15 subcohort participants who had missing data on one or more of the serologic biomarkers, we included 483 case and 485 subcohort participants in the present analysis.

At the baseline visit, all participants completed a self-administered questionnaire to provide information on general demographics, medical, smoking, dietary, and occupational histories. Height and weight were measured by trained nurses. The data on aspirin/disprin use for 436 cases and 380 subcohort participants and on family history (first-degree relatives) of colon cancer for 383 cases and 340 cohort participants were collected during the follow-up visit from November 1989 through February 1993. Serum samples were collected from each participant after an overnight fasting at baseline.

Serum CML-AGE and sRAGE levels were measured in duplicate by the Microcoat Biotechnologie Company by using the AGE-CML-ELISA kit (Microcoat Biotechnologie Company) and the human sRAGE Quantikine ELISA kit (R&D system Inc.), respectively. The AGE-CML-ELISA kit uses a CML-specific monoclonal antibody (mouse monoclonal 4G9; Alteon Inc.; ref. 13). Total sRAGE includes the C-truncated endogenous secreted form of RAGE (esRAGE), which is a splice variant of RAGE that lacks the transmembrane and cytoplasmic portion of the receptor (14), and proteolytic cleavage forms of RAGE (15). Case and subcohort samples were handled identically and placed randomly on each plate (each batch) in the same proportion, along with 10% blinded phantom quality control samples from a pooled sample. The intrabatch and interbatch coefficients of variation were 7% and 14% for CML-AGE and 3% and 6% for sRAGE, respectively. Serum concentrations of glucose and insulin were determined in 144 cases and 392 subcohort participants in a previous investigation (16). In this study, these 2 analytes were measured on an additional 364 cases that occurred after 1997 and 100 subcohort participants by using the same method in the same laboratory as the earlier study. A pilot study using previously tested 19 samples showed that the mean concentration of glucose was lower than the previous test and the mean concentration of insulin was higher than the previous test (P values for the Wilcoxon signed rank test was 0.10 for glucose and 0.03 for insulin).

The distributions of the selected characteristics between the cases and the subcohort participants were compared by using the Wilcoxon rank sum test for continuous variables and the χ² test for categorical variables. The residual method was used to adjust dietary intakes for total energy intake. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. The correlation between CML-AGE and sRAGE and with other exposures was examined by using Spearman's rank correlation coefficients in the subcohort.

We used weighted Cox proportional hazard regression models to calculate relative risks (RR) and 95% CI for colorectal cancer, as well as to conduct significance tests for trends and interactions. The quintile cut-points for serologic biomarkers were determined on the basis of the distribution among the subcohort participants and the lowest quintile was the reference category. The quartile cut-points were used in sensitivity analyses when the sample size became small. We used the follow-up time as the underlying time metric. The assumption of proportional hazards was tested by generating the interaction term of each exposure variable and person-years of follow-up in the model. In the weighted analysis, case participants were given a sample weight of 1 because they were sampled with certainty, whereas subcohort participants were given a sample weight of 49.4 (24,708/500). In addition, we evaluated the association between the CML-AGE and colorectal cancer risk by using the value adjusted for sRAGE by the residual method because sRAGE was thought to neutralize circulating CML-AGE. The adjusted CML-AGE was used in all the analyses. Potential confounding factors included all the variables given in Table 1 and randomization group (α-tocopherol, β-carotene, α-tocopherol and β-carotene, or placebo). Confounding effect was evaluated by using backward and forward methods with a variable included in the models if they changed the risk estimates by more than 10%. None of the variables changed the risk estimate for CML-AGE or sRAGE by more than 10%; however, we included age, years of smoking, and BMI in all models. We further included CML-AGE and sRAGE mutually and adjusted serum insulin and glucose in the models to assess whether the associations between CML-AGE or sRAGE and colorectal cancer were independent of other analytes. We found that the adjustment for serum glucose changed the risk estimates for CML-AGE and sRAGE by more than 10%. Therefore, all other potential confounding factors were reevaluated in the models included age, years of smoking, BMI, CML-AGE or sRAGE, and serum glucose. To further examine the interrelations among the 4 serologic markers, we estimated the associations between serum glucose and insulin and colorectal cancer risk with adjustment for serum CML-AGE or sRAGE with adjustment for the batch effect. The values of these 4 serologic biomarkers were log-transformed to normalize their distributions when used as continuous variables in the models. Dose–response trends across increasing quintiles (or quartile) of the biomarkers were tested by using a score variable based on the median value of each quintile (or quartile).

We evaluated the joint effects (i.e., departure from multiplicative models of interaction) of sRAGE with CML-AGE, insulin, or glucose (all dichotomous) for predicting risk of colorectal cancer. Because glucose levels assayed at 2 different times had slightly different distributions, specific medians among the subcohort participants were used to dichotomize glucose levels at 99 mg/dL for the earlier time point and 77 mg/dL for the later time point. The similar approach was used to dichotomize serum insulin. Interactions were represented as cross-product terms and the statistical significance of the interactions was tested by the Wald tests. Using the same approach, we evaluated the joint effects of CML-AGE or sRAGE with number of years of smoking, red meat intake (all used the median in the subcohort as the cutoff points), BMI (<25 vs. ≥25), use of aspirin/disprin (yes or no), randomization groups (placebo, α-tocopherol only, β-carotene only, or both), and anatomic subsite (proximal vs. distal tumors).

In the sensitivity analyses for the main effect and the joint effects, we restricted the study population to 144 cases and 392 subcohort participants whose serum and insulin levels were measured previously; we excluded participants with self-reported diabetes (n = 34), with family history of colon cancer (n = 23) or the information on family history of colon cancer missing (n = 245), or with less than 10 or 15 years of follow-up. All tests were 2-sided and P values less than 0.05 indicated statistical significance. SAS 9.0 (SAS institute, Inc.) and SUDAAN software were used for data analyses.

The follow-up was up to 21 years with a median of 15 years. Selected characteristics of 483 cases and 485 subcohort participants are described in Table 1. Compared with the subcohort participants, cases were older, had less heavy physical activity at work, had longer duration but lower intensity of smoking, and had higher daily intake of red meat and lower daily intake of calcium. CML-AGE and sRAGE were positively correlated (r = 0.48, P < 0.001). Serum sRAGE was negatively correlated with serum glucose (r = −0.11, P = 0.02) and serum CML-AGE was not correlated with serum glucose. CML-AGE had a weak negative correlation with BMI and a positive correlation with daily glucose intake. None was significantly correlated with red meat intake and years of smoking (data not shown).

The median levels (interquartile ranges) of CML-AGE were 540 ng/mL (478–631) for the cases and 561 ng/mL (471–668) for the subcohort; and for sRAGE these were 521 (388–707) pg/mL for the cases and 572 (417–742) pg/mL for the subcohort. The median levels of CML-AGE and sRAGE did not differ significantly between the cases and the subcohort participants (P = 0.12 and 0.08, respectively).

Table 2 shows that higher levels of sRAGE were associated with lower risk of colorectal cancer (fifth compared with first quintile; RR = 0.65; 95% CI, 0.39–1.07; P_trend = 0.03) after adjustment for age, years of smoking, BMI, and CML-AGE (model 2). Adjustment for serum glucose strengthened the inverse association (RR = 0.52; 95% CI, 0.30–0.89; P_trend = 0.009). There was no significant association between CML-AGE and colorectal cancer risk. We also estimated the risk by using adjusted CML-AGE value in that CML-AGE was adjusted for sRAGE. We observed that the second quintile of adjusted CML-AGE was associated with significantly increased risk of colorectal cancer and the significant risk decreased through the fourth quintile and attenuated for the highest quintile of adjusted CML-AGE. The adjustment for serum glucose strengthened the positive association between CML-AGE and pancreatic cancer by more than 10%. The adjustment for food consumptions, nutrient intake, aspirin/disprin use, or family history of colon cancer did not change the risk estimates for CML-AGE and sRAGE by more than 5%. Table 3 shows that the association of sRAGE and CML-AGE held true for both proximal and distal tumors.

We observed the similar magnitude of the positive associations between serum glucose and insulin and risk of colorectal cancer, as previously reported (16), among 144 cases and 392 subcohort participants for whom we had these data, after adjustment for the batch effect. Adjustment for CML-AGE increased the RR for glucose slightly, from 1.97 (95% CI, 0.98–3.96) to 2.17 (95% CI, 1.06–4.45, fourth compared with first quartile; data not shown).

Table 4 shows the statistically significant joint effects of sRAGE with CML-AGE (P_interaction = 0.01) and with serum insulin (P_interaction = 0.03) in association with colorectal cancer risk. The P value for the interaction between sRAGE and years of smoking was 0.07. Compared with higher sRAGE and lower CML-AGE, lower sRAGE and higher CML-AGE was associated with 73% increased risk of colorectal cancer. Compared with higher sRAGE and shorter years of smoking, lower sRAGE and longer years of smoking was associated with more than 1-fold increased risk of colorectal cancer. The interaction of sRAGE by insulin was on a submultiplicative scale. Compared with higher sRAGE and lower insulin, the risk associated with lower sRAGE and higher insulin was 1.62 (95% CI, 0.99–2.64), which was less than the product of the individual effect of higher insulin (RR = 1.34; 95% CI, 0.81–2.21) and lower sRAGE (RR = 2.20; 95% CI, 1.40–3.45). There were no significant interactions for CML-AGE or sRAGE by BMI, aspirin/disprin use, red meat intakes, anatomic sites, or randomization groups (P values for interactions ranged from 0.21 to 0.82) in relation to colorectal cancer risk.

The results for CML-AGE and sRAGE and colorectal cancer were similar in the sensitivity analyses. For example, when fifth quintile compared with first of sRAGE, the RR was 0.44 (95% CI, 0.22–0.86; P_trend = 0.02) among 350 cases and 418 subcohort participants who had been followed up for more than 10 years; the RR was 0.51 (95% CI, 0.29–0.89; P_trend = 0.04) among 470 cases and 464 subcohort participants without diabetes at baseline, and the RR was 0.44 (95% CI, 0.22–0.88; P_trend = 0.003) among 361 cases and 328 subcohort participants who had no family history of colon cancer. Among 144 cases and 392 subcohort participants whose glucose and insulin concentrations were assayed at the earlier time point, we saw a statistically nonsignificant associations between sRAGE and colorectal cancer (RR = 0.71; 95% CI, 0.36–1.40; highest compared with lowest quartile, P_trend = 0.30). The P value for the submultiplicative interaction between sRAGE and insulin was 0.25.

In this prospective study among Finnish male smokers, we found no significant association between serum CML-AGE and colorectal cancer risk. Prediagnostic serum sRAGE was inversely associated with colorectal cancer. Moreover, lower sRAGE in combination with higher CML-AGE or longer years of smoking was associated with higher risk of colorectal cancer compared with higher sRAGE in combination with lower CML-AGE or shorter years of smoking. There was submultiplicative interaction between serum sRAGE and insulin.

RAGE recognizes a wide range of environmental stressors and plays a role in innate immune responses and inflammation (17). The engagement of RAGE by CML-AGE results in enhanced generation of reactive oxygen species and activation of a diverse array of signaling cascades that lead to propagation of an inflammatory response by activation of nuclear transcription factors (17). Full-length RAGE has been described as a link between chronic inflammation and cancer development (18–20). Blockade of RAGE–high mobility group box-1 (HMGB1) interaction was shown to decrease the growth and metastases of both implanted and spontaneous tumors in susceptible mice (21). The role of RAGE–multiligand in fostering an inflammatory tumor microenvironment was reviewed recently (22). A handful of in vitro and animal studies have suggested that RAGE and its ligands play an important role in colorectal carcinogenesis by changing host immunity and tissue microenvironment (23–26). In clinical studies, coexpression of membranous RAGE with HMGB1 has been associated with malignant potential of colorectal adenomas (27) and metastatic potential of colorectal cancer (28).

We observed that risk of colorectal cancer for men with high-serum sRAGE was half that for men with low sRAGE and this association was observed for both proximal and distal tumors. Among nondiabetic subjects, higher sRAGE has been associated with lower risks of several age-related chronic diseases (29, 30). Among patients with type 2 diabetes, plasma sRAGE was highly inversely correlated with glycemic control, insulin resistance, C-reactive protein, and S100A12 (31). Along the same lines, our findings suggested that by neutralizing RAGE ligands, sRAGE may suppress the creation of a microenvironment for tumor growth fostered by engagement of RAGE and its ligands. Moreover, the significant increased risk of colorectal cancer was observed for men with lower sRAGE in the presence of longer years of smoking or lower CML-AGE. The submultiplicative interaction between sRAGE and insulin also supported the involvement of sRAGE in the colorectal carcinogenesis.

It is important to evaluate the associations between AGEs and cancer because of our daily exposure to dietary AGEs (2). AGEs induce permanent abnormalities in the extracellular matrix component and elicit oxidative stress, vascular inflammation, and thrombosis (32). AGEs can attenuate cellular insulin sensitivity in 3T3-L1 adipocytes (33). One study showed a strong to moderate AGE staining by immunohistochemistry in colon adenocarcinoma samples and the surrounding fibroblasts of 5 patients (34). Yamagishi and colleagues (35) hypothesized that AGEs explain the molecular link between hyperglycemia/diabetes and colorectal cancer. Our data do not strongly support this hypothesis because we did not observe a significant trend of increased risk of colorectal cancer with increasing CML-AGE. The lack of an association between CML-AGE and colorectal cancer incidence may be attributable to the detoxification or neutralization of CML-AGE by sRAGE or other receptors of AGEs, such as AGE receptor 1, that detoxify CML-AGE and counter-regulate their pro-oxidant effects (36). Nevertheless, we did observe increased risk-associated moderately higher levels of CML-AGE and sRAGE-adjusted CML-AGE. We speculate that when CML-AGE levels were high, more secreted sRAGE counteracts the effect of CML-AGE. The dynamic interaction between AGEs and the receptors in metabolism and their role in colorectal carcinogenesis need further characterization and elucidation.

The present investigation adds to the literatures relating circulating AGEs and sRAGE to several age-related diseases. Interestingly, these 2 markers are modulated by medications and lifestyle. Circulating levels of sRAGE increased among patients with type 1 diabetes who received angiotensin-converting enzyme inhibitor (ACEi; ref. 37) and among patients with hypercholesterolemia after the treatment with statin (38). Plasma AGEs levels were reduced among women with polycystic ovary syndrome after the treatment with metformin, which is an insulin sensitizer and potent glycation inhibitor (39). ACEi, statin, and metformin have been shown to have anticancer properties in observational studies (40–42). Moreover, AGEs levels are modulated by dietary intake (43). The modifiable nature of AGEs and sRAGE potentially will provide opportunity for cancer prevention if their roles in cancer development are elucidated.

This prospective study has several strengths. The fasting blood samples had been collected at least 5 years before the diagnosis of colorectal cancer and the follow-up was as long as 21 years; therefore, the results are less likely to be influenced by reverse causation. The exclusion of small number of participants with type 2 diabetes generated the same study findings. The extensive data obtained from the questionnaire allowed us to evaluate many potential confounders. Men with conditions (such as heart diseases and chronic renal failure) previously shown to influence serum sRAGE were not included in the ATBC Study and the observed associations would less likely to be confounded by these conditions. However, it is possible that unrecognized conditions, such as preexisting proinflammatory condition at blood collection, confound the association.

Our study also carries certain limitations. First, this study did not examine non–CML-AGEs or AGEs precursors in association with colorectal cancer risk. It has been reported that non–CML-AGE levels are associated with both fasting glucose and HbA1c levels in patients with diabetes (44). Second, our findings may not be applicable to nonsmokers and women and need to be confirmed in other study populations. Nevertheless, a positive correlation between CML-AGE and sRAGE that we saw has been reported in 2 Japanese studies (45, 46). Third, the data on family history of colon cancer and aspirin/disprin use were not complete for all the study subjects because the information was collected during the follow-up. Future studies with such information collected should evaluate their confounding effects adequately. Nevertheless, the incidence of colorectal cancer was low in 1985–1988 in Finland and the related influence on the study finding may not be substantial. Finally, this single study was not able to sort out the interrelations among CML-AGE, sRAGE, serum glucose, and dietary intakes in association with colorectal cancer. Although the adjustment of glucose changed the observed association by more than 10%, the interaction of sRAGE and glucose was not statistically significant.

In summary, we found that sRAGE was inversely associated with colorectal cancer risk. Although biologically plausible, higher CML-AGE levels were not associated with increased colorectal cancer risk. The role of the RAGE–ligand axis in cancer etiology deserves further investigation.

No potential conflicts of interest were disclosed.

L. Jiao and R.Z. Stolzenberg-Solomon designed the research. L. Jiao, S.J. Weinstein, R.Z. Stolzenberg-Solomon, D. Albanes, P.R. Taylor, B.I. Graubard, and J. Virtamo conducted the research. S.J. Weinstein, P.R. Taylor, D. Albanes, and J. Virtamo provided essential materials. L. Jiao, B.I. Graubard, and R.Z. Stolzenberg-Solomon carried out data analysis. L. Jiao, P.R. Taylor, and R.Z. Stolzenberg-Solomon wrote the article. L. Jiao and R.Z. Stolzenberg-Solomon had primary responsibility for final content. All authors read and approved the final manuscript.

The authors thank Kirk Snyder, Dominick Parisi, and Jason Ashby at the Information Management Services, Inc. for data management support; Karon Drew and the staff at NCI Frederick Central Repository Services for sample retrieval; Dr. William C. Kopp, Timothy Sheehy, and the staff at SAIC-Frederick for collecting pilot study samples and sample aliquot; Dr. Todd M. Gibson at Nutritional Epidemiology Branch for testing the batch effect of insulin and glucose concentrations; Dr. Hashem B. El-Sereg at Baylor College of Medicine for his support; and Tawanda Roy at Nutritional Epidemiology Branch for administrative assistance.

This work was supported by funding from the Intramural Research Program of the National Cancer Institute (P.R. Taylor, S.J. Weinstein, B.I. Graubard, D. Albanes, R.Z. Stolzenberg-Solomon) and an Intramural Research Award (L. Jiao) in the Division of Cancer Epidemiology and Genetics of the National Cancer Institute at the NIH. The Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study was supported by U.S. Public Health Service contracts (N01-CN-45165, N01-RC-45035, N01-RC-37004, and HSN261201000006C) from the National Cancer Institute, U.S. Department of Health and Human Services.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.