NMDA Receptors Regulate the Development of Neuronal Intrinsic Excitability through Cell-Autonomous Mechanisms

Maturation of neuronal and synaptic functions during early life is essential for the development of neuronal circuits and behaviors. In newborns synaptic transmission at excitatory synapses is primarily mediated by N-methyl-D-aspartate receptors (NMDARs), and NMDAR-mediated signaling plays an important role in synaptic maturation. Concomitant with synapse development, the intrinsic properties of neurons undergo dramatic changes during early life. However, little is known about the role of NMDARs in the development of intrinsic excitability. By using mosaic deletion of the obligatory GluN1 subunit of NMDARs in the thalamus of newborn mice, we showed that NMDARs regulate neuronal excitability during postnatal development. Compared with neighboring control neurons, neurons lacking NMDARs exhibit hyperexcitability and this effect is present throughout early life. Morphological analyses show that thalamic neurons without NMDARs have smaller soma size and fewer dendritic branches. Deletion of NMDARs causes a reduction of hyperpolarization-activated cation (HCN) channel function in thalamic neurons, and pharmacologically blocking HCN channels in wild type neurons mimics the effects of GluN1 deletion on intrinsic excitability. Deletion of GluN1 down-regulated mechanistic target of rapamycin (mTOR) signaling in thalamic neurons, and mosaic deletion of mTOR recapitulated the effects of GluN1 deletion. Our results demonstrate that NMDARs regulate intrinsic excitability and morphology of thalamic neurons through cell autonomous mechanisms that implicate mTOR signaling.


INTRODUCTION
Activation of N-methyl-D-aspartate receptors (NMDARs) by glutamate at excitatory synapses opens cation-selective channels with high permeability to Ca 2+ ions and generates excitatory postsynaptic currents (EPSCs) with a slow time course. NMDARs have critical functions in the development and plasticity of glutamatergic synapses (Malinow and Malenka, 2002;Song and Huganir, 2002). In the brain of newborns, glutamatergic synapses contain few or no α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs), and synaptic transmission is mediated primarily by NMDARs (Isaac et al., 1995(Isaac et al., , 1997Liao et al., 1995). Synaptic maturation is associated with an increase of AMPARs at the synapse, and this developmental change of AMPARs is regulated by NMDARs (Liao et al., 1999;Adesnik et al., 2008;Zhang et al., 2013). NMDARs are also required for pruning of redundant synaptic inputs during development (Ruthazer et al., 2003;Zhang et al., 2013).
Together with synaptic maturation, the intrinsic membrane properties of neurons undergo dramatic changes during early life. Cortical and thalamic neurons show negative shifts in resting potential and action potential threshold, an acceleration of action potential kinetics, and a reduction of input resistance during the first 3 weeks after birth (Ramoa and McCormick, 1994;Zhou and Hablitz, 1996;Franceschetti et al., 1998;Zhu, 2000;Zhang, 2004). Neuronal activity plays an important role in the development of intrinsic excitability (Spitzer and Ribera, 1998;Spitzer, 2006). In embryonic Xenopus spinal neurons, transient elevations of intracellular Ca 2+ caused by action potentials serve as a key messenger for the expression of ion channels (Gu and Spitzer, 1995). Visual stimulation in Xenopus tadpoles increases intrinsic excitability of optical tectal neurons (Aizenman et al., 2003). Activation of NMDARs causes Ca 2+ influx either directly through the channels or indirectly by triggering action potentials. However, the role of NMDARs in the development of intrinsic excitability of neurons is poorly understood.
Previous studies have shown that NMDARs regulate the development of neuronal morphology. In Xenopus embryos, pharmacological block of NMDARs inhibited activity-dependent dendritic arborization of optical tectal neurons (Rajan and Cline, 1998;Sin et al., 2002). The role of NMDARs in dendritic development has also been studied using knockout mice. NMDARs are generally heterotetramers consisting of two obligatory GluN1 subunits and a combination of two GluN2 subunits (GluN2A-D) (Cull-Candy et al., 2001). Selective deletion of GluN1 from all excitatory neurons in the cortex disrupted dendritic patterning of layer 4 neurons (Datwani et al., 2002). Similar results have been obtained with mosaic deletion of GluN2B, the primary GluN2 subunit in the developing brain (Espinosa et al., 2009), indicating that the effect of NMDARs on dendritic patterning is cellautonomous.
The hyperpolarization-activated current (I h ) plays important roles in the regulation of neuronal intrinsic excitability during development and in adult (Robinson and Siegelbaum, 2003;Bender and Baram, 2008;Biel et al., 2009;Noam et al., 2011). I h is a non-inactivating Na + /K + conductance activated by hyperpolarization at membrane potentials close to the resting membrane potential of neurons. Because of its unique properties, I h has an essential role in the regulation of resting membrane potential, input resistance, and other voltage-dependent mechanisms of neurons. I h -conducting HCN channels shows temporal and spatial patterns of expression in the brain (Moosmang et al., 1999;Bender et al., 2001;Notomi and Shigemoto, 2004). Developmental changes of HCN channel expression have been implicated in the maturation of network activity in the hippocampus (Bender et al., 2005). In thalamus, the upregulation of I h in thalamic neurons during early life is associated with increased expression of HCN1, 2 and 4 (Kanyshkova et al., 2009).
In the present study, we directly address the role of NMDARs in the development of intrinsic excitability of neurons in the thalamus. Because deletion of NMDARs from a large number of neurons can cause indirect effects due to changes in network activity, we use a genetic mosaic method to delete GluN1 from a fraction of neurons in thalamus, allowing us to assess cellautonomous functions of NMDARs. We find that NMDARs are required for the maturation of intrinsic excitability and morphology of thalamic neurons. This function of NMDARs is cell autonomous and implicates mTOR signaling and HCN channels.

Electrophysiology
Mice were anesthetized with tribromoethanol and decapitated. Brains were dissected quickly and sagittal sections 300 µm thick were prepared with a vibratome (Leica VT1200) and kept in artificial cerebral spinal fluid (ACSF) containing (in mM): 124 NaCl, 3.0 KCl, 1.5 CaCl 2 , 1.3 MgCl 2 , 1.0 NaH 2 PO 4 , 26 NaHCO 3 , and 20 glucose, saturated with 95% O 2 and 5% CO 2 at room temperature (21-23 • C). We used 1.5 mM Ca 2+ in ACSF because a previous study showed that the concentration of Ca 2+ in the cerebral spinal of young rats is about 1.5 mM (Jones and Keep, 1988).
For patch clamp recording, a slice was transferred to a submersion type chamber where it was continuously perfused with ACSF saturated with 95% O 2 and 5% CO 2 at 30-32 • C.
Neurons in the VPm were viewed with a 40x objective and Cre + neurons were identified by epifluorescence of tdTomato. Whole-cell recordings were obtained from the soma with a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA, United States). The pipette solution contained (in mM): 120 K-gluconate, 10 KCl, 4 ATP-Mg, 0.3 GTP-Na, 0.5 EGTA, and 10 HEPES (pH 7.2, 270-280 mOsm with sucrose). For current clamp recording, the series resistance (R s ) was fully compensated. For voltage clamp recording, R s , usually between 8 and 12 MΩ, was continuously monitored but not compensated. Data was discarded when R s changed by more than 30% during the experiment. Experiments were conducted using AxoGraph X (AxoGraph Scientific). Data were filtered at 4 kHz and digitized at 20 kHz. Data analysis was performed using AxoGraph X.
Intrinsic properties of neurons were analyzed using methods described previously (Zhang, 2004;Zhang and Arsenault, 2005). Resting membrane potentials were measured within 20 s of break in. Input resistance and time constant were calculated from voltage responses to 600-ms current steps of −100 pA. For the analyses of AP rheobase and F-I curves, current steps of 1 s in duration were applied once every 10 s, with increments of 10 or 20 pA. To minimize the effect of resting potential variation on rheobase and F-I curve, cells were held at −60 mV between current steps. In some experiments, synaptic blockers picrotoxin, kynurenic acid, and 6,7-Dinitroquinoxaline-2,3-dione (DNQX; all from Tocris) were applied through bath perfusion.
Voltage-dependency of I h activation was analyzed using a voltage-clamp protocol similar to that described in a previous study (Kanyshkova et al., 2009). This protocol takes into consideration voltage-dependent changes in the onset rate of I h . Initially held at −50 mV, the membrane potential of the cell was stepped to increasingly hyperpolarizing levels for several seconds to reach steady-state activation, then to −100 mV. The activation curve of I h was established by measuring the tail currents at 40 ms after stepping to −100 mV. The normalized I h , p(V), was calculated using the equation: where I min and I max are the amplitudes of tail currents for pre-pulses at −50 and −150 mV, respectively. The activation curves of I h were fitted to a Boltzmann equation: where V h is the membrane potential of half-maximal activation, and k is the slope factor.

Single Cell Labeling and Reconstruction
Neurons were labeled during whole-cell recording with 0.3% biocytin in the patch pipette. Slices were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (PB) overnight and processed as described previously (Zhang and Deschenes, 1997;Zhang, 2004). Briefly, slices were treated with 0.3% H 2 O 2 in 10% MeOH in 0.1 M PB to inactivate endogenous peroxidase. After wash, slices were processed for biocytin histochemistry using the ABC Elite Standard Kit (Vector Laboratories, Burlingame, CA, United States), 3,3-diaminobenzidine, and nickel ammonium sulfate. Cells were reconstructed and analyzed with the Neurolucida System (MicroBrightfield, Williston, VT, United States).
Data were analyzed using ImageJ. Changes in cell body size were quantified by measuring soma area. Confocal stacks of thalamic neurons immunostained for NeuN were z-projected at maximum intensity. Cre + neurons were identified by the expression of the reporter protein tdTomato. Cell bodies of individual neurons were outlined using the polygon selection tool and soma area was analyzed. The same method was used to quantify immunofluorescence of p-Akt and p-S6. The mean fluorescence intensity of immunofluorescence staining was measured for the soma.

Statistical Analysis
Statistical analyses were performed with JMP software (SAS, Cary, NC, United States). Data were presented either as median with interquartile range and the minimum and maximum (box and whisker) or as mean ± SEM. Statistical differences were determined using non-parametric Wilcoxon Rank Sum Test (two groups) or Krustal-Wallis Test (three or more groups). P-values were provided in the results or figure legends.

Mosaic Deletion Demonstrates Cell-Autonomous Role of NMDA Receptors in Regulating Intrinsic Excitability
N-methyl-D-aspartate receptors are ubiquitously expressed in the brain during development. Pharmacological block or genetic deletion of NMDARs from a large number of neurons disrupts activities of neuronal circuits. To distinguish cell-autonomous vs. non-autonomous effects, we performed mosaic deletion of NMDARs in the thalamus of newborn mice by using the conditional allele of the NMDAR subunit GluN1 (Grin1 fl ) (Zhang et al., 2013) and the transgenic serotonin transporter Cre strain (SERT-Cre) (Gong et al., 2007;Zhang et al., 2013). Our previous study showed that in SERT-Cre; Grin1 fl/fl mice, about 50% of VPm neurons express Cre and lack NMDARs, and the loss of NMDAR function occurs between P7 and P10 (Zhang et al., 2013). The expression of Cre in VPm neurons was random. To identify neurons with or without Cre expression, we used the Rosa26-tdTomato reporter (Rosa Ai14 ) (Madisen et al., 2010) to express the fluorescent marker tdTomato in Cre + neurons ( Figure 1A). We first recorded VPm neurons in acute slices obtained from SERT-Cre; Grin1 fl/fl ; Rosa Ai14 mice at P14-16, the age at which VPm neurons reach maturity in synaptic connectivity (Arsenault and Zhang, 2006;Wang et al., 2011). We applied NMDA to the recorded neurons to confirm the absence or presence of NMDARs (Figures 1B,C). Intrinsic excitability was analyzed with whole-cell patch clamp recording.
Previous studies have shown that SERT is transiently expressed in all VPm neurons during early life (Lebrand et al., 1998;Narboux-Neme et al., 2008), so it is unclear why only a fraction of VPm neurons in the SERT-Cre strain express Cre. To exclude the possibility that SERT-Cre is expressed in a subpopulation of VPm neurons with distinct intrinsic excitability, we recorded Cre + neurons in the VPm of SERT-Cre; Grin1 fl/+ ; Rosa Ai14 mice at P14-16. All Cre + neurons (Grin1 fl/+ ; Cre + ) in these Grin1 heterozygous mice had normal NMDAR function. We did not find any difference in intrinsic excitability between Grin1 fl/+ ; Cre + neurons (Figures 1F,G,I-L, in green) and Grin1 fl/fl ; Cre − neurons, indicating that in the presence of NMDARs, Cre + and Cre − neurons are indistinguishable in membrane property and excitability. Together, these findings suggest that the deletion of NMDARs alters excitability of VPm neurons through cell-autonomous mechanisms.
The experiments described above were conducted in the absence of any synaptic blockers. Deletion of NMDARs may alter spontaneous synaptic responses, which in turn, affect neuronal excitability. To examine this possibility, we performed recordings in the presence of glutamatergic blockers DNQX and kynurenic acid, and the GABAergic blocker picrotoxin. The differences in excitability between mutant and control neurons persisted in the presence of these synaptic blockers (Supplementary Figures S1A-E). These results show that the effects of NMDAR deletion on excitability in thalamic neurons are independent of spontaneous synaptic transmission.
Thalamic relay neurons fire bursts of APs when stimulated from hyperpolarized membrane potentials. We analyzed rebound bursts following hyperpolarizing current steps. Both mutant and control neurons showed rebound bursts when hyperpolarized to −70 mV (Figure 2A). There was no difference in the number of APs in a burst between mutant (3.1 ± 0.3, n = 11 cells) and control (2.8 ± 0.3, n = 12 cells; p = 0.39, Wilcoxon test). However, the average firing frequency was significantly reduced in mutant neurons ( Figure 2B).
Deletion of NMDARs may cause a delay in the maturation of neuronal excitability. To examine this possibility, we analyzed VPm neurons at P10-11 and P20-21. The effects of NMDAR deletion on excitability and membrane properties were already present at P10-11. Neurons deficient of NMDARs showed a decrease in rheobase ( Figure 3A) and an increase in the slope of the F-I curve ( Figure 3B), together with an increase in input resistance ( Figure 3C) and a decrease in whole-cell capacitance ( Figure 3D). These effects are comparable to those observed at P14-16 and P20-21 (Figures 3A-D). These results suggest that NMDARs are required for postnatal development of intrinsic excitability in thalamic neurons.
Our previous study showed that deletion of NMDARs downregulates synaptic AMPARs in VPm neurons (Zhang et al., 2013). The amplitudes of AMPAR-mediated EPSCs were reduced by 80% in VPm neurons deficient of NMDARs. To determine whether the effects of NMDAR deletion on intrinsic excitability are mediated by changes in AMPAR function, we used mice deficient of GluA3 (Gria3) and GluA4 (Gria4), two major AMPAR subunits for the thalamus. We have previously shown that deletion of both Gria3 and Gria4 reduces AMPAR-mediated EPSCs by 80% in VPm neurons without any changes in NMDARmediated EPSCs (Wang et al., 2011). Consistent with our previous finding, AMPAR-mediated EPSCs were dramatically reduced in VPm neurons from mice deficient of GluA3 and GluA4 ( Figure 4A). However, intrinsic excitability of VPm  The average firing frequency of bursts recorded from mutant (229 ± 9 Hz, n = 11) and control (308 ± 16 Hz, n = 12) neurons ( * * * p = 0.0007, Wilcoxon test). Data were collected from 8 mice aged P14-16.
neurons was not affected by the deletion of both GluA3 and GluA4 ( Figure 4B). There was no difference between mutant and control neurons in the rheobase, slope of the F-I curve, input resistance, or whole-cell capacitance (Figures 4C-F). These findings suggest that the effects of NMDAR deletion on intrinsic excitability are not caused by the reduction in AMPAR functions.

NMDARs Regulate Cell Size and Dendritic Morphology
Neuronal morphology is an important determinant of cellular excitability. The reduction of whole-cell capacitance in NMDARdeficient neurons indicates a decrease in cell size. We further examined the effect of NMDAR deletion on cell body size at P16-18. NeuN staining was used to label neuronal somata and soma area was quantified. In the VPm of mosaic mutant mice (SERT-Cre; Grin1 fl/fl ; Rosa Ai14 ), Cre + mutant neurons ( Figure 5A, left panel, cell body encircled by while dash line) had smaller soma size than neighboring Cre − control neurons (cell body encircled by yellow dash line) ( Figure 5B, red vs. gray). To determine whether SERT-Cre expression by itself may affect soma size, we also analyzed soma area in the VPm of heterozygous control mice (SERT-Cre; Grin1 fl/+ ; Rosa Ai14 ) at P16-18 ( Figure 5A, right panel). There was no difference in soma area between Cre + and Cre − in Grin1 heterozygous mice ( Figure 5B, green vs. blue). Importantly, there was no difference between control neurons (Cre − ) in Grin1 homozygous mice and neurons (Cre + or Cre − ) in Grin1 heterozygous mice ( Figure 5B). These results demonstrate that NMDARs regulate cell size in the thalamus through cell-autonomous mechanisms.
To determine the effects of NMDAR deletion on dendritic morphology, we labeled VPm neurons from SERT-Cre; Grin1 fl/fl ; Rosa Ai14 mice at P14-16 with biocytin in the whole-cell recording pipettes. Dendritic branches of individual VPm neurons were reconstructed and analyzed ( Figure 5C). The total dendritic length was significantly shorter in mutant neurons ( Figure 5D). Mutant neurons also had few dendritic branches than control neurons ( Figure 5E). We performed Sholl analysis of dendritic morphology. The distribution of Sholl intersections was significantly different between mutant and control neurons ( Figure 5F). The majority of Sholl intersections were much closer to the cell body for mutant than control neurons. These results suggest that NMDAR signaling regulates the growth of dendritic branches of thalamic neurons.

HCN Channels Are Implicated in the Regulation of Intrinsic Excitability by NMDARs
Hyperpolarization-activated cation channels have important functions in the regulation of neuronal excitability and these channels are up-regulated during early life in many parts of the brain including the thalamus (Kanyshkova et al., 2009). We first analyzed voltage sag generated by I h in VPm neurons. Control and mutant neurons were hyperpolarized by current steps from −60 to −75 mV ( Figure 6A). Mutant neurons had smaller voltage sag than control neurons (Figures 6A,B), indicating a down-regulation of HCN channel function in mutant neurons.
Next, we analyzed I h in VPm neurons using voltage clamp recording. Compared with control neurons, mutant neurons had much smaller I h (Figures 6C,D). In contrast, Grin1 deletion did not significantly affect the kinetics of I h activation (720.5 ± 28.5 ms, n = 20 cells for mutant vs. 651 ± 27.8 ms, n = 19 cells for control, p = 0.074, Wilcoxon test) or inactivation (240.0 ± 5.8 ms vs. 232.2 ± 6.0 ms, p = 0.49, Wilcoxon test). To determine whether NMDAR deletion alters voltage dependency of HCN channels, we analyzed the current-voltage (I-V) curve for steady-state I h using tail current measurement ( Figure 6E). The I-V relationship of steady-state I h was not significantly different between mutant and control neurons ( Figure 6F). The density of I h , estimated using the maximal tail current, was significantly reduced in mutant neurons (7.9 ± 0.5 pA/pF, n = 17 cells for mutant vs. 9.7 ± 0.5 pA/pF, n = 15 cells for control, * p = 0.034, Wilcoxon test). Together, these results suggest that deletion of NMDARs reduces the number of HCN channels in thalamic neurons.

NMDA Receptors Regulate Intrinsic Excitability through mTOR Signaling
Deletion of NMDARs disrupts many molecular pathways. Because of the reduction in cell size, we hypothesized that the mTOR pathway, known to play a critical role in cell growth (Hay and Sonenberg, 2004), is implicated in NMDARmediated regulation of intrinsic excitability. The PI3K/Akt pathway is upstream of mTOR activation. We first analyzed Akt phosphorylation in the VPm of SERT-Cre; Grin1 fl/fl ; Rosa Ai14 mice by measuring immunofluorescence of phosphorylated Akt (p-Akt) in single cells. We found that mutant neurons had significantly lower p-Akt signal than neighboring control neurons (Figures 7A,B). This result indicates that deletion of NMDARs down-regulated Akt function in thalamic neurons.
A key downstream effector of the mTOR signaling is the ribosomal protein S6. Activation of mTOR phosphorylates S6 through the S6 kinase (Hay and Sonenberg, 2004). We analyzed S6 phosphorylation in the VPm of SERT-Cre; Grin1 fl/fl ; Rosa Ai14 mice with immunofluorescence using an antibody against phosphorylated S6 (p-S6). Mutant neurons had much lower p-S6 signal than neighboring control neurons (Figures 7C,D). Together, our results show that deletion of NMDARs downregulates Akt/mTOR signaling in VPm neurons.
Next, we examined the effects of Mtor deletion on intrinsic excitability. We used the SERT-Cre and the conditional Mtor allele (Risson et al., 2009) (Figures 8A-J). Mutant neurons showed hyperexcitability with lower rheobase (Figure 8C), steeper slope of the F-I curve (Figure 8D), higher input resistance (Figure 8E), and lower whole-cell capacitance ( Figure 8F). Mutant neurons deficient also exhibited a more depolarized resting potential than control neuron ( Figure 8G) but no change in AP threshold ( Figure 8H). As in the case of NMDAR-deficient neurons, neurons deficient in mTOR had smaller I h (Figure 8I), but no change in the I-V relationship of I h activation ( Figure 8J). We also analyzed changes in cell size in fixed sections from SERT-Cre; Mtor fl/fl ; Rosa Ai14 mice at P17. Soma areas of mutant neurons were significantly smaller than those of control neurons (196.9 ± 2.9 µm 2 , n = 103 vs. 249.6 ± 5.6 µm 2 , n = 66; * * * * p < 0.0001, Wilcoxon test). To examine the possibility that the deletion of Mtor down-regulates NMDARs, we analyzed NMDAR-mediated EPSCs in VPm neurons in SERT-Cre; Mtor fl/fl ; Rosa Ai14 mice aged P15-16. There was no difference between mutant and control neurons in NMDAR-mediated EPSCs (Supplementary Figure S2), indicating that the deletion of Mtor does not alter NMDAR function in these cells. Together, these results support the role of mTOR signaling in NMDAR-mediated regulation of intrinsic excitability.

DISCUSSION
Previous studies have demonstrated that NMDARs have important functions in the development of dendrites and synapses. Here we provide genetic evidence that NMDARs are required for the development of intrinsic excitability of neurons. Mosaic deletion of Grin1 in thalamic neurons causes hyperexcitability and a reduction of cell size through cell-autonomous mechanisms. Our results suggest that NMDAR-mediated signaling is a key mechanism controlling the maturation of neurons.
Neuronal activity plays an important role in the maturation of intrinsic properties during development (Spitzer, 2006). As the predominant glutamatergic receptor in neonatal brain, NMDARs have essential functions in the regulation of activity of neuronal circuits during development. However, the role of NMDARs in the maturation of intrinsic neuronal excitability is not well (F) I-V curves of steady-state I h using the tail currents recorded from control (n = 12) and mutant (n = 11) neurons from 11 mice at P14-16. Data from control (black empty circles) or mutant (red dots) were fitted to a Boltzmann equation. V h , the membrane potential of half-maximal activation, was -88.5 ± 1.8 mV for control and -88.6 ± 3.8 mV for mutant cells (NS, p = 0.23, Wilcoxon test). (G) Effects of ZD7288 (20 µM, bath application) on firing patterns of a VPm neuron in a WT mouse at P16. (H) Rheobase of VPm neurons before and during ZD7288 application ( * * * p < 0.001, Wilcoxon test). (I) Slope of the F-I curve of VPm neurons before and during ZD7288 application ( * * * p < 0.001, Wilcoxon test). Data were from 14 neurons in 6 WT mice at P14-16. understood. Single-cell deletion of all AMPARs and NMDARs in the hippocampus of newborn mice caused hyperexcitation of CA1 pyramidal neurons (Lu et al., 2013), but it was unknown whether the effect was due to the lack of AMPARs or NMDARs. Deletion of NMDARs from most of the excitatory neurons in the forebrain using a Camk2a-Cre also led to increased intrinsic excitability of pyramidal neurons in adult mice (Tatard-Leitman et al., 2015). However, because of the late onset of the Camk2a-Cre expression (Tsien et al., 1996), the deletion of NMDARs in these mice probably occurs after 3 weeks of age, when the intrinsic properties of pyramidal neurons have reached the mature level. Furthermore, the Camk2a conditional Grin1 knockout mice also showed significant increases in cortical excitatory synaptic transmission and network activity, which could indirectly affect neuronal excitability. By using mosaic Grin1 deletion in the thalamus during early postnatal development, we find that ablation of NMDARs leads to hyperexcitability in mutant neurons without any effect on neighboring control neurons. Our results provide the first genetic evidence for cell-autonomous functions of NMDARs in the regulation of intrinsic neuronal excitability.
Intrinsic excitability of the neuron is determined by its morphology and the expression of various ion channels. Previous studies using single-cell gene deletion have shown either no or subtle effect of NMDAR deletion on neuronal morphology. Single-cell deletion of Grin1 from CA1 pyramidal neurons in the hippocampus had no effect on dendritic length or branch points (Adesnik et al., 2008). Mosaic deletion of Grin2B, the predominant GluN2 subunit in the developing brain, disrupted dendrite patterning of granule neurons in the dentate gyrus and spiny stellar neurons in cortical layer 4 (Espinosa et al., 2009). Neurons without Grin2b had more primary dendrites than control neurons although the total length of dendrites and branch numbers did not change. Increased dendritic branching was also observed in cortical layer 4 neurons in mice with cortex-restricted deletion of Grin1 (Datwani et al., 2002). It was unclear in these studies whether cell body size was affected by NMDAR deletion. Together, these previous results suggest that NMDARs have important functions in dendrite patterning but is not required for the growth of dendrites in these neurons.
In contrast, our results show that NMDARs are required for neuronal growth in thalamic neurons. VPm neurons deficient of NMDARs have smaller cell body size, fewer dendritic branches, and shorter dendrites. The reduction in cell body size is consistent with our findings of higher input resistance in mutant neurons. Deletion of Grin1 in cortical pyramidal neurons had no effect on input resistance (Tatard-Leitman et al., 2015), indicating that the morphology of these cells was not significantly affected by NMDAR deletion. On the other hand, selective deletion of Grin1 from GABAergic neurons in the cortex disrupted dendrite growth in GABAergic interneurons expressing Reelin (De Marco Garcia et al., 2015). Thus, the function of NMDARs in the development of neuronal morphology is likely to be cell-type specific.
Several studies have shown that HCN channels have important roles in activity-dependent regulation of intrinsic excitability in Mtor mutant neurons (n = 11) from 5 mice. V h was -87.6 ± 3.7 mV for control and -86.9 ± 3.3 mV for mutant cells (NS, p = 0.84, Wilcoxon test). the brain. Theta-burst stimulation of hippocampal pyramidal neurons, which induced long-term potentiation, also led to a persist decrease of intrinsic excitability (Fan et al., 2005). This reduction in intrinsic excitability required NMDAR activation and was mediated through upregulation of I h . HCN channels are highly expressed in dendrites (Tsay et al., 2007;Huang et al., 2009;Vaidya and Johnston, 2013). Sensory deprivation caused dendritic hyperexcitability in layer 5 pyramidal neurons of the somatosensory cortex and the effect was associated with a decrease of I h (Breton and Stuart, 2009), although whether this effect requires NMDARs is unknown. We find that deletion of NMDARs in thalamic neurons causes a decrease of I h and pharmacological block of I h in wild type neurons mimics the effects of NMDAR deletion on intrinsic excitability. Our results support a role of NMDARs in activity-dependent regulation of HCN channel function and intrinsic excitability.
The exact molecular mechanisms by which NMDARs regulate neuronal morphology and intrinsic excitability are not well understood. Previous studies have shown that NMDAR activation is required in activity-dependent up-regulation of Akt/mTOR signaling in hippocampal neurons (Cammalleri et al., 2003;Gong et al., 2006;Pirbhoy et al., 2016). We find that thalamic neurons deficient of NMDARs have reduced levels of p-Akt and p-S6, indicating that deletion of NMDARs causes a down-regulation of Akt/mTOR signaling in these neurons.
Supporting the role of mTOR signaling, we find that mosaic deletion of Mtor from the thalamus recapitulates the effects of NMDAR deletion on intrinsic excitability and cell body size. Our results suggest that a major cellular function of NMDARs during development is to regulate protein synthesis in activitydependent manner through mTOR signaling.
Several recent studies have shown that abnormal upregulation of mTOR signaling during development affects morphology and excitability of neurons. Tuberous sclerosis complex 1 (Tsc1) suppresses mTOR signaling, and loss of function mutations of Tsc1 cause tumors and a variety of neurological symptoms including intellectual disability, seizures, and autism (Caban et al., 2017). In the mouse, single-cell deletion of Tsc1 in the neocortex during embryonic development increased the level of p-S6 and soma size in the affected neurons (Feliciano et al., 2011). Increased soma size was also observed in the thalamus using mosaic deletion of Tsc1 (Normand et al., 2013). Deletion of Tsc1 at E12.5 in thalamic neurons reduced input resistance and the slope of the F-I curve (Normand et al., 2013), indicating that hyperactivity of mTOR signaling during embryonic development leads to hypo-excitability. We find that deletion of Mtor in thalamic neurons during development leads to opposite changes in membrane properties and cell firing. Our results demonstrate that mTOR is required for the development of neuronal excitability.
Mutations of GRIN1 have been reported in patients with neurodevelopmental disorders including early onset epilepsy, severe intellectual disability, motor dysfunction, and generalized cerebral atrophy (Ohba et al., 2015;Lemke et al., 2016). Although the mutations are predicted to disrupt NMDAR function, the pathophysiology of the diseases is unknown. Our results suggest that in addition to synaptic dysfunction, changes in neuronal morphology and intrinsic excitability may have important roles. Future studies are needed to analyze the effects of human GRIN1 mutations on function of neurons and neuronal circuits.

AUTHOR CONTRIBUTIONS
Z-WZ and GH designed research. GH performed research. GH and Z-WZ analyzed data. Z-WZ wrote the paper.

FUNDING
This work was supported by National Institutes of Health Grant NS064013. Core services at The Jackson Laboratory were supported by CORE grant (CA34196).