Small-Molecule Sigma1 Modulator Induces Autophagic Degradation of PD-L1

This article is featured in Highlights of This Issue, p. 185

Programmed death-ligand 1 (PD-L1, also known as B7-H1 or CD274) is a type I integral membrane glycoprotein that is cotranslationally inserted into, posttranslationally modified in, and transported through the secretory pathway of a range of cell types, including tumor cells (1–4). In this pathway, the endoplasmic reticulum (ER) is the primary site of synthesis, folding, and assembly of secreted and integral membrane proteins. Protein homeostasis in the ER relies on the timely convergence of multiple mechanisms that detect protein–concentration thresholds, provide quality control, and regulate the ebb and flow of ER proteins (4–6). Integral membrane proteins are intrinsically dependent on ER protein homeostasis machinery to reach the cell surface.

PD-L1 expressed at the surface of tumor cells can act as a T-cell inhibitory checkpoint molecule and can inactivate tumor-infiltrating immune cells that express cell surface programmed death-1 (PD-1; also known as CD279; refs. 1, 3, 7). Immune checkpoint inhibitors that block PD-L1/PD-1 interactions are promising therapeutic agents in strategies that exploit antitumor immune responses. However, a relatively small percentage of cancer patients respond to anti–PD-L1/PD-1 monotherapy, which thus far includes subsets of patients with melanoma, non–small cell lung cancer, metastatic bladder cancer, and renal cell carcinoma (7–10). Although high PD-L1 expression has been associated with poor prognosis in a range of cancers (11–13), high PD-L1 expression levels do not necessarily correspond with response to anti–PD-L1/PD-1 therapeutic agents (7, 14). For instance, anti–PD-L1/PD-1 agents to treat prostate and triple-negative breast tumors with high levels of PD-L1 have shown minimal efficacy in early clinical trials (7–10, 12–15). Therefore, the search for predictive biomarkers of response to PD-L1/PD-1 agents is an active area of investigation (14, 16), as is the search for novel agents and combination therapy approaches that modulate the tumor microenvironment to promote antitumor immunity (7, 14–17).

Full-length PD-L1 comprises 229 amino acids leading with an N-terminal signal sequence, IgV- and IgC- extracellular domains, which engage PD-1 on infiltrating immune cells, a single transmembrane domain, and a relatively short 31 residue cytoplasmic tail, which has no defined functional motifs and whose role in PD-L1 biology has not been determined (1–3, 18). The cellular factors and processes engaged in the synthesis, maturation, and transport of PD-L1 are also not well defined, and few PD-L1–regulatory proteins have been identified (1–3, 18).

Sigma1 is a unique ligand-operated integral membrane chaperone or scaffolding protein that is highly expressed in the ER of a range of cancer cell lines (19). Initially thought to be an opioid receptor (19), Sigma1 lacks homology with any known mammalian protein and is a distinct protein, unrelated to any traditional receptor (20, 21). A growing body of evidence demonstrates that inhibition of Sigma1 can suppress growth, proliferation, and induce apoptosis in multiple cancer cell lines (19) and demonstrates the regulation of cell-intrinsic properties of cancer cells by Sigma1; however, antitumor immunity activity induced by Sigma1 inhibition has not been determined (19). Emerging data suggest Sigma1 is a multifunctional chaperone (22) or scaffolding protein (19, 23) involved in maintaining ER protein homeostasis and supporting the increased demand for secretory pathway protein synthesis associated with tumor growth (19, 24). In this regard, we have shown that selective small-molecule Sigma1 modulators can be used to regulate protein translation and activate the unfolded protein response (UPR) and autophagy in a pharmacologically controllable manner (24, 25).

Autophagy describes a set of cellular sequestration and degradation mechanisms by which cells can maintain energy levels under conditions of metabolic stress as well as a mechanism by which large aggregates of misfolded proteins and cellular components are sequestered into membrane-bound vesicles called autophagosomes and subsequently targeted for lysosomal degradation (26, 27). Multiple types of autophagy have been identified corresponding to distinguishing mechanistic features, autophagic cargo proteins, and organelles involved. These include bulk macroautophagy, chaperone-mediated autophagy, secretory autophagy, ubiquitin-selective autophagy, lipophagy, mitophagy, and most recently ER-phagy (26–32). Emerging reports demonstrate that components of the secretory pathway membrane remodeling and protein trafficking machinery contribute to autophagy (28–32).

We previously reported a role for Sigma1 in autophagy in cancer cells and the ability to pharmacologically control autophagy with small-molecule Sigma1 modulators (24). In this study, we asked whether the ER protein homeostasis regulating properties of small-molecule Sigma1 modulators could be exploited to regulate the transport and stability of PD-L1 in cancer cells and thus suppress the surface expression of PD-L1 and potentially abrogate PD-L1/PD-1–mediated inactivation of immune cells.

IPAG (1-(4-Iodophenyl)-3-(2-adamantyl) guanidine) was purchased from Tocris. SA4503 was purchased from Axon Med Chem. Bafilomycin A1, wortmannin, rapamycin, and thapsigargin were purchased from Sigma-Aldrich. Cycloheximide was purchased from Alfa Aesar, and recombinant human Interferon (IFN) was purchased from Gibco. Endoglycosidase H (Endo H) and peptide:N-glycosidase F (PNGase F) were purchased from New England Biolabs. Trypan Blue was purchased from Invitrogen.

MDA-MB-231 and PC3 cells were purchased from ATCC and were authenticated by short tandem repeat profiling. ATG5^+/+ wild-type (WT) and ATG5^−/− knockout (KO) mouse embryo fibroblasts (MEF) described elsewhere were obtained from Dr. Noboru Mizushima via RIKEN (33). All cancer cell lines were maintained in RPMI1640 supplemented with 10% FBS (Corning), and ATG5^+/+ and ATG5^−/− MEFs were grown in DMEM supplemented with 10% FBS (Corning). Cells were seeded approximately 24 hours prior to the start of drug treatment in most assays. Human Sigma1 siRNA (sc-42250) and control-A siRNA (sc-37007) were purchased from Santa Cruz Biotechnology. siRNA transfections (100 nmol/35-mm well) were performed with INTERFERin (PolyPlus) transfection reagent according to the manufacturer's procedures. Forty-eight hours later, cells were reseeded, allowed to attach and recover for 24 hours, and transfected again. Cells were harvested 48 hours after the second transfection. Sigma1 shRNAs (TRCN0000296908 and TRCN0000291305), ATG5 shRNAs (TRCN0000330392 and TRCN0000151474), and ATG7 shRNA (TRCN0000007588) were purchased from Sigma-Aldrich. Infections were done with 48-hour and 72-hour lentivirus and ≤8 μg/mL polybrene (Santa Cruz Biotechnology). Cells were treated 5 days postinfection and harvested after indicated time point. Transient transfections of the pcDNA3.1, C-terminal–tagged Sigma1-HA (34), and C-terminal–tagged PD-L1-FLAG (Sino Biological Inc.) plasmid constructs were performed using jetPRIME transfection reagent (PolyPlus) per the manufacturer's protocol. In these experiments, 2 μg of DNA were transfected per 10-cm plate, and cells were harvested 48 hours posttransfection.

Cell lysis, protein extraction, SDS-PAGE, and immunoblotting were performed as described previously (24), with a few modifications. The Luminata Western HRP Substrate Chemiluminescence Kit (Millipore) was used to reveal immunoblotted proteins. The anti-Sigma1 antibody was generated in our laboratory as described elsewhere (24). The anti–β-actin (C4), anti-GFP (B-2), mouse anti-FLAG (OctA-probe H-5), anti–MHC-I (W6/32), and rabbit and mouse anti-HA (Y-11 and F-7) antibodies were purchased from Santa Cruz Biotechnology. The rabbit anti–PD-L1 (E1L3N XP), mouse anti–PD-L1 (405.9A11), anti–PD-L1 extracellular domain specific (D8T4X), anti-LC3B (D11 XP), anti-ATG5 (D5F5U), rabbit anti-FLAG epitope (DYKDDDDK tag), anti-GAPDH (D16H11 XP), and horseradish peroxidase–conjugated secondary antibodies were purchased from Cell Signaling Technology. The anti-Vinculin antibody (E00153) was purchased from Sigma-Aldrich.

Cells were seeded onto #1.5 (0.17 mm) borosilicate glass coverslips (Electron Microscopy Sciences) coated with 0.1 mg/mL 75,000 to 150,000 MW poly-d-lysine substrate (Sigma-Aldrich) 36 hours prior to fixation. The cells were washed with room temperature Dulbecco's modified PBS solution (DPBS) and fixed in 4% formaldehyde (Pierce) for 15 minutes. After DPBS washes, cells were permeabilized using ice-cold methanol at −20°C for 15 minutes and were washed. Next, cells were subjected to antigen retrieval using a 1% (v/v) citrate-based unmasking solution (Vector) for 10 minutes at 95°C. To reduce nonspecific secondary staining, cells were incubated with Image-iT FX solution (Molecular Probes) for 30 minutes at room temperature and washed with DPBS. Cells were blocked in 5% (v/v) normal goat serum (Invitrogen), 0.3% (v/v) Triton X-100 in DPBS for 1 hour. Primaries diluted in 1% (w/v) BSA, 0.3% (v/v) Triton X-100 in DPBS at 1:200 were added to the coverslips and incubated overnight at 4°C. After the primary stain, cells were washed with DPBS and stained with the corresponding secondary Alexa Fluor 488 goat anti-rabbit IgG (H+L) or Alexa Fluor 594 goat anti-mouse (Molecular Probes) in 1% (w/v) BSA, 0.3% (v/v) Triton X-100 in DPBS for 2 hours at room temperature at 1:1,000. Cells were washed with DPBS, nuclear counterstained with 0.1 μg/mL DAPI (Pierce) for 90 seconds, and then mounted onto glass slides using Prolong Gold (Molecular Probes). Coverslips were sealed using nail polish 36 hours after mounting. Images were acquired using the Olympus FV1000 inverted confocal microscope using a 60 × 1.42 NA oil immersion objective at a scanning resolution of 0.0294 μm/pixel. In the studies involving PD-L1 colocalization with GFP-LC3, five randomly selected microscopic fields were taken for each condition. For the PD-L1 cell surface localization studies, 11 and 12 randomly selected microscopic fields for vehicle and SA4503, respectively, were taken for each condition. Confocal images were median filtered (median intensities of 32 × 3-pixel squares centered at the target pixel were subtracted from the target pixel, respectively) as described previously (35). ImageJ was used to generate 3D surface plots.

The Promega PD-L1/PD-1 Blockade Assay Kit (J1250) was utilized to determine functional changes in PD-1/PD-L1 interactions with Sigma1 modulation. PC3 or MDA-MB-231 cells (1 × 10⁴) were plated per well on Wallac B&W 96-well Isoplates with white wells and bottoms and black matrix. Approximately 24 hours later, drug treatment of IPAG (0–20 μmol/L) was added for 16 hours. The following day, the drug-containing media were removed, and 1 × 10⁴ Jurkat cells were added to each of the treated wells. Six hours later, Promega Bio-Glo reagent was added to each well. After a 5-minute incubation, the plate was read with a GloMax 96 Microplate Luminometer (Promega).

Upon coculture of Jurkat and PC3 or MDA-MB-231 cells, the Jurkat cells are activated through the T-cell receptor complex. This activation results in the expression of genes that are downstream of the nuclear factor of activated T-cells (NFAT) response element. Here, the luciferase gene is downstream of the NFAT response element, resulting in only activated T cells showing luminescent signal. This activation of the NFAT response element can be inhibited via interaction between PD-1 and PD-L1. When PD-1 on Jurkat cells binds PD-L1 on cancer cells, the NFAT response element will become inactive, and the luminescence will abate. If an agent prevents the interaction of PD-1 and PD-L1, Jurkat cells will remain active, and the luminescent signal will increase in intensity when compared with the control.

MDA-MB-231 cells were harvested 24 hours after seeding by scraping with a rubber policeman and washed with DPBS. The cells were pelleted by centrifugation at 200 × g_max for 7 minutes at 4°C in a GH-3.8 rotor (Beckman Coulter), washed with DPBS, and pelleted once more. The cells were resuspended in ice-cold isotonic homogenization buffer [HB; 250 mmol/L d-sucrose, 150 mmol/L sodium chloride, 10 mmol/L HEPES pH 7.4, 1 mmol/L EDTA, supplemented with Halt Protease and Phosphatase Inhibitors (Pierce)] with the volume normalized against the pellet size. Cells were mechanically lysed using a syringe-driven stainless steel–encased Balch ball-bearing homogenizer (Isobiotec). The cells were passed though the tungsten carbide ball bore 14 times with a high tolerance clearance of 12 μm. The postnuclear supernatant (PNS) was generated by separation of the homogenate using centrifugation at 1,000 × g_max for 15 minutes at 4°C in an F45-30-11 rotor (Eppendorf). The iodinated density gradient media OptiPrep (Axis-Shield) was used to create a discontinuous density gradient. Stock (60% w/v) Optiprep was diluted in HB to obtain a 0.9 mL bottom layer of 50% (1.272 g/mL) Optiprep, a 0.8 mL middle layer of 30% (1.175 g/mL) Optiprep, and a 2 mL top layer of 10% w/v (1.079 g/mL) Optiprep loaded into OptiSeal polypropylene tubes (Beckman Coulter). A 1 mL aliquot of PNS was loaded on top of the OptiPrep and separated by ultracentrifugation at 48,000 × g_max for 19 hours at 4°C in a TLA 100.4 fixed angle rotor (Beckman Coulter). Fourteen 300 μL fractions were collected by tube puncture and gravity flow.

MDA-MB-231 cells were transiently transfected with pcDNA3.1 (empty vector plasmid), Sigma1-HA, and/or PD-L1-FLAG plasmid constructs using jetPRIME Transfection Reagent (PolyPlus). Forty-eight hours posttransfection, cells were harvested and lysed in HB as described above (see isopycnic density-gradient centrifugation). Subsequently, lysates were solubilized in an equal volume of 2× NP-40 wash buffer (300 mmol/L sodium chloride, 25 mmol/L Tris pH 7.6, 10% glycerol, 2% NP-40 detergent) supplemented with both Halt Protease and Phosphatase Inhibitors (Pierce). Detergent solubilized lysates were precleared on Protein G Magnetic Beads (Bio-Rad) coupled with mouse serum (Sigma-Aldrich). The precleared lysates were then layered on Protein G Magnetic Beads (Bio-Rad) coupled with mouse monoclonal anti-FLAG (H-5) antibody or mouse monoclonal anti-HA (F-7) antibody and incubated for 1 hour at 4°C with constant rotation. The Protein G Magnetic Beads were then washed with 1× NP-40 wash buffer (300 mmol/L sodium chloride, 25 mmol/L Tris pH 7.6, 10% glycerol, 1% NP-40 detergent). The immunoprecipitated protein complexes were eluted by heating at 95°C in Laemmli sample loading buffer supplemented with dithiothreitol (DTT) and β-mercaptoethanol.

Following 16-hour drug treatment, cells were harvested and lysed in HB as described above (see isopycnic density-gradient centrifugation). The PNS was separated into a cytosol supernatant and membrane pellet by ultracentrifugation at 100,000 × g_max for 1 hour at 4°C in a TLS-55 swinging bucket rotor (Beckman Coulter). To isolate plasma membrane from the total membrane, the membrane pellet was suspended in Upper Phase solution provided by the Membrane Protein Extraction Kit (BioVision Inc). The sample was mixed with an equal volume of Lower Phase solution from the same kit, vortexed and incubated on wet ice for 10 minutes. Centrifugation at 1,000 × g_max for 5 minutes at 4°C in an F45-30-11 rotor (Eppendorf) produced two phases separated by a visible interface. The upper phase was collected and saved on ice, while the lower phase was again mixed and incubated with upper phase solution before a repeat centrifugation. The collected upper phases were combined and mixed with fresh Lower Phase solution, incubated, and centrifuged. The upper phase was collected and contained the plasma membrane. The lower phases were combined and contained the intramembrane (ER/Golgi). Both phases were diluted in 5 volumes of ultrapure water and incubated at 4°C overnight, followed by centrifugation at 20,000 × g_max for 15 minutes at 4°C in an F45-30-11 rotor (Eppendorf). After aspiration, the resultant dry pellet was lysed in RIPA buffer and run on SDS-PAGE.

MDA-MB-231 cells were seeded, and 24 hours later, IPAG treatment was applied for 16 hours. At the time of harvest, cells were washed with PBS and then scraped into PBS. The cells were fixed in 4% formaldehyde for 10 minutes at 37°C and placed on ice for one minute. Cells were washed twice in incubation buffer (0.5% BSA in PBS) and aliquoted such that each tube contained 7.5 × 10⁴ cells. Cells were stained with the PD-L1 extracellular domain–specific antibody at 1:100 for one hour at room temperature. Cells were washed in incubation buffer then stained with anti-rabbit Alexa Fluor 488 secondary antibody (Cell Signaling Technology cat. no. 4412) at 1:500 for 30 minutes at room temperature. Cells were washed in incubation buffer and resuspended in 500 μL incubation buffer.

Cell fluorescence was analyzed using a Millipore Guava EasyCyte instrument, and data were analyzed using FlowJo V10. Cells were gated such that only single cells were analyzed. Mean fluorescence intensity (MFI) was measured for each population. Average and SEM values were calculated for the MFI of biological triplicates. Unpaired, two-tailed t test was performed to compare the treatment groups.

MDA-MB-231 and PC3 cells were treated 24 hours after seeding with 100 ng/mL IFNγ for 24 hours in the absence or presence of the Sigma1 inhibitor, IPAG (10 μmol/L) for the final 16 hours. Total cellular RNA was extracted from the treated cells using the RNeasy Kit (Qiagen) per the manufacturer's protocol. IFNγ-induced transcription of PD-L1 was confirmed by qRT-PCR. TaqMan primer probe sets were purchased from Life Technologies, and the genes and catalog numbers used for the qRT-PCR experiments are the following: PD-L1 (Hs00204257_m1) and GAPDH (Hs99999905-m1). qRT-PCR was performed using the 7900HT- Fast Real Time PCR System (Applied Biosystems), and the 9 reactions were performed in triplicate using the Brilliant II qRT-PCR Master Mix 1-Step Kit (Agilent Technologies) following the manufacturer's instructions. Data were normalized to GAPDH transcript levels and presented as fold increase or percent change of PD-L1 transcripts relative to DMSO-treated controls.

Cells were harvested and gently mixed at a 1:1 ratio of cell suspension:trypan blue. A hemocytometer was used to count the percentage of cells that excluded the trypan blue, which represents the percentage of live cells in the suspension. Values were generated from at least three independent determinations.

Tests of statistical significance of single comparisons were performed using Student t test. Statistical significance of multiple comparisons was determined by one-way ANOVA followed by Bonferroni posttest using Prism software (GraphPad).

PD-L1 is a type I integral membrane glycoprotein that is synthesized in, posttranslationally modified in, and transported through the cellular secretory pathway to reach the plasma membrane of tumor cells. The MDA-MB-231 triple-negative breast cancer cell line has been reported to express high levels of PD-L1 under standard cell culture conditions (11). By confocal microscopy, we confirmed that PD-L1 is enriched in the cytoplasm of MDA-MB-231 cells as well as present on the surface of these cells (Fig. 1A). Sigma1 is a unique integral membrane chaperone or scaffolding protein in the secretory pathway and is enriched in the ER of cancer cells. We performed isopycnic centrifugation to demonstrate that Sigma1 cofractionates with PD-L1 in MDA-MB-231 cells. We show that 50% of PD-L1 and 78% of Sigma1 coisolated in density gradient fractions 5, 6, and 7 (Supplementary Fig. S1). This suggests that Sigma1 and PD-L1 occupy membrane domains with similar density and thus similar cellular compartments. Therefore, we asked whether Sigma1 could physically interact with PD-L1.

Using carboxy-terminal (C-terminal) FLAG-tagged PD-L1 (PD-L1-FLAG) and C-terminal HA-tagged Sigma1 (Sigma1-HA), we found that Sigma1 coimmunoprecipitates with PD-L1 (Fig. 1B). The PD-L1-FLAG protein was detected as multiple bands, a predicted approximately 33-kDa unmodified species as well as a number of higher molecular weight (MW), slower migrating bands. PD-L1 is glycosylated as it is processed and matures through the secretory pathway, which is reflected in the heterogeneous, smeared migration pattern of PD-L1 bands on the immunoblot. We performed a peptide:N-glycosidase F (PNGase F) digestion of immunoprecipitated PD-L1-FLAG eluate and found that the PD-L1 smear was reduced to a single band of approximately 33 kDa, confirming that the higher MW, slower migrating bands are glycosyl–PD-L1 (Fig. 1C).

Interestingly, in the reciprocal co-IP, only the approximately 33-kDa nonglycosylated species and an approximately 45 kDa PD-L1-FLAG species coimmunoprecipitated with Sigma1-HA (Fig. 1D). PNGase F digestion of the immunoprecipitated samples reduced the higher MW PD-L1 band to approximately 33 kDa, confirming that the approximately 45 kDa band is a glycosylated form of the protein (Fig. 1E). Digestion of co-IP eluates with Endo H also reduced the approximately 45 kDa species to 33 kDa (Fig. 1F). Whereas PNGase F cleaves nearly all types of N-linked (Asn-linked) glycosylation, including complex glycans, Endo H cleaves only high mannose and some hybrid types of N-linked carbohydrates but not complex glycans. Thus, oligosaccharide structures beyond enzymatic modification by Golgi alpha-mannosidase II in the secretory pathway are resistant to Endo H cleavage. This differential sensitivity to PNGase F and Endo H is used to track the maturation of glycoproteins. The pattern observed here suggests that Sigma1 physically associates primarily with glycosyl-PD-L1 formed early in the proximal secretory pathway, ER, and early Golgi compartments (Fig. 1F).

We propose that Sigma1 is a ligand-operated scaffolding protein that promotes the stability, processing, assembly, and trafficking of specific proteins in the secretory pathway of cancer cells. In support of this hypothesis, we found that siRNA-mediated knockdown of Sigma1 resulted in a significant decrease in PD-L1 protein levels in triple-negative MDA-MB-231 breast cancer and androgen-independent PC3 prostate cancer cells (Fig. 2A). We also confirmed this effect with two distinct Sigma1 shRNAs, which produced similar trends in both of these cell lines (Fig. 2B). These data demonstrate that Sigma1 is required to sustain PD-L1 protein levels in these cancer cells. It is important to note that Sigma1 knockdown affects only a specific subset of proteins; shRNA-mediated knockdown of Sigma1 did not result in a significant decrease in HSP90 or Akt protein levels (Supplementary Fig. S2).

Subsequently, we asked whether a small-molecule inhibitor of Sigma1 could mimic the effects of Sigma1 knockdown. Here, we used IPAG, a prototypic selective small-molecule inhibitor of Sigma1 that we previously have shown engages the UPR, the ubiquitin proteasome system, and autophagy (23, 24). We found that IPAG induced a significant dose-responsive decrease of total PD-L1 protein in PC3 and MDA-MB-231 cells (Fig. 3A). In support of the selectivity of proteins affected/targeted by Sigma1 modulator treatment, we determined that MHC-I, a related secretory pathway associated membrane protein, was not eliminated under standard cell culture conditions in MDA-MB-231 cells and was only slightly decreased in PC3 cells (Fig. 3B). Thus, both RNAi knockdown and pharmacologic inhibition of Sigma1 selectively suppressed PD-L1 protein levels in cancer cells.

To determine the operative mechanism by which the Sigma1 modulator eliminated PD-L1, we evaluated the potential and relative contribution of protein synthesis and degradation by treating with IPAG in the presence of the translation inhibitor, cycloheximide. Whereas translation arrest by cycloheximide alone did not result in a decrease in PD-L1 protein levels over 8 hours, the combination of cycloheximide and IPAG resulted in significant decreases in PD-L1 protein levels after 4 and 8 hours of treatment (Supplementary Fig. S3). These data suggest that although IPAG-mediated translation arrest may contribute to decreased PD-L1 levels in response to long-term IPAG treatment, protein degradation plays a significant role in the elimination of PD-L1.

Mature PD-L1 expressed at the surface of tumor cells binds its cognate receptor PD-1 on the surface of infiltrating T cells and thus contributes to suppression of antitumor immunity. Blockade of PD-L1 binding to PD-1, that is, immune checkpoint inhibition, prevents inactivation of T cells and promotes antitumor immunity. Here, we utilized an NFAT transcriptional response element luciferase reporter assay (NFAT-luciferase) to evaluate the effects of Sigma1 inhibition on PD-L1/PD-1 interaction blockade (Fig. 3C). This assay involved coculture of PD-L1–expressing MDA-MB-231 or PC3 cancer cells with PD-1–expressing NFAT-luciferase transfected Jurkat T-cells. PD-L1/PD-1 binding results in T-cell inactivation and the abrogation of luminescent signal, whereas blockade of PD-L1/PD-1 binding prevents inactivation of genes regulated by the NFAT response element and results in luminescent signal (Fig. 3C).

MDA-MB-231 and PC3 cells were treated with multiple concentrations of IPAG (3, 10, and 20 μmol/L) for 16 hours, removed drug-containing media, and then cocultured with Jurkat NFAT-luciferase reporter T cells in the absence of drug. The transcriptionally induced bioluminescent signal increased in a concentration related manner in both MDA-MB-231 and PC3 cell cultures, indicating that IPAG disrupted PD-L1′s checkpoint activity (Fig. 3D). Importantly, we confirmed that the concentrations of IPAG and 16-hour treatment time used in these experiments did not induce cell death (Fig. 3E). On the basis of these data, we performed all subsequent mechanism-focused experiments well within the 16-hour time point to minimize potential confounds associated with cell death.

We next assessed how Sigma1 inhibition affects the expression of PD-L1 at the cell surface. Using flow cytometry, we demonstrated that treatment with IPAG (20 μmol/L) for 16 hours significantly decreased cell surface expression of PD-L1 on MDA-MB-231 cells (Fig. 3F and G). These data suggest that Sigma1 inhibition effectively disrupts PD-L1 function at least in part by decreasing its functional expression on the plasma membrane.

We also performed a biochemical membrane fractionation assay, which separates and isolates the cell surface plasma membrane from intracellular membranes, such as the ER and the Golgi, based on the affinity of membranes to distinct polymer phases (36). We used this technique to determine the extent to which mature PD-L1 at the plasma membrane and intracellular PD-L1 is suppressed by IPAG. IPAG treatment for 16 hours at a concentration of 10 μmol/L resulted in a salient decrease in the intracellular membrane PD-L1 and a more modest decrease in plasma membrane PD-L1 (Supplementary Fig. S4). These data suggest that IPAG treatment decreases nascent PD-L1 protein stability, thus preventing the trafficking of functional PD-L1 complexes to the cell surface.

SA4503 is a Sigma1-selective putative agonist or activator (positive modulator) (37). Cotreatment of SA4503 with IPAG blocked the decrease in PD-L1 protein levels, demonstrating Sigma1-specific pharmacologic activity (Fig. 4A). Interestingly, by confocal microscopy, treatment with SA4503 alone resulted in increased cell surface expression of PD-L1 (Fig. 4B and C), suggesting that PD-L1 stability and trafficking can be differentially promoted and inhibited by positive and negative modulators of Sigma1, respectively.

In some cells, SA4503 treatment was associated with intense localized intracellular clusters of PD-L1 staining. The nature of this clustered staining pattern is unclear.

We previously demonstrated that IPAG could induce autophagy in cancer cells (24). Therefore, we asked whether autophagy contributes to the decrease in PD-L1 protein levels in this study. An IPAG treatment time course (0, 4, 8, and 16 hours) experiment revealed increased LC3B corresponding with decreased PD-L1 levels (Fig. 5A). Thus, autophagosome formation coincides with decreasing PD-L1, suggesting a role for autophagy.

We confirmed that IPAG induced autophagy by treating WT and ATG5 KO MEFs with IPAG and demonstrating that LC3B was induced in the WT but not in the ATG5 KO MEFs (Fig. 5B). Subsequently, we confirmed that autophagy contributes to Sigma1 modulator induced PD-L1 degradation by blocking the process in MDA-MB-231 cells by RNAi knockdown of ATG5 (Fig. 5C; Supplementary Fig. S5A and S5B), of ATG7 (Supplementary Fig. S5C), and by pharmacologically inhibiting autophagy with wortmannin and bafilomycin A1 (Baf A1; Fig. 5D and E). First, we performed shRNA-mediated knockdown (KD) of ATG5, the product of an essential autophagy gene required for the formation of autophagosomes (26, 27, 31). Knockdown of ATG5 in MDA-MB-231 cells blocked IPAG-mediated degradation of PD-L1 (Fig. 5C; Supplementary Fig. S5A and S5B). This experiment showed that autophagosome formation is required for IPAG-induced degradation of PD-L1. We replicated the results of this study with two distinct ATG5 shRNAs (data shown for TRCN0000330392 in Fig. 5C and in Supplementary Fig. S5A, and similar results were obtained with TRCN0000151474, shown in Supplementary Fig. S5B). Furthermore, we confirmed this effect with another essential autophagy gene, ATG7. ATG7 shRNA knockdown also mitigated PD-L1 degradation by IPAG when compared with the control (Supplementary Fig. S5C), confirming the fundamental role of autophagy in IPAG-mediated degradation of PD-L1.

Furthermore, we showed that pharmacologic inhibition of early and late stages of autophagy prevented PD-L1 decrease during IPAG treatment. Wortmannin is a Vps34 inhibitor that blocks autophagosome formation (26, 27, 38). Consistent with ATG5 knockdown, cotreatment with IPAG and wortmannin abrogated the PD-L1–decreasing effect of IPAG (Fig. 5D). Bafilomycin A1 (Baf A1) inhibits vacuolar H⁺ ATPase, which blocks fusion of autophagosomes with lysosomes, and thus blocks autolysosomal degradation or autophagic flux (26, 27). We previously showed that IPAG induces autophagic flux in a number of cancer cell lines (24) and confirmed that IPAG induces autophagic flux here as well (Supplementary Fig. S6). Baf A1 blocked the decrease in PD-L1 in IPAG-treated MDA-MB-231 cells (Fig. 5E).

Next, we tested whether PD-L1 protein levels would be suppressed by other inducers of autophagy. Interestingly, induction of bulk macroautophagy by cell starvation in Hank's buffered saline solution (HBSS) or treatment with rapamycin (26, 27) did not alter PD-L1 protein levels (Fig. 5F and G). As IPAG induces ER stress–associated autophagy (24), we asked whether this was a general mechanism by which PD-L1 could be degraded. In contrast to IPAG, we found that a widely used chemical inducer of ER stress–associated autophagy, thapsigargin (TG), which blocks the sarco/endoplasmic reticulum Ca2⁺ ATPase (SERCA; refs. 26, 27), did not induce PD-L1 degradation under the experimental conditions evaluated herein (Fig. 5H). Altogether, these data suggest that Sigma1 modulators induce a form of selective autophagy.

Finally, we performed confocal microscopy to visualize autophagic sequestration and degradation of PD-L1 during IPAG treatment (Fig. 6). We treated MDA-MB-231 cells stably expressing a GFP-LC3 construct with IPAG alone or in the presence of Baf A1 (Fig. 6A). The formation of GFP-LC3 punctae indicates autophagosome formation. Baf A1 blocks autophagic flux and causes accumulation of autophagosomes and their cargo. This is detected as increased numbers and in some cases size of GFP-LC3 punctae. Consistent with the results described above, IPAG treatment produced GFP-LC3 labeled autophagosomes and induced their colocalization with PD-L1 (Fig. 6A). In contrast, treatment with Baf A1 alone resulted in increased GFP-LC3 autophagosomes but did not induce the same colocalization (Fig. 6A). Cotreatment with IPAG and Baf A1 resulted in increased numbers of GFP-LC3–positive punctae that colocalized with PD-L1, indicating an accumulation of PD-L1–containing autophagosomes (Fig. 6A–C). Under conditions of treatment with IPAG alone and cotreatment of IPAG and Baf A1, PD-L1 appears to be engulfed in autophagosomes (Fig. 6B and C). Altogether, these data suggest that autophagy is the operative mechanism by which IPAG induces intracellular degradation of PD-L1.

IFNγ is an immunostimulatory cytokine that has long been of interest for its potential in promoting antitumor immunity (39, 40). However, exposure to IFNγ also induces PD-L1 expression in cancer cells, thereby suppressing the antitumor immune response through adaptive immune resistance (39–41). As IFNγ is present in the tumor microenvironment, we asked whether small-molecule Sigma1 inhibitors could suppress/eliminate PD-L1 induction by IFNγ treatment. In our model, we found that IFNγ (100 ng/mL for 24 hours) induced PD-L1 protein levels (Fig. 7A–D) and transcripts (Fig. 7E and F) in MDA-MB-231 and PC3 cells. Cotreatment with IPAG (10 μmol/L for the final 16 hours of the 24-hour IFNγ treatment) decreased PD-L1 protein levels back to levels similar to or less than DMSO-treated conditions (Fig. 7A–D), with a slight decrease and a slight increase in PD-L1 transcripts in MDA-MB-231 and PC3 cells, respectively (Fig. 7E and F). It is important to note that, under IFNγ combined with IPAG conditions, PD-L1 transcript levels remain statistically significantly higher than DMSO control, whereas the protein level decreases to similar levels as the control. Therefore, under the experimental conditions used here, that is, 16-hour treatment with IPAG, the decrease in PD-L1 protein can be attributed to posttranslational effects, likely protein degradation.

To our knowledge, this is the first demonstration of a direct physical and functional interaction between Sigma1 and PD-L1, the first demonstration that Sigma1 modulators can be used to sequester and degrade secretory pathway proteins by a potentially novel form of selective autophagy, and the first demonstration of autophagic degradation of PD-L1. These data suggest that Sigma1 modulators can be used to regulate PD-L1 transport and stability through regulation of pharmacologically responsive protein complexes.

Traditionally, small molecules with affinity for Sigma1 have been classified as agonist or antagonist based primarily on data from rodent behavior assays (19). Differences between putative Sigma1 antagonists and agonists, however, have remained unresolved at the molecular and cellular level. Sigma1 does not have the properties of a traditional receptor, and therefore, designation of Sigma1 selective compounds as classically defined receptor antagonists or agonists may be inaccurate. Emerging evidence suggests that Sigma1 may function either as an allosteric cofactor protein associated with bona fide receptor systems or as a novel chaperone (19) or scaffolding protein. Therefore, it may be more appropriate to use the term Sigma1 modulator, and not agonist or antagonist, for compounds that pharmacologically alter Sigma1 activity. Here, we provide evidence in support of this concept with opposing actions of the putative Sigma1 antagonist IPAG (negative modulator or inhibitor) and the putative agonist SA4503 (positive modulator or activator). The Sigma1-selective actions of these compounds have been demonstrated elsewhere (24, 37, 42). In these studies, we also show that SA4503 blocked IPAG-mediated decreases in PD-L1 protein levels, pharmacologically demonstrating Sigma1-specific actions (Fig. 4A).

A number of publications report the cancer cell growth and survival inhibiting effects of compounds with affinity for Sigma1 (19); however, the specific mechanisms underlying these actions and the physiologic role of Sigma1 in cancer cells have remained unclear. Our data suggest that Sigma1 may play a role in regulating the components or compartments of the secretory pathway of cancer cells. Modulation of PD-L1 stability, transport, and activity by Sigma1 is consistent with such a physiologic role. Although the precise biochemical and molecular mechanisms governing Sigma1 actions in the secretory pathway of cancer cells remain to be determined, we posit that Sigma1 modulators coopt components of the ER protein homeostasis machinery to regulate the transport of PD-L1 transiting the secretory pathway to the plasma membrane. Demonstration that Sigma1 inhibitors can be used to sequester secretory pathway proteins into autophagosomes supports this notion. These results raise questions regarding what other secretory pathway-dependent proteins are regulated by Sigma1.

The biochemical and molecular mechanisms that govern PD-L1 transcription, translation, processing, assembly, transport, and functional binding partners (the intracellular proteins and cellular factors with which PD-L1 associates throughout its life cycle) remain poorly defined. Our data show that Sigma1 can physically interact with intracellular PD-L1 in pharmacologically responsive protein complexes that can be driven to autophagic sequestration and degradation of PD-L1. Our previous discovery that Sigma1 ligands can induce autophagy (24) and our observation that PD-L1 sequesters in punctate cytoplasmic clusters (Fig. 6) led us to investigate this process as a potential mechanism involved in Sigma1 inhibitor–induced PD-L1 degradation. By high-resolution confocal microscopy, we demonstrated that IPAG induces the sequestration and autophagic degradation of PD-L1 (Fig. 6). Inhibition of IPAG-mediated PD-L1 degradation by cotreatment with bafilomycin A1 (Baf A1) or with wortmannin (Fig. 5D and E) supports our hypothesis of autophagic degradation. Baf A1 blocks late-stage autophagic degradation, while wortmannin is a small-molecule inhibitor of Vps34, which blocks formation of autophagosomes (early-stage autophagy; refs. 27, 38).

ATG5 and ATG7 are essential autophagy genes required for autophagosome assembly and formation (31, 43). We confirmed that IPAG-induced autophagy requires ATG5 with ATG5 knockout MEFs (Fig. 5B). RNAi-mediated knockdown of ATG5 and ATG7 in MDA-MB-231 cells both prevented the elimination of PD-L1 in response to treatment with IPAG, thereby confirming autophagosomal degradation of PD-L1 (Fig. 5C; Supplementary Fig. S5).

Interestingly, treatment of cells with HBSS or rapamycin, both of which are commonly used to induce bulk macroautophagy (27, 44), did not result in decreased PD-L1 protein levels (Fig. 5F-G), nor did treatment with thapsigargin, which induces general ER stress–associated autophagy (Fig. 5H; refs. 27, 44). These data suggest that the Sigma1 modulator induces degradation of PD-L1 by a potentially novel form of selective autophagy.

In addition, because rapamycin, thapsigargin, wortmannin, and IPAG all suppress protein synthesis (25, 27, 44, 45), inhibition of protein translation may contribute to the suppression of PD-L1 levels during long-term treatment. However, it appears as though protein degradation is the operative mechanism by which PD-L1 protein levels are suppressed following IPAG treatment; this is supported by several lines of evidence demonstrating that whereas 16-hour treatment with IPAG results in decreased PD-L1 protein levels, treatment with rapamycin and thapsigargin did not (Fig. 5). This was further supported by the accelerated decrease in PD-L1 in IPAG-treated cells in which protein translation has been blocked by cycloheximide (Supplementary Fig. S3).

Our data suggest that IPAG induces autophagosomal degradation of nascent, ER-localized PD-L1, thus preventing surface expression of PD-L1. Salient elimination of the internal membrane fraction of PD-L1 and coimmunoprecipitation of Sigma1 with the lower MW species of N-glycosylated PD-L1 suggest that these proteins associate in the early/proximal secretory pathway, likely the ER. It is important to note, however, that the ER expands throughout the cytoplasmic space and associates with essentially all organelles within the cell (5). We do not rule out the possibility that Sigma1 may also bind to the cytoplasmic domain of plasma membrane–expressed PD-L1 and influence the levels and activity of cell surface PD-L1. Indeed, Sigma1 has been reported to bind to the cytoplasmic domains of other plasma membrane proteins (19) and plasma membrane components that interact with and contribute to autophagy as well as endosomal/endolysosomal internalization and degradation of cell surface proteins (31, 32).

The negative modulator of Sigma1, IPAG, suppresses PD-L1/PD-1 interaction associated signaling and transcriptional activity at 3 μmol/L, a concentration that does not alter PD-L1 levels (Fig. 3A and D), suggesting that Sigma1 regulates potential PD-L1 cofactors or interacting proteins required for its activity or modifies PD-L1 in a way that inhibits its function prior to or distinct from its destabilization and degradation. Emerging data suggest that Sigma1 associates with and can regulate multiple secretory pathway resident proteins in cancer cells. We have shown previously that Sigma1 modulators regulate ErbB2/HER2 and ErbB3/HER3 levels and activity in prostate and breast cancers (ref. 23 and our unpublished data, respectively).

This study addresses the intracellular processes regulated by Sigma1 modulators, and it provides evidence in support of the notion that the pharmacologic mechanism of action of Sigma1 modulator compounds involves modulation of Sigma1-associated functional protein interactions. Sigma1 engages in protein–protein interactions with multiple partner and client proteins. It is likely that ligand-induced changes in some of these protein associations may contribute to suppression of PD-L1. This may occur through direct regulatory mechanisms or indirectly through feedback regulation mechanisms. For example, Sigma1 has been implicated in immunomodulation (reviewed in ref. 19), and a role for Sigma1 in antitumor immunity has been proposed (reviewed in ref. 19). Sigma1 activators (agonists) have been shown to induce the production and secretion of IL10 and TGFβ and promote tumor growth in a syngeneic mouse lung cancer model (46). Interestingly, in this model, IL10 was only induced at the tumor site and increased TGFβ only occurred in tumor-bearing mice, suggesting increased cytokine production in response to the tumor or increased cytokine production by the tumor cells (46). Sigma1 inhibitors (putative antagonists) abrogated the production of IL10 and suppressed Sigma1 activator–induced tumor growth (46). Sigma1 modulators have also been reported to regulate the production of IFNγ (46). PD-L1 production and activity through modulation of cell surface expression are regulated in part by cytokines released in the tumor microenvironment during antitumor immune responses, including IFNγ, TNFα, TGFβ, IL17, and IL10 (7, 13, 14, 47, 48). Taken together, these data suggest that pharmacologic modulation of Sigma1 can regulate PD-L1 production and activity via immune response induced cytokine-mediated extracellular feedback loops as well as directly via protein–protein associations. These mechanisms regulate the tenor of the tumor immune microenvironment (49).

There is compelling rationale for and reemerging interest in combination therapies to sustain antitumor immune responses induced by stimulatory cytokines such IFNγ (39–41). The ability of Sigma1 modulators to suppress PD-L1 in cancer cells treated with immunostimulatory cytokines, such as IFNγ (Fig. 7), reveals potential therapeutic approaches involving cotreatment with IFNγ and Sigma1 modulators to prolong/sustain antitumor immune responses.

PD-L1 also costimulates IL10 secretion (50). Sigma1 activators (agonists) have been shown to block antitumor immunity, in part by induction of IL10 (46). Thus, Sigma1 regulation of PD-L1 may contribute to antitumor immunity through checkpoint inhibition as well as cosuppression of anti-inflammatory cytokines. Sigma1 inhibitors may promote antitumor immunity in part by eliminating PD-L1 as well as by modulating the production and secretion of cytokines from cancer cells.

Thus, control of PD-L1 through Sigma1 supports a role for Sigma1 in antitumor immunity and provides insight into the mechanism(s) by which Sigma1 compounds elicit their immunomodulatory effects. Our data suggest that selective small-molecule Sigma1 ligands may be used as novel modulators of the tumor immune microenvironment. The use of Sigma1 modulators to pharmacologically promote antitumor immunity presents an opportunity for novel drug combinations with checkpoint inhibitor agents to expand the population of patients that respond to PD-L1/PD-1–targeted therapies.

F.J. Kim is the scientific founder at, reports receiving commercial research support from, has ownership interest (including patents) in, and is a consultant/advisory board member for Context Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C.M. Maher, F.J. Kim

Development of methodology: C.M. Maher, J.D. Thomas, D.A. Haas, C.G. Longen, H.M. Oyer, F.J. Kim

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.M. Maher, J.D. Thomas, D.A. Haas, C.G. Longen, H.M. Oyer, J.Y. Tong, F.J. Kim

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.M. Maher, J.D. Thomas, D.A. Haas, C.G. Longen, H.M. Oyer, J.Y. Tong, F.J. Kim

Writing, review, and/or revision of the manuscript: C.M. Maher, F.J. Kim

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.M. Maher, H.M. Oyer, F.J. Kim

Study supervision: F.J. Kim

F.J. Kim was supported by an American Cancer Society Institutional Research Grant, Drexel University Clinical and Translational Research Institute Grant, Sidney Kimmel Cancer Center Pilot Study Award, and a Coulter-Drexel Translational Research Partnership Program Award.

The authors thank Drs. Paul McGonigle and James Barrett for critical reading of the manuscript.

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.