Exosome isolation from hepatocyte culture media was performed as described previously (15). Briefly, cell culture media was centrifuged at 500 × g for 10 min, 16,500 × g for 20 min, followed by filtration through a 0.2-μm filter (Life Sciences). Exosomes were pelleted at 120,000 g for 70 min with Type 70.1 Ti rotor (Beckman). The exosomes were further washed once with phosphate buffered saline (PBS) and centrifuged at 120,000 g for 70 min, then resuspended in a small volume of PBS for NanoSight™ assessment or in lysis buffer (1:10 dilution, Cell Signaling Technology, #9803) for Western blot. Plasma exosomes were isolated using the total exosome isolation kit (Invitrogen) according to manufacturer's instructions. The pellet was resuspended in sample dilution buffer for ELISA or lysis buffer for Western blot.

Male C57BL/6J wild-type (WT) mice were purchased from Jackson Laboratory. GsdmD knockout (KO) mice were obtained from Dr. Vishva Dixit (Genetech). TLR-4 KO (16), caspase-11 KO (5), and GsdmD KO (17) mice on C57BL/6 background were bred in Dr. Billiar's lab. We also generated mice with selective Hmgb1 deletion (5) in either myeloid cells (HMGB1^f/f Lyz2-cre⁺) or hepatocytes (HMGB1^f/f Alb-cre⁺). All animals were housed or bred in the specific pathogen-free animal facility at the University of Pittsburgh School of Medicine and were kept under a 12-h dark/light cycle, fed standard chow ad libitum.

Mice were intraperitoneally (i.p.) injected with 5 mg/kg LPS for the time indicated in the experiments. Knockdown of Rab27a in vivo was performed as previously (18). Briefly, 1 × 10⁹ plaque forming unit (PFU) adenoviruses Rab27a shRNA (Vector Biolabs, Malvern, PA) were injected into the penile vein of mice anesthetized by isoflurane. Two days after virus injection, mice were injected i.p. with LPS or saline. For exosome release inhibition in vivo, GW4869 dissolved in dimethyl sulfoxide (DMSO) (0.005%) was pre-injected into the penile vein at one dose of 2.5 mg/kg 1 h before LPS treatment.

Cells were isolated from mice by an in situ collagenase (type VI; Sigma) perfusion technique, modified as described previously (19). Cell viability was typical >95% by trypan blue exclusion. Hepatocytes (4 × 10⁵ cells/plate for six-well plates, 5 × 10⁶ cells/plate for 10-cm plates) were plated on gelatin-coated culture plates in Williams medium E with 10% calf serum, 15 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 10⁻⁶ M insulin, 2 mM L-glutamine, 100 U/ml penicillin, and streptomycin. Cells were allowed to attach to plates for at least 4 h before treatment.

Primary hepatocytes were treated with or without 1 μg/ml LPS in serum-free liver media (15 mM HEPES, 10⁻⁶ M insulin, 2 mM L-glutamine, 100 U/ml penicillin, and streptomycin) for 24 h. Culture media from two 10-cm plates for each group were harvested for exosome isolation. Proteins in the supernatant were extracted using methanol/chloroform. Total lysates were prepared using lysis buffer (1:10, Cell Signaling Technology). GW4869 and spiroepoxide were prepared as previously described (20). For exosome inhibition in vitro, GW4869 or spiroepoxide was added 2 h before LPS treatment (1 μg/ml LPS for 4 or 8 h). For knockdown of caspase-11 or GsdmD, 300 ng siRNA was diluted in 400 μl serum-free medium with 12 μl HiperFect™ transfection reagent (Qiagen) and mixed by vortexing. The mixture was incubated for 10 min at room temperature, added dropwise to the cells in 2-ml medium with 10% fetal bovine serum (FBS) (final siRNA concentration was 10 nM), swirled, and cultured the cells for 48 h.

Nuclear and cytoplasmic proteins were prepared using NE-PER™ Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher) according to the manufacturer's instructions. ER was prepared using an ER isolation kit (Sigma) according to the manufacturer's instructions. Nuclear/cytoplasmic or ER proteins were quantified using Pierce^TM BCA protein assay kit (Thermo Fisher).

Primary hepatocytes cultured on coverslips were treated as described and then fixed with 2% (w/v) paraformaldehyde in PBS for 15 min. Residual paraformaldehyde was removed by multiple washes with PBS, and cells were permeabilized with 0.1% Triton X-100 in PBS for 15 min at room temperature, washed with PBS and PBB (0.5% BSA in PBS) and blocked for 1 h with 20% normal goat serum (NGS, Sigma) in PBS. Then samples were incubated with the specific primary antibody for HMGB1 (Abcam, 1:200) in 1% BSA for 1 h, washed, and incubated with secondary antibody (goat anti-rabbit-Cy3, Jackson ImmunoResearch, 1:1,000). Additional in vitro experiments were performed with primary hepatocytes cultured on coverslips that were treated with Zombie Red™ viability dye (1:1,000, Biolegend) at room temperature in the dark for 30 min. All immunofluroescent staining sets involved staining the nuclei with Hoechst (1 mg/100 ml; Sigma) was applied at room temperature for 30 s followed by a single rinse of PBS to remove excess dye. In vitro samples cultured to collagen coated coverslips were adhered on the cell surface side of the coverslip to slides using Gelvatol [23 g of poly(vinyl alcohol 2000), 50 ml of glycerol, 0.1% sodium azide to 100 ml of PBS].

Liver tissue removed after perfusion with cold PBS and 2% paraformaldehyde was incubated for an additional 2 h to complete tissue fixation and then incubated for 24 h in 30% sucrose, followed by cryopreservation in liquid nitrogen cooled 2-methylbutane. Tissue sections of 6 μm were permeabilized with 0.3% Triton X-100 for 20 min, followed by staining according to the manufacturer's protocol of the in-situ Cell Death Detection Kit-TMR red (Roche). Samples were washed with PBS prior to being coverslipped using Gelvatol.

Regardless of the source of samples, all imaging conditions were maintained at identical settings with original gating performed using the primary delete control (no primary antibody). Large area images in X and Y were taken at a magnification of 20× with a two-fold digital zoom for the equivalent of nine fields/section with a Nikon A1 confocal microscope (purchased with 1S10OD019973-01 awarded to Dr. Simon C. Watkins). Quantification was performed in a blinded fashion using NIS Elements Software (Nikon). In brief, the Nikon NIS elements quantification software measure amount of cell death (either TMR or Zombie) fluorophore colocalized with the nuclear Hoechst fluorescences. The amount of HMGB1 content was measured for total HMGB1 fluorescences, as well as the amount of HMGB1 that colocalized with the nuclear content to enable the reporting of nuclear HMGB1, and cytosolic HMGB1 was analyzed as the amount of HMGB1 that did not colocalize with the nuclear HMGB1 content.

Mouse plasma was used for alanine aminotransferase (ALT) test. ALT levels were measured using the DRI-CHEM 4000 Chemistry Analyzer System (Heska). The ALT values were expressed as international units per liter.

Cells were plated on a 96-well black clear bottom plate. After LPS treatment, cells were washed and loaded with the ratiometric Ca²⁺ indicator Fura-2/AM in calcium-free Hank's balanced salt solution (HBSS) [at 37°C, 5% carbon dioxide (CO₂)] for 30 min, washed, and incubated for an additional 30 min prior to testing. Excitation was carried out at 340 and 380 nm, and emissions were collected at 510 ± 10 nm using BioTek SynergyMx multi-format microplate readers.

Exosome samples were analyzed as previously described (21). Briefly, exosomes isolated from 100 μl plasma were resuspended in 100 μl PBS and diluted 1:10,000 in particle-free water (W4502, Sigma). Exosomes isolated from 10⁷ cells were resuspended in 50 μl PBS and diluted 1:10,000 in particle-free water. After vortexing, the diluted samples were injected into the NTA LM-10 system continuously using a syringe pump. Particles were acquired by the machine, and data were analyzed with NTA particle analysis software.

HMGB1 ELISA Kit (IBL, Hamburg, Germany) was used to detect plasma HMGB1 levels according to the manufacturer's instructions. CD81 ELISA Kit (Cusabio, Wuhan, China) was used to detect plasma exosome samples according to the manufacturer's instructions.

Antibodies for Western blot analysis were as follows: anti-HMGB1 (1:1,000, Abcam), anti-caspase-11 (1:500, Sigma), anti-TSG101 (1:500, Novus), anti-CD81 (1:500, Novus), anti-Rab27a (1:1,000, Abcam), anti-Rab27b (1:1,000, Abcam), anti-beta actin (1:5,000, Abcam), anti-glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (1:5,000, Bio-Rad), anti-tubulin (1:5,000, Bio-Rad), anti-specificity protein 1 (SP1) (1:500, Santa Cruz), anti-phospho-camkk2 (1:1,000, Cell Signaling), anti-camkk2 (1:1,000, Novus), anti-calnexin (1:1,000, Novus), anti-ERp72 antibody (1:1,000, Cell Signaling). Secondary antibodies (1:10,000) were from Thermo Fisher Scientific. The procedure of Western blot analysis was as previously described (22). For in vitro experiments, hepatocytes were washed with PBS at the endpoint of experiments, collected in lysis buffer (Cell Signaling Technology) with phenylmethylsulfonyl fluoride (PMSF) and protease inhibitors, and centrifuged at 16,000g for 10 min, and supernatant was collected for Western blotting. For in vivo experiments, frozen liver (ischemic lobe) was homogenized in lysis buffer and centrifuged at 16,000g for 10 min, and supernatant was collected. Protein concentrations of the supernatants were determined with the bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific). Sodium dodecyl sulfate (SDS) loading buffer was then added to the samples. Denatured protein samples were analyzed by 10% or 15% SDS–polyacrylamide gel electrophoresis and then transferred onto a polyvinylidene difluoride membrane at 250 mA for 2 h. The membrane was blocked in 5% milk (Bio-Rad) in Tris-buffered saline (TBS) for 1 h and then incubated overnight with primary antibody in 1% milk in TBS overnight. Membranes were washed three times in TBS-Tween (TBS-T) for 10 min, incubated with horseradish peroxidase-conjugated secondary antibody for 1 h, and then washed three times for 10 min in TBS-T before being developed for chemiluminescence (Bio-Rad). Densitometry analysis was performed using the ImageJ Gel Analysis tool. GAPDH, β-actin, and tubulin are used as loading controls.

All data were analyzed using GraphPad Prism software (version 6.0). For in vivo and in vitro experiments, numerical measures will be compared using Student's t-test and were used for comparison between two groups or one-way ANOVA followed by post hoc Bonferroni test for multiple comparisons. If the data do not satisfy the assumptions necessary for this analysis, variables will be transformed, or a non-parametric alternative will be used. For the test above, there will be enough cultures/mice to attain a power of at least 80% at a significance level of 0.05. The required effect sizes for each analysis were estimated using the results from the preliminary studies. Based on these numbers, it was confirmed that the planned sample sizes are sufficient in attaining the desired power. Further, for statistical analysis of our in vitro studies, data will be expressed as mean ± SEM of three independent experiments performed in triplicate. A p < 0.05 was considered statistically significant for all experiments. All values are presented as the mean ± SEM.

We have previously demonstrated that hepatocytes release HMGB1 in sepsis and that this requires caspase-11 and GsdmD (5). Here, we confirmed that hepatocytes are the dominant source of the increases in circulating HMGB1 during endotoxemia. As shown in Supplemental Figure 1, cell-specific deletion of HMGB1 in hepatocytes, but not myeloid cells, prevented the rise in plasma HMGB1 observed in mice after LPS injection. As expected, LPS treatment in vivo led to an increase in liver levels of caspase-11 by 4 h that further increased at 8 h. The levels of caspase-11 at 24 h decreased to a similar level as at 4 h (Supplemental Figure 2). We confirmed that LPS treatment of cultured hepatocytes led to a caspase-11-dependent cleavage of GsdmD (Supplemental Figure 3). Deletion or knockdown of caspase-11 or GsdmD prevented LPS-induced HMGB1 release by cultured hepatocytes (Figures 1A,B,D,E; Supplemental Figure 6), while deletion of caspase-11 or GsdmD prevented the circulating rise in HMGB1 following LPS treatment in vivo (Figures 1C,F). The increase in extracellular HMGB1 induced by LPS treatment in vitro and in vivo was not due to cell death (Figure 2). LPS treatment in vivo failed to increase terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL)-positive cells in the liver or ALT levels in the plasma. Furthermore, LPS, at a concentration shown to induce HMGB1 release from hepatocytes in vitro, did not increase lactate dehydrogenase (LDH) release or Zombie-red uptake in cultured WT hepatocytes. Thus, hepatocytes actively release HMGB1 after exposure to LPS via a process that requires caspase-11 and GsdmD.

The active secretion of HMGB1 from cells requires at least two steps. First, HMGB1 accumulates in the cytoplasm, instead of the nucleus, and second, HMGB1, a leaderless protein, is released into the extracellular space (23). Therefore, we next assessed whether the nucleo-cytoplasmic translocation of HMGB1 in hepatocytes following LPS treatment required caspase-11 and/or GsdmD. As shown in Figure 3 and Supplemental Figure 7, LPS treatment led to an increase in cytosolic HMGB1 levels by 8 h. Whereas, deletion of caspase-11 or GsdmD had no impact on baseline levels of nuclear HMGB1 in hepatocytes, deletion of either gene prevented the accumulation of HMGB1 in the cytoplasm induced by LPS exposure. Specificity protein 1 (SP1) and GAPDH/tubulin were used to verify the compartment specificity of the proteins isolated from the nucleus and cytoplasm (Figure 5B; Supplemental Figure 9). These data establish that caspase-11 and GsdmD are involved in the nucleo-cytoplasmic translocation of HMGB1 that occurs after exposure of hepatocytes to extracellular LPS.

We have previously shown that hypoxia-induced HMGB1 release by hepatocytes requires camkkβ (24). Camkkβ belongs to the serine/threonine-specific protein kinase family and to the Ca²⁺/calmodulin-dependent protein kinase subfamily. Camkkβ catalyzes the phosphorylation of threonine residues located in the activation loop of the CaMKI and CaMKIV, enhancing their kinase activity. Therefore, we tested whether inhibition of camkkβ would reduce HMGB1 release from hepatocytes exposed to LPS. STO-609, a specific camkk inhibitor, blocked LPS-induced HMGB1 release (Supplemental Figure 4A). Interestingly, previous reports suggest that camkkβ (25) and its downstream targets, CaMKI (26, 27) and CaMKIV (28), may regulate HMGB1 nucleo-cytoplasmic translocation. Therefore, we investigated whether caspase-11 or GsdmD were required for camkkβ pathway activation. As shown in Figure 4 and Supplemental Figure 8, deletion of either caspase-11 or GsdmD reduced the phosphorylation of camkkβ in hepatocytes exposed to LPS (Figures 4A,B). Consistent with this result, we also found that intracellular free calcium was increased following LPS treatment in hepatocytes, and this required caspase-11 and GsdmD (Figures 4C,D). Furthermore, A23187, an ionophore that increases intracellular calcium levels, promoted HMGB1 release in response to LPS in caspase-11^−/− and GsdmD^−/− hepatocytes to a level similar to that seen in WT hepatocytes (Supplemental Figure 4B). Taken together, these data show that the Ca²⁺ signaling regulates LPS-induced HMGB1 release in hepatocytes, and this signaling pathway requires both caspase-11 and GsdmD.

Next, we investigated how GsdmD regulates intracellular calcium transients. Calcium storage is one of the functions commonly attributed to the ER in non-muscle cells (29). The N-terminal cleavage fragment of GsdmD can form pores in phospholipid membranes (13). Therefore, we sought evidence for GsdmD association with the ER in LPS-treated hepatocytes. The purity of the isolated ER was confirmed using electron microscopy and the ER protein markers, calnexin and ERp72 (Figures 5A,B). Western blot analysis demonstrated the presence of a GsdmD cleavage fragment in the ER lysate from LPS-treated WT hepatocytes but not caspase-11^−/− hepatocytes (Figure 5C). These findings raise the possibility that caspase-11 cleaves GsdmD, and the cleavage fragment then inserts in the ER membrane to release calcium into the cytoplasm.

We have recently provided evidence that HMGB1 is released into the extracellular space inside vesicles (5). To determine if these HMGB1-containing vesicles were exosomes, we confirmed that HMGB1 found in cell supernatant and plasma after LPS challenge was within CD81- and TSG101-positive vesicles (Figures 6A,B; Supplemental Figure 10). To further establish that HMGB1 is released via exosomes, NanoSight™ nanoparticle tracking analysis was used to show that mean size of particles isolated from hepatocyte supernatant in vitro and plasma were both in the range consistent for exosomes (40–100 nm) (Figures 6C,D). GW4869 and spiroepoxide, inhibitors of neutral sphingomyelinase associated with exosome release (20), reduced HMGB1 release from hepatocytes in a dose-dependent manner (Figures 6E,F). In vivo, GW4869 treatment before LPS challenge significantly suppressed HMGB1 levels in plasma (Figure 6G). Rab27a is required for exosome release (30). An adenovirus-expressing shRNA targeting Rab27a was used to suppress liver Rab27a (Supplemental Figure 5). Knockdown of Rab27a also significantly prevented HMGB1 increases in the plasma of LPS-treated mice (Figure 6H). Combined, these observations support the conclusion that HMGB1 is released into the extracellular space in exosomes.

We show above that the nucleo-cytoplasmic translocation of HMGB1 and the active extracellular release of HMGB1 in response to LPS exposure both require TLR4 and caspase-11/GsdmD. We hypothesized that the release of HMGB1 into exosomes would depend on the same pathways. Using KO mice, we confirmed that LPS-induced HMGB1 release in exosomes was TLR4, caspase-11, and GsdmD dependent both in vitro (Figures 7A,B; Supplemental Figure 11) and in vivo (Figures 7C–E). We next asked whether the TLR4 and caspase-11 pathways regulated total exosome release in response to LPS. LPS treatment markedly increased exosome numbers based on the Western blots for CD81 and TSG101. Only the deletion of TLR4 and not the deletion of caspase-11 or GsdmD prevented exosome release into the cell supernatant of cultured hepatocytes or into the plasma of mice after the LPS challenge (Figures 8A–E; Supplemental Figure 8). To further confirm these findings, we used ELISA for quantification of CD81, and the results were consistent with the Western blot analysis (Figure 8F). Thus, TLR4 regulates exosome formation, while caspase-11 and GsdmD are required only for HMGB1 delivery into exosomes.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.