Differential expression of MAGEA6 toggles autophagy to promote pancreatic cancer progression

The melanoma-associated antigen family A (MAGEA) antigens are expressed in a wide variety of malignant tumors but not in adult somatic cells, rendering them attractive targets for cancer immunotherapy. Here we show that a number of cancer-associated MAGEA mutants that undergo proteasome-dependent degradation in vitro could negatively impact their utility as immunotherapeutic targets. Importantly, in pancreatic ductal adenocarcinoma cell models, MAGEA6 suppresses macroautophagy (autophagy). The inhibition of autophagy is released upon MAGEA6 degradation, which can be induced by nutrient deficiency or by acquisition of cancer-associated mutations. Using xenograft mouse models, we demonstrated that inhibition of autophagy is critical for tumor initiation whereas reinstitution of autophagy as a consequence of MAGEA6 degradation contributes to tumor progression. These findings could inform cancer immunotherapeutic strategies for targeting MAGEA antigens and provide mechanistic insight into the divergent roles of MAGEA6 during pancreatic cancer initiation and progression.


Introduction
Cancer/testis antigens (CTAs) are encoded by genes whose expression is normally restricted to male germ cells but is de-repressed in various human cancers (Chen et al., 2017;Hofmann et al., 2008). Their unique expression patterns could have utility in cancer diagnosis, prognosis, and immunotherapy. The MAGE protein family of CTAs, including the MAGEA, B, and C Type one subfamily members, (Barker and Salehi, 2002;Simpson et al., 2005) are of particular interest. The aberrant expression of Type 1 MAGEs in cancer is a result of promoter hypomethylation due to genome-wide epigenetic reprogramming. The high tumor specificity of MAGEA expression led to multiple clinical trials targeting MAGEA genes with immunotherapy agents (Vansteenkiste et al., 2013;Zajac et al., 2017). However, responses have been modest (Connerotte et al., 2008).
MAGEA genes have been proposed to function as oncogenes; however, the underlying mechanisms are poorly understood . Consistent with a role in tumor progression, high MAGEA protein expression has been associated with advanced stage (Liu et al., 2008;Sang et al., 2017;Wang et al., 2016) and poor prognosis Xu et al., 2014;Zhai et al., 2016;Zhang et al., 2015) in multiple cancer lineages. Interactions between MAGEs and really interesting new gene (RING) domain containing E3 ubiquitin ligases (Doyle et al., 2010) enhance E3 ubiquitin ligase activity toward downstream substrates and lead to their degradation (Doyle et al., 2010), which may contribute to the oncogenic activity of MAGEs. A recent proteomic study revealed a negative association between autophagy activators and MAGEA protein expression in human melanoma (Shukla et al., 2018). In addition, MAGEA3 and MAGEA6 suppress autophagy in lung and colorectal cancer models (Pineda et al., 2015). The effects of MAGEA on autophagy provide a potential mechanism for oncogenic effects of MAGEA.
Autophagy promotes cell survival by sustaining cellular synthetic pathways and energy homeostasis under metabolic stress. It also preserves long-term cellular integrity by preventing the accumulation of misfolded proteins and malfunctioning organelles (White, 2012;Yang and Klionsky, 2010). Both pro-and anti-tumorigenic properties of autophagy have been reported and are largely determined by specific tumor stage and genetic context (Guo et al., 2011;Guo et al., 2013;Lock et al., 2011;Rao et al., 2014;White, 2012). During early steps in tumor development, autophagy acts as a tumor suppressor by facilitating DNA damage repair and reducing reactive oxygen species and thus maintaining genome integrity (Panda et al., 2015). At late stages of tumor development, autophagy increases the ability of tumor cells to cope with endogenous stress and increases chemotherapy or radiotherapy resistance (Pylayeva-Gupta et al., 2011;White, 2012) thus promoting tumorigenic behavior. Oncogenic KRAS mutations, which are present in more than 90% of pancreatic ductal adenocarcinoma (PDAC) tumors, represent the earliest driving event for PDAC (Almoguera et al., 1988). Knockdown of core members of the autophagy initiation complex in Kras G12D/+ transgenic mice increased accumulation of low-grade, pre-malignant pancreatic intraepithelial neoplasia lesions but rendered these lesions resistant to progression into invasive PDAC (Rosenfeldt et al., 2013;Yang et al., 2014). In contrast, pharmacologic inhibition of autophagy in established PDAC patient-derived xenografts induced apoptosis and reduced proliferation (Yang and Klionsky, 2010). Thus, autophagy appears to play a negative role in PDAC initiation but is important for late-stage tumor development.
Given the complex role of autophagy in PDAC, the role of interactions between MAGEA6, which has been found to be aberrant in PDAC, and the autophagy machinery in PDAC initiation remains unclear. In addition, there is an urgent need to understand the impact of cancer-specific mutations in MAGEA6 on tumor aggressiveness and on immunotherapy strategies (Caballero et al., 2010;Hagiwara et al., 2016). We therefore developed a comprehensive MAGEA cancer-specific mutation series to interrogate the impact of these mutations on protein expression and function. We also examined the role of PDAC-specific MAGEA6 variants using in vitro and in vivo PDAC models.

Mutational landscape of MAGEA genes in cancer
The MAGEA gene family is located on the X chromosome and consists of thirteen protein-encoding genes (MAGEA1 to A6, A8 to A12, A2B, and A9B) and one pseudogene, MAGEA7P. About 1000 unique non-silent MAGEA gene aberrations have been identified by ICGC (ICGC/TCGA Pan-Cancer Analysis of Whole Genomes Consortium, 2020), The Cancer Genome Atlas (TCGA), and Catalogue of Somatic Mutations in Cancer (COSMIC) (Tate et al., 2019) pan-cancer analyses ( Figure 1A). Most of these gene aberrations are missense mutations, and they are clustered on MAGEA1, A3, A4, A6, A8, A10, A11, and A12. As a group, MAGEA gene family members, are more frequently mutated in TCGA lung cancer patients and pancreatic cancer patients when compared to a randomized gene set of 1000 genes with similar size to MAGEAs ( Figure 1B), suggesting that mutation of MAGEA genes is functionally important in cancer.
MAGEA genes represent an attractive group of antigens for cancer immunotherapy owing to their upregulation in cancer tissues. Indeed, over the past 10 years, 42 phase I and II clinical trials (Supplementary file 2) have explored MAGEA3, A4, A6, A10, and A12 as targets for immunotherapy. More than 500 recurrent mutations have been identified in the five clinically studied MAGEA genes in TCGA, ICGC, and COSMIC pan-cancer analyses ( Figure 1C). These mutations were uniformly distributed along the entire gene without any obvious mutational hotspots.

Cancer-specific mutations of MAGEA genes reduce their protein expression
We leveraged our High-Throughput Mutagenesis and Molecular Barcoding (HiTMMoB) technology (Tsang et al., 2016) to mimic 82 recurrent mutation events (found in at least 150 cancer patients) in the five MAGEA genes that have been evaluated in clinical trials ( Figure 1C and (Figure 2A). The amino acids that led to lower protein levels when mutated were significantly more conserved than residues that did not alter protein expression when mutated ( Figure 2C), suggesting that the C-terminal region is important in regulating or maintaining protein stability.

MAGEA variants are degraded through the ubiquitin proteasome pathway
We speculated that the low expression of the cancer-associated MAGEA variants is ubiquitin proteasome dependent owing to their reported interaction with the various E3 ubiquitin ligase (Doyle et al., 2010). Indeed, while a subset of MAGEA variants had lower level of transcripts than wild-type MAGEA (Figure 3-figure supplement 1), in general their protein and mRNA expression levels did not correlate and changes in transcript levels were insufficient to explain changes in protein amounts ( Figure 3A). More importantly, the addition of proteasome inhibitor MG132 increased protein amounts of multiple MAGEA3, MAGEA6 and MAGEA10 variants ( Figure 3B and Figure 3figure supplement 2), indicating that the proteasome is a major mediator of degradation. However, the protein amounts of a number of the MAGEA variants were not altered by MG132 treatment, suggesting that alternative mechanisms contribute to the discrepancy between RNA and protein amount in these examples. Our data demonstrated that a high number of MAGEA variants are degraded through the ubiquitin proteasome pathway. Further evaluation is needed to determine the efficacy of immunotherapy targeting these low-expression MAGEA variants in cancer.
Cancer-specific mutations and glucose/glutamine depletion stimulate proteasome-dependent MAGEA6 degradation Recently, an International Cancer Genome Consortium (ICGC) study listed MAGEA6 among the top 16 significantly mutated genes across tumors in two of three computational methods used (Fisher combined p value test, p=0.07; likelihood ratio test, p=9.88E-05; convolution test, p=8.85E-05) (Biankin et al., 2012). The study identified two MAGEA6 mutations, MAGEA6 N254I (tumor specimen PACA-86-T) and MAGEA6 H305fs*>7 (tumor specimen ICGC_0050_TD), in PDAC patients. To examine the effects of the cancer-specific variants in the proper cell lineage, we stably expressed the MAGEA6 variants in non-transformed human pancreatic ductal epithelial cells (HPDE-iKRAS) (Tsang et al., 2016) and two PDAC cell lines, BxPC-3 and MIA PaCa-2 using lentivirus ( Figure 4A and Figure 4-figure supplement 1). Although quantitative real-time polymerase chain reaction (qRT-PCR) confirmed high mRNA expression of both wild-type and mutant MAGEA6, the protein expression of the mutants was significantly downregulated compared to wild-type. Notably, MAGEA6 H305fs*>7 loses its termination codon owing to a frameshift deletion, which allows part of the vector sequence to be translated and results in a 26-amino acid addition (ERGGRVIPSFLVQSWH YKKALLINLLQRTGHYQSK) at the C-terminus, distinct from the wild-type. This change is reflected in the size increase of the MAGEA6 H305fs*>7 protein ( Figure 4A and Figure 4-figure supplement 1).
The clone is referred to as MAGEA6 H305fs* hereafter to indicate the loss of the last 11 amino acids. Consistent with our MAGEA study in 293 T cells, addition of MG132 to the stable lines expressing MAGEA6 variants resulted in a dramatic increase in mutant protein levels, which reached levels similar to those of wild-type ( Figure 4A and  Figure 3B. strong HA-polyubiquitin signal on mutant MAGEA6 proteins, which was further enhanced by MG132 treatment ( Figure 4B). We also detected a strong polyubiquitination signal on MAGEA6 H305fs* and MAGEA6 N254I after MG132 treatment with an antibody targeting endogenous ubiquitin in HEK293T cells (Figure 4-figure supplement 2), indicating that the MAGEA6 polyubiquitination observed in our co-expression study was not an artifact of ubiquitin overexpression.
While studying MAGEA6 protein stability, we observed a robust reduction of MAGEA6 variant protein levels in HPDE-iKRAS cells after prolonged culture in Keratinocyte serum-free medium (KSFM) ( Figure 4C). We firstly examined the effect of cell confluency on MAGEA6 expression by seeding 1 Â 10 6 MAGEA6-expressing HPDE-iKRAS cells on 10 cm and 6 cm cell culture plates. Although maximum confluency occurred on the smaller plate approximately 24 hr earlier than on the larger plate (estimated HPDE-iKRAS doubling time: 0.973 d; R 2 : 0.9949) (Figure 4-figure supplement 3), we did not observe any difference between plates in MAGEA6 protein levels (Figure 4figure supplement 3), challenging the argument that MAGEA6 expression is confluency dependent, at least in this model.
Next, we explored the possibility that the KRAS/MAPK pathway mediates MAGEA6 degradation, as the KRAS/MAPK pathway is hyperactivated in 95% of PDACs . We leveraged our HPDE-iKRAS cell model, which expresses KRAS G12D in the presence of doxycycline (Tsang et al., 2016), to examine the role of KRAS/MAPK signaling in MAGEA6 expression regulation. As shown in Figure 4D, MAGEA6 protein remained low at 72 hr regardless of activation of KRAS/MAPK, as evidenced by the higher phosphorylation of ERK1/ERK2 Thr202 and Tyr204 (Sebolt-Leopold, 2000) under doxycycline treatment. Surprisingly, refreshing KSFM every 24 hr completely rescued MAGEA6 protein levels ( Figure 4D, right four lanes), suggesting that components in the medium preserve MAGEA6 protein.
Since cancer cells often require increased amounts of nutrients, such as glucose and glutamine, to sustain the enhanced metabolic pathways required for high proliferation rates, we tested whether glucose and glutamine contributed to maintaining MAGEA6 protein levels. In vitro culture of HPDE-iKRAS MAGEA6-expressing cells in Dulbecco Modified Eagle Medium (DMEM) without fetal bovine serum (FBS), glucose, or glutamine (DMEM-) for 5 hr robustly downregulated MAGEA6 expression ( Figure 4E). The reduced MAGEA6 expression was observed in the PDAC cell lines AsPC-1, BxPC-3, and MIA PaCa-2 ( Figure 4-figure supplement 4) in DMEM-. In contrast, DMEM-supplemented with carbon sources of glucose or glutamine partially rescued MAGEA6, while DMEM-containing FBS or both carbon sources fully recovered MAGEA6 levels in both short (5 hr) ( Figure 4E) and long (24 hr) ( Figure 4F) term culture conditions. Along with MAGEA6 expression reduction, a reduction in p70S6K (S6K) phosphorylation was seen on residue Thr389 in DMEM-and partially supplemented DMEM-( Figure 4E, lanes 2, 3, and 4). S6K Thr389 is a phosphorylation site targeted by mTORC1 kinase whose activity is suppressed when nutrients are depleted (Efeyan et al., 2012). The rapid, highly synchronized changes in MAGEA6 expression and S6K phosphorylation status confirmed that MAGEA6 expression is nutrient sensitive. FBS-, but not glucose-dependent MAGEA3/A6 protein stability has recently been shown to be regulated by the CRL4-DCAF12 E3 ubiquitin ligase in other cell lineages including HeLa (Ravichandran et al., 2019). The ability of glucose and other nutrient sources to stabilize MAGEA6 protein levels in our model may reflect differences in metabolic wiring between different cell lineages due to different genomic and epigenomic backgrounds.    To examine whether MAGEA6 protein stability is maintained by mTORC1, we treated HPDE-iKRAS MAGEA6-expressing cells with rapamycin and MK2206 to inhibit mTORC1 or its upstream kinase AKT1, respectively. We did not observe any change in MAGEA6 protein amount after either treatment (Figure 4-figure supplement 5), suggesting that MAGEA6 protein levels are not regulated by the AKT/mTORC1 pathway. Consistent with a role of ubiquitination for protein degradation regulation, addition of MG132 to the starved HPDE-iKRAS cells ( Figure 4G) and other PDAC lines (Figure 4-figure supplement 4) fully rescued MAGEA6 from degradation. In summary, our data show that cancer-specific mutations and carbon source depletion induce MAGEA6 degradation by proteasomes.
Wild-type MAGEA6, but not mutant MAGEA6, suppresses basal autophagy and starvation-induced autophagy activation In mammalian cells, mTORC1 and AMPK are key kinases that coordinate autophagy activity (Mihaylova and Shaw, 2011;Zoncu et al., 2011). mTORC1 activation promotes anabolic programs such as mRNA translation and represses catabolic programs such as autophagy when the nutrient supply is sufficient (Bar-Peled and Sabatini, 2014). The activation of mTORC1 can be informed by the phosphorylation status of its direct substrate S6K Thr389 (Zoncu et al., 2011). AMPK, on the other hand, is a pro-autophagy and energy-sensing kinase that is phosphorylated on Thr172 when activated (Hardie et al., 2016). In general, autophagy activity can be monitored by examining mTORC1 and AMPK pathways, counting LC3B puncta, and measuring autophagic flux, a frequently used technique to determine autophagy dynamics by comparing the amounts of stable protein substrates such as SQSTM1/p62 between samples in the presence and absence of lysosomal inhibitors (Bjørkøy et al., 2005).
In agreement with the reported autophagy inhibitory role of MAGEA3/6 (Pineda et al., 2015), HPDE-iKRAS, AsPC-1, and MIA PaCa-2 cells stably expressing wild-type MAGEA6 showed reduced AMPK phosphorylation on Thr172 and increased S6K phosphorylation on Thr389 ( Figure 5A and Figure 5-figure supplement 1). Immunofluorescence staining of LC3B in MAGEA6-expressing HPDE cells treated with lysosome inhibitor bafilomycin A1 (BafA1) dramatically reduced the cell population with a high number of LC3B puncta ( Figure 5B), suggesting slower autophagosome formation and thus lower basal autophagy activity levels in these cells. Autophagic flux study using BafA1 showed slower SQSTM1/p62 accumulation in MAGEA6-expressing cells compared with control cells ( Figure 5C, left), suggesting a slower turnover of SQSTM1/p62 and therefore further supporting the anti-autophagy role of MAGEA6. Importantly, cells expressing MAGEA6 H305fs* and MAGEA6 N254I , which are susceptible to proteasome mediated degradation, showed minimal changes in autophagy signaling ( Figure 5A In nutrient-deficient conditions, short-term culture of HPDE-iKRAS cells in DMEM-showed a gradual reduction in MAGEA6 expression, although MAGEA6 levels still remained sufficient to suppress AMPK Thr172 phosphorylation, reduce induction of LC3B expression ( Figure 5D) and to impair autophagic flux, as demonstrated by the slow accumulation of SQSTM1/p62 and LC3B in the presence of BafA1 ( Figure 5E). In contrast, cells expressing MAGEA6 H305fs* and MAGEA6 N254I displayed autophagy activation similar to that in control cells ( Figure 5F and Figure 5-figure supplement 3). Nevertheless, long-term culture (24 hr) of cells expressing wild-type MAGEA6 in DMEMcompletely abolished MAGEA6 expression and thus allowed autophagy re-initiation, as indicated by low S6K Thr389 and high AMPK Thr172 phosphorylation ( Figure 5G). In summary, MAGEA6 suppresses both basal and short-term starvation-induced autophagy in PDAC. Its autophagy-quenching effect is revoked when MAGEA6 is downregulated either by degradation-prone mutation or by long-term starvation-induced protein degradation.

MAGEA6 mutation status/protein amount manipulates autophagy to promote tumor progression at different stages
To investigate the oncogenic role of MAGEA6 in PDAC, we subjected non-transformed HPDE cells stably expressing wild-type MAGEA6 or GFP to bilateral subcutaneous implantation into athymic mice (5 Â 10 5 cells per flank; N = 5 for each cohort). In consonance with its role in autophagy suppression, MAGEA6 expression transformed HPDE cells, induced xenograft tumor growth  ( Figure 6A), and led to poor survival ( Figure 6B). In addition to autophagy suppression, MAGEA3/6 has been reported to enhance the degradation of the tumor suppressor p53 (Doyle et al., 2010), which may contribute to PDAC development. However, we did not find any association between endogenous MAGEA6 and p53 expression in HPDE-iKRAS or in six other PDAC cell lines ( Figure 6figure supplement 1). Furthermore, ectopic expression of MAGEA6 variants in the BxPC-3 and MIA PaCa-2 lines did not induce changes in p53 protein levels in the presence or absence of the proteasome inhibitor MG132 (Figure 4-figure supplement 1), indicating that MAGEA6 drives PDAC independently of p53. As stated earlier, autophagy exhibits both oncogenic and tumor-suppressing activities, largely depending on progression status of individual tumor. It is still unclear how autophagy is regulated to carry out these divergent activities during tumor development. The fact that MAGEA6 functions as a bona fide oncogene in PDAC but is subjected to degradation-prone mutation leads us to hypothesize that (1) MAGEA6 is an autophagy coordinator manipulating autophagy activity at different disease stages to promote tumor initiation and further development and (2) the nutrient-sensing stability of MAGEA6 provides autophagy regulation for better tumor survival under metabolically stressed conditions.
In agreement with the anti-tumorigenic role of MAGEA6 at a late disease stage, knockdown of MAGEA6 in the transformed PDAC cell line BxPC-3, which has high endogenous wild-type MAGEA6 expression, dramatically increased tumor volume compared with xenografts without MAGEA6 knockdown ( Figure 6C top and Figure 6-figure supplement 2). Immunohistochemical staining of the BxPC-3 tumors revealed a robust decrease in LC3B levels and, thus, elevated autophagy activity in the MAGEA6 knockdown cohort ( Figure 6C bottom). In support of this contention, knockdown of ATG7 and VPS34, key players for autophagy activation, suppressed tumor growth induced by MAGEA6 downregulation, consistent with autophagy activity playing a key role in promoting tumor growth in response to MAGEA6 depletion in BxPC-3 ( Figure 6D and Figure 6-figure supplement  2). We hypothesized that MAGEA6-dependent tumorigenicity is most critical at an early disease stage when low autophagy activity is preferred. We posited that the tumorigenicity of MAGEA6 and its autophagy suppression then subside as the tumor grows owing to spontaneous gene mutations and nutrient stress.
We next sought to examine the role of MAGEA6 over the course of cancer progression in patients. Since PDAC patients often present with late-stage disease and a low 5 year survival rate, we reasoned that lung cancer is a better option for evaluating the impact of MAGEA6 on different stages of cancer development (initiation and progression) because (1) a large number of patients with lung cancer showed high MAGEA6 expression and mutation events, and (2) approximately onethird of MAGEA6 mutation events (6/17) found in the ICGC lung cancer database displayed reduced protein expression ( Figure 6-figure supplement 3). A Kaplan-Meier survival analysis of 889 patients with lung cancer showed a significant association of high MAGEA6 expression and poor overall survival in cancers at stage I (hazard ratio = 1.96, p<0.0001) but not in those at stage II (hazard ratio = 1.1, p=0.6088) or stage III (hazard ratio = 1.06, p=0.8441) ( Figure 6E). Collectively, our in vivo data and the patient survival analysis strongly support the hypothesis that wild-type and mutant MAGEA6 play distinct roles to promote tumor initiation and development at different disease stages.

Discussion
Our study showed that a high number of MAGEA variants identified in pan-cancer analyses were associated with reduced protein expression, suggesting that the impact of these MAGEA gene mutations on potential and MAGEA directed cancer immunotherapy may need further evaluation. In our investigation of the relationship between MAGEA6 and autophagy, we found that intact MAGEA6 suppresses autophagy, which can be oncogenic in early-stage disease when low autophagy activity is preferred (Kimmelman, 2011). The degradation of MAGEA6 through ubiquitin proteasome pathway in late-stage disease revokes autophagy and further promotes cancer progression (Figure 7). Further study is required to examine whether TRIM28, or another E3 ligase, induces MAGEA6 degradation.
Our findings not only elucidate the regulation of tumor evolution but also reveal a novel crosstalk mechanism between autophagy and the ubiquitin proteasome system, which represent the two   Figure 6 continued on next page cornerstones of cellular catabolism. Indeed, several studies report evidence of crosstalk between these two catabolic pathways. For example, previous studies identified multiple protein substrates targeted by the autophagy and ubiquitin proteasome pathways (Milani et al., 2009;Pandey et al., 2007). Further analysis showed that impairment of proteasome activity led to activation of the autophagy pathway. The crosstalk between these two systems is believed to be a potential compensatory mechanism allowing cells to reduce levels of ubiquitinated substrates and provide tumorigenic advantages. Under the same theory, the reactivation of autophagy induced by low MAGEA6 expression may provide extra intracellular catabolic capability to continuously downregulate MAGEA6 targets and sustains tumor development. Further study of this crosstalk may help elucidate the importance of cellular catabolism for tumorigenesis.
Although our study mainly focused on the tumor biology of a few MAGEA6 variants, the MAGE family involves more than 40 genes classified as CTA genes, broadly expressed in many tumor types (Barker and Salehi, 2002;Simpson et al., 2005). Other MAGE members with similar biological properties may play a related role in cancer. For example, MAGEA2, A3, A6, and C2 were shown to bind to the TRIM28 E3 ligase with similar strength and to boost ubiquitin ligase activity against p53 (Doyle et al., 2010). The large number of CTA MAGEs and the high sequence homology among them may confer functional redundancy that permits ongoing positive selection toward diversification or acquisition of functional characteristics during tumor development. Indeed, MAGEA3 and MAGEA6 are believed to work complementarily to regulate TRIM28 E3 activity and autophagy suppression (Pineda et al., 2015). The co-expression of MAGEA genes and their mutation-dependent expression discrepancy may provide extra flexibility for tumors to adapt to the heterogeneity and diversity of stresses in the tumor microenvironment.
Autophagy has been studied extensively as a potential therapy target owing to its pro-tumor survival activity of conferring resistance against metabolic stress, radiation, and chemotherapies. Inhibition of autophagy using 3-methyladenine sensitizes cells to cisplatin and 5-fluorouracil in esophageal and colon cancers, respectively (Li et al., 2009;Liu et al., 2011). In PDAC cells, co-treatment of gemcitabine (the first-line therapy in the clinic) with 3-methyladenine, the AMPK inhibitor compound C, or the lysosomal inhibitor chloroquine potentiates gemcitabine-induced apoptosis . Recent PDAC studies also demonstrated autophagy activation when KRAS or the downstream ERK pathway is suppressed and pharmacologic inhibition of autophagy regulators synergistically enhanced the ERK inhibitors-induced antitumor activity in KRAS-driven PDAC (Bryant et al., 2019;Kinsey et al., 2019). Our finding that autophagy suppression is induced by MAGEA6 and reactivated upon MAGEA6 degradation may inform combination chemotherapeutic strategies that could improve outcome in PDAC. For example, PDAC tumors expressing high level of MAGEA6 may be more sensitive to cytotoxic agents due to the low autophagy activity while tumors harboring MAGEA6 degradation-prone mutations may show increased dependency on autophagy, and thus by more susceptible to treatment with an autophagy inhibitor.
In summary, our findings reveal a previously unrecognized, mutation-directed regulation of MAGEA protein stability that may have important implications for maximizing the clinical utility of MAGEA-targeted immunotherapy. These findings illuminate a molecular mechanism of metabolic stress-induced degradation of an autophagy suppressor, identify a mechanism for toggling autophagy during tumorigenesis, and provide insight into proteins that function as context-dependent oncogenes or tumor suppressors.

Materials and methods
More information of the reagents used in the study can be found in Supplementary file 1.

Cell culture
All cell lines were propagated at 37˚C in 5% CO 2 in a humidified atmosphere. HPDE cells and HPDE-iKRAS cells were cultured as described previously (Tsang et al., 2016). AsPC-1, BxPC-3, Capan-1, Capan-2, Panc-1, and MIA PaCa-2 were purchased from the American Type Culture Collection and cultured according to the manufacturer's recommendations. HEK293T cells were cultured in DMEM supplied with 10% FBS and 1% penicillin-streptomycin. All human cell lines were authenticated by fingerprinting using STR testing where possible and were verified to be free of mycoplasma contamination before use. In some experiments, as indicated, cells were propagated in doxycycline (100 ng/ ml), 20 nM BafA1 (Selleck Chemicals), 10 mM MG132 (Sigma-Aldrich), or glucose and glutamine-free DMEM supplemented with physiologic plasma concentrations of 5 mM glucose and/or 0.5 mM L-glutamine (Corning).

Cell assays
For cell proliferation assays, HPDE-iKRAS cells were plated onto white opaque 96-well microplates in quadruplicate. Cell density was assayed by CellTiter-Glo (Promega) using a Wallac Victor2 Multilabel Counter (PerkinElmer) at multiple the indicated time intervals. All data were assessed by exponential growth curve fitting analysis using Prism 4 (GraphPad).

Animal studies
Animal studies were conducted in accordance with an approved Institutional Animal Care and Use Committee protocol at Baylor College of Medicine. HPDE-iKRAS G12D cells transduced with GFP or MAGEA6 lentivirus (in pLenti6.3 backbone) or BxPC-3 cells transduced with shRNAs were subcutaneously injected 1 Â 10 6 ( Figure 6A, B and C) into both flanks and 2 Â 10 6 ( Figure 6D) cells per site into single flank of female nude animals (Harlan) following suspension in Hank's Balanced Salt Solution at 1:1 with Matrigel (BD Biosciences). The following MISSION pLKO shRNAs were used to knock down MAGEA6: shRNA #4: TRCN0000151826 and shRNA #10: TRCN0000155661. For ATG7 and VPS34 knockdown, we used MISSION pLKO shRNAs: TRCN0000007584 and TRCN0000196247 respectively. The slides were scanned by a Pannoramic 250 Flash whole slide digital scanner (3DHISTECH).

Immunohistochemistry
Murine lung tissues or subcutaneous tumors were excised, washed in PBS, and fixed in formalin for 48 hr. Once fixed, samples were dehydrated in graded ethanol series followed by xylene. Samples were then embedded in paraffin, sectioned onto slides (5 mm thick), and allowed to dry. For immunohistochemical staining, slides were deparaffinized using a standard procedure (xylene for two washes, 100% ethanol for two washes, 95% ethanol, 70% ethanol, and 50% ethanol). Slides were then processed using VECTASTAIN Elite ABC HRP Kit and DAB Peroxidase Substrate Kit (Vector Laboratories, PK-6101, SK-4100). Briefly, after the deparaffinizing procedure (described above), antigen retrieval was completed using 0.01 M sodium citrate buffer (pH 6.0), for 15 min at 95˚C, followed by blocking with 0.3% H 2 O 2 for 30 min. Slides were stained for LC3B antibody (1:4000, Novus Biologicals; NB100-2220) and processed using the VECTASTAIN and DAB kits in accordance with the manufacturer's instructions. LC3B staining was quantified using Immunohistochemistry (IHC) Image Analysis Toolbox on ImageJ following the instructions from the plugin developers.

Evolutionary conservation study
Conservation of protein sequences was analyzed via the ConSurf (http://consurf.tau.ac.il/); (Ashkenazy et al., 2016) online server. For proteins with a known 3D experimental structure, we used the known 3D structures as the starting point. Otherwise, we used protein sequences as input. ConSurf performed Basic Local Alignment Search Tool to collect homologous sequences with default parameters. Normalized conservation scores were reported for each protein sequence.

Kaplan-Meier survival analysis
Lung cancer patient survival was analyzed using KM plotter (Győ rffy et al., 2013) with MAGEA6specific probe: 214612_x_at. Patients with high and low MAGEA6 expression were divided by population median. Ethics Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee (IACUC) protocols (AN-5428) of Baylor College of Medicine. The protocol was approved by the Committee on the Ethics of Animal Experiments of Baylor College of Medicine. All surgery was performed under isoflurane anesthesia, and every effort was made to minimize suffering.  Table. . Supplementary file 2. List of clinical trials of cancer immunotherapy targeting MAGEA antigens. Clinical trials of MAGEA-targeted immunotherapy reported from clinicaltrials.gov.

Decision letter and Author response
. Supplementary file 3. MAGEA variants reported in skin, lung and pancreas cancer patients. Recurrent mutations of MAGEA3, A4, A6, A10 and A12 genes identified from skin, lung and pancreas cancer patients in ICGC, TCGA and COSMIC databases.
. Transparent reporting form Data availability All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for Figures 1.
The following datasets were generated: