A Central Role for Heme Iron in Colon Carcinogenesis Associated with Red Meat Intake

Colorectal cancer is the third most common type of cancer worldwide after lung and prostate cancer in men and after lung and breast cancer in women (1). Environmental factors, particularly diet, play roles in the development of colorectal cancer (2, 3). On the basis of epidemiologic studies, the World Cancer Research Fund panel considers the colorectal cancer risk associated with red and processed meat intake to be convincing and recommends limiting the consumption of red meat and avoiding the consumption of processed meat (2, 3). Our previous works showed that red and processed meats promote precancerous lesions (aberrant crypt foci, ACF, and mucin-depleted foci, MDF) in the colons of rats fed a low-calcium diet (4–6). These data strongly support the results from epidemiologic studies.

Three major mechanisms may explain the association between meat and colorectal cancer (7). First, potentially carcinogenic N-nitroso compounds can form in the gastrointestinal tract by N-nitrosation of peptide-derived amine or by nitrosylation yielding S-nitrosothiols and nitrosyl iron (FeNO). Collectively, these are measured as the apparent total N-nitroso compounds (ATNC; ref. 8). Second, meat cooked at high temperatures contains mutagenic heterocyclic amines (HCA) like 2-amino-1-methyl-6-phenylimidazo(4,5-b)pyridine (PhIP) and 2-amino-3,8-dimethylimidazo [4,5-f]quinoxaline (MeIQx); ref. 9]. Third, epidemiologic and experimental data support the hypothesis that heme iron present in red and processed meats promotes colorectal cancer (5–7, 10). This effect can be explained by the direct cytotoxic, genotoxic effects of heme on epithelial cells and by the catalytic effect of heme iron on the formation of ATNC and lipid peroxidation end-products like 4-hydroxynonenal (4-HNE; refs. 4, 5, 7, 11–14). Cross and colleagues investigated these 3 hypotheses in a cohort study and found a significant association between colorectal cancer and the intake of heme iron, nitrate from processed meat, and HCA (15). Nevertheless, numerous biases are possible in the determination of risk factors using epidemiologic approach, and contribution of each of these factors has never been evaluated experimentally in the same study.

The present study aimed at investigating the roles of these 3 potential mechanisms, namely heme iron, NOC, and HCA, in colorectal cancer in vivo at a precancerous lesion stage (MDF) in carcinogen-induced rats (see a flowchart in Supplementary Fig. S1). Doses were chosen to mimic red meat consumption. Subsequently, the results were confirmed at the tumor stage using C57BL/6J Apc^Min/+ mice, a genetic model of colorectal cancer. The tumor incidence was associated with genotoxicity endpoints in mucosa as γH2AX and anaphase bridges. Like tumors in the majority of human colorectal cancer cases and in Apc^Min/+ mice, the preneoplastic lesion MDF in rats show activation of the Wnt signaling pathway driven by mutations in Apc and/or in the β-catenin gene. We also used an intestinal cellular model with normal and premalignant cells (Apc^+/+ and Apc^−/+ cells) to complement the in vivo studies. In combination with animal models, such cellular models allowed us (i) to understand the effect of dietary compounds on cancer promotion at early stages of carcinogenesis and (ii) to explain and further investigate the effects observed in vivo.

Male 4-week-old F344 rats (n = 80; Charles Rivers), male and female 4-week-old C57BL/6J Apc^Min/+ mice (Jackson, Laboratory, n = 35) and Apc^+/+ mice (Charles River, n = 33), and male and female C57BL/6J mice (n = 10) were housed (2 rats per cage and 2 to 3 mice per cage) under standard laboratory conditions with free access to food and water. After acclimatization with AIN76 diet, rats were fed experimental diets for 100 days, Apc^Min/+ and Apc^+/+ mice were fed experimental diets for 49 days, and C57BL/6J mice were fed experimental diets for 14 days. Rats were killed by CO₂ asphyxiation, and mice were killed by cervical dislocation. Animal care was in accordance with the European Council and ARRIVE guidelines.

To assess the relative contributions of the 3 potential mechanisms (heme iron, NOC, and HCA), we conducted a 2 × 2 × 2 protocol on azoxymethane-induced F344 rats fed a diet containing 1% hemoglobin, HCA (PhIP, 50 μg/kg; MeIQx, 25 μg/kg), or both. To induce a strong endogenous NOC formation, drinking water was supplemented with sodium nitrate and nitrite (0.17 g/L of NaNO₂ and 0.23 g/L of NaNO₃) and compared with a nitrate-free water, according to the experimental groups described in Table 1A (16). Mice were fed a control diet or a 2.5% hemoglobin diet (Table 1B).

After 1 week on the experimental diet, rats received intraperitoneal injection of azoxymethane (Sigma; 20 mg/kg body weight).

The large intestines of rats and large and small intestines of mice were removed and fixed in 0.05% buffered formalin (Sigma).

MDF were scored in duplicate by 2 readers who were blinded to the origin of the colon following the high-iron diamine Alcian blue procedure (17) described by Santarelli and colleagues (4).

At sacrifice, the intestinal tract from duodenum to colon was removed. Sections of duodenum, jejunum, and ileum were harvested, opened along the longitudinal axis, and washed in PBS. After fixation in 10% formalin, mouse colons were stained for 6 minutes in a 0.05% filtered methylene blue solution, and small intestines were stained for 48 hours in a 300 ppm solution of methylene blue in formalin. One reader who was blinded to the origin of the sample scored tumors and determined their diameters using a binocular microscope at 25× magnification. All tumors in each section of the intestines were counted, the smallest tumors that could be counted were approximately 0.5 mm in diameter.

Feces were collected during the last 10 days and frozen at −20°C. Urine was collected on days 67 to 70 for rats and on days 44 to 45 for mice and frozen at −20°C before DHN-MA assay (Supplementary Materials and Methods).

Feces of 24 hours were collected. To prepare fecal water, distilled water (1 mL for rats or 0.85 mL for mice) was added to 0.3 g of dried feces. Fecal water was prepared as described by Pierre and colleagues (5).

The heme concentration in the fecal water was measured by fluorescence according to Van den Berg and colleagues (18) and as described by Pierre and colleagues (5). To determine the lipid peroxides in the lumen, thiobarbituric acid reactive substances (TBARS) were quantified in fecal water according to the technique of Ohkawa and colleagues (19) as described previously (20). The results are expressed as the MDA equivalent.

ATNCs include N-nitroso compounds, S-nitrosothiols, and FeNO nitrosyl heme. They were analyzed as described previously (11) with an Ecomedics CLD Exhalyzer (Ecomedics). The values measured in 100 μL of the sample are expressed as concentration (in μmol/L).

Apc^+/+ (derived from C57BL/6J mice) and Apc^−/+ (derived from C57BL/6J Apc^Min/+ mice) colon epithelial cells (21) express the heat-labile SV40 large T antigen (AgT tsa58) under the control of an IFNγ-inducible promoter. Both cell lines expressed cytokeratin 18, a marker of their epithelial phenotype (Forest 2003). Consequences of the Apc mutation were also detected in the Apc^−/+ cell line. As expected, actin network was disorganized in Apc^−/+ cells (Supplementary Fig. S2A; refs. 21, 22), accumulation of multinucleated cells was observed in Apc^−/+ cells (Supplementary Fig. S2A). As expected, the culture conditions affected cell proliferation due to the thermolabile tsA58 T antigen, which confers conditional immortalization: at 33°C with IFNγ, the large T antigen is active and drives cellular proliferation, and at 37°C, the temperature-sensitive mutation yields an inactive protein and cells act like nonproliferating epithelial cells (Supplementary Fig. S2B).

4-HNE derived from the oxidation of n-6 polyunsaturated fatty acids and 4-hydroxy-2-hexenal (4-HHE) derived n-3 polyunsaturated fatty acids were synthesized as described by Chandra and Srivastava (23). Malondialdehyde (MDA) derived from polyunsaturated fatty acids with 3 or more double bonds was prepared as described by Fenaille and colleagues (24).

Polymer resin (4-Fmoc-hydrazinobenzoyl AM NovaGel, Nobabiochem Merck Chemicals) contains hydrazine functional groups protected by Fmoc groups. To unprotect them, the resin was washed with 0.8 mL DMSO + 0.2 mL piperidine, vortexed for 1 minute, and allowed to settle for 15 minutes. The settled resin was washed twice with DMSO, 4 times with ethanol, and with distilled water. The amount of polymer resin used for each sample was based on using 100× the amount of MDA equivalents present in the fecal water sample. Polymer resin was added to fecal water, and the samples were agitated for 2 hours at 4°C. After letting the resin settle for 15 minutes, supernatant was transferred to a new tube with polymer resin and agitated for 2 hours at 4°C. After settling, the supernatant was diluted into culture medium without FCS and used for the MTT assay.

To determine cytotoxic activity of fecal water, of HNE (20 μmol/L), and of hemin (100 μmol/L), the MTT assay on Apc^+/+ and Apc^−/+ cells was used, as described previously (5). H2AX phosphorylation (γH2AX) is a rapid and sensitive cellular response to genotoxicity (25, 26). Genotoxicity and cytotoxicity of aldehydes were measured after 24 hours of treatment of Apc^+/+ and Apc^−/+ cells using a γH2AX in-cell Western blot assay according to Audebert and colleagues (27, 28). Graillot and colleagues (28) demonstrated that this assay can be used to measure cell viability via DNA quantification. Cells were seeded into 96-well plates at 5 × 10³ cells per well in DMEM supplemented with 10% (v/v) FCS, 1% (v/v) penicillin/streptomycin, and 10 U/mL IFNγ at the permissive temperature of 33°C. After 72 hours, cells were transferred to 37°C without IFNγ for 24 hours and then treated with aldehyde (5, 10, and 20 μmol/L) in duplicate. Culture medium without IFNγ and FCS was used for untreated control wells. This assay was repeated 4 times.

Apoptosis was measured in Apc^+/+ and Apc^−/+ cells using a luminescent assay (Caspase-Glo3/7; Promega). Cells were treated with aldehydes for 6 hours. After cell lysis, plates were incubated at room temperature for 2 hours, and the luminescence intensity of each well was determined using an INFINITEM200 plate reader (TECAN). This measure was performed in triplicate with aldehydes at 2.5, 5, 10, 20, 40, and 80 μmol/L.

Four-micrometer paraffin-embedded sections from formalin-fixed mouse small intestine (Swiss rolls) specimens were de-waxed in toluene and rehydrated. Sections were incubated in Dako peroxidase blocking solution (Dako S2023) and in goat serum (1:10, Dako X0907) for 20 minutes at room temperature. Sections were incubated with the rabbit polyclonal anti-γH2AX antibody (1:400, Cell Signaling Technology #9718) for 50 minutes at room temperature. The secondary antibody (biotinylated goat anti-rabbit, Thermo Scientific TR-060-BN) was applied for 30 minutes at room temperature followed by horseradish peroxidase (HRP)–streptavidin solution (DAB, Dako K0675) for 25 minutes. Peroxidase activity was revealed by DAB substrate (DAKO, K3468). Sections were counterstained with Harris hematoxylin, dehydrated, and coverslipped.

Enterocytes with nuclear γH2AX-positive foci or complete nuclear labeling were considered positive cells. Cells were assessed by counting the positive nuclei in segments of the small intestine specimen that were at least 200 glands long. The positive counts were expressed as counts per one villi-gland unit.

Chromosomal or mitotic alterations can arise from numerous events, including errors during cell division or repair of damaged DNA. As a consequence, the separating sister chromatids are often connected by DNA bridges in anaphase. Anaphase bridges (AB) were evaluated on 4-μm paraffin-embedded sections from formalin-fixed mouse small intestine (Swiss rolls). Sections were stained with hematoxylin and eosin and AB were evaluated under light microscope using 400× magnification. Four segments from the duodenum, jejunum, and ileum that were at least 100 consecutive glands long were selected for counting. Criteria for ABs included having a well-separated parallel anaphase plate displaying a perpendicularly aligned amphophilic (stretched) connecting filament (29). The scores were expressed as number of AB per villi-gland unit.

Results were analyzed using Systat 10 software for Windows, and all data are reported as mean ± SEM. For the in vivo experiments on chemically induced rats, the importance of each factor was tested independently of the experimental groups (ANOVA per factor). If a significant difference was found between groups (P < 0.05), each experimental group was compared with the control using Dunnett test, the difference between control and hemoglobin diets effect on Min mice tumors was analyzed using the Student t test. For the in vitro study, the dose effect of aldehydes was analyzed using one-way ANOVA. If a significant difference was found between groups (P < 0.05), each experimental group was compared with the control treatment using Dunnett test. Second, the effect of the mutation effect at each concentration of aldehyde was analyzed with the comparison between the Apc^+/+ and Apc^−/+ cell lines using the Student t test.

Only diets containing hemoglobin significantly increased the number of MDF per colon (P < 0.001) independent of the 2 other factors (Fig. 1A). Indeed, although nitrates/nitrites in drinking water induced a considerable increase in fecal ATNCs in all groups, they failed to increase the number of MDF per colon (Fig. 1A and D). Nevertheless, we noticed that the ATNC composition was different between groups, containing 30% to 80% FeNO and no S-nitrosothiols in the hemoglobin-fed groups compared with no FeNO and about 30% of S-nitrosothiols in other groups.

Diets containing hemoglobin significantly increase the amount of TBARS in fecal water (Fig. 1B) and the amount of urinary 1,4-dihydroxynonemercapturic acid (DHN-MA), a metabolite of the lipid oxidation product 4-HNE (Supplementary Fig. S3B). These oxidation biomarkers depended only on dietary and fecal heme (Supplementary Fig. S3A) and remained unchanged when the diet contained nitrates/nitrites or HCA without hemoglobin (Fig. 1B and Supplementary Fig. S3B).

Fecal water from rats fed hemoglobin-containing diets was more cytotoxic to the nonmutated Apc^+/+ cells than to premalignant Apc^−/+ cells (Fig. 1C). These data are consistent with previous results (20). Fecal water from rats fed HCA or nitrites/nitrates without hemoglobin was not cytotoxic to these cells (Fig. 1C).

With the trapping of aldehydes from fecal water of rats fed heme with a polymer resin with hydrazine functional groups, we found that in Apc^+/+ and Apc^−/+ cells, a 95% reduction in fecal water peroxidation was associated with a 75% reduction in cytotoxicity (Fig. 2A). Furthermore, we observed that only 4-HNE, but not heme, had differential cytotoxic effects in Apc^+/+ and Apc^−/+ cells that was similar to that observed with fecal water of rats fed heme (Fig. 2B).

We then measured the cytotoxic and genotoxic effects of 3 main lipid peroxidation end products (4-HNE, 4-HHE and MDA) on Apc^+/+ and Apc^−/+ cells using a γH2AX in-cell Western blot assay. HNE and HHE were more cytotoxic and more genotoxic to normal Apc^+/+ cells than to premalignant Apc^−/+ ones (Fig. 2C, P < 0.05). HNE at 10 and 20 μmol/L and HHE at 20 μmol/L were significantly more genotoxic to Apc^+/+ cells than to Apc^−/+ ones (Fig. 2C, P < 0.05). MDA was neither cytotoxic nor genotoxic in the tested cell lines (Fig. 2C). We confirmed these viability results with the MTT assay and with the CellTiter-Glo assay (Supplementary Fig. S4), and we used an expanded range of treatment concentrations (from 0 to 80 μmol/L) in these assays. We confirmed that HNE and HHE were more cytotoxic (from 10 to 80 μmol/L and from 40 to 80 μmol/L, respectively) to normal cells than to premalignant ones, whereas MDA had no effect (Supplementary Fig. S4).

We determined the caspase-3/7 activity and again found a significant difference between Apc^+/+ and Apc^−/+ cells after HNE and HHE treatment at 80 and 40 μmol/L, respectively (Fig. 2D, P < 0.05). Specifically, caspase-3/7 activity was higher in Apc^+/+ cells than in Apc^−/+ cells, and MDA treatment had no effect (Fig. 2D).

A diet containing 2.5% hemoglobin given to Apc^Min/+ mice significantly increased the intestinal tumor load (control diet: 67 ± 39 mm²; hemoglobin diet: 114 ± 47 mm², P = 0.004; Fig. 3A). These mice develop polyps mainly in the small intestine, and in our study, we did not observe any effects of hemoglobin diet in the colon. The hemoglobin diet also significantly increased the number of all tumors in the jejunum (Supplementary Fig. S5A). In the entire small intestine, the hemoglobin diet significantly increased the number of tumors with a diameter greater than 1 mm (1 mm < tumor size ≤ 2 mm, P = 0.04; 2 mm < tumor size, P = 0.006; Supplementary Fig. S5B). Giving the same hemoglobin diet to normal C57BL/6J Apc^+/+ mice induced no neoplasia. As observed in rats, the effect of dietary heme on mice tumors was associated with a significant increase in fecal heme (from 20 ± 22 to 198 ± 29 μmol/L) and with increases in lipoperoxidation biomarkers: fecal TBARS (Fig. 3B) and urinary DHN-MA (Supplementary Fig. S5C). We also assessed TBARS and the cytotoxicity of fecal water in the small intestine to measure biomarkers at the same location as tumors. As Min mice have a mutation in the Apc gene, we decided to use the mouse Apc^−/+ model to investigate the cytotoxic activity of fecal water in vivo. Heme diet was associated with a significant increase in fecal TBARS and cytotoxicity in the small intestine (Fig. 3C).

The induction of luminal lipid peroxidation by the hemoglobin diet (Fig. 3B and C) was associated with increased genotoxicity only in nonmutated C57BL/6J Apc^+/+ mice with a higher AB index in the epithelium (Fig. 4A and B). As expected the AB index was higher in C57BL/6J Apc^Min/+ mice than in C57BL/6J Apc^+/+ mice fed a control diet (Fig. 4A) because of the Apc mutation. In Apc^Min/+ mice, the hemoglobin diet had no additional genotoxic effect (Fig. 4A). The genotoxic effect of the hemoglobin diet in C57BL/6J Apc^+/+ was confirmed by γH2AX induction (Fig. 4C and D).

This study examines in vivo the relative contributions of the 3 main factors that may explain how consumption of red and processed meat promotes colorectal cancer. Cross and colleagues showed that these factors, that is, heme, HCA, and NOC, were associated with colorectal cancer in a prospective cohort study in humans (15). However, the identification of risk factors using an epidemiologic approach has to be correlated with the experimental approach to establish the causative effect of such factors. Here, heme iron was the only experimental factor associated with a significant increase in precancerous lesions (MDF) in rats. Heme iron showed no additive or synergic effects with nitrates/nitrites or with HCA. Using a complementary approach that included 2 animal models and a cellular model, we found that heme is the determining factor in the promotion of colorectal carcinogenesis and that the selective toxicity of heme-induced alkenals to nonmutated cells seemed to play an important role in this mechanism.

HCA are complete carcinogens that induce colon, mammary, and prostate tumors in rodents and monkeys (30). The absence of effects of HCA in this study could be explained by the dose we chose, which was based on the estimated dietary exposure to HCA in a diet that is high in red meat and was relevant of the human food exposure. Indeed, carcinogenic doses of HCAs in rodents are 1,000 to 100,000 times higher than levels found in human foods (31). Nitrite undergoes an enterosalivary cycle in humans but not in rats. We hypothesized that the addition of sodium nitrates/nitrites to the rodents' drinking water, which mimics human saliva, would increase the effects of heme iron in rats by boosting nitrosation in the gut. In humans, red meat consumption increases fecal ATNC concentrations (11), as in our study with rats. Nevertheless, we could not detect any association between the ATNC level (Fig. 1D) and carcinogenesis (Fig. 1A). The highest level of ATNC was seen in the control group given nitrates/nitrites-supplemented water; this group had the fewest MDF. The lack of a relationship between ATNCs and the number of MDF does not support a strong role for ATNCs in the promotion of colon carcinogenesis by red meat.

The present results strongly suggest that at concentrations that are in line with human red meat consumption, heme iron is associated with the promotion of colon carcinogenesis at a preneoplastic stage. Most human colon cancers have an Apc mutation, as do MDF in humans and rats (32). To unravel the mechanisms, we used a cellular model that represented the colorectal cancer stages that we investigated in vivo. We chose a cellular model that mimicked the early steps of carcinogenesis. This conditionally immortalized intestinal cellular model uses premalignant Apc^−/+ cells derived from C57BL/6J Apc^Min/+ mice and “normal” Apc^+/+ cells from C57BL/6J mice (33). Characterization of both cell lines showed the expected consequences of Apc mutation, such as actin network disassembly, aneuploidy, and multinucleated cells (Supplementary Fig. S2). These cell lines can therefore be used to study the mechanisms involved in the early steps of colorectal cancer and thus comprise a cellular model that is a relevant complement to our in vivo model.

In rats, promotion of colon carcinogenesis by dietary hemoglobin was associated with changes in noninvasive biomarkers: fecal water heme iron, TBARS, and cytotoxic activity. Only the hemoglobin diet increased TBARS levels in fecal water (Fig. 1B). We speculated that the cytotoxic effects of fecal water on normal and premalignant colonic cells in vitro mimic the in vivo situation with normal epithelium (Apc^+/+) and with Apc-mutated MDF. In this study, only fecal water from hemoglobin-fed rats was more cytotoxic to Apc^+/+ cells than to mutated cells (Fig. 1C). We propose that premalignant cell selection explains the heme-induced promotion of MDF. Aldehydes or heme iron itself, both present at high concentration in feces from hemoglobin-fed rats, might be responsible for this differential cytotoxicity. Using a resin to specifically trap fecal aldehydes, we showed that aldehydes alone are responsible for fecal water cytotoxicity. In addition, we observed that 4-HNE, but not heme iron, induced differential cytotoxicity in Apc^+/+ and Apc^−/+ cells similar to that observed with fecal water (Fig. 2B). Therefore, we propose that heme-induced lipid peroxidation in the gut explains the observed differential cytotoxicity and the colorectal cancer–promoting effects of heme that are observed in vivo.

To explore the link between aldehydes and the promotion of colon carcinogenesis, we tested the effects of 3 relevant aldehydes, 4-HNE, HHE, and MDA, in Apc^+/+ and Apc^−/+ cells. These α,β-unsaturated hydroxyalkenals are highly reactive compounds with proteins and nucleic acids (34), and they are potentially cytotoxic and genotoxic. In our cellular model, 4-HNE and HHE were more cytotoxic to normal cells than to premalignant cells and induced higher levels of apoptosis in normal cells than in premalignant cells. HNE was more cytotoxic than HHE, as reported previously (35). Furthermore, HNE, like HHE, was more genotoxic to normal cells than to premalignant ones (Fig. 2C), with a higher index of DNA double-strand breaks as revealed by the phosphorylation of histone H2AX. DNA double-strand breaks pose a critical hazard to the genome, and erroneous rejoining of DNA double-strand breaks can lead to mutation. These results thus suggest that at concentrations higher than 20 μmol/L, HNE and HHE will kill normal cells, whereas at lower concentration, they could create mutations in Apc^+/+ cells and might thus initiate carcinogenesis. Therefore, Apc-mutated cells are resistant to apoptosis and can survive to contact with cytotoxic and genotoxic aldehydes, which allows them to undergo further mutation and to become more malignant. Surprisingly, MDA was not toxic to the cells tested in this study but others found also that MDA had little or no toxicity in cells (35, 36). The results obtained by aldehyde trapping and in vitro with HNE and HHE confirmed our hypothesis that aldehydes are responsible for the differential cytotoxic effects of fecal water from heme-fed rats. Heme iron catalyzes the formation of aldehydes in the gastrointestinal tract, which would “select” premalignant cells and also increase the mutation frequency in normal cells (22).

This study shows that a hemoglobin-rich diet significantly increased the tumor load in the small intestine of Apc^Min/+ mice. In contrast, tumor load was not changed by heme diet in the colon of mice, despite the expected modulation of biochemical markers. The number of tumors in the colon of Min mice is low (<0.5 tumors per mouse), which reduces statistical power (37). These mice have a truncated Apc gene as in human familial adenomatous polyposis (FAP; ref. 38). Moreover, sporadic colorectal cancer tumors have the same early Apc mutation in 50% to 80% of cases (37). This mutation is also present in MDF (39). Our nutritional experiments in rats and mice were thus conducted in the defined genetic context of the Apc mutation. Promotion of carcinogenesis in rats and in Apc^Min/+ mice (Fig. 3A) was associated with 2 noninvasive biomarkers, fecal water TBARS and cytotoxic activity, in colon and in small intestine (Fig. 3B and C).

Furthermore, as we found that HNE is more genotoxic in vitro to wild-type cells than to Apc^−/+ cells (Fig. 2C), we decided to investigate the genotoxicity of dietary heme in vivo by measuring (i) the AB index in the epithelium of C57BL/6J Apc^+/+ mice and C57BL/6J Apc^Min/+ mice and (ii) by assessing γH2AX induction in the epithelium of C57BL/6J Apc^+/+ mice (Fig. 4). In these 2 studies, the induction of lipid peroxidation in the gut by heme (Fig. 3B and C) was associated with increased epithelial genotoxicity in Apc^+/+ mice but not in Apc^Min/+ mice. Together with in vitro data, these data show that dietary hemoglobin can induce DNA damage. We also observed ABs, which are biomarkers of chromosomal instability and a major consequence of Apc mutation. As expected, in mice fed the control diet, more ABs were seen in Apc^Min/+ mice than in Apc^+/+ mice (Fig. 4A). Moreover, the hemoglobin diet increased the AB index in Apc^+/+ mice (Fig. 4A). The hemoglobin diet induced the same number of ABs as the Apc mutation, suggesting that dietary hemoglobin generates strong initiators. Taken together, these data suggest that heme-induced aldehydes can induce mutations in vitro and in vivo and may initiate carcinogenesis.

In conclusion, we identified heme iron as the main factor responsible for the promotion of colorectal cancer by red meat and showed that aldehydes such as 4-HNE or HHE play roles in the underlying mechanism of action. Furthermore, we suggest that dietary heme could result in initiating agents in the gut. Improved dietary recommendations should focus (i) on the amount of heme iron in meat-based diets rather than on the modes of cooking or preparation and (ii) on dietary changes that could reduce the heme effect in the gut (i.e., on changes that limit the bioavailability of heme and of heme-induced peroxidation; ref. 40).

No potential conflicts of interest were disclosed.

The funders had no role in study design, data collection and analysis, the decision to publish, or preparation of the manuscript.

Conception and design: N. Bastide, F. Chenni, R. Santarelli, S. Taché, D. Corpet, F. Pierre

Development of methodology: N. Bastide, R. Santarelli, S. Taché, N. Naud, M. Baradat, I. Jouanin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Bastide, M. Audebert, R. Santarelli, S. Taché, N. Naud, R. Surya, D.A. Hobbs, F. Gueraud

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Bastide, F. Chenni, M. Audebert, R. Santarelli, S. Taché, N. Naud, M. Baradat, R. Surya, D.A. Hobbs, G.G. Kuhnle, F. Gueraud, F. Pierre

Writing, review, and/or revision of the manuscript: N. Bastide, F. Chenni, M. Audebert, R. Santarelli, F. Gueraud, D. Corpet, F. Pierre

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Chenni, R. Santarelli, N. Naud, I. Jouanin, D.A. Hobbs, F. Pierre

Study supervision: F. Pierre

Other (histopathology): I. Raymond-Letron

The authors thank F. Blas-Y-Estrada for animal care, M.L. Jourdain for cell-cycle data, and J-Ph. Nougayrède for immunofluorescent microscopy.

This research was supported by grants from the French National Research Agency (ANR-06-PNRA-5E14 HemeCancer and ANR-10-ALIA-014 SecuriViande).

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.