Methylation of the candidate biomarker TCF21 is very frequent across a spectrum of early-stage nonsmall cell lung cancers


  • Kristy L. Richards PhD, MD,

    1. Department of Genetics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    2. Department of Medicine, Lineberger Comprehensive Cancer Center, University of North Carolina School of Medicine, Chapel Hill, North Carolina
    3. Department of Genetics, Lineberger Comprehensive Cancer Center, University of North Carolina School of Medicine, Chapel Hill, North Carolina
    Search for more papers by this author
  • Baili Zhang MS,

    1. Department of Genetics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Menghong Sun PhD,

    1. Department of Pathology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Wenli Dong PhD,

    1. Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Jennifer Churchill BS,

    1. Department of Genetics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Linda L. Bachinski PhD,

    1. Department of Genetics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Charmaine D. Wilson MS,

    1. Department of Genetics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Keith A. Baggerly PhD,

    1. Department of Bioinformatics and Computational Biology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Guosheng Yin PhD,

    1. Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • D. Neil Hayes MD, MPH,

    1. Department of Medicine, Lineberger Comprehensive Cancer Center, University of North Carolina School of Medicine, Chapel Hill, North Carolina
    Search for more papers by this author
  • Ignacio I. Wistuba MD,

    1. Department of Pathology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    2. Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Ralf Krahe PhD

    Corresponding author
    1. Department of Genetics, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    2. Department of Thoracic/Head and Neck Medical Oncology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    • Department of Genetics, Unit 1010, The University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Blvd., Houston, TX 77030
    Search for more papers by this author
    • Fax: (713) 834-6319



The transcription factor TCF21 is involved in mesenchymal-to-epithelial differentiation and was shown to be aberrantly hypermethylated in lung and head and neck cancers. Because of its reported high frequency of hypermethylation in lung cancer, further characterization of the stages and types of nonsmall cell lung cancer (NSCLC) that are hypermethylated and the frequency of hypermethylation and associated “second hits” were assessed.


TCF21 promoter hypermethylation in 105 NSCLC including various stages and histologies in smokers and nonsmokers was determined. In addition, TCF21 loss of heterozygosity and mutational status were examined. Twenty-two cancer cell lines from varied tissue origins were also assayed. The NSCLC results were validated and expanded by examining TCF21 immunohistochemical expression on a tissue microarray containing 300 NSCLC cases.


Overall, 81% of NSCLC samples showed TCF21 promoter hypermethylation, and 84% showed decreased TCF21 protein expression. Multivariate analysis showed that TCF21 expression, although below normal in both histologies, was lower in adenocarcinoma than in squamous cell carcinoma and was not independently correlated with sex, smoking, and EGFR mutation status or with clinical outcome. Cell lines from other cancer types also showed frequent TCF21 promoter hypermethylation.


Hypermethylation and decreased expression of TCF21 were tumor specific and very frequent in all NSCLCs, even early-stage disease, thus making TCF21 a potential candidate methylation biomarker for early-stage NSCLC screening. TCF21 hypermethylation in a variety of tumor cell lines suggests it may also be a valuable methylation biomarker in other tumor types. Cancer 2011. © 2010 American Cancer Society.

Lung cancer is the number one cause of cancer mortality worldwide and kills more people than breast, colon, and prostate cancer combined.1 Unlike these other common cancers, however, lung cancer has no effective screening strategy to detect early-stage disease at a time when surgery may be curative. The need for such a strategy is obvious, and the many attempts to detect lung cancer early thus far have failed to show clinical benefit.2-4 These include screening CT scans, sputum cytology, screening chest x-rays, and serum markers.

Recently, promoter hypermethylation has been recognized as an important mechanism by which genes regulating cellular proliferation are silenced during cancer development.5, 6 Promoter hypermethylation involves DNA methylation of CpG islands in or near the promoter region of certain genes, rendering them transcriptionally silent. This down-regulation of gene expression of important cellular growth control genes has been shown to be important for cancer progression and outcome, with poorer outcomes associated with promoter hypermethylation of such important genes as RASSF1A, RARB, and HIF1.7-9

TCF21 is a recently recognized target of aberrant promoter hypermethylation in cancer, discovered in a genomic screen for regions of DNA that are hypermethylated in cancer.10 It was reported to be frequently hypermethylated in head and neck and lung cancers, and restoration of TCF21 expression inhibited tumor growth, both in a lung cancer cell line and in a mouse xenograft model. TCF21 is widely expressed; its normal function is to promote mesenchymal transition into epithelial cells.11 Reversal of this process, known as the epithelial-to-mesenchymal transition (EMT), has been implicated in tumor invasion and metastasis.12, 13 Therefore, the silencing of TCF21 may be a mechanism for tumor cells to gain these aggressive characteristics during the course of tumor progression. Given that TCF21 was reported to be frequently hypermethylated and silenced in nonsmall cell lung cancer (NSCLC), as well as its plausible biologic role in tumor progression, we sought to more precisely define the frequency of TCF21 promoter hypermethylation in NSCLC. We were especially interested in defining its frequency among different cancer stages and histologic subtypes. Here, we show that TCF21 is very frequently hypermethylated in a variety of NSCLCs and that protein expression of TCF21 is also very frequently reduced, either of which could be used for screening and/or diagnostic purposes as a biomarker of early disease.


Frozen Tumor Specimens, Cell Lines, and DNA Extraction

Patient NSCLC specimens were obtained from surgical specimens at both The University of Texas M. D. Anderson Cancer Center (42 matched tumor/normal samples, 7 unpaired tumor samples) as well as from the University of North Carolina Lineberger Comprehensive Cancer Center tumor bank of surgical specimens (56 unpaired tumor samples). In both institutions, informed consent was obtained prior to surgery for the use of specimens as part of an institutional review board-approved protocol, in accord with the Helsinki Declaration. Tissue was snap-frozen and used for later DNA extraction. Genomic DNA was extracted from the DNA-protein phase of TriZol-extracted tissues according to the manufacturer's suggestions (Invitrogen, San Diego, Calif). DNA was extracted using a PureGene kit (Gentra, Minneapolis, Minn) on cell pellets from 4 HNSCC cell lines (SCC-4, SCC-9, SCC-15, and SCC-25), 5 lung cancer cell lines (H1395, H520, H2170, SK-MES-1, and SW-900), 1 breast cancer cell line (MCF7), 1 cervical cancer cell line (HeLa), 2 brain cancer cell lines (SK-N-AS, and M059K), 1 uterine cancer cell line (AN3CA), 1 sarcoma cell line (HT1080), 1 kidney cancer cell line (HEK293), and 6 colon cancer cell lines (LoVo, SW48, HCT-15, DLD-1, COLO 320DM, and RKO) according to the manufacturer's suggestions. All cell lines are available from ATCC (Manassas, Va). Four normal pools, each comprising DNA from peripheral blood mononuclear cells (PBMCs) of 6 individuals, were generated representing different sexes and ages (females ≤40 years, females age >40 years, males ≤40 years, and males ≥40 years).

TCF21 Promoter Methylation

PCR and sequencing primers were designed the PSQ Assay Design software (Qiagen, Valencia, Calif). PCR was performed in a 25-μL reactions containing Qiagen HotStart Taq master mix (Qiagen) using 1 μL of bisulfate-converted DNA (about 10 ng/μL). Bisulfite conversion of genomic DNA was performed as previously reported.14 Briefly, 0.5-1.0 μg of genomic DNA was treated using a EZ-96 DNA Methylation Gold Kit (Zymo Research, Irvine, Calif), including DNA sulfonation, deamination, desalting, desulfonation, and recovery. Bisulfite-treated DNA was stored at −80°C until use. To reduce the cost per assay, an amplification protocol was developed using a biotinylated universal primer approach.14 Final primer concentrations were 10 nM of the reverse primer tailed with the universal primer (5′-GACGGGACACCGCTGATCG TTTACCAAAAAAAACCCCCTAA-3′), 100 nM of the untailed forward primer (5′-GGTAGGGTGGTTTTG AGTT-3′), and 90 nM of the universal biotinylated primer (5′-GGGACACCGCTGATCGTTTA-3′) in each reaction. The universal primer sequence is underlined. The predicted amplicon size was 153 bp. Amplification was carried out as follows: denaturation at 95°C for 5 minutes, followed by 50 cycles at 95°C for 30 seconds, 51°C for 1 minute, 72°C for 45 seconds, and a final extension at 72°C for 7 minutes.

Following PCR amplification, pyrosequencing was performed on a PSQ96HS system (Qiagen) according to the manufacturer's protocol including the use of single-strand binding protein (PyroGold reagents). The pyrosequencing primer was (5′-TTGAGTTTGGAGAAGG-3′). The results were analyzed using Q-CpG software (Qiagen), which calculates the methylation percentage (mC/[mC + C]) for each CpG site, allowing quantitative comparisons. The methylation index (MI) was calculated as the average value of mC/(mC + C) for all 9 of the interrogated CpG sites in the assay. Genomic DNA treated with M.SssI (New England Biolabs, Ipswich, Mass) was used as a universally methylated positive control; the same untreated genomic DNA amplified by whole genome amplification (GenomiPhi, GE Healthcare, Piscataway, NJ) was used as a universally unmethylated negative control.

Decitabine Treatment and Quantitative Real-Time RT-PCR

Three colon cancer cell lines (DLD-1, HCT-15, and RKO) with high levels (>85%) of TCF21 promoter hypermethylation were plated at a density of 500,000 cells/T75 flask. DLD-1 and HCT-15 cells were grown in RPMI-1640 supplemented with 10% FBS and 1% penicillin/streptomycin, RKO cells in EMEM supplemented with 10% FBS and 1% penicillin/streptomycin. Drug treatment with 1 μM decitabine (Sigma-Aldrich, St. Louis, Mo) was started 3 hours after seeding. Culture medium and drug were changed daily for treated and untreated cells. Cultures were grown for a minimum of 4 days until 80% confluence. Total cellular RNA was isolated using TRIZol reagent (Invitrogen). Input RNA (1 μg) was reverse-transcribed using a iScript cDNA Synthesis Kit (Bio-Rad, Hercules, Calif). TCF21 expression was assessed by TaqMan qRT-PCR using assays Hs00162646_m1 and Hs01546814_m1 (Applied Biosystems, Foster City, Calif) covering exons 1-2 and 2-3, respectively. qRT-PCR was carried out as follows in a 20-μL final reaction volume using 55 ng of RNA equivalents as cDNA input: initial denaturation at 95°C for 8.5 minutes, followed by 45 cycles at 95°C for 15 seconds, and 60°C for 1 minute, according to the manufacturer's suggestions. GUSB (Hs99999908_m1) was used as endogenous housekeeping gene control for normalization. Each assay was performed in triplicate. Relative expression was calculated using the ΔΔCt method and scaled.

LOH and Mutation Detection

Primers were designed for detection of 4 microsatellites within and flanking TCF21. Primer sequences are shown in Table 1. All forward primers were 5′-tailed with 5′-GACGGGACACCGCTGATCGTTTA-3′, and all reverse primers were 5′-tailed with 5′-GTTTCTT-3′. A universal primer with the sequence 5′-GGGACACCGCTGATCGTTTA-3′ end-labeled with either FAM, HEX, or NED was used in all microsatellite amplifications. PCR conditions for the 3-primer reactions were as described above for amplification using the universal biotinylated primer. Amplification products were pooled as appropriate and analyzed by capillary electrophoresis on an ABI 3100 Genetic Analyzer (Applied Biosystems).

Table 1. Primers for LOH and Mutation Detection in TCF21
AnalysisForward Primer (5′—3′)Reverse Primer (5′—3′)Amplicon Size (bp)

The coding region of TCF21 (exons 1 and 2) was sequenced in both directions in 4 fragments. In all, 45 lung cancer samples showing 0 or 1 hit were sequenced. Samples that had already been scored as having 2 hits were not sequenced. Primer sequences are shown in Table 1. All forward primers were 5′-tailed with the M13 forward sequence 5′-TGTAAAACGACGGCCAGT-3′, and all reverse primers were with M13 reverse 5′-CAGGAAA CAGCTATGACC-3′. After amplification, samples were treated with Exo-SAP (Amersham, Piscataway, NJ) and sequenced using Big Dye Terminator version 3.1 (Applied Biosystems, Carlsbad, Calif) under standard conditions, and products were purified by ethanol precipitation, dehydrated in a vacuum centrifuge, and resuspended in 20 μL of formamide before capillary electrophoresis on an ABI 3100 Genetic Analyzer. Sequences were aligned and visualized using Sequencher software (Gene Codes, Ann Arbor, Mich). Fragment 1 contained a polymorphic (CT)n simple tandem repeat of 8 to 12 units that, when polymorphic, was used to confirm retention of heterozygosity identified by the microsatellites.

Archival NSCLC Case Selection and Tissue Microarray Construction

We obtained archival formalin-fixed and paraffin-embedded (FFPE) material from surgically resected lung cancer specimens containing tumor and adjacent lung tissues from the Lung Cancer Specialized Program of Research Excellence (SPORE) Tissue Bank at The University of Texas M. D. Anderson Cancer Center, which was approved by the institutional review board. Tumor tissue specimens from 300 NSCLCs (191 adenocarcinomas and 109 squamous cell carcinomas) were histologically examined, classified using the 2004 World Health Organization classification system,15 and selected for tissue microarray (TMA) construction. After histologic examination, TMAs were constructed using triplicate 1-mm diameter cores from each tumor. Detailed clinical and pathological information, including demographic data, smoking history (never and ever-smokers) and status (never, former, and current smokers), pathologic TNM staging,16 overall survival, and time of recurrence, was available in most cases (Table 2). Patients who had smoked at least 100 cigarettes in their lifetime were defined as smokers, and smokers who had quit smoking for at least 12 months before lung cancer diagnosis were defined as former smokers.

Table 2. Clinical Characteristics of Patient Samples in the NSCLC Tissue Microarray and Correlation With TCF21 Expression
VariableNo.%Mean TCF21 ExpressionP-value
  • a

    Race: “Other” includes 13 African-Americans, 8 Asians, 9 Hispanics, and 1 Native American.

 Squamous cell carcinoma10936.354.9.003
 Caucasian, non-hispanic26989.741.6 
Tobacco history    
Pathological T classification    
Pathological N classification    
Pathological M classification    
Pathological stage    
Vital status    
Adjuvant therapy    

Immunohistochemical Staining and Evaluation

An antihuman TCF21 antibody was used for immunostaining (ab32981, Abcam). FFPE tissue histology sections (5 μm thick) were deparaffinized, hydrated, and heated in a steamer for 10 minutes with 10 mM sodium citrate (pH 6.0) for antigen retrieval. Peroxide blocking was performed with 3% H2O2 in methanol at room temperature for 15 minutes, followed by 10% bovine serum albumin in TBS-t for 30 minutes. Slides were incubated with primary antibody at 1:200 dilution for 65 minutes at room temperature. After washing with TBS-t, incubation with biotin-labeled secondary antibody for 30 minutes followed. Finally, samples were incubated with a 1:40 solution of streptavidin-peroxidase for 30 minutes. The staining was then developed with 0.05% 3′,3-diaminobenzidine tetrahydrochloride prepared in 0.05 mol/L Tris buffer at pH 7.6 containing 0.024% H2O2 and counterstained with hematoxylin. FFPE lung tissues having normal bronchial epithelia were used as a positive control. For a negative control, we used the same specimens used for the positive controls, replacing the primary antibody with PBS.

TCF21 immunostaining was detected in the cytoplasm of epithelial and tumor cells. Immunohistochemical expression was quantified by microscope observation by 2 pathologists (M.S. and I.W.) using a 4-value intensity score (0, 1+, 2+, and 3+) and the percentage of the extent of reactivity. A final score was obtained by multiplying both intensity and extension values (range 0-300), and 4 levels of expression were arbitrarily calculated based on that score: 1) negative (score 0-9); 2) low (score 10-100); 3) intermediate (score 100-199); and 4) high (score 200-300). Levels and scores were used for analysis.

EGFR Mutation Analysis

Exons 18 through 21 of EGFR were PCR-amplified using intron-based primers as previously described.17, 18 From microdissected FFPE cells, about 200 cells were used for each PCR amplification. All PCR products were directly sequenced using the PRISM dye-terminator cycle sequencing method (Applied Biosystems). All sequence variants were confirmed by independent PCR amplifications from at least 2 independent microdissections and DNA extraction and were sequenced in both directions, as previously reported.

Statistical Analysis

The clinical and pathological data were summarized using descriptive statistics and frequency tabulations. Wilcoxon rank-sum and Kruskal-Wallis tests were used to compare biomarker expression among different prognostic factor levels. The generalized linear model was used to assess the effect of prognostic factors on TCF21 expression in the multivariable setting. Fisher's exact test was used to compare the association between categorical variables. We examined the association between overall survival (OS) and recurrence-free survival (RFS) rates and TCF21 expression in NSCLC patients with stage I or II disease who had not undergone adjuvant chemotherapy. OS was defined as the time from surgery to death or the end of the study; RFS was defined as the time from surgery to recurrence or the end of the study. Univariate and multivariate Cox proportional hazards models were used to assess the effects of TCF21 protein expression on survival. Two-sided P values <.05 were considered statistically significant. All analyses were conducted using SAS (v 9.1, Cary, NC) and S-plus (v 8.0, Seattle, WA) software.


TCF21 Is Highly Methylated in Nearly All Cancer Cell Lines

To characterize TCF21 methylation levels in normal and malignant states, we examined various cancer cell lines from a spectrum of tissue types (brain, breast, cervix, colon, connective tissue, head and neck, kidney, lung, and uterus). We also assayed TCF21 methylation in normal PBMCs from younger and older individuals of both sexes because methylation levels can be influenced by age and/or sex. Universally methylated control DNA and genetically matched unmethylated control DNA defined the boundaries of detection of our assay (3%-93% methylation). Using pyrosequencing-based methylation analysis, we analyzed TCF21 methylation by averaging methylation levels of 9 promoter CpG sites. All but 1 cell line (SK-N-AS, a neuroblastoma cell line, 38%) was highly methylated, with levels at or approaching the upper limit of detection (Fig. 1). Normal PBMCs were essentially identical regardless of age or sex and demonstrated moderate levels of baseline methylation at approximately 20%.

Figure 1.

Methylation levels of individual cancer cell lines, normal PBMCs, and positive and negative methylation controls are shown. TCF21 promoter methylation levels are shown for 22 cancer cell lines and 4 pools of PBMCs of different sexes (male, female) and ages (≤40, >40 years). Control samples, fully methylated by treatment with SssI methylase or fully unmethylated by whole genome amplification, are also shown.

TCF21 Is Hypermethylated in >80% of NSCLC

To define the threshold for hypermethylation positivity, we began our analysis using genetically matched NSCLC and adjacent normal tissue pairs from the same patient. To assess the baseline levels of TCF21 methylation in lung tissue, we examined both normal adjacent tissue (NAT) from the tumor/normal (T/N) pairs (n = 42) comparing them with PBMC. Average methylation levels were 21.5% (SD = 4.6; n = 42) in NAT and 20.1% in the normal PBMC. Average TCF21 methylation level in the T samples was 41.3% (SD = 11.6; n = 42; Fig. 2A). The difference in average methylation between the N and T tissues was highly significant (P <1 × 10−13).

Figure 2.

Methylation levels and percentage of tumors with methylation levels >30% threshold are shown. (A) TCF21 promoter methylation levels are shown in a box-and-whisker plot for 42 normal adjacent lung tissues, 42 NSCLC tumors, 63 additional NSCLC tumors, all 105 NSCLC tumors combined, and 24 HNSCC tumors. (B) The bar graph represents the number of NSCLC and HNSCC tumors exceeding the 30% threshold for hypermethylation.

Using a threshold of 30% methylation, we found that 37 of 42 tumors (88%) were hypermethylated, whereas 41 of 42 matched normal samples (98%) were not. Using this cutoff to define hypermethylation, we then assayed a second set of 63 unpaired NSCLC samples. This second set of tumors contained a small number of large cell histologic subtypes, and some mixed histologic types (mostly adenosquamous). We found that 48 of these (76%) were hypermethylated (Fig. 2B). Overall, the average methylation level of all the tumor samples combined was 39.2% (SD = 11.7; n = 105). Using the threshold of 30% methylation, the overall frequency of hypermethylation in NSCLC was 81% (85 of 105).

Reactivation of TCF21 Expression by Demethylating Agent

To show that TCF21 promoter hypermethylation correlates with transcriptional silencing of the gene, we treated 3 colorectal cancer cell lines with high methylation levels (>85%) with the demethylating drug decitabine. We performed quantitative TaqMan mRNA real-time PCR to determine relative TCF21 expression with and without treatment. For all 3 cell lines culturing in the presence of the demethylating agent led to reactivation of TCF21 expression at the mRNA level, as assayed by 2 distinct quantitative real-time PCR assays (Fig. 3).

Figure 3.

Reactivation of TCF21 expression by the demethylating agent decitabine is shown. Quantitative TaqMan mRNA real-time PCR results interrogating the exon junctions of exons 1/2 and 2/3 in 3 colorectal cancer cell lines with high methylation levels (>85%) with and without decitabine treatment are shown.

Figure 4.

TCF21 protein expression is shown. (A) TCF21 immunohistochemical expression in lung cancer is shown. NSCLC samples were stained with an anti-TCF21 antibody and scored as none, low, medium, or high. Representative examples of SCC and adenocarcinoma samples with high TCF21 expression (left) and no TCF21 staining (right) are shown. (B) Shown is frequency of TCF21 expression in NSCLC on the TMA (300 patients). The percentage of samples in each expression category is shown. Reduced expression was defined as “none” or “low.”

Reduced Expression of TCF21 Protein in NSCLC

To determine whether TCF21 promoter hypermethylation also resulted in decreased TCF21 protein expression, we used a NSCLC TMA containing tumor samples from 300 patients. The microarray was stained with a TCF21 antibody, and protein levels were scored as none, low, intermediate, or high (Fig. 4A). Although normal adjacent lung tissue stained strongly for TCF21, 253 of 300 NSCLC samples (84%) showed reduced (low or none) staining (Fig. 4B).

Similar frequencies of TCF21 hypermethylation and decreased protein expression suggested that hypermethylation leads to reduced protein levels, which would be consistent with previously reported decreased mRNA levels resulting from TCF21 promoter hypermethylation.10 Because our TMA included only 9 overlapping samples between the TMA and TCF21 methylation sets, we assembled a smaller TMA with 31 samples overlapping (Table 3). Interestingly, TCF21 hypermethylation and reduced TCF21 protein expression were sometimes discordant (Table 3), suggesting that mechanisms other than hypermethylation could result in decreased protein expression.

Table 3. Methylation, LOH, and Protein Expression Data for TCF21 in Lung Cancer Samples
Sample IDHistology% MethylationHypermethylatedLOHNo. of HitsTCF21 Protein Expression on TMA
  1. Adeno, adenocarcinoma; SCC, squamous cell carcinoma; nd, not done; ROH, retention of heterozygosity; LOH, loss of heterozygosity; No. of Hits, total of hypermethylation and LOH events (if both were assayed): 0 is neither, 1 is either hypermethylation or LOH, and 2 is both; TMA, tissue microarray.

462Adeno37.21Yesnd Negative
612Adeno32.74Yesnd High
645Adeno31.39Yesnd Negative
759Adeno35.89Yesnd Low
801Adeno51.64Yesnd Low
842Adeno28.46Nond Low
758SCC43.11Yesnd Intermediate
870SCC20.62Nond High
793_2SCC22.67Nond Intermediate

TCF21 Loss of Heterozygosity and Sequence Analysis

Because some NSCLC samples showed loss of TCF21 protein expression without hypermethylation and the average levels of TCF21 hypermethylation were approximately 40%, which might not be expected to completely abolish protein expression, we examined potential “second hits” at the TCF21 locus (Table 3). First, we examined loss of heterozygosity (LOH) in 33 of the paired samples, using 4 microsatellite markers spanning the TCF21 locus and closely flanking region. LOH was seen in 14 of these samples (42%), with no significant differences between samples with and without hypermethylation (P = .172). In addition to LOH, we sequenced the TCF21 coding region in 45 lung cancer samples that showed either 0 or 1 hit by methylation or LOH analysis. Samples with both hypermethylation and LOH were not sequenced. No TCF21 coding mutations were found.

Reduced TCF21 Protein Expression Is Widespread and Independent of Stage and Other Clinical Features, But Correlated With Histology

To determine whether TCF21 expression was correlated with clinical features such as sex, race, stage, smoking status, histology, or prognosis, we performed univariate analysis (Table 2). Histology and TCF21 expression showed significant correlation (P = .003), as did smoking status (P = .048) and sex (P = .021). In a multivariate analysis with histology, sex, and smoking status, only histology was statistically significantly (P = .007) associated with TCF21 level, whereas smoking history and sex were not independently associated. Cox proportional hazards analysis was performed to assess the association between TCF21 with overall survival and recurrence-free survival, but neither association was significant in either a multivariate model or a univariate model (data not shown).

Given previously reported associations between smoking, sex, and histology with EGFR status,19 we then analyzed the 202-patient subset for which EGFR status was known for associations with TCF21 expression. When only adenocarcinomas were considered (n = 172), EGFR status was not associated with TCF21 expression (P = .138), nor was EGFR status associated with TCF21 expression in a univariate analysis with all 202 patients (P = .241). Therefore, the only significant correlation (P = .007) is that adenocarcinomas have lower TCF21 expression than do SCCs, although all histologies have significantly lower TCF21 levels than do normal tissue.


TCF21 Has the Highest Frequency of Promoter Hypermethylation in NSCLC of Any Gene Known to Date

Many genes have been reported to be hypermethylated in NSCLC.20-24 However, the frequency of these events has not been high enough in all NSCLC subtypes for utilization as a screening tool, requiring combinations of genes to approach a sensitivity high enough for a screening test. Despite numerous reports of hypermethylated genes in NSCLC, identified by a variety of approaches, none has a reported frequency of hypermethylation as high as TCF21, except for one that also examined TCF21 itself, and a recent publication limited to only the SCC subtype of NSCLC.20-23 This study was specifically focused on TCF21 in NSCLC and the susceptibility locus at 6q23-q25. Among 43 genes selected in the region, TCF21 had the highest rate of cancer-specific hypermethylation (81%),23 exactly matching our rate of TCF21 hypermethylation.

The high rates (80%-85%) of TCF21 promoter hypermethylation and decreased protein expression are high enough for TCF21 to be considered for development as a screening biomarker, either by increased methylation or decreased protein level. The sensitivity of TCF21 hypermethylation/decreased TCF21 protein expression compares favorably with that of prostate-specific antigen, the current screening biomarker for prostate cancer, which has been shown to be <4 (ie, in the normal range) in 15% of men with prostate cancer, a sensitivity of 85%.25 Of course, one of the main difficulties in lung cancer screening remains the acquisition of relevant tissue (in this case early lung tumors), but detection of TCF21 hypermethylation has been reported in biopsies and sputum samples, which is promising.26 If the sensitivity of TCF21 in sputum/bronchial brushings were not high enough to be used alone, it could be used as part of a panel of screening biomarkers.

Detection of TCF21 Hypermethylation by Highly Quantitative Method

One significant advantage of methylation detection by pyrosequencing-based methylation analysis (PMA) following bisulfite conversion is that quantitative levels can be measured across multiple sites, rather than the more qualitative output obtained with methylation-specific PCR (MS-PCR) or other qualitative or semiquantitative methods (eg, COBRA). PMA enabled us to reliably detect a difference between the 20% average methylation in N tissue and the 40% average methylation in T tissue. This difference would likely not have been detected with less quantitative methylation detection strategies. It is possible that other genes known to be hypermethylated in NSCLC may prove to be more sensitive and/or specific if more quantitative methods such as pyrosequencing are routinely applied. The 40% methylation level in NSCLC tissue raises the question of whether only 1 of the 2 TCF21 alleles is silenced by hypermethylation or whether 40% of cells have both alleles silenced, either of which could produce the observed result. It is interesting that hypermethylation of 40% of alleles is frequently associated with completely absent TCF21 protein expression, suggesting either that the second allele is silenced by a different mechanism than hypermethylation or that there is a threshold level of gene expression necessary to produce detectable TCF21 protein levels.

Reduction of TCF21 Protein Levels Similar to TCF21 Hypermethylation Rates

In addition to TCF21 hypermethylation, we also examined the downstream effect of this hypermethylation by examining protein expression directly. In both cases, we found TCF21 hypermethylation/decreased TCF21 protein levels at similar rates—81% and 84%, respectively. Given that decreased mRNA expression of TCF21 has been shown to result from promoter hypermethylation,10 the similar rates of hypermethylation and decreased protein expression are consistent with the notion that decreased mRNA expression results in decreased protein expression. However, because there were cases with low/absent protein expression despite normal TCF21 methylation levels, other regulatory mechanisms likely are in effect. LOH occurs at a rate of 42%. That LOH occurs in at least a few cases without TCF21 hypermethylation suggests inactivation of TCF21 in other ways. Because we did not detect any coding mutations, these could be promoter or other regulatory region DNA mutations. Alternatively, dysregulation by micro-RNA could be a factor. Interestingly, the sole predicted regulator of TCF21 is miR-92a,27 which is overexpressed in a variety of cancers.28, 29

TCF21 Is an Excellent Candidate Biomarker for Early Lung Cancer Detection

Several characteristics of TCF21 make it an attractive target for screening efforts in NSCLC. First, it is hypermethylated at similar frequencies in all histologic subtypes of NSCLC examined, including early- and late-stage cancers. Second, it has a higher frequency of hypermethylation than that of any other gene published to date in NSCLC, without subdivision by histologic subtype.10, 20-23 This high sensitivity is combined with a high specificity as well. We detected a false-positive rate of only 1 in 42 samples with NAT, for a specificity of 98%. In other reported control tissues, such as PBMCs and human bronchial epithelial cells (HBECs) from smokers, there were no false-positives (n = 20 in each case).23 The high specificity in normal adjacent tissue is especially noteworthy, in that there appears to be no evidence for a “field effect,” which can complicate screening in smokers, who often have cancers arising in a field of premalignant lesions, leading to false-positive screening results. Instead, the very low prevalence of TCF21 hypermethylation in NAT that we report suggests that TCF21 hypermethylation is restricted to cancerous tissue only.

In summary, we have established that TCF21 hypermethylation and reduced TCF21 protein are ubiquitous in NSCLC, occurring in 80%-85- of tumors across a wide variety of stages, histologies, and other clinical characteristics. Given the high rate of increased methylation and decreased protein expression, combined with its lack in normal adjacent tissue, we propose that TCF21 is an excellent candidate biomarker for further development as a lung cancer screening tool.


We thank Tamer Ahmed for technical assistance and Mario Sirito for helpful discussions.


Supported in part by the Kleberg Foundation, DoD W81XWH-05-2-0027, and NIH-NCI P01 CA34936 (to R.K.). These agencies had no involvement in the study design; in the collection, analysis, and interpretation of data; in writing of the manuscript; and in the decision to submit the manuscript for publication.