To measure callosal synaptic function and connectivity between hemispheres, mCherry-tagged Channelrhodpsin-2 (ChR2) was expressed in callosal projecting cortical neurons by injecting AAV9 into in one hemisphere of primary somatosensory cortex of postnatal day (P) one pups. At P18-25, acute coronal slices containing barrel cortex contralateral to the AAV injected side were prepared and whole cell patch clamp recordings of L2/3 pyramidal neurons were performed in the region innervated by fluorescently labeled callosal axons (Figure 1A–B). Monosynaptic excitatory postsynaptic currents (EPSCs) were elicited by stimulating ChR2 expressing callosal axons with brief (2 ms) blue LED light pulses in the presence of tetrodotoxin (TTX) and 4-aminopyridine (4-AP) as described previously (Petreanu et al., 2007; Rajkovich et al., 2017; Figure 1C). The blue LED was delivered through a 40 X lens centered around the soma of approximately 350 µm diameter in size and thus the amplitude of LED-evoked EPSCs likely reflects the overall strength of callosal synaptic inputs onto the recorded neuron. The amplitude of LED-evoked EPSCs in Fmr1 KO neurons was reduced by ~40 % compared to WT. Similarly, reduced EPSC amplitudes were observed by comparing raw values or when normalized to LED power (Figure 1D). To determine if weak callosal synaptic inputs were localized to a subcellular region on recorded L2/3 neurons (e.g. apical vs. distal dendrites), we utilized a method termed subcellular Channelrhodopsin2-assisted circuit mapping (sCRACM Petreanu et al., 2009). A blue laser was flashed across an array of locations, relative to the soma of the recorded neuron, in a pseudorandom order to elicit glutamate release from ChR2-expressing callosal axon terminals and EPSCs at localized dendritic sites (Figure 1E–F). The mean amplitude of EPSCs at each site was converted into a color map for each recorded neuron and these maps were then aligned by the position of soma and averaged within genotype. The resulting average map depicts the subcellular distribution of callosal synaptic input strengths onto WT or Fmr1 KO neurons (Figure 1G). Results reveal that callosal synaptic inputs onto WT L2/3 neurons are strongest at the proximal apical dendrites (Figure 1H), consistent with previous reports (Petreanu et al., 2009) and there is an interaction of Fmr1 and vertical position (*p < 0.05, F(15, 849) = 1.723; mixed-effects ANOVA). Comparison of mean EPSC amplitudes from proximal apical dendrites reveals a 50 % reduction in callosal synaptic input strength in Fmr1 KO mice (Figure 1H).

FMRP is known to have both pre- and postsynaptic, cell autonomous roles in excitatory synapse development, depending on the cell type and synaptic input (Patel et al., 2013; Patel et al., 2014; Pfeiffer et al., 2010). To determine the cell-autonomous and synaptic locus of FMRP function in development of callosal synapses, we deleted Fmr1 in a sparse population (~3%–5%) (Figure 2—figure supplement 1) of postsynaptic L2/3 pyramidal neurons by injecting AAV expressing GFP-tagged Cre (AAV9.GFP-Cre) into the left lateral ventricle of P1 pups (Kim et al., 2013) with a floxed Fmr1 gene (Fmr1^fl/fl or Fmr1^fl/y) (Figure 2A; Mientjes et al., 2006). In the same animals, AAV.ChR2-mCherry was injected into the right barrel cortex to label callosal axons as in Figure 1. This experimental design allowed us to measure the cell autonomous effect of postsynaptic FMRP deletion in L2/3 neurons in a primarily WT cortex on synapses from WT callosal axons. At P18-30, acute cortical slices were prepared and simultaneous whole cell recordings were performed on pairs of neighboring L2/3 pyramidal neurons in left barrel cortex with one being a GFP (-), or WT neuron, and the other a GFP (+), or Fmr1 KO neuron (Figure 2A–B). Monosynaptic EPSCs from callosal axons were elicited with either bulk LED stimulation or blue laser, to perform sCRACM, as in Figure 1 (Figure 2C, Figure 2—figure supplement 2A-B). Interestingly, at the earliest age for recording (P18-20), overall callosal synaptic input strengths were not different between WT and postsynaptic Fmr1 KO neurons (Figure 2D). However, at P23-30, postsynaptic Fmr1 KO neurons had a 40 % reduction in callosal synaptic input strength compared to neighboring WT neurons (Figure 2D). A two-way ANOVA indicates a significant interaction between genotype and age (Fmr1 x Age ***p < 0.001, F(1, 32) = 13.87)(Figure 2E). In addition, we observed a strong trend of callosal input strength to increase with developmental age in WT, but not Fmr1 KO, neurons (Figure 2E). Similar results were obtained with sCRACM which revealed a ~ 45 % reduction in callosal input strength onto proximal dendrites of Fmr1 KO neurons (Figure 2—figure supplement 2C-E). These data demonstrate that Fmr1 in postsynaptic L2/3 neurons cell-autonomously promotes the development and/or strengthening of callosal synaptic inputs. Weak callosal synaptic strength in Fmr1 KO neurons at P23-30 could be a consequence of deficient or delayed callosal synapse maturation and may normalize in the adult. To test this possibility, we repeated experiments deleting postsynaptic Fmr1 with a P1 AAV Cre-GFP injection and recorded LED-evoked EPSCs onto pairs of neighboring WT and Fmr1 KO L2/3 neurons in adult mice (P57-65). Similar to young mice, callosal mediated EPSCs were weak in Fmr1 KO neurons; reduced by 25%, in comparison to WT neurons (Figure 2F–G). These results indicate that weak callosal synaptic transmission persists into adulthood with postsynaptic Fmr1 deletion.

Weak callosal-mediated EPSCs onto postsynaptic Fmr1 KO L2/3 pyramidal neurons could be due to reductions in the number of synaptic connections, presynaptic release probability, the strength of individual synapses, or any combination of these. To further investigate the synaptic basis of weak callosal inputs in P23-30 slices, we recorded in strontium (Sr²⁺), substituted for Ca²⁺, in the external ASCF which results in asynchronous vesicle release and measurement of quantal synaptic events evoked from callosal axons by the LED (Oliet et al., 1996). Changes in the amplitude of quantal events reflect synapse strength, whereas changes in the frequency of events reflect changes in synapse number and/or presynaptic release probability. Postsynaptic Fmr1 KO neurons received >20% fewer evoked events in Sr²⁺ in comparison to neighboring WT neurons with no change in event amplitude (Figure 3A–C). The baseline frequency and amplitude of spontaneous EPSCs, prior to LED stimulation, were not different between genotypes. These results suggest the reduced EPSCs onto postsynaptic Fmr1 KO neurons are a consequence of reduced presynaptic release probability, functional synapse number or both. To further test this conclusion, we measured the coefficient of variance (C.V.) of LED-evoked callosal EPSCs onto WT and Fmr1 KO neurons (from experiments in Figure 2D, Figure 4 and Figure 5B). C.V. is inversely proportional to release probability and synapse number (Manabe et al., 1993) and was increased by 18 % in Fmr1 KO neurons (Figure 3—figure supplement 1). This result together with reduced frequency of events evoked in Sr²⁺ indicate that the weakening of callosal-mediated EPSCs in Fmr1 KO is due in part to decreased synapse number and/or presynaptic release probability.

To further differentiate between release probability and functional synapse number, we measured NMDA receptor (R) mediated EPSCs from callosal axons onto WT or postsynaptic Fmr1 KO L2/3 neurons. Because NMDARs are colocalized with AMPARs at excitatory synapses, reduced glutamate release probability or callosal synapse number onto Fmr1 KO neurons would be expected to result in weak NMDAR-EPSCs. To evoke isolated NMDAR EPSCs from callosal axons, we included the AMPAR antagonist DNQX, glycine, and low Mg²⁺ (0.1 mM) in the ACSF and voltage clamped neurons at –70 mV. In contrast to AMPAR-mediated EPSCs, amplitudes of NMDAR EPSCs evoked from callosal axons were not different between WT and neighboring postsynaptic Fmr1 KO neurons (Figure 3D). This observation suggests that the NMDAR content within callosal synapses are similar for WT and KO neurons and that there is no change in presynaptic release probability or synapse number for the callosal axons. Taken together with the reduced frequency of AMPAR quantal events and increased C.V. (Figure 3C, Figure 3—figure supplement 1), our results suggest that L2/3 Fmr1 KO neurons have an increased proportion of synapses with NMDARs, but not functional AMPARs, indicative of immature synapses.

For experiments described in Figures 1 and 2, we included TTX and 4-AP in the bath to isolate evoked EPSCs from ChR2-expressing callosal axons and block potential contamination from polysynaptic local L2/3 inputs. A caveat of this approach is that callosal synaptic transmission is not triggered by action potentials, but by direct depolarization and activation of voltage-gated Ca²⁺ channels at the callosal axon terminal. To determine if similar results are observed with action potential-evoked synaptic transmission, we repeated experiments without TTX/4AP, but increased Ca²⁺/Mg²⁺ and added the NMDAR blocker, CPP, to reduce polysynaptic responses from local circuits. We also used a blue laser to depolarize a local (30 µm) area of callosal axons while limiting activation of local circuits (Figure 5A). Quantification of mean EPSC amplitude within the region covering the soma and major dendrites shows that L2/3 pyramidal neurons with postsynaptic Fmr1 deletion have a 45 % reduction in callosal synaptic input strength as compared to neighboring WT neurons (Figure 5A–B).

L5 pyramidal neurons are another major target of callosal axons (Petreanu et al., 2007; Wang et al., 2007). To determine if weak callosal synaptic inputs to postsynaptic Fmr1 KO neurons are specific to L2/3, we performed the same experiment on L5 pyramidal neurons. Similar to L2/3, L5 pyramidal neurons with postsynaptic Fmr1 deletion have a 40 % reduction in the strength of callosal inputs stimulated around the soma and apical dendrites (Figure 5C–D). Together, these results show that postsynaptic deletion of Fmr1 in pyramidal cortical neurons generally weakens callosal synaptic inputs and this is observed with action potential driven synaptic transmission and across different layers.

Targeting, branching and elaboration of callosal axons into the contralateral neocortex occurs postnatally (~P5-15) and depends on activity of presynaptic and postsynaptic cortical neurons as well as whisker sensory experience (Huang et al., 2013; Mizuno et al., 2010; Wang et al., 2007). This suggests that sensory experience may interact with FMRP to regulate development of callosal synapses. To test this idea, we sparsely deleted Fmr1 in postsynaptic L2/3 neurons in the left barrel cortex and express ChR2-mCherry in callosal projecting neurons in the right hemisphere (as in Figure 4A). Beginning at P15, we unilaterally trimmed whiskers daily on the right whisker pad which would reduce the most direct ascending sensory-driven patterned activity to the left hemisphere containing L2/3 neurons with postsynaptic Fmr1 deletion (postsynaptic deprivation condition). In littermates, we trimmed whiskers on the left whisker pad which would primarily deprive the L2/3 callosal projection neurons of sensory driven patterned activity (presynaptic deprivation condition) (Figure 4A). Whisker trimming began at P15 to reduce effects of sensory deprivation on the early growth and branching of callosal projection axons and continued until the day before slice recordings. Recordings of EPSCs evoked from callosal axons by either LED stimulation (Figure 4B) or sCRACM (Figure 4—figure supplement 1) were obtained from pairs of neighboring WT and Fmr1 KO L2/3 neurons. Because we recorded from pairs of WT and postsynaptic Fmr1 KO neuron neighbors, any effects of sensory deprivation on callosal axon innervation, growth, or branching in a given cortical region would be expected to similarly affect each genotype. Thus, our results reflect the cell autonomous effects of Fmr1 on synaptic function or connectivity. In the ‘presynaptic deprivation’ condition, callosal synaptic input strength was weak (27 % reduction) onto postsynaptic Fmr1 KO L2/3 neurons as compared to WT, similar to that observed in whisker intact mice (Figure 2D). In contrast, in the ‘postsynaptic deprivation’ condition, callosal synaptic inputs strengths were similar between WT and Fmr1 KO neurons. To compare the callosal synaptic inputs strength within genotypes and across sensory deprivation paradigms, we normalized each LED induced EPSC to its stimulation power (Figure 4C). A two-way ANOVA revealed an interaction of deprivation condition and Fmr1 (*p < 0.05, F(1, 31) = 4.977, ANOVA). Surprisingly, callosal synaptic input strengths in WT neurons were not different between deprivation conditions. In contrast, in Fmr1 KO neurons callosal input strengths were weaker, decreased by ~45%, in the presynaptic deprivation condition, as compared to postsynaptic deprivation. These data suggest that whisker experience-driven activity of postsynaptic Fmr1 KO L2/3 neurons weakens callosal synaptic inputs.

A previous study reported weak L4 to L2/3 synaptic inputs in the barrel cortex of global Fmr1 KO mice (Bureau et al., 2008), suggesting that Fmr1 generally promotes excitatory synapse strength onto L2/3 neurons regardless of whether they are from local or long-range sources. To determine if postsynaptic Fmr1 promotes excitatory synapse development from local cortical circuits in a cell autonomous manner as it does for callosal inputs, we assessed local input strengths using Laser Scanning Photo-Stimulation (LSPS) with glutamate uncaging. Slices were bathed in MNI-caged glutamate and pseudorandom flashes of a UV laser (355 nm) beam at individual locations (20 µm diameter) within a 16-by-16 grid surrounding the recorded L2/3 neurons focally released glutamate to evoke action potentials and synaptic transmission from neurons at that location. The grid for UV laser flashing was positioned to stimulate neurons in Layers 2/3–5 of home and adjacent barrels. At P23-30, LSPS was performed on pairs of simultaneously recorded WT and neighboring L2/3 neurons with postsynaptic Fmr1 deletion (Figure 6A–B). The amplitude of monosynaptic EPSCs evoked from each position in the slice for an individual neuron were converted into a color map and then individual maps were aligned to the home barrel to create an average color map per genotype (Figure 6C–D). Responses from direct glutamate activation onto recorded neurons, as described in methods, were excluded from analysis, and represented by black pixels. EPSCs evoked from L4 home or adjacent barrels, L5A or L2/3 were unaffected on Fmr1 KO L2/3 neurons (Figure 6E), in contrast to the weak EPSCs from callosal inputs.

Bureau and colleagues demonstrated weak L4 to L2/3 synaptic strength in the global Fmr1 KO when measured at 2 weeks of age, but the difference was diminished at 3 weeks, suggestive of a developmental delay. To test if loss of postsynaptic Fmr1 weakens local L4 to L2/3 synaptic inputs at early developmental stages, we performed LSPS on pairs of WT and postsynaptic Fmr1 KO L2/3 pyramidal neurons at P14-17. Similar to results obtained at P23-30, we observed normal L4 and adjacent L2/3 synaptic input strengths onto postsynaptic Fmr1 L2/3 neurons recorded at P14-17 (Figure 6—figure supplement 1A). We also tested if embryonic deletion of postsynaptic Fmr1 was necessary to affect L4 to L2/3 synapse development, using slices from females with heterozygous (het) and mosaic expression of Fmr1, as we have described (Patel et al., 2013; Patel et al., 2014). Briefly, X-linked GFP mice were bred with Fmr1 KO males. Due to random X-chromosome inactivation in the embryo, female Fmr1 het mice offspring have a mosaic expression of GFP(+) WT and GFP(-) Fmr1 KO neurons. LSPS maps were performed on pairs of WT and Fmr1 KO neurons at P14-17 where we observed normal L4 input strengths onto L2/3 Fmr1 KO neurons (Figure 6—figure supplement 1B). In contrast to results with postsynaptic Fmr1 deletion, we observed weak L4-to-L2/3 synaptic strengths with LSPS in the global Fmr1 KO (Figure 6—figure supplement 1C) similar to that reported by Bureau et al., 2008. However, unlike Bureau et al, we observed weak L4 inputs at later ages (P18-25; Figure 6—figure supplement 1C). Taken together our results implicate FMRP in postsynaptic L2/3 neurons in the input specific development of callosal synapses. Although synaptic inputs from L4 to L2/3 are weak in the global Fmr1 KO, this is not a postsynaptic, cell autonomous function of FMRP, but instead is either a non-cell autonomous function of FMRP or a role in presynaptic L4 neurons.

Functional MRI studies show that Fmr1 KO mice have reduced corticocortical long-range connectivity, especially among the sensory and motor cortices (Haberl et al., 2015; Zerbi et al., 2018) as well as reduced corpus callosum structural integrity. By investigating callosal synaptic connections between bilateral barrel cortices, we find weak functional synaptic inputs at these long-range connections in L2/3 in Fmr1 KO mice (Figure 1). This result suggests that reduced functional coherence between bilateral barrel cortices in Fmr1 KO mice could be due to a disrupted communication between the cortices through the weakened monosynaptic transmission. Interhemispheric functional connectivity is thought to shape perceptual integration including those involved in speech comprehension and global form processing, domains that are impaired in ASD (Booth and Happé, 2018; Friederici et al., 2007; Happé and Frith, 2006; Peiker et al., 2015; Preisig et al., 2021; Simon and Wallace, 2016). Based on fMRI results in the Fmr1 KO (Haberl et al., 2015; Zerbi et al., 2018), it is likely that FMRP promotes other ‘long-range’ synaptic connections between cortical areas, such as ipsilateral connectivity between S1 and M1/M2, as well as cortical-subcortical structures.

Using in vivo sparse deletion and a simultaneous recording paradigm, we can stimulate the same set of callosal axons for a pair of WT and Fmr1 KO neurons and directly compare their synaptic inputs. Our observation of weaker callosal synaptic inputs onto cell autonomous Fmr1 KO L2/3 neurons (Figure 2) confirms results in Fmr1 global KO mice and further reveals an essential role of postsynaptic FMRP in promoting callosal synapses development. With postsynaptic Fmr1 KO we observe weak callosal synapses at P23-30, but not at P18-20. To determine if a similar developmental profile is observed in the global Fmr1 KO, we analyzed a subgroup of data collected at P18-20 and observe weak callosal synaptic strength in the Fmr1 KO at this early time point (WT = 181 ± 22 pA, n = 15; KO = 106.1 ± 17.77 pA, n = 7, *P < 0.05, unpaired t-test). The later developmental onset of weak callosal synaptic transmission with postsynaptic Fmr1 deletion may be due to later postnatal deletion of Fmr1 using P1 AAV-Cre-GFP injection which we estimate to occur about P7-P9 (Rajkovich et al., 2017). However, we cannot rule out a role for FMRP in presynaptic, callosal projecting neurons, or other cell types, such as oligodendrocytes, within the first postnatal weeks to establish callosal synaptic connections (Doll et al., 2020; Hanson and Madison, 2007; Patel et al., 2013).

Our results indicate that callosal synapses do not mature or are not maintained without FMRP in postsynaptic L2/3 neurons. In support of a role in maturation, callosal synaptic strength tends to increase from P18-20 to P23-30 in WT neurons but not in postsynaptic Fmr1 KO neurons (Figure 2E). Furthermore, we observe a selective weakening of AMPAR, but not NMDAR-, mediated synaptic transmission at callosal inputs onto Fmr1 KO neurons (Figure 3). This result, together with a decreased frequency of evoked quantal events in Sr²⁺ and increased coefficient of variation, suggests a reduced number of mature synapses with functional AMPARs. As synapses mature, they acquire NMDARs prior to AMPARs and NMDAR-only, immature synapses are often termed ‘silent’ synapses (Ashby and Isaac, 2011; Hanse et al., 2013). Although we did not measure ‘silent’ callosal synapses here, our findings would predict more ‘silent’ and immature callosal synapses in Fmr1 KO L2/3 neurons. In support of this idea, thalamocortical inputs to L4, another ‘long-range’ synaptic pathway, are delayed in their development in the Fmr1 KO, as measured by acquisition of AMPARs (Harlow et al., 2010). Another possible explanation for our results is the optogenetic stimulation paradigm we used to evoke glutamate release from callosal axons saturates synaptic NMDARs, but not AMPARs, due to the higher affinity of glutamate for NMDARs (Patneau and Mayer, 1990). Additional experiments and/or methods are needed to confirm a selective decrease in AMPARs at Fmr1 KO callosal synapses. The observations of reduced corpus callosum structural integrity and interhemispheric coherence with fMRI in adult Fmr1 KO suggest weak callosal connectivity in adults (Haberl et al., 2015; Zerbi et al., 2018). Weak callosal synapses persist in adults with postsynaptic Fmr1 deletion, suggesting a deficit in maturation or AMPAR insertion/stability, as opposed to developmental delay (Figure 2F–G). Reduced callosal axon diameter is observed in adult Fmr1 KO which may be a consequence of weak or immature callosal synapses and contribute to reduced interhemispheric coherence in FXS (Haberl et al., 2015).

Neocortical pyramidal neurons integrate excitatory synaptic inputs from local and long-range circuits, including ipsilateral and contralateral cortical areas (Feldmeyer, 2012; Gerfen et al., 2018). An imbalance in local and long-range connectivity has been hypothesized to contribute to ASD in humans; specifically, hyperconnectivity of local circuits and underconnectivity of long-range circuits or between brain regions (Belmonte et al., 2004; Courchesne and Pierce, 2005; O’Reilly et al., 2017; Rane et al., 2015). If or how ASD genes regulate the balance of local and long-range synaptic connectivity is unknown. Here we demonstrate that postsynaptic FMRP differentially regulates development and/or maintenance of synaptic inputs from local and long-range cortical sources. L2/3 neurons with postsynaptic deletion of Fmr1 have weak and immature callosal synaptic inputs but normal synaptic inputs from L4, adjacent L2/3 and L5 (Figure 6). Bureau et al. (Bureau et al., 2008) and we (Figure 6—figure supplement 1C) find that L4 to L2/3 synapses are weak in the global Fmr1 KO, but we do not observe this with embryonic or postnatal cell autonomous deletion of Fmr1 in L2/3 neurons. Together these results suggest a role for FMRP in presynaptic L4 neurons in synapse development onto L2/3 neurons, which is consistent with the reported deficits in L4 axon morphology in the global Fmr1 KO (Bureau et al., 2008).

Results in L5 also indicate differential regulation of local and long-range cortical connectivity by postsynaptic FMRP. Our previous work using multiple simultaneous recordings of locally connected L5A neurons, revealed hyperconnectivity of Fmr1 KO L5 neurons with their immediate neighbors ( < 40 µm apart) in S1 at 4 weeks of age (Patel et al., 2014). Hyperconnectivity of L5 local subnetworks resulted from deficient developmental pruning between Fmr1 KO L5 neurons and was observed in both the global Fmr1 KO and with postsynaptic Fmr1 deletion. Prefrontal L5 cortical neurons in Fmr1 KO mice are similarly hyperconnected (Testa-Silva et al., 2012). In contrast to local hyperconnectivity, here we find that postsynaptic deletion of Fmr1 in L5 neurons results in weak callosal synaptic inputs. Thus, Fmr1 KO L5 pyramidal neurons are hyperconnected locally and under-connected to contralateral cortex; an effect that is mediated by cell-autonomous and postsynaptic deletion of FMRP. Such an effect may promote the reported imbalances in local and long-range functional connectivity observed in ASD individuals. Postsynaptic deletion of FMRP in L2/3 neurons did not affect synaptic inputs from other layers or between columns within L2/3 (Figure 6) suggesting that FMRP may selectively promote pruning of local connections in L5 neurons. Alternatively, FMRP may prune synaptic connections within very local cortical subnetworks (within 40 µm) in both L2/3 and L5, but not connections between layers and columns.

The molecular mechanisms by which postsynaptic FMRP differentially regulates L4 and callosal inputs to L2/3 neurons is unclear. L4 neurons synapse primarily on basal dendrites of L2/3 neurons, whereas callosal inputs are primarily on apical dendrites (Bosman et al., 2011; Figure 1G). Therefore, localized expression and translational regulation of specific dendritic mRNAs by FMRP at either basal or apical dendritic compartments could differentially affect L4 and callosal inputs. Alternatively, FMRP translational regulation of postsynaptic cell adhesion molecules, such as neuroligins, could differentially impact specific presynaptic inputs based on their expression of binding partners such as neurexin splice variants (Südhof, 2017).

Differential regulation of local and long-range cortical synaptic connectivity is observed with another ASD-risk gene, Myocyte Enhancer Factor 2 C (Mef2c). Loss-of-function mutations in MEF2C, which encodes an activity-dependent transcription factor, are implicated in intellectual disability, ASD and schizophrenia (Assali et al., 2019; Mitchell et al., 2018; Rocha et al., 2016). In contrast to FMRP, postsynaptic deletion of Mef2c in L2/3 neurons results in fewer and weak inputs from local circuits (L4, L2/3, and L5), but strengthened callosal inputs (Rajkovich et al., 2017). The differential regulation of local and callosal synaptic connections by postsynaptic MEF2C and FMRP, albeit in different directions, suggest that imbalances in local and long-range synaptic connectivity may contribute to different genetic causes of neurodevelopmental disorders. Our results also implicate roles for transcription and translational control in the input-specific development of cortical circuits. Bidirectional regulation of local and long-range connectivity by Mef2c in L2/3 or by Fmr1 in L5 could be an effect of homeostasis or competition between local and long-range synaptic connections to maintain optimal cortical circuit function. For example, weakening of callosal synaptic input in Fmr1 KO neurons may be compensatory and an attempt to normalize hyperconnected or hyperexcitable local circuits.

The regulation of local or callosal connectivity by FMRP and MEF2C requires normal sensory experience suggesting that both of these genes function in experience and activity-regulated pathways necessary for cortical circuit development. Sensory deprivation by whisker trimming normalizes callosal synaptic inputs in L2/3 neurons with postsynaptic deletion of Fmr1 (Figure 4) or Mef2c (Rajkovich et al., 2017). Specifically, trimming whiskers contralateral to L2/3 neurons with postsynaptic Fmr1 deletion (postsynaptic deprivation) prevented callosal synaptic weakening. In contrast, ‘presynaptic deprivation’ or trimming whiskers contralateral to ChR2-expressing, callosal projecting neurons, had no effect (Figure 4). This result suggests that sensory-driven patterned activity of postsynaptic Fmr1 KO L2/3 neurons weakens callosal synapses. An intriguing possibility is that an activity-dependent long-term synaptic depression (LTD) process is enhanced at Fmr1 KO L2/3 neurons as observed in hippocampal CA1 (Huber et al., 2002). Sensory deprivation also induces homeostatic synaptic scaling in primary sensory cortices, as measured with spontaneous or miniature (m) EPSCs (Feldman, 2009; Hooks and Chen, 2020). Neither the frequency nor amplitude of mEPSCs was different between neighboring WT and Fmr1 KO L2/3 neurons in mice with normal sensory experience or ‘presynaptic deprivation’. However, with ‘postsynaptic deprivation’, mEPSC amplitude and frequency was increased in Fmr1 KO neurons relative to WT (Figure 4—figure supplement 2) suggestive of homeostatic up scaling in Fmr1 KO neurons that may contribute to callosal input strengthening in this condition.

Bilateral underconnectivity is common in ASD as well as disconnection of other long-range cortical connections with other brain regions such as hippocampus and cerebellum. Bilateral connectivity and synchrony between cortical regions are necessary for speech comprehension, sensory processing, and cognition (Bland et al., 2020; Castro et al., 2014; Friederici et al., 2007; Fries, 2009; Fries, 2015; Panzica et al., 2019), domains impaired in FXS and ASD. Our present findings contribute to the understanding of the cellular and synaptic mechanisms by which ASD-risk genes, such as FMR1, regulate long-range connectivity, how this is coregulated and balanced with local circuit connectivity and interacts with experience-dependent brain development. Such information may contribute to therapies to aid abnormal brain connectivity in ASD.

Fmr1 KO (Fmr1^-/y) and X-linked GFP mice were obtained from Jackson laboratories (Stock No: 003025 and 003116, respectively). Fmr1^fl/fl were obtained from Dr. David Nelson (Baylor College of Medicine) (Mientjes et al., 2006). Mice were maintained on a C57BL/6 J background and reared on a 12 hr light-dark cycle with access to food and water ad libitum. Male Fmr1 WT (Fmr1^+/y) with Fmr1 KO (Fmr1^-/y) littermates were used for experiments and both male and female pups were used for Fmr1 flox (Fmr1^fl/y and Fmr1^fl/fl). All animal experiments were conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) at University of Texas Southwestern Medical Center.

Commercially made AAV9.CMV.HI.eGFP-Cre.WPRE.SV40 (Addgene #105545) and AAV9.CAG.hChR2(H134R)-mCherry.WPRE.SV40 (Addgene #100054) were diluted to a titer ~10¹² vg/mL using sterile saline. Traces of Fast Green FCF dye (Sigma) were added to facilitate visualization of virus spreading. Neonatal mouse pups (P1) were first anesthetized by hypothermia, then fixed on a customized mold and placed on a stereotaxic frame. AAV9.eGFP-Cre (420–560 nL) was delivered to left ventricle at a depth of approximately 1.1 mm underneath skull through a beveled glass pipette using Nanoject II injector (Drummond Scientific, Inc). AAV9.ChR2-mCherry (400 nL) was delivered to superficial layers (0.5 mm underneath skull) of somatosensory cortex in right hemisphere at a speed of 1.2 µL/min using syringe pump (Harvard Apparatus, Inc).

Mice were anesthetized by intraperitoneal (i.p.) injection of Ketamine/Xylazine mixture and decapitated upon irresponsiveness to toe-pinch. Acute coronal slices (300 µm thickness) containing somatosensory barrel cortex were prepared using vibrating microtome (Leica VT1200S). During sectioning, tissue blocks were submerged in ice-cold dissection buffer containing (in mM): 110 choline chloride, 25 NaHCO₃, 25 dextrose, 11.6 ascorbic acid, 2.5 KCl, 1.25 NaH₂PO₄, 3.1 Na-pyruvate, 7 MgCl₂, and 0.5 CaCl₂, continuously aerated with 95%CO₂/5%O₂. For mice older than P23, transcardial perfusion of ice-cold dissection buffer was performed prior to decapitation to increase slice quality. Slices were then transferred to artificial cerebrospinal fluid (ACSF) solution containing (in mM): 125 NaCl, 25 NaHCO₃, 10 dextrose, 2.5 KCl, 1.25 NaH₂PO₄, 2 MgCl₂, and 2 CaCl₂ (aerated with 95%CO₂/5%O2) and recovered at 34 °C for 30 min followed by 30 min at room temperature.

After recovery, slices were transferred to a recording chamber at room temperature and perfused with ACSF aerated with 95%CO₂/5%O₂. Slices were visualized by infrared differential interference contrast (IR-DIC) optics (Olympus BX51W1). Whole cell recordings of L2/3 and L5 pyramidal neurons were obtained using borosilicate pipettes (4–7 MΩ) and a Multiclamp 700 A amplifier (Molecular Devices). Internal solution contained (in mM): 130 K-gluconate, 10 HEPES, 6 KCl, 3 NaCl, 0.2 EGTA, 14 phosphocreatine-tris, 4 Mg-ATP and 0.4 Na-GTP. All recordings were conducted in voltage clamp, holding at –70 mV unless otherwise specified, and data were collected and analyzed using custom Labview programs (Labview 8.6, National Instruments Inc). Spiking patterns upon current injection were used as criteria to identify excitatory neurons. For experiments where spiking was blocked by TTX, rise time to hyperpolarizing current ( > 50 ms) and kinetics of mPSC (width at half height >2 ms) were used as criteria (Povysheva et al., 2006). Excitatory neurons with resting membrane potential < –50 mV and a series resistance <35 MΩ were included in analysis. Voltages were not corrected for junction potential. For simultaneous patch clamp recordings, the distance between the pair of cells (center-to-center) is 10–40 µm.

Mice with either cortical injection of AAV9.ChR2-mCherry or unilateral ventricular injection of AAV9.eGFP-Cre and contralateral cortical injection of AAV9.ChR2-mCherry were used for this experiment. Fluorescence of GFP-positive soma and mCherry-labeled axons were visualized using a fluorescent mercury lamp (Excelitas Technologies Corp.). Slices containing clearly labeled mCherry+ axons from contralateral barrel cortex were used for recording. Slices with somatic infection of ChR2-mCherry due to virus leaked from contralateral hemisphere were discarded to avoid contamination from local inputs. The infection rate of recorded Cre-GFP+ neurons in barrel cortex was 3–5% (Figure 2—figure supplement 1). Only neurons residing in an area with densely labeled callosal axons were subject to recording to achieve a reliable magnitude and reduced variability of light-induced responses. For all LED experiments, responses were evoked by a 2 ms flash from a digitally controlled blue LED (final beam diameter: 350 µm; power: 0.1–4.6 mW; wavelength: 470 nm; M470L4-C1, Thorlabs Inc) through a 40 X water-immersed objective. The LED flash was centered on soma and proximal apical dendrites of recorded neurons.

To measure LED-evoked EPSCs, each cell (or cell pair) was stimulated 3–10 times with 20–30 s intervals. LED power was adjusted to obtain an EPSC amplitude of 100–1000 pA in WT neurons (for cell pairs). The external solution for both LED and sCRACM experiments contained ACSF with 1 µM TTX, 100 µM 4-aminopyridine (4-AP), 10 µM (±)–3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid (CPP) and 100 µM picrotoxin to isolate monosynaptic AMPAR-mediated excitatory inputs. Callosal input strength was measured as the peak amplitude of an average EPSC of 3–10 evoked EPSCs for each neuron.

To measure Sr²⁺ evoked quantal events, each cell pair was stimulated with an LED flash every 30 s, 12–30 times in ACSF containing 4 mM MgCl₂, 4 mM SrCl₂, 10 µM CPP and 100 µM picrotoxin. Events occurring within a 1 s window prior to the LED were defined as spontaneous events and those occurring 50–350 ms post-LED were defined LED-evoked events. The frequency and amplitude of events were analyzed using MiniAnalysis (Synaptosoft).

NMDAR-mediated EPSCs from callosal inputs were pharmacologically isolated in ACSF containing: 3 mM CaCl₂, 0.1 mM MgCl₂, 20 µM DNQX, 20 µM glycine, 100 µM picrotoxin, 1 µM TTX, 100 µM 4-AP. An LED flash was delivered every 30 s, 9 times. Evoked EPSCs were averaged and peak amplitude was measured.

After collection of LED-evoked responses, the objective was switched to 4 X to visualize a broader area. ChR2 expressing axons were then stimulated by a blue laser (1 ms; wavelength: 473 nm; power range: 0.7–16 mW; final beam diameter: 25 µm; CrystaLaser) scanning through a 12 × 12 grid (50 µm spacing) in a pseudorandom order to avoid repeated activation of neighboring locations. The grid was aligned along the pia and centered in the medial-lateral position on the recorded somas. The grid was repeatedly scanned 2–4 times at 40 s intervals. Laser-evoked EPSCs were collected from each grid spot. Most neuron pairs ( > 90%) were homogeneously distributed between 150 and 300 µm from pia surface.

For action potential-dependent callosal synaptic strength measurements, an independent cohort of mice was used. The recording ACSF was similar to sCRACM experiments, except TTX and 4-AP were omitted, and divalent cations were increased (4 mM MgCl₂; 3 mM CaCl₂) and CPP was added to reduce polysynaptic activation of local circuits. For L2/3 neurons, an 8 × 8 grid with 75 × 100 µm x-y spacing was used. For L5 neurons, an 8 × 8 grid with 75 × 125 µm x-y spacing was used.

For sCRACM, EPSCs evoked from two to four laser-stimulation at each grid spot from an individual neuron were averaged. The peak amplitude of the average EPSC (between 5–80 ms after laser onset) was determined as the input strength for that spot. A spatial map of input strengths was then generated for each individual neuron. The spatial maps for all neurons of each genotype were then aligned to soma location, oriented with respect to the pial surface, and averaged to generate an average spatial map of input strengths for each genotype aligned to the soma. A color representation of the average spatial map for each genotype were generated for the figures. Pixel size for the genotype-averaged color maps was halved through pixel interpolation (25 × 25 µm) to provide better spatial resolution for soma alignment. Vertical profiling of input strength was achieved by averaging inputs from each horizontal row and plotting against the vertical distance from soma. Input strength from a specific area was calculated by averaging input strengths within an area for each neuron and averaging according to genotype.

For action-potential-dependent ChR2 activation and circuit mapping, analysis was performed as for sCRACM, except EPSC amplitudes were averaged during 2–30 ms after laser onset to exclude contamination from polysynaptic responses.

LSPS experiments were performed similar to that described previously (Rajkovich et al., 2017; Shepherd et al., 2003). For all experiments, only brain slices with L2/3 apical dendrites parallel to the slice surface were used to ensure preservation of the planar barrel cortical geometry of cross-layer synaptic pathways spanning at least 3 barrel columns. Usually 2–4 brain slices per animal met such criteria. ACSF included 4 mM MgCl₂, 4 mM CaCl₂, and CPP (10 μM) to reduced polysynaptic local circuit activity (Rajkovich et al., 2017; Shepherd et al., 2003), and 4-Methoxy-7-nitroindolinyl-caged-L-glutamate, MNI glutamate (MNI, 0.3 mM, either Tocris-1490 or HelloBio-HB0423). A 1 ms UV laser flash (wavelength: 355 nm; power range: 30–40 mW; final beam diameter: 20 µm; DPSS Lasers Inc) was delivered at individual points within a 16 × 16 grid (50 × 60 µm x-y spacing) in a pseudorandom order. The grid was aligned along the pia surface and centered medial-laterally on the soma location. The entire grid was repeatedly scanned 2–4 times at 40 s intervals. Two to four maps were acquired for each neuron included in all datasets.

For each neuron, a single average map was calculated from acquired LSPS maps, where at each stimulation point the averaged light evoked EPSC area was calculated within a time window of 5–80 ms following the laser pulse. If a response was observed within 5 ms of LSPS and displayed kinetics visibly faster than the longer-latency EPSC then it was considered to be a non-synaptic, ‘direct’ response to uncaging of glutamate on the recorded neuron. Direct responses were removed from the map and not included in analysis. For responses with a major component of monosynaptic transmission and minor contamination from ‘direct’ activation, ‘direct’ response component was subtracted from the whole response by fitting a double-exponential decay equation. An IR-DIC image of the slice with patch pipettes in place and stimulation grid was acquired prior to LSPS for marking soma location and anatomical features of the slice (i.e. barrels, etc.). Finally, a color map for each neuron was created. All individual color maps within genotype were then overlaid upon spatial alignment with respect to the center of the ‘home’ barrel directly beneath recorded L2/3 neurons. Superimposition of the average maps was achieved by (1) transposing each map such that the home barrel center was located at the origin of alignment grid, (2) preserving the medial-lateral orientation of the brain slice, and (3) stretching the home barrel in x and y dimensions to normalize barrel size. Pixel size for the genotype-averaged color maps was halved through pixel interpolation (25 × 30 µm) to provide better spatial resolution for soma alignment. Black pixels within an averaged color map in the figures indicate deleted direct responses or pixels that did not meet the minimum sampling threshold (minimum of n = 8 neurons per stimulation point).

Unilateral whisker trimming was performed daily on mice starting at P15 until the day before experiment (P23-30) using a miniature electric shaver. All whiskers from the trimmed pad were maintained at a length <1 mm. For a litter of mice, half of the mice would be trimmed on the right whisker pad, ipsilateral to the AAV9.ChR2-mCherry injection (postsynaptic deprivation), while the other half would be trimmed on the left whisker pad, contralateral to the AAV9.ChR2-mCherry injection (presynaptic deprivation).

All the statistical tests and graphs were performed using GraphPad Prism 8 (GraphPad Software Inc). Prior to statistical tests of significance, each dataset underwent normality tests (D'Agostino-Pearson and Anderson-Darling) to determine if parametric or non-parametric tests should be used. All statistical tests are two-sided. Only datasets tested to be normally distributed by both tests were subject to parametric statistics – unpaired or paired t-test. Otherwise, Mann-Whitney test or Wilcoxon matched pairs test were used as indicated in the figure legends. For data collected from double-patch experiments, paired statistics were used as each pair of neurons would be considered correlated. For comparing more than two groups of data, two-way ANOVA or mixed effect analysis with multiple comparisons was used. Repeated measures ANOVA was used for vertical profile maps in Figures 1 and 2.