The autism-associated gene Scn2a contributes to dendritic excitability and synaptic function in prefrontal cortex

Scn2a loss impairs axonal and dendritic excitability in distinct developmental periods.

To determine how Scn2a haploinsufficiency affects cortical neuron excitability, acute slices containing medial prefrontal cortex (mPFC) were prepared from Scn2a^+/− and wild-type (WT) littermate control mice aged postnatal day (P)4 through P64. Scn2a^+/− mice have a 50% reduction in Na_V1.2 mRNA, have reduced Na_V-mediated currents in dissociated cell culture at P5–9 (Planells-Cases et al., 2000), and are convulsive-seizure free (Mishra et al., 2017; Ogiwara et al., 2018), consistent with the majority of Scn2a loss-of-function cases reported in ASD (Sanders et al., 2018). We focused our studies primarily on subcortically projecting layer 5b thick-tufted neurons that were identified based on electrophysiological properties (Clarkson et al., 2017; Dembrow et al., 2010; Gee et al., 2012) (see Methods), as dysfunction within this region and cell class has been implicated in ASD (Willsey et al., 2013). Neuronal excitability was assessed using a series of current steps to generate APs. AP threshold was depolarized and spike output was reduced relative to WT in the first postnatal week (Fig. 1A, C, F). Differences between Scn2a^+/− and WT were not present thereafter (Fig. 1B, C, F), consistent with increased Na_V1.6 expression in the AIS. In agreement, persistent sodium currents, which reflect the recruitment of AIS sodium channels with the lowest threshold (Taddese and Bean, 2002), were suppressed in Scn2a^+/− mice at P6, but not at P30 (Fig. S1A–G). Furthermore, axonal conduction, assayed by imaging AP-evoked calcium transients in boutons, was reliable in both immature and mature neurons (Fig. S2). Input resistance was comparable between Scn2a^+/− and WT at all ages, and transient and sustained potassium currents were not different in both immature and mature neurons (Fig. S1H–K). Therefore, Scn2a haploinsufficiency impairs neuronal excitability early in development without functional compensation from the remaining Scn2a allele or other ion channels involved in AP initiation.

Although impairments in AP threshold recovered after P7, a closer examination of the AP waveform revealed a striking reduction in the velocity of the AP (dV/dt) that became more pronounced as neurons matured (Fig. 1E–F) (>P22, WT: 578.2 ± 9.2 V/s, n = 44 cells; Scn2a^+/−: 417 ± 9.5, n = 33 cells; p < 0.001, Mann-Whitney). Similar relative deficits were identified in putative thin-tufted neurons, which have different AP waveform characteristics compared to thick-tufted neurons (Dembrow et al., 2010) (>P18: WT: 426.9 ± 24.7 V/s, n = 8 cells; Scn2a^+/−: 358 ± 21.1, n = 9 cells; p = 0.007, Mann-Whitney). Plotting AP velocity as a function of voltage (phase-plane) highlights two components of the rising phase of the AP (Fig. 1D), as measured with somatic current-clamp. The first relates to the initiation of the AP in the AIS and the second to the recruitment of somatic sodium channels (Bean, 2007). In Scn2a^+/− cells, both AIS and somatodendritic components of the AP were slower, suggesting that Na_V1.2 channels are distributed in both the AIS and somatodendritic compartment, and that their density is lower in both compartments. Similar AP speed deficits were observed in pyramidal cells located in other neocortical regions, including medial orbitofrontal cortex and primary visual cortex (Fig. S3A–D). In contrast to neocortical neurons, no alterations in AP threshold and peak speed were apparent in hippocampal CA1 pyramidal neurons (Fig. S3E–F). This is consistent with immunostaining indicating that hippocampal pyramidal cell dendrites predominantly express Na_V1.6 (Lorincz and Nusser, 2010).

Immunostaining for Na_V1.2 in neocortex suggests that Na_V1.2 channels are predominantly localized to excitatory neurons, with little to no membrane localization observed in parvalbumin or somatostatin positive interneurons (Li et al., 2014; Yamagata et al., 2017). These cell classes instead express Na_V1.1 channels, and exhibit hypoexcitability in Scn1a haploinsufficient mice aged P14–21 (Favero et al., 2018; Tai et al., 2014). To further test whether Na_V1.2 contributes to AP excitability in these interneuron classes, we crossed Scn2a^+/+ and Scn2a^+/− mice to mice engineered to express tdTomato in either parvalbumin or somatostatin positive cells (see Methods). Fluorescently guided whole-cell current-clamp recordings were then made from these populations in slices prepared from P34–40 animals (Fig. 2). Though AP waveforms differed across cell class, neither peak dV/dt (Fig. 2; PV peak dV/dt: WT: 336 ± 19 V/s, n = 9 cells, Scn2a^+/−: 326 ± 21 V/s, n = 9 cells; SOM peak dV/dt: WT: 362 ± 27 V/s, n = 12 cells, Scn2a^+/−: 313 ± 23 V/s, n = 11 cells; p = 0.43, F_genotype(1,37) = 0.65, 2-way ANOVA) or threshold (PV threshold: WT: −51.7 ± 1.6 mV, n = 9 cells, Scn2a^+/−: −51.9 ± 0.7 mV, n = 9 cells; SOM threshold: WT: −39.7 ± 0.7 mV, n = 12 cells, Scn2a^+/−: −41.6 ± 1.0 mV, n = 11 cells; p = 0.41, F_genotype(1,37) = 0.68, 2-way ANOVA) were affected by loss of Na_V1.2.

Overall, these data indicate that Scn2a haploinsufficiency alters the excitability of neocortical pyramidal cells, but not parvalbumin or somatostatin positive interneurons. Furthermore, within pyramidal cells, Na_V1.2 channels may be more consequential for somatodendritic excitability than axonal excitability. To explore the cellular consequences of reduced somatodendritic excitability, we modified an established computational model of cortical pyramidal cells (Hallermann et al., 2012) by varying the localization and relative density of Na_V1.2 and Na_V1.6 (Ben-Shalom et al., 2017). First, Na_V1.2 channels were localized only to the AIS, and AP waveform was compared between models with 100% and 50% Na_V1.2 expression (Fig. 3A). AP waveform was only modestly altered by such manipulations. To best account for empirical observations, we found that Na_V1.2 expression was required throughout the somatodendritic compartment, in equal levels with Na_V1.6 (Fig. 3A, Fig. S4A–B, D–E). In this configuration, we modeled AP waveform throughout the dendritic arbor and found that backpropagating APs (bAPs) from the AIS to the dendritic arbor were attenuated in a distance dependent fashion. Though AP amplitudes were reduced only 10% in the soma, they were attenuated to far greater extents in distal dendrites (41% of WT amplitude in distal tuft) (Fig. 3B, Fig. S5D), suggesting that cellular processes that depend on bAPs may be altered in Scn2a^+/− cells.

Backpropagating APs can provide instructive signals for dendritic integration and synaptic plasticity, largely through their ability to elicit calcium influx through dendritic voltage-dependent calcium channels and NMDA receptors (Feldman, 2012; Larkum et al., 1999a; Stuart and Häusser, 2001). To determine whether bAP-associated dendritic excitability was impaired in Scn2a^+/− neurons, we imaged calcium transients evoked by trains of high-frequency bursts (2APs at >100Hz) at various locations throughout the apical dendrite of layer 5 pyramidal cells. Bursts of APs, rather than single APs, were chosen because they have been shown to evoke dendritic calcium transients throughout apical dendrites, including distal tuft regions (Barth et al., 2008; Gulledge and Stuart, 2003; Larkum et al., 1999b; Short et al., 2017). In WT neurons, bursts reliably evoked calcium transients throughout the apical dendrite. By contrast, calcium transients in Scn2a^+/− neurons rapidly diminished in amplitude with increasing distance from the soma, becoming virtually absent in the most distal dendritic branches (Fig. 3C–D). To determine whether these effects were due to acute loss of Na_V channels, rather than a consequence of altered excitability during development, we partially blocked sodium channels in mature WT neurons with tetrodotoxin (TTX), using a concentration that mimicked the reduced AP speed observed in Scn2a^+/− neurons (5 nM, Fig. S4G–H). Consistent with acute effects of Scn2a haploinsufficiency, bAP-evoked calcium transients were reduced to a comparable extent in TTX-exposed WT neurons (Fig. 3C–D). Thus, Scn2a^+/− haploinsufficiency leads to a major, persistent deficit in dendritic excitability in adult L5 neurons that is likely due to loss of Na_V1.2 channels localized to the dendrite.

Excitatory synapses are impaired in mature Scn2a^+/− neurons.

Impairments of both axonal and dendritic excitability could affect synapses, which are a common locus of dysfunction in mouse models of ASD (Bourgeron, 2015; Monteiro and Feng, 2017; Tsai et al., 2012). We tested this by first measuring miniature excitatory and inhibitory postsynaptic currents (mEPSC, mIPSC) at P6 and P27, capturing developmental periods of axonal and dendritic excitability deficits, respectively. Neither mEPSCs nor mIPSCs were affected at P6. At P27, mEPSC frequency was reduced by 48% (WT: 9.3 ± 1.1 Hz, n = 23 cells; Scn2a^+/−: 4.9 ± 0.8, n = 19 cells; p < 0.001 Mann-Whitney), with no change in mEPSC amplitude (Fig. 4A–B) or mIPSC amplitude or frequency (S6A–B).

Lower mEPSC frequencies could reflect reductions in release probability or the number of functional synapses contributing to AMPA-mediated mEPSCs. The paired pulse ratio, which is sensitive to differences in release probability, was not different between Scn2a^+/− and WT cells (Fig. 4C, S5C–D). By contrast, the ratio of AMPA- to NMDA-mediated current in evoked EPSCs was reduced in Scn2a^+/− cells (Fig. 4D; WT: 5.5 ± 0.7, n = 8 cells: Scn2a^+/−: 3.2 ± 0.2, n = 12 cells; p < 0.01, Mann-Whitney). Taken together, the reduction in both mEPSC frequency and AMPA:NMDA ratio suggest that mature Scn2a^+/− neurons have an excess of AMPA-lacking spines that are more commonly observed at earlier developmental time points (Hanse et al., 2013; Kerchner and Nicoll, 2008).

Consistent with these functional observations, dendritic spine morphology was also altered in mature, but not immature, Scn2a^+/− cells, though overall dendritic arbor structure was unaltered (Fig. 5A–B). At P5–6, filipodial spines identified on developing apical dendrites were similar in morphology and density in wildtype and Scn2a^+/− neurons (Fig. 5C–D). In more mature neurons (>P23), spine density was comparable between WT and Scn2a^+/− on both apical and basal dendrites; however, Scn2a^+/− spines were longer and had smaller heads relative to their total head and neck volume (Fig. 5C–D), thus sharing features with spines more commonly found in less mature cells.

To determine whether impaired synaptic function was due to early developmental deficits in axonal excitability or persistent deficits in dendritic excitability, we engineered a mouse with one Scn2a allele under Cre-loxP control (Scn2a^+/fl) (Fig. S6A). Mice were first crossed to the CaMKIIα-Cre driver line, which expresses Cre in neocortical pyramidal cells only after P10 (Fig. S6B, Xu et al., 2000). As such, Scn2a^+/fl::CaMKIIα-Cre mice develop with normal AP threshold and spike output. In Scn2a^+/fl::CaMKIIα-Cre mice, peak dV/dt was not different at P18, likely due to a combination of low Cre expression and slow Na_V1.2 turnover. By P50, peak dV/dt matched constitutive Scn2a^+/− neurons (Fig. 6B). Furthermore, bAP-evoked dendritic Ca transients were suppressed to comparable extents as observed in constitutive Scn2a^+/− mice (Fig. 6C–D). Thus, these mice developed without early deficits in axonal excitability and exhibited impaired dendritic excitability only later in development.

Strikingly, AMPA:NMDA ratio was reduced in Scn2a^+/fl::CaMKIIα-Cre neurons when measured after P50, revealing that impaired dendritic excitability alone is sufficient to alter synapse strength (Fig. 6E; WT: 4.2 ± 0.3, n = 12 cells; floxed: 2.5 ± 0.4, n = 11 cells; p < 0.01, Mann-Whitney). To test whether these synaptic deficits result from a cell-autonomous loss of Scn2a, we injected a dilute Cre-expressing adeno-associated virus (AAV-EF1α-Cre-mCherry) into mPFC of P28 Scn2a^+/fl mice (Fig. 6A). Four weeks later, we found that both peak dV/dt and AMPA:NMDA ratio were reduced in Crepositive, but not Cre-negative neurons (Fig. 6B, E; WT: 5.0 ± 0.4, n = 10 cells; floxed: 2.4 ± 0.4 n = 9 cells; p < 0.01, Mann-Whitney). Modest increases in spine length were also noted in spines along the apical dendrites of Scn2a^+/fl::CaMKIIα-Cre cells (Fig. S7). These results indicate that the persistent, cell-autonomous, dendritic function of Na_V1.2 is required to maintain aspects of normal synaptic function in mature pyramidal cells.

These changes in synapse structure and function could arise if synapses lack instructive signals for maintaining synaptic strength. Indeed, physiologically relevant forms of synaptic plasticity depend on the coincident detection of bAPs and glutamate at the synapse, a process that may be affected in Scn2a^+/− cells (Feldman, 2012; Magee and Johnston, 1997; Markram et al., 1997, 2012). To test whether plasticity is impaired in these cells, we induced long-term potentiation (LTP) of putative apical dendrite synapses by pairing layer 1 fiber stimulation with bursts of APs (Fig. 7A, Kampa et al., 2006; pairing protocol: Tzounopoulos et al., 2004). In WT neurons, paired stimulation resulted in long-lasting synaptic potentiation. In contrast, LTP was not observed in either Scn2a^+/− neurons or in WT neurons in 5 nM TTX (Fig. 7B–D, EPSP slope, normalized to baseline: WT: 1.52 ± 0.10, n = 11 cells; Scn2a^+/−: 0.98 ± 0.11, n = 11 cells; TTX: 0.97 ± 0.06, n = 9 cells; p = 0.0002, Kruskal-Wallis test; WT vs Scn2a^+/−: p = 0.0007, WT vs TTX: p = 0.002, Dunn’s multiple comparisons test). Thus, Scn2a haploinsufficiency impairs synaptic plasticity, consistent with impairments in bAP-evoked calcium transients in the dendrites.

Scn2a^+/− mice have trends towards deficits in learning and social behavior

Given these physiological deficits and the association of loss-of-function SCN2A mutations with ASD and ID in humans, we next asked if Scn2a haploinsufficiency results in corresponding behavioral impairments in mice. As mouse models harboring other ASD-associated genetic variants exhibit a range of behavioral differences (Pasciuto et al., 2015; Silverman et al., 2010), we screened Scn2a^+/− mice and WT littermates of both sexes (5–13 mice per sex and genotype combination) through a behavioral panel designed to assess locomotion, anxiety, repetitive behavior, sociability, and learning (Fig. S8). In total, 10 assays were performed yielding 30 metrics that we assessed for genotypic differences. While no individual metric differed significantly between the wildtype and heterozygote mouse after correcting for multiple comparisons, we observed a trend towards impaired reversal learning in a water T-maze task in males (Fig. S8A). In this task, mice must first learn which arm of a maze contains a submerged platform (day 1–4, 4 trials/day), then reverse their association when the platform location is switched (day 5–8, 6 trials/day). A replication experiment of two further batches of male (23 vs. 24 mice total) and female (20 vs. 24 mice total) Scn2a^+/− and WT littermates showed a similar trend toward reversal learning impairments (Fig. S9A, p = 0.026, F_Genotype (1, 45) = 5.2, 2-way repeated measures ANOVA; alpha required for multiple comparisons in replication cohort: 0.003), while no trends were observed in females.

We also observed trends suggestive of social impairments in female mice (Fig. S9B–D). Specifically, in our initial screen we observed no difference in social approach behavior in a two-chamber social task, where mice had the option of interacting with a caged, sex-matched stimulus mouse in one chamber or an inanimate toy mouse in the other (Fig. S8B). In a second cohort, mice were paired with the same stimulus mouse for three trials to assess social approach and habituation. Both sexes and genotypes showed a comparable preference to the stimulus mouse vs the toy mouse across the three trials (Fig. S8B). Given the importance of social interaction in ASD, we also tested interactions when the toy mouse was replaced with a novel stimulus mouse on the fourth trial. Both male and female WT mice switched preference to the new stimulus mouse whereas female Scn2a^+/− mice showed no preference for the novel mouse (Δinteraction time:, Scn2a^+/+: −17.1 ± 4.1 sec, n = 13 mice; Scn2a^+/+: 9.3 ± 10.3 sec, n = 8 mice, p = 0.017; Mann-Whitney) (Fig. S9D). Female Scn2a^+/− mice were also found to enter the open arms of the elevated plus maze more often (percent time in open arms: Scn2a^+/+: 22.9 ± 2.1%, n = 24 mice; Scn2a^+/+: 34.3 ± 2.7% sec, n = 20 mice, p = 0.002; Mann-Whitney), with a trend towards spending more time in these open arms (number of open arm entries: Scn2a^+/+: 56.6 ± 3.3, n = 24 mice; Scn2a^+/−: 73.6 ± 5.5 n = 20, p = 0.012; Mann-Whitney) (Fig. S9F). Overall, these data suggest that Scn2a^+/− mice may display sex-specific impairments in behavioral flexibility and social discrimination.

Na_V1.2 channel expression in neocortical pyramidal cell dendrites

While the expression of Na_V1.2 in the AIS is well established (Gazina et al., 2015; Hu et al., 2009; Liao et al., 2010), there is less consensus on its distribution in pyramidal cell dendrites. The first immunostaining of Na_V1.2 in neocortex revealed somatic and apical dendritic localization in pyramidal cells, though this may represent a pool of newly synthesized channels within vesicles (Gong et al., 1999). Functionally, somatodendritic Na_V currents in neocortical pyramidal neurons are best described by Na_V1.2 channels (Hu and Bean, 2018; Hu et al., 2009). Here, we found that AP waveform was altered in constitutive and conditional Scn2a^+/− prefrontal thick-tufted layer 5 pyramidal cells in a manner best explained by somatodendritic loss of Na_V1.2, with no functional compensation from the residual Scn2a allele or Na_V1.6 (Fig. 1–2, S1, S4). Similar observations were made in prefrontal thin-tufted layer 5 pyramidal cells, and in thick-tufted layer 5 pyramidal cells localized to orbitofrontal and primary visual cortex (Fig. S3). Moreover, AP-evoked calcium transients in PFC pyramidal neurons were suppressed throughout the dendritic arbor of Scn2a^+/− cells (Fig. 3, 6), and this effect could be mimicked in wildtype neurons acutely by sub-saturating concentrations of TTX.

In mature hippocampal preparations, Na_V1.2 channels have been visualized in pyramidal cell initial segments, and also were co-localized with the presynaptic terminal marker VGlut1 in strata containing unmyelinated axons, but were not present in pyramidal cell dendrites (Lorincz and Nusser, 2010, but see Johnson et al., 2017). Consistent with this observation, AP waveform was not altered in recordings from CA1 hippocampal neurons (Fig. S3). This suggests that Na_V1.2 channels are selectively found in the dendrites of neocortical, but not archicortical, pyramidal cells. This hypothesis will require further work to measure dendritic excitability directly in hippocampal neurons, and to rule out compensation by other Na_V subtypes. Furthermore, there may be regional differences in expression in other neocortical regions between frontal and visual cortex, or in other laminae. Altogether, these data indicate that Na_V1.2 is functionally expressed in the soma and dendrites of neocortical pyramidal neurons, where they play an important role in excitability and synapse function.

Role of Na_V1.2 in dendritic and synaptic function

Scn2a haploinsufficiency selectively impairs dendritic excitability and synaptic function in mature PFC pyramidal cells, with no effects on AP initiation or propagation (Fig. 1, S2). This distinguishes Na_V1.2 from other Na_V isoforms expressed in the mature cortex that either support overall excitability in inhibitory neurons (Na_V1.1) or contribute to both action potential generation and dendritic excitability in pyramidal neurons (Na_V1.6) (Hu et al., 2009; Lorincz and Nusser, 2008; Ogiwara et al., 2007). In the mature brain, Na_V1.2 may instead play a major role in supporting the backpropagation of action potentials within neocortical pyramidal cell dendrites. These bAPs depolarize dendrites and can contribute to the generation of dendritic spikes, especially when paired with local synaptic stimulation (Larkum et al., 1999a), or when packaged in high-frequency bursts (Barth et al., 2008; Gulledge and Stuart, 2003; Larkum et al., 1999b). Here, we found that somatodendritic Na_V1.2 channels are critical for supporting bAPs from the AIS to the distal dendrites, and that Scn2a haploinsufficiency impairs the ability to induce a burst-based form of associative spike-timing dependent plasticity at apical tuft synapses (Figs. 3C–D, 5C, 7). Furthermore, excitatory synapse function was altered in ways most consistent with a postsynaptic, dendritic role for Na_V1.2 channels in mature neurons. Presynaptic metrics, including the likelihood of evoking calcium transients in axonal boutons and paired-pulse ratio, were unaffected by Scn2a loss (Fig. S2, S5). By contrast, miniature EPSC frequency and AMPA:NMDA ratio were reduced, consistent with the hypothesis that a higher proportion of silent synapses are present in Scn2a^+/− cells (Fig. 4). These effects were mirrored by structural differences in spines imaged on basal and apical dendrites of more mature neurons (Fig. 5C–D). Interestingly, these functional and structural aspects were dissociable, as induction of Scn2a haploinsufficiency late in life reduced AMPA:NMDA ratio with little change in spine shape (Figs. 6E, S7). This is consistent with other cell- and circuit-based perturbations in more mature neurons, where synapse strength can be altered markedly with relatively small changes in spine morphology (Dobi et al., 2011; Sala et al., 2003; Shen et al., 2014), and suggests that certain aspects of spine morphology are developmentally regulated before the ages at which Scn2a loss was induced in Scn2a^+/fl mice studied here.

While we focused on backpropagation of bursts and related plasticity in the tuft of Scn2a^+/− cells, reductions in Na_V1.2 density likely affect many other dendritic processes. For example, dendrites are capable of generating spikes independent of APs at the AIS through multiple mechanisms, including Na_Vs (Kim et al., 2015; Stuart et al., 1997). These dendritic spikes alone can be instructive for synaptic plasticity (Golding et al., 2002; Kampa et al., 2007). Furthermore, plasticity mechanisms that influence synaptic strength in dendrites more proximal to the soma are also likely affected, given the structural changes observed in basal dendrites and in EPSCs evoked with electrical stimulation in layer 5. High-frequency burst-based stimuli used here readily produced calcium transients in proximal dendritic locations in Scn2a^+/− cells. Other activity that drives dendritic calcium influx, including single APs or coordinated activation of excitatory inputs (Ariav et al., 2003), may instead be affected in these regions.

Measurements of persistent sodium current suggest that the loss of Na_V1.2 channels is not functionally compensated for by Na_V1.6. Consistent with this, mRNA for Na_V1.6, Na_V1.1, and Na_V beta subunits are unaltered in Scn2a^+/− mice (Ogiwara et al., 2018). We also found no differences in transient or sustained potassium currents or input resistance, suggesting that other channel types are similarly unaltered by Na_V1.2 reductions. However, interpretation of this is hampered by the inability to adequately clamp these currents via somatic recording. While empirical measurements indicate that we have reasonable voltage control over persistent sodium currents in the AIS (Fig. S1E–G), other events, including synaptic activity and potassium currents, may suffer from incomplete voltage control in these large dendritic cells (Beaulieu-Laroche and Harnett, 2018; Williams and Mitchell, 2008). Importantly, voltage errors are likely comparable across genotype, since we do not observe differences in dendritic morphology, input resistance, and spine density (Figs. 5, S1). Furthermore, effects of Na_V1.2 loss on bAP-associated dendritic calcium transients were recapitulated by acutely blocking a small fraction of Na_Vs in WT cells. Thus, even though we have not assessed the function of ion channels in the apical tuft directly, our data are most easily explained by a reduction in Na_V1.2 alone, with no compensation from other Na_Vs or other ion channel classes.

Relationship with other ASD-associated genes

While SCN2A’s ASD association has been established for some time (Sanders et al., 2012), the mechanistic underpinnings of this strong association have been a mystery. Since Na_V1.2 channels were best understood for their role in AP initiation (Hu et al., 2009), it was difficult to find parallels with other ASD-associated genes that are associated with synaptic function and gene transcription (De Rubeis et al., 2014). Data shown here instead suggest that SCN2A function in ASD converges at the dendrite with the large group of ASD-associated genes involved in synaptic function. Indeed, disruptions in both excitatory synaptic function and neuronal excitability have been reported in other ASD-associated genes (Bateup et al., 2011; Bozdagi et al., 2010; Contractor et al., 2015; Greer et al., 2010; Meredith and Mansvelder, 2010; Nestor and Hoffman, 2012; Williams et al., 2015; Yashiro et al., 2009; Yi et al., 2016). How the other ASD-associated genes with roles in regulating gene expression mediate ASD risk remains unknown (Sanders, 2015), however we note that SCN2A has been reported as a regulatory target of both FMRP and CHD8 (Darnell et al., 2011; Golden et al., 2019; Sugathan et al., 2014).

SCN2A was one of the first genes identified with a clear excess of missense mutations in addition to predicted protein truncating variants linked to ASD and ID (Ben-Shalom et al., 2017; Sanders et al., 2015). Recent efforts have identified other genes harboring ASD-associated missense variants, though the functional implications of such mutations are largely unclear (Geisheker et al., 2017). Na_V1.2 function, by contrast, can be well characterized, allowing for a better understanding of how missense mutations that alter channel activity affect neuronal function. Several recurring SCN2A missense variants block sodium flux (Begemann et al., 2019; Ben-Shalom et al., 2017). These may be functionally identical to protein truncating variants, since we and others find that Na_V1.2 function is reduced 50% in Scn2a^+/− mice (Planells-Cases et al., 2000), with no evidence for functional compensation. Other SCN2A missense variants observed in ASD patients alter channel voltage dependence or kinetics, reducing neuronal excitability in models of developing cortical pyramidal cells (Ben-Shalom et al., 2017). Whether they result in similar dendritic deficits in mature neurons remains to be tested.

Role of Scn2a in learning and behavior

The behavioral differences observed thus far in heterozygous Scn2a mice parallel aspects of behavioral phenotypes reported in other mouse models of ASD-associated genes (Pasciuto et al., 2015; Silverman et al., 2010). Here, we observed trends toward impaired behavioral flexibility in males and impaired social recall and reduced anxiety in females. Reversal learning effects we observed in male mice parallel spatial learning impairments observed previously in male Scn2a^+/− mice (Middleton et al., 2018). Whether female Scn2a^+/− are similarly resistant to these specific spatial learning impairments has not been explored. Male Scn2a^+/− mice were also recently found to display novelty-induced exploratory hyperactivity, social behavior impairment, enhanced fear conditioning, and deficient fear extinction (Tatsukawa et al., 2019). Of note, several phenotypic results, including open field, elevated plus maze, and social interaction, were not replicated here. This issue of reproducibility is unfortunately common in behavioral studies examining ASD-like phenotypes in mouse, and may relate to differences in genetic background, genetic drift over generations, or experimental conditions (Brunner et al., 2015; Chadman et al., 2008; Ey et al., 2012; Spencer et al., 2011).

Children with SCN2A loss-of-function variants display a range of phenotypes, including ASD, ID, and cortical visual impairments. Here, we found that neurons in visual cortex have similar deficits in AP waveform, possibly reflecting deficits in dendritic excitability that could impair visual processing (Xu et al., 2012). Furthermore, a fraction of these children also develop seizure phenotypes, typically after the first year of life (Sanders et al., 2018; Wolff et al., 2017). Scn2a^+/− mice have been reported to be seizure free (Mishra et al., 2017; Planells-Cases et al., 2000), or to have low-frequency, short duration spike-wave discharges (Ogiwara et al., 2018). The variability of seizure occurrence in human and mouse suggests that factors other than Scn2a loss alone, including genetic background, may contribute to seizure susceptibility.

In conclusion, we found that haploinsufficiency induction at two different developmental timepoints, both well after Na_V1.2 channels cease to contribute to AP initiation, was sufficient to alter synaptic function (Fig. 3, 6). While this result does not exclude additional effects on network function due to reduced axonal excitability in early cortical development, it does indicate that proper Na_V1.2 function that supports dendritic excitability is required for maintaining synaptic strength. Thus, restoring proper Na_V1.2 function, even in the mature brain, may be an enticing avenue for therapeutic intervention in loss-of-function SCN2A cases.

CONTACT FOR REAGENT AND RESOURCE SHARING

EXPERIMENTAL MODEL AND SUBJECT DETAILS

All experimental procedures were performed in accordance with UCSF and Gladstone Institutes IACUC guidelines. All experiments were performed on mice housed under standard conditions with ad libitum access to food and water. C57B6J (JAX: 000664), B6.Cg-Tg(Camk2a-cre)T29–1Stl/J (Jax: 005359), Ai14: 6;129S6-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J (JAX: 007908) were obtained from The Jackson Laboratory and bred in our onsite colony. Scn2a^+/− mice were provided by Drs. E. Glasscock and M. Montal (Mishra et al., 2017; Planells-Cases et al., 2000). Scn2a^+/fl mice were created by inserting LoxP sites around exon 2 of the Scn2a gene (Fig, S6A, Cyagen). A targeting vector was generated by PCR using BAC clone RP23–332C13 and RP24–42717 from the C57BL/6J library to create homology arms around a LoxP flanked cKO region that includes exon 2 of the Scn2a gene. The targeting vector included a Neo cassette flanked by Rox sites for positive selection, and DTA outside of the homology arms for negative selection. The linearized vector was subsequently delivered to ES cells (C57BL/6) via electroporation, followed by drug selection, PCR screening, and Southern Blot confirmation. Confirmed clones were then introduced into host embryos and transferred to surrogate mothers. Chimerism in the resulting pups was identified via coat color. F0 male chimeras were bred with C57BL/6 females to generate F1 heterozygous mutants that were identified by PCR.