Synaptic neoteny of human cortical neurons requires species-specific balancing of SRGAP2-SYNGAP1 cross-inhibition

Summary

Human-specific (HS) genes have been implicated in brain evolution, but their impact on human neuron development and diseases remains unclear. Here, we study SRGAP2B/C, two HS gene duplications of the ancestral synaptic gene SRGAP2A, in human cortical pyramidal neurons (CPNs) xenotransplanted in the mouse cortex. Downregulation of SRGAP2B/C in human CPNs led to strongly accelerated synaptic development, indicating their requirement for the neoteny that distinguishes human synaptogenesis. SRGAP2B/C genes promoted neoteny by reducing the synaptic levels of SRGAP2A,thereby increasing the postsynaptic accumulation of the SYNGAP1 protein, encoded by a major intellectual disability/autism spectrum disorder (ID/ASD) gene. Combinatorial loss-of-function experiments in vivo revealed that the tempo of synaptogenesis is set by the reciprocal antagonism between SRGAP2A and SYNGAP1, which in human CPNs is tipped toward neoteny by SRGAP2B/C. Thus, HS genes can modify the phenotypic expression of genetic mutations leading to ID/ASD through the regulation of human synaptic neoteny.

Graphical abstract

Highlights

• •Human-specific genes SRGAP2B/C are required for human cortical neuron neoteny
• •SRGAP2B/C slow down synaptogenesis by increasing the synaptic levels of SYNGAP1
• •A tug of war between synaptic SRGAP2A and SYNGAP1 sets the tempo of synaptogenesis
• •SRGAP2B/C act as genetic modifiers of the function of SYNGAP1, encoded by a major ID/ASD gene

Libé-Philippot et al. investigate the mechanisms underlying the prolonged synaptic development characteristic of human cortical neurons. They demonstrate that the human-specific genes SRGAP2B/C are required for human synaptic neoteny in vivo. They find that SRGAP2B/C slow down synaptogenesis by upregulating the synaptic levels of SYNGAP1, encoded by a major intellectual disability gene.

Introduction

A salient feature of human brain development is the considerably prolonged, or neotenic, tempo of cortical neuronal maturation, taking months to years in humans instead of days to months observed in other mammals, including non-human primates.^1,2,3,4,5,6 Cortical neuron neoteny affects multiple features of neuronal development and, in particular, synaptic development and maturation, which is thought to lead to prolonged periods of synaptic plasticity characterizing Homo sapiens.^2,7,8,9 Conversely, some neurodevelopmental diseases (NDDs) that impair higher cognitive functions, such as intellectual disability (ID) and autism spectrum disorder (ASD), may involve accelerated brain development.^10,11,12,13 Consistent with this hypothesis, synaptic neoteny was recently shown to be disrupted in human cortical neurons displaying SYNGAP1 haploinsufficiency,¹⁴ a prevalent cause of ID/ASD.

Despite the potential importance of synaptic neoteny for human brain evolution and diseases, the underlying mechanisms remain poorly understood. Human cortical pyramidal neurons (CPNs) develop along a prolonged neotenic timescale when xenotransplanted in the fast-developing mouse cortex, pointing to cell-intrinsic mechanisms.^{15,16,17,18,19,20} While mitochondria metabolism and epigenetic remodeling were recently shown to regulate the species-specific tempo of human CPN maturation,^21,22,23 one outstanding molecular candidate for the regulation of synaptic neoteny is the SLIT-ROBO Rho GTPase Activating Protein 2 (SRGAP2) gene family.²⁴ This family is composed of SRGAP2A, the ancestral gene shared by all mammals (named SRGAP2 in non-human species), human-specific (HS) gene duplicates SRGAP2B, SRGAP2C, and SRGAP2D, which emerged through segmental duplications during hominin evolution, over the past ∼2–3 million years.^25,26 Both HS paralogs SRGAP2B and SRGAP2C are dynamically expressed in the developing cortex and encode truncated versions of the F-Bin-Amphiphysin-Rvs (F-BAR) domain found in SRGAP2A,²⁴ while SRGAP2D is likely a pseudogene because of the lack of two internal exons leading to a premature termination codon. SRGAP2B and SRGAP2C can form unstable heterodimers with SRGAP2A²⁴ and are degraded by the proteasome in neurons.^27,28 Studies in mouse CPNs showed that SRGAP2A is a postsynaptic protein acting as a positive regulator of synaptic maturation.^24,29 Forced expression of SRGAP2B or SRGAP2C in developing mouse CPNs phenocopies a partial loss of SRGAP2A function, resulting in delayed synaptic maturation and increased density of excitatory and inhibitory synapses.^24,27,29 These results point to SRGAP2A and its HS paralogs SRGAP2B/C as potentially important regulatory components of synaptic development and evolution, but whether HS genes SRGAP2B/C contribute to neoteny in human neurons, in either the physiological or pathological condition, has never been tested.

Here, we explored the function and mechanisms of action of SRGAP2A and SRGAP2B/C in human CPNs in vivo, using xenotransplantation in the mouse neonatal cortex. We thus uncovered their essential role for synaptic neoteny characterizing human CPNs through the regulation of the ID/ASD synaptic gene synaptic Ras GTPase activating protein 1 (SYNGAP1).

Results

An in vivo model of loss of function of SRGAP2 genes in human CPNs

In order to specifically and simultaneously downregulate both SRGAP2B/C paralogs, but not SRGAP2A, in human CPNs, we used a knockdown (KD) approach using two independent lentiviral vectors expressing short hairpin RNAs (shRNAs), each targeting the shared 3′ untranslated region (3′ UTR) of SRGAP2B and SRGAP2C transcripts, which is not present in the ancestral SRGAP2A mRNA transcript (Figure 1A).²⁴ This approach was validated in deep-layer 5/6 human CPNs differentiated in vitro from pluripotent stem cells (PSCs) (Figures S1A and S1B), leading to a reduction in the levels of SRGAP2B/C proteins and thereby, as predicted by previous results,^27,30 to increased levels of the ancestral protein SRGAP2A (Figure S1C). Conversely, to examine the role of the ancestral SRGAP2A gene in human neurons, we used shRNA-mediated KD directed against SRGAP2A, which led to decreased SRGAP2A protein expression in human CPNs while leaving SRGAP2B/C unchanged (Figure S1D).

Figure 1.

SRGAP2B/C genes are required for dendritic spine neoteny of human cortical pyramidal neurons

(A) Description of the SRGAP2 gene family at the genomic and transcript levels; note the location of the mRNA sequences targeted by shRNAs and that SRGAP2B/C shRNAs do not target the SRGAP2C protein coding sequence (cDNA, in blue).

(B) Experimental design of human pluripotent stem cell (PSC)-derived neurons, infected in vitro with lentivirus (LV) expressing EGFP and shRNAs targeting Scramble, SRGAP2B/C, or SRGAP2A sequences ± SRGAP2C-HA cDNA, and then xenotransplanted in the mouse neonatal cerebral cortex followed by morphological analyses (see STAR Methods).

(C) Representative proximal dendritic branches of human PSC-derived neurons xenotransplanted in the mouse cerebral cortex, with LV-shRNA infection, followed from 2 to 18 months post-transplantation (MPT).

(D) Corresponding quantifications of the dendritic spine density (means ± 95% confident interval; two-way ANOVA tests with Tukey’s multiple comparison test; see Figures S5A and S5D) (numbers in Table S1; n = 12–54 neurons from 2–15 animals from 2–8 litters per stage per condition).

(E) Corresponding quantifications of the dendritic spine head width (means ± 95% confident interval; two-way ANOVA tests with Tukey’s multiple comparison test; see Figures S5C and S5E) (numbers in Table S1; n = 151–4,722 dendritic spines from 12–54 neurons from 2–15 animals from 2–8 litters per stage per condition).

^∗∗p < 0.01; ^∗∗∗∗p < 0.0001.

See also Figures S1–S5 and Table S1.

To assess the role of SRGAP2B/C in neuronal differentiation and synaptic development of human CPNs in vivo, we used a model of xenotransplantation in the neonatal mouse cerebral cortex that recapitulates some of the features of CPN development^15,20 (Figure 1B). Using this experimental approach, xenotransplanted human CPNs mature following a protracted timeline, taking months to develop complex dendritic morphology, mature dendritic spines, functional synapses, and cortical circuit integration.²⁰ Human cortical progenitors at early stages (4–5 weeks of differentiation from PSC stage^15,20) were infected in vitro with lentiviral vectors transducing the SRGAP2A or SRGAP2B/C shRNAs, as well as EGFP to enable identification and morphological analysis of the transplanted human neurons (Figure 1B). After 1 week of culture, the human cortical cells were treated with a γ-secretase inhibitor (DAPT) and antimitotic drug (Cytarabine, ARA-C) to enhance neurogenesis, decrease the number of progenitors, and enrich for cohorts of neurons of similar birth date, corresponding to similar (deep-layer) early fate as cortical temporal patterning is conserved in vitro.^15,20 1 day later, the human CPNs were transplanted into the host mouse at postnatal day (P)0 (newborn mouse).²⁰ We first analyzed the integration and cell fate of the transplanted human CPNs from an early stage (2 months post-transplantation [MPT]) through 18 MPT. SRGAP2B/C KD and control neurons appeared to integrate similarly in the mouse cortical gray matter. Human CPNs found in the mouse cortex were found to express mostly deep-layer 5/6 markers, while few of them expressed upper layer 2/3 markers (Figure S2C), suggesting that in our experimental setting, the majority of the xenotransplanted neurons found in the mouse cortex correspond to deep-layer fate neurons, as expected.²⁰ The layer fate and localization of the transplanted neurons were found to be similar across control, SRGAP2A KD, and SRGAP2B/C KD conditions (Figure S2).

These data indicate that SRGAP2B/C downregulation had no detectable impact on cortical neuron integration and fate determination, in line with the data obtained in mouse CPNs following downregulation of SRGAP2 or overexpression of SRGAP2C.^24,31

SRGAP2B/C are required for morphological neoteny in human CPNs in vivo

Next, we analyzed the morphological development of the xenotransplanted human CPNs at 2 and 6 MPT, a period of significant increase in dendritic arborization (Figure S3).²⁰ Downregulation of SRGAP2B/C in human CPNs at 2 MPT led to an increased dendritic complexity, quantified using Sholl analysis, compared with human CPNs transduced with control (scrambled) shRNA (Figures S3A and S3B). This phenotype appeared to be transient, as it could not be detected at 6 MPT or later (Figures S3E–S3H).

To assess the specificity of the phenotypes observed, we performed rescue experiments by co-infecting human CPNs prior to xenotransplantation with SRGAP2B/C shRNAs and a lentivirus expressing shRNA-resistant SRGAP2C fused to the HA-tag epitope (SRGAP2C-HA). This was sufficient to rescue the dendritic morphology effect of SRGAP2B/C KD, confirming the specificity of our findings (Figures S3C and S3D). We also observed a slight but significant decreased in dendritic complexity induced by SRGAP2C overexpression alone (Figures S3C and S3D), which could be due to a more powerful effect of SRGAP2C overexpression in downregulating SRGAP2A compared with SRGAP2A KD.

These results suggest that HS genes SRGAP2B/C transiently contribute to delayed dendritic maturation in human CPNs.

SRGAP2B/C are required for synaptic neoteny in human CPNs in vivo

Previous studies have shown that PSC-derived human xenotransplanted CPNs display a protracted pattern of development compared with those of other species,^15,16,18,20 but in these experiments, dendritic spine development of transplanted PSC-derived mouse CPNs was not assessed quantitatively. To this aim, we transplanted mouse PSC-derived deep-cortical-layer CPNs into the mouse cerebral cortex (Figures S4A and S4B) and assessed dendritic spine density (an index of synaptogenesis) and spine head size (an index of synaptic maturation) from 6 to 60 days post-transplantation (DPT) (Figure S4C). The transplanted mouse CPNs displayed a rapid increase in dendritic spine density and head size between 6 and 30 DPT, followed by stabilization between 30 and 60 DPT (Figures S4C–S4E). We next performed immunostaining for the postsynaptic protein postsynaptic density protein 95 (PSD95) and the presynaptic protein synaptophysin (SYP), followed by 3D reconstruction of the dendritic spines and PSD95/SYP clusters. By 30 DPT, most (>75%) of the dendritic spines displayed PSD95 puncta and were in contact with SYP-positive structures (Figures S4F and S4G), as expected for mature dendritic spines.³² Thus, transplanted mouse PSC-derived CPNs follow the same timeline of dendritic spine and synatogenesis as described for native (non-transplanted) mouse CPNs.^5,33,34

Similar analysis of dendritic spine development in human xenotransplanted CPNs (from 2 to 18 MPT) revealed a much more protracted pattern, confirming the cell-intrinsic neuronal neoteny described previously.^{15,16,17,18,20} Specifically, spine density and spine head size were very low at 2 MPT (compared with the peak observed for mouse PSC at 30 DPT) (Figures 1C–1E, S4C–S4E, S5A, and S5C). Human CPNs displayed a slow increase in dendritic spine density and head width up to 18 MPT (Figures 1D and 1E). Even at 18 MPT, spine density in human xenotransplanted CPNs had not yet reached a mature level of spine density. This is consistent with synaptic developmental timing of CPNs in the human brain, which takes 3–15 years depending on the cortical area.^5,35,36,37

Thus, the increase in time of dendritic spine density in transplanted mouse CPNs was >20-fold higher than in xenotransplanted control human CPNs (Figure S5B). These data confirm that transplanted human and mouse CPNs mature at their own species-specific pace.

We then tested the impact of SRGAP2B/C on dendritic spine development in xenotransplanted human CPNs. SRGAP2B/C KD neurons exhibited accelerated dendritic spine development (Figures 1C–1E and S5A–S5E; Table S1). While at 2 MPT the levels of dendritic spine density and head width were similar between all conditions, at later stages the SRGAP2B/C KD CPNs displayed an accelerated increase in dendritic spine density (Figures 1C–1E and S5A–S5E; Table S1). From 4 to 18 MPT, the increase in time of dendritic spine density in SRGAP2B/C KD CPNs was >2- to 3-fold higher compared with control human CPNs (Figures 1C, 1D, S5A, S5B, and S5D; Table S1). Similarly, the SRGAP2B/C KD CPNs displayed increased spine head width compared with control human CPNs (Figures 1C, 1E, S5C, and S5E; Table S1). Conversely, in SRGAP2A KD neurons, we observed a 2-fold slower increase in time in spine density (Figures 1C, 1D, S5A, S5B, and S5D; Table S1) and in spine head width compared with the control human CPNs (Figures 1C, 1E, S5C, and S5E; Table S1). At 18 MPT, SRGAP2B/C KD human CPNs displayed a 2- to 3-fold increase in spine density (∼1.2–1.5 spine per μm) compared with control human CPNs (∼0.5 spine per μm) (Figures 1D, S5A, and S5D; Table S1). In comparison, the dendritic spine density of deep-layer CPNs of the human prefrontal cortex at 1 year postnatally is around 0.3–0.5 spines per μm, similarly to our control values, and reaches values of around ∼1 spine per μm around 5–10 years after birth.³⁵ Thus, SRGAP2B/C-deficient human neurons display a strong acceleration of spine formation and maturation, as they reach high levels of dendritic spine density at least >12 months ahead of time.

To test the specificity of the shRNA KD approach, we performed rescue experiments as described above, using co-expression of the SRGAP2B/C shRNAs together with shRNA-resistant SRGAP2C-HA cDNA. At 6 MPT, expression of SRGAP2C was sufficient to rescue the increased spine density and dendritic spine head width induced by SRGAP2B/C KD, confirming the specificity of the KD approach (Figures S5F–S5I).

These results indicate that HS genes SRGAP2B/C are required for the species-specific neotenic dendritic spine development characterizing human CPNs in vivo, while SRGAP2A has opposite effects to enhance the pace of dendritic spine formation.

SRGAP2B/C are required for neoteny of functional synaptic maturation in human CPNs in vivo

We next determined the functional synaptic consequences of the accelerated development of dendritic spines in SRGAP2B/C KD neurons using patch-clamp recordings of human CPNs in ex vivo acute cortical slices prepared from xenotransplanted mice at both 2 and 6 MPT (Figure 2A; Figure S6).

Figure 2.

SRGAP2B/C are required for neotenic synaptic maturation of human cortical pyramidal neurons

(A) Experimental design of human PSC-derived neurons, LV infected in vitro expressing EGFP and shRNAs targeting Scramble or SRGAP2B/C sequences, and then xenotransplanted in the mouse neonatal cerebral cortex followed by electrophysiological analyses at 6 MPT.

(B) Representative traces of synaptic currents of human PSC-derived neurons xenotransplanted in the mouse cerebral cortex, with LV-shRNA infection, at 6 MPT.

(C and D) Corresponding quantification of the sEPSCs frequency and amplitude (Mann-Whitney tests) (n = 16–17 neurons per condition from 6–7 animals from 2 litters).

(E) Representative examples of AMPA and NMDA currents of human neurons, with LV-shRNA infection, at 6 MPT; the red dashed lines indicate the time points of quantification.

(F) Corresponding quantification of the AMPA/NMDA currents amplitude ratio (Mann-Whitney test) (n = 14–19 neurons per condition from 5–7 animals from 2 litters). Data are represented as single neurons with mean + SD.

(G and H) Cell capacitance (G) and input resistance (H) of human neurons, with LV-shRNA infection, at 6 MPT (Mann-Whitney test) (n = 16–18 neurons per condition from 6–7 animals from 2 litters).

^∗p < 0.05; ^∗∗p < 0.01.

See also Figure S6.

We first examined some biophysical properties, potentially related to observed changes in cell morphology. SRGAP2B/C KD neurons did not differ from control neurons in their input resistance (Figures 2H and S6G) but displayed a transient increase in membrane capacitance at 2 MPT compared with control (Figures 2G and S6F). The capacitance normalized at 6 MPT (Figure 2G), consistent with the SRGAP2B/C-dependent increase in dendritic arbor during this period (Figures S3E and S3F).

We then focused on spontaneous excitatory postsynaptic currents (sEPSCs), which typically increase in frequency and amplitude in xenotransplanted human CPNs over the first 6 MPT, reflecting excitatory (E) synapse formation and maturation, respectively.²⁰ We observed a significant increase of sEPSC frequency in SRGAP2B/C KD CPNs compared with control CPNs (Figures 2B and 2C), indicating increased number of functional synapses³⁸ as well as a significant increase in sEPSC amplitude in SRGAP2B/C KD neurons compared with control neurons (Figures 2B and 2D), suggesting accelerated rates of excitatory synaptic maturation.³⁸ In addition, we measured α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-D-aspartate receptor (NMDA) currents, as AMPA/NMDA current ratios represent a well-validated index of postsynaptic maturation of glutamatergic synapses.³⁸ This revealed a >2-fold increase in AMPA/NMDA amplitude ratio in SRGAP2B/C KD neurons compared with control neurons (Figures 2E and 2F), confirming that synapses are functionally more mature at 6 MPT following SRGAP2B/C loss of function compared with control human CPNs. The increase in functional synaptic development could be observed already at 2 MPT, thus preceding the increase in dendritic spine density or size (Figures S6A–S6E), suggesting an increased maturation of dendritic shaft synapses already at 2 MPT.

These results demonstrate that SRGAP2B/C-deficient human CPNs display more mature excitatory synapses than control human CPNs at the same time point, indicating the requirement of SRGAP2B/C for functional synaptic neoteny in human CPNs.

SRGAP2B/C regulate the synaptic levels of SRGAP2A and SYNGAP1

We next addressed the molecular mechanisms underlying the effects of SRGAP2B/C on the timing of synaptic development of human neurons in vivo. Work in heterologous expression systems and in mouse neurons showed that SRGAP2B/C are unstable proteins that can heterodimerize with SRGAP2A, leading to the formation of insoluble complexes and degradation by the proteasome.^27,28 We therefore examined SRGAP2A protein levels in human CPNs in vitro using cytosolic and synaptosome fractions prepared from purified CPN cultures (devoid of cortical progenitors) infected at day in vitro (DIV)45 (with shScramble, shSRGAP2A, or shSRGAP2B/C) and cultured until DIV70 (Figure 3A), a stage at which synaptic formation and function become prominent.¹⁵ SRGAP2B/C KD in human CPNs led to a significant increase in SRGAP2A protein levels within the cytosolic fraction (Figures S7C and S7D), similar to what we observed in whole-cell extracts (Figure S1C), but also in the synaptosome compartment (Figures 3B and 3C). Of note, SRGAP2B/C proteins were only detected in the cytosolic fraction and not in the synaptosome fraction (Figure S7E), suggesting that SRGAP2B/C mostly act on SRGAP2A outside of the postsynaptic compartment, although with consequences at the synaptic levels.

Figure 3.

SRGAP2B/C are human-specific regulators of SRGAP2A/SYNGAP1 antagonism at the synapse in vitro

(A) Experimental design: protein extraction at day in vitro 70 (DIV70) from human SYNGAP1^+/+ and SYNGAP1^+/− PSC-derived neurons with LV-shRNA infection at DIV45.

(B) Synaptic fraction of SYNGAP1^+/+ and SYNGAP1^+/− PSC-derived neurons at DIV70 with LV-shRNA infection at DIV45. The upper and lower parts of the blot (separated by a white lane) were cut prior to primary antibody incubation: the upper part was immunolabeled with antibodies against SYNGAP1, SRGAP2A, and PSD95, and the lower part with antibodies against SRGAP2 and SYP. (SRGAP2B/C) indicates the expected size of SRG2P2B/C, which was not detected in synaptosomes but in the cytosol fraction (see Figure S7C).

(C) Corresponding quantifications (ANOVA multiple tests; n = 3–5 experiments per condition). Data are represented as single values with mean + SD. ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001).

See also Figures S7 and S8.

Given the increase in SRGAP2A protein abundance at the synapse following SRGAP2B/C KD, we also probed for changes in other synaptic proteins. We first examined HOMER1 and PSD95, two key scaffolding proteins of AMPA and NMDA receptors at excitatory synapses, as well as the AMPA receptor subunit GluA1. We found no change in the synaptic abundance of these proteins following SRGAP2B/C or SRGAP2A KD at these early stages of neuronal development (Figures 3B, S7A, and S7B). We next turned to SYNGAP1, an abundant postsynaptic protein that was previously shown to act as a negative regulator of excitatory synaptic maturation in the mouse^39,40 and is encoded by a gene mutated in up to 1% of ID/ASD cases.^41,42,43 Importantly, we recently found using xenotransplantation that Syngap1^+/− human CPNs display accelerated dendritic spine and synaptic development compared with control human CPNs,¹⁴ similarly to what we observed here following SRGAP2B/C KD.

Surprisingly, SYNGAP1 protein levels were decreased in the synaptosome fraction of SRGAP2B/C KD human CPNs (Figures 3B and 3C), while its cytosolic levels remained unchanged (Figures S7C and S7D). Conversely, we found that SYNGAP1 protein levels were significantly increased in synaptosomes isolated from SRGAP2A KD neurons (Figures 3B and 3C), while no difference was detected in the cytosolic fraction (Figures S7C and S7D), suggesting that the changes observed are not linked to global rates of SYNGAP1 transcription or translation but rather to synaptic targeting. Similar results were obtained in CPNs differentiated in the same way from an independent human PSC line (AH4; Figure S7F).

These results indicate that SRGAP2A inhibits the synaptic accumulation of SYNGAP1 in human CPNs in vitro and that SRGAP2B/C, by downregulating SRGAP2A protein levels, have the opposite effect, increasing synaptic levels of SYNGAP1.

SYNGAP1 regulates the synaptic levels of SRGAP2A

We next tested the possibility that SYNGAP1 could conversely regulate the synaptic levels of SRGAP2A protein, using human CPNs generated from SYNGAP1 haploinsufficient (SYNGAP1^+/−) PSC.¹⁴ As expected, SYNGAP1^+/− human CPNs displayed a 50% reduction in SYNGAP1 levels in cytosolic compartments (Figures S7C and S7D) and >50% reduction in synaptic compartments (Figures 3B and 3C). Importantly, we also detected increased synaptic levels of SRGAP2A in SYNGAP1^+/− human CPNs compared with control SYNGAP1^+/+ human CPNs (Figures 3B and 3C), while no changes were detected in the cytosolic fraction (Figures S7C and S7D). In contrast, SRGAP2B/C levels did not display any detectable changes in SYNGAP1^+/− neurons compared with control human CPNs (Figures S7C and S7D).

Finally, in SYNGAP1^+/− human neurons, we found that SRGAP2A KD led to an increase in synaptic SYNGAP1 levels, reaching levels comparable to those found in SYNGAP1^+/+ control human neurons (Figures 3B and 3C), while no change was detected in the cytosolic fraction (Figures S7C and S7D).

These results indicate that SYNGAP1 inhibits the synaptic accumulation of SRGAP2A in human CPNs, thus pointing to a reciprocal cross-inhibition between SYNGAP1 and SRGAP2A.

SRGAP2/SYNGAP1 interplay at human CPN synapses involves small GTPase regulation

We next explored the molecular mechanisms underlying the synaptic SRGAP2A-SYNGAP1 antagonism and their downstream targets. To this end, we used a tailored gene replacement approach to identify which domain of SRGAP2A is required to regulate the synaptic levels of SYNGAP1. Specifically, we combined SRGAP2A KD loss of function with overexpression of cDNAs encoding either wild-type SRGAP2A or mutant forms of SRGAP2A abrogating interaction with its known effectors, followed by examination of synaptic SYNGAP1 levels. The SRGAP2A mutants tested (all resistant to the SRGAP2A shRNA) were the following: an Ena/Vasp Homology domain 1 (EVH1)-dead mutant disrupting Homer1-binding, a GTPase-activating protein (GAP)-dead mutant abolishing Rac1 GAP activity, and a SRC Homology 3 Domain (SH3)-dead mutant abrogating binding to Gephyrin and Robo^29,44,45 (Figure S8A). As expected, genetic replacement of wild-type SRGAP2A in SRGAP2A KD neurons reduced synaptic levels of SYNGAP1 to levels observed in control human CPNs (Figure S8A). Similarly, the EVH1-dead and SH3-dead mutant forms of SRGAP2A rescued the synaptic SRGAP2A and SYNGAP1 levels in SRGAP2A KD neurons (Figure S8A). However, the GAP-dead mutant failed to rescue the decrease of SYNGAP1 levels at the synapse induced by SRGAP2A KD (Figure S8A). These data indicate that SRGAP2A acts through its Rac1-specific GAP activity to antagonize the synaptic levels of SYNGAP1 (Figure S8A).

As SRGAP2A is a Rac1-GAP⁴⁶ and SYNGAP1 is a Ras-GAP,^47,48 we tested if the SYNGAP1/SRGAP2 cross-inhibition translates into an inhibition of Rac1/Ras GTPase activity. We used activated Rac1 and Ras pull-down to measure the levels of active Rac1-GTP and active Ras-GTP in human CPNs in vitro, in control, SYNGAP1^+/−, SRGAP2A KD, SRGAP2B/C KD conditions, and combinations thereof (Figure S8B). Ras-GTP levels were upregulated in SYNGAP1^+/− CPNs (Figures S8B and S8C), as expected. Conversely, Ras-GTP levels were decreased in SRGAP2A KD CPNs, while active Ras-GTP levels were increased in SRGAP2B/C KD CPNs (Figures S8B and S8C), thus consistent with the effects of these SRGAP2 manipulations on SYNGAP1 synaptic levels. On the other hand, while Rac1-GTP levels were increased following SRGAP2A KD and decreased with SRGAP2B/C KD, as expected, they were unchanged in SYNGAP1^+/− CPNs (Figures S8B and S8C).

These data point to the activity of Rac and Ras GTPases as a differential point of convergence downstream of SYNGAP1/SRGAP2 proteins during human cortical synaptic development.

The balance between SYNGAP1 and SRGAP2A/B/C determines the rates of human synaptic development

As SYNGAP1 deficiency was recently found to result in acceleration of human CPN dendritic spine and synapse formation,¹⁴ while SRGAP2A exerts opposite effects (Figures 1C–1E), we wondered whether the molecular antagonism between these two postsynaptic proteins influences synaptogenesis in human CPNs in vivo.

Specifically, we tested whether the accelerated pace of synaptic development characterizing SYNGAP1^+/− human CPNs¹⁴ could be rescued by downregulation of SRGAP2A, since SRGAP2A was upregulated in SYNGAP1^+/− human CPNs (Figures 3B and 3C). We thus performed xenotransplantation of SYNGAP1^+/− human CPNs, infected with control or shSRGAP2A, followed by morphological and functional characterization of synaptic maturation at 6 MPT compared with control (SYNGAP1^+/+) neurons (Figure 4A).

Figure 4.

SRGAP2A and SYNGAP1 epistasis regulates synaptic developmental tempo of human cortical pyramidal neurons in vivo

(A) Experimental design of SYNGAP1^+/+ and SYNGAP1^+/− human PSC-derived neurons, LV infected in vitro expressing EGFP and shRNAs targeting Scramble and SRGAP2A sequences, and then xenotransplanted in the mouse neonatal cerebral cortex followed by morphological and electrophysiological analyses at 6 MPT.

(B) Representative proximal dendritic branches of SYNGAP1^+/+ and SYNGAP1^+/− human PSC-derived neurons xenotransplanted in the mouse cerebral cortex, with LV-shRNA infection, at 6 MPT.

(C and D) Corresponding quantification of dendritic spine density and spine head width (two-way ANOVA test with Tukey multiple comparison tests) (for SYNGAP1^+/+ neurons: n = 30 neurons, 590–980 dendritic spines from 8 animals from 2 litters per condition; for SYNGAP1^+/− neurons: n = 30–34 neurons, 1,088–1,433 dendritic spines from 9 animals from 2 litters per condition) (part of SYNGAP1^+/+ data are included in Figures 1D and 1E) (see Figure S9A).

(E) Corresponding cumulative distribution of spine head width (Kolmogorov-Smirnov tests) (see Figure S9B).

(F) Representative examples of synaptic currents of SYNGAP1^+/+ and SYNGAP1^+/− human PSC-derived neurons xenotransplanted in the mouse cerebral cortex, with LV-shRNA infection, at 6 MPT.

(G and H) Corresponding quantifications of sEPSCs frequency and amplitude (Kruskal-Wallis tests with Dunn’s multiple comparison) (for SYNGAP1^+/+ neurons: n = 18 neurons from 6 animals from 2 litters; for SYNGAP1^+/− neurons: n = 10–23 neurons from 5–6 animals from 3 litters per condition) (SYNGAP1^+/+ shScramble data were already shown in Figures 2C and 2D).

(I) Cell capacitance of SYNGAP1^+/+ and SYNGAP1^+/− human PSC-derived neurons xenotransplanted in the mouse cerebral cortex, with LV-shRNA infection, at 6 MPT (Kruskal-Wallis tests with Dunn’s multiple comparison) (for SYNGAP1^+/+ neurons: n = 18 neurons from 6 animals from 2 litters; for SYNGAP1^+/− neurons: n = 10–23 neurons from 5–6 animals from 3 litters per condition) (SYNGAP1^+/+ shScramble data were already shown in Figure 2E).

(J) Representative examples of AMPA and NMDA currents of SYNGAP1^+/+ and SYNGAP1^+/− human PSC-derived neurons xenotransplanted in the mouse cerebral cortex, with LV-shRNA infection, at 6 MPT; the red dashed lines indicate the time points of quantification.

(K) Corresponding quantification of the AMPA/NMDA currents amplitude ratio (Kruskal-Wallis tests with Dunn’s multiple comparison) (for SYNGAP1^+/+ neurons: n = 15 neurons from 5 animals from 2 litters; for SYNGAP1^+/− neurons: n = 10 neurons from 3 animals from 3 litters per condition) (SYNGAP1^+/+ shScramble data were already shown in Figure 2G).

In (C), (G)–(I), and (K), data are represented as mean + SD. ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗∗p < 0.0001. See also Figure S9.

First evaluating dendritic spine maturation, as shown above (Figures 1C–1E), we found that SRGAP2A KD led to a decrease in dendritic spine density and spine head width at 6 MPT (Figures 4B–4E and S9). Conversely, SYNGAP1^+/− human CPNs displayed increased dendritic spine density, as previously observed,¹⁴ and larger dendritic spine head width (Figures 4B–4E and S9). Moreover, SRGAP2A KD in SYNGAP1^+/− human CPNs normalized to control levels both dendritic spine density and spine head size at 6 MPT (two-way ANOVA with Tukey multiple tests and multiple linear regression, p > 0.05) (Figures 4B–4E and S9).

We next assessed functional synaptic maturation of human CPNs following the same manipulations through patch-clamp recordings of xenotransplanted human neurons in ex vivo cortical slices isolated at 6 MPT. We found that SYNGAP1^+/− human CPNs displayed increased sEPSC frequency and amplitude, as well as an increased AMPA/NMDA ratio, consistent with increased rates of synaptic maturation compared with control SYNGAP1^+/+ human CPNs (Figures 4F–4K, as previously described¹⁴). These phenotypes were rescued following SRGAP2A KD (Figures 4F–4K), as there were no significant differences between SYNGAP1^+/+ shScramble and SYNGAP1^+/− shSRGAP2A (Kruskal-Wallis tests with Dunn’s multiple comparison tests, p > 0.05; Figures 4G, 4H, and 4K).

These results indicate that the tempo of synapse development in human CPNs is regulated by the balance between SRGAP2A and SYNGAP1.

SRGAP2/SYNGAP1 antagonism at the level of dendritic spines

To further understand the mechanisms underlying the antagonism between SYNGAP1 and SRGAP2A, we investigated whether they cross-regulate each other at the level of individual dendritic spines in vivo. To this aim, we infected SYNGAP1^+/+ or SYNGAP1^+/− human CPNs with control, SRGAP2A, or SRGAP2B/C shRNAs, followed by xenotransplantation (Figure 5A). At 6 MPT, we performed immunostaining for SRGAP2A or SYNGAP1 (Figure 5B), followed by deep-learning automated 3D segmentation of spine heads using a new computational pipeline (restoration enhanced spine and neuron [RESPAN])⁴⁹ and quantification of the fraction of SYNGAP1- and SRGAP2A-positive dendritic spines in all conditions (Figures 5C–5F).

Figure 5.

SRGAP2A and SYNGAP1 antagonism at the synapse of human cortical pyramidal neurons in vivo

(A and B) Experimental design of SYNGAP1^+/+ and SYNGAP1^+/− human PSC-derived neurons xenotransplanted in the mouse cerebral cortex, with LV-shRNA infection in vitro expressing EGFP and shRNAs targeting Scramble and SRGAP2A/B/C sequences (A), immunostained at 6 MPT for SYNGAP1 or SRGAP2A and EGFP to quantify the proportion of SYNGAP1-positive and SRGAP2A-positive dendritic spines (B).

(C) Representative EGFP dendritic branch from SYNGAP1^+/+ shScramble (control) neurons, together with spine head segmentation with RESPAN (see STAR Methods).

(D) Representative dendritic spines (single plans) EGFP-positive (green), positive and negative for SRGAP2A or SYNGAP1 (magenta), with corresponding segmentation masks obtained with RESPAN.

(E and F) Proportion of SYNGAP1-positive (E) or SRGAP2A-positive (F) dendritic spine per neuron (two-way ANOVA with Tukey’s multiple comparison test; for each condition, n = 4–9 neurons, from 3 animals from 1 litter). Data are represented as single values and median. ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001.

We first found that the proportion of SYNGAP1-positive spines was decreased in SYNGAP1^+/− CPNs compared to control neurons (Figure 5E), as expected. This proportion was similarly decreased in SRGAP2B/C KD CPNs (Figure 5E). The proportion of SYNGAP1-positive spines was similar between control and double SYNGAP1^+/−/SRGAP2A KD CPNs (Figure 5E). Conversely, the proportion of SRGAP2A-positive dendritic spines was decreased in SRGAP2A KD CPNs but was not different from the controls in the double SYNGAP1^+/−/SRGAP2A KD CPNs (Figure 5F). Finally, the proportion of SRGAP2A-positive spines was increased in SRGAP2B/C KD CPNs, as well as in the SYNGAP1^+/− CPNs (Figure 5F). These data thus largely confirm those obtained biochemically on in vitro CPNs (Figure 3), indicating that SRGAP2A and SYNGAP1 proteins cross-regulate negatively each other at the level of dendritic spines in vivo. Of note, unlike in the in vitro model, the levels of SYNGAP1 were not different in the SRGAP2A KD CPNs (Figure 5E), and the levels of SRGAP2A were not decreased in the SYNGAP1^+/−/SRGAP2A KD CPNs (Figure 5F). This discrepancy could relate to lower KD efficiency for SRGAP2A in vivo or to the fact that in vitro biochemistry experiments measured global levels of the proteins, while the in vivo approach measured them at the level of individual spines.

Finally, we tested SRGAP2A/SYNGAP1 interactions at synapses in vivo through immunoprecipitation experiments from synaptosome preparations of human adult cerebral cortex biopsies. This revealed no evidence of SRGAP2A/SYNGAP1 co-immunoprecipitation, further pointing to antagonistic localization of the two proteins in human synapses (Figure S8D).

Overall, these data support a model of reciprocal indirect cross-inhibition of accumulation of SRGAP2A and SYNGAP1 at excitatory synapses of developing human CPNs in vivo.

SRGAP2/SYNGAP1 antagonism during synaptic development in non-human mammals

To gain insights into the evolutionary implications of our findings, we tested whether the SRGAP2(A)/SYNGAP1 antagonism was conserved in other mammalian species that lack SRGAP2B/C genes, such as the mouse. We reasoned that if ancestral SRGAP2(A) promotes dendritic spine maturation by antagonizing postsynaptic SYNGAP1 accumulation, then the delayed synaptic maturation observed in mouse CPNs heterozygous for the SRGAP2A mouse orthologous gene Srgap2^24,29 should be suppressed in Srgap2/Syngap1 double heterozygous neurons. We first crossed two conditional knockout mouse lines for Srgap2 and Syngap1 and induced sparse, Cre-recombinase dependent, cell autonomous haploinsufficiency of Srgap2, Syngap1, or both, in layer 2/3 CPNs (Figures 6A and S10A). As reported previously,^24,29,39 morphological analyses at P21 revealed that dendritic spine head size was significantly smaller in Srgap2 heterozygous CPNs and significantly larger in Syngap1 heterozygous CPNs compared with wild-type layer 2/3 CPNs (Figures 6B–6D and S10D). Additionally, Srgap2 heterozygous CPNs displayed a higher dendritic spine density (Figures S10B and S10C). This effect was described previously and reflects the role of Srgap2 in limiting dendritic spine density, independently of its function in the tempo of dendritic spine development.^24,29 Finally, we observed an almost complete normalization of dendritic spine head size distribution in Srgap2^F/+;Syngap1^F/+ double heterozygote CPNs (Figures 6B–6D and S10D). These data indicate that the cross-inhibition of SRGAP2(A) and SYNGAP1 also regulates the tempo of spine maturation in the mouse cortex.

Figure 6.

SRGAP2A and SYNGAP1 epistasis regulates synaptic developmental tempo in mouse cerebral cortex

(A) Experimental design of in utero electroporation (IUE) at embryonic day 15.5 (E15.5) that targets the flex-TdTomato construct and Cre recombinase to layer 2/3 CPNs of mouse embryos. Morphometric dendritic spine analysis was then performed on postnatal day (P)21 coronal sections of the following genotypes: control (wild type), Srgap2^F/+ (haploinsufficiency), Syngap1^F/+ (haploinsufficiency), Srgap2^F/+;Syngap1^F/+ (dual haploinsufficiency), SRGAP2C^KI/+ (gain of function), and Syngap1^F/+;SRGAP2C^KI/+.

(B) Representative apical oblique dendritic branches of mouse L2/3 CPNs at P21 from the respective genotypes.

(C) Corresponding quantification of the spine head widths (Kruskal-Wallis tests with Dunn’s multiple comparison) (n = 1,030–2,008 dendritic spines, from 9–15 neurons, from 3–6 mice, from 2–3 litters per condition) (see Figure S10D).

(D) Corresponding cumulative distribution of the dendritic spine head widths (Kolmogorov-Smirnov tests) (see Figure S10D).

(E and F) Representative western blot (E) and quantification (F) of synaptic abundance of SYNGAP1, SRGAP2(A), and PSD95 in wild-type cerebral cortex at the indicated developmental time points (n = 21 mice from 4 litters, i.e., 3 mice per time point; mean ± SD). ns, not significant.

^∗p < 0.05; ^∗∗∗∗p < 0.0001.

See also Figure S10.

Next, we reasoned that if HS SRGAP2C promotes synaptic neoteny by promoting postsynaptic SYNGAP1 accumulation, then the delayed synaptic maturation observed in mouse CPNs with SRGAP2C knock-in^24,29 should be suppressed by Syngap1 haploinsufficiency. As reported previously,^24,29,39 analysis at P21 revealed that dendritic spine head size was smaller and dendritic spine density higher in SRGAP2C gain-of-function CPNs^24,29 (Figures 6B–6D and S10B–S10D). We observed an almost complete normalization of dendritic spine head size distribution and dendritic spine density in SRGAP2C^KI/+;Syngap1^F/+ double heterozygote CPNs, bringing the values very close to control CPNs as well as the Srgap2^F/+;Syngap1^F/+ double heterozygote CPNs (Figures 6B–6D and S10B–S10D).

Finally, to test whether the inverse relationship of SRGAP2A and SYNGAP1 synaptic targeting observed in the human CPNs was conserved in the mouse, we performed western blot analyses of the SRGAP2(A) and SYNGAP1 in synaptosome (Figures 6E and 6F) and cytosolic (Figures S10E and S10F) proteins extracts, from P7, a stage at which few dendritic spines are being formed, to P21, at which spines are considered mature in the mouse.^5,24 SYNGAP1 synaptic abundance increased gradually over the course of development while SRGAP2(A) decreased over this time period (Figures 6E and 6F), while a similar trend, but much less pronounced, was found in the cytosolic fraction (Figures S10E and S10F). Analysis of a recently published dataset⁵⁰ of postsynaptic protein abundance across developmental stages in human, macaque, and mouse cerebral cortex, which revealed the same inverted trajectories of SRGAP2(A) and SYNGAP1 synaptic abundance in all species (Figure S10G).

Overall, these data support a model of an evolutionarily conserved, mutually antagonistic interaction between postsynaptic proteins SRGAP2(A) and SYNGAP1 that sets the tempo of synaptic maturation in mammalian CPNs.

Discussion

Neoteny of cortical neuron development is thought to play a major role in the emergence of HS features of developmental plasticity and adult brain functional properties.^3,4,51 Despite recent progress, the molecular mechanisms underlying the prolonged development characterizing human CPNs remain poorly understood.^23,52 Here, we confirm that human xenotransplanted CPNs display a cell-intrinsic, species-specific pace of synaptic maturation, which requires the function of HS genes SRGAP2B/C. Moreover, we identified an evolutionarily conserved, mutually antagonistic interaction between postsynaptic proteins SRGAP2A and SYNGAP1 that sets the tempo of synaptic maturation in mammalian CPNs. We found that this balance is modulated in a species-specific manner by HS SRGAP2B/C genes, which by downregulating SRGAP2A, “tips the balance” toward SYNGAP1, thereby setting a neotenic tempo of synaptic maturation in human CPNs.

A main effect of downregulation of SRGAP2B/C in human-developing CPNs is the acceleration of dendritic spine formation and maturation. This may appear paradoxical, given the positive effects of inducing SRGAP2C on synaptic density in mature mouse CPNs (this study, Charrier et al.,²⁴ Fossati et al.,²⁹ and Schmidt et al.³¹). This likely reflects the fact that our in vivo xenotransplantation model, even after 18 MPT, corresponds to relatively early stages of human CPN development, when spine density reflects the ongoing rates of synapse formation. Allowing differentiation of human CPNs for significantly longer periods of time could reveal additional roles for SRGAP2B/C in regulating the final number of synapses in more mature human neurons. However, we are currently limited by the life span of the mouse as a host species, in which human control CPNs cannot reach the densities of dendritic spines observed in human individuals at 5–10 years old.

At the molecular level, we found that downregulation of SRGAP2B/C increases SRGAP2A synaptic protein levels in human CPNs, as previously suggested in heterologous systems.^27,28 Most strikingly, our study uncovers a molecular and cellular reciprocal antagonism between SRGAP2A and SYNGAP1, two synaptic proteins that exert opposite roles in developmental timing of synapse formation and maturation. Importantly, we found that this antagonism takes place at the level of spines or synapses, leading to intraneuronal heterogeneity in the expression of synaptic SRGAP2A and SYNGAP1.⁵³ The implications of these observations on synaptic function and plasticity remain to be explored, as well as the precise underlying molecular mechanisms. Although we do not exclude global transcription or translation regulation of these two genes, our data rather point to the synapse as the main site of cross-regulation between SYNGAP1 and SRGAP2A. Several mechanisms should be considered, including indirect effects on synaptic features such as activity, which regulates the targeting of SYNGAP1 through its phosphorylation by Ca²⁺/calmodulin-dependent protein kinase II (CAMK2).⁵⁴ Our data also suggest more direct effects through their regulation of small GTPases, which are known to control synaptic function and morphology.^55,56 The GAP activity of SYNGAP1 was recently shown to be dispensable for synaptic plasticity and learning in adult mice.^57,58 This suggests that the molecular mechanisms of action of SYNGAP1 may differ in developing and adult synapses, which will be interesting to determine in the future. Alternatively, competition for a common postsynaptic binding partner, yet to be determined, could be involved. An attractive candidate is Catenin Delta 2 (CTNND2), which was recently found to mediate the synaptic recruitment of SYNGAP1 at synapses and whose synaptic abundance in enhanced by SRGAP2C expression or SRGAP2A partial inactivation.⁵⁹

SYNGAP1 is key regulator of synaptic formation and function in cortical neurons,⁴⁰ and SYNGAP1 haploinsufficiency is responsible for up to 1% of all cases of non-syndromic ID, frequently associated with ASD.^40,60 Previous work has pointed to increased synapse formation in Syngap1^+/− juvenile mutant mice, suggestive of accelerated synaptic development.³⁹ Xenotransplantation experiments recently revealed that SYNGAP1 haploinsufficiency in human CPNs in vivo leads to disruption of human synaptic neoteny,¹⁴ as observed here following SRGAP2B/C loss of function. This further strengthens our finding that HS genes SRGAP2B/C promote neoteny by upregulating the synaptic levels of SYNGAP1 in human CPNs.

Precocious brain development was previously associated with specific forms of NDD, including ASD,^11,61 and in vitro PSC models have recently suggested altered timing of neuronal differentiation in some forms of ASD.⁶² The links uncovered here between SRGAP2B/C and SYNGAP1 strongly suggest that alterations in the neotenic pace of synaptic maturation in human CPNs may be a critical pathogenic mechanism in some forms of NDD, which will be important to further explore clinically and experimentally. Our data indicate that human CPNs are particularly sensitive to SRGAP2B/C levels of expression: after 18 months, downregulation of SRGAP2B/C in xenotransplanted human CPNs leads to levels of synaptic maturation similar to those observed in 5- to 10-year-old human brains.³⁵ This suggests that alterations of SRGAPB/C levels or function could have strong pathological consequences in the human brain in vivo. This could be explored further by examining the consequences of synaptic neoteny on human neuronal integration and circuit development and function, using xenotransplantation. While no evidence for SRGAP2B/C germline mutations or copy-number variations has been reported yet, the possibility of rare pathogenic variants remains to be explored.^63,64,65 By linking HS genes to SYNGAP1, our study identifies molecular mechanisms by which the phenotypic expression of genetic mutations leading to NDD can be human specific.

Limitations of the study

We compared the tempo of synaptic development of human xenotransplanted CPNs to mouse transplanted CPNs, but this comparison could be extended to more species (e.g., chimpanzees or macaque⁶⁶). We observed the subcellular location of SRGAP2A and SYNGAP1 proteins using single immunostaining on fixed tissues, which could be assessed in the future using dual genetic labeling and live in vivo imaging to assess the dynamic of their reciprocal distribution during development. The detailed molecular mechanisms of the functional cross-inhibition between SYNGAP1 and SRGAP2A remain to be fully understood.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Pierre Vanderhaeghen (pierre.vanderhaeghen@kuleuven.be).

Materials availability

All unique/stable reagents generated in this study are available from the lead contact. Some plasmids generated in this study have been deposited to Addgene (#220795–220798).

Data and code availability

• •All data are available in the manuscript or the supplemental information.
• •This paper reports original code used for spine segmentation and fluorescence quantification, which is available at https://github.com/lahammond/respan and described in more details in Garcia et al.⁴⁹
• •Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

Acknowledgments

We thank members of the P.V. laboratory for helpful discussions and precious help, including Niels Vidal (CBD). We thank patients for their participation in the study and Anaïs van Hoylandt (UZLeuven) for assistance. We thank Joris de Wit laboratory (CBD) for excellent advice as well as Bart de Strooper and Ludo Van Den Bosch laboratories (CBD) for sharing their chemiluminescence imaging systems. Some of the images were acquired on Zeiss LSM880 supported by Hercules AKUL/15/37_GOH1816N and FWO G.0929.15 to Pieter Vanden Berghe, KU Leuven, whom we also thank for training and technical advice. Some of the images were acquired with support from the Zuckerman Institute’s Cellular Imaging Platform. The authors gratefully acknowledge the VIB Bio Imaging Core for their support and assistance in this work. Drawings were created using BioRender.com. This work was funded by grants from the European Research Council (GENDEVOCORTEX), C1 KULeuven, the EOS Programme, EPINEURODEVO Grant ERANET NEURON, the Belgian FWO and FRS/FNRS, EU network NSC-Reconstruct, the Generet Foundation, and the Belgian Queen Elizabeth Foundation (to P.V.). B.L.-P. was supported by a postdoctoral fellowship of the FWO (12V1219N). Work in the Polleux lab is supported by the NIH (R35 NS127232) and a NOMIS Foundation Research Project. A.J.R. was supported by the NIH/NIGMS (T32 GM145440). Work in the C.C. lab is supported by Inserm, CNRS, ENS, labex Memolife, and the European Research Council (SYNPATH).

Author contributions

Conceptualization, B.L.-P., P.V., and F.P.; methodology, B.L.-P., R.I., P.V., F.P., K.W., S.B.G., L.H., and T.T.; investigation, B.L.-P., A.J.R., K.W., S.B.G., L.H., E.P.-H., P.V., and F.P.; formal analysis, B.L.-P., A.J.R., K.W., S.B.G., L.H., P.V., and F.P.; critical reagents, B.L.-P., R.I., M.D., A.v.B., R.L., S.B., V.G., D.R., C.C., and T.T.; funding acquisition, P.V. and F.P.; project administration, P.V.; supervision, P.V. and F.P.; writing – original draft, B.L.-P., P.V., and F.P.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal beta-tubulin-III	BioLegend	RRID:AB_2315514
Rabbit monoclonal TBR1	Abcam	RRID:AB_2936859
Rat monoclonal BCL11B/CTIP2	Abcam	RRID:AB_2064130
Rabbit polyclonal FOXG1	Takara	RRID:AB_2827749
Goat polyclonal SOX2	Santa-Cruz	RRID:AB_2286684
Rabbit polyclonal SRGAP2-Nter	Guerrier et al.46	N/A
Mouse monoclonal GAPDH	Sigma	RRID:AB_1078991
Chicken polyclonal EGFP	Abcam	RRID:AB_300798
Rabbit polyclonal SATB2	Abcam	RRID:AB_2301417
Rabbit polyclonal FOXP2	Abcam	RRID:AB_2107107
Mouse monoclonal POU3F2/BRN2	Santa-Cruz	RRID:AB_2737347
Rat monoclonal HA	Roche	RRID:AB_390918
Rabbit monoclonal SYP	Abcam	RRID:AB_2286949
Mouse monoclonal SYP	Synaptic System	RRID:AB_887823
Mouse monoclonal PSD95	Sigma	RRID:AB_10807979
Mouse monoclonal PSD95	Synaptic Systems	RRID:AB_10804286
Rabbit polyclonal SYNGAP1	ThermoFisher	RRID:AB_2287112
Rabbit monoclonal SRGAP2A	Abcam	RRID:AB_10971187
Mouse monoclonal HOMER1	Synaptic Systems	RRID:AB_2619855
Rabbit polyclonal GLUA1	Synaptic Systems	RRID:AB_2113441
Rabbit polyclonal CUX1	Proteintech	RRID:AB_2086995
Rabbit isotype control IgG	ThermoFisher	RRID:AB_2532938
Mouse monoclonal Beta-Actin	EMD Millipore	RRID:AB_2223041
Chemicals, peptides, and recombinant proteins
Noggin	R&D Systems	Cat #1967-NG
DAPT	Abcam	Cat #ab120633
ARAC	Merck	Cat #C3350000
Pan-Ras Activation Assay Kit	Cell Biolabs	Cat # STA-400
Rac1 Activation Assay Kit	Cell Biolabs	Cat # STA-401
Protein A magnetic beads	ThermoFisher	Cat# 10334693
Critical commercial assays
Accutase	ThermoFisher	Cat #A1110501
Matrigel	BD Biosciences	Cat #354277
NeuroCult dissociation kit	StemCell technologies	Cat #05715
Revert™ 700 Total Protein Stain	Licor	Cat # 926-11011
SuperSignal™ Western Blot Substrate	ThermoFisher	Cat #34094
Deposited data
Developmental Synaptic Protein Abundance	–	https://liwang.shinyapps.io/PSD_development_explorer/
Experimental models: Cell lines
Human: H9 ESC	WiCell	RRID:CVCL_9773
Human: AH4 IPSC	Espuny-Camacho et al.15	N/A
Human: SYNGAP1 +/KI	Vermaercke et al.14	N/A
Human: HEK-293T	ATCC	RRID:CVCL_1926
Mouse: E14Tg2a ESC	ATCC	RRID:CVCL_9108
Experimental models: Organisms/strains
Mouse: CD1	Charles River Laboratory	RRID:IMSR_CRL:022
Mouse: Rag2-/-	Jackson Laboratories	RRID:IMSR_JAX:008449
Mouse: SRGAP2C KI	Schmidt et al.27	N/A
Mouse: Syngap1+/F	Jackson Laboratories	RRID:IMSR_JAX:029303
Mouse: Srgap2a+/F	Shin et al.67	N/A
Recombinant DNA
pLV-H1-shScramble-hsynapsin-EGFP-WPRE	This Study	RRID: Addgene_ 220795
pLV-H1-shSRGAP2B/C#1-hsynapsin-EGFP-WPRE	This Study	RRID: Addgene_ 220796
pLV-H1- shSRGAP2B/C#2-hsynapsin-EGFP-WPRE	This Study	RRID: Addgene_ 220797
pLV-H1- shSRGAP2A-hsynapsin-EGFP-WPRE	This Study	RRID: Addgene_ 220798
pLV-hsynapsin-SRGAP2C-HA-WPRE	This Study	N/A
pLV-hsynapsin-SRGAP2A-HA-WPRE	This Study	N/A
pLV-hsynapsin-SRGAP2A_EVH1dead-HA-WPRE	This Study	N/A
pLV-hsynapsin-SRGAP2A_GAPDdead-HA-WPRE	This Study	N/A
pLV-hsynapsin-SRGAP2A_SH3dead-HA-WPRE	This Study	N/A
pCAG-cre	Suzuki et al.68	RRID:Addgene_122953
pEf1α-FLEX-tdTomato	Iascone et al.69	N/A
psPAX2	Addgene	RRID:Addgene_12260
pMD2.G	Addgene	RRID:Addgene_12259
pLV-hsynapsin-EGFP	Linaro et al.20	N/A
Software and algorithms
Imaris	Bitplan	https://imaris.oxinst.com/
Fidji	ImageJ	https://imagej.net/software/fiji/
RESPAN v0.9.81	Garcia et al.49	https://github.com/lahammond/respan
GraphPad	Prism	https://www.graphpad.com/

Experimental model and study participant details

Human tissue collection and preparation

The study on research involving human subjects was approved by the Ethics Committee Research of University Hospitals Leuven (UZ Leuven) (reference S61186). Prior to surgery, written informed consent was obtained. Information related to sex, age, race, socioeconomic status, surgical indication, and medication were retrieved from the medical reports. We used biopsies from 2 surgery patients: a temporal cortex biopsy from a 34-year-old female (mesial temporal ganglioglioma) and a frontal cortex biopsy from a 22-year-old female (craniopharyngioma). Human cortical samples were obtained from the temporal or frontal neocortex during neurosurgery. This cortex was resected as part of a standard lobectomy. In both patients, the neocortex was macroscopically normal and did not show signs of pathology upon histological examination. The pathological region was therefore not located in or near the cortical tissue used for the experiments.

Samples were collected at the time of surgery, immerged in ice-cold ACSF (NaCl 126 mM, NaHCO₃ 26mM, D-glucose 10mM, MgSO₄ 6mM, KCL 3mM, CaCl₂ 1mM, NaH₂PO4 1mM, 295-305mM, pH adjusted to 7.4, with 5% CO₂/95% O₂) and transferred immediately to the laboratory, as previously described,⁷⁰ and then flash frozen in an interval of 5-10 minutes and stored at -80°C.

Animals

All mouse experiments with xenotransplantation were performed with the approval of the KU Leuven Committee for Animal Welfare (protocol 2018/030). Mouse housing, breeding and experimental handling were performed according to the ethical guidelines of the Belgian Ministry of Agriculture in agreement with European community Laboratory Animal Care and Use Regulations (86/609/CEE, Journal officiel de l'Union européenne, L358, 18 December 1986). Animals were housed under standard conditions (12 h light:12 h dark cycles) with food and water ad libitum. Rag2^-/- mice used for xenotransplantations were purchased from Jackson Labs and bred and maintained in local facilities. The day of birth was defined as P0. The xenotransplantations were performed at P0 and animals have been processed between 2 and 18 months.

All mice used to perform in utero electroporation (IUE) were handled according to protocols approved by the institutional animal care and use committee (IACUC) at Columbia University. Animals were housed under standard conditions (12 h light:12 h dark cycles) with food and water ad libitum. IUE were performed at embryonic day (E)15.5. For SYNGAP1/SRGAP2A staining, ICR (CD1, Charles River Laboratory) were used as previously described.⁷¹ The conditional Srgap2 knockout mouse line and the conditional SRGAP2C knock-in line were generated as previously described.^31,67 The conditional SYNGAP1 knockout mouse line (Jackson Laboratories #029303) was obtained from The Jackson Laboratory.⁷² For the in utero electroporation experiments, C57BL/6J mice were crossed once with the outbred strain 129S2/SvPasCrl mice (obtained from Charles River) to produce F1 hybrid females. Then, timed pregnancies were done by crossing these F1 females with C57BL/6J mice to generate wild-type (control) animals, with Srgap2^F/F mice to generate heterozygous Srgap2^F/+ animals, and with Syngap1^F/F mice to generate heterozygous Syngap1^F/+ animals. To generate the double heterozygous Srgap2^F/+;Syngap1^F/+ animals and the double heterozygous Srgap2^F/+;Syngap^F/+, timed-pregnant female Srgap2^F/F or SRGAP2C^KI/KI mice were crossed with Syngap1^F/F mice. The data obtained from all animals were pooled without discrimination of sexes for the analysis. Data for this study are derived from a total of 237 mice of both sexes.

Cell lines and neuronal differentiation

Culture of human pluripotent stem cells (XX ESC H9, XX IPSC AH4) have been described previously.⁷³ Human H9 ESC SYNGAP1^+/- and AH4 IPSC were generated previously.^14,74 Culture of mouse pluripotent stem cells (E14TG2a, ATCC) have been described previously.^16,75,76

Human cells were maintained on irradiated mouse embryonic fibroblasts (MEF) in the ES medium until the start of cortical differentiation. Cortical differentiation from human ESC was performed as described previously¹⁵ with some modifications.^20,68 On DIV−2, ESCs were dissociated using Stem-Pro Accutase (Thermo Fisher Scientific, Cat#A1110501) and plated on matrigel- (hES qualified matrigel BD, Cat#354277) coated dishes at low confluency (5,000–10,000 cells/cm2) in MEF-conditioned hES medium supplemented with 10 μM ROCK inhibitor (Y-27632; Merck, Cat#688000). On DIV0 of the differentiation, the medium was changed to DDM,¹⁶ supplemented with B27 devoid of Vitamin A (Thermo Fisher Scientific, Cat#12587010) and 100 ng/ml Noggin (R&D systems, Cat#1967-NG), and the medium was changed every 2 days until DIV6. From DIV6, the medium was changed every day until DIV16. After DIV16, the medium was changed to DDM, supplemented with B27 (DDM/B27), and changed every day. At DIV25, the progenitors were dissociated using Accutase and cryopreserved in mFreSR (StemCell Technologies, Cat#05855).

Mouse cells were maintained on irradiated mouse embryonic fibroblasts (MEF) in the ES medium until the start of cortical differentiation. Cortical differentiation from mouse ESC was performed as described previously ^15,16,75,76 with some modifications. On DIV−1, ESCs were dissociated using 0.05% trypsin/0.5 mM EDTA and plated on matrigel- (hES qualified matrigel BD, Cat#354277) coated dishes at low confluency (5,000–10,000 cells/cm2) in MEF-conditioned hES medium supplemented with 10 μM ROCK inhibitor (Y-27632; Merck, Cat#688000). On DIV0 of the differentiation, the medium was changed to DDM,¹⁶ supplemented with B27 devoid of Vitamin A (Thermo Fisher Scientific, Cat#12587010), 100 ng/ml Noggin (R&D systems, Cat#1967-NG), and 1μM cyclopamine, and the medium was changed every day until DIV10. After DIV10, the medium was changed to DDM, supplemented with B27 (DDM/B27), and changed every day. At DIV12, the progenitors were dissociated using NeuroCult dissociation kit (StemCell technologies, Cat#05715) and LV infected (see below).

Differentiated cortical cells were validated for neuronal and cortical markers by immunostaining using antibodies for beta-tubulin-III (BioLegend, Cat#MMS-435P), TBR1 (Abcam, Cat#ab183032), BCL11B/CTIP2 (Abcam, Cat#ab18465), FOXG1 (Takara, Cat#M227), SOX2 (Santa Cruz, Cat#sc-17320) (Figure S1B).

HEK-293TT human embryonic kidney cells were obtained from American Type Culture Collection (ATCC cat# CRL-11268). HEK-293TT cells were grown in Dulbecco’s modified Eagle’s medium (DMEM; Invitrogen) supplemented with 10% fetal bovine serum (FBS; Invitrogen), 100mM Na-pyruvate, 8.9 mM NaHCO₃, and penicillin/streptomycin (Invitrogen) and split using TrypLE™ Express Enzyme.

Method details

DNA constructs

pLV-H1-shRNA-hsynapsin-EGFP-WPRE is a modified lentiviral plasmid derived from pH1-SCV2 described in Charrier et al.²⁴ The CAG promoter and the mVenus sequences were replaced by hsynapsin and EGFP sequences, respectively, using InFusion cloning (Clontech, cat#638909). hsynapsin and EGFP sequences were PCR amplified from pLenti-hSynI-EmGFP-WPRE described in Linaro et al.²⁰ All constructs were verified by DNA sequencing. The shRNA sequences were inserted by ligation at the NheI/BamhI sites. The shScramble seed sequence was 5’-ACA CCT ATA ACA ACG GTA G-3' described in Charrier et al.,²⁴ the shSRAGAP2C#1 sequence was 5’-ACACCTAAAGGTGCAAACATTAA-3' described in Charrier et al.,²⁴ the shSRAGAP2C#2 sequence was 5’-AAGGACAGGCATTGAATATCTTA-3' described in Charrier et al.,²⁴ the shSRGAP2A sequence was 5’-GCCACTCATCCCTGAAGAATC-3'. All constructs were verified by DNA sequencing.

pLV-hsynapsin-SRGAP2C-HA-WPRE, pLV-hsynapsin-SRGAP2A-HA-WPRE, pLV-hsynapsin-SRGAP2A-EVH1dead-HA-WPRE, hsynapsin-SRGAP2A-GAPdead-HA-WPRE, hsynapsin-SRGAP2A-SH3dead-HA-WPRE, were generated by insertion of SRGAP2 cDNAs already described^24,29 at a multiple cloning site (MCS) of a pLV-hsynapsin-MCS-WPRE plasmid using InFusion cloning (Clontech, Cat#638909). The MCS was prepared by annealing the following primers pair, 5’- CCGGTGCTAGCGTTAACTATATAGGCGCGCCG-3’/5’-AATTCGGCGCGCCTATATAGTTAACGCTAGCA-3’, inserted into the lentiviral backbone pLenti-hsynapsin-hChR2(H134R)-EYFP-WPRE (a gift from Karl Deisseroth (Addgene plasmid # 20945) by restriction digestion (AgeI/EcoRI) and ligation to obtain a pLV-hsynapsin-MCS-WPRE. SRGAP2C-HA was PCR amplified from the cDNA described in Charrier et al.²⁴ All constructs were verified by DNA sequencing.

The pCAG-Cre plasmid was previously described.⁷⁷ The FLEX-TdTomato plasmid was previously described.⁶⁹ The phsynapsin-EGFP was previously described.²⁰

Virus production

HEK-293T cells were transfected by packaging plasmids, psPAX2 (Addgene Cat#12260) and pMD2.G (Addgene Cat#12259), and the plasmid on interest in a lentiviral backbone. 3 days after transfection, culture medium was collected and viral particles were enriched by filter device (Amicon Ultra-15 Centrifuge Filters, Merck, Cat#UFC910008).

Titer check was performed on HEK-293T cell culture for every batch of lentiviral preparation. 100 000 cells were seeded on gelatin-coated coverslips (12-well plates) and infected with 6 serial dilutions of the lentivirus. Medium was changed 24h after and cells were fixed 72 hours after for 1 hour at room temperature with PBS PFA 4%, followed by immunostaining of the relevant marker (EGFP, HA) and Hoechst. From this, the volume of virus needed to infect a given number of cells was estimated (70% for xenotransplantations, 90% of in vitro experiments).

Whole cell protein extraction

Human cortical cells that were frozen at DIV25, were thawed and plated on matrigel-coated plates using DDM/B27 and Neurobasal supplemented with B27 (DDM/B27+Nb/B27) medium. Five days after plating (DIV30), cells were dissociated using Accutase and plated on new matrigel-coated plates at high confluency (100,000–600,000 cells/cm²) with lentiviral vector or not infected. The following day, medium was changed to DDM/B27+Nb/B27 medium. 14 days after thawing (DIV39), cells were treated with 10 μM DAPT (Abcam, Cat#ab120633) for 72 hours. The following day (DIV42), cells were treated with 10 μM DAPT and 5 μM Cytarabine (ARA-C) (Merck, Cat#C3350000) for 24 hours. The following day (DIV43), medium was changed to DDM/B27+Nb/B27 medium. At DIV45, cells were homogenized from 2 wells per condition in RIPA buffer (Tris-HCl 50 mM pH = 7.5, NaCl 150mM, NP-40 0.5%, sodium deoxycholate 0.5%, SDS 0.05%, protease inhibitors) using a plastic cell scraper. Samples have been extracted for 1 hours and centrifuged at 10000xg for 30 minutes at 4°C to pellet insoluble material.

Subsequently, samples were diluted to final 1x Laemmli buffer at 95°C for 5mn. They have been run in NUPAGE 10% Bis-Tris Protein Gel at the voltage of 90V for 2 hours in MOPS buffer and then transferred to PVDF Blotting Membrane at the voltage of 100V for 100 minutes. The membrane was blocked in the buffer (5% skim milk and 0.1% Tween20 in TBS) for 1 hour at room temperature and subsequently incubated in the blocking buffer containing rabbit anti-SRGAP2 Nter to label SRGAP2B/C and SRGAP2A (1:1000, described in Guerrier et al.⁴⁶ and mouse anti-GAPDH (1:5000, Sigma #G8795). Antibodies were incubated overnight at 4°C, followed by the incubation in the blocking solution containing secondary antibody anti-Rabbit and Mouse IgG antibodies conjugated with HRP at room temperature for 1 hour. Pierce ECL Western Blotting Substrate was used for signal detection. Protein abundance was then quantified using Fidji, and normalized to GAPDH levels and to the Non-infected condition.

Neonatal transplantation

Neonatal xenotransplantation was performed as already described^{15,16,20,75,76} with some modifications.

Human cortical cells that were frozen at DIV25, were thawed and plated on matrigel-coated plates using DDM/B27 and Neurobasal supplemented with B27 (DDM/B27+Nb/B27) medium. Six days after plating (DIV31), cells were dissociated using Accutase and plated on new matrigel-coated plates at high confluency (100,000–600,000 cells/cm²) with lentiviral vector. The following day, medium was changed to DDM/B27+Nb/B27 medium. At 14 days after thawing (DIV39), cells were treated with 10μM DAPT (Abcam, Cat#ab120633) for 72 hours. The following day (DIV42), cells were treated with 10μM DAPT and 5μM Cytarabine (ARA-C) (Merck, Cat#C3350000) for 24 hours. The following day (DIV43), medium was changed to DDM/B27+Nb/B27 medium. 19 days after thawing (DIV44), cells were dissociated using NeuroCult dissociation kit (StemCell technologies, Cat#05715) and suspended in the injection solution containing 20 mM EGTA (Merck, Cat#03777) and 0.1% Fast Green (Merck, Cat#210-M) in PBS at 100 000 cells/μl.

Mouse cortical cells that were dissociated using Neurocult dissociation kit at DIV12 on matrigel-coated plates using DDM/B27 and Neurobasal supplemented with B27 (DDM/B27+Nb/B27) medium and LV infected with pLV-hsynapsin-MCS-WPRE. Two days after (DIV14), cells were medium was changed to DDM/B27 medium and cells were treated with 10μM DAPT (Abcam, Cat#ab120633) for 24 hours. The following day (DIV15), cells were dissociated using NeuroCult dissociation kit and suspended in the injection solution containing 20 mM EGTA (Merck, Cat#03777) and 0.1% Fast Green (Merck, Cat#210-M) in PBS at 25 000 cells/μl.

Approximately 1-2 μl of cell suspension was injected into the lateral ventricles of each hemisphere of neonatal (P0) immunodeficient mice Rag2^−/− using glass capillaries pulled on a horizontal puller (Sutter P-97).

Xenotransplanted mouse cortex processing and immunostaining

Xenotransplanted animals were perfused transcardiacally with ice-cold sucrose 8% PFA 4%. Brains were dissected and soaked in the same fixative overnight, then stored in PBS azide. Then they have been sectioned in 80 μm thickness using vibratome. Slices were transferred into the blocking solution (PBS 0.3% Triton, 5% horse serum, 3% BSA) and incubated for 2 hours. Brain floating slices were incubated 3 days at 4°C with primary antibodies in the blocking condition: chicken anti-EGFP (1:1000; ab13970, Abcam), rabbit anti-TBR1 (1:1000, described above), rabbit anti-SATB2 (1:2000, Abcam, Cat# ab34735), rat anti-BCL11B/CTIP2 (1:500, described above), mouse anti-FOXP2 (1:1000, Abcam, Cat# ab16046), mouse anti-POU3F2/BRN2 (1:500, Santa Cruz #sc-393324), rat anti-HA (1:500; Roche #11867423001). After three PBS washes, slices were incubated overnight at 4°C with secondary antibodies in PBS: donkey anti-rabbit Cy3, anti-mouse a647, anti-rat a647, anti-chicken a488, anti-rat Cy3 (1:1000) and Hoechst (1:10000). For synaptic stainings, brain floating slices were incubated 3 days at 4°C in the blocking condition with primary antibodies: chicken anti-EGFP (1:1000), rabbit anti-SYP (1:500, Abcam, #ab32127), mouse anti-PSD95 (1:500, Sigma, #MABN68), rabbit anti-SRGAP2A (1:1000, Abcam, #ab124958), rabbit anti-SYNGAP1 (1:1000, ThermoFisher, #PA1-046). After three PBS washes, slices were incubated overnight at 4°C in the blocking condition with secondary antibodies in PBS: donkey anti-rabbit Cy3, anti-mouse a647, anti-goat Cy3 (1:500), anti-chicken a488 (1:1000) and Hoechst (1:10000). After three washes in PBS, brain sections were mounted on a slide glass with the mounting reagent (DAKO glycerol mounting medium) using #1.5 coverslips.

Image acquisition of human neurons

Confocal images were obtained with Zeiss LSM880 and LSM900 driven by Zen Black and Blue softwares equipped with objectives 10x, 20x, oil immersion 25x and oil immersion 40x & 63x, AiryScan system and argon, helium-neon and 405 nm diode lasers. Specifically, dendritic branches were acquired using an oil immersion x63 objective and AiryScan system with 0.1867357 μm Z steps and 1012x1032 pixel resolution. Neurons immunoreactive for TBR1 and located in cortical layers V–VI of the visual and somatosensory cortices have been considered. For the rescue experiment, EGFP-HA-immunoreactive neurons have been considered. 1 proximal dendritic branches from each neuron have been imaged, located 50 μm from the soma.

Image analysis of transplanted neurons

For human experiments, morphometric analyses of dendritic arbour and dendritic spines were performed by manual 3D reconstruction using Imaris software (Bitplane). The dendritic arbour was reconstructed using the Filament Tracer followed by Sholl analysis. Dendritic spines were quantified in proximal dendrites located 50 μm from the soma. Spine density was defined as the number of quantified spines divided by the length over which the spines were quantified. Head width was defined as the average length of the head that was perpendicular to the neck. The Cell Counter toolbox in Fiji was used to manually annotate cell fate markers.

For the analysis of the synaptic proteins of mouse transplanted neurons in vivo, we analyzed the SYP and PSD95 staining within or contacting the EGFP-positive dendritic spines using Imaris. Dendritic spines were reconstructed using the Filament Tracer and selected not to include the dendritic shaft. Then, we used the tool “spots close to surface” (ImarisXT extension) to filter the SYP spots located at 0.2 μm from the EGFP surface. PSD95 spots were identified using “Split spots into surface” tool.

For analysis of the synaptic proteins of human xenotransplanted neurons in vivo, dendritic spine and dendrite shaft automated segmentation were performed using Fiji and Restoration Enhanced Spine and Neuron Analysis (RESPAN)⁴⁹ using 3D stacks performed using Zeiss LSM880 Airyscan confocal microscope. First, SYNGAP1 and SRGAP2A positive domains were segmented by subtracting background fluorescence, which was determined for each image as the mean intensity of three regions of interest placed in featureless areas. This was followed by 3D recursive median filter with a radius of 2 pixels for 30 iterations biovoxxel/BioVoxxel-Toolbox: BioVoxxel Toolbox v2.6.0 (biovoxxel-toolbox_v2.6.0a), then local thresholding using Otsu’s method. The resulting binary masks and the corresponding raw fluorescence channel representing the dendrite were then batch-processed in RESPAN as previously described.⁴⁹ For spine and dendrite segmentation in RESPAN, an adaptive and self-configuring nnU-Net architecture⁷⁸ was trained using 47 image volumes acquired using a Zeiss Airyscan confocal and a Yokogawa W1 Spinning (with a lateral sampling of 65nm and an axial sampling 105nm) representing a total and 2489 spines accompanied by ground truth labels generated by a consensus of experts. The training was performed on a 24 GB NVIDIA (GeForce RTX 4090) GPU, with the 3D full resolution model being trained for 1000 epochs with an initial learning rate of 0.01 and parameters previously described.⁷⁸ Following training of the model, the output accuracy compared to ground truth labels in validation volumes was evaluated. The model achieved a pseudo-Dice score of 0.89 for spines and 0.96 for dendritic shaft, with 1 indicating perfect overlap. This model and the corresponding training plans are provided in Garcia et al.⁴⁹

Subsequent processing in RESPAN generated 3D morphological and spatial metrics for detected spines. To reduce false positives, only spines with a volume of 0.035 to 2 μm² that were within 4μm of the dendrite were included for subsequent analysis. Spines were considered positive for SYNGAP1 and SRGAP2A if the 3D spine mask contained masked voxels corresponding to these channels. The axial (z) resolution of our imaging approach (0.218 μm) limited the precision of our approach to quantify postsynaptic SRGAP2 and SYNGAP1 protein co-localization in segmented spines in the same way for all experimental conditions displayed in Figure 5.

The accuracy of RESPAN’s spine and dendrite detection was evaluated with RESPAN’s validation tool using ground truth validation data excluded from training. This evaluation measured precision, recall, and accuracy on a per-object basis using an Intersection over Union threshold of 50%. Using these metrics, the spine detection achieved a precision of 0.93, a recall of 0.86, demonstrating highly accurate spine detection.

Electrophysiological recordings and analysis

Whole cell patch-clamp recordings were performed on acute coronal slices prepared from 6 months old mice with xenotransplanted PSC-derived human neurons. Briefly, animals were anaesthetized intraperitoneally with Nembutal and transcardially perfused with ∼25 mL of ice-cold NMDG-based slicing solution. Brains were rapidly extracted and placed in ice-cold NMDG-based slicing solution containing (in mM): 93 N-Methyl-D-glucamine, 2.5 KCl, 1.2 NaH2PO4, 0.5 CaCl2, 10 MgSO4, 30 NaHCO3, 5 Na-ascorbate, 3 Na-pyruvate, 2 Thiourea, 20 HEPES and 25 D-glucose (pH adjusted to 7.35 with 10 N HCl, gassed with 95% O₂/5% CO₂). Coronal slices (250 μm) were cut in ice-cold NMDG-based slicing solution (using a Leica VT1200) and subsequently incubated for ∼6 minutes in the NMDG solution at 34°C. Slices were then transferred into holding aCSF, containing (in mM): 126 NaCl, 3 KCl, 1 NaH2PO4, 1 CaCl2, 6 MgSO4, 26 NaHCO3 and 10 D-glucose (gassed with 95% O₂/5% CO₂). Slices were stored at room temperature for ∼1 hour before experiments.

During experiments brain slices were continuously perfused in a submerged chamber (Warner Instruments) at a rate of 3-4 ml/min with 127 mM NaCl, 2.5 mM KCl, 1.25 mM NaH 2 PO 4, 25 mM NaHCO 3, 1 mM MgCl 2, 2 mM CaCl 2, 25 mM glucose at pH 7.4 with 5% CO 2 / 95% O 2. For sEPSC and AMPA/NMDA ratio recordings we added 20 μM bicuculline. Whole cell patch clamp recordings were done using borosilicate glass recording pipettes (resistance 3.5–5 MΩ, Sutter P-1000) filled with the following internal solution: 115 mM CsMSF, 20 mM CsCl, 10 mM HEPES, 2.5 mM MgCl 2, 4 mM ATP, 0.4 mM GTP, 10 mM Creatine Phosphate and 0.6 mM EGTA. Visually identifiable fluorescently labelled transplanted neurons were selected for recording. Whole-cell patch-clamp recordings were done using a double EPC-10 amplifier under control of Patchmaster v2 x 32 software (HEKA Elektronik, Lambrecht/Pfalz, Germany). Currents were recorded at 20 Hz and low-pass filtered at 3 kHz when stored. The series resistance was compensated to 75-85%. Spontaneous input was recorded using whole-cell voltage clamp recordings (V m =-70 mV). An extracellular stimulation pipette (borosilicate theta glass, Hilgenberg) was placed near the recorded neuron and used for initiating evoked AMPAR and NMDAR-mediated currents (80-120 μA, 1ms (Isoflex, A.M.P. Instruments LTD)). AMPAR-mediated evoked EPSCs were measured in whole-cell voltage clamp at a holding potential of -70 mV, while the NMDAR-mediated component was measured at +40 mV immediately after the initial AMPAR/NMDAR-mediated current (100-150 ms after electrical stimulation, as indicated by red dashed lines in the example traces). Evoked data were analysed using Fitmaster (HEKA Elektronik, Lambrecht/Pfalz, Germany), spontaneous input was analyzed using Mini Analysis program (Synaptosoft).

Human synaptosome protein extraction and whole cell Ras/Rac activation assays

Human cortical cells (frozen at DIV25) were thawed and plated on Matrigel-coated plates using DDM/B27+Nb/B27 medium at 37˚C with 5% CO₂. Seven days after thawing (DIV32), the cells were dissociated using Accutase and plated on Matrigel-coated plate at high confluency (450,000-700,000 cells/cm²). Four days later (DIV36), the medium was changed to DDM/B27+Nb/B27 medium with 10μM DAPT. Two days after DAPT induction (DIV38), the medium was changed to fresh DDM/B27+Nb/B27 medium. At DIV40, cortical cells were dissociated using NeuroCult Enzymatic Dissociation Kit following manufacturer’s instructions. For L1CAM+ MACS, dissociated cells were incubated with biotin conjugated anti-human CD171(L1CAM) (Miltenyi Biotec, Cat#130-124-046) in MACS buffer at 4˚C for 10 min. After washing using MACS buffer, human cells were incubated with anti-biotin microbeads (Miltenyi Biotec, Cat#130-090-485) in MACS buffer at 4˚C for 15 min. L1CAM positive selection were carried out with LS columns according to the manufacturer’s instructions. The sorted cells were plated on Poly-L-ornithine-, Laminin- and horse serum-coated 6-well plate at 110,000 cells/cm². The sorted cells were maintained in DDM/B27+Nb/B27 medium. Neurons have been infected at DIV45 with LV, then the medium was changed to DDM/B27+Nb/B27 medium and changed every 3-4 days until DIV70.

At DIV70, synaptosome proteins have been processed as following. Crude synaptosome extracts were prepared for each condition from 2 wells with 2M neurons, homogenized in Homogenization buffer (sucrose 1.28M, Tris(hydroxymethyl)aminomethane 20mM, MgCl₂ 4mM, protease inhibitors) using a plastic cell scraper. Homogenate was spun at 1000xg for 10 minutes at 4°C. Supernatant was spun at 14,000 x g for 20 minutes at 4°C. Supernatant was kept as the cytosolic fraction. P2 crude synaptosomes were re-suspended in Extraction Buffer (HEPES 50mM pH = 7.5, NaCl 150mM, EDTA 2mM, NP-40 1%, Triton 0.5%, Na₃VO₄ 1mM, NaF 30mM, protease inhibitors) and extracted for 1 hours and centrifuged at 10000xg for 30 minutes at 4°C to pellet insoluble material.

Subsequently, samples were diluted to final 1x Laemmli buffer at 95°C for 5mn. They have been run in NUPAGE 4-12% Bis-Tris Protein Gel at the voltage of 90V for 2 hours in MOPS buffer and then transferred to PVDF Blotting Membrane at the voltage of 100V for 100 minutes. Membranes have been stained using Revert™ 700 Total Protein Stain (Licor) following the protocol described by the supplier. The membrane was blocked in the buffer (5% skim milk and 0.1% Tween20 in TBS) for 1 hour at room temperature and subsequently incubated in the blocking buffer containing rabbit anti-SRGAP2 Nter to label SRGAP2C (1:1000, described above and inGuerrier et al.⁴⁶), mouse anti-GAPDH (1:5000, described above), rabbit anti-SYNGAP1 (1:5000; described above), rabbit anti-SRGAP2A (1:5000;described above), mouse anti-SYP (1:5000; Synaptic Systems #101011), mouse anti-PSD95 (1:5000; described above), mouse anti-HOMER1 (1:5000; Synaptic Systems #160011), and rabbit anti-GLUA1 (1:500; Synaptic Systems #182 003). Antibodies were incubated overnight at 4°C, followed by the incubation in the blocking solution containing secondary antibody anti-Rabbit and Mouse IgG antibody conjugated with HRP at room temperature for 1 hour. SuperSignal™ Western Blot Substrate were used for signal detection.

Signals have been measured using Fiji and normalized to the total protein staining and then to the WT shScramble value.

Ras/Rac activity were detected using the active pan-Ras or Rac1 pull-down and detection kits (Bio-Connect). Briefly, RAF1 RBD (for pan-Ras) and PAK1 PBD (for Rac1) agarose beads were mixed with 300 μg of whole-cell lysates (commercial extraction buffer with addition of protease inhibitors) and incubated on a wheel for 1 h at 4 °C, followed by three washes and elution with 2x Laemmli buffer at 95°C for 5mn. 5 μg of whole-cell lysates (i.e. total Ras or total Rac1) and pull-down samples have been run in NUPAGE 12% Bis-Tris Protein Gel at the voltage of 90V for 2 hours in MOPS buffer and then transferred to PVDF Blotting Membrane at the voltage of 100V for 100 minutes. The membrane was blocked in the buffer (5% skim milk and 0.1% Tween20 in TBS) for 1 hour at room temperature and subsequently incubated in the blocking buffer containing either mouse anti-pan-Ras or mouse anti-Rac1 primary antibodies. Antibodies were incubated overnight at 4°C, followed by the incubation in the blocking solution containing anti-mouse IgG antibody conjugated with HRP at room temperature for 1 hour. SuperSignal™ Western Blot Substrate were used for signal detection.

Signals have been measured using Fiji and normalized to the total Ras/Rac and then to the WT shScramble value.

In utero electroporation (IUE)

IUE was performed at embryonic day 15.5 (E15.5) on isoflurane-anaesthetized timed-pregnant female mice as previously described,⁷⁹ but with the following modifications. Endotoxin-free DNA containing 1 ug/ul of FLEX-TdTomato plasmid and 2-100 ng/ul of pCAG-Cre plasmid was injected into the ventricles of E15.5 embryos using a heat-pulled capillary attached to Picospritzer III (Parker). Electroporation was performed by applying 5 pulses of 42 V for 50 ms with 500-ms intervals using a 3-mm diameter platinum tweezer electrode (Nepa Gene) and a square wave electroporator (ECM 830, BTX). After placing embryos back into the abdominal cavity, the incision was closed using sutures and the mouse was allowed to recover on a heating pad.

IUE mouse processing and dendritic spine analysis

At P20-P21, mice were anesthetized with isoflurane, and intracardiac perfusion was performed using 4% paraformaldehyde (Electron Microscopy Sciences, cat no. 15714-S) in PBS as previously described,³¹ but with the following modifications. Brains were isolated and incubated overnight in the 4% paraformaldehyde in PBS solution at 4°C. The brains were then washed in PBS and sectioned along the coronal plane at 100 μm using a vibrating microtome (Leica VT1200S). Sections were collected spanning the approximate bounds of the somatosensory cortex. To confirm L2/3 cortical targeting, representative WT sections were blocked in 10% goat serum and 0.5% Triton X-100 in PBS for 1 hour at room temperature, and then incubated overnight at room temperature with the following antibody: rabbit anti-CUX1 (1:1000; Proteintech, #11733-1-AP) in 2% goat serum and 0.5% Triton X-100 in PBS. The stained slices were then washed in PBS with 0.5% Triton X-100 and incubated with the following secondary antibody: goat anti-rabbit Alexa Fluor 488 (1:200; Invitrogen, cat no. 11008) in 2% goat serum and 0.5% Triton X-100 in PBS. Sections were stained using DAPI (Invitrogen, cat no. D1306) and subsequently mounted on glass slides in Citifluor Mountant Solution (Electron Microscopy Sciences, cat no. AF100-5 & 17977-150). Slides were imaged on a W1-Tokogawa spinning disk confocal microscope using a Nikon 100x, 1.35 NA silicon-immersion objective (Nikon, cat no. MRD73950).

Morphometric analyses of dendritic spines were performed in the depth of the z stack for slices using Fiji as previously described,²⁴ but with the following modifications. Dendritic spines were quantified in oblique dendrites originating from the apical trunk. Spine density was defined as the number of quantified spines in the depth of the z stack divided by the length over which the spines were quantified. The dendritic segment length was measured with the Fiji segmented line tool in the maximum intensity projection. Head width was defined as the largest length of the head that was perpendicular to the neck. Only dendrites that were parallel to the plane of the slice were analyzed. Spine analysis was done in brain sections of comparable rostro-caudal position.

For synaptic stainings, brain floating slices were incubated 3 days at 4°C in the blocking condition with primary antibodies: chicken anti-EGFP (1:1000, described above), rabbit anti-SYP (1:500, described above), mouse anti-PSD95 (1:500, described above). After three PBS washes, slices were incubated overnight at 4°C in the blocking condition with secondary antibodies in PBS: donkey anti-rabbit Cy3, anti-mouse a647, anti-goat Cy3 (1:500), anti-chicken a488 (1:1000) and Hoechst (1:10000). Analysis was performed as described above for human xenotransplanted neurons.

Immunoprecipitation

Crude synaptosome extracts were prepared from thaw biopsy were homogenized using a Dounce homogenizer in Homogenization buffer (sucrose 1.28M, Tris(hydroxymethyl)aminomethane 20mM, MgCl₂ 4mM, protease inhibitors). Homogenate was spun at 1000xg for 10 minutes at 4°C. Supernatant was spun at 14,000 x g for 20 minutes at 4°C. Supernatant was kept as the cytosolic fraction. P2 crude synaptosomes were re-suspended in Extraction Buffer (HEPES 50mM pH = 7.5, NaCl 150mM, EDTA 2mM, NP-40 1%, Triton 0.5%, Na₃VO₄ 1mM, NaF 30mM, protease inhibitors) and extracted for 2 hours and centrifuged at 10000xg for 30 minutes at 4°C to pellet insoluble material.

Samples were incubated overnight on a wheel in a cold room with protein A magnetic beads coupled with 1ug of rabbit anti-SRGAP2A or rabbit IgG (ThermoFisher # 02-6102) antibodies. Beads were washed 4 times with the washing solution (HEPES 50mM pH = 7.5, NaCl 150mM, EDTA 2mM, NP-40 1%, Triton 0.5%, Na₃VO₄ 1mM, NaF 30mM) and one time with PBS. Subsequently, samples were eluted in 2x Laemmli buffer at 95°C. The input (in 1x Laemmli buffer) and immunoprecipitated samples were run in NUPAGE 4-12% Bis-Tris Protein Gel at the voltage of 90V for 2 hours in MOPS buffer and then transferred to PVDF Blotting Membrane at the voltage of 100V for 100 minutes. The membrane was blocked in the buffer (5% skim milk and 0.1% Tween20 in TBS) for 1 hour at room temperature and subsequently incubated in the blocking buffer containing rabbit anti-SYNGAP1 or rabbit anti-SRGAP2A antibodies overnight at 4°C, followed by the incubation in the blocking solution containing secondary antibody anti-Rabbit IgG antibody conjugated with HRP at room temperature for 1 hour. Pierce ECL Western Blotting Substrate was used for signal detection.

Mouse protein extracts and western blotting

Synaptosome isolation was performed from P7, P10, P12, P14, P16, P18, or P21 mouse brains as previously described⁸⁰ but with the following modifications. All centrifugation steps were performed at 4°C, and the rest of the protocol was performed on ice. Briefly, cortical hemispheres were dissected and homogenized in ice-cold HEPES-buffered sucrose (0.32 M Sucrose, 1 mM NaHCO 3, 1 mM MgCl 2, 10 mM HEPES pH 7.4), protease inhibitor cocktail (Pierce, cat no. A32955), and phosphatase inhibitor cocktail (Pierce, cat no. A32957) using a 5-mL glass Dounce homogenizer. The homogenate was centrifuged at 1000 rpm for 5 min. The resulting supernatant was centrifuged at 1400 rpm for 5 min to yield the nuclear pellet and the supernatant was then centrifuged at 15000 rpm for 15 min to yield the synaptosome pellet.

Using a BCA protein assay kit (Pierce, cat no. 23227), samples were prepared at concentrations of 1 mg/mL in Laemmli buffer (Bio-Rad, cat no. 1610747) containing 0.1 M Dithiothreitol and boiled at 95°C for 5 min. Proteins were separated on a 4-20% polyacrylamide gel (BioRad, cat no. 5678094) and then transferred to 0.2 um Nitrocellulose membranes (BioRad, cat no. 1704159) using the Trans-Blot Turbo Transfer System (BioRad, cat no. 1704155). Membranes were blocked in Intercept (TBS) Blocking Buffer (LiCor, cat no. 92760010) and incubated with the following primary antibodies overnight at 4°C: mouse anti-PSD-95 (1:1000; Synaptic Systems, cat no. 124 011), rabbit anti-SynGAP (1:1000; as described above), rabbit anti-SRGAP2(A) (1:1000; as described above), and mouse anti-beta-actin (1:10000; EMD Millipore, #MAB1501). The next day, the membranes were washed in TBS-T and incubated with goat-anti-rabbit IgG conjugated to IRDye800CW (1:5000; Li-Cor, cat no. 926-32211) and goat-anti-mouse IgG conjugated to IRDye680RD (1:5000; Li-Cor, cat no. 926-68070). Imaging of immunoblots was performed using an Odyssey CLx imaging system (Li-Cor). Signal intensity was quantified using Fiji and normalized to beta-actin.

Quantification and statistical analysis

Data are presented as single values with mean + SD (Figures 2, 3, 4C, 4G–4I, 4K, S1, S2, S5G, S6, S7, S8, and S10B), mean + SD (Figures 6F, S3, and S10F), box-and-whisker plot (Figures 4D, 6C, S4E, S5C, and S5H), cumulative distribution (Figures 4E, 6D, and S5I), every value point (Figures 5, S4D, S4G, S5A, and S10G), mean + 95% confidence interval of the mean (Figures 1, S4D, and S4E), or SuperPlots (Figures S5D, S5E, S9, S10C, and S10D) (see Figures for details).

Numbers of litters, animals, neurons and dendritic spines used are described in Figure legends and specifically for the Figures 1D, 1E, S5A, and S5B in Table S1. For xenotransplantation experiments experimental groups were composed of neurons processed and transplanted in the same time, meaning animals were littermates.

For human synaptosome blots, 3-5 protein extractions have been performed per condition from independent differentiation batches.

For human immunoprecipitation experiments, 2 experiments have been performed from different individuals. We used 3 biopsies from a 34-year-old female (temporal lobe epilepsy), 22-year-old female (frontal lobe epilepsy), 45-year-old female (temporal lobe epilepsy).

For protein extraction experiments from in vitro cultures, 3-6 independent cultures have been used. For mouse protein extraction experiments, 3 mice per timepoint were used.

Data shown in Figure S8I originate from https://liwang.shinyapps.io/PSD_development_explorer/ and have been published in Wang et al.⁵⁰ They contain 54 human samples (from gestation week 18 to 18 year-old) males and females, 21 macaque samples (from embryonic day E75 to 10 year-old) males and females, and >20 mouse samples (from P0 to P36) males and females.

For statistical analyses, neurons were taken as a n for dendritic spine density, Sholl analysis, electrophysiology and proportion of positive dendritic spines for synaptic markers; dendritic spines were taken as a n for head width measurements, except for SuperPlots in which average values per neurons and animals are shown; independent neural cultures were taken as a n for in vitro experiments; animals were taken as a n for cortical fate maker proportions and mouse protein abundance linear regression.

Unpaired Mann-Whitney test have been used for assessing the significance of differences in the analyses between two conditions using GraphPad, because we measured one continuous variable (i.e. sEPSC amplitude) between two independent groups (i.e. shScramble vs shSRGAP2B/C), with independent observations between the two groups. 2-way ANOVA multiple tests have been used using GraphPad when groups can be split into two independent variables, because we measured one continuous variable (i.e. dendritic spine density) between two independent variables (e.g. genotype and time or PSC genotype and shRNA condition), we used one value per neurons and from different animals and litters, we used the Geisser-Greenhouse correction which involves no assumption on the equal variability of the differences or the Dunnett’s correction when compared to one single condition (e.g. non-infected condition). Kuskal-Wallis with Dunn’s multiple comparison tests have been used using GraphPad to compare one continuous quantitative variable (e.g. sEPSC amplitude) between several independent groups, because we didn’t assume Gaussian distribution and we assume independence of the observations between the groups. Kolmogorov-Smirnov tests have been used using GraphPad to assess the significance of differences in the analyses of cumulative distributions, because we assumed dendritic spine head width are continuous and that the groups compared are independent. Multiple Linear Regression tests were used using GraphPad to assess the significance of differences in the analyses of dendritic spine head width, taken genotypes, neurons, animals and litters as main effects.

Published: October 14, 2024