Division of Haematology, Hanson Centre for Cancer Research, Institute of Medical and Veterinary Science, Box 14 Rundle Mall PO, Adelaide, SA 5000, Australia

Correspondence to: Leonie K Ashman, Division of Haematology, Hanson Centre for Cancer Research, Institute of Medical and Veterinary Science, Box 14 Rundle Mall PO, Adelaide, SA 5000, Australia

^aCurrent address: Program in Molecular Biology and Cancer, Samuel Lunenfeld Research Institute, Mount Sinai Hospital, 600 University Avenue, Toronto, Ontario, Canada M5G 1X5

^bG Caruana and AC Cambareri contributed equally to this work

Alternate splicing of mRNA encoding c-KIT results in isoforms which differ in the presence or absence of four amino acids (GNNK) in the juxtamembrane region of the extracellular domain of the receptor. In this study we show that these isoforms of human c-KIT, expressed at similar levels in NIH3T3 cells, display differential effects on various attributes of transformation. The GNNK- isoform strongly promoted anchorage independent growth (colony formation in semi-solid medium), loss of contact inhibition (focus formation), and led to tumorigenicity in nude mice. In contrast, the GNNK+ isoform elicited colony formation but relatively poor focus formation and no tumorigenicity. Saturation binding analysis indicated that the isoforms do not differ significantly in their affinity for the KIT ligand, Steel Factor (SLF). Negligible ligand-independent receptor phosphorylation was observed in either case but, after ligand stimulation, the GNNK- isoform displayed more rapid and extensive tyrosine autophosphorylation and faster internalization. Both isoforms recruited the p85 subunit of phosphatidylinositol 3-kinase and led to similar phosphorylation of its downstream effector c-Akt, but the GNNK- isoform gave rise to more MAP kinase phosphorylation. Thus the c-KIT isoforms display different signalling characteristics and have different transforming activity in NIH3T3 cells.

c-KIT; isoforms; transformation; ligand binding; signal transduction

Introduction

The receptor tyrosine kinase, c-Kit, is a known proto-oncogene. Truncation and mutation of feline or murine c-Kit (Besmer et al., 1986; Herbst et al., 1995a), or point mutation of human c-KIT (Furitsu et al., 1993) results in transformation and tumorigenicity. Ectopic expression of normal KIT can also contribute to tumorigenicity: expression of murine or human c-KIT in immortalized fibroblasts such as NIH3T3 cells leads to transformation which is partly factor independent (Alexander et al., 1991; Caruana et al., 1998).

Different isoforms of c-KIT exist as a result of alternate mRNA splicing events. In the mouse, two isoforms are known and are characterized by the presence or absence of a four amino acid sequence, GNNK, in the juxtamembrane region of the extracellular domain (Hayashi et al., 1991; Reith et al., 1991). These variants arise due to the use of alternative 5' splice donor sites at the exon/intron junction of exon 9 (Hayashi et al., 1991) and appear to be co-expressed in a variety of tissues (Reith et al., 1991). The GNNK+/- variants also exist in human c-KIT. Analysis by ribonuclease protection assay demonstrated the presence of the splice variants in a range of normal and myeloid leukaemic cell types with mRNA encoding the GNNK- form being more abundant (Crosier et al., 1993). These results were confirmed and extended in other studies (Furitsu et al., 1993; Piao et al., 1994; Zhu et al., 1994).

Little is known about whether functional differences exist between the isoforms. When transiently expressed in COS cells, GNNK-, but not GNNK+, murine c-Kit displayed some constitutive tyrosine phosphorylation and association with phosphatidylinositol 3-kinase (PI3-K) (Reith et al., 1991). However these isoforms had similar ability to restore survival, proliferation and adhesion responses to the KIT ligand, Steel Factor (SLF), in W^sh mast cells (Serve et al., 1995). We now report differences in the induction of various attributes of cell transformation in NIH3T3 cells stably expressing human c-KIT isoforms. Care has been taken to ensure that the levels of expression are comparable, since this is known to have a considerable effect on the outcome (Caruana et al., 1998). In addition, the affinities of the isoforms for SLF have been compared and the kinetics of receptor activation and downstream signalling have been examined.

Results

Generation and selection of NIH3T3 cells expressing isoforms of human c-KIT

Starting with a c-KIT cDNA clone encoding a protein corresponding to the published sequence (Yarden et al., 1987) i.e. GNNK+, a clone encoding the GNNK- isoform was generated using PCR methodology as detailed in Materials and methods and validated by sequencing. cDNAs were subcloned into the retroviral vector pRufMC1neo (Rayner and Gonda, 1994), packaged in 2 cells, and supernatants were used to infect early passage NIH3T3 cells. G418-selected cells were screened for expression of c-KIT by indirect immunofluorescence using monoclonal antibody (MAb) 1DC3 (Figure 1a). Analysis of mRNA showed similar levels of transcript of the predicted sizes (Figure 1b). That the correct isoforms were expressed was confirmed in each case by PCR analysis (Figure 1c).

To enable comparisons between populations expressing similar levels of the isoforms, and to compare the effects of different levels of the same isoform, pools of cells comprising the top 10, 5 and 2% of each population of infectants (labelled A, B and C respectively in Table 1) were isolated by sequential fluorescence activated cell sorting (FACS). These pools were expanded and, after checking their expression of c-KIT, cryopreserved. All subsequent experiments were performed on cells expanded for no more than 2 weeks from these frozen stocks, ensuring the oligoclonality and consistency of the cell populations. The relative levels of c-KIT were determined for each isoform using indirect immunofluorescence and flow cytometry. Results were standardized by saturation binding analysis using ¹²⁵I-SLF on one population for each isoform (see Figure 2a). In fact, there was good agreement between the results obtained from relative binding of a number of anti-KIT MAbs and SLF binding at saturation for the different isoforms indicating that the MAbs do not distinguish between the isoforms (data not shown). Mean copy numbers for the different populations of cells are included in Table 1 and were in the range 6.8 - 46´10³/cell. This can be compared with 20´10³/cell for CD34+ cells from normal human bone marrow (Cole et al., 1996).

Affinity of isoforms of c-KIT for SLF

Saturation binding of ¹²⁵I-SLF and Scatchard analysis was used to determine the relative affinity of the isoforms for ligand. Results of five to seven separate experiments on each isoform, carried out using NIH3T3 cells (n=2), PC12 cells (n=2 - 4) and FDCP1 cells (n=1) are summarized in Figure 2b, and show that there was no significant difference in dissociation constant (Kd=57±13 pM and 87±26 pM for the GNNK+ and GNNK- isoforms respectively) for SLF.

Transformation of NIH3T3 cells expressing c-KIT isoforms

The KIT-expressing cell populations described above were assayed for their ability to grow in an anchorage-independent fashion (colony formation in soft agar), and for the loss of contact inhibition (focus formation) in the presence or absence of 100 ng/ml SLF. Early passage NIH3T3 cells expressing murine GNNK- c-Kit (as previously described; Caruana et al., 1998) were included for comparison. A representative set of experiments is shown in Table 1; in each case the results were confirmed in at least one additional experiment. As previously described for human GNNK+ KIT expressed from pRSV009/A+ vector (Caruana et al., 1998), anchorage independent growth was observed in cells expressing both isoforms, and the yield of colonies was lower at high levels of expression. Similar to the previous study, colony formation was substantially independent of added SLF for human but not murine Kit.

Different results were obtained when cell populations were assayed for the loss of contact inhibition. The GNNK- isoform was much more efficient at inducing focus formation and this was substantially SLF-dependent, at least at lower receptor levels. In contrast to the results of the colony assay, increasing focus formation was observed with increasing receptor levels for both isoforms (Table 1). In comparison, murine GNNK- Kit resulted in growth of cells as a dense monolayer without the formation of distinct foci. An example of the focus formation assay is shown in Figure 3. Consistent with the loss of contact inhibition, cells expressing GNNK- or, to a lesser extent, GNNK+ c-KIT grew to a higher density than cells infected with empty vector (Table 1).

NIH3T3 cells expressing the most comparable levels of c-KIT protein for each isoform were tested for their ability to give rise to tumours in nude mice. Only cells expressing the GNNK- isoform were tumorigenic in vivo with a latency period of approximately 42 days (Table 1). As previously reported (Caruana et al., 1998), NIH3T3 cells expressing murine GNNK- c-Kit also induced tumours, in this case with an average latency of 50 days.

Kinetics of activation of c-KIT isoforms

To begin to address the biochemical basis for the distinct transforming effects of c-KIT isoforms, we examined the kinetics of receptor tyrosine phosphorylation and down-regulation in response to SLF. Since the different isoforms were shown to have similar affinity for SLF, we used a single stimulating dose of 100 ng/ml in these experiments. Cells expressing comparable levels (approximately 10⁴ copies/cell) of each isoform were cultured to approximately 70% confluence then starved of serum for 3 h prior to pulsing with SLF. After the indicated times, cells were chilled on ice and lysed in situ with ice-cold lysis solution. Lysates containing comparable amounts of protein were subjected to immunoprecipitation with the anti-c-KIT MAb, KIT4, followed by polyacrylamide gel electrophoresis (PAGE) under reducing conditions and Western blotting with MAb to phosphotyrosine, then c-KIT, and detection using the FluorImager. Co-immunoprecipitation of PI3-K, which is the dominant downstream effector molecule that associates with c-Kit (Reith et al., 1991; Lev et al., 1991; Herbst et al., 1995b), was detected by blotting with antibody to the p85 subunit. To enable comparison of data from different experiments, standard aliquots of NP40 lysates of starved and SLF-pulsed MO7e cells were subjected to immunoprecipitation and analysed in parallel with NIH3T3 infectants. Results of typical blots and data derived by quantitation using ImageQuant software are shown in Figure 4. Under the conditions used in this study, a very low level of receptor phosphorylation, and some association with p85, was seen in the absence of SLF. The presence or absence of GNNK had a substantial effect on the kinetics and extent of receptor phosphorylation, with the GNNK+ isoform displaying lower level but apparently more sustained phosphorylation. In five independent experiments, the mean ratio of the phosphotyrosine to c-KIT signals at the peak response was 5.7±1.6 for the GNNK- isoform compared with 0.82±0.44 for GNNK+. Furthermore, receptor phosphorylation peaked earlier for the GNNK- isoform (median 2.5 min; range 2 - 3 min post stimulation) than the GNNK+ isoform (median 7.5 min; range 2.5 - 10 min). In contrast, the maximum level of p85 recruitment to the two isoforms was similar (Figure 4). Blotting for c-KIT protein indicated that the GNNK- isoform was more rapidly down-regulated, accounting for the loss of associated p85 and phosphotyrosine at later time points.

Internalization of c-KIT following SLF stimulation

The apparent down-regulation of GNNK- c-KIT protein shown in Figure 4 could arise from degradation (for example cleavage of the extracellular domain (Yee et al., 1993; Brizzi et al., 1994)), or from internalization, cytoskeletal association and incomplete solubilization in NP40 detergent. We used confocal microscopy on permeabilized and non-permeabilized cell monolayers to monitor SLF-induced internalization of the receptor. As illustrated in Figure 5, GNNK- c-KIT was completely lost from the cell surface within 10 min of SLF stimulation, while the GNNK+ isoform was still evident at 20 min. Experiments on permeabilized cells indicated that loss of the receptor from the surface was substantially due to endocytosis. Some intracellular c-KIT, possibly newly synthesized, was seen in unstimulated cells expressing both isoforms, although there was a hint of GNNK- c-KIT in endocytic vesicks as indicated by punctate staining. Following SLF stimulation, substantial internalization of GNNK- isoform could be seen as early as 3 min, whereas a comparable level of endocytosis of GNNK+ required at least 10 min (Figure 6).

Downstream signalling from c-KIT isoforms

As shown in Figure 4, PI3-K was recruited to a similar extent to both receptor isoforms. As a measure of PI3-K activation, phosphorylation of a major target, c-Akt in NP40 lysates was examined by Western blotting with antibodies specific for the phosphorylated form. Both c-KIT isoforms were capable of activating c-Akt phosphorylation to a similar extent following SLF stimulation (Figure 7). As a measure of activation of the Ras-MAP kinase (MAPK) pathway, phospho-MAPK was also examined. The GNNK- isoform brought about fourfold stronger phosphorylation of MAPK than GNNK+ c-KIT, although activation appeared transient (Figure 7). Similar results were obtained in three independent experiments. This transience may have been due in part to translocation to the nucleus c.f. (Traverse et al., 1992) as nuclear phospho-MAPK was observed at 30 min post SLF stimulation by confocal microscopy. However no appreciable difference between cells expressing the different isoforms was apparent (data not shown).

Discussion

Alternative splicing of mRNA encoding human c-KIT occurs in an apparently tissue non-specific fashion (Crosier et al., 1993) and its significance is unknown. To investigate the function of the different isoforms of human c-KIT, we have expressed them in NIH3T3 fibroblasts which lack endogenous murine c-Kit. Since stable expression of human or murine c-Kit in these cells induces several characteristics of transformation (Caruana et al., 1998), the resultant populations could be used to compare cellular responses elicited by binding of the ligand SLF to the different receptor isoforms. Any differences could then be explored at the level of signal transduction. The GNNK- isoform was more strongly transforming than the GNNK+ isoform when expressed at similar levels (approximately 10⁴ copies/cell). Interestingly, dissociation of the different correlates of transformation was observed. While the GNNK+ isoform was at least as effective as GNNK- in inducing anchorage-independent growth (colony formation in soft agar) it was relatively poor at overcoming contact inhibition (focus-formation assay), and was non-tumorigenic in nude mice.

In order to gain some insight into the biochemical basis for these different cellular responses we examined receptor activation in response to SLF. Saturation binding analysis indicated that both isoforms have similar affinity for ligand. Therefore subsequent experiments were performed with a single saturating level of SLF (100 ng/ml). The presence or absence of the GNNK tetrapeptide had a profound effect on the kinetics and extent of receptor phosphorylation (Figure 4). Activation of GNNK- was rapid, peaking at 2 - 3 min, followed by down-regulation which involved internalization of the receptor. In contrast, the GNNK+ isoform displayed later peak tyrosine phosphorylation (around 7.5 min) and showed little dephosphorylation or down-regulation by 20 min. A marked difference was observed in the relative level of phosphorylation which was sevenfold higher for the GNNK- isoform. Whether this reflects a difference in efficiency of phosphorylation overall, or in the phosphorylation of specific sites is not known. Despite the low level of phosphorylation of GNNK+ c-KIT, PI3-K was recruited similarly to both forms of the receptor and similarly activated based on phosphorylation of its major down-stream effector c-Akt (Franke et al., 1997; Marte and Downward, 1997). Recruitment of PI3-K to another type 3 receptor tyrosine kinase (RTK), the platelet-derived growth factor receptor (PDGFR), was previously reported to have the least requirement of several substrates for receptor phosphorylation (Rankin and Rozengurt, 1994). In contrast, phosphorylation of MAPK paralleled that of c-KIT and was much stronger with the GNNK- isoform. The relationship between hyperphosphorylation of GNNK- c-KIT and its more rapid endocytosis is unclear. Experiments with endocytosis-defective cells indicated that maximal phosphorylation of the EGF receptor requires internalization. Furthermore, efficient MAPK phosphorylation and activation required receptor endocytosis while activation of phospholipase C was more efficient in endocytosis-defective cells (Vieira et al., 1996). Thus the different rates of internalization of the c-KIT isoforms may contribute to their altered specificity of signalling.

The results presented here differ from those of Reith et al. (1991) with murine GNNK+ and GNNK- isoforms transiently expressed in COS cells. These workers observed some constitutive tyrosine phosphorylation of the GNNK- isoform, but not GNNK+, in the absence of ligand. In contrast, receptor phosphorylation in the absence of SLF was very low for both isoforms in our study. One possible explanation is that the levels of receptor expression, which were not determined by Reith et al. (1991) were probably somewhat higher than in our study. However, a significant level of factor-independent receptor activation can occur for both the GNNK+ and GNNK- isoforms of human c-KIT, as is evident from the results of the colony assays described here (Table 1) where relatively high frequencies of factor-independent colonies were observed. Based on our previous work, these are unlikely to arise as a result of autocrine stimulation by murine SLF produced by NIH3T3 cells. The other unusual feature of this assay was the fact that higher levels of receptor expression tended to be inhibitory, confirming an observation we previously made with murine GNNK- Kit and human GNNK+ KIT in another vector system (Caruana et al., 1998). Strong stimulation of certain signalling pathways may be inhibitory to NIH3T3 proliferation through induction of p21^Cip1 (Woods et al., 1997; Sewing et al., 1997). In contrast to the colony assay, NIH3T3 infectants grown attached to the dish in the assay for focus formation showed much less evidence of factor-independent KIT activation and inhibition at higher receptor levels was not observed (Table 1). The conditions in this assay are comparable to those under which the biochemical assays were performed. These results indicate a complex interplay between signals via c-KIT and adhesion molecules. It was recently reported that proliferation of fibroblasts induced by Ras required Rho activation to prevent p21^Cip1 induction (Olson et al., 1998). Since Rho is also involved in cell adhesion processes, it may be that this pathway is operative in the focus formation assay, but that repression of p21^Cip1 is defective in contact-deprived cells, i.e. in the colony assay.

The differences between the isoforms in their transforming ability could be correlated with differences in signalling. For example, the GNNK- isoform, which induced anchorage-independence, loss of contact inhibition and tumorigenicity, displayed stronger receptor phosphorylation, more rapid internalization, and stronger activation of the MAPK pathway following SLF stimulation than the GNNK+ isoform. In contrast, the GNNK+ isoform induced anchorage-independent growth with similar efficiency to GNNK- c-KIT, but was much less effective in inducing other attributes of transformation. The fact that SLF binding to the GNNK+ isoform efficiently recruited and activated PI3-K suggests that this pathway is particularly important in preventing death of cells deprived of contact with extracellular matrix (`anoikis'; (Ruoslahti and Reed, 1994)). Indeed PI3-K induced activation of c-Akt has recently been demonstrated to be a major mechanism of promoting cell survival in response to growth factors (Kennedy et al., 1997) and cell adhesion (Khwaja et al., 1997). There are strong parallels between GNNK+ KIT and R-Ras in both signalling and their ability to induce different attributes of transformation in vitro. Activated R-Ras was as effective as H-Ras in promoting growth of NIH3T3 cells of soft agar, but was much less effective in inducing focus formation (Cox et al., 1994). Furthermore, activated R-Ras stimulates the PI3-K pathway but, in contrast to other Ras proteins, has little effect on the MAPK pathway in fibroblasts (Marte et al., 1997).

The molecular mechanisms by which these relatively minor sequence differences between the c-KIT isoforms lead to such remarkably different activation characteristics and biological behaviour are unknown. For example, it is unclear how the GNNK+/- variation in the juxtamembrane region of the extracellular domain influences the rate of receptor phosphorylation and internalization, but it seems likely that it must modulate interactions with other membrane proteins. Since the isoforms have similar affinity for ligand, it seems improbable that the tetrapeptide has an appreciable effect on receptor homodimerization. The data presented here imply major differences between the isoforms in their interaction with signal transducing molecules. It will be important to investigate their interactions with phosphatases such as SHP-1 which is associated with c-KIT (Yi and Ihle, 1993) and a range of proteins such as Shc (Matsuguchi et al., 1994), Tec (Tang et al., 1994), Lyn (Linnekin et al., 1997), PLC1 (Rottapel et al., 1991; Herbst et al., 1995b), p120^CBL (Wisniewski et al., 1996), JAK2 (Weiler et al., 1996) and Stat 1 (DeBerry et al., 1997) that are known to be recruited or activated on SLF binding to c-KIT.

Materials and methods

Production of cDNAs encoding the c-KIT isoforms

A construct encoding the GNNK+ isoform of c-KIT (corresponding to the published sequence; (Yarden et al., 1987)) cloned into pBluescript SK (pBSSK) was provided by Dr Douglas Williams, Immunex Corporation, Seattle, WA, USA. Using this clone as template, a 6 kb product encoding the GNNK- isoform was generated by PCR using oligonucleotide primers flanking the sequence encoding GNNK: sense primer 5'-AGCAAATCCATCCCCACACC-3' and anti-sense primer 5'-CTTTAAATGCAAAGTTAAAATAGGC-3' corresponding, respectively, to nucleotides 1562 - 1581 and 1525 - 1549 of the c-KIT sequence. PCR was performed using Pfu polymerase (Stratagene, La Jolla, CA, USA) and consisted of an initial denaturation cycle of 94°C for 7 min then 25 cycles of 1 min at 94°C, 1 min at 45°C and 12 min at 72°C. The 5' ends were phosphorylated and the ends ligated. The entire c-KIT product was verified by sequencing and subcloned into the retroviral vector pRUFMC1neo (Rayner and Gonda, 1994).

Generation of cell lines expressing c-KIT isoforms

The retroviral constructs were transfected into 2 packaging cells (Mann et al., 1983) and virus-containing supernatants from G418-resistant pools of cells were used to infect early passage NIH3T3 cells as previously described (Caruana et al., 1998). The viral titres of the supernatants ranged from 1 - 6´10⁵ c.f.u./ml. G418-resistant infectants were assayed for expression of c-KIT protein and mRNA by immunofluorescence and flow cytometry and Northern blotting, respectively, as previously described (Caruana et al., 1998).

To verify that the cell populations expressed the correct isoforms, first strand cDNA was reverse transcribed from 200 ng of poly(A)⁺ mRNA from each cell line using the First-Strand cDNA Synthesis Kit (Pharmacia Biotech, Uppsala, Sweden). Amplification was performed with Taq polymerase (Perkin Elmer Cetus, Norwalk, CT, USA) using the primer pair sense: 5'-GGGGGATCCGATGTGGGCAAGACTTCT-3' (nt 1506 - 1524); anti-sense: 5'-CAGCAAAGGAGTGAACAG-3' (nt 1582 - 1599). The PCR products were size fractionated on a 4% agarose gel for detection of the 12 bp insertion or deletion which give products of 93 bp and 81 bp respectively. Populations of cells expressing different levels of the c-KIT isoforms were obtained by fluorescence-activated cell sorting (FACS) using MAb 1DC3, as previously described (Caruana et al., 1998), with selection of the top 2 - 10% of cells in several rounds of sorting.

Expression levels and comparison of the affinity of different isoforms of c-KIT for SLF

Levels of expression were quantitated by flow cytometry and saturation binding analysis, also as previously described (Caruana et al., 1998). Recombinant human SLF (Stem Cell Factor, Amgen, Thousand Oaks, CA, USA) was iodinated to a specific activity of 13.7 Ci/g. Binding assays were performed on the NIH3T3 transfectants grown to 80% confluency in 24-well tissue culture plates over a concentration range of 6 pM to 3 nM ¹²⁵I-labelled SLF in binding medium (RPMI-1640 supplemented with 0.5% BSA and 0.1% NAN₃) with non-specific binding determined at each concentration with excess unlabelled SLF. After incubation at 4°C for 4 h, unbound ligand was removed and the wells washed twice in cold binding medium. Specifically bound ligand was determined after lysis of the cell monolayer with subsequent transfer and counting on a -counter (Cobra Auto Gamma, Packard, Meridien, CT, USA). Dissociation constants and copy number were determined by Scatchard analysis.

Assays of cell transformation

Cell populations were assayed for anchorage-independent growth by colony-formation in soft agar and for loss of contact inhibition by focus-formation as previously described (Caruana et al., 1998). Saturation density was determined by plating 5´10⁴ cells in 6 cm diameter dishes and grown to confluency. Cells were harvested and cell numbers and viability were determined using a haemocytometer and Trypan blue exclusion. To assess tumorigenicity, BALB/c nude mice were injected subcutaneously in the hind flank with 8´10⁶ cells. Tumour production and size were monitored twice weekly.

Immune precipitation and Western blot analysis

NIH3T3 transfectants were grown to 80% confluency in 9 cm dishes in DMEM/10% FBS, then starved of serum for 3 h. The cells were then pulsed with 100 ng/ml SLF at 37°C for the times indicated. Plates were rinsed twice in ice-cold PBS, then lysed with 1 ml ice-cold lysis buffer (1% NP40 in TSE: 50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA pH 8.0), with Complete^TM Protease Inhibitors (Boehringer Mannheim GmbH, Mannheim, Germany) and 20 M sodium orthovanadate (ICN Biomedicals, Aurora, OH, USA). Cells were scraped off the dish using a teflon scraper and incubated on ice for 30 min. Lysates were then centrifuged at 13 000 r.p.m. in a microfuge. Protein content for each lysate was determined using the BCA Protein Assay Reagent (Pierce, Rockford, IL, USA).

Immunoprecipitations were performed using 5 g anti-c-KIT MAb (KIT4, IgG2a) and 30 l Protein A Sepharose CL 4B (50% slurry) (Pharmacia). Sepharose pellets were washed in lysis buffer, then boiled 2 min in reducing loading buffer and separated by SDS - PAGE. Immunoprecipitated proteins were electroblotted to PVDF membrane (0.45 m, Micron Separations, Westborough, MA, USA) and probed with a cocktail of anti-phosphotyrosine Abs (4G10 (Upstate Biotechnology, Lake Placid, NY, USA) and PY20 (Transduction Laboratories, Lexington, KY, USA)), followed by anti mouse Ig-alkaline phosphatase conjugate (AMRAD, Melbourne, Australia). All incubations and washes were done using Tris buffered saline pH 7.4, 0.1% Tween 20 (TBST). Alkaline phosphatase was visualized using VISTRA ECF substrate (Amersham, Buckinghamshire, UK), and a FluorImager 595 (Molecular Dynamics, Sunnyvale, CA, USA) (570 nm BP filter). Blots were then stripped at 60°C for 30 min in stripping buffer (62.5 mM Tris HCl pH 6.8, 2% SDS, 100 mM -mercaptoethanol), washed in TBST extensively, and reprobed simultaneously with anti-c-KIT (MAb 1C1; kindly provided by Dr H-J Bühring, University of Tübingen) and anti-PI3-K (Upstate Biotechnology). For detection of phospho-MAPK and phospho-Akt, 20 l aliquots of NP40 lysates were electrophoresed, transferred to PVDF membrane and probed with rabbit antisera (New England Biolabs, Beverly, MA, USA) as above.

Receptor internalization

Analysis of cell surface and intracellular antigen distribution was performed by confocal microscopy. In all cases, the NIH3T3 cells were plated into Lab-Tek^Ò Chamber Slides^TM (Nunc, Naperville, IL, USA) at 2´10⁴/chamber in DMEM with 10% FBS and cultured overnight. The cells were starved of serum for 2 h, then pulsed with 100 ng/ml SLF at 37°C for the times indicated. Cells were fixed for 30 s in permeabilizing fixative (47.5% methanol, 47.5% acetone, 5% formaldehyde) or for 15 min in non-permeabilizing fixative (1% paraformaldehyde in PBS) at 4°C, rinsed five times in ice cold PBS, then murine anti-c-Kit or isotype-matched negative control MAbs, diluted in 10% normal rabbit serum (NRS), were added and incubated at 4°C for 2 h, followed by sheep anti-mouse Ig-FITC (AMRAD) secondary antibody (in 10% NRS) for 1 h at 4°C. After rinsing and fixation for 15 min in 1% paraformaldehyde slides were analysed on a MRC 600 confocal microscope (Bio-Rad Microscience, Cambridge, MA, USA) using 488 nm excitation.

Acknowledgements

We thank Ly Nguyen for skilled technical assistance, Paul Sincock for help with confocal microscopy, Sonia Young for contributing binding data on PC12 cells, and Regan Forrest for carrying out preliminary signalling experiments. This work was supported by a grant from the National Health and Medical Research Council of Australia of which LK Ashman is a Senior Research Fellow, and by an Australian Postgraduate Award to G Caruana.

Alexander WS, Lyman SD and Wagner EF. (1991). EMBO J. 10, 3683-3691. MEDLINE

Besmer P, Murphy JE, George PC, Qui F, Bergold PJ, Lederman L, Snyder HW, Brodeur D, Zuckerman EE and Hardy WD. (1986). Nature 320, 415-421. MEDLINE

Brizzi MF, Blechman JM, Cavalloni G, Givol D, Yarden Y and Pegoraro L. (1994). Oncogene 9, 1583-1589. MEDLINE

Caruana G, Cambareri AC, Gonda TJ and Ashman LK. (1998). Oncogene 16, 179-190. MEDLINE

Cole SR, Aylett GW, Casey G, Harvey NL, Cambareri AC and Ashman LK. (1996). Leukemia 10, 288-296. MEDLINE

Cox AD, Brtva TR, Lowe DG and Der CJ. (1994). Oncogene 9, 3281-3288. MEDLINE

Crosier PS, Ricciardi ST, Hall LR, Vitas MR, Clark SC and Crosier KE. (1993). Blood 82, 1151-1158. MEDLINE

DeBerry C, Mou S and Linnekin D. (1997). Biochem. J. 327, 73-80. MEDLINE

Franke TF, Kaplan DR, Cantley LC and Toker A. (1997). Science 275, 665-668. Article MEDLINE

Furitsu T, Tsujimura T, Tono T, Ikeda H, Kitayama H, Koshimizu U, Sugahara H, Butterfield JH, Ashman LK, Kanayama Y, Matsuzawa Y, Kitamura Y and Kanakura Y. (1993). J. Clin. Invest. 92, 1736-1744. MEDLINE

Hayashi S-I, Kunisada T, Ogawa M, Yamaguchi K and Nishikawa S. (1991). Nucl. Acids Res. 19, 1267-1271.

Herbst R, Munemitsu S and Ullrich A. (1995a). Oncogene 10, 369-379. MEDLINE

Herbst R, Shearman MS, Jallal B, Schlessinger J and Ullrich A. (1995b). Biochem. 34, 5971-5979.

Kennedy SG, Wagner AJ, Conzen SD, Jordan J, Bellacosa A, Tsichlis PN and Hay N. (1997). Genes Dev. 11, 701-713. MEDLINE

Khwaja A, Rodriguez-Viciana P, Wennström S, Warne PH and Downward J. (1997). EMBO J. 16, 2783-2793. Article MEDLINE

Lev S, Givol D and Yarden Y. (1991). EMBO J. 10, 647-654. MEDLINE

Linnekin D, DeBerry CS and Mou S. (1997). J. Biol. Chem. 272, 27450-27455. MEDLINE

Mann R, Mulligan RC and Baltimore D. (1983). Cell 33, 153-159. MEDLINE

Marte BM, Rodriguez-Viciana P, Wennström S, Warne PH and Downward J. (1997). Curr. Biol. 7, 63-70. MEDLINE

Marte BM and Downward J. (1997). T.I.B.S. 22, 355-358.

Matsuguchi T, Salgia R, Hallek M, Eder M, Druker B, Ernst TJ and Griffin JD. (1994). J. Biol. Chem. 269, 5016-5021. MEDLINE

Olson MF, Paterson HF and Marshall CJ. (1998). Nature 394, 295-299. Article MEDLINE

Piao X, Curtis JE, Minkin S, Minden MD and Bernstein A. (1994). Blood 83, 476-481. MEDLINE

Rankin S and Rozengurt E. (1994). J. Biol. Chem. 269, 704-710. MEDLINE

Rayner JR and Gonda TJ. (1994). Mol. Cell. Biol. 14, 880-887. MEDLINE

Reith AD, Ellis C, Lyman SD, Anderson DM, Williams DE, Bernstein A and Pawson T. (1991). EMBO J. 10, 2451-2459. MEDLINE

Rottapel R, Reedijk M, Williams DE, Lyman SD, Anderson DM, Pawson T and Bernstein A. (1991). Mol. Cell. Biol. 11, 3043-3051. MEDLINE

Ruoslahti E and Reed JC. (1994). Cell 77, 477-478. MEDLINE

Serve H, Yee NS, Stella G, Sepp-Lorenzino L, Tan JC and Besmer P. (1995). EMBO J. 14, 473-483. MEDLINE

Sewing A, Wiseman B, Lloyd AC and Land H. (1997). Mol. Cell. Biol. 17, 5588-5597. MEDLINE

Tang B, Mano H, Yi T and Ihle JN. (1994). Mol. Cell. Biol. 14, 8432-8437. MEDLINE

Traverse S, Gomez N, Paterson H, Marshall C and Cohen P. (1992). Biochem. J. 288, 351-355. MEDLINE

Vieira AV, Lamaze C and Schmid SL. (1996). Science 274, 2086-2089. Article MEDLINE

Weiler SR, Mou S, DeBerry CS, Keller JR, Ruscetti FW, Ferris DK, Longo Dl and Linnekin D. (1996). Blood 87, 3688-3693. MEDLINE

Wisniewski D, Strife A, Berman E and Clarkson B. (1996). Leukemia 10, 229-237. MEDLINE

Woods D, Parry D, Cherwinski H, Bosch E, Lees E and McMahon M. (1997). Mol. Cell. Biol. 17, 5598-5611. MEDLINE

Yarden Y, Kuang W-J, Yang-Feng T, Coussens L, Munemitsu S, Dull TJ, Chen E, Schlessinger J, Francke U and Ullrich A. (1987). EMBO J. 6, 3341-3351. MEDLINE

Yee NS, Langen H and Besmer P. (1993). J. Biol. Chem. 268, 14189-14201. MEDLINE

Yi T and Ihle JN. (1993). Mol. Cell. Biol. 13, 3350-3358. MEDLINE

Zhu WM, Dong WF and Minden M. (1994). Leuk. Lymph. 12, 441-447.

Figure 1 (a) Expression of c-KIT isoforms by pools of NIH3T3 infectants determined by immunofluorescence with anti-c-KIT MAb 1DC3 and flow cytometry. The dashed lines indicate cells infected with empty vector. (b) Expression of c-KIT mRNA by the infected cells. Poly(A)⁺ selected mRNA (2 g/track) was subjected to agarose gel electrophoresis and Northern blotting with a full-length GNNK+ cDNA probe and a 780 bp human GAPDH probe. The 6.3 and 5.1 kb bands are as predicted for the transcripts generated from the pRUFMC1neo constructs containing the c-KIT cDNA (3 kb). (c) Confirmation of mRNAs encoding the different isoforms in NIH3T3 infectants. Fragments containing the variant sequences were amplified by RT - PCR, using mRNA as in (b) and separated on a 4% agarose gel

Figure 2 Saturation binding analysis of the affinity of c-KIT isoforms for SLF. Binding analysis was carried out using ¹²⁵I-labelled SLF as described in Materials and methods. Scatchard plots shown illustrate a representative experiment with c-KIT expressing NIH3T3 cells. The bar graph summarizes results of five to seven experiments carried out using NIH3T3, FDCP1 and PC12 infectants. Bars indicate s.e.m. There were no significant differences between the affinity of the different isoforms (two-tailed t-test; P>0.1)

Figure 3 Focus formation by NIH3T3 cells expressing different isoforms of c-KIT. Pools of infected cells isolated by FACS and expressing different levels c-KIT (see Table 1 for details) were assayed for focus formation in the presence or absence of SLF. The confluent cultures were stained with Diff-Quick at 12 days to reveal foci. Cells infected with empty vector are labelled RUFMC1neo

Figure 4 Time course of tyrosine phosphorylation and PI3-K recruitment of c-KIT isoforms following SLF stimulation. Serum-starved cells pulsed with SLF (100 ng/ml) for the indicated times followed by lysis in situ. Lysates containing equivalent amounts of protein were immunoprecipitated with MAb Kit4, then separated by SDS - PAGE under reducing conditions, blotted with antibodies to phosphotyrosine, c-KIT or PI3-K and detected using the FluorImager. The upper panels show digitized images. M+ indicates results obtained with a standard aliquot of a cell lysate from SLF-stimulated MO7e cells. Lower panels show quantitation of these data using ImageQuant software: () GNNK- c-KIT; () GNNK+ c-KIT; (a) c-KIT tyrosine phosphorylation; (b) c-KIT protein; (c) PI3-K p85 subunit

Figure 5 Loss of surface c-KIT following SLF stimulation. Sub-confluent cultures were pulsed for the indicated times with 100 ng/ml SLF, fixed with 1% paraformaldehyde, labelled with MAb then examined by confocal microscopy and the staining quantitated

Figure 6 Endocytosis of c-KIT following SLF stimulation. Cultures were stimulated with SLF as for Figure 5 and cells were treated with the permeabilizing fixative (47.5% methanol, 47.5% acetone, 5% formaldehyde) and examined by confocal microscopy

Figure 7 Activation of MAPK and c-Akt following stimulation with SLF. Aliquots (20 l) of NP40 lysates containing equivalent amounts of protein from SLF-stimulated cells expressing GNNK+ or GNNK- c-KIT as described for Figure 4 were electrophoresed, Western blotted with antibodies to phospho-MAPK, total MAPK and phospho-Akt, and analysed with the FluorImager. Upper panels show digitized images, lower panels show quantitation of these data using ImageQuant software: () GNNK+ c-KIT; () GNNK- c-KIT; (a) phospho-MAPK; (b) phospho-Akt

 Properties of the NIH3T3 cell lines expressing the c-KIT isoforms