Thrombospondin-4 is a putative tumour-suppressor gene in colorectal cancer that exhibits age-related methylation

Quantitative expression (qRTPCR) analysis was performed on 55 matched normal and tumour samples. All tumour samples collected in a fresh state were macroscopically dissected to remove contaminating normal mucosa. Quantitative methylation (qMSP) analysis was also conducted on this patient cohort as well as biopsies of normal mucosa taken during clinically indicated colonoscopy in a series of 99 patients. Biopsies were taken from each of 4 locations within the colon (caecum, transverse colon, sigmoid colon and rectum). All polyps were removed and submitted for histology. All patients gave written, informed consent, and the study was approved by the RBWH and QIMR Human Research Ethics Committees. DNA was extracted from tumour samples using a salt precipitation method as previously described [20], and from biopsy specimens using a DNA column method (Qiagen, Hilden, Germany). CIMP status was determined in tumours by normalising the methylation levels of each of the 5 Laird CIMP markers (CACNA1G, IGF2, NEUROG1, RUNX3, SOCS1) [19] by the methylation levels observed for the highly methylated ALU gene to generate the PMR, or Percentage of Methylated Reference. Tumours were classified as CIMP-negative (CIMP-neg) if 0/5 markers were methylated, CIMP-low (CIMP-L) if 1/5 or 2/5 markers were methylated, and CIMP-high (CIMP-H) if 3, 4 or 5/5 markers were methylated. The study cohort consisted of 14 CIMP-High, 11 CIMP-Low and 30 CIMP-Negative tumours.

Total RNA was extracted using an RNeasy MidiPrep kit (Qiagen, Hilden, Germany) and cDNA was synthesised using random hexamers and SUPERSCRIPT III (Invitrogen, Carlsbad, California). A Taqman^® Gene Expression Assay (part# 4331182; assay ID Hs00170261_m1 Applied Biosystems, Carlsbad CA, USA) was performed on cDNA generating a 96 bp THBS4 product. Gene expression was normalised to β-actin (ACTB) expression using Taqman^® Gene Expression Assay (part# 4331182; assay ID Hs99999903_m1 Applied Biosystems, Carlsbad CA, USA) generating a 171 bp product. The qPCR was performed in duplicate on a RotorGene3000 (Qiagen) using Absolute QPCR Mix (AB1133A; Integrated Sciences, NSW, Australia) and cycling of 40 cycles at 60°C annealing. The mathematical model described by Pffafl was used to determine the expression of THBS4 relative to the housekeeping gene ACTB [21].

THBS4 immunohistochemistry was performed on a subset of 40 patients. Briefly, fixed tumours were embedded in paraffin blocks and 0.2 μM sections were cut and mounted onto Superfrost Plus slides. They were deparaffinized in xylol and rehydrated by gradient alcohol before a 15 minute incubation with 0.5% hydrogen peroxide in phosphate buffered saline (PBS) to quench activity of endogenous peroxidases. After washes in PBS, the first containing 0.05% Triton X-100, sections were incubated in 10% goat serum and 0.01% acetylated BSA for 60 mins, then incubated overnight at 4°C in mouse anti-human THBS4 monoclonal antibody (MAB2390, R&D Systems, Minneapolis, MN) at 1:1500 in PBS/5% goat serum/1%BSA. Sections were again washed and incubated with rabbit anti-mouse Envision (Dako, Denmark) for 30 mins. The chromogenic substrate was 3,3-diaminobenzidine and sections were counterstained with Meyers' prior to dehydration and mounting of slides in Depex. Negative (no antibody) controls were included in all runs. Once the normal staining pattern for THBS4 was established, positive control normal tissue sections were also included in all runs for consistency.

A mammalian expression construct (pcDNA3.1::THBS4) was designed to examine the effect of over-expressing THBS4 in CRC cell lines. Briefly, 1 μg of either pcDNA3.1(+) (Invitrogen, Carlsbad, CA, USA) or pcDNA3.1::THBS4 was transfected into cells using FuGENE6 (Roche Applied Science, Indianapolis, IN) at a ratio of 3:1. Three null-expressing cell lines (Lim1215, SW48 and HT29), one low-expressing cell line (SW480) and four high-expressing cell lines (DLD1, LS174T, HCT116 and RKO) were transfected in triplicate at an initial density of approximately 50%. Cells were allowed to recover for 48 hrs before application of selective media at a final concentration of 700 ng/μL G418 (GibcoBRL Life Technologies, Invitrogen, CA) for 10-14 days. At this time there were no surviving untransfected control cells, and transfected colonies were stained with 0.25% crystal violet/80% methanol. The HT29 transfection was repeated and individual colonies were expanded to confirm THBS4 transcript re-expression.

In order to determine the methylation status of the THBS4 promoter, qMSP was performed for 55 paired normal and tumour samples, and 13 cell lines. Briefly, 1.5 μg of DNA was modified with sodium bisulfite using the EpiTect Bisulfite kit (Qiagen, Hilden, Germany), diluted 1/8 and subjected to real time PCR in duplicate on a Rotorgene6000 (Corbett Research, QIAGEN, Germany) using Absolute QPCR Mix (AB1133A; Integrated Sciences, NSW, Australia) and cycling of 40 cycles at 60°C annealing (400 nM of F 5'-CGTTGTCGCGGAGTTTAGTA-3'; 600 nM of R 5'-ACGACGACGACGTTAACC-3'; 50 nM of Probe 5'-[DFAM]-ACCTCGATCGACGCCCGAAC[DBHQ1]). The level of methylation was determined by normalising the THBS4 methylation levels by the methylation levels observed for the highly methylated ALU gene to generate the PMR, or Percentage of Methylated Reference (400 nM of ALU-F 5'-GGTTAGGTATAGTGGTTTATATTTGTAATTT-3'; 400 nM of ALU-R 5'-ATTAACTAAACTAATCTTAAACTCCTAACCT-3'; 100 nM of ALU-Probe [6FAM]-CCTACCTTAACCTCCC-[MGBNFQ]; cycling as described for THBS4 qMSP above). Analysis was performed using MethyLight [22].

The Wilcoxon signed rank test was used to compare THBS4 expression as well as methylation between the normal mucosal and tumour pairs. The Kruskal Wallis test was used to test for a relationship between tumour THBS4 methylation and CIMP status as well as the difference in normal mucosal THBS4 methylation in patients with and without colorectal pathology. A non-parametric trend test was applied to confirm an ordinal gradient across multiple groups. THBS4 methylation within the normal colonoscopic biopsy samples was calculated at each of the four sites sampled, as well as analysed according to an average proximal (mean of cecum and transverse colon), distal (mean of sigmoid colon and rectum) and pancolorectal (mean of all four sites) result. Correlations were analyzed by the Spearman's rank (ρ) coefficient. Distal and proximal levels of methylation were analyzed by a paired t-test. All tests were performed using Stata Statistical Software, version 10 [StataCorp].

THBS4 expression normalised to ACTB was generally quite low in both peritumoural normal mucosa and in tumours (Figure 1). There was significantly higher expression of THBS4 in the normal mucosa compared to the tumour. Median THBS4 expression was 0.04 vs. 0.01 respectively (p = 0.0035).

Bild vergrößern

Figures 2A,B & 2C show the examples of THBS4 immunohistochemical staining. In normal tissue, (Figure 2A and the right side of 2B) THBS4 protein was absent from the majority of normal epithelial cells, but was occasionally expressed in a vesicular pattern. Low levels were seen in the cytoplasm toward the luminal surface. THBS4 expression tended to be even lower in tumours compared with normal mucosa, with the majority of tumours having no detectable THBS4 expression (left side of Figure 2B, Figure 2C). In general, THBS4 protein was most likely to be seen at the luminal surface of normal crypts or tumour glands. However, there was some heterogeneity within the tumour sections and Figures 2D,E and 2F show some interesting variations to the general protein localisation in tumour sections. In Figures 2D and 2E, THBS4 protein seemed to be significantly up-regulated and actively secreted or packaged in vesicular structures (Figure 2F). These expression patterns suggest that THBS4 protein may be secreted or packaged and may give clues as to its function, although further functional studies are now required to test this hypothesis. If THBS4 protein is being packaged and secreted, it is possible that this may be one reason such low levels are detected in normal mucosa. The low frequency of detectable staining in tumour samples did not allow meaningful comparison with other molecular analyses. Immunohistochemical staining was primarily used to provide novel insight into the cellular distribution of THBS4 expression patterns.

Bild vergrößern

Colony formation was significantly repressed in all eight cell lines examined following over-expression of THBS4. We observed that both the number of colonies and their size were greatly reduced upon forced over-expression of THBS4. Figure 3 quantifies the THBS4 repression of colony formation in all cell lines tested. The reduction in colony formation ranged from 34% to 89% compared with the vector-only controls. Forced over-expression of THBS4 in multiple colorectal cancer cell lines, regardless of their basal THBS4 expression, consistently reduced colony forming ability. This indicates that high levels of this protein is associated with reduced tumour cell growth. To confirm that THBS4 transcript was introduced by transfection with this construct, HT29 was transfected and 14 individual clones were expanded. Of these, 8 showed high levels of THBS4 transcript whilst the remaining 6 had negligible expression levels comparable with untransfected parental HT29 cells.

Bild vergrößern

Overall, colorectal cancers had significantly higher THBS4 methylation than the matched normals (Figure 4). The median PMR in normals was 0.06 vs. 2.6 in tumours (P < 0.0001). Tumours were classified into CIMP subgroups using the Laird marker panel [19]. When stratified by CIMP status, there was no difference in the methylation of normal mucosa between the different classes of tumours. However, THBS4 methylation in tumours was positively correlated with CIMP (Figure 5). Tumour PMRs showed a progressive increase with increasing level of CIMP, with the median PMR for CIMP-negative tumours being 0.69, 3.4 for CIMP-L tumours, and 5.8 for CIMP-H tumours (p = 0.033).

Bild vergrößern

Bild vergrößern

Using a PMR of 10 as a cut off [19], tumours were separated into THBS4 methylation positive (PMR ≥10) and THBS4 methylation negative (PMR < 10). THBS4 promoter hypermethylation (PMR ≥10) was observed in 4/30 (13.3%) CIMP-NEG tumours, 4/11 (36.4%) CIMP-L tumours, and 5/14 (35.7%) CIMP-H tumours. Thus, while the average tumour PMR correlates well with an increasing number of markers positive for CIMP, there were an equal proportion of CIMP-L and CIMP-H tumours that showed high THBS4 methylation.

In CRC cell lines, there was no significant correlation between THBS4 transcript expression when compared with THBS4 PMR (ρ = 0.42, p = 0.20) on a continuous scale (Table 1). There was also no difference in expression in the colorectal cancer cell lines tested when the PMR cut-off of ≥10 was used (p = 0.35, data not shown). However, of the 11 cell lines tested, only 3 had a PMR of < 10. Tumours were also analysed for THBS4 expression level. Tumours with THBS4 PMR ≥10 appeared to have lower THBS4 expression, but this difference was not significant, according to the Wilcoxon Rank-sum (Mann-Whitney) analysis (p = 0.26, data not shown). Thus, while there was a trend towards lower THBS4 expression in tumours with a PMR greater than or equal to 10 having decreased expression, this difference was not significant. When tumour expression levels were normalised to matched normal mucosa expression levels, 44 cancers showed reduced expression greater than 1.5 fold (average 186.5 fold down-regulation) whilst 15 cancers showed greater than 1.5 fold up-regulation (average 26.5 fold) and seven tumours did not vary in expression levels by more than 1.5 fold.

THBS4 methylation was detected within the normal mucosa in all 99 of the colonoscopy patients with similar levels of methylation within the proximal and distal colorectum (median PMR 0.66 vs. 0.65, respectively, p = 0.66). Additionally, there were no differences in THBS4 methylation between the sexes, either at individual sites or in terms of pancolorectal THBS4 methylation. There was however, a strong and direct correlation between patient age and pancolorectal THBS4 methylation (ρ = 0.50, p < 0.0001, Figure 6).

Bild vergrößern

THBS4 methylation was lower in the background mucosa of patients with colorectal adenomas compared to those with normal colonoscopies. Inclusion of the peritumoural normal mucosal samples and stratification by age, revealed a consistent and inverse association between adenomatous pathology and THBS4 methylation (Figure 7). Importantly, THBS4 methylation within the normal colorectal field showed an impressive and inverse biological gradient with the presence of colorectal pathology, with the lowest levels of THBS4 methylation within the mucosal field being associated with the most advanced pathology (Figure 7). This suggested that THBS4 methylation in the normal mucosa was not directly implicated in promoting colorectal neoplasia and could in fact be directly or indirectly protective. This is supported by our DNA methylation data in other "type A" genes [23].

Bild vergrößern