Fig 2: Depletion of Atad3a evokes a favorable tumor immune microenvironment and improves the efficacy of ICI.a–h BALB/c mice were inoculated orthotopically with 5 × 104 4T1 cells transfected with control shRNA (shControl) or shRNA for Atad3a (shAtad3a#1 and shAtad3a#2). a, b The endpoint tumor images (a) and volume (b) of tumors formed by control and Atad3a-knockdown cells in BALB/c mice (n = 6, one-way ANOVA). c Left, IHC staining of Atad3a and PD-L1 on serial sections of tumors formed by control and Atad3a-knockdown cells. Scale bars, 50 µm. Right, IHC score of Atad3a in control and Atad3a-knockdown tumors (n = 6 fields, t-test). d IHC score of PD-L1 in control and Atad3a-knockdown tumors (n = 6 fields, t-test). e Quantification of the percentage of tumor-infiltrating CD8+ T cells in tumors formed by control and Atad3a-knockdown cells by flow cytometry (n = 5, t-test). f Quantification of the percentages of tumor-infiltrating IFN?+CD8+ T cells and IFN?+CD4+ T cells by flow cytometry (n = 5, t-test). g Ratio of CD8+ cytotoxic T lymphocytes to CD4+CD25+Foxp3+ Treg cells (n = 5, t-test). h Quantification of the percentages of PD-1+TIM-3+CD8+ T cells (left) and PD-1+TIM-3+CD4+ T cells (right) by flow cytometry (n = 5, t-test). i–m BALB/c mice were inoculated orthotopically with 5 × 104 4T1 cells transfected with control shRNA (shControl) or shRNA specific for Atad3a (shAtad3a), Pink1 (shPink1) or both (shAtad3a + shPink1). i, Left, the endpoint images of tumors formed by control, Atad3a-knockdown, Pink1-knockdown or Atad3a and Pink1 double knockdown 4T1 cells in BALB/c mice. Right, immunoblot of Atad3a and Pink1 in these 4T1 cells. j The volume of tumors mentioned above (n = 6, one-way ANOVA). k Quantification of the percentage of tumor-infiltrating CD8+ T cells by flow cytometry (n = 5, one-way ANOVA). l Quantification of the percentage of tumor-infiltrating IFN?+CD8+ T cells by flow cytometry (n = 5, one-way ANOVA). m Quantification of the percentage of PD-1+TIM-3+CD8+ T cells by flow cytometry (n = 5, one-way ANOVA). n–s 4T1 tumors formed by control and Atad3a-knockdown cells were established orthotopically in BALB/c mice and received vehicle, anti-PD-L1 antibody (PD-L1 mAb), paclitaxel (PTX) or combined anti-PD-L1 antibody with paclitaxel treatment (PD-L1 mAb + PTX). IgG2b and saline were used as controls. n Experimental protocol. o, p The endpoint tumor images (o) and the volume (p) of tumors (n = 6, one-way ANOVA). q–s Quantification of the percentages of tumor-infiltrating CD8+ T cells (q), IFN?+CD8+ T cells (r) and PD-1+TIM-3+CD8+ T cells (s) in 4T1 tumors formed by control and Atad3a-knockdown cells received treatments as described above, determined by flow cytometry (n = 5, one-way ANOVA). t Schematic model. Patients with PD-L1-positive TNBC could be divided into two groups based on ATAD3A expression. Patients with ATAD3A-high tumors might respond more poorly to ICIs plus paclitaxel therapy, and inhibition of ATAD3A is required to improve clinical outcome. Patients with ATAD3A-low tumors might benefit significantly from ICIs plus paclitaxel combination therapy. See also Supplementary information, Figs. S7–S9.

Fig 3: Involvement of ATAD3A-PINK1 mitophagy in the distribution and degradation of PD-L1.a Left, immunostaining of PD-L1 (green) and TOM20-labeled mitochondria (red) in BT549 human TNBC cells with or without ATAD3A-knockdown (shATAD3A#1 and shATAD3A#2). Scale bars, 20 µm and 2 µm (inset). Right, the percentage of PD-L1 co-localized with TOM20 (n = 5 fields, t-test). b Immunoblot of PD-L1 in the cytoplasm and mitochondria of control and ATAD3A-knockdown BT549 cells. TOM20 and Tubulin were used as mitochondria and cytoplasm protein controls, respectively. Cyto, cytoplasm; mito, mitochondria. c Flow cytometry (left) and quantification (right) of surface PD-L1 in control and ATAD3A-knockdown BT549 cells (n = 3, one-way ANOVA). d Venn diagram depicting overlapped genes for the interaction protein of ATAD3A set (BioGRID, RP5-832C2.1), the protein localization to mitochondrion set (GOBP 0070585) and the intrinsic component of mitochondrial membrane (GOCC 0098573). e Immunoblot of PINK1 in control and ATAD3A-knockdown BT549 cells. f Immunoblot of PD-L1 and PINK1 in HEK293T cells overexpressing PD-L1 (OE-PD-L1) and control cells (OE-Control), assessed after immunoprecipitation with immunoglobulin G (IgG) or antibody to PINK1. g Protein direct interaction analysis of the intracellular domain of PD-L1 (ICD) and PINK1 in vitro. Purified Flag-labeled full-length PINK1 was incubated with Biotin-labeled PD-L1 ICD domain, followed by streptavidin pull-down and immunoblot. h Schematic diagram of Flag-labeled full-length (FL) and truncated mutants with indicated domains (amino acids 1–155, amino acids 156–320, amino acids 321–509, amino acids 510–581) of PINK1. MTS, mitochondrial targeting sequence; N-lobe, kinase domain N; C-lobe, kinase domain C; CTD, C-terminal domain. i Protein direct interaction analysis of the intracellular domain of PD-L1 (ICD) and truncated PINK1 mutants in vitro. Purified Flag-labeled full-length and truncated PINK1 were incubated with Biotin-labeled PD-L1 ICD domain, followed by streptavidin pull-down and immunoblot. The estimated size of PINK1-4 (amino acids 510–581) which did not express in HEK293T cells was labeled with asterisk. j Immunoblot of PD-L1 in the cytoplasm and mitochondria of MDA-MB-231 cells with or without PINK1-knockdown (shPINK1#1 and shPINK1#2). TOM20 and Tubulin were used as mitochondria and cytoplasm protein controls. Cyto, cytoplasm; mito, mitochondria. k Left, co-localization of PD-L1 (green) and TOM20 (red) in control, ATAD3A knockdown, PINK1 knockdown or ATAD3A and PINK1 double knockdown BT549 cells. Scale bars, 20 µm and 2 µm (inset). Right, the percentage of PD-L1 co-localized with TOM20 (n = 5 fields, one-way ANOVA). l Immunoblot of PD-L1 in the cytoplasm and mitochondria of control, ATAD3A-knockdown, PINK1-knockdown or ATAD3A and PINK1 double knockdown BT549 cells. m Immunoblot of indicated proteins in PD-L1-transfected HEK293T cells with or without PINK1 overexpression. n Immunoblot of PD-L1 in control and PINK1-knockdown (shPINK1#1 and shPINK1#2) BT549 cells. o Immunoblot of PD-L1 in BT549 cells transfected with control shRNA or shATAD3A (shATAD3A#1 and shATAD3A#2). p Immunoblot of total PD-L1 in control, ATAD3A-knockdown, PINK1-knockdown or ATAD3A and PINK1 double knockdown BT549 cells. q Immunoblot of PD-L1 in control and ATAD3A-knockdown BT549 cells treated with 20 µM CHX for indicated times. h, hours. r Quantification of PD-L1 intensity in immunoblot in control and ATAD3A-knockdown BT549 cells. s Immunoblot of PD-L1 in control and ATAD3A-knockdown MDA-MB-231 cells incubated with 20 nM BafA1 for indicated times. h, hours. Data are representative of at least two independent experiments and are shown as means ± SD. See also Supplementary information, Figs. S3 and S4.

Fig 4: ATAD3A prevents PD-L1 distribution to mitochondria.a Venn diagram depicting the overlapped common genes in the mitochondrion component set (GOCC 0005739), mitochondrion organization set (GOBP 0007005), and immune response set (GOBP 0006955). b Kaplan-Meier analysis of the OS of glioblastoma patients who received anti-PD-1 antibody treatment with high or low ATAD3A mRNA expression (n = 29). Data were derived from GSE121810 cohort. Cutoff of ATAD3A expression was determined by the median (Kaplan-Meier method and log-rank test). c Kaplan-Meier analysis of the OS of TNBC patients with high or low ATAD3A transcriptional level. Data were derived from TCGA database. Cutoff of ATAD3A expression was determined by the median (Kaplan-Meier method and log-rank test). d Kaplan-Meier analysis of the OS (left) and PFS (right) of TNBC patients with high or low ATAD3A expression from the multicentric clinical cohort (Kaplan-Meier method and log-rank test). These patients underwent radical mastectomy or modified mastectomy and received the standard chemotherapy. e, f Left, co-localization of PD-L1 (green) and TOM20-labeled mitochondria (red) in TNBC specimens with high (e) or low (f) ATAD3A expression pre- or post-chemotherapy including taxol in immunotherapy cohort. Arrowheads, co-localization. Scale bars, 20 µm and 2 µm (inset). Right, the percentage of PD-L1 co-localized with TOM20 (n = 9 fields, t-test). g The correlation between ATAD3A levels and the number of tumor-infiltrating CD8+ T cells in TNBC tumors (n = 67). IHC score was quantified by Image-Pro Plus software 6.0, three to five fields per section (Pearson correlation analysis). h Representative images (left) and IHC score (right) of ATAD3A expression in tumors from non-responders and responders before anti-PD-1 antibody plus nab-paclitaxel therapy in immunotherapy cohort. Scale bars, 50 µm (n = 30, Mann-Whitney U test). i Patients with different responses in the immunotherapy cohort. Subjects were colored by their responses to combination therapy. CR, complete response; PR, partial response; SD, stable disease; PD, progressive disease. y-axis was determined by IHC score of ATAD3A by logarithm to the base 2. Cutoff of ATAD3A expression was determined by the median (n = 30, Fisher’s exact test). j Kaplan-Meier analysis of the progression-free survival of TNBC patients with high or low ATAD3A expression receiving combination therapy (n = 16). Data are derived from the immunotherapy cohort (Kaplan-Meier method and log-rank test). See also Supplementary information, Fig. S2.