Colon carcinogenesis is a multistage process involving genetic alterations of various tumor suppressor genes and oncogenes. Repressor element 1 silencing factor (REST ), which was originally discovered as a transcriptional repressor of neuronal genes, plays an important role in neuronal differentiation. In a previous genetic screening for tumor suppressor genes in human cancers, REST was identified as a candidate tumor suppressor gene in colorectal carcinogenesis. However, the role of Rest in colon carcinogenesis in vivo remains unclear because of the embryonic lethal phenotype of the conventional Rest knockout mouse. In the present study, we conditionally deleted the Rest gene in the intestinal epithelium and investigated the effect of Rest ablation in mouse colon tumorigenesis. A conditional ablation of Rest in the colonic crypts led to a rapid upregulation of Rest-targeted genes, such as Syt4, Bdnf, and Tubb3, suggesting that Rest actually suppresses the expression of its target genes in the colon. However, Rest ablation did not lead to any significant effect on the development of colon tumors in two independent mouse models of colon carcinogenesis. In addition, despite the upregulation of neuronal genes in the colonic crypts, no neuronal differentiation was observed in the colonic crypts and tumors after the Rest ablation. These results indicate that the loss of Rest expression by itself does not promote the development of colon tumors in mice, and suggest that REST may exert a tumor suppressing activity in conjunction with the additional genetic/epigenetic abnormalities that occur during colon carcinogenesis. (Cancer Sci 2011; 102: 1659–1664)
Repressor element 1 silencing factor (REST; also called neuron-restrictive silencing factor [NRSF]) was originally discovered as a transcriptional repressor of a number of neuronal genes.(1,2) REST binds to a conserved 21–23 bp motif known as repressor element 1 within the control regions of target genes, and recruits multiple co-factors through repressor domains to alter epigenetic modifications, leading to the generation of a silencing complex. REST is ubiquitously expressed in non-neuronal cells,(3) and it prevents neuronal gene expression in non-neuronal cells.(4)
A link between REST dysfunction and carcinogenesis has been recognized(5–7) in a number of cancers such as prostate cancer,(8) breast cancer,(9–11) small cell lung cancer,(12–16) medulloblastoma,(17–19) and neuroblastoma.(20–23) In an RNAi-based screening for tumor suppressor genes, REST was identified as a candidate novel tumor suppressor gene.(24) Consistent with this notion, a REST mutation was identified in a colon cancer cell line, DLD-1, and the REST locus is deleted in approximately one-third of human colon cancers (14 of 42 primary tumors and 13 of 38 cell lines).(24) In addition, exogenous REST has been shown to suppress the growth of the colon cancer cells that lack REST expression, suggesting that REST actually plays a role in the tumor suppression in vitro. Although REST-mediated cellular transformation is proposed to be associated with the PI3K pathway, the precise mechanism(s) underlying the involvement of REST in colon carcinogenesis remain unclear. In particular, there is no in vivo evidence that establishes the function of Rest in tumor suppression.
In the present study, we examined the effect of genetic ablation of Rest during colon carcinogenesis in vivo. We herein show that genetic deletion of Rest results in derepression of the Rest-targeted neuronal genes in the colonic crypts, however, Rest ablation does not promote mouse colon carcinogenesis.
Materials and Methods
Animals. All animal experiments were approved by the Animal Research Committee of the Graduate School of Medicine, Gifu University (Gifu, Japan). In a previous study, homozygous Rest knockout (KO) mice showed embryonic lethality around E10.5, with a growth retardation phenotype.(25) In the present study, in order to investigate the effect of Rest deletion on colon carcinogenesis in vivo, we used mice expressing conditional knockout alleles of Rest.(26) In the Rest conditional mice, the endogenous Rest loci were replaced by the conditional KO alleles carrying the floxed last exon, which encodes the CoRest binding site that is essential for the generation of the silencing complex.(27) An ires-Gfp sequence was inserted into the 3′-UTR of the Rest gene to monitor the transcription of the modified allele (Rest2lox allele). The Rest2lox allele was recombined into the Rest1lox allele in the presence of Cre recombinase. Despite the presence of the remaining exons 1–3 of the Rest1lox allele, altered Rest transcripts were not detected in Rest1lox/1lox mouse embryonic stem cells,(26) thus suggesting the 1lox allele to be equivalent to the conventional KO allele.
ApcMin/+ mice, doxycycline-inducible Cre mice, and intestinal epithelium-specific Cre-expressing (Fabp-Cre) mice were described previously.(26,28,29) Doxycycline-inducible Cre mice harbor two transgenes, Rosa26-M2rtTA and Col1A1-tetO-Cre.(26) The experimental mice were obtained by breeding.
Experimental procedures. We tested the effect of Rest ablation in mouse colon carcinogenesis in two independent experiments (Fig. 1). The first experiment (protocol 1) was a chemically-induced colon carcinogenesis model using doxycycline-inducible Cre-expressing mice. The other experiment (protocol 2) used the ApcMin/+ mouse colon carcinogenesis model combined with the Fabp-Cre mouse. In protocol 1, Cre recombinase was induced by doxycycline treatment after carcinogen exposure, mimicking when the Rest gene is lost after the initiation phase of carcinogenesis. In contrast, in protocol 2, Cre-mediated genetic recombination in the intestinal epithelia started before the completion of intestinal morphogenesis, thereby mimicking when the Rest recombination occurs before the initiation of carcinogenesis.
In protocol 1, doxycycline-inducible Cre; Rest2lox/2lox mice were separated into a Rest KO group (n = 27) and a control group (n = 11). Five-week-old mice were given a single i.p. injection of azoxymethane (15 mg/kg body weight; Wako, Osaka, Japan), a colon-specific carcinogen. One week later, the mice were fed dextran sodium sulfate (20 mg/mL; Wako), a potent tumor promoter for colon tumorigenesis, in their drinking water for 1 week. The mice in the Rest KO group were fed 2 mg/mL doxycycline (Sigma, St. Louis, MO, USA) in their drinking water, supplemented with 10 mg/mL sucrose three times per week (at weeks 8, 14, and 20), whereas mice in the control group were fed tap water throughout the experiment. All mice were killed at 22 weeks of age.
In protocol 2, intestine-specific Rest KO mice (ApcMin/+; Fabp-Cre+; Rest2lox/2lox mice, n = 23), heterozygous Rest KO mice (ApcMin/+; Fabp-Cre+; Rest2lox/+ mice, n = 32), and control mice (ApcMin/+; Fabp-Cre−; Rest2lox/2lox mice and ApcMin/+; Fabp-Cre−; Rest2lox/+ mice, n = 26) were examined for the development of colon tumors. All mice were housed in rooms without any chemical treatment during the experiment and were killed at 20 weeks of age.
In both protocols, the colons were cut open longitudinally, then washed with PBS. Visible tumors (larger than 0.5 mm in their maximum diameters) on the colon mucosa were counted, and their maximum diameters were measured. Tumor samples were fixed in 10% buffered formalin for 24 h and embedded in paraffin. Sections were stained with H&E, then serial sections were used for the immunohistochemical analysis. Immunostaining was carried out using an avidin–biotin immunoperoxidase assay. The primary antibodies used in the immunostaining were anti-β-catenin (1:1000 dilution; BD Biosciences, San Diego, CA, USA), anti-chromogranin A (1:1000 dilution; Dako, Carpinteria, CA, USA), and anti-Ki-67 (1:100 dilution; Dako).
Crypt isolation. In order to examine the effect of Rest ablation in the colonic epithelium, we carried out crypt isolation to exclude the contaminating stromal cells in the colonic mucosa, as described previously.(29) The removed colon was cut into three equal segments. The distal segment was used for crypt isolation.
Confirmation of Rest recombination. To examine the recombination status of the conditional Rest allele in the colon, we carried out a Southern blot analysis of the Rest loci. Total DNA was extracted from isolated intestinal crypts of doxycycline-inducible Rest knockout mice at the indicated time intervals (Fig. 2a). DNA samples (10 μg each) were digested with Mfe1 (Bio-Rad, Hercules, CA, USA). The digested DNA samples were electrophoresed, transferred onto nylon membranes, and hybridized with a DIG probe against the Rest loci in PerfectHyb (Toyobo, Osaka, Japan).(26) Signals were detected by chemiluminescence with LAS-4000 (Fujifilm, Tokyo, Japan). We also carried out a PCR-based analysis to examine the recombination of the Rest gene in colon tumors. DNA was isolated from formalin-fixed paraffin-embedding blocks using the Pinpoint Slide DNA Isolation System (Zymo Research, Orange, CA, USA). The PCR was carried out using primers specific for mouse floxed Rest. The primers for the Rest2lox allele were forward (F) (5′-CCCTTATGGGTGCAAGTGTT-3′) and reverse (R) (5′-GGGGGACAAAGCCACTCTA-3′). The primers for Rest1lox allele were F (5′-GGGTGCAAGTGTTCTCTTGTCT-3′) and R (5′-CAAGTAACTAAAAATTAGGAACTACCG-3′).
Real-time PCR analysis. Total RNA was extracted from the isolated intestinal crypts as described previously (Fig. 2a), and from the colon tumors of ApcMin/+ mice using the RNeasy Mini kit (Qiagen, Valencia, CA, USA). Total RNA (0.5 μg each) was reverse transcribed using Superscript III Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA).
Quantitative real-time PCR was carried out with the Thermal Cycler Dice Real Time System Single (Takara, Kyoto, Japan) using the SYBR Green (Takara) method. The primer sequences used in the quantitative real-time PCR analyses for Bdnf, Tubb3, and Rest were obtained from PrimerBank (http://pga.mgh.harvard.edu/primerbank/). The primers for β-actin were F (5′-CATCCGTAAAGACCTCTATGCCAAC-3′) and R (5′-ATGGAGCCACCGATCCACA-3′). The primers for Gfp were F (5′-ACCAGCAGAACACCCCCATC-3′) and R (5′-AGCTCGTCCATGCCGAGAGT-3′). The primers for Syt4 were F (5′-TGCTTTTGGCCTCGTCTTCA-3′) and R (5′-GCGGTTTTACCCTTCACTTCAC-3′).
Statistical analysis. Statistical significance was evaluated using either Student’s t-test or Welch’s t-test for paired samples. The results of experiments are presented as the mean ± SEM. P < 0.05 was considered to indicate a significant difference.
Cre expression induces genetic ablation of Rest in colonic mucosa. In order to examine the efficiency of Rest recombination in protocol 1, we carried out a Southern blot analysis using genomic DNA from isolated colonic crypts of doxycycline-inducible Cre; Rest2lox/2lox mice (Fig. 2a). The Southern blot analysis revealed that the majority of the Rest2lox alleles in the colonic crypts were recombined into Rest1lox alleles in doxycycline-treated mice (the Rest1lox allele was confirmed in seven out of eight colonic crypts). However, partial recombination of the Rest allele was also observed in some cases in the non-treated group (5 out of 13 colonic crypts), probably due to the leaky expression of the Cre transgene (Fig. 2b).
In protocol 2, Rest recombination was examined using a PCR assay with genomic DNA from formalin-fixed paraffin-embedding specimens of Fabp-Cre+; Rest2lox/2lox mice. The recombined Rest gene (Rest1lox allele) was only detected in mice containing the Fabp-Cre allele, and the Rest2lox allele was also detected in these mice, indicating partial recombination, which is consistent with a previous experiment showing a recombination efficiency of approximately 50% in the colonic crypts by the Fabp-Cre allele (Fig. 2c).(28) These results indicate that genetic ablation of Rest was induced successfully in the colons of two independent Cre-expressing mouse models. In both protocol 1 and 2, despite the presence of the Rest1lox allele, the mice were healthy, and no detectable difference was observed in the histological analyses in comparison to the control mice.
Transcript levels of Rest and Rest-targeted genes in colonic mucosa. We next examined the expression levels of Rest in colonic tumors of ApcMin/+ mice. The quantitative real-time PCR analysis revealed that Rest expression is not different between the colon tumors and non-neoplastic normal mucosa, suggesting that loss of Rest is not involved in the colon tumorigenesis of ApcMin/+ mice (Fig. 3a). In order to elucidate the effect of genetic ablation of Rest in the colonic mucosa, we measured the mRNA expression levels of the Rest gene and Rest-targeted genes in the isolated colonic crypts of doxycycline-inducible Cre; Rest2lox/2lox mice by quantitative real-time PCR analysis (Fig. 2a). Consistent with the genetic recombination observed in the Southern blot analysis, the expression of Rest in the colonic crypts was significantly decreased in the doxycycline-treated mice in comparison with the non-treated mice (Fig. 3b). In line with the decreased expression of the Rest gene, the expression of Gfp was also downregulated in the doxycycline-treated crypts (Fig. 3c). In contrast, the expression of Syt4, Bdnf, and Tubb3, which are known to be Rest-targeted genes, were significantly upregulated in the doxycycline-treated crypts (Fig. 3d). These results indicate that the genetic recombination of Rest results in the rapid derepression of the Rest-targeted genes, suggesting that Rest plays a role in the repression of neuronal genes in the colonic crypts.
Macroscopic analysis of colon tumor development. Macroscopically, polypoid tumors were observed in colonic mucosa of mice during both experiments. In protocol 1, genetic recombination into the Rest1lox allele was only confirmed in Cre-induced tumors (n = 4). In protocol 2, the recombined Rest allele (Rest1lox allele) was detectable in the majority of Rest KO tumors (5 out of 6 tumors), whereas all control tumors (n = 9) retained non-recombined Rest alleles (Rest2lox/2lox) (Fig. 2c). In protocol 1, the multiplicity of colon tumors was 0.37 ± 0.14/mouse in the Rest KO mice and 0.18 ± 0.18/mouse in control mice (Fig. 4a). In protocol 2, the multiplicity and the maximum diameter of colon tumors were 4.96 ± 0.57/mouse and 3.31 ± 0.12 mm in the Rest KO mice, 4.81 ± 0.51/mouse and 3.37 ± 0.12 mm in the Rest heterozygous mice, and 4.35 ± 0.71/mouse and 3.68 ± 0.15 mm in the control mice, respectively (Fig. 4b,c). No macroscopic tumors were observed in the intestine-specific Rest KO mice without the ApcMin allele (Apc+/+; Fabp-Cre+; Rest2lox/2lox mice, n = 4). Furthermore, a lack of colon tumor development in Fabp-Cre+; Rest2lox/2lox mice was also confirmed in the mice older than 12 months (data not shown), thus suggesting that Rest ablation alone is not sufficient to initiate colon tumorigenesis.
Collectively, the multiplicity of colon tumors in the Rest KO mice was slightly higher than that in the control mice, but the difference between the Rest KO and control mice was not significantly different (protocol 1: Rest KO vs. control, P = 0.46; protocol 2: Rest KO vs. control, P = 0.51). In addition, no statistically significant difference was observed in the size of the colon tumors, although there was a tendency for the tumor size to correlate with the expression of Rest (protocol 2: Rest KO vs. control, P = 0.054).
Histological and immunohistochemical analyses of colon tumors. Colon tumors were processed for histological examinations. As Rest recombination occurs partially by the Fabp-Cre transgene in protocol 2, the genetic status of Rest was determined in each tumor by PCR using primers specific for the Rest2lox and Rest1lox alleles. Regardless of the Rest recombination, the histological analysis revealed that all colon tumors consisted of tubular dysplastic glands. There were no detectable histological differences between the colon tumors of Rest KO and control mice in either protocol (Fig. 5a).
We further analyzed these tumors by immunostaining for β-catenin, Ki-67, and chromogranin A. The accumulation of β-catenin protein is a critical event that occurs during colon carcinogenesis. Indeed, in the present study, the accumulation of β-catenin was observed in the dysplastic glands in the colonic tumors. However, the β-catenin immunostaining showed no detectable difference between colon tumors in the Rest KO and control mice in either protocol (Fig. 5a).
We also carried out Ki-67 immunostaining to compare the proliferative activities of tumor cells with different genetic status of Rest. We counted Ki-67 immunopositive tumor cells out of 1000 randomly selected colonic tumor cells from Rest KO and control mice, and calculated the ratio of Ki-67 positive tumor cells. The Ki-67 positive cell ratios in the colon tumors of Rest KO and control mice were 27.6 ± 3.39% (n = 5) and 31.4 ± 4.84% (n = 5) in protocol 1 (P = 0.53), and 44.2 ± 3.00% (n = 10) and 41.2 ± 2.53% (n = 11) in protocol 2 (P = 0.45), respectively (Table 1).
Table 1. Ki-67 and chromogranin A positive tumor cell ratio in colon tumors of Rest knockout (KO) and control mice
Chromogranin A (%)
Protocol 1, chemically-induced colon carcinogenesis model using doxycycline-inducible Cre-expressing mice. Protocol 2, ApcMin/+ mouse colon carcinogenesis model combined with Fabp-Cre mouse.
27.6 ± 3.39
0.28 ± 0.08
31.4 ± 4.84
0.10 ± 0.07
44.2 ± 3.00
0.07 ± 0.02
41.2 ± 2.53
0.12 ± 0.02
Rest has been shown to be a master negative regulator of neuronal differentiation. Indeed, we confirmed that genetic ablation of Rest leads to the derepression of neuronal gene expression. Therefore, we next examined the enteroendocrine cell differentiation of the intestinal cells after loss of Rest expression. Chromogranin A immunostaining was used to assess the endocrine differentiation of both non-neoplastic and tumor cells in the colon. Chromogranin A immunostaining showed no detectable differences related to the Rest recombination in both non-neoplastic and tumor cells (Fig. 5b). The positive cell ratios for chromogranin A in Rest KO and control tumors was 0.28 ± 0.08% (n = 5) and 0.10 ± 0.07% (n = 5) in protocol 1 (P = 0.12), and 0.07 ± 0.02% (n = 10) and 0.12 ± 0.02% (n = 11) in protocol 2 (P = 0.17), respectively (Table 1).
Colon carcinogenesis is a multistage process involving genetic and epigenetic alterations of various tumor suppressor genes and oncogenes. Our previous studies identified two distinct stages of colon tumorigenesis in ApcMin/+ mice (microadenomas and macroscopic tumors).(30) The formation of microadenomas is accompanied by the activation of the canonical Wnt pathway through the genetic loss of the Apc gene, leading to the accumulation of β-catenin. Recent evidence suggests that DNA methylation is closely associated with the progression of microadenomas into macroscopic tumors.(28,31–33) Although the identity of the target genes of DNA methylation involved in colon tumorigenesis remain unclear, these findings suggest that the progression of colonic tumors in ApcMin/+ mice requires synchronous alternations in the expression of multiple genes due to global changes of epigenetic modifications.
REST, a novel candidate tumor suppressor gene in colon carcinogenesis, maintains transcriptional silencing of various genes by recruiting multiple co-factors, including Co-Rest,(27) HDAC and Sin3 complex,(34–36) histone H3 K9 methyltransferase G9a,(37) histone H3 K4 demethylase LSD1,(38) and methyl DNA binding protein MeCP2.(39) Previous experiments indicated thousands of REST-targeted genes in embryonic stem cells and neuronal progenitor cells.(40) REST is thus suggested to alter the chromatin structure in conjunction with its co-factors, while also regulating the transcription of the REST-targeted genes through histone modification, chromatin remodeling, and genomic methylation. Given the fact that DNA methylation, one of the most important epigenetic modifications, plays a critical role in the transition from microadenomas to macroscopic tumors in the colon of ApcMin/+ mice, we hypothesized that the global changes in epigenetic modifications caused by Rest ablation might affect murine colon tumorigenesis.
In this study, we confirmed that the genetic ablation of Rest leads to the decreased expression of Rest and increased expression of Rest-targeted genes, suggesting that Rest represses the Rest-targeted genes in the colonic crypts. However, Rest ablation at both the pre-initiation and post-initiation phases of colon tumorigenesis showed no significant effect on tumor development in the colon. These findings indicate that loss of Rest expression by itself does not promote the development of colon tumors in mice. It is possible that Rest deletion alone might not be sufficient for the active changes of epigenetic modifications to induce the progression of colon tumorigenesis.
Rest has been regarded as a master negative regulator of neuronal differentiation in non-neuronal cells. Indeed, a targeted mutation of Rest in mice caused derepression of neuron-specific genes in a subset of non-neuronal tissues. Human colonic carcinoma expressing neuroendocrine genes (also called neuroendocrine carcinoma, NEC) is a highly aggressive carcinoma that comprises approximately 0.6% of colonic carcinomas.(41) Importantly, most neuroendocrine genes expressed in NEC are targets of the REST-repressing complex. Considering that the genetic deletion of Rest leads to the upregulation of neuronal genes in non-neuronal cells, we hypothesized that Rest deletion would induce neuronal differentiation in colonic tumors, thus leading to a NEC-like phenotype. In fact, previous studies have revealed that carcinomas with REST dysfunction frequently showed neuroendocrine characteristics.(8,12–16) However, in the present study, colon tumors lacking the Rest gene did not show the NEC-like histology. In addition, an increase in chromogranin A-positive cells, which is usually observed in NEC, was not detectable in the Rest-deleted tumors. Our results suggest that, although Rest ablation leads to the increased mRNA expression of neuronal genes, Rest ablation alone is not sufficient to induce neuronal differentiation in the colon. As the incidence of NEC is extremely rare compared to the relatively high incidence of REST deletion in colorectal cancer,(24,41) it is still possible that REST inactivation, in conjunction with additional genetic and/or epigenetic alterations, may be involved in the development of NEC. In this context, it would be interesting to examine the genetic status of REST in NEC in future studies.
In summary, we have shown that the genetic ablation of Rest does not affect the development of colon tumors in mice. These findings suggest that other genetic and/or epigenetic alterations might therefore be required to exert the tumor-promoting effect of Rest ablation during multistage tumorigenesis of the colon.
We thank Kyoko Takahashi, Ayako Suga, and Yoshitaka Kinjo for their technical assistance. This study was supported by grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and grants from PRESTO, and from the Ministry of Health, Labour and Welfare of Japan.
The authors have no conflict of interest to declare.