Oncogenic Kit signals on endolysosomes and endoplasmic reticulum are essential for neoplastic mast cell proliferation

Kit is a receptor-type tyrosine kinase found on the plasma membrane. It can transform mast cells through activating mutations. Here, we show that a mutant Kit from neoplastic mast cells from mice, Kit(D814Y), is permanently active and allows cells to proliferate autonomously. It does so by activating two signalling pathways from different intracellular compartments. Mutant Kit from the cell surface accumulates on endolysosomes through clathrin-mediated endocytosis, which requires Kit’s kinase activity. Kit(D814Y) is constitutively associated with phosphatidylinositol 3-kinase, but the complex activates Akt only on the cytoplasmic surface of endolysosomes. It resists destruction because it is under-ubiquitinated. Kit(D814Y) also appears in the endoplasmic reticulum soon after biosynthesis, and there, can activate STAT5 aberrantly. These mechanisms of oncogenic signalling are also seen in rat and human mast cell leukemia cells. Thus, oncogenic Kit signalling occurs from different intracellular compartments, and the mutation acts by altering Kit trafficking as well as activation.

T he Kit proto-oncogene encodes a type III receptor tyrosine kinase (RTK), a class of proteins that includes plateletderived growth factor receptors (PDGFR), Fms, and Fmslike tyrosine kinase 3 (Flt3) [1][2][3] . Kit is expressed on mast cells, interstitial cells of Cajal, haematopoietic cells, germ cells and melanocytes 4 . On stimulation with stem cell factor (SCF), Kit triggers many signalling events at the plasma membrane (PM), resulting in cell proliferation, survival and differentiation 5 .
Kit is composed of five N-glycosylated immunoglobulin domains in the N-terminal extracellular portion that bind SCF, as well as a transmembrane domain, and an intracellular C-terminal tyrosine kinase domain 6 . The binding of SCF autophosphorylates Kit on specific tyrosine residues. Kit then binds to other cytoplasmic enzymes containing Src hohomogy-2 (SH2) domain, and this complex phosphorylates other proteins [3][4][5][6] . This process activates the phosphatidylinositol 3-kinase-Akt (PI3K-Akt) pathway, the Ras-Raf-Erk cascade and Src kinases, which regulate cell growth, gene expression and cytoskeletal structures [7][8][9][10] .
In many mast cell neoplasms and gastrointestinal stromal tumours, Kit has gain-of-function mutations, causing permanent, ligand-independent activation of the receptor [11][12][13][14] . In human neoplastic mast cell disorders such as mastocytosis and mast cell leukemia, Kit often has an Asp816Val substitution (D816V) in the kinase domain 12,13 . Similar mutations are also found in rat mast cell leukemia (RBL-2H3; D817Y) and a mouse mastocytoma (P815; D814Y) 13 . Their permanent activation of the PI3K-Akt pathway causes neoplastic transformation. When mutant Kit activates the PI3K-Akt pathway, this suppresses apoptotic proteins such as Bim, resulting in cell survival [15][16][17][18] . When mutant Kit phosphorylates signal transduction and activator of transcription (STAT) proteins, they move from the cytoplasm to the nucleus and cause transcription of protooncogenes such as c-myc [19][20][21] .
Cancer-causing mutants of RTKs, such as Met(D1246N) and Flt3-internal tandem duplication (Flt3-ITD), can cause incorrect signalling not just from the PM, but also from intracellular compartments, because their activated kinase domain is always exposed [22][23][24][25][26] . These mutations also change the trafficking and degradation of the receptors, because they change glycosylation and ubiquitination, and receptors accumulate in the wrong compartments [22][23][24][25][26][27] . Signalling by oncogenic RTKs on intracellular compartments has been implicated in their transforming ability, but the mechanism of signalling by mutant Kit is not fully understood.
We recently established two mast cell lines, RCM and R, from mouse splenocytes. RCM cells proliferate without cytokines, generate tumours in vivo and express a mutant Kit, Kit(D814Y). R cells require cytokines to proliferate and express wild-type Kit (Kit(wt)). This scenario allows us to compare Kit(wt) with Kit(D814Y) in an identical cellular background.
To explore how Kit(D814Y) transduces oncogenic signals, we studied what pathways it activates, from various subcellular compartments, using immunofluorescence confocal microscopy, vesicle immunoprecipitation and chemical inhibition of intracellular trafficking.
In mice cells, Kit(D814Y) from the PM accumulates on endolysosomes through clathrin-mediated endocytosis (CME); this occurs in a kinase activity-dependent manner. It then forms a complex with PI3K, and activates Akt, leading to cell proliferation. Also, soon after Kit(D814Y) is synthesized, it appears in the endoplasmic reticulum (ER), where it causes oncogenic activation of STAT5. Two other mast cell lines, HMC-1 and RBL-2H3, from humans and rats, gave similar results. Our findings demonstrate that Kit signalling from subcellular compartments is necessary for the neoplastic proliferation of mast cells.

Results
Kit D814Y causes autonomous proliferation of mouse RCM cells. We recently established two mast cell lines from mouse splenocytes, RCM cells and R cells, bearing c-Kit and FceRI. RCM cells grow without cytokines and develop tumours in vivo (Fig. 1a). These cells show constitutively tyrosine-phosphorylated 145-and 160-kDa proteins, identified as the Kit tyrosine kinase (Fig. 1b,c; see also Fig. 4b). Furthermore, Kit's kinase domain has an Asp814Tyr (D814Y) mutation (Fig. 1d), which keeps the kinase permanently active 12,13,21 .
Immunoprecipitation assays confirmed that Kit(wt) in R cells and pt18 cells 28 was activated in a ligand-dependent manner, whereas Kit(D814Y) was phosphorylated and associated with the PI3K p85 subunit without SCF ( Fig. 1e; see also Fig. 3i), and thus was permanently active. Glutathione S-transferase-pulldown (GST-pulldown) assays showed that the C-terminal or N-terminal SH2 domains of p85 (SH2c or SH2n) were associated with SCF-stimulated Kit(wt) and Kit(D814Y) (Fig. 1f). Next, we treated RCM cells with PKC412, an inhibitor of Kit tyrosine kinase 29 . Similar to previous reports 16,29 , PKC412 inhibited the phosphorylation of Kit(D814Y) and p85's association with Kit(D814Y), and also inhibited cell proliferation in a dose-dependent manner (Fig. 1f,g and Supplementary Fig. 1; see also Fig. 4c). We next performed RNA interference experiments to suppress Kit(D814Y). Western blotting showed 80-85% less Kit(D814Y) protein after transfection with Kit1 and Kit2 small interfering RNAs (siRNAs) for 20 h (Fig. 1h, left). Knockdown of Kit(D814Y) also suppressed cell proliferation (Fig. 1h, right). Thus, the kinase activity is required for autonomous proliferation.
Kit D814Y localizes to endolysosomes. Kit mutants exogenously expressed in CHO, NIH3T3 or HEK293 cells are found mainly in the Golgi apparatus or ER in an immaturely glycosylated form [30][31][32] . To investigate Kit's glycosylation state, we treated Kit(D814Y) from RCM cells with endoglycosidase H, which digests immature high-mannose forms, but not mature complex-glycosylated forms. Figure 2a shows that most Kit(D814Y), like Kit(wt), was present in a complex-glycosylated form. Kit(wt) and Kit(D814Y) both shifted to a non-glycosylated form following the complete digestion of N-linked glycans by peptide-N-glycosidase F.
Next, we investigated the subcellular localization of Kit in paraformaldehyde-fixed (PFA-fixed) cells by immunofluorescence confocal microscopy. Consistent with previous reports 9,33,34 , in pt18 and R cells, most Kit(wt) was at the PM, whereas in RCM cells Kit(D814Y) was mainly on vesicular structures (Fig. 2b). We investigated these structures and calculated Pearson's correlation coefficients (Pearson's R) between Kit and various markers: calnexin (ER), TGN46 (Golgi), EEA1, CD63 (endosome), LAMP1 and cathepsin D (endolysosome). As cathepsin D and CD63 could not be detected by immunofluorescence in PFA-fixed cells, we fixed RCM cells with methanol to visualize those proteins. Although Kit's distributions were largely similar between the different methods ( Supplementary Fig. 2a,b), methanol caused fluorescence loss in the cytosol and nucleus (compare Fig. 2c with 2d). Kit(D814Y) co-localized with LAMP1 significantly, and with cathepsin D-positive vesicles somewhat, suggesting it is mainly present on endolysosomes (Fig. 2c,d,g). Moreover, green fluorescent protein-tagged (GFP-tagged) Kit(D814Y) was colocalized with LAMP1, and with fluorescent dextran, which is incorporated into endocytic compartments 35 (Fig. 2e). Kit(D814Y) surrounded cathepsin D as well as dextran (Fig. 2d,e, magnified images), indicating that the kinase domain is exposed on the cytoplasmic surface of endolysosomes.
Next, we compared the distribution of Kit(D814Y) and recycling endosomes using cholera toxin subunit B (CTXB; a recycling endosome marker) 36 . Kit's localization to recycling endosomes was significantly less than that to endolysosomes ( Fig. 2f,g), and a second marker of recycling endosomes, Rab11, confirmed this ( Supplementary Fig. 2c). These results indicate that Kit(D814Y) traffics to endolysosomes rather than recycling to the PM.    Kit D814Y moves to endolysosomes by kinase-mediated endocytosis. In RCM cells treated for 24 h with the kinase inhibitor PKC412, there was more Kit(D814Y) at the PM and less in endolysosomes (Fig. 3a). Treatment did not cause accumulation of Kit(D814Y) on either ER or Golgi. Following removal of inhibitor, Kit(D814Y) moved within 60 min from the PM, via endosomes, to endolysosomes ( Supplementary Fig. 3a). These results suggest that Kit's kinase activity is needed for endocytosis, but not for biosynthetic traffic to the PM. Figure 3b shows that 24-h treatment with an inhibitor of lysosomal proteases (NH 4 Cl) resulted in retention of Kit(D814Y) inside RCM cells, indicating that endocytosed Kit(D814Y) is targeted to lysosomes, and subsequently degraded. In the presence of cycloheximide, an inhibitor of protein synthesis, Kit(D814Y) was degraded within 5 h, whereas transferrin receptor (TfR), which undergoes recycling, was unchanged (Fig. 3c), suggesting that Kit traffics constitutively to lysosomes.
To study the role of clathrin in Kit(D814Y) endocytosis, we treated RCM cells with hypertonic sucrose, thereby blocking the formation of clathrin-coated pits 40 . Within 3 h, this reduced the protein levels of mature Kit(D814Y) and TfR without affecting the cytosolic protein p85 (Fig. 3d,e). Similar results were obtained from RCM cells treated with pitstop2, a selective inhibitor of CME 41 (Fig. 3f). To confirm that, with CME blocked, Kit and TfR were being processed by NCE, we treated cells for 3 h with pitstop2 plus filipin, or with sucrose plus filipin, which blocks NCE by disrupting lipid rafts 37,42 . Kit(D814Y) and TfR were then found at the PM, presumably protected from rapid degradation (Fig. 3g,h and Supplementary Fig. 3b). In cells treated with filipin alone, the protein levels and localization of Kit(D814Y) were unaffected (Fig. 3g,h and Supplementary Fig. 3b,c). Similar to sucrose and pitstop2, knockdown of the assembly polypeptide-2 a subunit (AP2a), which is required for CME 39 , decreased the protein levels of Kit(D814Y) and TfR ( Supplementary Fig. 3d). Thus, the major pathway for Kit(D814Y) endocytosis is CME.
In NCE, ubiquitination of receptors and other endocytic cargo is important for association with the ESCRT (endosomal sorting complexes required for transport) machinery, for rapid incorporation into lysosomes and for degradation 39,[43][44][45] . As Kit(D814Y) mainly undergoes CME not NCE, we hypothesized that it may not be fully ubiquitinated. When Kit(wt) was stimulated with SCF, in pt18 and R cells it became fully ubiquitinated, and was degraded within 30 min (Figs 3i and 6g; see also Fig. 3l), consistent with previous reports [31][32][33][34]45 . In sharp contrast, in RCM cells ubiquitination of Kit(D814Y) was much lower, regardless of the presence or absence of SCF. When CME was blocked, Kit(D814Y) became substantially ubiquitinated (Fig. 3j), presumably as it then trafficked via NCE. These results suggest that, in RCM cells, most Kit(D814Y) undergoes CME depending on its kinase activity, and accumulates on the cytoplasmic surface of endolysosomes, but not in a fully ubiquitinated state. In support of this, degradation of Kit(D814Y) was significantly slower than that of activated Kit(wt) in the presence of cycloheximide (Fig. 3k).
Next, to investigate the mechanism of endosome-to-lysosome trafficking of Kit, we knocked down tumour-susceptibility gene 101 (Tsg101), a component of ESCRT 43,44,46 . Figure 3l shows that in R cells treated with SCF for 30 min, Tsg101 knockdown inhibited the degradation of Kit(wt) but not of TfR (see also Supplementary Fig. 5e). These results suggest that Kit(wt) is incorporated into lysosomes in a manner that is dependent on ESCRT, similar to EGFR 39,43,46 . By contrast, knockdown decreased Kit(D814Y) in RCM cells (Fig. 3m). As ESCRT depletion enhances ESCRT-independent transport into lysosomes 46 , our results indicate that Kit(D814Y) is incorporated into lysosomes in a manner that is independent of ESCRT.
To investigate whether it was Kit(D814Y) that had activated Akt and STAT5, we inhibited Kit(D814Y) by PKC412 or siRNAmediated knockdown. Both treatments inhibited the activation of Akt and STAT5 (Fig. 4a,b). We further found that the PI3K p85 subunit was co-immunoprecipitated with Kit(D814Y), and that PKC412 blocked this association ( Fig. 4c; see also Fig. 1f,right). In this cell line, a PI3K inhibitor LY294002 suppressed Akt activation in a dose-dependent manner (Fig. 4d), as reported previously 7,19,47 . Thus, Kit(D814Y) activates Akt through association with PI3K.
Next, we tested whether Akt activation was required for autonomous proliferation. In RCM cells, the inhibitor Akti repressed the activation of Akt without affecting STAT5. It also caused a dose-dependent suppression of cell proliferation (Fig. 4e). Although the PI3K-Akt pathway has a role in RTK trafficking 35,48,49 , no redistribution was seen here on treatment with either LY294002 or Akti (Fig. 4f). This shows that the PI3K- Akt pathway has an essential role in cell proliferation but does not influence Kit(D814Y) trafficking.
We next suppressed STAT5 function by utilizing STAT5 inhibitor (STAT5i) or knockdown. Both treatments inhibited the proliferation of RCM cells (Fig. 4g,h). Knockdown had no effect on the activation of Akt. On the contrary, proliferation was unaffected by Erk1/2 inhibition with U0126 ( Supplementary  Fig. 4c). Taken together, these results suggest that Kit(D814Y) constitutively activates the PI3K-Akt pathway and STAT5, and that these activations are essential for autonomous proliferation. Activated RTKs, such as EGFR, Met, and PDGFR can transduce signals not only on the cell's outer membrane, but also on endocytic compartments 23,24,[49][50][51][52] . To examine whether endolysosomes serve as a platform for oncogenic Kit signalling, we purified endolysosomes from RCM cells by immunoprecipitation, using an anti-LAMP1 antibody. Figure 5a shows that Kit(D814Y), Akt and p85, but not STAT5, were found in the endolysosomal fraction. As active Akt is bound to phosphatidylinositol-3,4,5-triphosphate (PI(3,4,5)P 3 ), which is generated by PI3K, through its pleckstrin homology (PH) domain 3,53 , it is likely that Akt and Kit(D814Y)-PI3K become associated in endolysosomes. To test this hypothesis, we investigated whether PI(3,4,5)P 3 was generated on endolysosomes, by expression of a GFP-tagged Akt PH domain (PH-GFP) 54 . In RCM cells most PH-GFP was on vesicles but not at the PM, whereas PH(R25C)-GFP, a non-PI(3,4,5)P 3 -binding mutant, was in the cytoplasm (Fig. 5b). As PH-GFP could not be detected after immunofluorescence, we directly visualized endocytic compartments by fluorescent dextran. Figure 5c shows that they were significantly co-localized with PH-GFP. Furthermore, phosphorylated Akt was found on vesicles where Kit localized (Fig. 5d). These results suggest that Akt is activated by Kit(D814Y) predominantly on endolysosomes.
To examine whether the inhibition of Akt by BafA1 resulted from apoptosis, we performed immunoblotting for cleaved caspase-3, an apoptosis marker 20 . Figure 5g shows that BafA1 treatment for 24 h induced cleavage of caspase-3, but treatment for 3 h did not. The treatment for 3 h also inhibited Akt (Fig. 5g), indicating that the effect of BafA1 on Akt activity resulted from inhibition of Kit trafficking, but not from apoptosis.
Kit(wt) is also transported through the endocytic pathway after SCF binding 33,34 . To examine whether Kit(wt) also activated a PI3K-Akt pathway on endolysosomes, we treated R cells with BafA1 for 3 h and then stimulated them with SCF. Akt remained active up to 60 min after SCF stimulation, regardless of the presence or absence of BafA1 (Fig. 5h). Similar results were obtained from pt18 cells (Supplementary Fig. 5d). Figure 5i shows that in R cells stimulated with SCF for 30 min, p85 and Akt were absent from endolysosomes. Furthermore, in SCF-stimulated R cells, phosphorylated Akt and PH-GFP were not detected in endolysosomal vesicles (Fig. 5j,k). Also, accumulation of Kit(wt) on endosomes following Tsg101 knockdown did not activate Akt ( Supplementary Fig. 5e), suggesting that Kit(wt) activates the PI3K-Akt pathway transiently, presumably when bound to the PM. However, we could not exclude the possibility that Kit(wt) activates Akt soon after endocytosis.
Partially glycosylated Kit D814Y on the ER activates STAT5. Cytochalasin D, an inhibitor of endocytosis through actin depolymerization 35,36 , prevented the activation of Akt without influencing Kit(D814Y)'s kinase activity and p85's association with Kit(D814Y), but the treatment did not affect STAT5 activity ( Supplementary Fig. 5b,f). These results suggest that Kit(D814Y) activates STAT5 from a different subcellular compartment. Thus, we determined where in the cell Kit(D814Y) activates STAT5. First, we examined whether exocytic transport of Kit(D814Y) from the Golgi apparatus towards the PM was required. We treated RCM cells with monensin, which inhibits export from the Golgi by blocking intra-Golgi transport 56 . After 24-h treatment, Kit(D814Y) co-localized with the Golgi marker GM130 (Fig. 6a). Figure 6b shows that monensin treatment resulted in expression of partially glycosylated Kit(D814Y). Monensin had no effect on the autophosphorylation of Kit(D814Y), or on p85's association with Kit(D814Y) (Fig. 6b, left and Supplementary Fig. 5b). Moreover, monensin abolished Akt activation, presumably because Kit could no longer locate to endolysosomes. STAT5 activation was not affected (Fig. 6b, right), suggesting that Kit activates STAT5 on the Golgi and/or ER, not at the PM.
Brefeldin A (BFA) also inhibits protein export from the ER 26,30,58 . On 16-h BFA treatment, as with tunicamycin, partially glycosylated Kit(D814Y) accumulated on the ER and STAT5 became active (Fig. 6c,d). STAT5 was seen in the nucleus and on cytosolic reticular structures, but not at the PM (Fig. 6e, top). BFA treatment significantly enhanced the co-localization of Kit(D814Y) with STAT5 on the reticular structures (Fig. 6e,  bottom), further suggesting that Kit(D814Y) activates STAT5 on the ER. Figure 6f shows that the effects of BFA and tunicamycin  ARTICLE on Akt and STAT5 resulted from accumulation of Kit(D814Y) at the ER, not from apoptosis. In contrast to Kit(D814Y), in R and pt18 cells, SCF stimulation of Kit(wt) did not activate STAT5 (Fig. 6g), as previously described 27,29 . Accumulation of Kit(wt) on the ER did not affect STAT5 activation ( Fig. 6h and Supplementary Fig. 5g), indicating that STAT5 activation requires ER-localized Kit(D814Y). Taken together, these results suggest that newly synthesized partially glycosylated Kit(D814Y) on the ER activates STAT5 aberrantly. Treatment with monensin or BafA1 for 3 h inhibited Akt independently of apoptosis, and activated STAT5 transiently (Figs 5g and 6i; Supplementary Fig. 5h). These drugs did not enhance STAT5 activity in R cells (Fig. 6j and Supplementary  Fig. 5i). These results indicate that accumulated Kit(D814Y) on the Golgi and endosomes can activate STAT5 activation transiently. Thus, mechanisms of negative regulation for STAT5 may exist on the Golgi and endosomes.
The oncogenic role of Kit and its trafficking in rat and human cells. Next, we investigated whether the oncogenic role of Kit(D814Y) and its trafficking seen in RCM cells occur widely in neoplastic mast cells. The human and rat mast cell leukemia cell lines HMC-1 and RBL-2H3 endogenously express Kit with mutations in the kinase domain, these being Kit(D816V) and Kit(D817Y), respectively (Fig. 7a) 12,13 . In these lines, as with RCM, the kinase inhibitor PKC412 blocked Kit kinase activity and cell proliferation (Fig. 7b,c). Most mutant Kit was present as a complex-glycosylated form (Fig. 7d) that significantly co-localized with endolysosomal markers cathepsin D or LAMP1, rather than calnexin (ER) or GM130 (Golgi) (Fig. 7e,f). In HMC-1 cells, anti-Kit staining shows a pattern similar to that of expressed Kit(D814Y)-GFP but not to Kit(wt)-GFP ( Supplementary  Fig. 6a), confirming the endolysosomal Kit(D816V) staining.
We next examined Kit trafficking to endolysosomes in these cells. Figure 7g shows that PKC412 inhibited trafficking from the PM to endolysosomes, consistent with our findings on Kit(D814Y). Furthermore, for endocytosis, CME inhibition by sucrose reduced the protein levels of Kit but not of p85; in contrast, NCE inhibition by filipin did not affect the protein levels ( Fig. 7h and Supplementary Fig. 6b). This suggests that in HMC-1 and RBL-2H3 cells, as in RCM cells, Kit also undergoes CME. Taken together, these results suggest that, in HMC-1 and RBL-2H3, Kit's oncogenic role and intracellular trafficking are similar to those in RCM cells.
Kit signalling occurs in distinct compartments in human and rat cells. Next, we examined the subcellular location for signalling in HMC-1 and RBL-2H3. As in RCM, oncogenic activation of Akt and STAT5 was also seen and was prevented by PKC412 ( Fig. 8a and Supplementary Fig. 6c). Blockade of Kit trafficking to endolysosomes by BafA1 for 24 h also suppressed the activation of Akt but not of STAT5 ( Fig. 8b and Supplementary Fig. 6d), indicating that endolysosomal localization of Kit is essential for Akt activation. Blockade of ER export of Kit by BFA for 16 h suppressed Akt activation, but enhanced STAT5 activation (Fig. 8c). Collectively, these results suggest that in HMC-1 and RBL-2H3 cells, as in RCM cells, partially glycosylated Kit aberrantly activates STAT5 on the ER.

Discussion
In contrast to normal Kit, which signals from the PM, mutant oncogenic Kit signals from intracellular compartments (Fig. 8d). Newly synthesized, incompletely glycosylated mutant Kit initially localizes to the ER then activates STAT5. Subsequently, mutant Kit traffics to the PM through the Golgi along the secretory pathway and then immediately undergoes CME due to its kinase activity. It then accumulates in endolysosomes, but is not fully ubiquitinated. Mutant Kit is constitutively associated with PI3K, but the complex activates Akt only on the cytoplasmic surface of endolysosomes. In addition to previous reports 25,26,[30][31][32][33] , our study shows that the oncogenic signalling from mutant Kit is spatially distinct from normal signalling. By comparing RCM with R cells, we believe that our results show the significance of the unusual Kit localization seen in neoplastic mast cells.
Compartment-dependent oncogenic signalling by constitutively active growth factor receptors has occasionally been reported before. In multiple myeloma cells, deregulated FGF receptor 3 mutants accumulate at the Golgi, and initiate Erk1/2 and STAT1/3 signalling 22,27 . Oncogenic gp130 activates STAT3 not only on the ER but also on endosomes in hepatocytes 59 . In acute myeloid leukemia, the oncogenic Flt3 mutant Flt3-ITD activates STAT5 only on the ER 25-27 , as shown here for mutant Kit. Kit(D816V) directly tyrosine-phosphorylates STAT proteins in vitro 21 , suggesting that STAT5 might act as a substrate for ERlocalized mutant Kit. The Kit-STAT5 complex might escape from negative regulators such as protein tyrosine phosphatases on the ER but not on the Golgi, endosomes or the PM. The signalling platform may depend on the receptor and type of cell. These studies support our model that oncogenic Kit signalling occurs only on specific intracellular compartments in neoplastic mast cells. In leukaemia, mutant Kit participates in the PI3K-Akt pathway, in STAT5 activation, in the RhoA-Rho kinase pathway 47 and in activation of Src-like adaptor protein 60 , and there is great interest in further understanding the spatial organization of this signalling.
Mutant Kit associates with PI3K throughout its intracellular trafficking, but only activates Akt when on endolysosomes. The PI3K product PI(3,4,5)P 3 , which is required for Akt activation, is  ARTICLE believed to be generated solely at the PM 54,61,62 . However, PI(4,5)P 2 , a substrate of PI3K, is generated by phosphatidylinositide-4-phosphate-5-kinase on LAMP1-positive structures for proto-lysosome homeostasis 63,64 and plays a role in endosome-to-lysosome trafficking 52 . The following observation is also relevant: when SCF binds Kit at the PM, Kit activates inositol-5-phosphatase, which dephosphorylates PI(3,4,5)P 3 , and so terminates Akt signalling 10 . The significance is that, unlike normal Kit, mutant Kit cannot activate Akt at the PM. Further studies will be required to understand the mechanism by which the Kit-PI3K complex activates Akt selectively on endolysosomes.
In this study, CME stabilized mutant Kit on endolysosomes. This agrees with previous data showing that CME is essential for sustained signalling from Met, TGFbR and EGFR 24,[37][38][39]50 . However, mutant Kit was barely found in recycling endosomes, where the other receptors localize after ligand stimulation, and then recycle back to the PM. As very little SCF-bound normal Kit and mutant Kit recycle back to the PM 33,45 , CME must lead to at least two different pathways: the endolysosomal pathway leading to receptor destruction and the recycling endosome pathway, leading to receptor recycling. Ubiquitination of mutant Kit is insufficient for its rapid degradation. Defects in the tumour suppressor protein Cbl, an E3-ubiquitin ligase, also slow the degradation of Kit and Flt3, and prolong the activation of Akt, resulting in mastocytosis and myelo-proliferative disease 43,65,66 . When the CME pathway was blocked, mutant Kit was sorted into NCE and normally ubiquitinated. Neoplastic transformation might also involve an inappropriate sorting mechanism that recruits mutant Kit predominantly into CME not NCE. At present, we cannot explain why.
After SCF stimulation, Kit(wt) is transported into lysosomes in a manner dependent on ESCRT. This is consistent with previous reports that ESCRT transports ubiquitinated cargos 44,46,52 . ESCRT inhibition, however, does not cause accumulation of Kit(D814Y). Importantly, a recent study described that underubiquitinated cargos are incorporated into lysosomes in a manner independent of ESCRT 46 . Our results suggest that the mechanism of endosome-lysosome trafficking for mutant Kit is different from that for Kit(wt). Other mutant RTKs, such as EGFR(L858R) 23 , Met(D1246N) 24 and Flt3(D835Y) 25 , might also escape from ESCRT-dependent trafficking.
Imatinib, a Kit inhibitor used as a drug, is efficacious in most patients with gastrointestinal stromal tumours harbouring Kit mutations in the juxta-membrane region 4,11,14 . However, imatinib is ineffective for most human mast cell tumours with Kit mutations in the kinase domain 20,29,67 . Moreover, during imatinib treatment of the gastrointestinal tumours, mutations often develop in the Kit kinase domain 14 . Imatinib-resistant gastrointestinal tumour cells depend on activation of the PI3K-Akt pathway by Kit mutants for their proliferation, and so resemble mast cells 68 .
In this study, we showed that, in mast cells, mutant Kit trafficking to endolysosomes is critical for the activation of Akt. When acute myeloid leukemia cells with constitutively active Flt3 mutants, such as Flt3-ITD and Flt3(D835Y), are treated with statins (HMG-CoA reductase inhibitors), this reduces Akt activation by blocking receptor trafficking towards the PM 25 . Thus, Kit trafficking could be a new therapeutic target for mast cell tumours, and for imatinib-resistant gastrointestinal tumours. Combined therapy with an antibody and a kinase inhibitor seems attractive. As PKC412 treatment causes mutant Kit to remain at the PM through blocking the Kit kinase activity, it might enhance the activity of anti-Kit antibody 69 . Lapatinib, an EGFR kinase inhibitor, enhances the cytotoxic activity of trastuzumab, an anti-ErbB2 antibody in vivo 70 .
In conclusion, we show that compartment-dependent oncogenic Kit signalling is necessary for neoplastic mast cell proliferation. These findings provide new insights into the pathogenic role of Kit in neoplastic mast cell disorders. Improper trafficking and aberrant signalling are frequent features of constitutively active growth factor receptors, for which these data will shed light on the significance of the spatial organization of this oncogenic signalling.

Methods
Cells and stimulation. RCM (R cell, mutant Kit) and R cells were established from splenocytes of DO11.10 mice by repeated stimulation with ovalbumin peptides in vitro. The cell lines exhibited mast cell-like surface phenotype, c-Kit þ FceRI þ , and mast cell-like expression profiles of proteases. Moreover, RCM and R cells can secrete biologically active product on stimulation in vitro and in vivo. They do not express SCF. We were unable to find Kit(wt) by cDNA sequencing in RCM cells. For culture of R cells, we used culture supernatants from T-cell lines stimulated with an anti-T cell receptor antibody as a cytokine cocktail. R cells were cultured in 0.25% cytokine cocktail. RCM cells proliferated without the cocktail and developed tumours in vivo. HMC-1.2 (referred to henceforth as HMC-1), RBL-2H3 and pt18 cells were fro m Dr Hirohisa Saito and Dr Kenji Matsumoto (National Center for Child Health and Development), Dr Ko Okumura (Juntendo University) and Dr Ryo Goitsuka (Tokyo University of Science), respectively. These cells and RCM cells were cultured at 37°C in RPMI1640 medium supplemented with 10% fetal calf serum (FCS), penicillin, streptomycin and 50 mM 2-mercaptoethanol. HMC-1 cells were grown in suspension at 37°C in a-MEM containing 10% FCS, penicillin and streptomycin. For stimulation, cells were starved for at least 3 h and then treated with 50 ng ml À 1 recombinant mouse SCF (PeproTech).
Cell proliferation assay. Cells were cultured with inhibitors or cytokines for at least 16 h and then treated with [ 3 H]-thymidine deoxyribonucleotide (TdR) for 8 h.