Abstract

   The constrained cross-talk between myeloid cells and T cells in the
   tumor immune microenvironment (TIME) restricts cancer immunotherapy
   efficacy, whereas the underlying mechanism remains elusive. Parkin, an
   E3 ubiquitin ligase renowned for mitochondrial quality control, has
   emerged as a regulator of immune response. Here, we show that both
   systemic and macrophage-specific ablations of Parkin in mice lead to
   attenuated tumor progression and prolonged mouse survival. By
   single-cell RNA-seq and flow cytometry, we demonstrate that Parkin
   deficiency reshapes the TIME through activating both innate and
   adaptive immunities to control tumor progression and recurrence.
   Mechanistically, Parkin activation by AMP-activated protein kinase
   rather than PTEN-induced kinase 1 mediated major histocompatibility
   complex I down-regulation on macrophages via Autophagy related
   5–dependent autophagy. Furthermore, Parkin deletion synergizes with
   immune checkpoint blockade treatment and Park2^−/− signature aids in
   predicting the prognosis of patients with solid tumor. Our findings
   uncover Parkin’s involvement in suppressing macrophage antigen
   presentation for coordinating the cross-talk between macrophages and T
   cells.
     __________________________________________________________________

   Parkin suppresses antigen presentation in tumor-associated macrophage
   via degradation of MHC-I through autophagy.

INTRODUCTION

   Cancer immunotherapy has revolutionized the paradigms of conventional
   therapies. Despite durable curative effects achieved in immune
   checkpoint blockade (ICB) treatment for certain malignancies
   ([74]1–[75]4), few patients with solid tumor benefit from it due to
   side effect, poor efficacy, and high recurrence ([76]5, [77]6).
   Multidimensional profiling on the tumor immune microenvironment (TIME)
   of patients receiving ICB treatment through transcriptomic and
   proteomic analyses has revealed that exhausted T cells and
   immunosuppressive myeloid cells in the TIME largely account for
   patients’ low responsiveness to immunotherapies ([78]7–[79]10).

   Growing evidences support that the cross-talk between myeloid cells and
   T cells is crucial for tumor immunosurveillance and immune escape in
   the TIME ([80]11). As the largest population of myeloid cells,
   tumor-associated macrophages (TAMs) have been reported to capture and
   present tumor antigens to T cells, accompanied by the expression of
   soluble factors and costimulatory signals ([81]12, [82]13). However,
   TAMs in solid tumors usually exhibit immunosuppressive features. On one
   hand, TAMs are endowed with high expression levels of immune
   checkpoints [e.g., Programmed Death-Ligand 1 (PD-L1) and B7 homolog 4
   (B7-H4)], chemokines and cytokines [e.g., C-C motif chemokine ligand 2
   (CCL2), interleukin-6 (IL-6), IL-10, and transforming growth factor–β],
   and immunosuppressive molecules (e.g., arginine 1 and inducible nitric
   oxide synthase) ([83]14, [84]15). On the other hand, immunosuppressive
   TAMs reportedly form an “infiltration-rejection” tumor microenvironment
   due to their abundant infiltration, which severely attenuates the
   activation of tumor-specific T cells and the efficacy of
   immunotherapies ([85]16, [86]17). Hence, there persists a compelling
   need to find the mechanisms that link innate immunity to adaptive
   immunity, which regulate tumor immune escape.

   PARK2 (encoding Parkin), whose mutations are the most frequent causes
   of recessive Parkinson’s disease (PD), encodes a structurally
   autoinhibited E3 ubiquitin ligase ([87]18, [88]19). It is well
   established that Parkin can be phosphorylated by PTEN-induced kinase 1
   (PINK1) and recruited to mitochondria to ubiquitinate mitochondrial
   proteins, which ultimately removes damaged mitochondrial components by
   mitophagy ([89]20, [90]21). Beyond its vital role in mitochondrial
   quality control for cellular fitness, emerging research highlights
   Parkin’s involvement in inflammation: (i) Patients with Parkin
   mutations and Park2^−/− mice show elevated inflammatory factors such as
   IL-6 and tumor necrosis factor–α (TNF-α) ([91]22, [92]23); (ii)
   AMP-activated protein kinase (AMPK), which senses cellular adenosine
   5′-diphosphate (ADP)/adenosine 5′-triphosphate (ATP) ratio, has been
   reported to activate Parkin independent of PINK1 and modulate
   inflammation ([93]24, [94]25); and (iii) Parkin purportedly inhibits
   stimulator of interferon (IFN) genes–induced inflammation and
   mitochondrial antigen presentation ([95]23, [96]26), suggesting that it
   orchestrates both innate and adaptive immunities under inflammatory
   conditions. In terms of tumorigenesis, Parkin has been characterized as
   a potential tumor suppressor, with frequently diminished expression and
   mutations in various tumors ([97]24, [98]27–[99]29). Considering that
   the immunosuppressive milieu of TIME is accompanied by the fluctuations
   in ATP production, mitochondrial status, and complex inflammatory
   stimulation ([100]30–[101]32), it is intriguing to hypothesize that
   Parkin might regulate immune cell–mediated immunity in this context.
   Thus, whether and how Parkin modulates immune cells during tumor
   progression would be particularly attractive issues.

   Here, we highlight Parkin as a negative immune regulator, which
   restrains the cross-talk between macrophages and T cells to underpin
   tumor immune evasion. Through various in vivo murine tumor progression
   models, we found that both systemic Parkin deficiency and
   macrophage-specific ablation of Parkin could inhibit the progression of
   multiple cancers. Single-cell transcriptomic and flow cytometry
   analyses profiled distinct antitumor macrophage and effector T cell
   subsets in the TIME of Park2^−/− mice, which were further validated by
   adoptive transfer assays. Subsequent functional experiments showed that
   Parkin deficiency facilitates both innate and adaptive immune responses
   to remodulate the TIME and control tumor development and relapse.
   Mechanistically, Parkin suppresses major histocompatibility complex I
   (MHC-I) expression on TAMs through an unconventional AMPK-dependent
   pathway via the Autophagy related 5 (ATG5)–dependent autophagy
   activity, which was further validated by mice with conditional ATG5
   knockout (KO) in macrophages. We found that Park2^−/− signature is
   associated with improved prognosis in patients with solid tumor and
   targeting Parkin sensitizes ICB treatment, reinforcing Parkin as a
   promising target for cancer immunotherapy.

RESULTS

Parkin-deficient mice are resistant to tumor progression

   To determine Parkin’s impact in regulating antitumor immunity, we used
   a gene-deficient mouse strain lacking Park2 to establish subcutaneous
   tumor models. Park2^−/− (Parkin deficiency) mice showed stronger
   resistance to tumor progression than wild-type (WT) littermates in a
   syngeneic colon carcinoma MC38 model, in terms of tumor volume
   ([102]Fig. 1A and fig. S1, A and B), tumor weight (fig. S1C), and
   survival ([103]Fig. 1B). Similar results were observed when mice were
   inoculated with Hepa1-6 hepatocellular carcinoma cells ([104]Fig. 1, C
   and D, and fig. S1D), CT2A glioma tumor cells (fig. S1, E and F), or
   E0771 breast cancer cells (fig. S1G), suggesting that Parkin-deficient
   mice exhibit resistance to the development of multiple subcutaneous
   solid tumors.

Fig. 1. Loss of Parkin in mice leads to reduced tumor burden in multiple
cancers.

   [105]Fig. 1.
   [106]Open in a new tab

   (A and B) Tumor volume (A) and survival (B) of WT (n = 9) and Park2^−/−
   (n = 9) mice bearing subcutaneous MC38 tumors. (C and D) Tumor volume
   (C) and survival (D) of WT (n = 16) and Park2^−/− (n = 17) mice bearing
   subcutaneous Hepa1-6 tumors. (E) Survival analysis of WT (n = 8) and
   Park2^−/− (n = 10) mice in an in situ model of hepatocellular
   carcinoma. (F) Representative images of hepatocellular carcinoma
   tumors. (G) Plot showing the overall survival (OS) of WT (n = 12) and
   Park2^−/− (n = 10) mice intracranially inoculated with CT2A glioma
   cells. GBM, glioblastoma multiforme. (H) Representative histology of
   murine brains with intracranial tumors (tumor indicated by arrows).
   Images are representative of at least three images for one indicated
   animal. Scale bars, 2 cm. (I) t-distributed stochastic neighbor
   embedding (t-SNE) plot of 40,091 cells from WT or Park2^−/− mice
   bearing MC38 tumors (pooled cells of two mice per group) analyzed by
   single-cell RNA sequencing (scRNA-seq). DCs, dendritic cells. (J)
   Proportions of major cell types in (I). (K) Percentages of T cells,
   CD8^+ T cells, and CD4^+ T cells in subcutaneous MC38 tumors in WT
   (n = 5) or Park2^−/− (n = 4) mice. (L and M) IL-2 (L) and IFN-γ (M) in
   tumor interstitial fluids of WT (n = 7) and Park2^−/− (n = 8) mice. (N
   to P) Tumor volumes of WT and Park2^−/− MC38–bearing mice treated with
   (N) phosphate-buffered saline (PBS) and (O) anti-CD4 and (P) anti-CD8
   antibodies. The numbers of mice are indicated. The results are depicted
   as the means ± SEM, *P < 0.05, **P < 0.01, and ***P < 0.001. ns, not
   significant.

   In vivo liver metastasis model of colon carcinoma also showed that
   Parkin deficiency extended mouse survival (fig. S1H). We then
   established orthotopic solid tumor models to verify the effects of
   Parkin deficiency on tumor development more physiologically. Orthotopic
   injection of Hepa1-6 cells into liver capsules led to favorable
   survival in Park2^−/− mice compared with WT mice ([107]Fig. 1E). The
   livers of tumor-inoculated WT mice contained tumors with larger
   quantities and volumes compared with those of Park2^−/− mice, which had
   only small tumors and even tended to be normal ([108]Fig. 1F and fig.
   S2A). Hematoxylin and eosin (H&E)–stained liver sections from Park2^−/−
   mice revealed less lesions, confirmed by pale nuclei and pink cytoplasm
   (fig. S2, B and C). In parallel, in a murine glioma model, Parkin
   deficiency markedly impeded intracranial tumor growth and improved
   mouse survival ([109]Fig. 1G), in line with the H&E staining that
   Parkin deficiency alleviated intracranial tumor load ([110]Fig. 1H and
   fig. S2D). Collectively, these results indicate that loss of Parkin in
   mice prohibits the progression of multiple cancers.

T cells are essential for the superior antitumor effects in Park2^−/− mice

   To investigate the mechanisms responsible for Parkin deletion–mediated
   antitumor effects, we performed single-cell transcriptome sequencing
   [single-cell RNA sequencing (scRNA-seq)] on tumor-infiltrating CD45^+
   cells sorted from MC38 tumors in WT or Park2^−/− mice. A total of
   40,091 immune cells were obtained and partitioned into eight immune
   cell subsets (fig. S3A), which overlapped between Park2^−/− and WT mice
   in the t-distributed stochastic neighbor embedding (t-SNE) plot
   ([111]Fig. 1I). The TIME of Park2^−/− mice was significantly distinct,
   with the increased proportions of natural killer (NK) cells, T cells,
   monocytes, granulocytes, and macrophages, and the decreased proportions
   of B cells and mast cells compared with that of WT mice ([112]Fig. 1J
   and fig. S3B). Concordantly, the significantly increased proportions of
   the tumor-infiltrating T cells (TILs) in the TIME of Park2^−/− mice
   were confirmed by flow cytometry ([113]Fig. 1K and fig. S4A), while the
   proportions of other immune cell subsets remained similar between WT
   and KO mice (fig. S4, B to H).

   As the ultimate effector cells of antitumor immunity, CD4^+ and CD8^+ T
   cells directly recognize and kill tumor cells, thus controlling tumor
   progression ([114]33). Notably, both proportions of CD8^+ and CD4^+ T
   cells were significantly up-regulated in the TIME of Park2^−/− mice
   ([115]Fig. 1K), consistent with immunohistochemical staining of CD8
   (fig. S4, I and J) and CD4 (fig. S4, K and L) in tumor sections. We
   also observed significantly up-regulated IL-2 and IFN-γ in tumor
   interstitial fluids of Park2^−/− mice ([116]Fig. 1, L and M),
   suggesting enhanced effector functions of T cells in the TIME of
   Park2^−/− mice. We next set out to determine the contribution of T
   cells to Parkin deficiency–induced tumor regression. Depleting CD8^+ T
   cells completely abrogated the superior tumor suppression in Park2^−/−
   tumor–inoculated mice ([117]Fig. 1, N and O), while deletion of CD4^+ T
   cells partially rescued the decrease in tumor burden of Park2^−/− mice
   ([118]Fig. 1, N and P). Collectively, these data indicate that T cells
   are necessary for Parkin deficiency–mediated antitumor immunity.

Parkin deficiency indirectly activates TILs

   To further probe the functional status of TILs, we clustered the T cell
   population and obtained 13 subclusters in the scRNA-seq data ([119]Fig.
   2A and fig. S5A). We found that the percentages of early activated
   CD8^+ T cells (CCL8^hi), effector CD8^+ T cells [NCR^hi (Natural
   Cytotoxicity Triggering Receptor 1) and NKG7^hi (Natural Killer Cell
   Granule Protein 7)], and IFN-γ^+ CD8^+ T cells were significantly
   increased, while the percentage of regulatory T cells (T[regs]) was
   significantly decreased in the Park2^−/− group ([120]Fig. 2B). We also
   found abundant differentially expressed genes of T cells between WT and
   Park2^−/− groups. Genes related to T cell recruitment, activation, and
   tumor killing, including Ccl8, CD8a, Prf1, Klrc2, Nkg7, Ifng, and
   Il18rap, were significantly up-regulated in the Park2^−/− group. In
   contrast, genes related to T[reg] differentiation, recruitment, and
   immune suppression, such as Tnfrsf4, Ctla4, Foxp3, Il2a, and Il10, were
   significantly down-regulated in the Park2^−/− group ([121]Fig. 2C),
   substantiating a more tumoricidal phenotype. To further characterize
   the evolution and activation status of T cells, we performed trajectory
   analysis ([122]Fig. 2D and fig. S5, B and C) coupled with exhaustion
   and inhibitory scoring ([123]Fig. 2E and fig. S5D) on TILs. We found
   that T cells diverged into three developmental trajectories (left,
   upper, and upper-middle-lower) ([124]Fig. 2F). With the progression of
   the pseudotemporal, on both left and upper-middle-lower trajectories,
   the proportion of T cells in the KO group gradually decreased,
   accompanied by increasing trends of exhaustion and inhibition scores;
   while on the upper pathway, the proportion of T cells in the KO group
   gradually increased, albeit to no significant increase in the
   exhaustion and inhibition scores. We also observed markedly enriched
   cytotoxic, cytolytic, and cytokine effector signatures of
   tumor-infiltrating CD8^+ T cells in the Park2^−/− group compared with
   WT group ([125]Fig. 2G). In parallel, we examined the expression levels
   of IFN-γ, CD107a, granzyme B, and TNF-α in the in vitro reactivated
   TILs sorted from WT or Park2^−/− mice and observed marked elevation in
   both tumor-infiltrating CD8^+ ([126]Fig. 2, H and I, and fig. S6, A and
   B) and CD4^+ T cells (fig. S6, C to F) from Park2^−/− mice.
   Collectively, Parkin deficiency could reshape TILs into an enhanced
   activation and effector phenotype at transcriptional and protein
   levels.

Fig. 2. Parkin deficiency leads to greater antitumor activity of TILs.

   [127]Fig. 2.
   [128]Open in a new tab

   (A) Annotated Uniform Manifold Approximation and Projection (UMAP) plot
   of T cells from tumors in WT and Park2^−/− mice. T[H]1, T helper cell
   1; T[H]17, T helper cell 17; IST, Interferon-response CD4^+ T cell. (B)
   Percentages of early activated_CD8, effector_CD8, Ifng_CD8, and T[regs]
   in WT and Park2^−/− mice. (C) Volcano plot of differentially expressed
   genes (edgeR) of T cells from Park2^−/− and WT mice (thresholds: at
   least 1.2 fold change, adjusted P < 0.05). (D) Trajectory analysis of T
   cells in WT and Park2^−/− mice. (E) Expression of genes and a signature
   associated with T cell exhaustion differ in the three trajectory
   branches. (F) Relative proportions of T cells from WT and Park2^−/−
   mice in different trajectory branches. (G) Enriched signatures of
   cytotoxic, cytolytic, and cytokine effector CD8^+ T cells from WT and
   Park2^−/− mice. (H and I) Expression levels of IFN-γ (H) and CD107a (I)
   in TILs by fluorescence-activated cell sorting (FACS) analysis. T cells
   were isolated from MC38 tumors in WT (n = 4) or Park2^−/− (n = 6) mice
   and stimulated with phorbol 12-myristate 13-acetate and ionomycin. (J)
   Experiment design. sc, subcutaneously; iv, intravenously. (K and L)
   Subcutaneous tumor volume (K) and OS (L) of Rag1^−/− mice inoculated
   with MC38 cells and transplanted with WT or Park2^−/− splenic CD8^+ T
   cells from tumor-bearing mice. The numbers of mice are indicated. (M
   and N) Subcutaneous tumor volume (M) and OS (N) of Rag1^−/− mice
   inoculated with MC38 cells and transplanted with WT or Park2^−/−
   splenic CD8^+ T cells from naïve mice. The numbers of mice are
   indicated. (O and P) Expression levels of the IFN-γ in splenic CD8^+ T
   cells from tumor-bearing (O) or naïve mice (P) (n = 6 per group). The
   results are depicted as the means ± SEM, *P < 0.05 and ***P < 0.001.

   We then performed CD8^+ T cell adoptive transfer assays to verify the
   in vivo effects of superior activated T cells in Parkin-deficient mice
   ([129]Fig. 2J). Transfer of splenic CD8^+ T cells from tumor-bearing
   Park2^−/− mice into MC38-inoculated Rag1^−/− mice significantly
   attenuated tumor growth and prolonged survival ([130]Fig. 2, K and L),
   prompting us to explore whether Parkin directly regulates naïve T cells
   for their enhanced activation in tumors. To address this, we
   transferred splenic CD8^+ T cells from naïve WT and Park2^−/− mice into
   MC38-inoculated Rag1^−/− mice but observed no significant difference in
   tumor progression of recipient mice ([131]Fig. 2, M and N). We also
   observed comparable expression levels of T cell activation molecules,
   CD25, CD44, CD62L, CD69, and PD-1 between splenic T cells sorted from
   naïve WT and Park2^−/− mice (fig. S6, G and H), indicating that Parkin
   deficiency is likely to influence T cell activation via T
   cell–extrinsic effects. Notably, we noticed that splenic CD8^+ T cells
   from tumor-inoculated Park2^−/− mice ([132]Fig. 2O) rather than naïve
   Park2^−/− mice ([133]Fig. 2P) secreted more IFN-γ in vitro than WT
   counterparts, in line with enhanced T cell activation in the TIME from
   Park2^−/− mice ([134]Figs. 1M and [135]2, G to I), implying that Parkin
   deficiency indirectly affects T cell–mediated antitumor effects by
   regulating some “intermediate” cells in the context of tumor
   challenging.

Macrophages are profoundly activated to cross-talk with T cells in
Parkin-deficient TIME

   The TIME is an equilibrium where various immune cell subsets interface
   with each other to regulate immunosurveillance or immune escape
   ([136]34). As the largest population in the TIME, myeloid cells play
   pivotal roles in immunomodulation ([137]35). Considering that Parkin
   might indirectly influence T cell activation through myeloid cells, we
   extracted myeloid cells from the scRNA-seq results and obtained five
   subpopulations, among which macrophages accounted for the largest
   proportion ([138]Fig. 3A). Using the CellChat algorithm, we found that
   Parkin deficiency enhanced the intercellular communication between T
   cells (both CD4^+ and CD8^+ T cells) and myeloid cells (fig. S7, A and
   B), with macrophages consistently being the strongest communication
   counterparts with T cells ([139]Fig. 3, B and C, and fig. S7, C and D).

Fig. 3. Parkin deficiency enhances macrophage function and T cell activation.

   [140]Fig. 3.
   [141]Open in a new tab

   (A) Annotated UMAP plot showing scRNA-seq analysis of myeloid cells
   from tumors in WT and Park2^−/− mice. (B and C) Differential strength
   of interactions between CD8^+ T cells and myeloid cells in WT and
   Park2^−/− mice. (D) Heatmap of the expression levels of selected genes
   in TAMs of WT and Park2^−/− mice bearing MC38 tumors in scRNA-seq
   dataset. (E) Tumor volume of WT and Park2^−/− mice treated with
   anti–colony-stimulating factor 1 receptor (anti-CSF1R) antibodies or
   left untreated after inoculation of MC38 cells. The numbers of mice are
   indicated. (F) Tumor volume of control (Ctrl; n = 12) and
   Park2^fl/fl-Lyz2-Cre^+/− (n = 13) mice bearing subcutaneous MC38
   tumors. CKO, conditional KO. (G) Percentages of T cells, CD8^+ T cells
   in subcutaneous MC38 tumors in control (n = 5), or
   Park2^fl/fl-Lyz2-Cre^+/− (n = 5) mice. (H) Experiment design. (I) Tumor
   volume of WT mice inoculated with Hepa1-6 cells and transplanted with
   WT or Park2^−/− bone marrow–derived macrophages (BMDMs). The numbers of
   mice are indicated. (J) Transcripts from macrophages in tumors from WT
   and Park2^−/− mice in scRNA-seq dataset were assessed for enrichment of
   signatures of costimulation, antigen presentation, inflammation, and
   M1-type macrophages. (K) Bubble heatmap showing the interactions of
   selected ligand-receptor pairs between TAMs from WT or Park2^−/− mice
   and T cell subpopulations based on scRNA-seq analysis of the MC38
   model. The results are depicted as the means ± SEM, **P < 0.01 and
   ***P < 0.001.

   Given that Parkin might regulate antitumor effects of T cells via
   macrophages, we then asked whether Park2^−/− TAMs acquire a distinctive
   antitumor phenotype. Differential expression gene (DEG) analysis showed
   that Park2^−/− TAMs significantly up-regulated genes in antigen
   presentation (e.g., H2-K1), proinflammatory polarization (e.g., Il1β
   and Tnf), and chemokines (e.g., Ccl4, Ccl5, and Ccl8) and
   down-regulated genes in protumor polarization (e.g., Mgl1 and
   Mrc1)–related pathways ([142]Fig. 3D). When macrophages were depleted
   by anti–colony-stimulating factor 1 receptor (anti-CSF1R) antibodies,
   the difference in tumor volume ([143]Fig. 3E) and survival (fig. S7E)
   between WT and Park2^−/− tumor–bearing mice disappeared. To probe
   whether Parkin expressed in macrophages induces tumor progression, we
   bred Park2^fl/fl-Lyz-Cre^+/− mice whose Park2 was conditionally KO in
   macrophages and determined low Parkin’s expression levels in
   macrophages (fig. S7F). In the MC38 model, specific deletion of Parkin
   in macrophages impeded tumor growth ([144]Fig. 3F) and tumor weight
   (fig. S7, G and H) and facilitated T cell infiltration in the TIME
   ([145]Fig. 3G and fig. S7, I and J), which provides a strong support
   for our previous findings in Park2^−/− mice ([146]Fig. 1, A, B, and K,
   and fig. S1, A to C). Enhanced antitumor effects of Park2^−/−
   macrophages in vivo were also confirmed by an adoptive transfer assay
   ([147]Fig. 3H). We found that the transfer of Park2^−/− bone
   marrow–derived macrophages (BMDMs) from naïve WT mice into
   tumor-inoculated WT mice significantly inhibited tumor growth
   ([148]Fig. 3I) and prolonged survival (fig. S7K) in recipient mice,
   suggesting that Park2^−/− macrophages are critical for the enhanced
   antitumor activity. These data collectively illustrate that Parkin
   expressed in macrophages induces tumor progression and targeting Parkin
   markedly impedes cancer progression.

   We next sought to characterize the interaction between TAMs and T cells
   in the TIME of Park2^−/− mice. First, Park2^−/− TAMs scored
   significantly higher than WT TAMs in these signatures as inflammation,
   costimulation, antigen presentation, and proinflammatory polarization
   ([149]Fig. 3J). Second, Reactome pathway enrichment analysis of
   Park2^−/− TAMs showed up-regulated pathways including innate immune
   system and adaptive immune response (fig. S8A), indicating that
   Park2^−/− TAMs are prone to an activated state and favorable to
   reinvigorate adaptive immune response. Last, intercellular
   ligand–receptor analysis showed that Parkin deficiency enhanced the
   interaction between macrophages and T cells through up-regulating
   costimulatory, antigen-presenting, and adhesion-related surface
   molecules ([150]Fig. 3K). In addition, we focused on the effector_CD8^+
   T cells and early activated_CD8^+ T cell subpopulations that were
   enriched in the KO group and analyzed their intercellular interactions
   with macrophages. We found that Parkin deficiency significantly
   enhanced their interactions, confirmed by up-regulated cytokine ligand
   receptors such as IFN-γ and type II IFN receptors, cell adhesion ligand
   receptors such as intercellular adhesion molecule–1 (ICAM1), and
   chemokine ligand receptors such as CXCL16 and CXCR6 (fig. S8B).
   Collectively, these results suggest that Parkin deficiency profoundly
   reprograms macrophages to reinvigorate adaptive immunity.

Parkin deficiency facilitates macrophage antigen presentation to prime
antigen-specific T cells

   To investigate the mechanism by which Parkin regulates TAMs, we
   clustered the macrophage population in the scRNA-seq data and obtained
   10 subpopulations (fig. S9A), which reportedly exhibit antitumor or
   protumor effects, respectively ([151]36). The antitumor TAMs were
   classified into three subclusters (macro_CD80, macro_Ccl12, and
   macro_C1qc), among which the proportions of macro_CD80 and macro_C1qc
   clusters increased in the Park2^−/− group (fig. S9B). We next analyzed
   the enriched pathways shared by these two clusters. Kyoto Encyclopedia
   of Genes and Genomes (KEGG) pathway enrichment analysis showed that the
   shared pathways were “TNF signaling pathway,” “phagosome,” “nuclear
   factor κB (NF-κB) signaling pathway,” “cytokine-cytokine receptor
   interaction,” and “antigen processing and presentation,” with “antigen
   presentation”–related pathways being consistently enriched in Gene
   Ontology (GO) pathway enrichment analysis ([152]Fig. 4A). Computation
   of immune enrichment scoring on macro_CD80 and macro_C1qc macrophage
   clusters revealed significantly increased proinflammatory and antigen
   presentation signatures and pathway-related gene expression in the
   Park2^−/− group compared with WT group ([153]Fig. 4, B to D).

Fig. 4. Parkin suppresses antigen presentation of macrophages.

   [154]Fig. 4.
   [155]Open in a new tab

   (A) Enrichment analysis of up-regulated genes in Park2^−/− macrophages
   from macro_CD80 and macro_C1qc macrophage subgroups. GO_BP, Gene
   Ontology: Biological process; GO_CC, Gene Ontology: Cellular component;
   GP_MF, Gene Ontology: Molecular function. TAP, TAP protein is a
   heterodimeric peptide transporter consisting of the subunits TAP1 and
   TAP2. (B) ECDF plots of enrichment score for proinflammation and MHC-I
   antigen presentation signatures in WT (black) or Park2^−/− (red)
   macro_CD80 and macro_C1qc macrophages. (C) Dot plot of selected genes
   from M1 macrophage signatures. (D) Dot plot of selected genes from
   antigen presentation signature. (E) Representative images and
   proportions of OT-I T cells stimulated with BMDMs from (n = 3 per
   group). (F) MHC-I expression of TAMs in subcutaneous MC38 tumors (n = 4
   per group). (G and H) Representative images and proportions of OT-I T
   cells stimulated with TAMs for 72 hours (n = 3 per group) (G) or
   control and Park2^fl/fl-Lyz2-Cre^+/− mice (n = 5 per group) (H). (I)
   Representative images and proportions of IFN-γ^+ OT-I T cells after
   cocultured with BMDMs (n = 6 per group). (J) Secreted IFN-γ of OT-I T
   cells after cocultured with TAMs for 72 hours (n = 5 per group). (K)
   Heatmap of the expression levels of genes related to IFN-γ production
   of in MC38 tumors from WT and Park2^−/− mice in bulk RNA-seq dataset.
   (L) Tumor volume of WT and Park2^−/− MC38–bearing mice with IFN-γ
   blockade. The numbers of mice are indicated. (M) Productive clonality
   in WT or Park2^−/− TILs (n = 3 per group). (N) ECDF plot of enrichment
   score for memory T cell signature in WT (black) or Park2^−/− (red) T
   cells. (O) Dot plot of selected genes in memory T cell signature. (P)
   Tumor growth of WT (n = 9) or tumor-free Park2^−/− (n = 7) mice
   rechallenged with fivefold numbers of Hepa1-6 cells. The results are
   depicted as the means ± SEM, *P < 0.05 and ***P < 0.001.

   As the critical sentinels of immune system, macrophages can
   phagocytose, process, and cross-present exogenous antigens to
   antigen-specific T cells in the context of MHC-I ([156]37, [157]38).
   Parkin deficiency hardly affects macrophage phagocytosis as WT or
   Park2^−/− BMDMs phagocytosed tumor cells at comparable levels (fig.
   S10, A and B). In terms of cross-presentation, we found that the MHC-I
   signaling pathway network between macrophages and CD8^+ T cells was
   significantly up-regulated in the Park2^−/− group (fig. S10C).
   Considering increased MHC-I expression on Park2^−/− BMDMs when
   cocultured with MC38 and L1210 (fig. S10, D and E), we conducted in
   vitro antigen presentation experiments and found that ovalbumin
   (OVA)–impulsed Park2^−/− BMDMs presented more SIINFEKL (fig. S10F) and
   displayed a significantly stronger ability to promote OT-I T cell
   proliferation than WT BMDMs ([158]Fig. 4E). Similarly, we observed
   increased MHC-I expression on TAMs isolated from Park2^−/− mice at
   protein levels ([159]Fig. 4F) and stronger T cell priming ability of ex
   vivo isolated Park2^−/− TAMs than those from WT mice ([160]Fig. 4G), in
   line with the findings of the TAMs from Park2^fl/fl-Lyz2-Cre^+/− mice
   ([161]Fig. 4H and fig. S10G). Besides increased T cell priming, we
   found that these OT-I T cells primed by Park2^−/− ([162]Fig. 4I) or
   Park2^fl/fl-Lyz2-Cre^+/− TAMs secreted more IFN-γ than the control
   group ([163]Fig. 4J and fig. S10H), in line with elevated expression
   levels of IFN-γ production pathway–related genes in Park2^−/− immune
   cells ([164]Fig. 4K) and increased IFN-γ in Park2^−/− tumor
   interstitial fluids ([165]Fig. 1M). In vivo, IFN-γ is indispensable for
   Parkin deficiency–mediated antitumor effects, as IFN-γ blockade in the
   MC38 model almost completely abolished the difference in tumor growth
   between WT and Park2^−/− mice ([166]Fig. 4L).

   Antigen-presenting cells loaded with tumor antigens could prime
   antigen-specific T cells for clone expansion, and afterward, the
   proliferated antigen-specific T cells used T cell receptors (TCRs) to
   recognize, kill tumors, and even sustain long-term immune memory
   ([167]39). To address this, we performed TCR sequencing (TCR-seq) on
   TILs sorted from tumors of WT and Park2^−/− mice. We observed an
   increased clonality along with a reduced diversity of TCRs from
   Park2^−/− TILs ([168]Fig. 4M and fig. S10, I and J), supporting that T
   cells primed by Park2^−/− TAMs are prone to expand antigen-specific TCR
   clones. Moreover, immune enrichment scoring of scRNA-seq data showed
   that T cells from Park2^−/− tumor–inoculated mice acquired an enhanced
   memory signature ([169]Fig. 4, N and O). When we rechallenged fivefold
   numbers of tumor cells in the Park2^−/− mice that had recovered after
   primary tumor inoculation, all the rechallenged tumors were markedly
   eradicated in Park2^−/− convalescent mice ([170]Fig. 4P). Together,
   these results indicate that Parkin influences the effector and memory
   functions of T cells through regulating macrophage antigen
   presentation.

AMPK-dependent activation of Parkin reduces MHC-I expression on macrophage
surface via autophagy

   Parkin is autoinhibited with barely E3 ligase catalytic activity.
   Canonically, Parkin can be phosphorylated at S65 site of the
   ubiquitin-like domain by PINK1, resulting in its activation and
   translocation to defective mitochondria for mitophagy ([171]20,
   [172]21). We therefore examined whether Parkin regulates macrophage
   antigen presentation through PINK1-dependent activation and mitophagy
   in a subcutaneous MC38 model. Unexpectedly, the tumor progression and
   survival in Pink1^−/− mice were comparable with those in WT mice
   ([173]Fig. 5, A and B), suggesting that Parkin regulates macrophage
   antigen presentation through a PINK1-independent mechanism. To
   investigate the molecular mechanism of Parkin in antigen presentation,
   we constructed macrophage cell lines by overexpressing a set of Parkin
   variants in Raw264.7 cells and examined their MHC-I expression levels
   on plasma membrane (fig. S11, A to C). First, we found that
   overexpressing full-length Parkin significantly down-regulated both the
   surface and total cellular levels of MHC-I on macrophages ([174]Fig. 5,
   C to E), in agreement with our above findings that Parkin deficiency
   facilitates MHC-I expression ([175]Fig. 4F and fig. S10, D and E).
   Meanwhile, overexpressing Parkin catalytic dead mutant C430S (akin to
   C431S in human) ([176]18) barely affected MHC-I levels, indicating that
   Parkin suppressed antigen presentation through its ubiquitination
   activity ([177]Fig. 5F). However, overexpressing Parkin mutant S65A,
   which blocks the phosphorylation and activation by PINK1, still
   significantly down-regulated MHC-I expression such as WT Parkin
   ([178]Fig. 5F), consistent with the observation that PINK1 deficient
   mice failed to recapitulate Parkin deficiency–mediated antitumor
   effects in vivo ([179]Fig. 5, A and B).

Fig. 5. AMPK-dependent Parkin activation suppresses MHC-I presentation via
autophagy.

   [180]Fig. 5.
   [181]Open in a new tab

   (A and B) Tumor growth (A) and OS (B) of WT and Pink1^−/− mice bearing
   subcutaneous MC38 tumors (n = 5 per group). (C and D) Representative
   image (C) and surface levels of MHC-I (D) on control and Parkin
   full-length Raw264.7 cells (n = 6 per group). Comp-FL2-A, Compensated
   Signal Fluorescence Channel 2 Area; PE-A, Phycoerythrin Area Parameter.
   (E) Total cellular levels of Parkin and MHC-I on control and Parkin
   full-length Raw264.7 cells. GAPDH, glyceraldehyde-3-phosphate
   dehydrogenase. (F) Surface levels of MHC-I on control, Parkin
   full-length, Parkin S65A, S108A, and C430S mutant Raw264.7 cells (n = 9
   per group). (G) Expression and activity of the AMPK signaling in immune
   subpopulations. (H) Surface levels of MHC-I on Parkin full-length,
   Parkin S65A, S108A, and C430S mutant Raw264.7 cells treated without
   (NC) or with inhibitors of AMPK (AMPKi) (n = 3 per group). (I and J)
   mRNA expression levels of H2-K1 and H2-Q1 on control and Parkin
   full-length Raw264.7 cells (n = 3 per group). (K) Surface levels of
   MHC-I on control and Parkin full-length Raw264.7 cells treated without
   or with inhibitors of autophagy [CQ and bafilomycin A1 (BafA1)] (n = 6
   per group). (L and M) Western blot of MHC-I expression in control and
   Parkin full-length Raw264.7 cells treated with BafA1. (N) Tumor growth
   of WT and Atg5^fl/fl-Lyz2-Cre^+/− mice bearing subcutaneous MC38 tumors
   (n = 5 per group). (O) Representative image and percentages of T cells
   in MC38 tumors (n = 5 per group). (P) MHC-I expression of TAMs in MC38
   tumors from control and Atg5^fl/fl-Lyz2-Cre^+/− mice (n = 5 per group).
   (Q) Representative images and proportions of proliferating OT-I T cells
   stimulated with TAMs for 72 hours (h) (n = 4 per group). The results
   are depicted as the means ± SEM, *P < 0.05, **P < 0.01, and
   ***P < 0.001.

   The AMPK/ULK1 (Unc-51 Like Autophagy Activating Kinase 1) axis has been
   reported to phosphorylate Parkin’s Activation Critical Threshold domain
   (Parkin’s ACT) domain at S108 site for its activation ([182]25). We
   found that overexpressing Parkin mutant S108A rectified Parkin-induced
   MHC-I down-regulation ([183]Fig. 5F), suggesting that the AMPK/ULK1
   axis activates Parkin to decrease MHC-I expression. AMPK is rapidly
   activated following any stress that triggers a lowering of
   intracellular ADP/ATP ratio, which initiates ULK1 phosphorylation and
   the downstream components of autophagy cascade ([184]40, [185]41).
   Considering the findings that the activation of autophagy promotes
   MHC-I internalization and restricts antigen presentation ([186]42), we
   speculated whether AMPK/ULK1 activates Parkin to decrease MHC-I
   expression through elevated autophagy. Single-sample gene set
   enrichment analysis (ssGSEA) demonstrated that TAMs exhibited the
   strongest activity of AMPK pathway in the TIME ([187]Fig. 5G). Further,
   applying AMPK inhibitor on the raw cells overexpressing Parkin S65A
   mutant rather than the S108A or C430S mutant enhanced the surface
   levels of MHC-I ([188]Fig. 5H), confirming the AMPK-dependent Parkin
   activation in regulating antigen presentation. Overexpression of Parkin
   barely changed the mRNA expression level of MHC-I (H2-K1 and H2-Q1),
   suggesting that the decreased protein levels of MHC-I were independent
   of transcriptional regulation ([189]Fig. 5, I and J). In contrast, the
   decreased surface levels of MHC-I induced by Parkin overexpression in
   Raw267.4 cells were rescued by autophagy inhibitors chloroquine (CQ)
   and bafilomycin A1 (BafA1) ([190]Fig. 5K). Consistently, applying BafA1
   substantially increased total cellular levels of MHC-I in raw cells
   overexpressing full-length Parkin but scarcely changed those in control
   cells ([191]Fig. 5, L and M). Last, we use Atg5^fl/fl-Lyz2-Cre^+/− mice
   whose Atg5 is deleted in macrophages (fig. S12A) to examine whether
   specific deletion of autophagy in macrophage influences tumor
   progression and macrophage antigen presentation. In vivo MC38 model
   showed that specific deletion of Atg5 in macrophages impeded tumor
   growth ([192]Fig. 5N and fig. S12B) and facilitated T cell infiltration
   in the TIME ([193]Fig. 5O and fig. S12, C and D) and MHC-I expression
   on TAMs ([194]Fig. 5P). Antigen presentation assays also showed
   significantly stronger T cell priming ability of Atg5^flfl-Lyz2-Cre^+/−
   TAMs and BMDMs than the control group ([195]Fig. 5Q and fig. S12E).
   Collectively, these findings suggest that AMPK-dependent Parkin
   activation suppresses MHC-I expression on macrophage surface via
   autophagy.

Parkin-associated signatures correlate with poor prognosis

   Given the Parkin deficiency–mediated antitumor effects that we
   confirmed in preclinical tumor models, we next set out to assess the
   clinical relevance of Park2 in ICB treatment. First, using ssGSEA
   algorithm, we found that the expression level of Park2 was
   significantly inversely correlated with the abundance of activated
   CD8^+ T cells ([196]Fig. 6A) and effector memory CD8^+ T cells (fig.
   S13A) in various solid tumors. Then, we stratified patients with
   metastatic urothelial cancer (mUC) from the IMvigor210CoreBiologies R
   package (26359337) based on responders and nonresponders receiving
   PD-L1 blockade and found that nonresponders had higher Park2 expression
   than responders ([197]Fig. 6B). Meanwhile, significant associations
   were shown between low expression of Park2 and better overall survival
   in patients with mUC and progression-free survival in patients with
   melanoma receiving ICB therapies (fig. S13, B to D). Therefore, we
   wonder whether Parkin deficiency could improve the efficacy of ICB
   treatment in an MC38 model, which is responsive to anti–PD-L1 therapy.
   Anti–PD-L1 treatment indeed demonstrated better tumoricidal efficacy in
   Park2^−/− mice than WT mice ([198]Fig. 6C). Collectively, these results
   attest to the clinical relevance of Parkin deficiency in T cell
   infiltration and sensitivity to cancer immunotherapy.

Fig. 6. The absence of Parkin enhances the efficacy of ICB therapy, and
Parkin is a potential therapeutic target for cancer immunotherapy.

   [199]Fig. 6.
   [200]Open in a new tab

   (A) Correlation analysis between the expression of Park2 and the
   infiltration of activated CD8^+ T cells in various tumors. UCS, Uterine
   carcinosarcoma; THCA, Thyroid carcinoma; SARC, Sarcoma; PRAD, Prostate
   Adenocarcinoma; LGG, Low-grade gliomas; KIRP, Kidney renal papillary
   cell carcinoma; KIRC, Kidney Renal Clear Cell Carcinoma; HNSC, Head and
   neck squamous cell carcinoma; GBM, Glioblastoma multiforme; COAD, Colon
   adenocarcinoma; BRCA, Breast Invasive Carcinoma; BLCA, Bladder
   Urothelial Carcinoma. (B) Expression of Park2 in patients with mUC who
   received neoadjuvant anti–PD-1 therapies (GSE26359337). (C) Tumor
   volume of WT and Park2^−/− MC38–bearing mice administered
   intraperitoneally with anti-mouse PD-L1 or left untreated on days 9,
   12, 15, and 18. The numbers of mice are indicated. (D to K) Correlation
   of the Park2^−/− signature score with overall survival (OS), dextran
   sulfate sodium (DSS), and progression free interval (PFI) in adenoid
   cystic carcinoma (ACC) (D), breast cancer (BRCA) (E), colon
   adenocarcinoma (COAD) (F), liver hepatocellular carcinoma (LIHC) (G),
   prostate adenocarcinoma (PRAD) (H), rectum adenocarcinoma (READ) (I),
   skin cutaneous melanoma (SKCM) (J), and stomach adenocarcinoma (STAD)
   (K). Low Park2^−/− signature, blue curve; high Park2^−/− signature, red
   curve. The results are depicted as the means ± SEM, *P < 0.05 and
   **P < 0.01.

   As Parkin deficiency modulates macrophage antigen presentation and
   inflammatory response to facilitate T cell proliferation and
   activation, we suspected that Park2^−/− signaling could be beneficial
   for predicting patients’ prognosis in human tumors. To do so, we
   intersected genes up-regulated in macro_CD80 and macro_C1qc TAMs of the
   Park2^−/− group to establish an overall Park2^−/− activation signature
   (hereafter referred to as the Park2^−/− signature) based on scRNA-seq
   data. This signature consists of genes directly and indirectly
   controlled by Park2 in the transcriptional regulatory network of TAMs
   (fig. S13E). GO functional enrichment analysis showed that genes in the
   Park2^−/− signature were significantly enriched in pathways including
   chemokine binding, antigen peptide binding, and immune response–related
   pathways (fig. S13F). We then applied the Park2^−/− signature to 31
   types of patients with tumor from The Cancer Genome Atlas (TCGA)
   dataset. A pan-cancer survival analysis showed that up-regulation of
   Park2^−/− signature was associated with improved prognostic indicators
   overall survival (OS), dextran sulfate sodium (DSS), and progression
   free interval (PFI) in patients ([201]Fig. 6, D to K), indicating that
   Park2^−/− signature could serve as a biomarker for effectively
   predicting the prognosis for various patients with solid tumor.
   Overall, these results suggest that Parkin can be considered either a
   single target or a set of signature genes, providing perspectives for
   existing immunotherapies.

DISCUSSION

   Parkin has long been portrayed as a neuroprotector, as its dysfunction
   is largely ascribed to neurodegeneration in patients with PD ([202]19).
   Emerging evidences have implicated Parkin as a tumor suppressor
   regarding tumorigenesis ([203]24, [204]27–[205]29) but whether Parkin
   regulates immune cell–intrinsic immunity in the TIME remains largely
   unresolved. Here, we find that Parkin negatively modulates antitumor
   immunity in the TIME, as Parkin deficiency reshapes the
   immunosuppressive TIME and prohibits tumor progression ([206]Figs. 1 to
   [207]4). In this context, depletion of either Park2^−/− CD8^+ T cells
   ([208]Fig. 1, N and O) or macrophages ([209]Fig. 3, E and F) abolishes
   Parkin deficiency–mediated tumor regression. Our data shed light on the
   essential role of macrophages in Parkin-mediated immune response, as
   demonstrated by superior antitumor effects of Park2^−/− BMDMs
   ([210]Fig. 3, H and I) rather than naïve Park2^−/− splenic T cells
   ([211]Fig. 2, M and N) in adoptive transfer assays. Although
   normalization of adaptive immunity has primarily been perceived as a
   fundamental strategy for immunotherapy ([212]43, [213]44), it would be
   of great importance to targeting innate immunity to reverse the
   immunosuppression of the TIME. Ongoing efforts have focused on
   equipping TAMs with effector functions, including direct phagocytosis,
   activation of inflammation, and professional antigen presentation
   ([214]45, [215]46). In our work, Park2^−/− TAMs show enhanced
   inflammatory response and antigen presentation ([216]Fig. 4, A to D)
   and further facilitate T cell priming, clonal expansion, and long-term
   memory ([217]Fig. 4, E to G, I, and K to P), which were further
   validated in mice with macrophage-specific Parkin KO ([218]Fig. 4, H
   and J, and fig. S10, F to H). Overall, the above data underscore the
   interplay between TAMs and T cells in shaping antitumor immunity.

   Cancer and PD, two entirely different classes of disease, can converge
   on a single gene, PARK2. One possible reason is the modulation of
   inflammatory response, as patients with PD ([219]22) and tumor
   inoculated–Park2^−/− mice ([220]Figs. 3J and [221]4C) show elevated
   inflammatory factors. Previous studies have demonstrated that kinases
   PINK1 and AMPK, which sense deleterious leakage of mitochondrial DNA
   and cellular ADP/ATP ratio, respectively, could activate Parkin to
   regulate inflammatory response ([222]20, [223]21, [224]23–[225]25),
   implying a Parkin-related activation cascade in regulating antitumor
   immunity. Conventional PINK1/Parkin-mediated mitophagy seems to be
   negligible for immune cell activation in the TIME, as Pink1^−/− mice
   exhibit no tumor inhibition effect such as Park2^−/− mice ([226]Fig. 5,
   A and B) and mutation of the key PINK1 phosphorylation site in Parkin
   does not affect Parkin’s function in terms of MHC-I regulation
   ([227]Fig. 5F). Instead, we demonstrate that the regulation of MHC-I by
   Parkin requires the activation of the AMPK/ULK1 axis in conjunction
   with the downstream autophagy components ([228]Fig. 5, H to M), which
   were further validated in autophagy-deficient mice with a
   macrophage-specific manner ([229]Fig. 5, N to Q).

   Down-regulation of MHC-I expression has been reported to be kidnapped
   by tumor cells to avoid immunosurveillance ([230]47). Excessive
   autophagy purportedly degrades ubiquitinated MHC-I in tumor cells via
   the autophagy cargo receptor NBR1, which, in turn, impairs T cell
   response and favors immune evasion of pancreatic cancer ([231]42).
   Here, we find an unreported mechanism of Parkin in macrophages, which
   leverages AMPK activation and autophagy to regulate MHC-I surface
   expression, thereby broadening the role of AMPK/Parkin/autophagy axis
   beyond mitochondrial homeostasis. Meanwhile, MHC-I has been shown to be
   expressed at the surface of dopaminergic neurons in patients with PD,
   providing further support for the plausibility of MHC-I regulation in
   the development of PD ([232]48). Thus, it would be interesting to
   characterize the direct substrates of AMPK-induced Parkin–mediated
   ubiquitination in regulating MHC-I, which could provide insight into
   understanding both Parkin-mediated tumor development and PD pathology.

   Despite significant improvements made in ICB treatment, current
   immunotherapies urgently require promising targets and prognosis
   biomarkers ([233]49). Here, our work provides important implications
   for improving current ICB treatment. We demonstrate that the expression
   level of Park2 is negatively correlated with the infiltration of
   tumor-killing CD8^+ T cells in tumor tissues ([234]Fig. 6A and fig.
   S13A) and with the sensitivity of existing tumor immunotherapies
   ([235]Fig. 6, B and C). In addition, we established a Park2^−/−
   signature consisting of genes up-regulated in Park2^−/− TAMs and, as a
   proof of concept, confirmed its efficacy as a biomarker for predicting
   prognosis for ICB treatment ([236]Fig. 6, D to K). In light of in vivo
   antitumor effects of Park2^−/− BMDMs ([237]Fig. 3, H and I), targeting
   Parkin in chimeric antigen receptor macrophages (CAR-macrophages) holds
   promise for optimized cellular immunotherapy to overcome the limited
   infiltration of chimeric antigen receptor T cells into immunologically
   cold solid tumors ([238]50). Collectively, Parkin acts as a negative
   immune regulator to favor tumor progression and could serve as a
   promising target or a signature gene set to improve current
   immunotherapies.

   Overall, we demonstrate that targeting Parkin could not only enhance
   antigen presentation and inflammatory response of macrophages but also
   unleash the effector function of CD8^+ T cells against solid tumors.
   Our findings provide a fresh perspective on targeting innate immunity
   to tailor adaptive immunity, thereby eliciting a systemic antitumor
   immune response for cancer immunotherapy.

MATERIALS AND METHODS

Mice

   All mouse experiments were approved by the Institutional Animal Care
   and Use Committee of Sun Yat-sen University (approval number:
   SYSU-IACUC-2023-000800). Park2^−/− (B6.129S4-Prkn^tm1Shn/J) and
   Pink1^−/− (B6.129S4-Pink1^tm1Shn/J) mice were obtained from the Jackson
   Laboratory. WT (C57BL/6 J) and Rag1^−/− mice were purchased from
   GemPharmatech Co. Ltd., Nanjing, Jiangsu, China. Prkn-flox mice (strain
   S-CKO-11411) were purchased from Cyagen. Atg5-flox mice were provided
   by Z. Li from Southern Medical University. Lyz2-Cre mice were purchased
   from Shulaibao (Wuhan) Biotechnology Co. Ltd. Both male and female mice
   were used in this study. The sex of mice was not considered a covariate
   factor. All mice used in this study were between 8 and 24 weeks of age
   and were housed under specific pathogen–free condition at the animal
   facility of the Laboratory Animal Center of Sun Yat-sen University.

Cell lines and cell culture

   Murine colon adenocarcinoma MC38 (provided by R. Xu from Sun Yat-sen
   University Cancer Center), glioblastoma multiforme cancer cell line
   CT2A (provided by Q. Xie from West Lake University), hepatocellular
   carcinoma Hepa1-6, breast cancer cell line E0771 (provided by H. Zheng
   from Tsinghua University), fibroblast L929 cells (American Type Culture
   Collection), and macrophage cell line Raw264.7 (American Type Culture
   Collection) were cultured in Dulbecco’s modified Eagle’s medium
   containing 2 mM glutamine (Thermo Fisher Scientific) and were
   supplemented with 1% penicillin-streptomycin (Thermo Fisher Scientific)
   and 10% fetal bovine serum (FBS; TransGen Biotech). Cell lines were
   routinely tested for mycoplasma, and none of cell lines were positive
   in the reported experiments.

   Mouse BMDMs were produced as described previously ([239]51, [240]52).
   Briefly, mouse bone marrow cells were cultured in the medium with 10%
   FBS, 1% penicillin-streptomycin, and 30% (v/v) L929 cell–conditioned
   medium as a source of macrophage CSF. On days 4 and 7, the medium was
   half replaced with fresh medium. After 7 days, mature BMDMs were
   pretreated with different compounds at corresponding conditions.

Subcutaneous tumor models

   Tumor cells were digested into single cells by trypsin (Gibco) and
   washed three times with cold phosphate-buffered saline (PBS). A total
   of 2 × 10^6 MC38, 2 × 10^6 Hepa1-6, and 2.5 × 10^6 CT2A tumor cells
   were injected subcutaneously into the right flank of mice. A total of
   2 × 10^6 E0771 tumor cells were injected into mouse breast pads. The
   length (L), width (W), and height (H) of the tumors of mice were
   recorded starting at day 6. The tumor volume (V) was calculated using
   the formula V (mm^3) = L (mm) × W (mm) × H (mm) × 0.5. When tumors
   reached 1000 mm^3, mice were generally euthanized and recorded as dead
   for survival observation. For Hepa1-6 rechallenge tumor model,
   tumor-free mice were rechallenged with triple tumor cells 60 days after
   the first injection.

   For depletion of immune cells, 100 μg of anti-mouse CD4 (BioXCell,
   GK1.5) and anti-mouse CD8α (BioXCell, 2.43) antibodies were delivered
   by intraperitoneal injection every week; 200 μg of anti-mouse CSF1R
   (BioXCell, AFS98) antibodies were injected every 5 days from day 11.
   For combined therapy, when the tumor volume reached 100 mm^3, mice were
   injected intraperitoneally four times with anti-mouse PD-L1 (BioXCell,
   10F.9G2) antibodies at a dosage of 50 μg per mouse every 3 days.

Preparation of single-cell suspension for mouse samples

   To obtain tumor-filtrating immune cells, tumors were sliced into small
   pieces of 1 to 2 mm in petri dishes and then digested in digestion
   buffer [10 mM Hepes, collagenase D (1.25 mg/ml), collagenase V (0.85
   mg/ml), deoxyribonuclease I (30 U/ml), dispase II (1 mg/ml), and 10%
   FBS diluted in Hanks’ balanced salt solution] for 45 min on a rotor at
   37°C. The dissociated cells were passed through 70- and 40-μm cell
   strainers (BD) sequentially, centrifuged at 600g for 5 min at 4°C, and
   then prepared as single-cell suspension. Tumor-infiltrating immune
   cells were enriched from tumor tissues by magnetic separation using
   mouse CD45 MicroBeads (Miltenyi Biotec, catalog. no. 130-052-301).

Preparation of the tumor interstitial fluids

   The preparation process for tumor interstitial fluid is as follows:
   Isolate tumor tissue and put it in an Eppendorf tube. Add an
   appropriate volume of PBS at a ratio of 1 ml of PBS per gram of tumor
   tissue. Use a tissue grinder to homogenize the tumor tissue. After
   grinding, centrifuge at 5000 rpm for 5 to 10 min and collect the
   supernatant, completing the preparation of the tumor interstitial
   fluid.

Flow cytometry

   Anti-mouse CD16/32 and mouse immunoglobulin G2a antibodies were used to
   block nonspecific binding with Fc receptors before all surface
   staining. For staining, cells were stained with antibodies in PBS
   containing 2% FBS at 4°C for 30 min. The antibodies used for
   fluorescence-activated cell sorting (FACS) analyses are listed in table
   S9. The experimental operation followed the standard operating
   procedure for the CytoFLEX analyzer (Beckman Coulter), and FlowJo
   software was used for data analysis.

Immunohistochemistry

   The tumor tissues were fixed with 4% formaldehyde and stained with H&E
   at the Servicebio company. Then, the infiltration of CD4^+ or CD8^+ T
   cells was visualized by immunohistochemical staining and calculated
   with a 5× objective using ImageJ software.

Intracellular cytokine staining

   After surface staining, cells were washed by PBS and fixed with 100 μl
   of intracellular fixation buffer for 30 min at room temperature in the
   dark, followed by permeabilization of the cell membrane with 300 μl of
   permeabilization buffer. Last, cells were stained with antibodies
   against specific intracellular markers. The antibodies used for FACS
   analysis are listed in table S9.

T cell function assay

   For T cell activation assay, T cells were enriched from the spleens
   with EasySep Mouse T cell isolation kit (STEMCELL) according to the
   manufacturer’s instructions. Before the activation, anti-CD3ε (1 μg/ml)
   antibodies were precoated into the plate at 4°C for 12 hours. The
   anti-CD3ε antibodies and anti-CD28 (1 μg/ml) antibodies were used to
   stimulate T cells. After stimulation for 48 hours, the activation of T
   cells was detected by CytoFLEX analyzer (Beckman Coulter).

   For T cell proliferation assay, OT-I T cells were enriched from the
   spleens of OT-I mice with EasySep Mouse CD8^+ T cell isolation kit
   (STEMCELL) according to the manufacturer’s instructions. A total of
   1 × 10^5 BMDMs were seeded in 24-well flat-bottom plates overnight.
   Before cocultured with 5 × 10^5 carboxyfluorescein diacetate
   succinimidyl ester (CFSE)–labeled OT-I T cells, BMDMs were cultured by
   purified OVA peptide (1 μg/ml) for 6 hours. After incubation for 48
   hours, CFSE intensity was quantified by CytoFLEX analyzer (Beckman
   Coulter).

   For TIL reactivation assay, tumor-infiltrating CD4^+ and CD8^+ T cells
   were enriched from the tumor tissues by magnetic separation using mouse
   CD4^+/CD8^+ (TIL) MicroBeads (Miltenyi Biotec, catalog. no.
   130-116-480). TILs were reactivated by phorbol 12-myristate 13-acetate
   9 (20 ng/ml) and ionomycin (1 μg/ml) for 6 hours. Then, the activation
   of TILs was detected by CytoFLEX analyzer (Beckman Coulter).

T cell adoptive transfer assay

   Park2^−/− and WT splenic CD8^+ T cells were isolated and cultured as
   described above. Rag1^−/− mice were subcutaneously inoculated with
   2 × 10^5 MC38 cells. Then, Rag1^−/− recipient mice were injected
   intravenously with Park2^−/− and WT splenic CD8^+ T cells on day 1.
   Tumor volume was supervised every 2 days starting at day 6.

Macrophage adoptive transfer assay

   Park2^−/− and WT BMDMs were prepared as previously described and
   digested into single cells. Hepa1-6 tumor–bearing WT mice were injected
   with Park2^−/− and WT BMDMs (3 × 10^6 per mouse) intravenously on days
   3, 6, and 9 after tumor inoculation. Tumor volume was supervised every
   2 days starting at day 6.

In vitro phagocytosis assay

   Phagocytosis assays were performed according to the previous studies
   ([241]53, [242]54). A total of 5 × 10^4 Park2^−/− and WT BMDMs were
   seeded overnight in a 24-well tissue culture plate, treated or not
   treated with polyinosinic-polycytidylic acid (10 μg/ml),
   lipopolysaccharide (100 ng/ml), and IL-6 (100 ng/ml) for 24 hours. On
   the next day, 2 × 10^5 L1210 cells were labeled with 2.5 μM CFSE
   (Thermo Fisher Scientific) and incubated with Park2^−/− and WT
   macrophages in non-FBS medium plus anti-mouse CD47 monoclonal antibody
   (10 μg/ml) or control immunoglobulin G for 2 hours at 37°C. After
   incubation, macrophages were washed four times using PBS and imaged
   with an inverted microscope. The phagocytosis efficiency was calculated
   as equation = (the number of macrophages containing CFSE^+ target
   cells) / total macrophages × 100.

DNA constructs and retroviral infection

   To construct a series of Parkin variants, the full-length, point
   mutations S65A, S108A, and C430S of Parkin were cloned and fused to
   pMXs–green fluorescent protein (GFP), a retroviral vector. Raw264.7
   cell lines were respectively infected with these plasmids pMXs-GFP
   using Lipofectamine 3000 and sorted according to comparable GFP
   expression. Primers were designed for polymerase chain reaction
   amplification and sanger sequencing of Parkin full-length, point
   mutations S65A, S108A, and C430S variants (forward,
   5′-TCAAGGAGGACGGCAACATC-3′; reverse, 5′-CATCTGTTCTTGGCCCTGAG-3′).
   Detailed information about the construction strategies is provided in
   table S8.

Western blot

   Cells were lysed in radioimmunoprecipitation assay lysis buffer
   (Beyotime Biotechnology, P0013B). Protein concentration was assessed
   using the bicinchoninic acid protein assay kit (Beyotime Biotechnology,
   P0012S). Samples were resolved using 10% polyacrylamide gel and
   transferred to polyvinylidene difluoride membranes. Membranes were
   blocked with 5% bovine serum albumin/Tris-buffered saline with 0.1%
   Tween 20 detergent for 1 hour at room temperature and incubated
   overnight at 4°C with antibodies recognizing MHC-I (ABclonal, A1013),
   Parkin (Santa Cruz Biotechnology, sc-32282), and ATG5 (Cell Signaling
   Technology, 12994T). Membranes were then probed with horseradish
   peroxidase–linked secondary antibodies and developed Pierce ECL Protein
   Blotting substrate (Thermo Fisher Scientific) using ChemiDoc XRS+
   (Bio-Rad).

Enzyme-linked immunosorbent assay

   Mouse IFN-γ (Servicebio, EK280/3-01) and IL-2 (Servicebio, EK202/2-01)
   enzyme-linked immunosorbent assay kits were used to determine the
   concentrations of IFN-γ and IL-2 in tumor interstitial fluids following
   standard protocols.

scRNA-seq data processing and analysis

   Cell Ranger 5.0.1 (10x Genomics) was applied to align the gene
   expression libraries to the mm10 genome and generate gene-cell count
   matrices using default parameters. Then, the initial expression
   matrices were further filtered by Seurat 3.15 R package ([243]55)
   according to the following criteria: (i) Cells that contained less than
   500 molecular identifiers were filtered out; (ii) cells with less than
   200 or more than 6000 genes were removed; (iii) cells were excluded if
   contained higher than 15% mitochondrial counts or lower than 20%
   ribosomal counts; (iv) genes present in less than four cells were
   precluded; and (v) DoubletFinder 2.03 in R ([244]56) was used to
   exclude 2.5% potential doublets. In the end, a total of 40,119 cells
   with 22,620 genes were retained for the downstream analyses.

   Seurat was subsequently used to normalize the raw expression matrix,
   find highly variable features, identify integration anchors, integrate
   samples, and scale features. Principal components analysis was
   performed on the 2000 most highly variable features to project into the
   top 20 principal components, and dimension reduction was conducted via
   Uniform Manifold Approximation and Projection (UMAP) or t-SNE. Samples
   were further clustered by shared nearest neighbor–based Louvain
   algorithm as implemented in function FindClusters of Seurat. Cell
   clusters were annotated by the expression of known immune cell markers,
   including Cd3d, Cd3e, Cd3g, Cd28 (T cells), Cd8a, Cd8b (CD8^+ T cells),
   Cd4, Il7r, Ctla4 (CD4^+ T cells), Foxp3, Il2ra, Il2rb, Pdcd1 (T[regs]),
   Cd14, Sell (monocytes), Fcgr3, Adgre1, Itgam, Cd68, Cd163, Csf1r
   (macrophages), Itgax, Flt3, Batf, Thbd, Clec9a (dendritic cells),
   Tpsab1, Kit, Enpp3, Tpsb2 (mast cells), Fut4, Csf3r, Cxcr2, S100a8,
   S100a9 (neutrophils), Nkg7, Ccl5, Ncr1, Klrc1, Klrk1 (NK cells), Cd19,
   Ms4a1, Cd79a, Cd79b, and Blnk (B cells).

Differential expression and pathway activity analyses

   DEGs were identified via Wilcoxon test of FindMarkers and
   FindAllMarkers functions embedded in Seurat with a log[2]FC threshold
   of 0.25 and adjusted P ≤ 0.05. GO and KEGG analyses were conducted by
   clusterProfiler 4.05 R package ([245]57) based on the DEGs of each cell
   cluster. Gene set variation analysis 1.40.1 package in R ([246]58) with
   default setting was implemented in ssGSEAs to assign normalized
   enrichment score for individual cells.

Gene signatures

   The difference in signature scores between WT and Park2^−/− CD8^+ T
   cells was evaluated using Wilcoxon rank-sum test, visualized by violin,
   circle, or empirical cumulative distribution function (ECDF) plots.
   Detailed information about the gene signatures is provided in table S1.

Developmental trajectory inference

   Monocle 2.20.0 in R 28825705 was used to infer T cell trajectories, and
   dimensionality reduction and cell ordering were performed with default
   settings. The progenitor signature scores of all T cells were generated
   by ssGSEA to identify innate cell clusters, also considered the start
   of the pseudotime trajectory. Exhaustion and inhibitory scores for each
   T cells were added to Monocle and Seurat to produce further trajectory
   and ridge plots, respectively.

Determining potential ligand-receptor interactions

   R package CellChat 1.1.3 ([247]59) was used to study and visualize the
   communicating interactions between T cell and myeloid cell clusters,
   selecting the cell-cell contacts and mouse protein-protein
   interactions. To infer the up-regulated ligand-receptor pairs, the
   interactions of WT and Park2^−/− group populations were compared (table
   S4). Public ligand-receptor repository CellPhoneDB 2.1.1 ([248]60) was
   then applied to complement the cell-cell communications between CD8^+ T
   cell and macrophage subclusters (table S5).

TCR-seq data processing and analysis

   TCR-seq was conducted by Illumina HiSeq system (PE150 mode) using
   ImmuHub TCR profiling system (ImmuQuad Biotech). The V, D, J, and C
   segments were aligned with the international immunogenetic database to
   identify nucleotide and amino acid sequences of TCRα/β
   complementarity-determining region 3, excluding out-of-frame and stop
   codon sequences. The amounts of TCRα/β clonotype were determined by the
   adding numbers of each clone sharing the same nucleotide sequence of
   complementarity-determining region 3. The characteristics of the TCRα
   clonotypes were displayed in table S6.

Immune cell infiltration and correlation analysis

   The infiltration levels of immune cells were calculated as normalized
   enrichment score from TCGA pan-cancer RNA-seq data ([249]61) based on
   cell markers extracted from The Cancer Immunome Atlas ([250]62).
   Effector memory CD8^+ T cell scores were afterward analyzed for their
   correlation with the expression level of PARK2 in pan-cancer.

Assessment of predictive response efficacy for ICB responses

   Two ICB cohorts were collected to compare the expression level of PARK2
   between responders (complete response or partial response) and
   nonresponders (stable disease or progressive disease) to ICBs.
   IMvigor210 dataset containing patients with mUC treated with PD-L1
   inhibitors and Mariathasan et al. ([251]63) cohort with patients with
   melanoma receiving anti–CTLA-4 (Cytotoxic T-Lymphocyte Associated
   Protein 4) treatment ([252]64) were enrolled.

Survival analysis with the Park2^−/−signature in pan-cancer

   The Park2^−/− signature lists were defined as the intersect top 20
   up-regulated DEGs in KO group of the cluster macro_CD80 and macro_C1qc.
   Park2^−/− scores were first calculated as the sum TPM (transcripts per
   million, transcripts per kilobase of exon model per million mapped
   reads) value based on expression profiles from the TCGA ([253]61). The
   Park2^−/− signature gene lists and the computed scores were described
   in table S7. Pan-cancer survival Kaplan-Meier survival curves were
   plotted by survival 3.2-13 package in R, with the best divisions of
   PARK2 expression level separately counted by R survminer 0.49 package,
   and the cutoff for the survival analysis was median.

Statistical analysis

   All statistical analyses were conducted by GraphPad Prism 8.21 or R
   4.03 for evaluating statistical significance or plotting. Comparisons
   between two groups were computed by two-tailed paired or unpaired t
   test, and two-way analysis of variance (ANOVA) test was applied in
   comparisons with more than two groups. Correlation was assessed by
   Spearman or Pearson test, and cumulative survival was analyzed by
   log-rank test. P values were corrected by Benjamini-Hochberg method in
   multiple tests, and P ≤ 0.05 was considered significant for all
   statistical analyses.

Acknowledgments