Targeting autophagy potentiates antitumor activity of Met-TKIs against Met-amplified gastric cancer

Abstract
Met tyrosine kinase inhibitors (Met-TKIs) subjected to ongoing clinical trials are a promising option for Met-amplified gastric cancer (GC), but how to optimize their antitumor activity especially with combination schemes remains unclear. Since autophagy is known to be initiated by Met-TKIs, we investigated its underlying mechanisms and therapeutic potentials of Met-TKIs combined with autophagy inhibitors against Met-amplified GC. As expected, four Met-TKIs induced autophagy in Met-amplified GC cells marked by p62 degradation, LC3-II accumulation and increased LC3-positive puncta. Autophagy flux activation by Met-TKIs was further validated with combined lysosomal inhibitors, bafilomycin A1 (Baf A1) and hydroxychloroquine (HCQ). Molecular investigations reveal that autophagy induction along with mTOR and ULK1 de-phosphorylation upon Met-TKI treatment could be relieved by hepatocyte growth factor (HGF) and mTOR agonist MHY1485 (MHY), suggesting that autophagy was initiated by Met-TKIs via Met/mTOR/ ULK1 cascade. Intriguingly, Met-TKIs further suppressed cell survival and tumor growth in the presence of autophagy blockade in Met-amplified GC preclinical models. Thus, these findings indicate Met/mTOR/ULK1 cascade responsible for Met-TKI-mediated autophagy and Met-TKIs combined with autophagy inhibitors as a promising choice to treat Met-amplified GC.

Introduction
Despite recent improvements in anticancer ther- apeutics, clinically available drugs for gastric cancer (GC) are limited, and hence GC remains a leading cause of mortality in China1. Receptor tyrosine kinase Met (also known as hepatocyte growth factor (HGF) receptor) is a promising target for Met-addicted GC. The HGF/Met pathway broadly participates in GC survival, invasion and metastasis2, and aberrant activation of HGF/Met pathwayrepresented by Met overexpression and gene amplifica- tion frequently occurs in GC3. Met overexpression and amplification were found in 39% and 7% of advanced GC in our previous study, respectively4. Growing evidence suggests Met gene amplification rather than protein overexpression as a true oncogenic driver and a predictive marker for Met-TKIs in GC5–7. Several Met tyrosine kinase inhibitors (Met-TKIs) including crizotinib (Criz) and volitinib (Voli) against Met-amplified GC are being investigated for this reason.Targeted drugs usually elicit better antitumor activity when combined with chemotherapy or other inhibitors due to known or unknown mechanisms8. Therefore, understanding signaling pathway changes resulting from Met-TKI treatment are very critical to develop novel combination strategies for improving Met-TKI efficacy, especially in the Met-amplified subpopulation. Based onaccumulating data, autophagy is frequently induced by drug exposure and acts as an attractive molecular target to potentiate efficacy of anticancer treatment9–19. Autophagy, a cellular adaptive response to stresses including antic- ancer agents, is an evolutionally conserved proteolytic process involving lysosomal degradation and recycling damaged cellular components and energy to maintain homeostasis20. Of note, protective autophagy rising in most contexts poses an opportunity for autophagy inhibitor-based combination therapies. Autophagy block- ade has been applied concurrently with either che- motherapies or targeted therapies to optimize their efficacy in various cancers in preclinical studies9,10,12–16,19,21,22. Regarding GC, a previous study roughly revealed that targeting autophagy initiated by Met-TKIs improved Met- TKI efficacy in vitro23; however, Met-TKI-associated autophagy flux alterations, mechanisms underlying autophagy induced by Met-TKIs and therapeutic poten- tials of dual targeting Met/autophagy in Met-amplified GC, especially in vivo, remain far from clear. Hence, this study aims at these issues to deepen our understanding of potentials of optimizing Met-TKI efficiency with targeting autophagy in Met-addicted GC.

Results
Met-amplified GC cells6,24,25 were treated with various doses and duration of Met-TKIs. Met-TKIs actioned on Met-amplified GC cells, indicated by remarkable de- phosphorylation of Met (Fig. 1a, b). Of note, the total Met levels, both its pro-Met and Met form, tended to be reduced by Met-TKIs to some degree. Marked by degra- dation of p62 and accumulation of LC3-II, autophagy was initiated after Met-TKI treatment (Fig. 1a, b). LC3-positive puncta consistently increased compared to the control group (Fig. 1c). Thus, Met-TKIs induce- d autophagy in Met-amplified GC cells. As reported, LC3-II accumulates due to either increased autophagy flux or decreased autophagy degradation, which can be distinguished with combined lysosomal inhibitors26. Degradation of p62 in Met-amplified GC cells exposed to Met-TKIs was blocked while LC3-II accumulation and LC3-positive puncta increased in the presence of lysoso- mal inhibitors (bafilomycin A1 (Baf A1) and hydroxy- chloroquine (HCQ); Fig. 2a, b). These data suggest that autophagosome formation rather than blockade of autophagy degradation occurred upon Met-TKI treatment. Hence, Met-TKIs activated autophagy flux in Met-amplified GC cells.To excavate potential molecular mechanisms, we next evaluated phosphorylation status of mTOR (mammaliantarget of rapamycin, a common downstream target of HGF/Met signaling and core autophagy pathway) and its autophagy downstream effector ULK126. Met, mTOR, and ULK1 were de-phosphorylated when Met-TKIs induced autophagy, suggesting that mTOR inactivation and ULK1 activation were responsible for autophagy initiation by Met-TKIs (Fig. 3a).

As a ligand of Met, HGF can be used as a pMet activator, evidenced by reducing Met-TKI-caused Met de-phosphorylation (Fig. 3c). Of interest, HGF not only inhibited autophagy mediated by starvation (Fig. 3b) but also attenuated autophagy along with mTOR and ULK1 de-phosphorylation initiated by Met-TKIs (Fig. 3c). Meanwhile, these Met-TKI action- s could also be diminished by an mTOR agonist MHY1485 (MHY) (Fig. 3d). Taken together, Met-TKIs induced autophagy via Met/mTOR/ULK1 cascade in Met-amplified GC cells.Prevalently, autophagy is considered as a pro-survival mechanism in many cancers20. Given Met-TKI-mediated autophagy presented in our Met-amplified GC cells, dual inhibition of Met and autophagy was proposed. Com- pared to controls, Met-TKIs or the autophagy inhibitor HCQ alone inhibited cell viability to some extent, while the combination groups had a better response in Met- amplified GC cells (Fig. 4a). Given that HCQ can exert cytotoxicity independent of autophagy26, targeting autophagy by small interfering RNA (siRNA) interference was employed to further confirm augmented effects of autophagy blockade on Met-TKI cellular lethality. As anticipated, both Atg5 and BECN1 knockdown could sensitize the Met-amplified MKN45 cell to Met-TKIs (Fig. 4b–d). Due to cross-talks between autophagy and cell apoptosis27, we also detected the combination’s effect on apoptosis. As revealed, Met-TKI-mediated apoptosis increased in the presence of HCQ in terms of total apoptotic cells (Fig. 4e). However, the phenomena could arise in the context of increased necrotic cells by HCQ non-specific cytotoxicity and early-phase apoptotic cells were almost unaffected, which indicated that apoptotic cell death might not as necessarily as growth inhibition linked to strengthened cytotoxicity of Met-TKIs plus autophagy inhibitors in Met-amplified GC cells. Paralleled to the in vitro findings, clinically investigated Met-TKIs (Criz and Voli) or the autophagy inhibitor HCQ alone reduced tumor growth, while Met-TKIs combined with HCQ yielded the greatest tumor suppression in Met- amplified GC xenografts (Fig. 5a). Of note, the combina- tion of Voli and HCQ did not reach the level of statistical significance compared to Voli monotherapy in mice bearing SNU5 cells. Besides, autophagy occurred in tumors treated with Met-TKIs as indicated by p62degradation and LC3-II accumulation (Fig. 5b). These data suggest that Met-TKIs elicited greater antitumor activity in the presence of autophagy blockade in Met- amplified GC preclinical models.

Discussion
Continuous efforts to optimize Met-targeted ther- apeutic strategies led to the identification of Met ampli- fication as a possible driver event of GC5,7, which fostersclinical trials on Met-TKI therapeutic potentials against Met-amplified GC (www.clinicaltrials.gov). However, efficacy of Met-TKIs alone is limited in Met-addicted GC. Different approaches represented by combination designs, therefore, are being studied to improve overall response. To develop feasible drug combination strategies, mole- cular or signaling pathway changes after treatment mustbe known. Chemotherapies and targeted therapies have been validated to induce autophagy in solid tumors and in hematologic malignancies9–19,28–30. Protective autophagy frequently occurs after drug stimulation, indicative of potential benefits with combined autophagy inhibitors. Based on published results17,23,31, our data further display autophagy induction after exposure to Met-TKIs inMet-amplified GC preclinical models (Figs. 1 and 5b), with treatment dose and duration took into consideration. Of note, autophagy in the in vivo scenario will be strengthened by other evidence like HCQ-mediated LC3accumulation32,33 and immunohistochemistry analysis of autophagy-associated proteins26. Using lysosomal inhibi- tors recommended in a guideline26, autophagy flux acti- vation rather than repressed autophagy degradation byMet-TKIs was further elucidated (Fig. 2), which enabled autophagy modulators to be added in the treatment of Met-amplified GC. Of note, lysosomal inhibitors (Baf A1 and HCQ) only partially reversed p62 reduction induced by Met-TKIs (Fig. 2a). To further confirm decreased p62 levels dependent on autophagy, neighbor of BCRA1 gene1 (NBR1) that shares a similar function with p62 in autophagic degradation26,34 was used.

As expected, Met- TKIs downregulated NBR1 expressions in Met-amplified GC cells, which could be mitigated by lysosomal inhibi- tors (Fig. S2). However, these results did not exclude the possibility of p62 reduction by Met-TKIs through alter- native mechanisms such as proteasomal degradation andcaspase-induced cleavage26, which remains to be deci- phered using proteasome and pan-caspase inhibitors. Meanwhile, other assays for monitoring autophagy such as transmission electron microscopy and immuno- fluorescence experiments can be introduced since con- focal findings were less clear (Figs. 1c and 2b), especially for Met-TKI monotherapy groups. Besides, whether autophagy occurs in Met non-amplified GC cells is unclear. Despite limited effects on cell proliferation compared to Met-amplified GC cells, Met-TKIs initiated autophagy in Met non-amplified NCI-N87 cells (Fig. S1)6,24,25, whose values in the treatment of GC remains elusive. In order to optimize responses ofMet-addicted GC to Met-TKIs, only Met-amplified GC cells were used in this study.Molecular mechanisms underlying Met-TKI-induced autophagy in GC are unclear. De-phosphorylation of mTOR and ULK1 (activated and inactivated site, respec- tively) involves in autophagy26,35 and was observed upon Met-TKI treatment in Met-amplified GC cells (Fig. 3a). Met/mTOR/ULK1 cascade responsible for Met-TKI- mediated autophagy was thus further verified by com- bined HGF (pMet activator) and MHY (mTOR agonist), marked by antagonizing Met-TKI actions on autophagy, mTOR, and ULK1 de-phosphorylation (Fig. 3c, d). These phenomena consistently occur in cardiac disease and hepatocellular carcinoma35,36. However, the molecular events following ULK1 de-phosphorylation and signal transducer and activator of transcription 3 (STAT3) pathway’s role during the process as reported in lung cancer17 merit future investigations.

Autophagy after antitumor treatment may reduce effi- cacy of therapeutics9–17,19, in accordance with potentiated anticancer activity of Met-TKIs plus autophagy inhibition against Met-amplified GC preclinical models (Figs. 4 and 5a, c). These data suggest therapeutic potentials of the combination strategy in Met-amplified GC, which can be warranted by assessing impacts of this combination scheme on other events involving both HGF/Met path- ways and autophagy, such as cancer progression, angio- genesis, and antitumor immune response18,37–39. Of note, the superiority of Voli plus HCQ over Voli alone was presented in mice bearing SNU5 cells without statistical significance (Fig. 5a), which may result from striking efficacy of Voli monotherapy. On the other hand, available autophagy inhibitors can exert non-specific cytotoxicity, targeting autophagy to improve Met-TKI efficacy was further proven by knockdown of autophagy-related molecules (Atg5 and BECN1; Fig. 4b–d), and novel more specific or potent autophagy inhibitors like Lys05 and SAR40540,41 are expected to promote basic studies and enter clinical practice. Moreover, apoptotic cell death contributed to augmented efficacy of this combination strategy in terms of total apoptotic cells (necrotic cells included) but early-phase apoptotic cells were almost unaffected (Fig. 4e), inconsistent with a previous finding that autophagy inhibitor CQ mitigated Criz-induced apoptosis in Met-amplified GC cells31. This discrepancy might be explained by different treatment doses and durations of autophagy inhibitors, reflecting a dynamic process of autophagy and non-specific cytotoxicity of autophagy inhibitors to some extent. In line with the complexity, whether autophagy inhibited or enhanced apoptosis mediated by anticancer agents is controversialin different studies due to a variety of confounding factors.

Xenografts derived from Met-amplified GC cells were used for in vivo experiments but data would be more compelling if patient-derived xenograft (PDX) models were applied. Met amplification occurred in about 7% of patients with advanced GC in our retrospective study4 while merely about 1–2% in our ongoing prospective clinical trial (NCT01985555). Similar to the low Met amplification rate, none of our established 50 GC PDXs featured Met amplification43. Efficacy of Met-TKIs plus autophagy inhibitors can be further verified when Met- amplified PDXs are available. On the other hand, our previous studies revealed a pMet-positive rate of 6% in patients with advanced GC, a strong association between pMet expression and Met amplification, and Voli-elicited strong antitumor activity in a GC PDX model with pMet expression4,43. Indeed, GC cells (MKN45 and SNU5) were pMet positive and Met-TKI-based treatment obviously inhibited Met phosphorylation in this work (Figs. 1 and 3 and 5b). Thus, patients with Met amplification/pMet expression may be suitable for Met-TKI combined with autophagy inhibitors. Moreover, we notice that Met-TKIs caused Met de-phosphorylation accompanied by reduced expressions of pro-Met and Met (Figs. 1, 3 and 5b), indicating total Met protein levels affected by Met-TKIs. However, total Met protein levels can be decreased, unaffected, or increased when Met-TKIs yield anticancer activity and markedly inhibit Met phosphoryla- tion24,25,44,45. Met phosphorylation status, thus, outweighs total Met protein levels for Met-TKI antitumor actions, consistent with Met-TKI pharmacological mechanism. The significance of total Met expressions modified by Met-TKIs remains unrevealed.

Conclusions
Met-TKIs induced protective autophagy via Met/ mTOR/ULK1 cascade and Met-TKIs combined with autophagy inhibitors augmented antitumor activity in Met-amplified GC preclinical models, shedding light upon this combination strategy in the MHY1485 treatment of Met- amplified GC.