MT-KIT is primarily localized in the Golgi complex in GIST cell lines and tissues
To detect the cellular localization of MT-KIT (GIST cells: GIST430 and GIST882) and wild-type KIT (colon cancer cells: DLD-1 and Colo320DM), immunocytochemistry (ICC) analysis was performed. In GIST cells, fluorescence signal of MT-KIT was barely observed at the plasma membrane (PM) without permeabilization, whereas a strong perinuclear fluorescence signal of MT-KIT was detected following permeabilization (Fig. 1A). In colon cancer (CC) cells, wild-type KIT (WT-KIT) was detected only in the PM, regardless of permeabilization (Fig. 1B). The perinuclear MT-KIT did not colocalize with the ER marker calnexin, but clearly colocalized with the cis- (GM130), medial- (mannosidase ll), and trans-Golgi (Golgin-97) markers (Fig. 1C and D). Deglycosylation analysis of MT- and WT-KIT further revealed that KIT localized in the ER (endoglycosidase H-sensitive) was barely detected in GIST and CC cells, which supports the hypothesis that MT-KIT in GISTs is primarily localized in post-ER organelles (Fig. 1E). Golgi localization of MT-KIT was validated in tissues from 42 patients with GIST. Immunohistochemistry (IHC) analysis showed that perinuclear MT-KIT expression was detected in 29 patients (69%) (Supplementary Fig. S1 and Supplementary Table S2).
Although ICC analysis showed Golgi localization of MT-KIT, membranous MT-KIT expression in GISTs and CCs was quantitatively measured by biotin labeling and western blot analysis of PM proteins. Since the membranous MT-KIT expression in GIST cells was too low, the amount of GIST whole-cell lysate used for biotin-labeled protein pulldown was doubled than that in CCs. The ratios between membranous KIT levels compared with whole-cell lysates were approximately 0.40 and 0.20 for GIST430 and GIST882 cells, respectively, while those for CC cells were approximately 1.0 (Fig. 1F and G). The surface expression of MT-KIT in GIST430 cells was 2-fold higher than in GIST882 cells. Considering the heterozygous mutation in GIST430 cells, the proportion of GIST430 cells was relatively low. RT-PCR analysis in GIST430 cells revealed that mRNA expression of MT-KIT was 1.8-fold higher than that of WT-KIT, indicating that GIST430 cells predominantly expressed MT-KIT mRNA (Supplementary Fig. S2).
Membranous MT-KIT is not involved in GIST tumorigenesis
Despite a small fraction of the total MT-KIT, weak membranous KIT expression was detected in GIST cells (Fig. 1F). Therefore, a ligand tolerance assay was performed to test if the membranous MT-KIT responds to its ligand, the stem cell factor (SCF). As previously reported [10], SCF treatment rapidly downregulated WT-KIT expression in CC cells. However, MT-KIT levels barely changed in GIST cells following SCF treatment (Fig. 1H). The fate of membranous MT-KIT was tracked in a time-dependent manner to understand its role in GIST tumorigenesis. GIST882 cells harboring the homozygous KIT mutation were used for analysis. GIST882 cells were labeled with a fluorescent-dye-conjugated KIT antibody and examined over time using a confocal microscope. Membranous MT-KIT was internalized within 15 minutes and disappeared within 1 hour. Colocalization was not observed between MT-KIT and a Golgi complex marker, GM130, until MT-KIT disappeared, which excludes the possibility that Golgi retention of MT-KIT is mediated by the endosome-to-trans-Golgi retrieval pathway, a reported Golgi retention mechanism of the G-protein-coupled receptor [11] (Supplementary Fig. S3A). Next, the involvement of PM quality-control (PMQC) in the disappearance of MT-KIT was investigated. The labeling experiment was repeated to evaluate the colocalization of MT-KIT with an early endosome marker, EEA1, or a lysosomal marker, LAMP1, involved in the PMQC processes [12]. Colocalization of membranous MT-KIT with EEA1 and LAMP1 was clearly observed 15 and 30 minutes after tracking, and disappeared after 1 hour (Supplementary Fig. S3B and C).
To quantitatively validate the ICC results, western blot analysis of membranous KIT was performed in GISTs and CCs. In GISTs, the remaining membranous MT-KIT levels was measured through biotin-labeled protein pulldown assay and western blot at the indicated time-points. Bafilomycin A1 (BafA1), a lysosomal inhibitor, was used to validate lysosome-mediated protein degradation. CCs were included as controls to determine the original stability of WT-KIT. Since WT-KIT is mostly expressed in the PM, the remaining WT-KIT abundance was measured at the indicated time-points after cycloheximide (CHX) treatment. After 4 hours, >50% of the original WT-KIT level remained (Supplementary Fig. S4A), whereas in GIST882 cells, 8% remained. MT-KIT degradation was efficiently blocked by BafA1 treatment (Supplementary Fig. S4B). These findings suggest that membranous MT-KIT is hardly involved in GIST tumorigenesis due to its low membranous protein level caused by Golgi retention and plasma membrane quality control (PMQC).
Neither PM nor ER is the site for MT-KIT downstream signaling activation
Because of low protein levels and rapid degradation of membranous MT-KIT, we hypothesized that the Golgi-localized MT-KIT is the primary contributor to sustained activation of downstream signaling in GISTs. We first aimed to exclude the possibility that ER- or PM-localized MT-KIT initiates downstream signaling. Brefeldin A (BFA) was used to block ER-to-Golgi trafficking, thereby inducing ER retention of MT-KIT. ICC analysis showed that BFA led to ER retention of MT-KIT (Supplementary Fig. S5). Western blot analysis showed that BFA dramatically reduced phospho-KIT, phospho-AKT, and phospho-ERK levels, indicating that ER-retained MT-KIT was insufficient to activate downstream signaling (Fig. 2A). The involvement of membranous MT-KIT in downstream signaling was also investigated using 30N12, an inhibitor of trans-Golgi-to-PM trafficking [13]. To validate the effects of 30N12, western blot and ICC analysis of WT-KIT was performed in SCF-treated CC cells treated with or without 30N12. Treatment with 30N12 efficiently induced WT-KIT retention in the Golgi complex (Supplementary Fig. S6A), and reduced phospho-KIT, phospho-AKT, and phospho-ERK levels, that were upregulated by SCF treatment (Supplementary Fig. S6B). However, in GIST cells, 30N12 treatment did not change the phosphorylation levels of effector molecules, indicating that PM-localized MT-KIT was also insufficient to activate downstream signaling (Fig. 2B).
The Golgi complex is where MT-KIT delivers downstream oncogenic signaling
To investigate whether Golgi-retained MT-KIT primarily contributes to downstream oncogenic signaling activation, we investigated the interaction of MT-KIT with P85 and GRB2, the upstream effectors of the PI3K/AKT and MAPK/ERK pathways, respectively. Immunoprecipitation assays were performed using GIST cell lysates incubated with IgG (control) or KIT antibody. Western blot analysis showed that the endogenous MT-KIT bound to both P85 and GRB2 (Fig. 2C). The interaction of MT-KIT with GRB2 and P85 was further verified through ICC analysis. Since there are no commercial antibodies for ICC analysis of P85 and GRB2, HA-tagged P85 and GRB2 expression vectors were constructed. After vector transfection into GIST cells, cellular localization of MT-KIT, HA-P85, and HA-GRB2 were examined. GRB2 and P85 clearly showed colocalization with MT-KIT and GM130 (Fig. 2D and E). The localization of GRB2 and P85 after SCF treatment in CC cells was further investigated to compare the PI3K/AKT and MAPK/ERK pathway activation mediated by WT-KIT and MT-KIT. In contrast to GIST cells, ICC analysis showed that SCF treatment recruited P85 and GRB2 to the peri-PM region, where activated WT-KIT was localized (Supplementary Fig. S7). These data showed that MT-KIT directly initiates downstream signals from the Golgi complex, whereas WT-KIT initiates signals from the PM.
BLZF1, a Golgi-resident protein, is required for MT-KIT expression in the Golgi complex
Since the Golgi complex is a hub for constant protein trafficking, a complementary mechanism is necessary for the retention of MT-KIT. Therefore, we aimed to identify a possible MT-KIT-binding partner from the Golgi-related proteins that could tether MT-KIT to the Golgi complex during trafficking. Since there are only a few reported Golgi-related proteins [14-16], all possible candidates with commercially available antibodies were selected. Western blot was performed on ten Golgi proteins (GBF1, GM130, Golgin97, TGN38, GRASP65, GRASP55, BLZF1, STX3, STX6, and GOLPH3) in small-cell lung cancer (SCLC), leukemia, colon cancer, and GIST cell lines expressing WT-KIT or MT-KIT. All Golgi proteins showed variable levels in the cell lines, except for BLZF1 that selectively showed high protein levels in GIST cells (Fig. 3A). Moreover, immunoprecipitation with KIT antibody revealed that MT-KIT specifically bound to BLZF1 (Fig. 3B). The colocalization of MT-KIT and BLZF1 in the Golgi complex of GIST cells was validated by ICC analysis (Fig. 3C). On the other hand, colocalization in the Golgi complex of CC and leukemia cell lines (Colo320DM and HMC-1) was not detected (Supplementary Fig. S8). BLZF1 knockdown with short hairpin RNA (shRNA) dramatically downregulated MT-KIT expression, as demonstrated by ICC and western blot analyses (Fig. 3C-E). BLZF1 knockdown strongly inhibited cell growth in both imatinib-sensitive GIST430 and GIST 882 cells, and in imatinib-resistant GIST430-V654A and GIST48 cells (Fig. 3F and Supplementary Fig. S9). These findings collectively suggest that BLZF1 is required for Golgi retention and stable expression of MT-KIT, and consequently for GIST cell growth.
ATF6, a pro-survival ER stress sensor, is constantly activated in GISTs
It remains unclear how MT-KIT bypasses ERQC to reach the Golgi complex. As shown in Fig. 1E, MT-KIT in GISTs was primarily in the post-ER form, which indicates that MT-KIT is folded enough to avoid ER accumulation and stress induction. Since GIST cells grow in the muscle layer, they are constantly exposed to nutrient deficiency and hypoxia that also cause ER stress. Therefore, we hypothesized that GISTs might have acquired a UPR-related intrinsic mechanism to relieve ER stress during tumorigenesis. The activation status of the three sensors of the UPR pathways (ATF6, IRE1α, and PERK) was measured in the cell line panel shown in Fig. 3A. Western blot analysis showed that GIST cells exclusively and strongly expressed the cleaved form of ATF6 (cATF6), which is a transcriptional activator that upregulates the expression of chaperones and cell survival-related genes [6] (Fig. 4A). ICC analysis further showed that cATF6 was localized in the nucleus of GIST cells, whereas no nuclear fluorescence signal was detected in other cancer cells (Fig. 4B). These results suggest that GISTs may take advantage of ATF6 activation as a survival strategy against various cellular stressors.
Sustained activation of ATF6 is a novel mechanism for GIST cell survival
To investigate ATF6 involvement in the intrinsic tolerance of GISTs to ER stress, the GIST cell viability was measured 24 hours after treatment with an ER stress inducer thapsigargin (TG). TG treatment decreased GIST cell viability in a dose-dependent manner (0.1–5 µM) (Fig. 4C). Based on the results, mild (0.1 µM TG) and strong (5 µM TG) ER stress-inducing conditions were selected. Because of drastic cell death after 24 hours of treatment with 5 µM TG, the activation status of ATF6, IRE1α, and PERK was measured for up to 8 hours. The expression of CHOP (a cell death-related marker) and MT-KIT was also measured. Western blot analysis showed that strong ER stress rapidly and drastically reduced ATF6, cleaved-ATF6 (cATF6), and MT-KIT levels over time, whereas phospho-IRE1α, phospho-PERK, and CHOP levels drastically increased after 1 hour of treatment with TG and continuously increased for 8 hours (Fig. 4D). In the case of mild ER stress, ER stress-adaptive chaperones (BIP, GRP94, HSP70, and HSP90) were analyzed for up to 24 hours. ATF6, cATF6, and MT-KIT levels showed a moderate decrease until 8 hours of treatment with TG but were fully restored after 24 hours of TG treatment. On the other hand, phospho-IRE1α, phospho-PERK, and CHOP levels increased until 8 hours of TG treatment, but were reduced to basal levels after 24 hours. HSP90, BIP, and GRP94 levels gradually increased up to 24 hours of TG treatment, while CHOP levels showed a gradual decrease after it peaked at 4 hours post-treatment (Fig. 4E). To validate the direct involvement of ATF6 pathway in mild ER stress adaptation of GISTs, an ATF6 inhibitor, PF429242 (PF), was used to block ATF6 cleavage. Western blot analysis showed that PF effectively inhibited ATF6 cleavage, and was accompanied by a drastic decrease in HSP90, BIP, and GRP94 levels (Fig. 4F). Under PF-mediated ATF6 inhibition, even mild TG resulted in a dramatic cell death (Fig. 4G). The upregulation of HSP90, BIP, and GRP94 was significantly delayed in GIST cells treated with both PF and mild TG. The upregulation of CHOP remained even after 24 hours of TG and PF treatment (Supplementary Fig. S10). Next, the relationship between sustained ATF6 activation and MT-KIT folding and overexpression was investigated to analyze the interaction between the MT-KIT and chaperones. Immunoprecipitation analysis showed that MT-KIT specifically bound to HSP90 (Fig. 4H). These results suggest that ATF6 plays a pro-survival role against ER stress-mediated cell death, which benefits MT-KIT folding.
Pharmacological inhibition of ATF6 perturbs GIST cell growth irrespective of imatinib resistance and shows enhanced anti-tumor effects when combined with ER stress-inducing drugs
To verify the significance of ATF6 as a therapeutic target in GISTs, the antitumor effects of PF, Ceapin-A7, and melatonin, which are reported ATF6 inhibitors, were investigated. The efficacy of these inhibitors in combination with ER stress-inducing drugs, such as borteazomib and 17AAG, were further evaluated. To mimic a constant ER stress condition in vivo, analysis was performed with or without TG, and imatinib was used as a control drug. Cell viability analysis showed that ATF6 inhibitors alone efficiently suppressed GIST cell growth, irrespective of imatinib resistance, while the antitumor effect was synergistically enhanced when the ATF6 inhibitors are combined with ER stress-inducing drugs. The synergistic antitumor effect was also observed with imatinib (Fig. 5A and B; Supplementary Fig. S11). Treatment with each ATF6 inhibitor potently downregulated cATF6 and KIT levels (Fig. 5C; Supplementary Fig. S12). In the presence of TG, the anti-tumor effect of each ATF6 inhibitor, with or without ER stress-inducing drugs, was slightly enhanced (Supplementary Fig. S13). To verify this, GIST xenografts were generated using GIST430 and GIST430-V654A cells. The antitumor effect of PF with or without bortezomib was then assessed, and imatinib was used as a control drug. As expected, the GIST430 xenograft model was sensitive to imatinib, while the GIST430-V654A model was resistant to imatinib. Treatment with PF (30 mg kg-1) potently inhibited tumor growth irrespective of imatinib sensitivity. When combined with borteazomib (1 mg kg-1), the antitumor efficacy was synergistically enhanced, especially in the imatinib-resistant GIST430-V654A model (Fig. 5D). IHC was performed using mouse tissues to measure the protein levels of KIT, Ki67, and cleaved caspase-3. KIT and Ki67 levels gradually decreased in the following order: imatinib, PF, and PF with bortezomib, whereas cleaved caspase-3 showed the opposite (Fig. 5E and F).
Nuclear expression of ATF6 is frequently detected in GIST tissues
To demonstrate the clinical significance of ATF6 in GISTs, ATF6 expression in 42 patient tissues was analyzed. IHC analysis revealed that 33% of the patients (14/42, H-score ≤10) showed negative nuclear ATF6 expression, while 67% of the patients (28/42, H-score > 10) showed positive nuclear ATF6 expression (Supplementary Fig. S14 and Supplementary Table S2). Survival analysis indicated that patients with nuclear ATF6 expression had significantly shorter relapse-free survival (P = .033) and a tendency for shorter overall survival (P = .339) (Supplementary Fig. S15). Moreover, nuclear ATF6 expression showed significant correlation with recurrence or metastasis (P = .026; Supplementary Table S2).