Báo cáo khoa học: DYRK1A phosphorylates caspase 9 at an inhibitory site and is potently inhibited in human cells by harmine pptx

Depletion of DYRK1A from human cells by short interfering RNA inhibits the basal phosphorylation of caspase 9 at an inhibitory site, Thr125.. When co-expressed in cells, DYRK1A interacts

and is potently inhibited in human cells by harmine

Anne Seifert, Lindsey A Allan and Paul R Clarke

Biomedical Research Institute, College of Medicine, Dentistry and Nursing, University of Dundee, UK

DYRK1A is the most extensively characterized

mem-ber of the evolutionarily conserved dual-specificity

tyrosine-phosphorylation-regulated protein kinase

(DYRK) family, which is distantly related to

mitogen-activated protein kinases (MAPKs), cyclin-dependent

protein kinases (CDKs), CDK-like kinases (CLKs)

and glycogen synthase kinase 3 [1] The DYRK family

comprises several members in mammals, of which

DYRK1A and DYRK1B are predominantly localized

to the nucleus, whereas DYRK2 is cytoplasmic [2,3]

Functional studies in mammals and Drosophila suggest

a conserved regulatory role for DYRK1A in

neurogen-esis Mutant flies with reduced expression of minibrain

kinase, the Drosophila orthologue of DYRK1A,

dis-play specific reductions in the size of the optic lobes

and central brain hemispheres as well as distinctive

behavioural abnormalities [4] The human DYRK1A gene has been implicated as having a role in patho-genesis due to its location in the ‘Down’s syndrome critical region’ (DSCR) on chromosome 21 [5,6], which is present in three copies in Down’s syndrome individuals

The molecular mechanisms underlying Down’s syn-drome and its associated pathologies are likely to be complex An important role has been proposed for the transcription factor nuclear factor of activated T cells (NFAT), which is dysregulated by increased gene dos-age of DYRK1A and DSCR1, another DSCR gene that encodes an inhibitor of the protein phosphatase calcineurin⁄ PP-2B [7] Phosphorylation of NFAT by DYRKs counteracts its dephosphorylation by calci-neurin, thereby retaining NFAT in the cytoplasm and

Keywords

apoptosis; caspase; DYRK; harmine;

protein kinase

Correspondence

P R Clarke, Biomedical Research Institute,

University of Dundee, Level 5, Ninewells

Hospital and Medical School, Dundee DD1

9SY, UK

Fax: +44 1382 669993

Tel: +44 1382 425580

E-mail: p.r.clarke@dundee.ac.uk

(Received 28 August 2008, revised 8

October 2008, accepted 21 October 2008)

doi:10.1111/j.1742-4658.2008.06751.x

DYRK1A is a member of the dual-specificity tyrosine-phosphorylation-reg-ulated protein kinase family and is implicated in Down’s syndrome Here,

we identify the cysteine aspartyl protease caspase 9, a critical component of the intrinsic apoptotic pathway, as a substrate of DYRK1A Depletion of DYRK1A from human cells by short interfering RNA inhibits the basal phosphorylation of caspase 9 at an inhibitory site, Thr125 DYRK1A-dependent phosphorylation of Thr125 is also blocked by harmine, confirm-ing the use of this b-carboline alkaloid as a potent inhibitor of DYRK1A

in cells We show that harmine not only inhibits the protein–serine⁄ threo-nine kinase activity of mature DYRK1A, but also its autophosphorylation

on tyrosine during translation, indicating that harmine prevents formation

of the active enzyme When co-expressed in cells, DYRK1A interacts with caspase 9, strongly induces Thr125 phosphorylation and inhibits caspase 9 auto-processing Phosphorylation of caspase 9 by DYRK1A involves co-localization to the nucleus These results indicate that DYRK1A sets a threshold for the activation of caspase 9 through basal inhibitory phos-phorylation of this protease Regulation of apoptosis through inhibitory phosphorylation of caspase 9 may play a role in the function of DYRK1A during development and in pathogenesis

Abbreviations

CDK, cyclin-dependent kinase; CLK, CDK-like kinase; DYRK, dual-specificity tyrosine phosphorylation-regulated kinase; ERK, extracellular signal-regulated kinase; GFP, green fluorescent protein; MAPK, mitogen-activated protein kinase; NES, nuclear export signal; NFAT, nuclear factor of activated T cells; NLS, nuclear localization signal; siRNA, short interfering RNA; TPA, 12-O-tetradecanoylphorbol-13-acetate.

inhibiting its transcriptional activity [8] However, the

substrates and cellular functions of DYRK1A during

normal development and in pathological conditions

remain to be fully identified

Here, we identify a novel substrate for DYRK1A,

the cysteine aspartyl protease caspase 9, which is a

critical component of the intrinsic or mitochondrial

apoptotic pathway Caspase 9 is fully activated by a

variety of apoptotic stimuli that trigger the release of

cytochrome c from mitochondria Once in the cytosol,

cytochrome c induces the oligomerization of Apaf-1

and subsequent recruitment of procaspase 9 into a

high-molecular-mass multimeric complex, termed the

apoptosome [9] Apaf-1-induced dimerization of

caspase 9 leads to its activation and autocatalytic

pro-cessing [10] Active caspase 9 initiates a proteolytic

cascade by processing and activating downstream

effector caspases such as caspase 3 and caspase 7,

lead-ing to the organized disassembly of the cell [11]

Regulation of apoptosome formation is controlled

at the level of cytochrome c release from mitochondria

by pro- and anti-apoptotic proteins of the Bcl-2 family

[12] In addition, the pathway is controlled

down-stream of cytochrome c release at the level of the

apoptosome [13] Caspase 9 activation is subject to

modulation by protein kinases activated in signal

transduction pathways initiated by extracellular signals

or cellular stresses [14–16] We have shown previously

that extracellular signal-regulated kinase (ERK)1 and

ERK2 MAPKs, which are activated in response to

sur-vival signals, restrain caspase 9 activation by direct

phosphorylation on a critical inhibitory site, Thr125

[14,17] Furthermore, CDK1–cyclin B1 protects mitotic

cells from apoptosis induced by microtubule poisons

by phosphorylating caspase 9 on the same residue [18]

During these studies, we obtained evidence that

Thr125 may be subject to phosphorylation by

addi-tional protein kinase activities, because a basal level of

Thr125 phosphorylation persists when ERK1⁄ 2 and

CDK1–cyclin B1 are inhibited [14,18]

Here, we identify DYRK1A as an additional kinase

that targets Thr125 of caspase 9 in cells These results

suggest a function of DYRK1A in the regulation of

apoptosis that may be relevant to its roles during

development and in pathogenesis We also present

evi-dence that harmine, a potent and specific inhibitor of

DYRKs in vitro [19], efficiently inhibits DYRK1A

activity towards caspase 9 in cells and also blocks the

co-translational activating tyrosine

autophosphoryla-tion of DYRK1A, showing that this b-carboline

alka-loid can be used to test proposed cellular targets of

DYRK1A and potentially could be used to reverse the

effects of DYRK1A overexpression

Results

Identification of DYRK1A as a Thr125 kinase

in cells Previous studies have shown that caspase 9 is phosphor-ylated on a single major site, Thr125, catalysed by the proline-directed kinases ERK1⁄ 2 MAPKs and CDK1–cyclin B1 in response to growth factors and during mitosis, respectively However, residual phos-phorylation when ERK1⁄ 2 and CDK1 are inhibited suggests that an additional kinase also targets this site [14,18] In serum-starved U2.C9–C287A cells, a U2OS-derived cell line stably expressing catalytically inactive caspase 9 [18], the MEK1 inhibitors PD0325901 and U0126, which block ERK1⁄ 2 activation, did not reduce basal Thr125 phosphorylation (Fig 1A) By contrast, both inhibitors blocked ERK1⁄ 2-dependent phospho-rylation of Thr125 induced by the phorbol ester 12-O-tetradecanoylphorbol-13-acetate (TPA) (Fig 1B) Furthermore, although the CDK1–cyclin B1-dependent phosphorylation of Thr125 induced by the microtubule poison nocodazole, which arrests cells in mitosis, was inhibited by the CDK inhibitors alsterpaullone, pur-valanol A or roscovitine (Fig 1C), only roscovitine and purvalanol A, but not alsterpaullone, caused a reduc-tion in basal phospho-Thr125 levels (Fig 1A) These results suggest that the basal Thr125 phosphorylation in unstimulated cells involves a novel kinase that is not dependent on ERK1⁄ 2 or CDK activity, but is sensitive

to roscovitine and purvalanol A

DYRK1 has been reported to be inhibited by rosco-vitine and purvalanol A, but not by alsterpaullone [20,21] Consistent with the idea that DYRK1A might

be a physiologically relevant Thr125 kinase, the amino acid context of Thr125 fulfils the reported primary sequence requirements of DYRK kinases (Fig 1D) DYRK1A and related DYRKs are proline-directed kinases which require a proline immediately C-terminal

to the phosphorylated serine or threonine residue They also favour the presence of an arginine residue at the )2 or )3 position N-terminal to the phosphory-lated serine or threonine residue [22,23], although exceptions exist [24,25] DYRK1A in particular favours an additional proline at the )2 position [22,23], as found in caspase 9 (Fig 1D) Our initial results therefore lead us to investigate further the potential role of DYRK1A as a Thr125 kinase

To test the role of DYRK1A independently of chemical inhibitors, we depleted DYRK1A from U2.C9–C287A cells using RNA interference Transfec-tion of two distinct synthetic short interfering RNA (siRNA) duplexes targeting DYRK1A mRNA resulted

in a strong decrease of endogenous DYRK1A protein

levels (Fig 2A) These siRNA duplexes had no effect

on the expression of the most closely related family

member DYRK1B (85% identical amino acids in the catalytic domain) [3] when this kinase was expressed as

a green fluorescent protein (GFP) fusion protein,

– p-Casp9 (T125) Casp9

Actin

p-ERK1/2 ERK1/2

PD U0 Alst PA Rosc

Casp9 p-ERK1/2 ERK1/2 Actin

– PD U0

TPA

– p-Casp9 (T125)

Actin Casp9 p-Casp9 (T125)

Nocodazole

– Alst PA Rosc –

H.sapiens M.musculus R.norvegicus

Fig 1 Small-molecule inhibitors suggest DYRK1A as a potential kinase that phosphorylates caspase 9 on Thr125 (A) U2.C9–C287A cells stably expressing caspase 9(C287A) were serum-starved and treated with protein kinase inhibitors for 30 min as indicated (B) U2.C9–C287A cells were serum-starved and preincubated with inhibitors 30 min prior to addition of TPA for 15 min or (C) incubated with nocodazole for

16 h before addition of inhibitors for 15 min Inhibitors used were PD0325901 (PD; 0.1 l M) , U0126 (U0; 10 l M) , alsterpaullone (Alst; 10 l M) , purvalanol A (PA; 10 l M) or roscovitine (Rosc; 20 l M) Cell lysates were probed on blots with antibodies against the specified proteins (D) Comparison of the amino acid sequence surrounding Thr125 of human, mouse and rat caspase 9 with the reported DYRK and DYRK1A consensus phosphorylation sequences [22,23] Phosphorylated residues are highlighted in bold.

Actin DYRK1A

siRNA

GFP-DYRK1B-p69

Actin siRNA

siRNA

Actin

Casp9 p-ERK1/2 p-Casp9 (T125)

ERK1/2

Actin

Casp9 p-ERK1/2 p-Casp9 (T125)

ERK1/2

siRNA

Fig 2 DYRK1A depletion inhibits phosphorylation of caspase 9 on Thr125 in cells (A) Transfection with two DYRK1A-targeting siRNA duplexes ablates endogenous DYRK1A protein U2.C9–C287A cells were transfected with siRNA duplexes targeting DYRK1A (1A#1 and 1A#2) or luciferase (Luc) as control, serum-starved for 48 h after transfection and lysed 24 h later Endogenous DYRK1A was precipitated from cell lysates using Ni 2+ -NTA agarose (B) Transfection of U2.C9–C287A cells with DYRK1A targeting siRNA duplexes (1A#1 and 1A#2) has no effect on protein levels of overexpressed GFP–DYRK1B–p69, whereas transfection with DYRK1B targeting siRNA (1B) decreases GFP–DYRK1B–p69 protein levels siRNA transfections were carried out 24 h prior to DNA transfections Cells were lysed 72 h after siRNA transfections (C) Depletion of DYRK1A decreases Thr125 phosphorylation in serum-starved U2.C9–C287A cells Transfections were carried out as in (A) (D) Depletion of DYRK1A decreases Thr125 phosphorylation in U2.C9–C287A cells in the presence of serum Transfections were carried out as in (A) In all cases, proteins were detected on immunoblots probed with the indicated antibodies.

cating their specificity for DYRK1A (Fig 2B)

Deple-tion of DYRK1A by siRNA significantly inhibited

basal Thr125 phosphorylation in cells both in the

absence (Fig 2C) and in the presence (Fig 2D) of

serum, demonstrating a role for DYRK1A in the

phos-phorylation of caspase 9 in cells DYRK1A depletion

did not impinge on ERK1⁄ 2 phosphorylation

(Fig 2C,D) nor did it prevent the phosphorylation of

caspase 9 at Thr125 induced by the phorbol ester TPA

(data not shown), showing that DYRK1A is not

required for the ERK1⁄ 2-dependent phosphorylation of

caspase 9 In contrast to the effect of depleting

DYRK1A, a DYRK1B-specific siRNA (see Fig 2B for

validation of knockdown efficiency) had no effect on

Thr125 phosphorylation (Fig 2C,D) Therefore, we

conclude that DYRK1A is required for basal phosphor-ylation of caspase 9 at Thr125 and we can exclude DYRK1B as a significant Thr125 kinase in U2OS cells

Direct phosphorylation of caspase 9 by DYRK1A

To establish whether DYRK1A can phosphorylate cas-pase 9 directly, we carried out an in vitro kinase assay using [32P]ATP[cP] and catalytically inactive His6– caspase 9(C287A) as a substrate Active DYRK1A produced in bacteria catalysed the incorporation of radiolabelled phosphate into caspase 9, whereas a cas-pase 9 mutant in which Thr125 was mutated to alanine (T125A) was not phosphorylated (Fig 3A) Thus, DYRK1A phosphorylates caspase 9 directly in vitro

Casp9 FLAG IP: -FLAG

DYRK WT DYRK K188R C9 C9 + DYRK WT C9 + DYRK K188R C9 T125A C9 T125A + DYRK WT

Casp9 FLAG Actin Input

p-Casp9 (T125) Casp9 p-ERK1/2 ERK1/2

Actin FLAG

FLAG-DYRK1A EV

p-Casp9 (T125) Casp9 p-ERK1/2 ERK1/2

Actin GFP

GFP-DYRK1A GFP-DYRK1B

Autorad Coomassie

+DYRK1A

Fig 3 DYRK1A interacts with caspase 9 and induces its phosphorylation on Thr125 in cells (A) DYRK1A directly phosphorylates caspase 9

at Thr125 Recombinant His 6 –caspase 9 (His–C9) or His 6 –caspase 9(T125A) (both containing the catalytically inactivating C287A mutation) was incubated with recombinant DYRK1A in the presence of [ 32 P]ATP[cP] as indicated Samples were analysed by SDS ⁄ PAGE, followed by autoradiography (B) Expression of DYRK1A causes Thr125 phosphorylation in cells U2.C9–C287A cells were transiently transfected with empty vector (EV) or wild-type (WT), K188R or Y321F mutant DYRK1A in pcDNA3–FLAG (C) Expression of DYRK1B also induces Thr125 phosphorylation U2.C9–C287A cells were transiently transfected with empty vector (EV), pEGFP–DYRK1A or pEGFP–DYRK1B-p69 (D) DYRK1A interacts with caspase 9 in cells U2OS cells were co-transfected with empty vector, wild-type (WT) or K188R DYRK1A

in pcDNA3–FLAG and caspase 9(C287A) (C9) or caspase 9(T125A ⁄ C287A) in pcDNA3 A portion of each cell lysate was retained as an input sample and FLAG-immunoprecipitations were performed on the remainder *Indicates bands resulting from IgG; arrows indicate bands of interest In (B–D), cell lysates were prepared 24 h after transfection and proteins were detected on immunoblots probed with the indicated antibodies.

and that Thr125 is the sole phosphorylation site on

caspase 9 targeted by DYRK1A

Expression of exogenous FLAG-tagged DYRK1A

in U2.C9–C287A cells caused a strong increase in the

phosphorylation of caspase 9 on Thr125 (Fig 3B),

confirming the ability of DYRK1A to target caspase 9

in cells The FLAG–DYRK1A mutants K188R or

Y321F, the catalytic activity of which is reduced due

to a mutation in the ATP-binding site or the

activa-tion-loop, respectively [26,27], did not result in strong

Thr125 phosphorylation (Fig 3B), confirming that the

protein kinase activity of DYRK1A is required

Expression of DYRK1B as a GFP fusion protein

caused strong Thr125 phosphorylation like DYRK1A

(Fig 3C), showing that DYRK1B is capable of

cataly-sing Thr125 phosphorylation even though it is not

responsible for basal Thr125 phosphorylation in U2OS

cells

Many protein kinases interact with their substrates

in complexes that can be co-precipitated For example,

caspase 9 associates with CDK1–cyclin B1 in cells

dur-ing G2 and mitosis [18] To test for the interaction

between caspase 9 and DYRK1A, U2OS cells were

transiently co-transfected with expression vectors

encoding caspase 9 and FLAG–DYRK1A or FLAG–

DYRK1A(K188R) Both wild-type and K188R mutant

FLAG–DYRK1A were able to precipitate caspase 9,

indicating caspase 9 and DYRK1A do indeed

asso-ciate in cells, but DYRK1A kinase activity is not

required Furthermore, caspase 9(C287A⁄ T125A)

lack-ing Thr125 still precipitated with FLAG–DYRK1A

(Fig 3D), showing that Thr125 and its

phosphoryla-tion are dispensable for the interacphosphoryla-tion Taken together

with the ability of DYRK1A to catalyse the

phospho-rylation of caspase 9 at Thr125 in vitro, these results

strongly indicate that DYRK1A targets caspase 9

directly in cells

Harmine is a potent inhibitor of DYRK1A in cells

The b-carboline alkaloid harmine has recently been

reported as a specific DYRK inhibitor in vitro by Bain

et al [19] Harmine inhibits purified DYRK1A in the

nanomolar range, with DYRK2 and DYRK3 inhibited

 10-fold less potently, and little or no inhibition by

1 lm harmine of a panel of 67 other protein kinases

[19] We found that the basal phosphorylation of

Thr125 in U2.C9–C287A cells that is due to DYRK1A

was potently inhibited by harmine, with a partial

inhi-bition even at a concentration of 0.01 lm and almost

complete inhibition at 1 lm (Fig 4A) We did not

observe any impairment of the basal phosphorylation

of ERK1⁄ 2 by harmine, excluding the possibility that

the reduction of Thr125 phosphorylation is due to defective ERK1⁄ 2 activation (Fig 4A) In agreement,

we also found no effect of 1 lm harmine on TPA-induced ERK1⁄ 2 activation and the subsequent phos-phorylation of Thr125 (Fig 4B)

A previous study identified harmine as an inhibitor

of CDKs at micromolar concentrations, with an

IC50= 17 lm for CDK1–cyclin B [28] Bain et al [19], however, demonstrated no significant inhibitory activity against another cyclin-dependent kinase, CDK2–cyclin A, at 1 lm harmine Consistent with the latter study, we observed that 1 lm harmine had no effect on CDK1–cyclin B1-dependent Thr125 phos-phorylation induced by nocodazole (Fig 4C) These results confirm that harmine selectively inhibits the basal phosphorylation of caspase 9 at Thr125 that is due to DYRK1A, but not the ERK1⁄ 2-dependent phosphorylation induced by mitogens or the CDK1– cyclin B1-dependent phosphorylation induced by mitotic arrest When the phosphorylation of Thr125 in caspase 9 was strongly induced by the expression of FLAG–DYRK1A, this activity was also completely inhibited by harmine, although higher concentrations

of the inhibitor were required, presumably because of the elevated levels of DYRK1A in the cells (Fig 4D)

We also analysed the effect of 1 lm harmine on the phosphorylation of Thr125 on endogenous caspase 9 immunoprecipitated from HeLa cells Immunoblots showed a reduction of Thr125 phosphorylation

in response to harmine, confirming that endogenous caspase 9 is also phosphorylated in a DYRK1A-depen-dent manner (Fig 4E)

DYRKs are unusual dual-specificity kinases that require autophosphorylation of an essential Tyr–Xaa– Tyr motif in the activation loop to form a mature kinase that has specificity towards serine and threonine residues in substrate proteins [21,26,29] Therefore, we were interested in studying whether harmine, like pur-valanol A [21], inhibits the autophosphorylation of DYRK1A on tyrosine in addition to the phosphoryla-tion of exogenous substrates Using an assay developed

by Lochhead et al [21], we translated FLAG– DYRK1A in rabbit reticulocyte lysate in the absence

or presence of harmine and found that harmine also blocks the tyrosine autophosphorylation of DYRK1A, with almost complete inhibition at 1 lm (Fig 4F)

Inhibition of caspase 9 auto-processing

by DYRK1A Previous work has shown an inhibitory effect of Thr125 phosphorylation on caspase 9 activation, thereby blocking downstream caspase 3 activation and

apoptosis [14,18] Although proteolytic processing of

caspase 9 is not required for its activation, generation

of a processed p35 form is dependent on the catalytic

activity of the enzyme [30] When we expressed

catalyt-ically active caspase 9 in U2OS cells, generation of the

p35 auto-processed form was significantly reduced by

co-expression of DYRK1A (Fig 5) Auto-processing

of caspase 9 was not antagonized by DYRK1A if the

Thr125 residue of caspase 9 was converted to alanine

(Fig 5A) Furthermore, inhibition of caspase 9

auto-processing required the kinase activity of DYRK1A,

because co-expression of the DYRK1A–K188R

mutant did not block caspase 9 auto-processing

(Fig 5B) Thus, DYRK1A inhibits the auto-processing

of caspase 9 in cells through the phosphorylation of

Thr125 on caspase 9 This result indicates that phos-phorylation of Thr125 by DYRK1A inhibits caspase 9 activation, consistent with previous studies on the effects of phosphorylation of this site by other kinases [14,18]

DYRK1A phosphorylates caspase 9 in the nucleus DYRK1A has been reported as a predominantly nuclear kinase [2] We therefore wished to determine whether DYRK1A and caspase 9 co-localize As anti-cipated, DYRK1A expressed as a fusion protein with GFP was found to be predominantly nuclear by con-focal microscopy (Fig 6A) When caspase 9 was expressed as the catalytically inactive mutant C287A in

p-Casp9(T125)

Actin

Harmine( M ) – 1 0.1 0.01 0.001

Casp9 p-ERK1/2 ERK1/2

FLAG-DYRK1A EV

p-Casp9(T125)

Actin Flag Casp9

p-Tyr FLAG-DYRK1A

EV FLAG-DYRK1A Harmine PA

Inhibitor ( M )

Actin

Nocodazole

Casp9 p-Casp9(T125)

p-Casp9(T125)

Actin

Harmine( M ) – 1 0.1 0.01 0.001

TPA

Casp9 p-ERK1/2 ERK1/2

p-Casp9(T125) Casp9 Casp9 Actin

IP

Input

-caspase-9 1

–

Harmine( M ) – –

Harmine( M ) – –

Harmine( M ) – –

– – –

Fig 4 Harmine inhibits DYRK1A-dependent phosphorylation of caspase 9 on Thr125 in cells (A) Harmine inhibits the basal phosphorylation

of Th125 Serum-starved U2.C9–C287A cells were treated with indicated concentrations of harmine for 30 min (B) Harmine does not inhibit ERK1 ⁄ 2 activation or TPA-induced Thr125 phosphorylation Serum-starved U2.C9–C287A cells were incubated with harmine for 15 min prior

to addition of TPA for further 15 min (C) Harmine does not inhibit mitotic phosphorylation of Thr125 U2.C9–C287A cells were incubated in the presence of nocodazole for 16 h prior to addition of harmine for 30 min (D) Harmine inhibits Thr125 phosphorylation caused by DYRK1A overexpression in cells U2OS cells were transfected with pcDNA3–caspase 9(C287A) and pcDNA3–FLAG–DYRK1A or empty vector (EV) Twenty hours after transfection, cells were incubated with indicated concentrations of harmine for 30 min (E) HeLa cells were incubated in the presence of 1 l M harmine where indicated for 30 min, followed by immunoprecipitation of endogenous caspase 9 using a caspase 9-spe-cific antibody Mock immunoprecipitations were carried out using an anti-Myc IgG (F) FLAG–DYRK1A was in vitro translated in rabbit reticulo-cyte lysate in the absence or presence of indicated concentrations of harmine or purvalanol A (PA), followed by immunoprecipitation using anti-(FLAG agarose) EV indicates empty vector In all cases, proteins were detected on immunoblots probed with the indicated antibodies.

U2OS cells, it localized to both the cytoplasm and the

nucleus Nuclear speckle-like foci were detected by

the pThr125 antibody in the absence of transfected

caspase 9 (Fig 6A) These speckles were not removed

by siRNA-mediated depletion of endogenous caspase 9

(Fig S1); therefore the speckles are not likely to

corre-spond to phosphorylated caspase 9 and probably

origi-nate from another epitope However, in cells in which

caspase 9 was co-expressed with DYRK1A, the signal

detected by the pThr125 antibody was strongly

increased in the nucleus, and pThr125 epitopes also

appeared in the cytoplasm The increased signal

gener-ated by DYRK1A expression was due entirely to the

phosphorylation of caspase 9 at Thr125, because no

increased signal was detected when cells were

co-trans-fected with a non-phosphorylatable mutant of

cas-pase 9(T125A) (Fig 6A) The increase in Thr125

phosphorylation also required the kinase activity of

DYRK1A, because it was not induced by the

ATP-binding site mutant K188R (data not shown)

To test if caspase 9 phosphorylation takes place in the nucleus where DYRK1A is localized, we engi-neered GFP–caspase 9 fusion constructs tagged with either a nuclear localization signal (NLS) or a nuclear export signal (NES) When expressed in cells, the NLS–GFP–caspase 9 and NES–GFP–caspase 9 pro-teins localized to the nucleus and cytoplasm, respec-tively (Fig 6B) When co-expressed with DYRK1A, GFP–caspase 9 and NLS–GFP–caspase 9 exhibited a higher level of Thr125 phosphorylation than NES– GFP–caspase 9 (Fig 6C) In agreement with this result, we also found that the endogenous basal Thr125 kinase had a stronger preference for nuclear caspase 9 than for cytoplasmic caspase 9 (Fig 6D) This nuclear kinase activity towards caspase 9 was sen-sitive to harmine and thus due to DYRK1A (Fig S2) Together, these results demonstrate that DYRK1A and caspase 9 co-localize to the nucleus and that DYRK1A phosphorylates caspase 9 in the nuclear compartment

Discussion The dual-specificity tyrosine phosphorylation-regulated protein kinase DYRK1A plays important roles during development and in human pathologies However, lit-tle is currently known about the substrates through which it exerts these effects Here, we have identified the apoptotic protease caspase 9 as a substrate for phosphorylation by DYRK1A at a critical site, Thr125 Previously we have shown that phosphoryla-tion of this site inhibits the activaphosphoryla-tion of caspase 9 and restrains apoptosis in human cells [14,18] We propose that basal phosphorylation of Thr125 in caspase 9 by DYRK1A sets a threshold in the response to apoptotic stimuli that is augmented in proliferating cells through the activities of ERK1⁄ 2 and CDK1–cyclin B1 kinases [14,18]

Although DYRK1A appears to be synthesized as a constitutively active enzyme, work on the cytoplasmic Caenorhabditis elegans DYRK orthologue MBK-2 has shown a cell-cycle and developmental stimulus-depen-dent regulation of DYRK activity [31], and DYRK1A may also be regulated through alterations of its expres-sion during development and the cell cycle in mamma-lian cells [32,33] Thus, the level of phosphorylation of caspase 9 catalysed by DYRK1A and its significance for cell survival is likely to be modulated by changes

in DYRK1A expression in vivo

DYRK1Ahas been mapped to the Down’s syndrome critical region on chromosome 21 that is present as an additional copy in Down’s syndrome individuals [5,6] DYRK1A is overexpressed in Down’s syndrome brains

p-Casp9(T125)

Casp9

Actin

WT

FLAG

Procaspase9 Processed Casp9 Procaspase9 Processed Casp9

T125A

A

B

p-Casp9(T125)

Casp9

Actin DYRK1A

DYRK1A WT DYRK1A K188R

Procaspase9 Processed Casp9 Procaspase9 Processed Casp9

Fig 5 Phosphorylation of Thr125 by DYRK1A inhibits caspase 9

activation (A) U2OS cells were transfected with pcDNA3 vectors

encoding catalytically active caspase 9 (wild-type; WT) or

catalyti-cally active caspase 9(T125A) and FLAG–DYRK1A or empty vector

(EV) (B) U2OS cells were transfected with pcDNA3 vectors

encod-ing catalytically active caspase 9 (wild-type) and wild-type (WT) or

K188R FLAG–DYRK1A or empty vector (EV) Cell lysates were

prepared 7 h after transfection and blotted with the indicated

anti-bodies.

D

Casp9/

GFP-DYRK1A

Casp9/

GFP

Casp9(T125A)/

GFP

Casp9(T125A)/

GFP-DYRK1A

Casp9 Casp9

Casp9

Casp9

p-Casp9(T125) p-Casp9(T125)

p-Casp9(T125)

p-Casp9(T125)

GFP-DYRK1A GFP

GFP-DYRK1A GFP

DNA DNA

DNA DNA

NES-GFP-Casp9

NLS-GFP-Casp9

GFP-Casp9

GFP DNA

p-Casp9 (T125) GFP-Casp9 Actin FLAG

FLAG-DYRK1A

EV

p-Casp9(T125) GFP-Casp9 Actin

Fig 6 Phosphorylation of caspase 9 on Thr125 by DYRK1A in the nucleus (A) Immunofluorescence staining of U2OS cells transiently trans-fected with vectors encoding caspase 9(C287A), caspase 9(T125A ⁄ C287A), GFP or GFP–DYRK1A Cells were fixed 20–24 h after transfec-tion and stained with antibodies directed against total caspase 9 and caspase 9 phosphorylated on Thr125 DNA was DAPI-stained and cells were analysed by confocal microscopy Scale bars, 10 lm (B) Localization of GFP–caspase 9(C287A), NES–GFP–caspase 9(C287A) and NLS–GFP–caspase 9(C287A) in U2OS cells Cells were transiently transfected with pEGFP vectors encoding the respective fusion proteins, followed by fixation after 8–9 h Scale bars, 10 lm (C) Overexpressed DYRK1A predominantly phosphorylates nuclear caspase 9 U2OS cells were transfected as in (B), but in combination with empty vector (EV) or DYRK1A in pcDNA3–FLAG and lysed 8–9 h after transfection (D) Endogenous Thr125 kinase(s) preferentially phosphorylate(s) nuclear caspase 9 U2OS cells were transfected as in (B) In (C, D), cell lysates were blotted with antibodies against the specified proteins Note that samples in (C) were harvested 8–9 h after transfection, whereas cells in (D) were lysed 20–24 h after transfection This difference in expression time accounts for the absence of a p-Casp-9(T125) signal in lanes 1–3 in (C).

[34], suggesting a role in neurogenesis like its Drosophila

orthologue, minibrain (mnb) [4] Studies in both mouse

and Drosophila have found an important role for

DYRK1A⁄ minibrain kinase in determining the number

of neurons during post-embryonic neurogenesis:

muta-tion of minibrain causes reducmuta-tion of the size of the

optic lobes and central brain hemispheres [4], whereas

mice lacking one copy of the DYRK1A gene exhibit

region-specific reductions in brain size [35] Although

the molecular mechanisms underlying this phenotype

are not understood, regulation of apoptosis might be

involved This idea is particularly appealing because

caspase 9 also has an essential function in mouse brain

development [36,37]

DYRK1A is a predominantly nuclear kinase that is

localized to intranuclear splicing speckles, which are

sites of mRNA processing [38] Our pThr125 antibody

also detects these speckles (Fig 6; data not shown),

although the phosphoepitope is probably not due to

caspase 9 (Fig S1) This study shows that caspase 9 is

partially localized to the nucleus and its

phosphoryla-tion by DYRK1A is promoted by nuclear targeting

and diminished by cytoplasmic targeting Previously,

although caspase 9 was reported to be mainly localized

to mitochondria and cytosol when analysed by

subcel-lular fractionation of Jurkat cells [39], GFP–caspase 9

was found partially localized to nuclei in Jurkat and

HEK293 cells [40] Nuclear caspase 9 has also been

observed in mammary epithelial cells [41] Our results

show that caspase 9 is distributed in both the nucleus

and cytoplasm in U2OS cells Caspase 9 would

encounter DYRK1A in the nucleus and become

phos-phorylated, before being redistributed to the

cyto-plasm In this way, a nuclear kinase, DYRK1A, can

regulate the cytoplasmic activity of caspase 9 as an

ini-tiator of apoptosis It does, however, remain possible

that caspase 9 has a distinct function within the

nucleus that is controlled by DYRK1A

Identification of caspase 9 as a bona fide substrate

for DYRK1A in cells has enabled us to confirm the

b-carboline alkaloid harmine as an intracellular

inhibi-tor of DYRK1A, as suggested by its identification as a

potent and selective inhibitor of DYRKs in vitro [19]

b-Carbolines are present in Peganum harmala and

other plants which have been used as medicinal

pre-parations as well as hallucinogens in traditional rituals

Harmine has a long history of use as a

chemothera-peutic drug for a number of diseases, including

malar-ial infection and Parkinson’s disease [42] Harmine and

related b-carbolines have cytotoxic activity towards

human tumour cell lines in culture [43], suggesting a

possible use in anti-cancer therapy Several putative

molecular targets for harmine have been identified

[28,44], but inhibition of DYRK1A at low (sub-micro-molar) concentrations in cells strongly suggests that inhibition of this kinase is involved in the biological activity of harmine in vivo Interestingly, our results show that harmine not only inhibits the protein–ser-ine⁄ threonine kinase activity of the mature enzyme, but also the tyrosine autophosphorylation that is required for maturation of the active enzyme [21] This indicates that harmine also inhibits formation of active DYRK1A in cells Identification of harmine as a cell-permeable DYRK1A inhibitor is anticipated to facili-tate the identification of further DYRK1A substrates

in vivoand also suggests its potential use to reverse the pathological effects of DYRK1A overexpression

Experimental procedures

Plasmids and recombinant proteins Caspase 9 cDNA in pcDNA3 (Invitrogen, Carlsbad, CA, USA) or pET28a (Novagen, Madison, WI, USA) has been described previously [14] To generate an expression con-struct for GFP–caspase 9 fusion protein, caspase 9 cDNA was subcloned into pEGFP(C2) Vectors encoding NLS– GFP–caspase 9 and NES–GFP–caspase 9 were constructed

by insertion of the SV40T NLS or the NES from the pro-tein kinase A inhibitor between the initiating ATG and the second codon of EGFP using the QuikChange site-directed mutagenesis kit Wild-type and K188R mutant pEGFP(C1)–DYRK1A (rat) as well as pEGFP(C1)– DYRK1B-p69 (human) were kind gifts from W Becker (Aachen, Germany) An expression construct for FLAG– DYRK1A was generated by subcloning DYRK1A cDNA into pcDNA3–FLAG (kindly provided by D Meek, University of Dundee, UK) Expression of recombinant His6–caspase 9(C287A) and His6–caspase 9(T125A⁄ C287A) proteins was carried out as described previously [14] All expression constructs encoding proteins bearing amino acid substitutions were generated by site-directed mutagenesis using the QuikChange kit (Stratagene, Cedar Creek, TX, USA) according to the manufacturer’s instructions Recom-binant DYRK1A, expressed as a fusion protein with gluta-thione S-transferase (GST–DYRK1A) in Escherichia coli, was purchased from Millipore (Watford, UK)

Antibodies and reagents The following antibodies were used for western blotting and immunological staining according to standard proto-cols: caspase 9 mAb (Chemicon, Temecula, CA, USA), phospho-ERK1⁄ 2, phospho-Tyr-100 (both Cell Signalling Technology, Beverly, MA, USA), ERK1⁄ 2 (Millipore), GFP, DYRK1A G-19 (both Santa Cruz Biotechnology, Santa Cruz, CA), Actin, FLAG-M2 (both Sigma-Aldrich,

St Louis, MO, USA), myc-9E10 (Cancer Research UK,

London, UK) Generation and characterization of rabbit

anti-[phospho-caspase 9(T125)] IgG was described

previ-ously [14] Reagents used were: nocodazole (Sigma), TPA

and protein kinase inhibitors (Calbiochem, San Diego, CA,

USA), harmine (Sigma-Aldrich) and PD0325901 (kindly

provided by P Cohen, University of Dundee, UK)

DYRK1A kinase assay

Recombinant His6–caspase 9 (1.5 lg) was added to a total

reaction volume of 15 lL kinase assay buffer (50 mm Tris

pH 7.5, 10 mm MgCl2, 100 lm ATP, 1 mm dithiothreitol)

containing 1.5 lCi [32P]ATP[cP] (from a 10 mCiÆmL)1

stock with specific activity 3000 CiÆmmol)1) The kinase

assay was initiated by adding 30 ng of active recombinant

GST–DYRK1A Reaction mixtures were incubated at

30C for 30 min Reactions were stopped by boiling in

SDS⁄ PAGE sample buffer and half the volume of a

reac-tion was analysed by SDS⁄ PAGE, followed by

autoradio-graphy

Cell culture, DNA transfections and treatments

HeLa and U2OS cells (obtained from Cancer Research UK

Cell Services) were maintained in Dulbecco’s modified

Eagle’s medium supplemented with 10% fetal bovine

serum, 50 UÆmL)1 penicillin, 50 lgÆmL)1 streptomycin and

2 mm l-glutamine (Invitrogen, Carlsbad, CA) For U2.C9–

C287A cells [18], which stably express catalytically inactive

caspase 9(C287A), the growth medium was supplemented

with G418 sulfate (400 ngÆmL)1, Calbiochem) Where

indi-cated, serum starvation was performed by culturing cells in

Dulbecco’s modified Eagle’s medium containing 0% fetal

bovine serum for 20–24 h DNA transfections were carried

out using CsCl-purified plasmid DNA and Superfect

(Qia-gen, Valencia, CA, USA) according to manufacturer’s

pro-tocol To arrest cells in mitosis, cells were treated with

100 ngÆmL)1 nocodazole for 16 h To activate ERK1⁄ 2

MAPK signalling, cells were incubated with 1 lm TPA for

15 min The protein kinase inhibitors PD0325901 (0.1 lm),

U0126 (10 lm), alsterpaullone (10 lm), purvalanol A

(10 lm), roscovitine (20 lm) or harmine (routinely, 1 lm)

were added as indicated The specificity of these inhibitors

towards a panel of purified protein kinases is reported by

Bain et al [19] For analysis by immunoblotting, cells were

lysed in SDS⁄ PAGE sample buffer

RNA interference

For siRNA transfections, cells were transfected with

100 nm siRNA duplex and Lipofectamine 2000 following

the manufacturer’s instructions (Invitrogen) Then, cells

were trypsinised and cultured for 72 h before analysis The

following siRNA duplexes were used to deplete DYRK1A (sense strands): 5¢-UAAGGAUGCUUGAUUAUGAdTdT-3¢ (DYRK1A#1), 5¢-AAACUCGAAUUCAACCUUAdTdT-3¢ (DYRK1A#2) Other siRNAs used were 5¢-CGUACG CGGAAUACUUCGAdTdT-3¢ (Luciferase), 5¢-CGACCU GACUGCCAAGAAAdTdT-3¢ (Caspase 9) and DYRK1B SmartPool (Dharmacon) comprising four different duplexes: 5¢-GAAAUUGACUCGCUCAUUGrUrU-3¢, 5¢-ACACGG AGAUGAAGUACUArUrU-3¢, 5¢-GCCAGAGGAUCUA CCAGUArUrU-3¢, 5¢-GCACAUCAAUGAGGUAUACr UrU-3¢ Single siRNA duplexes were synthesized by MWG (Martinsried, Germany)

Caspase 9 and FLAG immunoprecipitations Cells were lysed in IP buffer (20 mm Tris pH 7.6, 137 mm NaCl, 2 mm EDTA, 1 mm Na3VO4, 50 mm NaF, 5 mm b-glycerophosphate, 1% Triton X-100, 1 lm okadaic acid,

1 mm phenylmenthanesulfonyl fluoride, 1 lgÆmL)1 each aprotinin, leupeptin and pepstatin A) Immunoprecipitation

of endogenous caspase 9 from HeLa cells was carried out

as described previously [14] For co-immunoprecipitation of FLAG–DYRK1A and caspase 9 from U2OS cells, cell lysate (0.5 mg) was incubated with 15 lL anti-(FLAG aga-rose) (Sigma) for 1 h at 4C Beads were washed three times in IP buffer and boiled in SDS⁄ PAGE sample buffer Samples were analysed by western blotting

Ni2+-pulldown of endogenous DYRK1A from cells DYRK1A contains an internal stretch of 13 consecutive histidine residues, enabling the endogenous DYRK1A pro-tein to bind Ni2+-NTA agarose [45] U2.C9–C287A cells were lysed in buffer A (6 m guanidine–HCl, 10 mm Tris, 0.1 m phosphate buffer, pH 8.0) supplemented with 5 mm imidazole for 5 min, sonicated and incubated with 30 lL Ni-NTA–agarose (Qiagen) for 4–5 h Beads were pelleted and washed once in buffer A, followed by one wash in buf-fer B (8 m urea, 10 mm Tris, 0.1 m phosphate bufbuf-fer,

pH 8.0), one wash in buffer C (8 m urea, 10 mm Tris, 0.1 m phosphate buffer pH 6.5, 0.2% Triton X-100) and one wash in buffer D (buffer C supplemented with 0.1% Tri-ton X-100) For elution, beads were boiled in SDS⁄ PAGE sample buffer Supernatant was analysed by western blot-ting and endogenous DYRK1A was detected with a DYRK1A-specific antibody

In vitro translation and immunoprecipitation of FLAG–DYRK1A from rabbit reticulocyte lysate

In vitro translation of FLAG–DYRK1A was performed in absence or presence of inhibitors using the TNT Quick cou-pled transcription⁄ translation protocol (Promega, Madison,

WI, USA) and pcDNA3–FLAG–DYRK1A as template