1382?The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
A zebrafish model of tauopathy
allows in vivo imaging of neuronal
cell death and drug evaluation
Dominik Paquet,1,2 Ratan Bhat,3 Astrid Sydow,4 Eva-Maria Mandelkow,4
Stefan Berg,3 Sven Hellberg,3 Johanna Fälting,3 Martin Distel,5
Reinhard W. Köster,5 Bettina Schmid,1,2 and Christian Haass1,2
1Deutsches Zentrum für Neurodegenerative Erkrankungen (DZNE) and 2Adolf-Butenandt-Institute,
Biochemistry, Ludwig-Maximilians-University, Munich, Germany.
3AstraZeneca Research & Development, Södertälje, Sweden. 4Max-Planck Unit for Structural Molecular Biology at DESY, Hamburg, Germany.
5Helmholtz Zentrum München, Institute of Developmental Genetics, Neuherberg, Germany.
Neurodegenerative diseases are the most frequent cause of
dementia in our aging society. For these disorders, which include
Alzheimer disease (AD) and frontotemporal dementia (FTD),
disease-modifying treatments represent a highly unmet medical
need. AD and FTD are characterized by posttranslationally modi-
fied amyloidogenic proteins, which form neurotoxic oligomers
and are finally deposited as insoluble aggregates (1). Examples of
the proteinaceous building blocks of these deposits are amyloid
β peptide in AD and TAU in AD and FTD (2, 3). The TAU pro-
tein is an important target for research and drug development,
since its pathological alterations strongly correlate with disease
progression in AD and FTD and other neurodegenerative dis-
eases (4) and TAU suppression improves memory function (5).
Furthermore, mutations in the TAU-encoding gene microtubule-
associated protein TAU (MAPT) have been found in patients with
FTD with Parkinsonism linked to chromosome 17 (FTDP-17)
(6). One of the first characteristic modifications of TAU is its
pathologic phosphorylation at multiple residues, which are dis-
tributed over the whole protein (3, 7). The initial steps, which
lead to TAU phosphorylation and subsequent aggregation, were
mainly studied in cell culture but are still poorly understood in
vivo. These early stages are difficult to investigate in transgenic
mice, since it is not feasible to image cells accurately over longer
periods during early disease stages (8). Furthermore, vertebrate
in vivo models for rapid high-throughput screening for inhibi-
tors of TAU pathology are not available.
Here, we propose the zebrafish as a small vertebrate model of
tauopathies and other neurodegenerative diseases. It offers genetic
tractability in combination with a translucent embryo allowing
imaging of disease progression at cellular and subcellular levels
in the living animal. Furthermore, the potential of using small
zebrafish embryos for large-scale drug screening has already been
demonstrated (9). Large amounts of small fish, which can be ana-
lyzed in a 96-well format, can be rapidly produced. Fish are kept
in water, which facilitates uptake of substances from the aqueous
medium, allowing for automated compound application. In addi-
tion, simple readouts of characteristics such as altered mobility
and fluorescence or changes in biochemical composition of pro-
teins can be analyzed on a large scale. Moreover, the endothelial
blood-brain barrier (BBB) begins to become functional 3 days
after fertilization in zebrafish and has a structure similar to that in
higher vertebrates, allowing one to test for bioavailability of drugs
in the nervous system (10).
We have optimized the transgenic expression of the human pro-
tein TAU-P301L (a mutation genetically linked to FTD; ref. 6) in
zebrafish neurons by a newly designed Gal4–upstream activating
sequence–based (Gal4/UAS–based) vector system, which also great-
Conflict?of?interest: D. Paquet, A. Sydow, E.-M. Mandelkow, M. Distel, R.W. Köster,
and B. Schmid declare no conflict of interest. R. Bhat, S. Berg, S. Hellberg, and J. Fält-
ing are employees of AstraZeneca R&D. C. Haass is a consultant for Elan Pharmaceu-
ticals and receives research support from Boehringer Ingelheim KG.
Nonstandard?abbreviations?used: AD, Alzheimer disease; BBB, blood-brain
barrier; CDK2, cyclin-dependent kinase 2; dN-GSK3β, dominant-negative
Xenopus GSK3β-K85R; FTD, frontotemporal dementia; GSK3β, glycogen syn-
thase kinase 3β; hpf, hours post fertilization; IF, immunofluorescence; IHC,
immunohistochemistry; Ins, insert; MCS, multiple cloning site; UAS, upstream
activating sequence; WB, Western blot.
Citation?for?this?article: J. Clin. Invest. 119:1382–1395 (2009). doi:10.1172/JCI37537.
?The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
ly facilitates identification of the transgenic fish by a simultane-
ously expressed fluorescent reporter. We have further advanced the
system to allow introduction of various other genes and promoters
to facilitate the generation of different zebrafish disease models. In
contrast to previous studies (11, 12), we were able to monitor early
pathology, including disease-specific hyperphosphorylation and
conformational changes of TAU, neuronal and behavioral abnor-
malities as well as increased neuronal cell death within the first
days of embryonic development in stable transgenic zebrafish. The
rapid appearance of a pathologic phenotype allows one not only
to use the embryos to study disease progression in a transparent
vertebrate but also to validate and even screen on a relatively large
scale for compounds that modify early phenotypes, such as TAU
hyperphosphorylation, in vivo.
Hyperphosphorylation of TAU is believed to be a key initiator of
detachment of normal TAU from microtubules and subsequent
oligomerization and aggregation. Several kinases, including gly-
cogen synthase kinase 3β (GSK3β), cyclin-dependent kinase 5
(CDK5), ERK2, and microtubule affinity–regulating kinase
(MARK), directly phosphorylate TAU and are therefore consid-
ered as key therapeutic targets (13). Inhibitors that selectively
block these enzymes could therefore be used to slow the progres-
sion of disease pathology. Over the last years, selective inhibitors
for kinases involved in TAU phosphorylation have been identified,
and some of these compounds have even been tested in TAU trans-
genic mice, where they showed some effects on TAU phosphoryla-
tion. The activity of these inhibitors, however, has been hampered
by poor water solubility and bioavailability and substantial side
effects (13). In addition, large quantities of the compounds are
needed, since long-term administration is usually required to
observe potential efficacy. Therefore, there is a significant need
for technologies that can rapidly and efficiently identify more
active and specific compounds with optimized bioavailability in
in vivo models that represent disease pathology. We have shown
previously that novel substances can be developed by structure-
based design, which uses crystal structures of proteins, to model
small molecules into the active site of the targeted enzyme (14).
This approach can deliver inhibitors with excellent potency and
specificity. However, it requires validating the bioavailability and
selective inhibition of the target molecule in an in vivo setting. We
have now developed several highly specific inhibitors by structure-
based design, which bind to the active site of GSK3β and block the
enzyme function with high potency in vitro. Two of the best inhib-
itors were then investigated for their potential efficacy in zebraf-
ish. We demonstrate here for the first time, to our knowledge, that
we could reduce the pathologic phosphorylation of the human
disease–associated protein TAU in vivo in a transgenic zebrafish
model by several published GSK3β inhibitors. Furthermore, we
tested 2 newly developed highly selective GSK3β inhibitors with
potent in vitro activity and could show with our model that one of
them is also very potent in a whole organism in vivo.
Our study not only provides a small vertebrate model of tauop-
athies and other neurodegenerative diseases, which can be used
to monitor progression of pathology in vivo, but also a powerful
combination of in silico prediction of compounds with in vivo
validation of their bioavailability in a small transgenic vertebrate
disease model. Finally, our newly developed zebrafish transgenesis
vectors will allow modeling of numerous human disorders that
are based on protein misfolding and overexpression, 2 phenomena
often connected to each other (15).
A Gal4/UAS–based bidirectional expression system in zebrafish. The
first goal of this study was to generate transgenic zebrafish stably
expressing the FTD-associated human protein TAU-P301L (6).
Inefficient transgenesis and low protein expression levels have in
the past hampered the efficient generation of zebrafish disease
models expressing human genes. To overcome these limitations,
we have generated vectors that combine several features to increase
transgenesis rates and protein levels, facilitate the expression of dis-
ease proteins of interest, and allow efficient identification of trans-
genic lines as well as simultaneous in vivo monitoring of pathology
and phenotypes. We cloned 2 constructs, Driver and Responder,
based on the medaka Tol2 transposable element (16), which greatly
increases the rate of transgenesis (see Methods for details), and inte-
grated the Gal4/UAS expression system (17) into the 2 vectors (Fig-
ure 1A). Furthermore, we introduced Gateway recombination sites,
which allow rapid introduction of other genes and promoters (18)
(see Supplemental Figure 1 for details; supplemental material avail-
able online with this article; doi:10.1172/JCI37537DS1). The Driver
construct contains the neuronal promoter HuC (19), controlling
the expression of a Gal4-VP16 fusion protein, which in turn effi-
ciently transactivates and amplifies protein expression from a UAS
on the Responder construct. To achieve transgene expression in 2
orientations, we flanked the UAS sequence with 2 short minimal
promoters. In our constructs, this cassette drives the expression of
human TAU-P301L in one direction and the expression of the fluo-
rescent reporter DsRed in the other (Figure 1A). This bidirectional
expression allows the identification of TAU-expressing cells in live
embryos by concomitant DsRed fluorescence.
Transgenic fish were generated by injecting circular Driver and
Responder constructs together with Tol2 transposase mRNA,
which is translated into active transposase (a protein not encoded
by the zebrafish genome) in embryonic cells to catalyze integration
of both constructs into the zebrafish genome for a short period
of time (16). Both constructs integrate randomly into a subset of
embryonic cells leading to mosaic TAU- and DsRed-expressing
embryos. DsRed-positive embryos are raised and outcrossed to
wild-type fish. The offspring of founder fish with germ-line trans-
mission can be easily identified, as the embryos express DsRed in
mature neurons, making PCR screenings dispensable (Figure 1B).
The expression of TAU and DsRed in the transgenic zebrafish fully
overlaps, as shown by immunofluorescence (IF) staining using the
pan-TAU antibody T46 (20) and DsRed antibodies (Figure 1C).
We raised 76 injected founder fish to sexual maturity and iden-
tified 15 (19.7%) with DsRed-positive offspring. We analyzed 3
generations descending from 1 of the founder fish for genetic
inheritance by counting DsRed-negative and -positive embryos
and always found about one-fourth of the offspring to be DsRed
positive, implying that these embryos carry Driver and Responder
constructs (Supplemental Figure 2A). This ratio indicates inde-
pendent inheritance of both constructs, with single or multiple
insertions at 2 different genomic loci. We verified this by analyzing
225 embryos of an F2 outcross by PCR for Driver and Responder
genotypes (Supplemental Figure 2B). In addition, the embryos
showed stable protein levels and expression domains over 3 genera-
tions (Supplemental Figure 2C), indicating stable inheritance and
activity of the Gal4/UAS transgene. Finally, we did not observe any
morphological alterations in the transgenic fish (data not shown)
and also excluded alterations in the function of the zebrafish BBB
(for details see Supplemental Figure 3).
1384?The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
The transgenic zebrafish rapidly develop disease-specific alterations. Phos-
phorylation of TAU at certain serine and threonine residues serves as
a biochemical marker for pathologic alterations in AD and FTD (3, 7).
We therefore determined whether and when the TAU transgenic
fish recapitulate this hallmark of disease progression. Strikingly, in
embryos that were only 32 hours old, we could already detect posi-
tive immunoreactivity in spinal cord neurons with the antibodies
AT180 (21), AT270 (21), 12E8 (22), PHF1 (23), 422 (24), and AT8
(25), which specifically detect abnormally phosphorylated residues
T231/S235, T181, S262/S356, S396/S404, S422, and S202/T205,
respectively (Figure 2A). These findings were corroborated by immu-
noblots of lysates from 48-hour-old transgenic and nontransgenic
fish. The pan-TAU antibody T46 reveals a broad band around
64 kDa, which corresponds to the largest CNS TAU isoform in a
mixed state of phosphorylation. Human TAU was also recognized
by antibodies against phosphorylated epitopes that are typically
elevated in AD (AT180, AT270, 12E8, PHF1, and 422) (Figure 2B).
In addition to pathologic hyperphosphorylation, TAU also chang-
es its conformation during disease progression (26), which finally
culminates in aggregation and formation of tangles. The early con-
formational changes can be monitored by immunostaining using
antibody MC1, which is specific for a pathologic conformation of
TAU (26). Strikingly, we already found MC1-positive TAU in neu-
rons of 32-hour-old embryos, again demonstrating rapid occurrence
of AD- and FTD-like pathology in our transgenic zebrafish model
(Figure 2C). We also determined whether and when these changes to
pathologic conformation led to the formation of tangles and there-
fore analyzed later stages. In 5-week-old fish, which are still quite
transparent, we could not only observe the expression of the DsRed
transgene in living fish (Figure 2D) but also depict TAU-express-
ing cells with the pathologic AT8 epitope in immunohistochemical
stainings (Figure 2E) and Gallyas silver–positive tangles (Figure 2F)
in paraffin sections of spinal cord tissue. To demonstrate specific-
ity of the stainings, we also stained control transgenic lines, which
express DsRed only, and nontransgenic siblings. No cross-reactivity
in neurons was observed (Figure 2F and Supplemental Figure 4).
Rapid progression of AD/FTD-like late-stage pathology in TAU transgenic
fish. We have observed remarkable differences in the amount of TAU-
expressing cells, which are stained by early pathology-marking anti-
bodies, such as AT180, AT270 and 12E8, versus late markers, such as
A Gal4/UAS–based bidirectional expression system in
zebrafish. (A) The Driver construct contains the neuronal
zebrafish promoter HuC driving the expression of Gal4-
VP16, which binds to the UAS on the Responder con-
struct. Here, it activates the bidirectional expression of
hTAU-P301L and DsRed via the minimal promoters.
UAS-dependent gene expression of TAU and DsRed
is indicated in living fish by DsRed fluorescence. Driver
and Responder constructs are flanked by Tol2 transpo-
son sites. (B) To generate transgenic fish, the Driver and
Responder constructs were mixed and injected together
with Tol2 mRNA. The mRNA is translated to active trans-
posase, which detects the flanking Tol2 elements and
catalyzes random integration into the zebrafish genome
in a subset of embryonic cells for a short time period, gen-
erating mosaic founder embryos. Mosaic DsRed-positive
larvae were raised and outcrossed with wild-type fish. A
subset of the offspring will be transgenic and can be eas-
ily identified and sorted by DsRed-positive neurons. Scale
bar: 1 mm. (C) Double immunostainings for total TAU
(T46 antibody) and DsRed of 32-hpf transgenic zebrafish
embryos expressing hTAU-P301L and DsRed. Trans-
genic embryos express both hTAU-P301L and DsRed
in spinal cord neurons, showing effective bidirectional
expression from the Responder construct. Lateral views
of the trunk above the end of the yolk extension, anterior
to the left. Scale bar: 20 μm.
? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
422 and AT8, in 32-hour-old transgenic zebrafish. While the epitopes
that are detected by early marker antibodies are present in most TAU-
expressing cells, the signals of the late markers can only be found in
a small subset of neuronal cells (Figure 2A). This observation raises
the question of whether there is a progression to advanced pathology
that these immunopositive neurons have already reached in contrast
to the surrounding neurons and whether other neurons will follow
later. To monitor progression of the pathological AT8 immunoreac-
tivity, we compared the number of positive neurons in 24-hour-old,
32-hour-old, 48-hour-old, and 7-day-old embryos from a single trans-
Expression of hTAU-P301L induces rapid pathological hyperphosphorylation, conformational changes, and aggregation of TAU in transgenic
zebrafish. (A) Double whole-mount immunostainings for phosphorylated and total TAU of 32-hour-old transgenic zebrafish embryos expressing
hTAU-P301L and DsRed. TAU is phosphorylated in spinal cord neurons at residues Thr231/Ser235 (AT180), Thr181 (AT270), Ser262/Ser356
(12E8), Ser396/Ser404 (PHF1), Ser422 (422), and Ser202/Thr205 (AT8). (B) WBs of total and phosphorylated TAU of 48-hour-old transgenic
zebrafish embryos expressing hTAU-P301L and DsRed or DsRed alone or nontransgenic siblings. Phosphorylated TAU is only detected in
TAU/DsRed transgenic embryos. No cross-reacting bands are detectable in controls at the same molecular weight. In addition to the specific
band above 64 kDa (arrow) in TAU-positive embryos, antibodies PHF1 and 422 detect a nonspecific band at lower molecular weight in all
embryos (asterisk). (C) TAU changes its conformation to a pathologic state, as shown in whole-mount immunostainings, by using the conforma-
tion-specific antibody MC1, in most neurons of 32-hour-old embryos expressing the TAU transgene. (D) Side views of 5-week-old living zebrafish
in bright field and DsRed illumination. The fish are still rather translucent, and the transgene-expressing cells can be detected by their red fluo-
rescence (arrowheads). (E) Immunohistochemical stainings of spinal cord paraffin sections of the same 5-week-old TAU transgenic zebrafish
for total TAU (antibody DA9) and pathologically phosphorylated TAU (AT8) (arrowheads). (F) In addition, tangles are observed by Gallyas silver
staining in sections of the same 5-week-old TAU transgenic zebrafish (arrowheads). Scale bars: 20 μm.
1386? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
genic line, with comparable levels of transgene expression. We could
indeed observe a rapid and robust increase in AT8-positive neurons
over this short period of time (Figure 3A). This was not the case for
AT270-positive cells (Figure 3B).
Increased cell death in hTAU-expressing zebrafish neurons. Neuronal
cell death is the ultimate reason for the neurological deficits of AD
and FTD patients. We have therefore engaged in analysis to deter-
mine whether neurons indeed degenerate when TAU is expressed
in transgenic zebrafish. Cell death can be monitored in living fish
by incubation in acridine orange (27), a dye that stains nucleic
acids in dying cells. We have compared 6-day-old fish expressing
TAU/DsRed to DsRed only–expressing fish and nonexpressing sib-
lings. Strikingly, we could detect a significant increase of cell death
in the whole spinal cord of TAU-expressing fish, as compared with
both controls (Figure 4, A–D).
To analyze neurodegeneration in more detail and to demonstrate
the suitability of zebrafish larvae to study cellular processes in a whole
living animal in vivo, we monitored the neurodegeneration in TAU
transgenic fish by confocal time-lapse imaging. We recorded neu-
rons in the spinal cord that express DsRed over a period of 12 hours
(typical examples of still images are shown in Figure 4E; see also
Supplemental Video 1) and searched for dying cells by monitoring
uptake of acridine orange. An intact neuron in the field of view first
altered its shape and started to round up (54.0 to 58.1 hours post fer-
tilization [hpf]). Subsequently, this cell fragmented and took up acri-
dine orange, indicating a breakdown of the cellular membranes (58.1
to 60.3 hpf), and eventually began to disappear (60.3 to 65.7 hpf).
To our knowledge, this is the first demonstration of in vivo cell
death imaging in the field of neurodegeneration.
Expression of hTAU-P301L causes neuronal and behavioral abnormalities.
It has been previously demonstrated that elevation of TAU leads to
inhibition of intracellular transport with toxic consequences, which
are particularly pronounced in long-projecting motoneurons (28, 29).
This is accompanied by elevated pathologic phosphorylation. To
investigate abnormalities in neuronal morphology, we stained TAU
transgenic fish and control fish with antibody znp1, which labels
Rapid progression of AT8-positive late-stage FTD/AD-like pathology in transgenic embryos. (A) Double whole-mount immunostainings for TAU
phosphorylated at the AT8 epitope (a late marker of pathology) and the AT270 epitope (an early marker) together with staining by antibody
K9JA, which shows expression of total TAU in 24-, 32-, and 48-hpf and 7-day-old transgenic zebrafish embryos. There are very few AT8-posi-
tive neurons in the spinal cord of 24-hour-old embryos, while this number increased significantly in 32-hour-old embryos and rose even further
in 48-hour-old embryos. Finally, all TAU-expressing neurons contained the AT8 epitope in 7-day-old larvae. (B) In contrast, immunoreactivity of
the early marker AT270 was already strong in many TAU-expressing cells at 24 hpf and became only slightly stronger at older stages. The level
of total TAU detected in the same embryos by double staining with the K9JA antibody was similar between 24 hpf and 32 hpf. After 48 hours, a
substantial part of TAU was transferred to the neuronal projections. In contrast, AT8-positive TAU remained mainly in the cell bodies. The AT8-
positive somata combined with a lack of strongly AT8-positive neuronal projections is consistent with the pathological accumulation of modified
TAU in neuronal cell bodies of patients. Lateral views of the trunk above the end of the yolk extension, anterior to the left. Scale bars: 50 μm.
? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
the synaptic protein synaptotagmin in extending axons of all pri-
mary motoneurons (30, 31). We measured the length of the first
4 outgrowing caudal primary (CaP) motoneurons anterior to the
end of the yolk extension, which leave the spinal cord ventrally and
grow around the muscle in a stereotypic, time-dependent manner
as judged by synaptotagmin (znp1) staining (32). The length of the
developing motoneurons, in which synaptotagmin is present, is sub-
stantially shorter at 28 hpf in TAU transgenic fish (Figure 5, A–C).
In a later stage at 48 hpf, distribution of synaptotagmin is still
altered (Figure 5, D and E), although the projections have grown
further. Seven-day-old larvae are able to swim and catch food (data
not shown). On the cellular level, the motoneurons in both control
and TAU fish have grown around the muscle to the midline (Figure
5, F and G); however, the fine projections to the muscle are still
reduced in the TAU transgenic fish in comparison with controls.
Consistent with the altered motoneuron morphology, we
observed behavioral deficits, such as slow or absent movement in
most of the larvae at 48 hpf, an age at which a stereotypic escape
response behavior can be evoked by applying a touch stimulus to
the animals. We have collected pools of transgenic fish express-
ing high levels of TAU/DsRed and quantified the escape response
of 50 randomly picked individual larvae at 48 hpf. While most
transgenic DsRed only–expressing fish with comparable expres-
sion levels or nontransgenic fish responded to a touch stimulus at
the dorsal tip of the tail with a stereotypic escape response, most
TAU-expressing fish show a significantly reduced or even absent
response (Figure 5, H, I, and J; see also Supplemental Video 2).
We hypothesize that the motoneuron defect is the reason for the
observed movement deficits at this age.
Generation of potent GSK3β inhibitors by structure-based design. Our
transgenic zebrafish rapidly recapitulated key disease markers for
tauopathies, including hyperphosphorylation. We therefore deter-
mined whether our fish would be a useful tool for developing com-
pounds that could delay or stop the pathology of tauopathies. To
identify inhibitors of TAU kinases by structure-based design, we
refined a previously described approach (14) and found approxi-
mately 2000 actives in a high-throughput screening campaign
against the human GSK3β enzyme. The actives were confirmed
for dose response and encompassed hits from multiple chemical
series. Several compounds from the pyrazine chemical series were
cocrystallized with the GSK3β protein. Data from x-ray analysis of
the binding of the inhibitors within the ATP pocket of the GSK3β
protein led to the optimization of potency and selectivity of the pyr-
azines. Since CDK2 is the closest homolog to GSK3β, we optimized
for CDK2 selectivity aided by structure-based design. By optimizing
against CDK2, pan-kinase selectivity (27 kinases) was also obtained.
This understanding of structural activity relationships subsequently
led to the design of potent and selective GSK3β inhibitors AR-164
and AR-534 (Figure 6, A–C). AR-164 and AR-534 inhibit recombi-
nant human GSK3β with Ki values (the concentration of inhibi-
tor that reduces the formation speed of the metabolite by 50% at
low substrate concentration) of 8.9 nM and 2 nM respectively. In
contrast, AR-164 and AR-534 do not significantly inhibit CDK2/
cyclin E (Kis of 1440 nM and 100 nM, respectively), demonstrating
at least a 50-fold selectivity versus CDK2. Both inhibitors were cell
permeable. The permeability coefficients, which were determined in
Caco2 cells and in an in vitro BBB assay, were 11 × 10–6 cm/s and
3.9 × 10–3 cm/min, respectively, for AR-164 and 20 × 10–6 cm/s and
In vivo imaging of neuronal cell
death in TAU-expressing zebrafish.
(A–D) Side views of 6-day-old liv-
ing zebrafish larvae expressing
TAU/DsRed (A, see single acridine
orange–positive neuron in inset,
depicted by arrowhead), only DsRed
(B), or no transgene (siblings, C)
stained with acridine orange. TAU-
expressing fish show substantial
cell death while DsRed-express-
ing and nontransgenic fish show
only a low, basal amount of dying
cells (quantified in D). AO, acridine
orange. Data represent mean ± SD.
***P < 0.01. Scale bars: 100 μm; 10 μm
(insets). (E) Still images of several
time points of a time-lapse video,
showing a close-up of TAU-express-
ing neurons in the spinal cord of trans-
genic zebrafish, which were stained
with acridine orange. An intact neu-
ron (white arrowhead) with an axon
first changes its shape and rounds
up. Subsequently, fragmentation and
uptake of acridine orange is observed
(yellow arrowhead; see also Supple-
mental Video 1). Scale bars: 10 μm.
1388? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
12 × 10–3 cm/min respectively for AR-534 (Figure 6, B and C). These
data suggest that the bioavailability of AR-534 in the brain is most
likely higher than that of AR-164. Both AR-164 and AR-534 inhibit
the phosphorylation of TAU at Ser396 in a dose-dependent fashion
in 3T3 fibroblasts engineered to stably express 4-repeat TAU pro-
tein, exhibiting IC50 values of 49 nM and 79 nM, respectively (Figure
6, D and E). Effects were compared with those of the previously pub-
lished GSK3 inhibitors SB-415286 (33) and lithium chloride, which
exhibited IC50 values of 6 μM and 1.5 mM, respectively, in our assay.
Although AR-164 and AR-534 potently block phosphorylation of
human TAU expressed in cell culture and display a sufficiently good
BBB permeability in our in vitro test, this does not necessarily pre-
dict that the compounds are also active in a whole organism. Vali-
dating the in vivo activity of the inhibitors in TAU transgenic mice,
which were the only available vertebrate models of tauopathies so
far, is an expensive and time-consuming process and not amenable
to high-throughput screening for optimizing compounds for deter-
mining structural activity relationships. We have therefore studied
the conservation of the targeted enzyme in zebrafish, in which drug
screening would be more feasible, and found that GSK3β is highly
conserved, with over 90% identity at the amino acid level (Figure
6F). Furthermore, the residues of the active site, which interact with
our inhibitors, are completely identical. This high conservation to
the human GSK3β kinase allows using the zebrafish as a model to
validate the in vivo activity of GSK3β inhibitors and helps to further
develop this lead structure.
Expression of hTAU-P301L causes abnormalities in neuronal morphology and behavior. (A–G) Whole-mount immunostainings for znp1, which
labels synaptotagmin in the extending axons of primary motoneurons. Expression of TAU causes a dramatically reduced extension of znp1-
positive motoneurons already at 28 hpf (A and B). The difference in motoneuron length is quantified (C) by measuring the length of the first
4 znp1-stained motoneuron projections (marked 1, 2, 3, 4 in A and B) before the end of the yolk extension (marked by dotted line). Triangles and
diamonds represent values from individual motoneurons; colored horizontal lines represent mean ± SD. ***P < 0.01. In 48-hour-old embryos,
the motoneurons have grown further in both TAU fish and controls, but the motoneuron extensions are still reduced (D and E). 5 days later, the
motoneurons have grown around the muscle in both TAU fish and controls (F and G). The fine projections of motoneurons (see enlarged insets,
arrowheads), are still highly reduced in TAU transgenic fish. Lateral views of the trunk above the end of the yolk extension, anterior to the left.
Scale bars: 50 μm. (H–J) The stereotypic escape response, which is normal in DsRed-expressing (H) and nontransgenic larvae at 48 hpf (data
not shown), is highly reduced or absent in TAU-expressing larvae (I; see also Supplemental Video 2). The phenotype was quantified in groups
of 50 TAU/DsRed versus DsRed transgenic larvae, which were pooled from several clutches and selected for strong and comparable DsRed
expression (J). Error bars represent mean ± SD; ***P < 0.01.
? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
Chemical structure, design, and characteristics of GSK3 inhibitors AR-164 and AR-534. (A) Left panel shows surface representation of the x-ray
structure of GSK3β with inhibitor AR-164 in the active site. Polar areas are colored blue and lipophilic areas red. The methylpiperazine sulfone
amide extends out toward the solvent area. Right panel shows top view of the active site of GSK3β. Dotted yellow lines represent hydrogen
bonds between the protein backbone and AR-164. The 6-membered aromatic pyrazine moiety binds together with its anilino function to the back-
bone of the kinase. The pyridine ring binds to the conserved salt bridge formed by Lys85 and Asp200 (x-ray resolution: 2.47 Å). AR-534 binds in
a similar way (not shown). (B and C) Chemical structure of AR-164 and AR-534 and compound characterization values including Ki (mean from 3
independent experiments performed in duplicate), selectivity over CDK2, solubility and permeability coefficient Pe (cm/min) over cell membranes
(Caco2 cells), and BBB. (D) AR-164 and AR-534 inhibit pSer396 TAU phosphorylation in 3T3 fibroblasts harvested at 4 hours after treatment
in comparison with total TAU, as analyzed by quantitative WB. (E) Graphical representation with IC50 values of AR-164 and AR-534 effects on
inhibition of TAU phosphorylation compared with SB-415286 and LiCl. (F) Alignment of protein sequences of human and zebrafish GSK3β. Over
90% of the amino acids are identical; the residues in the active site of the enzyme, which interact with the inhibitors, are completely conserved.
1390? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
Potent in vivo effects of a newly developed GSK3β inhibitor. After dem-
onstrating that the TAU transgenic zebrafish display a number of
pathological symptoms that are diagnostic for tauopathies, we used
the fish to investigate in vivo 2 known GSK3 inhibitors, SB-216763
and SB-415286 (33), as well as our 2 newly developed compounds,
AR-534 and AR-164, which have a strong and selective inhibitory
activity on GSK3β in vitro (see Figure 6). In addition, we treated
transgenic fish with LiCl as described (34), as a positive control.
GSK3 inhibitors phenocopy a genetic loss of GSK3β and reduce pathologic hyperphosphorylation of TAU in vivo. (A) GSK3 inhibitors have
specific effects on early wild-type zebrafish development when treatment occurs between 4 and 24 hpf. LiCl treatment caused strongly reduced
eye formation (black arrowheads), while SB-216763 and SB-415286 also perturbed formation of brain, somites, and tail (gray arrowheads). In
addition to these phenotypes, AR-534 also caused multiple ear formation (arrowheads in enlargement). In contrast, AR-164, which has in vitro
activity comparable to that of AR-534, did not cause any detectable change in phenotype even at 4-fold concentration. Embryos injected with
dN-GSK3β mRNA to suppress the function of endogenous GSK3β phenocopied alterations of compound-treated fish; however, only AR-534 was
potent enough to phenocopy the ear multiplication seen in embryos expressing high levels of dN-GSK3β (arrowheads in enlargement). Embryos
injected with wild-type GSK3β did not display a phenotype at comparable mRNA concentrations. Scale bars: 500 μm; 100 μm (insets). (B) GSK3
inhibitors reduce pathologic TAU phosphorylation in vivo when applied to TAU transgenic fish between 20 and 100 hpf. Phosphorylated TAU
was detected by WB using antibody PHF1 and normalized to total TAU (K9JA). The band intensities were compared with a DMSO control on the
same gel. The upper specific PHF1-band (arrow) was quantified. As already shown in Figure 2, the lower PHF1 band was unspecific (asterisk).
(C) Quantification of WB band intensities of phospho-specific TAU as a percentage of the total amount of TAU detected by K9JA shown in B.
? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
To evaluate whether the compounds affect zebrafish Gsk3β
activity, we first treated wild-type zebrafish embryos between 4
and 24 hpf and examined their developmental phenotype. GSK3β
is a component of the canonical Wnt pathway that is required for
many early patterning events in the developing vertebrate embryo.
Disturbances of this pathway therefore lead to characteristic devel-
opmental defects (35). LiCl treatment caused strongly reduced
eye formation, as previously described (36), while SB-216763 and
SB-415286 also perturbed the normal formation of brain, somites,
and tail in addition to the eye phenotype (Figure 7A). The inhibitor
AR-534 had an even more drastic effect on the fish, causing multiple
ear formation in addition to the described phenotypes. In contrast,
AR-164, which has a comparable in vitro activity to AR-534, did not
cause any detectable phenotype even at 4-fold higher concentra-
tion (Figure 7A). To compare the efficiency of the inhibitors and
the specificity of the evoked phenotypes with a genetic loss of the
endogenous enzyme function, we injected mRNA encoding domi-
nant-negative Xenopus GSK3β-K85R (dN-GSK3β), which is over 92%
identical to zebrafish Gsk3β, into zebrafish eggs. To monitor dose-
dependent effects, we coinjected dN-GSK3β mRNA and GFP mRNA
and used GFP fluorescence to select embryos with lower or higher
protein expression. Strikingly, embryos expressing lower amounts
of dN-GSK3β phenocopied the eye loss, which was also observed in
embryos treated with SB-216763 and SB-415286 and LiCl, whereas
only the embryos expressing higher amounts also phenocopied
the multiple ear formation monitored in AR-534–treated embryos
(Figure 7A). This result indicates that AR-534 has the highest spe-
cific in vivo efficacy of all compounds used in this study. Similar to
γ-secretase inhibitors, which cause very profound defects in Notch
signaling during early development of zebrafish (37), such inhibi-
tors may only be used during late adulthood when FTD and AD
develop, thus excluding their negative effects on embryonic develop-
ment. To confirm that treatment of larvae at later stages does not
cause deleterious developmental problems but reduces pathologic
TAU phosphorylation in vivo, we now treated TAU transgenic fish
for 3 days starting at 20 hpf, after early development was complet-
ed. Treatment starting at this time point did not cause any major
morphological phenotype to the larvae (data not shown). After the
treatment period, embryos were lysed and subjected to quantitative
Western blot (WB) analysis, in which the levels of phosphorylated
TAU were normalized to total TAU to exclude differences in expres-
sion levels or gel loading. SB-216763 and SB-415286 reduced phos-
phorylation of TAU at the disease-specific PHF1 epitope by 30% to
40%, a low activity in comparison with the nonspecific inhibitor
LiCl, which reduces phosphorylation by nearly 90% when used at
high concentration (Figure 7, B and C). AR-534 was more potent
than the existing specific inhibitors SB-216763 and SB-415286 and
reduced TAU phosphorylation by 70% (Figure 7, B and C). In con-
trast to the other inhibitors, AR-164 did not show any inhibitory
activity on TAU phosphorylation in vivo (Figure 7B), which is con-
sistent with both its lack of activity on early development and the
lower penetrability in the in vitro BBB screen, as shown in Figure 6.
We conclude that, consistent with its activity on early zebrafish
development, AR-534 is also the most potent specific inhibitor of
GSK3β–mediated TAU phosphorylation.
We provide a newly developed technology approach for the fast
and efficient generation of transgenic zebrafish to study human
diseases resulting from protein misfolding and/or overexpression
and to optimize promising drug candidates in vivo. The genera-
tion of transgenic zebrafish overexpressing high levels of human
proteins has so far been rather difficult. We have overcome these
limitations with the efficient Tol2 transposon and Gal4/UAS–
based expression of human TAU-P301L in transgenic zebrafish.
In contrast to previous studies (11, 12), we could monitor early
pathology, including disease-specific hyperphosphorylation and
conformational changes of TAU as well as neuronal and behavior-
al abnormalities within the first 2 days of embryonic development
in stable transgenic zebrafish. Furthermore, the larvae developed
substantial neurodegeneration after a few days. These phenotypes
appear much more rapidly in our fish model than in the existing
mouse models, although zebrafish kept under laboratory con-
ditions live as long as 3 to 5 years (38), which is comparable to
laboratory mice (39). Importantly, this early pathology is crucial
to fully exploit all advantages of the zebrafish system, such as opti-
cal clarity and ease of manipulation. The similarity of the evoked
phenotypes to human disease is completed by the appearance of
neurofibrillary tangles after 5 weeks, validating our fish as a well-
suited model for tauopathies in vivo.
We expect the transgenic zebrafish to become a valuable model
system for gaining further insights into the pathology of dementias
by exploiting the unmatched potential for the analysis of cellular
processes in the living organism, which the zebrafish offers in con-
trast with other vertebrate models. Although living cells can also be
monitored in transgenic mice, experiments in the zebrafish can be
performed in a noninvasive way in much higher spatial and tempo-
ral resolution. Whole-cell imaging of neurons comprising somata
and projections can be done in embryos and larvae over several days
without removing the specimen from the microscope (40). Since
FTD-like pathology can be detected already after 1 to 2 days for
early and late disease markers and this pathology progresses rap-
idly, we can monitor disease progression in a compressed timescale
in young zebrafish embryos and larvae, which are perfectly suited
for in vivo imaging of neurons due to their optical transparency.
In fact, confocal imaging even allowed us to follow neuronal cell
death in vivo. To our knowledge, this is the first demonstration of
in vivo cell death imaging in the field of neurodegeneration. More-
over, neurons can be imaged in their natural environment together
with neighboring cell types such as astroglia, oligodendrocytes, and
microglia, which have been shown to participate in the pathology
of several neurodegenerative diseases (1). These important environ-
mental influences are not faithfully recapitulated in tissue explants
or primary cell culture. The early phenotypes allow testing for pos-
sible genetic modifiers of pathology in vivo through transient over-
expression of wild-type or dominant-negative forms of proteins by
DNA or RNA injection or downregulation mediated by injection of
The rapid appearance of pathologic phenotypes allows use of the
transgenic zebrafish larvae not only to image and understand disease
processes in a living animal but also to validate and even screen on
a relatively large scale for compounds that modify the pathology in
vivo. The hyperphosphorylation of TAU, which our fish model devel-
ops within 32 hours after fertilization, is believed to be a key initiator
of detachment of normal TAU from microtubules and subsequent
oligomerization and aggregation. Preventing this phosphorylation by
blocking the involved kinases could potentially slow further progres-
sion of pathology and thus be an efficient treatment option for AD
and FTD (13). We have developed compounds that bind to the active
site of GSK3β and effectively block enzyme activity. The identifica-
1392? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
tion of potent cell-permeable kinase inhibitors with in vivo efficacy is
a key component of the drug discovery process. With the use of ratio-
nal drug design, optimization of the pyrazine chemical series led to
the identification of several GSK3 inhibitors. We demonstrated by in
vitro studies that 2 of these compounds have high selectivity for the
target enzyme, good bioavailability, and high potency to block TAU
phosphorylation in cell culture. However, it is frequently observed
that such inhibitors may not be active in vivo, since they, for example,
do not enter the cells, are transported out, or are rapidly metabolized
when tested in whole organisms (13). This emphasizes the necessity
of rapid, high volume in vivo drug testing in disease models, which is
not feasible in transgenic mice on a large scale. Transgenic zebrafish
offer advantages of both in vitro and in vivo systems and allow rapid
screening for pathology-modifying drugs on a large scale, since
embryos can be tested for rescue of the pathologic phenotypes after
compound treatment in a 96-well–plate scale (9).
We show that GSK3β inhibitors cause specific phenotypes when
applied during early embryonic development and that the evoked
phenotypes already indicate the in vivo potency of the inhibitor.
Moreover, the GSK3β inhibitors can effectively reduce the rapid
disease-specific hyperphosphorylation of TAU in our transgenic
zebrafish. Using our model, we have evaluated the activity of 2 newly
developed GSK3β inhibitors with comparable in vitro activity. One
of the inhibitors, AR-534, was also very active in vivo, with a much
higher effect than the previously described GSK3-specific inhibitors
SB-216763 and SB-415286, which we used for comparison. In con-
trast, the inhibitor AR-164, which is equally potent in cell culture, was
inactive in vivo. These data were consistent with the over 3-fold lower
performance of AR-164 in our BBB test. As this is an in vitro test, which
can only serve as a model of the in vivo situation, the bioavailability
of AR-164 could be even lower in our transgenic zebrafish, explain-
ing the lack of activity. It has been shown recently that zebrafish
form a functional BBB that is similar to that of higher vertebrates.
The BBB in zebrafish consists of endothelial cells, which can be char-
acterized by the expression of the marker proteins claudin-5 and ZO1
and starts to become functional at 3 days after fertilization (10). The
developing BBB might be less permeable to AR-164 versus AR-534
and could therefore be one reason for the absent in vivo activity of
AR-164. Alternatively, degradation or metabolism of AR-164 could
also account for the activity differences we observed. Taken together,
the different activities of AR-534 and AR-164 in the in vitro versus the
in vivo situation clearly illustrate the need for effective in vivo–screen-
ing tools, as this can help to identify promising compounds more
quickly and directly, while at the same time allowing one to eliminate
substances without reasonable in vivo activity early in the screening
process. Taken together, our transgenic zebrafish model could be a
promising tool for streamlining pharmacological screening for many
neurodegenerative diseases. In fact, we have already successfully used
this technology to generate transgenic zebrafish lines overexpress-
ing a number of other amyloidogenic proteins, such as amyloid
β peptide, TDP-43, and α-synuclein (data not shown).
All experiments were performed in accordance with animal protection
standards and were approved by the government of Upper Bavaria (Regier-
ung von Oberbayern, Munich, Germany). The zebrafish wild-type AB line
was maintained, mated, and raised as described (41). Embryos were kept in
E3 medium at 28.5°C and staged as described (42).
Our transgenesis constructs are based on the pT2KXIGdeltaIN plasmid (a gift
from K. Kawakami, National Institute of Genetics, Shizuoka, Japan), which
is derived from the medaka Tol2 transposable element (16). pT2KXIGdeltaIN
was cleaved by BglII and XhoI to remove the insert between the Tol2 sites,
and a multiple cloning site (MCS) was introduced. The MCS was cut out by
PvuII from pBS_I-SceI, which contains the pBluescript MCS flanked on both
sides by I-SceI restriction sites. To assemble the final transgenesis constructs,
fragments were first combined in pBS_I-SceI, as this vector contained more
suitable restriction sites, and later transferred to pT2KXIGdeltaIN.
Driver construct. The SV40 late polyA signal was PCR amplified from
pCS2+ using the primers pCS2+ poly(A)-F SacII (5′-AAAAAACCGCG-
GAGTCGTATTACGTAGATCCAGACATGA-3′) and pCS2+ poly(A)-R SacI
(5′-AAAGAGCTCCACACCTCCCCCTGAAC-3′), cleaved by SacI/SacII, and
ligated into pBS_I-SceI. The neuronal HuC promoter (19) was PCR ampli-
fied using the primers HuC-F XhoI/EcoRI (5′-GCTCGAGGAATTCACTA-
ATTTGAATTTAAATGC-3′) and HuC-R2 EcoRI/ClaI (5′-GGAATTCATC-
GATTCTTGACGTACAAAGATGATATTGATCTAGG-3′), cut by XhoI/ClaI,
and ligated into pBS_I-SceI_pA. Gal4-VP16 (17) was cleaved by BamHI/
SnaBI from pBS_HuC_Gal4-VP16, blunted, and ligated into pBS_I-SceI_
HuC_pA cut by EcoRV/SnaBI. The HuC-Gal4-VP16-pA fragment was trans-
ferred to pT2KXIGdeltaIN by I-SceI digest and ligation.
Responder constructs. We generated transgenic control fish expressing
DsRed but not TAU using a construct that contained E1b-UAS-E1b flanked
by DsRed and TAU, but TAU was isolated from the minimal E1b promoter
by an insert of 200 bp (Ins), which inhibits expression of TAU. DsRed.T4 (43)
was cleaved from pCS2+_DsRed.T4 by NcoI/SnaBI digest to replace GFP in
the vector pBS_E1b_UAS_E1b_GFP_pA. This vector contains UAS flanked
by E1b minimal promoters (17) on both sides. Ins-hTAU-P301L was cleaved
by BglII/BamHI from pNG2htau40/P301L and ligated at the BamHI site
of pBS_I-SceI_pA. E1b-UAS-E1b-DsRed-pA was cleaved by HindIII/SmaI
from pBS_E1b_UAS_E1b_DsRed.T4_pA, blunted and ligated into EcoRV-
cleaved pBS_I-SceI_Ins_hTAU-P301L_pA. The full expression cassette was
transferred to pT2KXIGdeltaIN by I-SceI digest and subsequent ligation.
The final Responder construct was used together with the Driver construct
to generate transgenic fish Tg(HuC:Gal4/UAS-DsRed).
To generate transgenic fish expressing DsRed and TAU, hTAU-P301L
was amplified from pNG2htau40/P301L by PCR using primers hTAU
ATG-F BamHI/NcoI (5′-GGGATCCCATGGCTGAGCCCCGCCAGGA-3′)
and hTAU TAG-R NcoI/EcoRI (5′-GGAATTCCATGGTCACAAACCCT-
GCTTGGCTA-3′), cleaved by NcoI and inserted into pBS_E1b_UAS_
E1b_GFP_pA cleaved by NcoI/SnaBI to remove GFP. DsRed together
with the SV40 late polyA signal were cleaved by ClaI/ApaI from pCS2+_
DsRed.T4 and introduced into pBS_E1b_UAS_E1b_hTAU-P301L_pA.
The insert in pT2KXIGdelta-hTAU-P301L-Ins-E1b-UAS-E1b-DsRed was
removed by ApaI/NheI digest and replaced by the new cassette cleaved
from pBS_pA_DsRed_E1b_UAS_E1b_hTAU40-P301L_pA. The distance
of TAU from E1b is only 35 bp in this construct; therefore, TAU is effi-
ciently expressed together with DsRed. The final Responder construct
was used together with the Driver construct to generate transgenic fish
Gateway constructs. Gateway recombinations were performed according to
the manufacturer’s instructions (Invitrogen). Details about Gateway clon-
ing can be found at www.invitrogen.com/site/us/en/home/Products-and-
To adapt our responder construct to Gateway cloning, hTAU-P301L
was replaced by a Gateway destination site flanked by attR1/2, which
was cleaved from pBS-DEST_attR1-2 (a gift from P. Lemaire, IBDML,
Marseille, France) by EcoRV. The resulting construct was termed Desti-
nation Responder vector.
? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
To adapt our Driver construct to Gateway cloning, the HuC promoter
was also replaced by a Gateway destination site flanked by attR1/2. The
resulting construct was termed Destination Driver vector. To test the func-
tionality of the vectors, the HuC and αTub promoter and hTAU-P301L
were amplified by PCR using the following primers: HuC-F XhoI/EcoRI
(5′-GCTCGAGGAATTCACTAATTTGAATTTAAATGC-3′) and HuC-R2
GATCTAGG-3′) for HuC; M13-FP (5′-TGTAAAACGACGGCCAGT-3′) and
αTub-R (5′-GGCACGCTGTGAAGAAAAAG-3′) for αTub; and hTAU ATG-F
BamHI/NcoI (5′-GGGATCCCATGGCTGAGCCCCGCCAGGA-3′) and
hTAU TAG-R NcoI/EcoRI (5′-GGAATTCCATGGTCACAAACCCTGCTT-
GGCTA-3′) for hTAU-P301L. PCR products were TOPO-cloned into entry
vectors using the pCR8/TOPO/GW Kit from Invitrogen and sequenced.
Promoters and TAU were then LR recombined into Destination Driver and
Responder vectors, using LR Clonase II from Invitrogen according to the
Plasmid preparation, mRNA synthesis, and embryo injection
Plasmids were prepared by extraction with a MACHEREY-NAGEL Maxi Kit
and further cleaned up by using the Qbiogene GENECLEAN Kit accord-
ing to the manufacturer’s instructions. The vectors pCS-TP (a gift from K.
Kawakami, National Institute of Genetics, Shizuoka, Japan), pCS2+_GFP,
pCS2+_dN-GSK3β, and pCS2+_GSK3β (both gifts from R. Rupp, Ludwig-
Maximilians-University) were linearized by NotI and cleaned up by phenol/
chloroform precipitation. mRNAs were synthesized from the linearized vec-
tors with the Ambion mMESSAGE mMACHINE kit (Applied Biosystems)
according to the manufacturer’s instructions and stored in small aliquots
at –80°C. Shortly before injection, circular Driver and Responder plasmids
were mixed with Tol2 transposase mRNA, all at a final concentration of
25 ng/μl, in dH20 containing diethylenepyrocarbonate-treated (DEPC-
treated) 0.2 M KCl and 20% phenol red (44). GFP mRNA was used at a con-
centration of 100 ng/μl and GSK3β mRNAs were used at a concentration of
50 ng/μl, all diluted in DEPC-treated dH2O containing 20% phenol red.
One-cell–stage embryos were collected, oriented on agar plates, and
injected with approximately 1 nl of the injection solution. Plasmid DNA
and transposase mRNA were injected through the yolk into the cytoplasm
of the first cell before the first cell division. GFP and GSK3β mRNAs were
mixed before use and injected into the yolk. At around 6 to 8 hpf, 2 random
fertilized embryos from each dish were tested for transposition by PCR
as described (44). At 30 hpf, embryos were selected for DsRed expression
under a Leica fluorescent stereomicroscope and raised at 28.5°C to adult-
hood. Adult founder fish were identified by outcrossing them to AB wild-
type fish and screening the F1 generation for DsRed-expressing embryos.
These embryos were raised to establish transgenic lines.
Identification of transgenes by PCR
The offspring of 225 embryos of an F2 outcross were analyzed by PCR.
Three independent PCRs were performed to show Gal4 on the Driver con-
struct, TAU on the responder construct, and actin as a loading control.
The following primers were used: Gal4: HuC 3′F 5′-TGGCGAAGACT-
GTCCTTTTT-3′, Gal4-R 5′-GGTCTTCTCGAGGAAAAATCAG-3′; hTAU:
TAU-OF 5′-AGGAGTTCGAAGTGATGGAAGAT-3′, TAU-IR 5′-GTGGC-
GATCTTCGTTTTACCAT-3′; and actin: actin-F 5′-TGTTTTCCCCTC-
CATTGTTGG-3′, actin-R 5′-TTCTCCTTGATGTCACGGAC-3′. For the
PCRs, the 4- to 24-hour-old eggs were fixed in methanol; then the DNA
was extracted by digestion with proteinase K for 4 hours at 55°C followed
by inactivation of the enzyme for 10 minutes at 95°C. The extracted DNA
was used as a template for PCR. To analyze the results, the products of the
3 PCRs for each possible genotype from 2 different embryos were mixed
together, loaded on a gel, counted, and imaged.
IF, immunohistochemical stainings, Gallyas silver stainings,
Whole-mount IF stainings were performed according to standard meth-
ods (45). For immunohistochemical and Gallyas silver stainings on par-
affin sections, the 5-week-old fish were deeply anesthetized and killed
in water with ice. Then the whole fish were fixed overnight in 4% PFA
and transferred to PBS and the sections and stainings were performed as
described for mouse tissue (46).
WBs were performed according to standard methods. Embryos or larvae
were frozen in liquid nitrogen and 10 to 50 μl of chilled Laemmli buffer
were added per fish. Fish were homogenized by sonication and incubated
at 95°C for 10 minutes. Embryo lysates were subjected to centrifugation
at 13,000 g for 1 minute, and 10 to 20 μl of the supernatant was loaded
on an 8% or 10% polyacrylamide gel. The following antibodies were used
for IF, immunohistochemistry (IHC), and WB: (a) TAU. T46 (20) (IF:
1:200, WB: 1:1000; Invitrogen), K9JA (IF: 1: 500, WB: 1:5000; no. A0024;
Dako), DA9 (IHC: 1:250; a gift from P. Davies, Albert Einstein College
of Medicine, New York, New York, USA), AT8 (25) (IF: 1:200, IHC: 1:50,
WB 1:1000; Pierce, Thermo Scientific), 422 (24) (IF: 1:200, WB 1:1000; a
gift from C. Czech, Hoffmann La Roche, Basel, Switzerland), AT180 (21)
(IF: 1:200, WB: 1:1000; Pierce, Thermo Scientific), AT270 (21) (IF: 1:200,
WB: 1:1000; Pierce, Thermo Scientific), 12E8 (22) (IF: 1:200, WB: 1:1000;
a gift from P. Seubert, Elan Pharmaceuticals, San Francisco, California,
USA), MC1 (26) (IF: 1:50; a gift from P. Davies, Albert Einstein College of
Medicine, New York, New York, USA), PHF1 (23) (IF: 1:100, WB 1:1000; a
gift from P. Davies, Albert Einstein College of Medicine, New York, New
York, USA); (b) Other. DsRed (IF: 1:200; Clontech), actin (WB: 1:1000;
Sigma-Aldrich), znp1 (30) (IF: 1:100; Developmental Studies Hybridoma
Bank), Alexa Fluor 488 goat anti-mouse IgG (1:500; Invitrogen), Alexa
Fluor 546 goat anti-rabbit IgG (1:500; Invitrogen), Alexa Fluor 488 goat
anti-rat IgG (1:500; Invitrogen).
Acridine orange stainings and time-lapse imaging
Living zebrafish larvae were anesthetized with Tricaine (Sigma-Aldrich),
incubated in a solution of 3 μg/ml acridine orange (Sigma-Aldrich) in E3
medium with Tricaine for 30 minutes (27), and washed twice in E3 with
Tricaine. For imaging, fish were embedded in 1.6% low melting agarose in
E3 and overlayed with E3/Tricaine. The green nuclei of the neurons were
counted in the whole spinal cord of 44 fish for each genotype using a Zeiss
compound microscope with a Zeiss Plan-Apochromat 10×/0.45 lens. For
time-lapse imaging, a z-stack of a chosen region of the spinal cord of the
transgenic fish with a thickness of 10 μm was imaged every 3 minutes for
10 to 16 hours on a Zeiss LSM 510 META inverted confocal microscope
(Zeiss, provided by the Hans and Ilse Breuer Foundation, Frankfurt am
Main, Germany) in a heated chamber. The images of the z-stack were com-
bined in a maximum projection, and the time series of image projections
was exported to a time-lapse video with 8 frames per second using Zeiss
LSM 510 Confocal software, version 4.2 SP1. The videos and still images
taken from the videos were processed with ImageJ 1.41 (http://rsb.info.nih.
gov/ij/) and iMovie 7 on Mac OSX 10.5.
48-hour-old living zebrafish larvae of several clutches were pooled together
and sorted for visible DsRed expression. The selected fish were dechori-
onated manually at least 3 hours before the experiment. To evaluate the
escape response, fish were touched with the tip of a fine needle for at least
2 times at the dorsal tip of the tail. An escape response in which the fish did
not move a distance of at least 3 times its own body length was considered
as reduced. A minimum of 3 groups of 50 fish were quantified for each
genotype. At least one experiment was recorded and processed as a video.
1394?The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
Living zebrafish embryos, larvae, and 5-week-old juveniles were examined
and imaged on a fluorescent stereomicroscope (Leica). For imaging fixed
stainings, heads of zebrafish embryos were removed, and the tails were flat
mounted in 1% low melting agarose in 50% glycerol/dH2O on coverslips
and visualized with an LSM 510 META inverted confocal microscope.
Motoneuron length measurements of confocal pictures were done using
the overlay function of the Zeiss LSM software with maximum projections
of z-stacks with constant stack height containing the whole neuronal pro-
jection. Images comparing 2 groups of fish were always done at the same
time using identical settings. Paraffin sections were imaged on a Zeiss com-
pound microscope. Images were assembled in Adobe Photoshop 8.0.
Analysis of compound characteristics
The GSK3β Kis were determined using a scintillation proximity assay with
a biotinylated peptide sequence from eIF2B and [γ-33P]ATP as substrates.
The CDK2 Kis were determined using a scintillation proximity assay with
a biotinylated peptide sequence from GST retinoblastoma and [γ-33P]ATP
as substrates. Scintillation proximity assays and kinetic analyses were per-
formed as described earlier (14). IC50 inhibition curves were analyzed by
nonlinear regression using GraphPad Prism 5 (GraphPad Software).
BBB permeability in vitro was analyzed using an in vitro cell culture
model for permeability across the BBB as described earlier (47).
Analysis of GSK3β inhibition in TAU transfected cells
3T3 fibroblasts were engineered to stably express 4-repeat TAU protein.
The cells were treated with vehicle or with increasing concentrations of
inhibitors and harvested at 4 hours after treatment. Cultures were then
washed twice with 5 mM MgCl2 PBS. Extracts for WB analysis were pre-
pared by homogenizing cells in ice-cold extraction buffer consisting of
20 mM HEPES, pH 7.4, 100 mM NaCl, 10 mM NaF, 1% Triton X-100, 1 mM
sodium orthovanadate, 10 mM EDTA, and protease inhibitors (2 mM
phenylmethylsulfonyl fluoride, 10 μg/ml aprotinin, 10 μg/ml leupeptin,
and 10 μg/ml pepstatin). The samples were homogenized at 4°C, and pro-
tein content was determined by the Bradford method. Total protein (25 μg)
was electrophoresed on 10% SDS-PAGE gel and transferred to a nitrocel-
lulose membrane (Schleicher & Schuell Bioscience Inc.). The experiments
were performed using the following primary antibodies: pS396, 1:1000,
and Tau5, 1:1000. The filters were incubated with the antibody at 4°C
overnight in 5% nonfat dried milk. A secondary horseradish peroxidase–
linked sheep anti-mouse (1:1000; Amersham Biosciences) or horseradish
peroxidase–linked donkey anti-rabbit (1:5000; Amersham Biosciences)
followed by ECL detection reagents (Amersham Biosciences) was used for
immunodetection. Quantitation of immunoreactivity was performed by
Stock solutions were prepared by dissolving compounds in DMSO (SB-216763
and SB-415286, 20 mM), 50% DMSO/dH2O (AR-534, 5 mM), or dH2O
(AR-164, 10 mM, and LiCl, 3 M). Stock solutions were diluted with
embryo medium E3 containing 1% DMSO to final dilutions of 50 μM for
SB-216763, 100 μM for SB-415286, 100 mM for LiCl, 25 μM for AR-534,
and 100 μM for AR-164. The unpermeable chorions of fragile 4-hour-old
embryos were only slightly opened to allow access to the compounds, while
rather robust 24 hpf embryos were completely dechorionated. Embryos
were incubated in 1 ml compound dilutions or control medium (E3 con-
taining 1% DMSO) in 12-well plates; the solutions were changed daily.
Embryos were either treated from 4 to 24 hpf or from 24 to 100 hpf.
Mean values and SD were calculated with Microsoft Excel, version 12.1.3.
Statistical analysis was performed using 2-tailed Student’s t test in Microsoft
Excel. Data are presented as mean ± SD. P < 0.05 was considered significant.
This work was supported by grants from the Deutsche Forschun-
gsgemeinschaft (SFB 596 to B. Schmid and C. Haass; the Center for
Integrated Protein Science to C. Haass and B. Schmid; and the Leib-
niz Award to C. Haass); Elitenetzwerk Bayern and Universität Bay-
ern (to D. Paquet); the Federal Ministry for Education and Research
(BMBF, BioFuture-Award 0311889 to R.W. Köster); the Studienstif-
tung des deutschen Volkes (to M. Distel); and the European Com-
munity’s Seventh Framework Programme (FP7/2007-2013) under
grant agreement no. 200611 (MEMOSAD to C. Haass, B. Schmid,
E.-M. Mandelkow, and D. Paquet). C. Haass is supported by a
research professorship at Ludwig-Maximilians-University. We thank
A. Hruscha, M. Teucke, S. Schätzle, H. Kaiser, D. Drexler, O. Petrova,
Y. Xu, E. Jerning, J. Neelissen, and Y. Nilsson for technical help; K.
Kawakami, R. Rupp, and P. Lemaire for providing constructs; and P.
Seubert, P. Davies, and C. Czech for providing antibodies. We thank
the Hans and Ilse Breuer Foundation for the confocal microscope
and K. Winklhofer for critically reading this manuscript.
Received for publication September 23, 2008, and accepted in
revised form February 25, 2009.
Address correspondence to: Christian Haass, Deutsches Zentrum
für Neurodegenerative Erkrankungen (DZNE) and Adolf-Buten-
andt-Institut, Biochemistry, Ludwig-Maximilians-University,
Schillerstr. 44, 80336 Munich, Germany. Phone: 49-89-2180-75472;
Fax: 49-89-2180-75415; E-mail: email@example.com.
1. Haass, C., and Selkoe, D.J. 2007. Soluble protein
oligomers in neurodegeneration: lessons from the
Alzheimer’s amyloid beta-peptide. Nat. Rev. Mol.
Cell Biol. 8:101–112.
2. Kosik, K.S., Joachim, C.L., and Selkoe, D.J. 1986.
Microtubule-associated protein tau (tau) is a major
antigenic component of paired helical filaments
in Alzheimer disease. Proc. Natl. Acad. Sci. U. S. A.
3. Grundke-Iqbal, I., et al. 1986. Abnormal phos-
phorylation of the microtubule-associated protein
tau (tau) in Alzheimer cytoskeletal pathology. Proc.
Natl. Acad. Sci. U. S. A. 83:4913–4917.
4. Braak, H., and Braak, E. 1991. Neuropathological
stageing of Alzheimer-related changes. Acta Neuro-
5. Santacruz, K., et al. 2005. Tau suppression in a neu-
rodegenerative mouse model improves memory
function. Science. 309:476–481.
6. Hutton, M., et al. 1998. Association of missense
and 5′-splice-site mutations in tau with the inher-
ited dementia FTDP-17. Nature. 393:702–705.
7. Mandelkow, E.M., and Mandelkow, E. 1998. Tau in
Alzheimer’s disease. Trends Cell Biol. 8:425–427.
8. Gotz, J., and Ittner, L.M. 2008. Animal models of
Alzheimer’s disease and frontotemporal dementia.
Nat. Rev. Neurosci. 9:532–544.
9. Zon, L.I., and Peterson, R.T. 2005. In vivo drug
discovery in the zebrafish. Nat. Rev. Drug Discov.
10. Jeong, J.Y., et al. 2008. Functional and developmen-
tal analysis of the blood-brain barrier in zebrafish.
Brain Res. Bull. 75:619–628.
11. Tomasiewicz, H.G., Flaherty, D.B., Soria, J.P., and
Wood, J.G. 2002. Transgenic zebrafish model of
neurodegeneration. J. Neurosci. Res. 70:734–745.
12. Bai, Q., Garver, J.A., Hukriede, N.A., and Burton,
E.A. 2007. Generation of a transgenic zebrafish
model of tauopathy using a novel promoter ele-
ment derived from the zebrafish eno2 gene. Nucleic
Acids Res. 35:6501–6516.
13. Mazanetz, M.P., and Fischer, P.M. 2007. Untan-
gling tau hyperphosphorylation in drug design for
neurodegenerative diseases. Nat. Rev. Drug Discov.
14. Bhat, R., et al. 2003. Structural insights and
biological effects of glycogen synthase kinase
3-specific inhibitor AR-A014418. J. Biol. Chem.
15. Winklhofer, K.F., Tatzelt, J., and Haass, C. 2008.
The two faces of protein misfolding: gain- and loss-
of-function in neurodegenerative diseases. EMBO J.
16. Urasaki, A., Morvan, G., and Kawakami, K. 2006.
? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 5 May 2009
Functional dissection of the Tol2 transposable ele-
ment identified the minimal cis-sequence and a
highly repetitive sequence in the subterminal region
essential for transposition. Genetics. 174:639–649.
17. Koster, R.W., and Fraser, S.E. 2001. Tracing trans-
gene expression in living zebrafish embryos. Dev.
18. Walhout, A.J., et al. 2000. GATEWAY recombina-
tional cloning: application to the cloning of large
numbers of open reading frames or ORFeomes.
Methods Enzymol. 328:575–592.
19. Park, H.C., et al. 2000. Analysis of upstream elements
in the HuC promoter leads to the establishment
of transgenic zebrafish with fluorescent neurons.
Dev. Biol. 227:279–293.
20. Kosik, K.S., et al. 1988. Epitopes that span the tau
molecule are shared with paired helical filaments.
21. Goedert, M., et al. 1994. Epitope mapping of mono-
clonal antibodies to the paired helical filaments of
Alzheimer’s disease: identification of phosphoryla-
tion sites in tau protein. Biochem. J. 301:871–877.
22. Seubert, P., et al. 1995. Detection of phosphorylated
Ser262 in fetal tau, adult tau, and paired helical
filament tau. J. Biol. Chem. 270:18917–18922.
23. Greenberg, S.G., Davies, P., Schein, J.D., and Binder,
L.I. 1992. Hydrofluoric acid-treated tau PHF pro-
teins display the same biochemical properties as
normal tau. J. Biol. Chem. 267:564–569.
24. Hasegawa, M., et al. 1996. Characterization of
mAb AP422, a novel phosphorylation-dependent
monoclonal antibody against tau protein. FEBS
25. Biernat, J., et al. 1992. The switch of tau pro-
tein to an Alzheimer-like state includes the
phosphorylation of two serine-proline motifs
upstream of the microtubule binding region.
EMBO J. 11:1593–1597.
26. Jicha, G.A., Berenfeld, B., and Davies, P. 1999.
Sequence requirements for formation of confor-
mational variants of tau similar to those found in
Alzheimer’s disease. J. Neurosci. Res. 55:713–723.
27. Furutani-Seiki, M., et al. 1996. Neural degeneration
mutants in the zebrafish, Danio rerio. Development.
28. Thies, E., and Mandelkow, E.M. 2007. Missorting
of tau in neurons causes degeneration of synapses
that can be rescued by the kinase MARK2/Par-1.
J. Neurosci. 27:2896–2907.
29. Terwel, D., et al. 2005. Changed conformation of
mutant Tau-P301L underlies the moribund tauop-
athy, absent in progressive, nonlethal axonopa-
thy of Tau-4R/2N transgenic mice. J. Biol. Chem.
30. Trevarrow, B., Marks, D.L., and Kimmel, C.B. 1990.
Organization of hindbrain segments in the zebrafish
embryo. Neuron. 4:669–679.
31. Fox, M.A., and Sanes, J.R. 2007. Synaptotagmin I and
II are present in distinct subsets of central synapses.
J. Comp. Neurol. 503:280–296.
32. Myers, P.Z., Eisen, J.S., and Westerfield, M. 1986.
Development and axonal outgrowth of identi-
fied motoneurons in the zebrafish. J. Neurosci.
33. Smith, D.G., et al. 2001. 3-Anilino-4-arylma-
leimides: potent and selective inhibitors of glyco-
gen synthase kinase-3 (GSK-3). Bioorg. Med. Chem.
34. Hong, M., Chen, D.C., Klein, P.S., and Lee, V.M.
1997. Lithium reduces tau phosphorylation by
inhibition of glycogen synthase kinase-3. J. Biol.
35. Doble, B.W., and Woodgett, J.R. 2003. GSK-3: tricks
of the trade for a multi-tasking kinase. J. Cell Sci.
36. Stachel, S.E., Grunwald, D.J., and Myers, P.Z. 1993.
Lithium perturbation and goosecoid expression
identify a dorsal specification pathway in the pre-
gastrula zebrafish. Development. 117:1261–1274.
37. Geling, A., Steiner, H., Willem, M., Bally-Cuif, L.,
and Haass, C. 2002. A gamma-secretase inhibitor
blocks Notch signaling in vivo and causes a severe
neurogenic phenotype in zebrafish. EMBO Rep.
38. Gerhard, G.S., et al. 2002. Life spans and senescent
phenotypes in two strains of Zebrafish (Danio
rerio). Exp. Gerontol. 37:1055–1068.
39. Roberts, R.C. 1961. The lifetime growth and
reproduction of selected strains of mice. Heredity.
40. Koster, R.W., and Fraser, S.E. 2001. Direct imaging
of in vivo neuronal migration in the developing
cerebellum. Curr. Biol. 11:1858–1863.
41. Mullins, M.C., Hammerschmidt, M., Haffter, P., and
Nusslein-Volhard, C. 1994. Large-scale mutagen-
esis in the zebrafish: in search of genes controlling
development in a vertebrate. Curr. Biol. 4:189–202.
42. Kimmel, C.B., Ballard, W.W., Kimmel, S.R., Ull-
mann, B., and Schilling, T.F. 1995. Stages of
embryonic development of the zebrafish. Dev. Dyn.
43. Bevis, B.J., and Glick, B.S. 2002. Rapidly maturing
variants of the Discosoma red fluorescent protein
(DsRed). Nat. Biotechnol. 20:83–87.
44. Kawakami, K. 2005. Transposon tools and methods
in zebrafish. Dev. Dyn. 234:244–254.
45. Nüsslein-Volhard, C., and Dahm, R. 2002. Zebrafish:
a practical approach. Oxford University Press. Oxford,
United Kingdom. 303 pp.
46. Mocanu, M.M., et al. 2008. The potential for beta-
structure in the repeat domain of tau protein deter-
mines aggregation, synaptic decay, neuronal loss,
and coassembly with endogenous Tau in inducible
mouse models of tauopathy. J. Neurosci. 28:737–748.
47. Cecchelli, R., et al. 1999. In vitro model for evaluat-
ing drug transport across the blood-brain barrier.
Adv. Drug Deliv. Rev. 36:165–178.