ARIH1 signaling promotes anti-tumor immunity by targeting PD-L1 for proteasomal degradation

Abstract

Cancer expression of PD-L1 suppresses anti-tumor immunity. PD-L1 has emerged as a remarkable therapeutic target. However, the regulation of PD-L1 degradation is not understood. Here, we identify several compounds as inducers of PD-L1 degradation using a high-throughput drug screen. We find EGFR inhibitors promote PD-L1 ubiquitination and proteasomal degradation following GSK3α-mediated phosphorylation of Ser279/Ser283. We identify ARIH1 as the E3 ubiquitin ligase responsible for targeting PD-L1 to degradation. Overexpression of ARIH1 suppresses tumor growth and promotes cytotoxic T cell activation in wild-type, but not in immunocompromised mice, highlighting the role of ARIH1 in anti-tumor immunity. Moreover, combining EGFR inhibitor ES-072 with anti-CTLA4 immunotherapy results in an additive effect on both tumor growth and cytotoxic T cell activation. Our results delineate a mechanism of PD-L1 degradation and cancer escape from immunity via EGFR-GSK3α-ARIH1 signaling and suggest GSK3α and ARIH1 might be potential drug targets to boost anti-tumor immunity and enhance immunotherapies.

Fig. 1. EGFR inhibitors were screened to reduce membrane PD-L1 levels.

a High-throughput screening of 2125 FDA-approved drugs or drug candidates. U937 cells were incubated with IFNγ (100 ng/mL) for 48 h, treated with the drugs at 10 μM for 12 h. Ruxolitinib (Rux) was used as a positive control. The hit compounds that induced the decrease of PD-L1 levels are shown in blue. The depth of blue represents decreased level of PD-L1. The heatmap represents the targeted pathways obtained from the high-throughput screening, based upon decreased membrane PD-L1 level detected by flow cytometry. b Immunoblotting of PD-L1 in U937 cells treated with ES-072 (ES) or AZD9291 (AZD) at indicated concentrations for 12 h. ES and AZD are EGFR inhibitors. c, d Median fluorescence intensity (MFI) (c) and relative quantification (d) of PD-L1 in U937 cells treated with 10 μM ES-072 or 10 μM AZD9291 for 12 h. Data represent means ± SEM, n = 18, 6 independent repeats, ****P < 0.0001. e–g Immunoblotting (e, f) and flow cytometry (g) analysis of PD-L1 levels in U937 cells treated with 10 μM ES-072 or 10 μM AZD9291 for the indicated times. h, i Immunoblottings (h) of PD-L1 and β-TrCP, relative quantification (i) of PD-L1 MFI in H1975 cells transfected with non-targeting siRNA (si-CTRL) or β-TrCP siRNAs (si-β-TrCP) and treated with or without 10 μM ES-072/AZD9291 for 12 h. Data represent means ± SEM, n = 9, 3 independent repeats, NS: no significant; ****P < 0.0001. j–l Immunoblotting (j) and flow cytometry analysis (k) with relative quantification (l) of PD-L1 in H1975 cells treated with 10 μM ES-072 or AZD9291, and/or 5 μM LY2090314 (LY) for 12 h. LY is a GSK3 inhibitor. Data represent means ± SEM, n = 9, 3 independent repeats, ****P < 0.0001. Source data are provided as a Source Data file.

Fig. 2. Phosphorylation on S279/S283 impairs the stability of PD-L1.

a Mapping PD-L1 phosphorylation sites following ES-072 treatment. HEK293T cells were transfected with Flag-PD-L1 and treated with 10 μM ES-072 for 48 h. Immunoprecipitated PD-L1 was analyzed by mass spectrometry following phosphopeptide enrichment. Peptide ionization data corresponding to Ser279/283 are shown. b Schematic diagram of phosphorylated sites on PD-L1. Full-length PD-L1 was separated into an extracellular domain (ECD) and intracellular domain (ICD). SP: signal peptide; TM: transmembrane domain. c Schematic diagram of phosphorylated sites, S279 and S283 in the ICD of PD-L1. Positions of phosphorylated sites were labeled in red. This image was created by the first author. d Immunoblots of PD-L1 (anti-Flag) in HEK293T cells transfected with Flag-PD-L1 (WT) or Flag-PD-L1 (mutants) following treatment with 25 ng/mL EGF and/or 10 μM ES-072 for 48 h, 2SA represents S279A/S283A. e Immunoblots of PD-L1 (anti-Flag) in HEK293T cells transfected with Flag-PD-L1 (WT/2SA) following treatment with 20 μg/mL cycloheximide (CHX) for indicated times. f Immunoblots of H1975 cell lysates following ES-072 treatment for 2 h, at indicated doses. g HEK293T cells were transfected with Flag-PD-L1 (WT/2SA) following treatment with 10 μM ES-072 for 2 h; PD-L1 was immunoprecipitated with anti-Flag and immunoblotted with indicated antibodies. Flag-tagged empty vector (Flag-EV) was transfected as a negative control. h Representative images of p-PD-L1 (Ser279) and PD-L1 immunohistochemistry (IHC) staining from EGFR wild-type vs. mutant human alveolar adenocarcinoma specimens. Scale bars represent 50 μm. i, j Quantification of IHC analysis for p-PD-L1 (Ser279) (i) and PD-L1 (j) in h. Data represent means ± SEM, n = 7 (i), n = 5 (j), **P < 0.01, P = 0.0013 (i); P = 0.002 (j). k Immunoblottings of p-PD-L1 (Ser279) and PD-L1 in EGFR wild-type (n = 4) vs. mutant (n = 4) human lung adenocarcinoma specimens. Source data are provided as a Source Data file.

Fig. 3. GSK3α-mediated phosphorylation of PD-L1 promotes PD-L1 degradation.

a Immunoblots of PD-L1 (anti-Flag) in HEK293T cells transfected with Flag-PD-L1 (WT) or Flag-PD-L1 (mutants) following treatment with 25 ng/mL EGF and/or 10 μM ES-072 for 48 h, 2SA represents T180A/S184A. b HEK293T cells were transfected with Flag-PD-L1 and treated with ES-072 (10 μM) for 48 h. Proteins that co-immunoprecipitated (Co-IP) with PD-L1 were analyzed by mass spectrometry. Kinases and kinase-related proteins are shown in heatmap. c Co-IP analysis for the interaction of GSK3α/GSK3β and PD-L1 in HEK293T cells transfected with Flag-PD-L1 and treated with or without 10 μM ES-072 for 12 h, Flag-tagged empty vector (Flag-EV) was transfected as a negative control. d Proximity ligation assay (PLA) analysis for the interaction of GSK3α and PD-L1 in HEK293T cells treated as c. PLA signals are shown in red and the nuclei in blue; scale bar, 20 μm. Quantification for the mean area (MA) of PD-L1/GSK3α PLA speckles is indicated by scattergram. Data represent means ± SEM, n = 50, ****P < 0.0001. e Immunoblots of p-GSK3α, GSK3α, p-GSK3β, GSK3β, p-AKT, AKT, p-EGFR, and EGFR in H1975 cells treated with 10 μM ES-072 for indicated times. f In vitro GSK3α kinase assay was performed in the presence or absence of ATP. The phosphorylation of PD-L1 peptides (PD-L1-WT/2SA) was detected by dot blot with anti-p-PD-L1 (Ser279) antibody, p-PD-L1 peptides were synthesized as a positive control. g In vitro GSK3α kinase assay was performed in HEK293T cells transfected with Flag-PD-L1 and GSK3α-HA. Total cell lysates were immunoprecipitated with anti-Flag or anti-HA. The phosphorylation of PD-L1 by GSK3α was detected using an anti-p-PD-L1 (Ser279) antibody. h Immunoblots of p-PD-L1, PD-L1, and GSK3α in H1975 cells transfected with GSK3α-siRNAs. i Immunoblots of PD-L1 and GSK3α in U937 cells transfected with GSK3α-siRNAs and treated with or without 10 μM ES-072 for 24 h. j Immunoblots of PD-L1 (anti-Flag) and GSK3α in HEK293T cells transfected with Flag-PD-L1 following treatment with GSK3α-siRNA and treated with or without 10 μM ES-072 for 24 h. Source data are provided as a Source Data file.

Fig. 4. ARIH1 mediates PD-L1 ubiquitination and degradation.

a As in Fig. 3b, except ubiquitination-related proteins are shown in the heatmap. b Co-IP analysis for the interaction of ARIH1 and endogenous PD-L1 in HEK293T cells transfected with HA-ARIH1; HA-tagged empty vector (HA-EV) was transfected as a negative control. c Recombinant PD-L1 was purified using a GST pull-down assay and incubated with HEK293T lysates, which were transfected with HA-ARIH1. The interaction between ARIH1 and PD-L1 was detected by Immunoblot assay. d Co-IP analysis for the interaction of ARIH1 and PD-L1 in HEK293T cells transfected with HA-ARIH1 and Flag-PD-L1, treated with or without 10 μM ES-072 for 24 h; Flag-tagged empty vector (Flag-EV) was transfected as a negative control. e, f PD-L1 level in 20 μg/mL cycloheximide (CHX)-treated HEK293T cells transfected with or without HA-ARIH1 (e) and ARIH1-siRNA (f). g, h Immunoblots of PD-L1 and ARIH1 (HA) in H1975 cells transfected with ARIH1-HA (g) or ARIH1-siRNAs (h). i HEK293T cells were transfected with Flag-PD-L1. Co-IP analysis for the interaction of K48-ubiquitin and PD-L1 in HEK293T cells transfected with ARIH1-siRNAs or HA-ARIH1 and treated with MG132 (10 μM, a proteasome inhibitor) for 6 h. j Immunoblots of PD-L1 and ARIH1 in U937 cells transfected with ARIH1-siRNAs, following treatment with 10 μM ES-072 for 24 h. k HEK293T cells were transfected with HA-K48-ubiquitin and Flag-PD-L1. Co-IP analysis for the interaction of K48-ubiquitin and PD-L1 in HEK293T cells transfected with ARIH1-siRNAs and treated with or without 10 μM ES-072 for 24 h. l Recombinant PD-L1 and ARIH1 were purified in transfected HEK293T cells, respectively. An in vitro ubiquitination assay of PD-L1 was performed with purified GST-UBA1 (E1), His-UBCH7 (E2) in the presence or absence of ubiquitin or HA-ARIH1 (E3). Source data are provided as a Source Data file.

Fig. 5. GSK3α-mediated phosphorylation of PD-L1 promotes PD-L1-ARIH1 interaction and ARIH1-induced degradation.

a Co-IP analysis for the interaction of K48-ubiquitin, ARIH1 (HA), and PD-L1 in HEK293T cells transfected with HA-ARIH1 and Flag-PD-L1 (WT, S279A, S283A, or 2SA), treated with MG132 (10 μM) for 6 h; HA-tagged empty vector (HA-EV) was transfected as a negative control. b Co-IP analysis for the interaction of ARIH1 and PD-L1 in HEK293T cells transfected with HA-ARIH1 and Flag-PD-L1 (WT or 2SA), treated with 10 μM ES-072 for 24 h, Flag-tagged empty vector (Flag-EV) was transfected as a negative control. c HEK293T cells were transfected with Flag-PD-L1. Co-IP analysis for the interaction of ubiquitin and PD-L1 in HEK293T cells transfected with ARIH1-siRNAs or HA-GSK3α and treated with MG132 (10 μM) for 6 h; non-targeting siRNA (si-CTRL) was transfected as a negative control. d Immunoblots of PD-L1 and ARIH1 (HA) in H1975 cells transfected with HA-ARIH1, treated with or without 5 μM LY for 6 h. e, f MFI (e) and relative quantification (f) of PD-L1 in HA-ARIH1-overexpressed H1975 cells, treated with or without 5 μM LY for 6 h. Data represent means ± SEM, n = 9, 3 independent repeats, ****P < 0.0001. g HEK293T cells were transfected with Flag-PD-L1. Co-IP analysis for the interaction of K48-ubiquitin, ARIH1 (HA), and PD-L1 in HEK293T cells transfected with HA-ARIH1, treated with or without 5 μM LY for 12 h and treated with MG132 (10 μM) for 6 h. Source data are provided as a Source Data file.

Fig. 6. ARIH1 promotes anti-tumor immunity via PD-L1 degradation.

a Immunoblot of PD-L1 and ARIH1 (HA) in H1975 cells transfected with HA-ARIH1 (WT or Y392C); HA-tagged empty vector (HA-EV) was transfected as a negative control. b Co-IP analysis for the interaction of K48-ubiquitin, ARIH1 (HA), and PD-L1 in HEK293T cells transfected with Flag-PD-L1 and HA-ARIH1 (WT or Y392C) in the presence of 10 μM MG132 for 6 h. c 4T1 cells were infected with an empty vector (CTRL) or two different ARIH1-overexpressing lentiviral preparations (OE#1 and OE#2). ARIH1 and PD-L1 levels were determined by immunoblotting. d–f Tumor growth (d, e) of CTRL (n = 10) and ARIH1-OE cells (n = 11) in BALB/c mice and final tumor weights (f). Data represent means ± SEM, ***P < 0.001 (P = 0.0001), ****P < 0.0001. g, h Flow cytometry analysis for the tumor levels of CD8⁺ T cells (g) and CD8⁺GzmB⁺ T cells (h). Data represent means ± SEM, CTRL (n = 10) and ARIH1-OE (n = 5), ***P < 0.001 (P = 0.0006), **P < 0.01 (P = 0.0046). i–k 4T1 tumor xenograft growth in BALB/c mice (I, j) and final tumor weights (k) following treatment with ES-072 and/or anti-CTLA4 (n = 6-7). Vehicle = sodium carboxymethyl (CMC-Na). ES = ES-072. Data represent means ± SEM, *P < 0.05 (P = 0.01), **P < 0.01 (P = 0.0013), ***P < 0.001 (P = 0.0001), ****P < 0.0001. l, m Flow cytometry analysis for the tumor levels of CD8⁺ T cells (l) (n = 4–7) and CD8⁺GzmB⁺ T cells (m) (n = 4–7). Data represent means ± SEM. l **P < 0.01 (P = 0.0017), ***P < 0.001 (P = 0.0001), ****P < 0.0001. m *P < 0.05 (P = 0.0286), **P < 0.01 (P = 0.0012), ***P < 0.001 (P = 0.00099). n Immunobloting of the indicated tumor lysates. Source data are provided as a Source Data file.

Fig. 7. Schematic model for GSK3α-promoted and ARIH1-mediated PD-L1 degradation.

GSK3α phosphorylates PD-L1 at Ser279 and Ser283. This phosphorylation promotes the binding of PD-L1 with ARIH1, leading to PD-L1 ubiquitination and proteasomal degradation. This image was created by the first author.