Molecular Medicine Unit, University of Leeds, St James's University Hospital, Leeds LS9 7TF, UK

Correspondence to: M A Hull, E-mail: M.A.Hull@leeds.ac.uk

In human colorectal adenomas or polyps, cyclooxygenase-2 is expressed predominantly by stromal (or interstitial) macrophages. Therefore, we tested the hypothesis that macrophage cyclooxygenase-2 has paracrine pro-tumorigenic activity using in vitro models of macrophage-epithelial cell interactions. We report that macrophages can promote tumorigenic progression of intestinal epithelial cells (evidenced by decreased cell-cell contact inhibition, increased proliferation and apoptosis, gain of anchorage-independent growth capability, decreased membranous E-cadherin expression, up-regulation of cyclooxygenase-2 expression, down-regulation of transforming growth factor- type II receptor expression and resistance to the anti-proliferative activity of transforming growth factor-₁) in a paracrine, cyclooxygenase-2-dependent manner. Pharmacologically relevant concentrations (1-2 M) of a selective cyclooxygenase-2 inhibitor had no detectable, direct effect on intestinal epithelial cells but inhibited the macrophage-epithelial cell signal mediating tumorigenic progression. Cyclooxygenase-2-mediated stromal-epithelial cell signalling during the early stages of intestinal tumorigenesis provides a novel target for chemoprevention of colorectal cancer (and other gastro-intestinal epithelial malignancies, which arise on a background of chronic inflammation, such as gastric cancer) and may explain the discrepancy between the concentrations of cyclooxygenase inhibitors required to produce anti-neoplastic effects in vitro and in vivo.

Oncogene (2002) 21, 7175-7186. doi:10.1038/sj.onc.1205869

adenomatous polyps; colorectal cancer; cyclooxygenase; macrophage

Introduction

Cyclooxygenase-2 (COX-2) is expressed in a stage-specific pattern during human colorectal carcinogenesis. At later stages of tumorigenesis, COX-2 is expressed in 80-100% of human primary and secondary colorectal cancers (Eberhart et al., 1994; Sheehan et al., 1999; Hull et al., 2000a), predominantly by malignant epithelial cells (Kutchera et al., 1996; Sheehan et al., 1999; Sano et al., 1999; Hull et al., 2000a). Over-expression of COX-2 in intestinal epithelial cells in vitro has been demonstrated to promote malignant behaviour, such as decreased cell adhesion and increased cell motility (Tsujii and Dubois, 1995; Tsujii et al., 1997). In keeping with these observations, increased epithelial cell COX-2 expression in primary colorectal cancers has been associated with reduced patient survival (Sheehan et al., 1999). However, at an earlier stage of carcinogenesis, in benign human colorectal adenomas or polyps (which represent a more relevant target for chemoprevention), COX-2 is expressed predominantly by stromal cells (as well as by dysplastic epithelial cells), the majority of which have been identified as CD68-positive inflammatory mononuclear cells or macrophages (Bamba et al., 1999; Muller-Decker et al., 1999; Chapple et al., 2000; Hull et al., 2000b; Soslow et al., 2000; Hardwick et al., 2001; Khan et al., 2001; Takeuchi et al., 2002). A similar pattern of COX-2 expression has also been noted in tumours from several murine intestinal tumorigenesis models (Oshima et al., 1996; Taketo, 1998; Hull et al., 1999; Shattuck-Brandt et al., 1999, 2000), although there appears to be increased heterogeneity in the COX-2-expressing stromal cell population in mice compared with humans (Taketo, 1998; Hull et al., 1999; Shattuck-Brandt et al., 1999, 2000). Stromal mononuclear cell (as well as epithelial cell) COX-2 expression has also been observed in other pre-malignant conditions of the human gastro-intestinal tract, including Barrett's oesophagus (Fu et al., 1999; Morris et al., 2001) and Helicobacter pylori-associated chronic gastritis (Wilson et al., 1998; Sung et al., 2000).

There is direct genetic evidence of an important role for COX-2 during adenoma development and progression, from murine models of familial adenomatous polyposis (FAP; Oshima et al., 1996; Chulada et al., 2000), as well as pharmacological evidence from these models (Oshima et al., 1996, 2000; Jacoby et al., 2000) and patients with FAP (Steinbach et al., 2000). Therefore, the stromal cell localization of COX-2 in intestinal adenomas implies a paracrine mechanism of action of COX-2 during the early stages of intestinal tumorigenesis in humans and mice. Observations from the Apc⁷¹⁶ mouse model of FAP suggest that stromal cell COX-2 promotes angiogenesis in small intestinal adenomas (Sonoshita et al., 2001; Seno et al., 2002). However, stromal cell COX-2 expression was also associated with increased epithelial cell proliferation rates in adenomas, in this model, implying the existence of direct signalling between stromal cells and epithelial cells as well as pro-angiogenic effects (Sonoshita et al., 2001).

Hence, we tested the hypothesis that paracrine COX-2-mediated macrophage-epithelial cell signalling promotes direct tumorigenic progression of intestinal epithelial cells, using in vitro models of macrophage-epithelial cell interactions.

Results

Models of paracrine, COX-2-mediated macrophage-epithelial cell signalling

As the predominant COX-2-expressing stromal cell in human colorectal adenomas is the CD68-positive macrophage (Bamba et al., 1999; Chapple et al., 2000; Hardwick et al., 2001), we used murine RAW264.7 macrophages (Raschke et al., 1978) as convenient model COX-2-expressing macrophages. Under basal culture conditions, non-activated RAW264.7 macrophages (NM) did not express COX-2 protein (Figure 1a). However, following incubation with IFN- and LPS, 'activated' RAW264.7 macrophages (AM) expressed COX-2 protein (Figure 1a) and synthesized PGE₂ (Figure 1b). In keeping with previously published studies (Hwang et al., 1997), we were unable to detect COX-1 protein in RAW264.7 cell lysates, although COX-1 mRNA expression was demonstrated by RT-PCR analysis of RAW264.7 cell total RNA (data not shown). PGE₂ production by AM was inhibited by the selective COX-2 inhibitor SC236 in a concentration-dependent fashion at concentrations of 2 nM and above (Figure 1b). We then used 2 M SC236 during macrophage activation after confirming that this concentration of SC236 had no direct effects on viability of either NM or AM (data not shown) and did not affect COX-2 protein expression by AM (Figure 1a).

We used non-transformed rat intestinal epithelial cells (IEC-6) as model intestinal epithelial cells (Quaroni et al., 1979). SC236 (up to 10 M) had no direct effect on either short- (Figure 1c) or long-term viability (see also Figure 4) of IEC-6 cells. These cells express variable levels of PGE₂ EP receptor isoform mRNAs analysed by RT-PCR (Figure 1d). Transcripts for receptors EP2 and EP4 (and to a lesser extent EP1) were strongly expressed, in contrast to EP3 mRNA that was only detected at low levels compared with rat kidney (Figure 1d).

A macrophage-conditioned medium (MCM) model was designed which allowed investigation of paracrine signalling between macrophages and intestinal epithelial cells over extended periods of time (Figure 2a). In addition, a short-term (14 d) indirect co-culture model was employed to confirm the observations made during MCM culture (Figure 2a). The experimental design of both models (Figure 2b) ensured that the effect of COX-2 inhibition by SC236 (SC) during macrophage activation and conditioning of the medium (SC-AMCM) could be distinguished from a direct effect of the COX-2 inhibitor (added after MCM production) on intestinal epithelial cells themselves (Con+SC and AMCM+SC).

Activated macrophages promote tumorigenic progression of non-transformed intestinal epithelial cells

Long-term culture of non-transformed intestinal epithelial (IEC-6) cells in activated macrophage-conditioned medium (AMCM) induced 'fibroblastoid' morphological change with cell-cell overlapping and decreased contact inhibition (Figure 3c, h, s) compared with control (Con) and non-activated MCM (NMCM)-cultured cells which always maintained a contact-inhibited 'cobblestone' monolayer morphology (Figure 3a, b, f, g, r). Cells cultured in ('last wash') medium, which was used to remove residual IFN- and LPS prior to conditioning of medium by macrophages, did not demonstrate any differences from control cells. Changes in AMCM-cultured epithelial cell morphology (decreased cell-cell contact and cell elongation) were apparent from day 4 onwards and progressed over time, so that by passage 6 (day 70 onwards), a homogeneous population of 'spindle-like' cells had developed with a propensity to focus formation on tissue-culture plastic (Figure 3h). Reproducible morphological changes occurred in MCM culture, over a similar time-course, in six replicate experiments. Similar morphological changes to those noted in the MCM experiments were also noted in IEC-6 cells co-cultured with activated macrophages (Figure 3k-m), although these changes were accelerated compared with the MCM experiments. For example, cell overlapping was prominent by day 14 of co-culture (compare Figure 3c and m). AMCM-cultured cells maintained cytokeratin protein expression in a similar pattern to control cells (Figure 3p-q), confirming their epithelial origin, but had significantly less membranous E-cadherin protein expression than control cells (Figure 3r-s). This is in keeping with the decrease in contact inhibition and increase in cell overlapping that we observed in AMCM cultures.

By passage 6, AMCM-cultured IEC-6 cells had developed the ability to form colonies in soft agar (mean±s.e.m. colony forming efficiency (CFE) 0.024±0.00064% (n=3)) and interlinking cell networks on Matrigel^Ô (Figure 3t). In contrast, control and NMCM-cultured cells consistently underwent apoptosis within 48 h, in both these environments. However, despite the ability of AMCM-cultured cells to grow in an anchorage-independent manner in soft agar, AMCM-cultured cells were unable to form subcutaneous or caecal serosal tumours in nude mice. AMCM-cultured intestinal epithelial cells were also unable to invade thin Matrigel^Ô layers unlike human colorectal cancer cells (Sheng et al., 2001).

We then compared gene expression patterns between control and AMCM-cultured intestinal epithelial cells in order to determine whether the overall profile of intestinal epithelial cell gene expression was consistent with tumorigenic progression (Figure 4a). Compared with control IEC-6 cells, transcript levels for plasminogen activator inhibitor-1 (PAI-1), macrophage migration inhibition factor (MIF) and heat shock protein 27 were significantly increased in AMCM-cultured cells (passage 6), whereas those for TGF- receptor type II (TGF- RII), ezrin, endothelin-1 and Wilm's tumour suppressor were decreased. Messenger RNA levels for several MAP kinase cascade genes (Ras, MEK1, ERK2) were also increased. In keeping with the increased proliferation rate, transcript levels for proliferating cell nuclear antigen, cyclin D1 and cyclin D3 were also elevated in AMCM-cultured cells. Confirmatory RT-PCR data are shown for PAI-1, MIF and TGF- RII (Figure 4b). Consistent with down-regulation of expression of TGF- RII mRNA (Figure 4b, c), AMCM-cultured cells expressed lower levels of TGF- RII protein (Figure 4d) and were resistant to the growth inhibitory effects of rhTGF-₁ (Figure 4e) compared with control cells. As COX-2 is an inducible gene which is expressed by epithelial cells in some human sporadic colorectal adenomas (Chapple et al., 2000), we also investigated COX-2 expression by MCM-cultured intestinal epithelial cells. COX-2 mRNA and protein levels were increased in AMCM-cultured intestinal epithelial cells compared with control cells (Figures 4b and 5). In keeping with increased COX-2 expression, PGE₂ production by AMCM-cultured epithelial cells was significantly increased (359.0±37.3 pg/10⁶ cells (mean±s.e.m.); n=3) compared with control and NMCM-cultured cells which produced no detectable PGE₂ over a 6 h period. PGE₂ production by AMCM-cultured cells, over the 6 h culture period, was completely inhibited in the presence of 1 M SC236.

Development of different aspects of the tumorigenic phenotype occurred at varying rates. Although changes in IEC-6 cell morphology and E-cadherin expression were evident within a few days of initiation of AMCM culture, other phenotype changes occurred over a different time course. For example, TGF- RII mRNA down-regulation and MIF mRNA up-regulation were only evident when tested on day 84 (Figure 4b) with no difference in mRNA levels between control and AMCM-cultured cells noted at day 56 (data not shown). Alterations in intestinal epithelial cell kinetics also occurred in a time-dependent manner with a progressive rise in proliferation and apoptosis rates in AMCM-cultured cells being apparent from passage 3 (week 6) onwards (data not shown). Activated MCM had no significant effect on intestinal epithelial cell proliferation or apoptosis rates during the first 14 days (passage 1) of MCM culture (Figure 6a, b). However, during prolonged MCM culture, a time-dependent increase in proliferation and apoptosis rates of AMCM-cultured cells was noted compared with control cells, so that by day 82 of MCM culture (passage 6), AMCM-cultured intestinal epithelial cell proliferation was approximately threefold higher than control cells (Figure 6c, d). These cells also displayed a significantly higher apoptosis rate than control cells (Figure 6d). Cells (passage 6) cultured in NMCM also displayed a significantly increased proliferative capacity compared with control cells (Figure 6d). However, NMCM-cultured cells (passage 6) underwent apoptosis at a rate comparable to control cells (Figure 6d).

The intestinal epithelial cell phenotype induced by AMCM is reversible

Cells that had been cultured in AMCM for 84 days were placed back into control medium (termed AC cells). AC cells were unable to survive in an anchorage-independent environment in soft agar unlike age-matched AMCM-cultured cells (passage 8) which retained the ability to grow and form colonies in soft agar. AC cells did not maintain COX-2 protein expression when AMCM was removed from the culture (Figure 5a). By contrast, repression of TGF- RII protein expression, at levels associated with AMCM-culture, was maintained in AC cells, despite the absence of AMCM for a period of 28 days (Figure 4d). Although AC cell monolayers became contact-inhibited without cell overlapping, they did not regain a characteristic, cuboidal appearance but remained elongated, with an appearance similar to rhTGF-₁-treated intestinal epithelial cells (Sheng et al., 1999).

COX-2 inhibition during macrophage activation abrogates tumorigenic progression of intestinal epithelial cells

COX-2 inhibition by SC236 during AMCM production (SC-AMCM) abrogated induction of a tumorigenic phenotype in intestinal epithelial cells. Importantly, an identical concentration of SC236 added to AMCM after production (AMCM+SC) did not reverse the 'tumorigenic' effect, ruling out direct activity of the selective COX-2 inhibitor, remaining in the MCM, on IEC-6 cells.

Intestinal epithelial cells cultured in AMCM, which was produced in the presence of SC236 (SC-AMCM), maintained a contact-inhibited cell monolayer with only occasional cell overlapping, although the cells did appear slightly 'fibroblastoid' compared with control cells (Figure 3d, i). By contrast, cells cultured in AMCM+SC appeared identical to AMCM-cultured cells except that apoptotic cells were slightly more prominent at later passages (Figure 3e, j). A similar pattern of intestinal epithelial cell morphology was noted in the co-culture model (Figure 3k-o).

SC-AMCM-cultured cells were unable to grow in soft agar (CFE 0%). However, direct addition of SC236 to AMCM-cultured cells (AMCM+SC) did not interfere with anchorage-independent growth capability in soft agar (mean±s.e.m. CFE 0.027±0.00088% (n=3)). SC-AMCM-cultured cells were also unable to survive on Matrigel^Ô although cells cultured in AMCM+SC formed cellular networks in an identical manner to AMCM-cultured cells (data not shown).

We also observed that changes in TGF- RII and COX-2 expression induced by AMCM, occurred in a macrophage COX-2-dependent manner (Figures 4c-e and 5a. Down-regulation of TGF- RII expression at both mRNA and protein levels by AMCM was abrogated by SC236 when it was present during macrophage activation (SC-AMCM), but not when SC236 was added directly to AMCM-cultured intestinal epithelial cells (AMCM+SC; Figure 4c-d). Importantly, sensitivity to the growth inhibitory effects of rhTGF-₁ was maintained in SC-AMCM-cultured cells in parallel with receptor expression levels (Figure 4e). The selective COX-2 inhibitor had similar effects on AMCM-induced COX-2 up-regulation by intestinal epithelial cells. SC-AMCM-cultured epithelial cells did not express COX-2 protein or synthesize PGE₂ (Figure 5a). By contrast, SC236 added directly to AMCM-cultured cells did not reduce COX-2 protein expression (Figure 5a) although SC236 did inhibit COX-2 activity, as PGE₂ synthesis by AMCM+SC-cultured cells was not detectable.

SC236 (1 M) had no direct effect on control or AMCM-cultured intestinal epithelial cell proliferation or apoptosis (Figure 6b, d). However, at later stages of MCM culture (passage 6), proliferation of SC-AMCM-cultured cells was reduced compared with cells cultured in AMCM and was similar to that of NMCM-cultured intestinal epithelial cells (Figure 6d). AMCM-induced apoptosis did not occur in a COX-2-dependent manner, as the presence of SC236 during macrophage activation did not inhibit the pro-apoptotic activity of AMCM on intestinal epithelial cells (Figure 6d).

Paracrine pro-tumorigenic signalling by COX-2-expressing macrophages is not explained solely by PGE₂

Prostaglandin E₂ is the most abundant prostanoid in colorectal neoplasms (Yang et al., 1998) and promotes proliferation and malignant potential of human colorectal cancer cells in vitro (Sheng et al., 2001). During long-term culture in AMCM, intestinal epithelial cells were exposed to PGE₂ concentrations of approximately 30 nM (10 ng/ml). Therefore we tested whether addition of exogenous PGE₂ mimicked culture in AMCM. The presence of PGE₂ (10-500 nM) for up to 84 days (passage 6) did not induce changes in the morphology, cell proliferation or cumulative apoptosis rate of intestinal epithelial cells compared with control cells (ethanol carrier only; data not shown). In addition, exposure to PGE₂ (up to concentrations of 500 nM for 84 days) did not alter TGF- RII or COX-2 protein expression by IEC-6 cells, analysed by Western blotting (data not shown).

Discussion

We have demonstrated that activated macrophages can induce an intestinal epithelial cell phenotype compatible with neoplastic progression in vitro, in a paracrine, COX-2-dependent manner (Table 1).

Evidence from these models supports the concept that COX-2-mediated signalling from macrophages is relevant to the early stages of human intestinal tumorigenesis, in particular, adenoma development and progression. The gene expression pattern of AMCM-cultured cells was compatible with colorectal adenoma development in man, including down-regulation of E-cadherin (Gagliardi et al., 1995) and TGF- RII (Manning et al., 1991; Zhang et al., 1997). The COX-2-induced intestinal epithelial cell phenotype stopped short of full in vitro transformation (measured by the ability to form tumours in nude mice) but AMCM-cultured cells developed low efficiency, anchorage-independent growth capability in soft agar. These findings are in keeping with the phenotype of the few human colorectal adenoma cell lines that have been described (Willson et al., 1987; Paraskeva et al., 1990). Increased cell proliferation is recognised to accompany malignant progression of intestinal epithelial cells during human colorectal carcinogenesis. In our MCM model, increased proliferation occurred by both COX-2-dependent and -independent mechanisms. COX-2 inhibition during macrophage activation reduced proliferation down to levels associated with NMCM culture, but these values were significantly higher than proliferation rates associated with control cells. It is likely that the COX-2-independent increase in proliferative capacity demonstrated by NMCM-cultured intestinal epithelial cells is related to 'conditioning' of medium with mitogenic factors by non-activated RAW264.7 macrophages (Rosenberger et al., 2000). This phenomenon has been described in models using a variety of target cells (Goldman and Bar-Shavit, 1979; Hauptmann et al., 1993). In keeping with our observations that macrophage COX-2 activity promotes intestinal epithelial cell proliferation in vitro, Sonoshita et al. (2001) have recently reported that stromal cell COX-2 drives increased epithelial cell proliferation in Apc⁷¹⁶ mouse adenomas. The mechanism by which AMCM induced apoptosis of intestinal epithelial cells remains unclear. However, increased apoptosis in AMCM-cultured cells was not inhibited by SC236, implying that this may be a COX-2-independent process. Interestingly, increased apoptosis has previously been demonstrated in human colorectal adenomas compared with normal mucosa (Sinicrope et al., 1996).

The morphological changes we described in intestinal epithelial cells, along with gene expression changes including down-regulation of membranous E-cadherin, are similar to the in vitro phenomenon of epithelial-mesenchymal trans-differentiation (EMT) which is widely believed to model gain of invasiveness and metastatic ability by carcinoma cells, at later stages of carcinogenesis (Birchmeier et al., 1996). However, AMCM-cultured cells still maintained cytokeratin expression and failed to develop invasive properties during long-term culture. Both these features are incompatible with our model being simply a manifestation of EMT.

Evidence that paracrine COX-2 activity induces an epithelial cell phenotype, compatible with partial transformation in vitro, suggests a paradigm shift for the role of COX-2 during the early stages of human colorectal carcinogenesis. In sporadic colorectal adenomas, COX-2 is localized to the stromal cell compartment, predominantly in macrophages (Bamba et al., 1999; Chapple et al., 2000; Hull et al., 2000b; Hardwick et al., 2001), and mediates tumour promotion in a paracrine fashion. At later stages of human colorectal carcinogenesis, COX-2 is localized predominantly to malignant epithelial cells (Kutchera et al., 1996; Sheehan et al., 1999; Sano et al., 1999; Hull et al., 2000a), in which autocrine and/or paracrine mechanisms of action could exist. COX-2 over-expression in intestinal epithelial cells in vitro has been associated with similar changes (e.g. decreased E-cadherin and TGF- RII expression) to those described in our model (Tsujii and Dubois, 1995), which suggests that even though COX-2 is expressed in a stage-specific manner during human colorectal carcinogenesis, it may have similar downstream effects on epithelial cells regardless of whether stromal or epithelial COX-2 expression predominates. In a similar fashion, COX-2, in either stromal (Sonoshita et al., 2001) or epithelial cells (Tsujii et al., 1998), has also been demonstrated to promote tumour angiogenesis in different experimental models.

The precise point at which the switch from predominant stromal to epithelial cell COX-2 localization occurs during colorectal carcinogenesis, is currently unknown although there is tentative evidence for a link with RAS mutation which is believed to occur at the 'late' adenoma stage of colorectal carcinogenesis (Sheng et al., 1997; Fujita et al., 2000). Our model suggests that an alternative mechanism exists whereby COX-2-expressing macrophages induce neighbouring epithelial cell COX-2 expression, in adenomas. In keeping with this, the limited amounts of COX-2 protein present were localized prominently to epithelial cells adjacent to COX-2-positive macrophages, in our recent study of COX-2 localization in human colorectal adenomas (Chapple et al., 2000). A switch from predominant stromal to epithelial cell COX-2 expression during adenoma progression may explain (along with variations in immunohistochemistry protocols) existing discrepancies in COX-2 localization data from different human adenoma series (Bamba et al., 1999; Muller-Decker et al., 1999; Chapple et al., 2000; Hull et al., 2000b; Soslow et al., 2000; Hardwick et al., 2001; Khan et al., 2001; Takeuchi et al., 2002).

Observations that the selective COX-2 inhibitor celecoxib, as well as other non-steroidal anti-inflammatory drugs such as sulindac, induce regression of existing adenomas, as well as prevent polyposis in FAP patients (Giardiello et al., 1993, 2002; Steinbach et al., 2000) and murine models of FAP (Jacoby et al., 2000), concurs with our finding that the intestinal epithelial cell phenotype induced by macrophages in our model, was reversible. If an initiation-promotion-progression paradigm is applied to human colorectal carcinogenesis (Pitot et al., 2000), reversibility in vitro implies a role for COX-2 during reversible tumour promotion and emphasises the importance of COX-2 as a chemoprevention target for a potentially reversible stage of colorectal carcinogenesis.

An important finding of our experiments was that pharmacologically relevant concentrations of a selective COX-2 inhibitor were capable of inhibiting tumorigenic progression of intestinal epithelial cells in an indirect fashion (via inhibition of macrophage-intestinal epithelial cell signalling) but not via a direct effect on intestinal epithelial cells. Most studies investigating the direct effects of non-steroidal anti-inflammatory drugs (NSAIDs) and selective COX-2 inhibitors on normal or transformed intestinal epithelial cells in vitro have reported that supra-physiological drug concentrations are required for anti-proliferative activity, although these drugs have potent anti-neoplastic activity at lower concentrations in vivo (Williams et al., 2000a; Marx, 2001). An indirect mechanism of action of COX inhibition on intestinal epithelial cells (via inhibition of paracrine stromal-epithelial cell signalling) is a plausible explanation (along with, for example, anti-angiogenic activity) for the discrepancy in drug concentrations required for anti-cancer effects in vitro and in vivo, without the necessity to invoke COX-independent mechanisms of action for all NSAIDs that have been demonstrated to have anti-neoplastic properties (Marx, 2001). Similar observations confirming the importance of stromal cell COX-2 inhibition have been noted in a murine Lewis lung carcinoma model, in which genetic deletion of COX-2 in host stromal cells was associated with reduced tumour growth compared with animals with wild-type COX-2 expression (Williams et al., 2000b).

Paracrine COX-2-mediated signalling during the early stages of tumorigenesis may also contribute to the pathogenesis of cancers in the colon, and elsewhere in the human gastrointestinal tract, where there is a close link between chronic inflammation, COX-2-positive mononuclear cells and carcinogenesis e.g. inflammatory bowel disease, Helicobacter pylori-associated chronic gastritis and Barrett's oesophagus (Cordon-Cardo and Prives, 1999; O'Byrne and Dalgleish, 2001). These pre-malignant conditions exhibit prominent stromal mononuclear cell COX-2 expression (Wilson et al., 1998; Fu et al., 1999) and carcinogenesis in these conditions is prevented to a similar extent by long-term use of NSAIDs (Farrow et al., 1998). Thus, paracrine COX-2-mediated stromal-epithelial cell signalling may be a generic mechanism linking chronic inflammation and the early stages of gastro-intestinal carcinogenesis and requires further investigation in other organ systems.

Prostaglandin E₂ alone did not account for the COX-2-mediated phenotype changes that we observed in non-transformed intestinal epithelial cells, although we cannot rule out that other COX-2-derived eicosanoids contributed to the paracrine signal in our model. Current evidence would suggest that EP2 and EP4 receptors, which are both expressed by IEC-6 cells, are most likely to mediate the effects of paracrine PGE₂ on intestinal epithelial cells, at both early and late stages of colorectal carcinogenesis (Sonoshita et al., 2001; Sheng et al., 2001; Mutoh et al., 2002). Work is currently underway in our laboratory to identify the precise nature of the COX-2-dependent paracrine signal from macrophages. This may define complementary, indirect chemoprevention targets alongside COX-2. Possible paracrine mediators, in addition to COX-2-derived eicosanoids, include mutagenic COX-2-derived products such as reactive oxygen species and/or malondialdehyde (Sharma et al., 2001). Alternatively, other pro-tumorigenic factors produced by macrophages (Rosenberger et al., 2000; Nau et al., 2002), under the control of COX-2, could mediate the paracrine signal. Multiple growth factors and cytokines produced by activated macrophages have been demonstrated to have pro-tumorigenic activity in vitro and in vivo including TGF-₁, tumour necrosis factor , MIF etc (Sheng et al., 1999; Hudson et al., 1999; Coussens and Werb, 2001). However, the contribution of individual macrophage-derived factors to stromal-epithelial cell signalling during the early stages of intestinal tumorigenesis, as well as the effect of COX-2 inhibition on their expression by macrophages, remain to be determined.

In conclusion, we have provided in vitro evidence that macrophage COX-2 promotes direct tumorigenic progression of intestinal epithelial cells, in a paracrine manner. The localization of COX-2 to macrophages in the majority of human colorectal adenomas as well as other pre-malignant conditions of the GI tract implies that COX-2-mediated, paracrine stromal macrophage-epithelial cell signalling is a valid novel target for GI cancer chemoprevention strategies.

Materials and methods

Cell culture

Mouse RAW264.7 macrophages (ECACC, Porton Down, UK; Raschke et al., 1978) and rat intestinal epithelial (IEC-6) cells (ATCC, VA; Quaroni et al., 1979) were cultured in control medium (RPMI 1640 with GlutaMax-I^Ô, 100 U/ml penicillin, 100 g/ml streptomycin (Invitrogen, Paisley, UK) and 10% FCS (Harlan Sera Lab., Loughborough, UK) at 37°C, in the presence of 5% CO₂. All experiments were initiated on the 4th passage of the IEC-6 cell stock. In some experiments, IEC-6 cells were cultured in the presence of PGE₂ (Sigma, Poole, UK) or an equivalent dilution of ethanol carrier. Cell cultures were consistently negative for Mycoplasma infection tested by PCR (Stratagene, La Jolla, CA, USA).

Production of macrophage-conditioned medium (MCM)

RAW264.7 cells (6´10⁶ per 150 cm²) were cultured for 24 h prior to priming with 100 U/ml IFN- (Sigma) for 4 h, followed by activation with 1 g/ml LPS (E. coli 026:B6; Sigma) for 4 h. Cells were washed three times (the final wash in control medium being reserved as a 'last wash' control) before addition of fresh medium (1 ml per 2´10⁵ cells) that was then conditioned for 24 h and filtered (0.2 m). In all experiments, macrophage activation was carried out in the absence or presence (2 M) of the selective COX-2 inhibitor SC236 (Pharmacia Corp., St Louis, MO, USA). In order to counteract the growth inhibitory effect of IFN- and LPS treatment on RAW264.7 cells, non-activated MCM was produced from RAW264.7 cells seeded at 2´10⁶ per 150 cm², to ensure that the medium was eventually conditioned by similar numbers (2´10⁵ cells/ml) of activated and non-activated RAW264.7 cells.

MCM-culture of IEC-6 cells

IEC-6 cells with an initial seeding density of 5´10⁴ per 75 cm² were cultured on tissue culture plastic in a 1 : 2 dilution of freshly prepared MCM in control medium. Cultures were carried out for 14 days each passage. At the end of each passage, cells were trypsinized and re-seeded at the same initial cell density. Medium was replaced with fresh MCM or control medium every 48-72 h.

Indirect co-culture of RAW264.7 and IEC-6 cells

RAW264.7 cells (4´10⁵) were seeded into individual wells of co-culture companion 6-well plates (BD Bioscience, Oxford, UK) prior to activation as described above. The seeding density for non-activated RAW 264.7 cells was 1.5´10⁵ per well. After washing, fresh control medium (2 ml) was placed on RAW264.7 cell cultures prior to insertion of a 0.4 m pore co-culture insert containing 5´10³ IEC-6 cells (in 2 ml of control medium). Inserts were placed in wells containing freshly prepared activated macrophages, non-activated macrophages, or control medium every 48-72 h. At the end of the culture period, insert membranes were fixed in 4% paraformaldehyde prior to staining with haematoxylin and eosin (H&E).

Cell counting, proliferation and apoptosis assays

Adherent and floating IEC-6 cell populations were counted separately by light microscopy in 0.4% trypan blue (Sigma) in PBS and by fluorescence microscopy of 4% paraformaldehyde-fixed, Hoechst 33358 (ICN, Costa Mesa, CA, USA)-stained (0.5 g/ml) cells, respectively. Counting of IEC-6 cells in 24-well plates was also determined by 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulphophenyl)-2H-tetrazolium (MTS) assay (CellTiter 96^Ò; Promega, Madison, WI) as per manufacturer's instructions. At designated time points (days 0, 4, 8 and 12), floating cells were collected and spun at 5000 r.p.m. for 10 min, prior to fixation in 4% paraformaldehyde. Morphologically apoptotic cells were identified as described (Smith et al., 2001). Cells with characteristic changes of apoptosis by fluorescence microscopy invariably made up >99% of all floating cell populations. Remaining adherent cells were washed twice with PBS before the addition of MTS reagent.

Anchorage-independent growth and invasion assays

Anchorage-independent growth assays of IEC-6 cells in soft agar (BD Bioscience) were performed as described (Sheng et al., 1999) in triplicate except that top and bottom agar layers incorporated a 1 : 2 dilution of each relevant MCM. Colonies greater than 200 m at day 14 were counted. IEC-6 cells were also cultured on Matrigel^Ô (BD Bioscience) diluted 1:2 with RPMI 1640 medium. Matrigel^Ô invasion assays were carried out as per manufacturer's instructions.

Tumorigenicity assay

2-10´10⁶ activated MCM-cultured or control cells were injected either subcutaneously or into the serosal surface of the caecum of Balb/c nu/nu mice (n=3). Animals were sacrificed between 4-8 weeks and tumour growth was analysed histologically.

Gene expression analysis

The Atlas^Ô rat cDNA expression array (BD Bioscience) was used for the comparison of the gene expression pattern of control IEC-6 cells and activated MCM-cultured IEC-6 cells at passage 6. Total RNA was purified by TRIZol^Ô reagent (Invitrogen) and pre-digested with DNase-I (Ambion, Cambridge, UK). RNA samples were checked for the absence of genomic DNA (<0.001%) by PCR. cDNA probes were synthesized from 15 g of DNase I-treated total RNA using [³²P]-dATP (Amersham Pharmacia Biotech, Little Chalfont, UK) according to the manufacturer's protocol. Membranes were hybridized overnight at 68°C and washed per protocol before exposure to phosphoimager screens. Images were analysed using AtlasImage^Ô 1.1 software (BD Bioscience). Intensity values were normalized against three housekeeping genes (-actin, poly-ubiquitin and 40S ribosomal protein S29). A difference in expression level of any individual mRNA was deemed significant if there was a greater than or equal to twofold change in arbitary intensity units above a threshold value of 40.

RT-PCR analysis of total RNA samples was carried out using 30 amplification cycles (45 s denaturation at 94°C, 45 s annealing at 55°C, and 60 s extension at 72°C) followed by a 10 min extension cycle at 72°C. EP receptor PCRs were performed for 35 cycles, as above, but with a 59°C annealing temperature for EP3. Primer sequences for rat cDNAs were as follows: -actin, forward 5'-TAAAGAGAAGCTGTGCTATGTTGC-3', reverse 5'-GATGGAATTGAATGTAGTTTCATGG-3'; TGF- RII, forward 5'-CGTTACACTCCTCTGTGTTACAGG-3' reverse 5'-AAGAACGACAAGAACATTACTCTGG-3'; PAI-1, forward 5'-CATCCTGGAACTGCCCTACC-3', reverse 5'-CCTGGTCATGTTGCTCTTCC-3'; MIF, forward 5'-ACCGGGTCTACATCAACTATTACG-3' reverse 5'-TGGTCTCAAACCATTTATTTCTCC-3'; COX-2, forward 5'-GTCCAGTTTTAAATGAACATGAAGG-3' reverse 5'-CCCAGTTTTAGATTTGTGTTTATGG-3'; EP1, forward 5'-AACTTCTTCGCCTCCTACCC-3' reverse 5'-AGCATCCCCTGTATCTGTGC-3'; EP2, forward 5'-GCTCAACTACGGGGAGTACG-3' reverse 5'-CGAAGGTGATGGTCATAATGG-3'; EP3, forward 5'-CACAGCAACCAGTCAAGTGC-3'; reverse 5'-CGAACACTGTCATGGTCAGC-3'; EP4, forward 5'-TGGCCATCGTAGTATTGTGC-3' reverse 5'-GAAGTAGGCGTGGTTGATG-3'.

Immunofluorescence

Cells grown on glass coverslips were fixed for 10 min at -20°C in 100% methanol. Blocking was performed using 5% (w/v) BSA in PBS for 30 min at 25°C. Coverslips were then incubated with either rabbit anti-cytokeratin (as supplied; Shandon, UK), mouse anti-human E-cadherin (2.5 g/ml; BD Bioscience) or goat anti-mouse COX-2 (5 g/ml; Santa Cruz Biotechnology, Santa Cruz, CA, USA) Abs for 1 h at 25°C. After washing three times with PBS including 0.02% (v/v) Tween 20 (PBS/T), FITC- or tetramethyl-rhodamine isothiocyanate-conjugated secondary Abs (5 g/ml) were incubated with coverslips for 60 min at 25°C. Following further PBS/T washes (´3), coverslips were mounted in 4,6-diamidino-2-phenylindole (DAPI)-incorporated (0.02 g/ml) VectaShield (Vector Laboratories, Burlingame, CA, USA). Immunofluorescence was visualized using a Zeiss Axioscope fluorescence microscope. Images were captured by a CCD camera and processed using the QuipFISH^Ô (Visys, Surrey, UK) programme. Confocal microscopy was performed using a Leica TCS SP laser scanning confocal microscope.

Western blot analysis

Cells were lysed in 50 mM Tris-Cl pH 8.0, 150 mM NaCl, 0.02% (w/v) NaN₃, 0.1% (w/v) SDS, 1% NP-40 and protease inhibitor cocktail (Complete, Roche). One hundred micrograms total protein samples were separated by 12% SDS-PAGE followed by wet transfer to PVDF membranes. Membranes were blocked in PBS incorporating 5% (w/v) non-fat skimmed milk for 1 h at 25°C prior to incubation with primary Ab for 2 h at 25°C. We used goat anti-mouse COX-2 (250 ng/ml), goat anti-mouse COX-1 (250 ng/ml), rabbit anti-human TGF- RII (200 ng/ml; all Santa Cruz) or mouse anti--actin (1:5000; Sigma) Abs diluted in PBS including 1% (w/v) non-fat skimmed milk. Three washes with PBS/T were followed by incubation with the appropriate horseradish peroxidase-conjugated secondary Ab (each 200 ng/ml; DAKO, Ely, UK) for 1 h at 25°C before final washes (´3) with PBS/T. Enhanced chemiluminescence was detected as described by the manufacturer (Perbio Science, Tattenhall, UK).

Measurement of PGE₂ levels

Cells were seeded as described for the MCM experiment and cultured for 12 days in appropriate medium. Old medium was aspirated, fresh control medium was added and then left for 6 h, prior to collection. PGE₂ levels in cell-free supernatants were measured by competitive ELISA (Amersham).

Acknowledgements

We acknowledge the kind assistance of the Molecular Medicine Unit Confocal Imaging Group with the immunofluorescence experiments and the expert assistance of Dr S Shnyder (University of Bradford, UK) with the caecal xenograft experiments. This work was funded by the MRC (UK), Yorkshire Cancer Research and the West Riding Medical Research Trust. MA Hull holds an MRC (UK) Clinician Scientist Fellowship.

Bamba H, Ota S, Kato A, Adachi A, Itoyama S, Matsuzaki F. (1999). Int. J. Cancer, 83: 470-475. Article MEDLINE

Birchmeier C, Birchmeier W, Brand-Saberi B. (1996). Acta Anat., 156: 217-226.

Chapple KS, Cartwright EJ, Hawcroft G, Tisbury A, Bonifer C, Scott N, Windsor AC, Guillou PJ, Markham AF, Coletta PL, Hull MA. (2000). Am. J. Pathol., 156: 545-553. MEDLINE

Chulada PC, Thompson MB, Mahler JF, Doyle CM, Gaul BW, Lee C, Tiano HF, Morham SG, Smithies O, Langenbach R. (2000). Cancer Res., 60: 4705-4708. MEDLINE

Cordon-Cardo C, Prives C. (1999). J. Exp. Med., 190: 1367-1370.

Coussens LM, Werb Z. (2001). J. Exp. Med., 193: F23-F26. Article MEDLINE

Eberhart CE, Coffey RJ, Radhika A, Giardiello FM, Ferrenbach S, DuBois RN. (1994). Gastroenterology, 107: 1183-1188. MEDLINE

Farrow DC, Vaughan TL, Hansten PD, Stanford JL, Risch HA, Gammon MD, Chow W-H, Dubrow R, Ahsan H, Mayne ST, Schoenberg JB, West AB, Rotterdam H, Fraumeni JF, Blot WJ. (1998). Cancer Epidemiol. Biomarkers Prev., 7: 97-102. MEDLINE

Fu S, Ramanujam KS, Wong A, Fantry GT, Drachenberg CB, James SP, Meltzer SJ, Wilson KT. (1999). Gastroenterology, 116: 1319-1329. MEDLINE

Fujita M, Fukui H, Kusaka T, Morita K, Fujii S, Ueda Y, Chiba T, Sakamoto C, Kawamata H, Fujimori T. (2000). J. Gastroenterol. Hepatol., 15: 1277-1281.

Gagliardi G, Kandemir O, Liu D, Guida M, Benvestito S, Ruers TG, Benjamin IS, Northover JM, Stamp GW, Talbot IC. (1995). Virchows Archiv., 426: 149-154.

Giardiello FM, Hamilton SR, Krush AJ, Piantadosi S, Hylind LM, Celano P, Booker SV, Robinson CR, Offerhaus GJ. (1993). N. Engl. J. Med., 328: 1313-1316. Article MEDLINE

Giardiello FM, Yang VW, Hylind LM, Krush AJ, Petersen GM, Trimbath JD, Piantadosi S, Garrett E, Geiman DE, Hubbard W, Offerhaus JA, Hamilton SR. (2002). N. Engl. J. Med., 346: 1054-1059. MEDLINE

Goldman R, Bar-Shavit Z. (1979). J. Nat. Cancer Inst., 63: 1009-1016.

Hardwick JCH, van den Brink GR, Offerhaus GJ, van Deventer SJH, Peppelenbosch MP. (2001). Oncogene, 20: 819-827.

Hauptmann S, Zwadlo-Klarwasser G, Jansen M, Klosterhalfen B, Kirkpatrick CJ. (1993). Am. J. Pathol., 143: 1406-1415.

Hudson JD, Shoaibi MA, Maestro R, Carnero A, Hannon GJ, Beach DH. (1999). J. Exp. Med., 190: 1375-1382. Article MEDLINE

Hull MA, Booth JK, Tisbury A, Scott N, Bonifer C, Markham AF, Coletta PL. (1999). Br. J. Cancer, 79: 1399-1405.

Hull MA, Fenwick SW, Chapple KS, Scott N, Toogood GJ, Lodge JP. (2000a). Clin. Exp. Metastasis, 18: 21-27.

Hull M, Chapple K, Coletta L, Poulsom R. (2000b). Gastroenterology, 119: 600-602.

Hwang D, Jang BC, Yu G, Boudreau M. (1997). Biochem. Pharmacol., 54: 87-96.

Jacoby RF, Seibert K, Cole CE, Kelloff G, Lubet RA. (2000). Cancer Res., 60: 5040-5044. MEDLINE

Khan KNM, Masferrer JL, Woerner BM, Soslow R, Koki AT. (2001). Scand. J. Gastroenterol., 36: 865-869.

Kutchera W, Jones DA, Matsunami N, Groden J, McIntyre TM, Zimmerman GA, White RL, Prescott SM. (1996). Proc. Natl. Acad. Sci. USA, 93: 4816-4820. Article MEDLINE

Manning AM, Williams AC, Game SM, Paraskeva C. (1991). Oncogene, 6: 1471-1476. MEDLINE

Marx J. (2001). Science, 291: 581-582.

Morris CD, Armstrong GR, Bigley G, Green H, Attwood SEA. (2001). Am. J. Gastroenterol., 96: 990-996.

Muller-Decker K, Albert C, Lukanov T, Winde G, Marks F, Furstenberger G. (1999). Int. J. Colorectal Dis., 14: 212-218.

Mutoh M, Watanabe K, Kitamura T, Shoji Y, Takahashi M, Kawamori T, Tani K, Kobayashi M, Maruyama T, Kobayashi K, Ohuchida S, Sugimoto Y, Narumiya S, Sugimura T, Wakabayashi K. (2002). Cancer Res., 62: 28-32.

Nau GJ, Richmond JFL, Schlesinger A, Jennings EG, Lander ES, Young RA. (2002). Proc. Natl. Acad. Sci. USA, 99: 1503-1508. Article MEDLINE

O'Byrne KJ, Dalgleish AG. (2001). Br. J. Cancer, 85: 473-483.

Oshima M, Dinchuk JE, Kargman SL, Oshima H, Hancock B, Kwong E, Trzaskos JM, Evans JF, Taketo MM. (1996). Cell, 87: 803-809. MEDLINE

Oshima M, Murai N, Kargman S, Arguello M, Luk P, Kwong E, Taketo MM, Evans JF. (2000). Cancer Res., 61: 1733-1740.

Paraskeva C, Corfield AP, Harper S, Hague A, Audcent K, Williams AC. (1990). Anticancer Res., 10: 1189-1200. MEDLINE

Pitot HC, Hikita H, Dragan Y, Sargent L, Haas M. (2000). Aliment. Pharmacol. Therap., 14: (Suppl. 1) S153-S160.

Quaroni A, Wands J, Trelstad RL, Isselbacher KJ. (1979). J. Cell Biol., 80: 248-265. MEDLINE

Raschke WC, Baird S, Ralph P, Nakoinz I. (1978). Cell, 15: 261-267.

Rosenberger CM, Scott MG, Gold MR, Hancock REW, Finlay BB. (2000). J. Immunol., 164: 5894-5904.

Sano H, Kawahito Y, Wilder RL, Hashiramoto A, Mukai S, Asai K, Kimura S, Kato H, Kondo M, Hla T. (1999). Cancer Res., 55: 3785-3789.

Seno J, Oshima M, Ishikawa T, Oshima H, Takaku K, Chiba T, Narumiya S, Taketo MM. (2002). Cancer Res., 62: 506-511.

Sharma RA, Gescher A, Plastaras JP, Leuratti C, Singh R, Gallacher-Horley B, Offord E, Marnett LJ, Steward WP, Plummer SM. (2001). Carcinogenesis, 22: 1557-1560.

Shattuck-Brandt RL, Lamps LW, Heppner Goss KJ, DuBois RN, Matrisian LM. (1999). Mol. Carcinog., 24: 177-187. Article MEDLINE

Shattuck-Brandt RL, Varilek GW, Radhika A, Yang F, Washington MK, DuBois RN. (2000). Gastroenterology, 118: 337-345. MEDLINE

Sheehan KM, Sheahan K, O'Donoghue DP, MacSweeney F, Conroy RM, Fitzgerald DM, Murray FE. (1999). JAMA, 282: 1254-1257. MEDLINE

Sheng GG, Shao J, Sheng H, Hooton EB, Isakson PC, Morrow JD, Coffey RJ, DuBois RN, Beauchamp RD. (1997). Gastroenterology, 113: 1883-1891.

Sheng H, Shao J, O'Mahony CA, Lamps L, Albo D, Isakson PC, Berger DH, DuBois RN, Beauchamp RD. (1999). Oncogene, 18: 855-867.

Sheng H, Shao J, Washington KM, DuBois RN. (2001). J. Biol. Chem., 276: 18075-18081. Article MEDLINE

Sinicrope FA, Roddey G, McDonnell TJ, Shen Y, Cleary KR, Stephens LC. (1996). Clin. Cancer Res., 2: 1999-2006.

Smith M-L, Hawcroft G, Hull MA. (2001). Eur. J. Cancer, 36: 664-674.

Sonoshita M, Takaku K, Sasaki N, Sugimoto Y, Ushikubi F, Narumiya S, Oshima M, Taketo MM. (2001). Nature Med., 7: 1048-1051.

Soslow RA, Dannenberg AJ, Rush D, Woerner BM, Khan KN, Masferrer J, Koki AT. (2000). Cancer, 89: 2637-2645. MEDLINE

Steinbach G, Lynch PM, Phillips RK, Wallace MH, Hawk E, Gordon GB, Wakabayashi N, Saunders B, Shen Y, Fujimura T, Su LK, Levin B. (2000). N. Engl. J. Med., 342: 1946-1952. Article MEDLINE

Sung JJY, Leung WK, Go MYY, To KF, Cheng ASL, Ng EKW, Chan FKL. (2000). Am. J. Pathol., 157: 729-735. MEDLINE

Taketo MM. (1998). Clinical significance and potential of selective COX-2 inhibitors Vane JR and Botting RM (eds) London: William Harvey Press, pp. 129-139.

Takeuchi M, Kobayashi M, Ajioka Y, Honma T, Suzuki Y, Azumaya M, Narisawa R, Hayashi S, Asakura H. (2002). Int. J. Colorectal Dis., 17: 144-149.

Tsujii M, DuBois RN. (1995). Cell, 83: 493-501. MEDLINE

Tsujii M, Kawano S, DuBois RN. (1997). Proc. Natl. Acad. Sci. USA, 94: 3336-3340. Article MEDLINE

Tsujii M, Kawano S, Tsuji S, Sawaoka H, Hori M, DuBois RN. (1998). Cell, 93: 705-716. MEDLINE

Williams CS, Watson AJ, Sheng H, Helou R, Shao J, DuBois RN. (2000a). Cancer Res., 60: 6045-6051. MEDLINE

Williams CS, Tsujii M, Reese J, Dey SK, DuBois RN. (2000b). J. Clin. Invest., 105: 1589-1594. MEDLINE

Willson JKV, Bittner GN, Oberley TD, Meisner LF, Weese JL. (1987). Cancer Res., 47: 2704-2713. MEDLINE

Wilson KT, Fu S, Ramanujam KS, Meltzer SJ. (1998). Cancer Res., 58: 2929-2934. MEDLINE

Yang VW, Shields JM, Hamilton SR, Spannhake EW, Hubbard WC, Hylind LM, Robinson CR, Giardiello FM. (1998). Cancer Res., 58: 1750-1753. MEDLINE

Zhang T, Nanney LB, Peeler MO, Williams CS, Lamps L, Heppner KJ, DuBois RN, Beauchamp RD. (1997). Cancer Res., 57: 1638-1643.

Figure 1 COX-2 and PGE₂ EP receptor expression. (a) Western blot analysis of COX-2 and -actin protein expression by non-activated RAW264.7 macrophages (NM) and RAW264.7 macrophages activated by IFN- and LPS in the absence (AM) or presence (SC-AM) of 2 M SC236. Up-regulation of COX-2 protein expression was maintained for 24 h during medium conditioning (data not shown). (b) Concentration-dependent inhibition of activated RAW264.7 macrophage PGE₂ synthesis by SC236. Data are expressed as the mean±s.e.m. PGE₂ production per 10⁶ cells over 24 h (n=3). (c) SC236 (1 M) does not affect IEC-6 cell viability. Cells (5´10⁴) were cultured in the presence of varying concentrations of SC236 for 96 h. Data are expressed as the mean+s.e.m. relative cell number (measured by MTS assay) compared with control cells (n=3). (d) RT-PCR analysis of PGE₂ EP receptor mRNA expression by IEC-6 cells. Rat kidney cDNA (Clontech) was used as a positive control

Figure 2 In vitro models of paracrine macrophage-intestinal epithelial cell signalling. (a) IEC-6 intestinal epithelial cells were cultured in the presence of either RAW264.7 macrophage-conditioned medium (MCM) or co-cultured with RAW264.7 macrophages. Fresh medium or macrophage cultures were replaced every 2-3 days. M, macrophages; Epi, epithelial cells. (b) Controls used in the MCM and co-culture experiments. Medium from the last wash (LW) of activated macrophages, prior to conditioning was used to exclude possible effects of residual trace amounts of IFN- and LPS following macrophage activation. The selective COX-2 inhibitor SC236 (SC) was added either during macrophage activation and medium conditioning (SC-AMCM) or directly to the medium (after conditioning) used for intestinal epithelial cell culture (Con+SC and AMCM+SC). Identical controls were utilized in the co-culture model. All epithelial cell cultures (Con+SC, SC-AMCM or AMCM+SC) were exposed to the same concentration of SC236 (1 M) following 1:2 dilution of MCM in control medium

Figure 3 Morphology of intestinal epithelial cells. Phase-contrast photomicrographs of IEC-6 cells cultured for 12 days (a-e) and 82 days (f-j) in the presence of control medium (a, f), NMCM (b, g), AMCM (c, h), SC-AMCM (d, i) or AMCM+SC (e, j). Photomicrographs of H&E-stained IEC-6 cells on insert membranes co-cultured for 14 days in the presence of macrophage-free control medium (k), NM (l), AM (m), SC-AM (n), AM+SC (o). Immunofluorescence for cytokeratin on control (p) and AMCM-cultured (q) cells (passage 4). Blue; nuclear DAPI staining, Red; cytokeratin staining. Immunofluorescence for E-cadherin on control (r) and AMCM-cultured (s) cells (passage 4). Cell network formation by AMCM-cultured cells (passage 6) on Matrigel^Ô at day 4 (t). All control, NMCM-cultured and SC-AMCM-cultured cells underwent apoptosis after 24-48 h on Matrigel^Ô. However, AMCM+SC-cultured cells remained viable with similar morphology to AMCM-cultured cells (data not shown)

Figure 4 Changes in intestinal epithelial cell gene expression during AMCM culture. (a) cDNA array analysis of mRNA levels in control (Con) and AMCM-cultured cells (passage 6). Numbers denote signals for PAI-1 (1) and TGF- RII (2) and MIF (3). (b) Confirmatory RT-PCR analysis of mRNA expression using the cDNAs tested on the arrays. (c) Densitometric analysis of RT-PCR for TGF- RII mRNA from different MCM cultures (passage 6). A representative gel and the mean+s.e.m. densitometric values for the TGF- RII mRNA level normalized to -actin mRNA (n=3) are shown. (d) Western blot analysis of TGF- RII protein expression by intestinal epithelial cells (passage 6). AC denotes cells cultured in AMCM for 84 days (passage 6), which were then cultured in control medium for 28 days before analysis. (e) The effect of rhTGF-₁ on control, AMCM-cultured and SC-AMCM-cultured intestinal epithelial cell proliferation. Equal numbers of cells (1´10⁵) were incubated with 4 ng/ml rhTGF-₁ for 48 h before cell counting. Data are expressed as the mean+s.e.m. percentage number of cells compared with untreated cells (n=3)

Figure 5 COX-2 expression by intestinal epithelial cells. (a) Western blot analysis of COX-2 protein expression (passage 6). AC denotes cells cultured in AMCM for 84 days (passage 6), which were then cultured in control medium for 28 days before analysis. (b) Immunofluorescence for COX-2 on control (Con) and AMCM-cultured cells (passage 6). Blue; nuclear DAPI staining, Red; COX-2 staining

Figure 6 Intestinal epithelial cell proliferation and apoptosis during MCM culture. (a) and (b) passage 1 (day 1-14), (c) and (d) passage 6 (day 71-84). (a) and (c) are proliferation curves. Filled symbols, control cells; open symbols, AMCM-cultured cells. (b) and (d) show the cell number at day 12 (open boxes) and the cumulative apoptotic cell count (filled boxes; determined from floating cell counts which were obtained every 48 h for 14 days). Data are expressed as the mean+s.e.m. of three independent MCM cultures

Table 1 The intestinal epithelial cell phenotype induced by IFN-/LPS-activated macrophages in a COX-2-dependent manner