Introduction

A disintegrin and metalloprotease with trombospondin motifs (ADAMTS) is a family of proteins that share a complex structural design characterized by the presence of a metalloproteinase domain linked to a variety of specialized ancillary domains including a series of thrombospondin-1 repeats [1–5]. Certain members of the ADAMTS family (ADAMTS1, ADAMTS4, ADAMTS5, ADAMTS8, ADAMTS9, and ADAMTS15) are called ‘aggrecanases’ as they can proteolytically process aggrecan within the interglobular domain by separating its globular G1 and G2 domains at a specific Glu373–Ala374 bond [1, 6, 7] or at one or more sites within the more C-terminal glycosaminoglycan-bearing region [8]. Cytokines such as IL-1β have been shown to markedly upregulate ADAMTS18 expression in OUMS-27 chondrosarcoma cells and in human chondrocytes [9]. Several reports have shown that the tumor suppressor activities of ADAMTS1 and ADAMTS8 derive from the angio-inhibitory capacities of their thrombospondin-1 domains [10, 11], whereas ADAMTS12 inhibits tumor growth through its ability to modulate the Ras-dependent extracellular signal-regulated kinase (ERK) signaling pathway [12].

Promoter hyper-methylation is one of the hallmarks of carcinogenesis associated with transcriptional silencing of genes encoding for diverse cellular pathways, and is considered to be an important epigenetic mechanism implicated in the regulation of normal gene expression. Such changes often affect 5′ regulatory CpG genomic regions and can be associated with aberrant expression of certain genes in cancer [13]. Genetic and epigenetic analysis supported the proposal that ADAMTS family members may act as anti-tumor proteases. Moreover, ADAMTS1, ADAMTS9, and ADAMTS15 have been identified as epigenetically silenced genes in several carcinomas [14–17]. Recently, ADAMTS18 has been characterized a putative tumor suppressive gene and is found to be epigenetically silenced in esophageal, nasopharyngeal, and some other carcinomas [18].

High-resolution melting (HRM) analysis is a novel tool for analysis of promoter methylation [19]. The new approach is based on the “melting” properties of DNA in solution [20], and was originally developed for SNP genotyping [21]. The principle of this method is that bisulfite-treated DNA templates with different contents of methyl-cytosine can be resolved by melting analysis due to differences in melting temperatures [22]. HRM relies on the precise monitoring of the change of fluorescence as a DNA duplex melts. This technique requires the use of standard PCR reagents and double-stranded DNA-binding dyes that can be used at saturating concentrations without inhibiting PCR amplification [21]. The melting analysis does not allow detailed information about the methylation of single cytosines within the sequence of interest, but can distinguish fully and partially methylated samples. The semi-quantitative measurement of methylation is important because low levels of methylation may not be biologically important [23, 24]. Also, quantification of promoter methylation may enable early detection of cancer and early metastatic spread [25].

In our study, we used HRM analysis to detect the methylation levels of ADAMTS18 gene in 100 gastric cancers, 100 colorectal cancers, 70 pancreatic cancers, and equal number of adjacent normal tissues. Methylation levels in all three types of cancers were significantly higher than that in normal tissues. This is the first study to estimate the prevalence of ADAMTS18 methylation based on large amount of tumor samples, showing that epigenetic regulation of ADAMTS18 was associated with carcinogenesis.

Materials and methods

Controls and patient samples

CpGenome Universal Methylated and unmethylated DNA (Chemicon, Millipore Billerica, MA, USA) were used as 100 and 0% methylated control DNA, respectively. Methylation standards were constructed by diluting 100% methylated bisulfite-modified control DNA in a pool of unmethylated bisulfite-modified DNA at ratios of 10% and 1%. These standards were included in each experimental run.

Surgically resected tumor tissues and adjacent normal tissues were collected from 100 primary gastric cancer patients, 100 primary colorectal cancer patients, and 70 primary pancreatic cancer patients. Staging was assessed after pathological examination of formalin-fixed specimens based on the 2002 TNM classification (6th edition of the staging criteria of the UICC and AJCC). The study was approved by the ethical committee of the Shenzhen Hospital, Peking University. The individuals gave their written informed consent. The investigations were conducted according to the Declaration of Helsinki principles.

Extraction of genomic DNA and sodium bisulfite modification

Genomic DNA was isolated from the tissues using the genomic DNA extraction kit (Innogent, Shenzhen, China) according to the manufacturer’s instruction. One microgram of genomic DNA was subjected to bisulfite conversion with the EZ DNA methylation kit (Zymo research, USA). The eluted DNA (40 μl volume) was used for the HRM analysis.

HRM analysis

PCR amplification and HRM were performed on the ABI7500 (Applied Biosystems) as adapted from the published protocol [19]. The primers were designed as outlined [19]. The sequences of the primers for ADAMTS18 (Chromosome 16: 76025677–76027811) are as follows: forward—GCGTTCGTATTTGATTATGG and reverse–TCGACGACACCTACCTTAA (134 bp). PCR was performed in a 20 μl volume containing: 1× buffer, 2 U Hotstart Taq DNA polymerase (Takara), 250 nM of each primer, 2.5 mM SYTO-9, and 10 ng bisulfite-treated DNA template, with 3 mM final MgCl2. Each reaction was performed in triplicate. The cycling conditions were as follows: 1 cycle of 95°C for 10 min, 60 cycles of 95°C for 10 s, 61°C for 10 s, and 72°C for 10 s; followed by an HRM step of 95°C for 1 min, 40°C for 1 min, 65°C for 15 s, and continuous acquisition to 95°C at 1 acquisition per 0.3°C. A standard curve with known methylation ratios was included in each assay and was used to deduce the methylation ratio of each tumor and normal sample. HRM data were analyzed using the High-Resolution Melting Software (Applied Biosystems). Output plots are in the form of normalized melting curves and difference plots. Statistical analysis was performed using chi-square test through the SPSS 10.0 software for Windows. P values less than 0.05 were considered statistically significant.

Bisulfite genomic sequencing analysis

Bisulfite-treated genomic DNA was amplified using the following primers: forward—GTTGGGGGAGTTAGTTGG and reverse—ACGACACCTACCTTAACCAC. Amplified PCR product was purified and cloned into pCR4-TOPO vector (Invitrogen, Carlsbad, CA). DNA sequencing was performed on at least five individual clones.

Quantitative RT-PCR

Total RNA was isolated from tissues by using AxyPrepTM Blood Total RNA MiniPrep Kit (Axygen) according to the manufacturer’s instruction. First strand cDNA was synthesized with RevertAidTM First Stand cDNA Synthesis Kit (Fermentas). Quantitative PCR was performed through BioRad Chromo4 real-time PCR system. The relative abundance of ADAMTS18 mRNA level was calculated by using the comparative C(T) method [26, 27] with GAPDH as the internal control. Logarithmic transformation (Log2 [1 + X]) was used to transform primary data to normal distribution. Data from three independent experiments were analyzed by student t test and P < 0.05 was considered statistically significant. The primers for ADAMTS18 are: forward: 5′CGCCAGCGGATTAAATGATGA and reverse: 5′CAAATGCTGAAAATCGGTAGTGC. The primers for GAPDH are: forward: 5′CAGCCTCAAGATCATCAGCA and reverse: 5′TGTGGTCATGAGTCCTTCCA.

Results

The sensitivity of HRM analysis for ADAMTS18

The sensitivity of the ADAMTS18 HRM analysis was tested by using dilutions of fully methylated DNA into unmethylated DNA. The HRM standard melting curve was derived from five samples with the following ratios of methylated DNA: 0, 1, 10, and 100% methylation. The inclusion of CpGs in the primer sequence makes it possible to direct the PCR bias toward the methylated templates by manipulating the annealing temperature of PCR amplification. At the annealing temperature of 61°C, methylation level as low as 1% can be easily detected (Fig. 1). The normalized melting profiles of the PCR product amplified from the same template were consistent between replicates and between different runs, and the shapes of normalized melting profiles were amplification independent as samples with different starting amount of template displayed very similar profiles (data not shown). To confirm the reliability of HRM, bisulfite genomic sequencing (BGS) was used in our study. Bisulfite-modified genomic DNA from representative samples (0 methylation, 0–1 methylation, 1–10 methylation and, 10–100% methylation) was amplified and cloned. Ten clones were randomly selected and subjected to sequencing. We found that data from HRM and BGS were consistent (Supplementary Fig. 1B).

Fig. 1
figure 1

Normalized HRM standard curves and difference plot of ADAMTS18 gene. Templates with different ratios (as indicated) of methylated DNA were amplified at the annealing temperature of 61°C, and subjected to HRM analysis. a Normalized melting curves, b difference plot

Full size image

Methylation levels of ADAMTS18 in tumor samples and normal tissues

Table 1 shows the methylation levels of ADAMTS18 in 100 gastric, 100 colorectal and 70 pancreatic cancer samples, and equal number of adjacent normal tissues. In all three types of cancers, methylation levels of ADAMTS18 in cancer samples were significantly higher than that in normal tissues.

Table 1 Methylation levels of ADAMTS18 in cancer samples and normal tissues
Full size table

Correlation between methylation and expression of ADAMTS18

In order to confirm that methylation of ADAMTS18 indeed correlate with the downregulation of its gene expression, we then used real-time PCR to detect the mRNA levels of ADAMTS18 in the cancer samples. Since low levels of methylation may not be biologically important [23, 24], the cancer samples were sub classified into three groups: <1% methylation, 1–10% methylation, and 10–100% methylation. The numbers of samples in the three groups are 64, 17, and 19 (gastric cancer), 51, 33, and 16 (colorectal cancer), and 43, 18, and 9 (pancreatic cancer), respectively (Table 2). As expected, in all three types of cancers, expression levels of ADAMTS18 were significantly lower in the groups of 1–10% and 10–100% methylation (Fig. 2), which is consistent with previous reports [13].

Table 2 Methylation levels of ADAMTS18 in different stages of cancers
Full size table
Fig. 2
figure 2

ADAMTS18 expression in different types of cancer samples. Total RNA from cancer samples were extracted and subjected to real-time PCR analysis. The relative abundance of ADAMTS18 mRNA level was calculated by using the comparative C(T) method after logarithmic transformation. Asterisk P < 0.05

Full size image

Correlation between methylation of ADAMTS18 and staging of cancers

According to the 6th edition of the staging criteria of the UICC and AJCC, the cancer samples used in our study were assessed after pathological examination based on the 2002 TNM classification. The number of samples in each stage is summarized in Table 2. There was no significant association between ADAMTS18 methylation status and TNM staging in all three types of cancers.

Discussion

Several methods have been developed for the analysis of methylation, each with their characteristic strengths and weaknesses. The most widely used method is methylation-specific PCR (MSP) that uses primers specific for methylated, bisulfite-modified DNA [28]. MSP is very sensitive but is not quantitative, thus can lead to the classification of gene methylation when only a small number of cells are positive. Genomic sequencing can be considered the gold standard [29, 30]. It provides the most detailed information but is relatively insensitive, and its expensive cost makes it generally unsuitable for screening. Pyrosequencing was recently introduced with higher sensitive, but is dependent on the availability of the proprietary instrumentation [31]. HRM analysis becomes a novel tool for analysis of promoter methylation [19]. The applications of sequencing and HRM in methylation studies utilize methylation-independent PCR (MIP) where the primers are designed to amplify the bisulfite-modified sequence regardless of its methylation status. However, MIP primers do not always lead to the proportional amplification of methylated and unmethylated sequences [32, 33]. So, inclusion of some CpGs seems necessary in the primer sequence to avoid the underestimate of the degree of methylation [33]. At lower annealing temperatures, the primers bind both methylated and unmethylated templates and PCR bias will favor the amplification of unmethylated sequences. At higher annealing temperatures, primer binding will favor methylated sequences. Thus the optimal annealing temperature is important for the effective amplification of templates independent of methylation status. HRM has several advantages over the other methods. It has high throughput and relatively cheap. More importantly, it can be used to estimate the proportion or extent of methylation when run with standards. This is especially useful when assessing clinical cancer samples for predictive markers such as ADAMTS18 where discrimination between different levels of methylation may have diagnostic and prognostic value. In our study, we have shown that HRM is applicable for the very sensitive detection of ADAMTS18 methylation in an unmethylated background. With HRM, we were able to detect the methylation level of ADAMTS18 as low as 1%.

ADAMTS18 has been characterized as a putative tumor suppressive gene, and is found to be epigenetically silenced in esophageal, nasopharyngeal, and some other carcinomas [18]. But only limited number of primary tumor samples was recruited in previous studies. This makes further evaluation necessary to better estimate the prevalence of ADAMTS18. Our study for the first time investigated ADAMTS18 methylation in large amount of primary cancer samples. The high prevalence of hypermethylation of ADAMTS18 in the gastric, colorectal, and pancreatic cancers (36, 49 and 39%, respectively) suggests that transcriptional silencing of ADAMTS18 by methylation is common and may be involved in the pathogenesis of many types of cancers. On the other hand, methylation levels among different cancer types are not consistent. In our cases, colorectal cancers have the highest percentage of methylation. This indicates that contribution of ADAMTS18 methylation to carcinogenesis relies on the unique property of different cancer types. Furthermore, a certain incidence of ADAMTS18 methylation was also detected in adjacent normal tissues, although the methylation levels in these normal tissues remain lower than 10% (Table 1). Aberrant hypermethylation could be caused by various factors in normal tissues such as diets or aging [34] and this might result in the low methylation frequencies in normal tissues. Another possibility is the contamination of trace amount of tumor tissues during the surgical resection process, since our normal tissues are resected from adjacent area of the same patient. In the future, application of microdissection would be very helpful to distinguish tumors from normal tissues.

A significant correlation between ADAMTS18 methylation and loss of expression of ADAMTS18 was observed in all three types of cancers by real-time PCR. This is not unexpected and is consistent with many previous studies. We also analyzed the staging of the cancers in relation to the methylation of ADAMTS18. To our surprise, no association was found between them. This can be explained by the fact that ADAMTS18 methylation is supposed to be the cause, rather than the outcome, of carcinogenesis. Thus, even in the stage I of cancers, ADAMTS18 gene is already in methylated state, as observed in our study. This suggests that ADAMTS18 methylation cannot be used as an auxiliary criterion for the staging of cancers, but rather, can be used as a marker for the diagnosis of early stage cancers.