Abemaciclib Is Effective Against Pancreatic Cancer Cells and Synergizes with HuR and YAP1 Inhibition Free

This article is featured in Highlights of This Issue, p. 1959

Pancreatic ductal adenocarcinoma (PDAC) is one of the most aggressive solid tumors. (1, 2) Currently, the only curative option is surgery, but most patients present with inoperable disease. (3). In the metastatic setting, the standard of care is a combination of nontargeted chemotherapies, FOLFIRINOX (folinic acid, 5-fluorouracil, irinotecan, oxaliplatin) or gemcitabine plus nab-paclitaxel (4, 5). While these regimens are more effective than the previous standard of care of gemcitabine alone, they only increase median survival by a few months (6). These statistics highlight the need for new and targeted therapies for the treatment of PDAC.

CDKN2A encodes the cell-cycle inhibitor proteins p16(INK4A) and the p14(ARF; ref. 7). Loss of CDKN2A due to the gene mutation or promoter hypermethylation can be found in 90%–98% of PDAC cells (8). CDKN2A is an endogenous inhibitor of cell-cycle progression which, when activated, leads to a G₁ cell-cycle arrest through the inhibition of CDK4/6 (9). In dividing cells, the presence of growth factors or other progrowth signals leads to an increase in the transcription of cyclin D1 (10). Cyclin D1 can then bind and activate CDK4/6 to monophosphorylate and partially inactivate Rb early in the G₁ phase of the cell cycle (11). The Rb–E2F complex is an important transcriptional repressor of DNA synthesis, and its inactivation is necessary for the progression to S-phase. Inactivation of Rb leads to the release of the E2F 1–3 transcription factors (7, 12), and upon release, there is a transcriptional upregulation of target cell-cycle genes such as cyclin E and cyclin A (13), allowing for progression of the cell cycle into S-phase (14). As described above, in PDAC, loss of p16 leads to unchecked cellular division and disruption of the G₁–S checkpoints (15). When p16 is silenced, Rb remains nonfunctional as a regulator allowing for unchecked proliferation.

While the p16 pathway is frequently disrupted in PDAC, Rb is commonly preserved in PDAC and the presence of pRb is an important predictor of response for CDK4/6 inhibitors (16, 17). As a strong premise for this study, others have demonstrated and advocated for the use of CDK4/6 inhibitors for the treatment of PDAC (2, 18, 19). To underscore this work, there is currently a phase Ib trial studying the effects of palbociclib with nab-paclitaxel in patients with metastatic pancreatic cancer (NCT02501902). In addition, the CDK4/6 inhibitor abemaciclib (LY2835219) is currently being investigated in a phase II trial, as a solo agent, and in combination with a PI3K inhibitor in patients with metastatic pancreatic cancer (NCT02981342). In regards to this study, abemaciclib has been described as a more selective inhibitor of CDK4/6 (20, 21), is FDA-approved for hormone receptor–positive breast cancer (14, 22), and is currently being tested for efficacy in lung cancer (22), esophageal cancer (23), and some other soft-tissue cancers (24), yet has limited preclinical evaluation in PDAC. Herein, we evaluate the efficacy of abemaciclib in preclinical PDAC models in vitro and in vivo and explore its mechanism of action, and screened for potential synergistic combinations.

Human pancreatic cancer cell lines were purchased from the ATCC, confirmed negative for Mycoplasma contamination and validated using short tandem repeat profiling. All cells were cultured in DMEM supplemented with 10% FBS, antibiotics and L-glutamine. Patient-derived cell lines (PDX) were obtained from Dr. Talia Golan (The Chaim Sheba Medical Center at Tel HaShomer, Tel Aviv, Israel; ref. 25). PDX cell lines were cultured in RPMI supplemented with 10% FBS, antibiotics and L-glutamine. Mia PaCa2 “chronic” abemaciclib therapy lines were created by treating cell lines with increasing concentration of abemaciclib over 10 months and resistance testing was performed monthly using Pico Green assay to calculate IC₅₀/GR₅₀. All cell lines used in experiments were under passage 20, except for the resistant lines. 3D mouse organoids were generated from previously reported mouse models of PDAC, KPC, and PKP (26).

Most compounds used in this study were purchased from MedChem Express and Sigma-Aldrich. CMLD-2 was purchased from Millipore Sigma (catalog# 538339), and CA3 was purchased from Selleckchem (catalog# S8661). Abemaciclib, gemcitabine, and palbociclib were diluted in water to a 10 mmol/L stock concentration and stored at −20°C until use. Paclitaxel, CMLD2, pyrvinium pamoate, verteporfin, and CA3 were diluted in DMSO to a 10 mmol/L stock concentration and stored at −20°C/−80°C until use.

Assessment of cellular viability and calculation of IC₅₀ were performed using a Pico Green assay. Cells were seeded at 1,000 cells/well (Mia PaCa2) or 2,000 cells/well (Panc-1, HS 766T) in a 96-well plate and treated with drug for 5 days. Upon collection Pico Green (Thermo Fisher Scientific, catalog #P7581) was used as per manufacturer's instructions. Fluorescence (504 nm/531 nm) was determined using Promega GloMax machine. For cell lines with different proliferation and doubling times, we calculated growth rate inhibition (GR) and GR₅₀ values using a published and validated GR calculator (http://www.grcalculator.org/grtutorial/Home.html; ref. 27).

For mouse organoid experiments, assessment of cellular viability was done using CellTiter Glo protocol (Promega, catalog# G7570). Cells were seeded at 1,500 cells/well in a 96-well plate and treated with drug for 5 days. At collection day, CellTiter Glo reagent was added directly to the wells, mixed for 10 minutes and incubated for 10 minutes. Luminescence was quantified using a Promega GloMax.

For colony-forming experiments, cells were seeded at 800 or 1,500 (Mia PaCa2), 2,000 (Panc-1), or 3,000 (HS 766T) cells/well respectively in a 6-well plate. Drug was replenished every 3 days for 10 days total. Colonies were fixed with 80% methanol and stained with 0.5% crystal violet/ 25% methanol solutions for 10 minutes each. Colonies were counted manually or using ImageJ software and graphed using Prism (version 7.03).

Cells were treated with abemaciclib for 72 hours and upon collection were incubated with 10 μmol/L BrdU (Sigma, #B5002) for 1 hour prior and then fixed in ice-cold 70% ethanol. Cells were washed and incubated in 2 mol/L HCl for 1 hour to denature DNA. After washing, HCl was neutralized with 0.1 mol/L sodium borate (pH 8.5) for 10 minutes. Cells were stained with anti-BrdU FITC-conjugated antibody (Thermo Fisher Scientific, #11-5071-42) for 1 hour in the dark, washed, and stained with propidium iodide (PI) for 10 minutes. Samples were run using BD Celesta machine and data were analyzed using FlowJo (version 10) software.

Cells were seeded at 3,000 cells/well in 24-well plates with glass coverslips inlay and treated with water or 0.5 μmol/L abemaciclib or H₂O₂ at 10 μmol/L, 100 μmol/L, or 500 μmol/L and collected at various days for detection of β-galactosidase (β-gal) using manufacturer's kit (Cell Signaling Technology, #9860). After treatment, cells were washed with PBS, fixed, permeabilized stained with β-gal, and mounted/stained with DAPI (Invitrogen/Thermo Fisher Scientific, #P36931) on a glass slide and imaged using Leica DM4B.

ROS assay was performed using DCFDA/H2DCFD Cellular ROS Detection Assay Kit (Abcam, catalog# ab113851). Cells were seeded at 3,000 cells/well and treated with water or 0.5 μmol/L abemaciclib or H₂O₂ at 10 μmol/L, 100 μmol/L, or 500 μmol/L. Upon collection samples were collected per kit protocol and fluorescence (485 nm/535 nm) was determined using a Promega GloMax.

Cells were seeded at 3,000 cells/well in 24-well plates. Upon collection the caspase 3/7 live stain was added (Thermo Fisher Scientific, #R37111) 1 drop per 0.5 mL of media was added and incubated for 1 hour. GFP and brightfield images were taken using the Evos FL cell imaging system.

Annexin V/PI staining was performed using a flow cytometry kit (Thermo Fisher Scientific, # V13242). Changes in complexity and morphology of cells treated with water or 0.5 μmol/L abemaciclib after three days of treatment was monitored by forward and side scatter (FSC-A, SSC-A) measurements using BD Celesta machine. All events were gated excluding debris with a minimum of 10,000 cells analyzed per sample. Data were analyzed using FlowJo software.

Western blot and RNA immunoprecipitation assays were performed as described previously (28). For IHC, phospho-Rb, total Rb, Ki67, and TUNEL stainings were performed on 10% paraffin-embedded tumor slides as per the manufacturer's protocol and imaged using Leica florescence microscope at 20× and 40× magnifications. IHC staining was performed by our pathology department, by W. Jiang and R. O'Neill, at Thomas Jefferson University (Philadelphia, PA).

For protein evaluation by Western blot analysis, the following primary antibodies were used: phospho-Rb (1:500, Cell Signaling Technology, catalog no. 9307), total Rb (1:500, Cell Signaling Technology, catalog no. 9309), HuR (1:1,000, Life Technology/Thermo Fisher Scientific, catalog no. 390600), YAP1 (1A12; 1:500, Cell Signaling Technology, catalog no. 12395), cyclin D1 (1:500, Santa Cruz Biotechnology, catalog no. 8396), p27 (1:1,000, Santa Cruz Biotechnology, catalog no. sc-1641), and α-tubulin (1:5,000, Invitrogen, catalog no. A11126).

For IHC analysis, phospho-Rb (Cell Signaling Technology, catalog no. 8516), total Rb (Cell Signaling Technology, catalog no. 9309), and Ki67 (Roche, catalog no. 790-4286), and TUNEL (Trevigen TACS 2 TdT Core Kit, catalog no. 4810-30-CK) staining antibodies were used. For TUNEL staining, only diaminobenzidine (DAB) staining was used.

Total genomic DNA was extracted using the DNeasy Blood & Tissue Kits (Qiagen, catalog no. 69506) and 100 ng DNA was used per RT-PCR reaction using the Platinum Taq DNA Polymerase kit (Thermo Fisher Scientific, catalog no. 10966026). Samples were run on 0.75% agarose gel and LB buffer and imaged with UV box and camera. PCR SYBR primers were created by IDT and shown in Supplementary Figs. S1F and S3D. DNA sequencing was performed by Genewiz using Sanger sequencing methodology.

Total RNA was extracted using the RNeasy Mini Kit (Qiagen, catalog no. 74106) and cDNA was made using 1,500 μg total RNA using Applied Biosystems High Capacity cDNA Reverse Transcriptase kit (Thermo Fisher Scientific, catalog no. 4368814) and quantitative PCR (qPCR) was performed using specific SYBR primers, as described above, and run on QuantStudio 3 (Thermo Fisher Scientific). Relative quantification was performed using the 2^−ΔΔC_t method and comparing to GAPDH or 18s.

For transfections, Mia PaCa2 cells were seeded at 5 × 10⁵ cells per 10-cm dish. The next day, transfection reagent (si negative control, siHuR or siYAP1) and Lipofectamine 2000 were added and transfected as described previously (29). Media were changed on cells the next day and 24 hours later the cells were split for either cell viability (Pico Green) experiments or Western blot experiments. Mia PaCa2 HuR CRISPR knockout cells were generated as described previously (28) and validated prior to use.

Athymic female nude mice were purchased from Envigo at 6 weeks old and 20–25 g in size. Tumor growth was initiated by subcutaneous injection of 2 × 10⁶ Mia PaCa2 cells in a 1:1 mixture of Matrigel and PBS on bilateral rear flanks of each subject animal. When mean tumor volumes reached approximately 100 mm³ in size, the animals were randomized by tumor size into three groups with 5 mice/group. For efficacy studies, abemaciclib was dosed orally at 75 mg/kg, 0.1 mL/dose, daily Monday–Friday and the vehicle-only control group was given oral 1% hydroxyethyl cellulose (HEC) vehicle according to the same schedules. Mice body weight and tumor volumes were recorded three times per week, and once tumors reached a volume of 1,500 mm³, the mice were euthanized. All animal studies were approved and performed in accordance with the Institutional Animal Care and Use Committee (IACUC) guidelines. Abemaciclib (MedChem Express, catalog no. HY-16297, lot no. 27360) was formulated daily in 1% HEC in 25 mmol/L phosphate buffer (PB; pH = 2).

Statistical analysis was performed using Prism (version 7.03) software. An independent two-tailed t test was utilized for comparison of mean percent change in tumor volume between placebo and treatment groups. A P < 0.05 is considered statistically significant. For the in vivo experiment, Grubb test was employed on all group values to assess for outliers and if any value was found to be positive for an outlier, it was excluded from statistical analysis. Grubb testing is available online along within GraphPad.

Abemaciclib has been demonstrated to be a more potent inhibitor of CDK4/6 compared with the CDK4/6 inhibitors palbociclib and ribociclib in vitro (20). Therefore, we sought to define the efficacy of this drug in several PDAC cell lines; Mia PaCa2, Panc-1 (30), and HS 766T cells (31), three PDAC cell lines with characteristic loss of CDKN2A (Supplementary Fig. S1A). Palbociclib was published to be effective in PDAC in the past (2); therefore, we compared abemaciclib to palbociclib as an established benchmark. We found that across these three PDAC cell lines, abemaciclib had IC₅₀ <1 μmol/L, and was comparable with palbociclib in all three cell lines (Fig. 1A). In long-term drug sensitivity assays, abemaciclib inhibited colony formation with IC₅₀s in the nanomolar range and was comparable with palbociclib at similar drug concentrations, as seen in Fig. 1B and Supplementary Fig. S1E. We also tested standard PDAC chemotherapeutics gemcitabine and oxaliplatin on PDAC cell lines and found abemaciclib was less effective at inhibiting cell survival as compared with gemcitabine, but more potent than oxaliplatin (Fig. 1C, HPNE comparison shown in Supplementary Fig. S1D).

To investigate the time dependency of abemaciclib effect on cell viability inhibition, we compared abemaciclib to no treatment over the course of 5 days. We found a statistically significant decrease in cellular viability of the abemaciclib-treated PDAC cell lines when compared with no treatment over days 3, 4, and 5 (Fig. 1D). This difference was most pronounced on day 5 in all three cell lines with a P < 0.0001. To validate these effects in more clinically relevant models, we determined the CDKN2A status of patient-derived xenografts (PDX, Supplementary Fig. S1A). The PDX lines were found to have a deletion of CDKN2A and abemaciclib also had similar effects on cell viability (Supplementary Fig. S1B). To confirm the pathway, we compared the efficacy of abemaciclib in our KPC/KTC mouse organoid lines, which are wild-type CDKN2A, to CDKN2A-null mouse organoids generated from PKP mice (ref. 26; Supplementary Fig. S1A). We found that abemaciclib had slightly lower IC₅₀ values in the CDKN2A-null organoids compared with CDKN2A-proficient KPC organoids (Supplementary Fig. S1C). In addition, abemaciclib had a significantly lower IC₅₀ in Mia PaCa2, which is CDKN2A deficient, compared with normal pancreatic cell line HPNE, which is has CDKN2A preserved (Supplementary Fig. S1D).

Abemaciclib's main mechanism of action has been demonstrated to be the inhibition of cell-cycle progression leading to senescence in Rb-positive cells (14, 21). In addition, abemaciclib has been reported in other cancer types to cause apoptosis (14, 23, 24). We sought to assess whether abemaciclib causes cell death, senescence, or both in our PDAC cells (Fig. 2A). We first assessed for cleaved caspase 3—an early marker of apoptosis upon abemaciclib treatment. We detected an increase in cleaved caspase 3 levels in Mia PaCa2 cells treated with abemaciclib compared with no treatment on days 1–3 of treatment, compared with staurosporine-treated cells that served as our positive control (Fig. 2B). To confirm these findings, we performed live-cell staining for caspase 3/7 activity and detected a significant increase in fluorescent caspase 3/7 cleavage activity (Fig. 2C) over days 1 and 2 in Mia PaCa2 cells (P < 0.001, P < 0.025, respectively). Similar results were obtained in Panc-1 and HS 766T cells (Fig. 2C). As a validating marker for early and late apoptosis, Annexin V staining was performed after 3 days of abemaciclib treatment. We detected an increase in apoptotic and dead cells when compared with no treatment in all cell lines. Upon further analysis, we found that there were more PDAC cells in early apoptosis (only Annexin V positive) compared with late apoptosis (both Annexin V and PI positive). Of note, Panc-1 had more cells in late apoptosis (both Annexin V and PI positive) after treatment (Fig. 2D; Supplementary Fig. S2A). Having demonstrated that there was significant induction of apoptosis at the time of treatment, we next sought to determine whether there was a concurrent arrest in cell-cycle progression in remaining live cells.

Abemaciclib's effect on cell-cycle progression was analyzed using phospho-Rb (pRb) immunoblot and BrdU incorporation analyses. BrdU is a florescent-labeled analogue of the DNA precursor thymidine and incorporates within dividing cells and can be used to track cells in G₁, S, and G₂ phases. Our results demonstrate that 3-day treatments with abemaciclib significantly reduced pRb expression in all three cell lines compared with no treatment (Fig. 3A). Abemaciclib treatment caused a dose-dependent inhibition of BrdU incorporation in Mia PaCa2, Panc-1, and HS 766T cell lines (Fig. 3B). In all three cell lines, there was approximately 50% decrease in BrdU incorporation starting at 0.010 μmol/L abemaciclib treatment (Fig. 3B) and at 0.5 μmol/L of abemaciclib over 75% of cells had a decrease in BrdU incorporation. Specifically, abemaciclib arrested cells in G₁–S phase (Fig. 3C) consistent with the downstream effects of Rb inactivation by CDK4/6 inhibitors (14, 32).

To determine whether abemaciclib induced senescence in PDAC cells, we first analyzed for β-galactosidase staining over a 7-day time period. Hydrogen peroxide (H₂O₂) served as an established positive control of early stress-induced senescence (refs. 33, 34; Supplementary Fig. S3A). To further validate our β-gal assay techniques, we tested previously published chemotherapeutics that have been demonstrated to induce a senescent phenotype [gemcitabine (35), irinotecan (36), 5-florouracil (37), oxaliplatin (38)], compared with paclitaxel that was not shown to induce senescence (Supplementary Fig. S3C). We found positive staining in cells as early as 2 days posttreatment with abemaciclib (Fig. 4A), consistent with senescence (14). In Mia PaCa2 and HS 766T cells, there was a peak at days 2–3 and then a plateau in β-gal–positive cells (Fig. 4C). Panc-1 cells, however, displayed a peak after 3 days of treatment. On days 5–7, approximately 75% of Mia PaCa2, 60% of HS 766T, and 40% of Panc-1 cells demonstrated strong β-gal staining indicating senescence (Fig. 4A–C). These cells also demonstrated classic senescent morphologic changes, appearing flat with spindle-shaped borders (ref. 39; Fig. 4B).

Another well-characterized marker of senescence is increased levels of senescence-associated secretion (SASP) markers. To assess changes in these markers, mRNA was extracted from cells treated with abemaciclib or H₂O₂ and analyzed for SASP markers: IL1a, MCP 1/CCL2, and DCR2 (35) over the same 7-day period. Abemaciclib treatment significantly increased mRNA expression of multiple SASP markers in all three cell lines (Fig. 4D). In the Mia PaCa2 cells, abemaciclib significantly increased MCP2/CCL2 mRNA expression as early as day 1. In Panc-1 cells, there was a significant increase in MCP2/CCL2 and DCR2 mRNA markers and in HS 766T only DCR2 mRNA marker was significantly elevated after abemaciclib treatment. This trend correlated with β-gal staining (Fig. 4C and D) and shows 60%–75% of PDAC cells undergoing senescence after abemaciclib treatment. Finally, we sought to understand whether abemaciclib's senescence induction was due to reactive oxygen species (ROS) generation, as is the case with H₂O₂ ROS induction of stress-induced senescence (33). We performed a DCFDA cellular ROS labeling and detection assay on PDAC cells treated with H₂O₂ or abemaciclib over 7 days. We found that H₂O₂ increased DCFDA florescence (positive control) 24 hours after treatment, but abemaciclib did not cause any significant increase in DCFDA florescence even at day 7 (Supplementary Fig. S3B).

In an in vivo study using Mia PaCa2 cells in a subcutaneous xenograft mouse model, we observed a significant decrease in tumor growth in the abemaciclib-treated arm versus control (P < 0.0001, Fig. 5A; Supplementary Fig. S4A) and the therapy was safe as demonstrated by mouse weights remaining within 10% of baseline (Fig. 5A). At the end of study, tumors were harvested and weighed (Fig. 5B), showing significant difference in tumor weights of abemaciclib-treated group compared with vehicle control.

Harvested tumors from the xenograft study were sectioned and stained for pRb, Rb, and Ki67. The abemaciclib-treated tumors demonstrated a decrease in pRb and total Rb expression compared with vehicle (Fig. 5C; Supplementary Fig. S4C). Pathologist scoring and quantification of IHC staining shows that in the abemaciclib-treated tumors, there were 46.3% cells with positive pRb staining, compared with no treatment that had 74.5% cells with positive pRb. In addition, compared with no treatment that had 78.8% cells with positive Ki67 staining, the abemaciclib-treated tumors had 59.5% cells with positive Ki67 staining (Fig. 5D). Overall, when normalized to no treatment, abemaciclib-treated tumors had approximately 38% reduction in pRb staining, and 25% reduction in pRb/total Rb staining along with Ki67 staining (Fig. 5D). Compared with vehicle alone, abemaciclib-treated tumors had weak TUNEL-positive staining (Fig. 5D). RNA extracted from these tumors were probed for SASP markers indicative of senescence and we found significant increases in IL1a, MCP1/CCL2, and DCR2 mRNA, in the abemaciclib-treated tumors when compared with vehicle alone (Fig. 5E). These results demonstrate that abemaciclib treatment induces apoptosis, decreases pRb and tumor growth, and induces senescence in this in vivo xenograft model.

To determine whether any of the drugs standardly used in PDAC synergized with abemaciclib, we evaluated synergism of abemaciclib with standardly used PDAC therapeutics using the Loewe synergism method (40). Our initial panel covered the components of FOLFIRINOX (Supplementary Fig. S5A), as well as gemcitabine and paclitaxel (Fig. 6A; Supplementary Fig. S5B); unfortunately, we did not find any significant synergism with these standard-of-care agents except for weak synergy with 5-FU. As we did not observe significant synergy with standard therapeutics, we wanted to determine whether other targeted therapeutics may synergize more potently with abemaciclib in PDAC.

Our lab has extensively characterized the importance of the mRNA-binding protein, Human Antigen R (HuR), as an essential mediator of PDAC cell survival and chemotherapeutic resistance (28, 41). Importantly, HuR is a critical regulator of cyclin D1, with loss of HuR leading to a dramatic decrease in cyclin D1 expression at the mRNA and protein level (42, 43). Therefore, we hypothesized that CDK4/6 inhibition combined with HuR inhibition could provide an enhanced response and increased sensitivity to abemaciclib. We found synergism with two published HuR inhibitors, pyrvinium pamoate (44) and CMLD2 (41) when dosed in combination with abemaciclib in Mia PaCa2 cells (Fig. 6A; Supplementary Fig. S5B) or with Panc-1 cells (Fig. 6B; Supplementary Fig. S5B) in a short-term Pico green assay.

Yes-Associated Protein 1 (YAP1) is a transcriptional coactivator that is a critical regulator of PDAC development (45) and is also a known regulator of cyclin D1 (46) and CDK6 (47, 48). It has also previously been demonstrated that inhibition synergizes with CDK4/6 inhibitors in esophageal cancer (48) and thus, we tested the ability of YAP-1 inhibitors to synergize with abemaciclib treatment in PDAC. We observed synergism in PDAC cells treated with abemaciclib and the YAP-1 inhibitors verteporfin (49) and CA3 (ref. 48; Fig. 6A and B; Supplementary Fig. S5B).

To assess for synergistic combinations in a long-term assay, we performed colony-forming assays with abemaciclib and the HuR and YAP1 inhibitors that were found to be synergistic in long-term drug sensitivity assay as well (Fig. 6C and D). We found that in both Mia PaCa2 and Panc-1, the combination of abemaciclib with either a HuR inhibitor or YAP1 inhibitor was able to decrease the number of colonies than monotherapy (Fig. 6C and D).

To validate the role of HuR loss and abemaciclib, we tested abemaciclib in our Mia PaCa2 HuR CRISPR KO cells (ref. 28; CRISPR KO validation data; Supplementary Fig. S6C) and found that there was a decrease in abemaciclib's IC₅₀ in two different CRISPR KO clones (ELAVL1^−/−) when compared with the parental cell lines (Supplementary Fig. S6A) confirming our HuR inhibitor data. In these HuR KO cells, there was a more significant decrease in pRb levels after abemaciclib treatment compared with parental cells (ELAVL1^+/+) that were treated with abemaciclib (Supplementary Fig. S6B).

As a secondary method to confirm that inhibition of HuR or YAP1 increased abemaciclib's sensitivity in PDAC cells, we transfected cells with either siHuR or siYAP1 oligonucleotides. When compared with si negative control, cells transfected with siHuR and siYAP1 had a lower IC₅₀s for abemaciclib treatment compared with si negative control cells, demonstrating an increase in sensitivity for abemaciclib due to HuR or YAP1 silencing (Supplementary Fig. S6D). We validated target knockdown by Western blot analysis and found there was a 67% knockdown of HuR and 90% knockdown of YAP1, as compared with control cells, three days after transfection had taken place (Supplementary Fig. S6E).

Abemaciclib treatment lead to a decrease in pRb in all the samples compared with their no treatment counterparts, and in the siRNA negative control sample, led to an increase in cyclin D1 (Supplementary Fig. S6E). Increased expression of cyclin D1 after CDK4/6 inhibition has been seen in other models (19) and has been linked to a possible mode of CDK4/6 resistance, but interestingly, after HuR or YAP1 knockdown, there is no increase in cyclin D1 protein levels in the abemaciclib-treated samples (Supplementary Fig. S6E). To confirm that cyclin D1 is regulated by HuR in our cell culture model, we performed an HuR RNA immunoprecipitation assay (RIP) and probed for p27, cyclin D1, and an established HuR target PIM1 (ref. 28; Supplementary Fig. S6F). Compared with IgG, we found significant increases in cyclin D1, p27, and our positive control PIM1 (P < 0.0001) in the HuR pulldown mRNA (Supplementary Fig. S6F).

To assess whether synergism between abemaciclib and HuR or YAP1 inhibition was through their shared regulation of cyclin D1 (42, 43, 46), we transfected cells with si cyclin D1 oligonucleotides. Because of the finding that cells transfected with si cyclin D1 grew slower than the negative control cells, we calculated GR₅₀s to compare between the two conditions. When compared with si negative control, cells transfected with si cyclin D1 had a lower IC₅₀ for abemaciclib treatment compared with siControl cells, demonstrating an increase in sensitivity for abemaciclib due to cyclin D1 silencing (Supplementary Fig. S6G). We validated target knockdown by Western blot analysis (Supplementary Fig. S6H) and found there was a 63% knockdown of cyclin D1 three days after the transfection had taken place (Supplementary Fig. S6H). In addition, in the si cyclin D1 abemaciclib-treated samples, there was no further increase in cyclin D1 levels when compared with the abemaciclib-treated si negative control sample (Supplementary Fig. S6H), validating what we had seen earlier with the siHuR and siYAP1 abemaciclib-treated cells (Supplementary Fig. S6E).

To model the potential effects and resistance mechanisms caused by long-term abemaciclib therapy in patients, Mia PaCa2 and Panc-1 cells were cultured with increasing concentrations of abemaciclib for 10–12 months to create abemaciclib-resistant clones. These cells had a 2- to 4-fold increase in GR₅₀ in the “chronic” abemaciclib therapy cells (meaning these cells were less sensitive) when treated with abemaciclib indicating a resistance phenotype (Fig. 7A–C). To evaluate whether these “chronic” abemaciclib therapy cells had any increased or decreased responsiveness to other chemotherapeutics, we calculated GR₅₀ on common components of FOLFIRINOX, along with HuR and YAP1 inhibitors used earlier. As seen in Fig. 7B and C, we found that both the Mia PaCa2 and Panc-1 “chronic” abemaciclib therapy cells were more resistant to gemcitabine and 5-FU but remained sensitive to pyrvinium pamoate (HuR inhibitor) and verteporfin (YAP1 inhibitor). In addition, when these “chronic” abemaciclib therapy cells were treated with abemaciclib and gemcitabine in combination, we found antagonistic drug combinations, but when we treated these “chronic” abemaciclib therapy cells with HuR inhibitor, CMLD2, or YAP1 inhibitors, CA3 or verteporfin, in combination and abemaciclib, there was increase of abemaciclib with HuR and YAP-1 inhibitors (Fig. 7D and E). These findings demonstrate that the Mia PaCa2 “chronic” abemaciclib therapy cells have more significant synergy with these compounds compared with that of the parental cell line in Fig. 6A and B. Finally, to understand whether these “chronic” abemaciclib therapy cells were able to maintain their resistance to abemaciclib after removal of drug therapy, we removed abemaciclib for 21 days and compared GR₅₀s in these cell lines. In both Mia PaCa2 and Panc-1 “chronic” abemaciclib therapy lines, we found that once we removed abemaciclib from the media these cells had phenotypes similar to their parental counterpart cells. We observed that as early as 14 days and 21 days without abemaciclib, these “chronic” abemaciclib (CA)-treated cells had decrease in GR50 to abemaciclib, as seen in Supplementary Fig. S7A. Although these “chronic” abemaciclib–treated cells stained positive for β-gal (Supplementary Fig. S7B), they did not have elevations in SASP markers (Supplementary Fig. S7C), indicating that the positive staining is likely due to stress-induced senescence.

The premise for studying CDK4/6 inhibition in PDAC cells has been well established (2, 7, 18, 19). Specifically, the CDK4/6 inhibitor, abemaciclib's effects in breast cancer and other solid tumor types have been well characterized (14, 20–22, 50), but little is known about its effect in PDAC. We demonstrate that abemaciclib treatment caused inhibition of pRb and induced senescence in vitro and a decrease in tumor volume in vivo. In addition, we found a reduction in total Rb in vivo, as was seen in other tumor models including breast cancer and melanoma (50).

As most recent positive clinical trials for patients with pancreatic cancer have involved combination therapies (5, 51), we performed a focused drug screen to identify agents that would synergize with abemaciclib in PDAC cells. Previously, it has been reported that palbociclib is synergistic with gemcitabine in PDAC (18) and abemaciclib is synergetic with gemcitabine in lung cancer (52). Our data demonstrate an additive nature between abemaciclib and gemcitabine in PDAC (Fig. 6A and B). In addition, synergism was observed between abemaciclib and HuR and YAP1 inhibitors. We confirmed targeting of these small molecules with siRNA silencing of HuR and YAP1 in PDAC cells. Mechanistically, rationale exists to explain the synergy of combining abemaciclib with inhibition of HuR and YAP1, as both HuR and YAP1 can regulate cyclin D1 (42, 43, 46). Muralidharan and colleagues (42) found that treatment with siHuR oligos in a lung cancer model lead to downregulation of cyclin D1 and upregulation of p27, both of which are important for cell-cycle regulation and interaction with the cyclin D1–CDK4/6 complex (42). In addition, Li and colleagues (48) has published that YAP1 inhibition, using the inhibitor CA3, was synergistic with CDK6 inhibition in esophageal cancer both in vitro and in vivo (48). The combined effects of abemaciclib and HuR inhibition or YAP1 inhibition on the same cyclin D1/ CDK4/6 axis could logically explain the synergism we observed. Along with validating that cyclin D1 pulls down with HuR via RIP (Supplementary Fig. S6F), we have observed that by targeting cyclin D1, we were able to sensitize cells to abemaciclib treatment (Supplementary Fig. S6G).

In addition to studying synergistic drug combinations in PDAC cells, we sought to make abemaciclib-resistant cell lines to understand resistance mechanisms and to model the potential effects of long-term abemaciclib therapy on patients. Knudsen and colleagues recently published a review discussing mechanism of resistance to CDK4/6 inhibitors (19). They found that although the mechanisms for CDK4/6 inhibitor resistance are not fully elucidated, there are important drivers in the AKT, mTOR, and cyclin D1 pathways to promote resistance (19). In breast cancer models, upregulation of AKT is linked to the accumulation of cyclin D1, and in pancreatic cancer models upregulation of cyclin D1 can be decreased by using mTOR inhibitors (19). Cyclin D1 has an important role in developing acquired resistance to CDK4/6 inhibitors, although this mechanism is unclear (19). To assess an abemaciclib-resistant phenotype, we attempted to create abemaciclib-resistant cell lines. After 10–12 months of abemaciclib treatment, these “chronic” abemaciclib therapy cells had a decrease in growth rate and were more resistant to both abemaciclib, gemcitabine, and 5-FU therapy, but not to HuR/ YAP1 inhibitors. This may be simply due to the cells not proliferating as rapidly, and therefore, these cells may be less sensitive to chemotherapeutics and antimetabolites that are ineffective in nondividing cells. Alternatively, these findings may offer further evidence that HuR/YAP1 inhibitors may work well with abemaciclib in treating patients with PDAC. Although the resistance to abemaciclib appeared to be a transient response, as these cells regained sensitivity to abemaciclib 2–3 weeks after the cells were removed from chronic abemaciclib exposure. Further investigation into the biology of these “chronic” abemaciclib therapy cells may give us new insights on possible drug combinations that can be used to break resistance in PDAC cells or provide information for next line treatment options (i.e., post-abemaciclib therapy).

Taken together, we present a promising preclinical evaluation of abemaciclib in both PDAC in vitro and in vivo models. Importantly, we were able to identify novel synergy between abemaciclib and HuR/ YAP1 inhibitors, as potential clinically relevant combination therapies to treat this deadly disease. Further investigations are warranted to properly characterize HuR/YAP1 inhibitors and these drug combination therapies, with the goal to test the best combination for the treatment of PDAC. Even though more diverse and sophisticated models could be used to evaluate abemaciclib-based therapies, we strongly believe these studies support the notion for a clinical trial evaluating abemaciclib in patients with PDAC.

T. Golan reports receiving commercial research grants from MSD Merck and, AstraZeneca, has received speakers bureau honoraria from Abbvie, and has a consultant/advisory board relationship with Abbvie, AstraZeneca, and Teva. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T. Dhir, C.W. Schultz, S.Z. Brown, J. Posey, J.R. Brody

Development of methodology: T. Dhir, C.W. Schultz, A. Jain, S.Z. Brown

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Dhir, C.W. Schultz, A. Haber, A. Goetz, C. Xi, W. Jiang, T. Golan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Dhir, C.W. Schultz, A. Jain, A. Goetz, W. Jiang, C.J. Yeo, M.J. Pishvaian, J.R. Brody

Writing, review, and/or revision of the manuscript: T. Dhir, C.W. Schultz, A. Jain, S.Z. Brown, A. Haber, G.H. Su, W. Jiang, C.J. Yeo, T. Golan, M.J. Pishvaian, J.R. Brody

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Dhir, C.W. Schultz, L. Xu, C.J. Yeo

Study supervision: G.H. Su, J.R. Brody

This study was performed with grant support from the Newell Devalpine Foundation, 1R01CA212600-01 (to J.R. Brody) and T32 training grant NIH/NIGMS T32GM008562 (to T. Dhir). This work was also supported, in part, by a National Cancer Institute of the National Institutes of Health under a Cancer Center Support Grant 5P30CA056036-17 (SKCC, TJU). This study was not funded by Eli Lilly and Company. IHC slides were processed and stained by Raymond O'Neill, Medical Technologist at Thomas Jefferson University Hospital (Philadelphia, PA).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.