Stromal Production of Prostacyclin Confers an Antiapoptotic Effect to Colonic Epithelial Cells

Introduction

Genetic and epidemiological studies have implicated COX-2 4 in the pathogenesis of CRC. We initially reported overexpression of COX-2 in 90% of CRCs and nearly 50% of colonic adenomas (1) . We subsequently identified a human CRC cell line, HCA-7, colony 29, that expresses COX-2 (2) . These cells form a uniform polarizing monolayer when cultured on Transwell filters, and the epidermal growth factor receptor is found predominantly at the basolateral surface, as it is in all polarized epithelial cells. Basolateral but not apical delivery of the epidermal growth factor receptor ligand transforming growth factor α results in up-regulation of COX-2 and production of PGs that are released exclusively into the basolateral medium of polarized HCA-7 cells. These cells produce PGE₂, PGD₂, thromboxane, and PGF_2α, but not PGI₂.

In contrast, most human CRC cell lines do not express COX-2, and it now appears that, at least early in the malignant process, COX-2 is expressed chiefly in the stroma (3, 4, 5) . Indeed, mounting evidence has highlighted the importance of the stromal compartment in epithelial tumors in general and intestinal tumors in particular (6, 7, 8, 9) . The present studies were performed to characterize PG production by human pericryptal fibroblast cell lines and to test the hypothesis that stromal PGs affect epithelial function and contribute to colon carcinogenesis. An in vitro model system was used to simulate in vivo growth conditions whereby pericryptal fibroblast lines and HCA-7 polarized epithelial monolayers were cocultured on the opposing surfaces of Transwell filters.

Materials and Methods

Reagents.

All culture reagents were purchased from Hyclone (Logan, UT), and all chemicals were purchased from Sigma (St. Louis, MO), unless otherwise indicated. Rofecoxib was a generous gift from Pharmacia Upjohn (Peapack, NJ). Carbaprostacyclin, docosahexaenoic acid, and arachidonic acid were purchased from Cayman Chemical (Ann Arbor, MI).

Cell Culture.

HCA-7 cells, passage 20–35, were cultured on 12-mm Transwell filters (pore size, 0.4 μm) under previously defined conditions (2) . Transepithelial electrical resistance across the Transwell filter was measured using a Millicell Electrical Resistance System (Millipore, Bedford, MA) to evaluate functional integrity of tight junctions. Experiments were conducted 7–10 days after seeding, when resistance was >400 ohms·cm².

The isolation and propagation of LIM-PF1 and LIM-PF2 normal pericryptal fibroblasts have been described by Rockman et al. (10) . In an identical manner, LIM-HPF3 pericryptal fibroblasts, as well as lines HPCF1 to HPCF6, were established from the normal-appearing colonic mucosa of seven different individuals with HNPCC by the laboratories of Whitehead and Boman, respectively. All of these fibroblast lines express α smooth muscle actin, a marker of pericryptal fibroblasts, as determined by Western blot with a mouse monoclonal antibody (Sigma-A2547). Fibroblasts were plated and grown to confluence on the bottom surface of Transwell filters, after which HCA-7 cells were plated on top of Transwell filters and maintained until they formed a confluent polarizing monolayer.

PG Analysis.

Eicosanoids were quantified by gas chromatographic/negative ion chemical ionization mass spectrometric assays using stable isotope dilution techniques as described previously (11) .

Apoptosis Assays.

Apoptosis was evaluated by TUNEL and Cytodeath M30 antibody assays (Roche Diagnostic Corporation, Indianapolis, IN). Total cell nuclei were stained with 4′,6-diamidino-2-phenylindole (Sigma). Cells were counted using a Zeiss Axiophot microscope, a SPOT camera (Diagnostic Instruments Inc., Sterling Heights, MI), and Metamorph software version 4.6r8 (Universal Imaging Co., Downingtown, PA). Data were evaluated by a mixed effects ANOVA method (12) and considered significant if differences between groups were P < 0.05.

Results

We examined eicosanoid production from fibroblast cell lines established from grossly normal colonic mucosa at the distal margin of two resected nonhereditary colonic tumors (LIM-PF1 and LIM-PF2, referred to as normal fibroblasts) or from the colonic mucosa of seven individuals with HNPCC (LIM-HPF3 and HPCF1 to HPCF6, referred to as HNPCC fibroblasts). In contrast to HCA-7 cells, the two normal fibroblast lines produced readily detectable amounts of PGI₂, as measured by its stable metabolite 6-keto-PGF_1α (Fig. 1) ⇓ . HNPCC fibroblasts produced, on average, 50-fold more PGI₂ than normal fibroblasts. All of these fibroblast lines generate significant amounts of PGE₂, PGD₂, thromboxane, and PGF_2α; however, the concentrations of these eicosanoids do not differ between normal fibroblasts and HNPCC fibroblasts (data not shown).

Given the observation that HCA-7 cells do not produce PGI₂ but that fibroblast cell lines do, we speculated that this eicosanoid may affect epithelial cell function. PGI₂ has been shown to stimulate growth and inhibit apoptosis in other epithelial cell systems (13 , 14) . Because PGI₂ is highly unstable, we used the stable synthetic PGI₂ analogue carbaprostacyclin to study its effects on HCA-7 cells. Carbaprostacyclin added to HCA-7 cells at concentrations between 10⁻⁹ and 10⁻⁴ m had no effect on proliferation of HCA-7 cells or two gastric cancer cell lines (AGS and MKN-28) as measured by bromodeoxyuridine incorporation, tritiated thymidine incorporation, or colony formation in soft agar (data not shown).

Because it has been shown previously that PGI₂ inhibits programmed cell death in hepatic epithelial cells (13) , we considered whether this eicosanoid may have an antiapoptotic effect on HCA-7 cells. Butyrate, a bacterially produced short chain fatty acid found naturally at millimolar concentrations in the gut, has been reported to induce apoptosis in two CRC cell lines, HT-29 and HCT-116 (15 , 16) . HCA-7 cells treated with 5 mm sodium butyrate for 18 h demonstrated a more than 10-fold increase in the number of TUNEL-positive cells (control, <0.5%; butyrate treatment, >5% total nuclei). We found that polarized HCA-7 cells pretreated with carbaprostacyclin showed a significant concentration-dependent decrease in butyrate-induced apoptosis as measured by TUNEL assay (Fig. 2) ⇓ or by Cytodeath M30 assay (data not shown). A similar reduction in butyrate-induced apoptosis was seen with another polarizing colon cancer cell line, HCT-8, treated with carbaprostacyclin (data not shown).

Because HNPCC fibroblasts produce significant quantities of PGI₂, we hypothesized that coculturing HCA-7 cells with HNPCC fibroblasts would protect HCA-7 cells from butyrate-induced apoptosis. Cells were cultured in such a way as to simulate stromal epithelial interactions. Fibroblast lines were grown on the bottom of Transwell filters, and polarized HCA-7 epithelial cells were grown on top. Levels of apoptosis for HCA-7 cells grown in the presence of normal fibroblasts (LIM-PF1 and LIM-PF2), which produce less PGI₂, were not statistically different from HCA-7 cells grown alone. However, HCA-7 cells grown in the presence of HNPCC fibroblasts (LIM-HPF3, HPCF1, and HPCF4) were significantly more resistant to butyrate-induced apoptosis than HCA-7 cells grown alone (Fig. 3) ⇓ .

Additional studies were undertaken to evaluate whether PGI₂ is a critical mediator of HCA-7 cell resistance to butyrate-induced apoptosis. Addition of the selective COX-2 inhibitor rofecoxib reduced PGI₂ production in HNPCC fibroblasts by >90% (Fig. 4A) ⇓ . Administration of rofecoxib to cocultures of HNPCC fibroblasts (LIM-PF3 or HPCF2) and HCA-7 cells abrogated the antiapoptotic effect conferred by HNPCC fibroblasts on HCA-7 cells (Fig. 4B) ⇓ . Addition of carbaprostacyclin to these cultures, in turn, countered the effect of the COX-2 inhibitor. Furthermore, conditioned medium from HNPCC fibroblasts did not protect HCA-7 cells against butyrate-induced apoptosis (data not shown), a finding that argues against a major protective role for the paracrine action of more stable peptide growth factors such as hepatocyte growth factor and keratinocyte growth factor, which have also been shown to protect against apoptotic stimuli (17 , 18) .

PGI₂ is thought to induce its biological activity by binding to and activating one of two receptors, the G protein-coupled cell surface PGI₂ (IP) receptor and nuclear PPAR δ (19) . HCA-7 cells have been shown to express PPAR δ (20) , and we were not able to detect expression of IP receptor in HCA-7 cells by Northern blot analysis (data not shown). Furthermore, we found that treating HCA-7 cells with an alternative ligand for PPAR δ, docosahexaenoic acid (21) , had a virtually identical effect to that observed with carbaprostacyclin (shown in Fig. 2 ⇓ ), reducing the number of apoptotic nuclei in a concentration-dependent manner (mean control, 6.1% versus 3.7% and 3.4% at 10⁻⁶ and 10⁻⁵ m, respectively; P < 0.05). Taken together, the similar apoptosis-attenuating effects of PGI₂ and docosahexaenoic acid and lack of expression of IP receptors strongly suggest that the PGI₂ effect in HCA-7 cells is mediated by PPAR δ.

Discussion

We find that fibroblasts derived from seven individuals with HNPCC produce an abundance of PGI₂. HNPCC patients suffer from a genetic predisposition to CRC due to germ-line mutations in DNA mismatch repair genes resulting in genomic instability in the colonic epithelium (22) . Could enhanced PGI₂ expression have a role in the susceptibility of HNPCC kindreds to CRC? Although there appears to be a shortened interval from adenoma to carcinoma in HNPCC individuals (1–3 years compared with 8–10 years for individuals with sporadic CRC), HNPCC individuals with CRC fare better clinically. COX-2 immunoreactivity has been reported to be reduced in the tumor epithelium of HNPCC individuals with CRC compared with individuals with sporadic CRC (23 , 24) . Based on our findings, it is tempting to speculate that early colonic lesions in HNPCC individuals are maintained and progress rapidly due to the nurturing influence of PGI₂ produced by the adjacent stroma. As these tumors progress and lose contact with their supporting stroma, they may be more susceptible to apoptotic stimuli in the absence of local PGI₂ production.

We propose a model in which overproduction of PGI₂ contributes to progression of neoplastic epithelium by preventing apoptosis in the face of accumulating genetic lesions and a noxious colonic environment. Studies will now be directed toward elucidating the mechanism(s) underlying the increased PGI₂ production in HNPCC-derived fibroblasts and how this antiapoptotic effect is executed. Preliminary studies reveal no evidence for microsatellite instability in these HNPCC-derived fibroblasts. 5 Of interest, Sonoshita et al. (25) recently observed that COX-2 is expressed in the fibroblasts and endothelial cells of intestinal polyps of individuals with familial adenomatous polyposis.

This study provides one mechanism by which the stroma, via its production of PGI₂, can regulate epithelial function, that is, by promoting epithelial survival. PGI₂ would be an ideal candidate molecule for such localized signaling, given its short half-life under physiological conditions (26) . The normal colonic epithelium is sheathed in a single layer of pericryptal fibroblasts. Kaye et al. (27) have described these pericryptal fibroblasts and colonocytes as forming a highly ordered unit. Whereas the epithelial cells proliferate at the crypt base and acquire more differentiated features as they migrate toward the luminal surface, the single layer of fibroblasts migrates in from the lamina propia and differentiates (5 , 28) . These fibroblasts then undergo very little proliferation or migration, possibly contributing local cues that promote epithelial survival, growth, and differentiation (29 , 30) . Future studies will address whether there are different patterns of spatial or temporal production of PGI₂ by pericryptal fibroblasts along the colonic crypt in normal and diseased states.