Frequent Inactivation of Axon Guidance Molecule RGMA in Human Colon Cancer Through Genetic and Epigenetic Mechanisms

GASTROENTEROLOGY 2009;137:176 –187

Frequent Inactivation of Axon Guidance Molecule RGMA in Human Colon Cancer Through Genetic and Epigenetic Mechanisms VIVIAN S. W. LI,* SIU TSAN YUEN,*,‡ TSUN LEUNG CHAN,* HELEN H. N. YAN,* WAI LUN LAW,§ BONNIE H. Y. YEUNG,* ANNIE S. Y. CHAN,* WAI YIN TSUI,* SAMUEL SO,储 XIN CHEN,¶ and SUET YI LEUNG* *Department of Pathology, The University of Hong Kong, Queen Mary Hospital, Pokfulam, Hong Kong; ‡Department of Pathology, St. Paul’s Hospital, Causeway Bay, Hong Kong; §Department of Surgery, The University of Hong Kong, Queen Mary Hospital, Pokfulam, Hong Kong; 储Department of Surgery, Stanford University, Stanford, California; and ¶Department of Biopharmaceutical Sciences, University of California, San Francisco, California

BASIC– ALIMENTARY TRACT

BACKGROUND & AIMS: Repulsive guidance molecule member A (RGMA) is a glycosylphosphatidylinositol-anchored glycoprotein and axon guidance molecule that signals through its receptor, neogenin (NEO1), a homologue of the deleted-in-colorectal cancer (DCC) gene. RGMA also functions as a bone morphogenetic protein (BMP) coreceptor. We studied the potential roles of RGMA and NEO1 in colorectal cancer (CRC) pathogenesis. METHODS: We analyzed expression of RGMA and NEO1, as well as their epigenetic and genetic changes, in a large series of CRC samples, normal colon tissues, adenomas, and cell lines. These studies were accompanied by in vitro functional assay. RESULTS: RGMA and NEO1 expression were significantly down-regulated in most CRCs, adenomas, and cell lines. RGMA was frequently silenced by promoter methylation in CRCs (86.7%), adenomas (90.9%), and CRC cell lines (92.3%) but not in normal colon tissues; allelic imbalance of RGMA and NEO1 was observed in 40% and 49% of CRCs, respectively. In CRC samples, reduced RGMA levels were significantly associated with mismatch repair deficiency or mutations in KRAS or BRAF. Exposure to 5-aza-2=deoxycytidine restored RGMA expression in CRC cell lines. Transfection of RGMA into CRC cells suppressed cell proliferation, migration, and invasion and also increased apoptosis in response to DNA-damaging agent. CONCLUSIONS: The frequent genetic and epigenetic inactivation of RGMA in CRCs and adenomas along with its in vitro function collectively support its role as a tumor suppressor in colon cells. These findings add to the expanding list of axon guidance molecules with disrupted function during colon carcinogenesis and create new opportunities for early detection and drug development.

C

olorectal cancer (CRC) is a common cancer worldwide in which development of effective means for early detection and treatment remains a challenge. Axon guidance molecules are a family of genes that function to guide neurons and their axons to appropriate targets through attractive or repulsive interactions with associated receptors. They are also expressed in diverse tissue outside the brain where they act to regulate developmen-

tal processes and morphogenesis. Moreover, emerging data suggest that interaction between axon guidance cue and their receptors plays critical roles in carcinogenesis through the regulation of angiogenesis, cell survival, and apoptosis, as well as cell positioning and migration.1,2 This offers new therapeutic opportunities because agents that can mimic or block the activity of axon guidance molecules have been developed, which may be applicable in the treatment of cancers. There are 4 main categories of highly conserved families of axon guidance molecules, including semaphorins, slits, netrins, and ephrins. Dysregulation of genes or corresponding receptors within these categories, such as netrin-1, deleted-in-colorectal cancer (DCC), UNC5H, SLIT2, and EPHB2, was frequently observed in various types of cancers, including CRCs.1,3 The netrins and their receptors are well known to play critical roles in colon carcinogenesis. It has been shown that both DCC and UNC5H receptors act as dependence receptors for netrin-1, such that they promote cell death when netrin-1 is absent but enhance cell survival when netrin-1 is present.3 Although overexpression of netrin-1 leads to intestinal tumor development and malignant transformation in animal models,4 DCC and UNC5H are frequently inactivated in human CRCs by genetic and epigenetic mechanisms.5,6 Repulsive guidance molecule (RGM), a glycosylphosphatidylinositol-anchored glycoprotein, is a novel axon guidance molecule first identified in embryonic chicken that harbors repulsive and axon-specific guiding activity for retinal axons.7,8 Three orthologs of RGM have been identified in mice and human beings, including RGMa, RGMb, and RGMc.9 Vertebrate RGMa shows the highest Abbreviations used in this paper: 5aza-dC, 5-aza-2=-deoxycytidine; BMP, bone morphogenetic protein; bp, base pair; CPT, camptothecin; CRC, colorectal cancer; DAPI, 4=,6-diamidino-2-phenylindole; DCC, deleted-in-colorectal cancer; LOH, loss of heterozygosity; MSP, methylation-specific polymerase chain reaction; MTT, methyl thiazolyl tetrazolium; NEO1, neogenin; qRT-PCR, quantitative reverse transcription– polymerase chain reaction; RGM, repulsive guidance molecule; RGMA, repulsive guidance molecule member A. © 2009 by the AGA Institute 0016-5085/09/$36.00 doi:10.1053/j.gastro.2009.03.005

homology with chicken RGM and is most extensively characterized functionally, wherein the ligand/receptor signaling with neogenin (NEO1) to guide axon growth is confirmed.9 In contrast, the receptors for and roles of RGMb and RGMc remain mostly unknown. RGMa and netrin-1 share the common receptor NEO1, with which the 2 ligands were proposed to interact independently to provide respective repulsive and attractive cues for the navigation of axonal growth in developing vertebrae forebrain.10 Apart from the role of RGMa in nervous system development, little is known about the function of the RGM family in the gastrointestinal tract until a recent study that showed the expression of RGMa and RGMb in the proliferative basal crypt epithelial cells of mouse intestine.11 NEO1 was first identified as a homologue of the DCC gene.12 It acts as the receptor for both RGMa and netrin-1 in mediating axonal guidance.10 Netrin-1 can bind DCC, UNC5H, and NEO1, whereas RGMa interacts with NEO1 only,3,7 indicating a specific and unique role between the latter 2 molecules. Although NEO1/netrin-1 interaction was reported to regulate mammary gland morphogenesis,13 little is known about the role of RGMa/NEO1 signaling in epithelial cells or its possible involvement in neoplastic processes. We have recently characterized the expression profiles of a series of CRCs, adenomas, and normal colons and noticed a significant down-regulated expression of both RGMA and NEO1 in most CRCs and adenomas (detailed analysis of expression array data will be described in a separate manuscript). Interestingly, both RGMA and NEO1 were highly expressed in the basal proliferative crypt compartment as recorded in our previously published expression profile of colon crypt maturation,14 which was consistent with the data derived from in situ hybridization study of the mouse intestine.11 Thus, the paradoxical down-regulation of RGMA and NEO1 in the proliferative colonic neoplastic cells is intriguing and may suggest an aberrant down-regulation through genetic or epigenetic mechanisms. Given the intimate relation of netrin-1, DCC, and UNC5H in CRC development, we sought to validate the expression of RGMA and NEO1 in colon crypt development and cancers and to characterize the genetic and epigenetic mechanisms for their down-regulation in CRC. Finally, in vitro assays were performed to examine the functional role of RGMA in colon cancer cell lines.

Materials and Methods For details, see the Supplementary Methods (see supplemental material online at www.gastrojournal.org).

Clinical Samples and Cell Lines Frozen tumor and normal colon samples were obtained from colectomy specimens from Queen Mary Hospital, The University of Hong Kong. The study was approved by the local institutional review board. DNA

DYSREGULATION OF RGMA IN CRC

177

was extracted from 61 CRCs/paired normal colons and 11 adenomas from polypectomy specimens. RNA was extracted from the same set of tumors and 24 normal colon mucosae of the 61 CRC cases. Gene expression profile data were generated on a subset of these tumors with the use of cDNA microarrays according to methods previously described.14 Expression data corresponding to RGMA and NEO1 were extracted from the expression data set. RNA was subsequently extracted from an additional 104 CRCs, for a total of 165 cases for RGMA expression analysis in relation to clinicopathologic and molecular parameters (Supplementary Table 1). Protein lysate was extracted from 3 CRC/paired normal colons for immunoblotting analysis. RNA derived from 4 pairs of normal colon top and bottom crypt compartments was obtained as previously described.14 Thirteen colon cell lines were obtained from the American Type Culture Collection (Manassas, VA). DNA from 28 healthy blood donors was used as the control for methylation analysis.

Real-Time Quantitative and Semiquantitative Reverse Transcription–Polymerase Chain Reaction Quantitative reverse transcription–polymerase chain reaction (qRT-PCR) for RGMA and NEO1 was performed as described14 with the use of the TaqMan Gene Expression Assay (Applied Biosystems, Foster City, CA) in the CRC/normal tissue samples. Primers specific for the transcript 1 and transcript 2 of RGMA, and for NEO1 were also designed for RT-PCR analysis (Figure 1A; Supplementary Table 2).

Methylation Analysis, Pyrosequencing, and Demethylation Treatment Sample DNA was treated with sodium bisulfite as described,15 and the methylation status was assessed by methylation-specific PCR (MSP) (Figure 1B; Supplementary Table 3). To quantify the percentage of methylated alleles in RGMA gene promoter, we pyrosequenced (Pyrosequening AB, Uppsala, Sweden) the bisulfite-converted DNA with the method previously described (Supplementary Table 4).15 DLD1 and SW48 were treated with demethylating agent 5-aza-2=-deoxycytidine (5aza-dC) as described and assessed for RGMA re-expression and promoter demethylation.16

Mutation Screening and Loss of Heterozygosity Analysis To detect mutation, the 4 exons of the RGMA gene were directly sequenced. Seven microsatellite markers on chromosome 15q, spanning RGMA (15q26.1) and NEO1 (15q24.1) loci, were selected for loss of heterozygosity (LOH) analysis (Supplementary Tables 5 and 6).

BASIC– ALIMENTARY TRACT

July 2009

178

LI ET AL

GASTROENTEROLOGY Vol. 137, No. 1

Figure 1. (A) Genomic structures of the 2 RGMA transcripts. Black bars are the RT-PCR primers; gray bars, TaqMan primers and probe. (B) Map of RGMA promoters, positions of CpG sites (vertical bars) and primers for methylation studies. 42bp indicates 42-bp deletion polymorphism.

Transfection and In vitro Assays for Cell Proliferation, Anchorage-Independent Growth, Colony Formation, Cell Migration, Invasion, and Apoptosis BASIC– ALIMENTARY TRACT

The full-length cDNA of RGMA transcript 1 was amplified and subcloned into pIRES-EGFP bicistronic plasmid (Clontech, Mountain View, CA). DLD1 stable transfectants were established as cloned or pooled stable lines, whereas transient transfection was performed in LOVO cells. Unless otherwise stated, all in vitro assays were performed in triplicate and repeated at least once for consistency. Transient colony formation assay was performed as stated previously.17 Anchorage-independent growth was studied by soft-agar assay. Cell proliferation was assessed by the colorimetric methyl thiazolyl tetrazolium (MTT; Sigma, St Louis, MO) assay. Cell mobility was assessed by the wound scratch assay. Migration assays were performed with the of transwell filters with 8-␮m pores. Invasion assay was performed with the same protocol except that the membrane was precoated with 10% matrigel (Becton Dickinson, Franklin Lakes, NJ). DLD1 and LOVO transfectants were induced to undergo apoptosis through treatment with camptothecin (CPT; Calbiochem, San Diego, CA) or staurosporine (Roche, Indianapolis, IN) (for DLD1 transfectants only). The apoptotic population was quantified by both nuclear 4=,6-diamidino-2-phenylindole (DAPI) staining and flow cytometry after propidium iodide staining to assess the sub-G1 fraction.

Statistical Analysis Comparison of difference between groups was performed using the Student’s t test (when there was equal variance between groups), Mann–Whitney U test, or paired t test (for paired samples) as appropriate.

Results RGMA and NEO1 Expressions Are Down-Regulated in CRCs Figure 2A summarized the relative expression levels of RGMA and NEO1 measured by cDNA microarrays in colon tops versus basal crypts and normal colon mucosae versus adenomas and cancers. RGMA was highly expressed in basal crypt (P ⬍ .001), and down-regulated in both adenomas (P ⫽ .005) and cancers (P ⬍ .001). Expression of NEO1 showed a weaker but similar trend (top versus bottom, P ⫽ .02; adenomas versus normals, P ⫽ .04; cancers versus normals, P ⬍ .001). qRT-PCR validated the high expression level of RGMA in the bottom crypts from 4 top/bottom pairs of normal colon mucosae (P ⫽ .003; Figure 2B). Examination of 61 CRCs and 24 paired normal colons showed that RGMA levels were reduced in most CRCs except for 1 or 2 cases, in which the levels were still lower compared with proliferative basal crypts (P ⬍ .001; Figure 2B). Expression of RGMA in 11 adenomas also showed a reduced level compared with normal colon mucosae (P ⬍ .001). Good correlation was observed between the expression levels as measured by cDNA microarray and qRT-PCR (n ⫽ 65; r ⫽ 0.634, P ⬍ .001). Analysis of the 13 colon cancer cell lines also showed reduced RGMA levels in all cell lines (P ⬍ .001). SW620 had the highest RGMA level among the cell lines, but the expression level was still lower compared with that in the proliferative colon basal crypt compartment. The uniformly reduced expression of RGMA across all colon cell lines, CRCs, and adenomas suggested that RGMA inactivation is essential in CRC development and may play a critical role in the early events of tumor transformation. Although we noted a weaker trend of up-regulated NEO1 expression in colon basal crypts, such trend could not be confirmed by qRT-PCR analysis (Figure 2B). How-

DYSREGULATION OF RGMA IN CRC

179

Figure 2. (A) RGMA and NEO1 expression by microarray. Data are normalized against the mean of normal mucosae for CRCs and adenomas, and mean centered for colon top versus bottom crypts. Red and green colors indicate above and below mean expression according to the color scale bar; gray color, missing data. (B) Differential expression of RGMA and NEO1 in colon crypts and CRCs by qRTPCR. The expression level is normalized against the mean of colon bottom crypt. *P ⬍ .05; **P ⬍ .001. (C) Reduced protein levels of RGMA and NEO1 in CRCs (T) compared with their corresponding normal colon mucosae (N).

ever, significant reduction was observed in NEO1 expression by qRT-PCR in most CRCs compared with normal colon mucosae (P ⬍ .001). NEO1 expression levels were also reduced in most adenomas (P ⬍ .001) and some colon cell lines (P ⫽ .005). The qRT-PCR expression data correlate well with the cDNA microarray data (n ⫽ 67; r ⫽ 0.533, P ⬍ .001). Reduced protein levels of both RGMA and NEO1 was confirmed by immunoblotting analysis in 3 randomly selected CRCs compared with their corresponding normal colon mucosae (Figure 2C).

Frequent Hypermethylation of RGMA Promoter in CRC Cell Lines Because methylation-induced gene silencing is frequently observed in other axon guidance molecules in human cancers, we first examined for the presence of methylation in RGMA and NEO1 gene promoters in 13 colon cancer cell lines. RGMA resides in chromosome 15q26.1 and has 2 transcript variants differing in their 5= sequences (Figure 1A). Transcript 1 corresponds to the reference sequence of RGMA, whereas transcript 2 is still not fully characterized regarding its 5= end (Figure 1A). Although the qRT-PCR analysis cannot distinguish between the 2 transcripts, we designed specific RT-PCR primers to amplify them separately. RT-PCR analysis

confirmed the qRT-PCR expression data, with the absence of both transcripts of RGMA in most colon cell lines (Figure 3A). The 4 cell lines (SW620, SW480, HCT116, and CACO2), showing weak expression of either or both transcripts, corresponded to those with the highest levels of RGMA as measured by qRT-PCR, albeit still lower compared with that in normal colon. We then designed 5 MSP primer pairs spanning different regions of the promoters of both transcripts (see Figure 1B for primer positions). MSP analysis showed very frequent and variable extent of methylation of the RGMA gene promoters in the entire panel of colon cell lines, whereas methylation was completely absent in leukocytes from the 28 normal samples (Figure 3A). In general, cell lines with the most extensive methylation were more likely to have complete silencing of both transcripts. Overall, MSP2 is the region showing the highest frequency of methylation. To further quantify the methylation percentage in the cell lines, we performed pyrosequencing to assess 6 CpG sites in the promoter region encompassed by MSP2 (⫺643 to ⫺611 base pair [bp] from the translation start site of transcript 1). We noted a rather consistent pattern of methylation levels across the 6 sites; thus, the average was taken for downstream analysis (Figure 3B). The

BASIC– ALIMENTARY TRACT

July 2009

180

LI ET AL

GASTROENTEROLOGY Vol. 137, No. 1

BASIC– ALIMENTARY TRACT

Figure 3. (A) RGMA and NEO1 expression by RT-PCR (top) and qRT-PCR in relation to promoter methylation by MSP and pyrosequencing in 13 colon cell lines. NC indicates normal colon mucosa. (B) Methylation status in 6 CpG sites by pyrosequencing in the RGMA promoter. Re-expression of the RGMA transcripts (C) and demethylation of the RGMA promoter in DLD1 and SW48 after 5aza-dC treatment by both pyrosequencing (C) and MSP (D). The # indicates heterozygous for 42-bp polymorphism.

mean methylation levels as measured by pyrosequencing showed a consistent trend concordant with the MSP results (Figure 3A). Most cell lines showed methylation level of greater than 80%, suggesting biallelic methylation. For cell lines with a lower level of methylation (LS174T, LS180, and CACO2), other yet unknown mechanisms may act in concert with methylation to downregulate RGMA expression. To confirm the significance of the methylation detected, we performed demethylation treatment for 2 colon cell lines. Culture of DLD1 and SW48 in the presence of 5aza-dC successfully rescued the expression of the 2 RGMA transcripts (Figure 3C), which was paralleled by the demethylation of the gene promoter regions as confirmed by MSP, and a significant decrease of methylation

level from greater than 80% to less than 41%, as quantified by pyrosequencing (Figure 3C and D; Supplementary Figure 2A). We next characterized the expression of NEO1 and observed varying but detectable expression across all cell lines (Figure 3A). MSP analysis, however, did not show any methylation in all 13 colon cancer cell lines (Supplementary Figure 1), suggesting that methylation is an improbable cause for NEO1 down-regulation in colon cancer cell lines.

Frequent Hypermethylation of RGMA Promoter in CRCs and Adenomas We next examined for the methylation of RGMA in 61 CRCs and their corresponding normal colonic

tissues with pyrosequencing. One case was excluded eventually because of incomplete bisulfite conversion. We noted homogeneous methylation across all 6 CpG sites (spanning region ⫺643 to ⫺611 bp) in most CRCs and adenomas, whereas methylation in normal colon was either undetectable, present in very low levels, or affecting only 1 or 2 CpG sites (Figure 4A). Representative pyrosequencing chromatograms were shown in Supplementary Figure 2B. Figure 4B summarized the average methylation level across the 6 CpG sites in normal colons, corresponding cancer tissues, adenomas, and cell lines. Methylation levels were significantly higher in adenomas, CRCs, and colon cancer cell lines, compared with normal colon (P ⬍ .001). In normal colon, the average methylation level ranged from 0% to 9.98%, with a mean of 4.08%; thus, samples were scored as hypermethylated if the average methylation level was greater than 10%. Overall, 52 (86.7%) of the 60 CRCs and 10 (90.9%) of the 11 adenomas were hypermethylated. The results suggested that RGMA hypermethylation is an early and common event in CRC development. As for NEO1, MSP analysis was performed in 11 adenomas and 59 CRCs. No methylation was detected (see representative MSP results in Supplementary Figure 1C). We next performed LOH analysis to examine the possible involvement of allelic loss in RGMA and NEO1 down-regulation in CRCs.

Identification for Allelic Imbalance in RGMA and NEO1 in CRCs Loss of chromosome 15q loci has been reported in some CRC cases.18,19 Interestingly, RGMA and NEO1 are located only 20 Mb apart at 15q26.1 and 15q24.1, respectively. We thus examined for LOH with the use of 7 microsatellite markers spanning a 20-Mb region in chromosome 15q, encompassing both RGMA and NEO1 (Supplementary Table 7). Figure 4C summarizes the results for LOH analysis of the 7 markers amongst the 61 CRCs. We noted frequent LOH involving 15q spanning both RGMA and NEO1. Overall, when the 4 markers (D15S127, D15S158, CA22, and tgtc) encompassing RGMA were used, 40% informative cases (22/55) showed LOH in ⱖ1 markers. Among the other 3 markers (D15S645, D15S192, and D15S205) flanking NEO1, 49% informative cases (24/49) showed LOH involving ⱖ1 markers. Combining the 2 sets of markers, 39.6% (19/48) CRCs showed LOH involving both RGMA and NEO1.

Identification for Somatic Mutations of RGMA in CRCs Complete sequencing for all 4 exons of RGMA in 61 CRCs showed 2 novel somatic missense mutations (3.3%; Supplementary Figure 3A), c.763G⬎A (G255R) and c.1180G⬎A (A394T), that change evolutionarily conserved amino acids in the C-terminal conserved domain of the gene (Supplementary Figure 3B). The first mutation (G255R) changes the neutral glycine to a basic arginine, whereas the second mutation (A394T), located in

DYSREGULATION OF RGMA IN CRC

181

the hydrophobic region of RGMA, switches the amino acid from a nonpolar hydrophobic alanine to a polar hydroxyl threonine. Although their functional significance yet needs to be confirmed, both mutations may possibly distort the protein structure and thus affect its function. We also noted 3 somatic mutations affecting the noncoding regions of the gene, as well as some novel germline polymorphisms (summarized in Supplementary Table 8 and Supplementary Figure 3C), whereas their significance remains unknown. Taking into account the epigenetic and genetic alterations, 57 (93.4%) of 61 CRCs harbored ⱖ1 RGMA inactivation events (Figure 4D), with 3% showing combination of methylation and mutation, 30% with methylation and LOH, 54% with methylation alone, and 7% with LOH alone. Combining with the observed down-regulated gene expression, our results suggest that the inactivation of RGMA is critical for CRC initiation and progression. As for NEO1, in which methylation is uncommon, LOH and other yet unknown mechanisms may contribute to its down-regulated expression.

Relation of RGMA Expression with Clinicopathologic and Molecular Parameters We subsequently expanded the patient cohort to analyze the clinicopathologic correlation with RGMA expression. qRT-PCR study of 165 patients showed significant association between RGMA expression and specific molecular genetic changes, such that a lower RGMA level was observed in CRCs with mismatch repair deficiency (P ⬍ .001) or mutation in either BRAF V600E or KRAS (P ⫽ .003) (Table 1). We also observed a trend toward reduced RGMA expression in right-sided (P ⫽ .068) and poorly differentiated (P ⫽ .074) tumors. Interestingly, in the smaller cohort of 60 CRCs assessed for RGMA promoter methylation, increased methylation was identified to associate with the presence of either BRAF V600E or KRAS mutation (P ⫽ .014) (Table 1). Although significant correlation of RGMA expression with tumor stage or patient survival was not exhibited, the reported strong association of mismatch repair deficiency or mutation of BRAF or KRAS with the CpG island methylator phenotype (CIMP)20,21 may collectively suggest RGMA inactivation by the methylator pathway of CRC evolution.

RGMA Expression Inhibits Colony Formation and Cell Growth To study the functional role of RGMA in CRC, we restored the expression of RGMA in 2 colon cell lines for multiscope assessment. DLD1 and LOVO, both with hypermethylation in the 5 MSP sites and absence of RGMA expression, were selected for transfection with pIRES2EGFP/RGMA or empty pIRES2-EGFP vector as control. RGMA was expressed stably in DLD1 cells as pooled clones (DLD1/RGMA pooled) and also in single clone (DLD1/RGMA16), along with appropriate mock controls. For LOVO cells, only transient transfection was

BASIC– ALIMENTARY TRACT

July 2009

182

LI ET AL

GASTROENTEROLOGY Vol. 137, No. 1

BASIC– ALIMENTARY TRACT

Figure 4. (A) Schematic summary of pyrosequencing results in 61 CRC/normal colon pairs and 11 adenomas. The - indicates the case was excluded because of incomplete bisulfite conversion; dotted line, the methylation cutoff for scoring tumors as positive for methylation (mean ⬎ 10%). (B) Mean methylation percentage of the RGMA promoter as measured by pyrosequencing in colon tissues and cell lines. **P ⬍ .001 (C) Schematic summary of LOH results from the 7 microsatellite markers on chromosome 15q. Numbers in the brackets represent base pair positions away from the RGMA gene. (D) The bar chart summarizes distribution of the different RGMA-inactivating events among the 61 CRCs.

performed (LOVO/RGMA) as establishment of stably transfected lines failed because of its excessive susceptibility to RGMA transfection. Expression of RGMA was confirmed by both RT-PCR and Western blotting (Figure

5A). Under transient transfection, RGMA was found to suppress colony formation for both DLD1 (P ⫽ .005) and LOVO (P ⬍ .001; Figure 5B) cells. In soft-agar assay, we noted a reduced anchorage-independent growth in

July 2009

DYSREGULATION OF RGMA IN CRC

183

Table 1. Relations of RGMA Expression and Methylation Level with Clinicopathologic Parameters and Molecular Genetic Data in Colorectal Cancers RGMA mRNA expression

RGMA methylation percentage

No. of cases Mean expression ⫾ SD P value, t test No. of cases Mean methylation ⫾ SD, % P value, t test Mismatch repair deficiency Present Absent KRAS mutation Present Absent BRAF V600E mutation Present Absent BRAF or KRAS mutationa Present Absent

48 117

⫺1.091 ⫾ 1.502 0.234 ⫾ 1.491

69 96

⫺0.440 ⫾ 1.467 0.056 ⫾ 1.678

11 154

⫺1.076 ⫾ 1.952 ⫺0.085 ⫾ 1.567

80 85

⫺0.527 ⫾ 1.544 0.202 ⫾ 1.594

⬍.001

.587 12 48

22.1 ⫾ 8.8 24.1 ⫾ 12.2

23 37

27.5 ⫾ 10.4 21.4 ⫾ 11.8

3 57

31.0 ⫾ 5.8 23.4 ⫾ 11.7

26 34

27.9 ⫾ 10.0 20.6 ⫾ 11.9

.051

.048

.048

.270

.003

.014

LOVO/RGMA cells compared with mock controls (P ⫽ .002), but the effect was not observed in DLD1-transfected stable clones (Supplementary Figure 4A). We also noticed the suppression of cell growth in LOVO/RGMA cells compared with mock controls by MTT assay (Supplementary Figure 4B), yet again the effect was not observed in DLD1 cells.

RGMA Suppresses Colon Cancer Cell Migration and Invasion To evaluate the effect of RGMA in cancer cell migration and invasiveness, we first performed the monolayer wound-healing assay. We noted a delay in the closure of the wound gaps for DLD1/RGMA16, DLD1/ RGMA pooled cells, as well as LOVO/RGMA cells, compared with their mock controls (Figure 5C). For the quantitative assessment of cell migration, we next performed the transwell migration assay with the use of the modified Boyden chamber. We noted a significant inhibition of cell migration in DLD1/RGMA16, DLD1/ RGMA pooled, and LOVO/RGMA cells compared with their corresponding mock controls (Figure 5D, left). To assess the ability of RGMA in modulating cancer cell invasion, we coated the transwell membrane with 10% matrigel (Becton Dickinson) and repeated the experiment. Once again, we found the suppression of invasion in DLD1/RGMA16, DLD1/RGMA pooled, and LOVO/ RGMA cells compared with mock controls (Figure 5D, right). The suppression of migration by wound-healing assay and invasion were weaker in DLD1/RGMA pooled cells than in DLD1/RGMA16 cells. This may be due to a more heterogeneous population and a weaker expression of the RGMA protein in the pooled stable cells.

Expression of RGMA Increases Sensitivity to Apoptotic Treatment To examine the role of RGMA in regulating apoptosis, RGMA-transfected cells were treated with the

DNA-damaging agent CPT (Calbiochem), whereas apoptosis was assessed by DAPI nuclear staining and further quantified by measuring the sub-G1 population with the use of flow cytometry. Both DAPI nuclear staining and sub-G1 population quantification showed a significant increase of apoptotic cells in RGMA-transfected cells after 48 hours of induction by CPT. The apoptotic cells (sub-G1 population) increased from 40% in DLD1/Mock3 to 73% in DLD1/ RGMA16 as measured by flow cytometry (P ⬍ .001). Cell counting by DAPI staining showed a similar trend, with 8.6% apoptotic cells in DLD1/Mock3 compared with 34% in DLD1/RGMA16 (Figure 6A and B). Note that CPT induced G2 arrest in LOVO/Mock, which is p53 intact, but not in p53-deficient DLD1 cells. RGMA overexpression, however, could dramatically abrogate G2 arrest and accelerate apoptosis (Figure 6C). After treatment with 15 nmol/L CPT, the sub-G1 population increased from 11% in LOVO/Mock cells to 35% in LOVO/RGMA cells (P ⫽ .001). This was further confirmed by DAPI staining, with which apoptotic cells increased from 25% in LOVO/Mock cells to 57% in LOVO/RGMA cells (Figure 6D). The DLD1/RGMA16 also showed increased apoptosis in response to staurosporine, a protein kinase antagonist and apoptosis inducer, compared with mock control (Supplementary Figure 5). In conclusion, dissimilar to netrin-1, which inhibits apoptosis when interacting with its dependence receptor DCC or UNC5H, our results suggest that RGMA functions to induce apoptosis on DNA damage and other cellular stress. Thus, RGMA and netrin-1, albeit sharing the same receptor NEO1, may play different functions in CRC development, probably via different mechanisms.

Discussion In this study, we have identified the reduced expression of RGMA and NEO1 in colorectal adenomas,

BASIC– ALIMENTARY TRACT

Other variables analyzed, included sex, age, tumor stage, tumor type, overall, and disease-free survival, are not significant. There is a trend toward reduced RGMA mRMA expression levels for right-sided tumor (P ⫽ .068) and poor tumor differentiation (P ⫽ .074). aBRAF V600E and KRAS mutations exist in mutually exclusive relation to each other in CRCs.

184

LI ET AL

GASTROENTEROLOGY Vol. 137, No. 1

BASIC– ALIMENTARY TRACT

Figure 5. (A) RGMA expression was confirmed by RT-PCR (top) and Western blot (bottom) after transfection. DLD1/Mock3 and DLD1/RGMA16 indicate cloned stable transfectants; pooled, pooled stable transfectants. Expression of RGMA inhibits colony formation (B), wound-healing ability (C), migration, and invasion by transwell assays (D). Values represent mean from triplicate wells ⫾ 1 SD and normalized to 1 relative to Mock.

carcinomas, and colon cell lines. Similar to the observations with other axon guidance molecules and corresponding receptors, methylation is the predominant mechanism for the down-regulation of RGMA. Along with frequent LOH and infrequent missense mutations, 93% CRCs have ⱖ1 epigenetic or genetic inactivation event for RGMA. Furthermore, increased methylation and reduced expression of RGMA showed significant association with the presence of either BRAF V600E or KRAS mutations, both of which are molecular parameters strongly associated with the CIMP phenotype,20,21 suggesting the potential of RGMA as one of the target genes

in the methylator pathway of CRC development. Combined with previous literature reporting the frequent inactivation of other axon guidance molecules and corresponding receptors in CRCs, including the DCC, UNC5A, UNC5B, UNC5C, SLIT2, SLIT3, EPHA7, EPHB2, and EPHB4 genes, the findings in this study highlighted the importance of axon guidance pathways in colorectal carcinogenesis.6,17,22–26 Given the large proportion of CRCs being inactivated for each individual gene (eg, 89% in either DCC or UNC5C6; 72% in SLIT223; 93% in RGMA), it is highly likely that most CRCs would have concurrent inactivation of multiple axon guidance pathways. Since

DYSREGULATION OF RGMA IN CRC

185

BASIC– ALIMENTARY TRACT

July 2009

Figure 6. Expression of RGMA enhances apoptosis after camptothecin (CPT) treatment in DLD1 stable clone (A and B) and LOVO transient transfectants (C and D). Apoptosis was measured by flow cytometry after propidium iodide (PI) staining and also counting of apoptotic cells after DAPI nuclei staining (white arrows). Values represent mean from triplicate experiments ⫾ 1 SD.

186

LI ET AL

BASIC– ALIMENTARY TRACT

axon guidance molecules function in a cooperative network to guide correct axon growth and tissue morphogenesis, multiple hits along this network may be necessary to incur the cellular disorientation that characterizes cancer cell growth. Interestingly, individual members of the axon guidance molecules and corresponding receptors may play opposing roles in carcinogenesis. For example, DCC and UNC5H function as tumor suppressors in the dependence receptor pathway, whereas their ligand netrin-1 is oncogenic.3,4 RGM has been shown to signal through NEO1 as a dependence ligand/receptor couple to promote survival of neurons in chick embryo,7 thus our observed tumor suppressive roles for RGMA contradict with those reported for netrin-1. However, because RGMa/NEO1 and netrin-1/NEO1 interactions, respectively, provide repulsive cue and attractive cues in axon guidance,10 it is plausible to hypothesize that loss in repulsion may contribute to an increase in mobility, migration, and invasion in cancer cells. Similar to the circumstance for RGMa/NEO1, a loss of repulsive interaction between Ephrin-B/EphB in CRCs, albeit through down-regulated EphB receptors in the latter situation, increases cancer cell mobility in vitro and promotes tumor progression and metastasis in vivo.2,17,22 Furthermore, it has been shown that a complex hierarchical interaction exists between different axon guidance molecules, such that signaling of one ligand/receptor pair may interfere with signaling from other pairs. For example, interaction of Slit2 with its receptor Robo would promote the binding of Robo to DCC, which would in turn suppress the responsiveness of DCC to netrin-1.27 Correspondingly, SLIT2 is a putative tumor suppressor frequently inactivated by methylation in CRCs, whereas its expression suppresses colon cancer cell growth in vitro.23 It would thus be interesting to study whether RGMA/NEO1 signaling may possibly modulate netrin-1/DCC signaling in a similar way. In addition, the role of RGMa/NEO1 signaling is context dependent. For example, RGMa/NEO1 signaling was shown to induce apoptosis in Xenopus early embryos.28 To date, little is known about the function of RGMA/NEO1 signaling in the intestine. Our data suggest that RGMA may function to restrict colon cell migration and to enhance apoptosis in response to DNA damage and other cellular stress. The basal crypt expression of RGMA may thus help to prevent the stem cell compartment from acquiring undesirable mutation or undergoing aberrant migration. RGMA may mediate its effect on cell migration and invasion by specific signaling pathways. For example, RGMa/NEO1 signaling would lead to the activation of RhoA to mediate growth cone collapse in neurons.29 Because the Rho/ROCK pathway is closely involved in cancer cell migration, RGMA may inhibit cancer cell migration and invasion under the same mechanism. Apart from its involvement in axon guidance, RGMa also acts as bone morphogenetic protein (BMP) corecep-

GASTROENTEROLOGY Vol. 137, No. 1

tor that enhances BMP signaling.30,31 Specifically, RGMa was shown to use both BMP receptor type II (BMPRII) and activin receptor type IIA (ActRIIA) for BMP2 and BMP4 signaling.31 BMP2 was shown to inhibit proliferation and to induce apoptosis and differentiation in colon cancer cell lines.32 Thus, RGMA may modulate BMP signaling to promote apoptosis in the presence of DNA damage. If RGMA indeed mediates part of its tumor suppressive effect by modulating BMP signaling, our data will provide a new mechanism to the inactivation of BMP signaling in CRCs on top of the previously reported ones; which include BMP2 silencing by promoter methylation, expression of BMP antagonist GREM1 in the cancer stromal cells, and mutation of BMPR1A or SMAD4 in juvenile polyposis.33–36 Overall, emerging data suggest the frequent involvement of 2 key pathways in CRC development, including the axon guidance family and BMP signaling, both infer therapeutic potentials. RGMA has the unique property of participating in both pathways and may therefore represent an important novel tumor suppressor gene in the early events of colon carcinogenesis. The common occurrence of RGMA promoter hypermethylation in both adenomas and cancers also renders it an ideal biomarker for the development of stool- or serum-based diagnostic strategy for early detection.

Supplementary Data Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at doi: 10.1053/j.gastro.2009.03.005. References 1. Chedotal A, Kerjan G, Moreau-Fauvarque C. The brain within the tumor: new roles for axon guidance molecules in cancers. Cell Death Differ 2005;12:1044 –1056. 2. Cortina C, Palomo-Ponce S, Iglesias M, et al. EphB-ephrin-B interactions suppress colorectal cancer progression by compartmentalizing tumor cells. Nat Genet 2007;39:1376 –1383. 3. Arakawa H. Netrin-1 and its receptors in tumorigenesis. Nat Rev Cancer 2004;4:978 –987. 4. Mazelin L, Bernet A, Bonod-Bidaud C, et al. Netrin-1 controls colorectal tumorigenesis by regulating apoptosis. Nature 2004; 431:80 – 84. 5. Fearon ER, Cho KR, Nigro JM, et al. Identification of a chromosome 18q gene that is altered in colorectal cancers. Science 1990;247:49 –56. 6. Shin SK, Nagasaka T, Jung BH, et al. Epigenetic and genetic alterations in Netrin-1 receptors UNC5C and DCC in human colon cancer. Gastroenterology 2007;133:1849 –1857. 7. Matsunaga E, Chedotal A. Repulsive guidance molecule/neogenin: a novel ligand-receptor system playing multiple roles in neural development. Dev Growth Differ 2004;46:481– 486. 8. Monnier PP, Sierra A, Macchi P, et al. RGM is a repulsive guidance molecule for retinal axons. Nature 2002;419:392–395. 9. Schmidtmer J, Engelkamp D. Isolation and expression pattern of three mouse homologues of chick Rgm. Gene Expr. Patterns. 2004;4:105–110.

10. Wilson NH, Key B. Neogenin interacts with RGMa and netrin-1 to guide axons within the embryonic vertebrate forebrain. Dev Biol 2006;296:485– 498. 11. Metzger M, Conrad S, varez-Bolado G, et al. Gene expression of the repulsive guidance molecules during development of the mouse intestine. Dev Dyn 2005;234:169 –175. 12. Vielmetter J, Kayyem JF, Roman JM, et al. Neogenin, an avian cell surface protein expressed during terminal neuronal differentiation, is closely related to the human tumor suppressor molecule deleted in colorectal cancer. J. Cell Biol. 1994;127:2009 –2020. 13. Srinivasan K, Strickland P, Valdes A, et al. Netrin-1/neogenin interaction stabilizes multipotent progenitor cap cells during mammary gland morphogenesis. Dev Cell 2003;4:371–382. 14. Kosinski C, Li VS, Chan AS, et al. Gene expression patterns of human colon tops and basal crypts and BMP antagonists as intestinal stem cell niche factors. Proc Natl Acad Sci U S A 2007;104:15418 –15423. 15. Chan TL, Yuen ST, Kong CK, et al. Heritable germline epimutation of MSH2 in a family with hereditary nonpolyposis colorectal cancer. Nat Genet 2006;38:1178 –1183. 16. Veigl ML, Kasturi L, Olechnowicz J, et al. Biallelic inactivation of hMLH1 by epigenetic gene silencing, a novel mechanism causing human MSI cancers. Proc Natl Acad Sci U S A 1998;95:8698 – 8702. 17. Guo DL, Zhang J, Yuen ST, et al. Reduced expression of EphB2 that parallels invasion and metastasis in colorectal tumours. Carcinogenesis 2006;27:454 – 464. 18. Paredes-Zaglul A, Kang JJ, Essig YP, et al. Analysis of colorectal cancer by comparative genomic hybridization: evidence for induction of the metastatic phenotype by loss of tumor suppressor genes. Clin Cancer Res 1998;4:879 – 886. 19. Tsafrir D, Bacolod M, Selvanayagam Z, et al. Relationship of gene expression and chromosomal abnormalities in colorectal cancer. Cancer Res 2006;66:2129 –2137. 20. Weisenberger DJ, Siegmund KD, Campan M, et al. CpG island methylator phenotype underlies sporadic microsatellite instability and is tightly associated with BRAF mutation in colorectal cancer. Nat Genet 2006;38:787–793. 21. Shen L, Toyota M, Kondo Y, et al. Integrated genetic and epigenetic analysis identifies three different subclasses of colon cancer. Proc Natl Acad Sci U S A 2007;104:18654 – 18659. 22. Batlle E, Bacani J, Begthel H, et al. EphB receptor activity suppresses colorectal cancer progression. Nature 2005;435:1126 – 1130. 23. Dallol A, Morton D, Maher ER, et al. SLIT2 axon guidance molecule is frequently inactivated in colorectal cancer and suppresses growth of colorectal carcinoma cells. Cancer Res 2003;63: 1054 –1058. 24. Dickinson RE, Dallol A, Bieche I, et al. Epigenetic inactivation of SLIT3 and SLIT1 genes in human cancers. Br J Cancer 2004;91: 2071–2078. 25. Wang J, Kataoka H, Suzuki M, et al. Downregulation of EphA7 by hypermethylation in colorectal cancer. Oncogene 2005;24: 5637–5647. 26. Davalos V, Dopeso H, Castano J, et al. EPHB4 and survival of colorectal cancer patients Cancer Res. 2006;66:8943– 8948.

DYSREGULATION OF RGMA IN CRC

187

27. Stein E, Tessier-Lavigne M. Hierarchical organization of guidance receptors: silencing of netrin attraction by slit through a Robo/ DCC receptor complex. Science 2001;291:1928 –1938. 28. Shin GJ, Wilson NH. Overexpression of repulsive guidance molecule (RGM) a induces cell death through Neogenin in early vertebrate development. J Mol Histol 2008;39:105–113. 29. Conrad S, Genth H, Hofmann F, et al. Neogenin-RGMa signaling at the growth cone is bone morphogenetic protein-independent and involves RhoA, ROCK, and PKC. J Biol Chem 2007;282: 16423–16433. 30. Babitt JL, Zhang Y, Samad TA, et al. Repulsive guidance molecule (RGMa), a DRAGON homologue, is a bone morphogenetic protein co-receptor. J Biol Chem 2005;280:29820 –29827. 31. Xia Y, Yu PB, Sidis Y, et al. Repulsive guidance molecule RGMa alters utilization of bone morphogenetic protein (BMP) type II receptors by BMP2 and BMP4. J Biol Chem 2007;282:18129 – 18140. 32. Hardwick JC, Van Den Brink GR, Bleuming SA, et al. Bone morphogenetic protein 2 is expressed by, and acts upon, mature epithelial cells in the colon. Gastroenterology 2004;126:111– 121. 33. Wen XZ, Akiyama Y, Baylin SB, et al. Frequent epigenetic silencing of the bone morphogenetic protein 2 gene through methylation in gastric carcinomas. Oncogene 2006;25:2666 – 2673. 34. Sneddon JB, Zhen HH, Montgomery K, et al. Bone morphogenetic protein antagonist gremlin 1 is widely expressed by cancer-associated stromal cells and can promote tumor cell proliferation. Proc Natl Acad Sci U S A 2006;103:14842– 14847. 35. Howe JR, Bair JL, Sayed MG, et al. Germline mutations of the gene encoding bone morphogenetic protein receptor 1A in juvenile polyposis. Nat Genet 2001;28:184 –187. 36. Howe JR, Roth S, Ringold JC, et al. Mutations in the SMAD4/ DPC4 gene in juvenile polyposis. Science 1998;280:1086 – 1088.

Received March 25, 2008. Accepted March 10, 2009. Reprint requests Address requests for reprints to: Suet Yi Leung, MD, FRCPath, Department of Pathology, The University of Hong Kong, Queen Mary Hospital, Pokfulam, Hong Kong. e-mail: [email protected]; fax: (852) 2872-5197. Acknowledgments The authors thank Patrick O. Brown and colleagues at the Stanford Functional Genomic Facility and the Stanford Microarray Database for their collaboration and support on microarray experiment. Conflicts of interest The authors disclose no conflicts. Funding This work is supported by grants from the Research Grants Council of the Hong Kong Special Administrative Region (Project No. HKU7622/05M and HKU7697/08M).

BASIC– ALIMENTARY TRACT

July 2009