MARK2 Regulates Chemotherapeutic Responses through Class IIa HDAC-YAP Axis in Pancreatic Cancer

Abstract

Despite paclitaxel’s wide use in cancer treatment, patient response rate is still low and drug resistance is a major clinical obstacle. Through a Phos-tag-based kinome-wide screen, we identified MARK2 as a critical regulator for paclitaxel chemosensitivity in PDAC. We show that MARK2 is phosphorylated by CDK1 in response to antitubulin chemotherapeutics and in unperturbed mitosis. Phosphorylation is essential for MARK2 in regulating mitotic progression and paclitaxel cytotoxicity in PDAC cells. Mechanistically, our findings also suggest that MARK2 controls paclitaxel chemosensitivity by regulating class IIa HDACs. MARK2 directly phosphorylates HDAC4 specifically during antitubulin treatment. Phosphorylated HDAC4 promotes YAP activation and controls expression of YAP target genes induced by paclitaxel. Importantly, combination of HDAC inhibition and paclitaxel overcomes chemoresistance in organoid culture and preclinical PDAC animal models. The expression levels of MARK2, HDACs, and YAP are upregulated and positively correlated in PDAC patients. Inhibition of MARK2 or class IIa HDACs potentiates paclitaxel cytotoxicity by inducing mitotic abnormalities in PDAC cells. Together, our findings identify the MARK2-HDAC axis as a druggable target for overcoming chemoresistance in PDAC.

Introduction

Mitosis must be precisely regulated and aberrations in mitosis often result in tumor initiation. Therefore, the mitotic window has been the target of anticancer drugs for decades (1, 2). Identifying regulators and/or signaling involved in the mitotic cell cycle may lead to the addition of druggable targets and the development of novel chemotherapeutics for combatting cancer.

Antitubulin agent paclitaxel (Taxol) is one of the antimitotic drugs and has been widely used in ovarian, breast, non-small cell lung cancers, and recently pancreatic ductal adenocarcinoma (PDAC) (2-4). Pancreatic cancer is projected to become the second leading cause of cancer-related death by 2030 (5). PDAC accounts for more than 90% of pancreatic cancer cases (6) and the majority of patients are diagnosed at a distant stage (about 53%), for which 5-year survival is only 3% (7). In approximately 80-90% of PDAC cases, tumors are already locally advanced at diagnosis and cannot receive curative resection treatment (7). Recently, nanoparticle albumin-bound (Nab) Paclitaxel combined with gemcitabine showed a significant improvement in the treatment of advanced PDAC patients (8-10) and the U.S. FDA established it as a first-line therapy for metastatic PDAC patients. However, approximately 69 to 75% of patients who receive adjuvant treatment relapse within two years (11, 12). The overall response rate of pancreatic cancer to current therapies is still very low due to the occurrence of drug resistance and attempts at reversing the resistance have not been successful in the clinic (13).

Microtubule affinity-regulating kinase proteins (MARK1-4) are critical for microtubule dynamics due to their modulation of MAPs (microtubule-associated proteins) such as Tau (14, 15). They are members of the AMPK/Snf1 family and control cell polarity and asymmetric cell division (15, 16). Deregulation of MARKs has also been associated with pathological conditions such as cancer and Alzheimer disease (16). For example, MARK1 is involved in cervical tumor cell migration (17) and MARK1-mediated phosphorylation is critical for DIXDC1’s tumor-suppressive function in lung cancer (18). MARK2 is an oncogene in non-small cell lung cancer associated with cisplatin resistance and the DNA damage response (19). MARK4 is upregulated in glioblastomas and implicated in prostate cancer, breast cancer, hepatocarcinoma, and leukemia (20-25). Several recent studies showed that MARK family members are important regulators of Hippo-YAP signaling (20, 26-28), which is critical in cancer development, drug resistance, and stem cell biology (29-33). However, how MARKs are regulated and their biological significance in response to antimitotic agents have not been defined.

MARKs regulate localization and activity of class IIa histone deacetylases (HDAC4/5/7/9) by direct phosphorylation (34). Since HDACs reverse the acetylation of nucleosomal histones by histone acetyltransferases and promote chromatin condensation and transcriptional repression, MARK phosphorylation promotes the cytoplasmic retention of HDACs and inhibits their deacetylation/transcription repression activity (34-36). HDACs regulate many cellular processes critical in tumorigenesis, including gene transcription, cell cycle progression, cell survival, DNA repair, protein trafficking, protein degradation, and cell migration (37). The HDAC inhibitor vorinostat (suberoylanilide hydroxamic acid, or SAHA) has been approved by the U.S. FDA for patients with cutaneous T-cell lymphoma. While HDAC inhibitor monotherapy showed clinical activity in hematological malignancies, results have been largely disappointing in most solid tumors including PDAC (38). Thus, a major effort for the HDAC inhibitors’ clinical application is focusing on combinations with other chemotherapeutic agents (39). Despite the important function of HDACs in tumorigenesis, how they sense the stress signal to antimitotic drugs (e.g. paclitaxel) is very unclear.

In order to understand how the kinome is involved in regulating antitubulin chemotherapeutics, we conducted kinome-wide screens using a Phos-tag-based approach. Our screens have a twofold purpose: 1) identifying novel mitotic kinases, as antitubulin agents arrest cells at mitosis, and 2) identifying novel kinases that determine antitubulin drug sensitivity. These screens identified MARK2 as a phospho-protein during mitosis and a critical determinant of paclitaxel cytotoxicity in PDAC cells. We further showed that MARK2 controls paclitaxel chemosensitivity in PDAC by phosphorylating class IIa HDACs. MARK2-HDAC controls a unique transcriptional program in response to antitubulin agents in a YAP-dependent manner. Our findings reveal the MARK2-HDAC-YAP axis as a druggable target for overcoming the resistance to chemotherapy in PDAC.

Materials and Methods

Cell culture and transfection

The immortalized human pancreatic nestin-expressing (HPNE) cell line was provided by Dr. Michel Ouellette (University of Nebraska Medical Center), who originally established the cell line and deposited it at American Type Culture Collection (ATCC) (74). Dr. Michael (Tony) Hollingsworth (University of Nebraska Medical Center) kindly provided the T3M4, S2.013, and Colo-357 pancreatic cancer cell lines. HEK293T, HeLa, HPAF-II, Capan-2, PANC-1, BxPC-3, and Hs776T cell lines were purchased from ATCC and cultured as ATCC instructed. The cell lines were authenticated at ATCC and were used at low (<30) passages. The mouse PDAC cell lines UN-KC-6141, UN-KPC-960, and UN-KPC-961 have been described (64). Attractene (Qiagen) was used for transient overexpression of proteins in HEK293T and HEK293GP cells, following the manufacturer’s instructions. HiPerFect (Qiagen) was used for siRNA transfections. SiRNA duplex for MARK2 was obtained from Sigma-Aldrich. Nocodazole (100 ng/ml for 20 h) and Taxol (100 nM for 20 h) (Selleck Chemicals) treatments were used to arrest cells in mitosis unless otherwise indicated. VX680, ZM447439, BI2536, purvalanol A, SP600125, rapamycin, and MK2206 were purchased from Selleck Chemicals. RO3306 and roscovitine were from ENZO life Sciences. U0126, SB203580, and LY294002 were from LC Laboratory. MK5108 (Aurora-A inhibitor) was from Merck. All other chemicals were from either Sigma-Aldrich or Thermo Fisher.

Expression constructs

The following plasmids were purchased from Addgene: pBa-eGFP-MARK2-WT (#66706), pEGFP-HDAC4-WT (#45636) and pEGFP-HDAC4-3SA (#45637), pLKO.1-H2B-RFP (#26001), pLenti PGK V5-LUC Puro (#19360), and GFP-Cyclin B1-R42A (#61849). Expression constructs GFP-YAP, Myc-WW45, Flag-MST1, Flag-MST2, Myc-LATS2 have been described previously (52). To make MARK2 or HDAC4 expression constructs, the full-length cDNA was cloned into the pSIN4-Flag-IRES-puro vector (43). Point mutations were generated by the QuikChange Site-directed PCR Mutagenesis Kit (Stratagene) and verified by Sanger sequencing.

shRNA-mediated knockdown and CRISPR/Cas9-mediated knockout

Downregulation of MARK2, HDAC4, or HDAC7 in HeLa, S2.013, Capan2, PANC-1, and UN-KC-6141 cells was achieved by lentivirus-mediated corresponding shRNA expression (75). All MISSION shRNA constructs were purchased from Sigma-Aldrich and targeting sequences are listed in Table S1. The shRNA construct targeting the 3’-UTR of human MARK2 was also used to knock down mouse Mark2 in KC6141 cells. Ectopic expression of MARK2, MARK2-3A, MARK2-KD, HDAC4, or HDAC4-3A was also achieved by a lentivirus-mediated approach (75). The HeLa cell line expressing TetOn-shCDK1 has been described (56).

To construct the EGFP-expressing all-in-one CRISPR/Cas9n plasmid targeting human MARK2, the sense and antisense oligonucleotides from Table S1 were synthesized, annealed and Golden Gate-assembled into the pX330A_D10A-1×2-EGFP and pX330S-2 vectors as described previously (43). The final construct was transfected into U2OS cells, and GFP-positive clones were selected by flow cytometry-based cell sorting.

Recombinant protein purification and in vitro kinase assay

GST-tagged MARK2, MARK2-3A (amino acids 388-788), HDAC4, HDAC4-3A (amino acids 200-680) and MST2-KD (kinase dead) proteins were bacterially expressed and purified on GSTrap FF affinity columns (GE Healthcare) following the manufacturer’s instructions. The GST-MARK2 proteins (0.5-1 μg) were incubated with recombinant 100 ng of CDK1–Cyclin B1 (SignalChem) in kinase buffer (New England BioLabs) in the presence of 10 μCi of [γ-³²P]ATP (3000 Ci/mmol, Perkin Elmer Life Sciences). GST-MST2-KD and GST-HDAC4 proteins (1 μg) were also used as substrates for MARK2 (150 ng) phosphorylation. Active MARK2, CDK5–p25, MEK1, ERK1, JNK1, JNK2, and p38α kinases were purchased from SignalChem. Phosphorylation (³²P incorporation) was visualized by autoradiography followed by Western blotting or detected by phospho-specific antibodies.

Western blotting, immunoprecipitation, and Phos-tag analysis

Cell lysate preparation, Western blotting analysis, immunoprecipitation, and lambda phosphatase treatment were done as previously described (76). Phos-tag was obtained from Wako Pure Chemical Industries, Ltd. (catalog no. 304-93521) and used at concentrations of 10-25 μM (with 100 μM MnCl₂) in 6-8% SDS-acrylamide gels.

Immunofluorescence (IF) staining, confocal microscopy, and immunohistochemistry (IHC) staining

Fluorescence staining was done as described (77). The stained cells were mounted with Fluoromount (Vector Laboratories) and visualized on an LSM800 Zeiss fluorescence microscope (Carl Zeiss). The ZEN 2.3 (blue edition) software (Carl Zeiss) was used to analyze and process all IF images. YAP cellular localization was visualized by YAP IF staining with an Alexa Fluor 647-conjugated YAP antibody (1:100, Cell Signaling Technology) following manufacturer’s instructions. The p-HDAC4 S246 antibody (1:100, Cell Signaling Technology) was also used for IF staining. IHC staining with cleaved caspase 3 (1:100, Cell Signaling Technology) in tumor tissues was performed according to a previously described protocol (59).

Antibodies

Rabbit polyclonal phospho-specific antibodies against MARK2 S456, S569, S619, and MST2 S15 were generated and purified by AbMart, Inc. The peptides used for immunizing rabbits were AKVPA-pS-PLPGL (MAR2 S456), RVPVA-pS-PSAHN (MARK2 S569), GVTPA-pS-PSGHS (MARK2 S619), and KLKKL-pS-EDSLT (MST2 S15). The corresponding non-phosphorylated peptides were also used for antibody purification and blocking assays. Anti-MARK2, Cyclin B1, β-actin were purchased from Santa Cruz Biotechnology. Anti-MARK1, MARK3, MARK4, p-MARK T208 (activation loop), HDAC1, HDAC2, HDAC4, HDAC5, HDAC6, HDAC7, p-HDAC4/5/7(S246/S259/S155), p-HDAC4/5/7(S632/S661/S486), cleaved-PARP (human specific), cleaved-PARP (rodent specific), cleaved-caspase 3, Erk1/2, Zyxin, Survivin, p-YAP S127, p-YAP S397, YAP, and p-Aurora-A/B/C were from Cell Signaling Technology. Anti-GST and LATS2 antibodies were from Bethyl Laboratory. Anti-Aurora-A and Flag antibodies were from Sigma. Anti-α-tubulin antibody was from Abcam Inc. All other antibodies used in this study are listed in Table S1.

qRT-PCR

The Direct-zol RNA Kit (ZYMO Research) was used for total RNA isolation. RNA reverse transcription and qRT-PCR were done as we have described (52).

Live-cell imaging

Cells were plated on black 96-well optical bottom plates (Thermo Fisher). Live-cell imaging was performed in a Cellomics Arrayscan VTI HCS Reader with 37°C, 5% CO₂ incubation using FluroBrite DMEM (Thermo Fisher). Cells were monitored for 24 h and pictures were taken every 5 min using an RFP filter. Measurements of cell cycle durations were done using the time-lapse sequences.

Luciferase reporter assays

Luciferase reporter assays were done as we previously described (52).

Caspase-Glo 3/7 assay

Control and MARK2-KD cells were plated in white-walled 96-well plates (NUNC, 136101, Thermo Fisher) for 24 hours and treated with paclitaxel for an additional 16 hours. Apoptosis was measured by Caspase-Glo 3/7 Assay Kit (Promega, G8090) according to the manufacturer’s instructions and quantified on a luminometer.

Clonogenic assay

Cells (2,000/well) were seeded at low density in six-well plates and incubated overnight. The cells were then treated with paclitaxel for 24 hours and were replaced with fresh medium without drug. After 14 days of incubation, colonies were fixed with 4% paraformaldehyde for 15 minutes at room temperature and were stained with 0.5% crystal violet for 30 minutes. Quantification was achieved by ImageJ software (ImageJ 1.53e, National Institutes of Health, USA).

PDAC organoid culture and drug treatment

The PDAC organoids were established from human pancreatic patient-derived-xenograft (PDX) tissues as described previously (78). 4.0×10³ cells of established organoids were embedded in GFR Matrigel (BD Biosciences) and cultured in 48-well plates (78). On the second day after seeding, organoids were drugged with paclitaxel (5-20 nM), vorinostat (5-20 nM) or the combination. After 72 hours of drug treatment, organoids were imaged with Cytation 3 (Biotek). Organoid viability was assessed with Cell Titer Glo 3D Cell Viability assay (Promega) according to the manufacturer’s protocol. Combination index (CI) was calculated by Compusyn software (https://www.combosyn.com/).

Animal studies

For in vivo xenograft studies, mouse PDAC cells (5.0×10⁶) were subcutaneously injected into each flank of six to eight-week-old male or female immunocompetent C57BL/6 mice (Charles River Laboratories, Wilmington, MA, USA). Cells were suspended in phosphate-buffered saline (PBS). Tumor sizes were measured twice a week using an electronic caliper, starting at ten days post injection. Tumor volume (V) was calculated by the formula: V = 0.5×length×width².

For orthotopic studies, UN-KC-6141 cells expressing luciferase (5000 cells/50 μl Hanks’ Balanced Salt Solution each mouse) were injected into the head of the pancreas of six to eight-week-old C57BL/6 mice (Charles River Laboratories). The orthotopic injection was performed as previously described (79). Approximately 10 min prior to imaging, mice were intraperitoneally injected with D-luciferin (150 mg/kg). Tumor burdens (primary and metastasis) were monitored by measuring bioluminescence emission using IVIS imaging system (Perkin Elmer) once a week starting at seven days post injection. Gemcitabine was purchased from SAGENT Pharmaceuticals (NDC 25021-234-10). Paclitaxel was purchased from Hospira, Inc. (NDC 61703-342-22). Vorinostat/SAHA was purchased from Selleck Chem and was dissolved in 2% DMSO, 30% PEG300, 5% Tween 80, and ddH₂O and was used every other day at 25 mg/kg via intraperitoneal injection. Mice were euthanized by CO₂ inhalation at the end of the experiment and the tumors were excised for subsequent analysis. All animals were housed in pathogen-free facilities. Animal experiments were approved by the University of Nebraska Medical Center Institutional Animal Care and Use Committee.

Statistical analysis

Statistical significance was analyzed using a two-tailed, unpaired Student’s t-test or a one-way ANOVA with a post-hoc Tukey HSD (Honestly Significant Difference) test for multiple comparisons. A P-value of <0.05 was considered to indicate statistical significance.

Results

A Phos-tag-based kinome screen for regulators of antitubulin drug response

To identify new kinases that are triggered by antitubulin agents, we used two approaches. First, we systematically investigated each individual kinase’s response (as shown by its expression level) to antitubulin drugs (paclitaxel and nocodazole) by Western blotting. Second, using a Phos-tag method, we determined each kinase’s phospho-regulation (phosphorylation status) in response to antitubulin drugs. The Phos-tag specifically binds phosphate ions and selectively separates phosphorylated proteins on SDS-PAGE gels (40). The human kinome has 518 protein kinases (41) including 90 protein tyrosine kinases (42). We screened 115 kinases (79 protein tyrosine kinases and 36 serine/threonine protein kinases) using primary antibodies obtained from Cell Signaling Technology, Bethyl Laboratories, and Santa Cruz Biotechnology (Table S1) (Fig. 1A and Fig. S1) (43-46). HeLa cells were used for the screen since they are easily arrested/synchronized in mitosis by paclitaxel and are widely used for mitotic studies.

Figure 1. A Phos-tag-based kinome wide screen identifies MARK2 as a phospho-kinase by CDK1 during antitubulin agent-induced mitotic arrest.

(A) HeLa cells were treated with DMSO, nocodazole (Noco, 100 ng/ml for 20 h, or Taxol (100 nM for 20 h). Total cell lysates were probed with the indicated antibodies on Phos-tag or regular SDS-polyacrylamide gels with the indicated antibodies.

(B) HeLa cells were treated with nocodazole as indicated and cell lysates were further treated with (+) or without (−) λ-phosphatase (ppase). Total cell lysates were probed with the indicated antibodies.

(C) HeLa cells were treated with nocodazole, with or without various kinase inhibitors as indicated. VX680 (2 μM), RO3306 (5 μM), BI2536 (100 nM), purvalanol A (10 μM), U0126 (20 μM), MK-2206 (10 μM), SB203580 (10 μM), SP600125 (20 μM), and rapamycin (100 nM) were used. Inhibitors were added (with MG132 to prevent Cyclin B1 from degradation and cells from exiting from mitosis) 2 h before harvesting the cells.

(D) GST-tagged MARK2 (amino acids 388-788) proteins were used for in vitro kinase assays with various purified kinases. MBPs (myelin basic protein) (Sigma) were included as positive controls, confirming the kinases are active.

(E) Identification of phosphorylation sites in MARK2.

(F) GST-tagged MARK2 or MARK2-3A (S456A/S569A/S619A) proteins were used for in vitro kinase assays with purified CDK1-Cyclin B1 kinase complex. RO3306 (5 μM) was used to inhibit CDK1 kinase activity.

(G) In vitro kinase assays with purified CDK1-Cyclin B1 complex and recombinant GST-MARK2 or GST-MARK2-3A and probed with phospho-antibodies.

(H) Endogenous MARK2 was immunoprecipitated from HeLa cells treated with nocodazole or Taxol and probed with the MARK2 p-S569 antibody.

(I) HeLa cells were transfected with 40 nM scramble (control) or siRNA against MARK2 for 48 h and were further treated with Taxol as indicated. Total lysates were probed with the MARK2 p-S619 antibody.

(J, K) HEK293T cells were transfected with Flag-MARK2. At 32 h post-transfection, the cells were treated with Taxol. Total cell lysates were subjected to Western blotting with the indicated antibodies. Non-p peptide: Western blotting in the presence of control (not phosphorylated) peptide; phospho-peptide: Western blotting in the presence of phosphorylated peptide (used for antibody generation).

(L) HEK293T cells were transfected and treated with Taxol together with or without various kinase inhibitors as indicated.

(M) CDK1 knockdown inhibited MARK2 phosphorylation. TetOn-inducible shRNA targeting CDK1 was expressed in HeLa cells. The cells were treated with or without doxycycline (Dox) for 2 days and were further treated with Taxol for an additional 20 h. Immunoprecipitated MARK2 and total cell lysates were subjected to Western blotting with the indicated antibodies.

(N) HEK293T cells were transfected as indicated. At 48 h post-transfection, total cell lysates were subjected to Western blotting with the indicated antibodies. F-CDK1 indicates the constitutively active form of Flag-tagged CDK1 (T14 and Y15 sites were mutated to non-phosphorylatable Alanine and Phenylalanine). G-CycB1: GFP-tagged constitutive active of Cyclin B1 (R42A non-degradable mutant).

Previous studies showed that ABL1 (47), AMPK (45), FAK (48), EGFR (49), SRC (50), and YES (46) are phosphorylated during mitotic arrest. We confirmed their mobility upshift (phosphorylation) during drug-induced mitotic arrest (Fig. S1), which validated the robustness of our screen. We also identified numerous novel alterations of protein kinases in response to antitubulin drugs, including ABL2 (ARG), ACK, EphA2, EphA3/4/5, EphB4, FGFR2, JAK1, MER, MET, PYK2, ROR2, AMPK, CHK1, IKKβ, MARK2/3, PKN1/2/3, PKR, and TAOK1. These kinases were upshifted (phosphorylated) during Taxol or nocodazole treatment, suggesting roles for these proteins in the response to antitubulin drugs (Fig. 1A and Fig. S1). Another significant change we noted in these cells was marked reduction of protein levels of FGFR4, TNK1, and AXL during antitubulin treatment (Fig. 1A and Fig. S1). MARK2 is known as an important regulator for microtubule dynamics and is expected to potentially affect mitosis and paclitaxel cytotoxicity. In addition, MARK family proteins are upstream regulators of the Hippo-YAP pathway. Therefore this study characterizes the increased phosphorylation of MARK2 and its role in mitotic progression and antitubulin chemotherapeutics.

CDK1 phosphorylates MARK2 in vitro

We further confirmed that the mobility upshift of MARK2 occurred in other cancer cells including PDAC cells (Fig. S2A). The phosphorylation at T208 (the autophosphorylation site in the activation loop of MARK2) was not altered during antimitotic drug-induced mitosis (Fig. S2B), suggesting the existence of phosphorylation site(s) other than T208 in MARK2. Lambda phosphatase treatment largely converted MARK2 mobility upshifted bands to fast-migrating bands, confirming MARK2 was phosphorylated during mitotic arrest induced by antimitotic agents (Fig. 1B).

We used various kinase inhibitors to identify the candidate kinase for MARK2 phosphorylation. Treatments with RO3306 (a CDK1 inhibitor) or purvalanol A (a CDK1/2/5 inhibitor) almost completely inhibited the mobility shift (Fig. 1C, lanes 4 and 6). These data suggest that CDK1 is likely the relevant kinase for MARK2 phosphorylation induced during paclitaxel or nocodazole treatment.

To determine whether CDK1 kinase can directly phosphorylate MARK2, we performed in vitro kinase assays with GST-tagged MARK2 proteins as substrates. Figure 1D shows that purified CDK1–Cyclin B1 kinase complex phosphorylated GST-MARK2 proteins in vitro. CDK2 and CDK5 were also able to phosphorylate MARK2 to a lesser extent (Fig. 1D). We also included MAPK-p38α, MEK1, ERK1, and JNK1/2 kinases in these assays as these kinases recognize the same phosphorylation sequence consensus as CDK1 (Fig. 1D). Little MARK2 phosphorylation was detected in the presence of these kinases (Fig. 1D).

CDK1 phosphorylates substrates at a minimal proline-directed consensus sequence (51). Database analysis (https://www.phosphosite.org) identified three Serine-Prolines (S456, S569, and S619) in MARK2 as possible CDK1 phosphorylation sites during mitosis (Fig. 1E). RO3306 treatment or mutating all these three sites to alanines completely blocked ³²P incorporation in vitro (Fig. 1F), suggesting that these sites are the main phosphorylation sites for CDK1. We next generated phospho-specific antibodies against these sites. CDK1-Cyclin B1 complex significantly increased phosphorylation of S456, S569, and S619 in GST-MARK2 in vitro, and the signals were abolished when the non-phosphorylatable mutant (3A: S456A/S569A/S619A) was used (Fig. 1G). Together, these data indicate that CDK1 phosphorylates MARK2 at S456, S569, and S619 in vitro.

CDK1 phosphorylates MARK2 in cells during mitosis

After confirming MARK2 phosphorylation by CDK1 in vitro, we next examined this phosphorylation in cells. Nocodazole or Taxol treatment significantly increased the phosphorylation of S569 and S619 in endogenous MARK2 (Fig. 1H, I). The signal was abolished by siRNA-mediated knockdown (Fig. 1I). Phosphopeptide-, but not regular non-phosphopeptide-, incubation completely blocked the phospho-signal of transfected MARK2, suggesting that these antibodies detect the phosphorylated form of MARK2 (Fig. 1J, K). Mutating all three serines to alanies also greatly diminished the phosphorylation of MARK2 (Fig. S2C), confirming the specificity of these phospho-antibodies.

Treatment with RO3306 or purvalanol A blocked MARK2 phosphorylation at S569 and S619 induced by Taxol, suggesting that phosphorylation of MARK2 is CDK1 kinase-dependent in cells (Fig. 1L, lanes 3 and 4). Furthermore, knockdown of CDK1 inhibited MARK2 phosphorylation in cells (Fig. 1M), and enhanced expression of hyperactive CDK1 or Cyclin B1 greatly increased MARK2 phosphorylation (Fig. 1N). Kinase activity of MARK2 is not required for its mitotic phosphorylation since MARK2-KD (kinase dead) was also phosphorylated at S569 and S619 during mitotic arrest (Fig. S2D). The non-phosphorylatable mutant (MARK2-3A) possesses similar T208 phosphorylation when compared with wild type MARK2, suggesting that mitotic phosphorylation of MARK2 does not impact its autophosphorylation in the activation loop (Fig. S2E). Together, these observations indicate that MARK2 is phosphorylated at S569 and S619 in cells during antitubulin drug treatment.

MARK2 regulates mitotic progression in a CDK1 phosphorylation-dependent manner

We next explored the possible role of MARK2 in regulating mitotic processes. Deletion of MARK2 (MARK2-KO) in U2OS osteosarcoma cells did not affect other MARK family members’ expression levels (Fig. S2F). We monitored mitotic progression in parental and MARK2-KO cells stably expressing RFP-H2B by utilizing fluorescence live-cell imaging. The parental U2OS cells condensed their chromatin and aligned their chromosomes in a tightly packed metaphase plate within 20 min after the nuclear envelope breaks down (NEBD). Anaphase onset occurred at approximately 52 min, followed by telophase, measured by chromatin decondensation, occurring at 60 min post-NEBD (Fig. 2A). In contrast, MARK2-KO cells showed a dramatic delay in chromosome alignment at metaphase for approximately 100 min (Fig. 2B, C). Accordingly the mitotic length was significantly increased in MARK2-KO cells compared with parental cells (Fig. S2G). These data suggest that MARK2 is required for metaphase-anaphase transition. Similar findings were observed in HeLa cells with MARK2 knockdown (Fig. 2D and Fig. S2H).

Figure 2. Phosphorylation of MARK2 is essential for precise mitosis.

(A, B) Inhibition of MARK2 delays mitotic progression. Live-cell images of U2OS-RFP-H2B or U2OS-RFP-H2B-MARK2-KO (knockout) cells entering and exiting mitosis.

(C, D) Quantification of mitotic length in HeLa and U2OS cells. Data were based on 54 (U2OS) or 49 (HeLa) mitotic cells for control and 45 (U2OS) or 46 (HeLa) for MARK2 knockout or knockdown cells. **: p<0.01, ***: p<0.001 (Student’s t-test).

(E) Establishment of U2OS MARK2-KO or HeLa MARK2-KD cells expressing Flag-MARK2-WT or Flag-MARK2-3A (S456A/S569A/S619A).

(F, G) Inhibition of MARK2 causes mitotic defects. Quantification of abnormal mitosis from live-cell imaging of HeLa cells expressing RFP-H2B. Data were expressed as mean ± SEM from three independent experiments (F). *: p=0.02, **: p<0.003, ***: p=0.001 (Student’s t-test). Representative images of chromosome misalignment and chromosome missegregation in MARK2-KD cells (G).

(H) Phosphorylation of MARK2 is required for normal mitosis. Quantification of mitotic defects from the indicated cell lines expressing RFP-H2B. Data were expressed as mean ± SEM from three independent experiments. Total cell counted: 37, 110, 57, 71 for control, KD, KD+WT, KD+3A, respectively. *: p=0.03, **: p=0.003 (knockdown vs control); p=0.005 (addback of WT vs knockdown) (Student’s t-test).

(I, J) Phosphorylation of MARK2 is essential for mitotic progression. Quantification of mitotic length in RFP-H2B-expressing HeLa cells. NEBD: nuclear envelope breakdown. ***: p=0.0003 or 0.0001 (I); p=0.0001 (J) (Student’s t-test). Total cell counted: 63, 61, 57, 63 for control, KD, KD+WT, KD+3A, respectively.

The necessity of MARK2 for proper prophase-to-anaphase progression prompted us to explore the role of CDK1-mediated MARK2 phosphorylation events. We generated stable cell lines expressing either MARK2-WT or MARK2-3A in the MARK2-KO U2OS or MARK2-knockdown HeLa cells (Fig. 2E). Inhibition of MARK2 resulted in mitotic defects including chromosome misalignment and missegregation (Fig. 2F, G). A significantly higher percentage of cells with MARK2 knockdown contain multiple nuclei (Fig. 2F). Importantly, addback of wild type MARK2 completely rescued the mitotic abnormality and morphology seen in MARK2-KO and knockdown cells (Fig. 2H). However, reexpression of the non-phosphorylatable mutant MARK2 (MARK2-3A) failed to restore the defects (Fig. 2H). Accordingly, MARK2, but not the MARK2-3A mutant, addback largely restored the mitotic delay to normal (Fig. 2I, J). These observations suggest that CDK1-mediated phosphorylation of MARK2 is critical for proper mitotic progression.

MARK2 regulates chemosensitivity in PDAC

To determine the biological significance of MARK2, we analyzed MARK2 expression in various types of cancer in public datasets and the most significant change (cancer vs. normal) of MARK2 signaling was found in PDAC (Fig. 3A and Fig. S2I). Importantly, high expression of MARK2 strongly correlates with poor survival in PDAC patients (Fig. 3B). MARK2 protein levels were also overexpressed in most pancreatic cancer cell lines compared with the non-cancerous immortalized human pancreatic nestin-expressing (HPNE) cells (Fig. 3C).

Figure 3. MARK2 inhibition promotes chemosensitivity in human and mouse PDAC cells.

(A, B) Clinical relevance of MARK2 in PDAC. MARK2 mRNA expression levels are upregulated in PDAC patient tumor tissue (n=179) compared with normal tissue (n=171) (A). MARK2 expression is positively correlated with overall survival rate in pancreatic cancer patients (p=0.0051) (B). Data were generated by an online software using TCGA datasets (gepia2.cancer-pku.cn).

(C) MARK2 protein expression in HPNE (non-cancerous pancreatic cell line) and human pancreatic cancer cell lines.

(D) Knockdown of MARK2 increased Taxol-induced apoptosis. Cells were treated with DMSO or Taxol (1 μM for 24 h). Total cell lysates were probed with the indicated antibodies. Cl-PARP: Cleaved PARP.

(E, F) MARK2 promoted cell survival under Taxol treatment in clonogenic assays. Cells were treated with DMSO or Taxol for 24 h as indicated and colonies were quantified after 14 days. Data were from three independent experiments. *: p=0.034, ***: p=0.006 (Student’s t-test).

(G) Phosphorylation is required for MARK2-driven resistance to Taxol in PDAC cells. MARK2-KO cells were reexpressed with Flag-MARK2 or Flag-MARK2-3A (S456A/S569A/S619A) and cells were treated with DMSO or Taxol (1 μM for 24 h).

(H) Enhanced expression of MARK2 promoted Taxol resistance in Colo357 cells. Cells expressing empty vector, MARK2-WT or MARK2-3A (S456A/S569A/S619A) were treated with DMSO or Taxol (200 nM for 48 h) and total lysates were probed with the indicated antibodies.

(I, J) Knockdown of MARK2 increased Taxol-induced apoptosis in mouse PDAC cells. Caspase 3/7 assay from mouse KC6141 cells treated with Taxol (10 nM for 16 h) (n=3) (I). **: p=0.007 (Student’s t-test).

(K, L) MARK2 inhibition enhanced gemcitabine cytotoxicity in human (K) and mouse (L) PDAC cells. Control and MAKR2-knockdown cell lines were treated with DMSO or gemcitabine (500 nM for 48 h for S2.013, 10 μM for 72 h for PANC-1, and 50 nM for 16 h for KC6141). Total cell lysates were probed with the indicated antibodies. Cl-Casp3: Cleaved caspase 3.

(M-O) MARK2 inhibition sensitized PDAC cells to chemotherapy. KC6141-shControl and -shMark2 cells were subcutaneously inoculated into C57BL/6 mice (both left and right sides). Drug treatment started at day 4 post-injection. Paclitaxel (12 mg/kg) and gemcitabine (50 mg/kg) were administered on every other day via intraperitoneal injection. The representative tumors in each group were excised and photographed at the endpoint (M). The shCtrl and shMark2 groups are identical in panels N and O. The P values are 0.002 (day 8), 0.001 (day 12), 0.001 (day 16), 0.0001 (day 20), 0.0001 (day 24) (N). ***: p=0.0001 (O).

To directly examine the role of MARK2 in PDAC cells, we knocked down MARK2 (MARK2-KD) in various cell lines (Fig. 3D). Significant changes were observed when cells were exposed to the antitubulin agent Taxol, which has been used together with gemcitabine as first-line chemotherapy for metastatic pancreatic cancer. PDAC cells (PANC-1, Capan-2, and S2.013) are relatively resistant to Taxol treatment (Fig. 3D). MARK2 knockdown greatly sensitized PDAC cells to Taxol-induced cell death (revealed by cleaved PARP) when compared to control cells (Fig. 3D). We further performed clonogenic assays to examine the effect of MARK2 on the survival rate in response to Taxol treatment. As shown in Figure 3E and 3F, knockdown of MARK2 significantly impaired cell survival after Taxol treatment compared with control cells.

We next determined whether antitubulin drug-induced phosphorylation is involved in regulating Taxol cytotoxicity in PDAC cells. Re-expression of MARK2, but not the MARK2-3A mutant, completely blocked cell death induced by Taxol in MARK2-knockdown cells, suggesting that CDK1 phosphorylation is essential for MARK2-driven Taxol chemosensitivity (Fig. 3G). Consistently, enhanced expression of wild type MARK2, but not the MARK2-3A mutant, promoted resistance to Taxol cytotoxicity in colo357 cells (Fig. 3H). In addition, knockdown of Mark2 in mouse PDAC cells (KC6141) also enhanced Taxol cytotoxicity (Fig. 3I, J). In the clinic, gemcitabine is used along with Nab-Taxol. Interestingly, knockdown of MARK2/Mark2 synergistically induced cell death with gemcitabine in both human and mouse PDAC cells (Fig. 3K, L). Next, control and Mark2-KD KC6141 cells were subcutaneously inoculated into immune-competent C57BL/6 mice and the mice were treated by intraperitoneal injection of PBS (vehicle control), Taxol or gemcitabine. No significant differences were observed in the volumes of tumors from control and MARK2-KD cells (Fig. 3M-O). Paclitaxel or gemcitabine treatment in control cells did not suppress tumor growth at the doses we used (Fig. 3M-O). Consistent with our observations from in vitro models (Fig. 4J, L), Mark2-KD cells formed significantly smaller tumors than the control cells under Taxol or gemcitabine treatment, suggesting that inhibition of Mark2 sensitizes PDAC cells to Taxol or gemcitabine cytotoxicity (Fig. 3M-O). Together, our studies have elucidated a novel regulatory mechanism and function of MARK2 in response to antitubulin drugs and indicate that MARK2 is a potent and novel regulator of chemoresistance in PDAC cells.

Figure 4. MARK2 promotes YAP activity by phosphorylating class IIa HDACs.

(A, B) Class IIa HDACs are phosphorylated during mitosis in a MARK2-dependent manner. HeLa cells (shCtrl and shMARK2) were treated with DMSO, nocodazole (Noco) or Taxol and total cell lysates were probed with the indicated antibodies (A). Phosphorylation of HDAC4 S246 is blocked upon MARK2 knockdown in human PDAC cells (B).

(C) Phosphorylation of HDAC4 S246 occurs during unperturbed mitosis. HeLa cells were synchronized by a double thymidine block and release method. Cells were stained with antibodies against p-HDAC4 S246 (red) together with α-tubulin (microtubule) and DAPI (DNA). A 20× objective lens was used to view various phases of the cells in a field.

(D) MARK2 directly phosphorylates HDAC4 at S246 and S632. GST-tagged HDAC4 or HDAC4-3A (S246A/S467A/S632A) (amino acids 200-680) proteins were used for in vitro kinase assays with purified GST-MARK2 kinase (SignalChem).

(E) HEK293T cells were transfected as indicated. At 48 h post-transfection, total cell lysates were subjected to Western blotting with the indicated antibodies. 3A: S456A/S569A/S619A. KD: kinase-dead (T208A/S212A).

(F-H) Phosphorylation of HDAC4 is essential for mitotic progression. Quantification of mitotic length in RFP-H2B-expressing HeLa cells. NEBD: nuclear envelope breakdown. **: p=0.002 or 0.005 (G); **: p=0.006 (H); *: p=0.03 (H) (Student’s t-test). Total cell counted: 68, 64, 64, 70 for control, KD, KD+WT, KD+3A, respectively.

(I, J) HDAC4 promotes YAP activation in a phosphorylation-dependent manner. (I) HEK293T cells were transfected as indicated. At 48 h post-transfection, total cell lysates were subjected to Western blotting with the indicated antibodies. 3A: S246A/S467A/S632A. (J) Luciferase reporter assays in HEK293T cells (n=3). **: p=0.004; ***: p=0.0009 (Student’s t-test).

(K-O) HDAC inhibition blocks YAP targets expression induced by Taxol. HeLa cells were treated as indicated. T: Taxol (100 nM for 16 h); S: SAHA (HDAC inhibitor, 500 nM for 48h). qRT-PCRs were performed to measure Zyxin, Survivin and LATS2 mRNA levels (K-M). Data were expressed as the mean ± SEM of three independent experiments. **: p=0.009 and p=0.006 (K); *: p=0.011 and **: p=0.009 (L); **: p=0.003 and p=0.004 (M); (Student’s t-test). Knockdown of HDAC4 or inhibition of HDACs inhibits Zyxin, Survivin, and LATS2 protein expression in PDAC cells (N, O).

(P, Q) HEK293T cells were transfected as indicated. LATS2 or YAP proteins were immunoprecipitated and the samples were probed with the indicated antibodies.

MARK2 phosphorylates MST2 and positively regulates YAP activity

Having established the role and regulation of MARK2 in Taxol chemotherapy, we next explored the downstream effector and mechanism of MARK2 in response to Taxol treatment. The Hippo (MST1/2-LATS1/2) kinase cascade phosphorylates primarily on the S127 site and inactivates YAP by sequestering it in the cytoplasm (52, 53). The transcriptional co-activator YAP binds to and functions through TEA-domain containing proteins (TEAD) transcription factors to regulate downstream targets including CTGF (the gene encoding connective tissue growth factor), Cyr61, LATS2, and Survivin (52, 54). YAP, Survivin, CTGF, LATS2, and Cyr61 are also critical regulators for antitubulin chemosensitivity (55-58). Consistent with previous studies in Drosophila (28), we found that MARK2 deletion greatly increased YAP cytoplasmic localization and p-S127 YAP levels (Fig. S3A-C), suggesting that MARK2 is a positive regulator of YAP. We and others showed that some of the YAP targets (Survivin, Zyxin, LATS2) were induced during antitubulin treatment (55, 58, 59). Importantly, we showed that Taxol-induced Survivin, LATS2, and Zyxin were largely blocked upon MARK2 deletion (Fig. S3D).

MARK2 interacts with MST1/2 and WW45 and reduces their inhibitory effects towards YAP activity (20, 28). The precise mechanism (including the phosphorylation sites on MST1/2/WW45) has not been defined. We found that MARK2 enhanced the mobility upshift of MST2 on a Phos-tag gel, suggesting that MST2 (not MST1) is phosphorylated upon MARK2 activation (Fig. S3E, F). CDK1 phosphorylation of MARK2 is not required for promoting MST2 shift/phosphorylation (Fig. S3E). Our in vitro kinase assay revealed MARK2 indeed phosphorylated MST2 (Fig. S3G). Sequence analysis identified a highly conserved site S15 on MST2 (not on MST1) as the MARK2 site. We generated a phospho-specific antibody against S15 and confirmed that MARK2 phosphorylated MST2 S15 in vitro (Fig. S3H). Enhanced expression of MARK2 stimulated S15 phosphorylation in cells in a mitotic phosphorylation-independent manner (Fig. S3I). Interestingly, increased phosphorylation at S15 was correlated with reduced MST2 kinase activity (measured by the p-T180 autophosphorylation) (Fig. S3I). In line with this observation, the non-phosphorylatable mutant MST2-S15A posseses enhanced kinase activity compared with wild type MST2 (Fig. S3J). Neither MST2 kinase activity (p-T180) or S15 was induced by antitubulin agents (Fig. S3J) (44). These observations suggest that MARK2 promotes YAP activation by phosphorylating MST2 at S15 and inactivating it. However, this mechanism is unlikely to be regulated by antitubulin drugs or in mitosis.

HDAC4 is phosphorylated by MARK2 in response to antitubulin treatment and in mitosis

Our findings suggest that MARK2 regulates YAP activity in response to antitubulin drugs/mitosis in a Hippo kinase/MST-independent manner. In addition to MST/WW45, MARK family proteins also phosphorylate MAP2/4 (15), HDACs (34), CLASP1/2, APC, and DIXDC1 (18). Class IIa HDACs (HDAC4/5/7/9) are phosphorylated by MARK2 at S246, S467, and S632 (numbering in HDAC4) and phosphorylation promotes HDACs’ cytoplasmic retention and inhibits their deacetylation/transcription repression activity (34-36). Interestingly, p-S246 levels were greatly increased during antitubulin treatment and these increases were largely blocked upon MARK2 deletion or knockdown in various cell lines (Fig. 4A, B). Increased phosphorylation at HDAC4 S246 was also evident in normal mitotic cells (Fig. 4C). MARK2 directly phosphorylates HDAC4 at S246 and S632 in vitro (Fig. 4D). Enhanced expression of MARK2 increased phosphorylation at S246 and the non-phosphorylatable MARK2-3A mutant was less active in promoting HDAC4 phosphorylation (Fig. 4E). Knockdown of HDAC4, like MARK2 inhibition, significantly delayed metaphase to anaphase transition and prolonged mitosis (Fig. 4F-H). Importantly, re-expression of wild type HDAC4, but not the non-phosphorylatable mutant HDAC4-3A (S246A/S467A/S619A) restored the mitotic defects in the knockdown cells (Fig. 4G, H). Together these studies identified class IIa HDACs as mitotic substrates for MARK2 and MARK2-meditated phosphorylation plays a vital role in mitotic progression.

HDAC4 promotes YAP activity in a mitotic phosphorylation-dependent manner

Given the connection between MARK2 and YAP/HDAC4, we hypothesized that HDAC4 also regulates YAP activity. Interestingly, exogenous HDAC4 significantly suppressed YAP S127 phosphorylation and enhanced expression of HDAC4-3A had no effect on S127 phosphorylation, suggesting that HDAC4 promotes YAP activity in a mitotic phosphorylation-dependent manner (Fig. 4I). In contrast, YAP S397 (which mediates YAP protein degradation) levels were not altered (Fig. 4I). Consistent with these observations, HDAC4, but not the HDAC4-3A mutant, stimulated YAP/TEAD transcription activity (Fig. 4J). Importantly, the HDAC inhibitor SAHA greatly inhibited the expression (as shown by reduction in both mRNA and protein levels) of YAP targets Survivin, LATS2, and Zyxin induced by Taxol treatment in PDAC cells (Fig. 4K-N). Knockdown of HDAC4 in S2.013 cells was sufficient to block the expression of LATS2, Survivin, and Zyxin induced by Taxol treatment (Fig. 4O). These data identified class IIa HDACs as positive regulators for YAP and suggest that HDAC-YAP controls a unique transcriptional program specific to the response to antitubulin treatment in PDAC cells.

We further explored the underlying mechanism through which HDAC4 regulates YAP. Consistent with the findings in Figure 4I, J, we found that the HDAC4-3A mutant had increased binding affinity with LATS2 when compared to wild type HDAC4 (Fig. 4P). In contrast, non-phosphorylatable HDAC4 (HDAC4-3A) reduced its association with YAP (Fig. 4Q). These observations suggest that HDAC4 promotes YAP activity by modulating the LATS2/YAP complex.

MARK2 regulates Taxol chemosensitivity by phosphorylating HDAC4 in PDAC cells

Alterations of HDAC4 have been implicated in several disease conditions including cancer (60-62). HDAC inhibitors have been shown to be promising in clinical settings and preclinical models of cancer treatment (61, 63). However, the functional significance, especially in Taxol chemosensitivity, of class IIa HDACs (HDAC4/5/7/9) in PDAC is largely unclear. Class IIa HDAC mRNAs are all overexpressed in PDAC compared with normal pancreatic tissue (Fig. S4A). Furthermore, HDAC4 and HDAC7 protein levels were also increased in most PDAC lines compared with the non-cancerous human pancreatic cells (HPNE) (Fig. 5A). HDAC5 and HDAC9 protein expression was extremely low in most cell lines. For other classes of HDACs (HDAC1/2/6), expression was not significantly upregulated in PDAC cell lines (Fig. 5A). Furthemore, HDAC4 and HDAC7 expression levels are positively correlated with MARK2 and YAP in PDAC patients (Fig. 5B, C). Knockdown of HDAC4 or HDAC7 in PANC-1 and S2.013 cells did not impair cell proliferaton nor cause apoptosis (Fig. 5D). Interestingly, knockdown of HDAC4 or HDAC7 greatly sensitized these cells to Taxol-induced cytotoxicity, as revealed by cleaved PARP (Fig. 5E, F). Consistent with these studies, combined SAHA or panobinostat (LBH-589) (pan-HDAC inhibitors) with Taxol treatment greatly induced cell death, while single agent treatment failed to promote significant apoptosis (Fig. 5G). Clonogenic assays confirmed that HDAC4 or HDAC7 knockdown significantly impaired cell survival after Taxol treatment compared with control cells (Fig. 5H, I). These observations suggest that HDAC4/7, like MARK2, are critical determinants for Taxol chemosensitivity in PDAC cells.

Figure 5. HDAC4/7 control chemosensitivity in human and mouse PDAC cells.

(A) Protein expression of HDACs and YAP HPNE (immortalized human pancreatic cells) and human pancreatic cancer cell lines. HDAC1/2, HDAC4/7, HDAC6 are class I, IIa, IIb HDACs, respectively. Class IIa HDAC5 and HDAC9 proteins were not detectable in most cell lines.

(B, C) MARK2, HDAC4/7, and YAP expressions are positively correlated in PDAC patients. Data were generated from an online analysis tool (gepia2.cancer-pku.cn) using TCGA database.

(D) Knockdown of HDAC4 did not affect the proliferation rate of S2.013 cells.

(E, F) Human PDAC cell lines with or without HDAC4 knockdown were treated with or without Taxol (1 μM for 24 h). Total cell lysates were probed with the indicated antibodies. Cl-PARP: cleaved PARP.

(G) Human PDAC cell lines were treated with Taxol in the presence/absence of pan-HDAC inhibitors for 24 h. LBH589 was used at 300 nM for S2.013, 200 nM for PANC-1, and 10 nM for Capan-2. SAHA was used at 3 μM for S2.013, 5 μM for PANC-1, and 2.5 μM for Capan-2.

(H, I) Knockdown of HDAC4 or HDAC7 impairs survival in clonogenic assays. Cells were treated with Taxol (50 nM, 100 nM for 24 h) and colonies were quantified from three independent experiments. ***: p<0.001, **: p=0.001 (Student’s t-test) (H). **: p=0.003, **: p=0.006 (Student’s t-test) (I).

(J, K) Phosphorylation is required for HDAC4-driven resistance to Taxol in PDAC cells. Establishment of HDAC4-knockdown cells expressing wild-type HDAC4 or HDAC-3A (S246A/S467A/S632A) (J). Cells from J were treated with DMSO or Taxol (1 μM for 24 h).

(L, M) HDAC4/7 knockdown enhanced Gemcitabine cytotoxicity in human PDAC cells. Control and HDAC4- or HDAC7-knockdown cell lines were treated with DMSO or gemcitabine (GEM, 500 nM for 48 h in S2.013 and 10 μM for 72 h in PANC-1).

(N, O) MARK2 regulates Taxol chemosensitivity through HDAC4. Establishment of MARK2-knockdown cells expressing HDAC4-3D (S246D/S467D/S632D) or HDAC4-3A (N). Cells were treated with DMSO or Taxol (1 μM for 24 h). Total cell lysates were probed with the indicated antibodies.

To examine the role of MARK2-mediated phosphorylation of HDAC4, we reexpressed HDAC4-WT or −3A in HDAC4 knockdown cells (Fig. 5J). As expected, addback of HDAC4-WT completely blocked the cell death induced by Taxol and reexpression of HDAC4-3A failed to do so (Fig. 5K), suggesting that MARK2 phosphorylation of HDAC4 promotes Taxol chemoresistance. Interestingly, knockdown of HDAC4 or HDAC7 also potentiated gemcitabine cytotoxicity (Fig. 5L, M). To further explore functional and genetic interactions between MARK2 and HDAC4, we ectopically expressed HDAC4-3D (phosphomimetic mutant) or −3A in MARK2 knockdown cells (Fig. 5N). MARK2 knockdown synergized cell apoptosis with Taxol as expected, and expression of HDAC4-3D, but not HDAC4-3A, largely blocked apoptosis in these cells (Fig. 5O), suggesting that MARK2 regulates Taxol chemosensitivity through phosphorylating HDAC4.

HDAC inhibition synergizes Taxol chemotherapy in PDAC in vivo

We found that HDAC7, but not HDAC1, HDAC2, HDAC4, and HDAC6, is upregulated in mouse KPC (concomitant expression of K-Ras^G12D and p53^R172H)/KC cells compared with normal mouse pancreas (Fig. 6A). PDAC cell lines derived from KPC and KC mice mimic the genetic compendium of human PDAC and importantly form both subcutaneous and orthotopic tumors in mice with intact immune systems (64). In line with the observations in human PDAC cells (Fig. 5), knockdown of mouse HDAC7 was sufficient to sensitize mouse PDAC cells (KC6141) significantly to Taxol or gemcitabine-induced apoptosis (Fig. 6B, C). To further confirm whether knockdown of HDAC7 is sufficient to potentiate Taxol and gemcitabine efficacy in vivo, KC6141 (shControl and shHDAC7) cells were subcutaneously inoculated into immune-competent C57BL/6 mice followed by PBS, Taxol or gemcitabine treatment. Knockdown of HDAC7, or treatment with Taxol or gemcitabine alone, did not affect KC6141 tumor growth (Fig. 6D, E). However, knockdown of HDAC7 in combination with Taxol or gemcitabine treatment greatly inhibited tumor growth (Fig. 6D, E). Next, we implanted KPC961 and KC6141 cells in mice followed by treatment with PBS, Taxol, SAHA, or Taxol combined with SAHA treatment. No significant differences were detected for the tumor volumes among PBS, Taxol, or SAHA treatment groups of mice bearing either KPC961 or KC6141 tumors (Fig. 6F-I). Paclitaxel and SAHA combination treatment significantly suppressed tumor growth compared with the rest of the groups (Fig. 6F-I). We also labeled KC6141 cells with luciferase prior to orthotopic implantation in animals. Again, SAHA+Taxol treatment substantially inhibited tumor growth (Fig. 6J-L). We observed metastasis in 4 of 6 control mice and no metastasis in all drug-treated mice (Fig. 6M). As expected, massive apoptosis was detected in SAHA/Taxol treated tumors (Fig. 6N). No overall toxicity was observed in animals (which had normal body weight and food intake), suggesting that these drugs were well tolerated in these animals. Furthermore, SAHA and paclitaxel treatment synergistically suppressed PDAC organoids growth (Fig. 6O, P). These observations suggest that HDAC inhibition overcomes Taxol chemoresistance in immune-intact animals and PDAC tumors. Altogether, our study suggests that MARK2 promotes Taxol resistance by phosphorylating HDAC4/7 and phosphorylated class IIa HDACs in turn stimulate YAP activity (Fig. 6Q).

Figure 6. Inhibition of HDAC promotes chemosensitivity in vivo.

(A) Protein expression of HDACs in normal mouse pancreas and mouse pancreatic cancer cell lines. HDAC1/2, HDAC4/7, HDAC6 are class I, IIa, IIb HDACs, respectively. Class IIa HDAC5 and HDAC9 protein levels were extremely low/not detectable in mouse pancreatic tissue and cell lines.

(B-E) Knockdown of HDAC7 enhances Taxol and gemcitabine cytotoxicity in mouse pancreatic cancer cell line KC6141. Cell lines were treated with DMSO, Taxol (10 nM for 16h) or gemcitabine (GEM, 50 nM for 16h). Cl-Casp3: cleaved caspase 3; Cl-PARP: cleaved PARP (B, C). Knockdown of HDAC7 synergizes with Taxol (D) or gemcitabine (E) chemotherapy in vivo. KC6141 cells were subcutaneously inoculated into C57BL/6 mice. Drug treatment started at day 4 post-injection. Paclitaxel (8 mg/kg) and gemcitabine (50 mg/kg) were used at every other day via intraperitoneal injection. The P values are 0.01 (day 8), 0.002 (day 12), 0.004 (day 16), 0.002 (day 20), 0.005 (day 24) (D); 0.0004 (day 16), 0.0003 (day 20), 0.0004 (day 24) (E). The P values are shown for comparing shCtrl+Taxol with shHDAC7+Taxol or shCtrl+GEM with shHDAC7+GEM groups using one-way ANOVA, Tukey’s test.

(F-I) Pan-HDAC inhibitor SAHA synergizes with Taxol treatment in vivo. KC6141 (F, G) and KPC961 (H, I) cells were subcutaneously inoculated into C57BL/6 mice. Animals were randomized (4 or 5 days post-injection) and treated with PBS, Paclitaxel (8 mg/kg), SAHA (25 mg/kg) or the combination. The P values are 0.02 (day 11), 0.002 (day 13), 0.004 (day 15) (G); 0.02 (day 16), 0.01 (day 20), 0.001 (day 22) (I). The P values are shown for comparing SAHA with SAHA+Taxol groups using one-way ANOVA, Tukey’s test.

(J-M) Pan-HDAC inhibitor SAHA synergizes with Taxol treatment in vivo (orthotopic model) (J, K). KC6141-luciferase cells were implanted into the head of the mouse pancreas and treatment was initiated at 7 days post-injection. The representative animals in each group at week 1 and week 5 were shown (J). At the end of week 5, tumors (primary and metastatic) were observed, removed and weighed (L, M). SAHA+Taxol treatment induced apoptosis (N). The P values are 0.003 (week 2), 0.009 (week 3), 0.0006 (week 4), 0.005 (week 5) (K); **: p=0.003 (L) (Student’s t-test).

(O, P) HDAC inhibitor SAHA synergizes with Taxol treatment in suppressing PDAC organoids growth. The PDAC organoids was treated as indicated for 72 h and representative images were shown (P). Ctrl: control (DMSO); T: Taxol (10 nM); S1: SAHA 10 nM; S2: SAHA 20 nM. **: p=0.002 (Student’s t-test). The combination indexes (CI) are 0.098 and 0.14 for Taxol 10 nM+SAHA 10 nM and Taxol 10 nM+SAHA 20 nM, respectively. A CI<0.25 is considered to indicate synergism for the two drugs involved in a treatment.

(Q) A proposed model. In response to Taxol treatment, MARK2 is phosphorylated. Phosphorylated MARK2 stimulates class IIa HDAC phosphorylation. Phosphorylated HDAC4 preferently binds to YAP and stimulates YAP transcriptional activity.

MARK2-HDAC inhibition induces mitotic defects to potentiate Taxol cytotoxicity in PDAC cells

Previous studies showed that in patient tumors, the mitotic index and the actual concentration of Taxol are much lower compared to those in cell culture (65, 66). Low concentrations of Taxol are sufficient to induce cell death due to chromosome missegregation on multipolar spindles (66). We wanted to explore whether inhibition of the MARK2-HDAC axis is sufficient to cause mitotic abnormality and subsequent cell death under the low concentrations of Taxol treatment. To test this hypothesis, we treated PDAC cells that have knockdown of MARK2 or HDAC4 or HDAC7 with 5 nM Taxol and monitored mitotic abnormality by visualizing the microtubules and chromosomes. Treatment with 5 nM Taxol or MARK2/HDAC4 knockdown induced modest mitotic defects, including lagging chromosomes at metaphase and chromosome misalignment in PANC-1 cells (Fig. 7A-E). Importantly, a significantly higher percentage of cells with mitotic defects (chromosome missegregation and multipolar spindles) was observed in MARK2, HDAC4, or HDAC7 knockdown PANC-1 cells with Taxol treatment (Fig. 7A-F). Consistent with these observations, combined treatment of HDAC inhibitor (SAHA or LBH589) with Taxol also resulted in significant mitotic defects (Fig. 7A-F). Similar findings were observed in S2.013 cells (Fig. 7G; Fig. S4B-E). In line with these findings, we further confirmed that inhibition of MARK2, HDAC4, or HDAC7 induced cell death even under low concentrations of Taxol treatment (Fig. 7H-J). Together, our data suggest a link between Taxol cytotoxicity and mitotic defects induced by MARK2-HDAC inhibition in PDAC cells.

Figure 7. MARK2-HDAC inhibition synergizes with Taxol treatment to induce mitotic defects in PDAC cells.

(A-E) Mitotic defects were quantified in PANC-1 cells under various treatments. Low concentrations of Taxol (5 nM) were used in all experiments. Under these conditions, cells proliferate without arrest. SAHA and LBH589 were used at 5 μM and 0.2 μM, respectively. Experiments were done at 24 h post-treatment. N=3 from 40-60 mitotic cells of each condition. P values on each column indicate the statistical analysis from comparisons with Taxol-treated cells (Student’s t-test).

(F) Representative confocal microscopy images of mitotic defects in PANC-1 cells.

(G) Mitotic defects were quantified in S2.013 cells. Treatments were the same as in PANC-1 cells.

(H-J) MARK2-HDAC inhibition synergizes with Taxol (low concentration) treatment to induce apoptosis in PANC-1 cells. Cells were treated with DMSO, Taxol (5 nM for 0 h, 24 h, 72 h and 120 h). Total cell lysates were probed with the indicated antibodies. C-PARP: cleaved PARP.

Discussion

MARK family members are protein kinases on which relatively few studies have been conducted, and their biological function, especially in cancer, is somewhat unclear. Emerging evidence showed that they control cell growth through the Hippo-YAP signaling pathway, although whether they promote or inhibit growth remains controversial. For example, MARK1/3/4, but not MARK2, have been identified as negative regulators of YAP oncogenic activity by promoting SCRIB association with MST1/2 and LATS1/2 (27). This observation is consistent with the fact that MARKs are phosphorylated and activated by tumor suppressor LKB1. In line with these studies, MARK1/4 have been shown to phosphorylate DIXDC1 to suppress cell invasion and metastasis (18). In contrast, several other studies showed that MARK1/4 and MARK3 promote YAP activation by inhibiting Hippo kinase signaling (20, 26, 28). Our data support a positive role of MARK2 in the regulation of YAP activity, likely through phosphorylating and inactivating MST2 (Fig. S3). Furthermore, MARK2 is overexpressed in human PDAC and negatively correlates with PDAC patients’ prognosis and its expression is positively correlated with YAP (Figs. 3A, B; 5B), supporting a protumorigenic function of MARK2 in PDAC. MARKs can be activated by upstream kinases via phosphorylation of the catalytic kinase domain, and numerous studies have identified many activation regulators of the MARK family, including MARKK/TAO-1, LKB1, CAMK1, GSK-3β, and αPKC (14). The current study identifies CDK1 as an upstream kinase that phosphorylates MARK2 specifically in mitosis, adding a new layer of regulation for MARK2. Differing from other kinase-mediated phosphorylation events, mitotic phosphorylation does not influence its kinase activity (T208 phosphorylation) (Fig. S2D, E) and yet still plays a critical role (Figs. 2, 3). Thus, future studies are needed to investigate how CDK1-mediated phosphorylation of MARK2 regulates its role in mitotic progression and chemosensitivity. Interestingly, we found that, in addition to MARK2, MARK3 is also up-shifted upon Taxol or nocodazole treatment (Fig. 1A). We are currently investigating how MARK3 is regulated and whether MARK3 also plays a role in regulating mitosis and antitubulin chemosensitivity similar to MARK2.

The poor prognosis of pancreatic cancer patients is mainly due to drug resistance. This study identifies class IIa HDACs as mitotic substrates for MARK2. MARK2 directly phosphorylates class IIa HDACs in response to antitubulin chemotherapy, and MARK2-HDAC controls a YAP-dependent transcriptional program induced by paclitaxel treatment (Figs. 4, 6Q). These observations added class IIa HDACs as new positive regulators of YAP. Given that MARK2-HDAC controls chemosensitivity in PDAC (Figs. 3, 5, 6), these findings provide new options for targeting MARK2-HDAC activity in reversing chemoresistance in PDAC. Pan-HDAC inhibitors are approved by the FDA as anticancer agents and several studies have showed that HDAC inhibitors reverse paclitaxel resistance in non-small cell lung cancer, papillary serous endometrial cancer, and ovarian cancer (67-69). HDAC inhibitors induce synergistic cytotoxicity with paclitaxel via regulation of HDAC class I members HDAC1 and HDAC6, stabilization of microtubules, and inhibition of paclitaxel-induced Survivin accumulation (67-69). Though we could not exclude the possibility of the involvement of other HDACs in paclitaxel chemosensitivity in PDAC cells, our current data support a role of class IIa HDAC (i. e. HDAC4/7) in regulating paclitaxol cytotoxicity downstream of MARK2. Therefore, targeting class IIa HDACs seems to be a more feasible way to overcome chemoresistance in PDAC. A common limitation of HDAC inhibitors in the clinic is dose-limiting toxicity, which has necessitated dose reductions or changes in dose scheduling (70). A strategy to maximize efficacy, reducing toxicity, and resistance by administering lower drug doses is combining anticancer drugs with other chemotherapeutic agents that have synergistic or additive antitumor effects (39, 71). When relatively low doses of a pan-HDAC inhibitor were used in our animal models, it was still sufficient together with paclitaxel or gemcitabine to suppress PDAC tumor growth without significant toxicity (Fig. 6). These results indicate that the combination treatment may help pancreatic cancer patients tolerate the drug (SAHA) by lowering the drug dose. Furthermore, the development of more selective inhibitors targeting a subset of HDACs can reduce toxicity while retaining antitumor activity. Thus, based on our results, it will be interesting to develop HDAC4 and HDAC7 selective inhibitors and test them in animal models of cancer.

Many studies have shown that the development of multidrug resistance, alterations in microtubule dynamics, and altered metabolism contribute to antitubulin drug resistance (72). However, attempts to reverse resistance by targeting these mechanisms have not successfully translated to the clinic. Thus, identification of new regulators and/or signaling pathways that are triggered by antitubulin agents may shed light on resistance mechanisms and lead to the development of novel prognostic or therapeutic approaches related to antitubulin chemotherapeutics. In the present study, we aimed to probe kinases, as a quarter of all drug discovery spending is used to target kinases, using a Phos-tag-based approach. In addition to MARK2/3, our screens identified many novel kinases, such as FGFR2/4, EPHA family, JAK1, MER, PKR, and TNK1, that are regulated during antitubulin treatment (Figs. 1A; S1). Interestingly, most of these kinases are heavily involved in cancer cell growth and tumorigenesis; however, their regulation and function in response to antitubulin agents have not been defined. Therefore, elucidation of the roles of these molecules may provide additional targets for reversing antitubulin chemoresistance. Large-scale phospho-proteomic studies have identified an array of phosphorylation events, including MARK2 and PKR phosphorylation, in response to antitubulin agents (73). However, hits from our studies (such as phosphorylation for FGFR2/4, EPHA family kinases, JAK1, and MER) were not identified in these proteomic studies. Thus, our current study reveals Phos-tag-based Western blotting analysis as an alternative tool for identifying regulators in response to antitubulin agents, and it is a strategy that can be applied in other systems as well.