SCH727965

CDK1/2/5 inhibition overcomes IFNG-mediated adaptive immune resistance in pancreatic cancer
Jin Huang,1,2 Pan Chen,3 Ke Liu,4,5 Jiao Liu,1 Borong Zhou,1 Runliu Wu,5 Qiu Peng,6 Ze-Xian Liu ,7 Changfeng Li,8 Guido Kroemer,9,10,11,12,13 Michael Lotze,14 Herbert Zeh,5 Rui Kang,5 Daolin Tang 1,5

⦁ Additional material is published online only. To view please visit the journal online ⦁ (http://dx.doi.⦁ org/10.1136/ gutjnl-2019-320441).
For numbered affiliations see end of article.

Correspondence to
Dr Daolin Tang and Dr Rui Kang, Department of Surgery, UT Southwestern Medical Center, Dallas, Texas, USA; [email protected] edu, [email protected]

JH, PC and KL contributed equally.

Received 9 December 2019
Revised 3 July 2020
Accepted 14 July 2020

© Author(s) (or their employer(s)) 2020. No commercial re-use. See rights and permissions. Published by BMJ.

To cite: Huang J, Chen P, Liu K, et al. Gut Epub ahead of print: [please include Day Month Year]. doi:10.1136/ gutjnl-2019-320441

ABSTRACT
Objective Adaptive immune resistance mediated by the cytokine interferon gamma (IFNG) still constitutes a major problem in cancer immunotherapy. We develop strategies for overcoming IFNG-mediated adaptive
immune resistance in pancreatic ductal adenocarcinoma cancer (PDAC).
Design We screened 429 kinase inhibitors for blocking IFNG-induced immune checkpoint (indoleamine
2,3-dioxygenase 1 (IDO1) and CD274) expression in a human PDAC cell line. We evaluated the ability of the cyclin-dependent kinase (CDK) inhibitor dinaciclib to block IFNG-induced IDO1 and CD274 expression in 24 human and mouse cancer cell lines as well as in primary cancer cells from patients with PDAC or ovarian carcinoma. We tested the effects of dinaciclib on IFNG-induced signal transducer and activator of
transcription 1 activation and immunological cell death, and investigated the potential utility of dinaciclib in combination with IFNG for pancreatic cancer therapy
in vivo, and compared gene expression levels between human cancer tissues with patient survival times using the Cancer Genome Atlas datasets.
Results Pharmacological (using dinaciclib) or genetic (using shRNA or siRNA) inactivation of CDK1/2/5 not only blocks JUN-dependent immune checkpoint expression, but also triggers histone- dependent immunogenic cell death in immortalised or primary cancer cells in response to IFNG. This dual mechanism turns an immunologically ’cold’ tumour microenvironment into a ’hot’ one, dramatically improving overall survival rates in mouse pancreatic
tumour models (subcutaneous, orthotopic and transgenic models). The abnormal expression of CDK1/2/5 and IDO1 was associated with poor patient survival in several cancer types, including PDAC.
Conclusion CDK1/2/5 kinase activity is essential for IFNG-mediated cancer immunoevasion. CDK1/2/5 inhibition by dinaciclib provides a novel strategy
to overcome IFNG-triggered acquired resistance in pancreatic tumour immunity.

INTRODUCTION
The past decade has witnessed the remarkable expansion of anticancer immunotherapies across a large range of tumour types. The effectiveness of the immunotherapies was first demonstrated with the administration of the cytokine interleukin-2 (IL2) in patients with melanoma and renal cancer.1

More recently, monoclonal antibodies that block immune checkpoint receptors or ligands, such as programmed cell death 1 (PDCD1/PD-1) and

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.
Huang J, et al. Gut 2020;0:1–10. doi:10.1136/gutjnl-2019-320441 1

GI cancer

programmed death ligand 1 (CD274), have led to breakthrough advances in improving outcomes of a sizeable fraction of patients with tumour.2 Unfortunately, some tumours are not responsive to immunotherapies either due to the lack of infiltrating T cells within the tumour microenvironment (TME) or the develop- ment of therapeutic resistance. In adaptive resistance, immune checkpoints are not constitutively expressed, but rather are induced by interferon gamma (IFNG) produced by T or natural killer (NK) cells that are attempting to attack neoplastic cells.3 A treatment option that overcomes this IFNG-elicited acquired resistance against immunotherapies could be promising.
Pancreatic ductal adenocarcinoma (PDAC) is one of the few KRAS-driven cancers for which survival has not improved over the past decades in spite of multiple clinical trials.4 In this study, we conducted a protein kinase inhibitor library screen to identify candidate kinases responsible for IFNG-induced expression of immune checkpoint-relevant proteins in a human PDAC cell line. This screen identified dinaciclib, an inhibitor of cyclin-dependent kinase (CDK)-1, CDK2 and CDK5 (hereafter referred to as CDK1/2/5), as a drug that interferes with IFNG- induced immune checkpoint expression and that additionally causes immunogenic cell death (ICD), a form of regulated cell death that is sufficient to activate an adaptive immune response in immunocompetent hosts.5 6 This dual mechanism enhances the anticancer activity of IFNG and improves overall survival rates in mouse PDAC models.

MATERIALS AND METHODS
Cell lines
Human or mouse cell lines were grown in Dulbecco’s Modi- fied Eagle Medium (DMEM) or Roswell Park Memorial Insti- tute (RPMI)-1640 medium with 10% fetal bovine serum, 2 mM L-glutamine and 100 U/mL of penicillin and streptomycin and maintained in a 5% CO2 incubator at 37°C. All cells were myco- plasma free and authenticated using short tandem repeat DNA profiling analysis.

Primary culture cells
Human clinical samples from patients with ovarian cancer or pancreatic cancer who underwent surgery were collected from the Xiangya Hospital. Collection of samples was approved by its institutional review board. After surgical resection, the tissue was promptly transmitted to a sterile dish with a medium containing penicillin/streptomycin. Primary tumour samples were minced with a GentleMACS Dissociator and received 5 min of centrifu- gation, 1000 revolutions/min, to remove supernatant. Cells were cultured in DMEM containing 10% fetal bovine serum (FBS), 100 U/mL of penicillin and streptomycin and insulin. Primary tumour cells were confirmed by CK19 (ductal cell marker) staining.

Animal models
We conducted all animal care and experiments in accordance with the Association for Assessment and Accreditation of Labo- ratory Animal Care guidelines (http://www.aaalac.org) and with approval from our institutional animal care and use committee. All experimental and control animals were matched on sex and age.
To generate a vaccination model, a total of 5×106 KPC or CT26 cells, untreated or treated with IFNG (10 ng/mL)/dinac- iclib (100 nM) for 24 hours, were combined with 200 µL of phosphate-buffered saline (PBS) and inoculated subcutaneously into the lower flank of 6-week-old female C57BL/6 (KPC) or

BALB/c (CT26) mice, whereas 5×105 untreated control cells were inoculated into the contralateral flank 7 days later.7 The percentage of tumour-free mice was monitored every week.
To generate murine subcutaneous tumours, 5×106 cells from indicated cancer cell lines in 100 µL PBS were injected subcu- taneously to the right of the dorsal midline in female 6-week to 8-week-old C57BL/6 mice. Once the tumours reached 100–150 mm3 at day 7, mice were randomly allocated into groups and treated with IFNG (2 µg/mice/intraperitoneal injection [i.p.], three times weekly), dinaciclib (8 mg/kg/i.p., three times weekly) or IFNG plus dinaciclib at day 7 for 2 weeks. Tumours were measured two times weekly and volumes were calculated using the formula length × width2 × π/6. To deplete CD8+ T cells, C57BL/6 mice were treated with 200 mg of control IgG (BioX- cell) or anti-CD8 (#BE0061, BioXcell) depleting antibody at day −1, +1, +7 and +15 relative to tumour cell implantation.8 To generate orthotopic tumours, female C57BL/6 mice were surgically implanted with 5×105 KPC cells in 10 µL PBS into the tail of the pancreas.9 10 One week after implantation, mice were randomly allocated into groups and treated with IFNG (2 µg/ mice/i.p., three times weekly), dinaciclib (8 mg/kg/i.p., three times weekly) or IFNG plus dinaciclib for 4 weeks. Animal
survival was monitored every week.
Pdx-1-Cre and K-rasG12D/+ transgenic mice on C57BL/6 back- ground were received from the Jackson Laboratory. Hmgb1flox/ flox mice on C57BL/6 background were obtained from Eugene
B. Chang. These mice were crossed to generate indicated KCH animals. Compared with the KPC model, the Hmgb1 depletion in the KCH model shows a faster K-Ras-driven PDAC phenotype.11 Reduced intracellular high-mobility group box 1 (HMGB1) expression also correlates with poor survival in patients with pancreatic cancer.11 At 1 month of age, mice were randomly allo- cated into groups and treated with IFNG (2 µg/mice/i.p., three times weekly), dinaciclib (8 mg/kg/i.p., three times weekly) or IFNG plus dinaciclib for 4 weeks. Animal survival was moni- tored every week.
Full reagents and methods are available in the online supple- mentary materials and methods.

RESULTS
Identification of dinaciclib as an immune checkpoint inhibitor PDAC is generally resistant to current strategies targeting the CD274 immune checkpoint pathway.12 13 In contrast, the pharmacological inhibition of indoleamine 2,3-dioxygenase
1 (IDO1), a cytosolic tryptophan-catabolising enzyme and immune checkpoint, yields promising responses in patients with metastatic PDAC (ClinicalTrials.gov identifier: NCT02077881). Using CFPAC1 cells, a human PDAC cell line harbouring hotspot (KRAS, TP53 and SMAD4) mutations, we screened a protein kinase inhibitor library containing 429 small-molecule compounds by western blot technology (figure 1A). A first screening using a higher working concentration (10 µM) found that 29 compounds, including 12 CDK inhibitors (41.3% of all compounds), completely blocked the IFNG-induced protein expression of both IDO1 and CD274 in CFPAC1 cells (figure 1B and online supplementary figure S1). A secondary screen for these 29 compounds using a lower concentration (1 µM or 100 nM) identified dinaciclib, a novel CDK inhib- itor with immunostimulatory activity that is currently evalu- ated in a phase III clinical trial for the treatment of chronic lymphocytic leukaemia, as the agent that most efficiently blocks IFNG-induced IDO1 and CD274 expression in CFPAC1 cells (figure 1B,C).

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.

2 Huang J, et al. Gut 2020;0:1–10. doi:10.1136/gutjnl-2019-320441

GI cancer

Figure 1 Identification of dinaciclib as an immune checkpoint inhibitor. (A) Flow chart describing western blot-based screening approach. (B) Schematic summary of the number of protein kinase inhibitors used at 10 µM, 1 µM or 100 nM, completely blocking interferon gamma (IFNG)-induced indoleamine 2,3-dioxygenase 1 (IDO1) and CD274 protein expression. (C) Western blot analysis of IDO1 and CD274 expression in CFPAC1 cells following treatment with IFNG (10 ng/mL) for 24 hours in the absence or presence of indicated protein kinase inhibitor (1 µM or 100 nM). (D) Heat map of mRNA expression levels of IDO1 and CD274 in 24 cancer cell lines following treatment with IFNG (10 ng/mL) in the absence or presence of dinaciclib (100 nM) for 24 hours. (E) Western blot analysis of IDO1 and CD274 expression in cancer cell lines (KPC, PANC02 and OVCAR5) as well as primary cancer cells from patients with pancreatic ductal adenocarcinoma cancer (PAN #1-3) or ovarian cancer (OVA #1-3) following treatment with IFNG (10 ng/mL) for 24 hours in the absence or presence of dinaciclib (100 nM). (F) Q-PCR analysis of indicated mRNA expression in CFPAC1 cells following treatment with IFNG (10 ng/mL) in the absence or presence dinaciclib (100 nM) for 24 hours. Data in (D) and (F) are from two independent experiments. Data in (C) and (E) are from three independent experiments. CDK, cyclin-dependent kinase.

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.
Dinaciclib used at 100 nM also blocked IFNG-induced mRNA expression of IDO1 and CD274 in 24 human and mouse cancer cell lines (figure 1D). Dinaciclib employed at 100 nM also blocked IFNG-induced protein expression of IDO1 and CD274 in malignant cell lines (KPC, PANC02 and OVCAR5), as well as in primary cancer cells from patients with PDAC or ovarian
Huang J, et al. Gut 2020;0:1–10. doi:10.1136/gutjnl-2019-320441
carcinoma (figure 1E). Dinaciclib failed to affect IFNG-induced mRNA upregulation of major histocompatibility complex, class I, A (HLA-A), C-X-C motif chemokine ligand 9 (CXCL9), CXCL10 and intercellular adhesion molecule 1 (figure 1F), all of which are regulators involved in IFNG-mediated antitumour immunity. In addition, IFNG failed to induce the expression of
3

GI cancer

killer cell lectin-like receptor K1 (KLRK1/NKG2D) and other checkpoint receptors (eg, hepatitis A virus cellular receptor 2 (HAVCR2/TIM3) and lymphocyte-activating 3 (LAG3)) in CFPAC1 cells (figure 1F). These findings demonstrate that dinaciclib selectively inhibits IFNG-elicited immune checkpoint expression (eg, IDO1 and CD274) in cancer cells.

CDK1/2/5 inhibition blocks JUN-dependent STAT1 expression and activation
Given that dinaciclib is a potent inhibitor for CDK1, CDK2, CDK5 and CDK9,14 we sought to define which CDK is required for IFNG-induced immune checkpoint expression. Dinaciclib caused cell cycle arrest in G2/M (online supplementary figure S2) and downregulated the expression of some CDKs (CDK7 and CDK9) and cyclins (CCNA1 and CCNB1) in CFPAC1 and OVCAR5 cells (online supplementary figure S3). The shRNA-mediated knockdown of single CDKs from the group of dinaciclib-inhibited enzymes (figure 2A) failed to completely block IFNG-induced IDO1 and CD274 expression in CFPAC1 cells (figure 2B). However, the shRNA-mediated knockdown of more than one dinaciclib-targeted CDK led to the conclusion that concomitant CDK1/2/5 depletion mimics the effects of dinaci- clib that abolish the IFNG-induced upregulation of IDO1 and CD274 at the mRNA (figure 2C) and protein levels (figure 2D). This result was further confirmed by means of siRNA-mediated gene depletion (online supplementary figure S4).
To determine how CDK1/2/5 drives immune checkpoint expression, we assayed the impact of CDK1/2/5 inhibition on the expression and activation of the transcription factor signal transducer and activator of transcription 1 (STAT1), knowing that STAT1 depletion by shRNA (figure 2E) or siRNA (online supplementary figure S5) limited IFNG-induced IDO1 and CD274 expression in cancer cells. The knockdown of a single CDK failed to block IFNG-induced STAT1 expression and phosphorylation (on Tyr701; figure 2B). However, dinaciclib (figure 2B) or joint CDK1/2/5 depletion (figure 2D) downreg- ulated the IFNG-induced STAT1 overexpression and hyper- phosphorylation. Bioinformatic analysis (https://www.qiagen. com/us/) predicted that JUN, forkhead box O1, the MYC proto- oncogene, the basic helix-loop-helix transcription factor, MYC- associated factor X, myocyte enhancer factor 2A (MEF2A) and STAT5B would be the top candidate transcription factors acting on the core STAT1 promoter. The knockdown of JUN by siRNAs, but not that of any of the other aforementioned candidates, diminished STAT1 expression and phosphoryla- tion, and blocked subsequent IDO1 and CD274 expression in IFNG-treated CFPAC1 (figure 2F,G) and OVCAR5 cells (online supplementary figure S6). Furthermore, shRNA-mediated JUN depletion in CFPAC1 cells demonstrated that JUN was required for the IFNG-induced mRNA expression of STAT1, IDO1 and PDL1 (online supplementary figure S6). Chromatin immunopre- cipitation analysis confirmed that JUN binds to the promoter of STAT1, but not IDO1 and CD274, in wild-type CFPAC1 cells treated with IFNG (figure 2H), supporting that JUN is the tran- scription factor that stimulates STAT1 expression. In line with the previous finding that JUN is a substrate of CDK,15 dinaciclib or CDK1/2/5 depletion prevented IFNG-induced JUN phos- phorylation at Ser63 and Ser73 (figure 2D and online supple- mentary figure S7). Consequently, the mutation of JUN on these phosphorylation sites abolished JUN binding to the STAT1 promoter (figure 2H) and abrogated the IFNG-mediated induc- tion of STAT1, IDO1 and CD274 mRNAs (figure 2I). These results reveal that CDK1/2/5-mediated JUN phosphorylation is
required for the transactivation of STAT1 by JUN, resulting in STAT1-dependent immune checkpoint expression.

CDK1/2/5 inhibition triggers caspase-dependent apoptosis Most cell cycle-targeting drugs can trigger cell death in a context- dependent manner.16 Compared with dinaciclib, IFNG itself did not cause cancer cell death in vitro (online supplementary figure S8A). In contrast, IFNG enhanced dinaciclib-induced cell death, which was blocked by the apoptosis inhibitor (Z-VAD-FMK), but not by inhib- itors of necroptosis (necrosulfonamide), ferroptosis (ferrostatin-1), nor autophagy (chloroquine; online supplementary figure S8A). As expected, these inhibitors efficiently blocked cell death in response to their corresponding positive stimuli (TZC, erastin and lapatinib, respectively; online supplementary figure S8B). These results indicate that IFNG/dinaciclib induced apoptotic cancer cell death. Consis- tent with this notion, genetic inhibition of apoptosis (eg, in BAX−/−/ BAK−/− cells), but not inhibition of necroptosis (eg, in RIPK1−/−, RIPK3−/− and MLKL−/− cells), autophagy (eg, in ATG3−/−, ATG5−/− and ATG7−/− cells) as well as removal of a ferroptosis-related gene (eg, in GPX4−/− cells), significantly blocked IFNG/dinaciclib-induced cell death in MEF cells (online supplementary figure S8C). Depletion of the apoptosis effector CASP3 by shRNAs (online supplementary figure S8D) also attenuated IFNG/dinaciclib-induced apoptosis in CFPAC1 and OVCAR5 cells (online supplementary figure S8E).
Next, we examined the impact of dinaciclib-mediated immune
checkpoint inhibition on apoptosis induction. Overexpression of IDO1 or kynurenine exposure, but not that of CD274, reversed IFNG/dinaciclib-induced apoptosis (online supplementary figure S8F). Correspondingly, CASP3 activity was inhibited with over- expression of IDO1 or kynurenine exposure, but not that of CD274, in response to IFNG/dinaciclib (online supplementary figure S8G). These findings suggest that dinaciclib-mediated IDO1 downregulation promotes IFNG/dinaciclib-induced apop- tosis through CASP3 activation.
The CDK inhibitor diniciclib inhibited mRNA expression of CD274 (online supplementary figure S8H) and IDO1 (online supplementary figure S8I), and enhanced the apoptosis (online supplementary figure S8J) of CFPAC1 cells induced by IFNG. Z-VAD-FMK blocked apoptosis, but failed to restore mRNA expression of CD274 and IDO1 in CFPAC1 cells after IFNG/ dinaciclib treatment (online supplementary figure S8H-J), indicating that CDK inhibitor-mediated suppression of IFNG- induced checkpoint expression in PDAC cells may not be due to the loss of cell viability.
We next examined the effects of dinaciclib on IFNG-induced CD8+ T cell activation and differentiation. IFNG/dinaciclib treatment did not induce apoptosis in splenic CD8+ T cells (online supplementary figure S8K), suggesting that the treatment with IFNG/dinaciclib is safe for normal CD8+ T cells. To deter- mine whether IFNG/dinaciclib treatment has a direct effect on T cell receptor activation, we conducted a dose–response exper- iment with plate-bound anti-CD3 for 24/48 hour. Compared with the positive control of soluble anti-CD28 stimulation, IFNG/dinaciclib could not significantly induce the expression of early activation markers (CD69, CD25 and CD44; online supplementary figure S8L). IFNG/dinaciclib failed to affect the expression of granzyme B (GZMB) and perforin 1 (PRF1) during cytotoxic T cell differentiation (online supplementary figure S8M). Overall, these findings indicate that IFNG/dinac- iclib may not directly affect the activation and differentiation of CD8+ T cells in vitro. Moreover, activated CD8+ T cells did not affect the expression of KLRK1/NKG2D, HAVCR2/TIM3 and LAG3 in KPC cells in the absence or presence of IFNG/

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.

4 Huang J, et al. Gut 2020;0:1–10. doi:10.1136/gutjnl-2019-320441

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.
GI cancer

Figure 2 Cyclin-dependent kinase (CDK1/2/5) inhibition blocks JUN-dependent signal transducer and activator of transcription 1 (STAT1) expression and activation. (A) Western blot analysis of CDK1-9 protein expression in indicated shRNA-mediated gene knockdown CFPAC1 cells. (B) Western blot analysis of indicated protein expression in indicated gene knockdown CFPAC1 cells following treatment with interferon gamma (IFNG; 10 ng/mL)
in the absence or presence of dinaciclib (100 nM) for 24 hours. The relative ratio of p-STAT1 to STAT1 is shown in the bottom panel (n=3, p<0.05, Kruskal-Wallis test). (C) Heat map of mRNA expression levels of indoleamine 2,3-dioxygenase 1 (IDO1) and CD274 in indicated gene knockdown CFPAC1 cells following treatment with IFNG (10 ng/mL) for 24 hours. (D) Western blot analysis of indicated protein expression in indicated gene knockdown CFPAC1 cells following treatment with IFNG (10 ng/mL) for 24 hours. The relative ratio of p-STAT1 to STAT1 is shown in the right panel (n=3, p<0.05, Kruskal-Wallis test). (E) Knockdown of STAT1 by two different shRNAs-blocked IFNG (10 ng/mL)-induced IDO1 and CD274 expression at 24 hours. (F) Western blot analysis of protein expression in indicated siRNA-mediated gene knockdown CFPAC1 cells. (G) Knockdown of JUN by siRNA-blocked IFNG (10 ng/mL)-induced STAT1, IDO1 and CD274 expression at 24 hours. (H, I) JUN-depleted CFPAC1 cells were transfected with
JUN-WT or JUN-mutant cDNA for 48 hours and then treated with IFNG (10 ng/mL) for 24 hours. Binding of JUN to STAT1, IDO1 or CD274 promoter (H) was then analysed by chromatin immunoprecipitation-qPCR (n=5, p<0.05, Kruskal-Wallis test; ns=no significance). In parallel, the mRNA expression levels of STAT1, IDO1 and CD274 were assayed (I). Data in (C), (H) and (I) are from two independent experiments. Data in (A), (B) and (D–G) are from three independent experiments.
Huang J, et al. Gut 2020;0:1–10. doi:10.1136/gutjnl-2019-320441 5

GI cancer

dinaciclib (online supplementary figure S8N), indicating that these checkpoints expressed in PDAC cells do not change during the T cell response.

CDK1/2/5 inhibition triggers histone-dependent ICD
Certain anticancer reagents can induce an immunogenic form of cell death associated with the translocation or release of damage-associated molecular patterns (DAMPs), which function as adjuvants to activate host antitumour immune responses.6 17 The level of DNA damage marker H2AX (online supplemen- tary figure S9A) and the release of nuclear DAMPs, such as HMGB1 and histones (especially H3 and H4; online supple- mentary figure S9B and S9C), were increased by IFNG/dinac- iclib compared with single treatments with IFNG or dinaciclib in human cancer cells (CFPAC1 and OVCAR5), as well as in murine KPC cells (a cell line derived from pancreatic tumours of K-RasG12D;Tp53R172H;Pdx1-Cre mice). In contrast, Z-VAD-FMK (but not the other cell death inhibitors) or the overexpression of IDO1 (but not the overexpression of CD274) blocked IFNG/ dinaciclib-induced HMGB1, H3 and H4 release (figure 3A), indicating that IDO1 expression itself is a negative regulator of DAMP release during apoptotic death.
To determine which nuclear DAMP release promotes the recruitment and activation of antigen-presenting cells in vitro, we used the mouse dendritic cell (DC) line DC2.4 and bone marrow-derived primary dendritic cells (BMDC) to assess response to supernatants of IFNG/dinaciclib-treated KPC cells. Like treatment with recombinant H3/4 proteins, these superna- tants induced cell migration (figure 3B) and mRNA expression of costimulatory molecules (CD80 and CD86; figure 3C) in DC2.4 and BMDC. Flow cytometry analysis also found that the surface expression of CD80 or CD86 was increased in DC2.4 cells (figure 3C). The antibody-mediated neutralisation of H3 and H4 (hereafter referred to as H3/4), but not that of HMGB1, reduced the supernatant-induced cell migration (figure 3B) and expression of CD80 and CD86 (figure 3C). Anti-H3/4 antibodies also abolished the capacity of IFNG/dinaciclib-treated KPC cells to prime T cells for the production of effector cytokines (IFNG, tumour necrosis factor (TNF) and IL2; figure 3D).
To determine whether IFNG/dinaciclib-induced ICD
suppressed tumour growth in vivo, IFNG/dinaciclib-treated KPC cells were injected subcutaneously into the right flank of immunocompetent C57BL/6 mice. Used as a tumour vaccine, the IFNG/dinaciclib-treated KPC cells were able to protect the majority of mice against rechallenge with live KPC cells injected into the opposite flank 1 week later. When IFNG/dinaciclib- killed KPC cells were coadministered together with anti-H3/4 antibodies, they failed to elicit protection against KPC cancers (figure 3E). In addition to the release of DAMP, the exposure of calreticulin (CALR) on the dying cell’s surface serves as an eat-me signal to mediate dinaciclib-induced ICD.18 Knockdown of CALR by siRNA in KPC cells also reversed IFNG/dinaciclib- induced ICD (figure 3E). Similar observations were obtained in BALB/c mice subjected to IFNG/dinaciclib-treated CT26 cells (figure 3F), a model of an immunologically hot tumour. Overall, these results indicate that IFNG/dinaciclib induces ICD in vitro and in vivo.

Anticancer activity of dinaciclib in combination with IFNG in vivo
Finally, we investigated the potential utility of dinaciclib in combination with IFNG for pancreatic cancer therapy in vivo, in three distinct preclinical mouse models: subcutaneous,
orthotopic and transgenic models. Compared with IFNG or dinaciclib alone, IFNG/dinaciclib administration blocked KPC or PANC02 tumour growth in subcutaneous models in C57BL/6 mice (online supplementary figure S10A, two-way analysis of variance (ANOVA) with repeated measures), correlating with decreased IDO1 (online supplementary figure S10B) and CD274 expression (online supplementary figure S10C), increased CASP3 activation (online supplementary figure S10D) and enhanced CD8+ T cell (but not CD4+ T, B, and NK cell) infil- tration (online supplementary figure S10E-H) of the tumours. Serum markers of tissue damage, such as blood urea nitrogen and alanine aminotransferase, were not changed by IFNG/dinac- iclib (online supplementary figure S10I), suggesting that this combination treatment is safe.
We further observed that treatment with IFNG/dinaciclib
prolonged the survival of C57BL/6 mice after the orthotropic implantation of syngeneic KPC cells into the pancreas (online supplementary figure S11A). Accordingly, tumour weight (online supplementary figure S11B) was reduced by the IFNG/ dinaciclib combination treatment, and this effect was associ- ated with decreased IDO1 (online supplementary figure S11C) and CD274 expression (online supplementary figure S11D), increased CASP3 activation (online supplementary figure S11E), and augmented CD8+ T cell (but not CD4+ T, B and NK cell) infiltration (online supplementary figure S11F).
Compared with KPC mice, conditional HMGB1 depletion in the pancreas (Pdx1-Cre;K-RasG12D/+;hmgb1−/−, termed KCH mice) significantly accelerates the initiation and progression of KRAS-driven PDAC in mice.11 19 IFNG/dinaciclib treatment prolonged the survival of KCH mice (figure 4A), as it reduced pancreatic ductal lesion formation and desmoplastic stromal responses (figure 4B,C). IDO1 (figure 4D) and CD274 expres- sion (figure 4E) was downregulated, whereas CASP3 activation (figure 4F) and CD8+ T cell (but not CD4+ T cell) infiltra- tion (figure 4G,H) were increased in tumours from KCH mice following IFNG/dinaciclib treatment. To further characterise the influence of IFNG/dinaciclib on the TME, we analysed cytokine mRNA profiles in tumour tissues by quantitative PCR (Q-PCR). These studies revealed that IFNG/dinaciclib treatment of estab- lished KCH tumours increased Th1 (eg, IFNG, TNF and IL2) and decreased Th2 (eg, IL4, IL5 and IL13) cytokine mRNA production, indicative of a generally proinflammatory cytokine profile within the TME (figure 4I). Consistently, circulating levels of IFNG, TNF and IL2 (figure 4J) were elevated by IFNG/ dinaciclib treatment.
We further assayed the activation status of the CD8+ T cells.
Cytotoxic T lymphocytes uses the perforin/granzyme cytotoxic pathway to kill tumours. The expression of GZMB and PRF1 was increased in CD8+ T cells from the IFNG/dinaciclib treat- ment group (figure 4K). Other T-cell activation markers, such as CD69, CD44 and CD25, were upregulated in CD8+ T cells from the IFNG/dinaciclib treatment group (figure 4K). The downregulation of CD62L was also observed in these CD8+ T cells (figure 4K). Interestingly, the expression of PDCD1/PD-1 and chemokine receptors (eg, C-X-C motif chemokine receptor 3 (CXCR3) and CXCR4) in CD8+ T cells was not significantly changed (figure 4K). Similar data on CD8+ T-cell activation was found in subcutaneous models (online supplementary figure S10J). These findings indicate the accumulation of activated CD8+ T cells in TME following IFNG/dinaciclib treatment.
In addition to their expression in cancer cells, CD274 and IDO1 are expressed by macrophages, which also play a role in regulating pancreatic tumour progression. Dinaciclib treatment inhibited IFNG-induced CD274 and IDO1 mRNA expression

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.

6 Huang J, et al. Gut 2020;0:1–10. doi:10.1136/gutjnl-2019-320441

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.
GI cancer

Figure 3 Cyclin-dependent kinase 1/2/5 inhibition triggers histone-dependent immunogenic cell death. (A) CFPAC1 cells were treated with interferon gamma (IFNG; 10 ng/mL)/dinaciclib (100 nM) in the absence or presence of Z-VAD-FMK (20 µM), ferrostatin-1 (500 nM), necrosulfonamide (1 µM), chloroquine (50 µM) or the indicated cDNA for 24 hours. Extracellular high-mobility group box 1 (HMGB1), H3, and H4 were assayed by ELISA (n=3, *p<0.05, Kruskal-Wallis test). (B) Anti-H3/4 Abs (anti-H3 Ab (10 mg/mL) and anti-H4 Ab (10 mg/mL)), but not anti-HMGB1 Ab (10 mg/mL) or control IgG (10 mg/mL), inhibited migration of DC2.4 or bone marrow-derived primarydendritic cells (BMDC) in response to IFNG/dinaciclib-treated KPC cells. Recombinant H3/4 protein (100 ng/mL) treatment was used as a control (n=5, *p<0.05, Kruskal-Wallis test). (C) In parallel, the mRNA or surface expression of CD80 and CD86 in DC2.4 or BMDC was assayed by Q-PCR or flow cytometry (n=3–5, *p<0.05, Kruskal-Wallis test). (D) IFNG/ dinaciclib-treated KPC cells were cocultured with mouse splenic CD8+ T cells at the ratio of 1:1 in the absence or presense of anti-H3/4 Abs, anti- HMGB1 Ab or control IgG at 10 mg/mL for 24 hours. The levels of IFNG, tumour necrosis factor (TNF) and interleukin-2 (IL2) in the cell culture medium were assayed (n=5, *p<0.05, Kruskal-Wallis test). (E, F) The depletion of H3/H4 (but not HMGB1 and control IgG) with specific blocking antibodies (20 mg/kg) or knockdown of CALR by siRNA abolished the capacity of IFNG/dinaciclib-treated tumour cells to vaccinate against KPC or CT26 tumour cells in C57BL/6 or BALB/c mice, respectively. The percentage of tumour-free mice is indicated (n=8–10 mice/group, *p<0.05, two-way analysis of variance test). Data in (E) and (F) are from two independent experiments. Data in (A–D) are from three independent experiments.

Huang J, et al. Gut 2020;0:1–10. doi:10.1136/gutjnl-2019-320441 7

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.
GI cancer

Figure 4 Anticancer activity of dinaciclib in combination with interferon gamma (IFNG) in vivo. (A) Combined administration with IFNG (2 µg/ mice/i.p., three times weekly) and dinaciclib (8 mg/kg/i.p., three times weekly) prolonged survival in KCH mice at 14 weeks of age (n=8–12 mice/ group, *p<0.05, log-rank test). (B–J) In parallel, pancreatic histology (B, C), immune checkpoint expression (D, E), CASP3 activity (F), CD4 and CD8 T-cell infiltration (G, H), gene mRNA expression (I) in isolated tumour or macrophages, serum cytokine level (J) and flow cytometry-based analysis of immune molecule expression (intracellular: granzyme B (GZMB) and perforin 1 (PRF1); surface: CD69, CD44, CD25, CD62L, PDCD1, CXCR3 and CXCR4) in isolated CD8 T cells (K) at 8 weeks of age were assayed (n=3–5 mice/group, *p<0.05, t or Kruskal-Wallis tests, bar=100 µm). (L) Kaplan-
Meier survival analysis of indoleamine 2,3-dioxygenase 1 (IDO1) gene expression in patient with pancreatic ductal adenocarcinoma using data from the Cancer Genome Atlas. (M) Schematic summary of the role of cyclin-dependent kinase 1/2/5 in the regulation of IFNG-induced adaptive tumour immune resistance. Data in (I–K) are from two independent experiments. Data in (A–H) are from three independent experiments. ARG1, arginase 1; IL2, interleukin-2; NOS2, nitric oxide synthase 2; TNF, tumour necrosis factor.

8 Huang J, et al. Gut 2020;0:1–10. doi:10.1136/gutjnl-2019-320441

GI cancer

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.
in isolated macrophages (F4/80+ cells) in KCH mice (figure 4I). The expression of nitric oxide synthase 2 (an antitumour M1 marker), but not arginase 1 (a protumour M2 marker), was upregulated in macrophages from the IFNG/dinaciclib treatment group (figure 4I). Together, these findings suggest that CDK may have a broad role in regulating immune checkpoint expression in the pancreatic TME.
To further investigate the requirement of CD8+ cells in promoting tumour rejection following IFNG/dinaciclib treat- ment, CD8+ T cells were selectively depleted using an anti-CD8 monoclonal antibody. Compared with the control IgG group, IFNG/dinaciclib-mediated tumour suppression in the subcu- taneous or orthotopic models was reversed by treatment with anti-CD8-depleting antibodies (online supplementary figure S12A,B, two-way ANOVA with repeated measures). IFNG/ dinaciclib-induced CD8+ T cells infiltration was diminished by treatment with anti-CD8-depleting antibodies (online supple- mentary figure S12C,D). Collectively, these preclinical findings demonstrated that the IFNG/dinaciclib combination therapy could possess significant CD8+ T cell-dependent antitumour activity in vivo.

DISCUSSION
Our current findings suggest that immunochemotherapy with IFNG and dinaciclib can boost an effective antitumour immune response to turn an immunologically ‘cold’ TME into a ‘hot’ one. Compared with other traditional methods of immunotherapy, our approach not only appears to overcome IFNG-mediated immune checkpoint expression, but also leads to immunogenic apoptotic cell death, which results in the intratumoural expres- sion of proinflammatory cytokines and infiltration by CD8+ cytotoxic T lymphocytes. Although an early study suggested that dinaciclib can trigger ICD in mouse colon cancer cells, the key mediator of this type of ICD remains unknown.18 Our results highlight that histone H3 and H4 are nuclear DAMPs respon- sible in this model for apoptotic ICD because that antihistone treatment blocks dinaciclib-induced ICD.
Individual tumours induce distinct pathways of immune check- point reprogramming to escape immune surveillance. Compared with other immune checkpoint molecules, IDO1 plays an important role in the pancreatic TME to promote immune resistance and tumour progression.20 Bioinformatic analyses of the Cancer Genome Atlas revealed that the abnormal expres- sion of CDK1 (online supplementary figure S13), CDK2 (online supplementary figure S14), CDK5 (online supplementary figure S15) and IDO1 (online supplementary figure S16) was associ- ated with poor patient survival in several cancer types, including pancreatic cancer (figure 4L). The IDO1 inhibitor indoximod, combined with gemcitabine/nab-paclitaxel, shows encouraging response rates in patients with metastatic pancreatic cancer (Clin- icalTrials.gov identifier: NCT02077881). In contrast to a report showing that CDK4/6 inhibitors stabilise the CD274 protein in certain cancers (breast, colon or melanoma),21 we found that the CDK1/2/5 inhibitor dinaciclib reduces IFNG induced the expression of CD274 and IDO1 in 24 human or mouse cancer cell lines. This effect was due to blockade of STAT1-dependent gene transcription, supporting a context-dependent role for the CDK family in the regulation of CD274 expression. Unlike of earlier works on CDK5-mediated CD274 expression in medul- loblastoma cells,22 we demonstrated that knockdown of single CDKs (eg, CDK1, CDK2, CDK5 and CDK9) fail to block IFNG- induced IDO1 and CD274 expression in PDAC cells, indicating a different role of CDK5 in these types of cancer.
Huang J, et al. Gut 2020;0:1–10. doi:10.1136/gutjnl-2019-320441
Our results suggest a treatment option to overcome IFNG- mediated adaptive tumour immune resistance by using the CDK1/2/5 inhibitor dinaciclib in the context of pancreatic cancer (figure 4M). This immunochemotherapeutic approach enhances tumour cell apoptosis and stimulates CD8+ T cell- dependent antitumour immunity, as it remarkably improves overall survival rates in multiple animal models of pancreatic cancer. Given the emerging roles of the CDK family in cancer immunity as well as the complexity of animal models often limiting detailed mechanistic interpretation of experimental findings, further studies are needed to explore the impact of dinaciclib alone or with other immunologically active agents in the treatment of pancreatic cancer using more clinically relevant models (eg, organoids). In addition, the safety and tolerability of the long-term use of IFNG/dinaciclib need to be further investigated, although our current animal study indi- cates that the combination of IFNG and dinaciclib do not lead to liver and kidney injury.

Author affiliations
1The Third Affiliated Hospital, Guangzhou Medical University, Guangzhou, Guangdong, China
2Department of Oncology, Xiangya Hospital, Central South University, Changsha, Hunan, China
3Department of Hepatobiliary Surgery, Hunan Cancer Hospital, Changsha, Hunan, China
4Department of Ophthalmology, the Second Xiangya Hospital, Central South University, Changsha, Hunan, China
5Department of Surgery, UT Southwestern Medical Center, Dallas, Texas, USA 6Cancer Research Institute, Central South University, Changsha, Hunan, China 7Cancer Center, Sun Yat-Sen University, Guangzhou, Guangdong, China 8Endoscopy Center, China-Japan Union Hospital, Jilin University, Changchun, Jilin, China
9Equipe Labellisée Par la Ligue Contre le Cancer, Université Paris Descartes, Paris, Île-de-France, France
10Metabolomics and Cell Biology Platforms, Gustave Roussy Cancer Campus, Villejuif, France
11Pôle de Biologie, Hôpital Européen Georges Pompidou, AP-HP, Paris, France 12Suzhou Institute for Systems Biology, Chinese Academy of Sciences, Suzhou, China 13Department of Women’s and Children’s Health, Karolinska University Hospital, Stockholm, Sweden
14Department of Surgery, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
Acknowledgements The authors thank Dave Primm (Department of Surgery, University of Texas Southwestern Medical Center) for his critical reading of the manuscript.
Contributors DT and RK designed the experiments. JH, PC, KL, JL, BZ, RW, QP, CL, Z-XL and DT conducted the experiments. DT wrote the paper. GK, HZ and ML assisted in data interpretation and edited the manuscript.
Funding JL is supported by grants from the National Natural Science Foundation of China (31671435, 81400132 and 81772508). CL is supported by a grant from the National Natural Science Foundation of China (81872323).
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.
ORCID iDs
Ze-Xian Liu http://orcid.org/0000-0001-9698-0610 Daolin Tang http://orcid.org/0000-0002-1903-6180

REFERENCES
⦁ Lotze MT, Matory YL, Ettinghausen SE, et al. In vivo administration of purified human interleukin 2. II. half life, immunologic effects, and expansion of peripheral lymphoid cells in vivo with recombinant IL 2. J Immunol 1985;135:2865–75.
⦁ Ribas A, Wolchok JD. Cancer immunotherapy using checkpoint blockade. ⦁ Science
2018;359:1350–5.
9

GI cancer

⦁ Topalian SL, Drake CG, Pardoll DM. Immune checkpoint blockade: a common denominator approach to cancer therapy. ⦁ Cancer⦁ ⦁ Cell⦁ 2015;27:450–61.
⦁ Waters AM, Der CJ. Kras: the critical driver and therapeutic target for pancreatic cancer. ⦁ Cold Spring Harb Perspect Med 2018;8. doi:10.1101/cshperspect.a031435. [Epub ahead of print: 04 Aug 2018].
⦁ Galluzzi L, Vitale I, Warren S, et al. Consensus guidelines for the definition, detection and interpretation of immunogenic cell death. ⦁ J Immunother Cancer 2020;8:e000337.
⦁ Tang D, Kang R, Berghe TV, et al. The molecular machinery of regulated cell death. ⦁ Cell ⦁ Res⦁ 2019;29:347–64.
⦁ Yang M, Li C, Zhu S, et al. Tfam is a novel mediator of immunogenic cancer cell death.
Oncoimmunology 2018;7:e1431086.
⦁ DeVorkin L, Pavey N, Carleton G, et al. Autophagy Regulation of Metabolism Is Required for CD8+ T Cell Anti-tumor Immunity. ⦁ Cell⦁ ⦁ Rep⦁ 2019;27:502–13.
⦁ Song X, Zhu S, Xie Y, et al. JTC801 induces pH-dependent death specifically in cancer cells and slows growth of tumors in mice. ⦁ Gastroenterology⦁ 2018;154:1480–93.
⦁ Xie Y, Zhu S, Zhong M, et al. Inhibition of Aurora Kinase A Induces Necroptosis in Pancreatic Carcinoma. ⦁ Gastroenterology⦁ 2017;153:1429–43.
⦁ Kang R, Xie Y, Zhang Q, et al. Intracellular HMGB1 as a novel tumor suppressor of pancreatic cancer. ⦁ Cell⦁ ⦁ Res⦁ 2017;27:916–32.
⦁ Winograd R, Byrne KT, Evans RA, et al. Induction of T-cell immunity overcomes complete resistance to PD-1 and CTLA-4 blockade and improves survival in pancreatic carcinoma. ⦁ Cancer⦁ ⦁ Immunol⦁ ⦁ Res⦁ 2015;3:399–411.
⦁ Foucher ED, Ghigo C, Chouaib S, et al. Pancreatic ductal adenocarcinoma: a strong imbalance of good and bad immunological cops in the tumor microenvironment. ⦁ F⦁ ront Immunol⦁ 2018;9:1044.
⦁ Paruch K, Dwyer MP, Alvarez C, et al. Discovery of dinaciclib (SCH 727965): a potent and selective inhibitor of cyclin-dependent kinases. ⦁ ACS⦁ ⦁ Med⦁ ⦁ Chem⦁ ⦁ Lett 2010;1:204–8.
⦁ Vanden Bush TJ, Bishop GA. Cdk-Mediated regulation of cell functions via c-Jun phosphorylation and AP-1 activation. ⦁ PLoS⦁ ⦁ One⦁ 2011;6:e19468.
⦁ Otto T, Sicinski P. Cell cycle proteins as promising targets in cancer therapy. ⦁ Nat⦁ ⦁ Rev ⦁ Cancer⦁ 2017;17:93–115.
⦁ Galluzzi L, Buqué A, Kepp O, et al. Immunogenic cell death in cancer and infectious disease. ⦁ Nat⦁ ⦁ Rev⦁ ⦁ Immunol⦁ 2017;17:97–111.
⦁ Hossain DMS, Javaid S, Cai M, et al. Dinaciclib induces immunogenic cell death and enhances anti-PD1-mediated tumor suppression. ⦁ J⦁ ⦁ Clin⦁ ⦁ Invest 2018;128:644–54.
⦁ Kang R, Tang D. The dual role of HMGB1 in pancreatic cancer. Journal of Pancreatology 2018;1:19–24.
⦁ Zhang T, Tan X-L, Xu Y, et al. Expression and prognostic value of indoleamine 2,3-dioxygenase in pancreatic cancer. ⦁ Chin⦁ ⦁ Med⦁ ⦁ J⦁ 2017;130:710–6.
⦁ Zhang J, Bu X, Wang H, et al. Cyclin D-Cdk4 kinase destabilizes PD-L1 via cullin 3- SPOP to control cancer immune surveillance. ⦁ Nature⦁ 2018;553:91–5.
⦁ Dorand RD, Nthale J, Myers JT, et al. Cdk5 disruption attenuates tumor PD-L1 expression and promotes antitumor immunity. ⦁ Science⦁ 2016;353:399–403.

Gut: first published as 10.1136/gutjnl-2019-320441 on 14 August 2020. Downloaded from http://gut.bmj.com/ on August 15, 2020 at Western Sydney University. Protected by copyright.SCH727965