Elevated mRNA Levels of DNA Methyltransferase-1as an Independent Prognostic Factor in PrimaryNonsmall Cell Lung Cancer

Hojoong Kim, MD, PhD1

Young Mi Kwon, MS1

Jin Seuk Kim, MS2

Joungho Han, MD, PhD3

Young Mog Shim, MD, PhD4

Joobae Park, MD, PhD2,5

Duk-Hwan Kim, MD, PhD2,5

1 Division of Pulmonary and Critical Care Medi-cine, Samsung Medical Center, SungkyunkwanUniversity School of Medicine, Seoul, Korea.

2 Center for Genome Research, Samsung Biome-dical Research Institute, Sungkyunkwan Univer-sity School of Medicine, Seoul, Korea.

3 Department of Pathology, Samsung MedicalCenter, Sungkyunkwan University School of Medi-cine, Seoul, Korea.

4 Department of Thoracic Surgery, SamsungMedical Center, Sungkyunkwan University Schoolof Medicine, Seoul, Korea.

5 Department of Molecular Cell Biology, Sung-kyunkwan University School of Medicine, Suwon,Korea.

BACKGROUND. Despite many reports about the involvement of DNA methyltrans-

ferases (DNMTs) in human cancers, including nonsmall cell lung cancer

(NSCLC), the clinicopathologic significance of DNMTs in primary NSCLC remains

to be elucidated.

METHODS. The relation between the mRNA levels of DNMTs (1 and 3b) and the

promoter methylation of the p16, RARb2, H-cadherin, GSTP1, RIZ, and FHIT

genes and the clinicopathologic features in 102 fresh-frozen tissues and paraffin

blocks were retrospectively studied. The mRNA levels of the DNMTs were as-

sessed via semiquantitative reverse-transcription polymerase chain reaction (RT-

PCR), and the methylation status of the CpG islands were determined by methyl-

ation-specific PCR.

RESULTS. The mRNA levels of DNMT1 and DNMT3b were elevated in 53% and

58% of 102 NSCLCs, respectively. Hypermethylation of p16, RARb2, H-cadherin,

GSTP1, RIZ, and FHIT occurred in 37%, 38%, 34%, 18%, 9%, and 31% of patients,

respectively. Univariate analysis showed that elevated DNMT mRNA levels were

not significantly associated with the hypermethylation of 6 genes. However, the

elevated mRNA levels of DNMT1 were determined to be significantly associated

with the hypermethylation of the p16 promoter (odds ratio [OR] ¼ 2.70, 95% con-

fidence interval [95% CI], 1.02–7.15; P ¼ .02), after controlling for age, gender,

pack-years smoked, histology, and pathologic stage. The hazard of failure in cases

with elevated mRNA levels of DNMT1 was 3.51 (95% CI, 1.18–12.76; P ¼ .02)

times higher than that in those without. The elevated mRNA levels of DNMT3b

were not ultimately associated with patient prognosis.

CONCLUSIONS. Elevated mRNA expression of DNMT1 may be an independent

prognostic factor in NSCLC and CpG island hypermethylation in NSCLC may be

maintained by a complex interaction of several factors rather than by a simple

transcriptional up-regulation of DNMT1. Cancer 2006;107:1042–9.

� 2006 American Cancer Society.

KEYWORDS: DNA methyltransferase, hypermethylation, nonsmall cell lung cancer,prognosis, tumor suppressor genes.

D NA methylation is known to be intimately involved in a host of

regulatory mechanisms, including embryonic development, im-

printing, the inactivation of the X-chromosome, and carcinogenesis.

Moreover, the aberrant methylation of CpG islands is known to exert

an important regulatory effect on gene expression, and has also

been determined to result in the transcriptional silencing of genes.1

The aberrant methylation of the normally unmethylated CpG island

located within the promoter region of a gene has been observed in

a variety of human cancers, including lung cancer.2 The methylation

Address for reprints: Duk-Hwan Kim, MD, PhD,Center for Genome Research, Samsung BiomedicalResearch Institute, Room B155, #50 Ilwon-dong,Kangnam-Ku, Seoul, Korea, 135-710; Fax: (02)3410-3649; E-mail: dukhwan.kim@samsung.com

Supported by the Seoul Research and BusinessDevelopment Program (grant 10582) and theScience Research Center/Engineering ResearchCenter (SRC/ERC) program of the Ministry ofScience & Technology/Korea Science and Engi-neering Foundation (MOST/KOSEF; grant R11-2005-017).

We thank Eunkyung Kim for assistance with datacollection and management and Hoon Suh forsample collection.

Received December 5, 2005; revision receivedApril 2, 2006; accepted April 21, 2006.

ª 2006 American Cancer SocietyDOI 10.1002/cncr.22087Published online 3 August 2006 in Wiley InterScience (www.interscience.wiley.com).

1042

status of more than 40 genes has been evaluated in

cases of lung cancer. Among these, the aberrant me-

thylation of the tumor suppressor genes p16INK4A,

fragile histidine triad (FHIT), retinoic acid receptor

beta (RARb), H-cadherin, DNA repair gene O6-

methylguanine-DNA methyltransferase (MGMT), and

the RAS effector homolog (RASSF1A) has been shown

to occur in a substantial percentage of nonsmall cell

lung cancers (NSCLCs).3,4 However, the mechanism

underlying the hypermethylation of CpG islands in

the genes relevant to lung cancer carcinogenesis is

yet to be elucidated.

DNA methylation is the consequence of the

transfer of a methyl group to the 50-position of cyto-

sine, and is accomplished by the action of a transact-

ing enzyme, referred to as DNA methyltransferases

(DNMTs). To our knowledge to date, three members

of the DNMT family, i.e., DNMT1, DNMT2, DNMT3a,

and DNMT3b, have been cloned in mammals. DNMT1

exhibits a higher degree of affinity for hemimethylated

CpG dinucleotides in vitro than unmethylated CpG,

and is believed to perform a vital function in the

maintenance of the preexisting methylation pattern.5

DNMT3a and DNMT3b are considered to be de novo

methyltransferases.6 Although the hypermethylation

of CpG islands has been suggested to be the result of

the up-regulation of DNMTs in human tumors, stu-

dies regarding the relation between the levels of

expression of the DNMTs and the aberrant methyla-

tion of CpG islands have produced contradictory

results in a variety of tumors.7–9 In addition, there has

not yet been any report issued regarding their relation

in primary NSCLC.

To investigate the factors involved in the hyper-

methylation of CpG islands in NSCLC, we attempted

to determine whether DNMT RNA expression levels

correlate with the hypermethylation of the CpG islands

of multiple tumor suppressor genes whose methylation

has been reported in cases of NSCLC, including p16,

RARb2, H-cadherin, GSTP1, RIZ, and FHIT. We also

attempted to understand the prognostic significance of

the DNMTs.

MATERIALS AND METHODSStudy Population and Tissue SpecimensA total of 102 primary NSCLC patients participated

in this study. All these patients had undergone cu-

rative surgical resections at the Department of Tho-

racic Surgery at the Samsung Medical Center in

Seoul, Korea, between July 1995 and February 2002.

Written informed consent was provided by all of

the study participants. Paired samples of primary

NSCLCs and matched noncancerous normal tissues

were obtained from each patient. Surgically re-

moved tissues were immediately snap-frozen in liq-

uid nitrogen and stored at �808C until use. The

occurrence of death was evaluated as of December

31, 2004. The data of patients who died of causes

other than lung cancer, did not die before the end

of the study, or who were lost to follow-up during

the study period were treated as censored data

when calculating survival.

RNA Extraction and cDNA SynthesisThe fresh-frozen tissues were sectioned using a micro-

tome; the serial sections were positioned on slides

before RNA extraction and then stained with hematox-

ylin and eosin in an attempt to evaluate the admixture

of tumorous and nontumorous tissues. Areas corre-

sponding to tumor and normal tissues were microdis-

sected carefully. The total RNA was isolated from the

fresh-frozen tissues, with an RNeasy total RNA kit

(Qiagen, Chatsworth, CA) in accordance with the man-

ufacturer’s instructions. The first strand of cDNA was

synthesized from 2.0 lg RNA in a total volume of 50 lLreactionmixture containing random hexadeoxynucleo-

tide primer and SuperScript II RNase H� reverse tran-

scriptase (GIBCO BRL, Gaithersburg, MD). The syn-

thesized cDNA was purified with Qiagen Qiaquick

columns and eluted with 10 mM Tris-HCl (pH 8.8).

One microliter of purified cDNA mixture was used in

the subsequent polymerase chain reaction (PCR) amp-

lification.

Semiquantitative Reverse Transcription-PCRThe amplification reactions were performed in a total

volume of 50 lL containing 1.0 lL of purified cDNA

mixture, primers (300 ng each), dNTPs (1.25 mM

each), MgCl2 (1.5–2.0 mM), Tris-HCl (6.7 mM, pH

8.8), and 1.25 units of Taq polymerase under the fol-

lowing conditions: an initial 4-minute incubation step

at 948C (hot start), 25–35 cycles (45 seconds at 948C,45 seconds at annealing temperature, 45 seconds at

728C), depending on the region to be amplified, fol-

lowed by a final 10-minute extension at 728C. The

PCR products were then resolved by electrophoresis

on 1.2% to 1.8% agarose gels, stained with ethidium

bromide (Life Technologies, Bethesda, MD), and

finally observed under ultraviolet illumination (Fig.

1A). The band intensities were quantitated using the

Gel documentation 2000 (Bio-Rad, Hercules, CA) and

analysis system (Quantity One 4.2 program in Bio-

Rad). Reverse transcription (RT)-PCRs were con-

ducted in triplicate for each of the primer sets and

the mean of the band intensities was used as the

quantitative value. Primer sequences and annealing

temperatures for the DNMTs and proliferating cell

Elevated mRNA Levels of DNMTs in NSCLC/Kim et al. 1043

nuclear antigen (PCNA) were previously published by

Sato et al.9 PCNA instead of 18S rRNA or actin was

used to correct for variations in the amounts of RNA

because maintenance methylation during the S-phase

is required and DNMT1 gene expression is prolifera-

tion-dependent.7,10

DNA Extraction from Paraffin BlocksFormalin-fixed, paraffin-embedded tissue blocks, each

containing at least 75% neoplastic tissue, were used in

this study. Serial 10-lm tissue sections were cut from

each of the paraffin blocks, then transferred to slides.

The sections were stained with hematoxylin and eosin

in order to delineate the tumor areas before DNA

extraction. Areas corresponding to tumors were micro-

dissected carefully from the surrounding normal stro-

mal tissues, placed in Eppendorf tubes, deparaffinized

overnight at 638C in xylene, and then vigorously vor-

texed. After 5 minutes of centrifugation at full speed,

the supernatant fluid was removed and ethanol was

added in order to remove any residual xylene, then

subsequently removed via centrifugation. After the

ethanol had been evaporated the tissue pellet was

resuspended in a lysis buffer ATL (Qiagen) and the

DNA was isolated according to the manufacturer’s

instructions (Qiagen).

Methylation-Specific PCRThe methylation status of the promoter regions of

the p16, RARb2, H-cadherin, GSTP1, RIZ, and FHIT

genes was determined by methylation-specific PCR

(MSP), as previously described by Herman et al.11

(Fig. 1B). Two primer sets were designed, 1 specific

to DNA unmethylated at the promoter region and

the other specific to methylated DNA. The primer

sequences and annealing temperatures used in the

MSP procedure were described previously.12,13 DNA

from the peripheral blood lymphocytes of a healthy

individual was treated with SssI methyltransferase

(New England Biolabs, Beverly, MA) and used as a

positive control for the methylated CpG islands. DNA

from normal lymphocytes was used as a positive

control for the unmethylated CpG islands.

Statistical AnalysisThe association between the elevated mRNA levels of

DNMTs and the clinicopathologic characteristics was

analyzed by the Wilcoxon rank sum test (or Student

t-test) and the chi-square test (or Fisher exact test)

for continuous and discrete variables. Multivariate logis-

tic regression was conducted in order to estimate the

relation between the hypermethylation of a gene and

the covariates determined by univariate analysis to

be statistically significant, as well as those that were

biologically important and the elevated mRNA levels

of DNMTs, and to calculate odds ratio (OR). The

effects of elevated mRNA levels of DNMTs on patient

survival were estimated by the Kaplan–Meier method

and the differences between the 2 groups were com-

pared using the log-rank test. A Cox proportional

hazards regression analysis was conducted in order

to estimate the hazard ratio after controlling for

potential confounding factors, including age, sex,

stage, smoking, and histology. All of the P values

shown are 2-sided and P values <.05 were considered

statistically significant.

RESULTSClinicopathologic CharacteristicsThe relations between the elevated mRNA levels of

DNMT1 and DNMT3b and the clinicopathologic chara-

FIGURE 1. Representative examples of reverse-transcription polymerasechain reaction (RT-PCR) and methylation-specific PCR (MSP). (A) Transcripts

of DNA methyltransferases (DNMT) 1, DNMT3b, and proliferating cell nuclear

antigen (PCNA) were detected by RT-PCR. Tumor (T) and matched normal tis-

sue (N) were run on 1.2% to 1.8% agarose gels. The numbers below the

bands indicate fold expression in tumor tissues compared with normal tis-

sues. (B) MSP on the RARb2, H-cadherin, RIZ, FHIT, p16, and GSTP1 genes

was performed using unmethylation-specific (U) and methylation-specific (M)

primer sets. Twenty lL of PCR product was run on a 2% metaphore agarose

gel, stained with ethidium bromide, and visualized under ultraviolet illumina-

tion. The numbers shown are sample identification numbers. Pos indicates

positive control.

1044 CANCER September 1, 2006 / Volume 107 / Number 5

cteristics are shown in Table 1. DNMT expression levels

(i.e., DNMT signal intensity normalized to that of con-

trol PCNA) were considered elevated if mRNA levels of

DNMTs in tumor tissues were greater than those of

any of the matched normal tissues.14 The mRNA levels

of DNMT1 and DNMT3b were found to be elevated in

53% of patients (54 of 102 patients) and 58% of

patients (59 of 102 patients), respectively. The expres-

sion levels of DNMT1 in cases with elevated mRNA

levels of DNMT1 were approximately 2.3 (range, 1.1–

8.7) times higher than those observed in normal tis-

sues. Forty-one of 54 patients with elevated mRNA

expression of DNMT1 were in the low range (<5-fold)

and 13 in the high range. For DNMT3b, the expression

levels in tumor tissues were approximately 2.7 (1.1–

10.2) times greater than those in normal tissues. The

patients with elevated mRNA levels of DNMT1 or

DNMT3b were slightly younger than those without, but

these differences were not determined to be statistically

significant. The elevated mRNA levels of DNMTs were

not associated with gender; the elevated mRNA levels

of DNMT1 were found to be more prevalent in females

than in males, but the elevated mRNA levels of

DNMT3b showed an opposite result. Smoking status

was also not associated with the elevated mRNA levels

of DNMT1 or DNMT3b. The overall prevalence of ele-

vated mRNA levels of DNMT1 or DNMT3b was similar

at all stages. The elevated mRNA levels of DNMT1 were

more frequently encountered in cases of adenocarci-

noma than in cases of squamous cell carcinoma (65%

vs. 47%), but this difference was not statistically signifi-

cant. The elevated mRNA levels of DNMT3b were also

not associated with tissue type (P ¼ .80). The elevated

mRNA levels of DNMT1 were determined to occur

more frequently in poorly differentiated cells rather

than in well- and moderately differentiated cell types

(P ¼ .19).

Elevated mRNA Levels of DNMTs and theHypermethylation of Multiple Tumor Suppressor GenesThe hypermethylation of p16, RARb2, H-cadherin,

GSTP1, RIZ, and FHIT was observed in 37%, 38%,

34%, 18%, 9%, and in 31% of patients, respectively.

The relation between the methylation status of indi-

vidual CpG islands and the elevated mRNA levels of

the DNMTs were investigated. However, the elevated

mRNA levels of DNMT 1 and DNMT3b did not show

a strong correlation with CpG island hypermethyl-

ation of any of the 6 genes studied (Fig. 2). Among

the 6 genes studied, the p16 gene exhibited the most

profound difference in the prevalence of aberrant

methylation between the groups with elevated mRNA

levels of DNMT1 and those without (42% vs. 31%,

respectively) (Fig. 2A), but the difference was not sta-

tistically significant.

Several factors, including smoking and age, are

known to be associated with DNAmethylation, although

the mechanism underlying the hypermethylation of

genes by these factors remains to be revealed. There-

fore, we conducted logistic regression analysis to con-

trol for the potential confounding effects of the

variables (Table 2). Patients with elevated mRNA levels

of DNMT1 had an increased risk of hypermethylation

at the promoter region of the p16 gene relative to those

without elevated mRNA levels of any of the DNMTs,

controlling for age, sex, histology, stage, and pack-

years smoked (OR ¼ 2.70, 95% confidence interval

[95% CI], 1.02–7.15; P ¼.02). Age and sex were adjusted

due to biologic significance, and smoking and histol-

ogy due to an association with p16 methylation in

NSCLC.15 These data indicate that the relation

between the elevated mRNA levels of DNMT1 and the

hypermethylation of genes differs on a genetic basis,

and may also be affected by other factors in addition

to the elevated mRNA levels of DNMT1. Although the

elevated mRNA levels of DNMT3b was found to have a

synergistic effect (OR ¼ 3.51, 95% CI, 1.33–16.07; P ¼

TABLE 1Clinicopathologic Characteristics (n = 102)

DNMT1

P

DNMT3b

PNo Yes* No Yes*

Agey 60 6 10 58 6 12 .48 616 9 58 6 12 .31

Sex

Male 32 31 26 37

Female 16 23 .34 17 22 .82

Pack-yearsy 55 6 36 60 6 38 .46 596 41 576 35 .95

Smoking status

Never 5 4 3 6

Exsmoker 28 24 19 33

Current 15 26 .23 21 20 .31

Pathologic stage

1 28 32 24 36

2 13 14 12 15

3 6 7 5 8

4 1 1 1.00 2 0 .51

Histology

Adeno 12 22 13 21

Squamous 28 25 24 29

Others 8 7 .24 6 9 .80

Differentiation

Well 5 6 4 7

Moderately 34 29 29 34

Poorly 9 19 .16 10 18 .66

DNMT indicates DNA methyltransferase; Adeno, adenocarcinoma; Squamous, squamous cell

carcinoma.

* Elevated mRNA expression of DNMT1 or DNMT3b.y Mean 6 the standard error.

Elevated mRNA Levels of DNMTs in NSCLC/Kim et al. 1045

.007) on the risk of p16 methylation in patients with

elevated mRNA levels of DNMT1, the elevated mRNA

levels of DNMT3b alone was not an independent risk

factor of p16methylation.

Survival AnalysisThe Kaplan–Meier estimator of the survivorship func-

tion was used in our determination of the effects of

the elevated mRNA levels of any of the DNMTs, or

the co-elevation in mRNA levels of the DNMTs, on

patient survival. The survival between groups with

elevated mRNA levels of DNMTs and those without

were compared by the log rank test. The data were

stratified on the basis of disease stage, as stage is an

independent risk factor in cases of NSCLC. The Kaplan–

Meier survival curves showed that the survival of pa-

tients with elevated mRNA levels of DNMT3b in TNM

Stage I or Stages II-III did not differ significantly from

the survival of patients without (data not shown).

However, the overall survival time in Stage I patients

(P ¼ .007) (Fig. 3A) and patients with Stages II-III dis-

ease (P ¼ .002) (Fig. 3B) was found to be significantly

shorter in patients with elevated mRNA levels of

DNMT1. The median survival time in Stage I and

Stages II-III patients with elevated mRNA levels of

DNMT1 was only 33 months and 17 months, respec-

tively. The median survival in patients with coeleva-

FIGURE 2. Prevalence of hypermethylation of multiple tumor suppressorgenes according to the expression status of DNA methyltransferases-1

(DNMT1) and DNMT3b. The black and gray bars indicate the groups with ele-

vated mRNA expression (A, DNMT1; B, DNMT3b) and those without, respec-

tively. There was no correlation noted between the hypermethylation of any of

the 6 genes and the elevated mRNA expression of DNMT1 or DNMT3b.

TABLE 2Multivariate Logistic Regression* of the Association between p16Methylation and Clinicopathologic Features (n = 102)

Variable OR 95% CI P

DNMTs

Noy 1.00

DNMT1 2.70 1.02–7.15 .02

DNMT3b 1.83 0.71–4.69 .21

DNMT1 *3b{ 3.51 1.33–16.07 .007

OR indicates odds ratio; 95% CI, 95% confidence interval; DNMT, DNA methyltransferase.

* Adjusted for age, sex, smoking, stage, and histology.y No mRNA elevation of either DNMT1 or DNMT3b.{ Coelevation of DNMT1 and DNMT3b.

FIGURE 3. Kaplan--Meier survival curve according to the elevated mRNAexpression of DNA methyltransferase-1 (DNMT1). Patients exhibiting the ele-

vated mRNA expression of DNMT1 manifested poorer survival compared with

those without the elevated mRNA expression of DNMT1 among (A) Stage I

patients (n = 60 patients) and (B) Stage II-III patients (n = 40 patients).

1046 CANCER September 1, 2006 / Volume 107 / Number 5

tion of DNMT1 and DNMT3b was slightly shorter in

Stage I and in Stages II-III patients (32 months and

16 months, respectively) compared with patients

with elevated mRNA levels of DNMT1 only, but this

difference was not statistically significant in either

Stage I patients (P ¼ .54) or Stages II-III patients (P ¼.71). This suggests that elevated mRNA levels of

DNMT3b do not significantly affect the survival in

patients with elevated mRNA levels of DNMT1.

Cox Proportional Hazards AnalysisCox proportional hazards regression analysis was per-

formed to determine whether elevated mRNA levels of

DNMTs are an independent prognostic factor, after

controlling for potential confounding factors, includ-

ing age, sex, stage, differentiation, histology, and pack-

years smoked (Table 3). DNA methylation status was

also controlled, as the promoter methylation of certain

genes is known to be associated with patient survival.

Patients without elevated mRNA expression in any of 2

DNMTs were used as a reference group. Patients with

elevated mRNA levels of DNMT1 were associated with

significantly unfavorable prognoses, and the elevated

mRNA levels of DNMT1 carried a 3.51 (95% CI, 1.18–

12.76; P ¼ .02) times higher risk of failure than that of

the reference group. The elevated mRNA levels of

DNMT3b did not affect patient prognosis (HR [hazards

ratio] ¼ 1.26, 95% CI, 0.26–5.14; P ¼.79). The hazard of

failure for patients in whom the mRNA levels of

DNMT1 and DNMT3b were co-elevated was 4.12 times

greater than that of the reference group (95% CI, 1.35–

18.35; P ¼ .01). However, the differences in survival

between the groups with elevated mRNA levels of

DNMT1 and the groups with coelevation of DNMT1

and DNMT3b were not statistically significant (P ¼.46). This indicates that elevated mRNA levels of

DNMT3b do not significantly affect the survival of

patients with elevated mRNA levels of DNMT1.

DISCUSSIONThe elevated mRNA levels of DNMT3b in the present

study were not significantly associated with hyper-

methylation of the 6 genes studied, but the elevated

mRNA levels of DNMT1 were associated with the hy-

permethylation of the p16 gene in multivariate logistic

regression analysis, thereby suggesting that other fac-

tors in addition to elevated mRNA levels of DNMT1

may be involved in CpG island hypermethylation of

tumor suppressor genes on a gene-specific basis in

primary NSCLC.

What, then, is responsible for this gene-specific

relation between CpG island hypermethylation and

the elevated mRNA levels of DNMTs in NSCLC? At

present, the molecular mechanisms underlying differ-

ential susceptibility to hypermethylation by DNMTs

remain somewhat unclear. The increased activity of

DNA methyltransferase is 1 of the factors associated

with increased susceptibility to the CpG island hyper-

methylation of a specific gene in cases of human can-

cer, and there may be a threshold of DNMTactivity for

the hypermethylation of a gene.16 Different degrees of

changes to the chromatin structure or histone modifi-

cations may also constitute sources of differential sus-

ceptibility to CpG island methylation via increases in

the activity of DNMT. Several groups17,18 have sug-

gested that chromatin structural modification is a pre-

requisite for the targeting of DNMT enzymes to

genetic sequences. A reduction in the level of DDM1

(decrease in DNA methylation), an SWI2/SNF2-like

protein, in the flowering plant Arabidopsis thaliana

was determined to induce a 70% reduction in the

amount of 5-methylcytosine, and to severely compro-

mise the rate of de novo methylation.18

In addition to chromatin remodeling, the modifi-

cation of histones that contribute to the maintenance

of chromatin structure around the promoter region

of a gene may also modulate genetic susceptibility to

hypermethylation.19,20 For example, a fifth histone,

H1, is known to be bound to the DNA on the inside

of the solenoid structure of condensed chromatin,

and appears to be necessary for the condensation of

nucleosome chains into higher-order structures. His-

tone H1 exhibits a clear preference for methylated

CpG, regardless of DNA sequences.21 One somatic

variant of histone H1, H1e, has been shown to in-

hibit the in vitro DNA methylation process via the

exertion of a specific inhibitory effect on DNA me-

thyltransferase activity in vitro by binding to CpG-rich

DNA around promoter regions.22 Therefore, alteration

in the function of H1e may serve to alter chromatin

structures, and this may perform a crucial function in

TABLE 3Survival Outcome by Multivariate Cox Proportional Hazards Analysis*in NSCLC (n = 102)

Variables HR 95% CI P

DNMTs

Noy 1.00

DNMT1 3.51 1.18–12.76 .02

DNMT3b 1.26 0.26–5.14 .79

DNMT1*3b{ 4.12 1.35–18.35 .01

NSCLC indicates nonsmall cell lung cancer; HR, hazards ratio; 95% CI, 95% confidence interval;

DNMT, DNA methyltransferase; CI, confidence interval.

* Adjusted for age, sex, smoking, stage, differentiation, and histology.y No mRNA elevation in either DNMT1 or DNMT3b.{ Coelevation of DNMT1 and DNMT3b.

Elevated mRNA Levels of DNMTs in NSCLC/Kim et al. 1047

the gene-specific hypermethylation affected by the ele-

vated mRNA levels of DNMT.

Local specific factors, including different degrees

of protection against methylation or tissue-specific

factors, might be responsible for the observed var-

iances in susceptibility. Cis-acting elements, which

block the spread of methylation from a methylation

center, thereby protecting CpG islands from methyla-

tion, may also be 1 of the factors responsible for

gene-specific methylation. Multiple SP-1 sites have

been detected at both the 50 and 30 boundaries, andwere theorized to potentially protect the islands from

the spread of methylation from a methylation cen-

ter.23 Therefore, we compared the local specific fac-

tors at the CpG islands of the 6 genes studied.

However, the frequency of Sp1 sites was not deter-

mined to differ significantly among the 6 CpG

islands, and no significant differences were observed

in the percentage of C þ G, the observed CpG/

expected CpG ratio, or the frequency of CpG per 100

basepairs (data not shown).

Exposure to tobacco smoke may induce selective

changes in a limited set of key regulatory transcription

factors, including the SP1 protein. The cis-acting fac-

tors that normally protect the islands from methyla-

tion might be disrupted by exposure to tobacco smoke.

Therefore, we conducted a multivariate analysis to

control for potential confounding factors, especially

smoking. The elevated mRNA levels of DNMT1 were

not associated with the hypermethylation of p16 in

univariate analysis, but showed an association with

p16 methylation in multivariate analysis, after adjust-

ing for possible confounding factors, especially smok-

ing. These observations indicate that environmental

factors can modulate the effects of DNMT1 on pro-

moter methylation.

The prognoses of patients with elevated mRNA

levels of DNMT1 in this study were poorer than those

without. These results indicate that elevated mRNA

levels of DNMT1 are an independent prognostic factor

in cases of NSCLC. However, a fold change of no less

than 2-fold may result from the imprecise nature of

semiquantitative RT-PCR. Therefore, we reanalyzed the

relation between patient survival and elevated DNMT1

expression at a cutoff value of a 2-fold change. But we

also found a significant difference in patient survival

between patients with elevated DNMT1 mRNA levels

more than 2-fold and those without elevated DNMT1

mRNA (P ¼ .04 in Stage I and P ¼ .01 in Stages II-III).

This may result from the small number of samples and

needs further study in a large sample to understand

the biologic significance of a fold-change of no less

than 2-fold and the effect of a different threshold of

DNMT1 overexpression on patient survival.

We also compared the mRNA levels between

semiquantitative RT-PCR and quantitative real-time

PCR for a small number of samples. Dichotomization

of patients by the presence or absence of mRNA ele-

vation in tumor tissue compared with normal tissue

was the same between the 2 methods, but the amount

of expression change of DNMTs 1 and 3b in both

patient groups was significantly different between the

2 methods. Accordingly, it is likely that the expression

levels of DNMTs measured by semiquantitative RT-

PCR may not be sufficient for studying an association

between promoter methylation or patient survival and

DNMTs expression at a different cutoff value (e.g., 1.5-

fold, 2-fold, 5-fold).

Taken together, it appears probable that aberrant

methylation of CpG islands in cases of NSCLC may

occur in a variety of complex ways, as the result of a

multiplicity of factors, rather than as the result of a

simple transcriptional up-regulation of any of the

DNMTs. Accordingly, it is likely that the conflicting

results about the relation between DNA methylation

and the elevated mRNA levels of DNMTs in previous

studies7–9,24,25 may have resulted from the study of dif-

ferent genes that have different gene-specific factors or

from the lack of controlling for several confounding

factors in data analysis. This study was severely limited

by the small number of the sample. Additional work

with a large sample will be required in order to pre-

cisely determine the factors responsible for the differ-

ential susceptibility to hypermethylation of a gene in

cases of NSCLC and to understand the effect of ele-

vated DNMT1 expression on patient survival at differ-

ent thresholds. DNMT3a expression was not studied in

this study because DNMT1 and DNMT3b are known to

cooperatively maintain DNA methylation and gene

silencing in human cancer cells.26 Although expression

of DNMTs at the protein level was not analyzed in the

present study due to lack of available sample, a correla-

tion of elevated DNMTmRNA levels with active protein

levels in NSCLC has been suggested.27,28

In conclusion, our results suggest that the deregu-

lation of DNMT1 may be an independent prognostic

factor in NSCLC and that CpG island hypermethylation

of tumor suppressor genes in NSCLC may be main-

tained by a complex interaction of several factors

rather than by a simple transcriptional up-regulation

of DNMT1.

REFERENCES1. Baylin SB, Herman JG, Graff JR, et al. Alterations in DNA

methylation: a fundamental aspect of neoplasia. Adv Can-

cer Res. 1998;72:141–196.

2. Esteller M, Corn PG, Baylin SB, Herman JG. A gene hyper-

methylation profile of human cancer. Cancer Res. 2001;61:

3225–3229.

1048 CANCER September 1, 2006 / Volume 107 / Number 5

3. Tsou JA, Hagen JA, Carpenter CL, Laird-Offringa IA. DNA

methylation analysis: a powerful tool for lung cancer diag-

nosis. Oncogene. 2002;21:5450–5461.

4. Zochbauer-Muller S, Minna JD, Gazdar AF. Aberrant DNA

methylation in lung cancer: biological and clinical implica-

tions. Oncologist. 2002;7:451–457.

5. Li E, Bestor TH, Jaenisch R. Targeted mutation of the DNA

methyltransferase gene results in embryonic lethality. Cell.

1992;69:915–926.

6. Okano M, Bell DW, Haber DA, Li E. DNA methyltransferase

Dnmt3a and 3b are essential for de novo methylation and

mammalian development. Cell. 1999;9:247–257.

7. Eads CA, Danenberg KD, Kawakami K, et al. CpG island

hypermethylation in human colorectal tumors is not asso-

ciated with DNA methyltransferase overexpression. Cancer

Res. 1999;59:2302–2306.

8. Kanai Y, Ushijima S, Kondo Y, et al. DNA methyltransferase

expression and DNA methylation of CPG islands and peri-

centromeric satellite regions in human colorectal and

stomach cancers. Int J Cancer. 2001;91:205–212.

9. Sato M, Horio Y, Sekido Y, et al. The expression of DNA

methyltransferases and methyl-CpG-binding proteins is

not associated with the methylation status of p14(ARF),

p16(INK4a) and RASSF1A in human lung cancer cell lines.

Oncogene. 2002;21:4822–4829.

10. Papadopoulos T, Pfeifer GP, Hoppe F, et al. Proliferation-

associated expression of DNA methyltransferase in human

embryonic lung cells. Virchows Arch B Cell Pathol Incl Mol

Pathol. 1989;56:371–375.

11. Herman JG, Graff JR, Myohanen S, et al. Methylation-

specific PCR: a novel PCR assay for methylation status of

CpG islands. Proc Natl Acad Sci U S A. 1996;93:9821–

9826.

12. Kim H, Kwon YM, Kim JS, et al. Tumor-specific methyla-

tion in bronchial lavage for the early detection of non-

small cell lung cancer. J Clin Oncol. 2004;22:2363–2370.

13. Du Y, Carling T, Fang W, et al. Hypermethylation in human

cancers of the RIZ1 tumor suppressor gene, a member of a

histone/protein methyltransferase superfamily. Cancer Res.

2001;61:8094–8099.

14. Kanai Y, Ushijima S, Kondo Y, Nakanishi Y, Hirohashi S.

DNA methyltransferase expression and DNA methylation

of CpG islands and peri-centromeric satellite regions in

human colorectal and stomach cancers. Int J Cancer. 2001;

91:205–212.

15. Kim DH, Nelson HH, Wiencke JK, et al. p16 (INK4a) and

histology-specific methylation of CpG islands by exposure

to tobacco smoke in non-small cell lung cancer. Cancer

Res. 2001;61:3419–3424.

16. Vertino PM, Yen RW, Gao J, Baylin SB. De novo methyla-

tion of CpG island sequences in human fibroblasts overex-

pressing DNA (cytosine-5-) methyltransferase. Mol Cell Biol.

1996;16:4555–4565.

17. Jeddeloh JA, Stokes TL, Richards EJ. Maintenance of genomic

methylation requires a SWI2/SNF2-like protein. Nat Genet.

1999;22:94–97.

18. Dennis K, Fan T, Geiman T, et al. Lsh, a member of the

SNF2 family, is required for genome-wide methylation. Genes

Dev. 1002;15:2940–2944.

19. Fahrner JA, Eguchi S, Herman JG, Baylin SB. Dependence

of histone modifications and gene expression on DNA hy-

permethylation in cancer. Cancer Res. 1002;62:7213–7218.

20. Kondo Y, Shen L, Issa JP. Critical role of histone methyla-

tion in tumor suppressor gene silencing in colorectal can-

cer. Mol Cell Biol. 2003;23:206–215.

21. Wellman SE, Sittman DB, Chaires JB. Preferential binding of

H1e histone to GC-rich DNA. Biochemistry. 1994;33:384–388.

22. Zardo G, Santoro R, D’Erme M, et al. Specific inhibitory effect

of H1e histone somatic variant on in vitro DNA-methylation

process. Biochem Biophy Res Commun. 1996;220:102–107.

23. Brandeis M, Frank D, Keshet I, et al. Sp1 elements protect

a CpG island from de novo methylation. Nature. 1994;371:

435–438.

24. Ting AH, Jair KW, Suzuki H, Yen RWC, Baylin SB, Schuebel

KE. CpG island hypermethylation is maintained in human

colorectal cancer cells after RNAi-mediated depletion of

DNMT1. Nat Genet. 2004;36:582–584.

25. Rhee I, Jair KW, Yen RWC, et al. CpG island is maintained

in human cancer cells lacking DNMT1. Nature. 2000;404:

1003–1007.

26. Rhee I, Bachman KE, Park BH, et al. DNMT1 and DNMT3b

cooperate to silence genes in human cancer cells. Nature.

2002;416:552–556.

27. Issa JPJ, Vertino PM, Wu J, et al. Increased cytosine DNA-

methyltransferase activity during colon cancer progression.

J Natl Cancer Inst. 1993;85:1235–1240.

28. Belinsky SA, Nikula K, Baylin SB, Issa JP. Increased cytosine

DNA-methyltransferase activity is target-cell-specific and

an early event in lung cancer. Proc Natl Acad Sci U S A.

1996;93:4045–4050.

Elevated mRNA Levels of DNMTs in NSCLC/Kim et al. 1049