N822K- or V560G-mutated KIT activation preferentially occurs in lipid rafts of the Golgi apparatus in leukemia cells

Background KIT tyrosine kinase is expressed in mast cells, interstitial cells of Cajal, and hematopoietic cells. Permanently active KIT mutations lead these host cells to tumorigenesis, and to such diseases as mast cell leukemia (MCL), gastrointestinal stromal tumor (GIST), and acute myeloid leukemia (AML). Recently, we reported that in MCL, KIT with mutations (D816V, human; D814Y, mouse) traffics to endolysosomes (EL), where it can then initiate oncogenic signaling. On the other hand, KIT mutants including KITD814Y in GIST accumulate on the Golgi, and from there, activate downstream. KIT mutations, such as N822K, have been found in 30% of core binding factor-AML (CBF-AML) patients. However, how the mutants are tyrosine-phosphorylated and where they activate downstream molecules remain unknown. Moreover, it is unclear whether a KIT mutant other than KITD816V in MCL is able to signal on EL. Methods We used leukemia cell lines, such as Kasumi-1 (KITN822K, AML), SKNO-1 (KITN822K, AML), and HMC-1.1 (KITV560G, MCL), to explore how KIT transduces signals in these cells and to examine the signal platform for the mutants using immunofluorescence microscopy and inhibition of intracellular trafficking. Results In AML cell lines, KITN822K aberrantly localizes to EL. After biosynthesis, KIT traffics to the cell surface via the Golgi and immediately migrates to EL through endocytosis in a manner dependent on its kinase activity. However, results of phosphorylation imaging show that KIT is preferentially activated on the Golgi. Indeed, blockade of KITN822K migration to the Golgi with BFA/M-COPA inhibits the activation of KIT downstream molecules, such as AKT, ERK, and STAT5, indicating that KIT signaling occurs on the Golgi. Moreover, lipid rafts in the Golgi play a role in KIT signaling. Interestingly, KITV560G in HMC-1.1 migrates and activates downstream in a similar manner to KITN822K in Kasumi-1. Conclusions In AML, KITN822K mislocalizes to EL. Our findings, however, suggest that the mutant transduces phosphorylation signals on lipid rafts of the Golgi in leukemia cells. Unexpectedly, the KITV560G signal platform in MCL is similar to that of KITN822K in AML. These observations provide new insights into the pathogenic role of KIT mutants as well as that of other mutant molecules. Electronic supplementary material The online version of this article (10.1186/s12964-019-0426-3) contains supplementary material, which is available to authorized users.


Background
KIT is a member of the type III receptor tyrosine kinase (RTK) family that includes platelet-derived growth factor receptor A/B (PDGFRA/B), fms, and fms-like tyrosine kinase 3 (FLT3) [1][2][3]. It is known to participate in tyrosine phosphorylation signals at the PM, ensuring cell growth and survival in hematopoietic cells, mast cells, interstitial cells of Cajal, germ cells, and melanocytes [4][5][6].
Previously, we reported that in MCL, KIT D816V (human) and KIT D814Y (mouse) activate AKT and the signal transducer and activator of transcription 5 (STAT5) in EL and on the ER, respectively (Table 1) [24,25]. Furthermore, previous studies showed that in cells other than those of MCL, such as GIST and NIH3T3 cells, JM-mut or AL-mut accumulates on the Golgi apparatus, where it initiates oncogenic signals (Table 1) [26][27][28][29]. These studies raised important questions as to whether mutant KIT initiates signaling from intracellular compartments in other cancers such as AML, and whether the mutation status of KIT affects the platform for oncogenic signaling.
KIT mutations have been found in approximately 30% of CBF-AML patients who have chromosome aberrations [31][32][33]. Recent studies showed that active KIT mutations are correlated with a poor prognosis in AML patients [31,32]. The major activating KIT mutations are found at D816 and N822 (26 cases and 14 cases in 63 KIT mutation-positive patients, respectively) [33]. Although spatio-temporal analyses of KIT D816V signals have been performed [24,25,28], it is unclear whether the N822K mutation in leukemia affects KIT localization and the signal platform.
We then investigated the relationship between KIT N822K localization and tyrosine phosphorylation signals in Kasumi-1 cells (an AML cell line) that endogenously express KIT N822K . Furthermore, we examined whether KIT V560G in HMC-1.1 (MCL) caused signaling on the Golgi, ER, PM, or EL. In Kasumi-1, KIT is preferentially found in EL. Newly synthesized KIT in the ER traffics to the PM through the Golgi and subsequently relocates to EL through endocytosis in a manner dependent on its kinase activity. Our immunofluorescence assay, however, showed that KIT autophosphorylation preferentially occurs on the Golgi. Indeed, KIT N822K activates AKT, ERK, and STAT5 on the Golgi in Kasumi-1 cells. Moreover, lipid rafts in the Golgi play a role in KIT signaling. Interestingly, KIT V560G in MCL transduces signals in the Golgi in a similar manner to KIT N822K in AML but not to KIT D816V in MCL. Our study demonstrates that both KIT N822K and KIT V560G are mainly present in EL, but that their signal platform in leukemia cells is the lipid rafts of the Golgi. Furthermore, blockade of mutant KIT incorporation into the lipid rafts may provide a new strategy for suppression of growth signals in leukemia cells.

Antibodies
The sources of purchased antibodies were as follows:

Immunofluorescence confocal microscopy
Cells in suspension culture were fixed with methanol for 10 min at − 20°C or with 4% paraformaldehyde for 20 min at room temperature, then cyto-centrifuged onto coverslips. GIST-T1 cells were cultured on poly-L-lysine-coated coverslips and fixed as above.

Western blotting
Lysates prepared in SDS-PAGE sample buffer were subjected to SDS-PAGE and electro-transferred onto PVDF membranes. Immunodetection was performed by ECL (PerkinElmer, Waltham, MA). Sequential re-probing of membranes was performed after the complete removal of primary and secondary antibodies in stripping buffer (Thermo Fisher Scientific), or inactivation of peroxidase by 0.1% sodium azide. Results were analyzed with an LAS-3000 with Science Lab software (Fujifilm, Tokyo, Japan) or a ChemiDoc XRC+ with Image Lab software (BIORAD, Hercules, CA).

Flow cytometry
Leukemia cells were directly collected from a cell suspension by centrifugation. GIST-T1 cells were scraped from culture dishes and then centrifugated. Cells were stained with anti-KIT (104D2) and Alexa Fluor 488-conjugated anti-mouse IgG in PBS supplemented with 0.5% calf serum and 0.1% NaN 3 at 4°C for 1 h each. Stained cells were fixed with 4% paraformaldehyde for 20 min at room temperature. Flowcytometric data were obtained by FACSCalibur (BD Biosciences, Franklin Lakes, NJ), and results were analyzed with FlowJo software (Tomy Digital Biology, Tokyo, Japan).

Analysis of protein glycosylation
Following the manufacturer's instructions (New England Biolabs, Ipswich, MA), NP-40 cell lysates were treated with endoglycosidases for 1 h at 37°C. The reactions were stopped with SDS-PAGE sample buffer, and products were resolved by SDS-PAGE and immunoblotted.

Statistical analyses
For statistical analysis, experiments were repeated as three biological replicates. Differences between two or more groups were analyzed by the two-tailed Student's t-test or one-way analysis of variance followed by Dunnett's multiple comparison post-hoc test, respectively. All significant differences stated indicate a 5% level of probability.

KIT N822K and KIT V560G mislocalize to EL in leukemia cells
To investigate the sub-cellular localization of endogenous KIT, we performed confocal immunofluorescence microscopic analyses in pt18 (mouse mast cell line, KIT wild-type (WT)), Kasumi-1 (human AML, KIT WT/N822K ), SKNO-1 (human AML, KIT N822K/N822K ), and HMC-1.1 (human MCL, KIT WT/V560G ) (Fig. 1a). As previously described [24], most KIT WT localized to the PM in pt18 (Fig. 1b, left). In contrast, KIT accumulated in vesicular structures in Kasumi-1, SKNO-1, and HMC-1.1 (Fig.  1b). We then performed co-staining assays to identify these structures. In Kasumi-1, KIT was co-localized with TFR (transferrin receptor, endosome marker) and LAMP1 (lysosome marker) rather than with calnexin (ER marker) or giantin (Golgi marker) (Fig. 2a). Similarly, in HMC-1.1, KIT in vesicular structures was colocalized with TFR and cathepsin D (lysosome marker) ( Fig. 2b; Additional file 1: Figure S1A). By calculating Pearson's R correlation coefficients (Pearson's R) for KIT versus organelle markers, the intensity from KIT-TFR was significantly greater than that from KIT-calnexin, −giantin, and -LAMP1 in Kasumi-1 (Fig. 2c, left graph), suggesting that KIT mainly localizes to endosomes. The quantification showed that in HMC-1.1, KIT was colocalized with TFR to a similar extent as with cathepsin D (Fig. 2c, right graph). In both types of cells, KIT was colocalized with EL markers rather than with ER/Golgi markers. We previously showed that in MCL and GISTs, KIT mutants are normally complexglycosylated in the Golgi [24,26]. To test for the KIT glycosylation state in Kasumi-1 and HMC-1.1, we treated KIT with endoglycosidase H, which digests immature high-mannose forms of KIT but not mature complex-glycosylated forms. Figure 2d shows that most KIT in these leukemia cells was in a complex-glycosylated form, similar to normal KIT [24,28,29]. KIT shifted to a non-glycosylated form following the complete digestion of N-linked glycans by peptide-Nglycosidase F. In SKNO-1, as with Kasumi-1 and HMC-1.1, KIT was complex-glycosylated (Additional file 1: Figure S1B). Pearson's R quantification from immunofluorescence images showed that in SKNO-1, KIT localized to endosomes to a similar extent as to lysosomes and it was found in EL rather than on ER/Golgi (Additional file 1: Figure S1C & D). These results suggest that complex-glycosylated KIT accumulates in EL in leukemia cells but not in the early secretory compartments. KIT mutants autonomously migrate from the PM to EL through endocytosis in a manner dependent on their kinase activity Next, we examined the role of KIT kinase activity on cell proliferation, growth signals, and KIT localization. As previously reported [36][37][38][39], Kasumi-1 and HMC-1.1 proliferate autonomously, and KIT tyrosine kinase inhibitors (TKIs), such as imatinib and PKC412, suppressed cell proliferation in a dose-dependent manner (Fig. 3a).
Immunoblotting showed that phosphorylation of KIT, AKT, ERK, and STAT5 occurred in the absence of TKIs, and PKC412 and imatinib reduced these phosphorylations (Fig. 3b), confirming that KIT activates AKT, ERK, and STAT5 in Kasumi-1 and HMC-1.1. Next, we investigated the localization of KIT in TKI-treated cells by immunofluorescence and flow cytometry. In Kasumi-1 cells treated with TKIs, KIT localized more in PM (outside ER staining, Fig. 3c & d) and less in endosomes ( Fig. 3e & f). Similar results were seen with HMC-1.1 and SKNO-1 (Additional file 1: Figure S1E). Collectively, in these leukemia cells, newly synthesized KIT in the ER moves to the PM along the secretory pathway and subsequently traffics to EL through kinase activitydependent endocytosis. In addition to our previous findings that KIT D816V and KIT ⊿560-578 in the PM are increased by TKI treatment [24,26,40], these results suggest that retention in the PM by TKIs is a ubiquitous feature of KIT mutants.

Autophosphorylation of KIT N822K and KIT V560G predominantly occurs on the Golgi in leukemia cells
We next examined the site of KIT activation in leukemia cells. To determine the signal platform for KIT, we immuno-stained for phospho-tyrosine residues in the kinase domain which would indicate KIT activation [7-9, 26, 27]. In Kasumi-1 cells, phospho-KIT Y703 (pKIT [Y703]) was clearly detected (Fig. 4a)  [Y703] was restricted to the perinuclear region in Kasumi-1 (Fig. 4a, top panels, arrowheads). Similar to KIT in Kasumi-1, KIT in HMC-1.1 was found in the perinuclear compartment (Fig. 4b).
Perinuclear KIT autophosphorylation was colocalized with GM130 (Golgi) rather than with PDI (ER), TFR (endosomes), or LAMP1 (lysosomes) ( Fig. 4c; Additional file 1: Figure S2A). Similar results were obtained with SKNO-1 (Additional file 1: Figure S2B & C). These results suggest that in leukemia cells, activation of KIT N822K and KIT V560G occurs predominantly on the Golgi although KIT itself is found mainly in EL.

KIT N822K and KIT V560G mainly activate downstream molecules on the Golgi in leukemia cells
We then examined whether KIT activated downstream molecules on the Golgi in leukemia cells. To resolve this question, we used inhibitors of intracellular trafficking, such as brefeldin A (BFA), 2-methylcoprophilinamide (M-COPA) (inhibitors of ER export to the Golgi) [27,34,35,41], monensin (an inhibitor of intra-Golgi trafficking) [26,42], and bafilomycin A1 (BafA1, an inhibitor of endosome-to-lysosome trafficking) [24,43]. In Kasumi-1 and HMC-1.1, treatment with BFA or M-COPA significantly increased colocalization of KIT with an ER marker, calnexin (Fig. 5a & Additional file 1: Figure S3A), confirming that the treatment inhibited ER export of KIT. Immunoblotting showed that KIT shifted to a lower molecular weight in BFA-or M-COPA-treated cells because of a defect in full glycosylation on the Golgi apparatus (Fig. 5b & c, top panels). KIT on the ER was dephosphorylated and unable to activate downstream molecules (Fig. 5b & c). Previous studies showed that a major target of BFA/M-COPA is Golgi-specific BFA-resistance guanine nucleotide exchange factor 1 (GBF1) that plays a role in the secretory pathway through activation of ADP ribosylation factor 1 (ARF1) [34,44,45]. Interestingly, knockdown (KD) of ARF1 and GBF1 with siRNAs did not cause a defect in full glycosylation of KIT or inhibition of signaling (Additional file 1: Figure S3B). Since BFA and M-COPA bind not only to the ARF1-GBF1 complex but also to other complexes [44,45], the blockers affect KIT trafficking in a manner independent of ARF1-GBF1 inhibition in the leukemia cells used in this study. Further study will be required for understanding how the inhibitors block KIT trafficking from the ER. Fig. 5d shows that inhibition of the Golgi export of KIT through blocking intra-Golgi trafficking did not suppress KIT signaling, suggesting that Golgi-localized KIT is sufficient for oncogenic signaling in Kasumi-1 and HMC-1.1. As shown in Additional file 1: Figure  S3C, KIT signals remained in BafA1-treated cells, indicating that endosome-to-lysosome trafficking is unnecessary for downstream activation. Taken together, these results suggest that the Golgi apparatus serves as the platform for KIT activation in leukemia cells. To support this conclusion, we stained for phospho-AKT (pAKT), pERK, and pSTAT5. As shown in Fig. 5e, these phosphorylations were found at the Golgi region in Kasumi-1 cells. Compared with pAKT, total AKT was barely seen in the Golgi (Additional file 1: Figure S3D, upper panels). In Kasumi-1, only part of AKT may be activated by KIT. Furthermore, total ERK and STAT5 were distributed in the Golgi region (Additional file 1: Figure S3D). These results support our hypothesis that KIT activates these downstream molecules on the Golgi in leukemia cells. In HMC-1.1, AKT, ERK, STAT5, and their phospho-forms showed a diffuse distribution compared with those in Kasumi-1 (Additional file 1: Figure  S3E). Since pAKT, pERK, pSTAT5, though small, were found in the Golgi region, they could be activated on the Golgi and subsequently move elsewhere.
Recently, we showed that in GISTs, KIT on the ER is dephosphorylated by protein tyrosine phosphatases (PTPs) [27]. We then considered the role of PTPs in KIT inactivation in the ER in leukemia cells. In M-COPA-pretreated Kasumi-1, a 3-h treatment with a PTP inhibitor (sodium orthovanadate, Na 3 VO 4 ) [46] restored pKIT [Y703], resulting in downstream reactivation (Fig.  5f, left), indicating that in Kasumi-1, PTPs play a role in KIT inactivation in the ER. In M-COPA-treated HMC-1.1, pKIT [Y703] and pSTAT5 were recovered by Na 3 VO 4 treatment, but AKT and ERK did not become (See figure on previous page.) Fig. 3 KIT migrates to EL through endocytosis in a manner dependent on their kinase activity. a Kasumi-1 or HMC-1. 1

cells were cultured for 24 h in the presence of KIT kinase inhibitors (imatinib, open circles; PKC412, closed circles). The graphs show the levels of [ 3 H] thymidine deoxyribonucleotide (TdR) incorporation into cells (counts per minute, c.p.m., growth index) at the indicated inhibitor concentrations. Results are means ± s.d. (n = 3). b
Kasumi-1 or HMC-1.1 cells were treated for 4 h with 1 μM PKC412 or 1 μM imatinib. Lysates were immunoblotted for KIT, phospho-KIT Y703 (pKIT Y703 ), AKT, pAKT, STAT5, pSTAT5, ERK, and pERK. c-f Kasumi-1 cells were treated for 12 h with 1 μM PKC412 (PKC) or 1 μM imatinib (IMA). c Cells were immunostained with anti-KIT (green) and anti-calnexin (ER marker, red). Confocal immunofluorescence images are shown. Insets show magnified images of the PM region. Bars, 10 μm. d Cell surface KIT levels determined by flow cytometry are shown. Non-permeabilized cells were stained with anti-KIT extracellular domain antibody. Green histogram, with KIT inhibitor treatment; white histogram, no KIT inhibitor; gray histogram, no anti-KIT antibody control. e Cells were immunostained with anti-KIT (green) and anti-TFR (endosome marker, red). Confocal immunofluorescence images are shown. Bars, 10 μm. f Pearson's R correlation coefficients were calculated by analyzing the intensity of KIT vs. TFR. Results are means ± s.d. (n = 22~29). ***P < 0.001. Note that these inhibitors lowered KIT in vesicular structures and increased KIT in the PM active on PTP inhibition (Fig. 5f, right). Negative regulation of AKT and ERK may differ among cell types. Taken together, these results suggest that ER-localized KIT is inactivated by PTPs. PTP1B, Src homology 2 containing PTP-1 (SHP-1), and SHP-2 have been reported as PTPs for KIT and FLT3 RTKs [47][48][49][50]. Thus, we knocked down these PTPs and treated cells with M-COPA to investigate the key PTPs for KIT in the ER. Additional file 1: Figure S4 shows that in M-COPA-treated cells, pKIT [Y703], pAKT, and pERK were not restored by PTP1B or SHP1/2 KD, suggesting that these PTPs are not responsible for this dephosphorylation in the ER. Interestingly, PTP1B but not SHP1/2 KD partially rescued pSTAT5 in M-COPA-treated cells (Additional file 1: Figure S4A, arrows). Although we were unable to identify KIT phospho-tyrosine sites that are dephosphorylated by PTP1B in this study, PTP1B in the ER may play a role in inactivation of the KIT-STAT5 axis.
SKNO-1 cells were similar to Kasumi-1 in phosphoregulation of KIT in intracellular compartments (Additional file 1: Figure S5A & B). However, AKT, ERK, and STAT5 were not activated by KIT N822K (Additional file 1: Figure S5C). SKNO-1 requires GM-CSF for proliferation, but the cytokine did not affect the activation of KIT, AKT, ERK, or STAT5 (Additional file 1: Figure  S5C, right panels). AKT and ERK were not found in specific compartments in SKNO-1 in the presence or absence of GM-CSF (Additional file 1: Figure S5D). At present, we are unable to find downstream molecules that are activated by KIT N822K in SKNO-1 cells. We will investigate the role of KIT N822K in SKNO-1 growth in the near future.

Lipid rafts play a key role in KIT signaling, which occurs on the Golgi apparatus
Recent studies showed that sphingomyelin-enriched membrane microdomains (lipid rafts) in the Golgi are needed for activation of an innate immunity molecule, STING [51,52]. Formation of normal lipid microdomains is inhibited by N-hexanoyl-D-erythro-sphingosine (cer-C6) through producing short chain sphingomyelin that disrupts the lipid order [51,53,54]. Figure 6a shows  1 cells (b) were immunostained for KIT (green), phospho-KIT Y703 (pKIT Y703 , red or green) together with GM130 (Golgi marker, blue), PDI (protein disulfide isomerase, ER marker, red), TFR (endosome marker, red), or LAMP1 (lysosome marker, red). Insets show magnified images. Bars, 10 μm. c Pearson's R correlation coefficients were calculated by analyzing the intensity of pKIT Y703 vs. organelle markers. Results are means ± s.d. (n = 12~21). ***P < 0.001. Note that pKIT Y703 was colocalized with GM130 rather than with PDI, TFR, or LAMP1 both in Kasumi-1 and HMC-1.1 cells that in Kasumi-1, cer-C6 treatment lowered the protein levels of KIT and inhibited KIT autophosphorylation and the activation of AKT, ERK, and STAT5 in a dosedependent manner. The treatment did not decrease but rather increased KIT on the Golgi (Fig. 6b & c). These results suggest that KIT N822K and KIT V560G require lipid rafts for their stability and activation in the Golgi in these leukemia cells.
Finally, we asked whether lipid rafts play a role in oncogenic signaling by all KIT mutants. GIST-T1 cells (KIT ⊿560-578 ) grow in a manner dependent on KIT signaling on the Golgi, whereas HMC-1.2 (mast cell leukemia, KIT V560G/D816V ) requires pAKT on EL and pSTAT5 on the ER [24][25][26][27] (Additional file 1: Figure  S6A & Table 1). In both cell types, TKIs increased PM distribution of KIT mutants (Additional file 1: Figure  S6B), supporting our data obtained with Kasumi-1 that mutant KIT localizes to intracellular compartments in a manner dependent on its kinase activity. In GIST-T1, cer-C6 inhibited the phosphorylation of KIT and downstream molecules (Fig. 6d). Unlike KIT in leukemia cells, that in cer-C6-treated GIST-T1 assumed an immature glycosylated form (Fig. 6e), indicating that KIT is complex-glycosylated in GIST after reaching lipid rafts. Similar to the results using Kasumi-1, the treatment did not decrease but rather increased KIT on the Golgi (Additional file 1: Figure S6C). On the other hand, in HMC-1.2, cer-C6 did not have an inhibitory effect on Golgi export of KIT D816V and growth signals ( Fig. 6f; Additional file 1: Figure S6D). Therefore, lipid rafts play a critical role in KIT signaling that occurs on the Golgi.

Discussion
In this study, we demonstrate that in leukemia cells, localization of KIT N822K and KIT V560G is clearly different from that of KIT WT in normal cells. We provide evidence that after secretory trafficking to the PM, these mutants localize to EL through kinase activity-dependent endocytosis. However, they are autophosphorylated predominantly on the Golgi apparatus, where they activate downstream molecules, such as AKT, ERK, and STAT5. Lipid rafts play a key role in KIT signaling on the Golgi. Moreover, ER-localized KIT is dephosphorylated by PTPs (Fig. 7). This phospho-regulation of KIT is similar to that in GISTs. Our observations show that in some cases, receptor mutants can initiate signals from the Golgi even if they are mainly present in EL (Table 1).
Recently, we reported that in MCL, AL-mut, such as KIT D816V (human) and KIT D814Y (mouse), accumulates in EL (Table 1) [24,25]. Unlike JM-mut, AL-mut in MCL activates AKT and STAT5 on EL and the ER, respectively. Previous studies showed that transfected AL-mut in cell lines other than MCL, such as NIH3T3 and GISTs, localizes to the Golgi, where it initiates oncogenic signals on the Golgi apparatus (Table 1) [26,28]. The host cell environment rather than the KIT mutation status may determine the mutant's subcellular localization. Considering that JM-mut in MCL activates a downstream pathway on the Golgi, these studies suggest that AL-mut activates downstream molecules on EL and the ER only when it is expressed in an MCL environment. Moreover, there is great interest in further investigation as to whether KIT D816V in AML causes growth signaling on the ER, Golgi, or EL. Further studies will be required for understanding the mechanism responsible for the difference in the signal platforms of KIT.
In innate immune cells, STING binds to exogenous DNA fragments in the ER, then moves to lipid rafts of the Golgi [51,52]. This requires palmitoylation of its cysteine residues for migration to the lipid rafts, where it can activate the TBK1-IRF3 pathway [51]. MET tyrosine kinase is activated and palmitoylated on its extracellular domain at the Golgi [69,70], and lipid rafts contribute to receptor distribution and stability [70,71]. These studies raise the interesting possibility that incorporation of KIT into the lipid rafts of the Golgi is involved in protein acylation. Investigation as to whether palmitoylation is necessary for KIT signaling on the Golgi is now under way.
PTPs play a role in inactivation of KIT in the ER. Our loss of function study showed that PTP1B, SHP-1, and SHP-2 are not major PTPs for KIT dephosphorylation in the ER, suggesting the role of other PTPs in KIT inactivation. In addition, negative regulators of KIT and downstream molecules could be abundant in intracellular compartments other than the Golgi. Further analyses of the localization and functions of these negative regulators should explain how KIT signaling is inactivated in the ER, PM, and EL both in leukemia and GISTs. In (See figure on previous page.) Fig. 6 Lipid rafts have a role in KIT signaling at the Golgi apparatus. a-c Kasumi-1 or HMC-1.1 cells were treated with 0~40 μM cer-C6 for 8 h (for inhibition of normal lipid raft formation). a Lysates were immunoblotted. b Cells treated with 40 μM cer-C6 for 8 h were immunostained for KIT (green), giantin (Golgi marker, red), or GM130 (Golgi marker, blue). Bars, 10 μm. c Pearson's R correlation coefficients were calculated by analyzing the intensity of KIT vs. giantin (Kasumi-1) or GM130 (HMC-1.1). Results are means ± s.d. (n = 16~22). **P < 0.01. d & e GIST-T1 cells were treated with 0~10 μM cer-C6 for 10 h. d Lysates were immunoblotted. e Lysates were treated with PNGase F or endoglycosidase H then immunoblotted with anti-KIT. CG, complex-glycosylated form; HM, high mannose form; DG, deglycosylated form. f HMC-1.2 cells were treated with 0~40 μM cer-C6 for 8 h, then immunoblotted Fig. 7 Model of mutant KIT trafficking and signals on the Golgi in leukemia cells. Newly synthesized mutant KIT (KIT N822K or KIT V560G ) in the ER traffics to the PM through the Golgi apparatus. They are normally complex-glycosylated in the Golgi. After reaching the PM, mutant KIT immediately undergoes endocytosis in a manner dependent on its kinase activity, then accumulates in EL. However, its full autophosphorylation mainly occurs on the Golgi, where it causes downstream activation. Lipid rafts play a role in KIT signaling. ER-localized KIT is inactivated by PTPs other words, the study will reveal the mechanism of deregulation of RTK on signal platforms.
In this study, KIT mutants were retained in the PM in TKI-treated cells, since TKIs inhibit endocytosis of KIT, which depends on the kinase activity. Furthermore, TKIs increase PM localization of EGFR in lung cancers and PDGFRA/KIT in GISTs [26,63,72]. A previous report showed that TKI treatment increases the FLT3-ITD PM level in AML, which enhances the effect of FLT3-directed immunotherapy in mice [73]. Moreover, anti-KIT antibody is efficacious for suppression of the autonomous growth of GIST cells [74,75]. From a clinical point of view, combined therapy with anti-RTK antibodies and TKIs seems attractive.
Small molecule TKIs and antibodies against RTKs have been developed for suppression of proliferative signals in cancers. In this study, blockade of the ER export of KIT with BFA/M-COPA decreased KIT's autophosphorylation in leukemia cells. Together with the results previous reports [25,27,76,77], our findings may offer a trafficking blockade of receptor mutants as a third strategy for inhibition of oncogenic signaling.

Conclusions
In conclusion, we show that in leukemia cells, N822Kand V560G-mutated KIT can initiate growth signals in lipid rafts of the Golgi apparatus. These observations provide new insights into the pathogenic role of KIT mutants as well as into that of other mutant signaling molecules. Moreover, from a clinical point of view, our findings offer a new strategy for leukemia treatment through that blocks the incorporation of KIT mutants into the lipid rafts of the Golgi.