Oncogene (1999) 18, 4798 ± 4807 ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00 http://www.stockton-press.co.uk/onc Contrasting e€ects of activating mutations of GaS and the thyrotropin receptor on proliferation and di€erentiation of thyroid follicular cells

M Ludgate1, V Gire1, M Crisp1, R Ajjan2, A Weetman2, M Ivan1 and D Wynford-Thomas*,1

1Cancer Research Campaign Laboratories, Department of Pathology, University of Wales College of Medicine, Cardi€ CF4 4XN, UK; 2Department of Medicine, University of Sheeld, Clinical Sciences Centre, Northern General Hospital, Sheeld, UK

The cyclic AMP pathway is a major regulator of To date, two mutations in Gas (GSP) have been thyrocyte function and proliferation and, predictably, identi®ed, Q227L and R201H, which activate the its inappropriate activation is associated with a sub-set of protein by `locking' it in its GTP-bound signalling human thyroid tumours. Activating mutations are, conformation. In contrast, ligand-independent activa- however, more common in the thyrotropin receptor tion of TSHR can occur through point mutation at (TSHR) than in its downstream transducer, Gas. To more than 20 di€erent sites, the majority clustered in investigate whether this re¯ects an inherent di€erence in its trans-membrane and intra-cellular domains (re- their oncogenic potency, we compared the e€ects of viewed in Paschke et al., 1994). In thyroid neoplasia, retrovirally-transduced mutant (A623I) TSHR or point mutations are found, as would be predicted, (Q227L) Gas (GSP), using the rat thyroid cell line mainly in the sub-set of benign tumours which display FRTL5 and primary human thyrocytes. In FRTL5, both increased growth and hyperfunction, i.e. autono- expression of GSP or mutant (m) TSHR induced a 2 ± 3- mously functioning `toxic adenomas' or `hot nodules'. fold increase in basal levels of cAMP. This was More rarely they can also occur in cancers (Said et al., associated with TSH-independent proliferation (assessed 1994; Russo et al., 1995a; Paschke and Ludgate 1997) by both cell number and DNA synthesis) and function (as which have elevated cAMP levels but, due presumably shown by increased expression of thyroglobulin (Tg) and to loss of di€erentiation, do not exhibit increased the sodium/iodide symporter). In primary cultures, thyroid hormone secretion. expression of mTSHR, but not GSP, consistently Interestingly, however, in all thyroid tumour types induced formation of colonies with epithelial morphology the prevalence of mutation of GSP appears to be much and thyroglobulin expression, capable of 10 ± 15 popula- lower than that of TSHR despite the fact that both tion doublings (PD) compared to less than three in might be expected to have the same ability to activate controls. Thus, while mTSHR and GSP exert similar the cAMP pathway. In four series totalling 114 cases of e€ects in FRTL5, use of primary cultures reveals a `hot nodule' (O'Sullivan et al., 1991; Russo et al., major di€erence in their ability to induce sustained 1995b, 1996; Parma et al., 1997) the frequency of GSP proliferation in normal human thyrocytes, and provides mutation varied from 0% (Fuhrer et al., 1997) to 38% the ®rst direct evidence that mTSHR is sucient to (O'Sullivan et al., 1991) with a weighted mean of 14%. initiate thyroid tumorigenesis. In contrast, TSHR mutation varied from 20% (Russo et al., 1996) to 82% (Parma et al., 1997) with a mean Keywords: TSH receptor; G protein; cyclic AMP; of 47%. Since this is probably an underestimate, given oncogene; thyroid the greater potential for missing mutations in TSHR, the latter must be at least 3 ± 4-fold more common than mutation of GSP. This di€erence in prevalence may of course itself Introduction merely re¯ect the greater range of functionally signi®cant targets for mutation in TSHR (at least 20 In physiological conditions, growth and function of codons) compared to GSP (only two). More interest- thyroid follicular epithelial cells are regulated princi- ingly, though, it could alternatively be due to an pally by binding of thyrotropin (TSH) to its receptor inherently greater biological potency of mTSHR as an (TSHR) which then activates adenylate cyclase and the oncogene, which may result from its ability to activate cAMP cascade via the hetero-trimeric transducer signal transduction pathways additional to cAMP; it is protein Gs (Vassart and Dumont, 1992). In a variety already known for example that at least some mutants of inherited and acquired pathological states, thyroid stimulate phospholipase C and the inositol phosphate hyperfunction and proliferation are associated with cascade (Parma et al., 1995). This issue has not been activating mutations in elements of this signal pathway, adequately addressed. notably TSHR itself (Van Sande et al., 1995; Russo et Most signalling studies in this ®eld have been al., 1996; Parma et al., 1997; Fuhrer et al., 1997) and performed, for ease of experimental manipulation, by the a sub-unit of the Gs protein (O'Sullivan et al., expression of mutant genes in non-thyroid rodent cell 1991; Russo et al., 1995a,b; Parma et al., 1997; Fuhrer lines (Parma et al., 1993; Paschke et al., 1994) which et al., 1997). are inappropriate for assessing their relative prolifero- genic e€ects on thyroid cells. Even where cells of thyroid origin have been used, this has nearly always been the rat thyrocyte cell line FRTL5. In the case of *Correspondence: D Wynford-Thomas Received 21 September 1998; revised 11 March 1999; accepted 11 GSP, expression of the Q227L mutant conferred March 1999 complete TSH independence for DNA synthesis GSP and TSHR mutants in thyrocyte function and proliferation MLudgateet al 4799 (Muca and Vallar, 1994) but, interestingly, required the e€ects of GSP and mutant TSHR on growth and addition of forskolin to achieve the same increase in function in FRTL5 cells, using a phenotypically neutral cell number as induced by TSH in control cells. Similar method of clonal selection, and furthermore, to results were obtained using the PDE inhibitor, compare their biological e€ects on human primary rolipram (Nemoz et al., 1995). TSH-independent thyrocytes. proliferation has also been seen with mTSHR, the extent varying in one study with the nature of the mutant (Porcellini et al., 1997), although it was not Results clear whether even the most potent construct (T632I) reached the level of TSH stimulated controls. In E€ect of GSP, wild-type and mutant TSHR expression another study, comparing the M453T mutant with on FRTL5 cells GSP, only the mutant receptor was able to induce clones with anchorage-independence and tumorigeni- Selection of clones following retroviral infection Mono- city in nude mice (Fournes et al., 1998), providing a layers of FRTL-5 (SB5 sub-clone; Burns et al., 1992) ®rst hint that mTSHR may be a more potent were infected with retroviral vectors encoding: (i) oncogene. mutant GSP (rat Q227L); (ii) mutant TSHR (human Published FRTL5 studies all su€er from limitations A623I); (iii) wild-type (wt) TSHR or (iv) neo-only however. Some were performed on pools of clones (control) genes. From each infection, several hundred (Porcellini et al., 1997) which, given the heterogeneity G418-resistant clones were obtained, from which six of the response expected in gene transfer experiments, were picked at random for preliminary analysis (data can lead to information being lost due either to over- not shown). The GSP clones were clearly very growth of more vigorously growing clones or by a heterogeneous in behaviour. In the absence of TSH, dilution e€ect from clones with low expression of the two had growth rates comparable to that of neo transgene. Others involved selection of individual controls in the presence of TSH, while the remaining clones based on their ability to overgrow the back- four showed varying degrees of TSH dependence, two ground culture in the absence of TSH (Fournes et al., of which also displayed increased cellular size (area) 1998), which introduces a strong bias for the most and altered cytoplasmic morphology. Three clones proliferative phenotype and risks masking di€erences were selected to represent this range of phenotypes: in potency between GSP and TSHR. Furthermore, gsp6 (TSH-independent); gsp1 (partially TSH-depen- most previous studies have examined only e€ects on dent); and gsp8 (partially TSH-dependent plus proliferation rather than di€erentiated thyrocyte morphological change). The six mTSHR clones were functions. Finally, in FRTL5, TSH has little e€ect on similarly heterogeneous, and again, three representative phospholipase C unless highly unphysiological concen- clones were selected for detailed study: MT1 (TSH- trations of hormone are used (Vassart and Dumont, independent); MT2 (partially TSH-dependent) and 1992), indicating that these cells may not be a good MT3 (partially TSH-dependent with morphological model system to investigate phenotypic responses change). The six wtTSHR clones were much more dependent on this pathway. homogeneous. None proliferated in the absence of Primary human cells are obviously the ideal TSH and there were no morphological changes. Three experimental model to test the ability of a putative were therefore selected at random: WT1, WT4 and oncogene to initiate tumour development and we have WT6 (in some experiments only the last two were successfully employed retrovirus vectors to demon- included). The neo-only clones were also homogeneous strate this for two other thyroid oncogenes, RAS and and so to limit the total number of clones which RET (Bond et al., 1994). However, in a recent study needed to be analysed in later experiments, these were (Ivan et al., 1997) using a retroviral vector expressing pooled to form a single control population (neo). Q227L GSP, we found that despite performing infections in a variety of conditions in which the E€ects on proliferation positive control ± mutant RAS ± generated large numbers of colonies, expression of GSP in human Cell number (Figure 1a and b; Table 1) Control (neo) thyrocytes in primary monolayer cultures induced only cells in the absence of TSH showed a consistent initial morphological responses (cytoskeletal changes and reduction of *40% in cell number at day 2 (due cellular hypertrophy), with very little proliferation. presumably to failure of some cells to attach). These in vitro ®ndings are in general in agreement with Thereafter there was a slow recovery, returning by in vivo studies in mice transgenic for the R201H GSP day 7 to the number seeded. In the presence of TSH mutant (Michiels et al., 1994) under the control of the (5 mU/ml) there was no initial cell loss, and near- bovine thyroglobulin promoter. These mice developed exponential growth leading to a 6 ± 7-fold increase by only late-onset sporadic thyroid nodules, rather than day 7. the expected generalized hyperplasia, suggesting either In the absence of TSH, the wtTSHR clones behaved that only a tiny sub-population of follicular cells are almost identically to the neo population. In contrast, sensitive to this growth stimulus, or more likely that both the GSP and mTSHR clones all showed some there is a need for a second oncogenic event. degree of TSH-independence. Although all but two Taken together these data suggest that GSP (gsp6 and 8) su€ered the initial cell loss, this was mutation alone may be insucient to account for the followed by a period of exponential growth with formation of a thyroid tumour and that the observed doubling times similar to that of TSH-stimulated neo predominance of TSHR mutations may indeed re¯ect cells (with the exception of gsp8). Cell numbers its greater biological potency. The purposes of this attained by day 7 were therefore intermediate between study were therefore to compare in more detail the those seen in neo controls in the absence and presence GSP and TSHR mutants in thyrocyte function and proliferation MLudgateet al 4800 of TSH. The highest values, displayed by clones gsp6 that no increase in cell number occurred during this and MT1, were not signi®cantly lower than that of neo time and presumably re¯ects either ongoing cell death plus TSH. or DNA synthesis without cell division (endoredupli- In the presence of TSH (Figure 1b), the initial loss in cation). In the presence of TSH, the BrdU LI increased cell number was less marked but interestingly nearly all to *70%. GSP, wt and mTSHR clones still showed lower In the absence of TSH, clones expressing wt TSHR numbers than neo controls at day 2. Thereafter, had an LI very close to that of the neo control. In though, all clones grew at rates similar to or greater contrast, all the GSP and mTSHR clones showed much than neo, and reached larger ®nal cell numbers at day higher values, in two cases (gsp6 and MT1), reaching 7, the highest values being seen in the three mTSHR levels not signi®cantly lower than that seen in neo in clones. the presence of TSH.

DNA synthesis (Table 1) In the absence of TSH, E€ect on tissue-speci®c di€erentiation *30% of the control (neo) cells incorporated BrdU during a 24 h labelling period corresponding to days Morphology In the absence of TSH, control neo cells 3 ± 4 of the growth curves. This was surprising given exhibited a ¯attened morphology which was restored to

a b

Figure 1 (a) Growth curves of FRTL5 (SB5) clones in the absence of TSH. Individual GSP, wtTSHR (WT) or mTSHR (MT) clones were seeded at 56104 per dish and counted on days 2, 4, 7 and 11. For comparison, results with the pooled neo population are also shown in the presence or absence of TSH (5 mU/ml); (b) Growth curves in the presence of TSH (legend as for (a)). Note logarithmic ordinate scale; s.e. values omitted for clarity

Table 1 E€ects of gsp, wt and mTSHR expression on proliferation and function of FRTL5 (SB5) cells compared to the e€ect of TSH stimulation Proliferation Function Cell noe Cell Noe BrdUf NISh NISh Clonea TSH cAMPb,c DTd (h) Day 2 Day 7 LI Tgg 7IBMX +IBMX Neo 7 100 70 64+5.8 114+8.2 31+2.0 38+0.8 4.0+0.12 320+13 Neo + 230+8 43 132+12 614+76 72+2.5 51+1.5 304+9.2 120+2.9 WT1 7 160+8 ND ND ND 29+1.5 ND 4.0+0.3 130+3.5 WT4 7 80+9 92 58+6.4 110+19 30+2.0 62+2.5 1.0+0.12 12+1.2 WT6 7 110+10 46 42+7.0 100+6.6 28+2.0 59+2.0 1.0+0.12 60+3.5 MT1 7 285+8 30 58+7.4 456+30 70+4.5 73+3.5 6.0+0.12 60+2.9 MT2 7 230+14 48 36+5.4 154+11 51+3.0 66+3.0 29+2.3 230+9.2 MT3 7 330+13 38 44+2.8 232+20 50+3.0 70+4.0 13+1.7 310+9.2 gsp1 7 220+9 45 56+9.4 212+24 47+3.5 48+8.0 15+1.2 200+4.6 gsp6 7 300+17 35 96+2.0 472+40 65+5.5 70+6.0 140+10 17+2.9 gsp8 240+13 114 142+2.4 276+46 50+4.0 63+5.0 16+0.58 160+4.0 a`Neo' are pooled clones. bAll results are expressed as mean of three or four replicates+s.e. cPer cent of basal level in neo controls. dDoubling time calculated from steepest part of growth curve. ePer cent of number seeded at day 0. fPer cent of nuclei labelled after 24 h incubation with BrdU. gPer cent of cells positive by immuno¯uorescence. hRatio of NIS: b-actin transcript abundance (6102) GSP and TSHR mutants in thyrocyte function and proliferation MLudgateet al 4801 the normal, more cuboidal, shape within 24 ± 48 h of inhibitor, IBMX, were largely as expected with an addition of TSH. In the absence of TSH, the wtTSHR increase in the NIS transcript in the majority of cases clones closely resembled the neo controls, but GSP (Figure 4b; Table 1). The exceptions were the two clones were less ¯attened and the mTSHR clones conditions showing highest levels in the absence of resembled TSH-stimulated controls. IBMX, i.e. gsp6 and neo plus TSH, in which a During the period of initial selection in G418, it was paradoxical reduction of NIS transcript levels was noticed that several groups of cells became enlarged observed. and ¯attened and displayed ®brillar changes in the cytoplasm reminiscent of `stress ®bres'. As selection E€ects of GSP, wt and mutant TSHR expression on proceeded, these changes disappeared in most clones, primary human thyrocytes but remained a prominent feature in gsp8 and MT3 (Figure 2). Primary thyrocytes in monolayer culture were infected with retroviral vectors either in 10% FCS or using a Basal cAMP The basal level of cAMP production in reduced-serum medium (1% FCS, supplemented with control neo cells in the absence of TSH was insulin and TSH) in an attempt to avoid the potential *30 fmoles/mg protein. Only one of the wtTSHR de-di€erentiating e€ect of serum on thyroid cells. In clones (WT1) had a signi®cantly higher level (160% of both cases, a well-characterized retroviral vector control; P50.01) (Table 1). In contrast, all of the encoding mutant H-RAS (Bond et al., 1994) was used GSP and mTSHR clones had basal levels ranging as a positive control. from 220 ± 330% of the neo control and similar or greater than those seen in neo cells in the presence of Colony formation in 10% FCS (Figure 5) In 10% TSH. FCS, the H-RAS vector generated approximately 20 well-di€erentiated epithelial colonies per dish of Thyroglobulin expression In the absence of TSH, 56105 cells infected using thyrocytes from three out weak cytoplasmic immuno¯uorescence was observed of four fresh and three out of eight stored samples with anti-Tg antibody in *40% of control neo cells, (from 12 independent subjects). Most colonies were which increased to *50% in the presence of TSH, able to proliferate for at least 20 PD. The neo-only accompanied by a marked increase in intensity (Figure vector as expected failed to generate any colonies, 3). In contrast, in all TSHR or GSP expressing clones, consistent with the very limited proliferative capacity even in the absence of TSH, the percentage of Tg- of normal human thyrocytes (Wynford-Thomas, positive cells was similar to or higher than that seen in 1993). GSP also failed to induce any colony controls in the presence of TSH (Table 1). Surprisingly, formation in 12 out of the 13 thyrocyte preparations this also included the two wtTSHR clones, whose infected and in the remaining case very few colonies cAMP levels were not signi®cantly greater than the were obtained. control values. The highest percentage of positive cells In contrast, the mTSHR vector generated 10 ± 15 however was observed in the GSP and mTSHR clones, colonies of epithelial morphology per dish from four which also showed more intense ¯uorescence than the out of ®ve fresh and two out of eight stored thyrocyte wtTSHR cells (Figure 3). samples. Typically, colonies grew for several weeks and reached a maximum size of 1000 ± 30 000 cells, Expression of the Na+/I7 symporter gene (Figure equivalent to 10 ± 15 PD (assuming no cell death). 4) In the absence of TSH, transcripts were often Cessation of growth was associated with alterations in below the limit of detection by RT ± PCR in control cellular morphology, characterized by spreading, neo and wtTSHR clones. In contrast, detectable levels ¯attening and appearance of cytoplasmic stress ®bres were present in all GSP and mTSHR clones. The with occasional pseudopodia (Figure 5). Similar results highest levels were seen in MT2 and gsp6, the latter were obtained when thyrocytes derived from the same approaching that of TSH-stimulated neo controls thyroids were infected with the wtTSHR vector, (Table 1). Results in the presence of the PDE although with a lower yield (2 ± 5 colonies per dish).

Figure 2 Phase contrast micrographs showing altered cellular morphology in FRTL5 clone gsp8 (2) compared to neo controls in the presence of TSH (1). (6100) GSP and TSHR mutants in thyrocyte function and proliferation MLudgateet al 4802

Figure 3 Thyroglobulin expression in FRTL5 detected by immuno¯uorescence (rhodamine label). All in the absence of TSH unless otherwise indicated. (1) negative control (neo with ®rst antibody omitted); (2) neo; (3) neo+TSH; (4) clone WT 4; (5) clone MT1; (6) clone gsp6. (6150)

It must be emphasized that these di€erences in yield and cytoskeletal changes. (Similar results were of colonies are not explicable by di€erences in the titre obtained irrespective of the inclusion of TSH at 0.03 of the corresponding vectors, at least as far as can be or 5 mU/ml up to 1 day following retroviral assessed using epithelial cell lines (FRTL5 and A431) infection.) Infection of stored follicles in 1% FCS as targets. was unsuccessful in all but one case due to poor cell attachment. Colony formation in 1% FCS In 1% FCS colonies were obtained from two out of ®ve freshly Expression of thyroglobulin To con®rm the cellular disaggregated thyrocytes with a similar yield to that identity of the colonies obtained from primary culture, seen with 10% FCS, i.e. approximately 20 H-RAS, expression of Tg was analysed by immuno¯uorescence. 10 mTSHR and two wtTSHR colonies per dish of In all colonies obtained with wt or mTSHR 56105 infected. Proliferative capacity however was cytoplasmic positivity was observed in the majority of greatly reduced, being limited to no more than ®ve cells (Figure 6). As observed with FRTL5, the intensity PD, even in the RAS colonies. In the case of TSHR of immunostaining was greater with mTSHR than with this was accompanied by an early onset of wtTSHR as was the percentage of positive cells morphological alterations including cellular ¯attening (74+4.5% and 51+4% respectively). GSP and TSHR mutants in thyrocyte function and proliferation MLudgateet al 4803

Figure 4 Duplex RT ± PCR analysis of sodium/iodide symporter (NIS) and b-actin transcripts in representative FRTL5 clones expressing mutant GSP, mutant TSHR or wild-type TSHR, in the absence (a) or presence (b)of2mM IBMX (all in the absence of TSH). A pooled control population (neo) in the presence or absence of TSH is shown for comparison. Note that the ®gure shows just one of three replicate analyses, and therefore does not match exactly the means shown in Table 1

Discussion

Activated GSP and TSHR both confer TSH-independent growth and function on FRTL5 cells As expected, in FRTL5 (SB5) cells, expression of either activated Gas or TSHR was able to stimulate cAMP production and thyroglobulin expression, to levels similar to that induced by TSH. For Tg, a signi®cant increase was also induced by over-expression of wtTSHR, re¯ecting its limited constitutive activity (Van Sande et al., 1995; Parma et al., 1993; Duprez et al., 1994) but interestingly this was not associated with any increase in cAMP. We also investigated expression of another recently cloned thyroid-speci®c gene ± the Na+/I7 symporter (Dai et al., 1996) ± although the non-availability of antibodies restricted this to assessment of transcript abundance. Increased expression was seen in some GSP and mTSHR clones but to a more limited extent than for Tg, remaining well below the TSH-stimulated value in all but one clone. Since the corresponding cAMP levels were higher than those of TSH-stimulated controls this Figure 5 Phase contrast micrographs of human primary indicates the need for additional signal pathways. thyrocytes expressing mutant H-RAS or mTSHR in 10% FCS. (1) Typical epithelial colony induced by mutant RAS; (2) typical Furthermore, the inability of mTSHR to mimic early mTSHR-induced epithelial colony; (3) mTSHR colony extracellular TSH stimulation (the only example of following several weeks in culture, showing cellular enlargement, this in our study) points to the involvement of ¯attening and prominent cytoskeletal striations. (6200) pathway(s) whose stimulation by ligand binding must be quantitatively or qualitatively di€erent from that following mutational activation. Nevertheless, NIS transcript levels approaching that induced by TSH GSP and mTSHR (but not wtTSHR) also conferred could be achieved by inclusion of IBMX, implying that TSH-independent proliferation, as assessed both in cAMP can be sucient at supra-physiological levels. terms of DNA synthesis and increase in cell number. Interestingly though, the paradoxical fall in NIS on Large variations were observed within each group, only addition of IBMX in conditions where it was already one GSP and one mTSHR clone approaching complete highly expressed (gsp6 and neo plus TSH) indicates TSH-independence. It is unlikely that this relates to that the optimal level of cAMP may be quite critical. di€erences in expression of the mutant genes since (in Excessive levels are perhaps inhibitory through contrast to one previous study (Porcellini et al., 1997)) activation of ICER, an alternatively spliced form of there was no consistent relationship to cyclic AMP the CREM transcription factor, which has already levels. Expression was not measured directly here been demonstrated in the FRTL5 cell line (Lalli and because basal cAMP was considered a more mean- Sassonecorsi, 1995). ingful biological endpoint, and the only way to GSP and TSHR mutants in thyrocyte function and proliferation MLudgateet al 4804

Figure 6 Thyroglobulin expression in colonies of human thyrocytes derived by infection with vectors expressing: (1) wtTSHR; and (2) mTSHR. Rhodamine immuno¯uorescence (6150)

e€ectively compare between GSP and mTSHR e€ects. over-expression of the wtTSHR. Presumably, as was The important ®nding for this study is that despite the seen with Tg expression in FRTL5, this re¯ects its inter-clonal variation, there was no signi®cant differ- weak but signi®cant constitutive activity (Van Sande et ence in the mean cAMP level or TSH-independent al., 1995) although, if so, it implies that the normal proliferative response between the GSP and TSHR cells are more responsive than FRTL5, since no TSH- clones, the respective basal cAMP levels (% of neo independent growth was observed with wtTSHR in the control) being 281+29% and 253+24%, with latter. Unfortunately, in the primary culture model, it corresponding BrdU LI values of 57+5.6% and was not possible to further investigate this by 54+6.5% (data derived from Table 1). measurement of receptor expression or cAMP levels. Similar TSH-independent stimulation of cAMP and proliferation have been reported separately both for Constitutive TSHR activity provides a signal not GSP (Muca and Vallar, 1994) and a range of TSHR mimicked by GSP or physiological TSH stimulation mutants (Porcellini et al., 1997) but few functional parameters were examined. The only previous study to The two major questions arising from the results with attempt a direct comparison between GSP and TSHR primary thyrocytes are: ®rstly, why GSP does not clones obtained from the same parent population produce an e€ect equivalent to that of mTSHR and (Fournes et al., 1998) concluded that GSP exerted a secondly, why is the latter so much more potent as a weaker e€ect than mTSHR (M453T) in terms of cAMP growth stimulus than stimulation by extracellular TSH. stimulation and doubling time in monolayer. However One potentially artefactual explanation for the the former result was obtained in the presence of ine€ectiveness of GSP is that forced expression of a IBMX, and the latter may have been in¯uenced by the mutant Gas sub-unit does not mimic exactly the method of clonal selection (TSH-independent prolifera- naturally occurring mutation in the endogenous gene tion rather than `neutral' selection in G418). because it does not lead to a corresponding release of On the basis of behaviour of FRTL5 cells in free b/g sub-units. The latter can act in their own right monolayer culture, the concensus from both the as signal transducers, interacting for example with PI- present and previous work is therefore that there is 3-kinase and tyrosine kinases (Van Biesen et al., 1996), little or no evidence for di€erence in the biological and could theoretically be required for the mitogenic potencies of activated GSP and TSHR on which to response in primary cells. However, this would not explain their di€erent mutational frequencies in thyroid explain the lack of sustained proliferation following tumours. administration of agents, notably cholera toxin (Wynford-Thomas, 1993) which activate endogenous Gs and closely mimic the e€ect of Gas mutation. GSP and mTSHR exert very di€erent e€ects on normal It seems most likely therefore that the essential signals human thyrocytes missing in the case of GSP are derived from a pathway In contrast to the above, our data on primary cells which diverges upstream of Gs, i.e. at the receptor itself. reveals a massive di€erence in the ability of mTSHR It is known that TSHR activates pathways additional to and GSP to induce sustained proliferation in normal cAMP, notably the phospho-inositide pathway (via Gq thyrocytes. Unlike the FRTL5 line, these cells are and phospholipase-Cb) (Vassart and Dumont, 1992) normally capable of only a few cell divisions in culture, and at least in one cell line model (Kupperman et al., even in the presence of TSH (Dumont et al., 1992; 1993) there is evidence that cross-talk with RAS is Wynford-Thomas, 1993). Whereas expression of GSP involved in the mitogenic response. What is puzzling nearly always failed to stimulate proliferation, expres- though is how expression of mTSHR can lead to sion of mTSHR consistently generated colonies of Tg- sustained proliferation, whereas stimulation of normal positive epithelial cells which were able to proliferate thyrocytes by extra-cellular TSH does not lead to more for up to 15 PD Interestingly, although fewer in than two or three PD either in vitro (Wynford-Thomas, number, similar colonies could also be obtained by 1993) or in the intact animal (Wynford-Thomas et al., GSP and TSHR mutants in thyrocyte function and proliferation MLudgateet al 4805 1982). One plausible explanation is that constitutive, previously (Bond et al., 1994; Ivan et al., 1997). To construct ligand-independent, receptor signalling may not invoke TSHR vectors, cDNAs for the entire coding sequences of the receptor downregulation which is a well recognized wild-type (wt) and A623I mutant receptor (kindly provided response to ligand-mediated activation (Rapoport et al., by Dr G Vassart, IRIBHN, Brussels) were subcloned into 1982; Lalli and Sassonecorsi, 1995). (Use of a plasmid pLNSX, which utilizes the SV40 early promoter for the gene of interest. (Attempts to clone mTSHR cDNA into heterologous promoter to drive TSHR in this and LTR-driven vectors were, for unknown reasons, unsuccessful. previous studies would, of course, circumvent the However, previous comparison of the two promoters driving transcriptional component of down-regulation (Lalli SV40T expression in primary thyrocytes did not reveal any and Sassonecorsi, 1995) but would not prevent receptor- detectable di€erences in strength ± Bond JA, unpublished.) G protein uncoupling (Rapoport et al., 1982)). The resulting constructs, and the pLNSX control were Although in FRTL5, down-regulation is also known transfected into the OE ecotropic packaging line and to occur (Lalli and Sassonecorsi, 1995), its negative retroviral supernatants from pooled G418 resistant clones in¯uence must be of insucient magnitude to prevent a subsequently used to infect the amphotropic packaging line, sustained proliferative response to TSH. Intriguingly Ccrip (Danos and Mulligan 1988). Supernatants from a panel though, di€erential sensitivity of ligand-dependent and of G418 resistant Ccrip producer clones were evaluated using the human epithelial cell line A431 as a recipient. Viral titre -independent stimulation to downregulatory control was assessed in terms of the yield of G418 resistant colonies could provide an explanation for the observation and, in the case of the TSHR vectors, stable expression was (Fournes et al., 1998) that anchorage-independent determined by measurement of speci®c [125I]TSH binding opposed to monolayer growth of FRTL5 was (Costagliola et al., 1994). The optimum producer clones, all strikingly stimulated by mTSHR, but not by TSH- with titres 4106 c.f.u./ml, were designated MLWT1 and stimulation of the wt receptor (nor by GSP). MLAI1 for the wt and A623I mutant TSHR respectively, and MLSX1 for the LNSX neo-only control. TSHR activation as an initiator of thyroid tumorigenesis Cells and culture conditions Irrespective of the above issues, our data show for the ®rst time that mTSHR is capable of driving clonal The SB5 subclone (Burns et al., 1992) of FRTL5, selected for expansion of normal human thyrocytes, providing its TSH dependency as previously described, which are direct experimental support for its role as an initiating referred to as `FRTL5' throughout. Cells were routinely cultured in a 2 : 1 : 1 mixture of DMEM, Ham's F12 and event in thyroid tumorigenesis, analogous to that MCDB 104 supplemented with 5% calf serum (Life already demonstrated by us for two other oncogenes Technologies, Paisley, UK), 5 mU/ml bovine TSH (Sigma; ± RAS and RET (Bond et al., 1994). As with the latter, UK), 10 mg/ml insulin, 1078 M hydrocortisone, 45 mg/ml proliferation eventually ceased, although after fewer ascorbic acid and 5 mg/ml transferrin (`5H medium'). For PD, following which cells enter a state of growth arrest experiments in basal conditions, cells were ®rst maintained in with morphological changes reminiscent of replicative the same medium but lacking TSH (`4H medium') for at least senescence (Wynford-Thomas, 1997a). The mechanism 4 days. of this intrinsic limitation of clonal expansion is Human primary thyrocytes were obtained by protease currently unknown although preliminary data suggest digestion of histologically-normal background tissue obtained that, at least in the case of RAS, increased expression of from thyroidectomies performed for solitary nodules as previously described (Williams et al., 1988). Monolayer INK4a the cyclin/CDK inhibitor p16 plays a role (Bond JA cultures were maintained in a 2 : 1 : 1 mixture of DMEM, -unpublished). Comparison of mTSHR colonies in 1% Ham's F12 and MCDB104 with 1% or 10% FCS (Life versus 10% FCS shows that the timing of this lifespan Technologies, Paisley, UK). In 1% FCS cultures, 10 mg/ml checkpoint can be modulated by the growth factor insulin and 5 mU/ml or 0.3 mU/ml bovine TSH were also environment. Although the colony sizes observed in included. Primary cultures from 13 di€erent thyroids were vitro would probably not be sucient to account for a used. Five were freshly disaggregated; in the remainder clinically-evident tumour, this raises the important follicles had been stored frozen in DMSO prior to use. possibility that the environment in the intact organ may permit signi®cantly more PD to be achieved. Retroviral infection Clearly though, further tumour development beyond the initial self-limiting clone must be dependent on Forty-eight hours after seeding in 5H medium, FRTL5 cells mutation of additional growth-regulatory genes, such were pretreated for 1 h with 8 mg/ml polybrene and then exposed to retroviral supernatants for 2 h before refeeding as p16ink4a (Wynford-Thomas, 1997a,b). If, as suggested with fresh medium. Cultures were passaged the following by our data, TSHR mutation usually generates smaller day, and after a further 24 h G418 selection commenced. Six clones in vivo than does RAS, the chance of such a G418-resistant clones from each GSP, wtTSHR and mTSHR second event will be correspondingly smaller. This may infection were isolated, at random, using cloning rings. be one explanation for the relative rarity of tumours Pooled G418 resistant clones derived from the MLSX1 initiated by TSHR mutation compared to RAS (when infected cells were used as a control neo-only population. all tumour types are considered together), despite the Primary human thyrocytes were similarly infected with much larger number of target sites for activating retroviral supernatants 48 h after plating, and G418 selection mutations in the TSHR gene. started after a further 48 h. For experiments performed in 1% FCS, cultures were maintained in the presence of insulin throughout, but TSH (0.3 or 5 mU/ml) was included only up Materials and methods to the day following retroviral infection.

Production of retroviral vectors Proliferation assays (FRTL5 only) The rat Q227L GSP and H-RAS amphotropic retroviral Growth curves 56104 cells were seeded as multiple vectors driven by a MoMuLV LTR have been described replicates in 24 well plates in 4H or 5H medium. Triplicate GSP and TSHR mutants in thyrocyte function and proliferation MLudgateet al 4806 wells were trypsinized and counted at 2, 4 and 7 days and results expressed as mean+s.e. of quadruplicate dishes (or also at 11 days for cultures in 4H. Results are expressed as colonies in the case of primary cells). the mean of triplicates+standard error (s.e.). Semi-quantitative RT ± PCR analysis of NIS expression DNA synthesis 56104 cells were seeded in 35 mm dishes, in 4H or 5H medium. Seventy-two hours later Cells were grown to con¯uence in 6-well plates, in 4H or 5H 10 mM BrdU was added for a further 24 h, following medium+IBMX. RNA was extracted (Chomczynski and which monolayers were ®xed in 3.7% formaldehyde Sacchi, 1987) and reverse-transcribed using an oligo-dT (15 min), permeabilized (0.2% Triton-X100) and immu- primer (Ajjan et al., 1998). PCR was performed in a 50 ml nostained using a mouse monoclonal anti-BrdU primary reaction containing 2.5 units Taq polymerase (Promega), antibody (Boehringer-Mannheim), in the presence of 0.1 mM of each dNTP, and 1.5 ml of cDNA, in a bu€er

10 U/ml DNAse followed by ¯uorescein-conjugated goat consisting of 1 mM MgCl2,10mM Tris HCl, 0.01% gelatin, anti-mouse IgG (Southern Biotechnology). The percen- 50 mM KCl, 0.1% Tween and 0.1% NP40. Ampli®cation was tage of labelled nuclei (LI) was determined from a count performed using 30 cycles of 948C, 1 min; 558C, 1 min; and of 4300 per dish and results expressed as means of 728C, 1 min. The NIS primers were designed using the quadruplicate dishes+s.e. published rat sequence (Dai et al., 1996) as follows: forward: 5'-CTG CGA CTC TCC CAC TGA-3' and reverse: 5'-CGC AGC TCT AGG TAC TGG TA-3'. In addition, the cAMP assay following b-actin primers were used as an internal standard: forward: 5'-GTG GGG CGC CCC AGG CACCA-3' and Five6104 cells were seeded in 24 well plates in 4H medium. 72 ± 96 h later, the medium was replaced with Ham's F12 reverse: 5'-CTC CTT AAT GTC ACG CAC GAT TTC-3'. containing 1% BSA with varying concentrations of bovine Control reactions omitting the reverse transcription step were TSH. After 4 h incubation cells were extracted in HCl, and also performed to exclude the possibility of genomic extracts evaporated to dryness before resuspension in bu€er contamination. Ten ml of the ampli®ed product were for cAMP determination using a commercial radio- separated by agarose gel electrophoresis, stained with immunoassay (Amersham, UK). Di€erences in growth rate ethidium bromide and scanned into a Bio-Rad gel of the clones were corrected by measuring total cellular documentation system for quanti®cation of band intensity. protein using a Bradford assay. Experiments were Results are expressed as the ratio of NIS to b-actin signals performed in quadruplicate and results expressed as a and expressed as the mean+s.e. of triplicate determinations. percentage (+s.e.) of the level in control cells in 4H medium. Statistics The signi®cance of di€erences between means was assessed by Detection of thyroglobulin by immunocytochemistry Student's t-test. Monolayers were ®xed with methanol (10 min, at 7208C), permeabilized with 0.1% Triton-6100 in PBS (15 min), and non-speci®c antibody binding blocked with 10% FCS in PBS (30 min). Cells were then incubated with a rabbit polyclonal Acknowledgements anti-human thyroglobulin antibody (Dako, Carpinteria CA, We are grateful to the Medical Research Council for grant USA) at a 1 : 500 dilution for 1 h, followed by rhodamine- support, to Brahms Diagnostics and to Dr Gilbert Vassart conjugated goat anti-rabbit IgG (Southern Biotechnology) (Brussels) for supply of reagents, to Michelle Haughton for (1 h). The percentage of cells containing immunodetectable technical support and to Theresa King for manuscript cytoplasmic Tg was scored for at least 300 cells per dish and preparation.

References

Ajjan R, Watson P, Findlay C, Metcalfe A, Crisp M, Fuhrer D, Holzapfee H, Wonerow P, Scherbaum W and Ludgate M and Weetman A. (1998). J. Endocrinol., 158, Paschke R. (1997). J. Clin. Endocrinol. Metab., 82, 3885 ± 351 ± 358. 3891. Bond JA, Wyllie FS, Rowson J, Radulescu A and Wynford- Ivan M, Ludgate M, Gire V, Bond JA and Wynford-Thomas Thomas D. (1994). Oncogene, 9, 281 ± 290. D. (1997). J. Clin. Endocrinol. Metab., 82, 2702 ± 2709. Burns JS, Shaw J, Williams ED and Wynford-Thomas D. Kupperman E, Wen W and Meinkoth JL. (1993). Mol. Cell. (1992). Int. J. Oncol., 1, 79 ± 92. Biol., 13, 4477 ± 4484. Chomczynski P and Sacchi N. (1987). Analyt. Biochem., 162, Lalli E and Sassonecorsi P. (1995). Proc. Natl. Acad. Sci. 156 ± 159. USA, 92, 9633 ± 9637. Costagliola S, Many M-C, Stalmans-Falys M, Tonacchera Michiels F, Caillou B, Talbot M, Dessarps-Freichey F, M, Vassart G and Ludgate M. (1994). Endocrinol., 135, Maunoury M, Schlumberger M, Merken L and Monier R. 2150 ± 2159. (1994). Proc. Natl. Acad. Sci. USA, 91, 10488 ± 10492. Dai G, Levy O and Carrasco N. (1996). Nature, 379, 458 ± Muca C and Vallar L. (1994). Oncogene, 9, 3647 ± 3653. 460. Nemoz G, Sette C, Hess M, Muca C, Vallar L and Conti M. Danos O and Mulligan RC. (1988). Proc. Natl. Acad. Sci. (1995). Mol. Endocrinol., 9, 1279 ± 1287. USA, 85, 6460 ± 6464. O'Sullivan C, Barton CM, Staddon SL, Brown CL and Dumont JE, Lamy F, Roger P and Maenhaut C. (1992). Lemoine NR. (1991). Mol. Carcinogen., 4, 345 ± 349. Phys. Rev., 72, 667 ± 697. Parma J, Duprez L, Van Sande J, Cochaux P, Gervy C, Duprez L, Parma J, Van Sande J, Allgeier A, Leclere J, Mockel J, Dumont J and Vassart G. (1993). Nature, 365, Schvartz C, Delisle MJ, Decoulx M, Orgiazzi J, Dumont J 649 ± 651. and Vassart G. (1994). Nature Genet., 7, 396 ± 401. Parma J, Duprez L, Van Sande J, Hermans J, Rocmans P and Fournes B, Monier R, Michiels F, Milgrom E, Misrahi M VanVliet G. (1997). J. Clin. Endocrinol. Metab., 82, 2695 ± and Feunteun J. (1998). Oncogene, 16, 985 ± 990. 2701. GSP and TSHR mutants in thyrocyte function and proliferation MLudgateet al 4807 Parma J, Van Sande J, Swillens S, Tonacchera M, Dumont Said S, Schlumberger M and Suarez HG. (1994). J. JE and Vassart G. (1995). Mol. Endocrinol., 9, 725 ± 733. Endocrinol. Invest., 17, 371 ± 379. Paschke R and Ludgate M. (1997). NewEngl.J.Med.,337, Van Biesen T, Luttrell LM, Hawes BE and Lefkowitz RJ. 1675 ± 1681. (1996). Endocrine Rev., 17, 698 ± 709. PaschkeR,TonaccheraM,VanSandeJ,ParmaJand Van Sande J, Parma J, Tonacchera M, Swillens S, Dumont J Vassart G. (1994). J. Clin. Endocrinol. Metab., 79, 1785 ± and Vassart G. (1995). J. Clin. Endocrinol. Metab., 80, 1789. 2577 ± 2585. Porcellini A, Ruggiano G, Pannain S, Ciullo I, Amabile G, Vassart G and Dumont JE. (1992). Endocrine Rev., 13, 596 ± Fenzi G and Avvedimento EV. (1997). Oncogene, 15, 611. 781 ± 789. Williams DW, Williams ED and Wynford-Thomas D. Rapoport B, Filetti S, Takai N and Seto P. (1982). FEBS (1988). Br. J. Cancer, 57, 535 ± 539. Letts, 146, 23 ± 27. Wynford-Thomas D. (1993). Cancer Surveys, 16, 115 ± 133. Russo D, Arturi F, Schlumberger M, Caillou B, Monier R, Wynford-Thomas D. (1997a). Eur. J. Cancer, 33, 716 ± 726. Filetti S and Suarez HG. (1995a). Oncogene, 11, 1907 ± Wynford-Thomas D. (1997b). Hormone Res., 47, 145 ± 157. 1911. Wynford-Thomas D, Stringer BMJ and Williams ED. Russo D, Arturi F, Suarez H, Schlumberger M, DuVillard (1982). Acta Endocrinol., 101, 562 ± 569. JA, Crocetti U and Filetti S. (1996). J. Clin. Endocrinol. Metab., 81, 1548 ± 1551. Russo D, Arturi F, Wicker R, Chazenbalk GD, Schlumber- ger M, DuVillard J-AD, Caillou B, Monier R, Rapoport B, Filetti S and Suarez HG. (1995b). J. Clin. Endocrinol. Metab., 80, 1347 ± 1351.