The incidence of esophageal adenocarcinoma is increasing at an alarming rate in western countries.1 Nearly all such cancers arise from Barrett's esophagus, a metaplastic condition in which intestinal-type epithelium replaces the normal squamous epithelium.2 The molecular mechanisms of this premalignant condition remain incompletely understood. Inactivation of tumor suppressor genes by promoter hypermethylation has been considered as a potentially important mechanism involved in the development of Barrett's esophagus.3 Several tumor suppressor genes, most notably p16, have been found to undergo epigenetic changes in Barrett's esophagus.3, 4, 5, 6
Aberrant promoter methylation of a new class of tumor suppressor genes,7 secreted frizzled-related protein (SFRP) genes, has been described in colon and gastric cancers.8SFRPs are secreted glycoproteins that work as inhibitive modulators of a putative tumorigenic pathway—the Wnt pathway, in which signals are transduced by Wnt ligands through frizzled (Fz) membrane receptors by competition with Fzs for Wnt ligands or by direct formation of nonsignaling complexes with Fzs themselves.9 Five SFRP genes (SFRP1-5) have been identified. SFRP underexpression and overexpression have been respectively associated with inhibition or acceleration of apoptosis in various disease states.10, 11, 12, 13, 14, 15, 16 As opposed to upregulation of SFRPs in apoptosis, abnormally low expression of SFRPs has been observed in surgically resected tissues from breast and cervical cancers.10, 11, 17, 18 Recently, frequent methylation associated with loss of SFRP expression has been observed in some cancers,8, 19 but SFRP methylation has not been studied in esophageal adenocarcinoma or in its precursor, Barrett's esophagus.
Our study was designed to (i) determine whether SFRP gene methylation is associated with oncogenesis in Barrett's esophagus; (ii) evaluate the potential of methylated SFRP genes to serve as biomarkers for Barrett's esophagus and cancer; and (iii) explore the functional role of methylation in silencing SFRP gene expression in Barrett's cancer.
AN, adjacent normal epithelium; BE, Barrett's esophagus; CA, adenocarcinoma; DAC, demethylation agent 5-aza-2′deoxycytidine; IHC, immunohistochemistry; MSP, methylation-specific PCR; NN, normal epithelium from normal individuals; RT-PCR, reverse-transcription polymerase chain reaction; SFRP, secreted frizzled-related proteins; TSA, trichostatin A.
Material and methods
In our study, 135 formalin-fixed and paraffin-embedded tissues were collected, including 40 esophageal adenocarcinomas, 37 Barrett's lesions, and 58 normal esophageal and gastric epithelia. Cancer and Barrett's samples were surgically resected specimens. Of the 58 normal epithelial samples, 28 were surgically resected specimens from Barrett's patients and 30 were biopsies from normal individuals. The samples were collected to represent each phase of the tumorigenesis of Barrett's esophagus. The clinical characteristics of these tissues are shown in Table I. All samples were obtained from patients diagnosed from 2001 to 2003 at Mayo Clinic (Rochester, MN). Our study was approved by the Institutional Review Board of Mayo Foundation.
Table I. Clinical Characteristics
Number of samples
Age (range, median)
Pathology and other information
From Barett's patients
From normal individuals
Three esophageal adenocarcinoma cell lines, including Bic-1, Seg-1 (gifts from Dr. David G. Beer, Department of Surgery, University of Michigan Medical School, Ann Arbor, Michigan) and OE33 (a gift from Dr. Deborah A. Lebman, Department of Microbiology and Immunology, Virginia Commonwealth University, Richmond, Virginia) were used in our study. These cells were derived from Barrett's-associated adenocarcinoma of the distal esophagus. Bic-1 and Seg-1 were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, and OE33 was grown in RPMI 1640 supplemented with 10% fetal bovine serum.
Microdissection and DNA extraction
Tissue sections were reexamined by a pathologist (L.J.B.) who circled histologically distinct lesions, including adenocarcinoma with different grades, Barrett's with different degrees of dysplasia and normal epithelia. Ten-micron paraffin-fixed tissue sections with histologically distinct lesions were carefully microdissected. Tissue and cell line DNA were extracted using Qiagen DNA Mini Kit (Qiagen, Valencia, CA).
Sodium bisulfite treatment, which converts unmethylated but not methylated cytosine residues to uracil, was performed as follows. Briefly, 1 μg of genomic DNA was denatured with 0.3 M NaOH in a total volume of 55 μL for 20 min at 42°C. A 600 μl mix of 10 mM hydroquinone (Sigma, St. Louis, MO) and 4 M sodium bisulfite (Sigma) was added to each denaturing reaction, and the mixture was incubated at 55°C for 4 hr in the dark. Five microliters of Glassmilk (Qbiogene, Carlsbad, CA) were added to bind the DNA for 10 min at room temperature. Glassmilk pellets were washed by 70% ethanol 3 times. Then 20 μM NaOH/90% ethanol was added to desulfate the bisulfite-treated DNA for 5 min at room temperature. After pelleting the Glassmilk, 90% ethanol was added to wash the DNA twice. After the last washing, ethanol has been completely removed; the bisulfite-treated DNA was eluted in 30 μL TE buffer (pH 7.5).
Methylation-specific PCR (MSP)
The methylation of 4 SFRP members, SFRP1, 2, 4 and 5, was explored. SFRP3 was not studied because it does not have a CpG island at the 5′ area.8 MSP was performed using a previously described method.20 Briefly, 1 μl bisulfite-modified DNA was amplified in a total volume of 25 μl containing 1× PCR buffer (Perkin-Elmer, Boston, MA), 1.5 mM MgCl2, 100 μM of each dNTP, 200 nM of each primer and 1 unit of AmpliTaq Gold polymerase (Perkin-Elmer). Amplification included hot-start at 95°C for 12 min, denaturing at 95°C for 30 sec, annealing at certain temperatures for 30 sec, extension at 72°C for 30 sec for 35 cycles and a final 10 min extension step. Bisulfite-treated human genomic DNA (Novagen, Madison, WI) and CpGenome™ Universal Methylated DNA (Chemicon, Temecula, CA) were employed as positive controls for unmethylation and methylation, respectively. All PCR reactions have been performed twice. Methylated (M) and unmethylated (U) primers and annealing temperatures are shown in Table II.
Table II. Primer Sequences and Related Information for SFRP Assays
Primer sequence (5′→ 3′)
PCR products were purified using Wizard® SV Gel and PCR Clean-Up System (Promega, Madison, WI) and then confirmed by automated sequencing in an ABI Prism 377 DNA Sequencers (Perkin-Elmer).
Bisulfite genomic sequencing
Bisulfite genomic sequencing was carried out according to reported methods.21, 22, 23 Bisulfite-modified DNA was amplified using primers (Table II) that flank the MSP region of each SFRP gene. One microliter bisulfite-modified DNA was amplified in a total volume of 25 μl containing 1× PCR buffer (Perkin-Elmer), 3.0 mM MgCl2, 100 μM of each dNTP, 200 nM of each primer and 1 unit of AmpliTaq Gold polymerase (Perkin-Elmer). Amplification included 95°C for 10 min, denaturing at 95°C for 30 sec, annealing at certain temperatures for 30 sec, extension at 72°C for 1 min for 38 cycles and a final 10 min extension step. The annealing temperature and the length of PCR products were listed in Table II. PCR products were purified using Wizard® SV Gel and PCR Clean-Up System (Promega) and then sequenced in an ABI Prism 377 DNA Sequencer (Perkin Elmer). The methylation levels of individual CpG sites were expressed as percentage of 5-methyl cytosine among the whole cytosine population (methylated plus unmethylated) at the same site.22 A single C at the corresponding was considered full methylation, a single T as no methylation and overlapping C and T as partial methylation.22
IHC was employed to study the protein expression. ABC immunostaining method (Vector Laboratories, Burlingame, CA) was performed according to the instruction of the manufacturer. Sections (5 μm thick) were cut and mounted on coated slides and then dewaxed and rehydrated in a xylene-ethanol series. Endogenous peroxidase activity was blocked by incubation in 0.3% H2O2 for 30 min. After washing in buffer for 5 min, slides were incubated in 2.5% normal blocking serum for 20 min and then incubated in polyclonal SFRP1 rabbit antibody (1:2,000, a gift from Dr. Pascale Dufourcq, Biochemistry Laboratory, University of Bordeaux, Pessac, France),24 polyclonal SFRP2 rabbit antibody (1:100, Santa Cruz Biotechnology, Santa Cruz, CA) or monoclonal SFRP4 antibody (1:2,000, a gift from Dr. Rajiv Kumar, Department of Medicine, Mayo Clinic, Rochester, Minnesota) for 30 min. SFRP5 antibody was not available. Slides were washed again and incubated in biotinylated secondary antibodies for 30 min. Then R.T.U. Vectastain Elite ABC reagent (Vector Laboratories) was added and incubated for 30 min after washing. Results were visualized in peroxidase substrate solution until desired stain intensity developed. After rinsing in tap water, sections were counterstained with hematoxylin, dehydrated and mounted. The immunostaining condition of each antibody has been tested in multiple sections of several samples. Previously confirmed positive and negative sections were stained at the same time as controls for each batch of slides. A section was considered as positive when more than 10% pathologically distinctive epithelial cells had at least a moderate staining intensity.25, 26, 27, 28
Demethylation and inhibition of histone deacetylation
The methylation status of SFRP genes in 3 esophageal adenocarcinoma cell lines was determined by MSP and bisulfite genomic sequencing. These cells were split to low density in 6-well plates, grown 12–24 hr and then treated using the following: (i) 5-Aza-2′-deoxycytidine (DAC; 5 μM; Sigma) or phosphate-buffered saline for 96 hr. Medium containing DAC or phosphate-buffered saline was changed every 24 hr. (ii) Trichostatin A (TSA; 300 nM for Bic-1, 150 nM for Seg-1 and 100 nM for OE33) (ICN Biomedicals, Irvine, CA) or an identical volume of ethanol as mock for 24 hr. (iii) DAC (5 μM) for 72 hr followed by TSA at concentrations mentioned above for an additional 24 hr. The dose and timing of DAC and TSA were based on prior tests showing optimal reactivation of gene expression, as well as published studies.29, 30
Real-time quantitative RT-PCR analysis
The mRNA expression of the 4 SFRP genes in the esophageal adenocarcinoma cell lines, which had received different combinations of DAC and/or TSA treatment or mock treatment, was quantified. Briefly, RNA from these cells was extracted with RNeasy Mini Kit (Qiagen) according to the manufacturer's protocol and quantified by spectrophotometer. Reverse transcription was performed on 2 μg total RNA using Omniscript RT Kit (Qiagen). The cDNA was amplified using a real-time PCR iCycler (BioRad, Hercules, CA). GAPDH was employed as an internal reference gene to normalize cDNA input.15, 31 The RT-PCR primers of SFRP genes and GAPDH were listed in Table II. The mRNA expression of SFRPs was defined as the ratio of the fluorescence emission intensity value of SFRP RT-PCR products to that of GAPDH PCR products, multiplied by 100. Real-time PCR assays were performed in a reaction volume of 25 μL containing 1× iQ™ SYBR® Green Supermix (BioRad), 40 nM each primer and 1 μL cDNA under the following conditions: 95°C for 3 min, followed by 40 cycles of 95°C for 15 sec and certain annealing temperatures for 45 sec.
Amplification was carried out in 96-well plates. Each plate consisted of bisulfite-treated DNA samples and multiple water blanks, as well as positive and negative controls. Each assay was performed in duplicate. Serial dilutions of positive controls were used to make standard curves for all plates. Melting curves were made after each PCR reaction to guarantee that only one identical product was amplified for all samples. Automated sequencing was also employed to confirm the PCR products.
χ2 test was used to analyze the different methylation frequencies of SFRP genes among 4 groups of esophagastric tissues (Table III) and the SFRP protein expression difference between cancers and Barrett's (Table IV). Fisher's exact test was performed to analyze the association of SFRP protein expression with cancer stage, grade or Barrett's dysplastic degree (Table IV) and the correlation of SFRP expression with methylation status (Table V). Significance was defined as p < 0.05. Statistical tests were performed using SAS statistical software (SAS Institute, Cary, NC).
Table III. Methylation of SFRP Genes in Different Esophageal Tissues
The differences between cancer and Barrett's esophagus were statistically significant for SFRP1, 2 and 4 proteins (p < 0.001).
SFRP1 had lower expression in Barrett's with higher dysplasia grade; SFRP2 had lower expression in cancer with higher grade and stage and Barrett's with higher dysplasia grade; and SFRP4 had lower expression in cancer with higher grade.
Table V. The Correlation of SFRP Expression with Methylation Status
N.S., not significant.
Methylation of SFRPs during the progression of Barrett's esophagus
SFRP1, 2, 4 and 5 were methylated in 93% (37/40), 83% (33/40), 73% (29/40) and 85% (34/40) of esophageal adenocarcinomas; 81% (30/37), 89% (33/37), 78% (29/37), 73% (27/37) of Barrett's tissues; 25% (7/28), 64% (18/28), 32% (9/28) and 21% (6/28) of normal mucosa adjacent to Barrett's tissues; 10% (3/30), 67% (20/30), 0% (0/30) and 13% (4/30) of normal biopsies from healthy individuals, respectively (Fig. 1; Table III). The methylation of each of the SFRP1, 2, 4, and 5 genes was significantly different in the 4 groups of tissues, including cancer, Barrett's, normal epithelia adjacent to Barrett's and normal epithelia from normal individuals (p < 0.001 for SFRP1, 4 and 5; p = 0.04 for SFRP2). Comparing methylation of individual SFRP genes, SFRP1, 4, 5 methylation differed significantly between cancer and adjacent normal epithelia (p = 0.001 for SFRP1, p < 0.001 for SFRP4 and 5), between cancer and normal epithelia from normal individuals (p < 0.001 for SFRP1, 4 and 5), between Barrett's and adjacent normal epithelia (p < 0.001 for SFRP1, 4 and 5), between Barrett's and normal epithelia from normal individuals (p < 0.001 for SFRP1, 4 and 5) and between normal epithelia adjacent to Barrett's and normal epithelia from normal individuals (p < 0.001 for SFRP4). No significant difference was found between cancer and Barrett's.
SFRP immunostaining in esophagogastric tissues
All SFRP proteins had expression on cellular membranes, but expression patterns of individual SFRP proteins were distinctive in different tissues. SFRP1 protein was abundantly expressed in both cytoplasmal and secretory granules; SFRP2 expression was intense on membrane but sparse in secretory granules; and SFRP4 expression was strong on basal and lateral membrane but weak on surface membrane (Fig. 2). SFRP4 expression was more intense in stroma than in epithelial cells, but stromal staining was not considered as positive signal in evaluating the expression status of epithelial cells. Because normal epithelial specimen both from Barrett's patients and from normal individuals showed the same SFRP expression patterns, but epithelia from esophagus and stomach exhibited different SFRP expression levels, we regrouped normal epithelia both from patients and from normal individuals into 2 types—31 normal gastric epithelia and 27 normal esophageal epithelia—by histologic origins. SFRP protein expression patterns differed in these 2 types of normal epithelia. SFRP1 was expressed in all normal gastric epithelia (100%, 31/31) but never in normal esophageal epithelia (0%, 0/27); SFRP2 was expressed in all gastric (100%, 31/31) and 78% (21/27) of esophageal epithelia; however, SFRP4 was always negative in gastric epithelial cells (0%, 0/31), but positive in all esophageal epithelia (100%, 27/27) (Fig. 2, Table IV). In neoplastic specimens, expression of SFRP1, 2 and 4 proteins was noted in 13% (5/40), 33% (13/40) and 10% (4/40) of cancers and in 50% (18/36), 83% (30/36) and 53% (19/36) of Barrett's, respectively (Fig. 2, Table IV). Significance difference between cancers and Barrett's was found for each of these 3 SFRP proteins (p < 0.001 for SFRP1, 2 and 4). When subtyping cancers by differentiation and stage and Barrett's by dysplastic degree, SFRP1 had lower expression in Barrett's with higher dysplasia grade (p = 0.05); SFRP2 had lower expression in cancer with higher grade (p = 0.002) and stage (p = 0.006) and Barrett's with higher dysplasia grade (p = 0.006); and SFRP4 had lower expression in cancer with higher grade (p = 0.006, Table IV).
When the frequency of SFRP protein expression was compared against the methylation status of corresponding genes in esophageal adenocarcinoma, SFRP2 and 4 expression was more frequently lost in cancers with methylation (p = 0.03 for SFRP2 and p = 0.001 for SFRP4, Table V). Significant difference was not observed with SFRP1. Those samples that were methylated by MSP and showed no loss of protein expression were reexamined by bisulfite genomic sequencing, and methylation was confirmed.
Re-expression of silenced SFRPs in esophageal adenocarcinoma cell lines
SFRP1 and 2 genes were methylated in all 3 cell lines, SFRP4 was not methylated in SEG-1 and SFRP5 was not methylated in BIC-1. Methylation status was determined by MSP and confirmed by bisulfite genomic sequencing (Fig. 3a). When SFRP genes were methylated, their corresponding mRNA expression was absent or faintly detected (Fig. 3b,c). However, after cells were treated by DAC demethylation, the silenced SFRP mRNA expression could be re-expressed or increased in most instances (Fig. 3b,c). The addition of TSA to DAC re-expressed SFRP2 and SFRP4 mRNA in OE33, which could not be induced by DAC alone, and markedly further elevated the mRNA expression level that had been re-expressed or increased by DAC (Fig. 3b,c). The effect of the DAC/TSA combination on SFRP5 expression was an exception (Fig. 3b,c). The influence of a single application of TSA on SFRP expression was minimal (Fig. 3b,c). SFRP4 was strongly expressed in SEG-1 that did not have SFRP4 methylation and SFRP5 was strongly expressed in BIC-1 that did not have SFRP5 methylation; the effect of DAC and/or TSA on their expression was minimal or negative (Fig. 3d).
Although various tumor suppressor genes have been reported to be methylated in Barrett's esophagus and esophageal adenocarcinoma,3, 4, 5, 6, 32 methylation of SFRP genes is a new observation and is notable for the extremely high frequency of its occurrence in both esophageal cancer and Barrett's. This investigation demonstrates that genes of the SFRP family are methylated in most cases of esophageal adenocarcinoma and its precursor, Barrett's esophagus. The promoter regions of SFRP genes were methylated in 73–93% of esophageal cancers and in 73–89% of Barrett's epithelia. However, SFRP methylation (excluding SFRP2) was much less common (21–32%) in normal-appearing epithelia adjacent to Barrett's and least common (0–13%) in normal esophageal or gastric epithelia from individuals without Barrett's. These data suggest that methylation of SFRP genes occur as an early event in the evolution of Barrett's esophagus and are preserved or increased through carcinogenic transformation.
As recently reviewed,33 biomarkers are needed for Barrett's esophagus and associated adenocarcinoma to facilitate early detection. The large majority of persons presenting with symptomatic esophageal adenocarcinoma are unaware that they harbored premalignant Barrett's esophagus for years prior to their cancer diagnosis, and simple means are needed to identify those in the general population with Barrett's esophagus if preventive interventions are to meaningfully impact cancer control. Because they are methylated commonly and early in Barrett's esophagus, SFRP genes represent potential markers for noninvasive screening. Methylated tumor suppressors34, 35 and genetic alteration, such as mutations and microsatellite instability,36, 37, 38 have been recovered in body fluids and have been considered as molecular markers for the early detection of human cancers, but their application has been limited due to low sensitivity when single markers have been considered. In contrast, SFRP gene methylation occurs in the majority of Barrett's related cancer. Recently, methylated SFRP2 was suggested as a marker candidate for colorectal cancer screening, as it was recovered in stool samples from a majority of patients with colorectal cancer.39 In our study, SFRP1, 4 and 5 appeared to be more discriminant for Barrett's esophagus and esophageal adenocarcinoma and may be more suitable candidate markers for these lesions. Subsequent studies will need to be conducted to determine their sensitivity in detecting Barrett's esophagus by stool assay or other approaches (e.g., esophageal swabbing).
Based on our study, methylation represents a likely mechanism of SFRP gene silencing in esophageal adenocarcinoma. Because Wnt/β-catenin pathway is associated with the tumorigenesis and SFRP proteins inhibit this important pathway by binding to both Wnt the ligand and frizzled the receptor,9 deficiency of SFRP protein could promote tumor formation. Recently, methylation was associated with the silencing of SFRP genes; however, its role in Barrett's cancer has not previously been explored. Our study suggests that methylation downregulates SFRP expression in esophageal adenocarcinoma based on the following observations: (i) SFRP protein expression is absent or markedly decreased in Barrett's esophagus and cancer but generally high in normal epithelia, which is the converse of SFRP gene methylation levels; (ii) SFRP protein expression is lower in esophageal cancer tissues with corresponding SFRP gene methylation than in those without SFRP methylation; (iii) mRNA expression of SFRP genes is minimal or absent in cancer cell lines with corresponding SFRP gene methylation but high in cell lines with unmethylated SFRP genes; and (iv) SFRP gene expression is reactivated in methylated cancer cell lines treated by demethylation. However, the protein expression of SFRP1 was not associated with the methylation status. Two potential explanations for this apparent discrepancy include (i) genetic alterations, such as mutation, may also play a role in silencing SFRP1 gene expression;19 and (ii) sample size of the unmethylated group was small and may not have provided adequate statistical power (only 3 SFRP1 umethylated cancer tissues).
Although there were similarities in methylation and expression levels among SFRP genes during the progression of Barrett's esophagus when evaluated in tissue homogenates, immunostaining revealed distinctive in situ protein expression patterns for each member of the SFRP family. SFRP1, 2 and 4 all had expression on cellular membrane of epithelial cells, which supports that SFRP's inhibitive function on Wnt pathway occurs on membrane.9 However, SFRP1 displayed intense staining in secretory granules, and SFRP4 protein was concentrated in stromal cells. Furthermore, in normal mucosa, SFRP1 was exclusively secreted by gastric glandular epithelia but not by esophageal squamous epithelia; SFRP4 was only expressed by esophageal squamous epithelia but not by gastric epithelia; and SFRP2 was expressed by both esophageal and gastric epithelia. The markedly different expression of SFRP proteins in 2 types of epithelia further suggests that these 3 SFRP proteins have distinctive functions and act differently in various types of epithelia. The cytoplasmic staining of SFRP proteins may reflect their precursor in cytoplasm before secretion because proteins are translated and processed in cytoplasm.
In our study, downregulation of SFRP protein expression was associated with higher dysplasia grade with Barrett's esophagus and with higher stage and grade with cancers, consistent with the stage association observed with other types of cancer.40, 41 This observation suggests that SFRP expression is lost progressively during the transition from normal epithelium to Barrett's esophagus and then to invasive cancer.
Our study did not explore genetic alterations of SFRP genes. However, SFRP gene mutations appear to be less common in cancer.19 For example, it is epigenetic inactivation of SFRP genes through methylation that allows constitutive Wnt signaling changes in colorectal cancer,42 and SFRP gene mutations do not appear to play a significant role.19
In summary, aberrant promoter methylation results in downregulation of SFRP gene expression and occurs commonly in Barrett's esophagus and esophageal cancer. SFRP methylation occurs as early as the transformation of normal epithelium to Barrett's esophagus. Methylated SFRP 1, 4 and 5 genes represent potentially discriminant tumor markers for Barrett's esophagus and its related neoplasia.
The authors thank Dr. P. Dufourcq for SFRP1 antibody, Dr. R. Kumar for SFRP4 antibody, Dr. D.G. Beer for SEG-1 and BIC-1 cells and Dr. D.A. Lebman for OE33 cell.