Decreased Expression of Gene Cluster at Chromosome 1q21 Defines Molecular Subgroups of Chemoradiotherapy Response in Esophageal Cancers

Purpose: Resistance to preoperative chemoradiotherapy (CTXRT) in 75% of patients with esophageal adenocarcinoma (EAC) underscores the need for identification of biomarkers of CTXRT response. We previously noted an association between decreased expression of epidermal differentiation complex (EDC) genes S100A2 and SPRR3 at chromosome 1q21 and CTXRT resistance. In the current study, we did an in-depth investigation of the expression of 1q21-1q25 region genes to uncover the role of the EDC and its flanking genes in CTXRT response.

Experimental Design: We compared 19 pretreatment EAC specimens with normal squamous mucosa for the expression of 517 genes at chromosome 1q21-1q25 and selected target genes based on their differential expression. Using the pathologic complete-response (pathCR) status of the resected specimens as a representation of CTXRT sensitivity, we assessed the association between the expression of target genes and CTXRT response and clinical outcomes.

Results: On the basis of the expression levels of IVL, CRNN, NICE-1, S100A2, and SPPR3, genes within and in close proximity to the EDC, cancers were segregated into high (subgroup I) or low (subgroup II) expressers. Four of the five pathCRs were high expressers. Thus, low expressers, with one exception, were all nonresponders. Patients in subgroup I also had longer survival than those in subgroup II, although this result was not statistically significant owing to the small study number.

Conclusions: The expression levels of genes mapping within and close to the EDC define CTXRT response subgroups in EACs.

Esophageal cancer remains one of the most fatal malignancies, with a 5-year survival rate of <20%. Over the past two decades, a remarkable shift has occurred in the epidemiology of esophageal cancer, resulting in an alarming increase in the incidence of adenocarcinoma of the proximal esophagus, with a relative decline in the incidence of esophageal squamous cell carcinoma (1). The incidence of esophageal adenocarcinoma (EAC) is increasing faster than that of any other type of cancer, at a yearly rate of ∼10%, and EAC ranks in the top 15 cancers among Caucasian men in the United States (1–3).

Most EACs arise in the background of Barrett's esophagus, which is characterized by the replacement of normal squamous epithelium with metaplastic columnar epithelium due to chronic reflux esophagitis (4). When EAC is local-regional, preoperative chemoradiation (CTXRT) is commonly recommended. However, the role of preoperative CTXRT remains controversial. When CTXRT is used, longer survival is noted in a small fraction (about 27%) of patients who achieve pathologic complete response (pathCR; refs. 5–9); however, patients who are likely to have a pathCR cannot be predicted by the pretreatment clinical parameters. Thus, it is of paramount importance to discover biomarkers that can predict response to CTXRT to individualize and optimize therapy for this group of patients. We believe that individualization of therapy may be possible by studying the molecular biology of EAC and the patient's genetics.

In an expression profiling pilot study, designed to identify molecular signatures predictive of response to therapy, we found substantially lower expression of the S100A2 (S100 calcium-binding protein A2) and SPRR3 (small proline-rich protein 3 or esophagin) genes in pretreatment cancer specimens of patients resistant to CTXRT (10). The loss of expression of both S100A2 and SPRR3 has been associated with premalignant and malignant states, including cancers of the lung, esophagus, and cervix (11–18). Both genes are located in an evolutionarily conserved genetic cluster, designated as the epidermal differentiation complex (EDC), at the 1q21 chromosomal band (19–21). This genetic region encompassing 2 Mb of genomic DNA harbors 3 gene families and approximately 43 genes that are involved in terminal squamous differentiation of the human epidermis (19). A recent study has suggested that the expressional down-regulation of some EDC complex genes, including S100A2 and SPRR3, may serve as a potential marker of progression to EAC (22).

We hypothesized that the EDC gene cluster is differentially expressed in CTXRT-sensitive and CTXRT-resistant EAC, and that this may reflect distinct biological EAC entities. To test this hypothesis, we compared pretreatment specimens of EAC and normal squamous esophageal mucosa for the expression of approximately 517 genes in the 1q21-1q25 chromosomal region that included the EDC cluster and its flanking regions and selected target genes based on their maximal differential expression. Then, using pathCR after therapy as an end point, we analyzed the association between the expression levels of 1q21-1q25 target genes and response to CTXRT and clinical outcome.

Patient selection and evaluation. Nineteen patients, including 16 from a previous report (10), with localized, histologically confirmed adenocarcinoma of the thoracic esophagus were included in the study. All 19 patients participated in a clinical trial approved by The University of Texas M. D. Anderson Cancer Center's Institutional Review Board. Patients with tumors classified as T2-3 with any N, patients with M1a cancer (celiac nodes associated with a gastroesophageal junction carcinoma), and patients with T1N1 carcinoma were considered eligible. All patients were evaluated before registration by a multidisciplinary team that included thoracic oncology surgeons, radiation oncologists, gastroenterologists, and medical oncologists. To be eligible, patients had to have cancer that was considered technically resectable. Patients with any evidence of metastatic cancer were not enrolled.

Treatment. The clinical protocol included three chemotherapy agents (docetaxel, irinotecan, and 5-fluorouracil) given before and during the preoperative radiotherapy regimen. Approximately 5 to 6 weeks following the end of chemoradiotherapy, patients were restaged fully, and surgery was done if they had no metastatic cancer and no contraindications for surgery. If a patient had an R0 resection, no further therapy was planned. Patients who underwent an R1 resection (microscopic carcinoma at the margin) or R2 resection (gross carcinoma after surgery) or who had M1 disease were offered palliative care.

Each patient was assessed at 3, 6, 9, and 12 months after surgery and then every 6 months for 2 additional years and then every year or until death. Local-regional recurrence was defined as recurrence within the surgical field or mediastinal nodes. Metastatic cancer was defined as evidence of cancer outside the regional area, such as in the bone, brain, liver, or lung.

Tissue specimens and collection. All tissue specimens were obtained during a diagnostic preoperative endoscopic procedure through a protocol approved by the M. D. Anderson Cancer Center Institutional Review Board and after informed consent was obtained from patients. Both normal squamous mucosal (NSM) tissue and cancer tissue were collected from each patient. The size of a typical biopsy specimen was ∼2.0 mm. Biopsy specimens were placed in cryogenic vials, snap frozen in liquid nitrogen, and stored at −80°C until further use. In this report, pretreatment cancers from 19 patients and NSM tissue from 7 of these patients with high-quality RNA were analyzed.

Histologic evaluation. For each specimen analyzed by microarray, an adjacent tissue biopsy was given to a pathologist (TTW or AR) to confirm the presence of cancer and its histology. Routine H&E-stained slides were used to evaluate the presence of cancer in pretreatment endoscopic biopsies and esophagectomy specimens. The pathologic response in the resected esophagus was assigned to one of two categories: no residual carcinoma in the esophagus (pathCR) or the presence of any cancer cells in the resected specimen (<pathCR). In this cohort of patients, all except three patients underwent surgical resection and, thus, had pathologic response data.

Oligonucleotide microarray analysis. A small sample protocol with 200 ng RNA using a second round of amplification was used to generate biotin-labeled cRNA, as described previously (10). Fifteen micrograms of cRNA was then fragmented and hybridized to the Affymetrix U133A GeneChip (Santa Clara, CA) as per the manufacturer's instructions. Generation of cRNA, hybridization of biotin-labeled cRNA to the oligonucleotide arrays, and image analysis were done according to protocols established in our institution's DNA Microarray Core Facility.

The Affymetrix U133A GeneChip contains 22,215 probe sets that correspond to 14,593 well-characterized human genes.⁷

Microarray Suite 5.0 software and custom tools developed by the M. D. Anderson Bioinformatics Department were used to analyze the data. The microarray data were processed using the position-dependent nearest neighbor model to normalize and extract gene expression values (23). The genes with below-median expression value were regarded as absent genes. The expression data of probe sets present in the 1q21-1q25 region representing 517 genes were extracted and processed using an unsupervised hierarchical clustering algorithm (24). The cluster analysis was done using the uncentered Pearson correlation as similarity metric and average linkage algorithm to combine cluster branches. Differentially expressed genes were identified using the standard t test.

Real-time quantitative PCR. We did the real-time quantitative PCR (qPCR) of nine genes that had average expression differences of 2-fold or greater between the NSM and cancer specimens in microarray analysis. The genes were RGS5 (regulator of G-protein signaling 5), ADAR (adenosine deaminase, RNA specific), ECM1 (extracellular matrix protein 1), IVL (involucrin), CRNN (cornulin), NICE-1, SPPR3, S100A2, and CRABP2 (cellular retinoic acid-binding protein 2). Due to insufficient RNA in five cancers (numbers 6, 20, 22, 23, and 38), we were able to evaluate only 14 cancers by real-time qPCR. For cancer specimens, triplicates of 100 ng of total RNA were reverse transcribed in a final volume of 20 μL using random primers and SuperScript II Reverse Transcriptase (Invitrogen, Carlsbad, CA). The final reaction products from triplicates were pooled and stored at −20°C until further use. For normal specimens, equal amounts of RNA from each of the seven NSM biopsies were pooled first and then 100 ng of the pooled RNA was reverse transcribed in triplicate as described above. For both cancer and normal samples, total RNA from the same aliquot of RNA that was used for microarray analysis was used for reverse transcription.

The TaqMan minor grove binder probe and the ABI Prism 7900 HT Sequence Detection System (PE Applied Biosystems, Foster City, CA) were used for detecting real-time PCR products. Primers and probes for the target and internal control gene (18S) were designed by PE Applied Biosystems and obtained via their Assays-on-Demand Gene expression products services. PCR assays included 10 μL of TaqMan Universal Master Mix No Amperase UNG (2×), 1 μL of 20× Assays-on-Demand Gene Expression Assay Mix, and 2 μL of cDNA diluted in Rnase-free water, in a final volume of 20 μL. The PCR thermal cycling conditions were as follows: 10 min at 95°C for AmpliTaq Gold activation and 40 cycles for the melting (95°C, 15 s) and annealing/extension (60°C, 1 min) steps. Each target was amplified individually and in duplicate.

Comparative C_T method (2^−ΔC_T) for relative quantification of gene expression. The expression of each target was calculated based on the difference between amplification of the individual target mRNA template and the internal control (18S) mRNA template using the ΔC_T method. The relative expression levels of target genes in comparison to the control gene were then calculated using the formula 2^−ΔC_T, where ΔC_T represents the difference between each target gene and the control gene (average C_T for the target minus average C_T for 18S RNA). To simplify the data presentation, we multiplied 2^−ΔC_T values by a factor of 1,000,000.

Statistical methods. The relative expression values obtained by real-time qPCR were used to determine the potential of the genes to discriminate pathCR from <pathCR. The log expression values of IVL, CRNN, NICE-1, CRABP2, ECM1, S100A2, and SPPR3 were used as predictors, and pathCR was used as the response variable. Univariate logistic regression analysis was done to identify markers most closely associated with pathCR.

Survival analyses were done for overall survival (OS) and disease-free survival (DFS) times. OS time was defined as the time from registration into the trial until death from esophageal cancer. When the date of death was not available, the date of the last follow-up was used instead. Data from patients who had not died by the time of analysis were censored for the purpose of statistical analysis. DFS time was defined as the time from registration into the trial to disease recurrence or last follow-up if the date of disease recurrence or death was not available. Data from patients who were alive without disease at the time of analysis were counted as censored. The association between molecular subgroups and OS or DFS was assessed by comparing the Kaplan-Meier survival curves with the log-rank test used to test differences in survival distribution.

Patient characteristics.Table 1 illustrates the patient characteristics. Patients were mostly men, and the clinical stages were as follows: stage IIA in 16%, stage III in 74%, and stage IVA in 10%. Seven (36.8%) of the 19 patients had EAC that developed in a Barrett's esophagus background. Of the 16 patients who underwent surgical resection following chemoradiotherapy, 5 (31%) had pathCR and 11 (69%) had <pathCR (6 patients had 1-10% residual carcinoma, 2 had 11-50% residual carcinoma, and 3 had 51-100% residual carcinoma).

The median time to local-regional or metastatic progression was 20.5 months (range, 2-39 months). The median survival time was 39 months (range, 6-45 months), with a 3-year OS rate of 58%.

Molecular subgroups by expression profiling of genes at chromosome 1q21-1q25. Unsupervised clustering analysis using expression data of genes mapped to the 1q21-1q25 chromosomal region segregated NSM specimens from cancers (Fig. 1A). Cancers, although more heterogeneous than NSM, separated into two major groups. One cluster designated as subgroup I included 9 cancers, and a second cluster designated as subgroup II consisted of 10 cancers. Cancers in subgroup I showed some heterogeneity but clearly segregated from NSM when subjected to cluster analysis separately without subgroup II cancers (Fig. 1B). In contrast, cancers in subgroup II clustered tightly and segregated distinctly from cancers in subgroup I, even when normal specimens were not included in the clustering analysis (Fig. 1C). It should be noted that matched NSM specimens were obtained from patients in both subgroups (three from subgroup I and four from subgroup II cancers).

Expressional mapping of region 1q21-1q25 and identification of target genes. Fifty-three genes were differentially expressed in cancer and NSM specimens (P < 0.01; t test). Table 2 shows the list of genes differentially expressed in the cancer specimens. Both cancer subgroups showed higher expression (≥2-fold) of 18 genes compared with NSM. Lower expression (≥2-fold) of 35 genes was seen in cancer versus NSM specimens (Fig. 2). The expression levels of several of these genes differed substantially among cancer subgroups. For instance, 14 genes (40%) showed a greater than 4-fold decrease in subgroup II cancers, whereas only 5 of the 14 genes showed such low expression in subgroup I cancers (Table 2).

Intriguingly, within the cancer subgroups, the expression pattern suggested a gradient of decreased expression culminating with the maximal down-regulation of genes included in the EDC cluster at 1q21. Thus, we focused our analysis on a 5-Mb DNA region located on 1q21-1q23 and encompassing the EDC genes. A schematic of the chromosomal region, with a detailed map of the EDC cluster and flanking genes, is shown in Fig. 3A. Using real-time qPCR, we analyzed the level of transcriptional expression of (a) three genes included in the EDC (IVL and SPPR3, both members of the cornified envelop precursor family, and S100A2) and (b) six genes flanking the EDC region (ECM1, ADAR, RGS5, CRNN, NICE-1, and CRABP2).

Real-time qPCR data confirmed the results obtained by microarray analysis. Compared with normal squamous epithelium, the expression levels of ECM1, IVL, CRNN, NICE-1, SPRR3, S100A2, and CRABP2 were lower in cancers, whereas the levels of ADAR and RGS5 were slightly higher in cancers.

Among cancer subgroups, subgroup II had substantially lower levels than subgroup I of expression of IVL, SPRR3, and S100A2, the three genes within the EDC (Fig. 3B). Among the genes within 0.5 Mb of the EDC cluster, the levels of CRNN and NICE-1 were also substantially lower in subgroup II compared with subgroup I. The levels of ECM1 and CRABP2, two genes flanking either side of the EDC (about 3 Mb apart), were, however, similar in both groups. A clear segregation of the two subgroups, without any overlap in the expression levels, was seen in CRNN and NICE-1. The dramatic differences in the expression levels of CRNN, NICE-1, IVL, SPRR3, and S100A2 segregated the cancers into high (subgroup I) and low expressers (subgroup II). It should be noted that substantial heterogeneity was found in gene expression levels among the cancers in both subgroups.

It has been suggested that within the 1q21 region, the EDC gene cluster and its closely flanking genes may have a coordinated transcription control mechanism (25–27). Hence, using the data obtained by real-time qPCR, we analyzed the correlation between the expression levels of 1q21 target genes using linear relationship analysis methods. Strikingly, as shown in Table 3, we found the presence of statistically significant coordinated down-regulation of the EDC genes and their neighboring genes, CRNN, NICE-1, and CRABP2.

Molecular malignant subgroups, clinical characteristics, and patient outcome. The 1q21-1q25 molecular malignant subgroups were not associated with patient characteristics, including age at diagnosis, clinical stage, and location of the primary tumor. It is noteworthy that five (71%) of the seven patients with EAC that developed in the background of Barrett's esophagus were in subgroup II (Table 1).

The molecular malignant subgroups seemed to be associated with response to CTXRT. Four (80%) of the five patients achieving pathCR were clustered in molecular subgroup I compared with only one patient (20%) in subgroup II, suggesting that subgroup II cancers were resistant to CTXRT. Although expression levels of CRNN, NICE-1, IVL, SPRR3, and S100A2 were substantially different in the two EAC subgroups (Fig. 3), a statistically significant relationship between these biomarkers and pathCR was not shown by univariate logistic regression analysis. Three biomarkers, IVL, NICE-1, and CRNN, on the basis of minimal overlap in the expression levels between the two subgroups of cancers, were dichotomized into high/low expression (high expression was defined as relative expression levels of >100 for CRNN and >5,000 for NICE-1 and IVL). Then, a two-sided Fisher's exact test was used to investigate the association of the dichotomized biomarkers with pathCR. A statistically significant association with pathCR was observed only for IVL (P = 0.05). However, the small sample size may have underpowered the study. Due to the heterogeneity in the expression levels of genes within the EDC, it was not possible to segregate responders from nonresponders within subtypes.

Patients in molecular subgroup II had worse OS, with a median time of 34 months (95% CI, 16-52 months) compared with 44 months (95% CI, 0-90 months) for those in subgroup I; however, owing to the small cohort size, no statistically significant difference was reached (P = 0.55; log-rank test). At 3 years, 32% of subgroup II survived compared with 78% of subgroup I. Similarly, molecular subgroup II was associated with shorter DFS, with a median DFS time of 14 months (95% CI, 2-26 months) compared with 37 months (95% CI, 0-82 months) for molecular subgroup I (P = 0.20; log-rank test). At 3 years, 44% of patients in subgroup II were disease-free compared with 78% in subgroup I.

Chemoradiation resistance and local recurrence or distant metastases are clinical features of over 75% of esophageal cancers treated with CTXRT. Thus, understanding the biological properties that render a cancer sensitive or insensitive to chemoradiation is crucial for tumor management.

To begin to understand biomarkers predictive of response to CTXRT, we previously conducted a transcriptional profiling study of pretreatment esophageal cancer biopsies derived from patients treated on a preoperative CTXRT protocol. Among the markers associated with the lack of response to CTXRT, we observed a down-regulation of genes located at the chromosomal region 1q21. The present analysis of the 517 genes within the 1q21-1q25 region clearly shows EAC resistant to CTXRT clustered together with a homogenous down-regulation of genes in this region compared with EAC sensitive to CTXRT.

The in-depth analysis of the genes included in the 1q21-1q25 region suggests a gradient of expression changes between cancer molecular subgroups. EAC sensitive to CTXRT harbored a transcriptional profile closer to that of NSM, whereas resistant EAC showed a dramatically different profile stemming mainly from the differential expression of the EDC gene cluster, including CRNN, NICE-1, IVL, SPRR3, and S100A2 (Fig. 3A and B). Relative expression levels of these genes, as determined by quantitative PCR, clearly segregate the two cancer types.

The proteins encoded within the EDC are implicated in the terminal differentiation of keratinocytes, and their expression is temporarily physiologically coordinated to mediate cessation of proliferation (e.g., S100As) and migration toward the superficial layers (e.g., IVL), with associated progressive cornification (e.g., SPRRs). The temporal expression of these proteins seems to be a biological program, coordinated by the subchromosomal position of gene territories and similar to that involving the MHC genes (27). The EDC cluster is also critical for the maturation and maintenance of the stratified squamous normal esophageal mucosa. Although the functions of the proteins encoded by the NICE-1, IVL, SPRR3, and S100A2 genes in terminally differentiating keratinocytes are well documented, the role of CRNN in the differentiation program is poorly understood. CRNN is either dramatically reduced or absent in primary esophageal cancer tissues, suggesting that it is esophageal specific and cancer related (28, 29). Our data, along with a previous report by Kimchi et al. (22), imply that the down-regulation of genes in this chromosomal region is involved in EAC development and progression. Our report also shows an association between these genes and CTXRT response.

Recognizing that our data are preliminary, we can still ponder some key observations stemming from the current study. These include (a) the degree of transcriptional suppression of genes mapping within and close to the EDC can segregate EAC into low and high expressers; (b) the high expressers, while showing transcriptional suppression in the EDC cluster region, seem to have a transcriptional signature closer to that of NSM; (c) the high expressers are more likely to be sensitive to CTXRT; and (d) the majority of the cancers with associated Barrett's metaplasia cluster in CTXRT-resistant cancer subgroup II. Thus, these observations suggest that EAC may include biologically and molecularly different entities: subgroup I maintaining similarity to NSM and subgroup II having a molecular signature similar to glandular epithelium. Of importance, cancers within subgroup I were more sensitive to CTXRT, similar to esophageal squamous cell carcinoma, for which the 3-year survival rate after CTXRT is higher than that for EAC (30, 31), and cancers within subgroup II were resistant to CTXRT, similar to EAC associated with Barrett's metaplasia (32). Our study, albeit small, supports the clinical observation that EAC associated with Barrett's metaplasia is more resistant to chemoradiation. Further studies are necessary to determine which molecular pathways are important for imparting chemoradiation resistance in subgroup II.

Cellular retinoic acid-binding protein II, encoded by CRABP2, is an intracellular lipid-binding protein that associates with retinoic acid with a subnanomolar affinity. Studies have shown that retinoic acid regulates the expression of markers of differentiation (33–35). Because it has been reported that the cellular retinoic acid-binding protein II enhances the transcriptional activity of the nuclear receptor, the retinoic acid receptor, by delivering retinoic acid to this receptor, the retinoic acid pathway may be affected by decreased expression of CRABP2 in these cancers.

Our results, along with those of other investigators, provide independent confirmation that EDC genes are differentially regulated in EAC and NSM and validate these genes as markers of EAC. To our knowledge, however, our study is the first to report coordinated down-regulation of keratinocyte differentiation genes and CRABP2 and the association between transcriptional suppression of differentiation-associated genes and resistance to chemoradiation in EAC. Studies using a larger sample set are in progress to validate IVL, CRNN, and NICE-1, in addition to previously shown SPRR3 and S100A2, as markers of resistance to CTXRT in EAC.

We thank Enrique Lopez-Alvarez and Jaime Bailey for their technical assistance, Dawn Chalaire for critical editing of the manuscript, and James Gilbert for his assistance in preparing the illustrations.