Autophagy promotes immune evasion of pancreatic cancer by degrading MHC-I

Abstract

Immune evasion is a major obstacle for cancer treatment. Common mechanisms of evasion include impaired antigen presentation caused by mutations or loss of heterozygosity of the major histocompatibility complex class I (MHC-I), which has been implicated in resistance to immune checkpoint blockade (ICB) therapy^1-3. However, in pancreatic ductal adenocarcinoma (PDAC), which is resistant to most therapies including ICB⁴, mutations that cause loss of MHC-I are rarely found⁵ despite the frequent downregulation of MHC-I expression^6-8. Here we show that, in PDAC, MHC-I molecules are selectively targeted for lysosomal degradation by an autophagy-dependent mechanism that involves the autophagy cargo receptor NBR1. PDAC cells display reduced expression of MHC-I at the cell surface and instead demonstrate predominant localization within autophagosomes and lysosomes. Notably, inhibition of autophagy restores surface levels of MHC-I and leads to improved antigen presentation, enhanced anti-tumour T cell responses and reduced tumour growth in syngeneic host mice. Accordingly, the anti-tumour effects of autophagy inhibition are reversed by depleting CD8⁺ T cells or reducing surface expression of MHC-I. Inhibition of autophagy, either genetically or pharmacologically with chloroquine, synergizes with dual ICB therapy (anti-PD1 and anti-CTLA4 antibodies), and leads to an enhanced anti-tumour immune response. Our findings demonstrate a role for enhanced autophagy or lysosome function in immune evasion by selective targeting of MHC-I molecules for degradation, and provide a rationale for the combination of autophagy inhibition and dual ICB therapy as a therapeutic strategy against PDAC.

Extended Data Figure 1 |. Heterogeneous distribution of MHC-I in KRas mutant cancers.

a, Immuno-isolation of intact lysosomes from HPDE and PDAC cell lines showing absence of non-lysosome markers as indicated. EE, early endosome; mito, mitochondria; ER, endoplasmic reticulum; lyso, lysosome. b, High power images showing MHC-I positive / LAMP1 positive (arrowheads), MHC-I positive / LAMP1 negative (arrows) and MHC-I negative / LAMP1 positive (asterisk) puncta. Scale bar, 5 μm. c,d, Localization of MHC-I (green) relative to LAMP1 (red) positive lysosomes (BEAS-2B, n = 14; A549, n = 17; H441, n = 15; H358, n = 13; HCT116, n = 13) (c) or LC3B (red) positive autophagosomes (BEAS-2B, n = 18; A549, n = 17; H441, n = 20; H358, n = 12; HCT116, n = 15) (d) in the indicated cell lines. Graphs show quantification of percentage co-localization. Cell lines indicated in red show significantly increased co-localization relative to BEAS-2B cells while cell lines indicated in blue show a modest increase (H358) or no difference (HCT116). Data are mean ± s.d. (c,d). Scale bar, 20 μm and 10 μm (insert). e, Flow cytometry-based analysis of surface MHC-I (H-2) in murine normal pancreas (C57Bl/6) and murine PDAC cells grown as organoids. (Top) Isotype-subtracted geometric mean fluorescence intensity (MFI). Each dot represents different animals/lines (n = 4). Data are mean ± s.e.m. (Middle) Representative flow cytometry plots. (Bottom) Representative images of organoids. f, Immunofluorescent staining images from a patient in Fig. 1g showing intracellular localization of MHC-I (green) in CK19 positive (red) ducts. Scale, 20 μm. A representative of at least two independent experiments is shown in a,b,e. P values determined by unpaired two-tailed t-tests (c-e). For gel source data of a, see Supplementary Fig. 1.

Extended Data Figure 2 |. Autophagy and lysosome inhibition restores MHC-I levels and plasma membrane localization.

a, Immunofluorescence staining of MHC-I following ATG3 knockdown. Scale, 50 μm. b, Representative flow cytometry plots for PaTu8902 cells following ATG3 (related to Fig. 2b) and ATG7 (see also panel d) knockdown. Representative plots from Fig. 2b and Extended Data Fig. 2d are shown. c, Effect of ATG7 knockdown on MHC-I (HLA-A,B,C) expression in PaTu8902 cells. d, Flow cytometry-based quantification of plasma membrane MHC-I (HLA-A,B,C) levels following ATG7 knockdown (n = 9 replicates from 3 independent experiments). e, Immunofluorescence staining of MHC-I following ATG7 knockdown. Scale bar, 50 μm. f,g, Surface MHC-I levels upon Atg3 (f) or Atg7 (g) knockdown in mouse PDAC cells. (left) Knockdown efficiency was confirmed by immunoblots. (middle) Cell surface MHC-I (H-2K^b/D^b) levels measured by flow cytometry (n = 8 replicates from 2 independent experiments). (right) Representative flow cytometry plots are shown. h, Treatment of KP4 cells with 150 nM Bafilomycin A1 (BafA1), a lysosomal V-ATPase inhibitor, for the indicated times causes an increase in HLA-A,B levels. i, Flow cytometry-based quantification of plasma membrane (PM) MHC-I in the indicated cell lines following treatment with BafA1 for 16 hrs (n = 9 replicates from 3 independent experiments). j, Surface MHC-I (H-2) levels measured by flow cytometry. Murine PDAC organoids were established from Atg5^+/+ and Atg5^−/− KPC cells. n = 4 biological replicates. Data are representative of three independent experiments. (right) Representative flow cytometry plots. k, Effect of BafA1 treatment on the expression levels of antigen presentation machinery. l, Quantitative proteomics analysis of Panc1 cells that were treated with CQ (10 μM) for the indicated periods. n = 3 biological replicates. m, Effect of BafA1 treatment on expression levels of MHC-I in the indicated cell lines. Cell lines denoted in green show a significant change across all HLA isoforms following BafA1 treatment. n,o, Effect of ATG3 knockdown in H441 cells on total MHC-I (n) and plasma membrane MHC-I (o) as measured by flow cytometry-based quantification (n = 9 replicates from 3 independent experiments). A representative of at least two independent experiments is shown in a-c,e-h,j,k,m,n. Data are mean ± s.d. and P values determined by unpaired two-tailed t-tests (d,f,g,i,j,l,o). For gel source data of a,c,f,g,h,k,m,n, see Supplementary Fig. 1.

Extended Data Figure 3 |. Inhibition of macroautophagy, but not LAP/LANDO, restores MHC-I levels.

Knockdown of FIP200, ATG14, Atg13, and Ulk1, but not Rubicon, increased MHC-I levels in PDAC cells. a,d,g,i,k, Knockdown efficiency was confirmed by immunoblot (a,i,k) and qPCR (d,g). Data are mean ± s.d. from three biological replicates per group (d,g). a,e, Whole cell abundance of MHC-I was assessed by immunoblot. c, Immunofluorescence staining of MHC-I (green) and LAMP2 (red). Scale bar, 50 μm. b,f,h,j,l, Cell surface MHC-I levels were measured by flow cytometry (b,f, n = 9; h,j, n = 12; l, n = 16). Data are pooled from at least three independent experiments. Graphs are mean ± s.d. a-f, PaTu8902 cells (human). g-l, HY15549 cells (mouse). A representative of at least two independent experiments is shown in a,c,e,i,k. P values determined by unpaired two-tailed t-tests. For gel source data of a,e,i,k, see Supplementary Fig. 1.

Extended Data Figure 4 |. The UBA domain of NBR1 is required for interaction with MHC-I.

a, (Related to Fig. 2c) Proximity-dependent biotinylation catalyzed by the TurboID biotin ligase conjugated to the C-terminus of HLA-A and Flag (HLA-A-TrID). Following addition of biotin, TurboID catalyzes the formation of biotin-5’-AMP anhydride which enables covalent tagging of endogenous proteins with biotin within a few nanometers of the ligase. b, (Related to Fig. 2c) HLA-A-TurboID was stably expressed in KP4 cells. Cells were treated with 10 μM of exogenous biotin for 30 min. After labeling, cells were lysed and biotinylated proteins were enriched with streptavidin conjugated beads. Biotinylated proteins were detected using streptavidin-HRP (b) or with antibodies against the indicated proteins (refer to Fig. 2c). Asterisks indicates ligase self-biotinylation. c, (Related to Fig. 2e) Endogenous ubiquitylated proteins were affinity captured from PaTu8902 cells with UBQLN1 UBA conjugated beads. Treatment of affinity captured samples for 1 hr with purified Usp2-cc (+) to induce deubiquitylation leads to loss of ubiquitylation. d, (Related to Fig. 2f) PaTu8902 cells stably expressing WT NBR1 (GFP-NBR1, n = 19 fields) or lacking the UBA domain (GFP-NBR1 dUBA, n = 16 fields) were co-stained for endogenous LC3B. Graph shows quantification of percent colocalization. Centerline indicates the median and whiskers represent the minimum and maximum values. Scale, 20 μm (insert 10 μm). e, (Related to Fig. 2g) Effect of NBR1 knockdown on respective HLA-A, B, and C levels in PaTu8902 cells. Note that blotting images for NBR1 and Tubulin are the same ones as in Fig. 2g. f, Immunofluorescence staining of MHC-I following NBR1 knockdown. Scale, 50 μm. A representative of at least three independent experiments is shown in b,c,e,f. For gel source data of b,c,e, see Supplementary Fig. 1.

Extended Data Figure 5 |. Autophagy inhibition restores MHC-I expression, leading to enhanced anti-tumour T cell response in vitro.

a,b, Autophagy flux (a) and cell surface MHC-I levels (b) in PDAC cells measured by flow cytometry. Mouse PDAC cells expressing the GFP-LC3-RFP reporter and doxycycline (Dox)-inducible mTurquoise2-Atg4B^C74A were grown as organoids for 8 days and treated with Dox (1 μg/ml) for the indicated hours. a, Autophagy flux represented by GFP/RFP ratio. Note that increased GFP/RFP ratio indicates reduced autophagy flux. b, Cell surface MHC-I (H-2K^b) levels. Representative flow cytometry plots were shown. Data are mean ± s.d. n = 3 biological replicates. Data are representative of at least four independent experiments. c, Fold changes of respective molecules on the cell surface quantified by flow cytometry. HY15549 cells expressing Dox-inducible mTurquoise2-tagged Atg4B^C74A were grown as organoids for 8 days and treated ± Dox (1 μg/ml) for 72 hrs. Positive surface expression of each molecule was confirmed using respective isotype controls. Molecules found in immunological synapses are underlined. TFRC, transferrin receptor. Data are mean ± s.d. n = 4 biological replicates. Representative data from two independent experiments are shown. d-f, (Related to Fig. 3a,b) Mouse PDAC cells expressing OVA and carrying doxycycline (Dox)-inducible mTurquoise2-Atg4B^C74A were grown as organoids and treated ± Dox (1 μg/mL) for 96 hrs. d, Autophagy inhibition was confirmed by immunoblot. mTurquoise2-Atg4B^C74A or endogenous Atg4B were detected by anti-ATG4B antibody. e,f, Flow cytometry plots for H-2K^b (e) and H-2K^b-SIINFEKL (f). Representative plots from Fig. 3a,b are shown. Grey, isotype control. g, Representative flow cytometry plots of the OT-I cells co-cultured with mouse PDAC cells from Fig. 3c. h, (Related to Fig. 3c) Quantitative reverse transcription PCR (qRT-PCR) analysis of OT-I cells that were co-cultured with PDAC cells for 48 hrs. Data are mean ± s.d. n = 3 biological replicates. For g, h, Dox(+) or Dox(−) indicates that PDAC cells were grown ± Dox (1 μg/mL) before co-culture. Dox was not added in co-culture. A representative of at least three independent experiments is shown in d-g. P values determined by unpaired two-tailed t-tests (a,b,h). ****P < 0.0001. For gel source data of d, see Supplementary Fig. 1.

Extended Data Figure 6 |. Autophagy inhibition modulates anti-tumour immunity in both orthotopic tumours and liver metastasis.

a, Immunoblots showing autophagy inhibition in mSt-Atg4B^C74A expressing cells. Mouse PDAC cells carrying doxycycline (Dox)-inducible mStrawberry (mSt) or mSt-Atg4B^C74A (4B) were treated with Dox (1 μg/mL) for the indicated days. mSt or mSt-Atg4B^C74A was detected by anti-RFP antibody. A representative of two independent experiments is shown. b-j, (Related to Fig. 3e–h) Mouse PDAC cells shown in (a) were orthotopically transplanted into syngeneic (C57Bl/6) mice. HY15549 cells (mSt, n = 8; 4B, n = 7) and HY19636 cells (n = 8 per group) were injected. b, Study design. c, Images of tumours at end point. d-g, HY19636 tumour weight (d), cell surface MHC-I levels (e) and PD-L1 levels (f) on PDAC cells and (g) tumour-infiltrating CD8⁺ T cells measured by flow cytometry. h,i, Representative H&E staining (h) and immunofluorescent staining (i) of HY15549 tumours (mSt, n = 8; 4B, n = 7). Scale bars, 100 μm. j, Quantification of tumour-infiltrating immune cells by flow cytometry (HY15549, n = 8 and 7; HY19636, n = 8 per group). Gating strategies are shown in Extended Data Fig. 7k and Supplementary Table 2. Treg, T regulatory cells; MDSC, myeloid-derived suppressor cells; TAM, tumour-associated macrophages. k-o, Autophagy-inhibition by Atg7 knockdown elicits similar anti-tumour T cell responses. k, Immunoblots for Atg7, LC3, and β-actin in PDAC cells (HY15549) expressing shRNAs against GFP or Atg7. A representative of at least two independent experiments is shown. l-o, Mouse PDAC cells shown in (k) were orthotopically transplanted into syngeneic mice (n = 7 per group). Images of tumours harvested on day 22 (l). Tumour weight (m). Tumour-infiltrating immune cells as measured by flow cytometry (n). Correlation between CD8⁺ T cell frequency among CD45⁺ cells and tumour weight (o). p-t, (Related to Fig. 3i–l) Autophagy inhibition modulates anti-tumour immunity in metastatic tumours in the liver. Mouse PDAC cells (HY15549) carrying Dox-inducible mSt or 4B were injected into the spleen of syngeneic (C57Bl/6) mice that were pre-fed with Dox-containing diet (n = 4 per group). PDAC cells were pre-treated with Dox (1 μg/mL) for 7 days before injection. Study design (p). Images of the liver (q). Representative images of H&E staining (r) and immunofluorescent staining (s) (n = 4 per group). Scale bars, 200 μm (r) and 100 μm (s). Quantification of immune cells in the liver metastasis as measured by flow cytometry (t). Data are mean ± s.e.m. n indicates individual mice. Statistical differences were determined by unpaired two-tailed t-tests (d-g,j,m,n,t), and Pearson correlation analysis (o). For gel source data of a,k, see Supplementary Fig. 1.

Extended Data Figure 7 |. Tumour regression upon autophagy inhibition is rescued by depletion of CD8⁺ T cells or ablation of cell surface MHC-I.

a-c, (Related to Fig. 3m) HY15549 cells with doxycycline (Dox)-inducible mStrawberry (mSt) or mSt-Atg4B^C74A (4B) were orthotopically injected into C57Bl/6 mice and fed with Dox-containing diet starting on day 5, and then received i.p. injection of anti-CD8 or isotype control IgG (n = 7 per group). a, Study design. b, Images of tumours. c, Tumour-infiltrating leukocytes as quantified by flow cytometry. d-f, (Related to Fig. 3n,o) HY15549 cells with Dox-inducible mSt or 4B were orthotopically injected into C57Bl/6 mice (WT) or Batf3^−/− mice (KO) (n = 8, 4, 8, and 5). d, Study design. e, Images of tumours. f, Quantification of tumour-infiltrating leukocytes by flow cytometry. g-j, (Related to Fig. 3p–r) HY15549 cells carrying Dox-inducible mSt or 4B were stably transfected with lentiviral vectors expressing control shRNA (shScr, solid line) or shRNA against B2m (shB2m, dashed line). g, Cell surface MHC-I as measured by flow cytometry. Cells were treated with IFN-γ (200 unit/mL) for 24 hrs before flow cytometry analysis. Representative data from three independent experiments are shown. h-j, Cells shown in (g) (4 × 10⁴ cells) were orthotopically transplanted into syngeneic (C57Bl/6) mice that were pre-fed with doxycycline diet (n = 8 per group). h, Study design. i, Images of tumours. j, Tumour-infiltrating leukocytes as quantified by flow cytometry. k, Gating strategies for flow cytometry analysis of tumours used in this study. See also Supplementary Table 2. Data are mean ± s.e.m. n indicates individual mice. P values determined by unpaired two-tailed t-tests (c,f,j).

Extended Data Figure 8 |. Separation of PDAC cells with distinct autophagy flux using the GFP-LC3-RFP reporter.

Heterogeneity in basal autophagy flux was explored using mouse PDAC cells (HY15549) expressing the GFP-LC3-RFP reporter. a,b, HY15549 cells were grown as organoids or transplanted into C57Bl/6 mice to form orthotopic tumours. a, Autophagy flux, as represented by GFP/RFP ratio, was measured by flow cytometry. Atg5^−/− MEF with the GFP-LC3-RFP reporter (black) was used as a control. Representative flow plots from three independent experiments are shown. b, Representative fluorescent images of orthotopic tumours. Cells with high autophagy flux show GFP-LC3 puncta formation (inset, arrowhead) and a decrease in total GFP-fluorescent signals, displaying red appearance in the merged image. Scale bar, 100 μm. c-h, Mouse PDAC organoids were dissociated into single cells and sorted into autophagy-high (AThi) or -low (ATlo) cells according to the GPF/RFP ratio. c, Sorting strategies. d, Top KEGG pathways enriched in AThi cells compared to ATlo cells. Gene set enrichment analysis (GSEA) was performed using RNA-sequencing (RNA-seq) data from sorted AThi and ATlo cells (n = 2 and 3 biologically independent samples), showing enrichment of the autophagy-lysosome gene signatures in AThi cells as compared with the ATlo cells. FDR, false discovery rate; NES, normalized enrichment score; Nom., nominal. e,f, Relative mRNA expression of autophagy/lysosome-related genes in the respective populations sorted from pooled populations (e) or a single-cell derived clone (f). n = 3 technical replicates. Representative results from four (e) and two (f) independent sorting experiments are shown. g-i, Clonogenic potential of sorted AThi and ATlo cells (g,h) and PDAC cells with Dox-inducible Atg4B^C74A (AY6284) (i). Representative data from at least two independent experiments are shown. n = 4 (g) and n = 3 (i) per group. Data are mean ± s.d. (e-i). P values determined by unpaired two-tailed t-tests (g-i).

Extended Data Figure 9 |. Basal autophagy flux determines immunogenicity of PDAC cells.

Mouse PDAC cells (HY15549) expressing the GFP-LC3-RFP reporter were sorted into AThi and ATlo cells (Extended Data Fig. 8c) and injected into the pancreas (a-g) or the spleen (h-k) of C57Bl/6 mice (a-f, h-k) or nude mice (g). Cells were sorted from pooled populations except for (b). a,b, Tumour weight on day 21. AThi and ATlo cells were sorted from either pooled populations (a) (n = 9 and 10) or a single-cell derived clone (b) (n = 10 per group). c-e, Tumours shown in (a) were analyzed. Cell surface MHC-I levels on PDAC cells measured by flow cytometry (c). Correlation between MHC-I levels on PDAC cells and tumour weight (d). Representative images of H&E (left) and immunofluorescent staining (right) (e). Scale bars, 100 μm. f, Quantification of tumour-infiltrating immune cells by flow cytometry (n = 8 per group). Orthotopic tumours harvested on day 21 were analyzed. g, Orthotopic tumours in nude mice harvested on day 19 (n = 8 and 7). h-k, Liver metastasis model. Study design (h). Weight of livers (i) and cell surface MHC-I levels on PDAC cells measured by flow cytometry (j) on day 17 (n = 5 per group). Representative immunofluorescence images of livers harvested on day 9 (n = 3 per group)(k). Frozen sections were stained with anti-CD8a antibody and DAPI. In these merged images, cells with high autophagy flux appear as red, reflecting the relative loss of GFP-fluorescence and lower GFP/RFP ratio, while cells with low autophagy flux appear as yellow to green, reflecting high GFP/RFP ratio. In the enlarged images, CD8a⁺ cells were indicated by white arrow heads. Scale bars, 100 μm. For a, c-f, and i-k, experiments were performed at least twice and representative data of one experiment are shown. Data are mean ± s.e.m. (a-c,f,g,i,j). n indicates individual mice. P values determined by unpaired two-tailed t-tests (a-c,f,g,i,j), and Pearson correlation analysis (d).

Extended Data Figure 10 |. Autophagy inhibition synergizes with dual ICB.

a-c, Anti-PD1 antibody treatment did not affect tumour growth in either control or autophagy-inhibited tumours. Mice bearing orthotopic PDAC tumours (HY15549) carrying doxycycline (Dox)-inducible mStrawberry (mSt) or mS-Atg4B^C74A (4B) were treated with Dox beginning on day 5 and received either isotype control IgG or anti-PD1 antibody (n = 7, 8, 8 and 7 per group). Study design (a). Images (b) and weight (c) of tumours. d-g, (Related to Fig. 4a–d) Mice bearing orthotopic PDAC tumours (HY15549) expressing Dox-inducible mSt or 4B were treated with Dox beginning on day 5 and received either isotype control IgG or dual ICB (anti-PD1/CTLA4 antibodies) (n = 7 per group). Representative images of immunofluorescence staining (d) and H&E staining (e). Scale bars, 100 μm. Quantification of tumour-infiltrating immune cells by flow cytometry (f,g). h, Cell surface MHC-I (H-2K^b/D^b) levels measured by flow cytometry. Mouse PDAC cells were treated with chloroquine (CQ) or Bafilomycin A1 (BafA1) at the indicated concentrations for 48 hrs (n = 4). Mouse PDAC cells were grown in 2D culture (CQ) or as organoids (BafA1). Representative results from at least three independent experiments are shown. i-m, Mice bearing orthotopic PDAC tumours expressing the GFP-LC3-RFP reporter were treated with PBS or CQ beginning on day 5 (n = 7 vs 6 for HY15549 and n = 8 vs 8 for HY19636). Study design (i). Images of tumours (j). Cell surface MHC-I and PD-L1 levels on PDAC cells measured by flow cytometry (k). Tumour weight (l). Quantification of tumour-infiltrating immune cells by flow cytometry (m). n, Representative fluorescence images of tumours expressing the GFP-LC3-RFP reporter from Fig. 4j. Numerical values represent mean fluorescent intensity of each field. o, Quantification of tumour-infiltrating immune cells by flow cytometry (n = 8, 8, 8, and 5; left to right). Tumours in Fig. 4k were analyzed. Data are mean ± s.e.m. (c,f,g,k-m,o) or ± s.d. (h). n indicates individual mice (c,f,g,k-m,o) or biological replicates (h). Except for the orthotopic implantation of HY19636 cells (j-m), all experiments were performed at least twice and representative data of one experiment are shown. P values determined by unpaired two-tailed t-tests (c,f-h,k-m,o). ****P < 0.0001.

Fig. 1 |. MHC-I is enriched in lysosomes of PDAC cells and displays reduced cell surface expression.

a, Levels of MHC-I (HLA-A,B,C) in HPDE and human PDAC cell lines. b, Localization of MHC-I (green) relative to LAMP1 (red) positive lysosomes. Graph shows the percentage co-localization (n = 14–20 fields). Scale, 20 μm. c, Presence of MHC-I in immuno-isolated lysosomes. d, Accumulation of MHC-I in immuno-isolated lysosomes following treatment with E64d/Pepstatin A for 6 hrs. e, Localization of MHC-I (green) relative to LC3B (red) positive autophagosomes. Graph shows the percentage co-localization (n = 14–20 fields). Scale, 20 μm. f, Flow cytometry-based analysis of intracellular versus plasma membrane (PM) MHC-I levels. Graph shows higher intracellular MHC-I relative to plasma membrane MHC-I in PDAC cells (n = 9 replicates pooled from 3 independent experiments per cell line). Data are mean ± s.d. g, Intracellular localization of MHC-I (green) in CK19 positive (red) ducts from patient PDAC specimens. Graph shows the percentage of ducts showing intracellular MHC-I localization. Scale, 20 μm. A representative of at least two independent experiments is shown in a,c,d. For box-and-whisker plots (b,e), centerlines indicate median and whiskers represent minimum and maximum values. P values determined by unpaired two-tailed t-tests. See Source Data for exact n for b,e. For gel source data of a,c,d, see Supplementary Fig. 1.

Fig. 2 |. NBR1 promotes MHC-I trafficking to the lysosome through an autophagy dependent pathway.

a, Effect of ATG3 knockdown on HLA-A,B,C levels in PaTu8902 cells. b, Flow cytometry-based quantification of plasma membrane (PM) MHC-I (HLA-A,B,C) levels (PaTu8902, n = 9, 8, 9; KP4, n = 9 per group; data pooled from 3 independent experiments) following ATG3 knockdown. c, KP4 cells expressing HLA-A-TurboID-Flag (HLA-A-TrID) were labeled with 10 μM biotin for 30 min. Biotinylation of proteins was detected following streptavidin pull down. Asterisk denotes self-biotinylation of HLA-A-TrID. Graph shows enrichment for each receptor (n = 4 independent experiments). d, Localization of MHC-I (red) relative to GFP-NBR1 (green) in HPDE and PaTu8902 cells. Arrowheads show examples of co-localization. Scale, 20 μm. Graph shows quantification of co-localization (HPDE, n = 23 fields; PaTu8902, n = 20 fields). e, Endogenous ubiquitylated proteins were affinity captured from PaTu8902 cells with UBQLN1 UBA conjugated beads. Arrowheads indicate MHC-I polyubiquitylation. Treatment of affinity captured samples for 1 hr with purified Usp2-cc (+) to induce deubiquitylation leads to loss of MHC-I polyubiquitylation. Control proteins: LC3B (no ubiquitylation) and EGFR (mono-ubiquitylation). f, Endogenous MHC-I co-localizes with WT GFP-NBR1 but not GFP-NBR1 lacking its UBA domain (dUBA). Graph shows quantification of colocalization (GFP-NBR1; n = 17 fields; GFP-NBR1 dUBA; n = 14 fields). Scale, 20 μm (inset 10 μm). g, Effect of NBR1 knockdown on MHC-I levels. h, Flow cytometry-based quantification of PM MHC-I (n = 9 replicates from 3 independent experiments) following NBR1 knockdown. Scale bar, 50 μm (d,f). A representative of at least two independent experiments is shown in a,e,g. Data are mean ± s.d. (b,c,h). For box-and-whisker plots (d,f), centerlines indicate median and whiskers represent minimum and maximum values. P values determined by unpaired two-tailed t-tests. For gel source data of a,c,e,g, see Supplementary Fig. 1.

Fig. 3 |. Autophagy inhibition enhances anti-tumour T cell response.

a,b, Surface H-2K^b (a) and H-2K^b-SIINFEKL (b) measured by flow cytometry. Mouse PDAC cells expressing OVA and Dox-inducible mTurquoise2-Atg4B^C74A were grown as organoids and treated ± Dox (1 μg/mL) for 96 hrs (n = 4 per group). c,d, Co-culture of OT-I cells with HY19636 cells shown in a,b. After 48 hrs, OT-I proliferation was measured by CFSE dilution (c) (n = 4 per group) and PDAC cell viability was measured by Cell-Titer Glo (d) (n = 6 per group). e-r, HY15549 cells carrying Dox-inducible mStrawberry (mSt) or mSt-Atg4B^C74A (4B) were orthotopically (e-h, m-r) or intrasplenically (i-l) injected into syngeneic mice (C57BL/6). MHC-I and PD-L1 expression on PDAC cells (f,g,j,k,p) and tumour-infiltrating CD8⁺ T cells (h,l,n,q) quantified by flow cytometry. e-h, Orthotopic tumours harvested on day 20 (mSt, n = 8; 4B, n = 7). i-l, Livers harvested on day 15 (n = 4 per group). Weight of tumours (e) and livers (i). Flow cytometry analysis (f-h, j-l). m, Weight of tumours following treatment with isotype control IgG or neutralizing monoclonal antibody against CD8 (n = 7 per group). n,o, Tumours in wild type mice (WT) or Batf3^−/− mice (KO) (n = 8, 4, 8, and 5; left to right). Quantification of CD8⁺ T cells (n). Tumour weight (o). p-r, Tumours expressing control shRNA (Scr) or shRNA against B2m (n = 8 per group) harvested on day 20. Flow cytometry analysis (p,q). Tumour weight (r). Data are mean ± s.d. (a-d) or s.e.m. (e-r). n indicates biological replicates (a-d) or individual mice (e-r). For a-m, p-r, experiments were performed at least twice and representative data of one experiment are shown. P values determined by unpaired two-tailed t-tests.

Fig. 4 |. Autophagy inhibition sensitizes PDAC to dual ICB.

a-d, Mice bearing orthotopic tumors (HY15549) expressing Dox-inducible mSt or 4B received isotype control IgG or dual ICB (anti-PD1/CTLA4 monoclonal antibodies; mAb) (n = 7 per group). Study design (a). Images (b) and weight (c) of tumours. Quantification of tumour-infiltrating CD8⁺ T cells by flow cytometry (d). e-k, Mice bearing orthotopic tumours (HY15549) expressing GFP-LC3-RFP reporter received CQ and ICB (n = 8 per group). Study design (e). Images (f) and weight (g) of tumours. No macroscopic tumour was identified in three animals receiving CQ + dual ICB (#3, 5, and 6). h, Response rates from two independent experiments. Response is defined as more than 80% reduction in tumour weight as compared with control tumours (PBS + IgG). i, Representative H&E images of the pancreas undergoing tumour regression (# 6). White dashed line indicates tumour remnants. Scales, 250 and 100 μm. j, Autophagy flux represented by GFP/RFP ratio per 20x field (n = 49, 39, 47 and 54; left to right). Increased GFP/RFP ratio indicates reduced autophagy flux. k, Quantification of tumour-infiltrating CD8⁺ T cells by flow cytometry (n = 8, 8, 8, and 5; left to right). l, In PDAC cells, surface MHC-I is downregulated by active degradation through the autophagy/lysosome system, contributing to the primary resistance to ICB. NBR1 binds to MHC-I, facilitating its trafficking to autophagosomes (left). Inhibition of autophagy or the lysosome restores surface MHC-I expression, leading to enhanced anti-tumour T cell immunity and improved response to ICB (right). Data are mean ± s.e.m. (c,d,g,k) or s.d. (j). n indicates individual mice (c,d,g,k) or individual 20x fields (j). All experiments were performed twice and representative data of one experiment are shown. P values determined by unpaired two-tailed t-tests.