Crosstalk of oncogenic and prostanoid signaling pathways

Introduction

The cyclooxygenases Cox-1 and Cox-2 are the rate-limiting enzymes in the synthesis of all prostanoids from arachidonic acid (Smith et al. 2000). While Cox-1 is expressed constitutively in a subset of cell types, Cox-2 is highly regulated by transcriptional and post-translational mechanisms in response to a plethora of stimuli. Induction of Cox-2 triggers the synthesis of different prostanoids that play essential roles in many physiological processes and responses, such as inflammation, pain, fever, and platelet aggregation. Cyclooxygenases catalyze a two-step reaction that converts arachidonic acid to prostaglandin H₂ (PGH₂) which in turn serves as the precursor for the synthesis of all biologically active prostanoids: PGD₂, PGE₂, PGF₂, prostacyclin (PGI₂), 15-deoxy-Δ^12,14-PGJ₂, and thromboxane A₂ (Fig. 1). Work of the past decade has clearly established Cox-2 and a subset of prostanoids and their receptors as crucial players in oncogenesis that regulate, and are regulated by, pathways with essential functions in oncogenesis. This article will provide a review of the literature covering this area of research.

Cox-2 and cancer

Cox-2 is overexpressed in many experimental and human tumors as a consequence of deregulated signaling pathways involved in the control of Cox-2 transcription, steady-state RNA levels, and/or translation (see also Fig. 2). Multiple pathways regulating Cox-2 expression have been described with a predominant role for Ras transmitted signals. Thus, Cox-2 transcription is upregulated through the Ras-Raf-MEK-ERK and Ras-Rac-MEKK1-JNKK-JNK cascades (Araki et al. 2003; Sheng et al. 1998; Subbaramaiah et al. 2002; Xie & Herschman 1995; Xie & Herschman 1996) (M. Kreutzer and R.M., unpubl. observation). While ERK phosphorylates and thereby activates the transcription factors PEA3 (an Ets family member) and C/EPB, JNK targets the transcriptional activator AP-1 (Jun/Fos heterodimers) interacting with an ATF/CRE site in the Cox-2 promoter (Reddy et al. 2000; Subbaramaiah et al. 2002; Xie & Herschman 1995). The Cox-2 promoter is also controlled by other oncogenesis-related mechanisms, including NFκB activation by hypoxia (Ji et al. 1998; Schmedtje et al. 1997), Wnt/Apc-controlled upregulation of the PEA3 and TCF-4 transcription factors (Araki et al. 2003; Howe et al. 2001), p53 (Han et al. 2002; Subbaramaiah et al. 1999) as well as RhoB and cAMP signaling (Shao et al. 2000). Furthermore, regulation of Cox-2 mRNA stability via the Ras-regulated Raf-MEK-ERK and PI₃K-PDK-AKT pathways has been reported (Lin et al. 2001; Sheng et al. 2000; Sheng et al. 2001).

There is an overwhelming body of evidence indicating that the deregulation or overexpression of Cox-2 expression in tumor and/or tumor stroma cells plays a major role in oncogenesis (Dannenberg & Subbaramaiah 2003; Gupta & Dubois 2001; Williams et al. 1999). This conclusion is mainly based on work with non-steroidal anti-inflammatory drugs (NSAIDs) that are known to inhibit either both cyclooxygenases, such as aspirin, or specifically Cox-2, such as the recently developed coxibs (e.g, rofecoxib, celecoxib) (Flower 2003). These studies have clearly demonstrated the potential of Cox-2 inhibiting NSAIDs to interfere with cell cycle progression and to induce apoptosis in tumor cells (Arico et al. 2002; Chen et al. 2003a; Cheng et al. 2002; Denkert et al. 2003; Detjen et al. 2003; Ding et al. 2000; Elder et al. 2000; Elder et al. 2002; Grosch et al. 2001; Hida et al. 2000; Hu et al. 2003; Jendrossek et al. 2003; Kardosh et al. 2004; Kundu et al. 2002; Leahy et al. 2002; Lee et al. 2002; Li et al. 2001; Liu et al. 1998; Minter et al. 2003; Moalic et al. 2001; Nakanishi et al. 2001; Narayanan et al. 2003; Peng et al. 2003; Poon et al. 2001; Richter et al. 2001; Romano & Claria 2003; Sawaoka et al. 1998; Souza et al. 2000; Totzke et al. 2003; Uefuji et al. 2000; Ueta et al. 2001; Wu et al. 2003; Yamazaki et al. 2002). In addition, NSAIDs have been shown to interfere with Cox-2 mediated processes that play essential roles in tumor progression, such as angiogenesis, tumor cell invasion and metastasis. The underlying molecular mechanism include:

1. Downregulation of cyclin-dependent kinases (Cdks), partly through Rb hyperphosphorylation and upregulation of Cdk inhibitors (Grosch et al. 2001; Kardosh et al. 2004; Nakanishi et al. 2001; Narayanan et al. 2003; Peng et al. 2003; Tseng et al. 2002; Yao et al. 2003; Yao et al. 2000);

2. Downregulation of anti-apoptotic proteins of the Bcl-2 family (Chen et al. 2003b; Gately & Kerbel 2003; Jendrossek et al. 2003; Lin et al. 2001; Liu et al. 1998; Nzeako et al. 2002; Richter et al. 2001; Uefuji et al. 2000; Ueta et al. 2001);

3. Modulation of the hyaluronate receptor CD44 and matrix metalloproteinase activities (Attiga et al. 2000; Dohadwala et al. 2002; Dohadwala et al. 2001; Liu et al. 2002; Pai et al. 2002; Pan et al. 2001); and

4. Inhibition of vascular endothelial growth factor (VEGF) synthesis (Kim et al. 2003; Li et al. 2002; Masferrer et al. 1999; Ogawa et al. 2003; Su et al. 2004; Williams et al. 2000).

It is therefore no surprise that NSAIDs have been shown to be potent suppressors of tumor growth in diverse mouse models, including human xenografts, chemically induced tumors as well as tumors developing in transgenic mice (see Table 1 for summaries and references). The crucial role of Cox-2 in oncogenesis has also been clearly demonstrated in multiple studies making use of mice with targeted disruptions of both alleles of the Cox-2 gene (see Table 2). Remarkably, an anti-tumorigenic effect has also been clearly documented in humans where the long-term use of aspirin appears to reduce the incidence of different types of cancer, in particular the development of colorectal carcinomas in patients with familial polyposis coli (Gupta & Dubois 2001; Markenson 1999; Reddy & Rao 2002; Williams et al. 1999).

Prostanoid signaling and oncogenesis

PGD₂, PGE₂, PGF₂, prostacyclin, and thromboxane A₂ interact with G-protein coupled membrane receptors, termed DP, EP-1 through EP-4, FP, IP and TP, respectively (Fig. 1). These receptors trigger multiple second messenger generating systems, including adenylate cyclase and phospholipase C (Versteeg et al. 1999). 15-deoxy-Δ^12,14-PGJ₂ also binds to and stimulates the nuclear “orphan” receptor and transcription factor “peroxisome proliferator-activated receptor-γ” (PPAR-γ), while the related PPAR-δ, also known as PPAR-β, interacts with prostacyclin (Fig. 1) (Berger & Moller 2002; Forman et al. 1996; Forman et al. 1995; Giguere 1999; Kliewer et al. 1999; Lim & Dey 2002; Michalik et al. 2004). Both PPARs form heterodimers with the retinoic acid receptor RxR and interact with composite DNA elements comprised of RxR and PPAR recognizing half-sites (Gearing et al. 1993; Kliewer et al. 1992; Marcus et al. 1993; Tontonoz et al. 1994).

Several specific components of the prostanoid signaling network have also been associated with oncogenesis, in particular PGE₂ and its membrane receptors EP-2, EP-3, and EP-4 as well as PPAR-γ and PPAR-δ. While PGE₂ signaling seems to play a predominant role in promoting tumor angiogenesis through upregulation of pro-angiogenic growth factors (Kurie & Dubois 2001), PPAR-δ and its ligand prostacyclin as well as PPAR-γ and its natural agonist 15-deoxy-Δ^12,14-prostaglandin J₂ have also been implicated in the regulation of tumor cell proliferation and apoptosis with opposing effects: while PPAR-γ activation results in anti-oncogenic effects, PPAR-δ seems to promote oncogenesis (Michalik et al. 2004). The published work addressing these issues is discussed in detail in the following sections. The only other Cox-2 product implicated in oncogenesis is TxA₂. This is suggested by the observation that the ectopic expression of TxAS increases the tumorigenicity of colon-26 cells in syngeneic mice (Pradono et al. 2002). Clearly, more experimental work is required to clarify a possible role of TxA₂ in oncogenesis.

Prostaglandin E₂ signaling in oncogenesis

Three of the four identified G-protein coupled PGE₂ receptors have been associated with cancer in different mouse models of oncogenesis (summarized in Table 2). Disruption of both EP2 alleles has been reported to inhibit tumor cell proliferation and tumor growth in Apc^Δ716 mice (Seno et al. 2002; Sonoshita et al. 2001). These mice harbor a mutant allele of the Apc tumor suppressor gene (see below) and develop multiple intestinal adenomas with high penetrance. In mice with an additional mutation in the Smad-4 gene, which codes for a component of TGF-β signaling cascade, adenomas progress to adenocarcinomas (Seno et al. 2002). Disruption of EP2 led to reduced polyp incidence and size and inhibited carcinoma growth beyond a small size. These tumors showed decreased microvessel density indicating a critical role for EP2 in tumor angiogenesis (Seno et al. 2002). In contrast, genetic inactivation of EP1, EP3 or EP4 had no effect on tumorigenesis (see Table 3) in Apc^Δ716 mice pointing to a specific role for EP2 in this model (Seno et al. 2002; Sonoshita et al. 2001). EP2 disruption has also been reported to inhibit the growth of syngeneic s.c. transplants of Lewis lung carcinoma and C26 colon carcinoma, but in these cases a major determinant of tumor growth inhibition was an increased anti-tumor CTL response (Yang et al. 2003).

In contrast to the lack of any effect of EP3 disruption on polyp formation in Apc^Δ716 mice (see above), its homozygous deletion has been reported to inhibit the growth of syngeneic s.c. transplants of Sarcoma-180, concomitant with reduced tumor-associated angiogenesis and reduced VEGF expression by stromal fibroblasts (Amano et al. 2003).

The role of EP4 has been studied in another mouse model of intestinal oncogenesis, i.e., the formation of polyps (aberrant crypt foci) in mice treated with the chemical carcinogen azoxymethane (Mutoh et al. 2002). Genetic inactivation of EP4 led to a 44% reduction in the numbers of polyps, while no significant effect on tumor growth was seen when other prostanoids receptor genes (EP2, DP, FP, IP or TP) were genetically inactivated. An inhibitory effect in the same animal model was also seen with a selective EP4 antagonist (Mutoh et al. 2002). In this case, the therapeutic target may be the tumor cell itself rather than tumor angiogenesis supporting stromal cells.

A direct effect of PGE₂ on tumor cells has also been described in another study using colon and gastric cancer cells as a model (Pai et al. 2002). In this study, PGE₂ has been found to transactivate the EGF receptor (EGFR), which requires activation of the tyrosine protein kinase c-Src, metalloproteinase activity and release of the EGFR ligand TGF-α. As a consequence of EGFR activation, the Erk signaling pathway is induced potentially leading to increased cell proliferation and promoting oncogenesis.

Prostacyclin and PPAR-δ signaling

PPAR-δ is a nuclear receptor that binds to, and is activated by, prostacyclin. However, a number of other natural fatty acids, including arachidonic acid and eicosapentaenoic acid, also interact with PPAR-δ, albeit with an apparently lower activation potential (Forman et al. 1997; Xu et al. 1999). It remains therefore unclear whether prostacyclin is the only physiologically relevant PPAR-δ ligand, but so far the role of other fatty acids or derivatives in PPAR-δ mediated oncogenesis-related processes has not been investigated. PPAR-δ heterodimerizes with the retinoic acid receptor RxR and interacts with a number of transcriptional cofactors and other nuclear proteins (Fig. 3), but their precise function in transcriptional regulation by PPAR-δ is unclear.

A role for PPAR-δ in tumorigenesis has been shown in intestinal cells where the Apc tumor suppressor gene product normally inhibits the TCF-4 mediated transcriptional induction of PPAR-δ (He et al. 1999), a transcriptional target of the Wnt pathway (Willert & Nusse 1998) (see also Fig. 4). In colorectal polyposis, Apc is genetically inactivated and transcription of PPAR-δ is increased due to deregulated TCF-4 activity. That this increase in PPAR-δ expression is functionally relevant is suggested by two independent observations. First, the disruption of both PPAR-δ alleles in the human colorectal carcinoma cell line HCT116 has been reported to inhibit tumor growth in immune-deficient mice (Park et al. 2001). The second piece of evidence has been obtained with Apc^Min mice, which harbor a dominant mutant Apc allele and develop intestinal adenomas with 100% penetrance. Even though the homozygous deletion of PPAR-δ in these mice does not lead to a lower overall tumor incidence, the growth of tumors beyond a diameter of approximately 2 mm is dramatically reduced (Barak et al. 2002), suggesting a role for PPAR-δ in tumor progression rather than initiation. In agreement with these findings is the recent observation that treatment of Apc^Min mice with the PPAR-δ agonist GW501516 led to a fivefold increase in the number of polyps larger than 2 mm, clearly implicating PPAR-δ in the regulation of intestinal adenoma growth (Gupta et al. 2004).

In line with a pro-oncogenic potential of prostacyclin is the observation that prostacyclin released by human colon carcinoma stromal fibroblasts promotes the survival of the tumor cells (Cutler et al. 2003), that keratinocytes from PPAR-δ null mice show an increased rate of apoptosis (Tan et al. 2001) and that apoptosis in mesenchymal renal medullary interstitial cells is inhibited by prostacyclin and PPAR-δ (Hao et al. 2002). Furthermore, the 15-lipoxygenase-1 product 13(S)-HODE, a PPAR-γ agonist (Nagy et al. 1998) and inhibitor of PPAR-δ expression (Shureiqi et al. 2003), induces apoptosis in colon cancer cells in a PPAR-δ dependent fashion (Shureiqi et al. 2003), confirming a crucial role for PPAR-δ in the 13(S)-HODE induced inhibition of proliferation.

Prostacyclin probably also plays a role in angiogenesis. Female Cox-2 null mice are unable to produce litters due to a failure in blastocyst implantation and decidualization (Dinchuk et al. 1995; Lim et al. 1997). These mice synthesize approximately 50% lower levels of prostacyclin, PGE₂, and TxB₂ at the implantation site (Lim et al. 1999). In addition, expression of the VEGF receptor Flk-1 at the implantation site is reduced. The implantation defect can be partially rescued by treatment with prostacyclin, which in this scenario probably acts through PPAR-δ. Interestingly, treatment with prostacyclin and PGE₂ also induces Flk1 expression at the implantation site (Lim et al. 1999). Collectively, these observations point to an involvement of prostacyclin-PPAR-δ signaling in angiogenesis during decidualization. A pro-angiogenic function or prostacyclin is also suggested by two other observations: perfusion of rat lung tissue with prostacyclin (or PGE₂) induces the synthesis of VEGF (Hoper et al. 1997), and the antisense-mediated inhibition of PGIS has been reported to interfere with capillary-like tube formation in HUVEC cultures (Spisni et al. 2001).

In apparent contrast to the observations summarized above, the ectopic expression of PGIS has been reported to inhibit chemically induced lung tumorigenesis in mice (Keith et al. 2002) and to promote apoptosis in the human embryonic kidney cell line 293 (Hatae et al. 2001). These apparent discrepancies may be due to different reasons. It is possible that intracellularly synthesized prostacyclin (by PGIS overexpression) and prostacyclin acting in a paracrine fashion have fundamentally different biological effects due to differences in subcellular distribution or interactions with other cellular components, such as fatty acid binding proteins (FABPs) serving as transport vehicles for nuclear translocation and modifiers of the transcriptional and biological response (Tan et al. 2002). Furthermore, cell type-specific differences in the effects of prostacyclin and PPAR-δ cannot be ruled out. In this context, the availability of other proteins interacting with, and modulating the function of, PPAR-δ, such as the retinoic acid receptor RxR or FABPs, may be crucial. Finally, PPAR-δ ligands other than prostacyclin most likely exist (Forman et al. 1997; Shaw et al. 2003; Xu et al. 1999), so that a functional link between prostacyclin and PPAR-δ is not obligatory. Clearly, more work is required to clarify the role of prostacyclin and PPAR-δ mediated signaling in tumorigenesis.

To date, only few target genes of PPAR-δ are known. Of particular interest with respect to tumorigenesis is the agonist-dependent PPAR-δ mediated upregulation of the Akt pathway in keratinocytes through the transcriptional induction of PDK-1 and ILK and the repression of the phosphatidylinositolphosphate phosphatase and tumor suppressor gene PTEN (Di-Poi et al. 2002). In vascular smooth muscle cells, PPAR-δ modulates the expression several cell cycle regulators (cyclin A, cdk2, p57^Kip) (Zhang et al. 2002) but this was not observed in skeletal muscle cells (Dressel et al. 2003). It is currently not known whether the regulation of Cdk activity by PPAR-δ plays a role in tumor cell proliferation. Finally, as PPAR-δ appears to be able to repress the transcriptional activity of PPAR-γ (Shi et al. 2002), targets of PPAR-γ (see below) might also represent indirect targets for PPAR-δ.

PPAR-γ signaling

PPAR-γ is another nuclear receptor that interacts with fatty acid derivatives. PPAR-γ shows similar protein-protein interactions as PPAR-δ (see above and Fig. 3), but is activated by different ligands, including the natural prostanoid 15-PGJ₂ and a number of isoform-specific synthetic agonists (thiazolidinedione anti-diabetic drugs), such as rosiglitazone (BRL 49653), troglitazone, ciglitizone and pioglitazone (Fig. 3). In addition, PPAR-γ specific antagonists have been identified, e.g., GW9662 and BADGE. A large body of evidence suggests that PPAR-γ has anti-oncogenic properties. Activation of PPAR-γ by 15-PGJ₂ or synthetic ligands inhibits tumor cell proliferation in vitro, suppresses tumor growth in various mouse models and induces tumor cell apoptosis in vivo and in vitro (see Table 4 for details and references). In addition, it has been reported that PPAR-γ is expressed at high levels in tumor endothelial cells, and that PPAR-γ ligands inhibit endothelial cells proliferation and interfere with tumor angiogenesis in mouse models (Panigrahy et al. 2002; Xin et al. 1999).

In Apc^Min mice, PPAR-γ agonists paradoxically enhance the formation of intestinal polyps (Lefebvre et al. 1998; Saez et al. 1998) whereas the growth of chemically induced polyps is inhibited (Osawa et al. 2003; Tanaka et al. 2001). A possible explanation is suggested by the observation that the heterozygous disruption of PPAR-γ leads to an increase in the level of β-catenin (Girnun et al. 2002), a central component of the Wnt-Apc pathway (see Fig. 4) (Willert & Nusse 1998). This induction of β-catenin depends on the presence of two intact Apc alleles, possibly because PPAR-γ targets one of the components acting upstream of β-catenin, such as axins, GSK-3 or Apc itself (Girnun et al. 2002). In analogy, ligand-activated PPAR-γ should be able to suppress β-catenin levels only in an Apc wildtype situation (Fig. 3), which would explain the discrepancies seen in different animal models of intestinal tumorigenesis.

The PPAR-γ mediated downregulation of β-catenin described above (Girnun et al. 2002) results in decreased transcriptional activity of TCF-4 and suppression of c-myc and cyclin D1 as target genes of the Wnt pathway (He et al. 1998; Tetsu & McCormick 1999; Yamakawa-Karakida et al. 2002). PPAR-γ thus impinges on oncogenesis by modulating the expression of essential components of the cell cycle control machinery. Intriguingly, β-catenin/TCF-4 signaling induces PPAR-δ in colorectal carcinoma cells (He et al. 1999). It is therefore tempting to speculate that PPAR-γ agonists exert their anti-tumorigenic effect in apart by downregulating PPAR-δ expression (Fig. 2 and Fig. 4). This may also explain the anti-proliferative and proapoptotic effect of the PPAR-γ agonist 13(S)-HODE described above (Shureiqi et al. 2003).

The microarray-based search for PPAR-γ target genes has provided additional hints how PPAR-γ might exert its inhibitory function during oncogenesis. Microarray studies with colon carcinoma cells have led to the identification of a number of bona fide PPAR-γ target genes with potential functions in growth, differentiation, and adhesion (Gupta et al. 2001) (see also Fig. 4). Although the role of these genes in colon tumorigenesis has not been addressed directly, one of them, RegIA, appears to be of particular interest. RegIA encodes a secreted protein that is overexpressed in colon carcinomas and has oncogenic properties in transgenic mice. It is thus conceivable that the observed downregulation of RegIA by PPAR-γ agonists might contribute to the inhibitory effect of these drugs on intestinal tumorigenesis.

PPAR-γ agonists have also been reported to modulate the expression and/or activity of a number of other cell cycle regulators and suppressors of apoptosis, including inhibition of cyclin D1 expression (Kitamura et al. 2001; Qin et al. 2003; Toyota et al. 2002; Wang et al. 2001), Rb hyperphosphorylation (Wakino et al. 2000), and inhibition of E2F activity (Altiok et al. 1997), induction of the Cdk inhibitors p18^ink4C, p21^WAF1 and p27^Kip1 (Koga et al. 2001; Morrison & Farmer 1999), upregulation of the growth inhibitory GADD153 gene (Satoh et al. 2002) and the TGF-β target gene TSC22 (Gupta et al. 2003), induction of the tumor suppressor and PI3 K antagonist PTEN (Farrow & Evers 2003; Patel et al. 2001), and inhibition of NfkB and Bcl-2 (Chen et al. 2002). Induction of ubiquitin-dependent protein degradation by PPAR-γ ligands has been described as an additional mechanism leading to reduced cyclin D1 (Qin et al. 2003) and cFLIP protein levels (Kim et al. 2002).

PPAR-γ have also been reported to suppress VEGF expression by tumor cells (Panigrahy et al. 2002), to inhibit the expression of the VEGF receptor genes Flt-1 and Flk-1 by endothelial cells (Xin et al. 1999), to downregulate matrix metalloproteinase (MMP) activity secreted by tumor cells and induce MMP inhibitors (Liu et al. 2003; Panigrahy et al. 2002; Sunami et al. 2002), supporting the notion that PPAR-γ plays a negative regulatory role in tumor metastasis and angiogenesis.

The results summarized above suggest that PPAR-γ agonists modulate the expression and activity of a large number of genes and proteins, and thereby affect multiple biological processes that are directly relevant to tumorigenesis. It is, however, unlikely that all these effects reflect bona fide functions of PPAR-γ. Thus, PPAR-γ independent effects by PPAR-γ agonists on gene expression, cell proliferation, and apoptosis have been described in several cases (Baek et al. 2004; Baek et al. 2003; Chawla et al. 2001; Clay et al. 2002; Laurora et al. 2003; Nosjean & Boutin 2002; Palakurthi et al. 2001). Of particular relevance in this context is the observation that thiazolidinediones inhibit translation initiation through inactivation of eukaryotic initiation factor 2 (eIF2) (Palakurthi et al. 2001) which in analogy to inhibitors of the mTOR kinase (like rapamycin) (Mills et al. 2001) might explain many of the anti-oncogenic effects of this class of PPAR-γ agonists. Even though these observations cast some doubt on the mechanistic interpretation of a number of published studies (including some of those listed in Table 4), the anti-proliferative, pro-apoptotic, and tumor suppressive effects of thiazolidinedione agonists, and thus their potential as anti-cancer drugs, are out of the question.

Observations made with human tumors support the notion that PPAR-γ has an inhibitory effect on oncogenesis. Thus, the PPAR-γ gene has been found to be altered by loss-of-function mutations in a fraction of human colon carcinomas (Sarraf et al. 1999). In human thyroid follicular carcinomas, the frequent t(2;3)(q13;p25) translocation results in fusion of the DNA binding domains of the thyroid transcription factor PAX8 to domains A to F of PPAR-γ (Kroll et al. 2000). The PAX8-PPAR-γ fusion protein inhibits thiazolidinedione-induced transactivation by PPAR-γ in a dominant negative manner. Furthermore, PPAR-γ transcriptional activity is down-regulated by oncogenic Ras signaling through Erk-mediated phosphorylation of a regulatory serine residue (Adams et al. 1997; Hu et al. 1996; Mueller et al. 1998). Thus, genetic alterations leading to a deregulated Ras pathway, such as constitutive receptor tyrosine kinase activation or Ras mutations, would potentially inhibit the tumor suppressive function of PPAR-γ (Fig. 2). In contrast, PPAR activity is inhibited by protein kinase A (PKA) mediated phosphorylation (Hansen et al. 2001; Lazennec et al. 2000). PKA can antagonize the Ras-Erk pathway and is generally associated with anti-oncogenic properties (Hafner et al. 1994). Opposing effects of Erk and PKA on the function of PPAR-γ are therefore conceivable in view of a tumor suppressor function of PPAR-γ.

On the basis of the observation that agonists for PPAR-γ induce terminal differentiation in normal preadipocytes and human liposarcoma cells in vitro (Tontonoz et al. 1997), three patients with advanced liposarcoma were treated with troglitazone (Demetri et al. 1999). Tumor biopsies showed that terminal adipocytic differentiation concomitant with a marked reduction of cell proliferation was induced in these tumors by troglitazone. Although larger clinical trials have to be carried out before any definitive statement regarding the benefit of troglitazone in the treatment of liposarcoma is possible, the encouraging pilot study by Demetri and colleagues suggests that PPAR-γ agonists may indeed turn out as suitable drugs for the treatment of a subset of human cancers.