SEARCH

SEARCH BY CITATION

Keywords:

  • SOX2;
  • MUC5AC mucin;
  • mucinous colorectal cancer;
  • serrated polyps;
  • gastric differentiation;
  • metaplasia

Abstract

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Mucinous colorectal cancers are characterized by abundant production of intestinal goblet cell mucin, MUC2 and frequent ectopic expression of gastric foveolar mucin, MUC5AC. SOX2, an HMG-box transcription factor expressed in gastric mucosa but not in intestine is thought to play an important role in regulating transcription and expression of gastric differentiation related genes. Herein, we investigated the possible role of SOX2 in MUC5AC transcription and in the development of mucinous cancers. We observed good correlation between SOX2 and MUC5AC message levels in most colon cancer cell lines. SOX2 expression significantly transactivated MUC5AC promoter/reporter constructs in 3 of 5 colon cancer cell lines. We also examined SOX2 expression in normal stomach and colon, nonmucinous and mucinous colorectal cancers, serrated polyps and conventional adenomas using immunohistochemistry and in situ hybridization. SOX2 was expressed in the nuclei of both gastric foveolar cells and fundic glands by immunohistochemistry and in the cytoplasm by in situ hybridization. SOX2 was not expressed in normal colon but was strongly expressed in serrated polyps, mucinous and signet ring cell carcinomas, but rarely in nonmucinous carcinomas and tubular adenomas. Concordant expression of SOX2 with MUC5AC was observed in these lesions. Our results suggest that SOX2 is important in the upregulation of gastric foveolar mucin, MUC5AC in colorectal mucinous and signet ring cell carcinomas. In addition, the expression of both SOX2 and MUC5AC in serrated polyps supports the hypothesis that these polyps may be predominant precursors of mucinous and signet ring cell carcinomas of the colorectum. © 2007 Wiley-Liss, Inc.

Tissue specific transcription factors play a pivotal role in determining cellular phenotypes during development.1 These factors bind to target DNA in a sequence-specific manner up or down stream of the promoter. SOX is a family of transcription factors containing a high mobility group (HMG) of DNA binding domains and is expressed in a wide variety of tissues and plays important roles in the regulation of organ development and cell type specification.2–4 Early in embryogenesis, histone modification has been implicated in the regulation of SOX2 and cell fate determination.5 Among the SOX gene family, SOX2 gene is expressed not only in embryonic stem cells and neural stem cells, but also in the neuronal tissues, lens and stomach.6, 7In situ analysis of the chicken cSOX2 showed localized expression in the embryonic endoderm with transcripts appearing before the start of morphogenesis and cytodifferentiation in the rostral gut epithelium from the pharynx to the stomach.8 SOX2 is also important for the development of stomach in mice.9 In human gastrointestinal tissues, SOX2 expression was found in the nuclei of foveolar epithelial cells of the stomach but at very low levels in the intestine.10 Therefore, SOX2 has been suggested to play important roles in regulating transcription and expression of gastric differentiation related genes.7, 10

The majority of colorectal cancers are nonmucinous; mucinous and signet ring cell carcinomas are distinct subtypes of colorectal cancers. Mucinous cancers are characterized by abundant production of goblet cell mucin, MUC2.11, 12 Mucinous cancers make up 10 to 20% of all colorectal cancers with characteristic clinical/pathological and molecular genetic and epigenetic characteristics.11, 12 At least 2 histogenic pathways for colorectal cancer have been proposed; a well established and predominant pathway involving the adenoma-carcinoma sequence and a recently proposed alternative pathway with a hyperplastic polyp-serrated adenoma-adenocarcinoma sequence.13–19

Mucin genes are expressed in a tissue and cell type specific manner in the normal gastrointestinal tract.20, 21 However, aberrant expression of mucin genes have been observed in colorectal cancers.22–25 When compared to nonmucinous colorectal cancers, mucinous cancers showed overexpression of MUC2 intestinal goblet cell mucin and more frequent occurrence of ectopic expression of gastric foveolar mucin, MUC5AC, indicating simultaneous activation of both mucin genes during tumorigenesis.25

It has recently been reported that SOX2 transcripts which are expressed in normal stomach and the gastric-type of gastric carcinoma are expressed at considerably lower levels in both the mixed gastric and intestinal-type and solely intestinal-type of gastric cancers concomitant with decreased expression of gastric foveolar mucin, MUC5AC.10, 26 In addition, overexpression of SOX2 induced the mRNA expression of endogenous MUC5AC in COS-7 cells indicating that SOX2 plays a role in gastric mucin gene regulation.10 These observations led us to hypothesize that SOX2 may play a role in metaplastic differentiation in mucinous and signet ring cell colorectal cancers leading to the expression of gastric foveolar mucin, MUC5AC.

In this study, we examined the expression of SOX2 and MUC5AC gastric foveolar mucin in colonic neoplasms using immunohistochemistry. In situ hybridization was also used for determining SOX2 mRNA levels. Our results indicate for the first time that SOX2 expression is elevated in mucinous and signet ring cell colorectal cancers concomitant with increased levels of MUC5AC expression. Furthermore, elevated expression of both SOX2 and MUC5AC was observed in hyperplastic polyps, serrated adenomas, villous adenomas but not in tubular adenomas supporting the hypothesis that villous adenomas and serrated polyps may be precursors of mucinous and signet ring cell cancers. Our results further suggest that SOX2 is an important factor in the upregulation of MUC5AC gene transcription in mucinous and signet ring cell cancers.

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Formalin-fixed paraffin embedded tissue blocks were obtained from the Department of Pathology, University of California, San Francisco and the San Francisco Veterans Affairs Medical Center. Fifteen samples each of normal colons, hyperplastic polyps, serrated adenomas, tubular adenomas, villous adenomas, nonmucinous cancers, mucinous cancers, 7 signet ring carcinomas and 5 normal stomachs were used in this study. We excluded polyps from patients who had familial adenomatous polyposis, hereditary nonpolyposis colorectal cancer or hyperplastic polyposis. This study was undertaken with approval of the Institutional Review Board for Protection of Human Subjects at the University of California, San Francisco and the San Francisco Veterans Affairs Medical Center.

Histological evaluation

Serial sections (5-μm thick) were cut from the paraffin blocks and were used for H&E staining and immunostaining. All H&E stained sections were evaluated by 2 authors (E.T.P. and S.K.). Polyps with dysplasia without serrated architecture were classified as tubular adenomas. Polyps with serrated architecture without dysplasia were classified as hyperplastic polyps. Serrated polyps without morphological dysplasia and with features of abnormal maturation or abnormal proliferation (basal crypt dilation, crypt branching, deep serrations, mitoses in the upper half of crypts, dystrophic goblet cells and horizontal crypts) were classified as sessile serrated adenomas27. Polyps with dysplasia and serrated architecture were classified as serrated adenomas based on the definition of Longacre and Fenoglio-Preiser.14, 27 Colorectal cancers were defined as mucinous if extracellular mucin represented more than 50% of the neoplasm, and signet ring cell carcinoma if signet ring cells comprised more than 50% of the neoplasm.

Immunohistochemistry

Slides were deparaffinized with xylene and rehydrated using a graded series of ethanol. The sections were treated with 3% hydrogen peroxide in methanol and blocked with 10% nonimmune goat antiserum. After antigen retrieval in a pressure cooker (Biocare Medical, Walnut Creek, CA) for 15 min in 10 mM sodium citrate, pH 6.0, the tissue sections were incubated sequentially with primary antibody (polyclonal anti-SOX2 antibody; 1:200; Chemicon, Temecular, CA), biotinylated secondary antibody (Zymed, South San Francisco, CA), streptavidin-peroxidase conjugate (Zymed) and 3,3′-diaminobenzidine substrate (Zymed). The sections were mounted with Permount following light counterstaining with Mayer's hematoxylin (Zymed). Immunohistochemistry for MUC5AC and MUC6 were carried out as described.25

In situ hybridization

For riboprobe preparation and labeling, EcoRI/XhoI fragment, encompassing bases 335 to 1,570 of the sequence, was subcloned into pBluescript and the resulting plasmid was linearized with either EcoRI or XhoI Digoxigenin labeled riboprobes encoding sense and antisense strands were generated by Lofstrand Laboratories (Gaithersburg, MD) by in vitro transcription using T7 polymerases. Lofstrand Laboratories also performed alkaline hydrolysis of the riboprobes to generate fragments averaging 200 bases in length.

Tissue sections were deparaffinized with xylene and rehydrated through standard ethanol/water solutions. The sections were processed for in situ hybridization as described previously.28 Hybridization with probe was conducted overnight at 55°C in 0.3 M NaCl, 20 mM Tris HCl (pH 8.0), 5 mM EDTA, 1× Denhardt′s solution, 10% dextran sulfate, 50% formamide, 1 mg/mL yeast tRNA and 750 mg/mL salmon sperm DNA. High stringency removal of the nonhybridized probe and detection of the annealed probe using sheep anti-digoxigenin FAb fragment conjugated to alkaline phosphatase and nitroblue tetrazolium chloride/5-bromo-4-chloro-3-indolyl-phosphatase and nitroblue tetrazolium chloride/5-bromo-4-chloro-3-indolyl-phosphate substrate were performed as described previously.25 The sections were counterstained with nuclear fast red.

Analysis of immunohistochemical and in situ hybridization data

Immunohistochemical and in situ hybridization results were evaluated by 2 investigators (E.T.P. and S.K.) independently. The results were evaluated for both percentage and intensity of stained cells. An intensity value of 0, 1, 2 and 3 correlated to no, weak, moderate and strong staining of the cells, respectively. When <10% of positive cells stained and/or the intensity was 0 or 1, immunostaining was considered to be negative. The rest of the cases were scored as positive. Normal stomach tissues were used as positive controls.

RNA isolation and RT-PCR analysis

RNA was isolated from cell lines using Tri reagent (Molecular Research Center, Cincinnati, OH). RT reactions were conducted using 1.5 μg RNA and random primers with Superscript II reverse transcriptase (Invitrogen, Bethesda, MD) as described.29 PCR amplifications29 using 0.03 μg RNA equivalent cDNA initiated with a 5 min denaturation at 94°C followed by 35 cycles of 30 sec at 94°C, 30 sec at 59°C and 30 sec at 72°C. A final elongation of 6 min at 72°C was then used. Primers for MUC5AC were as described.30 The forward primer for SOX2 was 5′-CCCCCCTGTGGTTACCTCTT-C, the reverse primer was 5′-TTCTCCCCCCTCC AGTTCG. Quantum RNATM 18s rRNA internal standards were from Ambion, Austin, TX and used at a ratio of 1:9 primer to competitor. Amplification reactions for MUC5AC, SOX2 and 18s RNAs were conducted simultaneously in the same thermal cycler to minimize variability. The amplified products were visualized by ethidium bromide staining following electrophoresis on 1.2% agarose gels. Band intensity was quantified using NIH Image software, values for MUC5AC and SOX2 were normalized by dividing by values obtained for the 18s internal standard. Controls for the possible presence of genomic DNA consisted of samples processed as described above except without reverse transcriptase. Samples processed in this way always failed to yield an amplicon band detected by gel electrophoresis. Determination of the methylation status of the MUC5AC gene promoter in cell lines was determined as described.30

Transfection of MUC5AC promoter/reporter constructs

MUC5AC promoter/reporter vectors have been described previously.31 An expression vector for SOX2 was prepared in the pcDNA6/V5-HisA vector (Invitrogen). The SOX2 insert was prepared by PCR using NIH Mammalian Gene Collection clone #2823425 (human SOX2) as a template. The forward primer was 5′-AAGCTTGAATTCGCTAGCCACCATGTACAACATGATGGAGACGGAGC and the reverse primer was 5′-CTCGAGTCATTCGAACATGTGTGAGAGGGGCAGTGTGCCG. These primers allowed the amplicon to be cloned directly into pcDNA6/V5-HisA using Nhel and BstBl restriction sites (underlined) present in the forward and reverse primers, respectively. The forward primer also incorporates an optimal translational initiation site (CCACC) upstream of the start codon (bolded) for the SOX2 open reading frame. The SOX2 sequence was confirmed by DNA sequencing. Note that this construct also encodes the V5-epitope and a 6X His Tag immediately downstream of the SOX2 coding sequence, a feature of the pcDNA6/V5-HisA vector. The analogous LacZ-expressing control vector, pcDNA6/V5His/LacZ was obtained from Invitrogen.

Transfections were conducted in 24-well plates as described32 using cells at ∼50% confluency. Each transfection utilized 1.5 μg MUC5AC promoter/reporter vector, 1.0 μg SOX2-pcDNA6/V5-HisA or control- pcDNA6/V5-His/LacZ vector, 0.2 μg pRLO Renilla luciferase internal standard vector (Promega, Madison WI) and 15 μg SuperFect transfection reagent (Qiagen, Valencia, CA). Following 2-days incubation in complete media, reporter assays were conducted using the Dual-Luciferase Assay system (Promega) as described.32 Statistical significance was assessed using the 2-tailed t-test. Cla colon cancer cell clones were generated by transfection with the SOX2-pcDNA6/V5-HisA construct using SuperFect followed by selection using 5 μg/mL Blasticicin (Invitrogen) in DMEM containing 10% fetal bovine serum.

Statistical analysis

Immunohistochemical and in situ hybridization data were analyzed using the χ2 test and Fisher's exact test using software from StataCorp (College Station, TX). Luciferase reporter assay data are presented graphically as the average of triplicates with statistical significance determined by 2-tailed, paired student's t-test.

Results

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Relationship between SOX2 and MUC5AC expression in normal human tissues

We examined the relationship between SOX2 and MUC5AC expression in normal tissues using the National Center for Biotechnology Information's (NCBI) Gene Expression Omnibus databank. Searches using MUC5AC and SOX2 revealed an analysis of 36 normal tissue RNA samples performed using Affymetrix U133 human Gene Chips (Accession number GDS1096).30 This analysis revealed that significant levels of SOX2 are present in several neuronal tissues (amygdala, brain, caudate nucleus, cerebellum, corpus, fetal brain, hippocampus, spinal cord and thalamus); in pituitary gland and breast; and also in stomach and trachea. The expression detected in the stomach and trachea is of note because these tissues were the only 2 normal tissues in which MUC5AC was detected in this data set.

Correlation between SOX2 and MUC5AC expression in colorectal cancer cells and clones

The databank analysis performed above together with the work of Tsukamoto et al.33 and Li et al.10 suggests that SOX2 expression may upregulate MUC5AC expression, at least in some cell types. To examine this further we studied the correlation between SOX2 and MUC5AC message levels in colon cancer cell lines. Figure 1a shows a comparison of SOX2 and MUC5AC levels in various colon cancer cell lines. In general, cell lines that express high normalized SOX2 levels also express high levels of MUC5AC (linear regression correlation coefficient r2 = 0.305). There are exceptions to this however, for example SW620 and CACO2 cells express high levels of SOX2 but only low levels of MUC5AC. This indicates that factors other than SOX2 are involved in MUC5AC regulation. We have previously determined that methylation of the MUC5AC gene promoter is one factor that determines MUC5AC message levels in pancreatic cancer cells.30 We therefore examined these colon cancer cell lines to determine MUC5AC promoter methylation status. It was determined that H498, RKO, SW620 and CACO2 had methylated MUC5AC promoters. When cell lines without MUC5AC promoter methylation were analyzed separately (201, VACO5, C, LS174T, C1a and RW2982) better correlation between SOX2 and MUC5AC expression levels was observed (r2 = 0.542). This compares to r2 = 0.392 when the set of cell lines with MUC5AC promoter methylation is analyzed separately. We also examined a set of C1a cell line clones permanently transfected with a SOX2 expression vector (Fig. 1b). These clones were found to express a range of SOX2 message levels that varied more than 5-fold. MUC5AC message levels were found to vary over a 3-fold range in this experiment and correlation with SOX2 message levels was apparent (r2 = 0.618).

thumbnail image

Figure 1. Correlation between SOX2 and MUC5AC expression in colorectal cancer cell lines and clones. Panel a, RT-PCR was used to determine message levels from RNA samples isolated from various colon cancer cell lines. Solid square and the solid line indicates cell lines with unmethylated MUC5AC promoters, open circles and the dashed line indicates cell lines with methylated MUC5AC promoters. Panel b, RT-PCR analysis of the correlation between MUC5AC and SOX2 levels in C1a clones.

Download figure to PowerPoint

Transactivation of MUC5AC promoter/reporter constructs by SOX2 expression vector

The data described above suggests that SOX2 expression may upregulate MUC5AC transcription. We identified many potential sites for SOX2 binding in MUC5AC promoter region (Fig. 2a). To determine whether SOX2 may upregulate MUC5AC transcription activity, we examined the ability of SOX2 expression vector to transactivate a 2,892 bp MUC5AC promoter/reporter construct following transfection into various colon cancer cell lines. This length of MUC5AC promoter contains 13 potential SOX2 binding sites defined as (A/T)(A/T)CAA(A/T)G with 1 mismatch allowed in a unique nucleotide (Fig. 2a).34 A comparable LacZ expression vector was used as control in these experiments. We found that cotransfection with SOX2 expression vector significantly transactivated the MUC5AC promoter/reporter vector in 3 of 5 colon cancer cell lines, LoVo, Caco2 and SW620 (Fig. 2b). Two additional lines (C1a and LS174T) did not show transactivation by SOX2 expression vector. This demonstrates the ability of SOX2 to activate MUC5AC transcription in some colon cancer cell lines. Other cell lines, notably Cla and LS174T did not demonstrate MUC5AC promoter transactivation by SOX2 in this experiment. This may be caused by SOX2 saturation in these cell lines as they both possess relatively high levels of SOX2 (Fig. 1a). It is also possible that these cell lines lack an essential cofactor for MUC5AC promoter transactivation by SOX2. We also tested constructs containing various length MUC5AC promoters for the ability to be transactivated by SOX2 in SW620 cells. The M5-2892 construct, which contains 13 potential SOX2 binding sites was transactivated to the highest degree, 4.99-fold (Fig. 2c). Construct M5-2164, which contains 8 sites, was transactivated 3.12-fold. All shorter constructs which contained 4 or fewer potential SOX2 sites were transactivated 2-fold to 2.55-fold by SOX2 expression vector (Fig. 2c). This includes the M5-324 construct, which does not contain any cognate SOX2 binding sites.

thumbnail image

Figure 2. Transactivation of MUC5AC promoter/reporter constructs by SOX2 expression vector. Panel a, Diagram showing the length in base pairs (bp) of the MUC5AC promoter/reporter constructs used in this study, constructs were described previously.28 Bent arrows indicate transcription initiation points. Panel b, Transactivation of the 2,892 bp MUC5AC promoter/reporter construct by SOX2 expression vector in various cell lines. Significant transactivation by SOX2 expression vector as compared to the LACZ expression vector control at p < 0.0001 is indicated by **; significant transactivation at the p < 0.002 level is indicated by*. Panel c, Transactivation of various length MUC5AC promoter/reporter constructs by SOX2 expression vector in SW620 cells. Fold induction by SOX2 vs. LACZ control is indicated at the left side of the figure.

Download figure to PowerPoint

Expression of SOX2 in normal gastric mucosa

Immunohistochemical analysis with SOX2 antibody showed strong staining in the nuclei of normal gastric foveolar cells (Fig. 3b) and fundic glands (Fig. 3c). In situ hybridization for the expression of SOX2 gene transcript showed positive signal in the cytoplasm of normal gastric foveolar cells and fundic glandular epithelium (Figs. 3e and 3f). The sense negative control probe did not stain these cells (not shown). Also not shown, pyloric gland cells showed positive signal for SOX2 gene transcripts.

thumbnail image

Figure 3. Expression of SOX2 protein and transcripts in normal stomach by immunohistochemistry (a, b, c) and in situ hybridization (d, e, f). Immunohistochemistry reveals intense nuclear staining in normal stomach. Original magnification is 100× (a); arrows indicate intense nuclear staining in gastric foveolar cells (b) and in fundic glands (c). In situ hybridization reveals intense cytoplasmic staining in normal stomach, 100× (d). Intense cytoplasmic staining of gastric foveolar cells is indicated in (e) and in fundic glands (f). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

Expression of SOX2 in colorectal neoplasms

We examined the expression of SOX2 in normal colon, colorectal cancers and polyps of various histological types using immunohistochemistry (Fig. 4). SOX2 immunoreactivity was absent in normal colon (not shown). In mucinous cancers, strong immunostaining was observed in the nuclei of cancer cells (Fig. 4b). Hyperplastic polyps (Fig. 4c), serrated adenomas (Fig. 4d), sessile serrated adenomas and villous adenomas also showed strong immunostaining in the nuclei of mucosal cells (Table I). SOX2 immunostaining was negative in nonmucinous cancers (Fig. 4a) and tubular adenomas. These results are summarized in Table I. A significantly higher frequency of SOX2 immunostaining was observed in the hyperplastic polyps (100%), sessile serrated adenomas (73%), serrated adenomas (67%), villous adenomas (87%), mucinous cancers (80%) and signet ring cell carcinomas (93%). By contrast, only 7% each of nonmucinous cancers and tubular adenomas showed positive immunostaining. In situ hybridization for SOX2 transcripts in colorectal neoplasms showed expression patterns very similar to those observed by immunohistochemistry except that the positive signal was observed in the cytoplasms (not shown). The percent positive cases for SOX2 in situ hybridization were the same as those shown in Table I except that sessile serrated adenoma and serrated adenomas showed positive signals in 12/15 cases (80%) and 11/15 cases (73%), respectively.

thumbnail image

Figure 4. Immunohistochemical analysis of SOX2 (a, b, c, d) and MUC5AC (e, f, g, h) expression in colonic neoplasms. Nuclear and cytoplasmic SOX2 expression seen in mucinous cancers (b) and nuclear expression in hyperplastic polyps (c) and serrated adenomas (d). Nonmucinous cancer shows no staining (a). Intense cytoplasmic staining is observed in mucinous cancer (f), hyperplastic polyps (g) and serrated adenomas (h). Non-mucinous cancer shows no staining (e). Original magnification is 100×. Inset magnification for a, e, g and h are 400×; b, c, d and f are 1000×.

Download figure to PowerPoint

Table I. Immunohistochemical Analysis of SOX2 Expression in Colorectal Neoplasms
TissueNo. of casesSOX2 expressionPositive cases SOX2 (%)
+++*+++
  • p value was obtained using the Fisher's exact test.

  • 1

    p < 0.001 (HP vs. TA).

  • 2

    p < 0.01 (SSA vs. TA).

  • 3

    p < 0.01 (SA vs. TA).

  • 4

    p < 0.01 (VA vs. TA).

  • 5

    p < 0.001 (MC vs. NMC).

  • 6

    p < 0.001 (SRC vs. NMC).

  • *

    Grade of staining intensity: −, negative; +, focal weak staining; ++, moderate staining; +++, strong staining.

HP115 11400 15 (100)
SSA215 7224 11 (73)
SA315 5325 10 (67)
VA415 8232 13 (87)
TA15 00114 1 (7)
NMC15 00114 1 (7)
MC515 11102 12 (80)
SRC615 11301 14 (93)

Concordance of SOX2 and MUC5AC expression in colorectal neoplasms

Since SOX2 proteins and transcripts are expressed in both gastric foveolar cells and fundic and pyloric gland cells, we compared the pattern and frequency of SOX2 expression in colorectal neoplasms with those of MUC5AC, a gastric foveolar cell marker and MUC6, a fundic and pyloric gland cell marker. Positive MUC5AC staining was seen in the cytoplasm of mucinous cancers (Fig. 4f), hyperplastic polyps (Fig. 4g), serrated adenomas (Fig. 4h), sessile serrated adenomas and villous adenomas, but not in nonmucinous cancers (Fig. 4e) and tubular adenomas.

Excellent concordance between SOX2 and MUC5AC expression was noted in individual specimens of hyperplastic polyps, sessile serrated adenomas, serrated adenomas, villous adenomas, mucinous cancers and signet ring cell carcinomas (Table II). However, no correlation between SOX2 and MUC6 expression was observed in these lesions. We have carried out immunohistochemistry on successive sections obtained from the same paraffin block. The results show that SOX2 and MUC5AC were expressed in the same region in most cases. It is hence likely that these are expressed in the same tumor cells.

Table II. Concordance of SOX2 and MUC5AC or MUC6 Expression by Immunochemistry in Colorectal Neoplasms
TissuesNo. of casesPositive cases (%)Concordance (%)1
SOX2MUC5ACMUC6SOX2 vs. MUC5ACSOX2 vs. MUC6
  • 1

    Concordance was defined as positive SOX2 expression and positive MUC5AC or MUC6 expression or negative SOX2 expression and negative MUC5AC or MUC6. TA and NMC were not included in this analysis as only 1 of 15 cases of lesions were found to be positive for SOX2 (Table I).

HP1515 (100)15 (100)1 (7)15 (100)1 (7)
SSA1511 (73)14 (93)0 (0)12 (80)2 (13)
SA1510 (67)13 (87)1 (7)12 (80)3 (20)
VA1513 (87)14 (93)0 (0)13 (87)2 (13)
TA151 (7)5 (35)0 (0)1 (7)0 (0)
NMC151 (7)6 (40)0 (0)1 (7)0 (0)
MC1512 (80)11 (73)4 (27)12 (80)5 (33)
SRC1514 (93)12 (80)2 (13)13 (87)3 (20)

Discussion

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

SOX2 has been reported to be important in gastric differentiation and altered levels of SOX2 expression has been observed in different histological types of gastric cancers depending on the degree of intestinal metaplasia in the tumor.7, 10, 26 Forced SOX2 expression in COS-7 cells also induced the expression of endogenous MUC5AC.10 Since MUC5AC, a gastric foveolar cell mucin, has previously been reported to be expressed in histological subtypes of colorectal neoplasms, we hypothesized that SOX2 may be involved in MUC5AC gene transcription in subtypes of colorectal neoplasms.25 To investigate these possibilities, we first examined whether the correlation between SOX2 and MUC5AC message levels exist in colon cancer cell lines. Good correlation between the expression levels of MUC5AC and SOX2 was found in most cell lines (Fig. 1a). Two cell lines, SW620 and Caco2, which expressed low levels of MUC5AC had high levels of SOX2 transcript. Since we have previously shown that methylation of MUC5AC gene promoter is one of the factors that determines MUC5AC message levels in cancer cells, we thought that it is important to study the correlation between MUC5AC expression and SOX2 expression in cells whose methylation status was known.30 This allowed us to compare the correlation in 2 groups, methylated and unmethylated. As we suspected, better correlation was obtained when the cell lines without methylation of MUC5AC gene promoter were compared (Fig. 1a). We also observed good correlation of MUC5AC and SOX2 message levels in Cla cell line clones that expressed varying levels of SOX2 message (Fig. 1b).

Having found many potential sites for SOX2 binding in the MUC5AC promoter, we examined the effects of SOX2 on MUC5AC gene transcription using MUC5AC promoter/reporter constructs in colon cancer cell lines. We found that SOX2 expression vector significantly transactivated the MUC5AC promoter/reporter vector in 3 of 5 colon cancer cell lines (Fig. 2b). Two cell lines, Cla and LS174T that expressed high basal levels of MUC5AC message (Fig. 1a) did not show significant transactivation by SOX2 expression vector. When the constructs containing various length MUC5AC promoters were examined for the ability to be transactivated by SOX2 in SW620 cells, the M5-2892 construct which contains 13 potential SOX2 binding sites was transactivated to the highest degree, 4.99-fold followed by the M5-2164 constructs containing 8 sites which transactivated 3.12-fold (Fig. 2c). All shorter constructs which contained 4 or fewer potential SOX2 sites were transactivated 2-fold to 2.55-fold. This indicates that SOX2 sites between bases 2,164 and 2,892 may be important in SOX2 induced MUC5AC gene transcription. The databank analysis on the microarray data showed high levels of SOX2 expression in neuronal tissues, pituitary gland, breast, stomach and trachea. The expression detected in stomach and trachea is of interest because these tissues were the only 2 normal tissues that expressed MUC5AC in this data base.

Taken together these data suggest that SOX2 may induce trans-acting factors or cofactors that activate MUC5AC transcription in addition to acting directly upon the MUC5AC promoter itself. SOX family of transcription factors regulate target genes by pairing with specific partner factors.3, 4, 6 This partnering allows SOX proteins to act in a cell specific manner, which is a key to their role in cell differentiation. For example, SOX2 regulates a set of genes such as undifferentiated embryonic stem cell transcription factor (utf-1) and Fgf-4 in embryonic stem cells (ES), embryonic carcinoma cells and inner cell mass by pairing with transcription factor Oct 3/4.6 SOX2 regulates a target gene, δ-crystalline to induce lens specific differentiation when partnered with δEF3 partner factor.6 There are several possible binding sites for both Oct3/4 and δEF3 on the MUC5AC promoter. However, identity of the partner proteins responsible for tissue and cell type specific differentiation for stomach, breast, trachea and neuronal tissues is not yet known.

Our immunohistochemical data showing SOX2 staining in the nuclei of gastric foveolar cells confirms previous reports.7, 10, 26 We also observed that SOX2 was expressed in the fundic and pyloric glandular epithelium as well. Thus, SOX2 may serve as a marker for gastric foveolar cell as well as fundic and pyloric glands. To our knowledge, this is the first in situ hybridization analysis for SOX2 in the human stomach. SOX2 expression in gastric types of gastric cancer but decreased expression of SOX2 in the mixed gastric and intestinal type and solely intestinal type of gastric cancers have previously been reported.7, 10, 26 However, the patterns and the levels of expression of SOX2 in colorectal neoplasms have not previously been reported. The present study shows for the first time that SOX2 expression occurs much more frequently in colorectal mucinous (80%) and in signet ring cell cancers (93%) compared to nonmucinous cancers (7%) (Table I).

It is also clear from this study that SOX2 is expressed in hyperplastic polyps, sessile serrated adenomas, sessile adenomas and villous adenomas but not in tubular adenomas (Table I). This pattern of expression correlated well with the MUC5AC expression in these lesions (Table II). These results provide further evidence for a causal link between SOX2 expression and MUC5AC induction. Furthermore, coexpression of SOX2 and MUC5AC in these lesions and mucinous and signet ring cancers provides added support for villous adenomas and serrated polyps being precursor lesions for mucinous and signet ring cell cancers.25 Since SOX2 is expressed in gastric foveolar cells as well as fundic and pyloric glands and SOX2 expression occurs ectopically in the subtypes of colorectal neoplasms, one would expect the ectopic expression of both MUC5AC, a gastric foveolar cell marker and MUC6, a fundic and pyloric gland marker in these neoplasms.20, 22, 25, 28 Interestingly, however, concordant expression of SOX2 was observed with MUC5AC but not with MUC6 in these lesions (Table II). The reasons for the differential expression of MUC5AC and MUC6 in SOX2 expressing colonic neoplasms are not clear. Since SOX2 requires pairing with specific partner proteins for activating target genes to induce cell- and tissue-type specific differentiation, it is possible that concurrent activation of SOX2 with gastric foveolar cell specific SOX2 partner factors, but not with the fundic and pyloric gland-specific partner factors may occur in these neoplasms.4, 6

Ectopic expression of cell or organ type specific markers has been observed in chronic inflammation and malignancy and has been characterized by metaplastic morphological changes.7, 10, 25, 26, 28, 35, 36 Intestinal metaplasia in the stomach is associated with chronic atrophic gastritis and is the most common precursor condition for gastric carcinoma.7, 10, 26 It is characterized by the transdifferentiation of gastric epithelial cells to an intestinal phenotype, accompanied by intestine-specific genes such as MUC2 and sucrase/isomaltase.10, 26 Transcription factors such as CDX1 and CDX2 have been suggested to play key roles in the induction of intestinal metaplasia in the stomach.36 Barrett's esophagus with intestinal metaplasia of the esophagus develops when the normal squamous esophageal epithelium is replaced by intestinal type columnar epithelium during the process of healing associated with chronic inflammation.37 It is also characterized by expression of the intestine specific genes such as MUC2 and TFF3.38 The ectopic expression of gastric foveolar cell specific mucin gene MUC5AC has recently been reported in pancreatic cancers and their precursor lesions as well as in serrated types of colorectal polyps and mucinous colorectal cancers.25, 28 The present study confirms these observations. The origin of precursor cells leading to metaplasia is not clear. Studies using cell markers suggest that metaplasia arises from pluripotent cells in the gastrointestinal mucosa.39, 40 However, a recent report on rodent model of inflammation suggests an alternative mechanism that the progenitor cells arose in the animals' bone marrow.41, 42 The mechanisms by which these cells differentiate and the factors leading to the proliferation of these metaplastic cells are largely unknown. Whether the tumors arise as a consequence of chronic inflammation or through other mechanisms, it is widely accepted that cancer stem cells represent a very small fraction of cells in the tumor and give rise to tumor cells of varying proliferative potential and heterogenous phenotypes.43–48

We have recently reported that HATH1, a transcriptional factor required for goblet cell biogenesis upregulates MUC2 goblet cell mucin gene expression in colon cancer cells and is strongly expressed at high levels concomitant with MUC2 expression in mucinous colorectal cancers as well as in serrated types of polyps and villous adenomas but not in nonmucinous cancers or tubular adenomas.49 The present study has demonstrated that SOX2 transcription factor upregulates MUC5AC gastric foveolar cell mucins in colon cancer cells and both SOX2 and MUC5AC are strongly and concordantly expressed in mucinous cancers, serrated types of polyps and villous adenomas. These results together suggest that simultaneous activation of intestinal goblet cell and gastric foveolar cell lineage differentiation occurs frequently in the pathogenesis of mucinous colorectal cancers and that these processes are mediated by activation of HATH1 and SOX2 transcription factors and associated factors. Lastly, the present study shows that SOX2 transcription factor upregulates MUC5AC gastric foveolar mucin gene transcription and that SOX2 is concordantly expressed with MUC5AC in these neoplasms, adding further support to the hypothesis that serrated polyps and villous adenomas may be precursors of mucinous and signet ring cell colorectal cancers.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

The authors thank Dr. Rina Wu (Zymed) for aid with immunohistochemistry.

References

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References