Connexin43 is overexpressed in ApcMin/+-mice adenomas and colocalises with COX-2 in myofibroblasts
Article first published online: 30 MAR 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 116, Issue 3, pages 351–358, 1 September 2005
How to Cite
Husøy, T., Knutsen, H. K., Cruciani, V., Ølstørn, H. B., Mikalsen, S.-O., Løberg, E. M. and Alexander, J. (2005), Connexin43 is overexpressed in ApcMin/+-mice adenomas and colocalises with COX-2 in myofibroblasts. Int. J. Cancer, 116: 351–358. doi: 10.1002/ijc.21025
- Issue published online: 10 JUN 2005
- Article first published online: 30 MAR 2005
- Manuscript Accepted: 22 DEC 2004
- Manuscript Received: 26 JUL 2004
- Research Council of Norway
- Norwegian Cancer Society
- intestinal cancer
The expression of gap junction proteins, connexins, in the intestine and their role in tumorigenesis are poorly characterised. Truncating mutations in the tumour suppressor gene adenomatous polyposis coli (APC) are early and important events, both in inheritable (familial adenomatous polyposis, FAP) and spontaneous forms of intestinal cancer. Multiple intestinal neoplasia (Min) mice, a FAP model with inherited heterozygous mutation in Apc, spontaneously develop numerous intestinal adenomas. We recently reported reduced expression of connexin32 in Paneth cells of Min-mice. We further examine the expression of connexin43 (Cx43) and other connexins as a function of heterozygous and homozygous Apc mutation in normal intestinal tissues and adenomas of Min-mice. Qualitative analysis of connexin mRNA in intestine revealed a similar expression pattern in Min- and wild-type (wt) mice. Connexin26 and connexin40 proteins were found in equal amounts in Min and wt epithelia of large and small intestine, respectively. Interestingly, the connexin43 level was increased in the stroma of Min-mice adenomas, in close proximity to epithelial cells with nuclear β-catenin staining. Cx43 and COX-2 were located to the same areas of the adenomas, and immunostaining exhibited coexpression in the myofibroblasts. Prostaglandin E2 induces Cx43 expression and COX-2 is the rate-limiting enzyme in the prostaglandin synthesis. However, the COX-2-specific inhibitor, celecoxib, did not reduce Cx43 expression. Although both Cx43 and COX-2 are target genes for β-catenin, they were overexpressed in stromal cells but not in epithelial tumour cells. We hypothesise that gap junctions may be of importance in the transfer of signals between epithelium and stroma. © 2005 Wiley-Liss, Inc.
Impairment of gap junctional intercellular communication (GJIC), either by mutation or loss of function, is involved in cancer development as well as other diseases.1, 2, 3 Gap junctions are the only channels that allow direct exchange of small metabolites between cells. The building proteins of gap junction channels, connexins (Cx), constitute a family with as many as 20 distinct isoforms in mammals. They are present in most tissues and are involved in a diversity of biologic processes. The functional importance of gap junctions in the intestine and their possible role in tumorigenesis have been poorly characterised, although gap junctions are found both in the epithelial and muscular tissues.4, 5
One of the main and very early events in intestinal carcinogenesis is truncating mutations in the tumour suppressor gene adenomatous polyposis coli (APC). Mutations in APC are both involved in inheritable and spontaneous forms of intestinal cancer, where >80% of sporadic adenomas have mutations in APC.6, 7, 8 Germline mutations in APC lead to familial adenomatous polyposis (FAP), an autosomal dominant condition characterised by the development of hundreds of colorectal adenomatous polyps in the second to third decade of life.9, 10
Apc plays important roles in a wide range of cellular functions such as proliferation, migration, differentiation and apoptosis.11 Truncated Apc disturbs these functions, but the underlying mechanisms are not fully understood. Apc is important in Wnt signalling, where it participates together with axin and glycogen synthase kinase 3β (GSK3β) to target the adhesion molecule and transcription factor β-catenin for degradation.12 In the absence of functional Apc, β-catenin accumulates in the nucleus where it can turn on transcription of several genes, including the gap junction protein Cx43, COX-2, cyclin-D1 and PPARδ.13, 14, 15, 16. However, it is believed that both Apc alleles must be mutated to activate the Wnt signalling pathway.17, 18 In adenomas of multiple intestinal adenoma (Min) mouse, a FAP model that spontaneously develops adenomas in the small intestine, the remaining wild-type (wt) Apc is often lost or mutated.19 Consequently, the Wnt-signalling pathway is activated in Min-mice adenomas, as shown by an increase in β-catenin and β-catenin/Tcf targets like cyclin-D1 and PPARδ (data not shown).20, 21
In addition to its role in Wnt-signalling, Apc can bind microtubules, and truncated Apc disturbs chromosome segregation, leading to chromosomal instability.22, 23 We reported that heterozygous mutation in Apc is sufficient to reduce microtubule polymerisation and Cx43 expression in intestinal cell lines.17 Subsequently, others have provided evidence in support of our findings on microtubules.24 Most interestingly, we recently found that Cx32 expression was markedly reduced in Paneth cells of Min-mice.25 Altogether, this suggests Cx to be downstream targets for mutations in one and both Apc alleles.
Our study has 2 objectives. First, we wanted to test the hypothesis that Cx43 as a target for β-catenin signalling is upregulated in adenomas of Min mice. The second objective was to investigate the general expression of Cx in the intestine of Min-mice compared to wt-mice at the mRNA level. For Cx where suitable antibodies were available, we further studied the expression at the cellular level as a function of heterozygous and homozygous Apc mutation in normal intestinal tissues and adenomas of Min-mouse, respectively.
Material and methods
A polyclonal rabbit anti-Cx43 antiserum was raised against the 20 C-terminal amino acids of Cx43. A tyrosine was added N-terminally to this peptide for coupling purposes. The antiserum recognises Cx43 from many cell types both in Western blots and immunofluorescence experiments.26 Polyclonal anti-Cx26 (cat. no. 51-2800) was from Zymed Laboratories (San Francisco, CA), and polyclonal anti-Cx40 (cat. no. Cx40A) was purchased from Alpha Diagnostic (San Antonio, TX). Antibody against β-catenin (cat. no. C19220) was obtained from Transduction Laboratories (Lexington, KY). Anti-β-actin (cat. no. M0851) and anti-von Willebrand Factor (vWF) (cat. no. A 0082) was purchased from DakoCytomation (Glostrup, DK). FITC conjugated anti-cleaved-caspase-3 (cat. no. 9667) was from Cell Signalling Technology (Hertfordshire, UK), and anti-COX-2 (cat. no. 160107) was from Cayman Chemical (Ann Arbor, MI). Anti-vascular endothelial growth factor (VEGF) (cat. no. sc-7269) was obtained from Santa Cruz Biotechnology (Santa Cruz, CA), and FITC conjugated anti-α-smooth muscle actin (αSMA) (cat. no. F 3777) was from Sigma (St. Louis, MO). Goat anti-mouse and goat anti-rabbit secondary antibodies conjugated to horseradish peroxidase (HRP) were obtained from Bio-Rad (Hercules, CA).
Animals and celecoxib treatment
The mice were bred at the Norwegian Institute of Public Health, Oslo, Norway, from inbred mice originally purchased from The Jackson Laboratory (Bar Harbor, ME). The Min/+ pedigree was maintained by mating C57BL/6J-+/+ (wt) females with C57BL/6J-Min/+ males, and procedures to secure inbreeding were followed. The Min/+ mice were identified by allele-specific polymerase chain reaction on DNA isolation from blood as described previously.27 Water and diet were given ad libidum.
Celecoxib from Pharmacia (Oslo, Norway) was diluted in maize oil to a concentration of 6 mg/250 μl and given to Min mice (n = 3, 13–16 weeks) by gavage (230 mg/kg bw/day) for 4 days. Control mice (n = 3) were given the same amount of maize oil. On the fifth day, the mice were put to death by cervical dislocation. Whole small intestine and colon were fixed in formalin for 72 hr before they were rolled up and embedded in paraffin. Adenomas and corresponding normal epithelia/mucosa were scraped of the muscular layer from fresh intestine, snap-frozen in liquid nitrogen and stored at −70°C until further use.
To increase the numbers of adenoma in the colon, we treated Min-mice with azoxymethane (AOM). AOM from Sigma was diluted in 0.9% NaCl, and 5 mg/kg bw (3 Min and 3 wt mice) or 7.5 mg/kg bw (2 Min and 2 wt mice) was injected (10 μl/g body weight) s.c. at 1 and 2 weeks after birth. The mice were inspected weekly. The mice were put to death by cervical dislocation at 18–21 weeks of age. The intestines were fixed for 24–48 hr before they were subjected to immunohistochemical analysis.
RT-PCR and sequencing of PCR products
RT-PCR was performed on RNA extracted from the Min- and wt-mice intestine using the Strata Prep total RNA Miniprep Kit (Stratagene, Amsterdam, the Netherlands). Three independent extractions were made from each tissue. Potential contaminating DNA was removed from the RNA using DNA-free kit (Ambion, Austin, TX). RNA was reverse transcribed by Superscript II (Invitrogen, Carlsbad, CA) and the cDNA was amplified by standard PCR procedures using the primer sets shown in Table I. Typically, PCR reactions consisted of 32–35 cycles (denaturing: 95°C for 45 sec; annealing: 46–60°C for 30 sec; elongation: 72°C for 30 sec). After cleaning and treatment with Exonuclease I and shrimp alkaline phosphatase (ExoSAP-IT; USB, Cleveland, OH), a dye terminator cycle sequencing reaction was performed using the DYEnamic ET dye terminator kit (Amersham Biosciences, Little Chalfont, UK). The products were thereafter analysed on a MegaBACE (Amersham Biosciences) to ensure the specificity of the primer sets. All kits were used according to the suppliers' recommendations.
|Cx||Upstream primer||Downstream primer||Prod. size (bp)|
|26||5′-TTA AGG ACA TCG AAG AGA TCA-3′||5′-AGA ATG CAA ATT CCA GAC AC-3′||268|
|29||5′-GGT TTT CGG CAA TGA T-3′||5′-AGA AGC TTG AGG CTT TTA GC-3′||278|
|29||5′-CGY TTC TGG GYC TTC CAR GTC AT-3′||5′-AAA CCC ARA AGC RCA AGC TCC A-3′||425|
|30||5′-TTC CTT TAC AAT GGG TAC CA-3′||5′-TTA AGC AGC ATG CAA ATC AC-3′||152|
|30.2||5′-TGG GGG AGT GGG CGT TCC TA-3′||5′-CGT GTT RCA CAC GAA CTC CTC-3′||169|
|30.3||5′-AGC CAR CAT GAA CTG GGS ATT TCT-3′||5′-TCC ATG CAY CTC TTG CCC ACC A-3′||651|
|31||5′-GCG CAT CTG GCT GTC AGT AG-3′||5′-CTA TGC TGG CGC ACT GTA CC-3′||440|
|31.1||5′-TGR GTC CAC CAT GAA CTG GAG-3′||5′-YTC TCA CCC TGG GYC ACA CAG GAA-3′||1134|
|32||5′-CAA CAC ATA GAG AAG AAA ATG CTA CGG-3′||5′-CAT GAA GAC GGT GAA GAC GGT T-3′||288|
|33||5′-GCC TTA CAC CAG CTC CTA GAA-3′||5′-GCC TTA AAT AAG ACA TAG AAC AGC-3′||689|
|36||5′-ATT GTR GGG GAG ACG GTG TA-3′||5′-TAG AYT GAC TTY CTC TTG GCC T-3′||821|
|37||5′-ACT CGA CCG TGG TGG GYA AGA TCT-3′||5′-ACT GGC CAT AGA GGA AGC CTG CCT-3′||471|
|39||5′-TTG GCA GGG TCA CCT ATC TA-3′||5′-CCG CGG TGT AGG GTG TAA GT-3′||185|
|40||5′-CAG GGC ACC CTA CTC AAC-3′||5′-TVT TTA GGC ACT GAT TGA AGT-3′||382|
|43||5′-CTG CCT TTC GCT GTA ACA CT-3′||5′-CGC TCA AGC TGA ACC CAT A-3′||399|
|45||5′-CCA ACC MAA ACC TAA GCA TG-3′||5′-ACA TAA AAC GGG TGG ACT TG-3′||156|
|46||5′-GGA SGT GTG GGG YGA YGA GCA RTC RGA CTT CAC-3′||5′-CCK CTC TTT CTT CTT CTC CTC CAT GCG CAC GAT-3′||202|
|46||5′-CAG GCC CAC RGA GAA GAC CAT CTT-3′||5′-CCT GCT TGA GCT TCT TCC AGC CCA-3′||113|
|46.6||5′-AAY CAY TCC ACC TTC GTG GGCA-3′||5′-GCC ATC TCA CAG AGG TTG AGG A-3′||252|
|50||5′-GCC ACA TCA TCT TCA AGA CC-3′||5′-TAG CCC TTG GCT TTC TGG AT-3′||340|
|57/62||5′-GAT AGG CCA ATA TAT TCT CTA TGG GT-3′||5′-GCA ATG CTG TGC ATR AAA AGC A-3′||141|
Formalin-fixed paraffin-embedded tissue blocks were cut at 4 μm and the sections were prepared for immunostaining. Epitope demasking was performed in a microwave oven for 12 min in Tris-EDTA pH 9.1 for β-catenin, Cx43, Cx40, Cx26 and COX-2. Immunostaining with anti-β-catenin (1:6,400), anti-Cx40 (1:100) and anti-Cx26 (1:40) antibody was performed using a DAKO Autostainer system according to the manufacturer's instruction. Anti-Cx43 (1:5,000) and anti-COX-2 (1:500) staining was done with Vectastain ABC kit (Vector Laboratories, Burlingame, CA).
Double immunofluorescence staining with anti-Cx40 (1:50) and FITC conjugated anti-active caspase 3 (1:10), anti-Cx43 (1:5,000) and anti-α-SMA-FITC (1:200), anti-COX-2 (1:500) and anti-α-SMA-FITC (1:200), and anti-Cx43 (1:5000) and anti-vWF-FITC (undiluted) were performed by first treating the sections with the nonconjugated antibody followed by anti-rabbit-Texas red (1:50) and then with the FITC-conjugated antibody. Anti-vWF was conjugated to FITC using the FluoroTag™ FITC Conjugation kit from Sigma.
Frozen adenomas and normal tissue were disrupted under liquid N2 and dissolved in sample buffer (6.25 mM Tris-base, pH 6.8, 5% glycerol, 2% sodium dodecyl sulphate, 13% 2-mercaptoethanol and bromophenolblue), followed by sonication and heating at 95°C for 5 min. The protein content was measured by a spectophotometer using the Bradford procedure. Equal amount of protein (about 40 μg) were loaded in each lane and separated on a discontinuous SDS-PAGE gel, followed by electroblotting onto nitrocellulose membrane. The membranes were blocked in 5% skimmed milk for 1 hr and probed with anti-Cx43 (1:500), anti-COX-2 (1:200), anti-VEGF (1:500) or anti-actin (1:250,000) at 4°C overnight. The membrane was then treated with HRP-conjugated secondary antibody for 1 hr, and the proteins were detected by chemiluminescence (Pierce, Rockford, IL). Quantification of band intensities was done by Kodak 1D image analysis software (Eastman Kodak Company, Rochester, NY). Statistical analyses of the band intensities were performed by a t-test using SigmaStat. Normal tissue samples (n = 5–6) were compared to tumour tissue samples (n = 6), irrespective of which animal it was taken from.
Expression of Cx in intestine of Min- and wt-mice
The expression at the mRNA level of all 19 mice Cx was investigated in Min-mouse and wt-mouse by RT-PCR in proximal and distal intestine and the colon. The intestine expressed Cx26, Cx29, Cx30.2, Cx30.3, Cx31, Cx31.1, Cx32, Cx36, Cx37, Cx40, Cx43, Cx45 and Cx50 (Table II). Only 1 minor difference was found in the Cx expression in the intestinal tissues of Min-mouse and wt- mouse at the mRNA level. However, as RT-PCR is not a quantitative method, only a clear presence (+) vs. a clear absence (−) was regarded as a difference in expression. Hence, any difference in Cx expression in normal tissue between Apc+/+ and ApcMin/+, as we previously reported,25 would be confined to specific cells and tissues within an organ.
Cx40 was expressed with similar intensity in a few cells in the top of the villi in the normal small intestine of both Min-mice (Fig. 1a) and wt-mice (not shown). The large intestine did not contain any Cx40-positive epithelial cells (not shown). As previously reported for other tissues, Cx40 was also found in the endothelium of blood vessels in Min-mice intestine (Fig. 1a*) and wt- mice (not shown). Enterocytes in the top of the villi are programmed to undergo apoptosis. To see whether expression of Cx40 was associated with apoptosis, we studied whether Cx40 was colocalised with active-caspase3 (Fig. 1c–f). However, Cx40 and active-caspase3 were only partially colocalised (Fig. 1f). Cx26 was found in the colon enterocytes of the normal Min-mouse intestine, with strongest expression in the upper part of the crypts (Fig. 1g) and in the apical part of the cells. A clear increase in Cx26 level was observed toward the proximal colon (Fig. 1g). There was no observable difference in the level of Cx26 expressed in the normal tissue between Min-mouse and wt-mouse. No Cx26 was found in the small intestinal epithelia of Min-mice or wt-mice (not shown). As we recently reported, Cx43 was located to the muscular tissue in the small and large intestine, with no difference in the level between Min- and wt-mice and no staining of the epithelial cells.25
Connexin localisation in intestinal adenomas of Min-mice
We further studied whether there could be an aberrant expression of Cx in the adenomas. Cx40 was found in a small fraction of the epithelial cells localised near the luminal part of the adenomas in the small intestine of Min-mice (Fig. 1b). Cx26 was found in epithelia in some colon adenomas with variable intensities (Fig. 1h).
β-catenin expression is induced in adenomas from Min-mice, and Cx43 is known to be a functional target for β-catenin/Tcf-mediated transcription. It was therefore of special interest to examine the expression and localisation of Cx43 in Min-mice adenomas. Cx43 expression was clearly increased in adenomas from small and large intestine of Min-mice (Fig. 2a,c,e,g), with localisation in the luminal part of the small intestinal adenomas (Fig. 2a,c). However, serial sections of Min-mouse adenomas showed that Cx43 and β-catenin were not expressed in the same cell types (Fig. 2). β-catenin was found in the cytoplasm and nuclei of the epithelial cells of intestinal adenomas (Fig. 2b,d,f,h),20 and Cx43 was induced in the connective tissue in close proximity to epithelial cells, where we generally found a more intense nuclear β-catenin staining in the luminal edge of the tumour (Fig. 2c,d,g,h).
Colocalisation of Cx43 and COX-2 and the effect of celecoxib
Another target for β-catenin-mediated transcription, COX-2, has previously been reported to be localised to the connective tissue in intestinal tumours from Min-mice.28 Since this reminds much of the distribution of Cx43 found in adenomas (Fig. 2), we examined the localisation of COX-2 in the adenomas. The COX-2 level was increased in the connective tissue in the luminal part of the adenomas (Fig. 3a,c), and in neighbouring sections it colocalised with Cx43 in the same areas of the adenomas, shown in a representative picture from several sections studied (Fig. 3b,d). The epithelial cells in adenomas did not stain for COX-2.
Prostaglandin E2 (PGE2), which is a product in the prostaglandin synthesis, induces Cx43 expression.29 COX-2 is a rate-limiting enzyme in the prostaglandin synthesis pathway. We therefore treated Min-mice with a dose (6 mg/mice) of the COX-2-specific inhibitor, celecoxib, reported to be effective,30 to investigate whether a reduction in PGE2 could reduce Cx43 expression in adenomas and normal intestinal epithelia from Min-mice by Western blotting. In agreement with the immunohistochemical results, adenomas exhibited higher amounts of both Cx43 and COX-2 (p = 0.004 for control and p = 0.03 for celecoxib treatment) than the normal tissue (Fig. 4). Quantification of the ratio between Cx43 level in adenoma and normal tissues from celecoxib-treated animals compared to control animals revealed no significant reduction in Cx43 expression after celecoxib treatment (Fig. 4a,b). As for Cx43, celecoxib treatment did not change the COX-2 level (Fig. 4a,b).
VEGF induces expression of Cx43 in cell lines and VEGF expression is also increased by β-catenin and prostaglandins.31, 32, 33 We therefore investigated the expression of VEGF in adenomas and normal tissue from the small intestine. No increase in VEGF was found in the adenomas compared to the normal intestinal tissue in the Min-mouse, and VEGF expression was not affected by celecoxib treatment (Fig. 4a,b).
Cell-specific induction of Cx43 and COX-2 in myofibroblasts in adenomas
Because COX-2 was reported to be expressed by myofibroblasts in the intestinal adenomas of Min-mice,28 we examined if this could also be the case for Cx43. Myofibroblasts express α-SMA. α-SMA is currently the best immunohistochemical marker for this cell type, although α-SMA is also expressed in endothelial cells. Cx43 (Fig. 5a) was clearly colocalised with α-SMA (Fig. 5b) in the luminal part of the adenomas as shown by merging Cx43 staining and α-SMA staining of the same section (Fig. 5c). This is even more clearly seen in pictures taken with a 100× objective, showing that exactly the same cells express α-SMA and Cx43 (Fig. 5d–f). The colocalisation of COX-2 and Cx43 in the same cells in the Min-mice adenomas was finally determined by showing that COX-2 also is detected in the α-SMA expressing myofibroblasts (Fig. 5g–i). No Cx43 was detected in the myofibroblasts of the normal intestinal tissue (Fig. 5j–l). To ensure that the Cx43-positive cells were not endothelium, we also double-stained tissue section with Cx43 and the endothelial marker vWF. No colocalisation was found between vWF and Cx43, showing that Cx43 is not expressed in endothelial cells in intestinal adenomas in Min-mouse (Fig. 5m–o).
We have recently reported reduced Cx expression caused by heterozygous mutation in the tumour suppressor gene Apc, both in intestinal cell lines and in Paneth cells in intestine of Min-mice.17, 25 A broad study of Cx expression in the intestine of Min-mice and wt-mice was therefore performed. No clear difference in Cx expression was observed at the mRNA level in any part of the intestine of Min-mouse compared to wt-mice. However, cell-specific difference in expression level, as we previously reported for the intestine,25 cannot be ruled out.
We found Cx40 localised to a few epithelial cells in the top of the villi in the normal intestinal tissue of Min-mice and in endothelial tissue as previously reported.34, 35 The position of the Cx40-positive cells could suggest a function in apoptosis, but only partial colocalisation with active-caspase 3 was found, indicating that Cx40 is not generally involved in apoptosis. However, altered gap junction function seems to be related to induction of apoptosis.36
We further studied the effect of homozygous mutations of Apc, as found in adenomas of Min-mice, on Cx expression and localisation. The protein level of Cx26 and Cx40 was more or less unchanged in adenomas of Min-mice compared to the normal tissue. Most interestingly, we found an increased expression of Cx43 in the stroma of intestinal tumours of the Min-mice. Furthermore, the increased Cx43 expression in stromal myofibroblasts was colocalised with COX-2, an expression that is well known to be increased in both humans and mice adenomas.37, 38 Similar to adenomas in Min-mouse intestine, Cx43 have also been reported to be increased in stromal cells of human breast carcinomas.39 Normally, Cx43 is only located in the muscular layer of the Min mouse intestine,25 as reported for several other mammals.40 Cx43 is also expressed in a few endothelial cells of the rat aorta.41 We found no Cx43 in the endothelium in either normal intestinal tissue or in adenomas of Min-mouse. However, a clear induction of Cx43 was found in myofibroblasts of adenomas, the same cells that are found to express COX-2.28
COX-2 expression has a large impact on adenoma growth in Min-mice, where treatment with a COX-2-specific inhibitor markedly reduced both the numbers and growth of adenomas.42 The cellular localisation of COX-2 has been a much debated question. COX-2 has in several studies been found in the stroma of human and mouse adenomas, with a similar appearance as that of Cx43 in the present study in Min-mice.43, 44 Both macrophages and fibroblasts have been reported to express COX-2 in intestinal adenomas, with partial discrepancy between reported localisation in humans and mice.43, 44, 45, 46 COX-2 was located to the myofibroblasts in adenomas of Min mice, a cell type also suggested to have increased expression of Cx43 in humans.28, 39 The expression of COX-2 and Cx43 could be connected in at least 2 ways. Several publications reported an increase in gap junction protein expression as a function of prostaglandin E2 (PGE2) exposure.29, 47, 48, 49 Because COX-2 performs the rate-limiting step in the production of prostaglandins, overexpression of COX-2 could therefore result in an increase in Cx43. However, COX-2 inhibition by celecoxib did not reduce Cx43 in adenomas, indicating that increased production of PGE2 is not the cause for aberrant Cx43 expression.
Both Cx43 and COX-2 are targets for β-catenin/Tcf-mediated transcription. We have previously shown an increased expression of β-catenin in the epithelial cells of the adenomas in Min-mouse.20 Although β-catenin and Cx43 were increased in different cell types, epithelium vs. myofibroblasts, they were always located in close proximity in the adenomas. Myofibroblasts are stromal cells that are abundantly present at the invasion front of primary colon tumours.50 Cancer cell-derived transforming growth factor-β1 (TGF-β1) can transdifferentiate fibroblasts into myofibroblasts.51 High levels of TGF-β are associated with colon cancer progression in human.52 Myofibroblasts are suggested to secrete signalling molecules that could stimulate invasion of cancer cells, since primary cultures of subepithelial myofibroblasts from human colon promote the migration of epithelial cells.53 How the increased β-catenin expression in the adenoma epithelium can influence the expression of proteins in the stroma or vice versa is not clear. Recently, β-catenin was found to regulate vascular endothelial growth in Min-mouse adenomas through increased expression of VEGF.54 VEGF regulates Cx43 expression in cardiac myocytes.31 However, we found no increase in VEGF in Min-mouse adenomas compared to normal tissue by Western blotting. The other way around, β-catenin expression in implanted mouse colon tumour cells was reduced after treatment with the COX-2-specific inhibitor rofecoxib.55 In addition, paracrine COX-2-mediated signalling by macrophages promotes tumorigenic progression of intestinal epithelial cells.56 At present, the functional significance of Cx43 expression in myofibroblasts is not known. We hypothesise that gap junctions may be of importance in the transfer of signals during tumour growth.
Altogether, Cx43 and COX-2 induction in myofibroblasts in close proximity to dysplastic β-catenin expressing epithelium supports the suggestion that there is mutual signalling between epithelial cells and stromal cells, which are important in intestinal tumour development.
We thank Ms. H.H. Haugen for excellent technical assistance. We also thank Dr. E. Rivedal for making the anti-Cx43 antiserum available to us. V.C. was partly supported by the Research Council of Norway.