DYRK1B (dual-specificity tyrosine phosphorylation-regulated kinase 1B) is amplified in certain cancers and may be an oncogene; however, our knowledge of DYRK1B has been limited by the lack of selective inhibitors. In the present study we describe AZ191, a potent small molecule inhibitor that selectively inhibits DYRK1B in vitro and in cells. CCND1 (cyclin D1), a key regulator of the mammalian G1–S-phase transition, is phosphorylated on Thr286 by GSK3β (glycogen synthase kinase 3β) to promote its degradation. DYRK1B has also been proposed to promote CCND1 turnover, but was reported to phosphorylate Thr288 rather than Thr286. Using in vitro kinase assays, phospho-specific immunoblot analysis and MS in conjunction with AZ191 we now show that DYRK1B phosphorylates CCND1 at Thr286, not Thr288, in vitro and in cells. In HEK (human embryonic kidney)-293 and PANC-1 cells (which exhibit DYRK1B amplification) DYRK1B drives Thr286 phosphorylation and proteasome-dependent turnover of CCND1 and this is abolished by AZ191 or DYRK1B RNAi, but not by GSK3β inhibitors or GSK3β RNAi. DYRK1B expression causes a G1-phase cell-cycle arrest, but overexpression of CCND1 (wild-type or T286A) fails to overcome this; indeed, DYRK1B also promotes the expression of p21CIP1 (21 kDa CDK-interacting protein 1) and p27KIP1 (CDK-inhibitory protein 1). The results of the present study demonstrate for the first time that DYRK1B is a novel Thr286-CCND1 kinase that acts independently of GSK3β to promote CCND1 degradation. Furthermore, we anticipate that AZ191 may prove useful in defining further substrates and biological functions of DYRK1B.
The DYRKs (dual-specificity tyrosine phosphorylation-regulated kinases) are an evolutionarily conserved family of protein kinases that are found within the CMGC group [CDK (cyclin-dependent kinase), MAPK (mitogen-activated protein kinase), GSK3 (glycogen synthase kinase 3) and CLK (CDC-like kinase)] of the eukaryote kinome . DYRKs contain a characteristic tyrosine-X-tyrosine (Y-X-Y) residue motif in their activation loop. Phosphorylation of the second tyrosine residue is critical for kinase activity  and occurs during translation when DYRKs autophosphorylate this tyrosine residue in cis ; however, the mature DYRKs phosphorylate substrates exclusively on serine or threonine residues . Mammalian DYRKs fall into two subgroups, class I (DYRK1A and DYRK1B) and class II (DYRK2, DYRK3 and DYRK4), and have roles in regulating transcription, cell-cycle progression, differentiation and survival. DYRK1B is particularly abundant in the testis and muscle where it is thought to play a key role in myoblast differentiation [4–6]. DYRK1B is amplified or overexpressed in certain cancers suggesting that it may be an oncogenic driver ; however, our knowledge of DYRK1B, its substrates and its role has been limited by the lack of potent and selective inhibitors. For example, whereas the DYRK inhibitor harmine exhibits selectivity for class I over class II DYRKs [8,9] it certainly has additional targets [8–10].
CCND1 (cyclin D1) is a key regulator of the G1–S-phase cell-cycle transition. In early G1-phase growth factor signalling leads to an ERK1/2 (extracellular-signal-regulated kinase 1/2)-dependent increase in CCND1 gene transcription . CCND1 forms complexes with CDK4 and CDK6 that phosphorylate the Rb (retinoblastoma) protein, thereby de-repressing E2F-dependent transcription of S-phase genes . CCND1 overexpression is implicated in the development and progression of numerous cancers and this may result from gene amplification or post-translational stabilization . CCND1 can bind to PCNA (proliferating-cell nuclear antigen) blocking DNA replication and so must be degraded for cells to progress through the S-phase . GSK3β phosphorylates CCND1 on Thr286 at the beginning of S-phase resulting in CCND1 nuclear export and subsequent proteasomal degradation [15,16]. The phosphorylation and turnover of CCND1 may be important in cancer since the non-phosphorylatable T286A mutant of CCND1 has a mainly nuclear localization and can transform murine fibroblasts, whereas wild-type CCND1 cannot . However, inhibition of GSK3β does not completely block CCND1 turnover suggesting the existence of other Thr286-CCND1 kinases [18,19]. Indeed, DYRK1B has been proposed to phosphorylate CCND1 and promote its turnover; however, DYRK1B was proposed to phosphorylate CCND1 on Thr288, distinct from the Thr286 site targeted by GSK3β [20,21].
In the present study we describe a novel small molecule DYRK1B inhibitor AZ191 that exhibits 10-fold selectivity for DYRK1B over DYRK1A and is biologically active in cells. We have used AZ191 in conjunction with in vitro kinase assays, phospho-specific immunoblot analysis and MS to show that DYRK1B promotes the turnover of CCND1 in a GSK3β-independent manner by phosphorylating CCND1 on Thr286, and not Thr288 as originally proposed . Finally, whereas DYRK1B promotes a G1-phase cell-cycle arrest, overexpression of CCND1 fails to prevent this, indicating that the decrease in CCND1 levels is not the sole cause of the G1-phase arrest. Indeed, DYRK1B expression also increases the levels of the CDK inhibitors p21CIP1 (21 kDa CDK-interacting protein 1) and p27KIP1 (CDK-inhibitory protein 1), further implicating DYRK1B in the regulation of the cell-cycle progression.
MATERIALS AND METHODS
AZ191 and AZ4216 were kindly provided by AstraZeneca. MG132 was purchased from Calbiochem and Chir99021 from Cayman Chemicals. Horseradish peroxidase-conjugated secondary antibodies were from Bio-Rad Laboratories, and detection was with the ECL system (GE Healthcare). siRNA oligonucleotides were obtained from ThermoScientific. All other reagents were from Sigma.
The anti-DYRK1B antibodies were produced in rabbits using the immunogen CGLRGVPQSTAASS for the Western blotting antibody and MAVPPGHGPFSGC for the antibody used in IP (immunoprecipitation) assays. Animal procedures were carried out in accordance with EU and U.K. Home Office regulations and in compliance with AstraZeneca bioethics policies. Antibodies specific for ERK1 (610031), HSP90 (heat-shock protein 90) (610418), β-catenin (610153) and p21Cip1 (556431) were from BD Biosciences; phospho-Thr286-CCND1 (2921) and phospho-Ser33/Ser37/Thr41-β catenin (9561) were from Cell Signaling Technology; FLAG (M2) was from Sigma; GST (glutathione transferase; 138) and phospho-Ser10-p27Kip1 (12939) were from Santa Cruz Biotechnology; GFP (11814460001) was from Roche; p27Kip1 (NA35) and CCND1 (CC12) were from Calbiochem; and HA (haemagglutinin) was provided by the Babraham Institute Monoclonal Antibody Facility.
pcDNA3-HA-CCND1 plasmids were reported previously . pcDNA3-FLAG-CDK4 was kindly provided by Dr Michelle Garrett (The Institute of Cancer Research, London, U.K.) and pEGFPC3-DYRK1A was kindly provided by Dr Walter Becker (Institut für Pharmakologie und Toxikologie, Aachen, Germany). DYRK1B was amplified by PCR from pOTB7-DYRK1B (from the Mammalian Gene Collection) and subcloned into pCMV-Tag2B, pEGFPC3 and into pcDNA4/TO. The sequences of all oligonucleotides used are available upon request.
Cells and cell culture
Cell culture reagents were purchased from Invitrogen. HEK (human embryonic kidney)-293 and PANC-1 cells were maintained in DMEM (Dulbecco's modified Eagle's medium) supplemented with 10% FBS, 2 mM L-glutamine, 100 units/ml penicillin and 0.1 μg/ml streptomycin. HTetR cells were maintained in DMEM supplemented with 10% FBS, 2 mM L-glutamine, 100 units/ml penicillin, 0.1 μg/ml streptomycin and 5 μg/ml blasticidin to maintain stable clone selection. HD1B cells were maintained in DMEM supplemented with 10% FBS, 2 mM L-glutamine, 100 units/ml penicillin, 0.1 μg/ml streptomycin, 5 μg/ml blasticidin and 100 μg/ml Zeocin™. Cells were routinely passaged until they reached 80% confluency.
In vitro DYRK kinase assay
EGFP–DYRK1A, EGFP–DYRK1B or EGFP–DYRK2 was immunoprecipitated from whole-cell lysate and pre-incubated in the absence or presence of various concentrations of AZ191 or harmine for 5 min at room temperature (20°C) and then assayed for kinase activity by incubating with 50 μM Woodtide, with two additional lysine residues attached to the N-terminus to allow it to bind to P81 paper (KKSSCYVDRKIYTYIQSRFY), 50 mM Tris/HCl (pH 7.5), 0.1 mM EGTA, 0.1% 2-mercaptoethanol, 10 mM MgCl2 and 0.1 mM [γ-32P]ATP in a total volume of 50 μl for 25 min at 30°C as described previously .
Preparation of cell extracts and Western blotting
Cells were lysed in ice-cold Tris-glycine lysis buffer, assayed for protein content and fractionated by SDS/PAGE as described previously . The SDS/PAGE gels were transferred on to Immobilon P membranes (Millipore), which were blocked in 0.1% Tween 20/TBS containing 5% (w/v) powdered milk and probed with the indicated antibodies. Immune complexes were visualized with the ECL system.
Pin1 (peptidylprolyl cis/trans isomerase, NIMA-interacting 1) pull-down assay
GST–Pin1 (wild-type and ∆WW domain mutant) was kindly provided by Dr Giannino Del Sal (Laboratorio Nazionale CIB, Trieste, Italy), immobilized on GSH–agarose beads and used to pull down phospho-Thr286-CCND1 as described previously .
Analysis of CCND1 phosphorylation by MS
HA–CCND1 was co-expressed in HEK-293 cells with either wild-type DYRK1B or kinase-dead DYRK1B. CCND1 was then extracted from whole-cell lysate by IP with HA–agarose (Santa Cruz Biotechnology) and fractionation by SDS/PAGE (10% gel). The gel was stained with Coomassie Blue and the CCND1 band was excised, digested with trypsin and analysed by MS essentially as described previously . For targeted LC–MS/MS (LC tandem MS) analyses, an inclusion list of theoretical doubly and triply charged ions corresponding to the non-, mono- and di-phosphorylated CCND1 peptide Ala268–Arg291 was used.
Generation of the HTetR and HD1B cell lines
HTetR cells express the Tet repressor protein and were generated by transfecting HEK-293 cells with pcDNA6/TR (Invitrogen) and selecting clones with blasticidin (Melford). HTetR cells were transfected with pcDNA4/TO-FLAG-DYRK1B and subclones were selected with resistance to blasticidin and zeocin (Invitrogen). Positive clones that displayed tetracycline-induced expression of DYRK1B were identified by Western blotting with anti-FLAG and anti-DYRK1B antibodies.
Ubiquitin pull-down assay
CCND1 polyubiquitination was assessed by its ability to be precipitated by a GST-fusion protein of the UBA (ubiquitin-binding domain) of DSK2 which preferentially binds Lys48-linked polyubiquitin chains . Cell extracts were generated as described in  and incubated end-over-end with GSH–agarose-immobilized GST–DSK2 at 4°C for 90 min, washed and fractionated by SDS/PAGE (12% gel) as described.
Assay of CCND1 turnover in cells
PANC-1 cells were pre-treated for 30 min with 1 μM AZ191, 1 μM harmine, 3 μM Chir99021, 3 μM AZ4216 or 10 μM MG132 prior to treatment with 10 μM emetine to block new protein synthesis. Cell extracts were fractionated by SDS/PAGE (10% gel) and immunoblotted as described previously .
Immune complexes were washed once with 1× buffer [100 mM NaCl, 50 mM Tris/HCl, 10 mM MgCl2 and 1 mM DTT (pH 7.9)] and resuspended in 100 μl of 1× buffer prior to the addition of 10 units of Calf Intestine Alkaline Phosphatase (Roche). Reactions were carried out at 37°C for 1 h with shaking.
Assay of cell proliferation
Flow cytometric analysis of cell-cycle profiles and [3H]thymidine incorporation were performed as described previously .
Cell growth curve
HTetR and HD1B cells were seeded at 1×104 cells/well in 12-well plates and cell number was determined by haemocytometer cell counting at 24 h intervals over the subsequent 6 days.
AZ191 is a potent and selective inhibitor of DYRK1B
The DYRKs can be subdivided into class I (DYRK1A and DYRK1B) and class II (DYRK2, DYRK3 and DYRK4). Harmine is a small molecule inhibitor that exhibits selectivity for class I over class II DYRKs [8,9]; it displays a limited selectivity for DYRK1A over DYRK1B and is frequently used as a probe for either, despite having additional off-target effects [8–10]. A small molecule inhibitor that displayed greater selectivity for DYRK1B over DYRK1A would certainly help in defining the cellular functions of DYRK1B. A set of 6-azaindoles from earlier protein kinase drug discovery projects at AstraZeneca were found to exhibit potent and selective inhibition of DYRK1B in kinase-selectivity screening. A screening cascade was established to further understand structure–activity relationships and drive optimization against DYRK1B from the initial hits. One of the more interesting compounds from this campaign was AZ191 (Figure 1A), described herein. Kinase selectivity for AZ191 was evaluated using the KINOMEscan™ screening platform, available via DiscoverX (http://www.discoverx.com; see the Supplementary Materials and methods section at http://www.biochemj.org/bj/457/bj4570043add.htm). AZ191 was tested against over 400 kinases, including 386 unique wild-type kinases, in a primary screen at a single concentration of 1 μM. The KINOMEscan™ selectivity score (S1) provides a measure of the number of non-mutant kinases with the percentage control signal <1/number of non-mutant kinases tested. For AZ191, the S1 score was 0.023, with only eight non-mutant kinase hits other than DYRK1B; these were CIT [citron (Rho-interacting, serine/threonine kinase 21)], DRAK1 (DAP kinase-related apoptosis-inducing protein kinase 1), DYRK1A, MEK5 (MAPK kinase 5), MKNK2 (MAPK-interacting serine/threonine kinase 2), PRKD2 (protein kinase D2), RPS6KA4 (ribosomal protein S6 kinase, 90 kDa, polypeptide 4) and YSK4 [yeast Sps1/Ste20-related kinase 4; also known as MAP3K19 (MAPK kinase kinase 19)] (Supplementary Table S1 at http://www.biochemj.org/bj/457/bj4570043add.htm).
AZ191 is a potent and selective inhibitor of DYRK1B
To confirm the potency and selectivity of AZ191 we performed in vitro kinases assays for DYRK1A, DYRK1B and DYRK2 using the synthetic peptide substrate Woodtide and [γ-32P]ATP. In this assay, harmine displayed the reported selectivity for class I over class II DYRKs, but was equipotent for DYRK1A and DYRK1B (Figure 1B). In contrast, AZ191 displayed high potency against DYRK1B (IC50 value of 17 nM) and was selective for DYRK1B over both DYRK1A and DYRK2 (Figures 1C and 1D).
AZ191 selectively inhibits the serine/threonine kinase activity, but not the tyrosine kinase activity, of DYRK1B
Harmine is a reversible inhibitor of the serine/threonine kinase activity of DYRKs, but is 20–60-fold less potent against the tyrosine kinase autophosphorylation activity DYRKs . To demonstrate if AZ191 exhibited similar properties, we expressed DYRK1B mutants in HEK-293 cells that were then treated with vehicle, AZ191 or harmine. As expected kinase-dead DYRK1B was not phosphorylated on Tyr273 (Figure 1E). Wild-type DYRK1B was phosphorylated on Tyr273, but neither harmine or AZ191 inhibited phospho-Tyr273 (Figure 1E). Thus at 1 μM AZ191, like harmine, selectively inhibited DYRK1B serine/threonine kinase activity (Figure 1C) with no effect on tyrosine kinase autophosphorylation at this concentration. In the present study we have used AZ191 to investigate the role of DYRK1B in CCND1 turnover.
DYRK1B phosphorylates CCND1 on Thr286, not Thr288
It was previously reported that DYRK1B phosphorylated CCND1 on Thr288 in the sequence ACT286PT288DVR ; the authors found that mutation of Thr286 increased, whereas Thr288 mutation reduced, DYRK1B-dependent phosphorylation of CCND1. This was surprising since DYRKs have been reported to be proline-directed kinases, preferring to phosphorylate serine or threonine residues with a proline in the +1 position (i.e. S-P or T-P) such as Thr286 . To re-investigate this we started by using in vitro kinase reactions. Purified recombinant DYRK1B was incubated with various GST–CCND1 fusion proteins and [γ-32P]ATP (Figure 2A). Mutation of Thr286 to an alanine residue (T286A) strongly inhibited CCND1 phosphorylation by DYRK1B, whereas mutation of Thr288 (T288A) had a small effect and the double mutant (T286A/T288A) behaved like T286A. The in vitro assay was repeated, but phosphorylation was detected by Western blotting using an anti-phospho-Thr286-CCND1 antibody. DYRK1B catalysed phosphorylation of wild-type GST–CCND1 at Thr286 and this was abolished in the T286A mutant, but also reduced in the T288A mutant (Figure 2B). Since this antibody recognizes phospho-Thr286-CCND1 these results suggested that Thr286 was the phospho-acceptor site and that mutation of the nearby Thr288 simply impaired substrate recognition.
Cyclin D1 is phosphorylated on Thr286, not Thr288, by DYRK1B
To extend these studies into cells we co-transfected HA–CCND1 constructs and FLAG–DYRK1B into HEK-293 cells. DYRK1B greatly enhanced the basal phosphorylation of CCND1 on Thr286 and this was abolished by either T286A or T288A mutations (Figure 2C). To further investigate DYRK1B-dependent phosphorylation of CCND1 we used the ‘Pin1 pull-down’ system [22,24]. Pin1 binds to phospho-serine–proline or phospho-threonine–proline motifs  and so should bind CCND1 that is phosphorylated on Thr286 (phospho-Thr286–Pro287), but not CCND1 that is phosphorylated on Thr288 (phospho-Thr288–Asp289). When expressed alone, no CCND1 was pulled down by GST–Pin1 beads; however, co-expression of DYRK1B increased Thr286 phosphorylation so that CCND1 was now precipitated by GST–Pin1 (Figure 2D). Little of the T286A or T286A/T288A CCND1 was precipitated by GST–Pin1, whereas there was some pull down of the T288A CCND1 mutant (Figure 2D), a result consistent with the previous CCND1 mutational studies (Figures 2A–2C). Given the recognized preference of Pin1 for phospho-serine–proline or phospho-threonine–proline motifs these results again suggested that CCND1 was phosphorylated at Thr286 and not Thr288.
Finally, HA–CCND1 was co-expressed with FLAG–DYRK1B in HEK-293 cells, immunoprecipitated with an anti-HA antibody, separated by SDS/PAGE, excised, digested with trypsin and analysed by MS. Only one phospho-peptide was detected and this was phosphorylated only on Thr286 (Figure 2E). Taken together, these data reveal that in vitro and in cells DYRK1B phosphorylates CCND1 at Thr286, and not Thr288 as proposed previously .
AZ191 is selective for DYRK1B over DYRK1A in cells
To test the potency and selectivity of AZ191 in cells, CCND1 was co-expressed with either DYRK1A or DYRK1B in HEK-293 cells which were subsequently treated with increasing doses of AZ191 or harmine prior to analysis of CCND1 phosphorylation by Western blotting. Harmine inhibited phosphorylation of CCND1 by both DYRK1A and DYRK1B (Figure 3A). In contrast, AZ191 displayed much greater potency for DYRK1B over DYRK1A, inhibiting CCND1 phosphorylation at doses as low as 30–100 nM (Figure 3B). Thus the selectivity of AZ191 for DYRK1B in vitro (Figure 1) was replicated in vivo for a known and validated substrate, CCND1.
AZ191 is a potent and selective inhibitor of DYRK1B in cells
DYRK1B drives CCND1 turnover independently of GSK3β activity in cells
To investigate the effect of DYRK1B on CCND1 in cells, we used the ‘Tet-on’ system, generating a control HEK-293 cell line that contains the Tet repressor protein, but no DYRK1B (HTetR) and a second cell line containing DYRK1B under the control of the Tet-operon (HD1B). In HD1B cells DYRK1B expression increased rapidly and was maximally elevated 4–6 h after Tet addition (Figure 4A). Co-ordinately, CCND1 levels dropped from 4 h onwards with a relative increase in the ratio of phospho-Thr286/total CCND1, indicating that inducible expression of DYRK1B promoted the phosphorylation and turnover of endogenous CCND1 (Figure 4A). Treatment of the control HTetR cell line had no effect on the levels of DYRK1B or CCND1. Phospho-Thr286-CCND1 levels decreased in the HTetR cells over time, probably reflecting alterations in cell-cycle progression as cells neared confluency.
DYRK1B kinase activity promotes CCND1 turnover independently of GSK3β
In contrast with experiments in which DYRK1B and CCND1 were both transiently overexpressed (Figures 2C and 2D) we observed a more modest increase in phospho-Thr286 on endogenous CCND1 in HD1B cells (Figure 4A). Since the phosphorylation of CCND1 at Thr286 is tightly linked to its proteasome-dependent degradation we speculated that this made it difficult to capture the accumulation of the phospho-Thr286 form. Indeed, treatment of HD1B cells with MG132 to inhibit the proteasome allowed us to observe a more robust and significant increase in phospho-Thr286 on endogenous CCND1 (Figures 4B and 4C). To verify that CCND1 was indeed ubiquitinated, whole-cell extracts from HEK-293 cells transfected with empty vector or HA–CCND1 were incubated with the UBA of DSK2 fused to GST in a pull-down assay. Wild-type DSK2 was able to capture polyubiquitinated CCND1, whereas the mutant DSK2 (∆), which is unable to bind ubiquitinated proteins, did not (Figure 4D). Since DSK2 specifically binds Lys48-linked polyubiquitin chains  this served to demonstrate that CCND1 was indeed subject to Lys48-linked polyubiquitination in HEK-293 cells.
Since GSK3β is a known Thr286-CCND1 kinase  it was important to determine if the DYRK1B-mediated phosphorylation of CCND1 was dependent upon GSK3β. HD1B cells were treated with AZ191 or one of two GSK3β inhibitors, Chir99021  or AZ4216, prior to the induction of DYRK1B. Expression of DYRK1B decreased total CCND1 levels, but maintained phospho-Thr286 levels, thereby increasing the phospho-Thr286/total CCND1 ratio; this was fully blocked by the DYRK1B inhibitor AZ191, whereas the two GSK3β inhibitors were ineffective (Figure 4E). Chir99021 and AZ4216 were active as they blocked the phosphorylation of, and thereby stabilized, β-catenin, a well-characterized GSK3β substrate . Thus GSK3β activity was not required for DYRK1B-mediated phosphorylation of Thr286-CCND1 (Figure 4E).
DYRK1B, not GSK3β, is the major CCND1 Thr286 kinase in PANC-1 cells
To investigate the turnover of endogenous CCND1 by endogenous DYRK1B we performed emetine chase experiments in PANC-1 cells, a pancreatic cancer cell line in which the DYRK1B gene is amplified . CCND1 turned over rapidly in PANC-1 cells with a half-life of approximately 30 min (Figure 5A), agreeing with previous studies , and this was blocked by the proteasome inhibitor MG132. Treatment with either AZ191 or harmine increased the half-life of CCND1, whereas the two GSK3β inhibitors, Chir99021 and AZ4216, had no effect (Figure 5A), although they were again effective at inhibiting GSK3β as judged by the loss of β-catenin phosphorylation (Figure 5B). Figure 5(B) also revealed that the two DYRK1B inhibitors did not inhibit GSK3β, another CMGC kinase, as judged by β-catenin phosphorylation.
DYRK1B, not GSK3β, is required for CCND1 turnover in PANC-1 cells
These results suggested that in PANC-1 cells DYRK1B, but not GSK3β, was controlling CCND1 turnover. To confirm this we compared the effects of AZ191 with that of DYRK1B siRNA knockdown. In cells transfected with control siRNA, a 60 min treatment with emetine resulted in the loss of CCND1 and this was again inhibited by AZ191 (Figure 5C, lanes 1 and 2 compared with 3 and 4). In parallel, DYRK1B siRNA knockdown also inhibited the turnover of CCND1, replicating the effects of AZ191 (Figure 5C, lanes 3 and 4 compared with 5 and 6). Furthermore, AZ191 had no further effect on CCND1 levels in cells in which its main target, DYRK1B, had been knocked down, consistent with AZ191 acting ‘on target’ (Figure 5C, lanes 7 and 8). Thus DYRK1B siRNA phenocopied the effects of AZ191, suggesting DYRK1B was the sole target of AZ191 in the regulation of CCND1 in these cells. Consistent with this, siRNA against GSK3β had no effect on the turnover of CCND1 (Figure 5C, lanes 1 and 2 compared with 9 and 10), replicating our results with GSK3 inhibitors (Figure 5A). Finally, AZ191 reversed CCND1 turnover in cells in which GSK3β had been knocked down (Figure 5C, lanes 11 and 12). Taken together these results confirm the selectivity of AZ191 and show that DYRK1B controls CCND1 turnover in PANC-1 cells independently of GSK3β.
DYRK1B phosphorylates CCND1 complexed with CDK4, as well as free monomeric CCND1
DYRK1B phosphorylated free monomeric CCND1 in vitro (Figures 2A and 2B). To address if DYRK1B could also phosphorylate CCND1 that was complexed with CDK4, HA–CCND1 and FLAG–CDK4 were co-expressed in HEK-293 cells prior to CDK4 isolation from cell extracts with an anti-FLAG antibody IP. No CCND1 was isolated from cells that did not express FLAG–CDK4 demonstrating the specificity of the IP procedure. However, in cells that co-expressed HA–CCND1 and FLAG–CDK4, the anti-FLAG antibody IP isolated both FLAG–CDK4 and HA–CCND1 (Figure 6A). Following expression of EGFP–DYRK1B, the phosphorylation of CCND1 complexed to CDK4 was increased both in cell lysates and in the anti-FLAG antibody IP indicating that DYRK1B can phosphorylate CCND1 complexed to CDK4 in cells. To assess direct phosphorylation of CCND1 bound to CDK4 by DYRK1B, HA–CCND1 was co-expressed with FLAG–CDK4 in HEK-293 cells and the CCND1–CDK4 complexes were isolated by an anti-FLAG antibody IP. These complexes were dephosphorylated using alkaline phosphatase and used as a substrate in a DYRK1B in vitro kinase assay. DYRK1B was able to re-phosphorylate CCND1 at Thr286 in this assay demonstrating that, in vitro, DYRK1B can phosphorylate CCND1 complexed to CDK4 (Figure 6B) as well as monomeric CCND1 (Figures 2A and 2B). The inclusion of a ‘no DYRK1B’ control (Figure 6B, last lane) ruled out the possibility of any other co-precipitating kinase phosphorylating Thr286. Thus DYRK1B can phosphorylate free monomeric and CDK4-bound CCND1 in vitro and can promote phosphorylation of CCND1 bound to CDK4 in cells.
DYRK1B can phosphorylate CCND1 in complex with CDK4
DYRK1B promotes a G1-phase cell-cycle arrest
Decreased levels of CCND1 are linked to diminished G1–S-phase transition and a G1-phase cell-cycle arrest  so we investigated the effect of DYRK1B expression on cell proliferation. Tet-induced DYRK1B expression in HD1B cells led to a 40% reduction in [3H]thymidine incorporation (Figure 7A) and reduced cell proliferation (Figure 7B), whereas the control HTetR cell line was non-responsive to Tet in both assays. Flow cytometry revealed that DYRK1B expression caused HD1B cells to undergo a sustained G1-phase cell-cycle arrest, whereas the HTetR cells did not (Figure 7C). DYRK1B expression did not increase the fraction of dead (sub-G1-phase) cells (results not shown). Thus Tet-regulated expression of DYRK1B promoted a G1-phase cell-cycle arrest.
DYRK1B inhibits G1–S-phase transition
Since DYRK1B drives CCND1 turnover in PANC-1 cells (Figure 5) we also examined the consequence of inhibiting endogenous DYRK1B in these cells. When serum-starved PANC-1 cells were released back into the cell cycle by the re-addition of FBS, inhibition of DYRK1B with either AZ191 or harmine increased cell-cycle progression (Figure 8A) and the fraction of cells in S-phase (Figure 8B) indicating that these cells were transiting from the G1- to the S-phase and on to the G2-phase at a higher rate than vehicle-treated cells. Adding back FBS to starved PANC-1 cells for 8 h caused re-expression of CCND1 that was phosphorylated at Thr286. Either AZ191 or harmine inhibited Thr286 phosphorylation at 8 h, consistent with CCND1 Thr286 phosphorylation being DYRK1B-dependent (Figure 8C). The effect of DYRK1B inhibitors on CCND1 phosphorylation and turnover was less apparent at 16 and 24 h (Figure 8C), but nevertheless DYRK1B inhibition caused a decrease in phospho-Thr286/total CCND1 at all time points (Figure 8D). At 8 h these cells showed no signs of increased expression of CCNA a useful marker for cells that have progressed through G1-phase and into or beyond the S-phase; CCNA increased at 24 h. Together, these results show that both Tet-regulated DYRK1B (Figures 7A–7C) and endogenous DYRK1B (Figures 8A–8D) act to inhibit cell-cycle progression through the G1-phase and probably early into the S-phase.
Inhibition of DYRK1B in PANC-1 cells inhibits Thr286 phosphorylation and promotes the G1–S-phase transition
DYRK1B expression increases the abundance of the cell-cycle regulators p21Cip1 and p27Kip1
To test the hypothesis that the DYRK1B-mediated turnover of CCND1 is required for the DYRK1B-induced cell-cycle arrest, cells were transiently transfected with wild-type or a T286A mutant of CCND1 to see if this would rescue the DYRK1B-induced G1-phase arrest. Overexpression of CCND1 in the HTetR cell line had no effect upon the cell-cycle profile (Figure 9A); furthermore, HD1B cells transfected with wild-type or T286A CCND1 exhibited the same magnitude of G1-phase arrest as the cells transfected with empty vector following inducible DYRK1B expression (Figure 9A). Thus the increased turnover of CCND1 alone does not account for the DYRK1B-induced G1-phase arrest.
The DYRK1B-induced G1-phase arrest is not solely mediated by a decrease in CCND1 and may involve an increase in the CDK inhibitors p21Cip1 and p27Kip1
The CDK inhibitors p21Cip1 and p27Kip1 can drive a G1-phase cell-cycle arrest and both are reported to be substrates for DYRK1B [33–36]. Indeed, Tet-treatment of HD1B cells (but not HTetR cells) increased the levels of DYRK1B, p21Cip1 and p27Kip1 (Figure 9B); this was strongly inhibited by AZ191, whereas harmine exhibited similar, but less pronounced, effects (Figure 9C). Whether expression of p21Cip1 and p27Kip1 in this cell system is due to direct phosphorylation by DYRK1B or a secondary effect of DYRK1B expression is currently unclear. The increase in p21Cip1 and p27Kip1 levels upon expression of DYRK1B was notably slower (from 12 h onwards) than the decrease in CCND1 expression (from 6 h onwards; Figure 4A). However, DYRK1B also increased phospho-Ser10-p27Kip1 in an AZ191-sensitive manner (Figure 9C) and this was apparent at 6–8 h, prior to the increase in p27Kip1 (Figure 9B), raising the possibility that this might be the driving event for stabilization of p27Kip1. Regardless, it is clear that expression of DYRK1B alters the levels of p21Cip1 and p27Kip1 and these are likely to account for the DYRK1B-inducible G1-phase cell-cycle arrest that is observed when CCND1 levels are maintained.
DYRK1B is involved in myoblast differentiation, stress responses and is also a candidate oncoprotein, and yet relatively little is known about DYRK1B substrates or cellular function. This reflects in part the lack of a suitable DYRK1B-selective inhibitor. In the present study we have demonstrated that AZ191 is a novel selective inhibitor of DYRK1B and have used it to show that DYRK1B phosphorylates CCND1 at Thr286 (not Thr288) to promote its destruction. Furthermore, we show that DYRK1B acts independently of GSK3β to phosphorylate CCND1 at Thr286. Finally, in PANC-1 cells DYRK1B, and not GSK3β, is the major Thr286-CCND1 kinase.
AZ191 as a probe for DYRK1B
Harmine is frequently used as an inhibitor of DYRK1B, but also inhibits DYRK1A and certainly has non-DYRK targets [8–10]. In addition, an alternative, as yet un-named DYRK inhibitor that is marginally more potent against DYRK1B than DYRK1A was described, athough not fully characterized [37,38]. In contrast, AZ191 is a novel DYRK1B-selective inhibitor with 5–10-fold selectivity over DYRK1A and ~100-fold selectivity over DYRK2, a class II DYRK. AZ191 is an ATP-competitive, cell permeable and reversible inhibitor that fully inhibits the serine/threonine kinase activity of DYRK1B at 1 μM without blocking auto-phosphorylation of the activation loop tyrosine residue. This mimics the mode of action of harmine ; however, unlike harmine, AZ191 is selective for DYRK1B over DYRK1A in vitro and in cells. Finally, AZ191 can inhibit the phosphorylation of CCND1 (a validated cellular target of DYRK1B, see below) in cells at doses as low as 30 nM. We therefore anticipate that AZ191 will serve as a useful probe for defining DYRK1B substrates and DYRK1B functions.
DYRK1B phosphorylates CCND1 at Thr286, not Thr288, in cells to promote its destruction
CCND1 bound to CDK4/6 promotes G1-phase progression, but it must also be degraded in the early S-phase to release PCNA and allow DNA replication . Phospho-Thr286-CCND1 levels rise in the S-phase leading to its cell-cycle-dependent degradation  and GSK3β is thought to be the S-phase kinase responsible for Thr286 phosphorylation . However, Yang et al.  demonstrated that activation of GSK3β signalling had little effect on CCND1 levels during the cell cycle and that GSK3β inhibition failed to alter total or phospho-Thr286 CCND1 levels; nor did overexpression of a constitutively active GSK3β alter CCND1 levels. These and other studies  suggest that another Thr286-CCND1 kinase can promote CCND1 degradation.
Our attention was drawn to DYRK1B, which had previously been implicated in the control of CCND1 turnover ; however, mutational studies suggested that it phosphorylated Thr288 and not Thr286 . This result was surprising for two reasons: first, Diehl et al.  had shown that the main phosphorylated residue on CCND1 was Thr286 since the T286A mutation abolished CCND1 phosphorylation ; secondly, DYRKs normally phosphorylate serine–proline or threonine–proline motifs , as is the case for Thr286, but not Thr288, in the CCND1 sequence LACT286PT288DVR. Nonetheless, a further study suggested that GSK3β phosphorylated CCND1 on Thr286 and DYRK1B phosphorylated on Thr288 . Notably these two studies [21,39] relied largely on mutagenesis to define phosphorylation sites. We were concerned that the close juxtaposition of Thr286 and Thr288 could result in artefacts in which loss of one site might influence recognition of the other site by DYRK1B. Indeed, in all our mutagenesis studies we observed that, whereas the T286A mutation abolished CCND1 phosphorylation, T288A also partially reduced phosphorylation; a result consistent with Thr286 as the phospho-acceptor site, with the integrity of Thr288 being important for substrate recognition. Indeed, DYRK1B enhanced the precipitation of CCND1 by GST–Pin1, consistent with phosphorylation at Thr286. To avoid such confusion we mapped the site of DYRK1B-induced CCND1 phosphorylation by MS; only one phospho-peptide was detected for the C-terminus of CCND1 and this peptide was phosphorylated on Thr286 not Thr288. Finally, DYRK1B increased Thr286 phosphorylation as detected by a phospho-Thr286-specific antibody. Together these results demonstrate that DYRK1B phosphorylates CCND1 at Thr286 and suggest that mutations in Thr288 that reduce phosphorylation by DYRK1B actually reflect reduced efficiency of Thr286 phosphorylation, possibly owing to reduced substrate recognition given its close proximity to Thr286. This proposal is internally consistent with all the results of the present study, with the widely accepted role of Thr286 in CCND1 turnover and with the ability of DYRK1B to promote CCND1 turnover itself.
DYRKs can act as priming kinases for GSK3  and some DYRKs can phosphorylate the same sites as GSK3β . It was therefore important to establish whether DYRK1B was acting directly to phosphorylate Thr286-CCND1 in cells or whether it was activating GSK3β. The results of the present study were clear and unambiguous: DYRK1B was able to phosphorylate CCND1 in an in vitro assay (where GSK3β was absent); inducible expression of DYRK1B failed to activate GSK3β and GSK3β inhibitors failed to reverse DYRK1B-induced turnover of CCND1, whereas AZ191 was effective. Indeed, as judged by AZ191 and siRNA, endogenous DYRK1B played the major role in proteasome-dependent CCND1 turnover in PANC-1 cells where GSK3β inhibition or GSK3β RNAi had no effect. Taken together these results show that even in cells with fully functional GSK3β signalling, DYRK1B phosphorylates CCND1 on Thr286 directly and not via activation of GSK3β. Thus DYRK1B is a novel Thr286-CCND1 kinase.
Regulation of cell cycle by DYRK1B
The results of the present study underline the role of DYRK1B in cell-cycle regulation. On the one hand, Tet-inducible expression of DYRK1B in HD1B cells elicited a sustained G1-phase arrest, whereas inhibition of endogenous DYRK1B in PANC-1 cells facilitated G1–S-phase progression following re-stimulation of starved cells with FBS. This raises two distinct questions: how does DYRK1B inhibit the G1–S-phase progression and why, as a candidate oncogene, does DYRK1B cause a G1-phase cell-cycle arrest?
With regard to how DYRK1B expression inhibits G1–S-phase progression, a previous study claimed that overexpression of T288A CCND1 was able to rescue the anti-proliferative effects of DYRK1B arguing that turnover of CCND1 was the major determinant of DYRK1B-mediated cell-cycle arrest ; however, this result is difficult to interpret when we now know that Thr288 is not the DYRK1B phosphorylation site. In the present study we repeatedly found that overexpression of CCND1 (wild-type or T286A) did not rescue the DYRK1B-induced inhibition of cell-cycle progression demonstrating unequivocally that loss of CCND1 was not the sole cause. Indeed, we also observed a slower increase in the expression of the CDK inhibitors p21Cip1 and p27Kip1, which are both well known for their ability to elicit a G1-phase cell-cycle arrest. p21Cip1 is reported to be phosphorylated at Ser153 by DYRK1B leading to its nuclear export ; however, this does not alter the ability of p21Cip1 to mediate cell-cycle progression. DYRK1B has also been proposed to phosphorylate Ser10 of p27Kip1 to increase p27Kip1 protein stability . Taken together with these studies the results of the present study suggest that DYRK1B expression or activation elicits a G1-phase cell-cycle arrest by a bipartite mechanism consisting of at least two kinetically and genetically distinct pathways: first, turnover of CCND1, which is a rapid event; and secondly, the slower expression of p21Cip1 or p27Kip1 which proceeds independently of the turnover of CCND1. There are clear parallels here with the response of cells to certain genotoxic stresses that cause rapid CCND1 turnover to initiate the G1-phase arrest, but require expression of p21Cip1 to sustain the arrest . Indeed DYRK1B has been proposed to be activated by 5-FU (5-fluorouracil) treatment to induce turnover of CCND1 and p27Kip1 accumulation .
The results of the present study also raise the wider question of why DYRK1B, a candidate oncogene that is selected for amplification in some tumours, can promote cell-cycle arrest as this would appear to be a tumour-suppressive mechanism. In this context it is worth noting that DYRK1B has also been proposed to promote cell survival . Several genes, notably members of the BCL2 (B-cell CLL/lymphoma 2) family, can act as oncogenes by promoting cell survival independently of the effects on cell proliferation. DYRK1B activation may be part of a co-ordinated stress response driving cell-cycle arrest to allow repair and/or survival. Indeed DYRK1B has been proposed to promote cell survival in response to cisplatin-induced apoptosis in NSCLC (non-small-cell lung carcinoma) cell lines , SW620 colon cancer cells and PANC-1 pancreatic cancer cells . DYRK1B has also been reported to reduce the level of ROS (reactive oxygen species) via up-regulation of SOD2 (superoxide dismutase 2) and SOD3, thereby increasing cell survival .
In summary, we have characterized AZ191 as a novel, potent and selective inhibitor of DYRK1B. Using AZ191, we have demonstrated that DYRK1B phosphorylates CCND1 directly on Thr286, not Thr288, leading to increased CCND1 turnover. In doing so DYRK1B acts independently of GSK3β; indeed, in some cells it may be the major Thr286 kinase. The DYRK1B-induced G1-phase arrest almost certainly involves the increased expression of p21Cip1 and p27Kip1 protein levels in addition to loss of CCND1. We anticipate that AZ191 will be useful in the future for defining new DYRK1B targets and biological functions.
Dulbecco’s modified Eagle’s medium
dual-specificity tyrosine phosphorylation-regulated kinase
glycogen synthase kinase 3
human embryonic kidney
mitogen-activated protein kinase
21 kDa CDK-interacting protein 1
proliferating-cell nuclear antigen
peptidylprolyl cis/trans isomerase, NIMA-interacting 1
CDK-inhibitory protein 1
Pamela Lochhead, Anne Ashford and Simon Cook conceived and designed the study, with input from Sylvie Guichard and Kevin Hudson. Jason Kettle and Kevin Hudson co-ordinated the drug-discovery effort that led to the identification of AZ191. The initial demonstration of CCND1 phosphorylation by DYRK1B was made by Pamela Lochhead. All other experiments in the paper were performed and analysed Anne Ashford with the exception of the MS, where she prepared samples and processing and analysis was performed by David Oxley. Anne Ashford prepared all the Figures. Anne Ashford, Pamela Lochhead and Simon Cook interpreted the results with contributions from Sylvie Guichard and Kevin Hudson. Anne Ashford and Simon Cook co-wrote the paper with critical comments and contributions from all authors.
We thank colleagues in the Cook Group for advice and many helpful discussions and Anne Segonds-Pichon (Babraham Bioinformatics Group) for advice on statistical analysis. We are also grateful to Walter Becker for providing the EGFP-DYRK1A expression plasmid.
This work was supported by the Biotechnology and Biological Sciences Research Council via a CASE PhD studentship [grant number 3068901 (to A.A.)] awarded to AstraZeneca and The Babraham Institute (to S.C.) and a Institute Strategic Programme Grant [grant number BBS/E/B/000C0417 (to S.C.)], and the Association for International Cancer Research via a project grant [grant number AICR09-0257(to S.C.] (supporting P.L.).