Introduction
Tumorigenesis is a multi-step process that results from progressive accumulation of genetic and epigenetic alterations in different genes. Chromosomal loss, which leads to the inactivation of tumor suppressor genes, is one of the most common genetic alterations detected in human cancer. Previous publications have documented that the p36 band of chromosome 1, 1
p36, is frequently deleted in a wide range of human cancers, including those of epithelial, neural and hematopoietic origin [
1]. Recently, Bagchi
et al. identified
CHD5, which localizes to the deletion region at 1
p36, as a tumor suppressor gene through functional analysis in a mouse model.
Chd5 knock-down was associated with hyperproliferation and reduced apoptosis and senescence, primarily through the p19
Arf/p53 pathway [
2].
CHD5 belongs to the chromodomain helicase DNA binding domain (CHD) family, which is a subclass of the Swi/Snf proteins [
3,
4]. Of nine members of the CHD family (CHD1-CHD9), two (CHD3 and CHD4) are components of the nucleosome remodeling and histone deacetylation (NuRD) complex and play an important role in chromatin remodeling [
5,
6]. As CHD5 shares the same functional domain with CHD3 and CHD4, it may also modulate chromatin remodeling and thus affect normal development and cancer.
Evidence that CHD5 functions as a tumor suppressor in human cancer has principally come from studies of neuroblastoma, in which
CHD5 mRNA expression was down-regulated likely through promoter methylation in tumors, and high expression of CHD5 was statistically associated with better patient survival [
7]. Furthermore, ectopic expression of CHD5 in neuroblastoma cell lines suppressed clonogenicity and tumor growth [
6,
7]. Hypermethylation of
CHD5 promoter has also been detected in gastric, colorectal and ovarian cancers, and somatic mutations have been detected in ovarian cancer [
8‐
11]. However, whereas 1
p36 is commonly deleted in human breast cancers, the role of CHD5 in breast cancer has not been evaluated.
We hypothesize that CHD5 is a tumor suppressor gene in breast cancer and tested this hypothesis in this study. We examined CHD5 for somatic mutation, copy number changes, mRNA and protein expression, and promoter methylation in primary tumors and cell lines from human breast cancer. In addition, we assessed its effect on cell proliferation in vitro and in vivo. We found that while CHD5 mutation was relatively rare, it had frequent down-regulation, hemizygous deletion and promoter methylation. Promoter hypermethylation correlated with lower levels of CHD5 mRNA expression, and demethylating treatment decreased promoter methylation and increased CHD5 expression, suggesting that promoter methylation is responsible at least in part for reduced CHD5 expression in breast cancer. Interestingly, reduced CHD5 expression significantly correlated with lymph node metastasis, recurrence and shorter patient survival in breast cancer. In addition, ectopic expression of CHD5 suppressed cell proliferation and tumor growth of the MDA-MB-231 cell line by arresting cell cycle at the G0/G1 phase, and inhibited invasiveness of MDA-MB-231 and Hs 578T cells in vitro. Our results strongly support a tumor suppressor role for CHD5 in breast cancer.
Materials and methods
Cell lines and tissues
In total, 32 breast cancer cell lines (BT-20, BT-474, BT-549, BT-483, CAMA-1, DU4475, HCC1143, HCC1395, HCC1500, HCC1599, HCC1806, HCC1937, HCC202, HCC2218, HCC38, HCC70, Hs 578T, MCF7, MDA-MB-134, MDA-MB-157, MDA-MB-175, T-47D, MDA-MB-231, MDA-MB-361, MDA-MB-415, MDA-MB-453, MDA-MB-468, SW527, UACC893, ZR-75-1, ZR-75-30, BRF-71T) and two immortalized non-neoplastic breast epithelial cell lines (184A1 and BRF-97T) were used in this study. Of these, BRF-97T and BRF-71T were purchased from Biological Research Faculty & Facility (BRFF, Ijamsville, MD, USA), and the rest were purchased from American Type Culture Collection (Manassas, VA, USA). The primary culture of human mammary epithelial cells (HMEC) was purchased from Cambrex (East Rutherford, NJ, USA). Cells were cultured according to the manufacturers' protocols.
Tumor tissues and adjacent normal tissues were obtained from 377 breast cancer patients who received surgery in the Cancer Hospital of Tianjin Medical University, Tianjin, China. Freshly frozen tissues from 107 patients were used to extract RNA or DNA samples, of which 58 were stored in RNAlater® solution and used for RNA extraction and CHD5 mRNA expression analysis, 30 had bisulfite-treated DNA from a previous study and were used for promoter methylation analysis (19 of them also used for RNA expression analysis), and 38 were used for mutation detection. Formalin-fixed paraffin-embedded primary tumor specimens from 289 patients were prepared in a tissue microarray format and used for CHD5 protein expression analysis by immunohistochemical (IHC) staining, which included 19 tumor specimens that were also used for mRNA expression and promoter methylation analyses. Samples were from the tissue bank at the Cancer Hospital of Tianjin Medical University and were used entirely based on availability. Use of human tissues and clinicopathological information was approved by the Institutional Ethics Committee. Genomic DNA was extracted using the DNeasy Blood & Tissue kit (Qiagen, Shanghai, China ), and total RNA was isolated using the RNeasy Mini kit (Qiagen).
The numbers of samples for different analyses were different because we used samples from our previous studies and many of them ran out of DNA, RNA or bisulfite-treated DNA. The majority of specimens used for IHC staining did not have DNA or RNA samples available. All samples with a result for any readout were included in this manuscript.
Real-time and semi-quantification PT-PCR
Reverse transcription reaction was performed using 2 μg total RNA with M-MLV reverse transcriptase (Promega, Madison, WI, USA). Real-time PCR reaction was performed in 25 μl reaction volume using SYBR Premix Ex Taq kit (TaKaRa, Dalian, China) and carried out on an iQ5 Multicolor Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). The CHD5 expression levels were normalized according to GAPDH in each sample. Primers used for CHD5 real time PCR were 5'-CAAGTGTAAAGGGAAGCGGAAGAAG-3' (forward) and 5'-CTTTTTATTCGGGGAGTAGTCAC-3' (reverse), and those for GAPDH were 5'-GGTGGTCTCCTCTGACTTCAACA-3' (forward) and 5'-GTTGCTGTAGCCAAATTCGTTGT-3' (reverse). CHD5 mRNA expression reading in HMEC was defined as 1, and CHD5 expression levels for other samples were normalized accordingly. Primer sequences for CHD5 semi-quantification PCR were 5'-TGAAGAAACTGCGGGATG-3' (forward) and 5'-TGCCGAACTTGAGGATGT-3' (reverse).
Mutation and methylation analysis
The 40 coding exons and adjacent splicing junctions of
CHD5 were amplified by PCR using previously reported primers and procedures [
12]. PCR products were purified and sequenced (Invitrogen, Beijing, China). When a sequence alteration was detected, PCR and sequencing were repeated to confirm it. For confirmed alterations in tumor samples, PCR and sequencing were performed in matched normal samples to determine whether an alteration was somatic or germline.
Promoter methylation analysis was conducted following a published method [
13]. Two pairs of primers, 5'- TTTAGGGGTTTGTATGGGTTTTAG-3' (F1)/5'-CCCTCTCCAAAAAAAATTAAAAAA-3' (R1) [
7] and 5'-TTTTTTTGGAGAGGGGGTTAGG-3' (F2)/5'-CTATAACAACCCCATCCCAT-3' (R2), were used. These primers amplify the CG rich region of the
CHD5 promoter from -841 to -246. Purified PCR products were cloned into pMD18-T vector (TaKaRa, Dalian, China). At least eight clones for each sample were sequenced. Methylation levels for each CpG site were indicated by the ratio of methylation positive clones to the total number of clones sequenced, and were categorized into three groups in cell lines: high (ratio > 0.5), low (0 < ratio < 0.5) and none (ratio = 0) (marked as dark, gray and white dots, respectively.). The average methylation level at a CpG site was also presented.
Copy number detection
Copy number changes in the
CHD5 gene were detected by real-time PCR using the SYBR
Premix Ex Taq kit (TaKaRa). Primer sequences for
CHD5 were 5'-CTGTTGCTGCAGTTCCTTCTC-3' and 5'-ATGAAGGACAGAACCTGCCTG-3'. The
RNaseP gene, which rarely has copy number changes in tumors [
14], was used as an internal control for a stable diploid copy number. Primer sequences for
RNaseP were 5'-CCTGGTACATGCCACTGATG-3' and 5'-AGTGTAGAGGGCAAGCCAGA-3'. Briefly, 10 ng genomic DNA from each sample was used as the template for PCR with either
CHD5 or
RNaseP primers in a volume of 20 μl. The final concentration for each primer was 0.25 μM.
CHD5 copy number was determined by dividing the ΔCt value for
CHD5 by that for
RNaseP. For cell lines, the average
CHD5/
RNaseP ratio among a pool of normal human genomic DNA (HGD) was defined as 1 and used as the value for a normal genome, and a deletion was defined by a ratio that was 0.5 or less. For primary tumors, a deletion was defined when the
CHD5/
RNaseP ratio in a tumor was about half or less of that in its matched non-cancerous tissue.
Western blotting
Western blotting was performed according to a published protocol [
15]. Antibodies for the following proteins were used: CHD5 (rabbit polyclonal, 1:1,000 dilution, Strategic Diagnostics, Newark, DE, USA), β-actin (A1978, Sigma-Aldrich, Beijing, China), p21 (#610233, BD Biosciences, Franklin Lakes, NJ, USA), p53 (sc-47698, Santa Cruz Biotechnology, Santa Cruz, CA, USA), vimentin (#3932, Cell Signaling Technology, Beverly, MA, USA), N-cadherin (#610920, BD Biosciences), E-cadherin (#3195, Cell Signaling) and ZEB1 (sc-25388, Santa Cruz Biotechnology).
Immunohistochemical (IHC) staining
Immunohistochemical analysis was performed as previously described [
16], using polyclonal rabbit anti-CHD5 antibody (1:2,000 dilution). The pre-immune serum was used as the negative control. In evaluating the specificity of the CHD5 antibody, we made cell pellets of ZR-75-1 (CHD5-negative) and T-47D (CHD5-positive) cells in agarose, fixed them in formalin, embedded in paraffin, prepared sections, and then conducted IHC staining. Detection of strong signals in T-47D but no signals in ZR-75-1 cells validated the antibody for IHC staining.
Protein expression levels of CHD5 were calculated by multiplying the intensity of nucleus staining (0 = no staining, 1 = low intensity, 2 = medium intensity, and 3 = high intensity) with the percentage of positively stained cells (0 = no, 1 = 1 to 25%, 2 = 26 to 50%, 3 = 51 to 100%). A combined index (intensity score × percentage score) of 0 or 1 was considered as negative staining (-), while an index of 2 or 3 as weak staining (+), 4 to 6 as moderate staining (++), and 9 as strong staining (+++). In statistical analysis, CHD5 expression levels were categorized into two groups: lack of expression (-) and weak to strong expression (+, ++, and +++).
The
CHD5 expression plasmid was constructed by cloning the coding region of
CHD5 into the pcDNA3.0 vector (Invitrogen). MDA-MB-231 cells were seeded in 6-well plates at a density of 5 × 10
5 cells per well and transfected by using the lipofectamine 2000 reagent (Invitrogen, Beijing, China). Forty-eight hours after transfection, one well of cells was harvested to confirm CHD5 expression, and the remaining wells were cultured in selection medium containing 800 μg/ml G418 (Sigma) for two weeks. Cells were then fixed, stained with sulforhodamine B (SRB), and measured for optical intensities, which indicated cell numbers, as described in our previous publication [
17].
RNA interfering
Small-interfering RNAs (siRNAs) for the CHD5 gene were purchased from Qiagen and Invitrogen. Two of them, named Q7 (Qiagen) and I20 (Invitrogen), were confirmed to be capable of knocking down CHD5 and thus were used. The target sequence of Q7 was 5'-ACGGTACATGATCCTCAACGA-3' and that of I20 was 5'-CAGCAGTTCTGCTTCCTCCTCTCAA-3'. A siRNA that does not target any known genes was purchased from Rui Bo (Guangzhou, China) and used as a control. The Lipofectamine RNAiMAX transfection reagent (Invitrogen) was used to transfect siRNAs into cells following the manufacturer's instruction.
Flow cytometric cell cycle analysis
MDA-MB-231 cells transfected with pcDNA3.0 vector or CHD5 expression plasmid were selected using G418 (800 μg/ml; Sigma) for two weeks. 1 × 106 cells of each group were collected, fixed with 70% ethanol for 24 hours, washed and re-suspended with phosphate-buffered saline (PBS), incubated with propidium iodide (20 mg/ml, Sigma) and Rnase A (20 mg/ml) for 30 min in the dark, and analyzed on a BD FACSCalibur flow cytometer (Franklin Lakes, NJ, USA).
Cell proliferation analysis
After transfection and selection with G418 for two weeks, MDA-MB-231 cells were harvested, diluted into five cells/ml using selection medium, plated into 96-well plates at 100 μl per well, and cultured for another two to three weeks. Cells surviving in a well were propagated and confirmed for CHD5 expression. Parental or clones of transfected cells were plated into 12-well plates at 3 × 104 cells/well, cultured for different times, fixed, stained with SRB and measured for cell numbers.
Hs 578T cells were also transfected with CHD5-expressing plasmid or control plasmids using lipofectamine 2000 reagent. After selection using G418 (400 μg/ml) for two weeks, the population cells of each group were plated into 12-well plates and used for cell proliferation analysis.
Tumor growth assay
One × 106 parental or clones of transfected MDA-MB-231cells were mixed with Matrigel (BD Biosciences, San Jose, CA, USA) and injected subcutaneously into four weeks old female BALB/c nude mice (eight mice for each group). Tumor volumes were measured once a week and calculated as below: L × W2 × 0.523 mm3, where L and W indicate the length and the width of a tumor. Mice were sacrificed at five weeks after injection, and tumors were surgically removed, weighed, and fixed in 4% formalin. Hematoxylin and eosin staining was performed for histological examination. The animal experiment was approved by the animal care and use center of Nankai University, Tianjin, China.
Invasion assay
The Boyden chamber assay was used for invasion assay. Briefly, 1 × 105 cells suspended in 200 μl serum-free medium were plated into the top chamber with 50 μl Matrigel-coated membrane (8 μM pole size, BD Biosciences, Shanghai, China). The chambers were then placed into 24-well plates with 600 μl serum-containing (10%) medium in each well. After 24 hours incubation, cells on the bottom side of the chamber membrane were fixed, stained with crystal violet, and photographed.
Statistics
The SPSS 15.0 software package (SPSS Inc, Chicago, IL, USA) was used for statistical analysis. Correlations between CHD5 expression and clinicopathological variables were evaluated by Chi-square (χ2) test. Progression-free survival (PFS) and overall survival (OS) rates were estimated by the Kaplan-Meier analysis. The Cox proportional hazard regression analysis was performed for the identification of relevant prognostic factors. For tumorigenesis experiments, tumor sizes or weights between two groups were analyzed using the Student's t test. The statistical analysis was two-sided and P-values less than 0.05 were considered as statistically significant.
Discussion
Deletion of 1
p36 is common in a variety of cancers, including breast cancer, suggesting that there is an important tumor suppressor gene in this region whose inactivation drives tumorigenesis. By analyzing mouse models generated by chromosome engineering to obtain different dosages of genomic interval D4Mit190-51, which corresponds to human 1
p36, Bagchi
et al. demonstrated that
CHD5 was the leading tumor suppressor gene in this interval [
2].
CHD5 has also been identified as a candidate tumor suppressor gene in neuroblastoma [
18]. In this study, we conducted a series of studies to examine whether
CHD5 is a tumor suppressor gene in human breast cancer. Whereas one mutation, a truncation mutation in the MDA-MB-231 breast cancer cell line, was found in
CHD5 among the samples examined,
CHD5 had frequent genomic deletion and promoter methylation in both breast cancer cell lines and primary tumors. Furthermore, promoter methylation correlated with transcriptional down-regulation of
CHD5 and demethylating treatment increased
CHD5 expression. CHD5 protein expression was also significantly reduced in primary breast tumors, and lack of CHD5 expression correlated with higher tumor grade, local recurrence, distant metastasis and worse patient survival. Functionally, CHD5 inhibited the proliferation and invasion of MDA-MB-231 and Hs 578T cells
in vitro and tumorigenesis of MDA-MB-231 cells
in vivo (Figures
1,
2,
3,
4,
5,
6,
7,
8). Taken together with previous findings that showed deletion of 1
p36 in breast tumors [
19,
20], our results suggest that
CHD5 is a tumor suppressor gene whose inactivation by mutation, chromosomal deletion or promoter methylation-mediated transcriptional down-regulation plays a role in the development and progression of breast cancer.
We noticed that in the T-47D breast cancer cell line, which expresses a higher level of
CHD5, RNAi-mediated knockdown of
CHD5 did not increase cell proliferation. Unexpectedly,
CHD5 knockdown decreased cell proliferation. While the reason for this unexpected finding remains to be determined, one possibility is that T-47D cells have acquired molecular alterations that have reversed the function of CHD5 from a proliferation suppressor to a proliferation enhancer. Such a functional reverse has been shown previously for some molecules. For example, TGF-β is a potent tumor suppressor in early stage tumorigenesis but becomes a promoter in late stage tumor progression [
21]; and deacetylation converts KLF5 from an anti-proliferative factor to a pro-proliferative in epithelial cells [
22].
A tumor-specific function-altering mutation is a well-established indicator for tumor suppressor genes. In the original study that identified 122 genes with somatic mutations in breast cancer through large-scale DNA sequencing, two of the 11 (18%) primary breast cancers had a heterozygous missense mutation in
CHD5 (133G > A, V45M; 1999A > G, R667G) [
12]. In our study, none of the 38 primary tumors had somatic mutations in
CHD5, while one of the 17 breast cancer cell lines examined had a frame shift mutation. Our results are consistent with another report in which no mutations were detected in 60 breast cancer samples while three mutations were identified in 123 ovarian cancer samples [
11], indicating that, although mutation of
CHD5 could occur in breast cancer, its frequency is rather low in primary tumors.
Among the SNPs detected in this study, some occurred frequently in cell lines but were not detected in primary tumors, including G529C (8/34 alleles or 24%) and T4715C (76%); others occurred frequently in the primary tumors but were not detected in any of the cell lines, including C2479T (30%) and G3436A (36%). These differences could be caused by differences in the ethnic background of patients whose primary tumors were used (Chinese) and patients from whom the cell lines were derived (non-Chinese). There is also a possibility that some of these SNPs contribute to successful establishment of breast cancer cell lines or even breast cancer progression, which remains to be determined.
Although somatic mutation of
CHD5 is rare, obvious down-regulation of
CHD5 mRNA and protein was frequent in breast cancer cell lines and primary tumors. Hemizygous deletion and promoter methylation were also common and could be responsible for transcriptional down-regulation of
CHD5 in breast cancer. In evaluating the role of promoter methylation in the down-regulation of
CHD5, we found that transcriptional down-regulation was significantly associated with promoter methylation and that demethylating treatment significantly increased
CHD5 expression, indicting a role of methylation in decreasing
CHD5 mRNA expression. These results are consistent with those from other types of tumors, such as neuroblastoma and ovarian cancer, in which frequent promoter methylation and reduced RNA expression of
CHD5 had also been detected [
7‐
10]. Therefore, methylation-mediated silencing, which is a common mechanism for the inactivation of tumor suppressor genes, occurs at
CHD5 in different types of human cancers including breast cancer.
Reduced
CHD5 mRNA expression was significantly associated with lymph node metastasis, while reduced CHD5 protein expression significantly correlated to recurrence, distant metastasis, progression-free survival and overall survival in breast cancer. Multivariate Cox regression analysis indicates that lack of CHD5 expression is an independent prognostic factor. These results are consistent with a previous report where allelic loss at 1
p was correlated with lymph node metastasis in breast cancer [
23] and that higher CHD5 expression was strongly associated with more favorable outcomes in neuroblastoma. Furthermore, the only breast cancer cell line harboring a truncated mutation at
CHD5, MDA-MB-231, is highly metastatic. Therefore, it is possible that inactivation of CHD5 contributes to metastatic progression of breast cancer. Supporting this hypothesis, expression of
CHD5 in the metastatic MDA-MB-231 and Hs 578T cells inhibited cellular invasiveness and down-regulated EMT markers.
Activation of the p19
Arf/p16
Ink4a locus by CHD5, which activates both p19
Arf/p53 and p16
Ink4a/Rb signal axes, has been established as a mechanism for how CHD5 executes its tumor suppressor function. However, the three breast cancer cell lines used in our functional studies, MDA-MB-231, Hs 578T and T-47D, harbor p53 mutations [
24], and no significant change in p53 expression was detected in MDA-MB-231 cells expressing ectopic
CHD5 or T-47D cells where
CHD5 was knocked down (Figure
8C, E). Rb expression was not changed in CHD5-expressing MDA-MB-231 cells either (data not shown). In addition, whereas the expression of p21, a target gene of p53 involved in cell cycle arrest, was increased in CHD5-expressing MDA-MB-231 cells (Figure
8C), its expression upon CHD5 knockdown in T-47D cells showed inconsistent changes, as p21 expression was decreased by one siRNA but not affected by another siRNA, although both siRNAs efficiently knocked down
CHD5 (Figure
8E). We thus speculate that CHD5 suppresses breast carcinogenesis by a p53- and Rb-independent mechanism. In fact, other mechanisms have been suggested in other types of tumors, as
CHD5 mutation or methylation is associated with KRAS or BRAF mutation in ovarian cancer [
11], and CHD5 expression correlates with MYCN amplification in neuroblastoma [
7,
25].
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
XW carried out most molecular, genetic and functional experiments. She also extracted DNA and RNA samples from some tissue specimens and cell lines, and drafted the manuscript. WL read tissue slides and scored CHD5 IHC staining with the guidance of LF. He also participated in animal experiments. XF conducted immunohistochemical staining of CHD5 in human tissues. DS conducted sequence data analysis and statistical analysis. LF participated in the mutation analysis of CHD5 in tumor samples and cell lines. ZZ and AL extracted DNA and RNA samples from some breast tissue specimens and cell lines. XS extracted DNA from some of the breast cancer cell lines and treated cell line DNA with sodium bisulfite for methylation analysis. LF is responsible for the collection and preparation of breast cancer tissues for analysis. ZZ and JTD designed the study, provided overall guidance and led the execution of all experiments. JTD is also responsible for the finalization of the manuscript. All authors read and approved the manuscript.