Introduction

Ras oncogene activation is frequent in human cancer, with particularly high incidence in certain epithelial cell types, including colon, pancreas and thyroid (Bos, 1989),and at least in the latter two cases, Ras mutation appears to be an initiating event in tumorigenesis (Almoguera et al., 1988; Lemoine et al., 1989b; Lemoine et al., 1990; Suarez et al., 1990). The majority of work carried out to identify the signalling pathways downstream of Ras has been carried out in fibroblast or rodent epithelial cell lines, which can exhibit very different responses to active Ras (al-Alawi et al., 1995; Khosravi-Far et al., 1996; Oldham et al., 1996; Graham et al., 1999). For example, active Ras causes the growth arrest of primary human fibroblasts (Serrano et al., 1997). However, in contrast and consistent with its putative role in tumour initiation, Ras oncogene activation switches on proliferation in normal human thyroid epithelial cells in vitro (Bond et al., 1994).

Whereas normal thyrocytes have an extremely low proliferative rate both in monolayer culture and in the intact tissue, the stable expression of mutant Ras induces a dramatic proliferative response that is sustained for up to 25 population doublings, closely resembling the phenotype of an early-stage thyroid tumour, follicular adenoma (Bond et al., 1994). We have previously demonstrated that at least two effector pathways, Raf/MAP kinase (MAPK) and phosphatidylinositol 3-kinase (PI3K), are necessary (although individually not sufficient) for the proliferative response to Ras in thyrocytes. PI3K activation appears to be required for the suppression of Ras-induced apoptosis, and the activation of the two pathways together leads to a weak synergistic effect, mimicking the proliferative effects of active Ras, although with greatly reduced efficiency (Gire et al., 1999, 2000).

As Raf and PI3K activation together could not reproduce the level of proliferation seen on introduction of H-RasV12 into primary thyroid epithelial cells, it was likely that at least one further Ras effector was involved, and a possible candidate was RalGEF (Marshall, 1996; Wolthuis and Bos, 1999). Using another Ras effector mutant (H-RasV12/G37) that selectively activates the RalGEF pathway, Miller et al. (1997, 1998) showed that they could induce proliferation but not transformation of Wistar rat thyroid epithelial (WRT) cells, and that expression of a dominant-negative Ral construct (RalN28) reduced this proliferation. Recent work by Hamad et al. (2002) comparing human and murine cell lines shows that activation of RalGEF is both necessary and sufficient for Ras-induced anchorage-independent growth as a measure of transformation by Ras.

While it is clear that different signalling pathways operate in different species, we sought therefore to identify whether activation of RalGEF was involved in the Ras-induced proliferation of primary human thyroid cells and whether it was able to synergize with the activation of PI3K and Raf/MAPK to emulate the effects of H-RasV12. In this paper, we show that (i) H-RasV12 activates the RalGEF pathway, (ii) this activation, although not sufficient alone, is necessary for the proliferative effect seen with mutant Ras in human thyroid epithelial cells and (iii) activation of all three pathways is able to reproduce the Ras-induced proliferation of primary thyrocytes.

Results

Evidence for the activation of RalGEF by H-RasV12 in primary human thyroid epithelial cells

To determine initially whether H-RasV12 activates the RalGEF pathway in thyroid epithelium, primary thyroid epithelial cells were infected with a retroviral vector encoding H-RasV12 (pLXSN H-RasV12), and 3 weeks after selection in G418, lysates were prepared from pooled colonies. The amount of Ral-GTP in each sample was assessed by RalBD pull-down assay, followed by Western blotting to detect RalA (see Figure 1). The total Ral content was assessed by Western blotting of the unfractionated lysate. The amount of Ral-GTP, corrected for total Ral protein, purified from the Ras-induced colonies was at least fivefold higher than that seen in the normal thyrocytes, hence confirming that expression of H-RasV12 does indeed lead to the activation of the RalGEF pathway in human thyrocytes.

Figure 1
figure 1

H-RasV12 activates RalGEF in primary human thyroid epithelial cells. A representative RalBD pull-down analysis from the lysates of pooled Ras colonies or normal thyrocytes with GST-RalBD (upper panel). Total Ral content was estimated by Western blot of the whole lysate (lower panel)

Full size image

Activation of the RalGEF pathway is not sufficient to induce thyroid epithelial colonies

Having established that the RalGEF pathway is activated by H-RasV12, we next investigated whether activation of the RalGEF pathway was alone sufficient to induce proliferation of thyroid epithelial cells. In order to do this, we utilized a Ras effector mutant, H-RasV12/G37, previously shown to activate specifically the RalGEF pathway without activating either Raf or PI3K (White et al., 1995), or a constitutively active RalGEF (Rlf-CAAX).

Infection of primary thyrocytes with retroviral vector encoding either H-RasV12/G37 (pBabepuro H-Ras G37) or Rlf-CAAX (in pMaRX) failed to produce any epithelial colonies in at least five separate experiments. This inability of RalGEF activation to induce epithelial colony formation is not unexpected, as we have already shown that both Raf/MAPK and PI3K activation are necessary for Ras-induced proliferation.

Effect of inhibition of RalGEF pathway on proliferation

In the absence of any available specific pharmacological inhibitors of the RalGEF pathway to investigate whether RalGEF is necessary for Ras-induced thyroid cell proliferation, we made use of a dominant-negative Ral construct, RalN28 (pMaRXpuroRalN28), which forms a nonproductive complex with the RalGEFs (Urano et al., 1996) impeding the activation of endogenous Ral.

First, in order to validate the effect of RalN28 expression, HT-Ori3 cells were coinfected with pLXSN H-RasV12 plus either pMaRXpuroRalN28 or pBabepuro empty vector. Pooled colonies from each were lysed and subjected to RalBD pull down as before, followed by Western analysis of the products. When corrected for total protein, there was an approximately threefold decrease in Ral-GTP in the RalN28 sample compared with the empty vector (Figure 2). We next examined the effects of RalN28 expression on H-RasV12-induced thyroid epithelial colony formation.

Figure 2
figure 2

Coexpression of RalN28 in H-RasV12-expressing HT-Ori3 cells reduces Ral-GTP levels. HT-Ori3 cells were infected with retroviral vectors encoding H-RasV12 plus either empty vector or RalN28. Immunoblot of Ral-GTP pull down from lysates of pooled colonies (upper panel). Total protein content was estimated by India ink staining of membrane after separation of the whole lysate by SDS–PAGE and then transfer to PVDF (lower panel)

Full size image

Coinfection of primary thyrocytes with pLXSN H-RasV12 plus pMaRXpuroRalN28 resulted in a 2.35±0.23-fold (mean±s.e.) reduction in colony number when compared with H-RasV12 plus an empty vector (P<0.01).

Those RalN28 colonies that grew showed an altered morphology compared with Ras colonies 14 days after selection (Figure 3a) and were characterized by marked vacuolation with a ragged appearance and poorly defined colony edges. Consistent with this appearance, the colonies were also reduced in size on expression of RalN28. BrdU labelling index (LI) was assessed as a measure of proliferation and found to be reduced to 4.9±1.2% (n=10) in H-RasV12 plus RalN28-expressing colonies compared to an average LI of 13.6±1.1% (n=5) for H-RasV12 colonies (P<0.001), indicating that the expression of dominant-negative Ral blocks the proliferative response of thyroid cells to active Ras.

Figure 3
figure 3

Expression of RalN28 in thyroid cells causes morphological changes, reduced proliferation and decreased RalGTP levels compared with H-RasV12 colonies. (a) Primary thyroid epithelial cells were coinfected with H-RasV12 plus either empty vector (a, b) or RalN28 (c, d). At 14 days after selection, cells were incubated for 24 h with BrdU and then fixed and stained with an anti-BrdU antibody and DAPI. Cells were visualized and photographed for phase contrast (a, c) and fluorescence images (b, d) (Magnification × 100). (b) Pooled colonies from normal thyrocytes, H-RasV12 or Ras plus RalN28-infected cells were lysed and subjected to RalBD pull-down assay (upper panel). Total protein content was estimated by India ink staining of membrane after separation and transfer of whole lysate (lower panel). All bands were quantified by densitometry and corrected values for Ral-GTP are shown below (as arbitrary units)

Full size image

H-RasV12 plus RalN28 colonies were also assayed for Ral-GTP content compared with Ras colonies. Primary thyroid epithelial cells were infected as before with pLXSN H-RasV12 plus either pMaRXpuroRalN28 or pBabepuro empty vector. After 14 days, pooled colonies were lysed and subjected to the RalBD pull down, followed by Western analysis (Figure 3b). As expected, the Ral-GTP signal was increased in the Ras colonies compared to the normal thyrocytes (∼2-fold increase when corrected for total protein). Coexpression of RalN28 with H-RasV12 caused the Ral-GTP level to decrease to 1.5-fold above normal. This incomplete inhibition is likely to reflect the ability of the analysed colonies to proliferate, albeit more slowly, despite the expression of RalN28. We would predict that the effect of RalN28 was more complete in those clones that failed to grow, but that we were of course unable to analyse.

Activation of MAPK, PI3K and RalGEF pathways mimic the proliferogenic effects of Ras in human thyroid epithelial cells

Owing to the inherently low efficiency of infection of thyroid epithelial cells with multiple retroviruses, followed by multiple drug selection, we adopted the technique of scrape-loading recombinant protein into the cells in order to assess the combined effects of the three different Ras effector mutants.

At 2 days after plating, monolayers of thyroid epithelial cells were scrape loaded with 1 μg/μl of purified recombinant H-Ras mutant protein (H-RasV12, V12/S35, V12/C40 and V12/G37), the individual mutants alone or all three together, and then replated on poly-D-lysine-coated dishes. At 48 h after replating, cells were incubated with BrdU for 24 h, then fixed and stained to assess LI as an indication of proliferation (Figure 4).

Figure 4
figure 4

Expression of the three Ras effector mutants mimics the proliferative effect of H-RasV12 expression. At 2 days after plating, thyrocytes were scrape loaded with 1 μg/μl purified recombinant Ras V12 or combinations of effector mutants and reseeded onto poly-D-lysine-coated dishes. At 48 h after replating, cells were labelled with BrdU for 24 h and then fixed and stained to assess BrdU LI (%). The data shown are the means of two independent experiments

Full size image

Scrape loading of H-RasV12 caused an approximately ninefold increase in BrdU LI compared with an IgG scrape-loaded control. Individually, the three effector mutants are unable to reproduce the effect of H-RasV12, with scrape loading of the V12/G37 mutant producing less than 10% of the BrdU LI seen with H-RasV12. Scrape loading of V12/S35 on its own was able to induce some proliferation, consistent with retroviral data, where the construct is able to turn on additional Ras effectors due to the induction of an autocrine loop acting through endogenous Ras (Gire et al., 1999). We previously showed by retroviral coexpression of V12/S35 and V12/C40 that these effectors synergize to produce a greater proliferative effect on thyrocytes, though still not reaching that achieved with H-RasV12. This is not apparent by scrape loading of the two mutants, possibly due to the transient nature of these experiments. Scrape loading of V12/G37 with either V12/S35 or V12/C40 showed an increased BrdU LI compared with V12/G37 alone, but still lower than that seen with V12/S35 alone. However, scrape loading of thyrocytes with the three effector mutants together leads to a BrdU LI more than 80% of that seen with H-RasV12 (Figure 4).

These data suggest that while RalGEF activation is not alone sufficient to cause the proliferation of thyroid cells seen with the introduction of mutant Ras, it can complement the other effector pathways and the three together can reproduce the proliferative effects of H-RasV12, at least as assessed by a short-term assay.

Discussion

By using a combination of Ras effector mutants and mutants of the downstream effectors, together with selective inhibitors, we have previously shown that both PI3K and MAPK activity are required for Ras-induced proliferation of human thyroid epithelial cells in primary culture (Gire et al., 1999, 2000). However, although the two together produce a synergistic effect, proliferation falls far short of that seen with the expression of H-RasV12, indicating that activation of at least one other pathway is involved.

We have shown in this paper that H-RasV12 is able to activate RalGEF in primary human thyroid epithelial cells, leading to an increase in Ral-GTP levels as demonstrated by a RalBD pull-down assay. However, by using a Ras effector mutant selective for activating the RalGEF pathway (H-RasV12/G37), or an activated mutant RalGEF (Rlf-CAAX), we were unable to induce colony formation in primary thyrocytes demonstrating that, in these cells, RalGEF activation is not sufficient for colony formation.

Inhibition of the RalGEF pathway by overexpressing a dominant-negative RalN28 caused a marked reduction in the yield of colonies induced by H-RasV12. In those that did grow, morphological changes and reduced proliferative rate were consistent with evidence for incomplete inhibition of Ras-induced Ral activation. Taken together, these results strongly suggest that RalGEF activation is necessary for Ras-induced proliferation in thyrocytes.

The scrape loading of all three Ras effector mutants together (G37, C40 and S35 activating the RalGEF, PI3K and Raf/MAPK pathways, respectively; White et al., 1995) led to a proliferative effect approaching that seen with H-RasV12. This demonstrates that activation of the RalGEF pathway is able to complement the effect seen with Raf and PI3K activation to mimic the proliferation caused by H-RasV12 in thyroid epithelial cells.

The majority of work carried out to investigate the function of Ras in tumour formation has been carried out in immortalized fibroblasts or rodent epithelial cell lines, which can be transformed by the expression of Ras (al-Alawi et al., 1995; Khosravi-Far et al., 1996; Oldham et al., 1996; Graham et al., 1999). Yet, like the difference in tumour-forming ability in different species, the requirement for the activation of the different effector pathways in Ras-induced proliferation varies greatly both between species and different cell types in the same species. Typically, it is thought that activation of Raf and thus the MAPK pathway is the most important factor. For example, in NIH 3T3 cells, activation of Raf/MAPK pathway is sufficient to transform them (Cowley et al., 1994; Khosravi-Far et al., 1996), whereas expression of constitutively active RalGEF constructs are not transforming (Feig, 2003). However, in the rat fibroblast cell line, REF-52, MAPK activation was not sufficient for growth factor independence (Joneson et al., 1996). Furthermore, in a model closer to our own, WRT cells, activation of the Raf/MAPK, PI3K or RalGEF pathways individually were all found to be sufficient for growth factor-independent proliferation (Miller et al., 1998; Cass et al., 1999; Cass and Meinkoth, 2000), demonstrating redundancy in Ras effector pathways in these cells. For example, expression of V12/G37 in WRT cells led to an increase in DNA synthesis (Miller et al., 1997). However, in a closely related thyroid epithelial cell line, FRTL5, MAPK activation caused only a weak increase in proliferation (Cobellis et al., 1998), again illustrating the differences in Ras signalling, even within different cell lines generated from the same species and tissue.

Having shown that MAPK, PI3K and RalGEF activation were all required to reproduce the Ras-induced proliferation of thyrocytes (Gire et al., 1999, 2000, this paper), and with previous work also suggesting that multiple pathways were involved in Ras-induced transformation (White et al., 1995), the results of a recent paper from Hamad et al. (2002) were somewhat unexpected. Hamad et al. (2002) showed that selective activation of RalGEF in a variety of human cell lines (embryonic kidney, fibroblasts, mammary epithelial cells and astrocytes) was both necessary and sufficient to induce transformation, partially mimicking the effect of Ras on these cells. They also showed that activation of Raf or PI3K was not sufficient for transformation of human cells and that there was no apparent synergy between these two pathways. This is in stark contrast to our results in primary thyrocytes, where individually the Ras effectors are unable to cause proliferation (apart from Raf, due to the autocrine loop), Raf and PI3K are able to synergize to cause some proliferation (Gire et al., 2000), but activation of all three effectors together is required to reproduce the effect of Ras on proliferation. Ras-induced proliferation of thyrocytes in culture closely resembles the effect of Ras in thyroid tumorigenesis in vivo (Bond et al., 1994) with no further mutations necessary to cause proliferation. However, the cell types used in the study by Hamad et al. (2002) do not normally display Ras mutation as an initiating event in tumour formation (Bos, 1989). Indeed, hTERT and T-Ag were introduced in order to overcome the cell cycle arrest caused by the introduction of Ras into most primary human cells (Serrano et al., 1997). A further major difference between our study and that of Hamad et al. (2002) is that we are examining the purely proliferative response of thyrocytes to Ras, while Hamad et al. (2002) looked at transformation of human cell lines (anchorage-independent growth and tumorigenesis).

Several Ral effectors have been identified to date, including RalBP1, filamin, PLD1 and the Sec5 subunit of the exocyst complex (reviewed by Feig, 2003). RalBP1 contains a Cdc42 GAP domain, and is able to act as a GTPase activator protein for both Cdc42 and Rac in vitro (Jullien-Flores et al., 1995) and therefore this could provide a link between Ral and cytoskeletal rearrangement. This link can be strengthened by the ability of Ral to bind the actin filament crosslinking protein, filamin, a process that leads to filopodia formation and where Ral appears to operate downstream of Cdc42 (Ohta et al., 1999; Sugihara et al., 2002).

Dominant-negative constructs of Ral or Rlf were able to block EGF-induced chemokinesis in a wound healing assay using T24 human bladder cancer cells, demonstrating that Ral is a key player in this process. Dominant-negative Rho A was also able to block this process. Ral interacts directly with both the Sec 5 and Exo 84 subunits of the exocyst complex (Moskalenko et al., 2002; Moskalenko et al., 2003), and through this plays a role in exocytosis and targeting of proteins to the basolateral membrane of polarized epithelial cells. Similarly, Rho A binds to the Sec3 subunit of the exocyst (Guo et al., 2001), and this perhaps explains the ability of dominant-negative constructs of either Ral or Rho to inhibit wound closure.

Ral also works in conjunction with another small GTPase, ARF6 to activate PLD1 (Luo et al., 1998; Xu et al., 2003), which is important in vesicle formation, ER to Golgi transport (Ktistakis et al., 1996; Roth et al., 1999), and receptor-mediated endocytosis (Shen et al., 2001). ARF6 can be activated downstream of PI3K through GEFs, which bind to PtdIns(3,4,5)P3, thus providing a possibility of crosstalk between pathways downstream of Ras (Venkateswarlu et al., 1998; Cullen and Venkateswarlu, 1999).

The mechanisms through which RalGEF activation may contribute to Ras-induced proliferation however are unknown, although a recent paper (Chien and White, 2003) suggested, by using specific siRNA knockouts, that Ral was able to mediate both proliferogenic and antiapoptotic responses. Ral activation has been shown in some cellular contexts to lead to the downstream activation of several transcription factors, including Stat3 (Goi et al., 2000), Jun (de Ruiter et al., 2000), NF-κB (Henry et al., 2000), AFX (De Ruiter et al., 2001) and TCF (Wolthuis et al., 1997), and the stimulation of these transcription factors can be blocked by dominant-negative Ral. Many of these factors are known to be involved in cell cycle stimulation; however, no Ral effectors have been identified as yet as intermediates between active Ral and these transcription factors, making the precise role of Ral in their activation difficult to define.

Further work is clearly required to identify the final targets downstream of Ras effector activation and elucidate their individual roles in the proliferation of thyrocytes in response to mutant Ras. Our finding that three effector pathways are required for Ras-induced proliferation in thyrocytes indicates a more robust control of cell proliferation in this cell type, compared with previous similar studies using rodent and human cell lines. This work also highlights the importance of appropriate cell context in models of Ras-induced tumour development.

Materials and methods

Cell culture

HT-Ori3 cells (Lemoine et al., 1989a) and monolayer cultures (>99% epithelial as judged by cytokeratin immunostaining) prepared from surgical samples of normal thyroid tissue (Williams et al., 1988) were maintained in a 2 : 1 : 1 mixture of Dulbecco's modified Eagles' medium, Hams' F12 and MCDB104 (Bond et al., 1994), supplemented with 10% foetal calf serum (Imperial Laboratories, London, UK).

Mammalian expression constructs

A retroviral vector encoding the Val12 mutant of human H-Ras (psi-Crip-DOEJ) was described previously (Bond et al., 1994). The Ras effector mutant, H-RasV12/G37, was provided by J Downward (LRI, London, UK) and then subcloned into pBabepuro (Morgenstern and Land, 1990). Rlf-CAAX and RalN28 were supplied by J Bos (Utrecht) and subcloned into pBabepuro and pMaRXpuro (Hannon et al., 1999), respectively.

Retroviral gene transfer

Primary cells or cell lines were plated at ∼5 × 105 cells/60 mm dish and infected 2 days later with retrovirus-containing medium from near-confluent producer cells, containing 8 μg/ml polybrene (Bond et al., 1994). After 3 days, cells were passaged into selective medium as indicated.

Preparation of recombinant proteins

Ras proteins were expressed from pGex vectors in Escherichia coli strain BL21DE3 (LysE) (pGex2T-V12-H-Ras (A Hall, LMCB, London, UK); pGex4T2-Ha-Ras V12S35, V12G37 and V12C40 (M White, UTSouthwestern, USA)) and purified by affinity chromatography on glutathione–sepharose (Pharmacia) as described previously (Gire et al., 1999). The final products were >90% pure as estimated by Coomassie blue staining following SDS–PAGE.

Scrape loading

Primary cultures were seeded at ∼5 × 105 cells/60 mm dish and 2 days later the medium was replaced with 150 μl buffer (10 mM Tris (pH 7.4), 114 mM KCl, 15 mM NaCl, 5.5 mM MgCl2) containing IgG or recombinant Ras protein. Cells were detached by gentle scraping with a ‘rubber policeman’ (Leevers and Marshall, 1992), and after 1 min, cells were reseeded onto poly-D-lysine-coated dishes (Sigma) in complete medium.

BrdU labelling

Cells were incubated with 10 μ M BrdU, fixed and permeabilized as described previously (Gire and Wynford-Thomas, 1998). The percentage of BrdU-labelled cells (LI) was determined for at least 250 cells per data point and expressed as a mean and standard error of triplicate determinations.

Preparation of cell lysates

Cells were scraped off subconfluent dishes using a ‘rubber policeman’ ice-cold STE buffer (10 mM Tris (pH 8), 1 mM EDTA, 150 mM NaCl). After pelleting, cells were lysed on ice in 50 mM Tris, 5 mM EDTA, 150 mM NaCl, 1% NP-40, 1 mM PMSF, 10 μg/ml aprotinin, 10 μg/ml leupeptin and then cleared by centrifugation. Total protein concentration was quantified using the Coomassie Plus protein assay kit.

Purification of GST-RalBD

BL21 bacteria were transformed with a vector encoding the GST-tagged Ral-GTP binding domain of RalBP1 (pGex4T3-RalBD, kind gift of J Bos) and induced with 0.2 mM IPTG (8 h at 30°C). Cleared bacterial lysates were incubated with glutathione-sepharose beads (Pharmacia) at 4°C for 60 min and then washed. GST-RalBD protein was eluted in 50 mM Tris (pH 7.4), 10% glycerol, 150 mM NaCl, 5 mM MgCl2, 5 mM DTT, 15 mM glutathione, dialysed and concentrated. The purity and concentration of the GST-RalBD protein was determined by the SDS–PAGE and using the Coomassie Plus protein assay kit.

Ral-GTP binding assay

This assay was carried out essentially as described by Wolthuis et al. (1998). Glutathione–sepharose beads were equilibrated with 50 mM Tris (pH 7.4), 200 mM NaCl, 2.5 mM MgCl2, 1% NP-40, 15% glycerol, 1 mM PMSF, 1 μ M leupeptin, 0.1 μ M aprotinin and 10 μg/ml soybean trypsin inhibitor. Purified GST-RalBD protein was added to the beads and incubated with agitation for 45 min at 4°C. After washing, total cell lysate (100–1000 μg, prepared as mentioned above) was added to the GST-RalBD-coupled beads. Samples were incubated with agitation at 4°C for 1 h and then washed thoroughly. Bound protein (containing RalGTP) was eluted by incubation in SDS–PAGE loading buffer for 5 min at 95°C. RalGTP pull down was determined by Western blotting with an anti-RalA monoclonal antibody (Transduction Laboratories, 50 ng/ml), detected with an HRP-conjugated secondary antibody and then visualized by ECL (Amersham). In parallel, unfractionated total cell lysates were subject to PAGE and Western blotting as above to determine total Ral protein content.