Golgi retention and oncogenic KIT signaling via PLCγ2-PKD2-PI4KIIIβ activation in GIST cells

Most gastrointestinal stromal tumors (GISTs) develop due to gain-of-function mutations in the tyrosine kinase, KIT. We recently showed that mutant KIT mislocalizes to the Golgi area and initiates uncontrolled signaling. However, the molecular mechanisms underlying its Golgi retention remain unknown. Here, we show that protein kinase D2 (PKD2) is activated by the mutant, which causes KIT’s Golgi retention. In PKD2-inhibited cells, KIT migrates from the Golgi region to lysosomes and subsequently undergoes degradation. Importantly, delocalized KIT is unable to trigger downstream activation. In the Golgi area, KIT activates the PKD2-phosphatidylinositol 4-kinaseIIIβ (PKD2-PI4KIIIβ) pathway through phospholipase γ2 (PLCγ2) to generate a PI4P-rich membrane domain, where the AP1-GGA1 complex is aberrantly recruited. Disruption of any factors in this cascade results in KIT release from the Golgi region, indicating that these PKD2-related pathways are responsible for the Golgi retention of KIT. Our findings unveil the molecular mechanisms underlying KIT mislocalization and provide evidence for a new strategy for inhibition of oncogenic signaling.


101
To determine the cause of KIT retention in the Golgi area, we treated GIST-T1 cells (KIT Ex11 , 102 IMA-sensitive), which were established from a GIST patient 23 , with more than 20 compounds 103 followed by immunofluorescence confocal microscopic analyses. As shown in Figure S1A, most 104 compounds had no effect on KIT localization in the Golgi area, similarly to our previous report 105 that PP2 (SRC tyrosine kinase inhibitor) and AKT inhibitor VIII did not affect the mutant 106 localization 16 . Interestingly, upon 8-h treatment with CRT0066101 (CRT), a PKD inhibitor, KIT 107 migrated from the Golgi region to non-Golgi punctate structures ( Figures 1A and S1A), as 108 confirmed by golgin97 or Golgi matrix protein 130 kDa (GM130) staining ( Figures 1B and S1B).

110
Immunoblotting revealed that the inhibitor decreased KIT protein levels in a time-dependent 111 manner without affecting mutant autophosphorylation ( Figure 1C). Inhibition of PKD activity 112 was confirmed by measuring the decrease in phospho-PKD substrate levels ( Figure 1C, bottom 113 panels). Importantly, because KIT in CRT-treated cells moved out from the Golgi region, the 114 mutant signaling site, the activation of AKT and STAT5 also went down ( Figure 1C). We next 115 hypothesized that CRT treatment caused translocation of KIT to lysosomes. Indeed, KIT was 116 found in lysosomal-associated membrane protein 1-positive (LAMP1-positive) lysosomes in 117 PKD-inhibited cells ( Figure 1D). Phospho-KIT Y703 (pKIT Y703 ), which mainly represents the 118 mutant form, was also located in lysosomes of CRT-treated cells ( Figures 1E and S1C).

119
Furthermore, the decrease of KIT level was suppressed by NH 4 Cl or chloroquine, which blocks 120 8 lysosomal protein degradation ( Figure 1F). In contrast, in the presence of CRT, the two lysosomal 121 inhibitors did not affect the levels of the Golgi/TGN proteins, GM130, golgin97, and syntaxin 6 122 ( Figure S1D), supporting that PKD inhibition selectively causes migration of KIT to lysosomes 123 and subsequent degradation. Next, we investigated whether similar results were obtained from

132
We previously showed that mutant KIT in MCL passes through the Golgi/TGN and 133 accumulates on endosome-lysosome membranes [13][14][15] . As shown in Figure 1H, unlike KIT in GIST 134 cells, KIT was not degraded by CRT treatment in the MCL cell line, HMC-1.2. Phospho-STAT5 135 expression was diminished by the treatment, but this was not due to KIT downregulation. Taken 136 together, these results suggest that PKD activity is essential for Golgi KIT retention in GIST cells.

137
In PKD-inhibited GIST cells, KIT migrates to the PM from the Golgi region, and undergoes 138 EIPA-sensitive endocytosis 139 Next, we investigated whether the tyrosine kinase activity of KIT is required for the mutant 140 trafficking to lysosomes in PKD-inhibited cells. We previously showed that TKIs (IMA or 141 9 PKC412) decrease the KIT level in the Golgi/TGN and increase it in the PM 15,16 (see also Figure   142 2A). When GIST-T1 cells were treated with CRT plus IMA for 16 hours, KIT was primarily found

159
Human PKD consists of three members, PKD1, PKD2, and PKD3 (refs. 30-32 ). In GIST-T1 cells, 160 all PKD members were expressed and found in the Golgi area ( Figures 3A and S3A). Thus, we 161 performed a small interfering RNA-mediated (siRNA-mediated) knockdown to determine the 162 10 PKD member critical for KIT retention. Interestingly, PKD2 depletion had similar effects as CRT 163 treatment, but neither PKD1 knockdown nor PKD3 knockdown affected KIT levels ( Figure 3B).

164
Moreover, immunofluorescence revealed that KIT was found in lysosomes, but not in the Golgi 165 area in PKD2 knockdown cells ( Figure 3C). The movement of KIT to lysosomes is an indication 166 of the delocalization of the mutant from the Golgi region. Figure 3D shows that KIT localized in

169
Next, we analyzed the activation of PKD by immunostaining for phospho-PKD2 S876 170 (pPKD2 S876 ), a marker of the kinase activation 31,32 . As shown in Figure 4A, pPKD2 S876 was mainly 171 found in the Golgi area, where KIT is localized. PKD2 phosphorylation in  172 cells was observed in the Golgi area, similarly to GIST-T1 ( Figure S4A). We were unable to detect 173 pPKD2 S876 in the Golgi area in GIST48 cells by our immunofluorescence, probably due to low 174 phosphorylation levels.

175
Commercially available anti-pPKD1 S910 antibody has been shown to cross-react with 176 pPKD2 S876 (ref. 33 ; Figure S4B). Our immunoblotting data for  showed that the position 177 of the pPKD1 S910 band was consistent with that of PKD2 rather than PKD1 ( Figure S4C).

178
Considering that only PKD2 knockdown affected KIT levels, PKD2 is likely the major activated 179 PKD member in GIST-T1. Moreover, anti-pPKD1 S910 antibody stained the Golgi area ( Figure   180 S4D). Phospho-PKD2 levels were decreased by CRT treatment (Figures 4B and S4E). Taken 181 together, the effect of CRT on KIT was mainly due to PKD2 inhibition in GIST-T1 cells.

182
Next, we examined the association between KIT and PKD2 using a co-immunoprecipitation 183 assay. KIT was co-immunoprecipitated with PKD2, and KIT inhibition by IMA blocked this 184 11 association ( Figures 4C and 4D). These results suggest that KIT binds to PKD2 in a kinase 185 activity-dependent manner. We were unable to detect tyrosine phosphorylation in PKD2 ( Figure   186 4C) and KIT was probably not phosphorylated by PKD ( Figure 4E), suggesting that they do not 187 directly phosphorylate each other. To determine whether KIT is associated with PKD2 in the 188 Golgi area, we performed a proximity ligation assay (PLA). When anti-KIT and anti-PKD2 189 antibodies were added, Duolink PLA signals were detected in the perinuclear region, which was 190 stained with a Golgi marker lectin-HPA ( Figure 4F). In addition, KIT inhibition with IMA 191 decreased PKD2 activity and phospho-PKD substrate levels ( Figures 4G and S4F). Collectively,

192
these results suggest that KIT induces PKD2 activation in the Golgi area.

193
As shown in Figures 4H, S4G, and S4H, neither CRT nor IMA affected the localization of 194 PKD2 to the Golgi area. Furthermore, PKD2 was not dissociated from KIT by CRT ( Figure 4I),

195
indicating that its association with KIT is activity-independent. In the presence of CRT, lysosomal    retention. In contrast, knockdown of OSBP1 or ARF1 did not affect KIT levels ( Figure 6A). As

228
Similar to the case of PKD2, PI4KIIIβ inhibition caused migration of KIT from the Golgi region 229 to lysosomes ( Figure 6D). As shown in Figure 6E, KIT was associated with PI4KIIIβ in a manner 230 dependent on the mutant kinase activity. In support of these data, PI4KIIIβ localized together with 231 KIT and PKD2 in the Golgi area ( Figure 6F). Finally, we tested whether PI4KIIIβ lies downstream

256
Next, we performed a co-immunoprecipitation assay. Figure 7E shows that KIT was associated 257 with γ-adaptin and GGA1 in a kinase-dependent manner. Furthermore, GGA1 was co-258 immunoprecipitated with γ-adaptin, and the interaction was decreased by IMA ( Figure 7F).

259
Considering our finding that both γ-adaptin knockdown and GGA1 knockdown caused migration 260 of KIT from the Golgi region, the GGA1-AP1 interaction may be important for Golgi retention 261 of KIT. Next, we investigated whether PKD inhibition affected the association between KIT and 262 PKD2 effectors. Unexpectedly, PKD2, PI4KIIIβ, GGA1, and γ-adaptin were co-263 immunoprecipitated with KIT, even when the PKD2 activity was inhibited ( Figure 7G). In 264 contrast, PKD2 inhibition with CRT markedly decreased GGA1 association with γ-adaptin 265 ( Figure 7H). PKD activity is probably not essential for the association of KIT with PKD2, 266 PI4KIIIβ, AP1, and GGA1, but is specifically required for the AP1-GGA1 interaction. Taken

276
Indeed, PKD2 was not co-immunoprecipitated with KIT in these cells ( Figure S7C). Unlike in

282
indicating that the regulation of GGA1-AP1 interaction in leukemia is different from that in GIST 283 cells.

284
We recently reported that the FLT3-ITD accumulates in the Golgi area in AML cells, similarly

369
In conclusion, we show that KIT mutants are selectively retained in the Golgi area through