Cortical wiring by synapse type-specific control of local protein synthesis

Abstract

Neurons use local protein synthesis to support their morphological complexity, which requires independent control across multiple subcellular compartments up to the level of individual synapses. Here we identify a signaling pathway that regulates the local synthesis of proteins required to form excitatory synapses on parvalbumin-expressing (PV⁺) interneurons in the mouse cerebral cortex. This process involves regulation of the TSC subunit 2 (Tsc2) by the Erb-B2 receptor tyrosine kinase 4 (ErbB4), which enables local control of mRNA translation in a cell type-specific and synapse type-specific manner. Ribosome-associated mRNA profiling reveals a molecular program of synaptic proteins downstream of ErbB4 signaling required to form excitatory inputs on PV⁺ interneurons. Thus, specific connections use local protein synthesis to control synapse formation in the nervous system.

The diversity of animal behaviors relies on the precise assembly of neuronal circuits, a process in which synapse formation plays a role. In the cerebral cortex, dozens of different types of excitatory glutamatergic pyramidal cells and inhibitory γ-aminobutyric acid-containing (GABAergic) interneurons are wired through distinct connectivity motifs (1, 2). For example, layer 4 excitatory neurons receive inputs from excitatory thalamic neurons and project to layer 2/3 excitatory neurons and feed-forward interneurons (3). Synapse specificity is established during development by dedicated transcriptional programs (4–10), but whether regulation of mRNA translation is also involved in this process remains to be elucidated.

Several pathways control protein synthesis (11, 12), including the mechanistic target of rapamycin complex 1 (mTORC1), a molecular complex composed of mTOR kinase and several other proteins that is activated by nutrients and growth factor signals and inhibited by the proteins Tsc1 and Tsc2 (13, 14). Multiple mTORC1 pathway proteins and the translation machinery have been identified in developing axons (15–17), and local protein synthesis occurs at excitatory and inhibitory synapses in the adult brain (18–24). To what extent local translation is differentially regulated in closely related cell types and at the level of specific synapses during the wiring of cortical circuits is unknown.

Specific synaptic defects in interneurons lacking Tsc2

To explore the role of protein synthesis in the wiring of different cell types in the cerebral cortex, we generated mice in which we deleted Tsc2 from the two largest groups of cortical GABAergic interneurons, parvalbumin (PV) and somatostatin (SST) expressing cells. To this end, we crossed Lhx6-Cre mice, which drives recombination in early postmitotic PV⁺ and SST⁺ interneurons, with mice carrying conditional (i.e., Cre-dependent) Tsc2 alleles (Tsc2^F/F) and a reporter for the visualization of recombined cells (see Methods). We chose these two cell types because although they derive from common progenitors in the medial ganglionic eminence (MGE) and the preoptic area (POA), populate the same layers of the neocortex, and are both reciprocally connected with pyramidal cells (25), they play very different roles in cortical information processing (26). We first confirmed that loss of Tsc2 (fig. S1, A and B) leads to overactivation of mTOR signaling in PV⁺ and SST⁺ interneurons by analyzing the levels of phosphorylated S6 ribosomal protein (P-S6rp), a critical downstream effector of mTORC1 (12). We observed increased levels of P-S6rp in conditional Tsc2 mutants compared to controls (fig. S1, C and D). We also found that the cell size of PV⁺ and SST⁺ interneurons is larger in conditional Tsc2 mutants than in controls (fig. S1, C and E), reinforcing the notion that mTOR signaling is indeed overactive in these cells (27, 28). Consistent with a previous report (29), we also noticed an increased density in an infrequent population of cortical interneurons that co-express PV⁺ and SST⁺ in conditional Tsc2 mutants compared to controls (fig. S2). In contrast, Tsc2 loss does not affect the density or laminar distribution of PV⁺ and SST⁺ interneurons (fig. S2).

We next investigated whether loss of Tsc2 affects synapse formation onto PV⁺ and SST⁺ interneurons in the neocortex. To this end, we assessed the number of excitatory synapses received by cortical PV⁺ and SST⁺ interneurons by quantifying puncta containing vesicular glutamate transporter 1 (VGluT1), a characteristic component of excitatory glutamatergic terminals, and PSD95, the primary scaffolding protein in the excitatory postsynaptic density, on the soma and dendrites of PV⁺ and SST⁺ interneurons (Fig. 1 and fig. S3). We found that the loss of one and, even more so, two Tsc2 alleles in PV⁺ interneurons led to an increase in the density of excitatory synapses received by these cells compared to control littermates (Fig. 1, A and B, and fig. S3, A and B). In contrast, conditional deletion of Tsc2 caused no changes in the density of excitatory synapses received by SST⁺ interneurons (Fig. 1, C and D, and fig. S3C). The size of VGluT1⁺ boutons and PSD95⁺ puncta was not altered by the loss of Tsc2 (fig. S4).

Fig. 1. Differential contribution of Tsc2 to synapse development.

(A) Schematic of synaptic markers analyzed (left). Confocal images (top) and binary images (bottom) illustrating presynaptic VGluT1⁺ puncta (magenta) and postsynaptic PSD95⁺ clusters (cyan) in PV⁺ (yellow) tdTomato⁺ (grey) interneurons from P18-21 control, heterozygous and homozygous conditional Tsc2 mutants. (B) Quantification of the density of VGluT1⁺PSD95⁺ synapses contacting PV⁺ interneurons (control, n =111 cells from 8 mice; heterozygous, n = 70 cells from 4 mice; homozygous, n = 46 cells from 3 mice). (C) Schematic of synaptic markers analyzed (left). Confocal images (top) and binary images (bottom) illustrating presynaptic VGluT1⁺ puncta (magenta) and postsynaptic PSD95⁺ clusters (cyan) in SST⁺ (yellow) tdTomato⁺ (grey) interneurons from P18-21 control, heterozygous and homozygous conditional Tsc2 mutants. (D) Quantification of the density of VGluT1⁺PSD95⁺ synapses contacting SST⁺ interneurons (control, n = 67 cells from 4 mice; heterozygous, n = 41 cells from 3 mice; homozygous, n = 60 cells from 5 mice). (E) Schematic of experimental design (top) and post-recording labelling of neurobiotin (NB, grey) -filled tdTomato⁺ (yellow) cells with PV (magenta) and SST (cyan) (bottom). (F) Example traces of sEPSCs recorded from PV⁺ interneurons from P18-21 control, heterozygous and homozygous conditional Tsc2 mutants. (G) Quantification of the frequency (left) and amplitude (right) of sEPSCs from PV⁺ interneurons (control n =14 cells from 5 mice, Lhx6-Cre;Tsc2^F/+ n=23 cells from 8 mice, Lhx6-Cre;Tsc2^F/F n =14 cells from 6 mice). (H) Example traces of sEPSCs recorded from SST⁺ interneurons from P18-21 control, heterozygous and homozygous conditional Tsc2 mutants. (I) Quantification of the frequency (left) and amplitude (right) of sEPSCs from SST⁺ interneurons (control, n = 9 cells from 5 mice; heterozygous, n = 10 cells from 9 mice; homozygous, n = 26 cells from 7 mice). One-way ANOVA followed by Tukey’s multiple comparisons test or Kruskal-Wallis followed by Dunn’s multiple comparisons test: *P < 0.05, **P < 0.01, ***P < 0.001. The dashed lines in the images shown in A and C outline the surface of the cells. Data are mean ± s.e.m. Scale bar, 10 μm and 1 μm (high magnification) (A, C), and 10 μm (E).

To explore whether the supernumerary excitatory synapses contacting PV⁺ interneurons in conditional Tsc2 mutants represent functional synapses, we analyzed synaptic function by whole-cell recordings. Analysis of miniature excitatory postsynaptic currents (mEPSCs) revealed no differences in the frequency or amplitude of these synaptic events between control and conditional Tsc2 mutants (fig. S5), which indicated that the supernumerary excitatory synapses contacting PV⁺ interneurons lacking Tsc2 could either have altered release properties or be functionally inactive. To distinguish between both possibilities, we assessed release probability by recording paired-pulse ratios in PV⁺ interneurons and found no differences between both genotypes (fig. S5). These observations suggested that the supernumerary excitatory synapses decorating PV⁺ interneurons in conditional Tsc2 mutants might not be mature.

We also observed cell type-specific alterations in intrinsic properties following the conditional deletion of Tsc2 (fig. S6A). These changes led to a reduction in the excitability of PV⁺ interneurons in conditional Tsc2 mutants compared to controls, while no difference was observed for SST⁺ interneurons (fig. S6B). Finally, the differential impact of the loss of Tsc2 in the integration of PV⁺ and SST⁺ interneurons in cortical networks was examined by recording spontaneous excitatory postsynaptic currents (sEPSCs) in both cell types (Fig. 1E). Tsc2 deletion in PV⁺ interneurons led to a significant increase in the frequency of sEPSCs with no changes in amplitude (Fig. 1, F and G), while neither the frequency nor the amplitude of sEPSCs changed in SST⁺ interneurons lacking Tsc2 (Fig. 1, H and I). Altogether, these results revealed that loss of Tsc2 differentially impacts the wiring of PV⁺ and SST⁺ interneurons, suggesting a cell type-specific role for Tsc2 in synapse formation.

We next wondered whether the function of Tsc2 in synapse development might be synapse type-specific. We reasoned that if Tsc2 plays a role in synapse formation in PV⁺ interneurons, loss of Tsc2 should also affect the development of the synapses made by these cells onto pyramidal cells. We focused on PV⁺ basket cell synapses, which target the soma of pyramidal cells and can be identified by the presynaptic expression of synaptotagmin-2 (Syt2) and gephyrin (Geph), a postsynaptic scaffolding protein of GABAergic synapses (30). We found no differences in the density of Syt2⁺Geph⁺ synaptic puncta contacting the soma of pyramidal cells in conditional Tsc2 mutants compared to control littermates (fig. S7). Thus, although the loss of Tsc2 causes a global disruption of mTOR signaling in PV⁺ and SST⁺ cells (e.g., increased cell size and P-S6rp), it seems to affect the wiring of interneurons in a cell type-specific and synapse type-specific manner.

Tsc2 functions downstream of ErbB4 in synapses

Since the loss of Tsc2 function only leads to changes in the excitatory synaptic input of PV⁺ interneurons, we reasoned that Tsc2 activity could be regulated locally by a signaling pathway specific to these synapses and necessary for their formation. In other cellular contexts, Tsc2 activity is inhibited by factors that stimulate cell growth through the phosphorylation of Akt (14, 31). The receptor tyrosine kinase ErbB4, which is enriched in PV⁺ but not SST⁺ interneurons, is required for the formation of excitatory synapses onto PV⁺ interneurons (32–34), activates Akt via PI3K phosphorylation (35), and associates with Tsc2 in synaptosome preparations obtained from the mouse neocortex during synaptogenesis (fig. S8). This led us to hypothesize that ErbB4 functions upstream of Tsc2 to regulate its activity during the development of these specific synapses (Fig. 2A). To begin testing this hypothesis, we analyzed Tsc2 phosphorylation in synaptosomes obtained from the neocortex of control and interneuron-specific Erbb4 conditional mutants (Fig. 2B). These preparations are enriched in synaptosomes (fig. S9 A and B), which are predominantly bipartite (fig. S9, C and D, fig. S10), have both the pre- and postsynaptic sides membrane-enclosed (fig. S9, E and F, fig. S11), and contain synaptic proteins (fig. S9G). We found reduced Akt-mediated phosphorylation of Tsc2 in cortical synapses from Erbb4 conditional mutants compared to controls (Fig. 2, C and D). We also found reduced phosphorylation of the mTORC1 effectors S6rp and eIF4E-binding protein 1 (4E-BP1) in synaptosomes from Erbb4 conditional mutants (Fig. 2, C and D), which indicates that the loss of ErbB4 function decreases mTORC1 synaptic activity due to the overactivation of Tsc2. None of these changes were observed in cytosolic fractions (Fig. 2, E and F), suggesting that ErbB4 is required to modulate Tsc2 signaling specifically at the synapse.

Fig. 2. ErbB4 regulates mTOR at excitatory synapses contacting PV⁺ interneurons.

(A) Hypothetical signaling pathway. P indicates phosphorylation; green arrow activation; red arrow inhibition; dotted arrows indicate indirect regulation. (B) Schematic of experimental design. (C) Phosphorylation and protein expression of Tsc2, S6rp, 4EBP1 and actin assessed by Western blot of cortical synaptic fractions from P30 homozygous conditional Erbb4 mice and their control littermates. (D) Quantification of phosphorylation of Tsc2, S6rp and 4EBP1 normalized to the total expression of the corresponding protein (top). Quantification of expression levels of Tsc2, S6rp and 4EBP1 normalized to actin (bottom) (Tsc2: control, n = 8 mice, homozygous, n = 6 mice; S6rp: control, n = 5 mice, homozygous, n = 4 mice; 4EBP1: control, n = 7 mice, homozygous, n = 5 mice). (E) Phosphorylation and protein expression of Tsc2, S6rp, 4EBP1 and actin assessed by Western blot of cortical cytosolic fractions from P30 homozygous conditional Erbb4 mice and their control littermates. (F) Quantification of phosphorylation of Tsc2, S6rp and 4EBP1 normalized to the total expression of the corresponding protein (top). Quantification of expression levels of Tsc2, S6rp and 4EBP1 normalized to actin (bottom) (Tsc2 and S6rp: control, n = 7 mice, homozygous, n = 6 mice; 4EBP1: control, n = 5 mice, homozygous, n = 4 mice). (G) Confocal images illustrating phosphorylation of S6rp (P-S6rp, grey) in Nrg3⁺ (magenta) PSD95⁺ (cyan) synaptosomes (top) and in VGluT1⁺ (magenta) PSD95⁺ (cyan) synaptosomes (bottom) from P21 homozygous conditional Erbb4 mice and their control littermates. (H) Quantification of P-S6rp staining intensity in Nrg3⁺PSD95⁺ (top) and VGluT1⁺PSD95⁺ synaptosomes (bottom). (I) Relative frequency distribution of P-S6rp staining intensity in Nrg3⁺PSD95⁺ synaptosomes (top) and in VGluT1⁺PSD95⁺ synaptosomes (bottom) (Nrg3⁺PSD95⁺: control, n = 22,063 synaptosomes from 6 mice, homozygous, n = 17,523 synaptosomes from 5 mice; VGluT1⁺PSD95⁺: control, n = 39,395 synaptosomes from 5 mice, homozygous, n = 48,057 synaptosomes from 6 mice). Two-tailed Student’s unpaired t-tests: *P < 0.05. Data are mean ± s.e.m. Scale bar, 1 μm.

We next investigated whether the mTORC1 inactivation observed in Erbb4 conditional mutants occurs at specific cortical synapses. Since ErbB4 is only expressed by specific classes of interneurons in the cerebral cortex (32, 36), we hypothesized that changes would be limited to synapses that contain ErbB4 receptors, such as the excitatory synapses received by PV⁺ basket cells. To test this idea, we isolated and plated cortical synaptosomes and analyzed the phosphorylation of S6rp in postsynaptic structures identified with PSD95. We identified excitatory synapses onto inhibitory neurons using neuregulin 3 (Nrg3), a marker of presynaptic excitatory terminals contacting PV⁺ interneurons (37, 38). We found decreased P-S6rp specifically in Nrg3⁺/PSD95⁺ synaptosomes from Erbb4 conditional mutants compared to controls, but no difference in ErbB4-independent glutamatergic synaptosomes (Fig. 2, G to I) or Syt2⁺/Geph⁺ synaptosomes (fig. S12), which correspond to the synapses made by PV⁺ basket cells onto pyramidal cells. These results indicate that ErbB4 regulates mTOR signaling in excitatory synapses received by PV⁺ interneurons.

To further support the idea that ErbB4 and Tsc2 function in the same signaling pathway controlling the formation of excitatory synapses onto PV⁺ cells, we performed a genetic interaction experiment. We generated Erbb4 conditional mutants carrying a conditional Tsc2 allele to compromise the function of Tsc2 in PV⁺ interneurons lacking ErbB4 and measured the density of excitatory synapses received by these cells. We found that deleting one Tsc2 allele from PV⁺ cells is sufficient to rescue the loss of excitatory synapses found in Erbb4 conditional mutants (fig. S13). Altogether, these results demonstrate that Tsc2 functions downstream of ErbB4 in regulating the excitatory synaptic input of PV⁺ interneurons.

Changes in the synaptic translatome of Erbb4 mutants

To identify the specific targets of ErbB4 that are involved in the development of synapses in PV⁺ interneurons, we first obtained the synaptic translatome of MGE/POA-derived interneurons from control and Erbb4 conditional mutants. To this end, we bred into these mice alleles carrying a mutation in the locus encoding the ribosomal protein L22 (Rpl22) that allows the conditional tagging of ribosomes with a hemagglutinin epitope (HA) tag (39). We then prepared synaptosomes from the neocortex of P15 mice, pulled-down ribosomes from this preparation using anti-HA beads to isolate ribosome-associated mRNA transcripts from MGE/POA-derived-interneurons in control and ErbB4 conditional mutants, and analyzed them by RNA sequencing (Fig. 3A and fig. S14, A to D).

Fig. 3. Synaptic ribosome-associated mRNAs altered in Erbb4 mutants.

(A) Schematic of experimental design. (B) Volcano plot displaying significantly differentially expressed ribosome-associated RNAs (DEG, orange) in P15 cortical synaptosomes from homozygous conditional Erbb4 mice compared to controls. Each dot represents one gene. FC: fold-change (FC > 1.5, P < 0.05). (C) Selected Gene Ontology (GO) Cellular Components (CC) terms significantly enriched in the dataset of downregulated genes in homozygous conditional Erbb4 mutants compared to controls. (D) Synaptic Gene Ontology (SynGO) postsynaptic cellular component categories significantly enriched in the dataset of downregulated genes in homozygous conditional Erbb4 mutants compared to controls. (E) Heatmaps showing the selection criteria for 15 “cell adhesion molecules” genes and 9 “AMPA receptors” genes. The asterisks indicate genes selected for validation.

We found that 70% of the differentially expressed genes were downregulated in Erbb4 conditional mutants compared to controls (Fig. 3B and fig. S14D). Among the downregulated genes, we observed a significant enrichment in genes associated with autism spectrum disorder (ASD) (fig. S15). Gene Ontology (GO) analysis revealed an enrichment in genes involved in processes such as “synapse organization”, “neurotransmitter receptor activity”, and “glutamatergic synapse” (Fig. 3C and fig. S14E), highlighting synaptic alterations in Erbb4 conditional mutants. We then identified genes coding for proteins with postsynaptic localization and/or function using Synaptic GO annotations (40). This analysis revealed that, within the postsynaptic category, the most enriched terms were “postsynaptic specialization” and “postsynaptic membrane” and, more specifically, genes encoding cell adhesion molecules and AMPA receptors (Fig. 3, D and E, and fig. S14F). Using a set of four additional criteria (see Methods), including their relative enrichment in PV⁺ cells, we selected seven candidates for functional validation: five cell adhesion molecules, SynCAM1 (encoded by Cadm1), Nptn, Nlgn3, TrkC (encoded by Ntrk3) and Clstn2, and two AMPA receptor-related proteins, GluA4 (encoded by GriA4) and Stargazin (encoded by Cacng2) (Fig. 3E). We confirmed that cortical PV⁺ interneurons express these mRNAs during synaptogenesis (fig. S16, A and B) and that they are present in the dendrites of these cells (fig. S17; see also Ref. 24). In addition, we found that these proteins cluster at the surface of PV⁺ cells in close apposition to innervating axon terminals expressing the presynaptic ErbB4 ligand Nrg3 (fig. S16, C to E). Altogether, our data reveal dysregulation of postsynaptic molecular complexes in PV⁺ cells lacking ErbB4.

ErbB4 regulates local protein synthesis at the synapse

Having identified ribosome-associated mRNAs that might be critical for the formation of excitatory synapses onto PV⁺ cells, we next wondered whether ErbB4 might regulate this process by modulating the local translation of these transcripts. mTORC1 signaling is critical for the modulation of protein synthesis (12), and our previous results revealed that mTORC1 synaptic activity decreases in the absence of ErbB4 (Fig. 2). ErbB4 could therefore regulate synapse formation by controlling local protein synthesis at the synapse through the inhibition of Tsc2. To test this hypothesis, we assessed whether activation of ErbB4 signaling in cortical synaptosomes increases the translation of the ribosome-associated mRNAs we identified downregulated in ErbB4 conditional mutants. To this end, we treated cortical synaptosomes with a soluble form of the epidermal growth factor (EGF)-like domain of neuregulin to activate ErbB4 receptors (41) (fig. S18, A to C) and examined candidate proteins by Western blot (Fig. 4A). We found increased protein levels for all but one of the downstream candidates of ErbB4 signaling (Fig. 4, B and C). We did not observe this increase when samples were pre-treated with the protein synthesis inhibitor cycloheximide (fig. S18, D and E), demonstrating that ErbB4 signaling induces the translation of these transcripts at the synapse. Altogether, these experiments revealed that ErbB4 regulates the local translation of synaptic proteins.

Fig. 4. ErbB4 regulates the local translation of synaptic proteins.

(A) Schematic of experimental design. (B) Protein expression of SynCAM1, Nptn, Nlgn3, TrkC, Clstn2, GluA4, Stargazin and actin assessed by Western blot of cortical synaptic fractions treated with neuregulin (Nrg) from P21 C57B6 mice. (C) Quantification of expression levels of SynCAM1, Nptn, Nlgn3. TrkC, Clstn2, GluA4 and Stargazin normalized to actin. One-tailed Student’s unpaired t-tests: *P < 0.05, **P < 0.01 (SynCAM1 and TrkC: +BSA, n =11 synaptosomes: +Nrg, n = 11 synaptosomes; Nptn and GluA4: +BSA, n = 9 synaptosomes, +Nrg, n = 10 synaptosomes; Nlgn3 and Clstn2: +BSA, n = 4 synaptosomes, +Nrg, n = 4 synaptosomes; Stargazin: +BSA, n = 7 synaptosomes, +Nrg, n = 8 synaptosomes). Data are mean ± s.e.m.

Mediators of excitatory synapse formation on PV⁺ interneurons

To assess whether the proteins being synthesized downstream of ErbB4 are involved in the formation of excitatory synapses on cortical PV⁺ interneurons, we performed interneuron-specific loss-of-function experiments in vivo using a conditional gene knockdown strategy (42). In brief, we designed conditional short-hairpin RNA vectors against the candidate genes (shCadm1, shNptn, shNlgn3, shNtrk3, shClstn2, shGria4, and shCacng2, with shLacZ as a control) and confirmed their effectiveness in vitro (fig. S19A). We generated adeno-associated viruses (AAV) expressing the most effective shRNA constructs, injected them into the neocortex of Lhx6-Cre neonates, and confirmed their ability to downregulate the expression of the corresponding target genes in vivo (fig. S19B). We then assessed the number of excitatory synapses received by PV⁺ interneurons expressing control and experimental shRNAs (Fig. 5A). We found that, compared to controls, reducing the expression of each of the seven targets in PV⁺ interneurons led to a decrease in the density of excitatory synapses received by these cells (Fig. 5, B to D). These results revealed a complex molecular program regulated by ErbB4 that includes SynCAM1, Nlgn3, TrkC, Clstn2, GluA4 and Stargazin and controls the formation of excitatory synapses onto PV⁺ interneurons.

Fig. 5. ErbB4 targets control excitatory synapse formation on PV⁺ interneurons.

(A) Schematic of experimental design. (B) Confocal images (top) and binary images (bottom) illustrating presynaptic VGluT1⁺ puncta (magenta) and postsynaptic PSD95⁺ clusters (cyan) in PV⁺ interneurons (grey) from P21 Lhx6-Cre mice injected with viruses expressing shRNAs targeting the genes of interest or with a control virus (shLacZ). (C) Quantification of the density of VGluT1⁺PSD95⁺ synapses contacting PV⁺ interneurons in knockdown and control mice. Two-tailed Student’s unpaired t-tests: *P < 0.05, **P < 0.01, ***P < 0.001; shCadm1 (n = 122 cells from 6 mice) and control (n = 113 cells from 6 mice); shNptn (n = 131 cells from 6 mice) and control (n = 119 cells from 6 mice); shNlgn3 (n = 86 cells from 4 mice) and control (n = 68 cells from 4 mice); shNtrk3 (n = 126 cells from 6 mice) and control (n = 105 cells from 5 mice); shClstn2 (n = 120 cells from 6 mice) and control (n = 112 cells from 6 mice); shGria4 (n = 121 cells from 8 mice) and control (n = 93 cells from 5 mice); shCacng2 (n = 115 cells from 6 mice) and control (n = 109 cells from 6 mice). (D) Proportion of synaptic loss in PV⁺ interneurons upon knockdown of ErbB4 downstream targets. Data are mean ± s.e.m. Scale bar, 1 μm.

Discussion

Although local protein synthesis is common to synapses in the adult brain (23, 43), the specificity of mRNA translation in different cell types and even in distinct connectivity motifs of the same neuron remains unexplored. Our work indicates that protein synthesis is regulated in a synapse type-specific manner during synapse formation. Tsc2, a regulator of mTORC1 signaling in multiple cellular contexts (12), promotes the development of excitatory synapses onto PV⁺ cells but not onto SST⁺ interneurons. This specificity is mediated by the activation of ErbB4, which controls excitatory synapse development through the inhibition of Tsc2 and the subsequent induction of a molecular program of mRNA translation involving the synthesis of several cell adhesion and glutamate receptor-related proteins, including TrkC, Clstn2, GluA4 and Stargazin, as well as other molecules previously linked to glutamatergic synapses contacting interneurons (44–46). This synapse type-specific regulation could also involve a redistribution of the protein supply to neighboring maturing synapses, as described for synaptic plasticity (47).

Our results reveal that local translation is involved in synapse formation. Work in Drosophila suggested that the local action of the phosphatase Prl-1 (phosphatase of regenerating liver 1) in synapse formation might be achieved by local translation (48). Here, we found that translation occurs locally in developing synapses in the mammalian neocortex and that disrupting the machinery regulating protein synthesis deregulates synapse formation. Moreover, we demonstrate that many of the proteins identified as being synthesized at developing synapses are involved in synapse formation. Therefore, local translation contributes to brain wiring not only during axon guidance (15–17, 49–51) but also in the final stages of neural circuit assembly.

Altered synthesis of synaptic proteins is a core pathophysiological mechanism in ASD (52–57). While the evidence linking ERBB4 with intellectual disability and ASD is scarce (58, 59), the enrichment of autism-associated genes among the genes downregulated in ErbB4 conditional mutants and the identification of NLGN3 as a target of ErbB4-Tsc2 signaling in the formation of excitatory synapses onto PV⁺ interneurons is intriguing because mutations in TSC2 and NLGN3 are associated with ASD (60–64). The excitation received by PV⁺ cells is modulated during experience (42, 65, 66), suggesting a prominent role in learning and memory. The synapse-specific molecular program unveiled in this study reinforces the idea that this connection is a sensitive hub for maladaptive network responses in neurodevelopmental disorders.

Materials and Methods summary

Conditional mouse mutants from MGE-derived interneurons were generated by crossing Lhx6-Cre mice with Tsc2^F/F or Erbb4 ^F/F mice. To generate cell type-specific knockdown in vivo, shRNAs against genes of interest were first tested for downregulation in vitro. Lhx6-Cre neonates were then injected intracranially with Cre-dependent AAVs expressing the shRNAs with the highest knockdown efficiencies. For histological analyses, brains were fixed in 4% paraformaldehyde (PFA) and sectioned frozen on a sliding microtome (Leica). 40 μm thick sections were used for immunohistochemistry, whereas 30 μm thick sections were used to detect gene expression using the RNAscope Multiplex Fluorescent Assay protocol (ACDBio). Images were acquired with an inverted SP8 confocal microscope (Leica) or an ApoTome microscope (Zeiss) and analysed with Imaris (Bitplane) and custom macros in FIJI (ImageJ). Synaptic function, paired-pulse ratio and intrinsic properties of interneurons were analysed by patch-clamp recordings on acute coronal slices using Mini Analysis (Synaptosoft) and Clampfit.

Synaptosomes were prepared with SynPER reagent (ThermoScientific) from cortical tissue and processed for several downstream experiments, including co-immunoprecipitation, ErbB4 pathway activation and protein synthesis, Western blot, plating and immunofluorescence, and transmission electron microscopy. For co-immunoprecipitation, synaptosomes were incubated overnight with antibodies or isotype controls, added on Dynabeads (Invitrogen), and then extensively washed. For ErbB4 activation, synaptosomes were treated with Nrg-EGF or BSA for control, with or without pre-incubation with cycloheximide. For Western blot, denatured protein extracts were separated by SDS-PAGE and transferred onto PVDF membranes, blocked, and incubated with primary antibodies, HRP-conjugated secondary antibodies, and chemiluminescent substrates. Membranes were then imaged with an Odyssey FC (Li-Cor). Densitometry analyses were performed with Image Studio Lite. For immunofluorescence, synaptosomes were plated on 8-well chamber slides coated with poly-D-lysine and fixed with 4% PFA. Samples were then blocked and incubated overnight with primary antibodies, followed by washes and incubation with secondary antibodies for 2 hours at room temperature. For electron microscopy, synaptosomes were fixed in 2.5% glutaraldehyde. Following washes, postfixation, dehydration and infiltration with Spurr epoxy resin-acetone mixture, samples were sectioned using an ultramicrotome (Leica) and examined on a JEM 1400 Flash transmission microscope (JEOL).

To isolate ribosome-associated RNAs of MGE-derived interneurons from cortical synaptic fractions, RiboTag (Rpl22^HA/HA) mice were crossed with Lhx6-Cre;Erbb4^F/F mice. Following the preparation of synaptosomes and anti-HA pulldown using magnetic beads (Pierce), RNA samples were sequenced using a HiSeq 2500 platform (Illumina). Differential gene expression was performed using DESeq2 on R, and candidate targets were selected using the SynGO tool and ranked for validation using a set of criteria that included expression and enrichment in MGE-derived interneurons.

Materials and Methods

Mice

All experiments were performed following the guidelines of King’s College London Biological Service Unit and of the European Community Council Directive 86.609/EEC. Animal work was carried out under license from the UK Home Office following the Animals (Scientific Procedures) Act 1986. Animals were maintained under standard laboratory conditions on a 12:12 light/dark cycle with water and food ad libitum. Both male and female mice were used indiscriminately throughout the study. Mice carrying loxP-flanked Tsc2 alleles (67) (JAX027458) were crossed with Lhx6-Cre mice (68) to generate the Lhx6-Cre;Tsc2^F/F mouse line. F1 mice were crossed with RCL^tdT mice (69) (JAX7909) to generate the Lhx6-Cre;Tsc2^F/F;RCL^tdT/+ mice. Mice carrying loxP-flanked Erbb4 alleles (70) were crossed with Lhx6-Cre mice to generate Lhx6-Cre;Erbb4^F/F mutants. Erbb4^F/F mice were crossed with Tsc2^F/F mice to generate Erbb4^F/F;Tsc2^F/F mice, which were then crossed with Lhx6-Cre;Erbb4^F/F mice to generate the Lhx6-Cre;Erbb4^F/F;Tsc2^F/+ mice. Erbb4^F/F mice were crossed with Rpl22^HA/HA (RiboTag) mice (39) (JAX011029) to generate Erbb4^F/F;Rpl22^HA/HA mice, that were then crossed with Lhx6-Cre;Erbb4^F/F mice to generate the Lhx6-Cre;Erbb4^F/F;Rpl22^HA/HA mice. Lhx6-Cre;RCL^tdT/+ mice were used to visualize dendrites of MGE/POA-derived interneurons. C57BL/6 mice (Charles River) were used for biochemistry and single-molecule fluorescence in situ hybridization experiments, and CD1 mice (Crl:CD1[ICR], Charles River) were used for in utero electroporation.

Histology

Immunohistochemistry

Mice were deeply anaesthetized with sodium pentobarbital by intraperitoneal injection and transcardially perfused with saline followed by 4% paraformaldehyde (PFA). Brains were postfixed for 2 h at 4 °C, cryoprotected in 15% sucrose followed by 30% sucrose and sectioned frozen on a sliding microtome (Leica SM2010R) at 40 μm. Free-floating sections were permeabilized with 0.25% Triton X-100 in PBS for 1 h and blocked for 2 h in PBS containing 0.3% Triton X-100, 10% serum, and 5% BSA. Sections were then incubated overnight at 4 °C with primary antibodies. The next day, sections were washed in PBS and incubated with secondary antibodies for 2 h at room temperature. When required, sections were counterstained with 5 μM DAPI in PBS. Sections were allowed to dry and mounted in Mowiol/DABCO. All primary and secondary antibodies were diluted in PBS with 0.3% Triton X-100, 5% serum, and 1% BSA. The following primary antibodies were used: guinea-pig anti-VGluT1 (1:2000, Chemicon #AB5905), mouse anti-PSD95 (1:500, NeuroMabs #75-028), chicken anti-PV (1:500, Synaptic Systems #195 006), mouse anti-PV (1:1000, Sigma-Aldrich #P3088), rat anti-SST (1:200, Millipore #MAB354), rabbit anti-dsRed (1:500, Clontech #632496), goat anti-mCherry (1:500, Antibodies Online #ABIN1440057), rabbit anti-Tsc2 (1:250, Cell Signaling Technology #4308), rabbit anti-P-S6rp (1:1000, Cell Signaling Technology #5364), mouse anti-Syt2 (1:250, ZFIN #ZDB-ATB-081002-25), mouse anti-Gephyrin (1:500, Synaptic Systems #147 011), rabbit anti-NeuN (1:500, Millipore #ABN78), rabbit anti-GluA4 (1:500, Millipore #AB1508), chicken anti-SynCAM (1:200, MBL International #CM004-3), goat anti-Nptn (1:500, Invitrogen #PA5-47726), rabbit anti-TrkC (1:500, Cell Signaling Technology #3376), rabbit anti-Cacng2 (1:100, Millipore #07-577), rabbit anti-Nlgn3 (1:400, Synaptic Systems #129 113), rabbit anti-Clstn2 (1:100, MyBioSource #MBS9208953). We used Alexa Fluor-conjugated (Invitrogen), DyLight-conjugated and Cy3-conjugated (Jackson ImmunoResearch) secondary antibodies. For biotin amplification, sections were incubated with biotinylated secondary antibody (1:200, Vector labs #BA-2000) followed by Alexa Fluor-conjugated Streptavidin (Invitrogen).

Single-molecule fluorescence in situ hybridization

All solutions for single-molecule fluorescent in situ hybridization were prepared in RNase-free PBS. Mice were perfused as described above. Brains were postfixed overnight at 4°C, cryoprotected in 15% sucrose followed by 30% sucrose, and sectioned frozen on a sliding microtome at 30 μm. Sections were mounted on RNase-free SuperFrost Plus slides (ThermoFisher) and probed against the candidate genes as well as Parvalbumin (Pvalb) using the RNAscope Multiplex Fluorescent Assay v2 protocol (ACDBio #323110). The following probes were used: Cadm1-C1 (Catalogue #492361), Nptn-C1 (Catalogue #1066021), Nlgn3-C1 (Catalogue #497661), Ntrk3-C1 (Catalogue #423621), Clstn2-C1 (Catalogue #542621), Gria4-C1 (Catalogue #422801), Cacng2-C1 (Catalogue #437221), Pvalb-C2 (Catalogue #421931).

Synaptosome experiments

Synaptosomes preparation

Cortices from both brain hemispheres were rapidly dissected in ice-cold PBS and immediately homogenized in SynPER Synaptic Protein Extraction Reagent (ThermoScientific) complemented with cOmplete protease inhibitors (Sigma-Aldrich) and phosSTOP phosphatase inhibitors (Sigma-Aldrich). Synaptosomes were prepared according to the manufacturer’s instructions and processed for further applications. For Western blot, synaptosomes were denatured at 95 °C for 10 min in Laemmli sample buffer (80 mM Tris-HCl pH 6.8, 100 mM DTT, 8.7% glycerol, 2% SDS, and 0.01% bromophenol blue). Protein concentrations were measured using Pierce 660nm Protein Assay (ThermoFisher).

Co-immunoprecipitation

Cortical synaptosomes from P15 C57BL/6 mice were diluted in 1 ml of co-immunoprecipitation buffer (20 mM Tris-HCl pH 8, 120 mM NaCl, 1% NP40, 1% glycerol, and 1x cOmplete protease inhibitors) and subsequently incubated overnight at 4 °C with 5 μg of one of the following antibodies: rabbit anti-Tsc2 (Cell Signaling Technology #4308), rabbit anti-ErbB4 (Santa Cruz #sc283) or rabbit IgG isotype control (Abcam #ab27478). 25 μl of Dynabeads Protein G slurry (Invitrogen) was washed in co-immunoprecipitation buffer and added to the samples for 3 h at 4 °C with gentle rotation. Beads were washed six times with co-immunoprecipitation buffer, resuspended in Laemmli sample buffer 1x, and co-immunoprecipitation samples were then denatured at 95 °C for 10 min.

Anti-HA pulldown

Cortical synaptosomes from P15 Lhx6-Cre;Rpl22^HA/HA mice were diluted in 1 ml of lysis buffer (20 mM Tris-HCl pH 8, 120 mM NaCl, 1% NP40, 1% glycerol, 1x cOmplete protease inhibitors). 100 μl of anti-HA magnetic beads (Pierce #88837) were washed in lysis buffer and incubated with the samples for 3-4 h at 4°C with gentle rotation. After incubation, beads were washed 6 times in ice-cold lysis buffer, resuspended in Laemmli sample buffer 1x, and samples were denatured at 95 °C for 10 min.

ErbB4 activation

Cortical synaptosomes were incubated at 35 °C with 450 rpm in PBS with 2 mM ATP (Sigma-Aldrich), 125 mM HEPES, 10 mM MgCl₂, 67 μM amino acid mixture (Promega), 1x cOmplete protease inhibitors EDTA-free, and 10 μM MG132 (Sigma-Aldrich). Synaptosomes were treated for 10 min (to assess ErbB4 activation) or 30 min (to assess synthesis of candidate targets) with 500 ng/ml Nrg-EGF (Peprotech #100-03) or 0.1% BSA for control. For cycloheximide treatment, synaptosomes were preincubated for 15 min with 50 μM of cycloheximide (Sigma-Aldrich). The reaction was stopped by adding 1% SDS and the samples were denatured at 95 °C for 10 min in Laemmli sample buffer. Protein concentrations were measured using Pierce 660 nm Protein Assay.

Plating and immunofluorescence

Cortical synaptosomes were incubated for 90 min at 4 °C with gentle shaking on Nunc Lab-Tek 8-well chamber slides (Sigma-Aldrich) coated with poly-D-lysine (ThermoScientific). To stain for membranes, synaptosomes were also incubated with DiI (Invitrogen #V22885). Synaptosomes were then fixed with 4% PFA for 15 min at room temperature and washed in PBS. Samples were permeabilized in PBS 0.5% Triton X-100 for 10 min, blocked with PBS 4% serum for 30 min at room temperature, and then incubated overnight at 4 °C with primary antibodies in PBS 4% serum. The next day, samples were washed in PBS and incubated with secondary antibodies for 2 h at room temperature. The samples were then washed in PBS and mounted in Mowiol/DABCO. The following primary antibodies were used: guinea-pig anti-VGluT1 (1:2000, Chemicon #AB5905), mouse anti-PSD95 (1:500, NeuroMabs #75-028), rabbit anti-P-S6rp (1:1000, Cell Signaling Technology #5364), goat anti-Nrg3 (1:500, Neuromics #GT15220), mouse IgG2a anti-Syt2 (1:250, ZFIN #ZDB-ATB-081002-25), mouse IgG1 anti-Gephyrin (1:500, Synaptic Systems #147 011), anti-synaptophysin (1:100, Abcam #ab14692). We used Alexa Fluor-conjugated (Invitrogen) and Cy3-conjugated (Jackson ImmunoResearch) secondary antibodies.

Transmission Electron Microscopy

Cortical synaptosomes from P21 C57B6 mice were fixed with 2.5% glutaraldehyde in 0.1M sodium cacodylate buffer (pH 7.3) for 3 h at 4°C. They were then washed in 0.1M cacodylate buffer before being postfixed with 1% osmium tetroxide (w/v), 1% potassium ferrocyanide (w/v) for 1.5 h at 4 °C. Samples were then washed thoroughly in distilled water, placed in an aqueous solution of 1% uranyl acetate, and stored overnight at 4°C. The next day, the samples were washed thoroughly in distilled water before dehydrating through a graded acetone series. They were then infiltrated with increasing concentrations of Spurr epoxy resin/acetone mixture before being placed into 100% resin overnight with rotation. Further infiltration with fresh Spurr resin was carried out before they were embedded and polymerized for 24 h at 70°C. Ultrathin sections (50-70nm) were prepared using a UC7 ultramicrotome (Leica), mounted on grids and contrasted using UranyLess (Electron Microscopy Sciences, #22409) and lead citrate (Electron Microscopy Sciences, #22410).

Imaging and analysis

Image acquisition

Confocal images were acquired with an inverted SP8 confocal microscope (Leica) or an ApoTome (Zeiss), using the LAS AF software and the ApoTome function in Zen2 software, respectively. Samples from the same experiment were imaged and analyzed in parallel, using the same laser power, photomultiplier gain and detection filter settings. Imaging PV⁺ and SST⁺ interneurons across cortical layers was performed at 8-bit depth, with 10x objective at 200 Hz acquisition speed. Imaging of synaptic markers and in situ hybridization was performed at 8-bit depth, with 100x objective and 2.2 digital zoom at 200 Hz acquisition speed. Imaging of Tsc2 intensity in cortical tissue was performed at 12-bit depth, with a 100x objective and 2.2 digital zoom at 200 Hz acquisition speed. Imaging of P-S6rp intensity in cortical tissue and synaptosomes was performed at 12-bit depth, with a 40x objective and 200 Hz acquisition speed, or a 63x objective, 2.2 digital zoom and 200 Hz acquisition speed, respectively. Imaging of other immunostainings on plated synaptosomes was performed at 8-bit depth, with a 63x objective, 2.2 digital zoom and 200 Hz acquisition speed. Imaging of post-recording immunostaining was performed at 8-bit depth, with a 20x objective at 200Hz acquisition speed. Electron microscopy samples were examined on a JEM 1400 Flash (JEOL) transmission microscope operated at 80kV, and images were acquired with a Flash Camera (JEOL) using Limitless Panorama (LLP) software.

Image analysis

Analyses in the conditional Tsc2 mutants were carried out in layers 2/3 of the somatosensory cortex. Analysis of cell volume and P-S6rp intensity was performed using Imaris 8.1.2 (Bitplane). Background subtraction and Gaussian filtering were first applied in all channels. Cell somas were reconstructed automatically as three-dimensional isosurfaces with the “create surface” tool using the PV staining (for PV⁺ interneurons) or the tdTomato staining (for SST⁺ interneurons). The volume of the reconstructed somas and the intensity of the P-S6rp channel (as integrated density corrected for soma size) inside each reconstructed soma were then quantified automatically.

Analysis of Tsc2 intensity was performed using a custom macro in FIJI (ImageJ). Background subtraction, Gaussian blurring, smoothing, and contrast enhancement were first applied in all channels. Cell somas were drawn automatically or manually based on the tdTomato staining to create a mask of the soma surface. The intensity of the Tsc2 channel (as integrated density corrected for soma size) was then quantified automatically for each reconstructed soma.

We used a custom macro in FIJI to analyze cluster/synaptic densities, as previously described (38, 42). Background subtraction, Gaussian blurring, smoothing, and contrast enhancement were first applied in all channels. Cell somas were drawn automatically or manually based on intensity levels of PV staining (for PV⁺ interneurons), tdTomato staining (for SST⁺ interneurons) or NeuN staining (for pyramidal cells) to create a mask of the soma surface and measure its perimeter. Cell dendrites from PV⁺ or SST⁺ cells were drawn manually based on intensity levels of tdTomato staining to create a mask of the dendrite surface and measure its perimeter. Presynaptic boutons and postsynaptic clusters were detected automatically based on thresholds of intensity. Thresholds for the different synaptic markers were selected from a set of random images before quantification, and the same threshold was applied to all images from the same experiment. The “Analyze Particles” and “Adjustable Watershed” tools were applied to the synaptic channels, and a mask was generated with a minimum particle size of 0.05. The soma or dendrite mask and the corresponding synaptic masks were merged to quantify the number of puncta contacting the soma and dendrite. Puncta were defined as presynaptic boutons when they were located outside the soma or dendrite and had ≥ 0.1 μm² colocalizing with the soma or dendrite perimeter. Puncta were defined as postsynaptic clusters contained inside a soma or dendrite and had ≥ 0.2 μm² colocalizing with the soma or dendrite perimeter. The size of VGluT1⁺ boutons and PSD95⁺ clusters were quantified automatically for each cell using the same images analyzed for synaptic density. Synapses were defined as presynaptic boutons and postsynaptic clusters contacting each other with a colocalisation area of ≥ 0.03μm² of their corresponding masks.

Analysis of P-S6rp intensity in synaptosomes was performed using a custom macro in FIJI. Background subtraction, Gaussian blurring, smoothing, and contrast enhancement were first applied in all channels. Pre- and postsynaptic puncta were detected automatically based on thresholds of intensity (thresholds for the different synaptic markers were selected from a set of random images before quantification, and the same threshold was applied to all images from the same experiment). The “Analyze Particles” and “Adjustable Watershed” tools were applied to the synaptic channels, and a mask was generated with a minimum particle size of 0.05. The two synaptic masks were merged to automatically quantify the number of synaptosomes. Synaptosomes were defined as pre- and postsynaptic puncta contacting each other with a colocalization area of ≥ 0.03 μm² of their corresponding masks. The intensity of the P-S6rp channel was automatically calculated in each defined synaptosome.

Analysis of in situ hybridization was performed using a custom macro in FIJI. Background subtraction, Gaussian blurring, smoothing, and contrast enhancement were first applied in all channels. Cell somas were drawn automatically or manually based on intensity levels of Pvalb in situ hybridization signal to create a mask of the soma surface and measure its perimeter. RNA particles were detected automatically based on thresholds of intensity (thresholds for the different target RNA were selected from a set of random images before quantification, and the same threshold was applied to all images from the same experiment). The “Analyze Particles” and “Adjustable Watershed” tools were applied to the target RNA channels, and a mask was generated with a minimum particle size of 0.05. The soma and target RNA masks were merged to automatically quantify the number of particles inside the soma.

Analysis of target staining intensity was performed using a custom macro in FIJI. Background subtraction, Gaussian blurring, smoothing, and contrast enhancement were first applied in all channels. Cell somas were drawn manually to create a mask of the soma surface. The intensity was measured automatically as raw integrated density.

Analysis of the density and laminar distribution of PV⁺ and SST⁺ interneurons was performed manually in FIJI from images of the somatosensory cortex. Cortical layers were identified using DAPI staining and their distinct histological characteristics.

Analysis of EM images was performed manually in FIJI. Intact synaptosomes were identified based on the structural characteristics of bipartite synaptosomes: a presynaptic bouton, defined by the presence of vesicles, attached to a membrane-enclosed postsynaptic density.

Western blotting

10-40 μg of denatured protein extracts were separated by SDS-PAGE using 10% acrylamide gels or 4-15% Mini-Protean precast gels (Bio-Rad) for 2 h at 120 V and transferred onto methanol-activated PVDF membranes at 350 mA for 2 h on ice. Membranes were blocked with either 5% non-fat milk or 5% BSA in TBS-T (20mM Tris-HCl pH 7.5, 150mM NaCl, and 0.1% Tween20) for 1 h at room temperature. Membranes were then probed with primary antibodies in blocking buffer overnight at 4 °C. The next day, membranes were washed in TBS-T and incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. Depending on the amount of protein of interest in the samples, membranes were incubated with either Immobilon Western Chemiluminescent HRP substrate (Millipore), SuperSignal West Pico or West Femto Chemiluminescent Substrates (ThermoFisher). Protein levels were visualized by chemiluminescence with an Odyssey FC (Li-Cor) and quantified with Image Studio Lite. Densitometry of the band of interest was normalized to that of actin for quantification. Densitometry of the band of interest was normalized to that of the corresponding total protein for phosphorylated proteins.

The following primary antibodies were used: rabbit anti-P-ErbB4 (Tyr1284, 1:1000, Cell Signaling Technology #4757), rabbit anti-ErbB4 (1:1000, Cell Signaling Technology #4795), rabbit anti-P-S6rp (Ser240/244, 1:5000, Cell Signaling Technology #5364), rabbit anti-S6rp (1:5000, Cell Signaling Technology #2217), rabbit anti-P-Tsc2 (Thr1462, 1:1000, Cell Signaling Technology #3617), rabbit anti-Tsc2 (1:1000, Cell Signaling Technology #4308), rabbit anti-P-4EBP1 (Thr37/46, 1:1000, Cell Signaling Technology #2855), rabbit anti-4EBP1 (1:1000, Cell Signaling Technology #9644), mouse anti-actin-peroxidase (1:20 000, Sigma-Aldrich #A3854), mouse anti-PSD95 (1:1000, NeuroMabs #75-028), rabbit anti-NeuN (1:1000, Millipore #ABN78), mouse anti-GAPDH (1:1000, Abcam #ab8245), rabbit anti-NMDAR2A (1:1000, Chemicon #AB1555P), rabbit anti-synaptophysin (1:500, Abcam #ab14692), mouse anti-HA (1:500, ThermoFisher #26183), mouse anti-p53 (1:1000, Cell Signaling Technology #2524), rabbit anti-GluA4 (1:1000, Millipore #AB1508), chicken anti-SynCAM (1:1000, MBL International #CM004-3), goat anti-Nptn (1:1000, Invitrogen #PA5-47726), rabbit anti-TrkC (1:1000, Cell Signaling Technology #3376), rabbit anti-Cacng2 (1:1000, Millipore #07-577), rabbit anti-Nlgn3 (1:1000, Synaptic Systems #129 113) and rabbit anti-Clstn2 (1:1000, MyBioSource #MBS9208953). The following HRP-conjugated secondary antibodies were used: goat anti-mouse-peroxidase (1:5000, Invitrogen #31444), donkey anti-rabbit-peroxidase (1:5000, Invitrogen #SA1-200), donkey anti-chicken-peroxidase (1:5000, Invitrogen #SA1-300), rabbit anti-goat-peroxidase (1:5000, Abcam #ab6741). For immunoprecipitation samples, the following secondary antibodies were used: anti-mouse-HRP (1:1000, Abcam #ab131368) and anti-rabbit-HRP (1:1000, Abcam #ab131366).

Electrophysiology

Slice preparation and patch-clamp recordings

Mice were anaesthetized with an overdose of sodium pentobarbital before decapitation. Coronal slices of 300 μm were cut using a VT1200S vibratome (Leica) in ice-cold artificial cerebrospinal fluid (ACSF) containing 87 mM NaCl, 11 mM glucose, 75 mM sucrose, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 0.5 mM CaCl₂, 7 mM MgCl₂, 26 mM NaHCO₃, oxygenated with 95% O2 and 5% CO2, and incubated for 1 h at 32 °C and subsequently at room temperature. Slices were transferred to the recording setup 15 min prior to recording and incubated at 32 °C while being continuously oxygenated with 95% O2 and 5% CO2 in recording ACSF containing: 124 mM NaCl, 1.25 mM NaH₂PO₄, 3 mM KCl, 26 mM NaHCO₃, 10 mM Glucose, 2 mM CaCl₂, 1 mM MgCl₂. Pipettes (3–5 MΩ) were made from borosilicate glass capillaries using a PC-10 pipette puller (Narishige) and filled with intracellular solution containing 130 mM potassium-gluconate, 5 mM KCl, 10 mM HEPES, 2.5 mM MgCl₂, 4mM Na2ATP, 0.4mM Na3GTP, 10 mM sodium-phosphocreatine, 0.6 mM EGTA (pH 7.2–7.3, 285–295mOsm) supplemented with either 0.2 mg/ml neurobiotin (Vector Laboratories) for current-clamp recordings, or 115 mM CsMeSO₃, 20 mM CsCl, 10 mM HEPES, 2.5 mM MgCl₂, 4 mM Na2ATP, 0.4 mM Na3GTP, 10 mM sodium-phosphocreatine, 0.6 mM EGTA (pH 7.2–7.3, 285–295 mOsm), and 0.2% neurobiotin for voltage-clamp recordings. Recordings were carried out in layers 2/3 of the somatosensory cortex. Traces were recorded using a Multiclamp 700B amplifier (Molecular Devices), sampled at 50 kHz and filtered at 3 kHz. mEPSCs were recorded in the presence of 1 μM Tetrodotoxin and 100 μM Picrotoxin (PTX) (both Tocris Bioscience) at a holding potential of −60mV. sEPSCs were recorded in the presence of 100 μM PTX at a holding potential of −60mV. Both mEPSCs and sEPSCs were analyzed using Mini Analysis (Synaptosoft). Intrinsic properties were analyzed using Clampfit 10.2. Paired pulse ratios (PPR) were recorded in the presence of 100 μM of PTX. Evoked excitatory postsynaptic currents (eEPSCs) were generated with a tungsten stimulator electrode (WPI) connected to an ISO-STIM 01D stimulator (NPI) and placed approximately 100 μm from the recorded cell. Ten sweeps of 2 pulses with an inter-stimulus interval of 50 ms were applied with an inter-sweep interval of 10s. Recordings were excluded if the first peak had an amplitude <50pA. PPR traces were analyzed using Clampfit 10.2. PPR was calculated as amplitude 2/amplitude 1.

Data analysis

Intrinsic properties were defined and analyzed as follows: RMP, the membrane potential recorded immediately after entering whole-cell configuration; TC, fitting of a single exponential to the response to a 10 mV hyperpolarizing step; IR, an average of the resistance calculated from 5 increasing 10 pA current steps; Ih sag, calculated from the average membrane potential at the end of a 1s current pulse, initially hyperpolarizing the cell to –100 mV; APmax, the maximum value reached by the AP; MFF, Maximum AP frequency elicited by increasing 100pA steps; fAHP, the potential difference between the threshold and minimum AP-value; and Capacitance, calculated from TC and IR.

Post-recording immunohistochemistry

After recording, slices were immediately fixed in PFA 4% for 30 min at 4 °C, washed in PBS and permeabilized with PBS containing 0.4% Triton X-100 for 1 h at room temperature. Slices were blocked in PBS with 0.3% Triton X-100, 10% serum, and 5% BSA for 3 h at room temperature and incubated with primary antibodies overnight at 4 °C. Slices were then washed in PBS and incubated with secondary antibodies overnight at 4°C. The next day, slices were washed in PBS, counterstained with 5 μM DAPI in PBS, and mounted in Mowiol/DABCO. All primary and secondary antibodies were diluted in PBS containing 0.3% Triton X-100, 5% serum, and 1% BSA. The following primary and secondary antibodies were used: rat anti-SST (1:200, Millipore #MAB354), chicken anti-PV (1:500, Synaptic Systems #195 006) rabbit anti-dsRed (1:500, Clontech #632496), goat anti-rat 647 (1:400, Invitrogen #A-21247), donkey anti-chicken 405 (1:200, Jackson ImmunoResearch #703-475-155) and donkey anti-rabbit Cy3 (1:500, Jackson ImmunoResearch #711-165-152). Neurobiotin was revealed with Streptavidin 488 (1:400 Invitrogen #S11223).

RNA isolation by RiboTrap pulldown

Cortices from P15 Lhx6-Cre;Erbb4^+/+;Rpl22^HA/HA and Lhx6-Cre;Erbb4^F/F;Rpl22^HA/HA mice were rapidly dissected in ice-cold RNase-free PBS and immediately homogenized in SynPER Synaptic Protein Extraction Reagent (ThermoScientific) supplemented with 1x cOmplete EDTA-free protease inhibitors (Sigma-Aldrich), 1 mg/ml Heparin (Sigma-Aldrich), 200 U/ml RNAsin (Promega), 100 μg/ml cycloheximide, and 1 mM DTT (Sigma-Aldrich). Synaptosomes were then prepared according to the manufacturer’s instructions. The final synaptosome pellets were resuspended in supplemented SynPER, and Igepal-CA630 (Sigma-Aldrich) was added to the samples to a final concentration of 1%. 100 μl of anti-HA magnetic beads (Pierce #88837) were washed in supplemented SynPER and added to the samples for 3-4 h at 4°C with gentle rotation. After incubation, beads were washed 3 times in ice-cold washing buffer (300 mM KCl, 1% Igepal-CA630, 50 mM Tris-HCl pH 7.5, 12mM MgCl₂, 1mM DTT and 100 μg/ml cycloheximide) and eluted in 350 μl of RLT Plus buffer from the RNAeasy Plus Micro kit (Qiagen) supplemented with 2-mercaptoethanol (Bio-Rad).

RNA sequencing and differential expression analysis

RNA purification, quantification, and quality check

RNA purification of immunoprecipitated RNA was performed using the RNeasy Plus Micro kit (Qiagen) following the manufacturer’s protocol. RNA quality was checked on a Bioanalyzer instrument (Agilent Technologies) using an RNA 6000 Pico Chip. Only RNA samples with RNA integrity number (RIN) values >7.4 were used for library preparation and sequencing.

Library preparation and Illumina sequencing

Four biological replicates were analyzed for each genotype. Library preparation and RNA sequencing experiments were performed by the Genomic Unit of the Centre for Genomic Regulation (CRG, Barcelona, Spain). Library preparation for all samples was performed with 1 ng of total RNA using the SMARTer Ultra Low RNAkit for ultra-low RNA. Samples were then sequenced paired-end using an Illumina HiSeq 2500 platform to approximately 60-80 million mapped reads per sample.

Quality control and differential gene expression analysis

RNA sequencing data analysis was performed by GenoSplice (www.genosplice.com). Sequencing, data analysis, reads repartition, and insert size estimation were performed using FastQC, Picard-Tools, Samtools and rseqc. Using STAR v2.4.0 (71) on the mm10 Mouse genome assembly with an average of 79.8±3.4% of uniquely mapped reads. The analysis of gene expression regulation was performed as described previously (72). In brief, for each gene present in the Mouse Fast DB v2021_1 annotation, reads aligning on constitutive regions (that are not prone to alternative splicing) were counted. Based on these read counts, normalization and differential gene expression were performed using DESeq2, as described before (73), on R v3.2.5. Genes were considered as expressed if their Reads Per Kilobase of transcript per Million mapped reads (RPKM) value was >99% of the background Fragments Per Kilobase of transcript per Million mapped reads (FPKM) value based on intergenic regions. Only genes expressed in at least one of the two compared experimental conditions were further analyzed. Results were statistically significant for p-values ≤ 0.05 and fold changes ≥ 1.5. Pathway and GO analyses were performed using WebGestalt (74) and were considered significant for p-values ≤ 0.05.

Enrichment analysis

A list of highly confident autism-associated genes (genes that score S, 1 or 2 and reported > 5 times) was retrieved from the SFARI Gene database (https://gene.sfari.org). To identify which of these human genes overlap with the list of genes downregulated in Erbb4 conditional mutants, we first mapped homologous genes using Ensembl’s BioMart database (https://www.ensembl.org/info/data/biomart/index.html) as a reference. Fold enrichment was calculated by comparing the proportion of overlapping genes to the expected chance of finding an autism-associated gene in the human coding genome. The statistical significance of the enrichment was determined using a two-sided Fisher’s exact test.

Candidate selection criteria

From the list of differentially expressed genes downregulated in Erbb4 conditional mutants compared to controls, we used the Synaptic Gene Ontology tool (SynGO, https://syngoportal.org) to highlight genes coding for proteins with synaptic localization and/or function (40). Among the “postsynapse” cellular component annotation, the terms “postsynaptic specialization” and “postsynaptic membrane” were significantly enriched when using the following settings: stringency = remove annotations with only proteomics evidence; minimum gene counts per term = 5. Merging both lists, we removed genes that were not expressed in PV⁺ interneurons at P10 (FPKM < 5 in Synapdomain (7); https://devneuro.org/cdn/synapdomain.php). The resulting list of genes was classified into further subcategories according to the gene ontology analysis or a manual MEDLINE search. The two subcategories with the highest number of genes were “cell adhesion molecules” and “AMPA receptor-related molecules”. We used a set of four additional criteria to select candidates for functional validation: (1) expression in PV⁺ interneurons at P10 (Synapdomain) (7), (2) expression in PV⁺ interneurons at P30 (SpliceCode) (75), available at https://scheiffele-splice.scicore.unibas.ch, (3) enrichment in PV⁺ versus SST⁺ interneurons (SpliceCode), and (4) literature on interneuron connectivity (MEDLINE search for “gene name” and “synapse” and “interneuron”). Each criterion was scored, and the sum of the four scores was used to rank the genes. In the “cell adhesion molecules” category, 8 genes had 3 or more points. Out of these 8 genes, we found three calsyntenin-encoding genes and two neuroligin-encoding genes. To avoid validating genes belonging to the same protein family, we selected Clstn2 and Nlgn3 for functional analyses. The other three selected genes were Nptn, Cadm1, and Ntrk3. In the “AMPA receptor-related molecules” category, the two top-ranking genes were Gria4 and Cacng2.

AAV experiments

Generation of AAV expression vectors

To downregulate the expression of the target genes in vivo, we took advantage of a cell type-specific knockdown strategy using a Cre-dependent cassette pDIO-DSE-mCherry-PSE-MCS as previously reported (7). We generated five shRNA constructs against each target gene using the Block-iT RNAi Designer web tool (Thermo Scientific) to identify target sequences within the coding region of the genes that show high knockdown effectivity in silico. The target sequences for each shRNA construct were the following: shLacZ (AAA TCG CTG ATT TGT GTA GTC), shCadm1 (shRNA-1: GGG AGG AGA TTG AAG TCA ACT; shRNA-2: GGT TCA AAG GGA ACA AGG; shRNA-3: GCA GTA TAA ACC GCA AGT GCA; shRNA-4: GCA TTT GAG TTA ACG TGT GAA; shRNA-5: GGC CAA ACC TGT TCA ATA), shNptn (shRNA-1: GCT GAG GAT TCA GGC GAA TAC; shRNA-2: GCA GGA TGC TAT GAT GTA CTG; shRNA-3: GGT GTG TTT GAG ATT TCT; shRNA-4: GGC TGA AAT CAT CCT TGT; shRNA-5: GGT GAT CAT TGT GTA TGA), shNlgn3 (shRNA-1: GCG AGG ACT TAG CGG ATA ATG; shRNA-2: GGA TAT GGT GGA TTG TCT TCG; shRNA-3: GCT ATG GCT CAC CTA CCT ACT; shRNA-4: GCA TGA CAT GTT CCA CTA TAC; shRNA-5: CAC CAT CAC TAT GAT TCC TAA), shNtrk3 (shRNA-1: GCC AGA GCC TTT ACT GCA TCA; shRNA-2: GGA CCA ATG TAC ATG CCA TCA; shRNA-3: GGA ACA TTG CAT TGA GTT TGT; shRNA-4: GCA CAG ATT TCT TTG ACT TTG; shRNA-5: GGC ATC ACT ACA CCA TCA TCG), shClstn2 (shRNA-1: GCA TCA CTA TGC CCT GTA TGT; shRNA-2: GGA GCA ACA TAT GAA CCA TAC; shRNA-3: GGA TAA AGT ATC ACT TCA ACC; shRNA-4: GGA CTT GGA TCC AAG GCA AGA; shRNA-5: GCA GGA GTC ATA AAC ATT TGG), shGria4 (shRNA-1: GCA ACT AGA GCT TGA; shRNA-2: GCA CGT CAA AGG CTA CCA TTA; shRNA-3: GGA TCT GAA ACA CCT CCA AAG; shRNA-4: GGA ATT GAC ATG GAG AGA ACA; shRNA-5: GCA ATG ACA CAG CTA TCG), shCacng2 (shRNA-1: GCC TCG AAG GGA ACT TCA AAG; shRNA-2: GCA AGC AAA TCG ACC ACT TTC; shRNA-3: GCT GAC ACC GCA GAG TAT TTC; shRNA-4: GCT GGC CGT GCA CAT GTT TAT; shRNA-5: GGA CAG GGA TAA CAG CTT TCT). For in vitro validation of knockdown effectivity of the shRNA constructs, the full-length coding regions of the target genes tagged with the hemagglutinin (HA)-tag (TAC CCC TAC GAC GTG CCC GAC TAC GCC) were designed and cloned in the pcDNA3.1(+) expression vector using GeneArt Gene Synthesis Services (Thermo Fisher). The reference coding sequences and the HA-tag insertion sites for the different target genes were as follows: HA-Cadm1 (NM_207676.2; HA-tag inserted after amino acid 420, alanine), HA-Nptn (NM_001357751.1; HA-tag inserted after amino acid 393, asparagine), HA-Nlgn3 (NM_172932.4; HA-tag inserted after amino acid 34, threonine), HA-Ntrk3 (NM_008746.5; HA-tag inserted after amino acid 1, methionine), HA-Clstn2 (NM_022319.2; HA-tag inserted after amino acid 22, glycine), HA-Gria4 (NM_019691.5; HA-tag inserted after amino acid 3, isoleucine), HA-Cacng2 (NM_007583.2; HA-tag inserted after amino acid 105, serine).

Cell culture and transfection

HEK293T and COS-7 cells were cultured in Dulbecco’s Modified Eagle’s medium supplemented with 10% fetal bovine serum, 2 mM glutamine, penicillin (50 units/ml) and streptomycin (50 g/ml). The cultures were incubated at 37 °C in a humidified atmosphere containing 5% CO₂. Cells were transfected using FuGENE transfection reagent (ThermoFisher) at a 3:1 DNA:FuGENE ratio. Cells were seeded on 12-well plates in triplicate experiments and co-transfected with an shRNA-expressing construct, the expression plasmid for the corresponding HA-tagged gene, and a plasmid for the expression of Cre recombinase. Co-transfection with a shLacZ construct was performed as a control for each target gene. 72 h after transfection, cells were scraped in RIPA lysis buffer containing 50 mM Tris-HCl pH 8, 150 mM NaCl, 0.5% sodium deoxycholate, 1% NP40, 0.1% SDS and 1x cOmplete protease inhibitors. Samples were denatured at 95 °C for 10 min in Laemmli sample buffer, and protein concentrations were measured using Pierce 660 nm Protein Assay.

AAV production

HEK293T cells were seeded on 15-cm plates and co-transfected with packaging plasmids AAV-ITR-2 genomic vectors (7.5 μg), AAV-Cap8 vector pDP8 (30 μg, PlasmidFactory GmbH, Germany, #pF478) using PEI (Sigma) at a 1:4 DNA:PEI ratio. 72 h post-transfection, supernatants were incubated with ammonium sulphate (65 g/200 ml supernatant) for 30 min on ice and centrifuged at 4000 rpm for 45 min at 4 °C. Transfected cells were harvested and lysed (150 mM NaCl, 50 mM Tris pH 8.5), followed by three freeze-thaw cycles and benzonase treatment (50U/mL; Sigma) for 1 h at 37 °C. Filtered AAVs (0.8 μm and 0.45 μm MCE filters) from supernatants and lysates were run on an iodixanol gradient by ultracentrifugation (Vti50 roto, Beckmann Coulter) at 50 000 rpm for 1 h at 12 °C. The 40% iodixanol fraction containing AAVs was collected, concentrated using 100 kDa-MWCO Centricon plus-20 and Centricon plus-2 (Merck-Millipore), aliquoted and stored at -80 °C. The number of genomic copies was determined by quantitative real-time PCR using the following primers targeting the WPRE sequence (Fw: GGC ACT GAC AAT TCC GTG GT, and Rv: CGC TGG ATT GAG GGC CGA A). After in vitro validation, the following shRNA constructs were selected for AAV production based on their high knockdown efficiency: shCadm1 (shRNA-1 and shRNA-5), shNptn (shRNA-1 and shRNA-3), shNlgn3 (shRNA-2), shNtrk3 (shRNA-3 and shRNA-4), shClstn2 (shRNA-1 and shRNA-4), shGria4 (shRNA-2), and shCacng2 (shRNA-2 and shRNA-4).

Surgeries

Stereotaxic injections

Stereotaxic injections were performed as previously described (7). For intracranial injections of shRNA-expressing AAV vectors, P2-3 Lhx6-Cre pups were anaesthetized with isoflurane and mounted on a stereotactic frame using a 3D printed isoflurane mask. A volume of 300nl was injected in the left hemisphere, at a rate of 100 nl/min, with viral titers ranging from 2.10¹² to 4.10¹³. The following coordinates were used to target MGE/POA-derived interneurons in layer 2/3 of the somatosensory cortex: anteroposterior +1.6/1.7 mm; mediolateral -1.8/-1.9 mm; depth -0.3 mm.

In utero electroporation

In utero electroporation was performed as previously described (38). Timed-pregnant CD1 females were deeply anaesthetized with isoflurane. Buprenorphine (Vetergesic, Ceva Animal Health Ltd) was administered for analgesia via subcutaneous injection, and ritodrine hydrochloride (Sigma-Aldrich #R0758) was applied to the exposed uterine horns to relax the myometrium. The DNA solution was mixed with Fast Green (Roche #06402612001), and 1-2 μl of the solution was injected into the lateral ventricle of embryonic day 14.5 embryos. To target DNA into cortical pyramidal cell progenitors of the subventricular zone, five electric pulses (45 V for 50 ms, with 950 ms intervals) were delivered through electrodes (CUY650P3, Nepa Gene) connected to an electroporator (NEPA21 Super Electroporator, Nepa Gene).

Statistical analyses

All statistical analyses were performed using Prism 9 (GraphPad Software). No statistical methods were used to predetermine sample sizes. Sample sizes were chosen based on previous publications in the field. Biological replicates (n values are different samples derived from different brains from different litters) were analyzed to assess the biological variability and reproducibility of data. Experimental mice from all genotypes or conditions were processed together. Samples were tested for normality using the Shapiro–Wilk normality test. Unless otherwise stated, data were analyzed by t-test or ANOVA followed by post hoc Tukey’s analysis for comparisons of multiple samples. Differences were considered significant when p values < 0.05. Data are presented as mean ± s.e.m. Statistical details of experiments are described in figure legends.

One sentence summary.

Local protein synthesis is regulated within connectivity motifs to control synapse formation in the nervous system.

Data and materials availability

Sequencing data have been deposited at the National Center for Biotechnology Information BioProjects Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE214258. All other data are available in the manuscript or the supplementary material.