ERK activation drives intestinal tumorigenesis in Apc(min/+) mice.

Toll-like receptor (TLR) signaling is essential for intestinal tumorigenesis in Apc(min/+) mice, but the mechanisms by which Apc enhances tumor growth are unknown. Here we show that microflora-MyD88-ERK signaling in intestinal epithelial cells (IECs) promotes tumorigenesis by increasing the stability of the c-Myc oncoprotein. Activation of ERK (extracellular signal-related kinase) phosphorylates c-Myc, preventing its ubiquitination and subsequent proteasomal degradation. Accordingly, Apc(min/+)/Myd88(-/-) mice have lower phospho-ERK (p-ERK) levels and fewer and smaller IEC tumors than Apc(min/+) mice. MyD88 (myeloid differentiation primary response gene 88)-independent activation of ERK by epidermal growth factor (EGF) increased p-ERK and c-Myc and restored the multiple intestinal neoplasia (Min) phenotype in Apc(min/+)/Myd88(-/-) mice. Administration of an ERK inhibitor suppressed intestinal tumorigenesis in EGF-treated Apc(min/+)/Myd88(-/-) and Apc(min/+) mice and increased their survival. Our data reveal a new facet of oncogene-environment interaction, in which microflora-induced TLR activation regulates oncogene expression and related IEC tumor growth in a susceptible host.

Introduction

The gastro-intestinal tract is constantly exposed to a vast number of commensal bacteria and their inflammatory products. Essential to intestinal homeostasis are pattern recognition receptors (PRR) such as TLR1. Engagement of TLR with their cognate ligands in the intestinal mucosa provokes the production of pro-inflammatory, pro-angiogenic and growth factors that support IEC differentiation and proliferation2. In a genetically susceptible host, an on-going intestinal inflammation provokes an uncontrolled growth of IEC leading to neoplasia3,4,5. Likewise, it was proposed that signaling through TLR regulates IEC tumor development, in mice heterozygous for a mutant form of the tumor suppressor gene, adenomatous polyposis coli (Apc)6. However, the molecular mechanisms and its relationship to intestinal inflammation have not been identified. The Apc+ mouse is an animal model of human familial adenomatous polyposis7. These mice develop multiple intestinal neoplasia (Min), after they lose the heterozygote wild type Apc allele and consequently die when they reach 6 months of age8.

The survival and growth of certain tumors are dependent on the continued activation of certain oncogenes. This phenomenon that was termed “oncogene addiction”, explains tumor suppression due to the inactivation of a single gene product9. The oncogene c-myc is critical for Apc-mediated tumorigenesis10,11. The genetic deletion of c-myc results in the inhibition of tumor growth 11 and as low as a two-fold reduction in c-myc expression in IEC is sufficient to inhibit tumorigenesis in Apc+ mice12-14. Here we identified that a MyD88-dependent activation of ERK in IEC is essential to drive intestinal tumor growth in Apc+ mice. Consequently, the inhibition of pERK abrogates the Min phenotype in these animals.

Results

MyD88 signaling is essential for polyp growth in Apc+ mice

TLRs signal mainly through either MyD88 or TRIF. To explore the potential impact of TLR signaling on IEC tumors we crossed Apcmin/+ mice to Myd88 or Trif (Lps2) mice 15. The average survival was 23 weeks for Apcmin/+ mice and 28 weeks for Apcmin/+/Lps2 mice. In contrast, all of the Apc+/Myd88 mice survived the 45-week study (Supp. Fig. 1A). We then determined the role of each adapter protein on tumor (polyp) formation at 20 weeks of age. Apc+/Myd88 mice had fewer polyps throughout the small and large intestines compared to Apc+ or Apc+/Lps2 mice (Supp. Fig. 1B and Supp. Fig. 1C), but they displayed circular raised lesions (microadenomas) in both the distal small intestine (DSI) and the colon (Supp. Fig. 1C-1F).

MyD88 signaling enhances IEC proliferation and suppresses IEC apoptosis in Apc+ mice

As the polyps in the Apc+/Myd88 mice failed to grow (Supp. Fig. 1C), we investigated whether the deletion of Myd88 affected IEC proliferation. The proliferation and the migration rate of IEC along the crypt-villus axis, as analyzed by BrdU incorporation, were decreased as compared to those in Apc+ mice (Fig. 1A). We also observed a significantly higher number of apoptotic IEC in Apc+/Myd88 (Fig. 1B) as well as increased levels of cleaved poly(ADP-ribose) polymerase (PARP), a substrate of caspase-316 (Fig. 1C). Taken together, these data indicate that TLR signaling via MyD88 enhances IEC proliferation and inhibits IEC apoptosis, and suggest that these two effects synergize in enhancing IEC tumor growth in the Apc+ mice.

Figure 1

Genetic disruption of Myd88 in Apc+ mice suppresses proliferation and enhances apoptosis of IEC

A. IHC and BrdU incorporation in IEC (DSI) after i.p. BrdU injection (scale bars - 20 μm, magnification ×200). BrdU-positive cells, per time point, were enumerated for each indicated position in a crypt (10 crypt-villi units/time point), position 0 being the base of the crypt 39. B. Apoptotic IEC (DSI) were determined by TUNEL assay (scale bars - 40 μm, magnification ×100). C. Cleaved product of poly(ADP-ribose) polymerase (PARP) in IEC (DSI) harvested from the indicated mice (n=2/group).

Myd88 signaling in IEC, but not in hematopoietic cells, controls IEC tumor growth in Apcmin/+ mice

In the intestinal mucosa, both IEC and bone marrow (BM)-derived cells have functional TLR that utilize MyD88 for signaling 17,18,19. To further identify the role of BM-derived cells in IEC tumorigensis, we generated BM chimeras: both Apc+ and Apc+/Myd88 recipients were reconstituted with BM harvested from either WT or Myd88 donors 20. Reconstitution of Apc+ recipients with either Myd88 or WT BM did not significantly alter polyp count and growth in either the DSI or the colon. Similarly, the number of polyps did not significantly change in Apc+/Myd88 recipients reconstituted with WT or Myd88 BM (Fig. 2A). These results indicate that polyp growth in Apc+ mice does not depend on TLR-MyD88 signaling in BM-derived cells and highly suggests its dependence on TLR-MyD88 activation of IEC.

Figure 2

Myd88 signaling in hematopoietic cells is not required for tumorigenesis in Apc+ mice

A. Polyp count in BM chimeras in the DSI and colon (P=n.s, n=7-9 mice/group). B. Polyp count in the small intestine in Apc+/Il1r1 and Apc+/caspase-1 mice at 20 weeks of age (n=7/group). C. Polyp count in Anakinra-treated Apc+ mice (DSI) (P=n.s., n=7/group).

To explore whether host-derived or microbial-derived TLR ligands play a role in IEC tumorigenesis, we crossed Apc+ mice with Il1r1 or with Caspase1 mice, which are limited in processing IL-1 and IL-18 21. As presented in Fig. 2B, there was no significant difference in the numbers of polyps in Apc+/Il1r1 or Apc+/Caspase1 as compared to Apc+ mice. In addition, administration of the IL-1R antagonist, Anakinra, did not affect the extent of IEC tumorigenesis in Apc+ mice (Fig. 2C). Collectively, these data strongly suggest that MyD88-dependent TLR activation by microbial ligands is responsible for IEC tumor growth in Apc+ mice.

A MyD88-dependent TLR signaling upregulates c-myc in IEC

The decrease in IEC proliferation and the increase in IEC apoptosis in Apc+/Myd88 mice suggested the involvement of a MyD88-dependent oncogene or mitogen in IEC tumorigenesis. Since c-myc is essential for tumorigenesis in Apc+ mice 11,12,14, we tested whether MyD88 regulates the expression of c-myc. MyD88-deficiency resulted in a significant decrease in the c-myc protein level in IEC. While c-myc was expressed throughout the crypt in both the DSI and the colon of Apc+ mice, its expression in Apc+/Myd88 mice was restricted to the base of the crypt (Fig. 3A). Immunoblotting analysis of c-myc in isolated IEC (DSI) confirmed the reduced expression of not only c-myc, but also pERK in Apc+/Myd88 mice (Fig. 3B and Supp. Fig. 2A). The decreased c-myc level in Apc+/Myd88 IEC was observed in both normal and tumor regions (Supp. Fig. 2B). However, the c-myc mRNA levels in IEC did not differ significantly between Apc+ and Apc+/Myd88 mice (Fig. 3C). Inactivation of Apc activates β–catenin, which induces transcription of c-myc. The deletion of MyD88 did not affect the β-catenin level in vivo (Fig. 3B) or Wnt3-induced activation of β-catenin in vitro (Fig. 3D). Collectively, these data indicate that MyD88 signaling affects tumorigenesis independently of the Wnt-APC-β-catenin pathway.

Figure 3

MyD88 regulates c-myc expression levels

A. IHC analysis of c-myc protein in IEC from the DSI and colon from 20-week old mice (scale bars, 10 μm, magnification ×200). B. IB analysis of the indicated proteins in IEC (DSI) of 20-weeks mice (n=2). C. Transcript levels of c-myc in IEC (DSI) (P=n.s, n=3/group). D. RKO cells transfected with either control or Myd88 siRNA, were stimulated with Wnt3a (100ng/ml) and subjected to IB analysis.

A posttranslational modification of c-myc by TLR-MyD88-ERK pathway stabilizes c-myc expression

The data suggested that MyD88-mediated signaling in IEC provokes tumor growth. Since IEC express functional TLRs 18,19 and Supp. Fig. 3A, we tested whether activation of a TLR-MyD88 pathway directly induces c-myc. Indeed, activation of TLR2 enhanced the protein level of c-myc in an IEC line RKO (Apc wild type) (Fig. 4A) in a MyD88-dependent manner (Supp. Fig. 3B). TLR5 (a MyD88-dependent TLR) activation in RKO produced a similar result (Supp. Fig. 3C). Consistent with the results obtained in vivo (Fig. 3C), the level of c-myc mRNA was not affected by TLR2 triggering, while the levels of IL-8 and IκBα were increased 19 (Fig. 4A).

Figure 4

TLR signaling via MyD88 stabilizes c-myc protein in IEC through activation of ERK

A. Upper panel: RKO cells were stimulated with P3C (2 μg/ml), lysed and analyzed by IB. Lower panel: Transcript levels after TLR2 stimulation (qPCR). B. Protein levels (IB) (Upper panel) and transcript level (qPCR) (Lower panel) in MG-132 treated (10 μM) RKO cells. C-myc was immunoprecipitated followed by IB with anti-ubiquitin (Ub) ab. C. RKO cells were treated with P3C (2 μg/ml) and ubiquitinated c-myc level was measured by IP followed by IB. D. Phospho-ERK and c-myc levels (IB) in U0126- or PD0325901-treated RKO cells.

The increase in c-myc protein level without a concomitant increase in mRNA level upon TLR stimulation, suggested that c-myc protein is subjected to post-translational modifications 22. Indeed, inhibition of proteasomal function by MG-132 enhanced the c-myc protein levels in RKO cells without affecting the mRNA level (Fig. 4B), indicating a steady state degradation of c-myc. We therefore tested whether TLR stimulation in IEC stabilizes c-myc protein. While the c-myc-ubiquitin conjugates were easily detected even in the absence of a proteasome inhibitor in RKO cells, they rapidly disappeared upon TLR2 stimulation with a concomitant increase in unconjugated c-myc protein (Fig. 4C). These data indicate that the TLR-MyD88-mediated signaling pathway stabilizes c-myc protein in IEC by inhibiting its proteasomal degradation.

MEK/ERK pathway phosphorylates c-myc on Serine 62, which stabilizes c-myc by preventing ubiquitin/proteasomal degradation 23,24,25. We examined whether Myd88-mediated activation of ERK is responsible for the stabilization of c-myc. Indeed TLR2 activation induced the phosphorylation of ERK as well as of c-myc on Serine 62 (Fig. 4A). In addition, blocking ERK activation with pharmacological inhibitors rapidly reduced c-myc level (Fig. 4D). Caco-2, another IEC line, expresses a truncated APC protein similar to that observed in Apc+ mice 26. We therefore tested whether MyD88-dependent ERK activation can stabilize c-myc in these Apc mutant cells. Indeed, TLR2 stimulation increased the c-myc protein level with concomitant ERK activation and a decrease in the polyubiquitinated c-myc (Supp. Fig. 4A and B). Similarly, stimulation of either TLR2 or EGFR in a non-transformed IEC line derived from the small intestine (RIE-1) also activated ERK and c-myc (Supp. Fig. 4C). These data indicate that c-myc level in Apc+ mice is maintained by two independent mechanisms, 1) a transcriptional activation of c-myc by β-catenin signaling initiated by Apc inactivation and 2) a post-translational stabilization of c-myc by MyD88-dependent ERK activation.

ERK signaling drives the Min phenotype

We tested whether a MyD88-independent activation of ERK increases c-myc protein level in Apc+ mice and restores the Min phenotype. As EGF activates ERK and enhances c-myc levels in non-transformed IEC (Supp. Fig. 4C), we treated Apc+/Myd88 mice with either EGF alone or with EGF plus a MEK1/2 inhibitor (PD0325901, PD). The latter is a specific and an effective pharmacological inhibitor of ERK 27 and is in phase II clinical trials. The administration of EGF significantly increased the number of polyps in the DSI (for comparison see Supp. Fig. 1B), and this induction was abrogated by PD treatment (Fig. 5A). Serum hemoglobin and body weights drop significantly in Apc+ mice over time, due to the increase in numbers and the sizes of the exophytic polypoid intestinal tumors minimizing food absorption, with subsequent intestinal obstruction and intestinal bleeding 28. The inhibition of tumor growth in Apc+/Myd88, PD-treated animals coincided with increased serum hemoglobin levels (Fig. 5B) and increased body weight (Fig. 5C), indicating that these were healthier animals. As expected, EGF administration enhanced levels of c-myc and pERK in IEC, which was reversed by PD treatment (Fig. 5D). Taken together, the inhibition of IEC tumors in PD treated mice further validated the regulatory role of ERK on tumorigenesis in Apc+/Myd88 mice (Fig. 3B) and could suggest that TLR-MyD88 pathway contributes significantly to ERK activation in Apc+ mice, under the steady state conditions.

Figure 5

Activation of ERK restores the Min phenotype in Apc+/Myd88 mice

A. PD reduces the number of polyps in EGF-treated Apc+/Myd88 mice (DSI) (n=8/group). B. Blood hemoglobin levels and (C) body weight of these mice. D. Top panel: H&E of DSI in control, EGF-treated, and EGF + PD-treated mice. The arrows indicate intestinal polyps. Middle panel: C-myc expression in IEC. Bottom panel: Phospho-ERK levels in IEC of these mice.

These results indicated a pivotal role for ERK activation in the Min phenotype. We therefore tested its tumorigenic role in 10-week old Apc+ mice. PD treatment for 14 weeks of these animals resulted in complete inhibition of polyp growth (Fig. 6A) with the concomitant increase in serum hemoglobin levels (Fig. 6B) and body weight (Fig. 6C). PD treatment inhibited the levels of both c-myc and pERK in IEC of these mice (Fig. 6D and 6E). Furthermore, PD treatment resulted in 100% survival whereas treatment of control animals with vehicle resulted in 100% mortality during the 17 weeks treatment period of Apc+ mice (Fig. 6F). To detect the long-term effects of PD treatment, we delivered it or vehicle to already 17-week PD-treated Apc+ mice, for additional 15 weeks. Continuous PD treatment inhibited tumorigenesis while its discontinuation provoked high tumor count (Fig. 6G). Collectively, these results indicate that the regulation of ERK pathway in Apc+ mice controls intestinal tumorigenesis and the subsequent manifestation of the Min phenotype, most likely via post-translational modifications of c-myc protein.

Figure 6

Activation of ERK is essential for the Min phenotype in Apc+ mice

A. Polyp count, B hemoglobin level, and C body weight in PD-treated Apc+ mice (n=6 for vehicle, n=9 for PD group). D. Upper panel: H&E of DSI in control and PD-treated Apc+ mice. Arrows indicate intestinal polyps. Lower panel: C-myc expression in IEC. E. IB analysis of c-myc and pERK levels in IEC (DSI) of these mice. F. Survival in PD-treated or vehicle-treated Apc+ mice for 17 weeks (n=8). G. The PD-treated Apc+ mice mentioned in F were split to PD- and vehicle-treated groups (n=4/group). Polyp count (DSI) was performed 15 weeks later. H. The microflora induces tumorigenesis in Apc+ mice by triggering TLR-ERK pathway in IEC. This stabilizes c-myc and inhibits its proteasomal degradation. Increased c-myc levels induce the Min phenotype. Additional signals such as growth factors, utilize the MEK-ERK pathway and similarly to TLR ligands, can enhance c-myc levels. Of note, sterile food and water still contain TLR ligands (e.g., LPS) that are capable of stimulating IEC. This mechanism may account for the Min phenotype observed in Apc+ mice housed under germ-free conditions 40.

Discussion

Overt inflammation can promote neoplasia 29,30,31,32. TLR activation of innate immune cells (e.g., macrophages) in the intestinal mucosa provokes the production of various pro-inflammatory mediators 3,5. This mechanism was proposed to enhance tumorigenesis in the Apc+ mice 6. However, our study indicates that MyD88 in non-hematopoietic cells, such as IEC, is required for intestinal tumor growth in the Apc+ mouse. Furthermore, we identified that TLR ligands presumably from intestinal flora (Fig. 4), and not from the host (Fig. 2), mediate IEC tumor growth under the steady-state conditions. In this setting, MyD88-mediated signaling, induces ERK activation that stabilizes and hence, increases the protein level of the oncogene c-myc in IEC 23. This sequence of events enhances IEC proliferation and reduces IEC apoptosis and therefore promotes IEC tumor growth in Apc+ mice.

The oncogene c-myc is a Wnt target gene 33,34. While β-catenin/TCF signaling induces c-myc transcriptionally, its expression levels are heavily regulated by ubiquitin-mediated proteasomal degradation 35,36 which can be antagonized by pERK phosphorylation of c-myc 23,24,25. Our findings indicate that Myd88-dependent, ERK activation is essential to stabilize c-myc levels (Fig. 4), that the activation of ERK by a MyD88-independent ligand, EGF 37, increases c-myc levels and restores the Min phenotype in the Apc+ mice (Fig. 5), and that treatment with a specific ERK inhibitor suppresses tumor development in both Apc+ and EGF-treated Apc+/Myd88 mice (Fig. 5 and Fig. 6). Collectively, these data indicate that 1) the loss of heterozygosity of Apc is insufficient to drive the Min phenotype in the Apc+ mouse, 2) that the synergy between c-myc transcription and post-translational modifications are required for tumor growth in this model, 3) activation of ERK is essential for IEC tumorigenesis in the Apc+ mouse and 4) that ERK functions as a major regulator of c-myc expression in the intestinal epithelium (Fig. 6H).

One mechanism that explains tumor suppression due to the inactivation of a single gene product is termed oncogene addiction. This phenomenon occurs when tumors require sustained activation of a single oncogene for their growth and survival, despite other oncogenic events 9. Our data reveal that the IEC tumor growth in the Apc+ mice is due to pERK “addiction”. ERK addiction was shown recently to drive the survival of certain intestinal epithelial cell lines in vitro 38, although via a different pathway. Activation of ERK in this setting is most likely induced by a TLR-MyD88-dependent pathway (e.g., microfora, Fig. 3) and by a TLR-Myd88-independent pathway (e.g, growth factors) (Fig. 5). Consequently, the inhibition of ERK prevents tumorigenesis in Apc+ mice, most likely via the generation of an unstable c-myc protein (Fig. 5-6) leading to low c-myc levels in IEC (Fig. 3). Although the regulation of the ERK-c-myc pathway is sufficient for the inhibition of the Min phenotype under the steady state conditions, and its reversal upon EGF administration, in Apc+ mice, we can't rule out other anti-apoptotic effects provoked by pERK 38 in IEC of these animals.

The dichotomy in tumor numbers between Apc+ and Apc+ mice (Supp. Fig. 1), as well as the biochemical evidence presented above in vitro (Fig. 4) and in vivo (Fig. 5-6), highly suggest the inductive role of microflora-derived MyD88 signaling on IEC tumorigenesis in Apc+ mice. These observations reveal a new facet of oncogene-environment interactions, which might explain why a germline mutation in Apc results primarily in tumors originating from the intestinal epithelium (Fig. 6H) and not in other organs. Since pERK is a major player in the induction of the Min phenotype (Fig. 5-6), we propose that interventions aimed at inhibiting ERK activation in IEC (Fig. 6) may help suppress the induction of IEC neoplasia in humans with variant Apc genes.

Materials and Methods

Materials

The following antibodies were obtained from Cell Signaling Technology (Danvers, MA): anti-phospho ERK1/2, anti-ERK1/2, anti-c-myc, anti-PARP, anti-β-catenin, anti-PCNA and anti-MyD88. Anti-ubiquitin antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA) and anti-phospho-c-myc (Ser62) antibody for IB and anti-c-myc antibody for immunohistochemistry from Abcam (Cambridge, MA). InSolution™ MG-132 was purchased from Calbiochem (San Diego, CA), the MEK1/2 inhibitor (U0126) from Promega (Madison, WI) and the MEK1/2 inhibitor, PD0325901, from Stemgent (San Diego, CA). Anakinra was purchased from Amgen (Kineret®, CA), recombinant mEGF from PeproTech, Inc. (Rocky Hill, NJ), the TLR2 ligand, Pam3Cys (P3C) from InvivoGen (San Diego, CA) and the Wnt3a from R&D system (Minneapolis, MN).

Mice

C57Bl/6J, Apc+ and Il1r1 mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Myd88 mice were kindly provided by Dr. S. Akira (Osaka University, Japan), and were backcrossed 10 generations onto C57Bl/6, Lps2 by Dr. B. Beutler (TSRI, San Diego, CA) and Caspase1 mice by Dr. R. Flavell (Yale University, CT). All these mice strain were crossed to Apc+ mice. All animal protocols received prior approval by the Institutional Review Board.

In vivo treatment with Anakinra

Eight to 10 week-old mice Apc+ mice were injected i.p with 50 mg/kg of Anakinra, 5 times/week for ten weeks and analyzed when they reached 20 weeks of age.

In vivo treatment with EGF

Eight to 10 week-old mice were injected i.p with EGF (2 μg/mouse), 3 times/week for 10 weeks and analyzed when they reached 20 weeks of age.

In vivo treatment with an ERK inhibitor

PD0325901 was dissolved initially in DMSO (50 mg/ml) as a stock solution. The stock solution was then diluted fresh in water containing 0.05% (Hydroxypropyl)methycellulose and 0.02% Tween 80. The formulation containing PD0325901 in 250 μl at the 25 mg/kg dose was administered by gavage three times a week to EGF-treated Apc+ mice or five times a week to Apc+ mice, for the duration of each study. Controls mice were treated with vehicle (gavage).

Bone marrow (BM) chimeras were generated by reconstituting irradiated (9 Gy of γ-radiation) 6-10 week-old Apc+ and Apc+ mice with BM cells (1.5 × 107, i.v.) from sex-matched WT or Myd88-/- donor mice. Chimerism was verified by qPCR of peripheral blood cells. Polyp counts were performed when mice reached 20 weeks of age.

BrdU staining was performed using a BrdU in situ staining kit (BD Biosciences, San Diego, CA). Mice were injected i.p. with 2 mg of BrdU solution. Intestinal tissue samples were fixed with formalin and embedded in paraffin. Immunostaining for labeled BrdU was performed according to the manufacturer's instruction. The enumeration of BrdU positioning was performed as described 39.

TUNEL assay was performed on paraffinized intestinal tissues according to the manufacturer's instruction (BD Biosciences). Nuclei were stained with Hoechst 33258 (Invitrogen, Carlsbad, CA).

Isolation of intestinal epithelial cells, RT-PCR, Immunoblotting and immunoprecipitation were performed as previously described 19.

Cell Culture

The human IEC cell lines RKO and Caco-2 were cultured in DMEM supplemented with 4.0 mM glutamine, 10% fetal calf serum, 50 U/ml penicillin and 50 μg/ml streptomycin.

siRNA-mediated knockdown

Myd88 siRNA or c-myc siRNA from Santa Cruz Biotechnology (Santa Cruz, CA). Briefly, siRNA (40 μM) in 50 μl of Opti-MEM (Invitrogen) was mixed with 5 μl of Dharmafect 4 (Dharmacon, Chicago, IL) in 50 μl of Opti-MEM. After 30 min incubation at RT, the transfection mixture was combined with 1 × 106 cells in culture medium. Non-targeting siRNA #2 (luciferase targeting siRNA) from Dharmacon was used as a control.

Histology and Immunohistochemistry (IHC)

DSI and colon were fixed in 10% formalin, paraffin embedded, and sectioned at 3 to 6 μm for H&E staining or immunostaining. The tissue sections were incubated with rabbit anti-c-myc ab (1:50), rabbit anti-pERK ab, or with control ab, overnight at 4°C. After washing with PBS, sections were incubated in HRP-conjugated secondary antibody for an hour and the staining was visualized with AEC peroxidase substrate kit (Vector Laboratories, Inc., Burlingame, CA), with hematoxylin nuclear counterstaining.

Blood hemoglobin was measured on a MS9 Blood Analyzer (Melet Schloesing Laboratories, El Cajon, CA) according to the manufacturer's instructions.

Statistical analysis was performed by Student's t test for paire samples or two-way ANOVA for multiple comparisons and by log-rank analysis for survival curves. Data are presented as means ± s.d.