Research Paper 51

β and CK1 by
Inhibition of cyclin-dependent kinases, GSK-3β
hymenialdisine, a marine sponge constituent
L Meijer1, A-MWH Thunnissen2*, AW White3, M Garnier1, M Nikolic4†,
L-H Tsai4, J Walter5, KE Cleverley6, PC Salinas6, Y-Z Wu7, J Biernat7,
E-M Mandelkow7, S-H Kim2 and GR Pettit3

Background: Over 2000 protein kinases regulate cellular functions. Screening for
inhibitors of some of these kinases has already yielded some potent and selective
compounds with promising potential for the treatment of human diseases.
Results: The marine sponge constituent hymenialdisine is a potent inhibitor of
cyclin-dependent kinases, glycogen synthase kinase-3β and casein kinase 1.
Hymenialdisine competes with ATP for binding to these kinases. A
CDK2–hymenialdisine complex crystal structure shows that three hydrogen
bonds link hymenialdisine to the Glu81 and Leu83 residues of CDK2, as
observed with other inhibitors. Hymenialdisine inhibits CDK5/p35 in vivo as
demonstrated by the lack of phosphorylation/down-regulation of Pak1 kinase in
E18 rat cortical neurons, and also inhibits GSK-3 in vivo as shown by the
inhibition of MAP-1B phosphorylation. Hymenialdisine also blocks the in vivo
phosphorylation of the microtubule-binding protein tau at sites that are
hyperphosphorylated by GSK-3 and CDK5/p35 in Alzheimer’s disease
(cross-reacting with Alzheimer’s-specific AT100 antibodies).
Addresses: 1CNRS, Station Biologique, BP 74,
29682 Roscoff cedex, Bretagne, France. 2University
of California, Department of Chemistry, Berkeley, CA,
USA. 3Arizona State University, Cancer Research
Institute, Tempe, AZ, USA. 4Howard Hughes Medical
Institute, Department of Pathology, Harvard Medical
School, 200 Longwood Avenue, Boston, MA, USA.
5Department of Molecular Biology, Central Institute of
Mental Health, 68159 Mannheim, Germany. 6The
Randall Institute, Developmental Biology Research
Centre, King’s College London, 26–29 Drury Lane,
London WC2B 5RL, UK. 7Max-Planck Unit for
Structural Molecular Biology, Notkestrasse 85,
D-22603 Hamburg, Germany.
Present addresses: *Laboratory of Biophysical
Chemistry, University of Groningen, Nijenborgh 4,
9747 AG Groningen, The Netherlands. †Department
of Experimental Pathology, GKT School of Medicine,
King’s College London, London, SE1 9RT, UK.

Conclusions: The natural product hymenialdisine is a new kinase inhibitor with Correspondence: Laurent Meijer
E-mail: meijer@sb-roscoff.fr
promising potential applications for treating neurodegenerative disorders.
Key words: Alzheimer’s disease, cyclin-dependent
kinases, glycogen synthase kinase, hymenialdisine, tau

Received: 4 August 1999

Revisions requested: 14 September 1999
Revisions received: 27 September 1999
Accepted: 6 October 1999

Published: 20 December 1999

Publication delayed at the author’s request

Chemistry & Biology 2000, 7:51–63

1074-5521/00/$ – see front matter

© 2000 Elsevier Science Ltd. All rights reserved.

Introduction filaments (PHFs) that consist mainly of hyperphosphory-

Alzheimer’s disease (AD) affects 5–10% of the population
over 65 years of age. The dementia associated with AD
results from the selective death of neurons, which is associ-
ated with several anatomo-pathological hallmarks such as
senile neuritic plaques and neurofibrillary tangles. Three
molecular actors clearly play a role in the development of
AD: the amyloid β peptide (Aβ), presenilins-1 and -2 and
the microtubule-associated protein tau. The senile neuritic
plaques form an extracellular core of deposited amyloid β
peptide, derived from proteolytic cleavage of the amyloid
precursor protein (β-APP; reviewed in [1]). The neurofib-
rillary tangles are intracellular aggregates of paired helical
lated tau proteins. Hyperphosphorylation occurs on more
than 20 sites and is carried out by two proline-directed
kinases, glycogen synthase kinase-3β (GSK-3β) and cyclin-
dependent kinase 5 (CDK5/p35) (reviewed in [2,3]), and
possibly also by casein kinase 1 (CK1) [4]. Finally, muta-
tions in the transmembrane proteins presenilin genes are
the most common cause of early onset familial AD
(reviewed in [1,5]).
The links between the three proteins, Aβ, presenilins and
tau, have remained elusive until recently (reviewed in
[5,6]). Presenilins appear to be unusual aspartyl proteases
52 Chemistry & Biology 2000, Vol 7 No 1

Figure 1

NH2 NH2 O NH2

Structure of hymenialdisine and related
N N NH
Br N metabolites isolated from marine sponges.
O HN
NH NH NH
R2 O Br O
Br
NH NH
N
R1 Br Br
NH2
Br
NH NH NH NH
NH
NH NH NH
Br
O O O NH 6 Dibromoageliferin
R1 R2 O
4 Stevensine/ odiline 5 Axinohydantoin R
1 Br H Hymenialdisine
2 H H debromohymenialdisine O
3 Br Br 3-bromohymenialdisine CH3
NH2 N
OH R
N N NH NH2
O O NH
NH Br NH NH
Br Br
NH NH R
N N Br N N
NH
Br N NH2
N NH H R
O NH NH
O
O O 10 Dibromocantharelline 7 Sceptrins
8 Dibromophakellstatin 9 Agelastatin A R= H / Br
R2 NH2 Br
N
Br NH2 NH
R1 NH O + COO
-
NH NH NH N
Br N O
NH R1 R2 O
NH
12 H H clathrodin 15 Agelongine
O O 13 H Br hymenidin
14 Br Br oroidin
11 Dispacamide A
Chemistry & Biology

that activate themselves autocatalytically. They cleave MAP-1B (reviewed in [2,17]), Pak1 kinase [18] and neuro-
β-APP to generate Aβ [7], particularly the most aggrega- filament subunits [19]. Intensive screening has identified
tion-prone and neurotoxic amyloid β peptide, Aβ42. Prese- a series of chemical CDK inhibitors (reviewed in [20–22]),
nilins also directly associate with β-catenin [8,9] and such as olomoucine [23], roscovitine [24,25], purvalanol
GSK-3β [10], two components of the Wnt signalling [26], flavopiridol [27], indirubins [28] and paullones
pathway. Activation of this pathway leads to inhibition of [29,30]. Some of these compounds display remarkable
GSK-3β and results in β-catenin stabilisation. The level of selectivity and efficiency.
β-catenin is indeed lower in extracts of brain with prese-
nilin-1 mutations compared with controls or sporadic In this article we report the discovery of hymenialdisine
Alzheimer’s cases [9]. Interestingly, expressed tau is phos- (HD, 1; Figure 1) as a very potent inhibitor of CDK1,
phorylated in cells co-transfected with mutant presenilin-1 CDK2 and CDK5, GSK-3β and CK1. HD has been iso-
but not in cells expressing wild-type presenilin-1 [10]. Aβ lated from several species of marine sponges along with a
treatment of cultured hippocampal neurons results in stim- variety of related metabolites (Figure 1) [31]. We also
ulation of GSK-3β activity [11]. All these results place tau report the crystal structure of a CDK2–HD complex, which
hyperphosphorylation downstream of presenilins and Aβ provides an important advance in understanding the mol-
action. Finally, the recent discovery of tau mutations in ecular mechanism of CDK inhibition. HD was found to
AD-related diseases (reviewed in [3]) and the recently inhibit the in vivo phosphorylation of specific neuronal pro-
described functions of tau in regulating intracellular traffic teins by GSK-3β and CDK5. In particular, the AD-charac-
along microtubules [12] have renewed the interest in teristic phosphorylation of tau is completely inhibited by
understanding the causal link between tau hyperphospho- HD in vivo. As a potent and rather selective inhibitor of
rylation and the pathways leading to the cellular modifica- kinases involved in AD and other degenerative disorders,
tions responsible for AD. HD could be a lead compound for evaluating the impor-
tance of tau hyperphosphorylation in neurodegenerative
Because protein kinases play an essential role in virtually diseases and for interfering with this molecular event.
all cellular processes and in most diseases, extensive
searches for selective inhibitors of these enzymes have Results
been carried out (reviewed in [13]). Our laboratory has β and CK1
HD potently inhibits CDK1, CDK2, CDK5, GSK-3β
focused its efforts on the cyclin-dependent kinase (CDK) While screening compounds isolated from marine inver-
family. CDKs are involved in cell-cycle control (CDK1–4, tebrates and plants for new CDK inhibitors [32], we dis-
6 and 7), thymocyte apoptosis (CDK2), neuronal func- covered HD (11) to be a potent inhibitor of CDK1/cyclin
tions (CDK5) and transcriptional control (CDK7–9) B (Figure 2, Table 1). HD belongs to a family of chemi-
(reviewed in [14–16]). In nervous tissues CDK5/p35 phos- cally and metabolically related, marine-sponge-derived
phorylates the microtubule-associated proteins tau and natural products (which contain both bromopyrrole and
 Research Paper Protein kinase inhibition by hymenialdisine Meijer et al. 53

Figure 2 Table 1

Kinase inhibition selectivity of hymenialdisine.

Cdk1/cyclin B kinase activity (% of maximum)

100 Enzyme IC50 (nM)

CDK1/cyclin B 22
80 CDK2/cyclin A 70
CDK2/cyclin E 40
60 CDK3/cyclin E 100
CDK4/cyclin D1 600
HD CDK5/p25 28
Diacetyl-HD
40 CDK6/cyclin D2 700
Diacetyldebromo-HD
Axinohydantoin Erk1 470
Dibromophakellstatin
20 Oroidin
Erk2 2000
Stevensine c-raf > 10,000
MAPKK 1200
0 c-Jun amino-terminal kinase 8500
0 0.1 1 10 100 1000 10,000 100,000
Concentration (nM) Chemistry & Biology Protein kinase C α 700
Protein kinase C β1 1200
Protein kinase C β2 1700
Inhibition of CDK1/cyclin B by HD analogues. CDK1/cyclin B was Protein kinase C γ 500
assayed as described in the Supplementary material section. Activity is
Protein kinase C δ 1100
presented as % of maximal activity (i.e. in the absence of inhibitors).
Protein kinase C ε 6500
Protein kinase C η 2000
Protein kinase C ζ 60,000
guanidine groups) [31,33–37]. In the presence of 15 µM cAMP-dependent protein kinase 8000
ATP, HD was found to inhibit CDK1/cyclin B, CDK2/ cGMP-dependent protein kinase 1700
cyclin A, CDK2/cyclin E, CDK3/cyclin E and CDK5/p35 GSK3-β 10
with IC50 values of 22, 70, 40, 100 and 28 nM, respec- ASK-γ (plant GSK-3) 80
tively (Table 1). As observed with olomoucine [23], Eg2 kinase 4000
roscovitine [24], indirubin-3′-monoxime [28], ken- CK1 35
CK2 7000
paullone [29] and, in contrast to flavopiridol [38], HD had
Insulin receptor tyrosine kinase 75,000
limited effect on CDK4/cyclin D1 and CDK6/cyclin D2 c-src tyrosine kinase 7000
(IC50 values of 600 and 700 nM, respectively). c-abl tyrosine kinase 4000
Topoisomerase I –(10,000)
HD was next tested on a variety of highly purified kinases Topoisomerase II α –(10,000)
(IC50 values are shown in Table 1). Kinase activities were Enzyme activities were assayed as described in the Supplementary
assayed with appropriate substrates and 15 µM ATP (this materials section, in the presence of increasing HD concentrations.
concentration was comparable to previously published IC50 values were calculated from the dose-response curves. –, no
studies) and in the presence of increasing concentrations effect at the highest dose tested (in parentheses).
of HD. Most kinases tested were poorly or not inhibited
(IC50 > 1 µM). However, two kinases, GSK-3β and CK1, HD is a competitive inhibitor of ATP
were strongly sensitive to HD (IC50 values of 10 and To investigate the mechanism of HD action, kinetic experi-
35 nM, respectively; Figure 2, Table 1). The HD-sensi- ments were performed by varying both ATP levels and HD
tive kinases were also assayed in vitro with physiologically concentrations (Figure 6a). Double-reciprocal plotting of
relevant substrates: a fragment of presenilin-2 [39] for the data demonstrates that HD is a competitive inhibitor for
CK1 (Figure 3), Pak1 [18] for CDK5/p35 (Figure 4), the ATP. The apparent inhibition constant (Ki) was 50 nM.
insulin-receptor substrate IRS-1 [40] (data not shown) or HD does not act by displacing cyclin B from CDK1
tau for GSK-3β (Figure 5). The sensitivity of the kinases (Figure 6b).
towards HD was comparable to the sensitivity of the same
kinases assayed with more artificial substrates. Crystal structure of the CDK2–HD complex
Crystal structures of CDK2 complexed with several
We next tested some natural HD-related compounds iso- inhibitors have been reported: isopentenyladenine [41],
lated from marine sponges and some synthetically modified olomoucine [41], des-chloro-flavopiridol [42], roscovitine
HD analogues on CDK1/cyclin B, CDK5/p35, GSK-3β and [25], staurosporine [43], purvalanol [26] and indirubin-
CK1 (Figure 2; Table 2). HD was the most active com- 3′-monoxime [28]. All these compounds bind in the ATP-
pound. Interestingly, dibromocantharelline (110) displayed a binding pocket located in the groove between the small
significant inhibitory effect towards GSK-3β (IC50 = 3 µM). and the large lobes of CDK2. To study in detail the
Hymenidin (113) was selective for CDK5. essential binding interactions of HD, and to compare
54 Chemistry & Biology 2000, Vol 7 No 1

Table 2

β and CK1 by HD analogues.

Inhibition of CDK1, CDK5 and GSK-3β

IC50 (nM)
Compound CDK1/cyclin B CDK5/p25 GSK3-β CK1

1 Hymenialdisine 22 28 10 35
Hymenialdisine–platinum complex 30 200 130 1000
Diacetylhymenialdisine 130 1500 160 550
Diacetyldebromohymenialdisine 1300 3000 600 1100
4 Stevensine/odiline 70,000 – (100,000) – (100,000) – (100,000)
5 Axinohydantoin 4000 7000 3000 4500
8 Dibromophakellstatin – (10,000) – (100,000) > 10,000 – (10,000)
9 Agelastatin A – (100,000) – (100,000) 12,000 – (100,000)
10 Dibromocantharelline > 100,000 > 100,000 3000 – (100,000)
11 Dispacamide A – (100,000) – (100,000) – (100,000) – (100,000)
12 Clathrodin > 100,000 > 100,000 10,000 – (100,000)
13 Hymenidin > 100,000 4000 12,000 > 100,000
14 Oroidin > 100,000 50,000 20,000 > 100,000
15 Agelongine – (100,000) – (100,000) – (100,000) – (100,000)
6 Dibromoageliferin > 100,000 > 100,000 11,000 – (100,000)
7 Sceptrin – (100,000) – (100,000) – (100,000) – (100,000)

Numbers refer to structures shown in Figure 1. Enzyme activities were calculated from the dose-response curves. –, no effect at the highest
assayed as described in the Supplementary material section, in the dose tested (in parentheses).
presence of increasing HD concentrations. IC50 values were

these with the interactions observed in the other 26.7%) and good geometry. Residues 36–44 and 149–163
CDK2–inhibitor complexes, we determined the structure of CDK2, which are part of two highly flexible loops, were
of a CDK2–HD complex at 2.1 Å resolution (Table S1 in left out from the final model because of weak or missing
the Supplementary material section). Treatment with a electron density. All nonglycine residues in the CDK2
cross-linking agent, prior to soaking, was necessary to model have mainchain torsion angles that lie well within
prevent the crystal from cracking. The final model con-
sists of 274 amino acid residues, one bound HD molecule, Figure 4
85 solvent molecules and four molecules of ethylene
glycol, with a crystallographic R-factor of 19.2% (Rfree of (a)
GST–Pak1K299
Figure 3
0–
0.1–
0.33–
1–
3.33–
10–
33.3–
100–
333–
1000–
3000–
10,000–
33,333–
100,000–

0–
5–
20–
Hymenialdisine concentration (nM) 30–
Roscovitine (µM)
(b)
ip/kinase
Pak1

–H4
PS–2.MBP–
1 2 1.6 1.4 2 1.4 2.7 3 2.8
Westerns

–p35

–Pak1
0–
33–
100–
333–
1000–
3333–
10,000–
33,333–
100,000–
0
0.1
0.33
1
3.33
10
33.33
100
333
1000
3333
10,000
0

Chemistry & Biology

Hymenialdisine concentration (nM)
Chemistry & Biology
HD inhibits the phosphorylation of presenilin-2 by CK1 in vitro. A
bacterially expressed fusion protein between presenilin-2 and maltose-
binding protein (PS-2–MBP) was phosphorylated in vitro with CK1 in
the presence of increasing HD concentrations and resolved by sodium
dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE),
followed by autoradiography.
HD inhibits phosphorylation of Pak1 by CDK5/p35 in vitro and in vivo.
Rat embryo cortical neurons were exposed to various HD
concentrations for 1 h. Pak1 was then immunoprecipitated and its
kinase activity towards histone H4 measured. (a) The level of H4
phosphorylation following SDS–PAGE of the substrate is shown.
Numbers correspond to the quantification of the autoradiography.
(b) Western blots show the amounts of Pak1 and p35. Note an
increase in p35 levels as a consequence of CDK5 inhibition.
 Research Paper Protein kinase inhibition by hymenialdisine Meijer et al. 55

Figure 5 Figure 6

(a) (a) nM HD
0.4
900 80
Tau– 65
600 50
0.3 300 35
0–
0.1–
0.33–
1–
3.33–
10–
33.33–
100–
333–
1000–
3333–
10,000–
0–
20
0
–60 0 60 120 0
Hymenialdisine concentration (nM) HD concentration (nM)

1/V
(b) 0.2
HD
OA

–
FL

Coomassie

0.1

K9JA

0.0
AT8 0.0 0.1 0.2 0.3 0.4
1/ATP
(b)

AT180
Cyclin B

PHF-1

AT100
Chemistry & Biology
HD inhibits tau phosphorylation by GSK-3β in vitro and in vivo.
(a) Bacterially expressed recombinant human tau was phosphorylated
in vitro with GSK-3β in the presence of increasing HD concentrations
and resolved by SDS–PAGE, followed by autoradiography. (b) Sf9
cells expressing htau23 were left untreated (–), or exposed to okadaic
acid (OA), HD or flavopiridol (FL) for 5 h. Cell lysates (3 µg htau23)
were resolved by SDS–PAGE, stained with Coomassie blue or
immunoblotted with various antibodies: K9JA (a pan-tau antibody)
recognizes all preparations that contain tau; AT8, AT180 and PHF1
are specific for different phosphorylated SP or TP motifs; AT100
recognizes tau phosphorylated at Thr212 and Ser214, a highly
specific reaction for the Alzheimer’s form of tau.
the energetically favorable regions of the Ramachandran
plot [44], except for two residues, Glu73 and Arg126, that
have backbone conformations falling just outside the
‘additional allowed’ region in φ–ψ space.
HD was unambiguously localised in a Fourier difference
map, confirming that this inhibitor also binds in the ATP-
binding pocket (Figure 7a). A representation of the
binding-site interactions is shown in Figures 7b and 8.
The pyrroloazepine double-ring system of HD fills a
shallow hydrophobic pocket formed by Ile10, Val18,
Ala31, Val64, Phe80 and Leu134, making several van der
Waals contacts with the sidechain atoms of these residues.
In addition, three hydrogen bonds are formed with the
CDK1
HD concentration (µM) 0 1 0 10 Chemistry & Biology
HD acts by competing with ATP. (a) Double reciprocal plots of kinetic
data from assays of CDK1/cyclin B protein kinase activity at different
concentrations of hymenialdisine. Enzyme activities were assayed as
described in the Supplementary material section. (a) 1/V versus 1/ATP
primary plot. ATP concentrations in the reaction mixture varied from
0.05 to 0.25 mM, concentration of histone H1 was kept constant at
0.7 mg/ml. Inset, secondary re-plots of slopes versus concentration
from primary plots. The apparent inhibition constant (Ki) is indicated by
an arrow. (b) HD does not release cyclin B from CDK1. p9CKShs1-
sepharose-immobilised CDK1/cyclin B was exposed to HD for 30 min,
washed and analysed by western blotting with anti-cyclin B and anti-
PSTAIRE antibodies.
backbone of CDK2, between the N1 atom of the pyrrole
ring and the carbonyl oxygen of Leu83, between the O1
carbonyl oxygen of azepine ring and the backbone amide
of Leu83, and between the N2 amide of the azepine ring
and the carbonyl oxygen of Glu81. The bromine atom
bound to the pyrrolo ring of HD points towards the
outside of the ATP-binding pocket, where it is partly
exposed to solvent, but also packed against the mainchain
carbonyl oxygen atoms of Ile10 and His84, and the
sidechains of Ile10 and Leu134. Binding of the guanidine
ring system of HD involves a few van der Waals contacts,
mainly with the sidechain of Val18, in addition to one
56 Chemistry & Biology 2000, Vol 7 No 1
Figure 7
(a)
Overall structure of a CDK2–HD crystal. (a) Stereo view of the
electron density for HD. The electron density at 2.1 Å resolution was
drawn from a Fo–Fc difference omit electron density map calculated
after simulated annealing refinement. The map was contoured at 2.0 σ.
(b) Backbone drawing of CDK2 with HD in the ATP-binding pocket
direct hydrogen bond between the N5 amino group of the
guanidine and one of the sidechain oxygen atoms of
Asp145, and two water-mediated hydrogen bonds
between the O2 of HD and the mainchain NH of Asp145,
and between the N5 of HD and the mainchain carbonyl of
Gln131. Comparison with the apo-CDK2 and CDK2–ATP
structures [45] reveals a large movement of Asp145 upon
binding of HD, consisting of a 1 Å shift of the Cα atom,
and a rotation of the sidechain of ~90° around the Cα–Cβ
bond away from the HD guanidine ring. All the other
residues in contact with the inhibitor have similar confor-
mations to those observed in the apo-CDK2 and
CDK2–ATP structures.
Comparison of CDK2–HD with other complexes
To provide a structural basis for understanding the potency
of HD we compared the structure of the CDK2–HD
complex with those of CDK2 complexed with ATP [45],
staurosporine, flavopiridol and with the purine analogues
olomoucine, roscovitine and purvalanol, as well as with the
structure of the cyclin A–CDK2–ATP complex [46].
The hydrophobic double-ring system of HD binds at
approximately the same position in CDK2 as the purine
ring of ATP in the CDK2–ATP complex, similar to the
positions of the double-ring systems in the other CDK2–
inhibitor complexes (Figure 9). Although the orientation of
(b)
Chemistry & Biology
between the smaller amino-terminal domain and the larger
carboxy-terminal domain. Secondary structural elements are indicated
by arrows for β strands and coils for α helices, and labelled as in the
apo-enzyme [74]. Drawn using MOLSCRIPT [75].
the different double-ring systems varies significantly among
the different inhibitors, it is restrained by the necessity to
provide optimal shape complementarity with the shallow
ATP-purine binding pocket while allowing the formation of
a number of hydrogen bonds with the backbone of residues
81–83 at the cross-over connection in CDK2. The hydro-
gen-bonding interactions in the CDK2–HD complex seem
to be the most favorable of all the CDK2–inhibitor com-
plexes studied so far. The hydrogen bonds between N2 and
the peptide oxygen of Glu81, and between O1 and the
peptide amide of Leu83 resemble closely those between
the adenine base of ATP and CDK2 and those in the com-
plexes with staurosporine and flavopiridol. The third hydro-
gen bond in the CDK2–HD complex with the mainchain
carbonyl of Leu83, is absent in these complexes, and can be
observed only in the complexes with the three purine-
based inhibitors olomoucine, roscovitine and purvalanol.
Similar to HD, these latter three inhibitors also form three
hydrogen bonds with the cross-over connection, but their
interaction with the Glu81 peptide oxygen is much weaker,
involving a rare C–H...O hydrogen bond with the acidic C8
atom of the purine ring [26].
The bromine atom of HD is bound close to a region in
CDK2 that in the other CDK2–inhibitor complexes is occu-
pied by a benzyl group. Binding of a hydrophobic group in
this region, where it can pack against the sidechains of
 Research Paper Protein kinase inhibition by hymenialdisine Meijer et al. 57

Figure 8

Protein–inhibitor interactions in the (a)

CDK2–HD complex. (a) Stereo diagram
showing the refined structure of HD in the
ATP-binding pocket of CDK2. Inferred
hydrogen bonds are shown as thin dotted
lines. Oxygen atoms are shown in red,
nitrogen atoms in blue and bromine in green.
(b) Schematic illustration of the interactions
between CDK2 and HD. Protein sidechain
contacts are indicated by lines connecting to
the respective residue box, whereas
interactions with mainchain atoms are shown
as lines to the specific mainchain atom. Van
der Waals contacts are indicated by dotted
lines and hydrogen bonds by broken lines.

(b) Phe80
Val64
Ala31

Glu81 O 2.
8 Lys33

H
N2
Wat566
C6 Val18
2.7 O1
C5 C7 2.5
NH

3.2
O2
Leu83 O C4 C8
C3 C10 H
3. C9 N
2
H N1
Cα
N3 H Asp145
C2 N4
Phe82 C1
H
C C11
3.1

0
3.
Br
N5 H
O
3.3 H
Wat580 3.
His84 3 Cα Gly13
2.8

Leu134 Wat556
O Gln131
O Ile10
Chemistry & Biology

Ile10, Phe82 and the backbone of residues 82–84, is impor-
tant for increasing the specificity of inhibitors for CDK2
[25,26,41]. Although the bromine in HD cannot provide the
same number of interactions as a benzyl ring, the presence
of this atom in HD probably contributes significantly to its
binding affinity and specificity towards CDK2, as can been
seen from the inhibitory activities of various HD analogues
in Table 2 and Figure 2.
The region of CDK2 occupied by the guanidine ring of
HD is also interesting. A superposition with the other
CDK2–inhibitor complexes shows that only the flavopiri-
dol and staurosporine inhibitors have groups bound in this
region of CDK2, which partly overlaps with the pocket
where the α-phosphate of ATP binds. Comparison of the
structure of the CDK2–HD complex with that of the
CDK2–flavopiridol complex [42] reveals a number of
striking similarities between the binding modes of these
structurally diverse inhibitors. The O2 carbonyl oxygen of
HD is located close to the position of the O7 hydroxyl
group of the flavopiridol, which emanates from the benzo-
pyran ring bound at the purine region of the ATP-binding
pocket. In both inhibitors the oxygen atoms form a water-
mediated hydrogen bond with the mainchain amide of
Asp145. Furthermore, the N5 amino group at the guani-
dine ring of HD is located near to the position of the posi-
tively charged amine group of the piperidinyl ring in the
CDK2–flavopiridol complex. Both atoms are in hydrogen-
bonding distance with the sidechain carboxylate of
Asp145. The energetically favorable interaction between
the positively charged amine group of the flavopiridol and
the negatively charged carboxylate of Asp145 would make
an important contribution to the binding strength of this
inhibitor to CDK2. A similar interaction seems possible in
the CDK2–HD complex, as under physiological condi-
tions the guanidine ring is likely to be at least partly proto-
nated at N3, thus providing a (partly) positive charge
delocalised between N3, C11, N4 and N5 [33]. Also the
58 Chemistry & Biology 2000, Vol 7 No 1
Figure 9
(a)
(b)
(c)
movement of Asp145 is conserved in both CDK2–
inhibitor complexes. It is also seen in the indirubin-5-
sulphonate–CDK2 structure [28]. Asp145 is part of the
conserved Asp–Phe–Gly (DFG) motif found in most
protein kinases [47]. Although significantly different from
those in the CDK2–ATP complex, the position and con-
formation of Asp145 in both CDK2–inhibitor complexes
is, in fact, very similar to those in the functionally more
relevant cyclin A–CDK2–ATP complex [46] (Figure 9a).
In vitro inhibition of presenilin phosphorylation
We next investigated the effects of HD on the in vitro and
in vivo phosphorylation of various protein substrates rele-
vant to AD. The large hydrophilic loop of presenilin-2
between transmembrane domains 6 and 7 is a substrate for
both CK1 and CK2 in vitro; this domain is phosphorylated
in vivo [39]. Using a presenilin-2–maltose-binding protein
(MBP) fusion protein as an in vitro substrate for CK1 we
Stereo views comparing the binding of various
CDK2 ligands or inhibitors. (a) Superposition
of HD (black bonds) on ATP (white bonds) in
apo-CDK2 and on ATP (yellow bonds) in the
cyclin A–CDK2 complex. (b) Superposition of
HD on olomoucine (white) and purvalanol
(yellow). (c) Superposition of HD on
flavopiridol. The CDK2 backbone atoms of
residues 81–84, and the sidechain of Asp145
in the CDK2–HD complex are shown in ball-
and-stick representation with black bonds.
The same residues from the superimposed
CDK2–ligand complex are shown as bonds.
In (a) the thin bonds refer to the apo-enzyme,
the thick bonds to the cyclin A–CDK2 dimers.
Chemistry & Biology
observed a dose-dependent inhibition of presenilin-2
phosphorylation by HD (Figure 3). MBP alone was not
phosphorylated by CK1 (data not shown).
In vitro and in vivo inhibition of Pak1 phosphorylation by
neuronal CDK5/p35
Among the physiological substrates of CDK5/p35 is the
neuronal kinase Pak1 [18]. Both Pak1 and p35 associate
with Rac, a small GTPase of the Rho family. Phosphoryla-
tion of Pak1 by CDK5/p35 results in inhibition of the Pak1
kinase activity [18]. Roscovitine inhibits CDK5/p35 and
the resulting down-regulation of Pak1 both in vitro and in
vivo [18]. These experiments were repeated with HD
(Figure 4). First CDK5/p35 was immunoprecipitated from
P02 rat cortices and its kinase activity towards
GST–Pak1K299 (a kinase-dead Pak1 mutant) in the pres-
ence of HD, roscovitine or dimethylsulfoxide (DMSO) was
assayed as described previously [18]. A dose-dependent
 Research Paper Protein kinase inhibition by hymenialdisine Meijer et al. 59

inhibition of CDK5 by HD was observed (IC50 values Figure 10

between 10 and 100 nM; Figure 4a). In vivo experiments
were next performed using cultured neurons from E18 rat (a) (b)
embryo cortices (Figure 4b). As previously shown [48], the
amount of p35 increases with the extent of CDK5/p35
kinase inhibition. An increase in Pak1 activity was
observed, consistent with an inhibition of endogenous
CDK5 activity (Figure 4b).

HD inhibits MAP-1B phosphorylation by GSK-3 in (c) (d)

cerebellar granule cell neurons
GSK-3β is inhibited by both WNT-7a and lithium in cere-
bellar granule cell neurons [49,50]. WNT-7a and lithium
induce axonal remodelling and loss of a phosphorylated
form of MAP-1B, a microtubule-associated protein
involved in axonal outgrowth [50]. As GSK-3β phosphory-
(e) (f)
lates MAP-1B at a site recognised by the antibody SMI-31,
inhibition of GSK-3β by WNT or lithium results in the loss
of a phosphorylated MAP-1B, MAP-1B-P [50]. To examine
the effect of HD on neuronal morphology and MAP-1B
phosphorylation cerebellar granule cells were cultured in
different concentrations of HD. In control cells with long
processes and very few filopodia (Figure 10a), MAP-1B-P is
(g) (h)
present along the entire length of the axon (Figure 10b). At
low concentrations (1 µM, 10 µM and 25 µM) HD had no
noticeable effect on the morphology of the cells
(Figure 10c), or the distribution of MAP-1B-P (Figure 10d).
However, 50 µM HD treatment induced axonal spreading
and branching and a shortening of axon length
(Figure 10e), with a concomitant loss of MAP-1B-P from (i) Concentration of HD (µM)
most of the axonal processes (Figure 10f). Treatment of 0 1 10 25 50 100
cultures with 100 µM HD caused a more dramatic change
in cell morphology characterised by extensive branching MAP-1B-P–
and spreading, shortening of axon length and an increased
number of filopodia (Figure 10g), together with loss of Chemistry & Biology
MAP-1B-P from processes (Figure 10h). The axonal
remodelling observed was associated with a loss of stable HD induces axonal remodelling and decreases GSK-3-induced
microtubules from spread areas of the axons (data not MAP-1B-P levels in axonal processes. (a,b) Double immunofluorescence
shown). HD induces the loss of MAP-1B-P in a dose- staining for GAP-43 (red) and MAP-1B-P (green) shows that MAP-1B-P
dependent manner as determined by western blotting is present along the axon. (c) A 20 h treatment with 10 µM HD has no
obvious effect on cell morphology, or (d) MAP-1B-P localisation. Arrows
(Figure 10i). This effect is similar to that observed with indicate the same cells. (e) 50 µM HD treatment induces axonal
lithium or WNT-7a treatment [50]. As we have shown that spreading, shortening of the axon and (f) loss of MAP-1B-P from axonal
HD inhibits GSK-3β directly, our results suggest that the processes. Arrows indicate the same cells. (g)100 µM HD treatment
loss of MAP-1B-P and axonal remodelling induced by HD causes a more dramatic change in cell morphology with spreading and
branching along the axon, an increased number of filopodia and
is a consequence of GSK-3β inhibition in cultured neurons. shortening of the axon. (h) MAP-1B-P is lost from most of the axonal
processes. Arrows indicate axons where MAP-1B-P is lost. Bar = 20 µm.
Inhibition of tau phosphorylation by GSK-3 (i) Western blotting analysis of cell lysates shows a gradual decrease in
The microtubule-binding protein tau is the substrate of MAP-1B-P in response to increasing doses of HD.
several kinases, including GSK-3β and CDK5/p35. Bacteri-
ally expressed recombinant human tau was indeed phos-
phorylated in vitro by GSK-3β and this phosphorylation was 50 µM flavopiridol (FL), a CDK inhibitor that also inhibits
inhibited in a dose-dependent manner by HD, with an IC50 GSK-3β (L.M., unpublished observations). Htau23 was
value of ~33 nM (Figure 5a). We next investigated the resolved by using SDS–PAGE followed by immunoblotting
effect of HD on the in vivo phosphorylation of human tau23 with various antibodies. K9JA (a pan-tau antibody) recog-
expressed in Sf9 cells (Figure 5b). Cells were left untreated nizes all preparations that contain tau. AT8, AT180 and
(–), or exposed to 0.2 µM okadaic acid (OA), 50 µM HD or PHF1 are specific for different phosphorylated Ser–Pro (SP)
60 Chemistry & Biology 2000, Vol 7 No 1
or Thr–Pro (TP) motifs: Ser202 and Thr205, Thr231 and
Ser235, and Ser396 and Ser404, respectively, (as numbered
in htau40, the longest human tau isoform). AT100 recog-
nizes tau phosphorylated at Thr212 and Ser214; this reac-
tion is highly specific for the Alzheimer’s form of tau but
occurs in Sf9 cells as well, provided both sites are phospho-
rylated [51]. The disappearance of AT100 signal following
treatment with HD or flavopiridol indicates that both com-
pounds are able to inhibit GSK-3β like activity in Sf9 cells.
Discussion
HD is a new and potent inhibitor of CDKs, GSK-3β and
CK1. We have characterised its molecular interaction with
CDK2, as well as its powerful effects on the in vitro and in
vivo phosphorylation of neuronal proteins. Particularly
interesting is the inhibitory effect of HD on the in vivo
phosphorylation of the AD protein tau.
HD has been found in species of marine sponges belong-
ing to the Agelasidae, Axinellidae and Halichondriidae fami-
lies [33–36,52]. These animals contain a variety of
substances that are clearly metabolically related to HD
(e.g. Figure 1). The synthetic and degradation pathways of
HD and how the related molecules fit in these pathways
still remain to be investigated. Some experimental evi-
dence suggests that oroidin and 4,5-dibromopyrrol-2-car-
boxylic acid, from Agelas clathrodes, act as ‘feeding
deterrent agents’, protecting the sponge from predation by
several species of fish [53].
Although HD inhibits CDKs by competition with ATP, it
belongs to a new class of chemical CDK inhibitors. Previ-
ously described antagonists all act by competing with ATP
binding. All interact with the ATP-binding pocket
through two or three hydrogen bonds with the Glu81 and
Leu83 residues of CDK2. The ATP-binding pocket of
CDKs (and presumably of other kinases) can therefore
accommodate an unexpectedly large variety of structures,
with high affinity and high selectivity.
The potency of HD as a CDK inhibitor can be explained
by analysing its binding interactions in the CDK2–HD
crystal structure. Most of these interactions involve atoms
of the HD pyrroloazepine ring system (which contribute
31 of the total 45 van der Waals contacts and three of the
four direct hydrogen bonds). The van der Waals packing
and hydrogen-bonding contacts of the HD double-ring
system with the same, mostly hydrophobic enzyme
residues that form the pocket for the adenine base in the
ATP complex structure is a common feature in all crystal-
lographically determined binding modes of CDK2
inhibitors, emphasising the anchoring nature of these
interactions. Affinity and selectivity of the inhibitors is
further built up from the binding of substituents that
emanate from the double-ring system, which are different
among the different inhibitors. One way to assess the
binding strength of a ligand from its binding mode in a
protein–ligand complex is the shape complementarity of
the contact surfaces of both partners. These values are cal-
culated from the structures of the different CDK2–
inhibitor complexes (Table S2 in the Supplementary
matieral section). For the three structurally related purine-
based inhibitors, there is a good correlation between the
inhibitory strength, and the shape complementarity with
the inhibitor-binding pocket. For the structurally more
diverse inhibitors des-chloro-flavopiridol and HD,
however, shape complementarity alone seems to be a poor
indicator of inhibitor strength. We believe that HD
achieves tight binding to CDK2 through the contribution
of strong specific hydrogen bonds, and a short-range elec-
trostatic interaction between the acidic Asp145 and the
basic guanidinium ring. In addition, a significant entropic
factor could be involved in the binding of HD to CDK2,
as the entropy penalty for the complex formation is
minimal due to the rigid nature of the inhibitor. Neverthe-
less, one can imagine that drug-design studies could sig-
nificantly improve the binding by optimising the shape
complementarity of HD, for instance by replacing the
bromine atom with more bulky groups. Unfortunately, at
this point we are unable to identify the common structural
features of the kinases for which HD is a potent inhibitor,
nor explain why other kinases are not sensitive. Alignment
of human CDK2, CK1 and GSK-3β (data not shown) is
not especially helpful. Only the crystal structure of a
CK1–HD or GSK-3β–HD will provide a clear view of the
key interacting residues underlying the selectivity.
Investigating the selectivity of a compound is not a trivial
matter. We feel that what is really important is the in vivo
selectivity, the range of targets that an inhibitor will actu-
ally interact with within a cell. One method we are cur-
rently developing is the purification of targets using
affinity chromatography on immobilised inhibitors. The
crystal structure of the HD–CDK2 complex provides pre-
cious information with respect to its orientation within the
ATP-binding pocket, showing which part of HD is acces-
sible to solvent. This is where a linker could be attached
to tether the inhibitor to a solid matrix while maintaining
free access to its kinase targets. Using this approach with
purvalanol, based on the CDK2–purvalanol crystal struc-
ture [26], we recently identified the intracellular targets of
purvalanol in a variety of cells and tissues (Knockaert et al.
and L.M., unpublished observations).
We are presently re-isolating HD-related compounds from
marine sponges to investigate their CDK, GSK-3β and
CK1 inhibitory properties. HD, the most abundant
metabolite (up to 0.5% of the sponge dry weight), can also
be used as a starting point for various chemical modifica-
tions, leading to new derivatives and this research is in
progress in our laboratories. Furthermore, HD has been
synthesised [54–56] and we are working on generating
 Research Paper Protein kinase inhibition by hymenialdisine Meijer et al.
new synthetic analogues. Finally, we have initiated a com-
binatorial chemistry approach to generate an even larger
number of HD-related molecules. These various
approaches, also partially guided by the CDK2–HD struc-
ture, should complement the initial structure–activity
study presented in this paper. Through these various
approaches, we hope to generate more selective and more
cell-permeable HD analogues.
The kinases responsible for the hyperphosphorylation of
tau and MAP-1B observed in AD certainly constitute rea-
sonable screening targets. By acting on GSK-3β, CDK5
and CK1, the major kinases involved in this process, HD
is a strong lead compound from which clinically useful
drugs could potentially be generated. In addition to AD,
HD could be used to target other neuronal disorders. For
example, CDK5 is present in Lewy bodies, the typical
inclusions observed in neurons in Parkinson’s disease [57]
and in glial cytoplasmic inclusions that characteristically
occur in the oligodendrocytes of patients with multiple
system atrophy [58]. CDK5 also accumulates in axonal
swellings in dogs with hereditary canine spinal muscular
atrophy [59]. Lithium is widely used for treatment of
mood disorders, but its exact mechanism of action is not
clearly defined. It appears to inhibit GSK-3β in a non-
competitive way [60], thereby reducing tau phosphoryla-
tion [61–63]. The efficiency of HD towards GSK-3β
(IC50 = 35 nM) compares very favourably with the relative
efficacy of lithium towards GSK-3β (IC50 = mM).
Insulin acts by activating a protein kinase cascade leading
to the inhibition of GSK-3β. This results in the stimula-
tion of glycogen synthase and protein synthesis (reviewed
in [64]). As a consequence, any GSK-3β inhibitor is a
potential insulino-mimetic and could be useful in treating
diabetes. This is certainly worth exploring with HD and
its analogues.
Finally, it has been reported that HD inhibits NFκB
activity through an alternate mechanism to inhibiting
protein kinase C or IκB phosphorylation [65,66]. HD has
also been reported to have anti-inflammatory properties
[67]. The possibility that CDKs, GSK-3β, CK1, or
another unknown kinase target of HD, are involved in
NFκB activation and in adjuvant-induced arthritis
deserves further investigation.
Significance
Among the estimated 2000 human protein kinases,
cyclin-dependent kinase 5 (CDK5) and glycogen syn-
thase kinase 3 (GSK3) appear to play an essential role in
the hyperphosphorylation of substrates involved in
Alzheimer’s disease. While screening through a large
collection of natural products derived from marine organ-
isms, we identified hymenialdisine (HD) as a potent and
rather selective inhibitor of these protein kinases. HD
61
acts by competing with ATP for binding at the catalytic
site of the kinase. HD was co-crystallised with CDK2,
and the resolution of this complex structure provides data
on the orientation and interactions of the inhibitor in the
ATP-binding pocket of the kinase. It also allows some
speculations on the further optimisation of HD analogues
as kinase inhibitors. The in vivo effects of HD on kinases
was demonstrated in several models. These results
should encourage the study of HD as a lead compound in
the treatment of neurodegenerative disorders.
Materials and methods
Preparation of CDK2–HD crystals
Human CDK2 was purified and crystallized as previously described
[68]. Initial attempts to obtain CDK2–HD crystals, by adding small
amounts of solid inhibitor to the crystal-containing drops, were unsuc-
cessful because of insolubility of the inhibitor. When soaking with HD
solubilized in the presence of 1% DMSO, however, it was observed
that crystals cracked and dissolved at inhibitor concentrations as low
as 10 µM, strongly suggesting that binding occurred (the same
DMSO-containing solution without inhibitor did not harm the crystals).
A procedure involving chemical cross-linking was employed to prevent
crystals from cracking. Crystals were first soaked in a solution contain-
ing 0.5 mM ATP, 1 mM MgCl2 for 2 h, then cross-linked with 0.1% glu-
taraldehyde, for 1 h at 4°C. After extensive washing the crystals were
transferred to an inhibitor solution in 0.2 M HEPES, 5% ethylene glycol
and 1% DMSO. This protocol allowed crystals to be soaked at
inhibitor concentrations up to 0.5 mM, for several days, without
showing any damage.
Determination of the CDK2–HD crystal structure
X-ray diffraction data to 2.1 Å resolution were collected on a single
CDK2–HD crystal using an R-Axis II image plate detection system,
mounted on a Rigaku rotation-anode generator. Data were collected at
120 K to prevent radiation damage of the crystal. Just prior to freezing,
the crystal was transferred to a cryo-protecting solution containing
25% ethylene glycol. Flash freezing was achieved in a dry nitrogen
stream using a Molecular Structure Corporation cryodevice. Freezing
altered slightly the unit cell dimensions and increased the mosaic
spread from 0.2–0.6°. The cross-linking by itself did not alter the dif-
fraction characteristics significantly. The intensity data were processed
with the DENZO and SCALEPACK programs [69]. The program
TRUNCATE, as implemented in the CCP4 suite [70], was used to
obtain the final set of structure factor amplitudes. A summary of the
data processing statistics is presented in Table S1 in the Supplemen-
tary material section. Refinement of the CDK2–HD complex was
started from the coordinates of the highly refined CDK2–ATP model.
All refinement steps were carried out using the program X-PLOR [71].
Molecular replacement followed by rigid body refinement was neces-
sary to successfully reorient and reposition the CDK2 molecule in the
unit cell of the frozen crystal. The model was then further refined using
several rounds of conjugated-gradient energy minimization. After this
stage clear electron density, calculated from 2Fo–Fc and Fo–Fc Fourier
maps, indicated the binding mode of the hymenialdisine inhibitor. Initial
coordinates and geometric restraint terms for hymenialdisine were
taken from the small molecule structure of Mattia et al. [72]. Refine-
ment of the CDK2–HD model was then pursued with several rounds of
both X-ray restrained energy minimization and molecular dynamics,
alternated with model building. Towards the end of the refinement
several water molecules and a few molecules of ethylene glycol were
added to the model. The stereochemistry of the CDK2–HD model was
verified using the software package PROCHECK [73].
Accession numbers
The co-ordinates of the CDK2–hymenialdisine structure have been
depostied in the PDB under the code 1DM2.
62 Chemistry & Biology 2000, Vol 7 No 1
Supplementary material
Supplementary material including details of diffraction data collection and
refinement statistics of the CDK2–HD complex, buried areas of CDK2
and ligands, chemicals and reagents used, buffers, kinase preparations
and assays, electrophoresis and blotting, in vitro and in vivo phosphoryla-
tion of presenilin-2, Pak1, MAP-1B and tau and MAP-1B immunochem-
istry is available at http://current-biology.com/supmat/supmatin.htm.
Acknowledgements
We are grateful to the following people for providing reagents: F. Cafieri, M.
Cobb, P. Davies, E. Fattorusso, M. Goedert, W. Harper, F. Hofmann, S.
Lohmann, T.J. Lukas, H. Mett, M. Meyerson, A. Mangoni, L. Pinna, E. Sausville,
O. Taglialatela-Scafati, H.Y.L. Tung, J.H. Wang, M. Watterson, G. Wilkin,
J.R. Woodgett and M. Yamashita. We thank Dr A. Larsen for performing the
topoisomerase I and II assays. We thank the fishermen of the ‘Station
Biologique de Roscoff’ for collecting the starfish, J. Orillon for the photo-
graphic work and O. Collin for computer expertise. This research was sup-
ported by grants from the ‘Association pour la Recherche sur le Cancer’ (ARC
9314) (to L.M.) and the ‘Conseil Régional de Bretagne’ (to L.M.), Outstanding
Investigator Grant CA-44344A1-05-10, U.S. National Cancer Institute, DHHS
(to G.R.P.), the Arizona Disease Control Research Commission (to G.R.P.)
and the Fannie E. Rippel Foundation (to G.R.P.). A.M.W.H.T. was supported
by a long-term fellowship from the Human Frontier Science Program.
References
1. Selkoe, D.J. (1998). The cell biology of β-amyloid precursor protein
and presenilin in Alzheimer’s disease. Trends Cell Biol. 8, 447-453.
2. Imahori, K., & Uchida, T. (1997). Physiology and pathology of tau
protein kinases in relation to Alzheimer’s disease. J. Biochem.
121, 179-188.
3. Mandelkow, E.-M. & Mandelkow, E. (1998). Tau in Alzheimer’s
disease. Trends Cell Biol. 8, 425-427.
4. Singh, T.J., Grundke-Iqbal, I. & Iqbal, K. (1995). Phosphorylation of tau
protein by casein kinase 1 converts it to an abnormal Alzheimer-like
state. J. Neurochem. 64, 1420-1423.
5. Haas, C. & Mandelkow, E. (1999). Proteolysis by presenilins and the
renaissance of tau. Trends Cell Biol. 9, 241-244.
6. Anderton, B.H. (1999). Alzheimer’s disease: clues from flies and
worms. Curr. Biol. 9, R106-R109.
7. Wolfe, M.S., Xia, W., Ostaszewski, B.L., Diehl, T.S., Kimberly, W.T. &
Selkoe, D.J. (1999). Two transmembrane aspartates in presenilin-1
required for presenilin endoproteolysis and γ-secretase activity. Nature
398, 513-517.
8. Yu, G., et al., & Fraser, P.E. (1998). The presenilin 1 protein is a
component of a high molecular weight intracellular complex that
contains β-catenin. J. Biol. Chem. 273, 16470-16475.
9. Zhang, Z.H., et al., & Yankner, B.A. (1998). Destabilization of β-catenin
by mutations in presenilin-1 potentiates neuronal apoptosis. Nature
395, 698-702.
10. Takashima, A., et al., & Wolzonin, B. (1998). Presenilin 1 associates
with glycogen synthase kinase-3β and its substrate tau. Proc. Natl
Acad. Sci. USA 95, 9637-9641.
11. Takashima, A., et al., & Yamaguchi, H. (1998). Activation of tau protein
kinase I/glycogen synthase kinase-3β by amyloid β peptide (25–35)
enhances phosphorylation of tau in hippocampal neurons. Neurosci.
Res. 31, 317-323.
12. Ebneth, A., et al., & Mandelkow, E. (1998). Overexpression of tau
protein inhibits kinesin-dependent trafficking of vesicles, mitochondria,
and endoplasmic reticulum: implications for Alzheimer’s disease.
J. Cell Biol. 143, 777-794.
13. Hudkins, R.L. (1999). ATP-site directed kinase inhibitors for
therapeutic intervention. Curr. Med. Chem. 6, 773-903.
14. Dunphy, W.G. (1997). Methods in Enzymology: Cell-Cycle Control.
Academic Press, New York.
15. Morgan, D. (1997). Cyclin-dependent kinases: engines, clocks, and
microprocessors. Annu. Rev. Cell Dev. Biol. 13, 261-291.
16. Vogt, P.K. & Reed, S.I. (1998). Current Topics in Microbiology and
Immunology: Cyclin-Dependent Kinase (CDK) Inhibitors. Springer
Verlag, Berlin.
17. Garcia-Perez, J., Avila, J. & Diaz-Nido, J. (1998). Implications of cyclin-
dependent kinases and glycogen synthase kinase 3 in the
phosphorylation of microtubule-associated protein 1B in developing
neuronal cells. J. Neurosci. Res. 52, 445-452.
18. Nikolic, M., Chou, M.M., Lu, W., Mayer, B.J. & Tsai, L.-H. (1998). The
p35/cdk5 kinase is a neuron-specific Rac effector that inhibits Pak1
activity. Nature 395, 194-198.
19. Qi Z., Tang D.M., Zhu X.D., Fujita D.J., Wang J.H. (1998). Association
of neurofilament proteins with neuronal Cdk5 activator. J. Biol. Chem.
273, 2329-2335.
20. Meijer, L. & Kim, S.H. (1997). Chemical inhibitors of cyclin-dependent
kinases. Methods Enzymol. 283, 113-128.
21. Garrett, M.D. & Fattaey, A. (1999). CDK inhibition and cancer therapy.
Curr. Opin. Gene. Dev. 9, 104-111.
22. Gray, N.S., Détivaud, L., Doerig, C. & Meijer, L. (1999). ATP-site
directed inhibitors of cyclin-dependent kinases. Curr. Med. Chem.
6, 859-875.
23. Vesely, J., et al., & Meijer, L. (1994). Inhibition of cyclin-dependent
kinases by purine derivatives. Eur. J. Biochem. 224, 771-786.
24. Meijer, L., et al., & Moulinoux, J.P. (1997). Biochemical and cellular
effects of roscovitine, a potent and selective inhibitor of the cyclin-
dependent kinases cdc2, cdk2 and cdk5. Eur. J. Biochem.
243, 527-536.
25. De Azevedo, W.F., Leclerc, S., Meijer, L., Havlicek, L., Strnad, M. &
Kim, S.-H. (1997). Inhibition of cyclin-dependent kinases by purine
analogues: crystal structure of human cdk2 complexed with
roscovitine. Eur. J. Biochem. 243, 518-526.
26. Gray, N.S., et al., & Schultz P.G. (1998) Exploiting chemical libraries,
structure, and genomics in the search for kinase inhibitors. Science
281, 533-538.
27. Losiewiecz, M.D., Carlson, B.A., Kaur, G., Sausville, E.A. &
Worland, P.J. (1994). Potent inhibition of cdc2 kinase activity by the
flavonoid L86-8275. Biochem. Biophys. Res. Commun.
201, 589-595.
28. Hoessel, R., et al., & Meijer, L. (1999). Indirubin, the active constituent
of a Chinese antileukaemia medicine, inhibits cyclin-dependent
kinases. Nat. Cell Biol. 1, 60-67.
29. Zaharevitz, D., et al., & Sausville, E.A. (1999). Discovery and initial
characterization of the paullones, a novel class of small-molecule
inhibitors of cyclin-dependent kinases. Cancer Res. 59, 2566-2569.
30. Schultz, C., et al., & Kunick, C. (1999). The Paullones, a series of
cyclin-dependent kinase inhibitors: synthesis, evaluation of CDK1/
cyclin B inhibition, and in vitro antitumor activity. J. Med. Chem.
42, 2909-2919.
31. Pettit, G.R., et al., & Boyd, M.R. (1997). Antineoplastic agents. 362.
Isolation and X-ray crystal structure of dibromophakellstatin from the
Indian Ocean sponge Phakellia mauritiana. J. Nat. Prod. 60, 180-183.
32. Pettit, G.R. (1996). Progress in the discovery of biosynthetic
anticancer drugs. J. Nat. Prod. 59, 812-821.
33. Sharma, G.M., Buyer, J. & Pomerantz, M.W. (1980). Characterization
of a yellow compound isolated from the marine sponge Phakellia
flabellata. J. Chem. Soc. Chem. Commun., 435-436.
34. Cimino, G., de Rosa, S., de Stefano, S., Mazzarella, L., Puliti, R. &
Sodano, G. (1982). Isolation and X-ray crystal structure of a novel
bromo-compound from two marine sponges. Tetrahedron Lett.
23, 767-768.
35. Kitagawa, I., Kobayashi, M., Kitanaka, K., Kido, M. & Kyogoku, Y.
(1983). Marine natural products. XII. On the chemical constituents of
the Okinawan marine sponge Hymeniacidon aldis. Chem. Pharm.
Bull. 31, 2321-2328.
36. Pettit, G.R., et al., & Camou, F. (1990). Antineoplastic agents. 168.
Isolation and structure of axinohydantoin. Can. J. Chem. 68, 1621-
1624.
37. Williams, D.H. & Faulkner, J. (1996). Isomers and tautomers of
hymenialdisine and debromohymenialdisine. Nat. Prod. Lett. 9, 57-64.
38. Carlson, B.A., Dubay, M.M., Sausville, E.A., Brizuela, L. & Worland, P.J.
(1996). Flavopiridol induces G1 arrest with inhibition of cyclin-
dependent kinase (CDK) 2 and CDK4 in human breast carcinoma
cells. Cancer Res. 56, 2973-2978.
39. Walter, J., Grünberg, J., Schindzielorz, A. & Haass, C. (1998).
Proteolytic fragments of the Alzheimer’s disease associated
presenilins-1 and -2 are phosphorylated in vivo by distinct cellular
mechanisms. Biochemistry 37, 5961-5967.
40. Eldar-Finkelman, H. & Krebs, E.G. (1997). Phosphorylation of insulin
receptor substrate 1 by glycogen synthase kinase 3 impairs insulin
action. Proc. Natl Acad. Sci. USA 94, 9660-9664.
41. Schulze-Gahmen, U., et al., & Kim, S.-H. (1995). Multiple modes of
ligand recognition: crystal structures of cyclin-dependent protein
kinase 2 in complex with ATP and two inhibitors, olomoucine and
isopentenyladenine. Proteins 22, 378-391.
42. De Azevedo, W.F., Mueller-Dieckmann, H.J., Schulze-Gahmen, U.,
Worland P.J., Sausville, E. & Kim, S.-H. (1996). Structural basis for
specificity and potency of a flavonoid inhibitor of human CDK2, a cell
cycle kinase. Proc. Natl Acad. Sci. USA 93, 2735-2740.
 Research Paper Protein kinase inhibition by hymenialdisine Meijer et al. 63

43. Lawrie, A.M., Noble, M.E., Tunnah, P., Brown, N.R., Johnson, L.N. & 66. Roshak, A., Jackson, J.R., Chabot-Fletcher, M. & Marshall L.A. (1997).
Endicott, J.A. (1997). Protein kinase inhibition by staurosporine Inhibition of NFκB-mediated interleukin-1b-stimulated prostaglandin
revealed in details of the molecular interaction with CDK2. Nat. Struct. E2 formation by the marine natural product hymenialdisine.
Biol. 4, 796-801. J. Pharmacol. Exp. Therap. 283, 955-96.
44. Ramachandran, G.N., Ramakrishnan, C. & Sasisekharan, V. (1963). 67. Di Martino, M., Wolff, C., Patil, A. & Nambi, P. (1995). Effects of a
Stereochemistry of polypeptide chain configurations. J. Mol. Biol. protein kinase C inhibitor (PKCI) on the development of adjuvant-
7, 95-99. induced arthritis (AA) in rats. Inflamm. Res. 2, S123-S124.
45. Schulze-Gahmen, U., De Bondt, H.L. & Kim, S.-H. (1996). High- 68. Rosenblatt, J., De Bondt, H.L., Jancarik, J., Morgan, D.O. & Kim, S.-H.
resolution crystal structures of human cyclin-dependent kinase 2 with (1993). Crystal structure of cyclin-dependent kinase 2. Purification
and without ATP: bound waters and natural ligand as guides for and crystallization of human cyclin-dependent kinase 2. J. Mol. Biol.
inhibitor design. J. Med. Chem. 39, 4540-4546. 230, 1317-1319.
46. Russo, A.A., Jeffrey, P.D. & Pavletich, N.P. (1996). Structural basis of 69. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction
cyclin-dependent kinase activation by phosphorylation. Nat. Struct. data collected in oscillation mode. Methods Enzymol. 276, 307-326.
Biol. 3, 696-700. 70. Collaborative Computational Project, Number 4. The CCP4 suite:
47. Hanks, S.K. & Hunter T. (1995). The eukaryotic protein kinase programs for protein crystallography. Acta Crystallogr. D 50, 760-763.
superfamily: kinase (catalytic) domain structure and classification. 71. Brünger, A.T. (1993). X-PLOR, Version 3.1. A System for Protein
FASEB J. 9, 576-96. Crystallography and NMR. Yale University Press, New Haven, CT.
48. Patrick, G.N., Zhou, P., Kwon, Y.T., Howley, P.M. & Tsai, L.-H. (1998). 72. Mattia, C.A., Mazzarella, L. & Puliti, R. (1982). 4-(2-amino-4-oxo-2-
p35, the neuronal activator of CDK5, is degraded by the ubiquitin- imidazolin-5-ylidene)-2-bromo-4,5,6,7-tetrahydropyrrolo-[2,3-c]azepin-
proteasome pathway. J. Biol. Chem. 273, 24057-24064. 8-one methanol solvate: a new bromo compound from the sponge
49. Lucas, F.R. & Salinas, P.C. (1997). WNT-7a induces axonal Acanthella aurantiaca. Acta Crystallogr. B 38, 2513-2515.
remodelling and increases synapsin I levels in cerebellar neurons. Dev. 73. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M.
Biol. 193, 31-44. (1993). PROCHECK: a program to check the stereochemical quality
50. Lucas, F.R., Goold, R.G., Gordon-Weeks, P.R. & Salinas, P.C. (1998). of protein structures. J. Appl. Crystallogr. 26, 283-291.
inhibition of GSK-3β leading to the loss of phosphorylated MAP-1B is 74. De Bondt, H.L., Rosenblatt, J., Jancarik, J., Jones, H.D., Morgan, D.O.
an early event in axonal remodelling induced by WNT-7a or lithium. & Kim, S.H. (1993). Crystal structure of cyclin-dependent kinase 2.
J. Cell Sci. 111, 1351-1361. Nature 363, 595-602.
51. Zheng-Fischhöfer, Q., Biernat, J., Mandelkow, E.-M., Illenberger, S., 75. Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both detailed
Godemann, R. & Mandelkow, E. (1998). Sequential phosphorylation of and schematic plots of protein structures. J. Appl. Crystallogr.
tau-protein by GSK-3β and protein kinase A at Thr212 and Ser214 24, 946-950.
generates the Alzheimer-specific epitope of antibody AT-100 and
requires a paired helical filament-like conformation. Eur. J. Biochem.
252, 542-552.
52. Braekman, J.C., Daloze, D., Stoller, C. & Van Soest, R.W.M. (1992).
Chemotaxonomy of Agelas (Porifera: Demospongiae). Biochem. Syst.
Ecol. 20, 417-431.
53. Chanas, B., Pawlik J.R., Lindel T. & Fenical W. (1996). Chemical
defense of the carribean sponge Agelas clathrodes (Schmidt). J. Exp.
Mar. Biol. Ecol. 208, 185-196.
54. Annoura, H. & Tatsuoka T. (1995). Total syntheses of hymenialdisine
and debromohymenialdisine: stereospecific construction of the
2-amino-4-oxo-2-imidazolin -5(Z)-disubstituted ylidene ring system.
Tetrahedron Lett. 36, 413-416.
55. Matsuki, S., Mizuno, A., Annoura, H. & Tatsuoka, T. (1997). Synthesis
of pyrroloazepines. Facile synthesis of 2-substituted pyrrole derivatives
by the phosgene method. J. Heterocyclic Chem. 34, 87-91.
56. Xu, Y.Z., Yakushijin, K. & Horne, D.A. (1997). Synthesis of C11N5
marine sponge alkaloids: (±)-hymenin, stevensine, hymenialdisine, and
debromohymenialdisine. J. Org. Chem. 62, 456-464.
57. Brion, J.P. & Couk, A.M. (1995). Cortical & brainstem-type Lewy
bodies are immunoreactive for the cyclin-dependent kinase 5. Am. J.
Pathol. 147, 1465-1476.
58. Nakamura, S., Kawamoto, Y., Nakano, S., Akiguchi, I. & Kimura, J.
(1998). Cyclin-dependent kinase 5 and mitogen-activated protein
kinase in glial cytoplasmic inclusions in multiple system atrophy.
J. Neuropathol. Exp. Neurol. 57, 690-698.
59. Green, S.L., Vulliet, P.R., Pinter, M.J. & Cork, L.C. (1998). Alterations
in cyclin-dependent protein kinase 5 (CDK5) protein levels, activity
and immunocytochemistry in canine motor neurone disease.
J. Neuropathol. Exp. Neurol. 57, 1070-1077.
60. Klein, P.S. & Melton, D.A. (1996). A molecular mechanism for the effect
of lithium on development. Proc. Natl Acad. Sci. USA 93, 8455-8459.
61. Hong, M., Chen, D.C., Klein, P.S. & Lee, V.M. (1997). Lithium reduces
tau phosphorylation by inhibition of glycogen synthase kinase-3.
J. Biol. Chem. 272, 25326-25332.
62. Munoz-Montano, J.R., Moreno, F.J., Avila, J. & Diaz-Nido, J. (1997).
Lithium inhibits Alzheimer’s disease-like tau protein phosphorylation in
neurons. FEBS Lett. 411, 183-188.
63. Lovestone, S., et al., & Anderton, B.H. (1999). Lithium reduces tau
phosphorylation: effects in living cells and in neurons at therapeutic
concentrations. Biol. Psychiatry 45, 995-1003.
64. Cohen, P. (1999). The Croonian lecture 1998. Identification of a Because Chemistry & Biology operates a ‘Continuous
protein kinase cascade of major importance in insulin signal Publication System’ for Research Papers, this paper has been
transduction. Phil. Trans. R. Soc. Lond. B 354, 485-495. published via the internet before being printed. The paper can
65. Breton, J.J. & Chabot-Fletcher, M.C. (1997). The natural product
hymenialdisine inhibits interleukin-8 production in U937 cells by be accessed from http://biomednet.com/cbiology/cmb — for
inhibition of nuclear factor-B. J. Pharmacol. Exp. Therap. 282, 459-466. further information, see the explanation on the contents pages.
 S1
Supplementary material
β and CK1 by
Inhibition of cyclin-dependent kinases, GSK-3β
hymenialdisine, a marine sponge constituent
L Meijer, A-M WH Thunnissen, AW White, M Garnier, M Nikolic, L-H Tsai,
J Walter, KE Cleverley, PC Salinas, Y-Z Wu, J Biernat, E-M Mandelkow,
S-H Kim and GR Pettit
Chemistry & Biology 2000, 7:51–63
Table S1 Supplementary materials and methods
Chemicals and reagents
Diffraction data collection and refinement statistics of the Sodium orthovanadate, EGTA, EDTA, RNAse A, Mops, β-glycerophos-
CDK2–HD complex. phate, phenylphosphate, sodium fluoride, glutathione-agarose, dithio-
threitol (DTT), bovine serum albumin (BSA), nitrophenylphosphate,
Space group P212121 leupeptin, aprotinin, microcystin pepstatin, soybean trypsin inhibitor,
Cell dimensions (Å) benzamidine, histone H1 (type III-S), myelin basic protein and casein
a 53.39 were obtained from Sigma Chemicals, [γ-32P]-ATP (PB 168) were
b 70.67 obtained from Amersham, roscovitine and flavopiridol from E. Sausville
c 72.26 at the National Cancer Institute.
Number of measurements (I/σ(I) > 1.0) 51723 The GS-1 peptide (YRRAAVPPSPSLSRHSSPHQSpEDEEE) was
Unique reflections 16264 synthesised by the Peptide Synthesis Unit, Institute of Biomolecular
Completeness of data to 2.10 Å (%) 98.1 Sciences, University of Southampton, Southampton SO16 7PX, UK.
Rsym* (%) 4.8
HD was isolated as described previously [S1,S2] and dissolved as a
Resolution range (Å) 30–2.10 10–50 mM stock solution in DMSO. It was diluted to 5–10 mM in
Rfactor† (%) 19.4 Me2SO just prior to use in aqueous buffers. Final DMS0 concentration
in the reaction mixture was below 1% (v/v).
Rfree‡ (%) 26.2
B values§ (Å2) Axinohydantoin [S2], HD–platinum complex, diacetylhymenialdisine,
Mainchain 32.8 diacetyl-debromohymenialdisine, dibromophakellstatin [S1], agelas-
Sidechains 35.5 tatin A [S3], dibromoageliferin, clathrodin, hymenidin, dibromocan-
Inhibitor 46.9 tharelline [S4] were prepared in the laboratory. Agelongine [S5],
Waters 35.2 dispacamide B [S6], sceptrin [S7], oroidin [S8] and stevensine
(odiline) [S5,S9] were purified from Agelas sp. and kindly provided by
Ethylene glycols 52.0
E. Fattorusso, F. Cafieri, A. Mangoni and O. Taglialatela-Scafati.
Rms deviations observed
Bond lengths (Å) 0.008 GST–retinoblastoma protein was expressed in bacteria and purified on
Bond angles (°) 1.32 glutathione-sepharose beads as described previously [S10]. Mono-
Number of water molecules 84 clonal anti-PSTAIRE and anti-cyclin B antibodies were generously
donated by M. Yamashita (Sapporo). p9CKShs1-sepharose beads were
*Rsym = Σ | I(h)- <I(h)>|/ Σ I(h), with I(h), observed intensity and <I(h>, prepared as described previously [S11]. AT-8, AT-180 and AT-100
mean intensity of reflection h over all measurement of I(h). antibodies were obtained from Innogenetics, SA, Ghent, Belgium),
factor = Σ|F0–Fc|/Σ (F0), the sums being taken over all reflections with
†R PHF-1 was a gift from P. Davies (Bronx, NY) and K9JA was from Dako
F/σ(F) > 1 cutoff. ‡Rfree = Rfactor for 10% of the data, which were not (Hamburg, Germany). SMI-31 was from Affiniti and antibodies against
included during crystallographic refinement. §B values = average B GAP-43 were a kind gift from G. Wilkin.
values for all nonhydrogen atoms.

Table S2

Buried areas of CDK2 and ligands.

CDK ligand Buried areas in CDK2 (Å) Buried areas in ligand (Å) Complimentary* % IC50 (µM)

ATP 435 340 78 –

Purvalanol 423 364 86 0.004
Hymenialdisine 335 228 68 0.022
Roscovitine 427 341 80 0.450
L868276 418 309 74 1.7
Olomoucine 360 273 76 7.0
Isopentenyladenine 290 203 70 50
*Complementarity = (buried area in ligand/buried area in CDK2) × 100. Contact surfaces calculated with the program MS [S20].
S2 Supplementary material
Buffers
Homogenization buffer. 60 mM β-glycerophosphate, 15 mM p-nitro-
phenyl-phosphate, 25 mM Mops (pH 7.2), 15 mM EGTA, 15 mM MgCl2,
1 mM DTT, 1 mM sodium vanadate, 1 mM NaF, 1 mM phenylphosphate,
10 µg leupeptin/ml, 10 µg aprotinin/ml, 10 µg soybean trypsin
inhibitor/ml and 100 µM benzamidine.
Buffer A. 10 mM MgCl2, 1 mM EGTA, 1 mM DTT, 25 mM Tris–HCl
pH 7.5, 50 µg heparin/ml.
Buffer C. homogenization buffer but 5 mM EGTA, no NaF and no pro-
tease inhibitors.
Hypotonic lysis buffer (HLB). 50 mM Tris-HCl pH 7.4, 120 mM
NaCl, 10% glycerol, 1% Nonidet-P40, 5 mM DTT, 1 mM EGTA, 20 mM
NaF, 1 mM orthovanadate, 5 µM microcystin, 100 µg/ml each of leu-
peptin, aprotinin and pepstatin.
Tris-buffered saline Tween-20 (TBST). 50 mM Tris pH 7.4,
150 mM NaCl, 0.1% Tween-20.
STM buffer. 10 mM Tris–HCl pH 8.0, 0.25 M sucrose, 10 mM MgCl2,
1 mM DTT, protease and phosphatase inhibitors [S12].
Kinase preparations and assays
Kinases activities were assayed in Buffer A or C (unless otherwise
stated), at 30°C, at a final ATP concentration of 15 µM. Blank values
were subtracted and activities calculated as pmoles of phosphate incor-
porated for a 10 min incubation (usually expressed in % of the maximal
activity, i.e. without inhibitors). Controls were performed with appropri-
ate dilutions of Me2S0. In a few cases phosphorylation of the substrate
was assessed by autoradiography after SDS–PAGE. IC50 values were
estimated from the dose-response curves. They provide a good esti-
mate, as long as the assay conditions are similar, but the accuracy
depends on the Km of each kinase. A more precise comparison of an
inhibitor’s effects on various kinases should be based on Ki values.
CDK1/cyclin B was extracted in homogenisation buffer from M phase
starfish (Marthasterias glacialis) oocytes and purified by affinity chro-
matography on p9CKShs1-sepharose beads, from which it was eluted by
free p9CKShs1 as described previously [S10,S11]. The kinase activity was
assayed in buffer C, with 1 mg histone H1/ml, in the presence of 15 µM
[γ-32P] ATP (3,000 Ci/mmol; 1 mCi/ml) in a final volume of 30 µl. After a
10 min incubation at 30°C, 25 µl aliquots of supernatant were spotted
onto 2.5 × 3 cm pieces of Whatman P81 phosphocellulose paper, and,
20 s later, the filters were washed 5 × (for at least 5 min each time) in a
solution of 10 ml phosphoric acid/liter of water. The wet filters were
counted in the presence of 1 ml ACS (Amersham) scintillation fluid.
GSK-3β was either purified from rabbit muscle or expressed in and
purified from insect Sf9 cells [S13]. It was assayed, following a 1/100
dilution in 1 mg BSA/ml 10 mM DTT, with 5 µl 40 µM GS-1 peptide as
a substrate, in buffer A, in the presence of 15 µM [γ-32P] ATP
(3,000 Ci/mmol; 1 mCi/ml) in a final volume of 30 µl. After 30 min incu-
bation at 30°C, 25 µl aliquots of supernatant were spotted onto P81
phosphocellulose papers and treated as described above.
CDK5/p25 was reconstituted by mixing equal amounts of recombinant
CDK5 and p25 expressed in Escherichia coli as GST fusion proteins and
purified by affinity chromatography on glutathione agarose (vectors kindly
provided by Dr J.H. Wang) (p25 is a truncated version of the CDK5 acti-
vator). Its activity was assayed in buffer C as described for CDK1/cyclin B.
CDK2/cyclin A, CDK2/cyclin E, CDK3/cyclin E, CDK4/cyclin D1 (pro-
vided by W. Harper), CDK6/cyclin D2 (provided by M. Meyerson), His-
tagged erk1 and erk2 (provided by M. Cobb), protein kinase C isoforms,
the catalytic subunit of cAMP-dependent protein kinase (provided by
S. Lohmann), cGMP-dependent protein kinase (provided by F. Hofmann),
Myosin light chain kinase (provided by T.J. Lukas and M. Watterson),
CK1 and 2 (provided by L. Pinna), ASK-γ (a plant homologue of GSK-3),
insulin receptor tyrosine kinase domain (CIRK-41; provided by
H.Y.L. Tung), c-raf, MAPKK (provided by D. Alessi), c-jun amino-terminal
kinase (obtained from Promega), c-src kinase and v-abl kinase (provided
by H. Mett) were assayed as described previously [S10].
Electrophoresis and western blotting
Proteins bound to p9CKShs1-Sepharose beads (starfish CDK1/cyclin B)
were denatured with 2 × Laemmli sample buffer. Samples were run on
10% SDS–PAGE. Proteins were transferred from the gel to a 0.1 µm
nitrocellulose sheet (Schleicher and Schuell) in a milliblot-SDE system
(Millipore) for 30 min at 2.5 mA/cm2 in transfer buffer. Subsequently,
the filter was blocked with 5% low fat milk in TBST for 1 h. The filter
was then washed with TBST and incubated for 1 h with the first anti-
bodies (anti-PSTAIRE, 1:2000; anti-cyclin B, 1:1000). After four
washes (1 × 20 min, 3 × 5 min) with TBST, the nitrocellulose sheet was
treated for 1 h with horseradish peroxidase-coupled secondary anti-
bodies diluted in TBST (1:1000). The filter was then washed
5 × (1 × 20 min, 4 × 5 min) with TBST and analysed by enhanced
chemioluminescence with ECL detection reagents and hyperfilm MP.
In vitro phosphorylation of presenilin-2
The large hydrophilic loop of presenilin-2 between transmembrane
domains 6 and 7 was expressed in E. coli as a presenilin-2 loop-
maltose-binding protein (MBP) fusion protein [S14]. Following affinity
chromatography purification presenilin-2–MBP was used as a substrate
for CK1. After 30 min incubation in the presence of various HD concen-
trations, the kinase reaction was stopped by addition of Laemmli
sample buffer. MBP–presenilin-2 was resolved using SDS–PAGE and
its phosphorylation level visualised by autoradiography.
In vitro and in vivo Pak1 phosphorylation by CDK5/p35
In vitro Pak1 phosphorylation. CDK5/p35 was immunoprecipitaed
from P02 rat cortices using the carboxy-terminal Santa Cruz antibody
(400 µg total protein per kinase assay). The immunoprecipitate was
split into equal aliquots and the kinase assays performed using
GST–Pak1K299 as a substrate. Roscovitine, HD or DMSO were
added to the beads just prior to the kinase assay, as described previ-
ously [S15]. Phosphorylation was monitored by autoradiography.
In vivo Pak1 phosphorylation. HD was added onto cultured neu-
rones obtained from E18 rat embryo cortices when the cultures were
4 days old. As a control DMSO was used to the maximal volume of
drug used (7.5 µl). The drug was left on the cells for 1 h. The cells were
then lysed on ice in STM buffer and the membrane fraction isolated
with STM containing 0.5% NP-40 [S15]. Pak1 immunoprecipitations
were carried out using a specific polyclonal antibody, followed by
kinase assays using histone H4 as a a substrate. The amount of H4
phosphorylation, reflecting the level of Pak1 activity, was determined by
measuring the intensity of the bands on autoradiography film and quan-
titating with NIH imaging program. As loading controls Pak1 and p35
western blots were made. The latter protein increases in amount with
the level of CDK5/p35 kinase inhibition, as shown previously [S16].
In vivo MAP-1B phosphorylation by GSK-3 and
immunochemistry
Cerebellar granule cells were isolated from newborn mice and purified
according to the method of Hatten [S17] using Percoll gradients. Cells
were plated onto dishes coated with poly-D-lysine (100 µg/ml) and
laminin (50 µg/ml) and grown in serum-free medium [S17] for 1 day,
the cultures were then treated with HD (1-100 µM) for 20 h. Cultures
were fixed with 4% formaldehyde in PBS and stored in PBS at 4oC.
Cells were permeabilised with 100% methanol and incubated with
primary antibodies against GAP-43 (a kind gift from Dr G. Wilkin) and
MAP-1B-P (SMI-31, Affiniti) overnight at 4oC. FITC and Texas Red
conjugated secondary antibodies were used (Vector). For western blot
analysis, cells were lysed in sample buffer and analysed by 8%
SDS–PAGE. Proteins were transferred to nitrocellulose membranes
and incubated with an antibody to MAP-1B-P (SMI-31) diluted in block

solution (0.1% Tween-20, 3% dried skimmed milk in TBS) for 2 h at
room temperature. HRP-conjugated secondary antibodies (Amersham)
were used and proteins were visualised using the ECL system (Pierce).
Blots were scanned using a flat bed scanner (UMAX Astra 1200S).
In vitro and in vivo Tau phosphorylation
Cells and viruses. Sf9 cells (InVitrogen, San Diego, CA) were grown
at 27°C in monolayer culture Grace’s medium (Gibco BRL, Gaithers-
burg, MD) supplemented with 10% fetal bovine serum and 50 µg gen-
tamycin/ml and 2.5 µg amphotericin/ml. BaculoGold was obtained
from PharMingen (San Diego, CA), pVL1392 from InVitrogen.
Tau transfection. The gene for htau23, the shortest human tau isoform
[S18] was excised from the bacterial expression vector pNG2 [S19]
with XbaI and BamHI, and inserted into the baculovirus transfer vector
pVL1392 cut with the same restriction endonucleases. The BaculoGold
system was used to construct the tau baculovirus containing vector. The
BaculoGold DNA is a modified type of baculovirus containing a lethal
deletion. Cotransfection of the BaculoGold DNA with a complementing
baculovirus transfer vector rescued the lethal deletion of this virus DNA
and reconstituted viable virus particles carrying the htau23 coding
sequence. Plasmid DNA used for transfections was purified using
QIAGEN cartridges (Hilden, Germany). Sf9 cells grown in monolayers
(2 × 106 cells in a 60 mm cell culture dish) were cotransfected with
baculovirus DNA (0.5 µg BaculoGold DNA) and with vector derivatives
of pVL1392 (2 µg) using a calcium phosphate coprecipitation method.
The presence of recombinant protein was examined in the infected cells
5 days post infection using SDS–PAGE and western blotting.
Tau phosphorylation in Sf9 cells. To determine the effects of kinase
inhibitors on tau phosphorylation, Sf9 cells infected with baculovirus
expressing htau23 were treated 36 h post-infection with 50 µM HD or
flavopiridol for 5 h before being harvested. To obtain control tau
samples with enhanced phosphorylation, the htau23 expressing Sf9
cells were treated with 0.2 µM okadaic acid for 5 h before harvest.
Tau western blotting. Sf9 cells were infected with recombinant virus
at a MOI of 1–5. Cell lysates were prepared in hypotonic lysis buffer
(HLB). After 15 min centrifugation at 16,000 g, the supernatant was
recovered and its NaCl concentration raised to 500 mM. It was then
boiled for 10 min and re-centriguged at 16,000 g for 15 min. Proteins
(3 µg) were resolved using SDS–PAGE, transferred to a PVDF mem-
brane, and Western blotted with the following antibodies: AT-8
(1:2000), AT-180 (1:1000), AT-100 (1:500), PHF-1 (1:600) and poly-
clonal anti-tau antibody K9JA.
Tau phosphorylation in vitro. Tau phosphorylation was performed
using purified GSK-3β and recombinant tau-32 (provided by Dr.
Goedert) as a substrate. After 30 min incubation in the presence of
various HD concentrations, under the GSK-3β assay conditions
described above, the kinase reaction was stopped by addition of
Laemmli sample buffer. Tau was resolved by SDS–PAGE and its phos-
phorylation level visualised by autoradiography.
Supplementary references
S1. Pettit, G.R., et al., & Boyd, M.R. (1997). Antineoplastic agents. 362.
Isolation and X-ray crystal structure of dibromophakellstatin from the
Indian Ocean sponge Phakellia mauritiana. J. Nat. Prod. 60, 180-183.
S2. Pettit, G.R., et al., & Camou, F. (1990). Antineoplastic agents. 168.
Isolation and structure of axinohydantoin. Can. J. Chem. 68, 1621-1624.
S3. D’ambrosio, M., et al., & Pietra, F. (1993). Agelastatin A, a new
skeleton cytotoxic alkaloid of the oroidin family. Isolation from the
Axinellid sponge Agelas dendromorpha of the Coral Sea. J. Chem.
Soc. Chem. Commun., 1305-1306.
S4. De Nanteuil, G., et al., & Laboute, P. (1985). Invertébrés marins du
lagon Neo-Caledonien.V. Isolement et identification des métabolites
d’une nouvelle espèce de spongiaire, Pseudaxinyssa Cantharella
[Marine invertebrates of the Neo-Caledonian lagoon. V. Isolation and
identification of metabolites from a new species of sponge,
Pseudaxinyssa Cantharella]. Tetrahedron 41, 6019-6033.
Supplementary material S3
S5. Cafieri, F., Fattorusso, E., Mangoni, A., & Taglialatela-Scafati O.
(1995). A novel bromopyrrole alkaloid from the sponge Agelas
longissima with antiserotonergic activity. Bioorg. Med. Chem. Lett.
5, 799-804.
S6. Cafieri, F., Fattorusso, E., Mangoni, A., & Tagliatela-Scafati, O. (1996).
Dispacamides, anti-histamine alkaloids from Carribean Agelas
sponges. Tetrahedron Lett. 37, 3587-3590.
S7. Walker, R.P., Faulkner, D.J., Van Engen, D., & Clardy, J. (1981).
Sceptrin, an antimicrobial agent from the sponge Agelas sceptrum.
J. Am. Chem. Soc. 103, 6772-6773.
S8. Garcia, E.E., Benjamin, L.E., & Fryer, R.I. (1973). Reinvestigation into
the structure of oroidin, a bromopyrrole derivative from marine sponge.
J. Chem. Soc. Chem. Commun., 78-79.
S9. Albizati, K.F., & Faulkner D.J. (1985). Stevensine, a novel alkaloid of an
unidentified marine sponge. J. Org. Chem. 50, 4163-4164.
S10. Meijer, L., et al., & Moulinoux, J.P. (1997). Biochemical and cellular effects
of roscovitine, a potent and selective inhibitor of the cyclin-dependent
kinases cdc2, cdk2 and cdk5. Eur. J. Biochem. 243, 527-536.
S11. Borgne, A., Ostvold, A.C., Flament, S., & Meijer, L. (1999). Intra-M
phase-promoting factor phosphorylation of cyclin B at the
prophase/metaphase transition. J. Biol. Chem. 274, 11977-11986.
S12. Nikolic, M., Chou, M.M., Lu, W., Mayer, B.J., & Tsai, L.-H. (1998). The
p35/cdk5 kinase is a neuron-specific Rac effector that inhibits Pak1
activity. Nature 395, 194-198.
S13. Hughes, K., Pulverer, B.J., Theocharous, P., & Woodgett, J.R. (1992).
Baculovirus-mediated expression and characterisation of rat glycogen
synthase kinase-3 beta, the mammalian homologues of the Drosophila
melanogaster zeste-white 3sgg homeotic gene product. Eur. J.
Biochem. 203, 305-311.
S14. Walter, J., Grünberg, J., Schindzielorz, A., & Haass, C. (1998).
Proteolytic fragments of the Alzheimer’s disease associated
presenilins-1 and -2 are phosphorylated in vivo by distinct cellular
mechanisms. Biochemistry 37, 5961-5967.
S15. Nikolic, M., Chou, M.M., Lu, W., Mayer, B.J., & Tsai, L.-H. (1998). The
p35/cdk5 kinase is a neuron-specific Rac effector that inhibits Pak1
activity. Nature 395, 194-198.
S16. Patrick, G.N., Zhou, P., Kwon, Y.T., Howley, P.M., & Tsai, L.-H. (1998).
p35, the neuronal activator of CDK5, is degraded by the ubiquitin-
proteasome pathway. J. Biol. Chem. 273, 24057-24064.
S17. Hatten, M.E. (1985). Neuronal regulation of astroglial morphology and
proliferation in vitro. J. Cell Biol. 100, 384-396.
S18. Goedert, M., Spillantini, M., Jakes, R., Rutherford, D., & Crowther, R.A.
(1989). Multiple isoforms of human microtubule-associated protein-
tau: sequences and localization in neurofibrillary tangles of Alzheimers-
disease. Neuron 3, 519-526.
S19. Biernat, J., Gustke, N., Drewes, G., Mandelkow, E.-M., & Mandelkow, E.
(1993). Phosphorylation of serine 262 strongly reduces the binding of
tau protein to microtubules: distinction between PHF-like
immunoreactivity and microtubule binding. Neuron 11, 153-163.
S20. Jones, T.A., Zou, J.-Y., Cowan, S.W., & Kjeldgaard, M. (1991).
Improved methods for building protein models in electron density
maps and the location of the errors in these models. Acta
Crystallogr. A 47, 110-119.