HDAC class I inhibitor domatinostat sensitizes pancreatic cancer to chemotherapy by targeting cancer stem cell compartment via FOXM1 modulation

The Human pancreatic cancer cell lines PANC1, ASPC1 and the hTERT-immortalized foreskin fibroblast BJhTERT were purchased from the American Type Culture Collection (ATCC, Rockville, MD, USA). PANC28 cell line was obtained from the laboratory of Dr. Marsha L. Fraizer and Dr. Douglas B. Evans. In adherent condition all cell lines were maintained as monolayer cultures and cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing 4.5 g/L glucose, glutamine, and non-essential amino acids and supplemented with 10% heat-inactivated fetal bovine serum and penicillin (100 IU/mL)–streptomycin (100 μg/mL). In spheroid-forming condition all cell lines were plated (4000 cells/ml) in low attachment plates and cultured 48 h in sphere medium (DMEM / F12 supplemented with BSA 0.1%, glucose 0.5%, heparin 4 μg/ml, L-glutamine 2.5 mM, PS 1X, FGF 20 ng/ml, EGF 20 ng/ml, B27 1X, insulin 20 μg/ml) to perform the assays shown. Cultures were maintained in a humidified atmosphere of 95% air and 5% CO2 at 37 °C. All cell lines were regularly inspected for mycoplasma. The cells have been authenticated with short tandem repeat profile generated by LGC Standards.

All media, sera, antibiotics, and glutamine for cell culture were from Lonza (Basel, Switzerland). Primary antibodies for western blotting were used according to the manufacturer’s protocol: β-Actin (#8227), poly-(ADPribose)-Polymerase (PARP)-Ab (#556494), phospho-Histone H2AX (γH2AX) (#05636), FOXM1 (#5436S), β-catenin (#8480S), Oct-4 (#2750S), were purchased from Cell signaling Technology (Danvers, MA, USA). γ-Tubulin (#sc-7396) were purchased from Santa Cruz Biotechnology (Dallas, TX, USA). Secondary antibodies were purchased as follows: polyclonal goat anti-rabbit IgG (H + L)-HRP conjugate (#1706515) and polyclonal goat anti-mouse IgG (H + L)-HRP conjugate (#1706516) were purchased from Abcam (Cambridge, UK); polyclonal rabbit anti-goat IgG-HRP conjugate (#sc-2768) were purchased from Santa Cruz Biotechnology (Dallas, TX, USA). Goat polyclonal Secondary Antibody to Mouse IgG - H&L - Alexa Fluor® 594 (#ab150120). Stem cell viability was evaluated by 3D Cell Viability Assay (Promega, Madison, WI, USA) according to the manufacturer’s protocol.

Domatinostat (4SC-202) was obtained from 4SC AG (Planegg-Martinsried, Germany) and dissolved in sterile DMSO for in vitro experiments and in methylcellulose for in vivo experiments; gemcitabine (Accord, Devon, UK) and nab-paclitaxel (Celgene, Milan, Italy) were provided by our pharmacy. Taxol (#PHL89806), 5′-deoxy-5-fluoro-uridine, (5’DFUR) (#F8791), SN-38 (#sc-203,697), oxaliplatin (#O9512) were purchased from Sigma- Aldrich (St. Louis,MO, USA). Bortezomib was obtained from cell signal technology (Danvers, Massachusetts, USA). Stock solutions were diluted to appropriate concentrations in culture medium before addition to the cells.

Cell proliferation was measured in 96-well plates in cells untreated and treated with described drugs as single agent or in combination. Cell proliferation was measured using a spectrophotometric dye incorporation assay (Sulforhodamine B) [18].

Drugs combination studies were based on concentration-effect curves generated as a plot of the fraction of unaffected (surviving) cells versus drug concentration after 96 h of treatment. Synergism, additivity, and antagonism were quantified after an evaluation of the CI, which was calculated by the Chou-Talalay equation with CalcuSyn software (Biosoft,Cambridge, UK), as described elsewhere [19]. A CI < 0.9, CI = 0.9–1.2, and CI > 1.2 indicated a synergistic, additive or antagonistic effect, respectively. The DRI determines the magnitude of dose reduction allowed for each drug when given in combination, compared with the concentration of a single agent that is needed to achieve the same effect.

Single cell suspensions were plated, as previously described [18] and treated or untreated with ≅ IC10 at 96 h concentrations of domatinostat (0.1 μM), gemcitabine (PANC1, 1.5 nM; PANC28, 5 nM; ASPC1, 1.5 nM) and Taxol (PANC1, 0.75 nM; PANC28, 0.15 nM; ASPC1, 0.325 nM). After 10 days, colonies were visualized and count as described previously [18].

Spheroid cultures, treated as indicated, were dissociated and live cells were sorted with a BD FACS Aria with a limiting dilution approach at 1, 2, 4, 8, 16, 32, 64 cells per well in ultra-low attached 96-well plates (Corning, NY, USA) in sphere medium. Stem cell frequency was evaluated after three weeks with the Extreme Limiting Dilution Analysis as described by Hu et al. [20].

Cell cycle analysis was performed at the indicated times in all cell lines treated with domatinostat and GT, alone or in combination, as previously reported [21].

Western blots were performed according to standard procedures [22]. Images were acquired using the Image Quant LAS 500 and the intensity was measured by Image Quant TL image software (GE Healthcare, Illinois, USA).

RNA was isolated by Trizol reagent (Invitrogen, CA, USA) as previously described [22]. Real-Time PCR by ABI Prism 7900 HT Sequence Detection System (Applied Biosystems, CA, USA) was performed using specific Taqman probes. All genes relative mRNA expression levels were calculated using the 2 ^-∆∆CT method and were normalized to that of β-actin as endogenous control gene β-actin.

6000 cell/well, plated on 96-wells, were treated with drugs as indicated in figure legends. Then cells were fixed in 4% paraformaldehyde (20 min at RT), blocked by 0.2% PBS/BSA solution (5 min at RT) and incubated with primary anti-FOXM1 antibody for 1 h at 37 °C. After washes, cells were incubated with anti-rabbit Alexa Fluor 595 (Thermo Fisher Scientific, Waltham, USA) overnight at 4 °C. Then the cells were washed and incubated 15′ with 4′,6-diamidin-2-fenilindolo (DAPI) (Thermo Fisher Scientific, Waltham, USA). Representative images were taken at 40X magnification by Opera Phenix microscope (PerkinHelmer,Waltham, MA USA) and the positive cells are counted by Harmony software (PerkinHelmer,Waltham, MA, USA).

ROS production was evaluated by culturing PANC1 and ASPC1 spheroids in Hydroethidine (Thermo Fisher Scientific, Waltham, MA, USA), according to the manufacturer’s protocol. Stained samples were evaluated by flow cytometry (FACS Canto, BD, Franklin Lakes, NJ, USA) and results analyzed with BD FlowJo software. The production of superoxide by mitochondria was evaluated by the fluorescent MitoSOX™ Red reagent (Thermo Fisher Scientific, Waltham, MA; USA), according to the manufacturer’s protocol and by microscopy. Briefly, PANC1 and ASPC1 spheroids were dissociated and plated on cover slips. Subsequently, the cells were treated as reported in figure legend with domatinostat (0.5 μM). Then, the media was removed and pre-warmed (37 °C) staining solution containing MitoSox probe was added for 15 min at 37 °C. The staining was photographed by Opera Phenix microscope (PerkinHelmer,Waltham, MA USA) and the positive cells are counted by Harmony software (PerkinHelmer,Waltham, MA, USA).

PANC-1 cells (12X10⁶) were crosslinked at room temperature for 10 min by adding formaldehyde to a final concentration of 1%. The action of formaldehyde was then neutralized by adding freshly dissolved Glycine to a final concentration of 125 mM. Cells were then centrifuged at 1200 rpm for 5 min and the pellet was washed twice in 1X cold PBS. Subsequently, cells were suspended in lysis buffer B (20 mM HEPES pH 7.7, 10 mM EDTA, o,5 mM EGTA, 0,25% triton X100) and C (50 mM HEPES pH 7.6, 150 mM NaCl, 1 mM EDTA, 0,5 mM EGTA) and alternately placed on a wheel at 4°C for 10 min. Another centrifugation was performed to collect the nuclei in buffer D (20 mM HEPES ph 7.6, 1 mM EDTA, 0,5 mM EGTA, 0,05% SDS and protease inhibitors). The nuclei were sonicated at maximum intensity by using Bioruptor Next Gen (Diagenode, Ougrée, Belgium). After sonication, samples were centrifuged at 13000 rpm for 10 min at 4°C and the supernatants were collected for ChIP assay analysis. The supernatants were transferred to a clean tube where the antibody was also added (about 3 μg per reaction). Immunoprecipitation was performed overnight at 4°C on a wheel and by adding to supernatant Protein A/G PLUS (Santa Cruz Biotechnology, Dallas, TX, USA, sc-2003), antibody, incubation buffer 1X (10 mM Tris pH 8, 150 mM NaCl, 1 mM EDTA, o,5 mM EGTA, o,15% SDS, 1% Triton X-100, protease inhibitors and 0,1% BSA). In parallel 10% of each sample was taken as an input indicator for further PCR analysis. The following day, samples were centrifugated at 1200 rpm for 5 min at 4°C and the supernatant was removed with several washes of 10 min at 4°C, using 500 μL of each wash buffer. After the final wash, 400 μL of elution buffer were added. Elution was carried out for 30 min at room temperature on a wheel. Subsequently, 125 mM NaCl was added to 400 μL of the sample. De-crosslinking continued overnight at 65°C. The day after, proteins were degraded by treatment with proteinase K, performed by incubating with 0.5 M EDTA, 1 M Tris pH 6.5 and proteinase K for 1 h at 45°C. DNA was then recovered with MinElute Reaction Cleanup Kit (Qiagen, Hilden, Germany). Real-time PCR analysis was then performed on these samples. The antibody used for this assay was: Anti-FOXM1 (Millipore, Burlington, MA, Stati Uniti). The following gene promoters were used: SOX2 FW: 5′-AGGGAGAGAAGTTTGAGCCC-3′; SOX2 REV:5′-GCGAGGAAAATCAGGCGAAG-3′; Rad51 FW:5′-GTAAAACTTGGCCCCTACACTG-3′; Rad51 REV:5′-ATAAGGTGCATCTCTCTCCCC-3′; OCT4 FW: 5’TGGAGGTGTGGGAGTGATTC-3′;OCT4 REV: 5-GACTACAGGCTTGGACCACT-3′; BIRC5 FW: 5′-TTTGCGAAGGGAAAGGAGGA-3′; BIRC5 REV: 5′-AATGAACAGGGGAGGGATGG-3′;CAT FW: 5′-TGGTCTACTTTGCAAGCTTGG-3′; CAT REV:5- AAGGTAATTGCAAGTGATTGGTT-3′; XRCC1 FW: 5′-GCGGGCGTAGTAAAAGACAG-3′; XRCC1 REV: 5′-TGAGGCCAAAAGAGAAGGGT-3′.

PANC1 and ASPC1 spheroid cultures were treated as reported in figure legends with domatinostat and/or NAC (Sigma-Aldrich, St. Louis, MO, USA) or Mitoquinone mesylate (MitoQ) (Selleck Chemicals LLC, Houston, TX, USA). Cells were then dissociated and stained with CD133-APC antibody (1:100, MiltenyiBiotec, Paris, France) in PBS for 20 min at 4°C, washed with PBS and re-suspended in PBS-Annexin V-FITC from BD (BD, Franklin Lakes, NJ, USA) for 15 min at 4°C for evaluation of apoptotic cells.

A network analysis was generated by (ingenuity pathway analysis) ipa software (GeneGo Inc., St. Joseph, MI, USA). IPA includes a manually annotated database of protein interactions and metabolic reactions obtained from the scientific literature. The networks were graphically visualized as hubs (proteins) and edges (the relationship between proteins).

The pCW57.1-FOXM1c plasmids were purchased from Addgene (Watertown, MA, USA). Adherent PANC1 cells were transfected using Lipofectamine 2000 Reagents (Thermo Fisher Scientific, Waltham, MA, USA), according to the manufacturer’s recommendation. After 4 h from transfection, cultures were used for western blot, real-time PCR, immunofluorescent experiments, cell survival assay as described above.

Genomics analysis were performed taking advantage of web servers for analyzing RNA sequencing expression data from the TCGA http://​gepia.​cancer-pku.​cn/​ and http://​r2.​amc.​nl.

All studies have been performed in accordance with the institutional guidelines and approval by local authorities (377/2019), in line with “Directive 2010/63/EU on the protection of Animals used for scientific purposes” and made effective in Italy by the Legislative Decree DLGS 26/2014.

PANC28 and PANC1 cells were respectively suspended in 5*10⁶ cells in 200 μl of PBS and Matrigel (BD Pharmingen, Milan, Italy) (1:1) and 5*10⁶ cells in 200 μl of PBS. The cells were subcutaneously injected in the flanks of 6-week-old female nude mice (Envigo Laboratories, Indianapolis, IN, USA). The mice were acclimatized in the Animal Care Facility of CROM–Centro Ricerche Oncologiche di Mercogliano. Tumor volume [1/2(length × width²)] was assessed using digital caliper. When the tumors became palpable, the mice were randomized into four experimental groups (n = 7). Mice were treated as followed: (a) vehicles; (b) gemcitabine (weekly 25 mg/Kg, i.p.) and nab-paclitaxel (weekly 20 mg/Kg, i.p.) re-suspended in salt solution 100 μl per dose; (c) domatinostat (20 mg/Kg 5 days/week, per os) re-suspended in Methocel 2% solution 250 μl per dose; (d) triple combination gemcitabine/nab-paclitaxel plus domatinostsat. Drug treatments were administered for 3 weeks. All mice received drugs vehicles. TGD and the percent change in the experimental groups was compared with that of the vehicle control groups as described before [23].

Both expression and localization of β-catenin were evaluated by IHC on formalin fixed paraffin embedded tumor samples derived from mice sacrificed at the end of PANC1 in vivo experiment. Briefly, the sections were incubated with primary antibody and then with biotin-conjugated secondary antibody, before incubation with specific streptavidin HRP-conjugated tertiary antibody (Thermo Fisher Scientific, Waltham, MA USA). Peroxidase reactivity was visualized using a 3,3′-diaminobenzidine (Abcam, Cambridge, UK). A single pathologist (R D.C.) performed a blinded analysis of the slides.

In order to explore the potential of domatinostat as an effective therapeutic approach to sensitize PDAC to chemotherapy, we performed an in vitro screening of drug combinations using chemotherapeutic agents currently employed for PDAC treatment, such as fluoropyrimidines (evaluating 5’DFUR, an intermediary prodrug of 5-fluorouracil), irinotecan (evaluating the active metabolite SN-38), oxaliplatin, gemcitabine and taxol. We tested three different pancreatic cancer cell lines (PANC1, PANC28 and ASPC1), showing a high similarity in domatinostat sensitivity, compared to striking differences in sensitivity to chemotherapeutics, tested in monotherapy (Suppl. Table S1 and Suppl. Fig. S1). We explored different cytotoxic ratios of domatinostat in combination with chemotherapeutic agents, either at equipotent doses (50:50 ratio) or using lower doses of chemotherapeutics (75:25 ratio), either simultaneously or sequentially (with a 24 h delay between the two agents). The combination index (CI) values, calculated at 50% (CI₅₀) of cell lethality demonstrated mostly synergistic (CI < 0.9) or additive (CI < 1.1) effects with all the anticancer agents tested, in all three cell lines (Fig. 1A-B and Suppl. Tables S2–5). Concomitant or sequential treatments were equally effective, although, CI values differ depending on the chemotherapeutic employed. Interestingly the synergistic interaction was also observed using lower doses of chemotherapeutics (75:25 ratio) (Suppl. Tables S3 and S5). Indeed, the evaluation of the dose reduction index (DRI) values, which represent the order of magnitude (fold) of dose reduction, obtained for the IC₅₀ (DRI₅₀) in combination treatment compared with single-drug treatment, confirmed, for all the chemotherapeutics evaluated, a significant potentiation of the antitumor effect when combined with domatinostat, in all three cell lines, with both simultaneous and sequential schedules (Fig. 1C-D, Suppl. Tables S2–5).

Next we explored combination treatment of domatinostat plus chemotherapy doublets, observing again significant synergistic anti-proliferative effects, either with gemcitabine/taxol (GT) (Fig. 1E-G and Suppl. Tables S2–3) or with fluoropyrimidine/irinotecan (5’DFUR/SN38) combinations (Suppl. Tables S4–5). Notably, in the hTERT-immortalized foreskin fibroblast BJhTERT cells, a non-tumorigenic cell line, we observed antagonist effects in all drug combinations tested (Suppl. Tables S2–5), suggesting a selective synergistic effect of domatinostat plus chemotherapy in tumor cells.

Since GT is the most common first-line option for the treatment of metastatic PDAC patients we further investigated the mechanism underlying the observed synergism by using domatinostat plus this chemotherapy doublet. Notably, for all the following experiments, if not differently mentioned, we tested domatinostat at 0.5 μM, a low dose if compared with reported preclinical studies with this agent [17, 24‐27].

We first confirmed the synergistic antitumor interaction by demonstrating in all three PDAC cell lines a clear statistically significant potentiation of apoptosis in combination treatment as compared to domatinostat or GT alone (tested at IC₅₀^96h), as shown by Annexin-V staining (Fig. 2A and Suppl. Fig. S2) and PARP-cleavage (Fig. 2B). These effects were not observed in BJhTERT cells. Accordingly, the synergistic pro-apoptotic effect was paralleled by a cell cycle perturbation effect characterized by a S-phase block induced by triple combination after 48 h of treatment in PDAC cells lines but not in normal BJhTERT cells (Suppl. Fig. S3).

Taking advantage of colony formation assay, we then evaluated domatinostat plus GT (IC₁₀^96h), either simultaneously or sequentially (with a 24 h delay between the two agents) on the three PDAC cell lines. As shown in Fig. 2C-E, we demonstrated a dramatic reduction of colony formation by triple domatinostat/GT combination in both treatment schedules, as compared to control, domatinostat or GT alone.

Finally, to determine the effect of domatinostat plus GT (IC₁₀^96h), on CSC compartment, we performed a limiting dilution assay. Notably, we demonstrated a statistically significant reduction of clonogenicity of PANC1, PANC28 and ASPC1 cells upon incubation for 24 h with triple combination vs domatinostat or GT alone (Fig. 2F).

Overall these data demonstrated the ability of domatinostat to sensitize pancreatic cancer cells to different chemotherapeutic agents employed in clinical practice to treat PDAC patients.

In order to better define the potential impact on CSC compartment we next explored the effect of domatinostat alone, or in combination with GT, on PDAC cell-derived self-assembled spheroids, a model characterized by a clear CSC-enrichment (Suppl. Fig. S4).

First, we observed that domatinostat alone, at low doses (0.5 μM and 1 μM), strongly reduced size, amount and viability of PANC1, PANC28 and ASPC1 spheroids (Fig. 3A-B, Suppl. Fig. S5A-B, Suppl. Fig. S6). Notably, we also demonstrated a clear reduction of CSC marker CD133 surface expression (Fig. 3C and Suppl. Fig. S5C) as well as of CD133 mRNA levels (Suppl. Fig. S5D) within 16 h of domatinostat treatment in all PDAC spheroids. Similarly, a reduction of CSC marker Oct-4 mRNA level was also demonstrated in all three spheroid models after 24 h of treatment with domatinostat (Fig. 3D and Suppl. Fig. S5E).

Furthermore, since it has been reported that oxidative stress modulation has a critical role in CSC by promoting proliferation, adaptation and resistance to chemotherapy [9], we also investigated the effects of domatinostat on CSCs redox homeostasis. In both PANC1 and ASPC1 spheroids we demonstrated a time-dependent increase in total ROS cellular amount induced by domatinostat within 4 h of treatment, reverted by concomitant treatment with the ROS scavenger N-acetylcysteine (NAC) (Fig. 3E). This effect was confirmed by concomitat ROS accumulation in mitochondrial compartment, the main source of ROS in living cells, as demonstrated by mitosox IF staining (Fig. 3F). Domatinostat-mediated ROS accumulation was paralleled within the same time frame by a pro-apoptotic effect in CSC subpopulation, as shown by Annexin V staining of CD133 positive cells. Notably, this effect was partially reverted by NAC, indicating that the induction of apoptosis was at least in part due to ROS accumulation (Fig. 3G). Intriguingly, apoptotic effect induced by domatinostat was almost completely reverted by concomitant treatment with the mitochondria ROS scavenger mitoquinone mesylate (MitoQ) (Fig. 3H). Overall, by comparing the efficacy of MitoQ vs NAC in reverting apoptosis, we can speculate that CSC are more sensitive to mitochondria ROS levels alteration comparing to cytoplasmic ROS levels.

Finally, we tested domatinostat in combination with GT in PANC1, PANC28 and ASPC1 spheroid models as compared to the differentiated counterpart cells grown in adherent conditions. Notably, spheroid models are significantly more resistant to GT treatment (G:12.5 nM; T: 1 nM; 72 h), however in combination with domatinostat (1 μM) we observed a similar significant decrease in viability in both spheroids and differentiated cells (Fig. 4). Interestingly, in two out of three cell lines (PANC28 and ASPC1) domatinostat alone was even more effective on spheroids compared to adherent growing cells.

Overall, these data demonstrated that domatinostat has an efficient antitumor effect on PDAC spheroids and thus on CSC subpopulation, related, at least in part, to the alteration of CSC mitochondrial and cellular oxidative homeostasis. Moreover, we might argue that the selective effect of domatinostat on CSC could be responsible for the clear potentiation of GT antitumor effect observed in combination treatment.

To better understand the molecular mechanism by which domatinostat induces a ROS-mediated cell-death in CSC, we performed RNA-seq data-mining from published results of domatinostat effects on PANC1 cells [17]. In details, we analyzed the expression of oxidative-stress related genes whose high expression we have previously demonstrated to be statistically significant associated with poor prognosis in solid tumors, including pancreatic cancer [9]. Among these genes, FOXM1 emerged as the top strongly down-regulated gene by domatinostat treatment (Suppl. Table. S6‐7). FOXM1, is a transcription factor with several functions, that has been reported to play a critical role in pancreatic cancer [9]. Indeed, by analyzing TCGA expression data in pancreatic cancer (PAAD dataset) we evidenced that FOXM1 high level is significantly related with bad overall survival (Fig. 5A), disease free survival (Fig. 5B) and with chemotherapy response (Fig. 5C) in PDAC patients.

We also found a higher FOXM1 protein expression in PANC1, PANC28 and ASPC1 spheroid models compared to differentiated cells, correlated with high β-Catenin and Oct-4 protein levels, thus suggesting and enrichment of FOXM1 expression in CSC subpopulation (Fig. 5D). Indeed FOXM1 governs the recruitment of β-catenin to the β-catenin-TCF4 transcription activation complex in the Wnt target gene promoter thus being involved in CSC phenotype [28]. We confirmed that domatinostat treatment leads to significant decrease of FOXM1 mRNA expression in all three PDAC cell lines (Suppl. Fig. S7A).

We next investigated FOXM1 localization upon domatinostat treatment from 2 up to 16 h, in PANC1 spheroids, demonstrating, by IF staining (Fig. 5E), a time-dependent decrease in nuclear FOXM1 localization with a peak between 6 and 8 h of treatment, reported also as decrease of nuclear-FOXM1 fraction and nuclear-FOXM1 spots (Fig. 5F and Fig. 5G). Furthermore WB analysis confirmed FOXM1-nuclear reduction, paralleled by protein cytoplasmic accumulation (Fig. 5H), thus suggesting that domatinostat affects FOXM1 activity also preventing its nuclear translocation. Notably, we observed a simultaneous similar β-catenin cytoplasmic localization following domatinostat treatment (Fig. 5H), confirming the tight functional connection between FOXM1 and β-catenin.

Moreover, since it has been previously demonstrated that FOXM1 protein levels is also regulated by ubiquitination and deubiquitination process and, thus, proteasome-dependent degradation [29], we evaluated domatinostat effect on FOXM1 protein expression in absence or presence of the proteasome inhibitor, bortezomib. As shown in Fig. 5I, concomitant treatment with bortezomib, completely reverted domatinostat-mediated inhibition of FOXM1 protein, suggesting that the inhibition of FOXM1 observed in PANC1 is partially due to domatinostat-increased FOXM1 protein degradation within 6 h from treatment (Fig. 5I).

Notably, domatinostat-mediated FOXM1 decrease was paralleled by the reduction of FOXM1 transcriptional target genes such as the oxidative stress-response antioxidants CAT, GPX2 and SOD2, the DNA damage-related RAD51 and XRCC1 genes and the stemness-related genes BIRC5 and SOX2 (Fig. 5L). Then, we performed a ChIP assay to investigate if domatinostat was able to reduce the FOXM1-binding to promoters of recovery stress (CAT), stemness (OCT4, BIRC5 and SOX2) and DNA damage (RAD51 and XRCC1) genes that we found as down-regulated at transcriptional level by domatinostat. In line with downregulation at transcriptional level, ChIP experiments using PANC1 spheroids followed by semiquantitative and quantitative PCR (Fig. 5M and Suppl. Fig. S7B) revealed the presence of FOXM1 on CAT, OCT4, BIRC5, SOX2, RAD51 and XRCC1 promoters in untreated conditions and its displacement after 16 h of treatment with domatinostat (1μM). Moreover, FOXM1 as well as β-catenin protein levels were reduced by domatinostat alone or in combination with GT (IC₅₀^96h) in PANC1 spheroids (Fig. 5N) and PANC28 and ASCP1 spheroids (Suppl. Fig. S8). This effect was paralleled by synergistic induction of DNA damage in triple combination as compared with domatinostat or GT alone, as demonstrated by increased expression of γ-H2AX (Fig. 5N and Suppl. Fig. S8).

To further confim the molecular mechanism behind the ability of domatinonstat to modulate stemness and oxidative stress homeostasis we performed an ingenuity pathway analysis (IPA) search on “BIRC5, NANOG, POUF1, CTNNB1 and SOX2”, as domatinostat modulated-stemness markers, and “GPX2, CAT, SOD2, RAD51 and XRCC1”, as domatinostat modulated-stress markers. IPA network revealed direct relationships between all the protein used as input, confirming a functional relationship between the targets of our treatment combination. Moreover, FOXM1 came out in the IPA upstream analysis as the most significant upstream regulator (Suppl. Fig. S9). Furthermore, an analysis on TCGA-PAAD data demonstrated a strong positive expression pattern correlation between FOXM1 and either DNA damage-related genes, such as EXO1, RAD51, XRCC2 and a stemness related gene, such as BIRC5 and the FOXM1-specific deubiquitinase, USP5 (Pearson’s R < 0.65) in pancreatic tumor tissues (Suppl. Table S6 and S7 and Suppl. Fig. S10), overall confirming and reinforcing our observations.

To confirm that the antitumor effect of domatinostat alone and in combination with GT is mechanistically connected with the modulation of FOXM1 and oxidative stress in CSC subpopulation, we then generated transiently FOXM1 over-expressing PANC1 cells (OE-FOXM1) (Suppl. Fig. S11). Notably, OE-FOXM1 cells showed higher mRNA levels of CSC marker Oct-4 and of oxidative stress-response FOXM1 transcriptional targets CAT, SOD2 and GPX2, compared to empty vector-transfected cells (EV-FOXM1) (Fig. 5O). Moreover, the cellular (Suppl. Fig. 12A) and mitochondrial (Fig. 5P) ROS accumulation induced by domatinostat was significantly reduced or completely abolished, respectively, in OE-FOXM1 compared to EV-FOXM1 cells. Notably, basal mitochondrial ROS levels in OE-FOXM1 cells were dramatically reduced, confirming a critical antioxidant role of FOXM1, particularly in the mitochondrial compartment.

Finally, coherently with the data presented above, OE-FOXM1 cells were less sensitive to either domatinostat or GT compared to EV-FOXM1 cells, and, more importantly, the synergistic antitumor effect of the triple combination was abolished in OE-FOXM1 cells (Fig. 5Q, Suppl. Fig. S12B).

Overall, FOXM1, a transcription factor correlated with PDAC patients’ bad prognosis, appears to play a critical role in the redox homeostasis of PDAC cells particularly in the CSC compartment. Indeed, FOXM1 down-modulation by domatinostat induced ROS accumulation targeting CSC subpopulation, thus leading to antitumor effect and sensitization to chemotherapy.

In order to confirm in vivo the synergistic antitumor effect observed in vitro we evaluated the activity of domatinostat in combination with chemotherapy in in PANC1 and PANC28 cell line xenograft mouse models. In detail, we evaluated combination treatment of domatinostat plus gemcitabine/nab-paclitaxel (GemNP) doublet, demonstrating in both PDAC models statistical significant decrease of tumor volume compared with control or single treatments (the treatment schedule is reported in Suppl. Fig. S13A) (Fig. 6A-B). Furthermore, by calculating the percent change in tumor volume from the time of initial treatment (day 0) to the end of the study (day 32 in PANC28 xenograft model and day 25 in PANC1 xenograft model), we confirmed that domatinostat plus GemNP combination significantly reduced the tumor burden in both models with a synergistic antitumor effect, compared to domatinostat or GemNP alone, particularly evident in PANC28 xenograft model (Fig. 6C-D).

The synergistic antitumor interaction was also tested by evaluating the tumor growth delay (TGD) induced by domatinostat plus GemNP, that reached a peak of more than 100%, indicating that the mean rate of tumor growth in the control was more than 3-fold higher than in the combination in both models (Fig. 6E-F). Notably, the maintenance of body weights (Suppl. Fig. S13 B-C) and the absence of other acute or delayed toxicity signs indicated a well tolerability of triple drugs combination.

We also validated in vivo the mechanistic findings evidenced in vitro on both FOXM1 modulation and CSCs targeting. In details, we found that the increased FOXM1 protein tumor expression observed in GemNP-treated mice was completely abolished by the concomitant treatment with domatinostat (Fig. 6G). Moreover, in line with in vitro data, FOXM1 down-modulation was paralleled also in vivo by a strong increase of γ-H2AX protein levels in both domatinostat and triple combination treatment group, indicating the induction of DNA damage (Fig. 6G).

Furthermore, we also found differential expression of β-Catenin in different treatment groups, with a prevalent membrane localization in domatinostat and triple combination, confirming in vivo the inhibition of Wnt-pathway by domatinostat (Fig. 6H).

Finally, although not statistically significant, we showed, also in vivo, a tendency of both FOXM1 and of CSC marker Oct-4 mRNA expression downregulation in xenograft tumors, induced by domatinostat alone or in combination with GemNP (Suppl. Fig. S14A and B).

All together, these data confirmed the potential of a treatment strategy based on the addition of domatinostat to standard chemotherapy regimen to improve pancreatic cancer patient’s outcome and to bypass chemoresistance mechanisms. Moreover, these results further confirm our hypothesis that domatinostat potentiates chemotherapy in PDAC by targeting CSCs compartment via FOXM1 modulation (Fig. 7).