Autism spectrum disorders (ASD) are an early onset, heterogeneous group of heritable
neuropsychiatric disorders with symptoms that include deficits in social interaction
skills, impaired communication ability, and ritualistic-like repetitive behaviors
1,2
. One of the hypotheses for a common molecular mechanism underlying ASD is altered
translational control resulting in exaggerated protein synthesis
3
. Genetic variants in chromosome 4q, which contains the EIF4E locus, have been described
in autistic patients
4,5
. Importantly, a rare single nucleotide polymorphism has been identified in autism
that is associated with increased promoter activity in the EIF4E gene
6
. Herein we show that genetically increasing the levels of eIF4E in mice
7
results in exaggerated cap-dependent translation and aberrant behaviors reminiscent
of autism, including repetitive/perseverative behaviors and deficits in social interactions.
Moreover, these autistic-like behaviors are accompanied by synaptic pathophysiology
in the medial prefrontal cortex, striatum, and hippocampus. The autistic-like behaviors
displayed by the eIF4E transgenic mice are corrected by intracerebroventricular (ICV)
infusions of the cap-dependent translation inhibitor 4EGI-1. Our findings demonstrate
a causal relationship between exaggerated cap-dependent translation, synaptic dysfunction,
and aberrant behaviors associated with autism.
eIF4E transgenic mice
7
exhibited increased levels of eIF4E across brain regions (Fig. 1a) without compensatory
changes in levels of other translational control proteins (Fig. 1b). We investigated
whether eIF4E was bound preferentially to either 4E-BP or eIF4G, which repress and
promote, respectively, the initiation of cap-dependent translation
8,9
. We found significantly higher levels of eIF4E/eIF4G interactions in the brains of
eIF4E transgenic mice (Fig. 1c and Supplementary Fig. 1a) with no alterations in the
interaction between eIF4E and 4E-BP (Fig. 1c, left and Supplementary Fig. 1a). To
confirm that the increased eIF4E/eIF4G interactions resulted in increased protein
synthesis, we infused puromycin into the lateral ventricle of cannulated mice and
labeled newly synthesized proteins using SUnSET
10,11
and observed increased de novo cap-dependent translation (Fig. 1d and Supplementary
Fig. 1b-g). Overall, our results indicate that overexpression of eIF4E results in
exaggerated cap-dependent translation in the brains of eIF4E transgenic mice.
We then determined whether eIF4E transgenic mice display repetitive and perseverative
behaviors, which are behavioral domains required for ASD diagnosis
2
. eIF4E transgenic mice exhibited repetitive digging behavior in the marble burying
test
12
and increased self-grooming
13
compared to wild-type littermates (Fig. 2a, 2b). eIF4E transgenic mice also displayed
cognitive inflexibility in both a water-based Y-maze task and a modified version of
the Morris water maze
14,15
. Learning ability in the acquisition and memory phases of these tasks was intact;
however, in the reversal phases, eIF4E transgenic mice were impaired in locating the
new platform positions (Fig. 2c, 2d and Supplementary Fig. 2e-h). We tested an additional
form of behavioral inflexibility by examining the eIF4E transgenic mice for extinction
of cued fear conditioning and found that they did not exhibit a significant reduction
in freezing responses following extinction training (Fig. 2e). These experiments suggest
that excessive cap-dependent translation in the brain affects the ability to suppress
previously codified response patterns and the ability to form new behavioral strategies
in response to changed environmental circumstances.
Abnormalities in social interaction skills are another behavioral defect displayed
by individuals with ASD
2
. In tests to examine social behavior
16–18
, the eIF4E transgenic mice did not show a preference for a nonspecific stranger versus
a novel, inanimate object (Fig. 2f, 2g). Moreover, eIF4E transgenic mice exhibited
diminished reciprocal interactions with a freely moving stranger mouse (Fig. 2h),
further evidence of deficits in social behavior. The deficits in social behavior of
the eIF4E transgenic mice are unlikely to be caused by a generalized increase in anxiety
(Supplementary Fig. 2c, 2d and 2j). Moreover, the eIF4E transgenic mice exhibited
mild hyperactivity (Supplementary Fig. 2a and 2b), but no impairments in motor coordination,
motor learning and sensorimotor gating (Supplementary Fig 2i and 2k, 2l). Taken together,
our behavioral analysis of the eIF4E transgenic mice indicates that increased cap-dependent
translation in the brain results in a distinct pattern of behavioral abnormalities
consistent with ASD.
Previous studies suggest that ASD symptoms such as cognitive inflexibility and deficits
in social behavior are generated by abnormalities in prefrontal and/or striatal circuits
19
. Consistent with this idea, the medial prefrontal cortex (mPFC) is implicated in
the modulation of social behaviors and social skills
20
, whereas motor, social, and communication impairments in boys with ASD are associated
with anatomical abnormalities in the striatum
21
. Therefore, we next examined whether the eIF4E transgenic mice exhibited specific
synaptic pathophysiologies in the mPFC and striatum.
In the eIF4E transgenic mice, examination of spontaneous synaptic “mini” events in
layers 2/3 of acute mPFC slices revealed an increase in the frequency but not amplitude
of excitatory events (mEPSC; Fig 3a), and an increase in the amplitude, but not frequency,
of inhibitory events (mIPSC; Fig 3b). No changes were observed in layer 5 (Supplementary
Fig. 3a, 3b). Thus, our data suggests an enhancement of excitatory input and post-synaptic
sensitivity for inhibitory events onto layer 2/3 pyramidal neurons, consistent with
the hypothesis that autism may arise from an imbalance between excitatory and inhibitory
synaptic transmission
22
.
To determine whether the increased frequency of spontaneous mEPSCs might result from
an enhanced number of synaptic contacts, we imaged dendritic spines using two-photon
laser-scanning microscopy (Fig. 3c, 3d and Supplementary Fig. 3c, 3d). We found a
significant increase (~12%) in spine density and observed a significantly smaller
spine volume in the eIF4E transgenic mice compared to wild-type littermates (WT=0.123±0.004
um3 and 4E Tg=0.110±0.004 um3, p=0.01 vs. WT, Student’s t-test).
Next, we examined whether increased expression of eIF4E also resulted in synaptic
pathophysiology in the striatum. We used high-frequency stimulation (HFS) to induce
long-term depression (LTD) in acute striatal slices
23
and found that eIF4E transgenic mice exhibited enhanced LTD compared to wild-type
littermates (Fig. 3e, Supplementary Fig. 3e, 3f). We hypothesize that the enhanced
LTD in eIF4E transgenic mice results in altered efficiency of striatal information
storage and processing, culminating in the inability to form new motor patterns and/or
to disengage from previously learned motor behaviors.
To determine whether the synaptic alterations described in the eIF4E transgenic mice
were selective for the fronto-striatal circuit, we examined synaptic plasticity in
the hippocampus
24
. We found that eIF4E transgenic mice exhibited enhanced mGluR-LTD compared to wild-type
littermates (Fig. 3f, Supplementary Fig. 3g, 3h), consistent with previous studies
showing that changes in brain protein synthesis are accompanied by altered (enhanced
or reduced) hippocampal mGluR-LTD
25,26
. Thus, consistent with the ubiquitous increase in brain expression of eIF4E, the
eIF4E transgenic mice display altered synaptic function and plasticity in several
brain regions (mPFC, striatum and hippocampus) implicated in behavioral abnormalities
associated with ASD.
Finally, we asked whether exaggerated cap-dependent translation was responsible for
the synaptic alterations and ASD-like behaviors displayed by the eIF4E transgenic
mice. We took advantage of 4EGI-1, an inhibitor of eIF4E/eIF4G interactions
8,11
, to block the synaptic and behavioral consequences of increased eIF4E expression.
Bath application of 4EGI-1 normalized the enhanced striatal LTD observed in the eIF4E
transgenic mice (Fig. 3g and 3h), suggesting that exaggerated striatal LTD (Fig. 3h)
is a direct consequence of increased binding of eIF4E to eIF4G (Supplementary Fig.
3i-k).
Next, we employed a subthreshold dose of 4EGI-1
11
to normalize the behavioral abnormalities in eIF4E transgenic mice without impairing
their wild-type littermates. eIF4E transgenic mice treated with 4EGI-1 exhibited a
decrease in repetitive behavior during the marble burying task, which started on day
four and persisted throughout day five (Fig. 4a). Moreover, we found that 4EGI-1 maintained
the behavioral effects observed in the marble-burying task (Supplementary Fig. 4a
and 4b). We also found that blockade of eIF4E/eIF4G interactions with 4EGI-1 significantly
improved the performance of eIF4E transgenic mice in the reversal phase of the Y-maze
test (Fig. 4b). These findings indicate that chronic treatment of eIF4E transgenic
mice with 4EGI-1 reverses their repetitive and perseverative behaviors. We also found
that infusions of 4EGI-1 rescued the social behavior deficits displayed by the eIF4E
transgenic mice in the three-chamber arena test, as they exhibited an increased preference
for a non-specific stranger compared to a novel object (Fig. 4c).
At the completion of the behavioral studies with 4EGI-1, we performed co-immunoprecipitation
experiments, confirming that 4EGI-1 reduced the increased eIF4E/eIF4G interactions
exhibited by the eIF4E transgenic mice (Fig. 4d and Supplementary Fig. 4c-e). Furthermore,
puromycin-labeling of newly synthesized proteins was reduced to wild-type levels,
indicating that 4EGI-1 was effective in attenuating the increased cap-dependent translation
in the eIF4E transgenic mice (Fig. 4e and Supplementary Fig. 4f and 4g). Together,
these results indicate that repeated treatment of eIF4E transgenic mice with 4EGI-1
reverses the increased binding of eIF4E to eIF4G, exaggerated cap-dependent translation,
and reversal of ASD-like behaviors.
Herein we have demonstrated that increased eIF4E expression and, consequently, dysregulated
translational control at the initiation phase of protein synthesis in mice results
in the appearance of synaptic dysfunction and aberrant behaviors consistent with ASD.
Based on our observations, we hypothesize that exaggerated cap-dependent protein synthesis
in the eIF4E transgenic mice and FXS model mice
27,28
results in enhanced translation of a specific subset of mRNAs. Thus, the identity
of both these mRNAs and the cis-acting elements in the 5’ UTR responsible for eIF4E-dependent
protein synthesis and their possible overlap with fragile X mental retardation protein
target mRNAs will be important investigations in future studies.
In closing, our studies with eIF4E transgenic mice indicate that ASD-like behaviors
can be induced by exaggerated cap-dependent translation in the brain. Moreover, we
demonstrated that aberrant repetitive, perseverative, and social behaviors displayed
by eIF4E transgenic mice are reversed by reducing eIF4E/eIF4G interactions, thereby
restoring translational homeostasis. Thus, our findings establish a causal link between
exaggerated cap-dependent translation and behaviors associated with autism. Finally,
our findings indicate that behavioral defects caused by exaggerated cap-dependent
translation, which also occurs in FXS
29,30
, a disorder with a high incidence of ASD, are not irrevocable and can be corrected
well into adulthood.
Methods Summary
All procedures involving animals were approved by the New York University Animal Care
and Use Committee and followed the NIH Guidelines for the use of animals in research.
For a detailed description of all the techniques used in this study, please refer
to Methods section. All the experiments were performed with the examiners blind to
genotype.
Methods
Housing
Generation of βT-Eif4e transgenic mice (eIF4E transgenic mice) has been described
previously
7
.
For all the experiments, we made use of littermates derived from crossing heterozygotes.
Mice were back-crossed to the N10 generation in C57BL/6J mice. Overall, eIF4E transgenic
mice were viable, fertile, and showed no gross anatomical abnormalities in the age
range used for this study. eIF4E transgenic mice and their wild-type littermates were
housed in groups of 3–4 animals per cage and kept on a regular 12 h light/dark cycle
(lights on at 7:00 am). Food and water were available ad libitum.
Surgery and drug infusion
Mice were anesthetized [ketamine (100 mg/kg) and xylazine (10 mg/kg)] and mounted
onto a stereotaxic apparatus. Cannulae (26-gauge) were implanted unilaterally at the
following coordinates (−0.22 mm anterioposterior, +1 mm mediolateral, and −2.4 mm
dorsoventral)
31
. Mice were allowed one week to recover after the surgery.
The infusions of the eIF4E/eIF4G inhibitor 4EGI-1 were performed as described previously
11
. 4EGI-1 dissolved in 100% DMSO was diluted in vehicle [0.5% (2-hydroxypropyl)-β-cyclodextrin
and 1% DMSO in ACSF]. 4EGI-1 (20 µM) or vehicle were infused over 1 min (0.5 µL/min;
Harvard Apparatus). On the last day of treatment, mice received infusion of 4EGI-1
alone or puromycin (25 µg in 0.5 µL) prior to 4EGI-1 infusions. All behavior and tissue
dissection occurred 1 hr after 4EGI-1 infusions.
Behavior
The following behavioral tests were performed on male eIF4E transgenic mice and their
wild-type littermates (2–6 months of age) as described previously: novelty-induced
locomotor activity
32
, open field
33
, elevated plus maze
33
, rotarod
34
, prepulse inhibition
33
, marble burying
14
, social behavior
16
, direct social interaction
35,36
, y-maze and the Morris water maze
7,35
.
For all experiments, mice were acclimated to the testing room 30 min prior to behavioral
training and all behavior apparatuses were cleaned between each trial with 30% ethanol.
The experimenter was blind to genotype and drug treatment while performing and scoring
all behavioral tasks. All behavioral tests were performed starting with the least
aversive task first (locomotor activity) and ending with the most aversive (either
water-based mazes or extinction of fear memory).
Western blots
Mice were killed by decapitation 1 hr after the infusion with either 4EGI-1 alone
or 4EGI-1 and puromycin. The striatum and prefrontal cortex were rapidly dissected,
placed on an ice-cold surface, and sonicated in 1% SDS and boiled for 10 min. Aliquots
(2 µl) of the homogenate were used for protein determination with a BCA (bicinchoninic
acid) assay kit (Pierce, Thermo Scientific, Rockford, USA). Equal amounts of protein
(20 µg) for each sample were loaded onto 10% polyacrylamide gels. Proteins were separated
by SDS–polyacrylamide gel electrophoresis and transferred overnight to polyvinylidene
difluoride membranes (Immobilon-Psq, Millipore Corporation, Billerica, USA). The membranes
were immunoblotted with antibodies against eIF4E (1:1000), eIF4G (1:1000), eIF4B (1:1000),
eIF4A (1:1000), 4E-BP (1:1000) (Cell Signaling Technology, MA, USA). Antibodies against
β-actin and tubulin (1:5000, Cell Signaling Technology, MA, USA) were used to estimate
the total amount of proteins. Detection was based on HRP-conjugated secondary antibody
(Promega, WI, USA) and chemiluminescence reagent (ECL or ECL plus; GEHealthcare, Buckinghamshire,
UK), and visualized using a Kodak 4000MM imager to obtain pixel density values for
the band of interest (Carestream, NY, USA). All images were obtained using maximum
sensitivity settings with no binning (0–65 K signal range). No images analyzed presented
saturating signals for the bands of interest (>65 K grayscale value). The amount of
each protein was normalized for the amount of the corresponding β-actin or tubulin
detected in the sample.
Immunoprecipitation
Tissue was homogenized in ice-cold lysis immunoprecipitation buffer containing (in
mM): 40 hepes (pH 7.5), 150 NaCl, 10 pyrophosphate, 10 glycerophosphate, 1 EDTA and
0.1% CHAPS, Protease Inhibitor II, Phosphatase Inhibitor Mixture I, II (Sigma-Aldrich,
MO, USA). Cleared homogenate (500 µg) was incubated with either anti-eIF4G (2.5 µg)
or eIF4E (2.5 µg) (Bethyl Laboratories, TX, USA) and gently shaken overnight at 4°C.
The antibody/lysate mix was incubated with 75 µL IgG bound to agarose-beads (Thermo
Scientific, IL, USA). The bead/sample slurry was incubated while rocking at 4°C overnight.
Supernatant was removed and saved, and immunoprecipitates were washed three times
in lysis buffer, and once in wash buffer (50 mM hepes pH 7.5, 40 mM NaCl, 2 mM EDTA).
SDS/PAGE buffer was added to the washed immunoprecipitates, which then were resolved
on 4 to 12% gradient gels. Efficiency of the immunoprecipitation was determined by
examining the supernatant and wash fractions obtained from the procedure on images
obtained from Kodak 4000MM imager (see Western Blots). Band density values for coimmunoprecipitated
eFI4E, eIF4G and 4E-BP were normalized to immunoprecipitated eIF4G or eIF4E.
SuNSET
A protein synthesis assay was performed as previously described using the SuNSET method
11
. Puromycin-treated samples were identified on blots using the mouse monoclonal antibody
12D10 (1:5000 from a 5-mg/mL stock). Because only a small fraction of the brain proteins
were labeled, signal from blots was identified using ECL-Advance (GEHealthcare, Buckinghamshire,
UK).
Electrophysiology
Hippocampal (400 µm), prefrontal and striatal slices (300 µm) for electrophysiology
were prepared as described previously
24
.
Solution to maintain slices
Cutting solution (CS; in mM): 110 Sucrose, 60 NaCl, 3 KCl, 1.25 NaH2PO4, 28 NaHCO3,
0.5 CaCl2, 7 MgCl2, 5 Glucose, 0.6 Ascorbate. Artificial cerebrospinal fluid (ACSF;
in mM): 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 25 D-glucose, 2 CaCl2, and 1 MgCl2.
Slices were incubated at room temperature and then were placed in the recording chamber
for additional recovery time of 60 min at 33°C.
Extracellular recordings
Extracellular field EPSPs were recorded as described previously
23,24
. In all the experiments, baseline synaptic transmission was monitored for at least
20 min before long-term depression (LTD) induction. Three trains of high-frequency
stimulation (3 sec duration, 100 Hz frequency at 20 sec intervals) were used to induced
LTD in striatal slices
23
while 10 min of incubation with DHPG (50 µM) was used to induce mGluR-dependent LTD
in hippocampal slices
24
. The slope of fEPSP was expressed as percent of the baseline average before LTD induction.
Intracellular recordings
Medial prefrontal pyramidal cells were illuminated and visualized using a x60 water-immersion
objective mounted on a fixed-stage microscope (BX61-WI, Olympus, Center Valley, PA),
and the image was displayed on a video monitor using a charge-coupled device camera
(Hamamatsu, Bridgewater, NJ). Recordings were amplified by multiclamp 700B and digitized
by Digidata 1440 (Molecular Devices, Sunnyvale, CA). The recording electrode was pulled
from a borosilicate glass pipette (3–5 MΩ) using an electrode puller (P-97, Sutter
Instruments, Navato, CA), was filled with an internal solution according to the specific
experimental requirement, and was patched onto the soma. Rs was compensated ~70% and
was readjusted before each experiment. A measured liquid junction potential was corrected
by adjusting the pipette offset. All voltage-clamp recordings were low-pass filtered
at 10 kHz and sampled at 50 kHz.
Internal solution for mEPSC (in mM): 120 cesium methane-sulfonate, 10 HEPES, 10 EGTA,
4 MgCl2, 0.4 NaGTP, 4 MgATP, 10 phosphocreatine, 5 QX-314 (pH adjusted to 7.3 with
CsOH, 290 mOsm). Bicuculline 50 µM and tetrodotoxin (TTX) 1 µM (Tocris, Ellisville,
MI) were added to the external ACSF bath solution.
Internal solution for mIPSC (in mM): 140 CsCl, 10 EGTA, 10 HEPES, 2 MgCl2, 2.0 Mg-ATP,
4 Na2-ATP, 0.4 Na2-GTP, 5 QX-314 (pH adjusted to 7.3 with CsOH, 290 mOsm), thus yielding
a chloride reversal potential of around 2 mV for the chloride currents. 40 µM 6,7-dinitroquinoxaline-2,3-dione
(DNQX), 50 µM 2-amino-5-phosphonopentanoate (D-AP-5) and 1 µM TTX were added to the
ACSF bath solution.
In these conditions, mEPSC and mIPSC were recorded in voltage clamp at −70 mV and
measured for 120 sec or 60 sec, respectively.
Dendritic spine morphology
Dendritic spine density experiments were performed as previously described
37,38
. Briefly, two-photon imaging was accomplished with a custom microscope and high-resolution
stacks (x = 0.13 µm, y = 0.13 µm, z = 0.2 µm per voxel) of dendritic segments throughout
the entire cell were taken for morphological analysis in NeuronStudio. Spine head
volume was calculated using a rayburst algorithm. Images were deconvolved prior to
volume measurements using custom routines written in MATLAB (Mathworks).
Supplementary Material
1