Mouse colorectal cancer cells (CT26) or triple-negative breast cancer cells (4T1) and human leukemic cells (Jurkat, MOLT-4) were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). Mouse colorectal cancer cells (C51) were from MP Colombo (Istituto Nazionale Tumori, Milan, Italy). All cell lines were periodically tested to avoid mycoplasma contamination using a universal mycoplasma detection kit (ATCC). Cells were cultured in RPMI (CT26, C51, Jurkat, MOLT-4) or DMEM high glucose (4T1) media supplemented with 10% of fetal bovine serum (FBS) (Dominique Dutscher, Bernolsheim, France) at 37°C with 5% of CO₂. Cells were treated with 10 µM of H89 (Cayman Chemicals, Tallinn, Estonia) at different time points.

Seven-week-old female BALB/c mice or Swiss nude mice were purchased from Charles River Laboratory (Saint-Germain-Nuelles, France) or the animal care facility at the University of Burgundy (Dijon, France). Mice were inoculated with subcutaneous injection in the right flank, and under isoflurane anesthesia, with 5 × 10⁵ colon CT26, C51, or 4T1 cell lines. After 6 days, when tumors reached ≈50 mm³, animals were randomly assigned to different groups (5 or 7 mice per cage) and treated with H89 (10 mg/kg in 100 µl of NaCl) with intraperitoneal (i.p.) injection twice a week. For TL CD8⁺ depletion, an anti-CD8a antibody (BioXCell, Euromedex, Mundolsheim, France) 500 µg in 100 µl NaCl or IgG2A control isotype (BioXCell) 500 µg in 100 µl of NaCl was injected (i.p.) together with the first H89 administration and renewed once a week. For IL-15RA blockade, an anti-TMB1 (BioXCell (catalog number BE0298) 50 µg in 100 µl of NaCl) or IgG2A control isotype (BioXCell 50 µg in 100 µl of NaCl) was injected (i.p.) the day before the first H89 administration and renewed twice a week. For the H89/5-FU combination, 5-FU (5 mg/kg in 100 µl NaCl) was injected (i.p.) weekly at the same time as H89 administration. For the oral administration, H89 was given twice a week by gavage (5 mg/kg in NaCl). For prophylaxis experience, H89 was administrated by i.p. injections (10 mg/kg in NaCl) 3 days before CT26 implantation. Tumor growth was monitored three times a week. For AOM/DSS-induced carcinogenesis, colorectal polyps were induced by azoxymethane (AOM) and dimethylhydrazine (DMH) i.p. injections before dextran sulfate sodium (DSS) administration in drinking water according to the protocol by Rosenberg et al. (23). For this experience, mice were treated with H89 (5 mg/kg in 100 µl NaCl by gavage) twice a week. Animals’ health status was monitored by observing weight variations three times a week, and mice were killed if their weight loss was higher than 20% of their initial weight. After 90 days, mice were killed and colorectal polyps were quantified.

NK cell depletion was performed using an anti-Asialo GM1 antibody (FUJIFILM Wako Chemicals, Neuss, Germany). BALB/c mice received the anti-Asialo GM1 antibody (10 µl/100 µl NaCl by i.v. injection according to the supplier’s instructions) on days −5, −3, and −1 before and then every 5 days after CT26 cell injection. H89 was injected three times from D-3 to D-1. Tumor growth was monitored three times a week.

The animal care staff observed the animals’ state each day, and we monitored their behavior. When the volume of the tumor reached 2,000 mm³, animals were killed. The experiments were approved by the ethics committee “C2EA Grand Campus Dijon No. 105” and by the ministry of research with the protocol codes 28546 and 33240.

CT26 tumor-bearing mice were treated with H89 (10 mg/kg in 100 µl of NaCl, i.p.) for 10, 14, or 21 days (10 mice/time point). At each time point, mice were killed and tumors were dissociated with collagenase and DNase solution before lysis of red blood cells. For membrane staining, cells were incubated with Flow Cytometry Staining Buffer (FCSB) for 15 min and then for 30 min at 4°C with membrane antibodies. For intracellular staining, cells were fixed and permeabilized with the Cytofix/Cytoperm™ kit from BD Biosciences (Le Pont de Claix, France) before incubation with antibodies for 30 min at 4°C. One million of cells were stained with viability FVS 700 (BD Biosciences) and TL-specific antibody panels (Supplementary Tables S2–S4) and analyzed by flow cytometry. All the acquisition were performed using a BD FACS Canto or an LSRII cytometer using the BD FACSDiva software (BD Biosciences). Data were analyzed using the FlowJo software v10.

Total RNA extraction was performed as per the manufacturer’s instructions. Sequencing, data quality, and reds repartition were performed by the BGIseq500 platform (BGI-Shenzhen, China). Differential expression analysis was performed with the DESeq2 R package (https://doi.org/10.1186/s13059-014-0550-8). Treatment groups were compared. Raw p-values associated with each gene were adjusted using Bonferroni correction, as advised. Gene Set Enrichment Analysis (GSEA) was used to identify biological pathways that are enriched in the lists of differential gene characteristics of each group of patients. Pathways from the Hallmark database were used. GSEA was conducted using the clusterProfiler R package (https://doi.org/10.1089/omi.2011.0118).

Naïve CD4⁺ or CD8⁺ T cells were isolated from the spleen and lymph nodes of female Balb/c mice using Miltenyi isolation kits (Miltenyi Biotech, Paris, France). Naive CD4⁺ T cells (CD62L⁺) were differentiated into Th1 or T_reg cell subtypes using cytokines and blocking antibody cocktails: anti-IL-4 (10 µg/ml) and IL-2 (10 µg/ml) for Th1 and anti-IFN-γ (10 µg/ml), anti-IL-4 (10 µg/ml), and TGFβ (4 ng/ml) for T_reg cells (24). During T-cell differentiation, Th1 and T_reg cells were treated with 1 µM of H89 for 72 h and T-cell differentiation was assessed by RT-qPCR. Total CD4⁺ T cells (CD62L⁻) were obtained using the negative fraction and were treated with 1 µM of H89 for 2 h to assess intracellular potassium [K⁺]_i concentration, as previously reported (25). CD8⁺ T cells were cultured for 72 h with 1 µM of H89 to assess the level of Akt phosphorylation and granzyme B production. All T cells were activated using anti-CD3/anti-CD28 Dynabeads (Gibco - Thermo Fischer Scientific, Waltham, MA, USA) and cultured in antibiotics-enriched RPMI medium [10% FBS with 1% penicillin, streptomycin, and amphotericin B (PSA)].

B16-OVA melanoma cancer cells were seeded in a 96-well plate (30,000 cells per well) in 200 µl of RPMI 10%FBS 1%PSA medium after cell trace™ violet labeling (Invitrogen - Thermo Fischer Scientific). Splenic CD8⁺ T cells were isolated from OT-I female mice using the CD8a⁺ T-cell Isolation Kit and cocultured with B16-OVA cells at different ratios of B16-OVA : CD8⁺ T cells) and treated, or not, with 5 µM of H89. After 48 h of coculture, B16-OVA cell viability was analyzed by flow cytometry using a Fixable Viability stain 700 (FVS 700) (BD Biosciences) and using the cell trace™ violet labeling to differentiate B16-OVA cells from CD8⁺ T cells.

Jurkat and MOLT-4 cells were lysed with a buffer containing imidazole-HCl, EDTA, and EGTA (Sigma-Aldrich, Saint-Quentin-Fallavier, France) with a protease and phosphatase inhibitor cocktail (Roche Life Science, Penzberg, Germany). After protein quantification with the Bio-Rad DC protein assay as recommended by the manufacturer (Bio-Rad, Marnes-la-Coquette, France), the enzymatic activity of the phosphatase PP2A was assessed using the PP2A Immunoprecipitation Phosphatase Assay Kit (Millipore, Guyancourt, France) as per the manufacturer’s procedures.

Th1 and Treg cells as well as mice tumors were collected using the Trizol reagent (Ambion - Thermo Fischer Scientific) to perform total RNA extraction. Five hundred nanograms of total RNA were reverse transcribed into cDNA using a M-MLV reverse transcriptase, recombinant RNasin^® Plus RNase inhibitor, and random primers (Promega, Charbonnières-les-Bains, France). cDNA level was assessed by real-time PCR using PowerUpTM SYBR™ Green (Applied biosystems - Thermo Fischer Scientific) on ViiA™7 Real-Time PCR System (Applied biosystems). Relative mRNA expression was quantified using the 2^−ΔCt formula after actin normalization (ΔCt). Primers were purchased at (Eurogentec, Seraing, Belgium) and are listed in Supplementary Table S1.

For PD-1, PD-L1, and CD80 expression analysis, CT26 or MOLT-4 cells, treated or not in vitro with H89, were stained using the tumor membrane staining protocol described above. All the antibodies used in this paper are listed in Supplementary Table S5. All the acquisitions were performed using a BD FACS Canto or an LSRII cytometer using the BD FACSDiva software (BD Biosciences). Data were analyzed using the FlowJo software v10.

CT26 tumors collected from BALB/c mice were cut in 4-µm-thick slices after formalin fixation and paraffin embedding. After demasking in a water bath with Tris-EDTA buffer (20 min, 95°C, pH 9), the slices were left to cool for 10 min at RT. Inhibition of endogenous horseradish peroxidase (HRP) was performed with H₂O₂ (3%) in PBS followed by saturation in TBS-Tween BSA 3%. Slices were then incubated for 1 h at RT with an anti-CD4 (dilution 1:800) or anti-CD8 antibody (dilution 1:500) (HistoSure HS-360108 or 361008) and then with a secondary antibody (Impress anti-rabbit IgG 5+L, MP-7401-50) ( Vector® Laboratories, Burlingame, CA, USA) with a substrate kit HRP (Vector^® Laboratories, NovaRED Substrate Kit SK-4800). The samples were then exposed to a Harris hematoxylin bath for 10 s before dehydration, followed by mounting in an organic medium.

Cell lysates were prepared and quantified, and proteins were separated in a 10% SDS-PAGE gel and transferred on a nitrocellulose membrane, as previously described (26). Membranes were incubated overnight at 4°C with anti-pAkt or anti-Akt antibodies (Cell signaling, Danvers, MA, USA) and HSC70 as an endogenous loading control (Santa Cruz Biotechnology, Dallas, TX, USA). Membranes were washed and incubated for 1 h with HRP-conjugated secondary antibody (Jackson Immunoresearch Laboratories, West Grove, PA, USA) and revealed using the Clarity™ Western ECL Substrate (Bio-Rad) and the ChemiDoc imaging system (Bio-Rad).

Statistical analyses were performed using paired or unpaired Student’s t-test one or two-tailed with significance determined at p ≤ 0.05 for in vitro and ex vivo experiments. For in vivo experiments, we used two-way ANOVA with Bonferroni’s correction for multiple comparisons. The GraphPad Prism 7 software was used to all statistical analyses, and differences were considered statistically significant at ^*p ≤ 0.05, ^**p ≤ 0.01; ^***p ≤ 0.001; ^****p ≤ 0.0001.

Based on our previous results, which showed that H89 reduces intestinal inflammation in a murine model of DSS-induced colitis (27), we investigated its prophylactic potential. In a mouse model of AOM/DSS colitis-associated intestinal carcinogenesis, we showed that H89 decreases tumor incidence as attested by a significant reduction in the total number of colorectal polyps (Figures 1A, B). In parallel, using a CT26 tumor-bearing mouse model, mice receiving H89 for 3 days in a row prior to implantation of CT26 cells, showed a significant decrease in tumor incidence (Figure 1C). We hypothesized that such an effect might involve an early innate immune response as NK cells. As shown in Figure 1D, we observed a significant antitumor effect in mice receiving anti-Asialo GM1 (used to deplete NK cells) compared with the control. Such effect is less pronounced than in H89-treated mice. However, the combination of H89 with anti-Asialo GM1 tends to inhibit H89-mediated tumor growth prevention but not to a significant extent. This result suggested that the preventive effect of H89 may involve other innate immune cells since the combination also significantly prevented tumor growth (Figure 1D).

Besides its prophylactic potential, we investigated whether H89 may have an antitumor therapeutic effect. We investigated the role of the adaptive immune system on mouse models of colon cancer. Intraperitoneal or per os administration (one route of administration used in anticancer treatments in humans) of H89 in BALB/c mice bearing CT26 or C51 cells significantly decelerated tumor progression (Figures 2A–C). Such antitumor effect of H89 was not restricted to colon cancer and was also observed in a triple-negative breast cancer mouse model, known for its aggressiveness and resistance to treatment (Figure 2D). We also investigated whether H89 may potentiates chemotherapy used in CRC such as 5-FU. Importantly, even though each agent used alone reduced CT26 tumor growth, their combination significantly enhanced their antitumor capacity (Figure 2E). We further tested whether H89-mediated tumor growth inhibition depends on the immune system. We then induced tumors in Swiss nude mice, lacking T cells, and treated them with H89. The monitoring of tumor growth showed that the antitumoral activity of H89 is stopped in Swiss nude mice (Figure 2F). We also showed that the depletion of CD8⁺ TLs blunted the antitumor efficacy of H89 (Figure 2G). We also observed that the anti-CD8⁺ antibody alone increased tumor growth compared with the vehicle, indicating the suppressive function of CD8⁺ TLs on tumor growth. Altogether, our results reveal that H89 antitumor efficacy is dependent on an adaptive immune response.

We investigated the effect of H89 on the intratumoral infiltration of CD8⁺ TLs and their state of activation and function. Quantification of intratumoral CD8⁺ TLs by flow cytometry analysis revealed that H89 significantly increased the amount of total CD8⁺ TLs (CD45⁺/CD3⁺/CD8⁺ cells) on day 21 but not on days 10 and 14 posttumor cell injection (Figure 3A). We found that H89 significantly increases the number of effector memory CD8⁺ TLs on day 14 but not the central memory TLs (Figure 3B; Supplementary Figure S1). Such effects were confirmed by IHC staining (Figure 3C). To depict how H89 mediates tumor recruitment of CD8⁺ TLs, we analyzed by RT-PCR the mRNA expression level of the chemokine CXCL10, known to attract CD8⁺ TLs in the TME. The results showed that the H89 significantly increased the expression of the CXCL10 gene transcript on day 14 (Figure 3D). We also analyzed the impact of H89 on the activation of CD8⁺ TLs by measuring interferon-gamma (IFN-γ) and granzyme B expression. We showed that H89 significantly increases the amount of IFN-γ mRNA within the tumor and IFN-γ production in CD8⁺ TLs (Figures 3E, F). Regarding granzyme B, H89 significantly raised its expression on day 10 (Figure 3G). A similar effect was obtained when CD8⁺ TLs isolated from the spleen lymph nodes of BALB/c mice were treated in vitro with H89 for 72 h (Figure 3H). We next used the OT-I/B16-OVA mouse model to investigate whether the activation of CD8⁺ by H89 correlates with their cytotoxic activity. Thus, we investigated the administration of H89 on ex vivo cocultured CD8⁺ TLs, purified from OT-I mice, with melanoma B16-OVA for 48 h significantly decreased melanoma B16-OVA cell viability in 1:5 and 1:10 ratios (target:effector) (Figure 3I).

In parallel, we investigated CD4⁺ TL infiltrates in H89-treated and H89-untreated mice bearing CT26 tumors. We showed that H89 significantly increases the amount of intratumoral CD4⁺ TLs (CXCR3⁺/CCR6⁻) (Th1) expressing IFN-γ (CD4⁺/IFN-γ⁺ TLs) (Figures 4A, B). In addition, we confirmed the intratumoral presence of Th1 cells by RT-qPCR analysis of T-bet expression, a well-known key transcription factor of Th1 cell differentiation. Indeed, H89 significantly increased the expression of T-bet on days 14 and 21 (Figure 4C). Conversely, CT26 tumors treated with H89 exhibited a decreased level of CD4⁺ T_regs (CD25⁺/CD127⁻) (Figure 4D). However, we failed to detect significant differences in Th2 (IL-4⁺ TLs) and Th17 (IL-17⁺ TLs) from untreated versus H89-treated tumors (Supplementary Figure S2A, B). We also evaluated the ability of H89 to induce the differentiation of Th1 and Treg cells in vitro. Differentiation of CD4⁺ naïve T cells isolated from the spleen and lymph nodes of BALB/c mice showed that H89 promotes Th1 differentiation as attested by a significant increase of T-bet transcription (Figure 4E). Conversely, H89 did not promote T_reg cell differentiation attested by the inhibition of FoxP3 gene expression (Figure 4F).

We analyzed by flow cytometry the impact of H89 on some immunosuppressive molecules expressed at the surface of TLs or cancer cells. We found that H89 significantly decreased the expression of PD-1 on total CD4⁺ TLs (isolated from the spleen of BALB/c mice) and on the CD8⁺-like cell line MOLT-4 (Figures 5A, B). We also analyzed another ICP, the intracellular K⁺ as a marker of TL activation. Indeed, naive or exhausted TLs are characterized by a high intracellular level of K⁺; this level decreases when TLs are activated (25). We then evaluated the impact of H89 in [K⁺]_i regulation in CD4⁺ cells. We observed that CD4⁺ TLs exhibited a low level of [K⁺]_i in presence of H89 (Figure 5C), attesting to their activation. Since some receptors such as PD-L1, another immunosuppressive actor, and CD80 (exhibiting activating and immunosuppressive functions according to the conditions) can be expressed on the cellular surface of cancer cells, we have tried to find out the influence of H89 on the expression of these receptors on the surface of CT26 cells. As depicted in Figure 5, H89 significantly affected both PD-L1 (Figure 5D) and CD80 (Figure 5E) expression.

To further decipher the molecular mechanisms related to the antitumor effect of H89, CT26 tumors on day 14 postinjection were harvested for transcriptomic analysis. Analysis of the transcriptome profiling of CT26 tumors from H89-treated and untreated mice led to the selection of 69 genes significantly differentially expressed based on an absolute log fold-change greater than 1 and an adjusted p-value below 0.05 (Supplementary Figure S3). Among them, 7 genes were downregulated, and 62 were overexpressed in treated mice (Supplementary Figure S3). Several genes significantly upregulated with H89 are related to immune function, such as IL-15RA, IL-18R1, STAT2, CCR9, and P2ry14, a G protein-coupled receptor. While some cancer-associated genes were downregulated, such as potassium channel 6.1 (Kir6.1), the atypical cadherin, FAT2, and keratin 18 (Krt18) (Figure 6A). Furthermore, GSEA analysis (FDR < 0.05) revealed distinct hallmark pathways between H89-treated (D14 of treatment) and untreated tumors, in particular, H89-mediated amplification of hallmarks of IFN-α and IFN-γ response (an antitumor phenotype) compared with untreated tumors with elevated protumoral hallmarks such as MYC pathway and oxidative phosphorylation (Supplementary Figure S3B).

Based on these transcriptomic data, we particularly focused our attention on the H89-mediated modulation of the IL-15/IL-15RA axis. Interestingly, the IL-15/IL-15RA signaling pathway is important in the activation of immune cells, including CD8⁺ TLs, and is associated with a favorable prognosis in CRC (10, 28). We first tested whether H89 also induced IL-15 in addition to IL-15RA. We found that H89 increases both mRNA and protein levels of IL-15 in tumors (Figure 6B). IL-15 mRNA levels were also increased in vitro in CT26 treated with H89 for 24 h (Figure 6C). Furthermore, we investigated the involvement of the IL-15/IL-15RA pathway in H89-mediated tumor growth inhibition in CT26 tumor-bearing mice. We found that blocking IL-15 signaling with an anti-TMB1 antibody abrogated the antitumor effect of H89 (Figure 6D). Since the Akt signaling pathway is mainly under the control of IL-15 (29), this prompted us to assess the phosphorylation status of Akt. We found that H89 significantly increases the phosphorylation of Akt (at Ser 473) in CD8⁺ TLs (Figure 6E), MOLT-4 cells (CD8⁺-like phenotype) (Figure 6F), and Jurkat cells (CD4⁺-like phenotype) (Figure 6G). Since Akt is a target of the serine/threonine protein phosphatase 2A (PP2A) (30), we investigated whether H89-mediated phosphorylation of Akt could be associated with the inhibition of PP2A activity. We showed that H89 significantly reduces PP2A activity in Jurkat (Figure 6H) and MOLT-4 cells (Figure 6I).