Temporal requirements of the fragile X mental retardation protein in the regulation of synaptic structure

The fragile X syndrome (FraX) mental retardation and autism spectrum disorder, which has a prevalence of ∼1:4000 males and 1:6000 females, is among the most common inherited neurological diseases(Koukoui and Chaudhuri, 2007; Penagarikano et al., 2007). Loss of fragile X mental retardation protein (FMRP) expression is the sole cause of the disease state. FMRP is an mRNA-binding protein that regulates mRNA stability and translation for a number of synaptic and cytoskeleton-associated proteins (Castets et al., 2005; Lu et al.,2004; Muddashetty et al.,2007; Reeve et al.,2005; Todd et al.,2003; Zalfa et al.,2007; Zhang et al.,2001). FMRP function regulates the activity-dependent control of synaptic connections via intersection with group 1 metabotropic glutamate receptor (mGluR) signaling (Antar et al.,2004; Bear et al.,2004; Ferrari et al.,2007; Huber et al.,2002; Nosyreva and Huber,2006; Pan and Broadie,2007; Pan et al.,2008). The synapsopathic clinical manifestations of the disease include mild to severe mental retardation(Penagarikano et al., 2007),delayed and depressed developmental trajectories(Bailey et al., 2001a; Bailey et al., 2001b), deficits in short-term memory (Cornish et al.,2001; Munir et al.,2000), hyperactivity (Einfeld et al., 1991), disordered sleep(Gould et al., 2000), seizures(Sabaratnam et al., 2001), and the cytological presentation of long, immature-looking cortical dendritic spines, indicating inappropriate development and/or failure of pruning and synapse elimination (Hinton et al.,1991; Rudelli et al.,1985).

Understanding FraX pathogenesis, and subsequently designing effective FraX interventions, requires knowledge of the temporal requirement(s) of FMRP function. A fundamental need is to determine whether FraX is primarily a`developmental disease', reflecting a transient role for FMRP in progressive neuronal maturation, a `plasticity disease', reflecting a maintained,constitutive requirement for FMRP at maturity, or some combination of a two-phase requirement giving rise to different FraX symptoms. This study aims to begin resolving this vital question using our well-characterized Drosophila FraX disease model(Zhang et al., 2001). Drosophila Fmr1 (dfmr1)-null mutants are fully viable but display impaired learning and memory(Dockendorff et al., 2002),arrhythmic circadian motor activity(Dockendorff et al., 2002; Inoue et al., 2002),over-elaboration of neuronal structure(Michel et al., 2004; Morales et al., 2002; Pan et al., 2004; Zhang et al., 2001), and altered neuronal function (Zhang et al.,2001). The primary synaptic model is the neuromuscular junction(NMJ), which displays increased synapse arborization and branching, increased synaptic bouton number, and elevated neurotransmission. As in mammals, Drosophila FMRP (dFMRP) functionally interacts with mGluR-mediated synaptic glutamatergic signaling in the regulation of postsynaptic glutamate receptor expression (Pan and Broadie,2007) and in sculpting presynaptic architecture(Pan et al., 2008).

In this study, we use the conditional, transgenic Gene-Switch (GS) method(Osterwalder et al., 2001) to drive wild-type dFMRP expression in dfmr1-null mutants. This approach allows targeted dFMRP expression during discrete temporal windows, enabling the definition of critical periods of function. We show that constitutive neuronal dFMRP expression rescues all NMJ synaptic structural defects,demonstrating a strictly presynaptic dFMRP requirement, with a mechanistic role in microtubule cytoskeleton regulation. By contrast, targeted presynaptic dFMRP expression does not rescue neurotransmission function in the null mutant, indicating a separable postsynaptic dFMRP requirement. Temporally, we show that transient early-development expression of dFMRP strongly rescues all facets of synaptic architecture, demonstrating an early role for dFMRP in establishing synapse morphology. We also show that acute dFMRP expression at maturity weakly rescues a subset of synaptic structure defects, showing that dFMRP can mediate some structural plasticity and that late-stage intervention might be beneficial.

Fly stocks were maintained at 25°C on standard medium. The w¹¹¹⁸ line served as the genetic background control for the dfmr1 null. Recombinant parental lines harboring the dfmr1-null allele (dfmr1^50M)(Zhang et al., 2001) and either a wild-type dfmr1 transgene under UAS control (UAS-9557-3)(Zhang et al., 2001) or the neuronal-specific driver elav-Gene-Switch construct (GSG-301)(Osterwalder et al., 2001)were generated using standard genetic techniques. Henceforth, `GS' as a genotype descriptor refers to the following: dfmr1^50M,elav-GSG-301/dfmr1^50M, UAS-9557-3. For RU486(mifepristone; Sigma, St Louis, MO) dosing, the drug was dissolved in 80%ethanol (EtOH) and mixed with food to the desired concentration (shown in figures as GS+RX with RU486 in μg/mL). For vehicle control, the equivalent volume of EtOH was used to identically treat the GS line (GS+E). GS animals were either constitutively raised on supplemented/control food or transferred at staged times as indicated.

The central nervous system (CNS), including the brain and the ventral nerve cord (VNC), was dissected free from staged and treated larvae in Ca²⁺-free modified Jan's standard saline(Jan and Jan, 1976). Dissected CNS samples were homogenized and boiled in 1× NuPage sample buffer(Invitrogen, Carlsbad, CA) supplemented with 40 mM DTT. The total protein from 2-6 brains per sample (depending on larval age) was loaded onto 4-12% Tris-bis acrylamide gels and electrophoresed in NuPage MES buffer (Invitrogen) for 1 hour at 200 V. Transfer to nitrocellulose was carried out for 1 hour at 100 V in NuPage transfer buffer (Invitrogen)/10% methanol. Processing was completed using the Odyssey near-infrared fluorescence detection system (Li-COR,Lincoln, NE) to enable quantitative western blot analysis. Antibodies used included: anti-dFMRP [1:3000; 6A15 (monoclonal), Sigma], anti-α-Tubulin[1:100,000; B512 (monoclonal), Sigma] and Alexa-Fluor 680-conjugated goat anti-mouse (1:10,000; Invitrogen-Molecular Probes). Raw integrated intensities were calculated for dFMRP for the lower molecular weight band of the doublet and for the α-Tubulin band. The ratio of dFMRP:α-Tubulin was used to normalize for loading.

Staged animals were dissected in standard saline and then fixed for 40 minutes with 4% paraformaldehyde/4% sucrose in PBS (pH 7.4). Preparations were rinsed with PBS, then blocked and permeablized with 0.2% Triton X-100 in PBS(PBST) containing 1% bovine serum albumin (BSA) and 0.5% normal goat serum(NGS) for 1 hour at room temperature. Primary and secondary antibodies were diluted in PBST containing 0.2% BSA and 0.1% NGS and incubated overnight at 4°C and 2 hours at room temperature, respectively. Antibodies used included: anti-dFMRP (1:500; 6A15), anti-Discs large (DLG) [1:200; 4F3(monoclonal), Developmental Studies Hybridoma Bank (DSHB), University of Iowa], anti-Futsch [1:200; 22C10 (monoclonal), DSHB], anti-horseradish peroxidase (HRP) [1:250; (polyclonal), Sigma], and Alexa-Fluor-conjugated secondaries (1:250; Invitrogen-Molecular Probes). All fluorescent images were collected using a Zeiss LSM 510 META laser-scanning confocal microscope.

The muscle 4 NMJ of abdominal segment 3 was used for all quantification. Values were determined for both left and right hemisegments; averaged for each n=1. Synapse junctional area was measured as the maximal cross-sectional area in a maximum projection of each collected z-stack. A synaptic branch was defined as an axonal projection with at least two synaptic boutons. Two bouton classes were defined: (1) type Ib(>2 μm diameter) and (2) mini/satellite (≤2 μm diameter and attached to a type Ib bouton of mature size). Each class is reported as number per terminal. ImageJ(http://rsb.info.nih.gov/ij/)was used for fluorescence intensity thresholding, automated regional outline and area calculation.

Staged animals were dissected in standard saline (containing 0.2 mM CaCl₂). NMJ preparations were loaded with FM1-43 (10 μM;Invitrogen-Molecular Probes) using depolarizing 90 mM KCl standard saline(containing 1.8 mM CaCl₂) for 5 minutes, washed and imaged. Preparations were then unloaded with the same stimulation for 2 minutes in the absence of FM1-43, washed and imaged. For quantification, only the muscle 4 NMJ in abdominal segments A2-A4 was used. Average fluorescence intensity values and bouton areas were measured from three NMJs per animal, with six individual boutons per NMJ assayed, and then averaged to generate a single data point (n=1 from 18 boutons). The fluorescence intensity units(FIU) measured per bouton are shown, together with the ratio of FM1-43 unload:load fluorescence intensity.

Two-electrode voltage-clamp recordings were made at 18°C from muscle 6 in abdominal segments A2-A4 of wandering third instar larvae to examine miniature excitatory junctional currents (mEJCs)(Zhang et al., 2001). Borosilicate glass electrodes were filled with 3 M KCl in standard saline containing: 128 mM NaCl, 2 mM KCl, 4 mM MgCl₂, 70 mM sucrose, 5 mM HEPES (pH 7.2) and 0.2 mM CaCl₂. Tetrodotoxin (3 μM; Sigma) was added to block action potentials. Each n=1 results from 240 seconds of gap-free recording from independent animals. Traces were filtered using a low-pass 8-pole Bessel filter with -3 dB cut-off of 0.5 kHz. Data were analyzed using Clampfit 9.2 (Molecular Devices, Sunnyvale, CA) using template-based event detection. All traces were analyzed for mean peak amplitude (nA) and frequency (Hz).

Statistical analysis was performed using GraphPad InStat 3 (GraphPad Software, San Diego, CA). Generally, unpaired t-tests were used to compare means of control and dfmr1-null, and Tukey-Kramer multiple comparisons tests were applied to all GS categories. In FM1-43 experiments,Dunnet's multiple comparison tests were also used to compare each value independently with the stated control. Significance levels in figures are represented as P<0.05 (^*); P<0.01(^**); P<0.001 (^***). Error bars represent s.e.m.

The conditional Gene-Switch (GS) method utilizes a modified UAS-GAL4 system that provides a means of tightly regulating tissue-specific transgene expression in order to define spatiotemporal requirements(McGuire et al., 2004; Osterwalder et al., 2001). GS-GAL4 lines are dependent on the progesterone analog RU486 (mifepristone) to drive hormone-responsive UAS-construct expression(Fig. 1A). We have driven a UAS-dfmr1 transgene with neuronal elav-GS GAL4 in the dfmr1-null mutant background to determine spatial and temporal requirements for dFMRP at the NMJ synapse. Thus, we assayed, in a targeted fashion, the presynaptic roles of dFMRP in synapse assembly and function following constitutive expression and timed intervention windows.

The critical first step was to determine RU486 dosage sufficient to match dFMRP levels in the wild type, so that the transgenic protein is not under- or overexpressed. RU486 dosage-dependence tests were conducted by assaying dFMRP expression in the larval CNS by western blot(Fig. 1B,C) and by in situ immunohistochemistry (Fig. 1D). Analyses were performed on wild-type control (w¹¹¹⁸), dfmr1-null and dfmr1-null animals harboring both the elav-GS GAL4 driver and UAS-dfmr1 transgene (henceforth referred to as GS animals), constitutively fed with RU486 (GS+R) or with ethanol vehicle only (GS+E). Feeding with 0.5 μg/mL RU486 yielded dFMRP levels closely approximating that of the wild-type control (98±5%),whereas 1 and 2 μg/mL RU486 induced progressive overexpression(190±32% and 255±25% compared with control, respectively; 2μg/mL versus control, P<0.05; Fig. 1C). Thus, in GS animals,RU486 drives dose-dependent expression of dFMRP in the nervous system, and a dosage of 0.5 μg/mL fed constitutively generates dFMRP expression indistinguishable from those of the wild type.

In the Drosophila FraX model, loss of dFMRP causes NMJ over-elaboration (including excessive branching, overgrowth and supernumerary synaptic boutons), accumulation of mini- or satellite boutons, and altered functional neurotransmission properties due to both pre- and postsynaptic defects (Zhang et al., 2001). To quantify branching, individual projections harboring more than two boutons were counted as synaptic branches (control 3.0±0.3 branches versus dfmr1-null 4.8±0.2 branches, n=12, P<0.001; Fig. 2A,B). To quantify synaptic area, the junction was delimited by either HRP (presynaptic) or DLG (postsynaptic) and the area measured (for example, HRP - control 160±8 μm² versus dfmr1-null 241±12 μm², n=12, P<0.001; Fig. 2C,D). Lastly, type Ib bouton numbers per terminal were counted,partitioned into either mature boutons (control 19±1 boutons versus dfmr1-null 29±1 boutons, n=12, P<0.001; Fig. 2F) or mini/satellite boutons based on size and relative positioning (control 0.6±0.2 mini-boutons versus dfmr1-null 3.7±0.4 mini-boutons, n=12, P<0.001; Fig. 2E,G).

We tested for rescue of architectural defects upon constitutive presynaptic dFMRP induction. NMJs were double-labeled with anti-HRP, to delineate the innervating presynaptic neuron, and with anti-DLG, to reveal the postsynaptic domain of the target muscle (Fig. 2A). GS animals fed with vehicle (EtOH) fully phenocopied the dfmr1-null with respect to all structural abnormalities(Fig. 2). By sharp contrast,constitutively RU486-fed animals were completely rescued, with entirely normal synaptic architecture. Presynaptically targeted dFMRP was assayed with two RU486 dosages: 0.5 μg/mL (wild-type dFMRP level; Fig. 1C) and 2.0 μg/mL(significantly elevated dFMRP; Fig. 1C). As predicted, the wild-type-control-matched dFMRP expression provided the most exact rescue of synaptic structure features. First, synaptic branch number was perfectly rescued from the elevated branching that characterizes the null mutant [GS+E 4.5±0.3 branches versus GS+RU486(0.5 μg/mL) 3.0±0.3 branches, n=12, P<0.001; Fig. 2B]. Second, both pre- and postsynaptic areas were restored to control levels [for example, HRP junctional area - GS+E 255±16 μm² versus GS+RU486 (0.5μg/mL) 150±7 μm², n=12, P<0.001; Fig. 2C,D]. Finally, the overproliferation of synaptic boutons was rescued for both the large, mature boutons [GS+E 28±1 boutons versus GS+RU486 (0.5 μg/mL) 17±1 boutons, n=12, P<0.001; Fig. 2F] and the small,immature satellite boutons [GS+E 3.4±0.3 mini-boutons versus GS+RU486(0.5 μg/mL) 0.3±0.1 mini-boutons, n=12, P<0.001; Fig. 2G]. These findings demonstrate an entirely presynaptic requirement for dFMRP in synaptic structuring.

Having confirmed GS system utility for examining morphological rescue strategies, we next moved to presynaptic dFMRP induction during discrete temporal windows. To test developmental roles in the establishment of synaptic organization, early GS induction of dFMRP was first examined. GS animals were treated with RU486 within 3 hours of hatching for a brief period of 12 hours(Fig. 3A). Neuronal dFMRP levels were then immediately analyzed for comparison to that of the wild-type control (Fig. 3B). RU486 fed at 0.1 μg/mL yielded dFMRP levels closest to that of the control at 142±48%, whereas higher concentrations of RU486 induced steep,dosage-dependent overexpression of dFMRP (0.25 and 0.5 μg/mL induced 397±128% and 556±268%, respectively; Fig. 3C).

The duration of dFMRP expression was analyzed by monitoring dFMRP levels at timed periods following withdrawal of RU486(Fig. 3D). At 24 hours post-treatment, there was a ∼70% reduction relative to the initial,induced dFMRP level, irrespective of RU486 dosage (0.1 μg/mL RU486 45±23%, and 0.25 μg/mL RU486 107±17%, as compared with control; Fig. 3E). Loss of dFMRP was progressive, with levels after 0.1 μg/mL RU486 treatment declining to 35±8% at 48 hours, 10±2% at 60 hours and undetectable by 72 hours post-treatment(Fig. 3E). These studies indicate that dFMRP can be rapidly and strongly induced within hours, but that the inherent stability of dFMRP causes persistence during a period of gradual decline. Analysis of normalized dFMRP:α-Tubulin intensity values indicates an apparent dFMRP half-life of 25.5±1.7 hours. This measured relative stability of dFMRP therefore limits the resolution of temporal expression windows.

Both FraX patients and Fmr1-knockout (KO) mice display an increased number of immature-appearing synaptic spines, similar to those occurring during neocortical development(Comery et al., 1997; Galvez and Greenough, 2005; Hinton et al., 1991; Irwin et al., 2001; Nimchinsky et al., 2001; Rudelli et al., 1985). In Fmr1 mutant mice, these aberrant synapses are developmentally transient and disappear, but then reappear later in the mature animal(Galvez and Greenough, 2005; Nimchinsky et al., 2001). These developmental dynamics suggest specific functions for FMRP during defined windows, especially early in synaptogenesis. To test this hypothesis in our model, dFMRP expression was induced for a brief 12-hour window immediately following hatching (Fig. 3A) and mature third instar larvae were collected 72 hours post-treatment [∼108 hours after egg lay (AEL)], with NMJ synapses co-labeled for HRP and DLG to compare RU486-treated with the EtOH-treated control (Fig. 4A). RU486 concentrations of 0.1 μg/mL and 0.25 μg/mL were analyzed, representing normal and overexpression levels, respectively.

Significant and complete rescue of the dfmr1-null structural phenotypes was observed with transient early expression of dFMRP(Fig. 4B-E). The synaptic overbranching that is characteristic of the dfmr1-null was fully rescued at the higher dFMRP level [GS+E 4.8±0.2 branches, n=14, versus GS+RU486 (0.25 μg/mL) 3.4±0.1 branches, n=12, P<0.001; Fig. 4B]. The greater synaptic bouton number was similarly restored to the wild-type level (GS+E 27±1 boutons, n=14, versus both 0.1μg/mL and 0.25 μg/mL GS+RU486 21±1 boutons, n=12, P<0.01; Fig. 4C). Finally, the increased synaptic junction area was rescued by both dFMRP levels; for example, the HRP presynaptic area was comparable under both conditions [GS+E 206±8 μm², n=14, versus both GS+RU486 (0.1 μg/mL) 163±8 μm², n=12, P<0.01, and GS+RU486 (0.25 μg/mL) 178±7μm², n=12, P<0.05; Fig. 4D,E]. These findings indicate a specific early developmental requirement for dFMRP in sculpting synaptic architecture. Further, the persistence of normal synaptic structure at the mature NMJ in the absence of dFMRP suggests that it is not required for the maintenance or stability of synaptic structure once established.

A crucial question for FraX patients is whether late correction of the FMRP deficit can ameliorate disease symptoms. Having identified the early role for dFMRP in the establishment of NMJ synaptic morphology, we next examined whether reintroduction of dFMRP into mature animals could rescue synapse structural defects. GS animals were treated with either RU486 or EtOH for 12 hours at a mature larval time point (96-108 hours AEL; Fig. 5A) and then immediately analyzed. Protein level comparisons showed that RU486 fed at 1-2 μg/mL drives dFMRP levels that are indistinguishable from that of the wild-type control (99±9% and 90±6%, respectively), whereas RU486 at 5μg/mL elevates dFMRP to 227±21% of control (compared with both 1 and 2 μg/mL, P<0.01 and P<0.001, respectively; Fig. 5B,C). Both the low (1μg/mL) and high (5 μg/mL) RU486 induction levels were used to examine the effect on synaptic architecture.

Examination of NMJ structure following late intervention showed slight but significant rescue of a subset of defects(Fig. 5D). There was no rescue of either branching or synapse area (data not shown). However, the increased synaptic bouton number (GS+E 27±1 boutons, n=12) was significantly reduced following the 12-hour acute intervention [GS+RU486 (1μg/mL) 22±1 boutons, n=12, and GS+RU486 (5 μg/mL)23±1 boutons, n=11, P<0.01; Fig. 5E]. Thus, late intervention using the GS system to induce targeted presynaptic dFMRP mediates a weak, partial rescue of synaptic structural phenotypes. This finding indicates both maintained morphological plasticity at the NMJ synapse and at least modest reversal potential for specific mutant phenotypes.

In addition to synaptic architecture, we examined the synaptic organization of the known dFMRP target Futsch, the microtubule-associated MAP1B homolog(Hummel et al., 2000). Futsch is negatively regulated by dFMRP, and dfmr1; futsch double-mutants display normal NMJ architecture (Zhang et al., 2001). Futsch is required for dendritic, axonal and synaptic development (Bettencourt da Cruz et al.,2005; Roos et al.,2000; Ruiz-Canada et al.,2004). In wandering third instar larvae, dfmr1-null synapses were found to contain significantly elevated numbers of Futsch-positive cytoskeletal loops within synaptic boutons (control 2.0±0.4 loops, n=11, versus dfmr1-null 4.3±0.4 loops, n=12, P<0.001; Fig. 6A,B). In control animals,the loop structures were usually restricted to terminal boutons. By contrast, dfmr1-null NMJs displayed Futsch-positive loop structures abnormally interspersed throughout the entire synaptic arbor(Fig. 6A). GS vehicle-fed animals phenocopied the dfmr1-null, and the defect was partially rescued by constitutively expressing dFMRP [GS+E 4.7±0.2 loops, n=12, versus both GS+RU486 (0.5 μg/mL) 3.5±0.4 loops, n=12, P<0.05, and GS+RU486 (2 μg/mL) 3.2±0.3 loops, n=13, P<0.01; Fig. 6B].

Examining temporal windows of GS intervention, we observed that 12-hour RU486 treatment at either early or mature larval time points restored a more normal cytoskeletal arrangement when examined at 108 hours AEL. In the early treatment window, complete rescue was observed only at the higher RU486 dosage[GS+E 7.8±0.4 loops, n=11, versus GS+RU486 (0.25 μg/mL)5.2±0.3 loops, n=12, P<0.001; Fig. 6C]. In the late treatment, progressive rescue was achieved in a dose-dependent manner with both RU486 concentrations tested [GS+E 8.8±0.6 loops, n=15,versus GS+RU486 (1 μg/mL) 6.1±0.2 loops, n=10, P<0.01, and GS+RU486 (5 μg/mL) 5.8±0.4 loops, n=11, P<0.001; Fig. 6C]. These findings suggest that the synaptic organization of the Futsch-positive microtubule cytoskeleton remains plastic throughout development and at maturity, and that dFMRP has a constitutively significant role in modulating this mechanism.

We next examined the potential for dFMRP induction to rescue synapse functional properties. dfmr1-null NMJ synapses exhibit a ∼2-fold increase in neurotransmission strength(Zhang et al., 2001), but it has not been shown whether the change is due to elevated glutamate release or increased glutamate receptor function. To make this distinction, we used the lipophilic styryl dye FM1-43, which incorporates into the presynaptic vesicle pool, so that its loss or retention can be visualized to monitor activity-dependent vesicle cycling(Brumback et al., 2004; Fergestad and Broadie, 2001; Rohrbough et al., 2004). After a round of stimulated dye loading and unloading(Fig. 7A), a significantly greater amount of the dye was released from mutant than control vesicles(control 0.46±0.04 unload:load, n=8, versus dfmr1-null 0.31±0.04 unload:load, n=8, P<0.01; Fig. 7B). The average area assayed per bouton was equivalent (control 9.3±1.5μm², n=5, versus dfmr1-null 9.2±1μm², n=5), ensuring that the differences observed were due to dye cycling rates and not sampling variability. The average fluorescence intensity unit (FIU) values obtained after loading were comparable (control 153±6 FIU versus dfmr1-null 153±5 FIU, n=5; Fig. 7C),but unloading was more rapid in the mutant (control 81±7 FIU versus dfmr1-null 43±4 FIU, n=5, P<0.01; Fig. 7C). Thus, the depressed level of retained dye in dfmr1 mutants indicates an enhanced rate of vesicle exocytosis and provides a mechanistic explanation for the elevated evoked synaptic transmission (Zhang et al., 2001).

Surprisingly, constitutive presynaptic dFMRP expression provided no rescue of the FM1-43 functional transmission defect in GS animals [GS+E 0.29±0.02 unload:load, n=13; GS+RU486 (0.5 μg/mL)0.30±0.03 unload:load, n=6; GS+RU486 (2 μg/mL)0.29±0.01 unload:load, n=9; Fig. 7B]. Again, actual fluorescence intensity values illustrate the heightened unloading phase and the lack of any phenotypic rescue upon GS induction [load: GS+E 124±8,GS+RU486 (0.5 μg/mL) 130±12, GS+RU486 (2 μg/mL) 127±5, n=5; unload: GS+E 33±4, GS+RU486 (0.5 μg/mL) 31±2,GS+RU486 (2 μg/mL) 36±3, n=5; Fig. 7C]. Of note, GS animals under all conditions appeared to load at slightly lower absolute levels;however, only the difference seen between w¹¹¹⁸ control(153±6) and GS+E (124±8) is statistically significant(P<0.05). These findings indicate an apparent postsynaptic requirement for dFMRP, with retrograde signaling control of presynaptic function. They also indicate that dFMRP displays differential spatial requirements in relation to synaptic structure and function.

To further examine functional requirements, two-electrode voltage-clamp(TEVC) recordings were made to monitor miniature excitatory junctional currents (mEJCs), as a direct measure of glutamate release(Fig. 8A). The mEJC amplitude was comparable between all genotypes, with a modest difference between control(1.01±0.04 nA, n=10) and dfmr1-null (0.84±0.04 nA, n=10, P<0.05; Fig. 8B). By contrast, substantial differences were observed in mEJC frequency, with a nearly 3-fold increase in the dfmr1-null compared with control (control 1.2±0.2 Hz versus dfmr1-null 3.4±0.5 Hz, n=10, P<0.01; Fig. 8C). Elevated mEJC frequencies were also present in the GS animals (GS+E 2.6±0.2 Hz, n=10) and, upon presynaptic dFMRP induction, the condition was markedly exacerbated, not rescued toward control [GS+RU486 (0.5 μg/mL)5.0±0.7 Hz; GS+RU486 (2 μg/mL) 6.4±0.9 Hz, n=10; Fig. 8C]. These findings provide strong support for a postsynaptic dFMRP requirement in regulating synapse function.

FraX is caused by the loss of FMRP; however, the spatial and temporal requirements for FMRP are largely unknown. The definition of these requirements in nervous system connectivity, transmission and plasticity is essential for understanding FraX pathophysiology and informing the design of effective FraX therapeutic intervention strategies. In this study, we tackle these questions using the established Drosophila FraX model, focusing on the NMJ model synapse. We use the inducible GS system to control spatial and temporal dFMRP expression in an otherwise null mutant. We show that constitutively expressed presynaptic dFMRP fully rescues synapse architecture and cytoskeleton patterning, but does not restore normal synaptic function. We therefore conclude that dFMRP has crucial presynaptic roles controlling synaptic morphology; however, the regulation of neurotransmission strength is mediated by postsynaptic function. We demonstrate that a brief dFMRP pulse early in development is sufficient to meet presynaptic needs controlling synapse configuration, but that a brief pulse at maturity also partially restores normal synapse structure. We therefore conclude that dFMRP plays a primary role early in synapse assembly, but the synapse maintains morphological plasticity. Moreover, late dFMRP reintroduction has some ability to counteract the effects of dFMRP loss throughout development. These conclusions suggest that FraX interventions should be targeted to young children in order to maximize efficacy, but also that restoration, or perhaps circumvention, of FMRP function at maturity could also be beneficial as a FraX treatment strategy.

Although neurological consequences are readily identifiable in the FraX disease state, the causative spatial defects associated with FMRP loss are largely undefined. The most-studied structural anomaly in FraX patients and Fmr1 KO mice is increased density of morphologically immature dendritic spines (Comery et al.,1997; Galvez and Greenough,2005; Hinton et al.,1991; Irwin et al.,2001; Nimchinsky et al.,2001; Rudelli et al.,1985). Although this is commonly treated as a solely postsynaptic defect, spines are synaptic contacts, by definition, so there should be an equivalent defect in opposing presynaptic boutons. Thus, the FMRP requirement could be presynaptic, postsynaptic or both. Drosophila Fmr1 mutants similarly display synaptic architecture defects in several classes of neurons,including motoneurons (Zhang et al.,2001), lateral and dorsal cluster neurons(Morales et al., 2002), and mushroom body Kenyon cells, a site of learning and memory consolidation(Michel et al., 2004; Pan et al., 2004). At the NMJ,synaptic defects have been attributed to both pre- and postsynaptic roles of dFMRP, as both neuronal and muscle overexpression result in altered architecture (Zhang et al.,2001).

Functionally, the Fmr1 KO mouse displays decreased cortical long-term potentiation (LTP) (Larson et al., 2005; Li et al.,2002) and increased hippocampal long-term depression (LTD)(Huber et al., 2002; Nosyreva and Huber, 2006). The LTD alteration is suggested to be postsynaptically mediated because FMRP normally functions to suppress translation of dendritic mRNAs to attenuate LTD(Bear et al., 2004); however,no spatial dissection of this function has been reported. In Drosophila, loss of dFMRP results in altered neurotransmission in both visual system synapses and at the NMJ(Zhang et al., 2001). In the most-studied NMJ location, elevated neurotransmission has been suggested to be primarily due to a presynaptic role for dFMRP, as neuronal overexpression results in dramatic elevation in spontaneous vesicle fusion events(Zhang et al., 2001). However,variability in functional outputs might reflect differential FMRP roles in trans-synaptic signaling, and decreased postsynaptic sensitivity is often offset by a compensatory increase in presynaptic release(Davis et al., 1998; Frank et al., 2006; Paradis et al., 2001; Petersen et al., 1997; Plomp et al., 1992; Sandrock et al., 1997).

The current study clearly demonstrates a presynaptic role for dFMRP in regulating NMJ synaptic architecture, including terminal area, synaptic branching, and the formation of synaptic boutons, all of which are negatively regulated by dFMRP function in the presynaptic cell. Perhaps most interesting is the accumulation of mini/satellite boutons in the absence of presynaptic dFMRP function. Work by us and others has suggested that these tiny boutons represent a developmentally arrested state at an otherwise normal stage of bouton maturation (Ashley et al.,2005; Beumer et al.,1999; Dickman et al.,2006; Torroja et al.,1999). Since the dfmr1-null also shows a supernumerary abundance of mature synaptic boutons, the accumulation of mini-boutons suggests that the absence of presynaptic dFMRP function triggers the initiation of a disproportionate number of bouton formation events, but that other proteins required for bouton maturation are limiting. Therefore, dFMRP has a primary role in restricting bouton deposition. An early-development pulse of dFMRP prevents the aberrant accumulation of mature boutons, but does not prevent accumulation of mini-boutons. Thus, dFMRP is constitutively required in the presynaptic cell to arrest this nascent stage of synaptic bouton formation.

We established previously that dFMRP acts as a translational repressor of the MAP1B homolog Futsch (Zhang et al.,2001). dfmr1-null phenotypes are mimicked by presynaptic overexpression of Futsch, and genetic control of Futsch overexpression in the dfmr1-null background completely rescues NMJ overgrowth phenotypes. Subsequent mouse studies revealed the same MAP1B upregulation and associated enhanced microtubule stability in Fmr1 KO neurons(Lu et al., 2004). At the Drosophila NMJ, Futsch binds microtubule hairpin loops in a dynamic subset of synaptic boutons (Roos et al.,2000). In other systems, the appearance of these microtubule structures is linked to the stalling of axonal growth cones(Dent and Kalil, 2001; Tanaka and Kirschner, 1991; Tsui et al., 1984), which predicts a similar role in synapse growth(Roos et al., 2000). In dfmr1-nulls, however, there is an increased number of Futsch loops throughout the overgrown NMJ synaptic arbor, a defect rescued by targeted presynaptic dFMRP expression. Furthermore, developmental analyses reveal that Futsch loops are more abundant during earlier stages of synapse assembly than at maturity (compare larvae at 108 hours AEL with wandering third instars). In dfmr1-nulls, Futsch loops are significantly more abundant, both during active synapse growth and at maturity. These results suggest that the dynamic growth capacity of the synapse is reflected by the number of Futsch loops as a function of developmental time and, thus, that dfmr1mutants are arrested in a premature growth state.

Presynaptic induction of dFMRP totally fails to rescue the elevated cycling of synaptic vesicles in the dfmr1-null mutant, indicating that this defect has its origin in a postsynaptic function of dFMRP. The known postsynaptic function at the Drosophila NMJ is selection of the appropriate glutamate receptor classes, with relative abundances dramatically skewed by the absence of dFMRP (Pan and Broadie, 2007). Therefore, our results suggest that defects in the postsynaptic glutamate receptor field cause a compensatory change in the presynaptic vesicle cycling underlying glutamate release, presumably via a trans-synaptic retrograde signal (Davis et al., 1998; Frank et al.,2006; Paradis et al.,2001; Petersen et al.,1997). In support of this hypothesis, a recent study showed that presynaptic Fmr1 genotype influences synaptic connectivity in a mosaic mouse model, with neurons lacking FMRP being less likely to form functional synapses (Hanson and Madison,2007). Conversely, acute postsynaptic expression of FMRP in Fmr1 KO neurons results in a decrease in the number of functional and structural synapses relative to neighboring untransfected neurons, indicating phenotypic rescue (Pfeiffer and Huber,2007). Although it is possible that pre- and postsynaptic effects are independent of one another, trans-synaptic compensation warrants consideration in dissecting the spatial requirement of FMRP in modulating synaptic function.

In mice, the appearance of Fmr1 mutant phenotypes is age-dependent, with at least some defects appearing transiently. In layer V barrel cortex, KO neurons display abnormal dendritic spine length/density during cortical synaptogenesis early in development (postnatal week 1);however, these differences are undetectable by week 4(Galvez and Greenough, 2005; Nimchinsky et al., 2001). Functionally, mutant mice display brain-region-specific defects in both LTD and LTP (Huber et al., 2002; Larson et al., 2005; Li et al., 2002; Nosyreva and Huber, 2006; Wilson and Cox, 2007). However, it is important to realize that such defects may reflect either an acute FMRP function in the adult animal or, just as easily, a transient role of FMRP during development that pre-establishes the ability of synapses to manifest plasticity at maturity. Indeed, synaptic plasticity defects appear to be much more severe during transient developmental windows, with less severe defects at maturity (Huang et al.,2006).

The GS system allows rapid induction of dfmr1 transcription, but an inherent limitation to the approach rests with the protein half-life. We show here that the dFMRP protein appears relatively stable, with a half-life of ∼26 hours. Thus, we can switch `on' dFMRP within a few hours, but the switch `off' is governed by the protein decay profile over the course of several days, limiting temporal resolution. For this reason, we restricted analyses to two intervention windows. First, we employed a 12-hour induction immediately after hatching, carefully determining the induction strength to ensure a match with endogenous dFMRP protein levels. With this treatment,dFMRP levels decrease exponentially with time; the protein was almost undetectable by ∼60 hours post-treatment. Second, a brief 12-hour induction at the terminal endpoint of larval life was executed. Again, we carefully controlled induction specific to this mature time point, to match levels of the introduced protein to dFMRP levels in the control. With this protocol, the animals completely lacked all dFMRP throughout development, with acute protein reintroduction immediately prior to analysis at maturity.

In the early induction paradigm, transient dFMRP expression yields almost complete rescue of dfmr1-null synaptic structural defects, including expansion of the synaptic terminal area, overbranching of synaptic processes,and the formation of excess synaptic boutons. The resolution towards wild-type architecture indicates a primarily early role for dFMRP in molding the NMJ,and suggests that dFMRP-mediated imprinting/patterning of synaptic development allows for appropriate and sustainable synaptic structure. These findings support the conclusion that dFMRP is required early for proper initiation of synaptogenesis and not synaptic maintenance. This conclusion is perhaps not unexpected for a protein that acts as a translational regulator of synaptogenic proteins, which will have their own perdurance at the synapse once properly regulated by early dFMRP function. In addition, however, acute dFMRP induction at maturity provides partial rescue of synaptic architecture defects. This effect is a true rescue, i.e. resolution of overgrowth phenotypes that are fully manifest at the start of the transgene expression window. This finding demonstrates that the established NMJ synapse displays morphological plasticity and can be remodeled. These findings are in agreement with presynaptically mediated remodeling in which synapse retraction occurs,trailing a postsynaptic `footprint' (Eaton et al., 2002; Pielage et al.,2005). Such regression would be requisite in mediating the observed rescue of dfmr1-null overgrowth via bouton destabilization and elimination.

Much work continues to be focused on alleviating FraX symptoms and targeting the causative molecular insults resulting in the disease state. Recent work in the Fmr1 KO mouse has shown rescue, at cellular and behavioral levels, via constitutive reduction in mGluR5(Dolen et al., 2007) and p21-activated kinase (Hayashi et al.,2007). Translation of these exciting new findings into clinical treatments will be better informed with the temporal requirement of FMRP clearly defined. In future work, we will test the efficacy of targeted temporal interventions in both mGluR-mediated signaling upstream of FMRP function, as well as translational consequences downstream of FMRP function.

We are especially grateful to Dr Haig Keshishian (Yale University) for the elav-Gene-Switch line that made this work possible. We thankfully acknowledge the Bloomington Drosophila Stock Center and the Iowa Developmental Studies Hybridoma Bank for genetic lines and antibodies,respectively. We thank Dr Elisabeth Dykens (Vanderbilt Kennedy Center for Research on Human Development) for her role in providing clinical co-mentorship to C.L.G. during this project; members of the Broadie laboratory, especially Dr Charles Tessier, for insightful discussion; Drs Sarah Repicky, Jeffrey Rohrbough and Kevin Haas, and Ms. Ashleigh Long for technical training, suggestions and support with regard to the electrophysiology studies presented. This work was funded by the National Institutes of Health through the NIH Roadmap for Medical Research Training Grant T32 MH075883 to C.L.G. and R01 grant GM54544 to K.B.