Genomic imbalances in esophageal carcinoma cell lines involve Wnt pathway genes

Abstract

AIM: To identify molecular markers shared across South African esophageal squamous cell carcinoma (ESCC) cell lines using cytogenetics, fluorescence in situ hybridization (FISH) and single nucleotide polymorphism (SNP) array copy number analysis.

METHODS: We used conventional cytogenetics, FISH, and multicolor FISH to characterize the chromosomal rearrangements of five ESCC cell lines established in South Africa. The whole genome copy number profile was established from 250K SNP arrays, and data was analyzed with the CNAT 4.0 and GISTIC software.

RESULTS: We detected common translocation breakpoints involving chromosomes 1p11-12 and 3p11.2, the latter correlated with the deletion, or interruption of the EPHA3 gene. The most significant amplifications involved the following chromosomal regions and genes: 11q13.3 (CCND1, FGF3, FGF4, FGF19, MYEOV), 8q24.21(C-MYC, FAM84B), 11q22.1-q22.3 (BIRC2, BIRC3), 5p15.2 (CTNND2), 3q11.2-q12.2 (MINA) and 18p11.32 (TYMS, YES1). The significant deletions included 1p31.2-p31.1 (CTH, GADD45α, DIRAS3), 2q22.1 (LRP1B), 3p12.1-p14.2 (FHIT), 4q22.1-q32.1 (CASP6, SMAD1), 8p23.2-q11.1 (BNIP3L) and 18q21.1-q21.2 (SMAD4, DCC). The 3p11.2 translocation breakpoint was shared across four cell lines, supporting a role for genes involved at this site, in particular, the EPHA3 gene which has previously been reported to be deleted in ESCC.

CONCLUSION: The finding that a significant number of genes that were amplified (FGF3, FGF4, FGF19, CCND1 and C-MYC) or deleted (SFRP2 gene) are involved in the Wnt and fibroblast growth factor signaling pathways, suggests that these pathways may be activated in these cell lines.

Key Words: Esophagus, Cancer, Single nucleotide polymorphism arrays, Fluorescent in situ hybridization

INTRODUCTION

Esophageal squamous cell carcinoma (ESCC) is a major cause of cancer-related death in the world and it is characterized by a peculiar epidemiology with worldwide geographic pockets of high incidence. In South Africa, the region of the Eastern Cape shows the highest incidence and ESCC represents the leading cancer affecting men and the second most common cancer in woman with a prevalence of 31.3 and 18 per 100 000 individuals, respectively[1]. The Gauteng (Johannesburg) and the Western Cape regions are also affected but to a lesser extent[2,3]. The tumors are diagnosed at advanced stages of the disease and patients have a poor survival rate. The etiology of this cancer is unresolved and while the most common risk factors associated with ESCC include smoking and alcohol consumption, these factors are surprisingly lacking in some parts of the world that have a high incidence such as in Iran and northern parts of China[4]. Additional risk factors have been proposed to play a role in some regions. In particular, exposure to fumonisin, a Fusarium fungi toxin that grows on maize, was reported in South Africa and China[5,6] as well as human papillomavirus (HPV) infection[7]. Poor nutrition is associated with ESCC in most parts of the world[4], and chronic inflammation was described in endemic parts of South Africa[8]. The respective part played by environmental risk factors and potential genetic susceptibility remains unclear and it is possible that different combinations of factors may be at play in different parts of the world.

It is widely accepted that recurrent chromosomal breakpoints in malignancies often pinpoint genes involved in the initiation or progression of cancer[9]. A major limitation in assessment of the chromosome complement in ESCC specimens is the difficulty in obtaining metaphases from fresh tumors, and established cell lines provide a unique resource for such investigations. The number of ESCC cell lines that have been reported to date remains limited and were all established in China[10-12] and Japan[13]. These cell lines have been investigated with one or several low resolution molecular cytogenetic techniques including cytogenetics, fluorescence in situ hybridization (FISH), multicolor FISH (M-FISH) or SKY and conventional comparative genomic hybridization (CGH). Various clonal aberrations have been identified and the most common abnormalities across studies involved over representation of chromosomes 1q, 3q, 11q and 8q as well as breakpoints in the centromeric or near centromeric regions of chromosomes 1, 3 and 8[10-12,14].

Five cell lines have previously been established from South African ESCC patients[15,16] but apart from cell line SNO, which was karyotyped, these cell lines were never characterized for their genetic constitution. We used conventional cytogenetics, FISH, and M-FISH to identify common chromosome structural abnormalities across these cell lines. Affymetrix 250K single nucleotide polymorphism (SNP) arrays were performed to investigate DNA copy number changes and the common aberrations detected by M-FISH and conventional cytogenetics. Here we describe clonal aberrations shared by these cell lines and highlight preferential targets for chromosomal rearrangements and copy number changes. These are the first ESCC cell lines from Africa to be genetically characterized to our knowledge.

MATERIALS AND METHODS

Cell lines

The five esophageal carcinoma cell lines used in this study were previously isolated from male patients with moderately differentiated ESCCs[16]. These cell lines were previously described and are designated as WHCO1, WHCO3, WHCO5, WHCO6 and SNO cell lines[15,16].

Cytogenetics

Cell lines were cultured at 37°C, 5% CO₂ in Dulbecco’s Modified Eagles medium: HAMS F12 (3:1) (GIBCO^®, Invitrogen Corporation, USA) containing 10% fetal calf serum, 100 μg/mL streptomycin (ICN, Costa Mesa, CA, USA) and 100 IU/mL of penicillin (ICN, Costa Mesa, CA, USA). When the cultures were half confluent, the cells were incubated with a final concentration of 0.44 μg/mL of Karyomax^® Colcemid^® (Invitrogen Corporation, USA) for 4 h. Cells were then harvested by standard cytogenetic procedure. Slides were either prepared for metaphase analysis using GTG banding by standard procedures or for FISH preparations.

M-FISH

M-FISH was performed using the SpectraVysion™ Assay (Vysis^®, Abbott Molecular Inc, IL, USA) according to the manufacturer’s protocol. The slides were analyzed on an Olympus BX41 fluorescent microscope with six single band pass filters for visualization of the six fluorophores. Genus™ CytoVision 3.0 software (Applied Imaging Corporation, San Jose, California, USA) was used for image acquisition and analysis. Ten metaphases were analyzed per cell line.

FISH

FISH was performed on metaphase chromosomes from all cell lines in order to confirm or refine translocation derivative breakpoints and clarify their composition. Probes specific for the short and long arms of chromosome 3, the short arm of chromosome 1 and the long arm of chromosome 22 (Qbiogene, Strasbourg, France) as well as the Vysis^® Cep 3 Alpha and Cep 1 Alpha SpectrumOrange probes (Abbott Molecular Inc., IL, USA) were hybridized to further map the breakpoints. The Vysis^® LSI IGH and Vysis^® LSI RARA, both dual color break apart rearrangement probes (Abbott Molecular Inc., IL, USA) were used to confirm the involvement of chromosomes 14 and 15 in translocation derivatives seen in cell lines WHCO1 and WHCO3 respectively. In order to establish if the EGFR (epidermal growth factor receptor) gene was involved in a marker chromosome 7 the Vysis^® LSI EGFR SO/CEP 7 SG probe (Abbott Molecular Inc., IL, USA) was hybridized to metaphase chromosomes.

FISH was also performed on interphase nuclei from all cell lines using the Vysis^® LSI C-MYC SpectrumOrange probe and the Vysis^® LSI t(11;14) dual color probe (Abbott Molecular Inc., IL, USA). These probes target the C-MYC gene (Spectrum Orange) on 8q24 and the CCND1 gene (Spectrum Orange) on chromosome 11q respectively and were used to confirm SNP array copy number results. A hundred interphase nuclei were analyzed in each cell line. All FISH experiments were performed according to the manufacturer’s instructions and analyzed on an Olympus BX41 fluorescent microscope equipped with appropriate fluorescence filters. The Genus™ CytoVision 3.0 software was used for image acquisition and analysis.

DNA isolation

DNA was extracted from the cell lines and from six blood specimens obtained from healthy and population-matched volunteers using standard phenol-chloroform extraction methods. DNA was quantified on the ND-1000 Spectrophotometer (Nanodrop^® Technologies, Rockland, DE) and quality assessed by gel electrophoresis.

SNP arrays and data analysis

Cell lines and control DNA were hybridized to the Affymetrix^® GeneChip^® Human Mapping 250 K Nsp Arrays (Santa Clara, CA). The GeneChip^® mapping 500 K protocols were used. The control group used as a reference for analysis was DNA samples extracted from whole blood from black male volunteers.

Data analysis was performed using the Affymetrix^®, Genotyping Console™ 2.0 and the copy number analysis tool (CNAT 4.0) software (Affymetrix^®, Santa Clara, USA). Subsequently data were analyzed with third party software, Genepattern[17]. The signal intensities from the CEL files were normalized by the PM-MM (perfect match minus mismatch) probe intensity and quantile normalization against the median intensity of the controls. The raw copy number was then estimated as a ratio against the signal intensities of the normal reference samples. The log2ratios were smoothed using GLAD (Gain and Loss analysis of DNA)[18].

To determine the significant common regions of amplification and deletion (i.e. driver aberrations as opposed to random passenger aberrations) across the five cell lines, the GISTIC (Genomic Identification of Significant Targets in Cancer) algorithm[19] was applied to the smoothed data. The algorithm first scores the regions of copy number change according to their frequency and amplitude, which indicates the likelihood of these regions to represent a driver aberration (G-score). The statistical significance of each G-score is calculated by comparison of these scores against a null model of random aberrations. This significance is represented as a q-value (False discovery rate), which is the likelihood that the data was generated by chance. The most probable locations for oncogenes or tumor suppressor genes are identified by calculating the minimal common regions of aberration, which are most significantly altered i.e. high amplitude change. The regions of aberrations with high G-scores and minimal q-values (less than 0.25) are therefore more likely to contain target genes that have a significant role in carcinogenesis. The chromosomal regions are then classified as high or low level where low level amplification is considered for a log2ratio greater than 0.1 but smaller than 0.9 and high level amplification is considered for a log2ratio > 0.9 (0.9 corresponds to at least 3.7 copies per diploid cell). Similarly low level deletions (hemizygous) were considered for a log2ratio of -0.1 and homozygous deletions were considered for a log2ratio < -1. The GISTIC algorithm has been used for copy number analysis in previous studies[20,21].

RESULTS

Cytogenetics and M-FISH

Cytogenetic analysis revealed complex numerical and structural chromosome aberrations in all cell lines with a high variability observed between cells from the same cell line (Figures 1 and 2). The G-banded karyotype of cell line WHCO1 and a corresponding M-FISH karyotype that illustrates the degree of variation from one metaphase to another are shown in Figure 1. The M-FISH data confirmed the complexity of the karyotypes and revealed the genetic composition of recurrent chromosome markers that could not be identified on G-banded metaphases (Figure 2); some of these markers are further discussed below. The detailed composite karyotypes obtained from twenty metaphases in each cell line are summarized in Table 1. All cell lines were near diploid except for cell line WHCO5 which was near tetraploid (Figure 2). Across the five cell lines a total of 97 translocations, 19 trisomies and 11 monosomies were detected. The breakpoints amounted to 203, with 78 of these clustering around the centromeric regions of chromosomes. The chromosomes involving the highest number of abnormalities were chromosomes 1, 3, 5, 7, 8, 9, 10, 11, 13, 14, 15, 18, 19, 20 and 22. G-banded metaphases also revealed the presence of isochromosomes involving the D group acrocentric chromosomes in all the cell lines. In particular isochromosome 13q, i(13)(p10) was common to cell lines WHCO3, WHCO5 and WHCO6, isochromosome 14q, i(14)(p10) was common to cell lines WHCO1, WHCO5 and WHCO6, and isochromosome 15q, i(15)(p10) was seen in cell lines WHCO3 and WHCO5.

Table 1 Composite karyotypes.

Cell line	Composite combined cytogenetics and M-FISH karyotype
WHCO1	42~73, XY, +X, +X [4], +1 [10], del 1(?) [4], + der(1)t(1;14)(p11;q11) x2 [18], t(1;14;22)(?;?;q?) [4], t(1;22)(?;q?) [6], + 2 [8], t(2;8)(?;?) [4], +3 [8], +del 3(p21p13) x2 [4], der(3;22)(q10;q10) [4], -4 [8], t(5;21)(?;?) [8], + der(5)t(5;?)(q23;?) [4], + del(6)(q?21) x2 [4], t(6;12)(?q;?p) [16], t(6;13)(?q25;?q14) x2 [18], +7 [4], + del(7)(q21) [4], + der(7;14)(p10;q10) [20], der(7)t(7;18)(p21;q23) [16], der(7)t(7;19)(p22;p13) [4], der(8)t(8;?19)(q?24;q?11) [4], t(8;22)(?;?) [18], +9 [10], + der(10)t(9;10)(q13;q?11.2) [4], +11 [8], t(11;17;20)(?;?;?) [4], +12 [10], t(13;22)(?;?) [4], add(14)(q?) [4], i(14)(q10) [6], der(15;19)(q10;q10) [20], der(16)t(5;16)(?;?) [8], +18 [4], der(19)t(5;19)(p?12;q11) [12], der(19) t(5,19)(q33;q13.4) [18], + der(19)t(9;19)(q?13;q?13) x2 [4], t(19;21)(?;?) [14], +20, +20 [20], + der (20)t(1;11;20)(?;?;?) [4], +1~7mar [20] [cp20]
WHCO3	46~50, X, -Y, +2 [12], der(5)t(5;8,18)(q?;?;?) [20], + del 7(q22) [4], der(7)t(7;9)(?;?) [6], t(7;9;16;18)(?;?;?;?) [14], t(7;15)(?;?) [4], +12 [6], der(12)t(6;12)(?;?) [16], i(13)(q10) [6], t(13;14)(?;?) [6], t(13;14;20)(?;?;?) [6], +14 [12], + der(15)t(1;15)(q11;p11) [6], + der(15)t(1;15;11)(?;?;?) [10], der(15;22)(q10;q10) [20], i(15)(q10) [4], der(16)t(3;16;22)(p?11.2;?;q?10) [20], +17 [10], der(?20)t(9;13;20)(?;?;?) [10], der(21)t(13;21)(?;?) [12] [cp20]
WHCO51	99~108, XY, t(X;4;10.22)(?;?;?;q?) [14], t(1;19)(?;?) [8], t(1;18)(?;?) [6], del 2 [2], der(2)t(2;9)(q12;q13) [10], t(2;9)(?q31;?q34) [8], +der(3)t(1;3)(p11-12;q11) [12], t(3;11;13;22)(?;?;?;?) [8], t(3;11;22)(?;?;?) [16], t(3;22)(p11;q11) [12], der(5;20)(p10;p10) [4], t(6;13)(?;?) [4], del(7)(q31) [4]; t(8;14;18)(?;?;?) [16], t(8;18)(?;?) [10], der(9)t(9;14)(?;?) [16], t(9;15)(q?;q?) [10], t(9;19)(?;?) [6], t(12;19)(?;?) [4], i(13)(q10) [6], i(14)(q10) [8], der(15)t(Y;15)(?;?)[12], der(15)t(7;15)(?;?) x2 [14], i(15)(q10) [4], der(19)del(19)(q13.?2)t(5;19)(p?12;p11) [12], del(20)(q11.2) [12] [cp20]
WHCO6	44~54, Y, -X, der(1)t(1;8)(p11;q11) [14], t(3;10)(?;?) [4], t(3;10)(?q13.3;p10) [6], t(4;10)(?;?) [4], t(5;10)(?;?) [10], t(5;22)(?;q?) [4], +6 [2], del(6)(q?21) [4]; t(6;11)(p12;q13) [6], der(22)t(6;22)(?;?) [6], +7[7], del(7)(q31) [8]; +8 [8], t(9;15)(?;?) [10], t(10;14)(?;?) [6], -11 [6], t(11;22)(p?;q?) [6], +12 [10], i(13)(q10) [6]; i(14)(q10) [6], +16 [8], t(17;19)(?;?) [10], +18 [4], del(20)(q?11.2) [6]; -21 [12], t(21;22)(?;?) [6], der(22)t(6;22)(?;?) [6], [cp14]
SNO	29~43, XY, +X, del X(?) [14], der(Y;15)(q10;q10) [12], t(1;16)(?;?) [14], der(2)t(X;2)(?;?) [16], der(2)t(1;2)(?;?) [18], der(3)t(1;3)(p11-12;q11) [20], t(3;9)(?;?) [4], t(3;10)(?;?) [14], t(3;12)(?;?) [20], t(3;20)(?;?) [8], t(4;9;11)(?;?;?) [20], t(4;11)(?;?) [12], der(5)t(1;5)(?;?) [20], -6 [18], del(6)(q?23) [20], der(7)t(3;7)(?q25;p22) [18], t(7;20;11;8;2)(?;?;?;?;?) [18], der(8)t(2;8)(?;?) [14], t(8;18)(?;?) [8], -9 [10], t(9;18)(?;?) [4], t(10;22)(q?;q?) [12], t(11;13)(?;?)[8], t(12;15)(?;?)[8], t(12;21)(?;?)[20], t(13;11;20)(?;?;?) [14], der(14;22)(p10;q10) x2 [20], t(14;19)(?;?) [20], -16 [14], der(16)t(9;16)(q?22;q?13) [20], der(17)t(6;17)(?;?) [12], -18 [16], -19 [16], der(19)del(19)(q13.?2)t(5;19)(q?12;q10) [16], +20 [10], del(20)(q?12) [16]; der(20;21)(q10;p10) [6], -21 [18], -22 [20] [cp20]

¹Only structural rearrangements are listed due to the complexity of this cell line.

Figure 1 G-banded and multicolor fluorescence in situ hybridization representative karyotypes of cell line WHCO1. A: G-banded karyotype; B: Multicolor fluorescence in situ hybridization (M-FISH) karyotype, the arrows indicate the marker chromosomes, der(1) t(1;14)(p11;q11) and del(7)(q21).

Figure 2 Multicolor fluorescence in situ hybridization representative karyotypes showing complex rearrangements in four cell lines. A: Cell line WHCO3, the white arrow indicates the der(15)t(1;15)(q11;p11); B: Cell line WHCO5, the white arrow points to the der(3)t(1;3)(p11-12;q11); C: Cell line WHCO6; D: Cell line SNO, the white arrow points to the der(3)t(1;3)(p11-12;p11) and the yellow arrow indicates the marker chromosome 7, mar(7), which involves the EGFR (epidermal growth factor receptor) locus (Figure 6).

The combined results of cytogenetics, FISH and M-FISH revealed a common translocation derivative, der(3)t(1;3)(p11;q11) that combined chromosome 3 and chromosome 1 short arms in cell lines WHCO5 and SNO (Figures 2, 3, 4). The SNP array copy number data did not bring further information on these breakpoints, possibly due to a lack of SNP probes in this region. Interestingly, the 1p11 breakpoint was also involved in cell lines WHCO1 and WHCO6 in differing unbalanced translocations (Figure 3). The M-FISH and FISH results with probes for the respective partner chromosomes and centromeric Cep1, confirmed the interpretation of these derivatives as der(1)t(1;14)(p11;q11) in cell line WHCO1 and der(1)t(1;8)(p11;p11) in cell line WHCO6 (results not shown). In contrast, a translocation derivative also involving the chromosome 1 pericentromeric region in cell line WHCO3 (Figure 3) was shown not to involve 1p11 but 1q11, and was interpreted as der(15)t(1;15)(q11-12;p11) (result not shown).

Figure 3 Partial G-banded karyotypes of the five cell lines showing shared chromosomal aberrations across cell lines. Chromosome derivatives from unbalanced translocations involving (A) chromosome 1p11-12 breakpoints in four cell lines, 1q11 in cell line WHCO3, (B) deletions of chromosome 7q in four cell lines, and (C) deletions of chromosome 20q in three cell lines.

Figure 4 DAPI stained metaphasic chromosomes showing the der(3)t(1;3)(p11-12;q11) in cell lines SNO and WHCO5. A, B: Arm-specific paint for chromosomes 1p (green) and 3p (red) in cell lines SNO and WHCO5 respectively showing the derivatives der(3)t(1;3)(p11-12;q11) arrowed; C, D: Arm specific paint for chromosome 1p (green) and Cep 3 α probe (red) in cell lines SNO and WHCO5 respectively showing that the centromere of chromosome 3 is retained on the derivatives der(3)t(1;3)(p11-12;q11).

Translocation derivatives involving chromosomes 3 and 22 were seen in cell lines WHCO3 and WHCO5 and were interpreted as der(16)t(3;16;22)(p11;?;q?11) and t(3;22)(p11;q11) respectively (Table 1). The array data showed corresponding hemizygous deletions at 3p11.2, in cell line WHCO3. Interestingly deletions at 3p11.2 were also observed in cell lines WHCO1, SNO and WHCO6. The minimal region of overlap was 343 kb (88 184 220-88 527 215 bp) in size and involved c3orf38 and CGG triplet repeat binding protein (CGGBP1) genes. In addition the EPHA3 gene was deleted in cell lines WHCO6 and SNO (Figure 5). The EPHA3 gene encodes a tyrosine kinase, which is mutated in lung and breast cancers[22-24].

Figure 5 CNAT plot of single nucleotide polymorphism hybridization results showing the map of the region of deletion on chromosome 3p11. 2-12.1 in cell lines WHCO1, WHCO3, WHCO5, WHCO6 and SNO. The red bars indicate the segment of deletion as determined by CNAT 4.0. The dotted lines indicate the copy number status for each single nucleotide polymorphism probe for each cell line. The blue block indicates the minimal common region of deletion and the red arrow indicates the two genes in this region, c3orf38 and CGGBP1.

Cell lines WHCO1, WHCO3, WHCO5 and WHCO6 all showed a deletion of chromosome 7 long arm with varying breakpoints, q21 to q31, on G-banded metaphases (Figure 3). However GISTIC analysis of the SNP array data identified a significant common focal deletion (q-value of 0.09) of approximately 5.16 Mb at 7q33-q34 (133 721 542-138 880 555 bp) in only three of these cell lines. This region contained 26 genes including the potential target gene homeodomain-interacting protein kinase-2 (HIPK2) the product of which activates p53 expression and is pro apoptotic[25].

Although deletions at 20q11.2 were detected in cell lines WHCO5, WHCO6 and at 20q12 in cell line SNO on G-banded metaphases (Figure 3), copy number analysis (CNAT) revealed that there was in fact amplification of 20q sequences in all cell lines implying that complex rearrangement occurred for these sequences to be relocated elsewhere in the genome. Two minimal regions of amplification were identified, the 20q11.21-q11.22, of approximately 1.12 Mb in size (31 799 867-32 906 584 bp), which contained 14 genes and a smaller region at 20q13.12 of 149.48 kb which contained five genes. Both these regions were amplified in all cell lines, yet these amplifications were not found to be significant on GISTIC analysis.

Cell line SNO showed a large marker chromosome 7 on metaphases analyzed by M-FISH. This marker was interpreted as a possible inverted duplication of chromosome 7p sequences (Figure 2D). FISH with an EGFR probe, revealed a high amplification of the EGFR gene in 14% of the cells (Figure 6). A high level amplification at 7p13-7p11.2 (genomic location of EGFR) was confirmed by GISTIC (q-value 0.14) on array analysis in this cell line, while low level amplification was observed in the remaining four cell lines in agreement with the presence of 4 to 7 copies on FISH analysis.

Figure 6 DAPI stained metaphase from cell line SNO showing fluorescence in situ hybridization results with the locus specific EGFR probe (red) and Cep7 α (green). The arrow indicates the chromosome 7 marker with a homogeneously stained region that contains the EGFR locus.

Copy number changes

In view of the high cell to cell heterogeneity in each cell line, and in order to separate the potential driver aberrations out of the background of aberrations that may have occurred by chance, we used the software GISTIC specifically designed for this purpose. Target genes, defined as genes whose alteration confers a cell growth advantage, are likely to reside in the regions amplified or deleted to the highest degree in a majority of cells and within a common region of overlap amongst all cell lines. Fourteen common regions of amplification and 20 regions of deletion were identified (Figure 7, Tables 2 and 3). The most significant chromosomal regions of amplification were, in descending order of significance: 11q13.3 and 8q24.21, in four and five cell lines respectively, 11q22.1-q22.3 and 5p15.2 both detected in two cell lines. Chromosomal regions of less significant amplification were 10p12.33-q21.3, 18p11.32, 20p11.1-p11.22, 22q11.21-q11.22, 22q12.3, 9q31.1, 22q11.21, 3q11.2-q12.2 and 1p34.2 (Figure 7). Similarly, chromosomal regions of deletion were seen in the following order of significance: 1p31.1-p31.2, 2q22.1, 3p12.1-p14.2, 4q22.1-q32.1, 8q11.1-p23.2, 14q21.2, 18q21.1-q21.2 and 21p13-q21.3 detected in all or four cell lines. Other regions of less significant deletion, and only seen in one or two cell lines, included: 10p12.31-p14, 14p13-q32.33, 3p26.3-q29, 9p24.2-p24.3, 6q24.3-q27, 12p13.33-q24.33, 7q33-q34, 8p23.3-q24.3, 10q11.23-q22.1, 11p11.12-q12.2 and 13q21.33-q34 (Figure 7). Together the regions of amplification and deletion encompassed a total of 4595 genes.

Table 2 Amplification peaks as detected by the GISTIC algorithm.

Cytoband	Location (kb)	Approx size (Mb)	Frequency (n/5)	Mean log2ratio1	q-value	Genes2
1p34.2	Chr1:39027237-40780163	1.75	3	1.10	0.18	MYCL1
3q11.2-q12.2	Chr3:95917505-101945216	6.02	4	1.40	0.17	MINA
5p15.2	Chr5:10051329-11800765	1.75	2	1.36	0.09	CTNND2
7p11.2-p13	Chr7:45294289-57299457	12.01	1	1.24	0.13	EGFR
8q24.21	Chr8:127445828-129661846	2.22	5	2.10	0.0007	MYC
9q31.1	Chr9:100905143-102152148	1.25	1	2.57	0.15
10q12.33-q21.3	Chr10:17811791-65388337	47.58	1	3.05	0.13
11q13.3	Chr11:68753086-69985447	1.23	4	1.81	5.76E-05	CCND1, CTTN, FGF3, FGF4, FGF19, MYEOV
11q22.1-q22.3	Chr11:100815801-103042620	2.23	2	2.35	0.03	BIRC2, BIRC3, YAP1
18p11.32	Chr18:1-1118244	1.12	2	1.04	0.13	TYMS, YES1
20p11.1-p11.22	Chr20:22140447-26145930	4.01	2	1.24	0.13	PYGB
22q11.21	Chr22:16558724-17937900	1.38	2	1.20	0.21	BID, CLDN5
22q11.21-q11.22	Chr22:18577713-20667607	2.10	2	1.20	0.13	CRKL, MAPK1
22q12.3	Chr22:31889314-32003182	0.11	2	1.15	0.13	LARGE

¹The mean log2ratio for the samples with a log2ratio > 0.9 (equivalent to 3.7 copies per diploid cell);

²The selected genes from the peak region.

Table 3 Deletion peaks as detected by the GISTIC algorithm.

Cytoband	Location (kb)	Size (Mb)	Frequency (n/5)		Mean log2ratio1	q-value	Genes2
			Hemi	Homo
1p31.1-p31.2	Chr1:66691991-71187083	4.50	3	1	-3.03	0.02	CTH
2q22.1	Chr2:141590067-141951947	0.36	3	2	-1.66	0.19	LRP1B
3p26.3-q29	Chr3:1-199344050	199.34	2	1	-2.90	0.022
3p12.1-p14.2	Chr7:60424050-85108679	24.70	4	1	-2.99	0.02	FHIT, ADAMTS9
4q22.1-q32.1	Chr4:91972774-162358674	70.40	3	2	-2.19	0.02	CASP6, SMAD1
6q24.3-q27	Chr6:147967444-170914576	22.95	2	1	-2.22	0.07
7q33-q34	Chr7:133721542-138880555	5.16	2	1	-1.72	0.09	HIPK2
8p23.2-q11.1	Chr8:4078057-47043375	43.00	4	1	-2.47	0.02	BNIP3L, INDO
8p23.3-q21.13	Chr8:1-146308819	146.31	1	1	-2.27	0.09
9p24.2-p24.3	Chr9:1151516-2459741	1.31	2	1	-2.53	0.03
10p12.31-p14	Chr10:10309026-19155158	408.85	2	1	-2.74	0.02
10q11.23-q22.1	Chr10:50393324-70694787	20.30	1	1	-2.35	0.12
11p11.12-q12.2	Chr11:50256798-61426521	11.20	1	1	-1.90	0.12
12p13.33-q24.33	Chr12:1-132078379	132.10		1	-2.80	0.08
13q21.33-q34	Chr13:68772537-113042980	44.30	1	1	-1.70	0.24
14q21.2	Chr14:42828345-44176016	1.35	3	1	-4.50	0.02
14p13-q32.33	Chr14:1-105311216	105.30		1	-5.10	0.02
18q21.1-q21.2	Chr18:46081464-51919972	5.80	5		-1.03	0.02	DCC, SMAD4
21p13-q21.3	Chr21:1-29932926	29.90	3			0.12	BAGE

¹The mean of the log2ratio of those samples with log2ratios < -1.3 (< 0.9 copies per diploid cell);

²The selected genes within the deletion peaks.

Figure 7 Plots of recurrent genomic amplifications (A) and deletions (B) detected in the esophageal squamous cell carcinoma cell lines from GISTIC analysis of single nucleotide polymorphism array data. The x-axis shows the G-score (top) and false discovery rate (q value; bottom). The green line indicates the false discovery rate cut off of 0.25. The circles indicate the peaks of the most significantly aberrant chromosomal regions.

Significant gains

We selectively describe below the five regions of amplification that were the most significant on GISTIC analysis (q-value < 0.25) (Figure 7). Chromosomes 11q13.3 (68 753 086-69 985 447 bp) and 8q24.21 (127 445 828-129 661 846 bp) were the two most amplified and most significant regions with a q-value of 5.76E^-05 and 0.0007 respectively. The 11q13.3 region was 1.23 Mb in size and was highly amplified in cell lines WHCO3, WHCO5, WHCO6 and SNO while amplified to a lesser degree in cell line WHCO1. This region harbors seven candidate genes including the cyclin D1 (CCND1), the cortactin (CTTN), the protein tyrosine phosphatase, receptor type, polypeptide, interacting protein α 1 (PPFIA1), FGF3, FGF4 and FGF19 and the myeloma overexpressed (MYEOV) genes, which could all play a role in ESCC oncogenesis. FISH validated these findings and confirmed the amplification of CCND1 (Table 4, Figure 8). The common amplicon of 2.22 Mb at 8q24.21 was highly amplified in cell lines WHCO1 and WHCO3 and moderately amplified in cell lines WHCO5, WHCO6 and SNO. This amplicon involved the oncogene v-myc myelocytomatosis viral oncogene homolog (C-MYC) and the family with sequence similarity 84, member B (FAM84B) gene. Locus specific FISH confirmed the amplification of C-MYC in cell lines WHCO1, WHCO3, WHCO5, WHCO6 and SNO (Table 4, Figure 8).

Table 4 Fluorescence in situ hybridization for detection of CCND1 and C-MYC amplification.

Cell line	FISH signals	Amplitude1	Log2Ratio
CCND1
WHCO1	5-15	2	1.56
WHCO3	> 20	2	2.57
WHCO5	10-20	1	0.78
WHCO6	4-8	1	0.55
SNO	15-20	1	0.79
C-MYC
WHCO1	4-6	1	0.12
WHCO3	15-> 20	2	1.83
WHCO5	10-> 20	2	1.83
WHCO6	10-> 20	2	2.06
SNO	4-> 20	2	1.71

¹Amplitude threshold where log2ratio < 0.1 = 0, log2ratio > 0.1 < 0.9 = 1 and log2ratio > 0.9 = 2 as determined by SNP array copy number analysis. FISH: Fluorescence in situ hybridization.

Figure 8 DAPI stained interphase nuclei hybridized with locus specific probes for C-MYC Spectrum Orange (red) (A-D) and CCND1 (red)/IGH (green) (E-H). C-MYC amplification can be seen in cell lines WHCO1 (A), WHCO3 (B), WHCO5 (C) and SNO (D). CCND1 amplification was detected in cell lines WHCO3 (E), WHCO5 (F), WHCO6 (G) and SNO (H).

A second focal region of high amplification on chromosome 11 was observed at 11q22.1-q22.3 (2.23 Mb in size) in cell lines WHCO5 and SNO. This region included the regulators of apoptosis BIRC2 (cIAP1) and BIRC3 (cIAP2), the matrix metalloproteinases (MMP) and the Yes associated protein (YAP-1) genes all potential target genes. The BIRC2 gene was previously described as a target of amplification /increased expression in cervical cancers[26] and the YAP-1 gene product is a cellular adaptor protein, which can induce BIRC2 expression. YAP-1 was reported to be over expressed in hepatic and mammary cancers[27]. In turn, the MMP genes, which include MMP1, MMP7, and MMP13, have been shown to be co-expressed in early stage ESCC correlating with a poorer prognosis[28].

A 1.75 Mb region on chromosome 5p15.2 (10 051 329-11 800 765 bp) was highly amplified in cell lines WHCO6 and WHCO5 and moderately amplified in cell lines WHCO1 and SNO. This region hosts the potential target gene, delta catenin (CTNND2), overexpressed in prostate cancer[29].

Four cell lines, WHCO6, WHCO3, WHCO5 and SNO, had focal gain on chromosome 3q. The minimal common region of amplification mapped at 3q11.2-12.2, and was 6.02 Mb in size (95 917 505-101 945 216 bp) (q-value = 0.17). This region is commonly amplified in a variety of cancers[20,30] including esophageal squamous carcinoma[31] and it involved the potential oncogene, MYC induced nuclear antigen (MINA).

The 18p11.32 sub-band was highly amplified in cell lines WHCO3 and WHCO6 and moderately amplified in the remaining 3 cell lines. This region of 1.12 Mb in size (1-1 118 244 bp) has previously been described in ESCC[32] and involves the potential oncogenes TYMS and YES-1. Both genes have been implicated in gastro intestinal cancer. The TYMS gene codes for a thymidylate synthase involved in DNA synthesis and targeted by the chemotherapy agent fluorouracil (5FU). TYMS overexpression leads to 5FU treatment resistance[32] and affects colorectal cancer treatment[33]. YES-1 is a homologue of the Yamaguchi sarcoma virus v-yes amplified and overexpressed in gastric cancers and ESCC[32].

Significant losses

The most significantly deleted chromosomal regions were 1p31.1-p31.2, 2q22.1, 3p12.1-p14.2, 4q22.1-q32.1, 8p23.2-q11.1, 14q21.2 and 18q21.1-q21.2 (in descending order of significance) in all or four of the cell lines (Figure 7). Less significant regions detected in less than four cell lines are depicted in Table 3. Three small regions of deletions involved chromosomes 1p, 2q and 18q. The 4.5 Mb deletion on chromosome 1 short arm, 1p31.2-p31.1 (66 691 991-71 187 083 bp) was seen in four cell lines (homozygous deletion in cell line SNO and hemizygous in cell lines WHCO1, WHCO5, WHCO6) with a high significant q-value of 0.02. Three genes with a reported tumor suppressor activity were involved in this deletion, the cystathionine γ-lyase (CTH), the growth arrest and DNA damage-45 α (GADD45α) and the DIRAS family, GTP-binding RAS-like 3 (DIRAS3) genes. The DIRAS3 gene was shown to be down regulated in hepatocellular carcinoma and breast cancers[34,35] and is postulated to have a tumor suppressive activity. Both the CTH and GADD45 genes were shown to negatively control cell growth[36,37].

Chromosome 2, sub-band q22.1 (141 590 067-141 951 947 bp) was lost in all cell lines (homozygous deletion in cell lines WHCO3 and WHCO5 and hemizygous in the other 3 cell lines) (q-value = 0.019). The low density lipoprotein 1B (LRP1B) tumor suppressor gene, deleted in lung cancer[38,39] maps in this region.

A chromosome 18q sub-band, q21.1-q21.2 (46 081 464-51 919 972 bp) (5.8 Mb) was hemizygously deleted in all five cell lines, involving both the SMAD4 and deleted in colorectal carcinoma (DCC) genes.

Three large regions of deletion involved chromosomes 3p, 4q and 8p. First, the 24.7 Mb region of deletion at 3p12.1-p14.2 (60 424 050-85 108 679 bp) was significant (q-value of 0.02) in all cell lines (homozygous deletion in cell line SNO). This region houses the FRA3B associated gene, Fragile Histidine triad (FHIT), whose deletions were previously detected by MLPA analysis in these cell lines[40]. The potential tumor suppressor ADAMTS9 gene, a metalloproteinase family member involved in inhibition of angiogenesis[41] was also involved in this deletion.

Second, the 4q22.1-q32.1 (91 972 774-162 358 674 bp) (75 Mb) region was homozygously deleted in cell lines WHCO3 and WHCO5 and hemizygously deleted in the three other cell lines with a q-value of 0.02. This region encompasses many genes but of interest are the Bone morphogenetic protein receptor 1B (BMPR1B), the caspase 6 (CASP6), the secreted frizzled-related protein 2 (SFRP2) and the SMAD protein 1 (SMAD1) genes, all potential tumor suppressor genes.

Lastly, a chromosome 8p23.2-q11.1 (4 078 057-47 043 375 bp) (43 Mb) deletion was seen in all cell lines (homozygous in cell line WHCO3 and hemizygous in the other 4 cell lines) (q value of 0.02), including five potential target genes, the BCL2/adenovirus E1B 19kDa interacting protein 3-like (BNIP3L), the leucine zipper tumor suppressor 1 (LZTS1) and the three tumor necrosis factor related superfamily genes TNFRSF10A, TNFRSF10B and TNFRSF10C.

DISCUSSION

We have characterized the karyotype and genomic constitution of five ESCC cell lines established in SA using a combination of traditional cytogenetics, M-FISH, FISH and SNP arrays. The number of ESCC cell lines genetically described worldwide is limited. Only eight ESCC cell lines have been investigated previously with traditional cytogenetics to our knowledge[10,12,42] and 10 with M-FISH[11,43,44], these are the only two techniques that can detect recurrent translocation breakpoints. In this study the five cell lines had complex karyotypes and were hyperploid with WHCO5 being near tetraploid. There was a high level of intra cell-line heterogeneity. The chromosomes most frequently involved in translocations were chromosomes 1, 3, 5, 7, 8, 9, 10, 11, 13, 14, 15, 18, 19, 20 and 22. These features were comparable to ESCC cell lines previously characterized[10-12,42,43] and across all studies that involved karyotyping, including the karyotyping of fresh ESCC tumor samples[14], chromosomes 1, 3 and 8 were the most commonly affected by translocation breakpoints[10-12,43].

Forty percent of translocation breakpoints occurred in near centromeric regions (Table 1). These were represented by unbalanced whole arm chromosome translocations, frequently involving chromosomes 1 and 3, and by isochromosomes for the D group acrocentric chromosomes 13, 14 and 15. Indeed, frequent centromeric breakpoints have been described in squamous carcinoma including ESCC[10,12] with up to 60% of all breakpoints being in centromeric regions[43] supporting the idea that centromeric disruption is a frequent event in epithelial cancers. It has been suggested that environmental factors may preferentially interact with centromeric sequences[45], and clastogenic compounds, such as mitomycin C, induce breaks in centromeric of chromosomes 1, 9 and 16[46]. Smoking is a major risk factor associated with ESCC and nicotine is known to induce single strand DNA breaks[47]. Although active smokers exhibit an increased number of breaks at fragile sites[48], it is not known if centromeric regions are also targeted.

Two chromosomal breakpoints were shared across cell lines. First, chromosome 1p11 was involved in a translocation in four cell lines, translocation t(1;3)(p11.2-12; q11) in cell lines SNO and WHCO5 and translocations with differing partners in cell lines WHCO1 and WHCO6. Second, chromosome 3p11-12 was involved in translocations in two cell lines and deletions in, or near the EPHA3 locus was seen in four cell lines.

This gene codes for a receptor tyrosine kinase, with a tumor suppressor activity[49]. It was found to be mutated in lung and breast cancers[22-24] and interestingly, found to be deleted in 18.2% of ESCC patients in a previous study[31].

Breakpoints in, or near the centromeric regions of chromosomes 1 and 3 have previously been reported in several ESCC cell lines[10-12], as well as in fresh ESCC tumor samples that were karyotyped[14], and in ESCC cell lines obtained by in vitro transformation with HPV[50]. This strongly points to 1p11 and 3p11 translocation hotspots in ESCC that may affect genes and/or regulatory sequences not identified by SNP array. Deletions affecting or near the EPHA3 gene may point to a role for this gene which was previously found to be deleted in 18.2% of ESCC patients[31].

We knew from previous studies that these ESCC cell lines, all overexpress the EGFR gene[16]. EGFR DNA amplification observed in cell line SNO is likely to contribute to EGFR overexpression whilst other factors are likely to be involved in the other four cell lines where low levels of EGFR amplification were observed.

In view of the high clonal heterogeneity observed in each cell line, we used the GISTIC software in addition to CNAT to analyze SNP array data and evidence the significant targets of amplification and deletions. The most stable genetic rearrangements are thought to reflect a proliferative cell growth advantage. In this context, the GISTIC algorithm allowed us to prioritize amplicons and regions of deletions in term of their likelihood to host driver genes. The five most interesting significant regions of amplification included chromosomal regions: 11q13.3, 8q24.21, 11q22.1-q22.3, 5p15.2 and 3q11.2-q12.2in decreasing order of significance.

The 8q24.21, 11q13.3 and 3q11.2-q12.2 regions have all previously been reported in a variety of carcinomas[51-53] and they often co-exist with one another. Genomic amplification at 8q24 occurs in a large variety of cancers[51,53-56], and most amplicons described in the literature involve both the C-MYC and FAM84B genes, as was observed in this study in four ESCC cell lines. In previous reports the target of amplification has been attributed to either both genes[54], or to one or the other[55,56] based on their respective increased transcription.

The 11q13.3 amplicon covered a large region containing a number of potential target genes. The CCND1 and MYEOV genes (11q13.3) were co-amplified in four cell lines. Co-amplification of these genes has been reported in multiple myeloma, breast cancer and ESCC[14]. CCND1 is a downstream effector in the Wnt2/β-catenin pathway and the most frequent target of amplification in several ESCC studies[14,53,57]. While the MYEOV gene has been associated with cell proliferation in colon cancer[58], its amplification in ESCC is not always matched by increased transcription due to its silencing by epigenetic mechanisms[59]. The cortactin gene, CTTN, involved in cell motility[60], was previously shown to be over expressed in ESCC pre-cancerous lesions, as well as in carcinogen induced murine ESCC supporting a role for this gene in ESCC carcinogenesis[61]. The three fibroblast growth factor (FGF) genes, FGF3, FGF4 and FGF19 were part of the 11q13.3 amplicon. FGF and Wnt signaling pathways cross talk in a number of carcinogenesis scenarios[62]. Activated FGF receptors activate the FRS-GR2-GAB1-PI3K-AKT signaling cascade, and downregulate GSK-3β protein activity, thus hampering β-catenin phosphorylation and degradation[62]. In particular, FGF19 ligand downregulates GSK-3β activity, which results in the release and nuclear accumulation of β-catenin. Nuclear β-catenin activates the transcription of downstream genes including C-MYC and CCND1[63].

In addition to the amplification of Wnt pathway activators, the SFRP2 tumor suppressor gene locus was deleted at chromosome 4q22.1-q32.1 (Figure 7). The SFRP2 gene encodes a frizzled-related protein and is part of the SFRP family of Wnt inhibitors. Loss of SFRP2 is detected in medulloblastoma and is suggested to contribute to carcinogenesis through loss of inhibition of the Wnt pathway[64].

The copy number data therefore suggests that the Wnt signaling pathway may be at work in these ESCC cell lines through one or the combined effects of genes activating the β-catenin transcriptional activity and/or the FGF signaling pathways as well as deletions of genes, at 4q22.1-q32.1, inhibiting this pathway. Amplicons at both 8q24 and 11q13.3-13.4 have been described in a variety of squamous cell carcinoma[54,65-67] suggesting that the activation of pathways through the combined effects of genes at 8q24 and 11q13.3-13.4 contributes to the development and aggressiveness of SCC.

In addition to the SFRP2 gene, the three tumor suppressor genes BMR1B, SMAD1 and CASP6 were also targets of deletion at 4q22.1-q32.1. Both BMR1B and SMAD1 genes have previously been reported to have decreased expression in gliomas correlating with poor survival[68], and BMPR1B decreased expression in breast cancer is associated with increased cell proliferation and poor prognosis[69]. The CASP6 gene encodes the pro-apoptotic caspase-6 protein[70].

Although large 3q amplicons are commonly observed in squamous carcinoma[71,72] in this study the 6 Mb, 3q11.2-12.2 amplicon was focal and involved the MINA gene. This gene has previously been reported to be overexpressed in 83% of ESCC in one study and its inhibition was shown to suppress ESCC cell proliferation[73].

A 43 Mb region of deletion at 8p23.2-q11.1 (4 078 057-47 043 375 bp) was observed in the five cell lines and involved five potential target genes. These included the BNIP3L gene deleted or downregulated in prostate cancer and malignant melanomas, respectively[65,74], the LZTS1 gene, deleted in oral squamous cell carcinomas and downregulated in breast carcinomas[75,76]; and the three TNFR genes, TNFRSF10A, TNFRSF10B and TNFRSF10C, whose epigenetic inactivation was reported in gastric cancers[77]. An 8p loss, was previously detected by conventional CGH in a study performed on 29 South African black and colored ESCC patients[78]. Chromosome 8p22 loss has also been reported in prostate, breast, lymphoma, hepatocellular and colorectal cancers[79-83].

Active smoking is linked to increased fragile site expression[84] and is also one of the primary risk factors associated with ESCC in South Africa[85]. We previously hypothesized that deletions affecting anti-oncogenes located at fragile sites may contribute to the etiology of ESCC in South Africa and reported FHIT intragenic deletions in these cell lines and a small cohort of patients[40]. FHIT gene deletions were confirmed here supporting its role in ESCC carcinogenesis.

Deletion at 18q23 involved the SMAD4 and DCC genes in all cell lines. Both genes have previously been reported to be down regulated in ESCC either by deletion, mutation or methylation[86]. Decreased expression of the SMAD4 gene, a tumor suppressor of the transforming growth factor β family signaling pathway, has been associated with ESCC tumor invasion[87]. The DCC gene was shown to be frequently methylated in ESCC tumor specimens[88].

In summary, breakpoints at 1p11 and 3p11 were recurrent in the five ESCC cell lines and may point to genes such as EPHA3 that may be involved in ESCC carcinogenesis. Copy number alterations involved both amplicons previously reported in squamous cell carcinoma (8q24, 11q13 and 3q11) as well as novel regions of significant amplification (11q22.1-q22.3, 5p15.2 and 18p11.32). The finding that a significant number of genes that were amplified (FGF3, FGF4, FGF19, CCND and C-MYC) or deleted (SFRP2 gene) are involved in the Wnt and FGF signaling pathways suggest that these pathways may be activated in these ESCC cell lines. These results warrant expression studies of these genes in both cell lines and patients’ specimens. Of interest, should FGF gene expression be increased, ESCC patients may benefit from the respective FGF targeted therapies recently developed[89,90].

COMMENTS

Background

Esophageal squamous cell carcinoma (ESCC) is a major cause of cancer death in the world and has a peculiar epidemiology with worldwide geographic pockets of high incidence particularly in Asia and Africa. In South Africa, in the Eastern Cape region, ESCC represents the leading cancer affecting men and the second most common cancer in woman. There is a need to identify biomarkers to better understand the pathophysiology of this cancer and inform future diagnostic and therapeutic strategies for these patients. Established cell lines provide a unique resource to investigate both the presence of chromosomes translocations and copy number imbalances.

Research frontiers

A limited number of ESCC cell lines, all established in Asia, were reported to date. These have been investigated with molecular cytogenetic techniques of limited resolution and no key pathways have been reported previously.

Innovations and breakthroughs

This is the first comprehensive molecular cytogenetic study of five ESCC cell lines established in South Africa. The authors combined high-resolution whole genome array copy number analysis with conventional cytogenetics and multicolor fluorescence in situ hybridization to assess common chromosomal imbalances. Interestingly, a significant number of genes that were amplified (FGF3, FGF4, FGF19, CCND1 and C-MYC) or deleted (SFRP2 gene) are involved in the Wnt and FGF signaling pathways. In addition, a deletion within or near the EPHA3 gene was present in 4 of these cell lines, corresponding to a translocation breakpoint at 3p11.2 shared in some cell lines.

Applications

These results suggest that the Wnt and FGF pathways may be involved in the initiation or progression of ESCC. They also point to the EPHA3 gene as an added potential key gene. Further study on patients’ specimens and functional studies will determine the significance of these genes in ESCC pathogenesis.

Terminology

High-resolution 250K single nucleotide polymorphism (SNP) arrays cover at least one SNP per 100 kb of DNA, using an average of 24 probes per SNP. The copy number data are derived from the summary of non-polymorphic SNPs and examined as a ratio to a reference genome. Changes in intensity ratios are indicative of amplification/deletions. The Wnt and fibroblast growth factor (FGF) signaling pathways are involved in tissue homeostasis as well as in cell proliferation and differentiation. Although their mechanisms differ, these two pathways cross talk through GSK3β inhibition.

Peer review

The manuscript by Brown and coworkers demonstrates that a significant number of genes that were amplified or deleted are involved in the Wnt and FGF signaling pathways in five cell lines established from South African ESCC patients. They suggest that these pathways are activated in these cell lines. The overall goal of the paper is relevant. The data presented are solid and credible. The results are interesting and clinically important.