Background

Although immunotherapy, especially immune checkpoint inhibitors, has shown clinical success, its efficacy is often limited.1 Failure to effectively activate the immune system to generate durable antitumor activity often leads to incomplete tumor elimination and tumor recurrence.2 A major factor in initial resistance to immunotherapy is the so-called “immunological cold” tumors, characterized by the absence of tumor-infiltrating lymphocytes.3 To overcome this challenge, immune-based combination therapies have emerged to establish a T cell-inflamed tumor microenvironment (TME) and enhance antitumor immune responses.4

A key functional feature of T cell-inflamed TME is the type I interferon (IFN) signature,5 which represents a key part of the initial phase of innate immunity. Type I IFNs play a crucial role in activating both innate and adaptive immunity, including dendritic cell maturation, antigen processing and presentation, and cross-priming of T cells.6 7 Studies have shown that type I IFN signature positively correlates with T-cell infiltration and clinical outcomes in various types of cancers,8 suggesting that activation of type I IFN signaling may represent a crucial mechanistic event in response to immunotherapy. Additionally, type I IFNs are potent inducers of immune checkpoint proteins,9–11 which are commonly used as predictive biomarkers of immunotherapeutic response. However, the expression of type I IFNs is often silenced or restricted within the TME.7 The stimulator of interferon genes (STING) is a well-characterized mediator in type I IFN production, which is activated by cytosolic DNA. Corresponding with type I IFN silencing, STING signaling is also frequently functionally suppressed in a wide variety of cancers.12 Deficiency in STING signaling limits innate sensing and is associated with poor prognosis and worse response to immunotherapy.13–15 Moreover, many conventional chemotherapeutics, targeted anticancer agents and immunological adjuvants are only fully efficient in the presence of intact type I IFN signaling.7 Therefore, enhancing STING and type I IFN signaling could be a promising strategy to improve the efficacy of immunotherapy.

Immunostimulatory chemotherapeutics are promising partners for combination regimens involving immune checkpoint inhibitors.16 One dogma might involve the induction of immunogenic cell death (ICD) that triggers type I IFN signaling and T-cell infiltration to enhance antitumor immunity.17 ICD is considered one of the most promising approaches to achieveing total tumor cell elimination and long-term immunological memory.18 Although previous studies have proposed several immunogenic drugs (eg, oxaliplatin, cyclophosphamide) to trigger ICD and type I IFN production in tumor cells and improve the therapeutic efficacy of immunotherapy, most chemotherapeutic agents render cell apoptosis via immunological silencing that results in less inflammation.18–2018 to 20 Intrinsic apoptosis, initiated by mitochondrial outer membrane permeabilization (MOMP), is considered as one of the major mechanisms underlying the antitumor activities attributed to chemotherapeutic agents.21 22 Following MOMP, mitochondrial DNA (mtDNA) is released into the cytosol and sensed by the second messenger cyclic GMP-AMP (cGAS)/ STING /interferon regulatory factor 3 (IRF3) pathway, producing type I IFNs. However, MOMP also causes cytochrome c release, which subsequently triggers caspase activation and facilitates the cleavage of cGAS and IRF3.23 The apoptotic caspase-mediated cleavages rapidly impede innate sensing and suppress mtDNA-induced STING-mediated type I IFN production, thereby preserving the immunologically quiescent state of apoptosis.24 25 Although mtDNA release appears to be a routine event during intrinsic apoptosis, the mtDNA-induced type I IFN secretion is only apparent in the absence of caspases.26 27 Interestingly, accumulating studies indicate that activated caspases are dispensable for cell death and the apoptotic clearance of cells in vivo.28 For example, Caspase-9-deficient cells exhibit only short-term resistance to apoptotic stimuli and do not determine drugs-treated cell death.27 This raises an important question of whether blocking chemotherapy-induced caspase activation could switch the intrinsic apoptosis from the “immunosuppressive” to “immunogenic” state to facilitate better tumor control.

In this study, we proposed a cell-based anticancer drug library screening approach for exploring the immunogenic apoptosis pathway and therapeutic targets under apoptotic caspase inhibition. We identified heat shock protein 90 (Hsp90) as a novel and potent target for activating immunogenic apoptosis and antitumor immunity. Blocking Hsp90-based chemotherapy-induced Caspase-9 activation provoked tumor intrinsic mtDNA sensing and T cell-inflamed TME for better tumor control. Notably, the combination treatment exhibited potent synergistic effects with programmed death-ligand 1 (PD-L1) blockade, even leading to complete tumor regression. Our findings highlight the potential of caspase inhibition in enhancing Hsp90-based chemotherapy-induced antitumor immunity and provide a novel therapeutic strategy to improve innate sensing and expand the benefits of immunotherapy.

Materials and methodsMice and cell culture

C57BL/6J female mice, 6 weeks old, were purchased from Shanghai Lingchang Biotechnology. The animal study was reviewed and approved by the Institutional Animal Care and Use Committee of Shanghai Jiao Tong University School of Medicine (Approval number: A-2020–001). CTIAC11 cells were generously provided by the Rolf Brekken laboratory at the University of Texas Southwestern Medical Center. MC38, CT26, HEK293T, HT29, and Panc02 cells were obtained from the Cell Bank of the Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. All cells used in this study were tested to be free of Mycoplasma contamination.

Reagents and antibodies

The reagents used included luminespib (747412-49-3, Topscience); VER50589 (747413-08-7, Topscience); emricasan (254750-02-2, Topscience). Anti-PD-L1 antibody (Clone: 10F.9G2), anti-mouse CD8 antibody (Clone: 116–13.1), anti-mouse IFNAR-1 antibody (Clone: MAR1-5A3) and isotype control were obtained from Bio X Cell. The fluorochrome-labeled anti-mouse antibodies used for flow cytometry were BV421-IFN-γ (505830; BioLegend); PerCP-Cy5.5-CD45 (103132; BioLegend); Fixable Viability Stain 780 (565388; BD Biosciences); CD8-APC (100712; BioLegend); CD4-V605 (562658; BD Pharmingen); TNF-α-PE (506306; BioLegend). The antibodies used for immunoblotting were anti-cGAS antibody (31659; CST); anti-STING antibody (13647; CST); anti-MAVS antibody (4983; CST); anti-MyD88 antibody (4283; CST); anti-GAPDH antibody (2118; CST); anti-Caspase-8 antibody (4790; CST); anti-Caspase-3 antibody (9662; CST); anti-Caspase-9 antibody (9504; CST); anti-Apaf-1 antibody (8969; CST); anti-Phospho-TBK1 antibody (5483; CST); anti-survivin antibody (2808; CST); anti-AIF antibody (5318; CST); anti-endonuclease G antibody (4969; CST); anti-tubulin antibody (5568; CST); anti-β-Actin antibody (3700; CST); anti-Bak antibody (3814; CST; 06–536, Sigma-Aldrich); anti-Bax antibody (2772; CST; Sc-23959, Santa Cruz); anti-Tom20 antibody (42406; CST); anti-Mcl-1 antibody (5453; CST); anti-Cleaved Caspase-9 antibody (9509; CST); anti-Cleaved Caspase-3 antibody (9661; CST); anti-PD-L1 (ab213480; Abcam). The anti-mouse antibodies used for immunohistochemistry were anti-granzyme B antibody (44153; CST); anti-CD8α antibody (44153; CST); anti-CD45 antibody (70257; CST). The anti-HSP70 antibody (10 995–1-AP; Proteintech) was used for flow cytometry.

In vitro high-throughput drug screening

The commercial anticancer drug library was purchased from Topscience (L2100). RAW-Lucia interferon-stimulated gene (ISG) cells with an IRF-inducible Lucia luciferase reporter construct were acquired from InvivoGen. Raw-Lucia ISG cells were seeded at a density of 30,000 cells per well in a 96-well flat bottom plate (Corning). After adhering to the plate, Raw-Lucia ISG cells were treated with drugs (1 µM) for 24 hours. Subsequently, 50 µL supernatant was transferred to a 96-well opaque white plate, and 50 µL QUANTI-Luc (InvivoGen) was added to detect luciferase activity.

CRISPR/Cas9-mediated gene knockout

Single guide RNAs (sgRNAs) were designed using the web-based tool CRISPR design (http://crispr.mit.edu/) and cloned into the vector lenti-CRISPR V.2 (Plasmid 49535, Addgene). The sgRNA targeting sequences are listed in online supplemental table 1.

In vivo tumor models

1×106 MC38 cells were injected subcutaneously into the flank of mice. For luminespib and emricasan treatment, luminespib and emricasan were dissolved in phosphate-buffered saline containing 10% Kolliphor EL (61791-12-6, SIGMA). Tumor-bearing mice were randomized and treated with luminespib (intraperitoneal injection, 75 mg/kg, every 2 days) and/or emricasan (intratumoral injection, 10 mg/kg, daily). Mice were killed when the tumor volume reached 2000 mm3. To block IFN-α/β receptor, tumor-bearing mice were injected intratumorally with anti-IFNAR1 antibody (150 µg per mouse, every 3 days). To deplete CD8+ T cells, tumor-bearing mice were injected intraperitoneally with anti-mouse CD8 antibody (100 µg per mouse, every 3 days). To block programmed death-ligand 1 (PD-L1) /programmed death-1 (PD-1) signaling pathway, tumor-bearing mice were injected intraperitoneally with anti-mouse PD-L1 antibody (100 µg per mouse, every 4 days).

RNA extraction and quantitative real-time PCR

Total RNA from cells was extracted using the TRIzol Reagent (15596026, Invitrogen) and reverse-transcribed with the PrimeScript RT reagent Kit (RR047A, Takara). Real-time PCR was performed on an ABI7900HI (Applied Biosystems). Gene expression was normalized to β-actin or L32. The primer sequences are shown in online supplemental table 2.

siRNA-mediated gene silencing

For transfection of siRNAs, siRNAs were transfected into cells with Lipofectamine 3000 (Invitrogen) according to the manufacturer’s protocol. After an additional 48 hours, cells were harvested for analysis. siRNA sequences were shown in online supplemental table 3.

Protein extraction and immunoblotting

Cells were lysed with RIPA buffer (PC101, Epizyme) supplemented with phosphatase (GRF102, Epizyme) and protease inhibitors (GRF101, Epizyme). Cytoplasmic and mitochondrial proteins were extracted following the manufacturer’s instructions (C500051, Sangon Biotech). The protein samples were denatured using SDS-PAGE Sample Buffer (P0015F, Beyotime Biotechnology) by heating. Subsequently, protein samples were separated on 6%–15% PAGE gels, transferred to nitrocellulose membranes (Millipore), incubated with primary antibody overnight at 4°C and secondary antibodies for 2 hours at room temperature. Protein signals were detected with the Tanon Image Analysis System. To assess the activation of Bcl2-associated X and Bcl2 antagonist/killer (BAX and BAK), protein samples were incubated with anti-BAX 6A7 or anti-BAK 23–38 and then with Protein A/G agarose beads (Millipore). Beads were washed five times and proteins were denatured by heating for immunoblotting.

Transcriptomic analysis

MC38 cells were treated with indicated drugs for 24 hours. Total RNA was extracted for transcriptome sequencing. Sequencing libraries were generated using NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, USA) following the manufacturer’s recommendations. The resulting clean reads were mapped to the mouse reference genome sequence (GRCm39), and the expression matrix was obtained. To identify differentially expressed genes (DEGs), the R package “DEseq2” was used, with genes having an absolute log2 fold change greater than 2 and false discovery rate (FDR) adjusted p value<0.01 considered as DEGs. Gene enrichment analysis was performed using the Gene Set Enrichment Analysis (GSEA) software (V.4.3.2) and the gene set database of Molecular Signatures Database.

Statistical analysis

Statistical analysis was performed using R software (V.4.2.3), GSEA (V.4.3.2) or GraphPad Prism (V.9.0), and all experiments were repeated at least three times independently. All values were presented as the means±SEM. The statistical significance was examined through the Student’s t-test, two-way ANOVA or log-rank test. A p value of <0.05 was considered significant. Some related quantitative statistical analysis was also provided in online supplemental file 2.

ResultsIdentification of Hsp90 inhibitor as an innate immune activator under caspase inhibition

To identify novel anticancer drugs with the potent potential to activate innate immunity, we first conducted a cell-based anticancer drug library screening approach using a well-established reporter system in RAW-Lucia ISG cells29 (figure 1A). This reporter system is sensitive to murine type I IFNs and can be applied to assess the drug’s capability in inducing type I IFN production.30 31 One thousand two hundred and fourteen conventional chemotherapeutic and targeted antineoplastic drugs were included in this anticancer drug library. Notably, several well-known inducers of ICD, including tacedinaline,32 teniposide,33 topotecan hydrochloride,34 imatinib mesylate,35 and epothilone B,36 demonstrated the ability to promote type I IFN production (online supplemental table 4), thereby providing robust validation of the effectiveness of our screening system. Our results suggested that the addition of a caspase inhibitor emricasan, significantly improved the ability of screened drugs to induce the production of type I IFNs, particularly for drugs that do not inherently stimulate this response (figure 1B and online supplemental figure S1A). Importantly, emricasan alone did not exhibit effective type I IFN-generating activity, confirming its synergistic effect in our screening system.

Figure 1

Caspase inhibition promotes Hsp90 inhibitor-induced tumor-intrinsic innate sensing. (A) Scheme of the in vitro high-throughput drug screen workflow. RAW-Lucia ISG cells were seeded in a 96-well flat bottom plate, and treated with each drug at a concentration of 1 µM. After 24 hours, the supernatants were harvested and subsequently transferred to a 96-well opaque white plate. QUANTI-Luc substrate was added to detect the luciferase activity and assess the levels of type I IFNs. (B) The scatter plot showing the screening results normalized to control or emricasan. Hsp90 inhibitors were highlighted in red on the plot. (C) Heatmap based on luciferase activity showing the top 10 most significant compounds in the presence of emricasan. (D) RAW-Lucia ISG cells were planted in 96-well plates and treated with luminespib (Lum)±emricasan (Em, 10 µM). After 24 hours, luciferase activity was assessed. (E) MC38 cells were treated with vehicle control (Ctrl), luminespib (Lum, 0.5 µM), emricasan (Em, 10 µM), luminespib and emricasan, respectively. After 24 hours, the supernatants were collected for IFN-β protein detection by ELISA (left); cells were collected for IFN-β detection by RT-qPCR (right). (F) HT29 cells were treated with vehicle control (Ctrl), luminespib (Lum, 0.5 µM), emricasan (Em, 10 µM), luminespib and emricasan, respectively. After 24 hours, cells were collected for IFN-β mRNA detection by RT-qPCR. (G, H) MC38 cells (G) and HT29 cells (H) were treated with indicated compounds for 24 hours. cells were collected for Cxcl10 and Isg15 mRNA detection by RT-qPCR. (I) Heatmap showing the top 10 significantly enriched genes that belong to type I IFN signaling between the control and luminespib plus emricasan groups. The heatmap was made by calculating the Z-score values. (J) In MC38 cells, the activation changes of IFN-α/β signaling pathway in luminespib plus emricasan group compared with control group were revealed by GSEA. (K) MC38 cells were treated with indicated compounds for 24 hours. Cells were collected for the detection of IFN-β mRNA by RT-qPCR. The compounds used were Z-VAD-FMK (Z-VAD, 20 µM), VER-50589 (VER, 2.5 µM), luminespib (Lum, 0.5 µM), and emricasan (Em, 10 µM). Data are shown as mean±SEM (n≥3). P value was calculated by unpaired Student’s t-test in (E–H and K). (**p<0.01, ***p<0.001). Hsp90, heat shock protein 90; IFN, interferon; ISG, interferon-stimulated gene; mRNA, messenger RNA; RT-qPCR, reverse transcription quantitative real-time PCR; GSEA, Gene Set Enrichment Analysis; FDR, false discovery rate.

Based on the screening results, we found that 8 of the top 10 candidate drugs were Hsp90 inhibitors (figure 1B,C and online supplemental figure S1A), which indicates the potent ability of Hsp90 inhibitors to stimulate type I IFN production. Notably, luminespib exhibited the most potent stimulatory effect on the production of type I IFNs (figure 1C), and its effect was dose-dependent (figure 1D and online supplemental figure S1B). We further confirmed the results in murine colorectal cancer cells MC38 and CT26, pancreatic cancer cells CTIAC11 and Panc02, and human colorectal cancer cell HT29. Luminespib or emricasan alone induced a marginal expression of IFN-β in all cell models (figure 1E,F and online supplemental figure S1C). However, combination treatment provoked a marked increase in IFN-β expression (figure 1E,F and online supplemental figure S1C), indicating synergistic effects between luminespib and emricasan. Consistently, the downstream signaling proteins (CXCL10 and ISG15) showed similar results (figure 1G,H, online supplemental figure S1D,E). Additionally, our transcriptomic (RNA sequencing) analysis also revealed a significant enrichment of type I IFN signaling after combination treatment (figure 1I,J and online supplemental figure S1F). To exclude any potential off-target effects of luminespib and emricasan, we tested another screened Hsp90 inhibitor VER50589, and another pan-caspase inhibitor Z-VAD-FMK, respectively. In line with the findings mentioned earlier, the combination of VER50589 and emricasan, as well as that of luminespib and Z-VAD-FMK, significantly enhanced the expression of IFN-β and its downstream signaling proteins (figure 1K, online supplemental figure S1G,H), confirming the role of Hsp90 and caspase in the process. Taken together, these results demonstrate that caspase inhibition promotes Hsp90-based chemotherapy-mediated type I IFN production.

Caspase inhibition promotes Hsp90 inhibitor-induced mtDNA release and triggers innate sensing through cGAS/STING pathway

To investigate the mechanistic basis of how luminespib in combination with caspase inhibitor activates type I IFNs, we first assessed the effects of combination treatment on the major steps of type I IFN signaling. Activation of cell-intrinsic type I IFNs is initiated by cytosolic nucleic acid sensing signaling,37 38 primarily via three distinct pathways (figure 2A): (1) sensing of cytosolic double-stranded DNA by cGAS; (2) sensing of cytosolic double-stranded RNA by retinoic acid-inducible gene I and melanoma differentiation-associated protein 5; (3) sensing of cytosolic single-stranded RNA by toll-like receptors. Notably, TANK-binding kinase 1 (TBK1) acts as a crucial signaling hub downstream of all three pathways, and is responsible for activating IRF3, leading to the production of type I IFNs and subsequent expression of ISGs. Our results showed that luminespib combined with caspase inhibitors (emricasan or Z-VAD-FMK) markedly activated TBK1 in MC38 and Panc02 cells (figure 2B and online supplemental figure S2A). Importantly, knockout of IRF3 abrogated the promoting effect of combination treatment on IFN-β production (figure 2C and online supplemental figure S2B), confirming the regulatory role of IRF3. To explore which upstream pathway is required for type I IFN production after combination treatment, we knocked out the key mediator of each signaling pathway in MC38 cells (online supplemental figure S2B). Similar to wild type, knockout of MyD88 or MAVS had minimal impact on IFN-β expression and TBK1 activation (figure 2C,D). In contrast, knockout of cGAS or STING showed similar results as knockout of IRF3, and largely diminished the activation of IFN-β and TBK1 induced by combination treatment (figure 2C,D). These results suggest that the combination treatment induces the production of type I IFNs via cGAS/STING/TBK1/IRF3 signaling.

Figure 2

Hsp90 inhibitor combined with caspase inhibition promotes tumor-intrinsic mtDNA sensing. (A) Schematic diagram of the main pathways involved in intracellular production of type I interferons. (B) MC38 cells were treated with luminespib (Lum, 0.5 µM) and emricasan (Em, 10 µM) or Z-VAD-FMK (Z-VAD, 20 µM) for 24 hours, the protein level of phosphorylated TBK1 (p-TBK1) was determined by western blot. Poly(I:C) (2 µg/mL) was used as a positive control. (C, D) Indicated cells were treated with luminespib and emricasan for 24 hours. The mRNA expression of IFN-β was determined by RT-qPCR (C) and the protein level of p-TBK1 was determined by western blot (D). (E) Representative confocal images showing cytosolic DNA (PicoGreen, green), mitochondria (Mito-Tracker, red), and nuclei (DAPI, blue) in MC38 cells after indicated treatments. Scale bar, 5 µm. (F, G) MC38 cells were treated with indicated compounds for 24 hours, cytosolic DNA was extracted, and the levels of gDNA (F) and mtDNA (G) were determined by RT-qPCR. (H, I) MC38 or ddC-treated MC38 (MC38-ddC) cells were treated with indicated compounds for 24 hours, and the mRNA expression of IFN-β was determined by RT-qPCR (H) and the protein level of p-TBK1 was determined by western blot (I). Data are shown as mean±SEM (n≥3). P value was calculated by unpaired Student’s t-test in (C and F–H). (***p<0.001). RT-qPCR, reverse transcription quantitative real-time PCR; ddC, dideoxycytidine; dsDNA, double-stranded DNA; dsRNA, double-stranded RNA; ssRNA, single-stranded RNA ;gDNA, genomic DNA; IFN, interferon; IRF3, interferon regulatory factor 3; MDA5, melanoma differentiation-associated protein 5; mRNA, messenger RNA; mtDNA; mitochondrial DNA;cGAS, the second messenger cyclic GMP–AMP; RIG-I, retinoic acid-inducible gene I; STING, stimulator of interferon gene; TBK1, TANK-binding kinase 1; TLR, toll-like receptor; WT, wild type.

Endogenous DNA sources that trigger STING activation include damaged genomic DNA (gDNA) and mtDNA.39 Indeed, we observed that emricasan had no obvious effect on cytosolic DNA (figure 2E and online supplemental figure S2C), while luminespib significantly increased cytosolic DNA regardless of emricasan treatment (figure 2E and online supplemental figure S2C). The excessive cytosolic DNA induced by luminespib was mainly derived from the accumulation of gDNA and mtDNA (figure 2F,G). However, addition of emricasan reduced cytosolic gDNA levels but further promoted the mtDNA release compared with the group treated with luminespib alone (figure 2F,G). These results indicate that mtDNA may be the major DNA source for DNA sensing in combination treatment, and luminespib could function as an mtDNA inducer. To investigate whether mtDNA is required for the production of type I IFNs, we depleted mtDNA using dideoxycytidine (online supplemental figure S2D), and found that depletion of mtDNA dramatically abolished IFN-β production and TBK1 activation after combination treatment (figure 2H,I), thus confirming the involvement of mtDNA in type I IFN production. Overall, these results demonstrate that inhibition of both Hsp90 and caspases induces mtDNA release, triggers cGAS/STING/TBK1/IRF3 pathways, and promotes the production of type I IFNs.

Blocking Caspase-9 signaling relieves the restriction of DNA sensing induced by Hsp90 inhibitor

MOMP is a crucial event that triggers mtDNA release into the cytosol. Previous studies have suggested that Hsp90 inhibitor can transcriptionally downregulate myeloid cell leukemia-1 (MCL-1),40 an anti-MOMP protein that binds and inhibits the executioners of MOMP, including BAK and BAX.41 Therefore, we hypothesized that luminespib could induce MOMP by activating BAK and BAX, thus promoting mtDNA release. As expected, both luminespib alone and combination treatment significantly downregulated MCL-1 and upregulated BAK and BAX (figure 3A and online supplemental figure S3A). To further validate the activation of BAX and BAK, we employed conformation-specific antibodies targeting the activation epitope. Our results demonstrated significant activation of both BAK and BAX on treatment with the Hsp90 inhibitor alone, as well as in the combination treatment (figure 3A). Importantly, both luminespib alone and combination treatment stimulated the release of mitochondrial content, as evidenced by loss of mitochondrial cytochrome c expression and staining (figure 3B,C), thereby confirming the onset of MOMP. Furthermore, we observed a significant enrichment of the apoptosis pathway and an increase in the activity of Caspase-3/7 and the expression of cleaved Caspase-3 in luminespib-treated cells (figure 3D–F and online supplemental figure S3B). However, the activation of Caspase-3/7 was blocked by emricasan (figure 3E,F). Given that caspases can mediate the cleavage of cGAS/STING/TBK1, this could explain why luminespib induces the production of type I IFNs only in the presence of caspase inhibitor.

Figure 3

Tumor-intrinsic Caspase-9 signaling restricts Hsp90 inhibitor-mediated DNA sensing. (A) MC38 cells were treated with indicated treatments for 24 hours. The protein level of BAX and BAK was determined by western blot. The activation of BAX and BAK was analyzed by immunoprecipitation with the conformation-specific anti-Bax and anti-Bak antibodies. The Tom20 protein was used as a mitochondrial marker. (B) Cytochrome c (Cyt-C) was determined by western blot. (C) Representative confocal images showing Cyt-C (green), Tom-20/mitochondria (red), and nucleus (blue) in MC38 cells after indicated treatments. Scale bar, 10 µm. (D) GSEA analysis revealing an upregulation of apoptosis signaling pathway in luminespib-treated group compared with control. (E, F) MC38 cells were treated with indicated compounds for 24 hours. The Caspase-3/7 activity was measured by Caspase-Glo 3/7 assay kit (E). The protein level of cleaved Caspase-3 was determined by western blot (F). (G, H) Indicated cells were treated with luminespib for 24 hours. The protein level of p-TBK1 was determined by western blot (G). The mRNA expression of IFN-β was determined by RT-qPCR (H). (I, J) MC38 cells were treated with indicated compounds for 24 hours. Caspase-9 activity was measured by Caspase-Glo 9 assay kit (I). The protein level of cleaved Caspase-9 was determined by western blot (J). (K, L) Indicated cells were treated with luminespib for 24 hours. The protein level of p-TBK1 was determined by western blot (K). The expression of IFN-β was determined by RT-qPCR (L). Data are shown as mean±SEM (n≥3). P value was calculated by unpaired Student’s t-test in (E, H, I and L). (***p<0.001). BAK, Bcl2 antagonist/killer; BAX, Bcl2-associated X; IFN, interferon; mRNA, messenger RNA;NES, normalized enrichment score ;GSEA, Gene Set Enrichment Analysis; FDR, false discovery rate; RT-qPCR, reverse transcription quantitative real-time PCR; TBK1, TANK-binding kinase 1; WT, wild type.

Emricasan is a pan-caspase inhibitor and blocks both intrinsic and extrinsic apoptosis. We next determined which apoptosis signaling is activated by luminespib. We knocked out Caspase-9 and Caspase-8 (online supplemental figure S3C), which are the key mediators of intrinsic and extrinsic apoptosis respectively, to mimic the function of emricasan. Similar to wild type, knockout of Caspase-8 failed to induce TBK1 activation as well as the expression of IFN-β and downstream signaling after luminespib treatment (figure 3G,H and online supplemental figure S3D). Conversely, deficiency of Caspase-9 exhibited similar effects to emricasan after luminespib treatment (figure 3G,H and online supplemental figure S3D), suggesting that Hsp90 inhibitor selectively activates Caspase-9 and subsequently mediates the restriction of type I IFN production. Indeed, we observed increased activity and a cleaved form of Caspase-9 following luminespib treatment (figure 3I,J). Moreover, we knocked out the upstream adapter protein APAF-1 and downstream executor Caspase-3 of Caspase-9 (online supplemental figure S3C). We found that loss of both proteins led to TBK1 activation as well as the expression of IFN-β and downstream signaling after luminespib treatment (figure 3K,L and online supplemental figure S3E), further confirming the role of the intrinsic apoptosis in luminespib-induced type I IFNs. Above all, our results suggest that the Hsp90 inhibitor induces the activation of tumor-intrinsic Caspase-9 signaling, which in turn restricts the production of type I IFNs.

Blockade of Hsp90 and Caspase-9 induces caspase-independent cell death and enhances tumor immunogenicity

Although luminespib induces apoptosis, our RNA sequencing analysis showed that emricasan did not significantly impact the signaling pathways involved in the cell cycle and cell death following luminespib treatment (figure 4A), indicating that emricasan fails to prevent cell death. To validate our findings, we conducted experiments on both mouse and human cells, and found that emricasan did not reduce the cell death mediated by luminespib (figure 4B). Additionally, the levels of survivin, which is a client protein of Hsp90 and essential for cell survival, decreased following luminespib treatment, regardless of emricasan treatment (figure 4C). To further verify the role of Hsp90 in cell death, we tested another Hsp90 inhibitor, VER50589, and obtained similar results (online supplemental figure S4). These results suggest that caspase inhibition does not weaken the cytotoxicity of Hsp90 inhibitor, but instead shifts Hsp90 inhibitor-induced apoptosis towards caspase-independent cell death (CICD).

Figure 4

Targeting heat shock protein 90 and caspases induces caspase-independent cell death and enhances tumor immunogenicity. (A) Heat map showing the changes in cell cycle and programmed cell death signaling pathway in MC38 cells after indicated treatment. (B) MC38, Panc02, and HT29 cells were treated with luminespib (0–2.5 µM) for 48 hours. Then cell viability was determined by Cell Counting Kit-8 (CCK-8) assay. (C) The protein level of survivin was determined by western blot. (D) MC38 cells were transfected with indicated siRNA, the efficiency of gene knockdown was detected by western blot and cell viability was determined by CCK-8 assay. (E) MC38 cells were treated with indicated compounds (Nec-1, necrostatin-1; Fer-1, ferrostatin-1; Chl, chloroquine) for 48 hours and cell viability was determined by CCK-8 assay. (F) Representative confocal images showing HMGB1 (green), DilC18/cell membrane (red), and nucleus (blue) in MC38 cells after indicated treatments. Scale bar, 10 µm. (G–J) MC38 cells were treated with indicated compounds for 24 hours. Then, the levels of HMGB1 (G) and ATP (H) released into the culture supernatant were quantified by an ELISA and an ATP test kit, respectively. Cells were harvested and cell surface expression of calreticulin (I) and Hsp70 (J) was measured by flow cytometry. (K) Scheme of in vivo tumor vaccination-rechallenge model. MC38-OVA cells were treated with lethal doses of luminespib, cisplatin, or combination of luminespib and emricasan in vitro, respectively. These treated cells were then inoculated subcutaneously into the flank of C57BL/6 mice. After 7 days, mice were rechallenged with live MC38-OVA cells by injection into the contralateral flank. (L) The percentage of rechallenged tumor-free mice was shown (tumor volume below 50 mm3 was recorded as tumor-free). Data are shown as mean±SEM (n≥3). P value was calculated by unpaired Student’s t-test in (G–J) or log-rank test in (L). (ns, not significant, ***p<0.001). AIF, apoptosis-inducing factor; HMGB1, high-mobility group box protein 1; MFI, Mean Fluorescence Intensity.

Apart from caspase-dependent apoptosis, MOMP can induce caspase-independent apoptosis by releasing other pro-apoptotic factors from the mitochondria, such as apoptosis-inducing factor (AIF) and endonuclease G.42 To investigate the role of AIF and endonuclease G in the combination treatment-induced cell death, we silenced their expression in MC38 cells using siRNA (figure 4D). However, knockdown of AIF or endonuclease G did not affect cell death following luminespib treatment regardless of emricasan treatment (figure 4D), ruling out the involvement of the two proteins. We next tested whether combination treatment induced other forms of CICD, such as necroptosis, ferroptosis or autophagy, using small molecular inhibitors. Strikingly, all tested inhibitors failed to reverse cell death (figure 4E), thus excluding the involvement of the three forms of CICD. During intrinsic apoptosis, MOMP disrupts mitochondrial function and, even in the absence of caspase activity, energy production eventually wanes and cells die.43 44 Therefore, luminespib-induced MOMP may activate a point-of-no-return cell death mechanism in the presence of emricasan, which may be responsible for the observed cell death.

ICD has emerged as a pivotal component of therapy-induced antitumor immunity.45 To evaluate the ability of luminespib combined with emricasan to stimulate ICD of tumor cells, we investigated the fundamental molecular events of ICD in vitro, including the release of high-mobility group box protein 1 (HMGB1), the secretion of ATP, and surface exposure of calreticulin (CRT) and heat shock protein 70 (Hsp70). Our results showed that, unlike emricasan, luminespib alone promoted the secretion of HMGB1 and ATP, as well as the expression of CRT and Hsp70 on the cell surface (figure 4F–J), while combination with emricasan significantly magnified the promoting effects (figure 4F–J). To further validate the immunogenicity-inducing effect of combination treatment in vivo, we employed a classical tumor vaccination-rechallenge model in immunocompetent C57BL/6 mice. MC38 cells were pretreated with luminespib and emricasan to induce CICD. Cisplatin, an inefficient ICD inducer,46 was used as a negative control. Dying cells were collected and injected as a vaccine into syngeneic mice. After 7 days, mice were rechallenged with live untreated MC38 cells on the opposite flank (figure 4K). As shown, immunization with luminespib-treated MC38 cells provided moderate protection against rechallenged tumor (figure 4L). However, mice immunized with combination-treated MC38 cells exhibited delayed tumor growth and 80% tumor-free survival (figure 4L). These results indicate that luminespib is insufficient to immunize mice and induce effective immunogenicity, but the combination with emricasan strengthens luminespib-mediated ICD-associated features and enhances specific antitumor immunogenicity.

Caspase-9 inhibition synergistically promotes Hsp90 inhibitor-mediated antitumor immunity

Type I IFN-mediated innate sensing and ICD-induced adaptive immunity are key elements in generating effective antitumor immune responses.6 17 In light of this, we evaluated the antitumor efficacy of Hsp90 inhibitor combined with Caspase-9 inhibition in the syngeneic tumor model. MC38 cells were subcutaneously implanted into the right flank of mice, and luminespib was intraperitoneally injected once every 2 days. To minimize off-target side effects of emricasan and prime immune cells locally, emricasan was intratumorally administrated every day for seven doses (online supplemental figure S5A). Both single treatments of luminespib or emricasan delayed tumor growth (figure 5A and online supplemental figure S5B). Notably, the combination treatment yielded more robust therapeutic effects with a tumor growth inhibition