healthmedicinet

Study design

Biopsy specimens fresh frozen from different parts of the colon and rectum, together
with blood samples, were obtained from 55 consecutive LS mutation carriers who underwent
regular colonoscopy screening or colectomy at two Finnish hospitals during 10/2011–5/2013
(Table 1). This prospective series was used to study DNA methylation changes in normal colonic
mucosa with respect to aging and previously diagnosed cancer. As only a few individuals
developed colorectal lesions (mostly hyperplastic polyps) during the 1.5-year interval,
all archival tubular and villous adenomas and carcinomas previously diagnosed in the
same individuals were gathered (retrospective series in Table 1) and used to investigate DNA methylation changes occurring in the adenoma carcinoma
progression sequence. Colonic and rectal biopsies from 22 familial adenomatous polyposis
(FAP) mutation carriers who participated in colonoscopy screening (Table 1) were studied for comparison.

Table 1. Characteristics of sample series

MMR status and analysis of MMR gene promoter methylation as the “second hit”

To test if the predisposing MMR gene had undergone somatic inactivation of the remaining
wild-type allele, colorectal specimens were evaluated for MMR protein expression by
immunohistochemical (IHC) analysis. All adenomas with high-grade dysplasia and all
carcinomas from LS patients showed loss of MMR protein corresponding to the gene mutated
in the germline whereas only 68 %, 67 % and 50 % of adenomas with low-grade dysplasia
showed loss of expression in MLH1, MSH2, and MSH6 mutation carriers, respectively (Table 2). The difference (adenomas with low-grade dysplasia vs. adenomas with high-grade
dysplasia and adenomas with low-grade dysplasia vs. carcinomas) was statistically
significant for MLH1-associated tumors. Overall, the results suggest that silencing of the relevant MMR
protein expression is a relatively late event in LS tumorigenesis. MSI analysis with
the mononucleotide repeat markers BAT25 and BAT26 showed that all adenomas and carcinomas with absent MMR protein were microsatellite
unstable with one exception, an adenoma with low-grade dysplasia from a MSH2 mutation carrier. The low tumor cell percentage (20 %) in that particular sample
was likely to explain stable microsatellites. All eight low-grade dysplasia adenomas
retaining MMR protein expression were microsatellite stable.

Table 2. Proportion of decreased MMR protein expression in Lynch syndrome adenomas and carcinomas

Promoter methylation as a possible “second hit” was assessed in tumors lacking MMR
protein. Methylation of MLH1 was mostly detected in the distal promoter (region A 12]). This region was methylated in 30 %, 31 % and 60 % of adenomas with low-grade dysplasia,
adenomas with high-grade dysplasia and carcinomas, respectively (Fig. 1). On the contrast, methylation of the proximal promoter (region C), most commonly
associated with MLH1 protein loss, was only observed in one adenoma having high-grade
dysplasia. No methylation was detected in promoter regions of MSH2 or MSH6. Taken together, promoter methylation constituted a putative “second hit” in 1/56
(2 %) tumors with silenced MMR protein.

Fig. 1. Frequency of hypermethylated CIMP markers in the LS retrospective series. Numerical
values of percentages are given above each bar. Hypermethylation thresholds were calculated
according to stringency level I. Pairwise comparisons were calculated by Fisher’s
exact test (two-sided p values), and the p values were adjusted for multiple comparisons by Bonferroni correction

CIMP

To assess whether coordinated methylation of multiple CpG islands that are normally
unmethylated plays a role in colorectal tumorigenesis, methylation-specific multiplex
ligation-dependent probe amplification (MS-MLPA) was used to study eight genes firmly
associated with CIMP (CACNA1G, IGF2, NEUROG1, RUNX3, SOCS1, CDKN2A, MLH1, and CRABP1). The MS-MLPA CIMP probe mix contains 3–6 probes for each CIMP marker gene, and the
average methylation dosage ratios (Dm) obtained for each probe and type of specimen
are given in Additional file 1: Table S1. Hypermethylation in tumor tissues was evaluated relative to probe-specific
thresholds derived from normal mucosa. The hypermethylation thresholds for the retrospective
(FFPE) series are given in Additional file 2: Table S2 and those for the prospective (fresh frozen) series in the legend of Additional
file 3: Figure S1. A gene was considered hypermethylated when at least one fourth (25 %)
or more of probe target sites were methylated 13].

The prospective LS series revealed increased methylation in adenomas and carcinomas
vs. normal colonic mucosa for the CIMP markers IGF2, NEUROG1, and CRABP1 (p values mostly non-significant due to small numbers of specimens; Additional file
3: Figure S1A). Hyperplastic polyps, too, showed frequent hypermethylation with IGF2 and NEUROG1. LS vs. FAP-associated adenomas and matching normal mucosa showed comparable frequencies
of hypermethylation.

The analyses were extended to the retrospective LS series with higher number of tumors
available (Fig. 1). Adenomas with high-grade dysplasia and carcinomas showed the highest frequencies
of hypermethylation (defined with stringency level I of Additional file 2: Table S2). The frequencies of hypermethylation for IGF2 and NEUROG1 were significantly increased in all tumor types when compared to normal colon. Notably,
the difference was significant already in adenomas with low-grade dysplasia (89 %
vs. 21 %, p??0.001, for IGF2 and 56 % vs. 8 %, p??0.001, for NEUROG1). Only carcinomas showed significantly elevated hypermethylation frequencies for
MLH1 and CRABP1.

As no consensus exists on how to score CIMP and no single panel is superior to others
13], three different marker panels were considered. Colorectal specimens from the retrospective
series were divided into CIMP(+) or CIMP(?) categories by using the Ogino 5/8, Ogino
6/8, and Weisenberger 3/5 panels (“Methods” section). Here, stringency level II (Additional
file 2: Table S2) was used to calculate probe-specific thresholds to avoid hypermethylation
in normal mucosa. The frequency of CIMP(+) specimens increased from normal mucosa
to adenomas with low-grade dysplasia to adenomas with high-grade dysplasia to carcinomas
regardless of the marker system used (Fig. 2). The Weisenberger panel yielded somewhat higher CIMP(+) frequencies compared to
the Ogino panels. Accordingly, 15 %, 23 % and 50 % of the adenomas with low-grade
dysplasia, adenomas with high-grade dysplasia, and carcinomas, respectively, were
CIMP(+) when using the Weisenberger 3/5 panel, compared to 7 %, 23 % and 40 % with
the Ogino 5/8 criteria, and 4 %, 23 % and 25 % with the Ogino 6/8 criteria (Fig. 2). Formal statistical significance was reached in the normal mucosa vs. carcinomas
comparison according to Ogino 5/8 (p?=?0.006) and Weisenberger 3/5 (p??0.001) criteria and borderline significance in the adenomas with low-grade dysplasia
vs. carcinomas comparison according to Ogino 5/8 (p?=?0.054) criteria. CIMP(+) vs. CIMP(?) tumors were diagnosed at similar average ages
(48 vs. 49 years for adenomas and 49 vs. 48 years for carcinomas) excluding age as
a possible confounding factor in the analyses (see below).

Fig. 2. Frequency of CIMP(+) specimen calculated by three different criteria. Numerical values
of percentages are given above each bar. Hypermethylation thresholds were calculated
according to stringency level II. Pairwise comparisons were calculated by Fisher’s
exact test (two-sided p values), and the p values were adjusted for multiple comparisons by Bonferroni correction

Methylation analysis of candidate genes

A custom MS-MLPA kit was designed to study methylation of seven candidate genes previously
associated with early colon oncogenesis in an experimental mouse model (DKK1, SFRP1, SFRP2, SFRP5, CDH1, HOXD1, and SLC5A8 (Additional file 4: Table S3) 11]. The average Dm obtained for each probe and type of specimen are shown in Additional
file 1: Table S1. Probe-specific hypermethylation thresholds were determined as described
for CIMP markers (stringency level I), separately for the retrospective series (Additional
file 2: Table S2) and the prospective series (legend of Additional file 3: Figure S1).

The prospective LS series indicated significantly higher frequencies of hypermethylation
for SFRP1 (95 %, p?=?0.006) and SFRP2 (67 %, p?=?0.012) in carcinomas vs. normal colonic mucosa (Additional file 3: Figure S1B). Additionally, 50 % of hyperplastic polyps revealed hypermethylation
for SFRP1 relative to normal mucosa (p?=?0.030). Hypermethylation frequencies in colonic tissues were comparable in LS vs.
FAP.

In the retrospective LS series, hypermethylation frequencies for SFRP2 were significantly higher in all tumor types when compared to normal mucosa (Fig. 3). Importantly, this included adenomas with low-grade dysplasia already (50 % vs.
13 %, p?=?0.042). For SFRP1, significantly increased frequencies of hypermethylation were only observed in adenomas
with high-grade dysplasia and carcinomas (Fig. 3).

Fig. 3. Frequency of hypermethylated candidate genes in the LS retrospective series. Numerical
values of percentages are given above each bar. Hypermethylation thresholds were calculated
according to stringency level I. Pairwise comparisons were calculated by Fisher’s
exact test (two-sided p values), and the p values were adjusted for multiple comparisons by Bonferroni correction

Correlation of candidate gene methylation with mRNA expression in cancer cell lines

To evaluate functional significance of promoter methylation, MMR-deficient colorectal,
endometrial, and ovarian cancer cell lines (Additional file 5: Table S4) were treated with the demethylating agent 5-aza-CdR and the histone deacetylase
inhibitor TSA, followed by RNA profiling on microarrays. Consistent (1.8–7.2-fold)
treatment-induced upregulation of SFRP1 was seen in HCT15, HCT116, and HEC59 analogous to LS-associated CRC and endometrial
cancer. SFRP2 was significantly upregulated (1.9-fold) in HCT116. Upregulation was accompanied
by reduced promoter methylation by MS-MLPA.

When methylation (Dm) values were plotted against mRNA expression in all studied cancer
cell lines (treated and untreated) and the corresponding normal tissues, a significant
inverse correlation was observed between SFRP1 (r?=??0.688, p?=?0.001) and SFRP2 (r?=??0.657, p?=?0.002) mRNA expression and methylation (Additional file 6: Figure S2). Moreover, SFRP1 and SFRP2 expressions were significantly lower in the untreated cancer cell lines compared
to the corresponding normal controls (data not shown). Our data suggest that DNA methylation
plays a significant role in the expressional regulation of these genes.

Hypermethylation of CIMP and candidate genes vs. expressional status of MMR proteins

Eleven adenomas with low-grade dysplasia retained MMR protein expression (Table 2) suggesting that the second hit to inactivate the responsible MMR gene had not yet
occurred. DNA was available for eight such adenomas, and 3/8, 4/8, 6/8, and 2/8 revealed
hypermethylation of IGF2, NEUROG1, SFRP1, and SFRP2, respectively. The hypermethylation frequencies were essentially comparable to those
in low-grade dysplasia adenomas with silenced MMR protein (15/18, 4/18, 11/18, and
10/18 for the respective loci). The results suggest that hypermethylation of the investigated
genes can occur prior to somatic inactivation of the predisposing MMR gene.

Effect of age at biopsy on normal colonic tissue methylation

Since DNA methylation tends to increase with age 14], the level of methylation in each prospectively collected colorectal mucosa specimen
was evaluated against the chronological age of the individual at the time of biopsy.
A moderate-to-strong positive correlation was detected between age at biopsy and normal
colonic mucosa Dm values corresponding to IGF2 probes I (r?=?0.694, p??0.0001), II (r?=?0.726, p??0.0001), and III (r?=?0.742, p??0.0001) and NEUROG1 probes I (r?=?0.566, p??0.0001), III (r?=?0.703, p??0.0001), and IV (r?=?0.655, p??0.0001) (Additional file 7: Figure 3SA–B). Additionally, moderate correlation between age at biopsy and SFRP1 (r?=?0.554, p??0.0001), SFRP2 (r?=?0.550, p??0.0001), and SLC5A8 (r?= 0.554, p??0.0001) methylation was observed in normal colonic mucosa (Additional file 7: Figure S3C). This indicates that aging itself increases methylation of the CIMP
markers in the histologically normal mucosa. No age-related correlation was observed
for MLH1 region C methylation.

Analysis of field defects in histologically normal colonic mucosa

To investigate if aberrant DNA methylation might form carcinogenic “fields” in the
histologically normal mucosa, colonic mucosa biopsies of the prospective LS series
were evaluated for hypermethylation of CIMP markers and candidate genes. The individuals
were divided into four groups depending on age at biopsy and absence vs. presence
of (previous or concurrent) CRC (groups 1–4). The first two groups included mutation
carriers 50 years old and below and consisted of 22 individuals without CRC (group
1) and 6 individuals with CRC (group 2). The remaining two groups included mutation
carriers above 50 years and consisted of 17 individuals without CRC (group 3) and
10 individuals with CRC (group 4). The interval between CRC diagnosis and time of
biopsy was 5.1 years (range 0–11.5) in group 2 and 5.5 years (range 0–12.4) in group
4. Average age at biopsy was comparable in group 1 (35, range 26–50) vs. group 2 (43,
range 39–48) and in group 3 (61, range 51–75) vs. group 4 (63, range 51–74), excluding
the age effect as a possible confounder in the respective comparisons.

When the effect of CRC on CIMP marker methylation was examined within the age groups
(?50 and above 50), no significant differences were observed (Additional file 8: Figure S4). However, age effect was evident from comparisons of the “under 50, no
CRC” with the “over 50, no CRC” groups and “under 50, CRC” with the “over 50, CRC”
groups, which revealed significant increases in methylation for several IGF2 and NEUROG1 probes (Additional file 8: Figure S4).

The candidate genes, too, showed an age effect, but despite it, significant increases
in methylation were observed for SFRP1 (p??0.0001) and SLC5A8 (p?=?0.007) in the “over 50 group,” when stratified by the presence vs. absence of previously
diagnosed CRC (Fig. 4). Moreover, our results from duplicate MS-MLPA assays (“Methods” section) as well
as from dilution experiments (Additional file 9: Figure S5) showed that the observed changes in methylation were unlikely to be explained
by technical variation or other similar reasons. Thus, our finding may indicate a
potential field defect in normal mucosa.

Fig. 4. Effect of aging and previously diagnosed CRC on SFRP1 and SLC5A8 normal colonic mucosa Dm values. Study groups: patients ?50 years without (?=?22) and with (?=?6) a CRC diagnosis, and patients 50 years without (?=?17) and with (?=?10) a CRC diagnosis. Statistical testing was performed with one-way ANOVA and Tukey’s
post hoc test or Kruskal-Wallis one-way ANOVA