Parkin Induces Ubiquitination and Large Extracellular Vesicle Release of HMGB1 to Activate Antitumor Immunity

Abstract

Parkin is a mitochondria-associated E3 ubiquitin (Ub) ligase that mediates mitophagy and organelle quality control. More recently, Parkin has been implicated in stimulating antitumor immunity and reprogramming the tumor immune microenvironment. Here, we showed that Parkin ubiquitinates the alarmin molecule, high mobility group box-1 (HMGB1) on Lys146 (K146) using predominantly K48 linkages. By molecular modeling, the in-between-ring (IBR) domain of Parkin (Gln326-Leu358) made extensive contacts with the amino-terminus A box of HMGB1 (Met1-Ser42), forming a mitochondria-associated Parkin-HMGB1 complex that juxtaposes K146 to Ub active site residues Gly76 and Arg74. Instead of proteasomal degradation, Parkin ubiquitination of K146 enabled the loading of HMGB1, but not HMGB1 K146A mutant, onto autophagy- and mitochondria-derived large extracellular vesicles (LEV). In turn, released Parkin-HMGB1 LEV stimulated a potent interferon (IFN) and cytokine response in recipient cells, expanding CD8+ T cell subsets with effector (CD69+/KLRG1+), self-renewal (TCF-1+/PD-1+), and cytotoxic (KLRG1+/GrzB+) properties. Conditional expression of Parkin induced HMGB1 release, activated intratumoral CD8+ T cells, and suppressed syngeneic tumor growth in vivo in a response that was abolished by HMGB1 silencing. These data identify that Parkin-LEV regulated release of HMGB1 reprograms antitumor immunity via stimulation of IFN signaling and expansion of specialized CD8+ T cell subsets.

INTRODUCTION

Interferons (IFN) and IFN-stimulated genes (ISG) are potent mediators of innate immunity, opposing pathogenic infections, including viruses (1). This pathway involves signaling by plethora of alarmin molecules (2) originating from infectious agents (3), mitochondrial damage (4) or multifunctional, cytokine-like effectors (5) that drive transcription of IFN, ISG and plethora of immune-inflammatory mediators (6).

Aside from pathogens, IFN signaling plays an important (7), but still poorly understood role in cancer. We know that this response is context-dependent (8), and paradoxically associated with tumor suppression (9,10) via recruitment of CD8 T cells (11) and activation of dendritic cells (DC) (11), or, conversely, tumor promotion, inducing CD8 T cell exhaustion (12) and insensitivity to therapeutic immune checkpoint inhibition (13). What controls the pro- vs. anti-tumorigenic responses is mostly unknown and a potential link between IFN signaling and endogenous tumor suppression mechanism(s) has not been investigated.

A candidate for this process is Parkin (PRKN), a mitochondria-associated E3 ubiquitin (Ub) ligase implicated in organelle quality control via mitophagy (14). Although this pathway has been linked to familial Parkinson’s Disease (PD), it is also clear that PRKN has functions outside the CNS, in particular tumor suppression (15). Accordingly, reintroduction of PRKN in transformed cells inhibits multiple tumor traits of cell division (16), metabolism (17,18) and cell motility (19). However, this may not be the only mechanism of tumor suppression as recent data have shown that PRKN also stimulates IFN signaling, reprograms an antitumor immune microenvironment and opposes tumor growth, in vivo (20).

In this context, PRKN-expressing cells have been shown to release High Mobility Group Box 1 (HMGB1) (20), a pleiotropic cytokine-like alarmin molecule (5) that drives multiple immune-inflammatory responses by stimulating cytokine and IFN gene expression (21,22). However, whether this response is important in PRKN tumor suppression is unclear: HMGB1 extracellular release has been linked to passive, unregulated discharge from dying cells (23), a role of post-translational modification(s), including ubiquitination, in this process has not been determined (24), and HMGB1-associated inflammation has been mostly associated with tumorigenesis (25) and inhibition of anticancer immunity (26).

In this study, we investigated the mechanism of PRKN release of HMGB1 in IFN signaling and stimulation of antitumor immunity.

MATERIALS AND METHODS

Cells and cell culture

Human prostate cancer PC3 cells (Cat. CRL-1435, RRID: CVCL_0035) and murine prostate cancer TRAMP-C2 cells (Cat. CRL-2731, RRID: CVCL_3615) were purchased from the American Type Culture Collection (ATCC) and maintained in culture with 5% CO₂ at 37°C in a humidified incubator. PC3 cells were grown in RPMI medium (Corning) supplemented with 10% fetal bovine serum (FBS) (Gemini) and 1% Penicillin/Streptomycin (Corning). TRAMP-C2 cells were grown in DMEM medium (Corning) supplemented with 5 μg/ml human insulin, 10 nM dihydrotestosterone (DHT), 10% FBS, and 1% Penicillin/Streptomycin. The murine prostate cancer MPTEN1 cell line has been described previously (20). To generate inducible PRKN-overexpressing MPTEN1 clones, a tetracycline based inducible system (Tet-ON system) was used (20). Lentivirus particles containing pLV[Exp]-CMV>tTS/rtTA and pLV[Exp]-TRE>PRKN were generated by VectorBuilder. Clones were selected in puromycin (1 μg/ml) and hygromycin B (100 μg/ml) and PRKN was conditionally induced by Doxycycline (Doxy, 0.1 μg/ml). Clones of PC3 cells conditionally expressing PRKN in response to Doxy treatment were described previously (20). Cell passaging was limited to fewer than 40 passages, and cell lines were authenticated by STR profiling using the AmpFISTR Identifiler PCR Amplification Kit (Life Technologies) at the Wistar Institute’s Genomics Shared Resource. Cultured cells were verified to be mycoplasma-free using the Mycoplasma Primer Set (Bioo Scientific) every 2 months.

Antibodies and reagents

Primary antibody sources and dilution ratios for Western blotting and immunofluorescence studies are presented in Supplementary Table 1. Antibodies used for flow cytometry studies are presented in Supplementary Table 2. For Western blotting, HRP-conjugated goat anti-rabbit IgG secondary antibodies were purchased from GE-Amersham Biosciences (Cat. NA934, RRID: AB_772206) and HRP-conjugated goat anti-mouse IgG secondary antibodies were purchased from Cell Signaling Technology (Cat. 7076S, RRID: AB_330924). For immunofluorescence studies, STAR 580 fluorochrome-conjugated goat anti-rabbit IgG and STAR 635 fluorochrome-conjugated goat anti-mouse IgG were purchased from Abberior. For analysis of HMGB1 and HMGB2 protein stability, cycloheximide was purchased from MedChemExpress (MCE). Autophagy inhibitors, 3-methyladenine (3-MA) and bafilomycin-A1 (Baf-A1) were purchased from MCE.

Plasmids and transfections

A cDNA for Myc-tagged human WT PRKN was purchased from GeneCopoeia. eGFP-tagged WT human PRKN and Flag-tagged WT human HMGB1 were obtained from Addgene (#45875, and #31609). cDNA clones encoding E3 ligase loss-of-function PRKN S65A and C431S mutants, as well as ubiquitination-defective HMGB1 K112A and K146A mutants or HMGB2 K147A mutant were generated using PfuUltra II fusion HS DNA Polymerase (Agilent Technologies). For transient gene expression via plasmid transfection, X-TremeGene HP DNA Transfection Reagent (Roche) was used. Briefly, the various cell types were seeded at 70% confluency in complete medium for 24 h, mixed in serum-free medium with supplier-recommended amounts of plasmid DNA and transfection reagent and incubated for 15 min at 22°C. Small interfering RNA (siRNA) sequences used in this study are presented in Supplementary Table 3. For gene knockdown experiments, PC3 cells were plated at 60% confluency in serum-containing medium. After 24 h, the cultured medium was changed to Opti-MEM (Gibco) followed by siRNA transfection using lipofectamine RNAiMAX Reagent (Thermo Fisher), according to the manufacturer’s instructions. In some experiments, clones of parental or conditionally PRKN-expressing MPTEN1 were stably transduced with control non-targeting shRNA (shCtrl) or HMGB1-directed shRNA (shHMGB1) at multiplicity of infection (MOI) 10. Cells were selected in G418-containing medium (500 μg/ml) and validated for stable protein knockdown by Western blotting.

Flow cytometry

Spleens isolated from the various animal groups harboring subcutaneous MPTEN1 tumors were mechanically dissociated, filtered on a 70 μm filter and red blood cells were lysed with ACK (Lonza). Subcutaneous flank tumors were processed using the mouse Tumor Dissociation KIT (Miltenyi Biotec). After the enzymatic digestion, cells were filtered, and an antibody mix (50 μl/sample) in Brilliant Stain Buffer (BD) was added for surface staining for 15 min at 4°C. After washes in 1% FBS, 5 mM EDTA in PBS, samples were stained with eBioscience^™ Foxp3/Transcription Factor Staining Buffer Set (ThermoFisher) for additional 45 min at 4°C with gentle agitation. Aqua or Zombie UV LIVE/DEAD^™ Viability/Cytotoxicity Kit was added (1:300). For intracellular staining, samples were incubated with antibody mix (50 μl/sample in Foxp3 permeabilization buffer) for 45 min at 4°C in agitation. Compensation was set with UltraComp eBeads^™ Compensation Bead (for use with antibodies) and ArC^™ Amine Reactive Compensation Bead Kit (for use with LIVE/DEAD^™; Fixable dead cell stain kits). Samples were acquired on a BD FACSymphony^™ Cell Analyzer (Becton Dickinson) and data were analyzed by FlowJo 10.7 software (Tri-Star). The markers used to quantify specific NK and B cell subsets are as follows: active NK (CD69⁺/KLRG⁺ in single live cells CD3⁻/CD19⁻/NK1.1⁺); mature NK (CD69⁻/KLRG⁺ in single live cells CD3⁻/CD19⁻/NK1.1⁺); active B cells (CD69⁺ cells in single live CD3⁻/C19⁺cells).

Protein analysis

Cells were lysed in lysis buffer (150 mM NaCl, 50mM Tris-HCl pH 8.0, 1% NP-40) containing EDTA-free protease Inhibitor cocktail (Roche) and phosphatase inhibitor cocktail (Roche), followed by incubation on ice for 30 min with vortexing every 5 min. Protein extracts were obtained by centrifugation at 13,000 rpm for 15 min at 4°C. Protein samples were quantified using a BCA assay (Thermo Fisher), and equal protein amounts were loaded into each well of SDS polyacrylamide gels and separated by electrophoresis. Proteins were transferred onto PVDF membrane (Millipore sigma) using a wet transfer system (Bio-rad). Membranes were blocked in 5% milk containing 0.1% Tween-20 in TBS (TBS-T) for 20 min at 4°C. After washes in 0.1% TBS-T, membranes were incubated with primary antibodies of various specificity with continuous agitation in 2% TBS-T plus bovine serum albumin (BSA) for 16 h at 4°C, washed and further incubated with HRP-conjugated secondary antibodies of appropriate specificity in 5% milk containing 0.1% TBS-T for 1 h. After washes, proteins were detected on film (Tomas Scientifics) using ECL substrate (GE - Amersham Biosciences).

Conditioned medium protein analysis

Aliquots of conditioned medium (CM) from an equal number of cells expressing vector or PRKN were collected and centrifuged at 2,000 rpm for 10 min to remove dead cells and debris. The collected CM was then concentrated using a 3 kDa cutoff Amicon Ultra-15 Centrifugal Filter (Millipore Sigma) at 4,000 rpm for 40 min at 4°C. The protein concentration of the concentrated CM was quantified using a BCA assay (Thermo Fisher) and then mixed with 4X Leammli sample buffer (Bio-rad) to a final concentration of 1X. Subsequently, the samples were boiled for 5 min at 95°C and loaded onto SDS-PAGE for protein analysis.

Extracellular vesicles (EV).

PC3 cells transiently or conditionally expressing PRKN were cultured in RPMI medium without FBS for 24 h. After centrifugation at 2000 g for 10 min at 4°C to remove floating cells and debris, the supernatant was harvested and subsequently centrifuged at 10,000 g for 30 min at 4°C to isolate large extracellular vesicles (LEV). Pelleted LEV were washed with PBS, pH 7.4, and further centrifuged at 10,000 g for 30 min at 4°C. The resulting supernatant was subject to ultracentrifugation at 100,000 g for 30 min at 4°C to isolate small extracellular vesicles (SEV). Pelleted SEV were washed with PBS, 7.4 and further centrifuged at 100,000 g for 30 min at 4°C for further investigation. Both SEV and LEV samples were lysed in lysis buffer containing EDTA-free protease inhibitor and characterized using antibodies to SEV markers, CD9 (RRID: AB_2798139), TSG101(RRID: AB_2208090), and HSP90 (RRID: AB_397799) or LEV marker, Calnexin (RRID: AB_2228381). In some experiments, LEV samples were permeabilized in 0.5% Triton X-100 and analyzed by confocal fluorescence microscopy.

To quantify size distribution of EV released by vector- or PRKN-expressing cells, two independent methods were used. For EV characterization using the ZetaView NTA system (Particle Metrix), the sensitivity and shutter parameters were set to 80 and 75. For experiments using Flow Nanoanalyzer (NanoFCM), silica beads ranging from 155 nm to 850 nm in diameter were used as size standards to generate a standard curve in scatter mode. The side-scattered light intensity of each particle was measured, and its size was estimated based on the standard curve.

PRKN-HMGB1 docking

The PIPER docking procedure (27–30) used for prediction of PRKN-HMGB1 interactions was a multi-stage process that utilized the Fast Fourier Transform (FFT) correlation approach. Initially, a rigid body global search was conducted, where the FFT correlation method evaluated the energy function in the six-dimensional space of the mutual orientations of the two proteins. This step allowed for the exploration of various binding configurations. Following the initial search, the structures were reduced using rigid body filters based on electrostatics calculations and empirical potentials. This reduction aimed to eliminate less favorable poses, streamlining the number of structures for further analysis. The retained structures were then clustered using pairwise root mean square deviation (RMSD) as the distance measure, which helped identify similar binding modes. In the final stage, the structures underwent refinement using a medium-range optimization method known as Semi-Definite programming-based Underestimation (SDU). This refinement process improved the accuracy of the predicted binding poses.

Molecular Dynamics simulations

Molecular dynamics (MD) simulations were conducted using the Desmond MD method (31,32), designed to investigate the behavior of complex systems at the atomic level. In this study, simulations were executed with Desmond software. The job type was configured for MD simulation employing an NPT ensemble to maintain constant pressure and temperature throughout the simulation. This ensemble examines systems at equilibrium, where the number of particles (N), pressure (P), and temperature (T) are kept constant. The simulations were performed at a physiological temperature of 300.0 K, which is representative of conditions found in living organisms. The total duration of the simulation was set to 200 ns, allowing for an extensive observation of the system's dynamics. The system maintained a neutral charge of 0. Pressure was regulated at 1 atm (101.325 kPa) using the Martyna-Tobias-Klein (MTK) barostat algorithm, a reliable method for pressure control in MD simulations. Prior to the main simulation, the system underwent an initial relaxation phase using a steepest descent minimization algorithm. This step, which involved 1,000 minimization steps with a convergence threshold of 10⁻⁵ kcal/mol, was essential for alleviating any steric clashes or high-energy conformations. The equations of motion during the simulation were integrated using the Verlet integration algorithm, an MD method that utilizes a leap-frog approach for accurate integration. Electrostatic interactions were calculated using the Ewald summation method, which is effective for periodic systems. The trajectories of the simulation were recorded every 50 ps, resulting in a comprehensive dataset of 4,000 frames over the course of the 200 ns simulation. This extensive data collection allowed for detailed analysis of the protein's dynamics and structural stability throughout the simulation period.

HDOCK docking of PRKN-HMGB1-Ub complex

The tertiary structure of human ubiquitin binding-PRKN based on X-ray diffraction data (PDB: 5N2W) was used for molecular docking (33). The tertiary structures of human HMGB1 and HMGB1 K146A mutant were predicted using the HDOCK server (34). Molecular docking predictions between PRKN and either HMGB1 or HMGB1 K146A mutant were performed using the HDOCK server (35). PRKN was used as the receptor and HMGB1 was used as the ligand in docking simulations. For each prediction, 100 models were generated, and the top 10 models were selected based on docking scores. Based on the results of MD simulations performed in this study, representative models of PRKN-HMGB1 and PRKN-HMGB1 K146A complexes were selected, and the predicted protein structures were visualized using ChimeraX. The HDOCK predictions for a PRKN-HMGB1 binding interface aligned with the MD simulations, involving the C-terminal In-Between-Ring (IBR) region of PRKN and the N-terminal A Box domain of HMGB1.

Immunoelectron microscopy

PC3 cells expressing vector, PRKN, or PRKN C431S mutant were fixed in 3% paraformaldehyde/0.1% glutaraldehyde in PBS for 1 h at 4°C. Following fixation, the cells were pelleted by centrifugation and rinsed 3 times for 5 min each in PBS. The cell pellets were encapsulated in 2% SeaPrep agarose, cooled for 15 min at 4°C, rinsed 3 times in PBS, trimmed into 1 mm³ pieces, and rinsed twice for 30 min each with 0.05 M NH₄Cl at 4°C to block potential free aldehyde groups. The blocks were stored in PBS at 4°C overnight, followed by low temperature dehydration and embedding in the hydrophilic resin Lowicryl K4M at -35°C according to standard protocols. The resin blocks were polymerized by exposure to UV light, and thick sections were cut with glass knives and stained with 2% Toluidine blue. The blocks were then further trimmed, and thin sections (60–80 nm thick) were cut with a diamond knife and retrieved onto formvar-coated copper grids.

For immunostaining, the grids were floated section-side down on a drop of PBS containing 0.5% ovalbumin for 20 min, followed by transfer to a drop of primary antibody (rabbit polyclonal anti-HMGB1 (Proteintech) at 2 μg/ml, or rabbit polyclonal anti-LC3-II (Novus Biologicals) at 100 μg/ml and incubated for 16 h at 4°C. Following rinses in PBS plus 0.5% ovalbumin (3 x 5 min each), the grids were floated on droplets of protein A-gold (10 nm particles; O.D.= 0.07 at 525 nm) for 1 h at 22°C. Finally, the grids were rinsed several times in PBS plus 0.5% ovalbumin and distilled water and contrasted with 3% aqueous uranyl acetate and lead citrate, rinsing, and air drying. The stained grids were imaged with a JEOL 1400 transmission electron microscope (JEOL LTD USA; Danvers, MA) operating at 60kV, and digital images were acquired in 16-bit .TIF format with an AMT XR11 mid-mount ccd camera. For image quantification, fifty-six TIF images were analyzed from 16 PRKN-expressing cells, and the number of gold particles located over 530 mitochondrial profiles were counted.

RT-qPCR

Total RNA from cultured cells was extracted using the Quick-RNA Microprep Kit (Zymo Research) according to the supplier’s instructions. The RNA was quantified using NanoDrop (Thermo Fisher), and 2 μg of total RNA from each sample was used for reverse transcription. cDNAs were generated with TaqMan^™ Reverse Transcription Reagents (Applied Biosystems) according to the supplier’s instructions using a thermocycler (Bio-Rad). For RT-qPCR amplification reactions, cDNA, forward and reverse primers for the individual gene of interest, and PowerUp^™ SYBR^™ Green Master Mix (Applied Biosystems) were used, and the reaction was carried out using an ABI Fast 7500 instrument (Applied Biosystems). The primer sequences used in this study are presented in Supplemental Table 4. The expression level of the genes of interest was quantified by comparing the cycle threshold (Ct) values to the Ct of 18S rRNA, and the relative cDNA expression levels were calculated using the ΔΔCt method.

Ubiquitination analysis

Cell lysates were prepared in 100 μl IP buffer (150 mM NaCl, 20 mM Tris, 0.5% NP-40, pH 7.4) containing EDTA-free protease inhibitor cocktail and phosphatase inhibitor cocktail with 2% SDS for denaturation. The lysates were boiled for 10 min at 95 °C and then sonicated at 20% amplitude for 5 sec. To reduce the concentration of SDS to 0.2%, 900 μl of IP buffer was added. The mixture was centrifuged at 13,000 rpm for 15 min at 4°C, and the supernatant was collected for protein quantification. Each 1 mg of collected lysates was incubated with 1 μg of antibody to FLAG in continuous agitation for 16 h at 4°C. Subsequently, 50 μl of pre-cleared Protein A/G Agarose beads (Santa Cruz) were added and incubated in agitation for 2 h at 4°C. Samples were washed three times with IP buffer, mixed with 50 μl 2X Leammli sample buffer (Bio-rad), and boiled for 5 min at 95 °C. After separation by SDS-PAGE, the samples were incubated with antibodies to pan-ubiquitin (Ub), Ub-K48 linkages or Ub-K63 linkages and analyzed by Western blotting.

Proximity Ligation Assay

Protein-protein interactions for endogenous HMGB1 and Myc-PRKN were assessed via the Duolink In Situ PLA kit (Sigma-Aldrich). Briefly, PC3 cells were fixed with 4% PFA for 10 min at 22°C, washed with 1X PBS three times for 3 min and permeabilized in the presence of 0.25% Triton X-100 for 7 min at 22°C. Samples were blocked using blocking solution for 1 h at 37°C in a humidified chamber. Solutions of primary antibody to Myc (1:1000, RRID: AB_331783) or HMGB1 (1:50, RRID: AB_2232989) were incubated with the various samples for 16 h at 4°C in a humidified chamber. Following primary antibody incubation, samples were washed in 1X wash buffer A for 5 min at 22°C twice. PLA probe incubation was done using PLUS rabbit and MINUS mouse probes at a 1:5 dilution in antibody diluent for 1 h at 37°C in a humidified chamber. Samples were washed in 1X wash buffer A twice for 5 min at 22°C followed by ligation. Ligase was added to 1X ligation buffer at a 1:40 dilution, and slides were incubated for 30 min at 37°C in a humidified chamber. Samples were then washed twice for 5 min at 22°C in 1X wash buffer A, followed by amplification of the PLA reaction using Polymerase at a 1:80 dilution in amplification buffer for 100 min at 37°C in a humidified chamber. Following amplification, samples were washed 2x for 10 min in 1X wash buffer B at 22°C, then once in 0.01X wash buffer B. Slides were mounted using mounting media with Duolink In Situ Mounting Media with DAPI (DUO82040–5ML) and signals were quantified by fluorescence microscopy.

Immunofluorescence

PC3 cells (2x10⁵) were plated on a 24-well plate coated with 12ø optical grade cover glasses (Thermo Fisher), transfected with PRKN cDNA for 48 h and fixed in PBS containing 4% PFA (Thermo Fisher) for 10 min. Autofluorescence from the cells was blocked using 20 mM glycine for 20 min at 22°C. Cells were permeabilized with 0.1% Triton X-100 in PBS for 10 min and blocked for non-specific binding using 10% BSA in 0.1% PBS-T for 1 h at 22°C. The cells were incubated with primary antibodies to HMGB1 (1:100, RRID: AB_2232989) and LC3 (1:50, RRID: AB_2800018) in 10% BSA plus 0.1% TBS-T with constant agitation for 16 h at 4°C. Subsequently, secondary antibodies of appropriate specificity (1:500) were added for 1 h at 22°C and cover glasses were counterstained with Hoechst 3342 (Cell Signaling) at a 1:500 dilution for 5 min at 22°C. The cover glasses were mounted using Prolong mounting reagent (Thermo Fisher) onto glass slide for imaging. Images were acquired using a Leica TCS SP8 confocal microscope (Leica) with a magnification of 63X, a pixel size <40 nm in x and y, and a z-stack of 0.13 μm. Images were deconvolved using Hyugens software (SBI).

Mitochondrial fractionation

For subcellular fractionation experiments (36), the various cell types were harvested, washed with ice-cold PBS and spun down at 500 g for 5 min at 4°C. Pellets were resuspended in cytosolic extraction buffer (150 mM NaCl, 50 mM HEPES, 20 μg/ml digitonin, and 1 M hexylene glycol) and incubated on ice until cell membranes were degraded. Cell degradation was confirmed by Trypan blue staining. Cells were centrifuged at 700 g for 10 min at 4°C, and pellets suspended in lysis buffer plus protease inhibitors were used as nuclear fraction. The corresponding supernatants were centrifuged at 2000 g for 10 min at 4°C, and pellets suspended in lysis buffer containing protease inhibitor were used as the mitochondrial compartment. The supernatants from the mitochondrial isolation process were used as the cytosolic compartment. The various subcellular fractions were quantified using a BCA protein assay and used for subsequent analysis.

Membrane fractionation

For membrane fractionation experiments (37), cultured PC3 cells were harvested and suspended in three times the cell pellet volume of B1 buffer containing 20 mM HEPES, pH 7.2, 400 mM sucrose, and 1 mM EDTA. Cells were homogenized using a 22 G needle. To collect the sedimented membranes, cells were subjected to sequential differential centrifugation at 3,000 g for 10 min, 25,000g for 20 min, and 100,000 g for 30 min. Each membrane pellet from the different centrifugation steps was normalized by the amount of phosphatidylcholine (PC) using a phosphatidylcholine colorimetric quantification assay (Cayman) and normalized for protein analysis.

For further fractionation steps, a sucrose and OptiPrep^™ (Millipore sigma) gradient density centrifugation protocol was used. Briefly, the pellets obtained after the 25,000 g centrifugation step were mixed with 0.75 ml of 1.25 M sucrose in 10 mM Tris-HCl buffer (pH 7.4). Sequentially, 0.5 ml of 1.1 M sucrose and 0.5 ml of 0.25 M sucrose were overlaid and centrifuged at 120,000 g for 2 h. The interface fraction between 0.25 M and 1.1 M (L fraction) and the pellet (P fraction) were collected. The L fraction was further suspended in 0.5 ml 19% OptiPrep solution containing 20 mM Tricine-KOH, pH 7.4, 250 mM sucrose, and 1 mM EDTA. An OptiPrep gradient solution containing 0.25 ml of 22.5%, 0.5 ml of suspended sample, 0.45 ml of 16%, 0.45 ml of 12%, 0.5 ml of 8%, 0.25 ml of 5%, and 0.1 ml of 0% OptiPrep solution was generated and centrifuged at 150,000 g for 3 h. Subsequently, equal amounts of 6 fractions from the mixture were collected, diluted with B88 buffer (20 mM HEPES-KOH, pH 7.2, 250 mM sorbitol, 150 mM potassium acetate,5 mM magnesium acetate) and the pellets were collected by centrifugation at 100,000 g for 1 h. Pellets were suspended with lysis buffer containing protease inhibitor cocktail and phosphatase inhibitor cocktail for protein analysis. Phosphatidylcholine concentration of each fraction was quantified and normalized for further experiments.

Animal studies

C57BL/6J (B6WT) and IFNAR1 KO mice were purchased from Jackson laboratory. Colonies were maintained and genotyped as recommended by the supplier. Experiments were carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (NIH), and protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of The Wistar Institute. Sample size was determined by power analysis and all animals were included in the analysis. MPTEN1 cells (1x10⁶) engineered to express Doxy-regulated PRKN expression and transduced with shCtrl or shHMGB1 were injected s.c. in the right flank of C57BL/6J mice in a 1:1 mixture of PBS:Matrigel. When superficial flank tumors reached a volume of 120–150 mm³, groups of animals were administered Doxy (500 ng/ml) in the drinking water and tumor burden was quantified with a caliper at increasing time intervals. At the end of the experiment, tumors were collected and processed for immunohistochemistry and flow cytometry.

Statistics

Statistical analysis was carried out using GraphPad Prism 10 software (RRID: SCR_002798). All biological experiments were repeated at least three times, and each experiment was performed in technical duplicates. For the analysis of two separate groups, a Mann-Whitney two-tailed t test with 95% confidence was used. For analysis of more than three groups, a one-way ANOVA test was used.

Data Availability

All data presented in this manuscript are available upon request from the corresponding author.

RESULTS

PRKN ubiquitination of HMGB proteins.

To identify novel targets of PRKN ubiquitination independently of mitophagy, we re-examined a recent SILAC proteomics screen that was carried out in prostate cancer PC3 cells in the absence of mitochondria-damaging stimuli (17). This analysis identified High-Mobility Group Box 1 (HMGB1) residues Lys146 (K146, Figure 1A) and Lys112 (K112, Supplementary Figure 1A) as sites of PRKN ubiquitination, in vivo. In addition, a homologous Lys147 (K147) was identified as a site of PRKN ubiquitination in HMGB2 (Supplementary Figure 1B). In the multidomain organization of HMGB1, K112 and K146 are contained in the B-Box DNA binding region.

Figure 1. PRKN-HMGB1 ubiquitination.

(A) MS/MS spectra of ubiquitination peptides of HMGB1 in PC3 cells transfected with PRKN (light label) or vector (heavy label) identified by SILAC-based ubiquitin remnant motif enrichment (K-ε-GG). The Lys146 (K146) ubiquitination site is in red. b- and y-ions are indicated by horizontal bars above and below the sequence, respectively. (B) PC3 cells transfected with Flag-HMGB1 and reconstituted with Myc-PRKN were immunoprecipitated (IP) with IgG or an antibody to Flag and analyzed with antibodies to ubiquitin (Pan-Ub), Ub-K48- or Ub-K63 by Western blotting. (C and D) The conditions are as in (B) except that PC3 cells expressing Flag-HMGB1 were reconstituted with WT PRKN or PRKN C431S mutant (C) or WT HMGB1 or HMGB1 K112A or K146A mutant (D), followed by IP with IgG or an antibody to Flag and Western blotting. For panels B-D, the position of ubiquitinated HMGB1 species is indicated by a side bracket. (E) PC3 cells expressing vector or PRKN were analyzed for co-association with endogenous HMGB1 by Duolink proximity ligation assay (PLA) and fluorescence microscopy. The number of co-associated spots per condition is indicated (mean±SD, p<0.0001). Scale bar, 30 μm. Inset, magnification of indicated areas. (F and G) PC3 cells separately expressing Flag-HMGB1 or GFP-PRKN were analyzed by Western blotting (F) and IP with an antibody to Flag followed by by Western blotting (G). (H and I) PC3 cells expressing vector or PRKN were analyzed for co-localization with TOM20 and HMGB1 by confocal microscopy (H) and a Pearson Correlation Coefficient (PCC) of protein colocalization was quantified (I). Bottom, quantification of signal intensity. Representative images. Scale bar, 20 μm. A p value by two-tailed unpaired t test is indicated. (J and K) Subcellular fractions containing nuclei (Nucl), cytosol (Cyto) or mitochondria (Mito) were isolated from PC3 cells transiently (J) or conditionally (Doxy-induced) (K) expressing PRKN and analyzed by Western blotting. Bottom, densitometric quantification of HMGB1 protein bands.

To characterize PRKN ubiquitination of HMGB proteins, we reconstituted PC3 cells with Flag-HMGB1 or Flag-HMGB2 together with Myc-PRKN (Supplementary Figure 1C). In these experiments, immune complexes of Flag-HMGB1 (Figure 1B) or Flag-HMGB2 (Supplementary Figure 1D) reacted with an antibody to ubiquitin (Ub), as well as Ub-K48 linkages, and, to a lesser extent, Ub-K63 linkages (Figure 1B, Supplementary Figure 1D). Immune complexes precipitated with IgG were unreactive (Figure 1B, Supplementary Figure 1D). To confirm the specificity of PRKN-dependent ubiquitination, we separately expressed Myc-PRKN or Flag-HMGB1 in PC3 cells (Supplementary Figure 1E). Similar to the data presented in Figure 1B, immune complexes precipitated with an antibody to Flag contained Ub bands of ~90 kDa, suggestive of poly-ubiquitination, and ~35 kDa, suggestive of mono-ubiquitination, only in the presence, but not in the absence of Myc PRKN (Supplementary Figure 1F). Finally, we examined another, independent murine prostate cancer cell type, TRAMP-C2, which was stably transduced to conditionally express PRKN in response to doxycycline (Doxy) treatment (Tet-ON system) (20). Immune complexes of Flag-HMGB1 precipitated from Doxy-treated TRAMP-C2 cells reacted with an antibody to Ub (Supplementary Figure 1G), whereas IgG immunoprecipitates or treatment with vehicle were unreactive (Supplementary Figure 1G).

To examine the requirement of PRKN E3 ligase activity in this response, we next expressed WT PRKN or PRKN E3 ligase-defective S65A or C431S mutant in PC3 cells (Supplementary Figure 1H). Consistent with the data above, HMGB1 precipitated from PC3 cells in the presence of WT PRKN reacted with an antibody to Ub or Ub-K48 (Figure 1C). Conversely, reconstitution of PC3 cells with PRKN C431S mutant did not result in HMGB1 ubiquitination (Figure 1C). As a complementary approach, we next mutated to Ala the predicted PRKN ubiquitination sites in HMGB1, K112A or K146A and expressed the various Flag-tagged recombinant proteins in PC3 cells (Supplementary Figure 1I). In these experiments, PRKN ubiquitinated WT HMGB1 as well as HMGB1 K112A mutant precipitated from PC3 cells (Figure 1D). In contrast, immune complexes containing HMGB1 K146A showed reduced reactivity with an antibody to Ub, Ub-K48 or Ub-K63 (Figure 1D). Similarly, PRKN ubiquitinated WT HMGB2 but not K147A HMGB2 mutant (Supplementary Figure 1D). Based on these results, we next focused on K146 as a major PRKN ubiquitination site on HMGB1.

Mitochondria-localized PRKN-HMGB1 complex.

We next asked if PRKN formed a complex with HMGB1 to enable K146 ubiquitination. First, recombinant PRKN and endogenous HMGB1 showed increased co-association in PC3 cells by proximity ligation assay, compared to control cultures (Figure 1E). As an independent approach, we separately expressed GFP-PRKN or Flag-HMGB1 in PC3 cells and carried out co-immunoprecipitation experiments (Figure 1F). Here, immune complexes of Flag-HMGB1 contained GFP-PRKN, whereas single transfection conditions or immunoprecipitation with non-binding IgG had no effect (Figure 1G). = We next studied a potential subcellular localization(s) of a PRKN-HMGB1 complex. We found that endogenous HMGB1 co-localized with mitochondria-associated TOM20 in PRKN-expressing PC3 cells, by confocal microscopy (Figure 1H and I). In contrast, HMGB1 did not co-localize with TOM20 in the presence of vector (Figure 1H and I). To characterize a potential redistribution of the various pools of HMGB1 in the presence of PRKN, subcellular fractionation experiments were next carried out. Here, transient (Figure 1J) or conditional (Doxy-induced) (Figure 1K) expression of PRKN induced comparable redistribution of the nuclear and cytosolic pools of HMGB1 to mitochondria, co-localized with PRKN (Figure 1J and K).

Predicted structure of a PRKN-HMGB1 complex.

By molecular modeling, an interaction matrix across 30 docking poses identified an extensive network of residues potentially involved in a PRKN-HMGB1 complex (Figure 2A). This analysis identified the docking pose 13 of HMGB1 with a mean Tanimoto similarity of 0.1412, a maximum similarity of 1.0, and a minimum similarity of 0.0 as a candidate for PRKN interaction. Accordingly, a predicted PRKN-HMGB1 (pose 13) binding interface consists of 66 amino acids (Figure 2B), with 35 residues mostly from the PRKN in-between-RING (IBR) domain (His279, Pro281, Gln282, Cys323, Gln326, Met327, Gly328, Gly329, Leu342, Pro343, Glu344, Pro345, Asp346, Cys347, Arg348, Lys349, Val350, Thr351, Glu353, Gly354, Gly355, Gly357, Leu358, Arg366, Glu367, Lys369, His373, Gly375, Glu376, Cys377, Ser378, Ala379, Val380, Phe381, and Glu382) and 31 residues mostly localized in the NH₂ A-box of HMGB1 (Met1, Arg10, Gly11, Met13, Ala17, Gly2, Val20, Gln21, Arg24, Glu25, Lys28, Lys29, Lys3, Lys30, His31, Pro32, Asp33, Ala34, Ser35, Val36, Phe38, Gly4, Phe41, Ser42, Pro6, Ala66, Ala69, Arg70, Arg73, Glu74, and Pro9). By molecular dynamics (MD) simulations (200 ns), a predicted PRKN-HMGB1 complex showed structural integrity with a root mean square deviation (RMSD) of 3.2 Å-6.4 Å (Figure 2C).

Figure 2. PRKN-HMGB1 complex.

(A) Predicted PRKN-HMGB1 contact interaction matrix across 30 docking poses. Color coding, residue sequence numbers. Only interacting residues are shown. (B) Predicted PRKN (green)-HMGB1 (grey) binding interface comprising 31 residues from the A-Box of HMGB1 and 35 residues from the PRKN IBR domain. (C) RMSD plot obtained during 200 ns molecular dynamics (MD) simulations. (D) Predicted HDOCK alignment of K146 of HMGB1 and Ub active site residues, R74 and G76. (E and F) Predicted HDOCK model of WT HMGB1 (E) or K146A mutant HMGB1 (F) binding to PRKN. (G and H) PC3 cells expressing PRKN and reconstituted with WT HMGB1 or K146A HMGB1 mutant were analyzed by confocal fluorescence microscopy (G) and a PCC of PRKN co-localization with WT or mutant HMGB1was quantified (H). Scale bar, 10 μm. Representative images. A p value by two-tailed unpaired t test is indicated. (I and J) PC3 cells expressing Flag-HMGB1 or Flag-HMGB1 K146A mutant were reconstituted with Myc-PRKN, IP with IgG or an antibody to Flag (I) or Myc (J) and immune complexes were analyzed by Western blotting. Bar graphs, densitometric quantification of WT HMGB1 (WT) or HMGB1 K146A mutant (Mut) protein band. (K) Mitochondrial fractions (Mito) of PC3 cells expressing Flag-HMGB1 or Flag-HMGB1 K146A mutant and Myc-PRKN were analyzed by Western blotting. Bar graph, densitometric quantification of WT HMGB1 (WT) or HMGB1 K146A mutant (Mut) protein band.

We next integrated these molecular simulations with protein complex predictions using the HDOCK server. This analysis suggested close alignment between HMGB1 K146 and Ub active site residues R74 and G76, implicated in Ub activation and substrate recognition, respectively (Figure 2D). Under these conditions, mutagenesis of K146 was predicted to introduce extensive conformational changes in HMGB1 that interfered with its ability to bind PRKN, as suggested by higher ligand RMSD for HMGB1 K146A mutant (96.5 Å) and less favorable docking score (-221.56), compared to WT HMGB1 (RMSD, 60.41 Å; docking score, -235.35) (Figure 2E and F). Consistent with this prediction, HMGB1 K146A mutant expressed in PC3 cells showed reduced co-association with Myc-PRKN, compared to WT HMGB1, by confocal microscopy (Figure 2G and H). Similarly, immune complexes of Flag-HMGB1 K146A (Figure 2I), or, reciprocally, Myc-PRKN (Figure 2J) precipitated from PC3 cells showed reduced co-association with PRKN or HMGB1 K146A mutant, respectively. Finally, mitochondrial extracts from PRKN-expressing PC3 cells contained reduced levels of HMGB1 K146A mutant, compared to WT HMGB1 (Figure 2K).

PRKN regulates HMGB1 extracellular release.

In cycloheximide (CHX) block and release experiments, expression of WT PRKN in PC3 cells did not affect the half-life (t½ ~8.5 h) of HMGB1 (Supplementary Figure 2A) or HMGB2 (Supplementary Figure 2B), compared to control cultures. In contrast, transient or Doxy-induced expression of PRKN resulted in nearly complete depletion of HMGB1 from PC3 cell extracts (Supplementary Figure 2C), in agreement with recent observations (20). Conversely, PRKN expression in PC3 cells did not significantly affect the total cellular content of HMGB1 K146A mutant, compared to vector (Figure 3A). Concomitantly with depletion in total cell extracts, PRKN expression in PC3 cells was associated with increased release of HMGB1 in the conditioned medium (CM) (20), whereas the extracellular release of HMGB1 K146A mutant was significantly reduced (Figure 3B).

Figure 3. PRKN-mediated extracellular release of HMGB1.

(A) PC3 cells expressing PRKN were reconstituted with Flag-HMGB1 (WT) or Flag-HMGB1 K146A mutant (Mut) and analyzed by Western blotting. TCE, total cell extracts. *, nonspecific band. (B) PC3 cells expressing PRKN were transfected with HMGB1-directed siRNA (siHMGB1) and reconstituted with Flag-HMGB1 or Flag-HMGB1 K146A mutant (Mut) and analyzed by Western blotting. CM, conditioned medium. Both endogenous HMGB1 and Flag-HMGB1 bands are indicated. (C-E) PC3 cells expressing vector or PRKN were transfected with ATG5- (C), FIP200- (D) or Rab27a (E)-directed siRNA (siATG5, siFIP200, siRab27a) and aliquots of TCE or CM were analyzed by Western blotting. Bottom, densitometric quantification of HMGB1 protein bands in the CM under the various conditions. (F) PC3 cells expressing vector or PRKN were analyzed for co-localization with endogenous LC3 and HMGB1 by confocal fluorescence microscopy. Scale bar, 10 μm. Representative fields are shown. A PCC is indicated (mean±SD, p<0.0001). Bottom, quantification of signal intensity. (G and H) PC3 cells expressing vector (V) or PRKN (P) were analyzed for total LC3 pucta (G) or LC3-TOM20 colocalization (H) by fluorescence microscopy. For (H), a PCC was quantified. p values by two-tailed unpaired t test are indicated. (I) PC3 cells expressing PRKN or PRKN C431S mutant were analyzed with an antibody to HMGB1 by immuno-gold transmission electron microscopy (TEM). Arrows, localization of gold particles. Scale bars, 500 nm. Representative images. (J) The experimental conditions are as in (J) and the total number of membranous structures/mitochondria (n=533) positive or negative for HMGB1 labeling by TEM was quantified. Total number of cells, n=16; total number of images, n=56; average number of particles/labeled mitochondria, n=3.4.

Although the extracellular release of HMGB1 has been linked to discharge from damaged and dying cells, we found that expression of PRKN or PRKN C431S mutant did not affect key hallmarks of cellular damage and immunogenic cell death (ICD), including mitochondrial inner membrane potential (Supplementary Figure 2D), membrane redistribution of calreticulin (Supplementary Figure 2E) or release of extracellular ATP (Supplementary Figure 2F). Consistent with this, PRKN expression induced a negligible increase in apoptosis (~10%) in PC3 cells, compared to vector transfectants (Supplementary Figure 2G). Conversely, PRKN expression was associated with upregulation of autophagy markers in PC3 cells, including phosphorylation of ULK1 (Ser555 and Ser757), Raptor (Ser792), Beclin-1 (Ser93) and AMPK (Thr172) (Supplementary Figure 2H).

Based on these results, we next asked whether activation of autophagy was required for the extracellular release of HMGB1 mediated by PRKN. In these experiments, siRNA silencing of upstream core autophagy components ATG5 (38) (Figure 3C) or FIP200 (39) (Figure 3D) abolished PRKN-induced release of HMGB1 in the CM of PC3 cells. Similar results were obtained with 3-methyladenine (3MA), a pharmacologic inhibitor of autophagosome formation, which also suppressed PRKN release of HMGB1 (Supplementary Figure 2I). Consistent with these data, silencing of Rab27a, which has been implicated in autophagosome fusion with the plasma membrane during secretory autophagy (40), significantly reduced the amount of HMGB1 released in the CM of PRKN-expressing PC3 cells, compared to control siRNA (Figure 3E). Conversely, treatment with glycyl-L-phenylalanine 2-naphthylamide (GPN), which induces lysosome disruption (Supplementary Figure 2I) or Bafilomycin A (BafA), an antagonist of autophagosome-lysosome fusion (Supplementary Figure 2J) did not affect PRKN extracellular release of HMGB1.

Characterization of membranous constituents implicated in PRKN release of HMGB1.

Next, we further characterized the secretory pathway(s) implicated in PRKN HMGB1 extracellular release. First, we observed that in the presence of PRKN, endogenous HMGB1 colocalized with autophagy-associated LC3, compared to vector control, by confocal microscopy (Figure 3F). Although the total amount of LC3 puncta per cell was not appreciably different in control or PRKN-expressing cells (Figure 3G), PRKN expression significantly increased the co-localization of LC3 puncta with mitochondrial TOM20, compared to vector control (Figure 3H). In addition, PRKN expression was associated with extensive immuno-gold labeling for HMGB1 of membrane-containing subcellular structures, corresponding morphologically to mitochondria (Figure 3I and J, Supplementary Figure 3A). These subcellular membranous structures/mitochondria also reacted with an antibody to autophagy-associated, LC3-II (Supplementary Figure 3B), whereas expression of PRKN C431S mutant considerably reduced immuno-gold labeling for HMGB1 (Supplementary Figure 3A) or LC3-II (Supplementary Figure 3B).

We next used a biochemical approach to independently complement these data and quantitatively characterize the membranous structures implicated in PRKN-HMGB1 extracellular release. First, we carried out a three-step membrane fractionation protocol to examine a potential vesicle-associated mechanism of HMGB1 secretion (Supplementary Figure 4A). In this analysis, PRKN co-associated with HMGB1, LC3 and mitochondrial TOM20 in a 25K membrane fraction, compared to vector (Supplementary Figure 4B and C). This 25K membrane fraction was also positive for nuclear (Lamin A/C), cytosolic (Clathrin) and lysosomal (LAMP1) markers but devoid of Golgi proteins (GM130) (Supplementary Figure 4B and C). Although PRKN was also present in 3K and 100K membrane fractions, these contained low (3K) to undetectable (100K) levels of HMGB1 (Supplementary Figure 4B and C). Next, we used sequential sucrose gradient centrifugation and OptiPrep fractionation to isolate all membrane-containing vesicles produced by PRKN-expressing PC3 cells (Figure 4A). Analysis of these isolated fractions demonstrated that PRKN was heterogeneously distributed in membranous vesicles of different density but was invariably co-associated with LC3 (Figure 4B and C, top). Specifically, F2 and F3 membranous vesicles (density, 1.099–1.108) contained varying levels of PRKN co-associated with LC3, calnexin, TOM20 and high levels of HMGB1 (Figure 4B and C, top).

Figure 4. Characterization of membranous constituents in PRKN-HMGB1 extracellular release.

(A) Schematic diagram of sequential sucrose- and OptiPrep-gradient density isolation of total cellular vesicles from PRKN-expressing PC3 cells. Created in BioRender. Altieri, D. (2025) https://BioRender.com/9w6x2fw (B) The conditions are as in (A) and the indicated gradient density fractions (F1-F7) of total cellular vesicles were analyzed by Western blotting. (C) Densitometric quantification of the indicated protein bands in OptiPrep gradient density fractions isolated from total cellular vesicles (top) or large extracellular vesicles (LEV) (bottom). Representative experiment out of three independent determinations. (D) ZetaView distribution of LEV produced by PC3 cells expressing vector or PRKN. The median LEV size per condition is indicated. (E) LEV isolated from PC3 cells expressing vector or PRKN as in (D) were fractionated by gradient density ultracentrifugation followed by Western blotting. (F) LEV isolated from PRKN-expressing PC3 cells were analyzed for co-localization of PRKN and HMGB1 by confocal fluorescence microscopy with quantification of individual LEV size. Scale bar, 1 μm. A representative microscopy field is shown. Right, quantification of signal intensity. (G) LEV isolated from PC3 cells expressing WT or K146A HMGB1 mutant were analyzed by Western blotting. (H) LEV or small extracellular vesicles (SEV) isolated from PC3 cells expressing vector or PRKN were analyzed by Western blotting.

PRKN-HMGB1 release in large extracellular vesicles.

Based on these data, we next asked if the membranous constituents implicated in PRKN HMGB1 release constituted a novel population of extracellular vesicles (EV). Adhering to international guidelines for EV studies (41), we found that transient or Doxy-induced conditional expression of PRKN in PC3 cells was associated with the release of both small EV (SEV, average diameter of ~150 nm) as well as large EV (LEV, average diameter of ~250 nm), by ZetaView quantification (Supplementary Figure 4D). Comparable results of SEV and LEV particle size were obtained with an independent approach by flow cytometry, with no significant differences between vector or PRKN-expressing PC3 cells (Supplementary Figure 4E). Under these conditions, gradient density fractionation of LEV released by control or PRKN-expressing cells (Figure 4D) identified F4-F5 fractions (density, 1.134–1.151 g/ml) that contained co-associated PRKN, HMGB1, calnexin, TOM20 and LC3 (Figure 4E and Figure 4C, bottom). Conversely, an F6 fraction (density, 1.172 g/ml) still contained high levels of PRKN co-associated with LC3, TOM20 and CLNX, but was devoid of HMGB1 (Figure 4E and Figure 4C, bottom). Similarly, PRKN co-localized with HMGB1 in LEV isolated from PRKN-expressing cells by confocal microscopy (Figure 4F, Supplementary Figure 4F), whereas LEV isolated from PC3 cells expressing vector did not contain PRKN (Supplementary Figure 4F). In these experiments, antibody reactivity for both HMGB1 and PRKN was observed with permeabilized LEV, but not non-permeabilized LEV (Supplementary Figure 4G), confirming the intra-LEV protein localization. Biochemically, LEV collected from PRKN-expressing PC3 cells reconstituted with HMGB1 K146A mutant contained TOM20 and LC3, but no HMGB1 (Figure 4G). Finally, in side-by-side comparison, LEV isolated from PRKN-expressing cells contained PRKN, HMGB1, calnexin and TOM20, whereas SEV isolated under the same conditions contained PRKN and TSG101, but were entirely devoid of calnexin, HMGB1 and TOM20 (Figure 4H).

PRKN-HMGB1 LEV stimulation of an IFN response.

Next, we hypothesized that a regulated release of HMGB1 through PRKN LEV secretion could stimulate immune-inflammatory responses. Consistent with recent observations (20), co-incubation of CM from PRKN-expressing PC3 cells with parental, PRKN-negative recipient cells stimulated IFN gene expression (Figure 5A). Conversely, CM from PC3 cells expressing PRKN C341S mutant had limited effect (Figure 5B). By multiplex Luminex quantification, the CM of PRKN-expressing PC3 cells contained high levels of several immune-inflammatory mediators, including chemoattractants, IFN-inducible T cell α chemoattractant (ITAC), MIP1α, MIP1β, and fractalkine, B (IL5) and T (IFNγ, IL2) cell activating cytokines, and mediators of acute phase/inflammatory responses (IL6, IL1β, IL13) (Figure 5C).

Figure 5. PRKN extracellular release of HMGB1 stimulates IFN gene expression.

(A and B) Aliquots of CM from PC3 cells reconstituted with WT PRKN (A) or PRKN C431S mutant (B) were co-incubated with PRKN-negative recipient PC3 cells followed by analysis of IFN gene expression by RT-qPCR (mean±SD, n=3). (C) The conditions are as in (A) and aliquots of CM from PC3 cells expressing vector or PRKN were analyzed by multiplex Luminex assay for the indicated cytokines/chemokines. Right, densitometric quantification of band intensity (mean±SD, n=3). (D and E) PC3 cells expressing PRKN were silenced for endogenous HMGB1 and reconstituted with WT HBGB1 or HMGB1 K146A mutant followed by Western blotting (D) or analysis of IFN gene expression by RT-qPCR (E) (mean±SD, n=3). (F) Aliquots of CM from PC3 cells reconstituted with WT or K146A mutant HMGB1 as in (D and E) were incubated with parental PC3 cells followed by analysis of IFN gene expression by RT-qPCR. Mean±SD (n=3). (G) LEV isolated from PRKN-expressing cells were labeled with DiD, incubated with recipient PC3 cells and analyzed by fluorescence microscopy. Scale bar, 10 μm. Representative merged images are shown. The concentrations of LEV added per condition are indicated. (H) LEV (left) or SEV (right) isolated from PC3 cells expressing vector or PRKN were co-incubated with recipient PC3 cells followed by analysis of IFN gene expression by RT-qPCR. Mean±SD (n=6). (I) LEV isolated from PC3 cells expressing PRKN or PRKN C431S mutant (Mut) were incubated with recipient PC3 cells followed by analysis of IFN gene expression. For panels A, B and F, numbers are p values by unpaired two-tailed t test. For panels H and I, ***, p<0.0001 by two-tailed unpaired t test.

Next, we carried out two sets of experiments to test the requirement of HMGB1 and PRKN ubiquitination in this response. First, siRNA silencing of endogenous HMGB1 (Figure 5D) abolished PRKN-induced IFN gene expression (Figure 5E). Conversely, reconstitution of HMGB1-silenced PC3 cells with WT HMGB1 restored IFN gene expression, whereas reconstitution with HMGB1 K146A mutant had no effect (Figure 5E). Similar results were obtained with the CM of PRKN-expressing PC3 cells. In these experiments, co-incubation of recipient cultures with CM from PC3 cells reconstituted with WT HMGB1 induced IFN gene expression, whereas CM from PC3 cells expressing HMGB1 K146A mutant had no effect (Figure 5F). Finally, we examined the specificity of HMGB1 in this response, and exposed PRKN-negative PC3 cells to increasing concentrations of soluble, recombinant HMGB1. Co-culture with recombinant HMGB1 induced concentration-dependent IFN gene expression in recipient cultures (Supplementary Figure 5A). Mechanistically, siRNA silencing of a key HMGB1 receptor on recipient PC3 cells, Receptor for Advanced Glycation End Products (RAGE) (Supplementary Figure 5B) abolished recombinant HMGB1-induced IFN response (Supplementary Figure 5C).

Next, we asked whether HMGB1 packaged in PRKN-associated LEV (Supplementary Figure 5D) directly stimulated immune-inflammatory responses in recipient cells. First, DiD-labeled LEV isolated from PRKN-expressing cells were readily taken up by parental PC3 cells in a concentration-dependent manner, by fluorescence microscopy (Figure 5G, Supplementary Figure 5E). Under these conditions, PRKN-derived LEV potently stimulated IFN gene expression in parental PC3 cells (Figure 5H, left). Conversely, LEV isolated from PC3 cells expressing PRKN C431S mutant had no effect (Figure 5I). Consistent with the specificity of LEV in this response, SEV produced by PRKN-expressing PC3 cells did not stimulate IFN gene expression in recipient PC3 cells (Figure 5H, right).

PRKN HMGB1-mediated CD8 T cell activation.

Co-incubation with CM harvested from Doxy-induced PRKN-expressing TRAMP-C2 cells significantly reduced the fraction of naïve CD8 T splenocytes from syngeneic C57BL/6J mice (Figure 6A). This response was associated with increased expression of effector memory (EM) markers, upregulation of the TCF1 transcription factor (Figure 6A and B), which critically regulates T cell stemness and response therapeutic immune checkpoint inhibitors (42), and expansion of double-positive CD8 T cell subsets expressing implicated in effector functions (CD69⁺/KLRG1⁺), self-renewal (PD1⁺/TCF1⁺) and antitumor cytotoxicity (KLRG1⁺/GrzB⁺) (Figure 6A and B). These cells also exhibited lower expression of the immune inhibitory receptor, TIM-3, which participates in T cell exhaustion (43), whereas the levels of PD-1 and LAG3 or Ki67-associated cell proliferation, were not significantly affected (Supplementary Figure 6A). Under these conditions, silencing of HMGB1 in PRKN-expressing TRAMP-C2 cells abolished CD8 T cell activation and suppressed the expansion of effector and cytotoxic subsets (Figure 6A and B, Supplementary Figure 6A). Conversely, reconstitution of these cells with WT HMGB1, but not HMGB1 K146A mutant restored CD8 T cell activation induced by PRKN CM (Figure 6A and B, Supplementary Figure 6A).

Figure 6. Parkin-HMGB1 stimulation of T cell activation.

(A) CM from TRAMP-C2 cells conditionally expressing PRKN (Doxy) were reconstituted with HMGB1 or HMGB1 K146A mutant (Mut), incubated with syngeneic CD8 splenocytes from C57BL/6J mice and analyzed for the indicated markers by multiparametric flow cytometry. The percentage of cells is indicated. EM, effector memory. (B) The conditions are as in (A) and syngeneic CD8 splenocytes incubated with TRAMP-C2 CM were analyzed for expression of TCF1 and double positive TCF1/PD-1 and KLRG1/GrzB subsets, by flow cytometry. (C) The conditions are as in (B) and CD4 T cells harvested from splenocytes of C57BL/6J mice were incubated with CM from PRKN-expressing cells and analyzed for the indicated markers by multiparametric flow cytometry. (D and E) Aliquots of CM from reconstituted TRAMP-C2 cells as in (A) were incubated with CD8 (D) or CD4 (E) T cells from syngeneic IFNAR1^−/− mice and modulation of the indicated markers was quantified by multiparametric flow cytometry. For panels B-E, each point corresponds to an individual mouse determination (n=6). Numbers are p values by two-way ANOVA. ns, not significant.

To test the specificity of these findings, we next examined the response of other syngeneic immune subsets to PRKN-HMGB1 CM. Co-culture with PRKN CM reduced the fraction of naïve CD4 T cells, increased the expression of TCF1 and expanded the double positive PD-1⁺/TCF1⁺ subset, whereas changes in the CD69⁺/KLRG1⁺ subpopulation did not reach statistical significance (Figure 6C). The levels of immune inhibitory receptors, TIM3, LAG-3 and PD-1 or Ki67 were not affected (Supplementary Figure 6B). Furthermore, PRKN CM did not modulate activation markers of NK and B cell subsets isolated from WT splenocytes (Supplementary Figure 6C). Similar to the results obtained with CD8 T cells (Figure 6A and B), silencing of HMGB1 abolished CD4 T cell activation induced by PRKN CM, in a response rescued by reconstitution of PRKN-expressing cells with WT HMGB1, but not HMGB1 K146A mutant (Supplementary Figure 6B). Finally, HMGB1 silencing with or without reconstitution with WT or mutant HMGB1 had no effect on NK or B cell subsets in the presence of PRKN CM (Supplementary Figure 6C).

To test the role of membrane IFN α and β receptor subunit-1 (IFNAR1) in PRKN-HMGB1 T cell activation, we next isolated splenocytes from IFNAR1^−/− mice. Deletion of IFNAR1 on recipient splenocytes abolished PRKN-HMGB1-induced CD8 (Figure 6D and E, Supplementary Figure 6D) as well as CD4 (Figure 6E, Supplementary Figure 6E) T cell activation and suppressed the expansion of effector or cytotoxic T cell subsets (Figure 6D, Supplementary Figure 6D and E). HMGB1 silencing in PRKN-expressing cells with or without reconstitution with WT HMGB1 or HMGB1 K146A mutant had no effect on the expression of activation markers in IFNAR1^−/− CD8 (Figure 6D), CD4 (Figure 6E), NK or B cells (Supplementary Figure 6F), irrespective of HMGB1 silencing and reconstitution with WT or K146A mutant (Supplementary Figure 6E and F).

PRKN-HMGB1 antitumor immunity, in vivo.

Finally, we examined the role of HMGB1 in PRKN antitumor immunity, in vivo. Engraftment of prostate cancer MPTEN1 cells stably expressing Doxy-regulated PRKN (20) onto the flanks of syngeneic C57BL/6J mice gave rise to exponentially growing flank tumors (Figure 7A). Addition of Doxy to the drinking water of tumor-bearing animals inhibited tumor growth, and induced tumor regression in certain animals (Figure 7A). Histologically, MPTEN1 tumors harvested from Doxy-administered animals contained high levels of as well as HMGB1, compared to control mice (Figure 7B). Consistent with the results of T cell activation, in vitro, intratumoral CD8 T cells from Doxy-treated mice exhibited increased expression of TCF1 and expansion of effector (CD69⁺/KLRG1⁺), self-renewing (PD-1⁺/TCF1⁺) and cytotoxic (KLRG1⁺/GrzB⁺) double positive subsets (Figure 7C). In contrast, intratumoral CD4 T cells exhibited more limited changes that did not reach statistical significance (Figure 7D). Also, the expression of other immune activation markers, LAG3, TIM3 and PD-1, or proliferation-associated Ki67 in intratumoral CD8 (Supplementary Figure 7A) or CD4 (Supplementary Figure 7B), was not significantly affected in vehicle or Doxy-treated animals. Changes in the activation profile of immune subsets were restricted to the tumor microenvironment as CD8 or CD4 splenocytes harvested from tumor-bearing animals did not show modulation of activation markers with or without Doxy administration (Supplementary Figure 7C and D).

Figure 7. Parkin-HMGB1 antitumor immunity.

(A) Prostate cancer MPTEN1 cells transduced with Doxy-regulated conditional PRKN expression were engrafted onto the flanks of C57BL/6 mice and tumor growth was quantified with a caliper at the indicated time intervals. Mice were administered vehicle or Doxy in the drinking water (500 ng/ml) when tumors reached a volume of ~120–150 mm³ (arrow). Each line is an individual tumor. (B) MPTEN1 tumor samples harvested from vehicle (Veh) or Doxy-treated mice as in (A) were analyzed for expression of PRKN (top), HMGB1 (middle) or HMGB1 in the presence of control shRNA or HMGB1-directed shRNA (bottom) by immunohistochemistry. Representative images. Scale bar, 50 μm. (C and D) Intratumoral CD8 (C) or CD4 (D) T cell subsets were harvested from mice as in (A) and analyzed for the indicated markers by multiparametric flow cytometry. Each symbol is an individual mouse determination. Numbers correspond to p values by two-way ANOVA. ns, not significant. (E) MPTEN1 cells expressing Doxy-regulated PRKN were stably transduced with control shRNA (shCtrl) or HMGB1-directed shRNA (shHMGB1) and cells harvested after three passages (Psg) were analyzed with or without Doxy by Western blotting. (F) MPTEN1 cells as in (E) were engrafted onto the flanks of immunocompetent C57BL/6 mice and tumor growth was assessed with a caliper at the indicated time intervals. All animals in both groups were administered Doxy in the drinking water when tumors reached ~120–150 mm³. Each point is a single animal determination. (G and H) The conditions are as in (F) and intratumoral CD8 (G) or CD4 (H) T cells were analyzed for the indicated markers by multiparametric flow cytometry. Each symbol is an individual mouse determination. Numbers correspond to p values by two-way ANOVA. ns, not significant. For panels G and H, each point is an individual determination. Numbers are p value by two-way ANOVA. ns, not significant.

To test a mechanistic role of HMGB1 released by PRKN-expressing cells in antitumor immunity, we next generated independent clones of MPTEN1 cells carrying stable shRNA silencing of endogenous HMGB1 (Figure 7E). When examined in vivo, stable depletion of HMGB1 (Figure 7B, bottom) reversed the inhibition of MPTEN1 tumor growth in Doxy-administered mice (Figure 7F). Instead, but consistent with the data above, administration of Doxy suppressed tumor growth in animals engrafted with MPTEN1 transduced with control shRNA (shCtrl) (Figure 7F). Under these conditions, deletion of HMGB1 prevented the activation of intratumoral CD8 T cells, suppressing the expansion of double positive subsets implicated in effector (CD69⁺/KLRG1⁺), self-renewing (PD-1⁺/TCF1⁺) and cytotoxic (KLRG1⁺/GrzB⁺) functions, compared to shCtrl (Figure 7G). A similar trend in reduced levels of double-positive intratumoral CD4 cell subsets under conditions of HMGB1 silencing did not reach statistical significance (Figure 7H).

DISCUSSION

In this study, we have shown that PRKN controls a novel, Ub-dependent mechanism of regulated HMGB1 extracellular release. This pathway depends on PRKN ubiquitination of K146 of HMGB1, formation of a PRKN-HMGB1-Ub complex at mitochondria and HMGB1 loading onto novel autophagy- and mitochondria-containing large extracellular vesicles (LEV) for extracellular release. In turn, PRKN-HMGB1 LEV potently modulate an immune tumor microenvironment, activating cytokine and IFN signaling and selectively expanding specialized CD8 T cell subsets with effector, self-renewing and cytotoxic properties.

Recruited to mitochondria following loss of inner membrane potential, PRKN has been linked to organelle quality control by eliminating subpar and defective mitochondria via a Ub-dependent autophagy-lysosome machinery (44), i.e. mitophagy (45). Defects in this pathway have been linked to neurodegeneration (46), but it is now clear that PRKN is a multifunctional E3 Ub ligase, affecting plethora of cellular responses. These are especially important in cancer, where PRKN has emerged as a nearly ubiquitous tumor suppressor, lost in most malignancies by mutagenesis (15) or methylation silencing (20). Whether there is a role for mitophagy in PRKN tumor suppression is not known. Reciprocally, there is evidence that mitophagy may favor tumor progression (47) and chemotherapy resistance (48), whereas E3 active PRKN can be recruited to mitochondria in the absence of mitochondrial damage (17). Together, this suggests the existence of mitophagy-independent PRKN functions potentially linked to tumor suppression.

In this context, the landscape of PRKN ubiquitination in cancer is unique, and encompasses novel protein targets such as the alarmins, HMGB1 and HMGB2 (20). Here, PRKN ubiquitinated K146 in the B-box of HMGB1, or the homologous K147 in HMGB2, through K48, and, minimally, K63 Ub linkages. In turn, this resulted in the formation of a mitochondria-associated PRKN-HMGB1-Ub complex, predicted to involve the A-box of HMGB1, a domain that negatively regulates inflammation (49) and the PRKN IBR region (50), a zinc-dependent folding unit that recruits E2 proteins, UbcH7 or UbcH8 (51). More work is needed to elucidate the requirements of a PRKN-HMGB1 complex and a potential role of K146 in this association. However, in the predicted structure, K146 of HMGB1 is juxtaposed to Ub active site residues G76 and R74 that play a key role in substrate recognition (52) and protein-protein interactions (53). Together, this suggests that mitochondria-associated active PRKN binds and ubiquitinates HMGB1 on K146, leading to the formation of a PRKN-HMGB1 complex, stabilized by Ub association. Such model is consistent with other data of PRKN Ub-dependent protein-protein interactions (33) and with the observation that an HMGB1 K146A mutant has reduced ability to bind PRKN and localize to mitochondria (this study).

Instead of proteasomal destruction as seen with other PRKN substrates (17,18,54), PRKN ubiquitination of HMGB1 enabled a novel pathway of regulated HMGB1 extracellular release. Although essential for its cytokine-like, inflammatory functions (5), the process of HMGB1 release has remained controversial (5), variously associated with passive discharge from damaged or dying cells (55) or, alternatively, through a mechanism of secretory autophagy (56,57), also invoked for other “leaderless” inflammatory proteins, such as IL-1β (58). Different from these scenarios, PRKN-induced HMGB1 release occurred in the absence of mitochondrial dysfunction, was not accompanied by significant cell death, and was specific for HMGB1, as other alarmin molecules associated with immunogenic cell death, such as extracellular ATP or membrane calreticulin were not affected. Instead, we found that a PRKN-HMGB-Ub complex was loaded onto a novel population of LEV, resulting in regulated extracellular release.

The detailed biogenesis of LEV in these settings remains to be elucidated. The data presented here suggest that LEV and not SEV are selectively implicated in HMGB1 release, and their membranous constituents are predominantly derived from mitochondria and autophagosomes, but not lysosomes. Although a role of autophagy in EV biogenesis and secretion (59) is established, the biochemical properties of PRKN-HMGB1 LEV suggest a different secretory pathway from other models of PRKN regulation of EV trafficking (60), secretory autophagy induced by lysosome inhibition (40) or endolysosome-dependent HMGB1 release (57). In addition, the participation of mitochondria in the biogenesis of PRKN-HMGB1 LEV and downstream signaling (see below) also differs from other mechanisms of mitochondria-derived vesicles (61) implicated in quality control to limit DAMP exposure (62) or in response to lysosomal impairment (63).

Once released in the extracellular space, PRKN-HMGB1 LEV were efficiently taken up by recipient cells and stimulated a potent IFN and cytokine response. This preferentially, if not selectively expanded CD8 T cell subsets with effector (CD69⁺/KLRG1⁺), cytotoxic (KLRG1⁺/GrzB⁺) and self-renewal/stemness (TCF1⁺/PD-1⁺) properties (20) in an IFNAR1-dependent mechanism. These specialized T cell subsets have been shown to play a key role in antitumor immunity, limiting the process of T cell exhaustion (64) and improving the response to therapeutic immune checkpoint inhibitors (42). There is also evidence for an important role of extracellular vesicles in immune reprogramming (65), with modulation of inflammation (66), cross presentation of tumor antigens (67) and T cell responses (68), and both HMGB1 and HMGB2 have been implicated in CD8 T cell activation (69) and self-renewal/stemness mechanisms (70), in agreement with the data presented here.

In line with these observations, PRKN-HMGB1 signaling reprogrammed an antitumor immune microenvironment, which was associated with potent inhibition ofsyngeneic tumor growth, in vivo. Conversely, HMGB1 deletion reversed antitumor immunity and prevented the emergence of specialized CD8 T cell subsets in the microenvironment. Together with other data that PRKN promotes DC antigen presentation for antitumor responses (71), these results reinforce a model of PRKN dual mode tumor suppression, shutting off intrinsic tumor traits of mitosis (16), metabolism (17,72,73) and cell motility (19), while also enabling Ub-dependent HMGB1 release, IFN and cytokine signaling and expansion of antitumor CD8 T cell subsets (this study and (20)).

In sum, given the ubiquitous loss of PRKN in cancer (17), the PRKN-HMGB1-LEV pathway described here likely functions as a tumor suppressor in multiple, genetically disparate malignancies in humans. Mechanistically, we speculate that a regulated, LEV-mediated pathway of HMGB1 extracellular release, as opposed to passive, unregulated discharge from dying cells (23), may explain its dichotomous functions, in vivo, enabling specific antitumor immunity (69) or, conversely, driving nonspecific inflammation and tissue damage (74), respectively. In this context, PRKN dual mode tumor suppression and stimulation of antitumor immunity may be therapeutically actionable in the clinic. At least in selected tumor types, methylation silencing of the PARK2 gene can be reversed by clinically approved demethylating therapy, restoring PRKN expression and activation of antitumorigenic IFN signaling and CD8 T cell expansion, in vivo (20).

Supplementary Material

This manuscript contains supplementary material and methods, supplementary figures 1–7, supplementary tables 1–4 and supplementary references.

STATEMENT OF SIGNIFICANCE.

Parkin ubiquitinates the alarmin molecule HMGB1 to enable its regulated release in large extracellular vesicles that activate interferon signaling, expand specialized CD8+ T cell subsets, and promote antitumor immunity.