Pancreatic ductal adenocarcinoma (PDAC) is driven by metabolic changes in pancreatic cells caused by oncogenic mutations and dysregulation of p53. PDAC cell lines and PDAC-derived xenografts grow as a result of altered metabolic pathways, changes in stroma, and autophagy. Selective targeting and inhibition of one of these may open avenues for the development of new therapeutic strategies. In this study, we performed a genome-wide siRNA screen in a PDAC cell line using endogenous autophagy as a readout and identified several regulators of autophagy that were required for autophagy-dependent PDAC cell survival. Validation of two promising candidates, MPP7 (MAGUK p55 subfamily member 7, a scaffolding protein involved in cell–cell contacts) and MDH1 (cytosolic Malate dehydrogenase 1), revealed their role in early stages of autophagy during autophagosome formation. MPP7 was involved in the activation of YAP1 (a transcriptional coactivator in the Hippo pathway), which in turn promoted autophagy, whereas MDH1 was required for maintenance of the levels of the essential autophagy initiator serine–threonine kinase ULK1, and increased in the activity upon induction of autophagy. Our results provide a possible explanation for how autophagy is regulated by MPP7 and MDH1, which adds to our understanding of autophagy regulation in PDAC.
This study identifies and characterizes MPP7 and MDH1 as novel regulators of autophagy, which is thought to be responsible for pancreatic cancer cell survival.
Pancreatic ductal adenocarcinoma (PDAC) is a cancer of unmet need with a median patient survival of only 6–9 months (1). Key PDAC features include a high rate of activating KRAS mutations, a hypervascular and hypoxic microenvironment, and reprogramming of cellular metabolism (2).
A number of studies have linked autophagy to PDAC survival and progression. Autophagy is constitutively activated in PDAC cell lines and tumors (3). This is required for tumor cell survival, as demonstrated by either pharmacologic or genetic autophagy inhibition resulting in loss of viability in cell models and PDAC xenografts (4). Despite current advances, PDACs become resistant to most therapies, and new treatment avenues are urgently needed.
Macroautophagy (hereafter autophagy) is an essential, evolutionarily conserved membrane-mediated process that delivers cytoplasmic constituents in double-membrane autophagosome vesicles to lysosomes for degradation, energy release, and component recycling (5). Mammalian autophagy is mediated by at least 18 autophagy (ATG) proteins acting in a concerted hierarchy (6). The mammalian ATG8s, which include LC3B, are used to monitor autophagosome formation and autophagic flux (7). Autophagy has a homeostatic role under basal conditions, eliminating damaged organelles and misfolded proteins that may otherwise diminish cellular fitness and integrity. Autophagy is also induced in response to stimuli, including starvation, hypoxia, and other stresses, to ensure cell survival.
The role of autophagy in cancer is complex and context-dependent with evidence for both tumor survival and tumor-suppressive roles. Autophagy is thought to be required for anticancer immunosurveillance and inhibition of malignant transformation. However, in established tumors, autophagy can also provide a means for tumor cell survival and therapy resistance (8). In PDAC, autophagy is constitutively activated (9) and required for tumor development, metabolism, and growth (4, 10).
Although the importance of autophagy in cancer is established, the process is still not fully understood, and it remains unclear how manipulation of autophagy in PDAC should be optimally deployed for clinical benefit. There is therefore a need for discovery of novel autophagy regulators, as the enhanced understanding of autophagy may yield new therapeutic targets.
To identify the novel targets, we performed a genome-wide loss-of-function screen to identify genes that regulate autophagy in PDAC cells. Regulators identified through the screen will elucidate molecular mechanisms governing tumor cell survival and autophagy. We report that levels of MPP7 and MDH1 influence autophagic flux under basal and hypoxic conditions. Our ATG protein pathway mapping experiments show MPP7 and MDH1 influence the earliest stages of autophagy. Furthermore, our results suggest that YAP1 may contribute to the positive regulation of autophagy by MPP7, and reveal a functionally important role for MDH1 in maintaining cellular ULK1 levels. Exploiting the apparent sensitivity of PDAC to autophagy inhibition provides a rational basis for the consideration of these novel autophagy regulators as therapeutic targets.
Materials and Methods
Cell culture and reagents
Cell lines were ordered from ATCC (BxPc-3, Panc 02.03, Panc 10.05, Panc-1, PL45), Riken (KLM-1, KP4-3, PK-1, PK-45H), JCRB (SUIT-2, KP-4, KP-2), and DSMZ (HUPT3, YAPC, PA-TU-8902, PA-TU-8999, DAN-G), expanded, and fluorescence-based Mycoplasma testing was performed, followed by agar culture confirmation. A large batch of cells was frozen. Cells were passaged twice a week and used until 20 passages. All PDAC cell lines were grown in full medium: DMEM with 10% FCS and 4 mmol/L l-glutamine.
PK-1 cell lines stably expressing Tet-On–inducible myc-tagged MDH1 and Tet-On–inducible HA-tagged MPP7 were maintained in DMEM with 10% tetracycline-free FCS, 4 mmol/L l-glutamine, and 1 μg/mL puromycin.
Where indicated, cells were treated with 100 nmol/L Bafilomycin A1 (BafA1; Calbiochem), 100 nmol/L Torin1 (Cayman Chemical), 100 nmol/L epoxomycin (Sigma) or for 24 hours with 1 μg/mL doxycycline (Takara Bio). For autophagy induction by amino acid starvation, cells were washed three times in Earle's Balanced Salt Solution (EBSS) and incubated in EBSS for 4 hours.
Lipofectamine 3000 (Thermo Fisher Scientific) was used for transient transfection of PK-1 cells according to the manufacturer's instructions. DNA plasmids were used at a concentration of 1.5 μg/mL of transfection mix.
siRNA transfection was carried out for 72 hours with a final concentration of 50 nmol/L siRNA, using Lipofectamine RNAiMAX (Thermo Fisher Scientific) according to the manufacturer's instruction. All siRNA oligos used were purchased from Dharmacon. For knockdown and rescue experiments, reverse siRNA transfection was performed on day 1, and transfection with DNA 48 hours later, 24 hours prior to cell harvesting. Immunoblotting was carried out as described previously (11).
Confocal and epifluorescence microscopy were carried out as described previously (12) and LC3 puncta formation was quantified by Imaris image analysis software.
The siGENOME SMARTpool human siRNA genome library and deconvoluted siGENOME SMARTpools (Dharmacon) were used in triplicate at a final siRNA concentration of 18.75 nmol/L, with 0.0075 μL/well of INTERFERin-HTS reagent (Polyplus), and 600 cells/well in a final volume of 50 μL/well in black and clear-bottom 384-well plates (Greiner). siRNA libraries were aliquoted on a Beckman FX liquid handling station and all other liquid addition steps performed with a FluidX Xrd-384 plate dispenser. Following reverse siRNA transfection, plates were incubated at 37°C in 5% CO2 for 72 hours, with BafA1 at a final concentration of 50 nmol/L present for the final 6 hours of transfection. Cells were then fixed in 4% paraformaldehyde (Sigma) for 30 minutes, followed by incubation with methanol for 15 minutes and staining with primary [anti-LC3 5F10 antibody (nanoTools Antikörpertechnik; used at 1:100 dilution in 1% BSA)] and secondary antibody (anti-mouse Alexa Fluor 488 Life Technologies at 1:1,000 dilution in 1% BSA).
Image capture and data analysis
To collect antibody labeling and nuclear staining, a Thermo Fisher Scientific Cellomics Arrayscan Vti (×20 magnification, 0.4 NA, BGRFR filter set, X1 camera, LED illumination) and accompanying image analysis software were used (8 fields scanned per well). The parameters recorded were: number of nuclei (object count) from the DAPI channel; number of LC3 spots/cell; total spot count; total spot area; average spot area; total spot intensity; and average spot intensity. More detailed descriptions of algorithm parameters are available upon request.
Raw data for the individual parameters were normalized to correct for plate, row, and column effects, then divided by the individual plate median absolute deviation, thus creating a Z-score. Replicates were summarized by taking the median Z-score for each siRNA pool and normalized scores for the three parameters were then combined into a ranked list, thereby determining the top 200 spot decreasers for the primary siGENOME screen. For the secondary deconvolution screen, results are expressed as percentage of control with respect to the RISC-free control (median of three replicates).
For the promising targets generated from the genome-wide siRNA screen experiments, molecular interaction information comprising protein–protein interactions (PPI) and transcriptional regulatory interactions (TRI) was retrieved from the Autophagy Regulatory Network (ARN; ref. 13). PPIs and TRIs were also retrieved from IntAct (14), HPRD (15) and HTRI (16), ORegAnno (17) databases, respectively. In addition, the annotation (core autophagy protein, posttranslational regulator, transcriptional regulator, etc.) of the target proteins was also derived from ARN. To select the targets that could have the highest effects on autophagy, they were ranked on the basis of the cell viability Z-score (≥−1.5 and ≤1.5), expression status in pancreatic cancer (source: Human Protein Atlas; ref. 18), presence of core autophagy interactors, and visual imaging. From the abovementioned analysis, the top 25% of the hits identified as being relevant modulators of autophagy in pancreatic cancer were selected. On the basis of the effects on long-term cell viability, targets with lethal effects were discarded. To further shortlist the targets, we developed an aggregated score for each target to determine its cumulative effect on autophagy. The aggregated score was based on read-outs from the experimental screens (measuring the differential effects on viability of autophagy-dependent and autophagy-independent PDAC cells) in addition to possible modes of action inferred from the interaction networks. We used PPIs and TRIs to determine whether there are any direct or indirect connections from the shortlisted targets to any of the 37 core autophagy proteins. The aggregated score was then calculated for each shortlisted target based on the expression (source: Human Protein Atlas; ref. 18) and methylation levels (source PCMdb; ref. 19) in pancreatic cancer, acumen cell number, and cell viability scores of the intermediary proteins connecting the targets with the core autophagy proteins. The average aggregated score for each target was then calculated by dividing the aggregated score by the total number of intermediary proteins through which the target could potentially mediate its effects on the core autophagy proteins. Targets with no direct or indirect interactions (via intermediary proteins) with core autophagy proteins were discarded from further analysis.
High-throughput growth assay
Single siRNA oligos (50 nmol/L final concentration) were reverse transfected in PK-1, PA-TU- 8902, KP-4, and SUIT-2 cells in a 96-well format using 0.03 μL Lipofectamine RNAiMAX/well. Three cell numbers were plated for each cell line. For PK-1 cells, 750, 1,000, and 2,000 cells were seeded/well, for PA-TU-8902 500, 1,000, and 2000 cells/well, for KP-4 250, 500, and 1,000 cells/well, and for SUIT-2 125, 250, and 500 cells/well. Each hit-targeting siRNA was tested in quadruplicate wells for each cell number. Control siRNAs were tested in duplicate wells. Seven days after transfection, cells were fixed using 4% PFA and cell proliferation was assessed by Hoechst staining and scanning on TTP Labtech's acumen Cellista.
Seventy-two hours posttransfection, qPCR samples were harvested and RNA was extracted using the RNeasy Mini Kit (Qiagen), followed by cDNA production using SuperScript II Reverse Transcriptase (Life Technologies). qPCRs were performed with Fast SYBR Green Master Mix (Thermo Fisher Scientific) and relative mRNA expression levels were calculated.
MDH1 activity assay
Cells were lysed in NP-40 buffer supplemented with a protease inhibitor (Roche) and immunoprecipitated with anti-myc agarose beads (Chromotek) for 2 hours. The immunoprecipitated beads were washed three times with NP-40 buffer, and the on-beads activity of myc-MDH1 was determined by measuring the decrease in the fluorescence of NADH (Ex. 350 nm; Em. 470 nm) on an EnVision 2102 multilabel plate reader. The reaction was monitored at 15-second intervals for 20 minutes. This protocol was adapted from Kim et al., 2012 (#23072).
Annexin V assay
PK-1 and KP-4 cells were harvested into a single-cell suspension, washed once in PBS, and resuspended in 1× Annexin V binding buffer (10× commercial buffer diluted in distilled water, BD Pharmingen, catalog no. 51-6710AK). To each 500 μL of the cell suspension, 2.5 μL of FITC-conjugated Annexin V antibody was added and incubated for 15 minutes at room temperature. To Identify dead cells, 4′,6-Diamidine-2′-phenylindole dihydrochloride (DAPI, Sigma, catalog no. D8417) was added to the samples to a final concentration of 2 μmol/L just before running them on the flow cytometer.
Annexin V signal was detected using a CS&T-calibrated BD LSRFortessa system (Becton Dickinson) configured with a 355 nm, 405 nm, 488 nm, 561 nm, and 633 nm laser excitation lines and associated detector arrays. Annexin V-FITC was detected in the 530/30 blue detector and DAPI was detected in the 450/50 UV detector. The acquisition gates were set using FACS DIVA software (v 8.0.1, Becton Dickinson) and a minimum of 30,000 single events were collected. Postacquisition analysis was performed using FlowJo software (version 10.1, Tree Star, Inc.). Briefly, intact cells were selected on FSC-A v SSC-A plot. Single cells were then selected using the area and the height from SSC. Live, apoptotic, and necrotic/dead cells were identified using a spider quadrant gate.
Organoid formation assay
The primary murine PDAC cells were isolated from Pdx1-Cre, KRasG12D, p53 fl/fl mouse pancreata (20) when tumors were palpable, and grown as three-dimensional organoids in Matrigel as described previously (21). All experiments performed in animals were approved by the Francis Crick Institute's Animal Welfare and Ethical Review Body, and conformed to UK Home Office regulations under the Animals (Scientific Procedures) Act 1986 including Amendment Regulations 2012. Cells were cultured in advanced DMEM/F12 medium with 10% FCS, penicillin/streptomycin (1×), 10% Noggin-condition media, 10% R-spondin–conditioned media, B27 Supplement (1×), nicotinamide (10 mmol/L), N-acetyl cysteine (1.25 mmol/L), EGF (0.05 μg/mL), FGF (0.05 μg/mL), and Gastrin (0.01 μmol/L). pGIPZ shRNAs targeting MPP7 and Atg5 (Dharmacon) were subcloned into the doxycycline-inducible pTRIPZ vector. Stable KPC organoids expressing each construct were then created using lentiviral transduction, passaged three times, and equal cell numbers for each construct were sorted on a BD FACSAria Fusion (BD Biosciences) into 96-well plate wells containing Matrigel. Organoid growth was assessed 6 days after the FACS sort, and organoid number quantified with ImageJ.
Genome-wide screen identifies novel regulators of autophagy in PDAC cell lines
To identify novel regulators of autophagy in PDAC, a multistep screening and validation approach was undertaken (see workflow in Fig. 1A). The aim of the screen was to identify the proteins essential for autophagy, measured by LC3 puncta levels upon BafA1 treatment, which is an inhibitor of the vacuolar-type H+-ATPase found in endosomes and lysosomes (7). BafA1 was used as a tool to measure the accumulated LC3 as it prevents its degradation in autophagosomes. LC3 puncta form upon activation of autophagy by the canonical pathway initiated by ULK1 activation and production of phosphatidylinositol-3-phosphate (PI3P; ref. 5), and recruitment of WIPI2 (12), leading to lipidation of LC3 (thus forming LC3-II) on autophagosomes. Under steady-state conditions, basal autophagy proceeds using the same machinery, but typically at a reduced level.
A selection was made of suitable cell lines for high-throughput genome-wide screening, and subsequent validation and hit selection. We aimed to identify an appropriately adherent cell line with high autophagic flux. Seventeen PDAC cell lines were assayed in normal growth medium with and without BafA1. LC3 puncta were detected by antibody labeling, and LC3-II was detected by Western blot analysis. Autophagic flux was determined by the fold induction of LC3 puncta and LC3-II upon BafA1 treatment (Fig. 1B and C). On the basis of its high autophagy flux and suitability for imaging in the high-throughput format, the KP-4 cell line was selected for the siGENOME screen.
The siGENOME screen was performed in triplicate under 6 hours of BafA1 treatment, using a whole-genome SMARTpool siRNA library. After fixation, multiple images of each well were taken, and endogenous LC3 puncta and nuclei were identified (Fig. 1D). Effects of each siRNA SMARTpool on LC3 puncta number were assessed by measuring related parameters, such as number, intensity, and area. Knockdown of known essential autophagy regulators such as ATG4B and BECN1 decreased the measured parameters as expected, for example, spot total area per object (STAPO) Z-score (Fig. 1E). Normalized Z-scores for the various LC3 (Ch2) parameters were combined for each gene to give a weighted binary score and allow us to create a ranking of autophagy regulators. Numerical values of the normalized Z-scores for these parameters for each siRNA SMARTpool of the top 200 positive autophagy regulators are presented in Supplementary Table S1.
Next, in the deconvolution screen, the top 200 puncta-decreasing targets from the genome-wide primary screen were screened again with four siRNAs for each target to confirm the primary screen results (Supplementary Table S2). The results of the deconvolution screen were subsequently integrated with the results of the primary screen. To identify robust autophagy modulators, we used systems and network biology concepts to rank the shortlisted hits. While the LC3 puncta-related parameters were derived from the primary screens, we carried out a molecular interaction network–based bioinformatic analysis to identify how the selected targets could affect autophagy either directly via interactions with core autophagy proteins or indirectly via other intermediary proteins. As a source of molecular interactions specific to autophagy, we used the ARN resource (13). In addition, expression of the analyzed targets in pancreatic cancer was retrieved from the Human Protein Atlas database (18) and used to incorporate context, specificity in the ranking process. Inferences from the visual interrogation of the primary image data were also incorporated to increase confidence in the primary screen hits. We selected the top 25% of the hits from the above analysis to further proceed with the shortlisting and experimental verification. Hence, using multiple indicators such as LC3 puncta-related parameters, molecular network–related properties, context specificity, and visual imaging, we created a shortlist of relevant and probable autophagy modulators in the context of pancreatic cancer and their possible interactions.
Identification of autophagy regulators that affect autophagy-dependent PDAC cell proliferation
A number of previous studies suggested a role for autophagy in pancreatic cancer cell survival (4, 10). We determined whether the survival of the 17 PDAC cell lines we studied was dependent upon autophagy. To achieve this, CRISPR knockout of the essential autophagy genes Atg5 and Atg7 was performed, followed by cell proliferation measurements. While multiple cell lines show dependence on autophagy, PK-1, PA-TU-8902, and KLM-1 show the greatest dependency on autophagy for survival, while the remaining cell lines tested, KP-4, SUIT-2, and others (Fig. 1F–M; Supplementary Fig. S1A–S1L), are much less dependent on autophagy for survival.
A 7-day proliferation assay was then carried out in the autophagy-dependent PK-1 cell line to establish how siRNAs knock-down of the top 25% autophagy regulators we identified would affect cell survival (Fig. 2A). Note, two siRNAs were selected from the deconvolution screen plus a number of controls. qPCR was carried out in parallel with the growth assay, and siRNAs that depleted the target mRNA levels with 50% efficiency or higher were considered to have sufficient knockdown efficiency (Supplementary Table S3). Depletion of the majority of autophagy regulators caused some decrease in the survival of autophagy-dependent PK-1 cells compared with RISC-free siRNA (Fig. 2A). Eleven targets were identified to have sufficient knockdown and cell survival, reduced by 50% or more (Fig. 2B).
Subsequently, we determined which of the 11 targets regulate autophagy in the autophagy-dependent PK-1 cell line. Note, siRNAs for the 11 targets were chosen based on qPCR data, and not all targets had 2 suitable siRNAs (GALNT1, GRIA3, MPP7, OPHN1, and UBR1 were only targeted by a single siRNA). After BafA1 treatment, the effect of hit depletion on LC3 puncta number and intensity was quantified (Fig. 2C; Supplementary Table S4). We next assayed whether the targets were differentially required for survival in autophagy-dependent versus autophagy-independent cells using a long-term growth assay. Survival after siRNA depletion was tested in KP-4 and SUIT-2, two autophagy-independent, and PK-1 and PA-TU-8902, two autophagy-dependent cells lines (Fig. 2D). Western blot analysis of LC3-II levels and qPCR to determine knockdown efficiency for 11 targets is summarized in Supplementary Fig. S2.
Using the three criteria shown in Fig. 2B–D, MDH1 and MPP7 were identified as among the best regulators. Knockdown of MPP7 and MDH1 (02) reduced the proliferation of PK-1 cells (Fig. 2B), decreased LC3 puncta (Fig. 2C), and finally showed an enhanced loss of proliferation on the autophagy-dependent cell lines (PK-1 and PA-TU-8902) compared with the autophagy-independent cell lines (KP-4 and SUIT-2). In addition, an important role for MDH1 in PDAC has been described (22–24), while MPP7 was relatively uncharacterized. Two other shortlisted candidates (UBR1 and GALNT1) proved to be less robust in the follow-up assays.
MPP7 regulates basal and hypoxia-induced autophagy in PDAC
MPP7, MAGUK p55 subfamily member 7, has been shown to be required for epithelial tight junction formation (25) and for myoblast cell proliferation (26). A single siRNA duplex was used to knockdown MPP7 in PK-1 cells, which inhibited LC3 lipidation both during amino acid starvation and BafA1 treatment (Fig. 3A and B). Transient expression of an siRNA-resistant mutant of MPP7 was able to rescue the levels of LC3-II after MPP7 knockdown (Fig. 3C). These data confirm that MPP7 is a positive regulator of starvation-induced autophagy in PDAC cells. Using a stable, inducible PK-1 cell line expressing HA-tagged MPP7, we could modulate MPP7 levels. A significant increase in LC3 puncta was seen upon MPP7 overexpression, compared with the doxycycline-treated controls (Fig. 3D and E).
We determined whether MPP7 is required for hypoxia-induced autophagy as PDAC tumors exhibit a hypoxic microenvironment due to poor vasculature (27). High levels of hypoxia predict poor patient prognosis (28), hence a potential drug target may be especially useful. Hypoxia is a known activator of autophagy (29), and in our experiments, both LC3-II levels and LC3 puncta increased in hypoxia-exposed cells (Fig. 3F–I). MPP7 depletion results in a striking decrease of both LC3-II levels and LC3 puncta in hypoxia-exposed cells (Fig. 3F–I), suggesting that MPP7 positively regulates not only basal, but also hypoxia-induced autophagy.
To identify the stage of autophagy MPP7 exerts its regulatory effect, ATG pathway mapping was performed under amino acid starvation. During the initiation of starvation-induced autophagy, inactivation of mTORC1 and activation of ULK1 kinase occurs (30). ULK1 kinase activity is repressed by mTORC1 in nutrient-rich conditions, whereas starvation inhibits mTORC1 kinase activity, resulting in decreased ULK1 Ser757 phosphorylation and consequent ULK1 activation (31). To assess whether MPP7 is required for mTORC1 deactivation, phosphorylation of ULK1 Ser757 was measured during starvation. We found that MPP7 depletion appears to result in a slightly faster rate of ULK1 Ser757 dephosphorylation relative to controls (Fig. 3J).
Downstream of mTORC1 and ULK1/2 in the autophagic pathway, WIPI2, a PI3P effector, is recruited to the forming autophagosome and mediates the recruitment of ATG12–ATG5–ATG16L1 complex, which is required for LC3 lipidation (12). WIPI2 then dissociates from formed autophagosomes. WIPI2 puncta formation is used to assess the recruitment of class III PI3K lipid kinase complex I (7), a critical early requirement for autophagosome formation. MPP7 depletion significantly reduces WIPI2 puncta number under conditions of starvation (Fig. 4A and B), providing further support that MPP7 may regulate autophagy at the initiation stage, and in particular PI3P levels.
MPP7 regulates autophagy through YAP1 activation
On the basis of bioinformatics analysis of MPP7 in the ARN (13), we predicted that YAP1 (Yes-associated protein 1), a transcriptional regulator involved in cell proliferation and apoptosis suppression, may be involved in the regulation of autophagy by MPP7. Previous findings indicate that MPP7 is required for YAP1 accumulation in the nucleus, where it is transcriptionally active (26). Furthermore, YAP1 increases cellular autophagic flux in breast cancer cells, promoting breast cancer cell survival (32). We confirmed that YAP1 is required for both basal and starvation-induced autophagy in PK-1 cells (Fig. 4C and D), as YAP1 depletion coincides with a reduction in LC3 lipidation, both in fed and starved BafA1-treated cells. In addition, YAP1 depletion reduces hypoxia-activated autophagy (Fig. 4E).
We observed depletion of MPP7 results in accumulation of YAP1, phosphorylated at S127 (Fig. 4F), which is the cytoplasmic, inactive form of YAP1, confirming MPP7 is required for YAP1 activation (26). Overexpressed YAP1 in MPP7-depleted cells resulted in a rescue of autophagic flux (Fig. 4G and H). Interestingly, the regulation of YAP1 activity and phosphorylation by MPP7 seems to be autophagy dependent, as ATG13 depletion appears to deactivate YAP1 (Fig. 4I). Furthermore, in stable cell lines expressing WT and inactive S94A YAP1, inactive S94A YAP1 is unable to rescue autophagy (Fig. 4J). In summary, our data demonstrate that MPP7 may positively regulate YAP1 activity in PDAC cells, and this may contribute to the positive regulation of autophagy by MPP7.
MDH1 regulates basal and hypoxia-induced autophagy in a PDAC cell line
MDH1 is known to be activated in human PDAC samples (23) and supports pancreatic cancer growth (22), and our findings indicate that MDH1 is required for both autophagy and survival in autophagy-dependent PDAC cells (Fig. 2).
Like MPP7, MDH1 is required for both basal and starvation-induced autophagy: LC3 lipidation is reduced in starvation and BafA1 treatment when MDH1 is depleted (Fig. 5A and B). Conversely, induction of myc-MDH1 using a stable doxycycline-inducible PK-1 cell line significantly increased the number of LC3 puncta, both with and without depletion of the endogenous protein (Fig. 5C–E). MDH1 depletion significantly reduced hypoxia-induced autophagy, as demonstrated by a decrease in both LC3 lipidation and LC3 puncta (Fig. 5F–I).
MDH1 regulates autophagy by maintaining ULK1 levels, and MDH1 is activated by autophagic stimuli
We next analyzed the effect of MDH1 depletion on autophagy during amino acid starvation. Similar to MPP7, ULK1 pSer757 levels were reduced in amino acid starvation (Fig. 6A). Pathway mapping analysis was carried out for MDH1 and similarly to MPP7, an MDH1 reduction coincided with reduced numbers of WIPI2 puncta under amino acid starvation, suggesting that MDH1 could also regulate autophagy, in part, through regulation of PI3K activity (Fig. 6B and C).
However, we observed that ULK1 levels were reduced by depletion of MDH1 (Fig. 6A, D, and E). Conversely, induction of myc-MDH1 increased ULK1 levels (Fig. 6F and G). Interestingly, under conditions of MDH1 depletion, inhibition of proteasome activity by epoxomycin treatment resulted in increased ULK1 levels, suggesting MDH1 may regulate ULK1 proteasomal degradation (Fig. 6H). In support of this conclusion, epoxomycin treatment rescued autophagy under conditions of MDH1 depletion (Fig. 6I and J).
We tested whether MDH1 regulation of ULK1 proteasomal degradation is specific to ULK1, or whether MDH1 also regulates proteasomal degradation of other proteins such as p53 and Mcl-1 (Fig. 6K). We found that while proteasomal degradation of p53 does not appear to be regulated by MDH1, levels of Mcl-1, another protein involved in autophagy, decrease on MDH1 depletion, and is rescued by proteasome inhibition. This suggests that the loss of MDH1 results in alterations of signaling upstream of PI3K, and downstream of mTORC1, but MDH1 may regulate degradation of other proteins than ULK1.
Cytoplasmic MDH1 controls the production of malate by reduction of oxaloacetate, using the reducing power of NADH (33). MDH1 exists in two forms–the monomeric form, which is enzymatically inactive, and the active dimeric form (34). We found that autophagy induction through amino acid starvation and Torin treatment (to inactivate mTORC1) resulted in higher levels of the active form of MDH1. We found both higher levels of the dimerized form of MDH1 (Fig. 7A–C), and an increased enzymatic activity of MDH1 as measured by assessing NADH conversion to NAD (Fig. 7D and E). Hypoxic conditions also caused an increase in MDH1 dimer formation (Fig. 7F and G). These results suggest that MDH1 enzymatic activity is linked to its role in positively regulating autophagy.
MDH1 is monomethylated on Arg248, and this posttranslational modification prevents dimerization and inhibits the catalytic activity of MDH1 (24). Mutation of the arginine (R248) to a phenylalanine (F) to mimic methylation (35) has previously been shown to reduce the catalytic activity of MDH1 (24). Given the activation of MDH1 during autophagy induction, we determined whether the methylation of MDH1 affects its ability to regulate autophagy. Doxycycline-inducible myc-tagged MDH1 R248F cell lines were compared with the inducible myc-MDH1–expressing cells. Whereas myc-MDH1 was able to rescue autophagy and ULK1 levels in siRNA-depleted cells (Fig. 7H and I), expression of the less catalytically active R248F mutant did not rescue autophagy levels (Fig. 7J and K), suggesting that the catalytic activity of MDH1 is required for its role in autophagy.
Levels of MDH1, MPP7, and downstream autophagy regulators ULK1 and YAP1 affect PDAC cell survival
Using an organoid model of mouse PDAC (KRasG12D, p53 fl/fl), and organoid formation as a readout for “cancer stemness” and self-renewal (20), doxycycline-inducible knockdown of MPP7 and MDH1 reduced both organoid growth and autophagy levels (Fig. 7L–N). This provides further evidence for MPP7 and MDH1 being required for PDAC cell survival, not just of immortalized human PDAC cell lines, but also of primary organoids, a physiologically more relevant model system (21).
Furthermore, colony formation assays show that MPP7 and MDH1 depletion and depletion of their downstream autophagy regulators YAP1 and ULK1 reduces proliferation of the autophagy-dependent PK-1 PDAC cell line. The reduction in viability of the autophagy-independent KP-4 cell line is much less striking when these proteins are depleted (Supplementary Fig. S3A and S3B). To support this, Annexin V assays demonstrate that the population of necrotic and apoptotic cells increases when the autophagy regulators MDH1, MPP7, YAP1, and ULK1 are depleted (Supplementary Fig. S3C and S3D).
Increasing evidence that PDAC cells require autophagy for survival is supported by studies where pharmacologic or genetic inhibition of autophagy caused xenograft regression and cell death (4). However, the mechanisms through which autophagy in PDAC is regulated remain to be determined.
As autophagy is a vital process for cellular homeostasis and plays an important role in malignancies including PDAC, it is of both basic scientific and translational interest to decipher additional autophagy regulators. Previously, high-throughput screening has proven to be a fundamental tool enabling key regulatory protein identification (36, 37). These screens have been performed using overexpressed markers, such as GFP-LC3, which might result in a bias toward overexpression artifacts.
We identified and validated a number of new autophagy regulators in four PDAC cell lines through an unbiased loss-of-function screens with a quantifiable endogenous autophagy read-outs. Our hit-selection strategy involved choosing autophagy regulators that were required for autophagy-dependent PDAC cell survival and not autophagy-independent PDAC cell survival. This approach revealed new roles for MDH1 and MPP7. Extensive verification of MPP7 and MDH1 as autophagy regulators used a range of experimental approaches and established that MDH1 and MPP7 impact autophagy under both basal and hypoxic conditions. It is important to note that while this study has been performed in PDAC cell lines and organoids, we have not tested the effect of MDH1, MPP7 and downstream proteins in other cancer cell types, or in healthy cells, so we cannot rule out that these proteins regulate autophagy in other contexts.
We propose that MDH1 functions to maintain cellular ULK1 levels, the key autophagy initiator, by activation of the class III PI3K and by mediating ATG9 trafficking (38, 39). Our results further suggest that MDH1 prevents proteasomal degradation of ULK1, which has previously been shown to be a mechanism of controlling autophagy (40, 41).
MDH1 is overexpressed in several cancers (41) and contributes to PDAC cell proliferation and metabolism via NAD production to support glycolysis (23, 24). We present a new role for MDH1 in PDAC, where MDH1 contributes to PDAC survival by promoting autophagy. Furthermore, the catalytic activity of MDH1 may be linked with its ability to indirectly activate autophagy (Supplementary Fig. S3E). In addition to its direct effect via ULK1, MDH1 becomes activated upon autophagic stimuli such as starvation and hypoxia (Supplementary Fig. S3F).
Our study reveals MPP7 as a novel positive regulator of PDAC cell survival and autophagy, both in immortalized human pancreatic cancer cells and primary PDAC organoids. A mechanism through which MPP7 regulates autophagy involves YAP1, a transcriptional activator with an established function in PDAC. YAP1 levels are elevated in pancreatic cancer compared with normal pancreas (42), and YAP1 promotes pancreatic cancer in mice via activation of JAK–STAT3 signaling (43). We propose that MPP7 promotes YAP1 activation in PDAC cells, which is supported by data in skeletal muscle stem cells (26). Further support of our hypothesis that YAP1 is required is that two members of the HIPPO pathway, MOBKL1A (ranked 136) and TEAD3 (ranked 67), are in the top 200 autophagy regulators (see Supplementary Table S2). Other data to support of a role for the HIPPO pathway is an identification of LC3 as a substrate for the HIPPO kinases MST1/MST2 (44). The role of YAP1 in autophagy requires further investigation, in particular, how MPP7 regulates YAP1, and whether YAP1 activates autophagy via its transcriptional activator role or an unidentified cytoplasmic function.
In conclusion, our study highlights MPP7 and MDH1 as key autophagy regulators in PDAC. In hypoxia, both MDH1 and MPP7 are required for autophagy, and MDH1 becomes activated in response to hypoxia. This is of particular translational relevance as poor vasculature and the resulting hypoxic microenvironment are seen in human PDAC (27). Given the importance of autophagy in PDAC cell survival and PDAC progression, this information provides a significant step forward in developing approaches to exploit the sensitivity of PDAC malignancy to autophagy inhibition.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: M. New, T.V. Acker, K.M. Ryan, S.A. Tooze
Development of methodology: M. New, T.V. Acker, T. Korcsmaros, K.M. Ryan, M. Howell, S.A. Tooze
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. New, T.V. Acker, J.-I. Sakamaki, M. Jiang, J. Long, V.M.-Y. Wang, A. Behrens, J. Cerveira, H.B.J. Jefferies, K.M. Ryan, M. Howell, S.A. Tooze
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. New, T.V. Acker, J.-I. Sakamaki, M. Jiang, R.E. Saunders, P. Sudhakar, T. Korcsmaros, H.B.J. Jefferies, K.M. Ryan, M. Howell, S.A. Tooze
Writing, review, and/or revision of the manuscript: M. New, T.V. Acker, P. Sudhakar, T. Korcsmaros, H.B.J. Jefferies, K.M. Ryan, S.A. Tooze
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.V. Acker, H.B.J. Jefferies, K.M. Ryan, M. Howell, S.A. Tooze
Study supervision: M. New, K.M. Ryan, S.A. Tooze
We thank Scott Vande Pol (Department of Pathology, University of Virginia) for pcDNA3-flag-MPP7; Eric Sahai and Anna Dowbaj (Tumour Cell Biology Laboratory, Francis Crick Institute) for EYFP-YAP1-WT and EYFP-S94A and advice; H.D. Youn (Cancer Research Institute, Seoul National University, Seoul, Korea) for pcDNA 3.1-MDH1-Myc; Louise Fets and Dimitrios Anastasiou for advice on MDH1 enzyme activity assay; and Thomas Mercer and Alex van Vliet for advice and reading the manuscript. We acknowledge the support of Astellas Pharma Inc., and in particular, Kyoko Saita, Hiromichi Kimura, and Takatoshi Soga for scientific discussions. We also thank Chris Baker (Cancer Research Technology) for project management and scientific discussions. The work in Tooze lab (S.A. Tooze, M. New, T.V. Acker ) and the High Throughput Screening Facility (M. Howell, R.E. Saunders, M. Jiang) was supported by Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001187, FC001999), UK Medical Research Council (FC001187, FC001999), and Wellcome Trust (FC001187, FC001999). The work in Ryan lab (K.M. Ryan) was funded by the Cancer Research UK (C596/A17196). M. New, T.V. Acker, J.-I. Sakamaki, and J. Long were funded by Astellas Pharma as part of a sponsored research collaboration. The work in the Behrens lab (A. Behrens and V.M.-Y. Wang) was supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001039), the UK Medical Research Council (FC001039), and the Wellcome Trust (FC001039). T. Korcsmaros was supported by a fellowship in computational biology at Earlham Institute (Norwich, UK) in partnership with the Quadram Institute (Norwich, UK), and was strategically supported by Biotechnological and Biosciences Research Council UK (BB/J004529/1 and BB/P016774/1).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.