Temporal Requirements of the Fragile X Mental Retardation Protein in Modulating Circadian Clock Circuit Synaptic Architecture

Loss of fragile X mental retardation 1 (FMR1) gene function is the most common cause of inherited mental retardation and autism spectrum disorders, characterized by attention disorder, hyperactivity and disruption of circadian activity cycles. Pursuit of effective intervention strategies requires determining when the FMR1 product (FMRP) is required in the regulation of neuronal circuitry controlling these behaviors. In the well-characterized Drosophila disease model, loss of the highly conserved dFMRP causes circadian arrhythmicity and conspicuous abnormalities in the circadian clock circuitry. Here, a novel Sholl Analysis was used to quantify over-elaborated synaptic architecture in dfmr1-null small ventrolateral neurons (sLNvs), a key subset of clock neurons. The transgenic Gene-Switch system was employed to drive conditional neuronal dFMRP expression in the dfmr1-null mutant background in order to dissect temporal requirements within the clock circuit. Introduction of dFMRP during early brain development, including the stages of neurogenesis, neuronal fate specification and early pathfinding, provided no rescue of dfmr1 mutant phenotypes. Similarly, restoring normal dFMRP expression in the adult failed to restore circadian circuit architecture. In sharp contrast, supplying dFMRP during a transient window of very late brain development, wherein synaptogenesis and substantial subsequent synaptic reorganization (e.g. use-dependent pruning) occur, provided strong morphological rescue to reestablish normal sLNvs synaptic arbors. We conclude that dFMRP plays a developmentally restricted role in sculpting synaptic architecture in these neurons that cannot be compensated for by later reintroduction of the protein at maturity.

clearly dependent on the FMR1 gene family across a wide range of species.
The Drosophila circadian clock circuitry is particularly well characterized (Helfrich-Forster, 2005;Chang, 2006;Nitabach and Taghert, 2008). Within this circuit, the small ventrolateral neurons (sLN v s) are a key subset of clock neurons that possess suffi cient circadian pacemaker capacity to modulate morning activity peaks in light:dark cycles and drive rhythmicity in constant darkness (Renn et al., 1999;Grima et al., 2004;Stoleru et al., 2004). These critical clock neurons express the neuropeptide pigment dispersing factor (PDF) (Helfrich-Forster, 1995, 2005, which has long been proposed to serve as the clock output mediating coordination of downstream neurons. Characterization of pdf-null animals reveals impaired light cycle entrainment with advanced evening activity peaks and disrupted anticipation, while free-running behavior in constant darkness shows progressive deterioration of rhythmicity (Renn et al., 1999). Further, disrupted cycling of PDF within the terminal sLN v projections induced by transgenically-mediated electrical hyperexcitability causes altered behavioral rhythms with multiple periods (Nitabach et al., 2006), and work using a membrane-tethered toxin to infl uence neuronal ion channel properties revealed a 4-h phase advance in the PDF peak accumulation in the sLN v s, which coordinately shifts morning anticipatory behavior INTRODUCTION Fragile X Syndrome (FraX) is a common heritable neurological disease manifesting mental retardation, autism, attention defi cit disorder and disrupted activity patterns (Bassell and Warren, 2008;Gatto and Broadie, 2009;Hagerman et al., 2009). FraX patients display characteristic hyperactivity and disordered sleep associated with altered melatonin levels (Elia et al., 2000;Gould et al., 2000;Miano et al., 2008). Similarly, FraX genetic disease models also show dramatically impaired circadian activity profi les. Drosophila mutants lacking their sole fragile X mental retardation 1 (FMR1) family member (dfmr1) display signifi cantly altered motor activity, disrupted sleep and striking circadian rhythm phenotypes, hallmarked by arrhythmicity and erratic locomotion (Dockendorff et al., 2002;Inoue et al., 2002;Morales et al., 2002;Bushey et al., 2009;Gatto and Broadie, 2009). In mammals, a tripartite gene family (FMR1, FXR1 and FXR2) encodes proteins with >60% amino acid identity and high functional domain conservation Zhang et al., 1995). Mice defi cient for either Fmr1 or Fxr2 alone maintain fairly normal rhythmicity, albeit with free-run periodicity shortening . However, combinatorial Fmr1/Fxr2 double null animals are entirely arrhythmic and further lack basic entrainment capacity . Thus, the circadian regulation of behavior is . The PDF neuropeptide facilitates circadian oscillations in individual neurons, while driving synchrony in neuronal subgroups Yoshii et al., 2009).
Null dfmr1 sLN v s exhibit striking architectural defects, including overgrowth, overextension and mistargeting beyond the normal branching and defasiculation point in the protocerebrum, projecting a more expansive terminal synaptic arborization (Dockendorff et al., 2002;Morales et al., 2002;Reeve et al., 2005Reeve et al., , 2008Sekine et al., 2008;Gatto and Broadie, 2009). Conversely, targeted dfmr1 overexpression in sLN v s results in a dramatic collapse of their dorsal synaptic arbor (Reeve et al., 2005), indicating that the levels of the dFMR1 product (dFMRP) protein alone can bidirectionally modulate synaptic complexity. Thus, dFMRP plays a critical role in controlling the synaptic connectivity within the clock circuit that drives circadian activity patterns.
A fundamental need is to determine the critical period(s) of FMRP function. Treatment design, implementation and effi cacy depend upon determining whether FraX is primarily a disease of development or maintained function. FMRP has been proposed to be required very early, during neurogenesis and pathfi nding, later during stages of synaptogenesis and neuronal circuitry refi nement (e.g. activity-dependent synaptic pruning), or at maturity for use-dependent circuit remodeling (Nimchinsky et al., 2001;Galvez and Greenough, 2005;Larson et al., 2005;Yun et al., 2006;Bureau et al., 2008;Gatto and Broadie, 2008;Tessier and Broadie, 2008). Clearly, these potential roles may not be mutually exclusive. To distinguish these possibilities, we have used the conditional, transgenic Gene-Switch (GS) method (Osterwalder et al., 2001;Roman et al., 2001) to neuronally target wild-type dFMRP in null dfmr1 mutants (Gatto and Broadie, 2008). Conditional expression was driven during windows throughout development and at maturity, followed by examination of the resultant sLN v synaptic architecture using a PDF-adapted modifi ed Sholl Analysis. Neither early developmental nor adult-onset dFMRP expression was suffi cient to prevent or reverse, respectively, the sLN v synaptic defects. Surprisingly, inappropriate dFMRP expression during early pupal metamorphosis actually exacerbated dfmr1-null phenotypes. Successful and striking structural resolution of synaptic connectivity defects was obtained only via targeted reintroduction of dFMRP during late pupal brain development. These results reveal a highly specifi c, dFMRPdependent window of sLN v clock neuron development during the late stages of synaptogenesis and/or synaptic refi nement.

DROSOPHILA GENETICS
Drosophila stocks were maintained at 25°C on standard cornmeal/ agar/molasses medium supplemented with yeast. The w 1118 line served as the genetic background control for the null dfmr1 allele (dfmr1 50M ) Gatto and Broadie, 2008). Stocks carrying PDF-Gal4 and UAS-mCD8::GFP, both from the Drosophila Stock Center (Bloomington, IN, USA), as well as UASsynaptotagmin-GFP (Zhang et al., 2002), were introduced into the dfmr1-null background using standard genetic techniques. As previously described (Gatto and Broadie, 2008), recombinant parental lines harboring the null dfmr1 allele  and either a wild-type dfmr1 transgene under UAS control (UAS-9557-3)  or the neuronal-specifi c driver elav-Gene-Switch construct (GSG-301) (Osterwalder et al., 2001) were generated using standard genetic techniques. Accordingly, "eGS" 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, USA) dosing (eGS + RU), the drug was dissolved at 10 mM in 80% ethanol (EtOH) and either mixed with food to the desired concentration for larval (0.5 µg/mL) or adult (30 µg/mL) feeding, or it was used directly for immersion treatment of both larvae and pupae as indicated. For pupal day 1 (P1) induction, late-stage wandering third instar larvae were washed briefl y with EtOH, soaked in RU486 for 2 min, and then transferred to fresh vials for puparium formation. Induction from P2-P4 required an anterior incision in the pupal case with a 30-gauge needle and subsequent submersion in RU486 for 2 min. Pupae were then transferred to humid chambers to prevent desiccation. For vehicle control, equivalent volumes of 80% EtOH were used to identically treat the eGS line (eGS + E).

WESTERN BLOT ANALYSES
The central nervous system, including the brain and the ventral nerve cord, was dissected free from staged and treated larvae in Ca 2+free modifi ed standard saline, while the brain proper was dissected from both pupae and adults (Gatto and Broadie, 2008). Dissected samples were homogenized and boiled in 1X NuPage sample buffer (Invitrogen, Carlsbad, CA, USA) supplemented with 40 mM DTT. The total protein from four brains per sample was loaded onto 4-12% Bis-Tris gels and electrophoresed in NuPage MES Buffer (Invitrogen) for 1 h at 200 V. Transfer to nitrocellulose was carried out for 1 h at 100 V in NuPage transfer buffer (Invitrogen) with 10% MeOH. Processing was completed using the Odyssey near infrared fl uorescence detection system (Li-COR, Lincoln, NE) to enable quantitative Western blot analysis. Antibodies used included: anti-dFMRP (1:2000; 6A15 monoclonal, Sigma), anti-α-tubulin (1:200,000; B512 monoclonal, Sigma), and Alexa-goat-anti-mouse-680 (1:10,000; Invitrogen-Molecular Probes). Raw integrated intensities were calculated for the lower molecular weight band of the dFMRP doublet and for the α-tubulin band. The ratio of dFMRP: α-tubulin normalized for loading and allowed comparison of eGS neuronally induced dFMRP relative to endogenous dFMRP in age matched wild-type (w 1118 ) animals.

PDF-ADAPTED MODIFIED SHOLL ANALYSIS
A PDF-adapted version of the Modifi ed Sholl Analysis (Fernandez et al., 2008) was developed. A traditional Sholl array using a series of 10 µm-spaced concentric rings was centered upon the dorsal horn bifurcation of the sLN v terminal arbor and extended forward for 100 µm with assays conducted at zeitgeber time 2-4, i.e. 2-4 h after lights on. To provide for compatibility with the GS system, this assay examined PDF reactive puncta (≥1 µm in diameter) within each 10 µm concentric ring. Puncta were counted as a refl ective index of spatial synaptic arbor complexity. Values were determined for each concentric ring throughout the arbor in both left and right hemispheres and averaged for n = 1 for each individual animal.

STATISTICS
Statistical analyses were performed using GraphPad InStat 3 (GraphPad Software, Inc., San Diego, CA, USA). Unpaired, nonparametric, two-tailed Mann-Whitney tests were used to compare means of control and dfmr1, as well as means of eGS treated with vehicle or RU486 in the PDF-adapted Modifi ed Sholl Analysis. While unpaired, parametric student's t-tests could have been applied in certain instances, the cumulative consideration of data sets examining each concentric ring in a given treatment paradigm did vary with regard to normality. Thus, it was deemed more appropriate to globally apply the more conservative, nonparametric measures. Signifi cance levels in fi gures are represented as p < 0.05 (*); p < 0.01 (**); p < 0.001 (***). All error bars represent standard error of the mean (SEM).

OVER-ELABORATION OF sLN V SYNAPTIC ARBORS IN dfmr1-null MUTANTS
In the Drosophila FraX model, dFMRP loss results in overgrowth of sLN v synaptic arbors in the dorsal brain (Dockendorff et al., 2002;Morales et al., 2002;Reeve et al., 2005Reeve et al., , 2008Sekine et al., 2008;Gatto and Broadie, 2009). The increased architectural complexity can be readily visualized with expression of the UAS-mCD8::GFP membrane marker transgene driven by the LN v -specifi c PDF-GAL4, comparing control and dfmr1-null mutant ( Figure 1A). Co-immunostaining for the LN v -specifi c neuropeptide PDF reveals a punctate pattern of varicosities distributed throughout the sLN v synaptic arbor interspersed along the axonal branches. Likewise, expression of the UAS-synaptotagmin-GFP transgene, an integral synaptic vesicle marker, specifi cally labels synaptic boutons within the sLN v arbor revealing the same characteristic synaptic overgrowth in dfmr1-null mutants ( Figure 1B). Double-labeling shows an array of synaptotagmin-GFP boutons within the synaptic arbor, overlapping and co-localized with the anti-PDF array. These analyses show that PDF labeling serves as an excellent indicator of synaptic boutons in both control and dfmr1 nulls ( Figure 1B). Co-localization of PDF with synaptotagmin-GFP was seen in 77 ± 3% of puncta in control and 71 ± 2% in dfmr1 mutants (p = 0.13, n = 7 for both). PDF was independently present in 19 ± 3% of control puncta vs. 23 ± 2% in dfmr1 (p = 0.21), and synaptotagmin-GFP was observed alone in 4 ± 1% of control and 6 ± 0.3% of dfmr1 puncta (p = 0.13).
We conclude that examining the spatial distribution of PDF puncta serves as a refl ective index of both the axonal arbor morphological complexity and synaptic bouton distribution in the sLN v s.

GENE-SWITCH CONDITIONAL CONTROL OF dFMRP DURING DEVELOPMENT AND AT MATURITY
To examine the temporal requirement(s) of dFMRP in regulating synapse architecture and to determine whether dFMRP reintroduction could rescue synaptic defects, we exploited the transgenic conditional GS system (Osterwalder et al., 2001;Roman et al., 2001). The GS method employs a hormone-responsive modifi ed UAS-GAL4 system to drive inducible, targeted, tissue-specifi c transgene expression. We controlled a UAS-dfmr1 transgene with the pan-neuronal elav-GS GAL4 by treating with the activating progesterone analog, RU486 (mifepristone), in the dfmr1-null mutant background to examine the infl uence of dFMRP reintroduction on sLN v morphology and synaptic elaboration (Figure 1) (Gatto and Broadie, 2008). In defi ning treatment paradigms, analyses were performed on wild-type control (w 1118 ), dfmr1-null and dfmr1-null animals harboring both the elav-GS GAL4 driver and UAS-dfmr1 transgene (eGS animals, henceforth) treated with RU486 (eGS + RU) or only ethanol vehicle (eGS + ET) as indicated (Figure 2A). We fi rst confi rmed that RU486 induced elav-GS GAL4 does drive dFMRP transgene expression pan-neuronally in the brain, albeit not uniformly (Figure 2A). Close examination likewise confi rmed elav expression in the sLN v s nuclei (Figure 2B), establishing that sLN v s are elav-positive. These fi ndings indicate that the GS method allows control of dFMRP expression in neurons widely throughout the brain, including within the sLN v neurons.
The expression of dFMRP is strongly modulated as a function of age both during development and adult aging, with peak expression at the late stages of brain maturation (Gatto and Broadie, 2008;Tessier and Broadie, 2008;Bushey et al., 2009). Therefore, RU486 treatment windows were assessed at all time points during development and at maturity for both endogenous and induced dFMRP expression levels ( Figure 2C). Progressive interventions were performed to drive dFMRP transgene expression from the earliest stages of development, through intermediate and later stages, culminating with adult-onset expression. The earliest induction was to constitutively feed RU486 throughout larval development (instars 1-3) and then remove animals from treatment. Controlled RU486 dosage was used to match endogenous dFMRP levels, measured in wandering third instar larvae at 94 ± 11% (n = 3) of the corresponding control levels ( Figure 2C). Following removal of the drug, dFMRP level rapidly declined during pupal stages (P1-P4) and was almost entirely absent by eclosion into the adult allowing transient early dFMRP expression.

FIGURE 2 | Gene-Switch conditional induction of dFMRP expression during staged periods throughout development and at maturity. (A)
The transgenic crossing and treatment scheme used to mediate temporal control of dFMRP expression in the dfmr1 50M -null background using the Gene-Switch (GS) system. Homozygous dfmr1-null animals harboring the neuron-specifi c elav-GS driver (GSG-301) and UAS-dfmr1 transgene responder (UAS-9557-3) (eGS) were treated with vehicle (EtOH) alone or RU486 to generate the two matched populations for comparisons. Whole brain imaging (inset, right) illustrates typical patterns of dFMRP immunostaining after adult-onset induction with RU486. (B) Representative images of adult Drosophila brains expressing a UAS-mCD8::GFP (green) transgene under the control of the PDF-GAL4 driver, co-immunostained for elav (red). Low magnifi cation image on the left; higher magnifi cation images on the right. Elav is expressed within the sLN v cell bodies (arrow, left), in a punctate pattern within the neuronal nuclei (right). (C) Representative Western blots illustrating the dFMRP induction achieved using stage-specifi c RU486 treatment paradigms, indicated to the left of each series, assessed at ∼24 h intervals from initiation. Genotypes as indicated: w 1118 (control), homozygous dfmr1 50M null allele (dfmr1) and dfmr1 50M , elav-GSG-301/dfmr1 50M , UAS-9557-3 (eGS). Treatments as indicated: eGS with ethanol vehicle (eGS + ET) and RU486 (eGS + RU). Blots were probed for dFMRP, with α-tubulin employed as a loading control. Sequential RU486 treatments were used during each day of pupal metamorphosis to drive dFMRP expression starting at pupal day 1 (P1), P2, P3 and P4 ( Figure 2C). Induction at P1 was achieved by bath application of RU486 on the late wandering third instar larvae, just prior to puparium formation. This treatment drives dFMRP expression starting in P1 and extending into the early stages of pupal development (Figure 2C). We previously reported a quantifi ed half-life of dFMRP driven by larval RU486 induction of ∼25 h (Gatto and Broadie, 2008). In the pupae, induced dFMRP persists longer, presumably owing to restrictions on drug clearance and metabolism in the pupae, so that pupal RU486 treatments result in broader windows of expression. In each of the later pupal interventions, the pupal case cuticle was cut to introduce RU486 via brief (2 min) bath application for absorption ( Figure 2C). Treatments were performed just prior to the onset of day P2, P3 and P4 to drive dFMRP expression, which was then monitored over subsequent 24 h intervals progressing to adult eclosion.
The fi nal intervention was to drive dFMRP expression only at maturity in an otherwise dfmr1-null mutant condition. Larvae and pupae were reared until eclosion in the absence of RU486, and no dFMRP expression was detectable ( Figure 2C). Following eclosion, adult animals were then fed RU486 to drive dFMRP levels matching wild type after 24 h of treatment. Adult onset expression was maintained for an additional 72 h course to assess the consequences of continued adult dFMRP expression ( Figure 2C).

EARLY DEVELOPMENTAL AND ADULT-ONSET dFMRP EXPRESSION DO NOT ALLEVIATE dfmr1-null DEFECTS
Using the conditional GS system, we systematically investigated when reintroduction of dFMRP could restore the gross sLN v synaptic architecture defects of the dfmr1 null (Figure 1). We examined how each inductive paradigm might modulate synaptic structure, as quantifi ed with our modifi ed Sholl Analysis of synaptic bouton spatial distribution ( Figure 1C). We fi rst examined animals in which normal dFMRP levels were driven throughout larval life, the period of neurogenesis and neural fate specifi cation. This developmentally early intervention provided no detectable restorative effect on the organization of the sLN v s arborization or synaptic bouton disposition as compared to the eGS + ET animals, which phenocopy the dfmr1 null alone ( Figure 3A). When dFMRP is introduced during larval stages only, the resultant adults display PDF puncta within the sLN v arbors that remain unchanged with respect to total number (eGS + ET: 88 ± 6, n = 10 vs. eGS + RU: 85 ± 6, n = 10; p = 0.97) and spatial distribution throughout the Sholl array (Figure 3A, right). We conclude that providing dFMRP transiently during early stages of neural development provides no benefi t for correcting the hallmark synaptic defects caused by the loss of dFMRP as assayed at maturity. We next assayed whether adult-onset dFMRP induction could rescue synaptic defects in dfmr1-null animals ( Figure 3B). This determination is particularly important for the design of FraX intervention strategies, as it addresses the question of whether the synaptic overgrowth characteristic of the disease is reversible at a state of maturity. eGS dfmr1-null animals were raised in the absence of RU486 induction preventing dFMRP expression during development and then switched to an RU486-containing diet at eclosion. eGS animals with sustained induction for the following 3 days showed no detectable reversal of the overgrown sLN v synaptic terminal area, the over-elaborated synaptic branch arborization, or the over-proliferation of synaptic boutons as compared to the eGS + ET animals, which phenocopy the dfmr1 null alone ( Figure 3B). When reintroduced only in the adult, the aged adult animals expressing dFMRP display PDF synaptic puncta within the sLN v s arbors that remain unchanged with respect to total number (eGS + ET: 77 ± 4, n = 15 vs. eGS + RU: 73 ± 4, n = 15; p = 0.63) and spatial distribution throughout the Sholl array (Figure 3B, right). We conclude that providing dFMRP at maturity, following completion of neural development, provides no benefi t for correcting the hallmark synaptic defects caused by the loss of dFMRP.
Despite reintroduction of dFMRP throughout either the larval or adult phase, no palliative effect was observed in the dfmr1-null sLN v synaptic connections. Therefore, we next tested whether phenotypic reversal of these defects would be identifi able upon dFMRP elevation in these cells. We compared two conditions: (1) UAS-mCD8:: GFP alone, and (2) UAS-mCD8::GFP co-expressed with UAS-dfmr1, specifi cally driven in sLN v s ( Figure 3C). dFMRP in the control is expressed at a uniform low level in neurons, including the sLN v clock neurons ( Figure 3C, top row, red), whereas targeted dFMRP, exploiting the PDF-GAL4 driver, elevates expression only in these cells ( Figure 3C, middle row, red, arrows). This targeted, constitutive expression of dFMRP within the sLN v s results in collapse and retraction of the terminal synaptic arbor, evident in the reduction of synaptic area encompassed, loss of synaptic varicosities and complete lack of outward projecting axonal branches (Figure 3C, green). Examination of the PDF distribution within these collapsed terminal arbors revealed that, although the gross anatomy of the synapse was radically altered, discrete PDF-positive synaptic puncta remained apparent ( Figure 3C, bottom row, red). These results suggest that the presynaptic potential of the sLN v s remains intact. In contrast to the changes resulting from constitutive lateral neuron-specifi c dFMRP overexpression, timed pan-neuronal dFMRP introduction with the GS system, either early or late, causes no such change in the synaptic arborization of the dfmr1 null sLN v s, which remain over-extended and over-elaborated. Even over-expression of dFMRP in the adult (Figures 2A,C, after full 72 h RU486 treatment: 246 ± 14% control level, n = 4) failed to cause a reduction in the synaptic arbor of the null mutant, showing that the over-expression effect is likewise restricted to a specifi c developmental window. These results together strongly suggest that GS-induced dFMRP expression has not yet been provided at the appropriate stage.

A CRITICAL WINDOW OF dFMRP REQUIREMENT DURING LATE BRAIN DEVELOPMENT
The above results suggest that the requirement for dFMRP function in synaptic patterning likely occurs in the period of late brain development during pupal metamorphosis. This timing is consistent with elevated endogenous dFMRP expression during the late pupal period, and the role of dFMRP during this period in activity-dependent synaptic refi nement (Tessier and Broadie, 2008). As the pupal period lasts 4 days at 25°C (P1-P4), we subdivided this period by inducing dFMRP at the start of each consecutive day in turn in otherwise dfmr1-null mutant animals ( Figure 2B). The consequences on sLN v clock neuron synaptic arborizations were then evaluated in mature adults (d1-2) in each case. The results of these studies are shown in

www.frontiersin.org
August 2009 | Volume 3 | Article 8 | 9 Gatto and Broadie Synaptic structuring transiently requires FMRP We began with RU486 drug delivery to drive dFMRP induction at the earliest stage of pupal development, with induction at pupal day 1 (P1; Figure 4A). This early phase dFMRP expression provided no correction of the dfmr1-null sLN v synapses compared to the eGS + ET animals, which phenocopy the dfmr1 null alone. Indeed, this inappropriate early induction exacerbated the mutant phenotype, resulting in an elevated number of PDF synaptic boutons redistributed within 30% of the sLN v axonal arbor based on the Sholl Analysis (Figure 4A, right). While the total number of boutons was unchanged (p = 0.32; n = 11 eGS + ET, n = 14 eGS + RU), the spatial distribution of boutons was signifi cantly altered, with an increased number of boutons in the 40 µm concentric ring (eGS + ET: 14 ± 1 vs. eGS + RU: 18 ± 1; p = 0.025), 70 µm ring (eGS + ET: 3 ± 0.5 vs. eGS + RU: 7 ± 1; p = 0.037), and 80 µm ring (eGS + ET: 2 ± 0.6 vs. eGS + RU: 5 ± 1; p = 0.035). The overall frequency of boutons in the more distal domains was shifted from 82% to 93% at 70 µm and from 63% to 93% at 80 µm in eGS + ET compared to eGS + RU (Figure 4A, right). Similarly, induction at P2 provided no improvement of the dfmr1-null sLN v architecture compared to the eGS + ET animals ( Figure 4B). P2 induction of dFMRP provided no signifi cant infl uence upon the synaptic arborization, with no change in the total number of PDF boutons (eGS + ET: 81 ± 4, n = 15 vs. eGS + RU: 77 ± 7, n = 15; p = 0.27) or their spatial distribution (Figure 4B, right). We conclude that induction of dFMRP during the fi rst half of pupal development does not restore normal synaptic connectivity in the null mutant animals.

DISCUSSION
Loss of FMRP compromises the synaptic connectivity of the circadian clock circuit and impairs the behavioral output of rhythmic circadian activity patterns, making this system ideal for investigating the temporal requirements for FMRP function (Elia et al., 2000;Gould et al., 2000;Dockendorff et al., 2002;Inoue et al., 2002;Morales et al., 2002;Reeve et al., 2005Reeve et al., , 2008Miano et al., 2008;Sekine et al., 2008;Zhang et al., 2008;Bushey et al., 2009). Using our well-established Drosophila FraX model Bolduc et al., 2008;Gatto and Broadie, 2008;Pan et al., 2008;Tessier and Broadie, 2008;Repicky and Broadie, 2009) and the wellcharacterized Drosophila clock circuitry (Helfrich-Forster, 2005;Chang, 2006;Nitabach and Taghert, 2008), this study was aimed at identifying the critical period(s) of dFMRP function in the synaptic organization of clock neurons. The GS method provides tight temporal control of transgene expression specifi cally within neurons (Osterwalder et al., 2001;Roman et al., 2001), allowing us to dissect the exact timing of the dFMRP requirement. We have identifi ed a single responsive time window, pupal days 3-4, during which reintroduction of dFMRP provides rescue of the arbor overgrowth and excessive synapse establishment characterizing the dfmr1-null mutant. Introduction of dFMRP either before or after this period provides no benefi t. We conclude, therefore, that dFMRP function in sculpting synaptic connections is locked into this transient developmental time window. As these studies were conducted with pan-neuronal expression, we cannot yet distinguish between a cell autonomous function of dFMRP in sLN v clock neurons compared to potential roles of dFMRP in neighboring neurons.
We have previously investigated the temporal requirements of dFMRP at the peripheral glutamatergic neuromuscular junction (NMJ) (Gatto and Broadie, 2008). In dfmr1-null mutants, this synaptic arbor is over-elaborated with expanded synaptic area, excess synaptic branching and supernumerary synaptic boutons , thus closely resembling the phenotype in the central brain clock neurons shown here. Employing GS inductive paradigms to drive dFMRP in the dfmr1-null mutant background, these studies similarly demonstrate a temporally restricted role for dFMRP in sculpting synaptic architecture (Gatto and Broadie, 2008). Re-introduction of dFMRP only during the period of normal NMJ synaptogenesis enabled the correct development and maintenance of standard, sustainable terminal structure. In contrast, dFMRP introduction at maturity provided very little rescue of the mutant overgrown synaptic defects, with only weak, partial reduction in the number of synaptic boutons (Gatto and Broadie, 2008). The partial rescue achieved likely refl ects the fact that synaptogenesis continues throughout the larval period in Drosophila, and therefore development and maturity are not as well separated as in the brain clock circuit. Overall, these previous fi ndings at the NMJ strongly support our current results of a transient development requirement for dFMRP within the sLN v clock neurons.
Interestingly, a host of FraX-associated phenotypes are known to be developmentally transient, also suggesting temporally-restricted FMRP function. In the Drosophila FraX model, dfmr1-null mushroom body neurons display heightened axonal and dendritic complexity, with excessive synaptic connectivity (Pan et al., 2004). These defects become evident only very late in pupal brain development, with the appearance of abnormally long axonal branches and associated excessive synaptic processes (Tessier and Broadie, 2008). Upon initial use following eclosion, activity-dependent synaptic pruning occurs in wild-type but completely fails in dfmr1null mutant neurons. Following this normal dFMRP-dependent developmental window, delayed and inappropriate pruning occurs in the absence of dFMRP (Tessier and Broadie, 2008). This refi nement process takes days in Drosophila, but occurs over the course of weeks in mice. Similar to the Drosophila fi ndings, in Fmr1-null mouse layer V somatosensory barrel cortex, the increase in synaptic dendrite spine length and density manifests during postnatal week 1, attenuates by week 2 and is undetectable by week 4 (Nimchinsky et al., 2001;Galvez and Greenough, 2005). Likewise, Fmr1-null ascending projections connecting layers III and IV display aberrant transmission strength, experience-dependent plasticity and axonal arborization at postnatal week 2, which also resolves by postnatal week 3-4 (Bureau et al., 2008). These studies suggest a transient FMRP role in synaptic development and refi nement. It should be noted that differences could yet arise between the Drosophila and murine FraX models with regard to the temporal requirements for FMRP. Mice harbor the two Fmr1 paralogs, Fxr1 and Fxr2 Zhang et al., 1995), which could infl uence the effects of Fmr1 deletion during development. However, the protracted postnatal period required for circuit refi nement in mammals needs to be carefully considered to defi ne developmental vs. adult FMRP requirement. We eagerly await the establishment of similar mouse conditional Fmr1 expression models to provide further insight into the etiology of FraX.
In the Drosophila FraX model, the P3-P4 critical period definition for sLN v synaptic structuring is entirely consistent with our recent fi nding that dFMRP levels are transiently elevated during this same period (Tessier and Broadie, 2008), with subsequent rapid decrement following maturation (Tessier and Broadie, 2008;Bushey et al., 2009). A similar transient period of FMRP elevation is reported in rodents, likewise corresponding to the restricted period of synaptogenesis and synaptic refi nement (Khandjian et al., 1995;Singh et al., 2007;Singh and Prasad, 2008). During the corresponding period, dFMRP has been shown to regulate a similar stage-dependent mechanism of activity-dependent axonal pruning and synapse elimination in the Drosophila mushroom body (Tessier and Broadie, 2008). Assuming a conserved mechanism within the clock circuit, this role would explain the temporally-restricted dFMRP function in establishing appropriate sLN v synaptic connections. Consistent with this transient critical window for dFMRP function, assays to ameliorate dfmr1-null clock defects by decreased metabotropic glutamate receptor (mGluR) signaling, via treatment with mGluR antagonists such as 2-methyl-6-phenylethynyl-pyridine (MPEP), fail to resolve circadian impairments with treatment either during larval development or in the adult (McBride et al., 2005). We suggest that such previous pharmacological treatment attempts have inadvertently missed the therapeutic target window of late brain development.
The PDF cells have been described as "GABA-responsive wake-promoting" neurons in the clock circuit (Parisky et al., 2008). Suppression of neuronal activity in these cells increases sleep, while coordinately increasing LN v excitability decreases sleep. sLN v s express the Rdl GABA A receptor, and sLN v -specifi c Rdl knockdown impairs sleep while Rdl over-expression yields more rapid sleep onset and enhanced sleep (Parisky et al., 2008). These fi ndings are particularly relevant to the Drosophila FraX model, as it has been recently reported that dfmr1 nulls display increased sleep, while dFMRP over-expression decreases sleep (Bushey et al., 2009). However, dfmr1 nulls have been shown to under express all known Drosophila GABA A receptor subunits (D'Hulst et al., 2006). Consistently, the Fmr1 mutant mouse has reduced GABA A receptors (D'Hulst et al., 2006;Gantois et al., 2006), associated with both structural and functional alterations of the GABAergic circuits in the neocortex and hippocampus, respectively (D'Antuono et al., 2003;Selby et al., 2007). These changes may correlate with recent accounts of neuronal hyperexcitability in both Drosophila and mouse FraX models (Gibson et al., 2008;Repicky and Broadie, 2009). A small molecule screen for rescue of dfmr1-mutant phenotypes identifi ed GABA, nipecotic acid (GABA reuptake inhibitor) and creatinine (suspected GABA A receptor activator), again implicating GABAergic signaling (Chang et al., 2008). In dfmr1 nulls, GABA treatment specifi cally relieved misregulation of the MAP1B homolog, Futsch, mushroom body β-lobe midline crossing and courtship defects. Together, these data suggest that manipulation of GABA signaling