Fragile X Mental Retardation Protein Requirements in Activity-Dependent Critical Period Neural Circuit Refinement

Summary

Activity-dependent synaptic remodeling occurs during early-use critical periods, when naïve juveniles experience sensory input. Fragile X Mental Retardation Protein (FMRP) sculpts synaptic refinement in an activity sensor mechanism based on sensory cues, with FMRP loss causing the most common heritable autism spectrum disorder (ASD), Fragile X syndrome (FXS). In the well-mapped Drosophila olfactory circuitry, projection neurons (PN) relay peripheral sensory information to the central brain mushroom body (MB) learning/memory center. FMRP-null PNs reduce synaptic branching and enlarge boutons, with ultrastructural and synaptic reconstitution MB connectivity defects. Critical period activity modulation via odorant stimuli, optogenetics and transgenic tetanus toxin neurotransmission block show elevated PN activity phenocopies FMRP-null defects, while PN silencing causes opposing changes. FMRP-null PNs lose activity-dependent synaptic modulation, with impairments restricted to the critical period. We conclude FMRP is absolutely required for experience-dependent changes in synaptic connectivity during the developmental critical period of neural circuit optimization for sensory input.

eTOC

Doll et al. discover activity-dependent neural circuit refinement restricted to an early-use sensory critical period requires FMRP, lost in Fragile X syndrome. Developmental hyperexcitation phenocopies synaptic connectivity defects. Sensory-, optogenetic- and neurotransmission-dependent remodeling during the critical period all require FMRP.

Introduction

Fragile X syndrome (FXS) is the most common heritable autism spectrum disorder (ASD)[1]. Leading hypotheses of FXS pathogenesis emphasize faulty neural circuit refinement during early postnatal critical periods of neurodevelopment [2]. In addition to low IQ and learning disabilities, FXS patients exhibit hypersensitivity to multiple sensory modalities [1], and up to 20% comorbidity for childhood epilepsy [3]. Fragile X Mental Retardation Protein (FMRP) expression delineates early-use critical periods of neural circuit optimization in FXS disease models [4,5]. FMRP is proposed to function within the activity sensor mechanism, mediating activity-dependent synaptic remodeling based on early sensory experience [6,7]. Here, we test this hypothesis in the Drosophila FXS model [8], which closely replicates human disease state symptoms including learning/memory deficits, hyperactivity, interrupted sleep and impaired social interaction [9,10]. This disease model has provided key mechanistic insights into FXS neuropathology through application of a sophisticated genetic toolkit within well-mapped circuitry to unlock FMRP activity-dependent roles in neural refinement.

Neural circuit remodeling occurs in critical periods of early sensory input brain development, when neurons are primed to alter synaptic architecture, connectivity and function in response to initial experience [11]. The Drosophila brain olfactory learning and memory circuit provides an excellent model of experience-dependent refinement during a defined critical period immediately following eclosion, amenable to sensory conditioning paradigms [12–14] and targeted optogenetic manipulation [5,15]. We have discovered neuron type-specific requirements for FMRP in this circuit, with FMRP loss causing excessive activity in excitatory (E) neurons and reduced function in inhibitory (I) neurons, supporting the E/I imbalance theory of FXS and related ASD states [16,17]. Importantly, E/I phenotypes are phenocopied by targeted optogenetic modulation of activity states during the critical period, suggesting tight temporal limits to activity- and FMRP-dependent mechanisms [5]. We have tested FMRP requirements within the antennal lobe medial projection neuron 2 (mPN2)[18] in regard to dendritic arbor development [15] and functional activity-dependent calcium signaling [19]. We now dissect sensory input-, activity- and FMRP-dependent remodeling of mPN2 synaptic input to the Drosophila brain learning/memory center during the critical period.

In this study, we discover synaptic remodeling and activity-dependent refinement restricted to the early-use critical period in the central brain mushroom body (MB) calyx, where mPN2 has presynaptic input onto Kenyon cells (KCs)[20]. In dfmr1 null animals, we find synaptic dysmorphia with both confocal and transmission electron microscopy imaging of synaptic microglomeruli, with reduced mPN2 branching, malformed boutons and reduced connectivity onto postsynaptic KCs. Using GFP reconstitution across synaptic partner (GRASP) studies, we find fewer mPN2 contacts in the dfmr1 mutants. We discover critical period olfactory experience causes dramatic changes in mPN2 synaptic architecture, phenocopying FXS disease model defects. Using cell-targeted optogenetics to either stimulate or suppress mPN2 critical period activity, we find bidirectional activity-dependent control of branching and synaptic bouton refinements both absolutely require FMRP. Using conditional neurotransmission blockade with transgenic tetanus toxin, we reveal a striking expansion of mPN2 innervation, showing synaptic function restricts FMRP-dependent synaptic remodeling. This combination of temporal- and cell-targeted transgenic tools allows systematic dissection of activity-dependent critical period mechanisms of neural circuit refinement in response to initial sensory experience at an entirely new level. We conclude that FMRP is a required mediator of experience-driven neural circuit optimization.

Results

FMRP mediates critical period refinement of mPN2 synaptic innervation in the MB calyx

The antennal lobe medial projection neuron 2 (AL-mPN2) somata are located in the ventrolateral portion of the Drosophila brain subesophogeal ganglion (SEG), with dendritic arbors in the ventrolateral 1 (VL1) glomerulus of the antennal lobe (AL) and axons extending to both mushroom body (MB) and lateral horn (LH; Fig. 1A, B)[15,18]. FlyLight transgenic driver R65G01-Gal4 specifically targets this bilateral pair of neurons (Fig. 1B)[21], providing single-cell resolution for synaptic connectivity characterization and targeted activity manipulations. Ventral branches emanate from the mediolateral mPN2 axon within the MB calyx, providing presynaptic input into Kenyon cell (KC) microglomeruli (Fig. 1A, B). Wildtype control mPN2 neurons generally develop two long and two short ventral branches, which are populated by synaptic boutons in the form of globular varicosities (Fig. 1A, B). These presynaptic boutons contain the core active zone scaffold Bruchpilot (Brp; Fig. 1B)[22] within the MB calyx microglomeruli, whereas the apposing KCs contain the core postsynaptic scaffold Discs Large (Dlg; Fig. 1B)[23]. Use of R65G01-Gal4 to drive plasma membrane UAS-mCD8::GFP labels single mPN2 synaptic architecture within the MB calyx (Fig. 1B). We quantified the length of ventral synaptic branches and the cross-sectional area of synaptic boutons to assess the impact of FMRP on critical period synaptic development.

Figure 1. FMRP controls mPN2 innervation of MB calyx during the critical period.

(A) Schematic of AL-mPN2 connectivity within the olfactory learning and memory circuit. mPN2 receives Ir75 olfactory sensory neuron dendritic input in the antennal lobe VL1 glomerulus and projects an axon to 1) MB calyx and 2) lateral horn. In the calyx (inset, right), mPN2 projects ventral branches innervating Kenyon cells within microglomeruli. AL-mPN2, antennal lobe medial projection neuron 2; MB, mushroom body; LH, lateral horn; iACT, inner antennal cerebral tract; Ir75d, ionotropic receptor 75d. (B) mPN2 dendrites (top, left) and Ir75d axons (bottom, left) connect in the VL1 glomerulus. mPN2 axons in the MB calyx, showing presynaptic Brp (top, right) and postsynaptic Dlg (bottom, right) forming microglomeruli. (C) Genetic background control (w¹¹¹⁸) and dfmr1 null (dfmr1^50M) mPN2 innervation of MB calyx, showing synaptic branches (brackets) and boutons (arrows) with R65G01-Gal4>UAS-mCD8::GFP expression. (D) Quantification (mean ± standard error) of synaptic branch length (left) and bouton area (right) in both genotypes over developmental time at 0, 1 and 7 days post-eclosion (dpe). Significance determined from two-tailed Mann-Whitney tests: p***<0.001.

We focused on an early-use critical period of synaptic refinement during 0–1 days post-eclosion (dpe)[13,15], just prior to the striking decrease in FMRP expression [4,5]. At this timepoint, dfmr1 null mPN2 neurons display reduced branching and expanded synaptic bouton area compared to genetic background (w¹¹¹⁸)(Fig. 1C). At 0 dpe, control branches average 7.26±0.37μm (n=124 branches), compared to 4.83±0.16μm in dfmr1 mutants (n=132 branches), a ~33% decrease in the FXS condition (p***<0.0001). This defect is also evident at 1 dpe, with control branches averaging 7.53±0.34μm (n=130), compared to 4.99±0.19μm in dfmr1 mutants (n=124, p***<0.0001; Fig. 1D). In addition to reduced branching, dfmr1 null neurons have enlarged synaptic boutons. At 0 dpe, the average maximum cross sectional area in controls is 2.04±0.04μm² (n=241 boutons) compared to 2.38±0.05μm² in mutants (n=264, p***<0.0001; Fig. 1D). This defect is more prominent at 1 dpe, with control boutons averaging 2.14±0.04μm² (n=301) versus 2.68±0.04μm² in mutants (n=267, p***<0.0001; Fig. 1D). Importantly, both phenotypes are restricted to the critical period, as mature (7 dpe) branching is indistinguishable between dfmr1 nulls (7.27±0.35μm, n=127 branches) and controls (7.07±0.42μm, n=81; p=0.72), and bouton area likewise does not differ between genotypes (control, 2.47±0.05μm², n=169 boutons; dfmr1, 2.37±0.04μm², n=266 boutons, p=0.06; Fig. 1D). These results demonstrate a restricted temporal requirement for FMRP in mPN2 critical period synaptic refinement within the MB calyx.

FMRP required for critical period synaptic molecular maturation in the MB calyx

We next investigated FMRP-dependent synaptic maturation during the critical period by assaying the molecular composition of mPN2-KC synapses (Fig. 1A, B). We assessed presynaptic maturation by examining expression of the active zone scaffold Brp [22] in staged animals during the critical period and at maturity. Genetic background control (w¹¹¹⁸) and dfmr1 null mutants (dfmr1^50M) expressing R65G01-Gal4>UAS-mCD8::GFP were labeled with Brp^nc82 antibody [22], with expression levels quantified inside individual mPN2 boutons in the MB calyx. FMRP strikingly regulates critical period presynaptic molecular assembly, as distinct Brp punctae populate each mPN2 presynaptic varicosity in all control boutons, whereas dfmr1 null mutant boutons show reduced, diffuse Brp expression with fewer punctae (Fig. 2A, B). Every control bouton shows distinct Brp domains, whereas mutant boutons clearly have fewer, less defined areas, with greatly reduced overall Brp expression. At 0 dpe, null dfmr1 mPN2 boutons display a ~50% reduction in Brp expression (n=75 boutons) compared to controls (n=80; p***<0.0001; Fig. 2A–C). At 1 dpe, mPN2 boutons manifest a persistent Brp reduction, with dfmr1 nulls showing a ~25% deficit (n=105) compared to background controls (n=85; p***<0.0001; Fig. 2C). However, this impaired presynaptic maturation is strictly limited to the early-use critical period, as mature mPN2 boutons (7 dpe) express indistinguishable levels of Brp in control boutons (n=55) compared to dfmr1 null boutons (n=60; p=0.82; Fig. 2C). Thus, there is a striking defect in presynaptic molecular maturation in the absence of FMRP that is restricted to the early-use critical period.

Figure 2. FMRP promotes mPN2 synaptic differentiation during the critical period.

(A) Genetic background (w¹¹¹⁸) control mPN2 synaptic boutons at 1 day post-eclosion, labeled by R65G01-Gal4>UAS-mCD8::GFP (green) and the presynaptic marker Brp (red). (B) Null dfmr1 (dfmr1^50M) mutant mPN2 boutons display a developmental delay in expression of Brp relative to genetic controls. (C) Brp expression within individual mPN2 axonal boutons over developmental time (mean ± standard error). Five boutons were selected from each neuron per developmental time point. Expression levels normalized to genetic controls. Significance determined through two-tailed unpaired t-test (0 dpe) or two-tailed Mann-Whitney test (1 and 7 dpe). p***<0.001.

FMRP regulates synaptic connectivity only during the critical period

The critical period specific defects in mPN2 synaptic branching, bouton formation and presynaptic assembly led us to next directly investigate synaptic connectivity with MB Kenyon cells. We used the GFP Reconstitution Across Synaptic Partners (GRASP) paradigm [24,25], employing together a postsynaptic KC-specific “bait” construct (mb247spGFP11; UAS-spGFP1-10) crossed to R65G01-Gal4 to drive expression of the UAS-spGFP1-10 fragment in presynaptic mPN2 (Fig. 3). The mb247-expressing KCs produce one part of the GFP molecule, with the rest generated in mPN2 neurons, allowing GFP reconstitution only upon synaptic contact (Fig. 3). Both wildtype control and dfmr1 null crosses were examined during the critical period (0–1 dpe) and at maturity (7 dpe), using an anti-GFP antibody specifically engineered to recognize only the entire, intact GFP molecule. In line with the experimental design, we find the GRASP signal strictly confined within the MB calyx, in areas of direct synaptic contact between mPN2 and KCs (Fig. 3A–C). As expected, there is little or no detectable GRASP signal in the adjacent lateral horn (LH) innervated by the exact same mPN2. We examined the number and area of mPN2-KC GRASP contacts within the MB calyx as a function of development time (Fig. 3).

Figure 3. FMRP regulates mPN2-KC synaptic connectivity during critical period.

(A) GFP reconstitution across synaptic partners (GRASP) between control (w¹¹¹⁸) mPN2 neuron (presynaptic axon, red) and mb247-expressing Kenyon cells in MB calyx. GFP (green) is restricted to MB calyx (dashed line, left) and not evident in the adjacent lateral horn (LH). Right: High magnification reveals GRASP contact points (arrows) distributed through innervating boutons. (B) GRASP analysis in dfmr1 null (dfmr1^50M) animals. GRASP contacts remain confined to the MB calyx (left panel), yet mPN2-KC contacts occur clustered proximally in axon (bar, right panel), are reduced, and do not appear to occur in all boutons (arrowhead). (C) Schematic showing GRASP experimental design: KCs express spGFP11 (mb247>spGFP11) and mPN2 expresses spGFP1-10 (R65G01-Gal4>UAS-spGFP1-10). Reconstitution of intact GFP molecule (green signal) occurs only with immediate contact proximity in the calyx. (D) Quantification (mean ± standard error) of total GRASP contact number per calyx (left) and area (right) in control and dfmr1 at 0, 1 and 7 days post-eclosion (dpe). Significance determined through two-tailed unpaired t-test or two-tailed Mann-Whitney tests: p***<0.001.

In controls, the GRASP signal localizes to mPN2 boutons along ventral branches (Fig. 3A, arrows), and is restricted to contact points between presynaptic mPN2 and postsynaptic KCs (green), but does not extend into the mPN2 axon (red). In dfmr1 nulls, GRASP contacts are more spatially restricted, proximal to the distal axon (Fig. 3B, bar), and some mutant boutons completely lack GRASP contacts (arrowhead). At 0 dpe, there is a significant reduction in mPN2-KC GRASP contacts in dfmr1 mutants (control, 11.06±0.76, n=18 calyces; dfmr1, 6.41±0.27, n=34), a >40% reduction (p<0.0001, t-test; Fig. 3D). At 1 dpe, there is a persistent loss of synaptic contacts in the absence of FMRP (control, 11.35±0.47, n=20; dfmr1, 7.87±0.32, n=30; p<0.0001, t-test). Despite a loss of contact points, the size of GRASP contacts is dramatically increased in dfmr1 mutants. At 0 dpe, control contacts average 1.98±0.04μm² in cross-sectional area (n=199 contacts) compared to 2.46±0.05μm² in dfmr1 nulls (n=218; p<0.0001, t-test; Fig. 3D). A similar defect persists at 1 dpe, with controls averaging 2.15±0.04μm² (n=226) versus 2.62±0.05μm² in dfmr1 nulls (n=236 contacts, p<0.0001, M-W test; Fig. 3D). Importantly, these striking GRASP defects are restricted to the early-use critical period. Both the number (control, 10.8±0.45, n=25 calyces; dfmr1, 11.04±0.43, n=26; p=0.70, t-test) and size (control, 2.34±0.04μm², n=270 contacts; dfmr1, 2.27±0.04μm², n=287; p=0.42, M-W test) of GRASP contacts are indistinguishable between genotypes at maturity (7 dpe; Fig. 3D). These results reveal a core role for FMRP in synaptic connectivity refinement, with dfmr1 null mPN2 neurons developing fewer and larger connections with downstream KCs specifically during the critical period.

FMRP limits synaptic bouton growth and promotes synaptic active zone formation

The dysmorphic synaptic contacts between mPN2 neurons and MB KCs in dfmr1 mutants led us to examine synapses at the ultrastructural level (Fig. 4). The MB calyx neuropil is comprised of an outer fibrous layer and a central glomerular layer forming a “fan” shape neuropil (Fig. 4A). The glomerular region is innervated by large PN synaptic boutons arising from axons of the inner antennocerebral tract (iACT) just ventral to the calyx (Fig. 4A)[26]. PN boutons are encased in numerous Kenyon cell dendritic profiles, occasionally interspersed with extrinsic neuron boutons that are much smaller than PN boutons (Fig. 4A)[26]. Using ultrathin coronal sections with transmission electron microscopy, we compared MB calyx architecture in genetic background control (w¹¹¹⁸) and dfmr1 nulls (dfmr1^50M) specifically during the critical period immediately following eclosion (0 dpe). Figure 4A shows the MB calyx architecture, including the prominent inner antennocerebral tract (iACT, yellow) containing the mPN2 axons, large KC somata that ring the calyx (red) and distinctive PN synaptic boutons within the calyx (green). Even at the lowest magnification, PN boutons are clearly enlarged in dfmr1 null mutants (Fig. 4B). At higher magnification, we quantified both PN synaptic bouton area and active zone number, as denoted by the classic t-bar morphology (Fig. 4c)[27].

Figure 4. FMRP limits synaptic bouton growth and promotes synapse formation.

(A) MB calyx transmission electron micrographs of genetic background control (w¹¹¹⁸) and dfmr1 null mutant (dfmr1^50M) animals, pseudocolored to show cellular architecture: MB calyx neuropil (green), Kenyon cell bodies (red), and antennocerebral tract (ACT, yellow) carrying mPN2 axons. (B) Higher magnification micrographs showing PN-KC synaptic contacts within the MB calyx in control (left) and dfmr1 null (right). Individual synaptic boutons containing synaptic vesicle (SV) pools, t-bar active zones (arrows) and mitochondria (M). (C) High magnification images showing examples of single t-bar active zones in the control (top) and dfmr1 null mutant (bottom). (D) Quantification of PN bouton area (left) and T-bar active zone number per unit area (right), comparing control and dfmr1 null mutants. All micrographs taken from 0 day post-eclosion brains. Significance determined through two-tailed Mann-Whitney tests: p***<0.001.

FMRP strikingly regulates the size of PN synaptic boutons and the distribution of presynaptic active zones (Fig. 4A, B). PN boutons present as large synaptic vesicle (SV) filled varicosities, characterized by a high abundance of large mitochondria, and multiple presynaptic t-bar active zones (Fig. 4B, C). The most obvious phenotype in dfmr1 null calyces is a prominent increase in PN synaptic bouton size (Fig. 4A, B). The average area of dfmr1 null boutons is 21.79±1.02μm² (n=93 boutons) compared to control boutons of 14.02±0.68μm² (n=89), a highly significant 35% increase (p<0.0001; Fig. 4D) that is closely consistent with results from above confocal microscopy assays (Fig. 1). High magnification electron micrographs also show that synaptic active zone number and distribution are dramatically reduced in dfmr1 null boutons (Fig. 4B, C). The mutants contain fewer synaptic t-bars per synaptic bouton area (0.07±0.01, n=93 boutons) compared to matched controls (0.14±0.01, n=89), a ~50% reduction in the absence of FMRP (p<0.0001; Fig. 4D, right). Taken together, electron microscopy analyses show FMRP acts to restrict synaptic bouton size and promote active zone density, providing direct evidence of reduced synaptic connectivity between PNs and KCs within the MB calyx during the critical period.

FMRP required for critical period olfactory sensory experience remodeling of synapses

Critical period synaptic refinement entails experience- and activity-dependent processes, and FMRP is hypothesized to be an integral component of these remodeling mechanisms [28]. We therefore first employed olfactory exposure to test the impact of sensory experience on mPN2 synaptic connectivity during the critical period compared to maturity. Owing to recent work fine-mapping the olfactory circuitry [29,30], we were able to define a specific odorant (pyrrolidine) for mPN2 studies. The mPN2 dendritic arbor lies within the VL1 glomerulus (Fig. 1A)[15,18], and the ionotropic receptor 75d (Ir75d)-expressing olfactory sensory neurons (OSNs) in antennal coeloconic sensilla ac1, ac2 and ac4 provide presynaptic input to VL1 (Fig. 1B); these OSNs demonstrate a strong, highly specific response to pyrrolidine [29]. To examine olfactory experience-dependent refinement of mPN2 connectivity in the MB calyx, staged control and dfmr1 null animals expressing R65G01-Gal4>UAS-mCD8::GFP were exposed for 24 hours to vehicle alone (mineral oil) or pyrrolidine (1:100 in mineral oil), followed by quantification of mPN2 synaptic branching and bouton formation in the calyx (Fig. 5).

Figure 5. FMRP required for critical period sensory remodeling of mPN2 synapses.

(A) mPN2 innervation of the MB calyx during the 1 dpe critical period for w¹¹¹⁸ control (left) and dfmr1 null (right). Arrows indicate synaptic boutons in the two genotypes. Images show R65G01-Gal4>UAS-mCD8::GFP expression in inverted black-and-white to better reveal synaptic architecture. (B) Comparable mPN2 images after 24-hours of pyrrolidine odor exposure during the early critical period (0–1 dpe) in control (left) and dfmr1 null (right) animals. The dramatically shortened control synaptic branches often possess foot-like protrusions (arrowheads). (C) Quantification of synaptic branch length and bouton area, with and without 24-hour pyrrolidine odor exposure. Left panels: critical period at 1 dpe comparing control (no odor) and odor exposed (0–1 dpe) w¹¹¹⁸ and dfmr1 null. Right panels: maturity at 7 dpe comparing control (no odor) and odor exposed (6–7 dpe) for both genotypes. Significance determined from Dunn’s multiple comparisons tests: p*<0.05, p***<0.001. See also Figures S1 and S2.

Sensory exposure to pyrrolidine causes striking FMRP-dependent changes in mPN2 connectivity during the critical period (0–1 dpe), but not at maturity (7 dpe; Fig. 5). In the genetic background control (w¹¹¹⁸), 0–1 dpe pyrrolidine exposure causes a ~35% decrease in branching (w¹¹¹⁸ control, 8.29±0.27μm, n=180 branches; w¹¹¹⁸ + odor, 5.42±0.18μm, n=196; p***<0.0001, Dunn’s test; Fig. 5A–C), an effect that phenocopies critical period differences between control and dfmr1 null mPN2 neurons (Fig. 1). The boutons of odor-exposed mPN2 neurons often display foot-like protuberances (Fig. 5B, arrowheads) and lack the characteristic morphology of wildtype boutons. In contrast, dfmr1 null neurons do not demonstrate any detectable sensory experience-dependent changes (Fig. 5A, B). In quantifying synaptic branching, vehicle control (4.8±0.21μm, n=96 branches) and odor-exposed (4.83±0.21μm, n=82) dfmr1 null neurons are indistinguishable (p>0.99, Dunn’s test, Fig. 5C). Pyrrolidine exposure also increases mPN2 synaptic bouton area in wildtype (control, 2.03±0.03μm², n=447 boutons; odor, 2.16±0.03μm², n=459, p=0.02, Dunn’s test), but has no impact in dfmr1 null mutants (control, 2.46±0.05μm², n=211; odor, 2.52±0.05μm², n=183; p>0.99 Dunn’s test; Fig. 5A–C). Moreover, pyrrolidine exposure from 0–1 dpe in w¹¹¹⁸ causes a ~20% decrease in Brp expression in boutons compared to unexposed controls (n=135 boutons each; p***<0.0001, M-W test), whereas dfmr1 null boutons showed no odor-induced changes in Brp (n_control=40, n_odor=50, p=0.83, M-W test; Fig. S1).

At maturity (7 dpe; Fig 5C), olfactory exposure has no affect on control branching (vehicle, 8.04±0.32μm, 124 branches; odor, 8.67±0.39μm, 145; p>0.99, Dunn’s test) or synaptic bouton size (vehicle control, 2.07±0.03μm², n=285 boutons; odor, 2.05±0.03 μm², n=352; p>0.99, Dunn’s test; Fig. 5C). Likewise, pyrrolidine has no effect on mPN2 bouton area in dfmr1 null mutants (vehicle, 2.4±0.05μm², n=223 boutons; odor, 2.51±0.03μm², n=187; p=0.79, Dunn’s test), although there is a significant decrease in branch length (vehicle, 6.01±0.24μm, 96 branches; odor, 4.81±0.25μm, 77; p=0.013, Dunn’s test; Fig. 5C, right). We noted morphological differences between w¹¹¹⁸ and dfmr1 at maturity (7 dpe), which are not present in standard conditions (Fig. 1). Finally, to control for odor specificity, we exposed 0–1 dpe control and dfmr1 null animals to the non-specific ethyl acetate odorant, which activates the DM1 glomerulus in the dorsal AL [31]. We find no significant changes in mPN2 innervation following ethyl acetate exposure in either genetic controls (branch length, p=0.36, n_control=133, n_odor=149 branches; bouton area, p=0.16, n_control=299, n_odor=299 boutons, M-W tests) or dfmr1 nulls (branch length, p=0.07, n_control=90, n_odor=79 branches, M-W test; bouton area, p=0.21, n_control=155, n_odor=145 boutons, t-test; Fig. S2). Taken together, specific odorant sensory experience powerfully drives mPN2 synaptic remodeling in the MB calyx in an FMRP-dependent mechanism restricted to the early-use critical period, resulting in synaptic connectivity changes strikingly reminiscent of dfmr1 null neurons.

FMRP required for activity-dependent synaptic remodeling during the critical period

The application of optogenetics to bidirectionally modulate neural activity during development represents a powerful approach to dissect activity-dependent circuit maturation. We therefore first employed targeted transgenic expression of an enhanced channelrhodopsin variant (CsChrimson)[32], providing cell autonomous stimulation of mPN2 during critical period development with timed blue-light (470 nm) exposure [15]. Background control and dfmr1 null animals expressing R65G01-Gal4 to drive UAS-CsChrimson::mVenus were exposed to 24 hours (0–1 dpe) of blue-light stimulation (5Hz, 20ms pulses). As a stimulation control, the exact same light-exposed genotypes were raised on vehicle (EtOH) control food that did not contain the essential all-trans retinal (ATR) cofactor, which must be exogenously supplied for channelrhodopsin function in Drosophila. Representative images and quantified results from the UAS-CsChrimson experiments are displayed in the top half of Figure 6.

Figure 6. FMRP- and activity-dependent bidirectional changes in mPN2 synapses.

(A) mPN2 neurons innervating MB calyx expressing R65G01-Gal4>UAS-CsChrimson, but lacking the essential ATR cofactor required for channelrhodopsin function (control), display the usual synaptic differences between w¹¹¹⁸ and dfmr1 null mutant at 1 dpe after 24 hours of blue light stimulation. (B) Stimulated (ATR fed) neurons display reduced synaptic branch lengths and enlarged boutons in w¹¹¹⁸ (left), but no effect in dfmr1 (right) after 24-hours of blue light stimulation (0–1 dpe). (C) Quantification of branch length (left) and bouton area (right) from critical period light stimulation. Significance from Dunn’s multiple comparisons tests: p***<0.001. (D) mPN2 innervating MB calyx expressing R65G01-Gal4>UAS-eNpHR3.0 but lacking ATR (control) display typical w¹¹¹⁸ vs. dfmr1 null synaptic differences at 1 dpe after 24-hours of amber light. (E) Activity repressed (ATR fed) neurons exhibit the opposing consequence of increased MB calyx innervation in w¹¹¹⁸ (left), but no effect in dfmr1 nulls (right) after 24-hours of amber light (0–1 dpe). (F) Quantification of synaptic branch length (left) and bouton area (right) following critical period optogenetic activity repression. Significance determined from Dunn’s multiple comparisons tests: p*<0.05, p**<0.01, p***<0.001. See accompanying data on optogenetics at maturity in Figures S3 and S4.

Developmental excitation of genetic background control mPN2 neurons causes decreased MB calyx synaptic branching (Fig. 6A, B), strikingly similar to the effects of critical period sensory odorant exposure (Fig. 5) and the dfmr1 null phenotype (Fig. 1). Blue-light stimulation of CsChrimson in control mPN2 neurons results in a decrease in synaptic branching, from 9.57±0.31μm in vehicle only (n=221 branches) to 7.53±0.22μm in ATR-raised experimental animals (n=239; p=0.0002; Fig. 6C). Optogenetic stimulation also causes a significant increase in mPN2 synaptic bouton area, from 2.05±0.03μm² (n=592 boutons) in vehicle control to 2.25±0.03μm² (n=524; p<0.0001) in ATR-fed animals (Fig. 6C). Both of these effects closely phenocopy the critical period synaptic dysmorphia in mPN2s lacking FMRP (Fig. 1). In stark contrast, optogenetic stimulation of dfmr1 null neurons does not result in any detectable change in synaptic branching (dfmr1 control, 5.5±0.21μm, n=158 branches; ATR, 5.68±0.19μm, n=197; p>0.99) or synaptic bouton size (control, 2.48±0.04μm², n=328 boutons; ATR, 2.49±0.04μm², n=382; p>0.99; Fig. 6C). In addition, stimulation of mPN2 neurons at maturity (7 dpe) does not result in significant differences in bouton area (w¹¹¹⁸: n_control=271, n_ATR=286 boutons, p>0.99; dfmr1: n_control=292, n_ATR=306, p>0.99) or branch length (w¹¹¹⁸: n_control=115, n_ATR=110 branches, p=0.63; dfmr1: n_control=133, n_ATR=124, p>0.99; Fig. S3). Therefore, cell-autonomous optogenetic stimulation of mPN2 neurons during critical period development reveals an FMRP-dependent mechanism for activity-dependent synaptic refinement and, importantly, excess developmental stimulation phenocopies the FXS disease model, supporting the FXS hyperexcitation hypothesis.

FMRP required for bidirectional activity-dependent changes in synaptic remodeling

Developmental hyperexcitation paradigms, including sensory experience and targeted channelrhodopsin stimulation, remodel wildtype neurons in the early-use critical period to phenocopy dfmr1 null mPN2 neurons, suggesting that FMRP restricts developmental excitation [16,19]. To test this hypothesis further, we utilized single cell targeted halorhodopsin optogenetics to suppress mPN2 firing during the critical period. We used R65G01-Gal4 to drive the hyperpolarizing halorhodopsin UAS-eNpHR3.0 [33] to directly suppress mPN2 excitability during critical period development, using pulsed amber-light (590 nm) exposure to cell-autonomously suppress activity [15,19]. Genetic background controls and dfmr1 null animals from R65G01-Gal4>UAS-eNpHR3.0 crosses fed on either vehicle (EtOH) or ATR cofactor were exposed to amber-light pulses (5Hz, 20ms) for 24 hours (0–1 dpe) to reduce excitability in mPN2 neurons during the critical period. Representative images and quantified results from the R65G01-Gal4>UAS-eNpHR3.0 experiments are displayed in the bottom half of Figure 6.

Hyperpolarized mPN2s display increased synaptic branching within the MB calyx compared to controls (Fig. 6D, E). Vehicle control mPN2 neurons develop branches averaging 6.63±0.21μm (n=192 branches) in length, compared to an average length of 8.17±0.23μm in ATR-fed experimental animals (n=171, p<0.0001; Fig. 6F, left). Importantly, critical period activity suppression in dfmr1 null mutants results in no detectable change in mPN2 synaptic branching (vehicle control, 5.79±0.21μm, n=180 branches; ATR-fed, 5.57±0.19μm, n=188 branches, p>0.99; Fig. 6D–F). Developmental hyperpolarization has no effect on the average area of wildtype boutons (vehicle control, 2.02±0.03μm², n=460 boutons; ATR-fed, 2.04±0.03μm², n=486; p>0.99). Interestingly, however, optogenetic hyperpolarization during the critical period causes a significant reduction in synaptic bouton size in dfmr1 null mPN2 neurons (vehicle control, 2.59±0.04μm², n=240 boutons; ATR-fed, 2.25±0.03μm², n=332; p<0.0001; Fig. 6F). Finally, we repeated hyperpolarization at maturity (7 dpe), but do not find any significant changes in branching within genotypes (w¹¹¹⁸: n_control=108, n_ATR=86, p>0.99; dfmr1: n_control=130, n_ATR=110 branches, p>0.99), although there is a significant increase in bouton size in wildtype only (w¹¹¹⁸: n_control=214, n_ATR=172, p<0.0001; dfmr1: n_control=258, n_ATR=228 branches, p>0.99; Fig. S4), consistent with published reports [34]. In conclusion, developmental hyperpolarization causes increased synaptic branching in mPN2 neurons in a mechanism that requires FMRP.

FMRP required for neurotransmission-dependent critical period synaptic remodeling

Chemical neurotransmission mediating the intercellular communication between synaptic partners is required for activity-dependent neural circuit remodeling [35]. To test the direct role of synaptic function in mPN2 connectivity refinement in the MB calyx, we drove the enzymatically active tetanus toxin light chain (UAS-TNT) to sever communication between mPN2 and downstream KCs [36]. This neurotoxin protease cleaves the essential synaptic vesicle v-SNARE synaptobrevin, leading to a complete blockade of neurotransmission [37]. For the conditional, temporal control of UAS-TNT during the critical period, we used the temperature-sensitive Gal80 repressor (Gal80^ts) to regulate R65G01-Gal4 mediated expression in mPN2 neurons [15,19,38]. We drove UAS-TNT in genetic background control (w¹¹¹⁸) and dfmr1 null (dfmr1^50M) mPN2 neurons using Gal80^ts; R65G01-Gal4>UAS-TNT in both genotypes. As a control, we used Gal80^ts; R65G01-Gal4/+ lacking the UAS-TNT transgenic construct. All four genotypes were raised at permissive 18°C, and then shifted to restrictive 29°C at the last pupal day (pupal day 4), leading to TNT expression in the experimental animals. We examined mPN2 synaptic branching and bouton development in four conditions: control and dfmr1 genotypes, with and without TNT (Fig. 7A, B), to test the impacts of neurotransmission and FMRP on mPN2 synaptic architecture in the MB calyx assayed at the end of the early-use critical period (1 dpe).

Figure 7. FMRP enables neurotransmission-dependent mPN2 synaptic refinement.

(A) Temporally-controlled, conditional expression of tetanus neurotoxin (TNT) in mPN2 neurons innervating the MB calyx (Gal80^ts; R65G01-Gal4>UAS-TNT; UAS-GCaMP5G) in the w¹¹¹⁸ genetic background. Genetic control neurons without tetanus toxin (control, left) display typical mPN2 innervation, whereas neurotransmission blockade (UAS-TNT, right) causes a dramatic expansion. (B) The same experiment in dfmr1 null (dfmr1^50M) mutants. mPN2 innervation of the calyx in the absence (control, left) and presence of tetanus toxin (UAS-TNT, right). (C) Quantification of synaptic branch length (left) and bouton area (right) in all four conditions. All animals were raised at restrictive 18°C and shifted to permissive 29°C at pupal day 4, leading to expression of TNT. All analyses were performed at 1 dpe. Significance determined from Dunn’s multiple comparisons tests: p*<0.05, p**<0.01, p***<0.001.

Critical period neurotransmission is required to restrict mPN2 branching in the MB calyx (Fig. 7A). Compared to control animals, conditional TNT synaptic transmission blockade causes a dramatic expansion of mPN2 innervation. Qualitatively, wildtype controls show striking growth following critical period synaptic silencing, whereas dfmr1 null neurons exhibit minimal response (Fig. 7A, B). Note that both genotypes develop longer mPN2 synaptic branches in this experiment, but the proportional dfmr1 phenotype remains the same (compare to Fig. 1). Control mPN2s with induced critical period TNT expression develop synaptic branches averaging 13.19±0.51μm (n=146 branches), a >25% increase with neurotransmission blockade compared to controls (9.51±0.49μm, n=87; p=0.0003; Fig. 7C). Critical period expression of tetanus toxin does not significantly impact synaptic branching in dfmr1 null mPN2, with branches averaging 7.83±0.41μm (n=127 branches) compared to 8.21±0.33μm in mutants with timed TNT expression (n=142, p>0.99). However, TNT expression has no significant effect on either wildtype or dfmr1 null mPN2 bouton size (w¹¹¹⁸, p=0.19; dfmr1, p=0.11; Fig. 7C). Similar to halorhodopsin hyperpolarization, the critical period neurotransmission block causes enhanced MB calyx innervation. These results show that neural excitation and downstream synaptic transmission are both required to restrict synaptic remodeling in a mechanism that absolutely requires FMRP.

Discussion

Neural circuit remodeling during developmental critical periods requires reception of sensory experience (activity) and the responsive orchestration of synaptic refinement to optimize behavioral performance [11,17]. FMRP is hypothesized to mediate these activity-dependent critical period processes in an activity sensor mechanism [5,6,39], and as an activity-dependent translational regulator [40]. To test these hypotheses, we dissected FMRP requirements in the well-mapped Drosophila olfactory learning/memory circuit [41], focusing on projection neurons linking upstream sensory neurons [29] to the downstream central brain mushroom body mediating learning acquisition and memory consolidation [42]. Mushroom body Kenyon cells also associate sensory input with a valence signal from dopaminergic neurons, connecting sensory experience to the reward pathway [43,44]. Null dfmr1 mutants exhibit deficits in olfactory learning and memory [9,10,45,46], Kenyon cell architecture [46,47], projection neuron dendritic arborization [15] and activity-dependent calcium signaling [19]. In the FXS condition, transiently altered synaptic connectivity between projection neurons and target Kenyon cells profoundly impacts establishment of specific associations between sensory input, learning/memory and resultant behavioral output. We predict the seemingly ephemeral changes have lasting impacts into maturity, when differences in synaptic architecture are minimal [5,48], but strong behavior deficits persist [49,50]. We hypothesize subtle differences in circuit connectivity, or consequent functional synaptic deficits arising from transient critical period defects, must be manifest in impairments in emergent circuit properties at maturity that result in persistent behavioral deficits.

Synaptic connectivity investigations show two primary defects in FMRP-deficient mPN2 neurons: 1) truncated synaptic branches in the posterior mushroom body calyx, and 2) enlarged synaptic boutons on postsynaptic Kenyon cells (Fig. 1). Importantly, both defects manifest only during the early-use critical period and are not detectably present at maturity, after FMRP expression has precipitously declined [4,5]. Milder, persistent synaptic architecture defects are detected in some cases, dependent on the genetic background. Null dfmr1 mutant boutons also display a critical period-restricted reduction in presynaptic active zone scaffold Brp (Drosophila ELKS protein) only during the critical period (Fig. 2), showing that FMRP regulates a core organizing component of presynaptic maturation [27] selectively during this transient time window. Using transgenic GFP reconstitution to test synapse connectivity [25], we find FMRP-deficient mPN2 neurons develop impaired synaptic partner interactions with reduced mPN2-KC contacts (Fig. 3). GRASP synaptic defects likewise are restricted to the early-use critical period. Electron microscopy during the critical period reveals greatly enlarged synaptic boutons [47] with reduced active zone density [51] in dfmr1 null mutants compared to age-matched controls (Fig. 4). These ultrastructural results are consistent with the light microscopy findings, revealing expanded synaptic bouton area (compare to Figs. 1, 3) coupled with reduced synaptic density (compare to Fig. 2) during the critical period. Taken together, these combined approaches reveal compromised synaptic connectivity in the Drosophila disease model, consistent with defects in the mouse FXS model [16,28], which transiently occur only during the early-use critical period.

We next explored activity-dependent FMRP roles in the critical period. We find that critical period exposure to sensory olfactory experience causes dramatic changes in mPN2 mushroom body synaptic connectivity (Fig. 5), reminiscent of odorant-induced critical period changes in antennal lobe synaptic glomeruli [12,14]. Synaptic remodeling is FMRP-dependent, and critical period activity phenocopies dfmr1 null defects (Fig. 5). Induced changes are specific to the pyrrolidine-sensitive VL1-mPN2 glomerulus, as other odorants (i.e. ethyl acetate) do not alter mPN2 synapses. Importantly, olfactory experience at maturity has no effect on wildtype mPN2s, but does cause minor changes in dfmr1 null mPN2s, which supports the ‘shifted critical period’ ASD hypothesis [2,52]. FMRP and activity may function in parallel pathways, but the fact that FMRP is activity regulated [5,6], and mediates activity-dependent processes [39], strongly suggests a direct activity-dependent FMRP mechanism for critical period synaptic refinement. We find mPN2-targeted optogenetic stimulation during the critical period [15] phenocopies FXS model synaptic defects, with reduced branching and enlarged synaptic boutons (Fig. 6), reminiscent of defects in downstream Kenyon cells [5]. Similar cell-autonomous optogenetic stimulation causes erroneous axon pathfinding [53] and diminished axon outgrowth [54]. Importantly, both sensory stimulation via peripheral odorant exposure and direct mPN2 stimulation via channelrhodopsin optogenetics phenocopy FXS model defects. All activity-dependent changes require FMRP and are tightly restricted to the early-use critical period. Together, these results support the FXS hyperexcitation theory [17] and highlight a critical period deficit in the suppression of excitatory synapses.

In contrast to stimulation paradigms, cell-targeted halorhodopsin suppression of neuronal activity causes increased mPN2 synaptic branching in the MB calyx (Fig. 6). This result demonstrates bidirectional capacity for mPN2 to manifest activity-dependent changes in synaptic connectivity during the early-use critical period [15]. This phenotype is comparable to the overgrown axonal projections that result from developmental application of the GABA antagonist picrotoxin [55], suggesting that activity normally limits synaptic connectivity. Surprisingly, hyperpolarization of wildtype mPN2 neurons also caused increased synaptic bouton size at maturity, as reported previously by Kremer and colleagues [34], albeit not during the critical period. It is therefore clear that neuronal hyperpolarization impacts synaptic connectivity and architecture in a distinct mechanism compared to excess excitation. However, it is not clear what role the FMRP activity sensor plays when neuronal activity is dampened. Indeed, we were surprised that halorhodopsin hyperpolarization influences dfmr1 null mPN2 synaptic bouton area (Fig. 6), suggesting that neurons lacking FMRP retain some capacity to function in activity-dependent synaptic bouton refinement during critical period development. There is evidence that FXS disease model dysfunction can be alleviated through increased activation of the inhibitory neural circuitry: for example, pharmacological enhancement of GABAergic signaling is sufficient to rescue some FXS hyperexcitation [56] and can rescue biochemical, morphological and behavioral phenotypes in the Drosophila FXS disease model [57]. Thus, excitation/inhibition balance appears important for sculpting synaptic circuit connectivity during the critical period.

The blockade of mPN2 neurotransmission by conditional, targeted expression of the tetanus neurotoxin (TNT) leads to striking synaptic overgrowth in wildtype neurons that represents an opposite extreme in comparison to dfmr1 null phenotypes (Fig. 7). Suppressed circuit activity (via both halorhodopsin and tetanus toxin manipulations) may spur increased process exploration or connectivity with potential synaptic targets in the mushroom body calyx, further suggesting that reduced branching in FMRP-deficient mPN2 neurons may stem from excess excitation during critical period development. TNT neurotransmission blockade similarly causes aberrant competition for glomerular space during olfactory circuit targeting [58], and enlarged downstream postsynaptic terminals within motor circuits [59]. In dfmr1 null mutants, neurotransmission blockade has little impact on mPN2 presynaptic architecture (Fig. 7), demonstrating yet another level of activity-dependent FMRP requirement. As we do not yet possess tools to assay mPN2 postsynaptic partners, we have no insight into postsynaptic Kenyon cell differentiation downstream of the TNT neurotransmission blockade. Our planned future work to manipulate neuronal excitability and neurotransmission strength should provide more precise understanding of FMRP function in limiting excitatory synapse connectivity in the developing brain circuitry. The clear requirement for FMRP in activity-dependent synaptic refinement during the early-use critical period, evidence of temporally-shifted critical periods in the FXS condition, and the promise of new paradigms to rebalance excitatory/inhibitory synaptic connectivity all hold tremendous future therapeutic potential for combating the FXS disease state.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to the corresponding author, Kendal Broadie (kendal.broadie@vanderbilt.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Drosophila Genetics

All stocks were reared on standard cornmeal/agar/molasses Drosophila food at 25°C, unless stated otherwise. Multiple recombinant lines of the characterized dfmr1 null allele dfmr1^50M [8] and dfmr1 genetic background line (w¹¹¹⁸) were generated using standard genetic techniques with the following transgenic lines: 1) R65G01-Gal4 driver line for targeted mPN2 control [21], 2) both UAS-mCD8::GFP (plasma membrane) and UAS- DenMark (dendrite) reporters [60], 3) the 20XUAS-IVS-GCaMP5G calcium reporter [61], 4) both the tagged UAS-IVS-CsChrimson::mVenus [32] for target optogenetic excitation and UAS-eNpHR3.0::eGFP [33] for optogenetic activity inhibition, 5) conditional tubP- Gal80^ts; TM2/TM6 [38] for temporal control of Gal4, 6) Ir75d-Gal4 for labeling the target olfactory sensory neuron (OSN) presynaptic to VL1 glomerulus (Drosophila Stock Center, Bloomington, IN, USA), 7) mb247-spGFP11; UAS-spGFP1-10 for GRASP experiments [25], and 8) the insulated UAS-TNT [37] for synaptic silencing. For the conditional Gal80^ts experiments, both control and experimental animals were raised at the same permissive temperature (18°C, Gal80 ^ts active) until the last pupal day (P4), and then shifted to the restrictive temperature (29°C, Gal80 ^ts inactive) until 1 day post-eclosion (dpe). The developmental stages of manipulation and analysis are indicated for each individual experiment. Animals of both sexes were used in all analyses.

METHOD DETAILS

Immunocytochemistry and Imaging

Immunocytochemistry was done as we previously described [15]. Brains were dissected in 1xPBS, fixed for 30 minutes in 4%PFA/4%sucrose, washed 3×20 minutes with 1xPBS, and blocked for 1 hour in 1xPBS/1%BSA/0.5%NGS/0.2%Triton-X 100. Brains were then placed in primary antibody on a shaking platform at 4°C overnight. Primary antibodies used include the following: mouse anti-FMRP (6A15, 1:500; Sigma), rabbit anti-GFP (ab290, 1:2000; Abcam, Cambridge, UK), mouse anti-Bruchpilot (nc82, 1:100; Developmental Studies Hybridoma Bank (DSHB), Iowa City, IA), and mouse anti-Discs Large (4F3, 1:50; DSHB). For the GRASP experiments, a new polyclonal chicken anti-GFP recognizing only the complete GFP molecule was generated in the Vanderbilt Antibody and Protein Resource (used at 1:200). Following 3×20 minutes washes in 0.2% PBS-Triton, brains were incubated in secondary antibody for 2–3 hours. Secondary antibodies used include the following: anti-rabbit-IgG AlexaFluor 488, anti-chicken-IgG AlexaFluor 488 as well as anti-mouse-IgG AlexaFluor 568 (Molecular Probes, Eugene, OR), all used at a 1:500. Following 3×20 minute washes in 0.2%PBS-Triton, 1×20 minutes in 1xPBS and 1×20 minutes in dH₂0, brains were mounted in Fluoromount G (Electron Microscopy Sciences, Hatfield, PA). Brain images acquired on a Zeiss Meta 510 laser-scanning confocal microscope with 40–63X objectives. The central brain region was imaged in collected Z-stacks of 1μm optical sectional depths.

Electron Microscopy Processing

Electron microscopy studies were carried out as previously described [26], with slight variations. Staged animals were immersed in 70% EtOH for 1 minute. Brains were dissected in phosphate-buffered saline (PBS, pH 7.4) followed by immediate fixation in 2.5% gluteraldehyde in 0.1M sodium cacodylate buffer (SC, pH 7.4) over-night (O/N) at 4°C. Samples were then washed 3X 20 minutes with SC at room temperature (RT) followed by a final wash in SC O/N at 4°C. Samples were then fixed with 1% osmium tetroxide in SC for 1 hour at RT. Samples were then washed 3X 20 minutes with SC at RT. Samples were treated with 2% uranyl acetate in water for 2 hours at RT, followed by 3X 20 minute washes in water at RT. Samples were then placed through an EtOH dehydration series (30, 50, 70, 90, 95, 2X 100%) for 10 minutes each at RT. Propylene oxide (PO) was then used as a transitional solvent with 50/50 EtOH/PO and 2X 100% PO for 10 minutes each at RT, then 75/50 PO/Ebed-812 (Electron Microscopy Services) resin for 30 minutes at RT, 50/50 PO/Ebed-812 for 1 hr at RT, and 50/50 PO/Ebed-812 O/N at RT. Finally, 3X changes of fresh Epon-812 were allowed to infiltrate the sample for 2 hours, 4 hours and then O/N at RT.

Electron Microscopy Block Sectioning and Imaging

Flat block molds (Electron Microscopy Sciences) were half filled with Epon-812 resin and heated at 60°C until the resin reached a tacky consistency (approximately 4 hours). Samples were then placed in blocks, the brains oriented anterior up and dorsal forward (toward the block face). Blocks were filled with resin and allowed to fully polymerize for 48 hours at 60°C. After polymerization, thick sections (500nm) were cut using a Leica UCT Ultracut microtome until the dorsal most brain region appeared (first sight of brain tissue in sections). Thick section (500nm) cutting was then continued until the MB calyx was reached (approximately 40μm). Thin sections (65nm) were then cut for another 20μm, and collected on formvar-coated grids (Electron Microscopy Sciences) at 10 sections per grid. Images were acquired using a Philips CM10 transmission electron microscope (TEM) at 80kV. Measurements were taken with ImageJ freehand selection tool. 5–10 PN boutons were quantified within each MB calyx. To eliminate the chance of reimaging synaptic boutons, only one section per grid was imaged, with at least 5 grids (~3μm) of space between each grid examined. 3 animals were used for each EM grid, with 5 MB calyces analyzed per genotype. For sample size, 16 control and 22 dfmr1 sections were examined, with 70 control and 80 dfmr1 boutons, respectively.

Sensory Olfactory Exposure

Staged animals at either eclosion (0 dpe) or maturity (7 dpe) were exposed to odorants for 24 hours at RT. Pyrrolidine was selected based on specific activation of the Ir75d receptor [29], which is expressed in olfactory sensory neurons that innervate the VL1 glomerulus [30], presynaptic to the mPN2 dendritic arbor [15, 18]. Ethyl acetate was used to activate the DM1 glomerulus [31]. Olfactory exposure was done within a sealed 50 mL conical tube. Mineral oil control (500μL) or pyrrolidine (1:100 dilution; #394238, Sigma-Aldrich) was placed in a closed 1.5 mL microcentrifuge tube, with the cap center punctured by a 20 gauge needle. Genetic background (w¹¹¹⁸) control and dfmr1 null (dfmr1^50M) mutants expressing R65G01-Gal4>UAS-mCD8::GFP were used to assess PN architecture and connectivity following odorant exposure.

Optogenetics

Pairwise crosses were made between R65G01-Gal4 driver and UAS channel lines (UAS-IVS-CsChrimson::mVenus, UAS-eNpHR3.0::eGFP) in genetic background control (w¹¹¹⁸) and dfmr1 null mutant (dfmr1^50M) backgrounds. Offspring from all four genotypes were fed from hatching on food supplemented with either 10μL EtOH (vehicle control) or 100μM all-trans retinal (ATR), a co-factor essential for channel function [62]. Developmentally staged animals were placed in 30mm petri dishes with Whatman paper strips saturated in a 20% sucrose solution containing either vehicle control or ATR. All control and experimental animals were placed in a LED exposure chamber with two Luxeon Rebel Endor Star 3X 15-Watt LED arrays (blue light for ChR2 variants, amber light for eNpHR3.0; LED Supply, Randolph, VT). At 15V, the LED arrays generate ~100μW/cm² of blue light at the working distance of 2cm. Animals were exposed to 24 hours of 20ms light pulses at 5Hz frequency. Acutely dissected brains from light-exposed animals were then processed for imaging as described above.

QUANTIFICATION AND STATISTICAL ANALYSIS

Synaptic branch length was determined using the ImageJ Simple Neurite Tracer plugin. Branch length outliers were determined and excluded by using Robust Regression and Outlier Removal (ROUT) in Prism, with Q=1%. Individual synaptic boutons were defined as UAS-mCD8::GFP varicosities >0.75 μm² (minimal area) driven by the mPN2 specific R65G01-Gal4 driver. Only mPN2 boutons with clearly defined boundaries were included in analyses. Synaptic bouton area (maximum cross section) was determined using the ImageJ freehand selection tool. Anti-Brp fluorescence intensity was assayed under identical confocal imaging settings in every given experiment, with all data normalized to genetic control levels for each independent trial. All statistical analyses were performed using Prism software (GraphPad Software, San Diego, CA). No statistical methods were used to determine sample sizes before experiments, reflecting the standards in the field. Multiple independent replicates were performed for all experiments, with each trial representing a separate genetic cross for the experimental genotypes. For all data analyses, Gaussian distribution was determined using the D’Agostino-Pearson omnibus normality test. Data from two group comparisons were analyzed with either a two-tailed unpaired t-test (Gaussian distribution) or the two-tailed Mann-Whitney test (M-W test; non-Gaussian distribution). Data from all three or more comparisons were always analyzed with one-way analysis of variance (ANOVA), with the Tukey’s multiple comparisons post-test (Gaussian distribution) or the Kruskal-Wallis test (non-Gaussian distribution), followed by the Dunn’s multiple comparison test comparing each mean to every other column. Statistical details are included in each section, with n values and n representation indicated for each experiment. The statistical tests utilized are indicated either in the figure legends or within manuscript text (when multiple tests were used in a single figure). For most figures, data are presented in box-and-whisker plots (minimum, median, maximum and quartiles). For normalized Brp quantification (Fig. 2 and Fig. S1), data are presented as mean ± standard error of the mean. Significance levels are shown as p>0.05 (not significant, n.s.), p<0.05 (*), p<0.01 (**) and p<0.001 (***).

Highlights.

• Fragile X syndrome disease model shows critical period restricted circuit defects

• Critical period activity-dependent synaptic remodeling absolutely requires FMRP

• Critical period hyperexcitation phenocopies Fragile X syndrome synaptic defects

• FMRP enables critical period sensory activity refinement of synaptic connectivity