M. L. Castrén, Institute of Biomedicine, Physiology, University of Helsinki, Box 63, Haartmaninkatu 8, FIN-00014 Helsinki, Finland. E-mail: Maija.Castren@helsinki.fi
Fragile X syndrome (FXS) is a common cause of inherited intellectual disability and a well-characterized form of autism spectrum disorder. As brain-derived neurotrophic factor (BDNF) is implicated in the pathophysiology of FXS we examined the effects of reduced BDNF expression on the behavioral phenotype of an animal model of FXS, Fmr1 knockout (KO) mice, crossed with mice carrying a deletion of one copy of the Bdnf gene (Bdnf+/−). Fmr1 KO mice showed age-dependent alterations in hippocampal BDNF expression that declined after the age of 4 months compared to wild-type controls. Mild deficits in water maze learning in Bdnf+/− and Fmr1 KO mice were exaggerated and contextual fear learning significantly impaired in double transgenics. Reduced BDNF expression did not alter basal nociceptive responses or central hypersensitivity in Fmr1 KO mice. Paradoxically, the locomotor hyperactivity and deficits in sensorimotor learning and startle responses characteristic of Fmr1 KO mice were ameliorated by reducing BNDF, suggesting changes in simultaneously and in parallel working hippocampus-dependent and striatum-dependent systems. Furthermore, the obesity normally seen in Bdnf+/− mice was eliminated by the absence of fragile X mental retardation protein 1 (FMRP). Reduced BDNF decreased the survival of newborn cells in the ventral part of the hippocampus both in the presence and absence of FMRP. Since a short neurite phenotype characteristic of newborn cells lacking FMRP was not found in cells derived from double mutant mice, changes in neuronal maturation likely contributed to the behavioral phenotype. Our results show that the absence of FMRP modifies the diverse effects of BDNF on the FXS phenotype.
Fragile X syndrome (FXS) is a common cause of inherited intellectual disability (see review Garber et al. 2008). The neurobehavioral symptoms of FXS include hyperactivity, defects in sensory integration, communication difficulties, poor motor coordination, social anxiety and restricted repetitive and stereotyped patterns of behavior. About 30% of FXS males fulfill the standardized criteria of autism (Brown et al. 1986; Hagerman et al. 2010; Hernandez et al. 2009) and 75% meet the diagnostic criteria for attention deficit and/or hyperactivity disorder (ADHD) (Baumgardner et al. 1995). The syndrome is caused by a loss of functional fragile X mental retardation protein 1 (FMRP), a RNA-binding protein that interacts with many pre- and postsynaptic transcripts, regulates their translation and is essential for the normal maturation and function of synapses (Darnell et al. 2011). Altered spine morphology, abnormalities in the neural stem cell differentiation, and many features in the behavioral phenotype of FXS individuals correlate with findings in Fmr1 knockout (KO) mice that serve as a model for human FXS (Bakker et al. 1994). Epilepsy found in around 20% of FXS males may associate with brain-derived neurotrophic factor (BDNF) gene polymorphisms linked with reduced BDNF secretion (Louhivuori et al. 2009).
BDNF is a member of the neurotrophic factor family that binds to the tropomyosin-related kinase B (TrkB) receptor and is widely expressed in the mammalian brain (Ernfors et al. 1992; Hofer et al. 1990; Klein et al. 1990). Brain-derived neurotrophic factor promotes the survival and differentiation of a variety of neuronal populations in the developing and adult brain. Brain-derived neurotrophic factor-TrkB signaling is essential for synapse maturation and several forms of synaptic plasticity (Poo 2001) and acts as a neurotransmitter modulator in adult brain (Lu et al. 2009). Mice lacking completely BDNF die before the second postnatal week (Ernfors et al. 1994). Bdnf+/− mice have roughly 50% reduction of BDNF levels, are viable, and show changes in eating behavior, obesity, hyperactivity and aggressiveness (Kernie et al. 2000; Linnarsson et al. 1997; Lyons et al. 1999). The loss of forebrain BDNF alters pharmacological responses (Monteggia et al. 2004), suggesting that reduced BDNF expression increases vulnerability to distinct perturbations. There is evidence that BDNF/TrkB signaling plays a role in the pathophysiology of autism spectrum disorder and ADHD (Correia et al. 2011; Miyazaki et al. 2004; Nelson et al. 2001; Perry et al. 2001; Shim et al. 2008). Enhanced TrkB receptor signaling in neural progenitors lacking FMRP as well as regional and subcellular alterations of BDNF expression in Fmr1 KO mouse brain (Louhivuori et al. 2011) suggest that BDNF/TkB signaling contributes to the susceptibility of ADHD and autism in the FXS phenotype.
Here, we examined the effects of reduced BDNF expression on the molecular and behavioral phenotype of adult FXS mice. The constitutive reduction of BDNF expression in novel transgenic Fmr1 KO mice who carry a deletion in one copy of the Bdnf gene allowed us to identify changes in the behavioral phenotype caused by alterations associated with reduced BDNF expression both in the developing and mature brain of Fmr1 KO mouse.
Male Fmr1 KO mice (B6.129P2-Fmr1tm1/Cgr/J) and heterozygote Bdnf+/− mice (B6.129S4-Bdnftm1Jae/J) on the C57BL/6J genetic background were purchased from Jackson Laboratory (Bar Harbor, ME). Bdnf+/− mice were crossed and maintained on the C57BL/6JOlaHsd substrain in the Animal Centre of University of Helsinki. The Fmr1 gene is located on the X chromosome in both mice and men and a deletion of the gene in one allele results in FXS in males. Fmr1+/− female mice were crossed with a male Bdnf+/− mouse to generate double transgenic (dMT) mice, single-mutant mice and wild-type (WT) littermates. Mice in F6 generation were used to breed the mice for these studies to ensure the C57BL/6JOlaHsd genetic background. For the behavioral studies, we crossed male dMT mice from the F6 generation with female WT and Fmr1+/− littermates. Only male mice were used in the studies. Wild type littermates were used as controls in all experiments. At the beginning of behavioral testing littermates were 3–4 months old. Group-housed mice were maintained under 12-h light–dark cycle (lights on from 0600 to 1800 h) with food and water available ad libitum. The behavioral experiments were carried out during light phase (between 0900 and 1600 h). In the behavioral experiments group sizes were WT, n = 9 mice; Bdnf+/−, n = 10 mice, Fmr1 KO, n = 10, dMT, n = 9, except in pain studies n = 5 in each mouse group. All animal experiments were done in accordance with the guidelines of the National Institute of Health Guide for the Care and Use of Laboratory Animals and carried out under protocols approved by the Experimental Animal Ethical Committee of Southern Finland. Efforts were made to minimize animal suffering and to use only the number of animals necessary to produce reliable scientific data. In tests of pain responses, we used experimental procedures that allowed minimizing the intensity and duration of pain.
All parameters were scored by two investigators blind to the genotype of the mice. Body weight of the mice was recorded. Behavioral tests were conducted in the order presented below and the interval between the subsequent tests was 2–3 days:
Open field test
The mice were released in the corner of novel open field arena (30 × 30 cm; MED Associates, St. Albans, VT, USA) surrounded by frames with infrared light barriers for detection of animal's position. Horizontal and vertical activity was recorded for 30 min (light intensity ∼150 lx). Peripheral zone was defined as a 6-cm wide corridor along the wall.
Standard hot plate (TSE, Bad Homburg, Germany) was used for the assessment of nociceptive sensitivity. The mouse was confined on the plate heated to 52°C by Plexiglass cylinder (diameter 19 cm and height 31 cm). Hind paw response (licking or shaking) was noted as the pain threshold in seconds. Maximum time of the test was 60 seconds.
Acoustic startle reflex
Acoustic startle reflex experiments were performed in startle reflex apparatus (MED Associates). The isolation chambers were equipped with an acoustic stimulator and a platform with a transducer amplifier. Data acquisition was performed using Med Associates 5 software (Startle Reflex System). Mice were placed in the startle chamber with a background white noise of 65 dB and left undisturbed for 5 min. Mice were tested in 36 trials of white noise acoustic startle stimulus (20 milliseconds). Nine stimulus levels (68, 72, 74, 78, 82, 86, 90, 100 and 110 dB) were performed in pseudo randomized order. The inter-trial interval ranged between 8 and 15 seconds. The startle response was recorded for 65 milliseconds from the onset of the startle stimulus. The maximum startle amplitude recorded during the sampling window was used as the dependent variable. The startle response was averaged over four trials at each startle stimulus level.
The mice were trained on an accelerating rotarod (Ugo Basile, Comerio, Italy) equipped with automatic fall detector over a period of 2 days with three trials each day. The speed of rotation increased at a constant rate (from 4 to 40 r.p.m.) in 5 min and the cutoff time was 6 min.
Morris water maze
A custom-made circular water pool (120 cm diameter) with black walls was equipped with an escape platform (10 cm diameter) located 0.5 cm below the water surface in the center of one of four equally sized quadrants and computer interfaced video tracking system (Noldus Information Technology, Wageningen, The Netherlands). For each trial, a mouse was released to swim in random positions and the time to reach the escape platform was measured. The mice were given two training blocks with three trials per day. The interval between trials was about 3 min and between training blocks about 5 h. The platform was kept in a constant location for 3 days and thereafter moved to the opposite quadrant for 2 days. The transfer tests were conducted approximately 18 h after the 6th and 10th training sessions. The swim-paths were recorded and analyzed by the water maze software (EthoVision XT 7; Noldus Information Technology). The spatial memory was estimated by the time spent in the zone around the platform and in the corresponding zones of the three remaining quadrants. In addition, the swimming distance was recorded.
The experiments were carried out employing a computer-controlled fear conditioning system (TSE). Training was performed in a clear acrylic cage (23 × 23 × 35 cm) within a constantly illuminated (125 lx) fear conditioning box. A loudspeaker provided a constant, white background noise (68 dB) for 120 seconds followed by 10 kHz conditioned stimulus tone (75 dB, pulsed 5 Hz) for 30 seconds. The tone was terminated by unconditioned stimulus, foot-shock (0.6 mA, 2 seconds, constant current) delivered through a stainless steel floor grid (floor bars 4 mm diameter, distance rod center to rod center 8.9 mm). Two conditioned–unconditioned stimulus pairings were separated by a pause for 30 seconds. Contextual memory was tested 24 h after the training. The animals were returned to the conditioning box and total time of freezing (defined as an absence of any movements for more than 3 seconds) was measured by infrared light barriers scanned continuously with a frequency of 10 Hz.
Formalin-induced spontaneous pain
Formalin (2.5%, 20 µl) was administered by intraplantar injections to a proximal site in the hind paw. Spontaneous pain behavior was assessed by calculating the duration of time the animal spent expressing spontaneous pain-like behavior (guarding, lifting, shaking or licking the injected paw). The percent of time spent expressing pain-like behavior was assessed separately during the early or first phase of formalin-induced pain (0–5 min) and the late or second phase of formalin-induced pain (15–30 min).
To assess tactile allodynia-like behavior induced by formalin (Wei et al. 2011), the frequency of withdrawal responses to the application of monofilaments (von Frey hairs) to the hind paw was examined 30 min following formalin administration. Since the focus of monofilament testing was on secondary (central) hypersensitivity, monofilaments were applied to the skin area adjacent to the formalin-treated skin, in which primary afferent nerve fibers were intact. A series of monofilaments that produced forces varying from 0.02–1 g (North Coast Medical Inc., Morgan Hill, CA, USA) was applied in ascending order five times to the plantar skin adjacent to the formalin-treated site at a frequency of 0.5 Hz. A visible lifting of the stimulated hind limb was considered a withdrawal response. An increase in the response rate represents mechanical hypersensitivity.
Brain-derived neurotrophic factor ELISA
For the BDNF expression studies hippocampi were collected from mice sacrificed by cervical dislocation followed by anesthesia with CO2. Samples were frozen on dry ice, and stored at −70°C until use. The BDNF expression was determined using BDNF enzyme-linked immunosorbent assay (ELISA) (Quantikine human BDNF kit; R&D Systems, Abingdon, UK) as described previously (Louhivuori et al. 2011).
BrdU birthdating study
To study survival of newborn cells in the hippocampus, bromodeoxyuridine (BrdU) was administered intraperitoneally at the dose of 75 mg/kg four times every 2 h (300 mg/kg total) during the day after the last behavioral test (Fig. 6c). Four weeks after injections, animals were deeply anesthetized and sacrificed by cervical dislocation followed by anesthesia with CO2. Hippocampi were dissected and divided to ventral and dorsal parts. The BrdU labeling was detected from the tissue as described previously (Wu & Castrén 2009). Deoxyribonucleic acid was extracted from the hippocampi, denatured and dot-blotted onto membrane. The BrdU incorporation was detected by immunostaining with mouse BrdU-specific monoclonal primary antibody (1-299-964; Roche, Espoo, Finland).
Neural progenitor cultures were propagated from the wall of lateral ventricles of male Fmr1 KO and dMT pups and their WT littermates at embryonic day 14 as previously described (Castrén et al. 2005). For live Cell-IQ imaging studies 20–30 neurospheres (diameter around 225 µm) were collected from the free-floating neurosphere cultures and placed on cover slips coated with 10 µg/ml poly-dl-ornithine (Sigma, Helsinki, Finland) and differentiated for 1 day without the mitogens.
Live Cell-IQ imaging
Early neurosphere differentiation and neurite formation were investigated using a Cell-IQ® imaging system (Chip-Man Technologies Ltd., Tampere, Finland) which combines phase contrast microscopy to automation. An image was taken every 15 min from each grid area.
Data obtained from behavioral tests were analyzed with Statview 5.0 for Windows software (SAS, Cary, NC), unless specified otherwise, using a two-way analysis of variance (anova) for repeated measure analysis or one-way anovas followed by Fishers's protected least significance post hoc test or Tukey test. In Cell-IQ image series, Image J 1.45 cell tracking was used to analyze neurite lengths. Microsoft Excel 2007 and Origin 7.5 were used to perform one-way anova followed by Tukey test. Values are presented as mean ± standard error of mean (SEM) or mean ± standard deviation as indicated in figure legends.
The BDNF expression and weight gain in Fmr1 null mice with one copy of the Bdnf gene
To study the effects of reduced BDNF expression on the behavioral phenotype of FXS, we generated dMT male Fmr1 null mice with one copy of the Bdnf gene by crossing Bdnf+/− mice with Fmr1 KO mice. At the age of 4–5 months, the expression of BDNF protein varied with the genotype (Fig. 1a; one-way anova, F3,37 = 11.64, P < 0.0001). Post hoc tests indicated that the BDNF protein expression was significantly reduced in all transgenic mice when compared with WT controls (Fig. 1a). The expression in the hippocampus of Bdnf+/− mice was reduced by 52% when compared with WT controls (Fig. 1a; P < 0.01). The hippocampal BDNF expression of Fmr1 KO mice was found to show bimodal regulation: BDNF levels were significantly increased at the age of 2 and 4 months (P < 0.01) but reduced by 27% at the age of 4–5 months (in the end of the behavioral studies) and further reduced by 53% at the age of 6–7 months (P < 0.05), when compared with the expression in WT mice (Fig. 1b). In dMT mice, hippocampal BDNF levels were 57% lower than those of WT controls (P < 0.001) and 28% lower than those in Fmr1 KO mice (P < 0.05) but were not significantly (P > 0.05) different from those of Bdnf+/− mice (Fig. 1a).
During the observation period that lasted up to 1 year, the mice showed a significant increase in body weight (Fig. 1c; two-way anova, F2,75 = 245, P < 0.0001). The main effect of the genotype (F3,75 = 33.4, P < 0.0001) and the interaction between the genotype and duration of the observation period (F6,75 = 10.4, P < 0.0001) were significant. Post hoc tests indicated that the gain of body weight during the observation period was significantly more pronounced in the Bdnf+/− male mice than in WT animals indicating that Bdnf+/− male mice displayed abnormalities of eating behavior (Fig. 1c). Surprisingly, the body weight of dMT mice did not differ from WT controls but differed significantly from the body weight of Bdnf+/− mice (Fig. 1c), indicating that the absence of FMRP counteracted the effects of reduced BDNF expression on the eating behavior.
The effects of reduced BDNF on learning of Fmr1 KO mice
BDNF expression is essential for hippocampus-dependent spatial learning that is commonly tested by the hidden platform in the Morris water maze (MWM). We observed that the time needed to find the hidden platform was significantly different among the genotypes (Fig. 2a; two-way anova, F3,204 = 7.54, P < 0.0001). Moreover, the latency to find the hidden platform was decreased with repetition of the trials (Fig. 2a; two-way anova, F5,204 = 31.25, P < 0.0001), independent of the genotype (F15,204 = 0.81, (P > 0.05). Throughout the six trials of the hidden platform test, WT mice tended to need less time to find the platform than the other genotypes tested in this study. After six sessions with a stable location of the hidden platform, the platform was moved to the opposite quadrant. When the hidden platform was removed, the performance of the animals, as revealed by the distance needed to reach the platform, varied significantly among the genotypes (Fig. 2b; one-way anova, F3,34 = 4.0, P < 0.05). Post hoc tests indicated that Fmr1 KO and Bdnf+/− mice did not differ significantly from the performance of WT littermates, but dMT mice showed an increased cumulative search error after removal of the platform (Fig. 2b). After removal of the platform, the time spent in the quadrant in which the platform was expected to be varied significantly among the genotypes (Fig. 2c; one-way anova, F3,34 = 3.14, P < 0.05). According to post hoc test the dMT group spent less time in the quadrant where the platform was expected to be than WT and Bdnf+/− mice (Fig. 2c). The abnormalities were not caused by differences in the swimming abilities since velocities in the different mouse groups were similar (data not shown). Furthermore, the reversal trials in the MWM did not reveal any differences between the different mouse groups (data not shown).
Following testing in the MWM, we examined the effects of reduced BDNF on conditioned fear learning. In the conditioning paradigm, mice learn to associate a particular compartment (contextual memory) with an unconditioned aversive stimulus, an electric shock. When tested 24 h after conditioning, contextual fear learning of dMT mice was dramatically impaired as measured by freezing when compared with WT, Fmr1 KO and Bdnf+/−mice (Fig. 3; one-way anova, F3,34 = 2.96, P < 0.05, post hoc test, P < 0.05). Freezing of Bdnf+/− and Fmr1 KO mice did not differ from WT controls under our experimental setting (P > 0.05). Together, these data suggest that the deficiency of both FMRP and BDNF in the dMT mice causes learning impairment that is more pronounced than the impairments observed in single-mutant mice.
Reduced BDNF expression ameliorated locomotor hyperactivity, deficits in startle responses and sensorimotor learning in Fmr1 KO mice
In the open field test, the locomotor activity of male Fmr1 KO mice was increased and seen as a longer distance traveled by Fmr1 KO mice when compared to the distance moved by mice in other groups (Fig. 4a; WT, 47.68 ± 3.49 m; Fmr1 KO, 60.33 ± 4.68 m; Bdnf+/−, 41.29 ± 6.00 m; dMT, 46.95 ± 1.97 m; one-way anovaF3,34 = 3.67, P < 0.05). Under the conditions used, we did not observe any increase in the locomotor activity of Bdnf+/− mice (Fig. 4a; P > 0.05). Surprisingly, the total distance traveled by dMT mice did not differ from the distance traveled by WT littermates (Fig. 4a; P > 0.05), suggesting that the reduced BDNF expression ameliorated the hyperactivity of the Fmr1 KO mice.
Motor behavior was assessed in the rotarod test. Repetition of the rotarod test improved the performance, as indicated by the prolongation of the drop latency from the rod (Fig. 4b; two-way anova, F5,204 = 4.0, P < 0.005) which was independent of the genotype (interaction between the latency and genotype: F15,204 = 0.93). The genotype had a significant main effect on the rotarod performance (Fig. 4b; two-way anova, F3,204 = 10.63, P < 0.0001). Post hoc tests indicated that the Fmr1 KO mice stayed less time on the rotating rod with a gradually increasing speed when compared to the performance of WT controls, whereas the rotarod performance of Bdnf+/−or dMT mice did not differ from the performance of WT littermates (Fig. 4b). These data suggest that deficits in motor coordination and skill learning caused by the absence of FMRP can be prevented by reducing BDNF expression.
Next, we studied the effects of reduced BDNF expression on sensory stimuli in dMT mouse by investigating startle responses to auditory stimuli and suppression of a startle that is preceded by an innocuous tone in prepulse inhibition (PPI). The startle response was significantly different among the genotypes (Fig. 4c; two-way anova, F3,306 = 11.72, P < 0.0001). The difference in the startle responses among genotypes tended to be larger at high sound intensities (Fig. 4c; two-way anova, F24,306 = 1.93, P < 0.01). In contrast to startle response, we found no significant differences in PPI among the genotypes (data not shown). According to post hoc tests, startle responses in Fmr1 KO mice were significantly reduced when compared with WT mice at sound intensities of 86, 90 and 100 dB (P < 0.05). The startle responses of Bdnf+/− mice did not differ from WT controls at any of the sound intensities. Moreover, no difference was found in the startle responses between dMT and WT mice, independent of the sound intensity, indicating that impaired startle reflexes in FXS mice were normalized by reducing BDNF expression.
The effects of reduced BDNF expression on pain behaviors of Fmr1 KO mice
Variation in the genotype failed to induce differences in baseline withdrawal responses to mechanical stimulation (Fig. 5a; mixed model two-way anova, F3,72 = 1.34, P > 0.05) or heat (Fig. 5b; one-way anova, F3,34 = 0.65, P > 0.05). The mechanically evoked responses to stimulation of the uninjured skin area adjacent to a formalin-treated site were significantly increased in all experimental groups indicating development of secondary (central) hypersensitivity (the mean formalin-induced increase in the cumulative response rate to monofilament stimulation was 196 ± 21%, n = 20). The formalin-induced secondary hypersensitivity to mechanical stimulation varied significantly with the genotype (Fig. 5c; mixed model two-way anova, F3,72 = 4.37, P < 0.01); in Fmr1 KO and dMT mice the formalin-induced secondary hypersensitivity tended to be weaker than in WT controls, although post hoc tests failed to indicate significant differences among the groups (Fig. 5c). Brain-derived neurotrophic factor is known to modulate central sensitization at pathological states (Kerr et al. 1999) but a decrease in the BDNF expression in mice with a genetic deletion of a Bdnf gene was without any effects on central hypersensitivity (Fig. 5c). The first phase of formalin-induced sustained pain behavior was not different among the genotypes (Fig. 5d,e; one-way anova, F3,16 = 0.66, P > 0.05), whereas the second phase of the formalin-induced sustained pain behavior varied significantly among the genotypes (Fig. 5d,e; one-way anova, F3,16 = 5.96, P < 0.01). According to post hoc tests, the second phase of the formalin-induced sustained pain behavior was augmented in the dMT group indicating that the second phase of formalin-induced pain behavior was enhanced in Fmr1 KO mice by reduced BDNF expression (Fig. 5d,e).
The effects of reduced BDNF expression on newborn neuronal cells lacking FMRP
We studied the effects of reduced BDNF expression on adult neurogenesis in Fmr1 mice in vivo by investigating the survival of BrdU-labeled newborn hippocampal cells. The BrdU staining in the ventral part of the hippocampus was significantly different among the genotypes (Fig. 6a; one-way anova, F3,16 = 5.82, P < 0.05). According to post hoc tests, BrdU staining was decreased in the ventral part of the hippocampus of Fmr1 KO mice 4 weeks after BrdU injections when compared to WT controls (Fig. 6a) indicating reduced survival of hippocampal newborn cells in the absence of FMRP. An identical decrease in the BrdU staining was seen in the ventral part of the hippocampus of Bdnf+/− mice when compared to WT controls (Fig. 6a). Combined reduction of BDNF and FMRP in dMT mice resulted in a decrease in BrdU staining as seen in single transgenics (Fig. 6a), suggesting that the effects of BDNF and FMRP were mediated in a common pathway. The BrdU expression in the dorsal part of the hippocampus did not differ between different genotypes (Fig. 6b; one-way anova followed by Tukey test, F3,12 = 0.87, P > 0.05).
We previously showed that newborn neurons differentiated from mouse and human FMRP-deficient neural progenitors show a short neurite phenotype (Castrén et al. 2005). We propagated neural cortical progenitors from mice with different genotypes and live imaging of differentiating neurospheres allowed us to analyze the neurite formation in individual differentiating cells during the first 24 h of neurosphere differentiation. In Cell-IQ system, images of cells migrating out of neurosphere cell clusters were captured every 15 min. Image analysis revealed that a larger number of newborn cells differentiated from cortical progenitors propagated from brains of Fmr1 KO mice had shorter neurites than those propagated from WT controls (Fig. 6c,d; one-way anova followed by Tukey test, F12,13 = 12.25, P < 0.001) and less cells with medium (F12,13 = 5.45, P < 0.05) or long length neurites (F12,13 = 8.99, P < 0.01) whereas the neurite phenotype of newborn cells differentiated from progenitors of dMT mice did not differ from the phenotype of WT cells according to the post hoc tests (Fig. 6c,d). These data show that the neurite phenotype of newborn cells caused by the absence of FMRP was prevented by reducing BDNF, suggesting that reduced BDNF and the loss of FMRP counteracted each other's effects.
We have generated transgenic Fmr1 KO mice with a deletion in one copy of the Bdnf gene and examined the effects of reduced BDNF expression on the Fmr1 KO mouse behavioral phenotype. We observed that the adult dMT mice deficient in both BDNF and FMRP show impaired contextual fear learning and spatial learning deficits that were more pronounced than those with a deletion of only one of these genes. In addition, spontaneous pain behavior was increased in dMT mice but reduced BDNF expression did not alter basal nociceptive responses or the constitutive central hyposensitivity of Fmr1 KO mice. To our surprise, the absence of FMRP prevented changes in eating behavior and increases in body weight normally seen with Bdnf+/− mice (Kernie et al. 2000; Rios et al. 2001). Furthermore, the locomotor hyperactivity, and deficits in sensorimotor learning and auditory responses normally seen in Fmr1 KO mice (Chen & Toth 2001; Frankland et al. 2004; Nielsen et al. 2002; Qin et al. 2005; Spencer et al. 2011) were ameliorated by reduced BDNF expression. In addition, decreases in BDNF expression reduced the survival of newborn cells in the ventral part but not in the dorsal part of the adult hippocampus both in the presence and absence of FMRP. We also found that decreasing BDNF expression ameliorated abnormalities in neural progenitor differentiation in FXS and prevented the short neurite phenotype of newborn cells during early differentiation. This suggests that changes in maturation of neuronal cells contribute to alterations in the behavioral phenotype.
FMRP is an RNA-binding protein that binds a subset of dendritic mRNAs and plays a role in the localization, translation and stability of mRNAs in neurons (see review Bassell & Warren 2008). We have previously shown that an increase in dendritic BDNF mRNA in neurons is associated with a decrease in cortical BDNF protein levels in Fmr1 KO mouse brains (Louhivuori et al. 2011) indicating reduced translation of BDNF in cortical neurons in the absence of FMRP. Differential regulation of BDNF mRNA expression in cortical and hippocampal neurons may explain anatomical variations in BDNF protein levels. In contrast to the reduced cortical BDNF expression, the hippocampal BDNF expression was shown to be significantly increased during early adulthood in Fmr1 KO mice (Louhivuori et al. 2011). This study revealed that the BDNF expression declines in the hippocampus of Fmr1 KO mice after the age of 4 months when compared with WT controls, suggesting alterations of responses to neuronal degeneration in FXS. As both activity-dependent local translation of BDNF and retrograde BDNF-TrkB signaling are considered key players in synaptic plasticity (Poo 2001) reduction of dendritic BDNF translation in Fmr1 KO mice crossed with Bdnf+/− mice may explain the deterioration in cognitive capacity observed in dMT mice.
The role of FMRP in hippocampus-dependent learning and neurogenesis in adults was recently confirmed by ablation of Fmrp in adult mouse neural progenitors that resulted in impaired learning and reduced hippocampal neurogenesis (Guo et al. 2011). Testing of learning and memory of Fmr1 KO mice in the MWM has yielded conflicting results which may be explained by genetic and environmental factors and different behavioral testing strategies (Bakker et al. 1994; Dí Hooge et al. 1997; Dobkin et al. 2000; Gantois et al. 2001; Kooy et al. 1996; Paradee et al. 1999; Peier et al. 2000). We observed a very mild spatial learning defect in Fmr1 KO mice. Also results of studies on complex behavior of heterozygous Bdnf mice vary (Linnarsson et al. 1997; Montkowski et al. 1997) but clear correlations between hippocampal learning and regulation of BDNF activity are shown in several studies. Selective deletion of forebrain BDNF from early development or in the adulthood as well as a deletion of hippocampal-specific BDNF impairs spatial learning in the MWM (Gorski et al. 2003; Heldt et al. 2007; Monteggia et al. 2004). Furthermore, intrahippocampal administration of BDNF improves performance of WT animals in the MWM (Cirulli et al. 2004) whereas blocking the action of BDNF by infusing neutralizing antibodies during pre-training causes learning impairment (Mu et al. 1999). Reductions in both FMRP and BDNF expression in dMT mice caused learning deficits that became evident during the last trials of the hidden platform test and the probe trial. Adult Bdnf+/− mice with highly reduced BDNF levels show similar behavioral phenotype (Linnarsson et al. 1997), with data suggesting that the effects of reduced BDNF on hippocampus-dependent learning are enhanced in the absence of FMRP. Consistent with the findings in the MWM, the hippocampus-dependent associative learning of dMT mice was highly impaired indicating that reduced BDNF expression together with the absence of FMRP caused in dMT mice contextual fear learning deficits that are not normally seen in Fmr1 KO mice (Dobkin et al. 2000; Peier et al. 2000; Van Dam et al. 2000).
We found that the aberrances in locomotor activity, startle responses and sensorimotor learning normally seen in Fmr1 KO mice were ameliorated by reducing BDNF expression in dMT mice, although these behaviors were unaffected in Bdnf+/− mice. Locomotor hyperactivity in both novel and familiar environments and reduced startle responses at higher intensities are consistent findings in Fmr1 KO mice (Bakker et al. 1994; Chen & Toth 2001; Frankland et al. 2004; Hayashi et al. 2007; Nielsen et al. 2002; Peier et al. 2000; Qin et al. 2005; Spencer et al. 2011) but rotarod performance of Fmr1 KO mice has not been extensively investigated (Peier et al. 2000). The effects of reduced BDNF on mouse behavior may be influenced by the parental genetic background, gender and age of mice (Crusio et al. 2009). An age-dependent reduction of striatal BDNF expression was recently shown to correlate with deficits in the rotarod performance in Bdnf+/− mice on the C57BL/6 genetic background at 3 months of age (Boger et al. 2011) and Bdnf+/− mice show increased variance of activity and may be hyperactive when stressed (Kernie et al. 2000; Rios et al. 2001). Interestingly, BDNF overexpression was recently shown to decrease acoustic startle but it did not affect locomotor activity or coordination (Papaleo et al. 2011). The absence of locomotor hyperactivity and deficits in the rotarod performance and startle responses in adult male Bdnf+/− mice shown in this study are in line with previous studies (Boger et al. 2011; Dluzen et al. 2001; Gorski et al. 2003; MacQueen et al. 2001). The nigrostriatal system has an influence on the rotarod performance and both BDNF and FMRP are implicated in the modulation of dopamine responses that are related to motor coordination in the striatum (Boger et al. 2011; Wang et al. 2010). However, studies investigating the relationship of catecholamine systems to present findings have yet to be conducted. Hippocampus-dependent and striatum-dependent systems form at least partially independently working systems that complete each other in control of behavioral learning and performance according to the multiple memory systems model (Lee et al. 2008). The systems compete with one another during some learning situation and allow optimal responding in a particular situation (Poldrack & Packard 2003). Reduced BDNF may by modifying these simultaneously and in parallel operating systems worsen hippocampus-dependent learning but ameliorate hyperactivity and deficits in startle responses and sensorimotor learning in Fmr1 KO mice.
Conditionally, ablated BDNF expression in the adult mouse hippocampus impairs the survival of newborn cells (Choi et al. 2009). We observed a reduction of cell survival in the ventral part of the hippocampus in Bdnf+/− (Sairanen et al. 2005), Fmr1 KO (Eadie et al. 2009), and dMT mice in agreement with reduced hippocampal BDNF expression in these mice. However, the genetic deletion of one Bdnf allele prevented the short neurite phenotype characteristic of cells lacking FMRP during early neuronal differentiation, suggesting that at least some beneficial effects of reduced BDNF on the behavioral phenotype of Fmr1 KO mice could be caused by alterations in the differentiation and maturation of neuronal cells.
BDNF is essential for the survival of certain sensory neurons during postnatal development (Valdés-Sánches et al. 2010) but it is also independently of its survival-promoting function required for the normal function of slowly adapting mechanoreceptors (Carroll et al. 1998). Reduced BDNF expression in Bdnf+/− mice results in partial reduction of nociceptive afferent neurons as well as an increased latency of heat-evoked pain responses (Valdés-Sánches et al. 2010). However, we observed that basal responses to mechanical or heat stimulation of Bdnf+/− mice did not differ from WT controls. Similarly, basal responses of Fmr1 KO mice were intact in the present as well as an earlier study (Price et al. 2007) and the absence of FMRP together with further reduced BDNF expression did not result in abnormalities in the basal responses to noxious stimuli in dMT mice. Contrary to a previous study (Price et al. 2007), we failed to observe a decrease in the second phase of formalin-induced pain behaviors of Fmr1 KO mice, nor did Bdnf+/− mice show changes in formalin-induced spontaneous pain behaviors when compared with WT controls. However, the second phase of the formalin-induced pain behaviors of dMT mice was significantly increased indicating that reduced BDNF expression together with the loss of FMRP resulted in a phenotype not seen in single mutants. Fmr1 KO mice also exhibited a reduction in central hypersensitivity not seen in Bdnf+/− mice and the reduced BDNF expression did not alter this. The reduction in central hypersensitivity to mechanical stimulation is in line with a delayed development of centrally mediated tactile allodynia described earlier in Fmr1 KO mice with an experimental neuropathy (Price et al. 2007). Recent studies suggest that FMRP plays a role in homeostatic synaptic plasticity and repressing activity-dependent translation of BDNF in neurons may generate a feedback loop altering neuronal activity at the level of the neuron rather than single synapses (Darnell et al. 2011). A reduction in central hypersensitivity in FXS could also reflect alterations of neuronal activity that leads to disturbances in the activation of higher level integrative brain regions.
Hyperactivity in the phenotype of Fmr1 KO mouse resembles ADHD frequently found in FXS individuals. Inappropriate levels of hyperactivity and inattentiveness in ADHD have been shown to correlate with an increase in circulating BDNF (Shim et al. 2008). This is in line with present findings which indicate that locomotor hyperactivity of FXS mice is prevented by reducing BDNF expression. Motor learning requires attention in early phases. Alterations in motor coordination and activity are features associated with autism as well as Fmr1 KO mice (Bernardet & Crusio 2006). Blood BDNF levels are also reported to be abnormally high in autism and mental retardation (Miyazaki et al. 2004; Nelson et al. 2001) but the mechanisms underlying BDNF increases are not known. Reduced startle reflex as seen in Fmr1 KO is found in a number of people with schizophrenia and in psychopathic individuals (Docherty 1996; Patrick et al. 1993). Altogether, our findings indicate that BDNF is involved in synaptic dysfunction that underlies behavioral phenotype in FXS (Belmonte & Bourgeron 2006) and suggest in line with numerous other molecular and genetic studies that BDNF/TrkB signaling plays a role in the pathogenesis of developmental pervasive disorders (Belmonte & Bourgeron 2006; Correia et al. 2011; Miyazaki et al. 2004; Nelson et al. 2001; Perry et al. 2001). However, the final human FXS phenotype may be modulated in a complex manner by several environmental factors and gene polymorphisms. This may influence the outcome of clinical studies with different experimental approaches. For example, a Finnish population-based study of FXS men found associations between polymorphisms in the BDNF gene with epilepsy (Louhivuori et al. 2009) whereas a Spanish study that used a more heterogenous population found no association of the single Val66Met polymorphism in the BDNF gene with epilepsy (Tondo et al. 2011).
In summary, our research shows that BDNF displays numerous and complex effects on the behavioral phenotype of FXS. While reduction of BDNF expression has negative effects on cognitive behavior in adult male Fmr1 KO mice, it has beneficial effects on motor activity and motor skill learning. Thus, some behavioral abnormalities in Fmr1 KO mice may be induced by increased BDNF signaling and could theoretically be counteracted by reducing extracellular BDNF levels. However, subcellular alterations as well as regional and age-dependent changes in the BDNF expression in the brain of Fmr1 KO mice open questions regarding the mechanisms underlying the phenotypic outcomes in dMT mice. Additional studies using cell cultures and conditional knockout mice are needed to further characterize the role of BDNF in the FXS phenotype during different developmental stages and ageing. Understanding of the diverse actions of BDNF and their final contribution to the behavioral phenotype of animal models will facilitate evaluation of appropriate targets to treat behavioral problems of individuals with neurodevelopmental diseases such as FXS.
We are very grateful to Outi Nikkilä for help in mouse genotyping and Dr Voikar Vootele for critical review of the behavioral studies. Mouse Behavioral Unit is supported by Biocenter Finland. This work was supported by grants from the Academy of Finland, Rinnekoti Foundation and University of Helsinki. The authors declare no competing financial interests.