Selective Alanine Transporter Utilization is a Therapeutic Vulnerability in ARID1A-Mutant Ovarian Cancer

Abstract

Subunits of the SWI/SNF chromatin remodeling complex are altered in ~20% of human cancers. Exemplifying the alterations is the ARID1A mutation that occurs in ~50% ovarian clear cell carcinoma (OCCC), a disease with limited therapeutic options. Here, we showed that ARID1A mutations create a dependence on alanine by regulating alanine transporters to increase intracellular alanine levels. ARID1A directly repressed alanine importer SLC38A2 and simultaneously promoted alanine exporter SLC7A8. ARID1A inactivation increased alanine utilization predominantly in protein synthesis and passively through the tricarboxylic acid cycle. Indeed, ARID1A-mutant OCCCs were hyper-sensitive to inhibition of SLC38A2. In addition, SLC38A2 inhibition enhanced chimeric antigen receptor-T cell assault in vitro and synergized with immune checkpoint blockade using an anti-PD-L1 antibody in a genetically engineered mouse model of OCCC driven by conditional Arid1a inactivation in a CD8⁺ T-cell dependent manner. These findings suggest that targeting alanine transport alone or in combination with immunotherapy may represent an effective therapeutic strategy for ARID1A-mutant cancers.

Introduction

ARID1A encodes a key subunit of the SWI/SNF chromatin-remodeling complex and acts as a tumor suppressor (1). Inactivating mutations in ARID1A are frequently observed in ovarian clear cell carcinomas (OCCC; >50%) and ovarian endometrioid carcinomas (OEC; >30%) (2,3). Notably, over 90% of the ARID1A mutations identified in ovarian cancers are frameshift or nonsense alterations that lead to loss of ARID1A protein expression (2–4). OCCC is typically resistant to standard chemotherapy used for ovarian cancer treatment. Among patients diagnosed at advanced, metastatic stages, OCCC is associated with the poorest prognosis among all epithelial ovarian cancer subtypes (5,6). These observations underscore a significant unmet clinical need for the development of effective therapeutic strategies targeting ARID1A-mutated OCCC.

ARID1A mutation drives metabolic reprogramming as evidenced by a decrease in glycolysis and an increase in the tricarboxylic acid cycle (TCA) in part through glutamine metabolism (7,8). Alanine is a non-essential amino acid and can, therefore, be synthesized by mammalian cells by the enzyme alanine aminotransferase, also known as glutamate-pyruvate transaminase (GPT) through pyruvate transamination (9). Alanine synthesis and catabolism are compartmentalized with GPT expressed in the cytosol, and GPT2 localized in mitochondria (9,10). Certain types of cancer cells catabolize alanine to fuel the TCA cycle (11,12). In contrast, T cells appear to exclusively use extracellular alanine for protein synthesis, and alanine deprivation suppresses the proliferation and effector functions of T cells (13).

There are inconsistent reports in the literature regarding whether the inactivation of the SWI/SNF complex sensitizes tumors to immune checkpoint blockade (ICB) (14–16). Notably, ARID1A mutation sensitizes ovarian cancers to ICB, such as anti-PD-L1 (17). However, ICB alone exhibits a modest effect on improving survival of mice carrying Arid1a-inactivated ovarian tumors (7,18). Thus, to achieve a complete eradication of ARID1A-mutated ovarian cancer, combinational therapeutic strategies are needed.

As previously described (19,20), transporters are membrane-bound proteins that mediate the translocation of substrates across biological membranes. There are two main transporter superfamilies: the ATP-binding cassette (ABC) superfamilies and the solute carrier (SLC) superfamilies (19,20). ABC transporters utilize energy from ATP hydrolysis and function to efflux transport, while SLC transporters are involved in cross-membrane importing and exporting of small molecules such as amino acids. SLC transporters are emerging as therapeutic targets in diseases such as cancer (19,20). There is evidence that links the SWI/SNF complex to SLC transporters (21). However, the potential role of ARID1A in regulating SLC transporters remains to be elucidated. In addition, SLC transporters such as SLC38A2 regulates the competition for nutrients such as glutamine between tumor cells and conventional dendritic cells (22). This raised the possibility that therapeutic targeting SLC transporters may enhance the efficacy of cancer immunotherapies such as ICB.

Here, we report that inhibition of alanine uptake suppresses the growth of ARID1A-mutated tumor cells and synergizes with ICB through modulating CD8⁺ T cells in the tumor microenvironment.

Materials and Methods

Cell Lines

Human ovarian clear cell carcinoma (OCCC) cell line RMG1 (JCRB, Cat# JCRB0172, RRID:CVCL_1662) and human embryonic kidney cell line HEK-293T (ATCC, Cat#: CRL-3216, RRID: CVCL_0063) were obtained in 2014 and cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37 °C supplied with 5% CO₂. OCCC cell lines OVCAR429 (Dr. Ie-Ming Shih, Johns Hopkins University, RRID:CVCL_3936), TOV21G (ATCC, Cat#: CRL-11730, RRID:CVCL_3613), OVISE (JCRB, Cat#: JCRB1043, RRID:CVCL_3116) and OVTOKO (JCRB, Cat#: NIHS0301, RRID:CVCL_3117) cells were obtained in 2014 and cultured in RPMI-1640 with 10% FBS and 1% penicillin/streptomycin. Rhabdoid tumor cell line G401 (JCRB, Cat# JCRB9065, RRID:CVCL_0270) was cultured in McCoy’s 5a medium with 10% FBS and 1% penicillin/streptomycin at 37 °C supplied with 5% CO₂. Primary cultures of human ovarian clear cell carcinoma with (VOA4841) and without (XVOA295) ARID1A expression were as described previously (23). VOA4841 and XVOA295 cells were obtained in 2017 and cultured in RPMI-1640 supplemented with 10% FBS and 1% penicillin/streptomycin. All the cell lines were used within a month after thawing. All the cell lines were authenticated using short tandem repeat DNA profiling. Mycoplasma was tested monthly using mycoplasma PCR detection kit (Sigma-Aldrich, Cat#: MP0035).

Human Samples

Ovarian cancer tissue microarrays (TMA) of OCCC samples were kindly provided by Dr. Ronny Drapkin from The University of Pennsylvania. The selection criteria were a pathologic diagnosis of clear cell carcinoma with involvement of ovary. This was confirmed by pathology review and the diagnosis the patients were treated for clinically. The tumors were from primary surgeries with written informed consent. Metabolomics data of OCCC tumors were kindly provided by Dr. Ernst Lengyel from The University of Chicago.

Animal Models

For mouse models, 2 different strains were used in this study. The transgenic mouse model of Arid1a^−/−;Pik3ca^H1047R genetic clear cell ovarian tumor were generated by crossing Arid1a^flox/flox mice with R26-Pikca^H1047R as we previously published (23). 6–8 weeks old female NSG mice (NOD.Cg-Prkd^cscid Il2rg^tm1Wjl/SzJ, RRID: IMSR_JAX:005557) from Jackson Laboratory were housed and maintained in individual microisolator cages in a rack system capable of managing air exchange with filters.

ARID1A CRISPR Knockout

ARID1A knockout RMG1 and OVCAR429 cells were generated as previously described.(23,24) For ARID1A knockout, pSpCas9(BB)-2A-GFP (Addgene, Cat#: 48138, RRID:Addgene_48138) and pFETCh_Donor (Addgene, Cat#: 63934, RRID:Addgene_63934) plasmids were purchased from Addgene. Guide RNA sequence (5’- TGTCCCACGGCTGTCATGAC -3’) targeting the terminal codon of ARID1A was inserted into pSpCas9(BB)-2A-GFP. About 500 base pairs of homologous arms at both sides of guide RNA targeting site were cloned and inserted into pFETCh-donor. ARID1A knockout clones were selected by puromycin (1 μg/mL) and validated by immunoblot.

Colony Formation Assay

3,000 to 5,000 cells were seeded into 24-well tissue culture plates. For treatment, cell medium was changed every three days with appropriate drug doses for a total of 10–14 days. Afterward, colonies were stained with 0.05% crystal violet (Sigma-Aldrich, Cat#: C0775), and the signal intensity was quantified using the National Institutes of Health ImageJ software (version 1.53a).

Cell Viability AlamarBlue Assay

3,000 cells were seeded into each well of a 96-well plate. After 48 hours, the medium was replaced with fresh medium containing 10 % dialyzed fetal bovine serum (FBS), with either normal (10 mM) or low (1 mM) glucose, with or without 2 mM glutamine, and varying concentrations of alanine (1, 10, 100 mM). Empty wells containing the same volume of complete medium were used as blank background. After 72 hours, 10% AlamarBlue reagent (Fisher, Cat#: DAL1100) was added into each well and then incubated for 2 hours at 37°C. AlamarBlue fluorescence was then measured using a microplate reader (BioTek Synergy H1) at excitation and emission wavelengths of 540 nm and 590 nm, respectively.

Immunoblotting

Whole-cell lysate was extracted using RIPA buffer (50 mM Tris pH 8.0, 150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 1 mM EDTA, 10 mM dithiothreitol (DTT) and 1 mM PMSF) on ice. For SLC38A2 and SLC7A8 examination, cell lysates were not boiled. Proteins were separated by SDS-PAGE gel and transferred to a PVDF membrane (Millipore, IPVH00010). The membranes were blocked with 5% non-fat milk and incubated with primary antibodies overnight at 4 °C with gentle agitation using the following antibodies and dilutions: rabbit anti-β-actin (1:2000; Cell Signaling, Cat#: 4970, RRID:AB_2223172), rabbit anti-sodium potassium ATPase (1:2000; Abcam, Cat#: ab76020, RRID:AB_1310695), rabbit anti-GPT (1:500; proteintech, Cat#: 16897–1-AP, RRID:AB_2230815), rabbit anti-GPT2 (1:500; proteintech, Cat#: 16757–1-AP, RRID:AB_2112098), rabbit anti-ARID1A (1:1000; Cell Signaling, Cat#: 12354, RRID:AB_2637010), rabbit anti-BRG1 (1:1000; Cell Signaling, Cat#: 49360, RRID:AB_2728743), mouse anti-ARID1B (1:1000; Abgent, Cat#: AT1190a, RRID:AB_1551334), rabbit anti-SNF5 (1:1000; Cell Signaling, Cat#: 91735, RRID:AB_2800172), rabbit anti-SLC38A2 (1:500; MBL International, Cat#: BMP081, RRID:AB_10597880), rabbit anti-SLC7A8 (1:500; Sigma-Aldrich, Cat#: AV43930, RRID:AB_1857211), and rabbit anti-GLUT1 (1:500; Sigma-Aldrich, Cat#: 07–1401, RRID:AB_1587074). Objective signals were amplified with HRP-conjugated secondary antibodies (1:2000; Cell Signaling, Cat#: 7076, RRID:AB_330924 and 1:3000; Cat#: 7074, RRID:AB_2099233) and detected by chemiluminescent substrate (Thermo Fisher, Cat#: 34094).

In vitro T cell Killing Assay

The elimination index was determined according to a previous study (25). Briefly, anti-CD19 CAR T cells were co-cultured with CD19 overexpressed target cells at an E:T ratio of 1:3 in RPMI 1640 fully supplemented in the absence of cytokines. After 48 h, cells were stained with 10 μg/mL propidium Iodide (Thermo Fisher, Cat#: P3566), and live/dead cells were analyzed by flow cytometry (BD Biosciences, LSRII). The elimination index was calculated as follows: 1 - (ratio of live target cells with CAR T cells / ratio of live target cells in the control group).

Chromatin Immunoprecipitation (ChIP)

Cells were crosslinked with 1% formaldehyde for 10min and then quenched by 0.125M glycine for 5min at room temperature. Fixed cells were lysed with ChIP lysis buffer 1 (50mM HEPES-KOH pH 7.5, 140 mM NaCl, 1mM EDTA pH 8.0, 1% Triton X-100, and 0.1% DOC) on ice and lysis buffer 2 (10 mM Tris pH 8.0, 200 mM NaCl, 1 mM EDTA, and 0.5 mM EGTA) at room temperature. Chromatin was digested with micrococcal nuclease (MNase Cell Signaling, Cat#: 10010) in digestion buffer (10 mM Tris pH8.0, 1 mM CaCl2, and 0.2% Triton X-100) at 37 °C for 15 min. Nucleus products were broken down by Bioruptor pulse at high frequency. The following antibodies were used for ChIP: rabbit anti-ARID1A (Cell Signaling, Cat#: 12354, RRID:AB_2637010; 10 μL/IP), rabbit anti-BRG1 (Cell Signaling, Cat#: 49360, RRID:AB_2728743; 20 μL/IP), rabbit anti-SNF5 (Cell Signaling, Cat#: 91735, RRID:AB_2800172; 20 μL/IP) and mouse anti-Pol II (Santa Cruz, Cat#: sc-47701, RRID:AB_677353; 5 μg/IP). ChIP DNA was purified by the ChIP DNA clean and concentrator kit (Zymo Research, Cat#: D5205) and analyzed by qPCR. Primers targeting the SLC38A2 and SLC7A8 promoters were used for ChIP–qPCR are purchased from Integrated DNA Technologies: human SLC38A2 promoter forward (5’-GAAGGCCGAAATGGGACGAT-3’), human SLC38A2 promoter reverse (5’- GCTTGCTTGGTGGGGTAGGA-3’), human SLC7A8 promoter forward (5’- CAACACCGAAAAGAAACACCC-3’) and human SLC7A8 promoter reverse (5’-GATACCACAGGCACTGACCAA-3’).

Reverse Transcription and Quantitative Real-Time PCR (RT-qPCR).

Total RNA was extracted using Trizol reagents (Thermo Fisher, Cat#: 15596026) according to the manufacturer’s protocol. RNA was reverse transcribed with the High-Capacity cDNA Reverse Transcription kit (Thermo Fisher, Cat#: 4368813). RT-qPCR was performed using the QuantStudio 3 Real-Time PCR System (Thermo Fisher). Following primers for PCR were purchased from Integrated DNA Technologies; human SLC38A2 (forward: 5’-CTGAAAGACCGCAGCCGTAG-3’ and reverse: 5’-TATCCAAAGAGGGCGGCAAG-3’), human SLC7A8 (forward: 5’- GTTGTGGGAGCCCTCTGCTATG-3’ and reverse: 5’-CCAGCACAGCAATCCACAGC-3’), human GPT2 (forward: 5’- ATCCTCACGCTGGAGTCCATGA-3’ and reverse: 5’- ATGTTGGCTCGGATGACCTCTG-3’) and human B2M (forward: 5’- TGGAGCATTCAGACTTGTCTTTCA-3’ and reverse: 5’- CACGGCAGGCATACTCATCTT-3’).

Immunohistochemistry (IHC) Staining

Ovarian cancer tissue microarrays (TMA) were kindly provided by Ronny Drapkin from The University of Pennsylvania. For immunohistochemical staining, TMA or mice tumor sections were incubated with primary following antibodies at 4 °C overnight: rabbit anti-ARID1A (1:500; Cell Signaling, Cat#: 12354, RRID:AB_2637010), rabbit anti-SLC38A2 (1:500; MBL International, Cat#: BMP081, RRID:AB_10597880), rabbit anti-SLC7A8 (1:200; MBL International, Cat#: BMP041, RRID:AB_1953129), rabbit anti-Cleaved Caspase-3 (1:50; Cell Signaling, Cat#: 9661, RRID:AB_2341188), rabbit anti-Ki67 (1:500; Abcam, Cat#: ab16667, RRID:AB_302459), rabbit anti-serine 10 phosphorylated Histone H3 (p-H3S10) (1:200; Abcam, Cat#: ab5176, RRID:AB_304763). The tissue sections were incubated with appropriate secondary antibodies (Dako, Cat#: K4011) for 1 hour at room temperature. Cell nuclei were stained using Mayer’s Hematoxylin (Dako, Cat#: S3309). Staining density was quantified using histologic score (H-Score).

Immune Cell Profiling

Tumors were chopped and digested with Mouse Dissociation Kit (Miltenyi Biotec, 130–096-730) according to the manufacturer’s instructions. Single cells were then harvested with 70 mm strainer and used for staining. Live/dead cells were discriminated by DAPI (Thermo Fisher, Cat#: 62247). Fc blocking (BD Biosciences, Cat#: 553142, RRID:AB_394657) was followed by cell surface staining in FACS buffer (1% FBS in PBS buffer) using antibodies against CD45 (1:100; BD Biosciences, Cat#: 564279, RRID:AB_2651134), CD11b (1:100; BD Biosciences, Cat#: 568345, RRID:AB_2941960), CD4 (1:100; Biolegend, Cat#: 100414, RRID:AB_312699), CD8a (1:100; BD Biosciences, Cat#: 750023, RRID:AB_2874241), Ly-6G (1:100; Biolegend, Cat#: 127643; RRID:AB_2565971), Ly-6C (1:100; BD Biosciences, Cat#: 569012), PD-1 (1:100; Biolegend, Cat#: 135217, RRID:AB_10900085), F4/80 (1:100; Biolegend, Cat#: 123114, RRID:AB_893478). Intracellular staining was carried out using True-Nuclear Transcription Factor Buffer Set (Biolegend, Cat#: 424401) according to manufacturer instructions. Briefly, cells were fixed for 45 min at room temperature in the dark. Following fixation, cells were washed with 1x permeabilization buffer and then incubated with anti-Foxp3 antibody (1:100; BD Biosciences, Cat#: 563487, RRID:AB_2738236). Data was acquired using flow cytometry (BD FACSymphony A3) and analyzed using FlowJo software (version 10.0).

Extracellular Acidification Rate (ECAR) and Oxygen Consumption Rate (OCR) Assay

ECAR and OCR were determined using a Seahorse XFe96 analyzer (Agilent). Cells were seeded (25,000 cells per well for RMG1 or 15,000 cells per well for OVCAR429) in an XF96 cell culture microplate, cells were cultured in the media containing 10 mM glucose and 2 mM glutamine for 48 hours. Subsequently, 1 mM alanine was added to the media, and the cells were cultured for an additional 6 hours. Media was exchanged to XF media (10 mM glucose and 2 mM glutamine) 30 min before the assay. XF Glycolysis Stress Test Kit (Agilent, Cat#:103020–100) was used to measure the glycolytic capacity. Glucose, oligomycin and 2-deoxy glucose (2-DG) were diluted into XF media and loaded into the cartridge to achieve final concentrations of 20 mM, 2 μM, and 50 mM, respectively. ECAR was determined according to the manufacturer’s instructions. XF Cell Mito Stress Test Kit (Agilent, Cat#:103010–100) was used to measure cellular mitochondrial function. Oligomycin, FCCP and antimycin/rotenone were diluted into XF media and loaded into the cartridge to achieve final concentrations of 2.5, 0.5, 0.5 μM, respectively. OCR was determined according to the manufacturer’s instructions.

To determine the short-term effects of alanine supplement. RMG1 cells (25,000 cells per well) were seeded in an XF96 cell culture microplate, cells were cultured in the media containing 10 mM glucose and 2 mM glutamine for 48 hours. Media was exchanged to XF media (no glucose and no glutamine) 30 min before the assay. XF Cell Mito Stress Test Kit (Agilent, Cat#:103010–100) was used to measure cellular mitochondrial function. Alanine, oligomycin, FCCP and antimycin/rotenone were diluted into XF media and loaded into the cartridge to achieve final concentrations of 1 mM, 3 μM, 1 μM and 0.5 μM, respectively. OCR was determined according to the manufacturer’s instructions.

LC–MS/MS and HRLC/IC-MS

The following methodological details were provided by Wistar Proteomics and Metabolomics Facility. LC–MS/MS was performed for alanine tracing experiments (intracellular metabolites). Cells were cultured with the DMEM media containing 10% dialyzed FBS and 1 mM glucose, and labeled with 1 mM ¹³C₃-L-alanine (Cambridge Isotope, Cat#: CLM-2184-H) for 18 h. Cells were spun down and pellets were resuspended in ice-cold extraction solution containing LC–MS-grade methanol and ultrapure water at a ratio of 80:20 (v/v). Samples were vortexed for 5 min at 4 °C before centrifugation. Final cleared metabolite extracts from cells were transferred to silanized glass vials and loaded onto an autosampler for analysis by LC–MS. LC–MS analysis was performed on a Q Exactive HF-X Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher) equipped with a HESI II probe and coupled to a Vanquish Horizon UHPLC system (Thermo Fisher). Then, 4 μl of sample was injected and separated by HILIC chromatography on a ZIC-pHILIC 2.1 mm i.d. × 150 mm column (EMD Millipore). The mobile phase A was 20 mM ammonium carbonate, 0.1% ammonium hydroxide (pH 9.2) and mobile phase B was acetonitrile. LC was run at a flow rate of 0.2 ml min−1 and gradient used was as follows: 0 min, 85% B; 2 min, 85% B; 17 min, 20% B; 17.1 min, 85% B; and 26 min, 85% B. The column was maintained at 45 °C and the mobile phase was also pre-heated at 45 °C before flowing into the column. The relevant MS parameters were as listed: sheath gas, 40; auxiliary gas, 10; sweep gas, 2; auxiliary gas heater temperature, 350 °C; spray voltage, 3.5 kV for the positive mode and 3.2 kV for the negative mode. Capillary temperature was set at 325 °C and funnel RF level at 40. Samples were analyzed in full MS scan with polarity switching at scan range 65 to 975 m/z; 120,000 resolution; automated gain control target of 1E6; and maximum injection time of 100 ms. Identification and quantitation of metabolites was performed using TraceFinder 4.1 and Compound Discoverer 3.0 (Thermo Fisher).

HRLC/IC-MS was performed by MD Anderson Cancer Center Metabolomics Core for intracellular alanine, pyruvate tracing and alanine tracing (secreted metabolites) experiments. For intracellular alanine determination, cells were cultured in DMEM media containing 10% regular FBS and 10 mM glucose for 18 h, for pyruvate tracing, cells were cultured in no glucose DMEM containing 10% dialyzed FBS and labelled with 1 mM ¹³C₃-L-pyruvate (Cambridge Isotope, Cat#: CLM-2440) for 18h, for alanine tracing (secreted metabolites), cells were cultured with the DMEM media containing 10% dialyzed FBS and 1 mM glucose, and labeled with 1 mM ¹³C₃-L-alanine (Cambridge Isotope, Cat#: CLM-2184-H) for 18 h. To determine the relative abundance of polar metabolites in cell and media, extracts were prepared and analyzed by ultra-high-resolution mass spectrometry (HRMS). For ICMS analysis, metabolites were extracted using ice-cold 80/20 (v/v) methanol/water with 0.1% ammonium hydroxide. Extracts were centrifuged at 17,000 g for 5 min at 4°C, and supernatants were transferred to clean tubes, followed by evaporation to dryness under nitrogen. Dried extracts were reconstituted in deionized water, and 10 μL was injected for analysis by ion chromatography (IC)-MS. IC mobile phase A (MPA; weak) was water, and mobile phase B (MPB; strong) was water containing 100 mM KOH. A Thermo Scientific Dionex ICS-6000+ system included a Thermo IonPac AS11 column (4 μm particle size, 250 x 2 mm) with column compartment kept at 35°C. The autosampler tray was chilled to 4°C. The mobile phase flow rate was 360 μL/min, and the gradient elution program was: 0–5 min, 1% MPB; 5–25 min, 1–35% MPB; 25–39 min, 35–99% MPB; 39–49 min, 99% MPB; 49–50, 99–1% MPB. The total run time was 50 min. To assist the desolvation for better sensitivity, methanol was delivered by an external pump and combined with the eluent via a low dead volume mixing tee. Data were acquired using a Thermo Orbitrap IQ-X Tribrid Mass Spectrometer under ESI negative ionization mode. For HILIC analysis (Alanine, Aspartate, Glutamate and Glutamine), same samples were injected by liquid chromatography (LC)-MS. LC mobile phase A (MPA) was 95/5 (v/v) water/acetonitrile containing 20mM ammonium acetate and 20 mM ammonium hydroxide (pH~9), and mobile phase B (MPB) was acetonitrile. Thermo Vanquish LC system included a Xbridge BEH Amide column (3.5 μm particle size, 100 x 4.6 mm) with column compartment kept at 30°C. The autosampler tray was chilled to 4°C. The mobile phase flow rate was 300 μL/min, and the gradient elution program was: 0–3 min, 85% MPB; 3–10 min, 85–30% MPB; 10–20 min, 30–2% MPB; 20–25 min, 2% MPB; 26–30 min, 2–85% MPB. The total run time was 30 min. Data were acquired using a Thermo Orbitrap Exploris 240 Mass Spectrometer (Thermo Fisher) under ESI positive/negative ionization (polarity switching) mode at a resolution of 120,000.

All the raw files were imported to Thermo Trace Finder 4.1 software (Thermo Fisher) for final analysis. The fractional abundance of each isotopologue is calculated by the peak area of the corresponding isotopologue normalized by the sum of all isotopologue areas.

Measuring Alanine Levels in Human OCCCs

Human OCCC samples were obtained from patients undergoing primary debulking surgery at the University of Chicago. All patients were consented before surgery for tissue collection protocol, and patient-derived tissues were obtained fresh by the Human Tissue Resource Center at the University of Chicago according to approved institutional review board protocol (IRB-13372B). The paraffin-embedded tumor samples corresponding to the fresh frozen samples were obtained from the archives of the Department of Pathology according to approved institutional review board protocol (IRB-13372B).

Alanine was extracted from tumor samples similar to the procedure described previously (26). The frozen tumors were pulverized using a mortar and pestle on dry ice. The metabolites were extracted using the ice-cold 4/4/2 acetonitrile/methanol/water (20 μL solvent per mg of tissue, LC-MS grade solvents), homogenized (Omni International, TH115-PCR5H, stainless steel probe), vortexed and subjected 2 times- to sonication for 2 minutes in ice-cold water bath, freeze in liquid nitrogen for 1 minute, thaw on ice and subsequently vortex for 5 min at 2000 rpm and 4° C using Thermomixer. Samples were incubated on ice for 20 minutes, centrifuged at 20,000g for 20 minutes at 4°C and 500μL of supernatant from each sample was dry down using the Genevac EZ-2.4 elite evaporator. The dry-down samples were stored at −80°C until the analysis. On the day of the analysis, the samples were re-suspended in 500 μL of 60/40 acetonitrile/water. The LC-MS was performed at the University of Chicago Medicine Comprehensive Cancer Center Metabolomics Platform as described previously (26).

Kaplan-Meier Survival Curve Analysis

To analyze the overall survival (OS) of cancer patients receiving immunotherapy, a web-based survival analysis tool (https://www.kmplot.com) was used (27). The analysis included the total of 933 patient samples from the immunotherapy database encompassing various cancer types, including bladder cancer, esophageal adenocarcinoma, glioblastoma, hepatocellular carcinoma, head and neck squamous cell cancer, melanoma, non-small-cell lung cancer, small cell lung cancer and urothelial cancer. Patients in the database received different immune checkpoint inhibitors, including anti-PD-1 (nivolumab or pembrolizumab), anti-PD-L1 (atezolizumab or durvalumab), or anti-CTLA-4 (ipilimumab or tremelimumab). The median-grouping (low, high) was chosen to divide patients into two groups based on gene expression of SLC38A2 or SLC7A8.

Measurement of ¹⁴C Activity by Liquid Scintillation Counting

Cells were seeded (300,000 per well for RMG1, 400,000 per well for VOA4841 or 400,000 per well for XVOA295) in a 6-well plate. After 48h, media was replaced with no glucose DMEM containing 10% dialyzed FBS and 0.1 μCi/mL ¹⁴C₁-L-alanine (Moravek, Cat#: MC2279). The cells were then incubated for an additional 6 h. Whole cell lysates were collected using 100 μL of RIPA lysis buffer. Protein samples were purified using a chloroform-methanol precipitation method: Cells were lysed with 100 μL of RIPA lysis buffer, followed by centrifugation to remove cell debris. 400 μL of methanol, 100 μL of chloroform and 300 μL H₂O were added to the supernatant sequentially, with vortexing performed after each addition. Then the mixture was centrifuged at 15,000 × g for 1 min, producing three distinct layers: a large aqueous layer on top, a circular flake of protein in the interphase, and a smaller chloroform layer at the bottom. The top aqueous layer was carefully removed, and 400 μL of methanol was added to the remaining mixture. The sample was then centrifuged at 15,000 × g for 1 min. The supernatant was discarded, and the protein flake was air-dried. The dried protein flake was re-dissolved in 100 μL of 1% SDS (50 mM Tris-HCl, pH 8.0). Whole cell lysates and protein samples were mixed with liquid scintillation cocktail (Revvity, Cat#: 6013321) at a 1:100 ratio and analyzed using a Tri-Carb 4910TR Liquid Scintillation Counter (PerkinElmer).

OCCC Orthotopic Mouse Models

The protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of The Wistar Institute or The University of Texas MD Anderson Cancer Center. Mice were maintained at 22–23°C with 40–60% humidity and 12 hours light/12 hours dark cycle. Briefly, 1×10⁶ cells (RMG1, ARID1A knockout RMG1 and TOV21G) with inducible shSLC38A2 were injected into the ovarian bursa sac of 6–8 weeks old female NSG mice. Once the tumors were palpable (7 days after implantation for RMG1 and TOV21G, 12 days after implantation for ARID1A knockout RMG1), tumor bearing mice were randomized into two groups (n=7 per group). The mice in each group were intraperitoneally injected with vehicle or 20 mg/kg doxycycline (Dox) every two days for three weeks. In addition, dox-treated group was continuously provided with dox containing drinking water (2 mg/L). Mice were then sacrificed, and tumor burden was examined using tumor weight as a surrogate in each treatment groups.

OCCC Transgenic Mouse Model

The protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of The Wistar Institute or The University of Texas MD Anderson Cancer Center. In the conditional Arid1a−/−;Pik3ca^H1047R genetic OCCC mouse model, for Slc38a2 knockout in vivo, guide RNA (5’- TTTCCCTCTGTGTAGCATCC -3’) targeting murine Slc38a2 gene was inserted into pCRISPR-Cre plasmid. After lentivirus packaging and concentration, lentivirus of CRISPR-Cre or Slc38a2-CRISPR-Cre were intrabursally injected into the ovarian sac of 6–8 weeks old Arid1a−/−;Pik3ca^H1047R transgenic mice. After four weeks, the OCCC tumors bearing mice were randomized for IgG (Bio X Cell, Cat#: BE0090, RRID:AB_1107780,10 mg/kg, twice per week) or anti-PD-L1 antibody (BioXcell, BE0101, RRID:AB_10949073, 10 mg/kg, twice per week) treatments. After 3 weeks of treatment, mice were then sacrificed, and tumor burden was examined using tumor weight as a surrogate in each treatment groups.

For CD8⁺ T cell depletion, an anti-CD8 antibody (BioXCell, Cat#: BE0117, RRID:AB_10950145, 10 mg/kg, twice per week) was used to deplete CD8⁺ T cells. An isotype-matched IgG (Bio X Cell, Cat#: BE0090, RRID:AB_1107780,10 mg/kg, twice per week) was used as a negative control. For survival experiments, the guidelines of The Wistar Institute IACUC were used for endpoint assessment, such as when tumor burden exceeded 10% of body weight.

Ultrasound Tumor Monitoring

In Arid1a−/−;Pik3ca^H1047R genetic OCCC mouse model, following the ovarian bursa sac injection, all mice were given 10–14 days to recover from the surgery. Tumor initiation was then screened weekly using the Vega Preclinical Ultrasound System (Revvity) under anesthesia with 2.5% isoflurane (28). Briefly, mice were placed flat on an acoustically transmissible membrane (proprietary material, Revvity) and a thin layer of water was used to couple the membrane to the tissue. The mice were flat on the membrane and were imaged in the supine positions using a linear array transducer pulsing at 24 MHz. The transducer was pointing up and in contact with the imaging membrane and was robotically scanned to acquire 3D wide-field scans with a step size of 0.1 mm. Acquisition lasted less than 3 min per mouse (with an additional 2–3 min of prep time for depilation). Ultrasound images and tumor volumes were analyzed using SonoEQ 2.0.2 software (Revvity). Tumor initiation was considered if volume ≥1 mm³ and sustained or increased in volume over the course of the experiment.

Co-efficiency of Drug Interaction (CDI) Analysis

CDI was calculated as follows: CDI = ((cA+cB)–cA×cB)/cAB, where cA represents the inhibitory rate of inhibitor A, cB represents the inhibitory rate of inhibitor B, and cAB represents the inhibitory rate of combination treatment of inhibitor A and B (29). Synergy is defined as CDI lower than 1.0, and antagonism is defined as CDI significantly greater than 1.0.

Mass Spectrometry Imaging (MSI)

MSI was performed as previously described (30). Briefly, prior to metabolite analysis, sections were deparaffinized with xylene, 2 × 3 min, with no further rehydration. Fiducial points were etched onto the slides with a diamond scribe, and images were acquired of the sections using an Epson Perfection V600 Photo flatbed document scanner (Epson US) at 4800 dpi. Tissue sections were coated with 4 mg/mL 1,5-diaminonaphthalene in 50% ethanol, 50% 0.1 M HCl using an HTX M5 Robotic Reagent Sprayer (HTX Imaging) over 10 passes, with a nozzle temperature of 75°, a flow rate of 0.10 mL/min, a track speed of 1200 mm/min, and a track spacing of 3 mm. Mass spectrometry images were acquired on a Bruker timsTOF fleX QTOF mass spectrometer (Bruker) at 20 μm resolution with 200 laser shots summed per pixel. Data were acquired in negative ion mode over the m/z range 50–600. Voltages and transfer and storage times were appropriately optimized. FlexImaging 7.0 was used to align each slide to the optical image acquired on the Epson scanner using the etched fiducial points. The aligned image was then used to guide data collection from the tissue sections. Prior to each image acquisition, the instrument was mass-calibrated using red phosphorus. After collection, data files were loaded into SCiLS Lab 2024a Pro (Bruker) for visualization. Peaks from each imaging dataset were manually picked, excluding known matrix peaks and non-monoisotopic peaks. Metabolites were putatively identified using the SCiLS MetaboScape 2023b plugin. Images were exported from SCiLS as .imzML files for further analysis and integration with other imaging modalities.

Sequential Immunofluorescence Staining (seqIF) and Image analysis

SeqIF was performed using the COMET instrument (Lunaphore Technologies) as previously described (30,31). Briefly, FFPE Tissue Microarray slides were deparaffinized in xylene followed by rehydration in a graded alcohol series and blocked with 3% hydrogen peroxide for 10 min. Antigen retrieval was performed with EZ-AR2 Elegance buffer (BioGenex, Cat#: HX032-YCD) at 107 °C in an EZ-Retriever system V.3 (BioGenex) for 15 min. The processed slides were then transferred to a Multistaining Buffer (Lunaphore Technologies, Cat#: BU06) bath until use. The microfluidic chip (9 × 9 mm imageable area) was clamped against the FFPE tissue section on a standard microscope slide forming a closed reaction chamber. The reagents were delivered through microfluidic channels under highly controlled conditions. Automated multiplex sequential immunofluorescence staining and imaging was performed on the COMET platform (Lunaphore Technologies). Slides underwent iterative staining and imaging, followed by elution of the primary and secondary antibodies. Primary rabbit anti-CD3e (1:600; Cell Signaling, Cat#: 73484), rat anti-CD4 (1:100; Thermo fisher, Cat#: 14–9766-82, RRID:AB_2573008), rabbit anti-CD45 (1:200; Cell Signaling, Cat#: 98819), rat anti-CD8a (1:100; Thermo fisher, Cat#: 14–0195-82, RRID:AB_2637159), rabbit anti-PD-1 (1:200; Cell Signaling, Cat#: 55789), mouse anti-Ki67 (1:100; Cell Signaling, Cat#: 62548, RRID:AB_2935678), rat anti-B220 (1:250; Thermo fisher, Cat#: 14–0452-82, RRID:AB_467254), rat anti-Ly-6C (1:50; Bio-rad, Cat#: MCA2389GA, RRID:AB_844551), rabbit anti-CD11b (1:3000; Abcam, Cat#: AB209970, RRID:AB_2915959) and rabbit anti-CD206 (1:400; Cell Signaling, Cat#: 87887SF) were diluted to desired concentrations based on preliminary titration tests to optimize signal-to-noise ratio in multistaining buffer with 3% BSA and 1% horse serum (Sigma-Aldrich, Cat#: H1338). Secondary anti-rabbit Alexa Fluor 555 (1:200; Invitrogen, Cat#: A21430, RRID:AB_10374475), anti-rabbit Alexa Fluor 647 (1:400; Invitrogen, Cat#: A21246, RRID:AB_2535814), anti-rat Alexa Fluor 555 (1:200; Invitrogen, Cat#: A21434, RRID:AB_141733), anti-rat Alexa Fluor 647 (1:400; Invitrogen, Cat#: A21247, RRID:AB_141778) and anti-mouse Alexa Fluor 555 (1:200; Invitrogen, Cat#: A48287, RRID:AB_2896353) were used in multistaining buffer, respectively regardless of species reactivity. Primary antibody incubation was carried out for 8 min for each cycle, and all secondary antibodies were incubated for 2 min. Antibodies were then eluted following each cycle for 4 min. DAPI (Thermo fisher, Cat#: 62248) was used either alone, or in conjunction with secondary antibodies, at a 1:2000 dilution in multistaining buffer.

Image analysis was performed using Visiopharm image analysis software version 2023.09 x64 (Visiopharm). Fluorescent images for each sample were first aligned to the corresponding MSI images H&E images utilizing the Tissuealign module to obtain a 3-dimensional image. The H&E layer was used for tissue segmentation to separate tumor and non-tumor areas. After tissue segmentation, cell boundaries were determined by a pretrained machine learning algorithm that used DAPI channel to automatically identify nuclei and cells. Identified cells were then phenotyped using Visiopharm’s unbiased autoclustering module using only the top 20% of pixel values per cell.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism software (version 8.0). Analysis of variance with Fisher’s least significant difference was used to identify significant differences in multiple comparisons. Log-rank test was used to compare the survival distributions among experimental groups. Experiments were repeated at least twice. Quantitative data are expressed as mean ± SD unless otherwise stated. No statistical method was used to predetermine sample size. No data were excluded from the analyses. All analyses were performed blindly but not randomly. All mice for animal experiments were randomized.

Data Availability

The previously published ChIP-seq data that were re-analyzed here are available in the Gene Expression Omnibus under accession codes GSE120060(24), GSE69566(32), GSE124225(33), GSE123284 (34). Previously published RNA-seq data that were re-analyzed here are available in the Gene Expression Omnibus under accession codes GSE120060 (24) and are available at https://github.com/kbolton-lab/Bolton_OCCC (35). Previously published metabolomics data are available in the MassIVE under accession code MSV000086347 (7).

This paper does not report original code.

All raw data generated in this study are available upon request from the corresponding author.

Results

ARID1A Inactivation Increases Intracellular Alanine Levels

Given that SLCs are emerging cancer therapeutic targets, we determined whether ARID1A regulates SLCs. Toward this goal, we examined the changes in SLC gene expression based on a previously published RNA-seq dataset comparing control ARID1A wildtype and ARID1A knockout RMG1 cells (Fig. 1A) (24). This analysis revealed that neutral amino acid importer SLC38A2 is upregulated, while neutral amino acid exporter SLC7A8 is downregulated by ARID1A knockout (Fig. 1B). To determine which neutral amino acid is regulated by SLC38A2 and SLC7A8 in response to ARID1A knockout, we examined steady state metabolites in two isogenic ARID1A wildtype and ARID1A knockout OCCC cell lines, namely RMG1 and OVCAR429 (7). The analysis revealed that alanine was the only consistent change in neutral amino acid between these two isogenic cell lines (Fig. 1C; Supplementary Fig. S1A). Notably, alanine levels were significantly higher in ARID1A mutant (n=4) compared with ARID1A wildtype (n=7) primary human OCCC specimens (Fig. 1D).

Figure 1. ARID1A inactivation increases intracellular alanine leveling by concomitant upregulating importer SLC38A2 and downregulating exporter SLC7A8.

A, Expression of ARID1A in the indicated clear cell ovarian cancer (OCCC) cell lines determined by immunoblot.

B, Heatmap of differentially expressed SLC family member genes between control and ARID1A knockout RMG1 cells in a previously published RNA-seq analysis (GEO: GSE120060). Transporters were considered differentially expressed if fold change (FC) ≥ 2.5 and FDR ≤ 0.001. n = 3 biological independent experiments.

C, Volcano plot showing changes for metabolites between control and ARID1A knockout RMG1 cells in a previously published metabolomics data (MassIVE: MSV000086347). Plot shows average of 3 biological independent experiments. Metabolites were considered significantly changed if fold change (FC) ≥ 1.5 and P ≤ 0.01. P-values were calculated using two-tailed t-test.

D, Peak areas of alanine were determined by LC/MS +/- HILIC in OCCC tissues with wildtype or mutant ARID1A. P-values were calculated using two-tailed t-test. n = 7 independent samples in the ARID1A wildtype group, n = 4 independent samples in the ARID1A mutant group. Error bars represent mean with SD.

E and F, Expression of ARID1A and SLC38A2 in the indicated RMG1 cells determined by immunoblot (E). Please note that the multiple bands in SLC38A2 immunoblots represent modified forms of SLC38A2 through glycosylation. The relative levels of intracellular alanine were measured by HRLC/IC-MS in the indicated cells (F). P-values were calculated using two-tailed t-test. n = 3 biological independent experiments. Error bars represent the mean with SD.

G and H, Expression of ARID1A and SLC7A8 in the indicated RMG1 cells determined by immunoblot (G). Cells were incubated for 18 h in the presence of 1 mM ¹³C₃-pyruvate and the metabolites in media were collected for analysis. ¹³C labeled alanine secreted from the indicated cell lines were measured by HRLC/IC-MS (H). P-values were calculated using two-tailed t-test. n = 3 biological independent experiments. Error bars represent the mean with SD.

We next sought to directly demonstrate the role of SLC38A2 and SLC7A8 in regulating intracellular alanine levels. Toward this goal, we knocked down SLC38A2 in ARID1A knockout cells and measured intracellular alanine levels. SLC38A2 knockdown significantly mitigated the increase in the levels of intracellular alanine observed in ARID1A knockout cells (Fig. 1E and F). As a control, SLC38A2 knockdown in ARID1A knockout cells did not affect intracellular levels of other substrates of SLC38A2, namely glutamine, glycine and proline (Supplementary Fig. S1B). Conversely, we ectopically expressed the exporter SLC7A8 and measured secreted ¹³C-labeled alanine in the supernatant. Indeed, SLC7A8 ectopic expression significantly restored the decrease in alanine secretion induced by ARID1A knockout (Fig. 1G and H). In contrast, ectopic expression of SLC7A8 in ARID1A knockout cells did not significantly increase the levels of other SLC7A8 substrates in the culture medium, namely leucine, tryptophan, glutamine and valine (Supplementary Fig. S1C). Thus, we conclude that ARID1A inactivation increases intracellular alanine levels by simultaneously upregulating importer SLC38A2 and downregulating exporter SLC7A8.

Alanine Regulates the Growth of OCCC Cells in an ARID1A Status-dependent Manner

We next determined whether alanine plays a role in the growth of OCCC cells with or without ARID1A knockout. We examined the cell growth by colony formation assay in both normal and nutrient-limited conditions mimicked by 10mM or 1mM glucose in the culture medium with or without supplementation of 1mM alanine. Notably, we utilized an alanine-free medium and dialyzed FBS for cell culture. In 10mM glucose condition, alanine supplementation did not have any effect on cell growth regardless of ARID1A status (Supplementary Fig. S2A and S2B). However, in nutrient-limited 1mM glucose conditions, alanine supplementation significantly rescued the growth inhibition in ARID1A knockout but not wildtype cells (Fig. 2A; Supplementary Fig. S2A). Similar findings were made in two isogenic ARID1A wildtype and knockout OCCC cell lines (Fig. 2A and Supplementary Fig. S2A) and three ARID1A-mutated OCCC cell lines (Fig. 2B and Supplementary Fig. S2B). This suggests that this is not a cell-line-specific effect. This result suggests that ARID1A inactivation enables cells to use alanine to regulate cell growth in nutrient-limited conditions. Notably, concomitant loss of SMARCA4/2 reduced glucose uptake and glycolysis, while increasing import of glutamine (21). This correlated with a decrease in GLUT1 expression. Consistently, we observed a decrease in GLUT1 expression by ARID1A knockout (Supplementary Fig. S2C and S2D). Supplementation of alanine at a dose (up to 140mM) that competes with glutamine import suppresses SMARCA4/2 deficient cancer cells (21). To reconcile this discrepancy, we cultured control and ARID1A knockout cells under normal (10 mM) or low (1 mM) glucose conditions, with or without 2 mM glutamine, and at varying concentrations of alanine (1, 10, 100 mM). Regardless of glutamine concentration, supplementation of alanine at a physiological concentration (1 mM) (36) increased the viability of ARID1A knockout cells, but not control wildtype cells, under glucose-low condition (Fig. 2C). In the presence of glutamine, ARID1A knockout cells are hypersensitive to super-physiological dose of alanine (100mM) under both normal and low glucose conditions (Fig. 2C). This is consistent with the finding that a high dose of alanine competes with glutamine and inhibit its uptake in cells with SMARCAR 4/2-loss (21). This is also consistent with our prior finding that ARID1A-mutated OCCCs exhibit a dependency on glutamine uptake (7), which was validated here by showing that the viability of ARID1A knockout cells is significantly lower compared with wildtype cells when cultured in glutamine-free medium (Fig. 2C). However, in the absence of glutamine, a high dose of alanine failed to further reduce the viability of ARID1A knockout cells (Fig. 2C). This supports that the inhibitory effect of a high dose alanine is due to its competition with glutamine uptake (21). Thus, the discrepancy can be reconciled by doses of alanine used in these studies.

Figure 2. Alanine regulates the growth of OCCC cells in an ARID1A status dependent manner.

A, Growth of control and ARID1A knockout RMG1 cultured in DMEM containing 10 mM glucose or 1 mM glucose with or without 1 mM alanine supplementation was determined by colony formation assay. P-values were calculated using two-tailed t-test. n = 4 biological independent experiments. Error bars represent mean with SEM.

B, Growth of ARID1A-mutated TOV21G cells cultured in RPMI-1640 containing 10 mM glucose or 1 mM glucose with or without 1 mM alanine supplementation was determined by colony formation assay. P-values were calculated using two-tailed t-test. n = 4 biological independent experiments. Error bars represent mean with SEM.

C, Growth of control and ARID1A knockout RMG1 cultured in DMEM containing 10 mM glucose or 1 mM, with or without 2 mM glutamine, and at varying concentrations of alanine (1, 10, 100 mM) was determined by alamarBlue cell viability assay. P-values were calculated using two-tailed t-test. n = 5 biological independent experiments. Error bars represent mean with SD.

D and E, Expression of ARID1A and SLC38A2 in control and ARID1A knockout RMG1 cells expressing the indicated shSLC38A2s or control determined by immunoblot (D). Growth of the indicated cells was determined by colony formation assay (E). P-values were calculated using two-tailed t-test. n = 3 biological independent experiments. Error bars represent mean with SEM.

F and G, Expression of ARID1A and SLC38A2 in the indicated primary OCCC cultures determined by immunoblot (F). Growth of the indicated cells were determined by colony formation assay (G). P-values were calculated using two-tailed t-test. n = 3 biological independent experiments. Error bars represent mean with SEM.

H, Growth of RMG1 cells treated with vehicle control, 10 μM BD98 (a ARID1A-specific BAF complexes inhibitor) or 1 μM BRM014 (a BRM/BRG1 ATPase inhibitor) with or without SLC38A2 knockdown was determined by colony formation assay. P-values were calculated using two-tailed t-test. n = 3 biological independent experiments. Error bars represent mean with SD.

I and J, Expression of ARID1A and SLC7A8 in control and ARID1A knockout RMG1 cells with or without ectopic FLAG-tagged SLC7A8 expression (I). And growth of the indicated cells was determined by colony formation assay (J). P-values were calculated using two-tailed t-test. n = 4 biological independent experiments. Error bars represent mean with SEM.

Next, we knocked down SLC38A2 and determined that ARID1A inactivation exacerbated the growth inhibition induced by SLC38A2 knockdown (Fig. 2D and E; Supplementary Fig. S2E–S2H). Notably, a similar observation was made in primary cultures of OCCC with or without ARID1A expression (Fig. 2F and G) (23). Indeed, inhibitors of the SWI/SNF complex, namely BD98 that specifically inhibits ARID1A-containing complexes (37) and BRM014, a dual inhibitor of BRG1 and BRM1 ATPase activity (38) sensitized the cells to SLC38A2 knockdown (Fig. 2H). However, BD98 and BRM014 failed to further sensitize ARID1A knockout cells to SLC38A2 knockdown (Supplementary Fig. S2I). This supports the notion that the observed effects are on target and depend on the functional ARID1A-containing SWI/SNF complex. Conversely, we ectopically expressed SLC7A8 and observed that ARID1A inactivation enhanced the growth inhibition induced by ectopic SLC7A8 expression (Fig. 2I and J; Supplementary S2J–S2K). This suggests that restoration of SLC7A8 may represent a therapeutic strategy for ARID1A-mutated OCCCs. We conclude that alanine regulates the growth of OCCC cells, correlating with the upregulation of SLC38A2 and downregulation of SLC7A8 induced by ARID1A inactivation.

SLC38A2 and SLC7A8 are Direct Target Genes of the ARID1A-containing SWI/SNF Complex

We next determined the mechanism by which ARID1A regulates SLC38A2 and SLC7A8. We validated the upregulation of SLC38A2 and downregulation of SLC7A8 at both mRNA and protein levels by ARID1A knockout in both RMG1 and OVCAR429 cells (Fig. 3A–C and Supplementary Fig. S3A). Ectopic wildtype ARID1A expression rescued the changes in SLC38A2 and SLC7A8 expression induced by ARID1A knockout (Fig. 3C and Supplementary Fig. S3A–C). Likewise, wildtype ARID1A restoration in ARID1A mutant OCCC OVISE cells downregulated SLC38A2 and upregulated SLC7A8 (Fig. 3D). Similarly, restoration of SNF5 in SNF5 mutant rhabdoid tumor cell line G401 downregulated SLC38A2 and upregulated SLC7A8 (Fig. 3E). Notably, knockdown or knockout of other subunits of the SWI/SNF complex such as ARID1B, SNF5 and BRG1 upregulated SLC38A2 and downregulated SLC7A8 (Fig. 3F and G; Supplementary Fig. S3D). Consistently, inhibition of the SWI/SNF activity by BD98 and BRM014 upregulated SLC38A2 and downregulated SLC7A8 (Fig. 3H; Supplementary Fig. S3E).

Figure 3. SLC38A2 and SLC7A8 are direct target genes of the SWI/SNF complex.

A and B, Expression of SLC38A2 (A) and SLC7A8 (B) in control and ARID1A knockout RMG1 and OVCAR429 were determined by qRT-PCR. P-values were calculated using two-tailed t-test. n = 4 biological independent experiments. Error bars represent mean with SEM.

C, Control and ARID1A knockout RMG1 cells with or without wildtype ARID1A restoration were examined for expression of ARID1A, SLC38A2 and SLC7A8 by immunoblot.

D, ARID1A-mutated OVISE cells with or without wildtype ARID1A restoration were examined for expression of ARID1A, SLC38A2 and SLC7A8 by immunoblot.

E, SNF5-mutated G401 cells with or without wildtype SNF5 restoration were examined for SNF5, SLC38A2 and SLC7A8 expression by immunoblot.

F and G, RMG1 cells with BRG1 knockdown (F) or SNF5 knockdown (G) were examined for SLC38A2 and SLC7A8 expression by immunoblot.

H, Expression of SLC38A2 and SLC7A8 were examined by immunoblot in RMG1 cells treated with vehicle control, 10 μM BD98 (a ARID1A-specific BAF complexes inhibitor) or 1 μM BRM014 (a BRM/BRG1 ATPase inhibitor) for 4 days.

I and J, The indicated ChIP-seq and input tracks in the SLC38A2 (I) and SLC7A8 (J) gene loci in control and ARID1A knockout RMG1 cells from a previously published GEO dataset (GSE120060).

K and L, The association of ARID1A, BRG1, SNF5 and RNA Pol II with the SLC38A2 (K) and SLC7A8 (L) gene promoters in parental and ARID1A knockout RMG1 cells were examined by ChIP–qPCR analysis. An isotype-matched IgG was used as a negative control. P-values were calculated using two-tailed t-test. n = 4 biological independent experiments. Error bars represent mean with SEM.

M-O, Representative images of ARID1A, SLC38A2 and SLC7A8 in ARID1A-positive or -negative cases from a OCCC tissue microarray (M). Expression of SLC38A2 and SLC7A8 were scored as high or low based on the median of histological score (H-score) in the indicated ARID1A-positive and -negative OCCCs (N). n = 20 independent ARID1A-positive samples, n = 20 independent ARID1A-negative samples. Correlation between ARID1A and SLC38A2 or SLC7A8 in ARID1A-positive samples (O). n = 23 (20 independent ARID1A-positive OCCC samples, one endometrioid carcinoma and 2 independent endometriosis. The P-values were calculated by Fisher’s exact test in M and N, and by Pearson correlation in O.

We next re-analyzed a previously published ARID1A ChIP-seq dataset in ARID1A-wildtype and ARID1A knockout cells (24). The analysis revealed that SWI/SNF complex subunits, such as ARID1A and BRG1, are associated with promoters of the SLC38A2 and SLC7A8 genes (Fig. 3I and J). Similar findings were observed in other databases (Supplementary Fig. S3F). Notably, ARID1A knockout increased the active transcription epigenetic markers such as lysine 27 acetylation on histone H3 (H3K27ac) and lysine 4 tri-methylation on histone H3 (H3K4me3) in the locus of SLC38A2 gene, which correlated with an increase in RNA polymerase II (Pol II) in this locus (Fig. 3I). In contrast, ARID1A knockout decreased these active epigenetic markers in the locus of SLC7A8 gene, which correlated with a decrease in RNA Pol II (Fig. 3J). Thus, ARID1A represses SLC38A2 and promotes SLC7A8 transcriptionally. We next validated the association of ARID1A, BRG1 and SNF5, core subunits of the SWI/SNF complex, with the promoters of SLC38A2 and SLC7A8. ARID1A knockout reduced the association of both BRG1 and SNF5 with the promoters of the SLC38A2 and SLC7A8 genes (Fig. 3K and L; Supplementary Fig. S3G and S3H). Consistent with the notion that ARID1A inactivation upregulates SLC38A2, ARID1A knockout increased the association of RNA Pol II with the SLC38A2 gene promoter (Fig. 3K; Supplementary Fig. S3G). In contrast, ARID1A knockout decreased the association of RNA Pol II with the SLC7A8 gene promoter (Fig. 3L; Supplementary Fig. S3H), which correlated with downregulation of SLC7A8 by ARID1A knockout. This data supports that ARID1A-containing SWI/SNF complex functions as a repressor of SLC38A2 and an activator of SLC7A8.

We next correlated the protein expression of ARID1A with SLC38A2 and SLC7A8 in a panel of 40 cases of human OCCCs (Fig. 3M). We observed that ARID1A positive OCCCs correlated with a significantly lower level of SLC38A2 (Fig. 3N). In contrast, ARID1A-negative expression correlated with a significantly lower level of SLC7A8 (Fig. 3N). In ARID1A positive 20 human OCCCs, one endometrioid ovarian carcinoma and 2 endometriosis tissues, we observed a negative correlation between expression of ARID1A expression and SLC38A2 and a positive correlation between expression of ARID1A and SLC7A8 (Fig. 3O). Finally, in a previously published large OCCC dataset (35), SLC38A2 is expressed significantly higher in tumors with alterations in the SWI/SNF complex. In contrast, SLC7A8 is expressed significantly lower in these tumors (Supplementary Fig. S3I and S3J). Together, we conclude that the ARID1A-containing SWI/SNF complex functions as a repressor of SLC38A2 and as an activator of SLC7A8.

ARID1A Inactivation Increases Alanine Utilization in Protein Synthesis and the TCA Cycle

We next determined how the ARID1A status differentially affects alanine utilization. Toward this goal, we examined the expression of GPT and GPT2, two enzymes involved in alanine catabolism. Notably, GPT is barely detectable by immunoblot in all the OCCC cell lines tested (Supplementary Fig. S4A). In contrast, GPT2 is expressed at various levels in different cell lines (Supplementary Fig. S4A). However, ARID1A knockout did not affect GPT2 expression (Supplementary Fig. S4A–S4C). We next determined the effects of GPT2 inhibition by genetical knockdown or a small molecule inhibitor L-cycloserine on the growth of OCCC cells with or without ARID1A knockout. Notably, ARID1A knockout did not affect the growth inhibition induced by GPT2 inhibition (Supplementary Fig. S4D–S4F). Likewise, wildtype ARID1A restoration in ARID1A mutant OVISE cell line with low GPT2 expression did not affect growth inhibition induced by GPT2 knockdown (Supplementary Fig. S4A and S4G–S4I). Similar observations were made in cells cultured in both dialyzed and normal FBS (Supplementary Fig. S4H and S4I). These findings suggest that the increase in alanine uptake by ARID1A inactivation was not utilized for catabolism that depends on GPTs.

We next measured glycolysis by seahorse analysis in control and ARID1A knockout cells with or without alanine supplementation in the nutrient-limited culture medium. Although ARID1A knockout decreased glycolysis as determined by seahorse analysis, alanine supplementation did not affect glycolysis (Supplementary Fig. S5A–S5C). In contrast, ARID1A inactivation increased the TCA cycle activity and alanine supplementation further increased the TCA cycle as determined by oxygen consumption rate (OCR) (Fig. 4A; Supplementary Fig. S5D). This is consistent with the findings that the cytosolic GPT is barely expressed, while mitochondrial GPT2 is expressed in OCCC cells (Supplementary Fig. S4A). Thus, our data showed that alanine uptake is regulated by SLC38A2 (Fig. 1E and 1F), and the alanine supplementation increases OCR in ARID1A-inactivation cells (Fig. 4A; Supplementary Fig. 5D). These findings are consistent with the notion that the observed increase in OCR may be attributed to the upregulation of SLC38A2. Accordingly, we knocked down SLC38A2 in control and ARID1A knockout cells. Indeed, the increase in OCR induced by ARID1A knockout was reduced by SLC38A2 knockdown (Fig. 4B; Supplementary Fig. S5E). However, since GPT2 knockdown did not differentially affect the growth of control and ARID1A knockout cells (Supplementary Fig. S4), this raises the possibility that alanine utilization in regulating the TCA cycle is passive. To directly test this possibility, instead of supplementing alanine in the cell culture medium continuously, we acutely injected the alanine immediately before measuring OCR using the seahorse. Under these conditions, alanine supplementation failed to significantly increase OCR in ARID1A knockout cells (Fig. 4C–E). Notably, SLC38A2 knockdown decreased OCR in these conditions (Fig. 4C–E). Together, these results support the notion that the increase in alanine uptake by ARID1A knockout contributes to the TCA cycle in a passive manner but not glycolysis.

Figure 4. ARID1A inactivation increases alanine utilization in protein synthesis and the TCA cycle.

A, Oxygen Consumption Rate (OCR) determined by a seahorse mitochondrial stress test in control and ARID1A knockout RMG1 cells cultured in medium with or without 1 mM alanine. Error bars represent mean with SD of 4 biological independent experiments.

B, Basal OCR in control and ARID1A knockout RMG1 cells expressing the indicated shSLC38A2s or control was determined by using a seahorse mitochondrial stress test. P-values were calculated using two-tailed t-test. n = 6 biological independent experiments. Error bars represent mean with SEM.

C-E, OCR determined by a seahorse mitochondrial stress test in control (C) and ARID1A knockout (D) RMG1 cells expressing shSLC38A2#1 or control with or without acute alanine injection. Basal OCR for the indicated groups were calculated (E). P-values were calculated using two-tailed t-test. n = 5 biological independent experiments. Error bars represent mean with SD.

F, Expression of SLC38A2 in control and ARID1A knockout RMG1 cells expressing an inducible shSLC38A2#1 with indicated doxycycline treatments for 72 h determined by immunoblot.

G and H, The indicated RMG1 cells were treated with 1 μg/ml doxycycline or vehicle for 72 h, followed by incubating in DMEM containing 1 mM glucose and 1mM ¹³C₃-alanine for 18 h. Intracellular metabolites were extracted to measure the indicated metabolites by LC–MS/MS (G) or secreted metabolites in the media were collected for analysis by HRLC/IC-MS (H). Mass isotopologs (M + X) analysis of the indicated metabolites are shown as percentage of indicated number of carbons labeled with heavy isotype.

I, The indicated RMG1 cells were incubated in DMEM containing no glucose and 0.1 μCi/mL ¹⁴C₁-L-alanine for 6 h. Radio signal was detected in total cell lysates or precipitated proteins by liquid scintillation.

P-values were calculated using two-tailed t-test. n = 3 biological independent experiments. Error bars represent mean with SD.

We next determined alanine utilization by ¹³C₃-alanine isotope tracing. To limit the potential adaptive response induced by constitutive SLC38A2 knockdown, we generated a doxycycline-inducible SLC38A2 shRNA in wild-type and ARID1A knockout RMG1 cells. We validated the inducible knockdown of SLC38A2 in both wildtype and ARID1A knockout RMG1 cells (Fig. 4F). Notably, inducible SLC38A2 knockdown suppressed cell growth more effectively in ARID1A knockout compared with control cells (Supplementary Fig. S5F), which correlates with an accumulation of cells at the G0/G1 phase of the cell cycle and a decrease in the S phase of the cell cycle (Supplementary Fig. S5G). However, markers of apoptosis, such as annexin V, were not induced by SLC38A2 knockdown in ARID1A knockout cells (Supplementary Fig. S5H). Consistently, inducible SLC38A2 knockdown significantly reduced OCR in ARID1A knockout but not wild-type control cells (Supplementary Fig. S5I). Cells were next incubated with ¹³C₃-alanine to infer alanine metabolism and associated metabolic pathways. The ¹³C₃-alanine stable isotope tracer analysis revealed that ARID1A knockout increased intracellular alanine, which was reduced by SLC38A2 knockdown. Notably, the major labelled metabolite was N-acetylalanine (Fig. 4G), indicating that alanine was mainly used for protein synthesis. In contrast, pyruvate and lactate as well as TCA cycle intermediates such as citrate and α-ketoglutarate and glycolysis intermediates such as 3-phosphoglycerate and phosphoenolpyruvate were less or not labeled (Fig. 4G). We next examined alanine metabolites in the extracellular culture medium. Validating our approach, we observed a decrease in alanine in cell culture medium in ARID1A knockout cells. Consistent with the notion that alanine is mainly used for protein synthesis, the levels of labeled pyruvate, lactate and N-acetylalanine are low in the culture medium (Fig. 4H). To directly determine the contribution of alanine to protein synthesis, we performed ¹⁴C₁-L-alanine labeling. Indeed, ¹⁴C₁-L-alanine label was increased in the precipitated proteins by ARID1A knockout, which was reduced by SLC38A2 knockdown (Fig. 4I). Similar findings were made in primary OCCC cultures as well (Supplementary Fig. S5J). Together, we conclude that the increase in alanine uptake by ARID1A inactivation is predominantly utilized for protein synthesis and to a lesser extend in the TCA cycle in a passive manner.

SLC38A2 Inhibition Suppresses the Growth of ARID1A-inactivated OCCCs

We next determined whether SLC38A2 is a therapeutic target for ARID1A-mutated OCCC. Toward this goal, we generated a SLC38A2 inducible knockdown ARID1A-mutated TOV21G cell line (Fig. 5A and B) and used orthotopic xenograft models. Briefly, the orthotopically transplanted cells were allowed to establish the orthotopic tumors (Supplementary Table S1). Mice were then randomized for with or without doxycycline treatment for 3 weeks. We used tumor weight as a surrogate for tumor burden. Notably, SLC38A2 knockdown significantly reduced the burden of orthotopic xenografts formed by ARID1A-mutated TOV21G cells (Fig. 5C and D). The observed tumor suppressive effect by inducible SLC38A2 knockdown is ARID1A status dependent. For example, inducible SLC38A2 knockdown significantly reduced the burden of orthotopic xenografts formed by ARID1A knockout RMG1 cells (Fig. 5E and F). In contrast, SLC38A2 knockdown did not significantly affect the growth of tumors formed by ARID1A wildtype control RMG1 cells (Fig. 5G and H). SLC38A2 knockdown significantly reduced the expression of cell proliferation marker Ki67 and mitotic marker serine 10 phosphorylated histone H3 (p-H3S10) in tumors formed by ARID1A-mutated TOV21G or ARID1A knockout but not control wildtype RMG1 cells (Figure 5I and J; Supplementary Fig. S6). However, expression of apoptosis markers such as cleaved caspase 3 was not affected by SLC38A2 knockdown (Figure 5I and J; Supplementary Fig. S6). This is consistent with the in vitro finding that SLC38A2 knockdown significantly reduced S phase of the cell cycle in ARID1A knockout, but not wildtype cells. Thus, we conclude that inhibition of SLC38A2 suppresses the growth of ARID1A-inactivated OCCCs.

Figure 5. SLC38A2 inhibition suppressed growth of ARID1A-inactivated OCCCs in vivo.

A, Expression of SLC38A2 determined by immunoblot in ARID1A-mutated TOV21G cells expressing an inducible shSLC38A2 with or without the indicated doses doxycycline treatment for 72 h.

B, Growth of the indicated ARID1A-mutated TOV21G cells expressing an inducible shSLC38A2 with or without the indicated doses doxycycline treatment was determined by colony formation assay. P-values were calculated using two-tailed t-test. n = 4 biological independent experiments. Error bars represent mean with SEM.

C and D, Orthotopic xenografts formed by ARID1A-mutated TOV21G cells expressing an inducible shSLC38A2 treated with vehicle or doxycycline. Shown are images of reproductive tracks with tumors from indicated groups at the end of treatment (C). Tumor weight was measured as a surrogate for tumor burden (D). P-values were calculated using two-tailed t-test. n = 7 mice per group. Error bars represent mean with SEM.

E and F, Orthotopic xenografts formed by ARID1A knockout RMG1 cells expressing an inducible shSLC38A2 treated with vehicle or doxycycline. Shown are images of reproductive tracks with tumors from indicated groups at the end of treatment (E). Tumor weight was measured as a surrogate for tumor burden (F). P-values were calculated using two-tailed t-test. n = 7 mice per group. Error bars represent mean with SEM.

G and H, Orthotopic xenografts formed by ARID1A wildtype RMG1 cells expressing an inducible shSLC38A2 treated with vehicle or doxycycline. Shown are images of reproductive tracks with tumors from indicated groups at the end of treatment (G). Tumor weight was measured as a surrogate for tumor burden (H). P-values were calculated using two-tailed t-test. n = 7 mice per group. Error bars represent mean with SEM.

I and J, Tumors formed by the indicated control or ARID1A knockout RMG1 cells were subjected to H&E staining and immunological staining for SLC28A2, cell proliferation marker Ki67, mitotic marker serine 10 phosphorylated histone H3 (pH3S10) or apoptosis marker cleaved caspase 3 on serial sections (I) and the histological score (H-score) of the indicated markers was quantified from three separate fields from seven tumors from seven individual mice in each of the indicated treatment groups (J). Scale bar = 100 μm. P-values were calculated using two-tailed t-test. Error bars represent mean with SD.

SLC38A2 Inhibition Synergizes with Immune Checkpoint Blockade in an Arid1a-inactivated Genetic OCCC Mouse Model

In addition to cancer cells, T cells appear to exclusively use extracellular alanine for protein synthesis and alanine deprivation suppresses the proliferation and effector functions of T cells (13). This raises the possibility that SLC38A2 and SLC7A8 expression in ARID1A-inactivated cancer cells affect T cell-based anti-tumor immunity. Consistently, SLC38A2 expression negatively correlated with response to ICBs such as anti-PD-1 (nivolumab or pembrolizumab), anti-PD-L1 (atezolizumab or durvalumab), or anti-CTLA-4 (ipilimumab or tremelimumab) in a pan-cancer database that include 9 cancer types as determined by overall survival of patients treated with ICB (Fig. 6A). In contrast, SLC7A8 expression positively correlated with response to ICB in the same patient cohort (Fig. 6B). To directly test whether SLC38A2 expression in cancer vs. T cells affect T cell killing, we used T cells expressing chimeric antigen receptors (CAR T cells) targeting human CD19 (hCD19). We engineered control and ARID1A knockout cells expressing CD19 (Supplementary Fig. S7A). Killing assays using anti-CD19 CAR T cells showed that knockdown of SLC38A2 significantly enhanced ARID1A knockout cells to T cell assault (Fig. 6C; Supplementary Fig. S7B). Conversely, overexpression of SLC38A2 in anti-CD19 CAR T cells significantly enhanced its killing activity against ARID1A knockout cells (Fig. 6D and E; Supplementary Fig. S7C). Notably, ARID1A knockout augmented the elimination rate of CAR T cells (Fig. 6C and Supplementary Fig. S7B). This is consistent with a recent report that ARID1A inactivation in tumor cells activate the STING-type I IFN signalling to promote CD8+ T cell-based anti-tumor immunity (39).

Figure 6. SLC38A2 inhibition and anti-PD-L1 are synergistic in suppressing ARID1A-inactivated OCCCs.

A and B, Kaplan–Meier analysis of overall survival (OS) based on SLC38A2 (A) or SLC7A8 (B) mRNA levels in cancer patients receiving immunotherapy. n = 933 independent samples. P-values were calculated by Log-rank test.

C, Killing of control and ARID1A knockout RMG1 cells expressing CD19 with or without SLC38A2 knockdown was measured after coculture with anti-CD19 CAR T cells for 48 h at the 1:3 E:T ratio. P-values were calculated using two-tailed t-test. n = 3 biological independent experiments. Error bars represent mean with SD.

D, Expression of SLC38A2 and loading controls Na+/K+ ATPase in anti-CD19 CAR T cells with or without SLC38A2 expression was determined by immunoblot.

E, Killing of control and ARID1A knockout RMG1 cells expressing CD19 was measured after coculture with anti-CD19 CAR T cells with or without SLC38A2 overexpression for 48 h at the 1:3 E:T ratio. P-values were calculated using two-tailed t-test. n = 3 biological independent experiments. Error bars represent mean with SD.

F, Expression of Slc38a2 in Arid1a^−/−;Pik3ca^H1047R tumors developed from mice injected with a lentivirus encoding both Cre-recombinase and a sgRNA targeting mouse Slc38a2 or control was determined by immunoblot.

G-J, Mice bearing Arid1a^−/−;Pik3ca^H1047R OCCCs were randomized for the four indicated treatment groups. Images of reproductive tracts with tumors from the indicated groups at the end of treatment are shown (G). Tumor weight was measured as surrogate for tumor burden (H). Ascites produced in the indicated treatment groups (I) was quantified (J). Statistical co-efficiency of drug interaction (CDI) analysis revealed that the CDI for the combination were 0.88 for tumor weight and 0.89 for ascites volume (<1, indicative of a synergistic effect). P-values were calculated using two-tailed t-test. n = 5 mice per group. Error bars represent mean with SD.

In addition, ARID1A mutation confers sensitivity to ICBs such as anti-PD-L1(17,40). Thus, we sought to determine whether SLC38A2 inhibition in tumor cells synergizes with ICB in ARID1A inactivated OCCC. Toward this goal, we used a conditional genetic Arid1a^flox/flox /Pik3ca^H1047R OCCC mouse model as previously published (23,41). Specifically, we knocked out SLC38A2 in OCCC developed from the Arid1a^−/−;Pik3ca^H1047R genetic mouse model by injecting a lentivirus simultaneously encoding small guide RNA (sgRNA) to mouse Slc38a2 gene and a Cre-recombinase (Fig. 6F). Consistent with the results obtained using orthotopic immunocompromised xenograft models, SLC38A2 knockout significantly reduced the growth of OCCCs and ascites developed from the immunocompetent Arid1a^−/−;Pik3ca^H1047R genetic mouse model (Fig. 6G–J). We next treated control and SLC38A2 knockout OCCCs developed in the Arid1a^−/−;Pik3ca^H1047R models with IgG control or anti-PD-L1 antibody. Indeed, SLC38A2 knockout synergizes with anti-PD-L1 in reducing tumor burden (synergy index: 0.88<1) and ascites production compared with either of the individual treatments (Fig. 6G–H). Secondary to tumor burden, we observed a similar synergy in ascites production (synergy index: 0.89<1) (Fig. 6I–J). This result suggests that inhibition of SLC38A2 in tumor cells remodels the tumor immune microenvironment to synergize with ICB in ARID1A inactivated OCCC.

To test this possibility, we performed spatial metabolites and immune analysis. Our results showed that SLC38A2 knockout in cancer cells decreased the alanine ratio in tumor vs. non-tumor region (Fig. 7A–C). Consistently, alanine levels were increased in CD45⁺ cells (Fig. 7A–C). As a control, glutamine distribution in CD45⁺ cells were not significantly changed by SLC38A2 knockout in cancer cells (Supplementary Fig. S8A and S8B). Spatial immune profiling revealed that the combination significantly increased the infiltrating CD8⁺ T cells and there was a trend toward an increase in infiltrating CD4⁺ T cells (Fig. 7D and Supplementary Fig. S8C), which as validated by flow cytometry analysis (Supplementary Fig. S8D). In addition, there was an increase in proliferation markers such as Ki67 in CD8⁺ T cells (Supplementary Fig. S8E). This is consistent with previous reports that T cells exclusively rely on extracellular alanine for both proliferation and effector function (13). As a control, there was no change in other immune populations such as B cell, monocytes and M2 macrophage (Supplementary Fig. S8F). Notably, SLC38A2 knockdown did not affect PD-L1 expression in cancer cells (Supplementary Fig. S8G) or PD-1 expression in CD8⁺ or CD4⁺ T cells (Supplementary Fig. S8H). Together, these findings suggest that CD8⁺ T cell may play a key role in the observed tumor suppressive effects. Accordingly, we used an anti-CD8 antibody to deplete the CD8⁺ T cells. Notably, anti-CD8a antibody significantly abrogated the observed improvement in the survival of tumor bearing mice treated with the combination (Fig. 7E). These data show that inhibition of SLC38A2 synergizes with immune checkpoint blockade in suppressing the growth of ARID1A-inactivated OCCCs.

Figure 7. CD8 T cells contribute to the anti-tumor effects of SLC38A2 inhibition and anti-PD-L1 combination.

A, Arid1a^−/−;Pik3ca^H1047R tumors dissected the indicated treatment groups were subjected to H&E staining, sequential immunofluorescence staining (seqIF) and mass spectrometry imaging (MSI). Representative images showing H&E images were analyzed by Visiopharm image analysis software and used for tissue segmentation to separate tumor and non-tumor areas. SeqIF images were analyzed by Visiopharm image analysis software and used to visualize the spatial distribution of immune cells (e.g. CD4⁺ and CD8⁺ T cells), along with their exhaustion (PD-1) and proliferation (Ki67) markers. MSI images were analyzed by SCiLS Lab 2024a Pro and used to map spatial distribution of metabolites (e.g. alanine). The analyzed seqIF, H&E and MSI images were aligned using Visiopharm image analysis software.

B, The ratio of alanine mean intensity in the tumor region of interest (ROI) area to that in the non-tumor ROI area. The alanine mean intensity in each ROI area were measured by Visiopharm for the indicated treatment groups. P-values were calculated using two-tailed t-test. n = 5 mice per group. Error bars represent mean with SD.

C, Mean intensity of alanine in CD45⁺ cells were measured by Visiopharm for the indicated treatment groups. P-values were calculated using two-tailed t-test. n = 5 mice per group. Error bars represent mean with SD.

D, Numbers of infiltrating CD8⁺ T cells were counted by Visiopharm and normalized to the whole tissue ROI area in the indicated treatment groups. CD8⁺ T cells were identified by the overlapping immunofluorescence signals of CD45 and CD8a. P-values were calculated using two-tailed t-test. n = 5 mice per group. Error bars represent mean with SEM.

E, Mice bearing Arid1a^−/−;Pik3ca^H1047R OCCCs were randomized for the three indicated treatment groups. After completing treatment, mice were followed for survival and the Kaplan–Meier survival curves for each of the indicated groups are shown (n = 5 mice per group), P-values were calculated by Log-rank test.

Discussion

We show that ARID1A inactivation increases intracellular alanine levels through upregulation of importer SLC38A2 and downregulation of exporter SLC7A8. Interestingly, alanine was predominantly used for protein synthesis in ARID1A inactivated cells. Although alanine contributes to the TCA cycle in these cells, it does so in a passive manner. Consistently, the knockdown of GPT2 did not selectively affect the growth of ARID1A inactivated cells, and acute alanine supplementation failed to increase OCR in ARID1A inactivated cells. ARID1A inactivation decreases glycolysis, which is consistent with previous reports that ARID1A inactivation increases glutamine uptake and utilization (7). Likewise, SMARCA4-mutant lung cancer cells with intact SMARCA2 create a dependence on OXPHOS (42). Furthermore, concomitant loss of SMARCA4/2 reduced glucose uptake and glycolysis, while increasing import of glutamine via upregulating SCL38A2 (21). However, supplementation of alanine at a higher concentration (35 – 140mM) that competes with glutamine import suppresses SMARCA4/2 deficient cancer cells. In contrast, supplementation of alanine in the physiologically relevant dose (1mM) (36) in a nutrient-limited condition rescued the growth of ARID1A-inactivated OCCC cells. In addition, SLC38A2 regulates the competition for glutamine between tumor cells and conventional dendritic cells (22). Notably, SLC38A2 knockout did not affect intracellular glutamine in ARID1A knockout cells (Supplementary Fig. S1B) or change the distribution of glutamine in vivo (Supplementary Fig. S8A and S8B). Thus, the substrates for the upregulated SLC38A2 could be context-dependent. Beyond alanine’s role in tumor cells, our study conceptually established an interplay between ARID1A-deficient tumor and its immune microenvironment through SLC38A2 expression to affect responses to immunotherapy including ICB and CAR T cells.

Notably, GPT expression is barely detectable in most of OCCC cell lines tested. However, there was a low percentage of lactate labeling in the tracing experiments using RMG1 cells that express very low levels of GPT. This suggests that in cells with GPT expression, alanine may be utilized for glycolysis. However, ARID1A inactivation does not differentially create a dependence on GPT2 despite the increase in intracellular alanine levels. Thus, ARID1A inactivation increases alanine uptake and consequently intracellular alanine levels by upregulating SLC38A2 and downregulating SLC7A8. The abundant availability of alanine allows its utilization in the mitochondria to fuel the TCA cycle via conversion to pyruvate by GPT2. However, this is not essential because GPT2 inhibition and thus blocking its utilization in the TCA cycle does not affect the growth of ARID1A inactivated cells. This is consistent with our finding that alanine was predominantly utilized for protein synthesis. Thus, ARID1A inactivation increases intracellular alanine that is predominantly used for protein synthesis and, to a less degree, for the TCA cycle in a passive manner.

SLC38A2 expression in tumor cells suppresses CAR T cell-mediated killing. Conversely, SLC38A2 expression in CAR T cells enhances their killing of tumor cells. Our future studies will directly establish the competition for alanine by CAR T cells and tumor cells. Consistently, T cells depend on exogenous alanine for proliferation and activation (13). SLC38A2 expression negatively correlates with the survival of patients receiving ICB in a database consisting of different cancer types. The limited number of patients in the individual cancer types prevented us from performing a robust statistical analysis. Nonetheless, our finding suggests that the competition for alanine by T cells and tumor cells regulates the anti-tumor immunity. Indeed, SLC38A2 knockout synergizes with anti-PD-L1 in a CD8⁺ T-cell dependent manner in an Arid1a inactivation-driven genetic OCCC mouse model. Spatial analysis revealed that SLC38A2 knockdown in OCCC cells decreased the ratio of alanine distribution in tumor vs. non-tumor regions, which correlates with an increase of alanine in CD45⁺ cells. However, a limitation of the present study is that we did not directly observe an increase in alanine levels in CD8⁺ T cells due to a combination of technical limitations and biological reasons. Technically, the alanine signal in the spatial analysis is relatively low. Biologically, the abundance of the infiltrating CD8⁺ T cells is relatively low and particularly in the control untreated tumors, which prevented us from directly overlapping alanine and CD8⁺ T cells in the spatial analysis. Likewise, we were unable to detect N-acetylalanine in the spatial analysis to directly demonstrate a decrease in alanine utilization in tumor cells. Regardless, the decrease in the ratio between tumor cells and non-tumor cells for alanine signaling in SLC38A2 knockout OCCC cells supports the importance of SLC38A2 in regulating alanine uptake. The fact that two inhibitors of the SWI/SNF complex upregulates SLC38A2 in tumor cells cautions the implication of this class of inhibitors in the tumor immune microenvironment. A limitation of our study is the lack of specific SLC38A2 inhibitor. One challenge associated with inhibiting SLC38A2 will have to do with the potential compensatory mechanisms. As such, it would be advantageous to simultaneously inhibiting the function of SLC38A2 and boosting the function of SLC7A8. In addition to PD-1/PD-L1 inhibitors, anti-CTLA-4 or a combination can be considered for this purpose. In summary, our data indicate that targeting alanine uptake through suppressing SLC38A2 alone or in combination with immune checkpoint blockade represents an effective therapeutic strategy for ARID1A-mutant cancers.

Statement of Significance:

ARID1A mutations regulate expression of alanine transporters to control alanine distribution between cancer cells and the associated tumor microenvironment, which may be exploited therapeutically alone or in combination with immunotherapy.