Neuroligin 1 modulates striatal glutamatergic neurotransmission in a pathway and NMDAR subunit-specific manner

Together with its presynaptic partner Neurexin 1 (Nxn1), Neuroligin 1 (NL1) participates in synapse specification and synapse maintenance. We and others have shown that NL1 can also modulate glutamatergic synaptic function in the central nervous system of rodent models. These molecular/cellular changes can translate into altered animal behaviors that are thought to be analogous to symptomatology of neuropsychiatric disorders. For example, in dorsal striatum of NL1 deletion mice, we previously reported that the ratio N-methyl-D-aspartate receptor (NMDAR) mediated synaptic currents to α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate receptor (AMPAR) mediated synaptic currents (NMDA/AMPA) is reduced in medium spiny neuron (MSNs). Importantly, this reduction in NMDA/AMPA ratio correlated with increased repetitive grooming. The striatum is the input nucleus of the basal ganglia (BG). Classical models of this circuitry imply that there are two principal pathways that render distinct and somewhat opposite striatal outputs critical to the function of these nuclei in modulating motor behavior. Thus, we set out to better characterize the effects of NL1 deletion on direct and indirect pathways of the dorsal striatum by genetically labeling MSNs participating in the direct and indirect pathways. We demonstrate that a decrease in NMDAR-mediated currents is limited to MSNs of the direct pathway. Furthermore, the decrease in NMDAR-mediated currents is largely due to a reduction in function of NMDARs containing the GluN2A subunit. In contrast, indirect pathway MSNs in NL1 knockout (KO) mice showed a reduction in the frequency of miniature excitatory neurotransmission not observed in the direct pathway. Thus, NL1 deletion differentially affects direct and indirect pathway MSNs in dorsal striatum. These findings have potential implications for striatal function in NL1 KO mice.

Together with its presynaptic partner Neurexin 1 (Nxn1), Neuroligin 1 (NL1) participates in synapse specification and synapse maintenance. We and others have shown that NL1 can also modulate glutamatergic synaptic function in the central nervous system of rodent models. These molecular/cellular changes can translate into altered animal behaviors that are thought to be analogous to symptomatology of neuropsychiatric disorders. For example, in dorsal striatum of NL1 deletion mice, we previously reported that the ratio N-methyl-D-aspartate receptor (NMDAR) mediated synaptic currents to α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate receptor (AMPAR) mediated synaptic currents (NMDA/AMPA) is reduced in medium spiny neuron (MSNs). Importantly, this reduction in NMDA/AMPA ratio correlated with increased repetitive grooming. The striatum is the input nucleus of the basal ganglia (BG). Classical models of this circuitry imply that there are two principal pathways that render distinct and somewhat opposite striatal outputs critical to the function of these nuclei in modulating motor behavior. Thus, we set out to better characterize the effects of NL1 deletion on direct and indirect pathways of the dorsal striatum by genetically labeling MSNs participating in the direct and indirect pathways. We demonstrate that a decrease in NMDAR-mediated currents is limited to MSNs of the direct pathway. Furthermore, the decrease in NMDAR-mediated currents is largely due to a reduction in function
The basal ganglia (BG) are an essential component of larger parallel circuits implicated in modulation of lower-order motor control and higher-order habit formation (Villablanca and Marcus, 1975;Yin and Knowlton, 2006;Graybiel, 2008). BG circuitry is segregated in two main pathways with differential gene expression (Gerfen et al., 1990;Lobo et al., 2006) and, most importantly, with outputs of more or less opposite effects on motor and cognitive ''programs'' (Villablanca and Marcus, 1975;Yin and Knowlton, 2006;Graybiel, 2008). The output projections from the striatum are originated in medium spiny neurons (MSNs) of either the striatonigral (direct) pathway or the striatopallidal (indirect) pathway. Imbalances in the output of these pathways are thought to be critical in the generation of repetitive/perseverative-like behaviors in animals and humans.
In an effort to better define synaptic dysfunction in striatal circuits in the setting of NL1 deletion, we crossed genetic markers of direct and indirect pathway MSNs into our NL1 KO mice. These markers allow us to examine pathway-specific effects of NL1 deletion on dorsal striatum circuitry. Interestingly, we have demonstrated pathway-specific alterations in glutamatergic inputs onto direct and indirect pathway MSNs. Specifically, alterations in NMDAR-mediated currents are limited to the direct pathway while different synaptic abnormalities were demonstrated in the indirect pathway. We further refine the mechanism of reduced NMDAR function by implicating GluN2A-containing receptors. Here we replicate our previous work (Blundell et al., 2010) and extend those findings to show pathway-specific and subunit-specific alterations in NMDAR currents as well as other aspects of striatal synaptic function.

Mice
All experiments were performed in accord with the University of Texas Southwestern Medical Center Animal Care and Use Committee in compliance with National Institutes of Health guidelines for the care and use of experimental animals. The NL1 KO mouse was generated as described previously (Varoqueaux et al., 2006). The background of our original NL1 mutant line is a hybrid 129S6/SvEVTac X C57BL/6J cross. These mice were crossed once to mice expressing distinct fluorescent markers in the direct pathway (NL1/DR1 mice, td-Tomato driven by the dopamine-1 receptor promoter; Shuen et al., 2008) and in the indirect pathway (NL1/DR2 mice, enhanced green fluorescent protein driven by the dopamine-2 receptor promoter; Gong et al., 2003). In an effort to preserve the original hybrid background, we then backcrossed each of these ''labeled'' mice for five generations into the original hybrid NL1 mutant background. In order to obtain mice in which both D1 and D2 MSNs were labeled, we crossed heterozygous NL1 mice with the fluorescent markers to one another for three generations. NL1 heterozygous F3 mice expressing markers in both pathways were then used as breeders to generate WT and NL1 KO individuals for experiments as well as NL1 heterozygous mice to replace breeders. These crosses created the added benefit of uniform genetic background across all experiments.
Recently the use of bacterial artificial chromosome (BAC) transgenic mice expressing EGFP driven by the DR2 promoter had been called into question because, when comparing to wildtype (WT) mice, an increased expression of the D2 receptor as well as a mild hyperactive phenotype were found in these mice but not in BAC transgenic mice expressing either EGFP or td-Tomato in the direct pathway Kramer et al., 2011). Later reports from several groups suggest that strain-specific, high-sensitivity may have been the culprit for those findings (Chan et al., 2012;Nelson et al., 2012). In the present study, comparisons are made between WT and NL1 KO mice that both express either D1 or D2 labeled BAC transgenes. Thus, any effects of these transgenes are equivalent in both WT and NL1 KO groups. Furthermore, our use of hybrid genetic backgrounds may obviate this potential confound (Gertler et al., 2008;Grueter et al., 2010;Lobo et al., 2010;Chan et al., 2012;Nelson et al., 2012).

Genotyping
Genotyping was performed by PCR of genomic DNA extracted from tails. For NL1 genotyping the following primers were used: MB0606 (WT-specific), 5 -CGA GAG TCA GGT AAA TTG AAC ACC AC-3 ; T676 (KO-specific), 5 -GAG CGC GCG CGG CGG AGT TGT TGA C-3 ; and T1660 (common), 5 -GTG AGC TGA ATC TTA TGG TTA GAT GGG-3 . Importantly, given that amplification of the mutant PCR fragment was significantly more efficient than the amplification of the WT PCR fragment, we used a 3:1-4:1 ratio of WT-specific primer vs. the KO-specific primer. PCR products were run in a 1.6-1.8% agarose gel. WT and KO PCR product sizes were: ∼550 bp and ∼415 bp, respectively. The PCR protocol was: 1 step at 98 • C for 3 min; 30 cycles of 98 • C for 15 s, 65 • C for 30 s and 72 • C for 20 s; a final step at 72 • C for 5 min. Genotyping for DR1-tdTomato transgene or for DR2-EGFP transgene, was done using same procedures as in Ade et al. (2011).

Striatal Neurotransmission
Horizontal-oblique slices were prepared from NL1/D1 and D2 mouse brains, and acute slice striatal recordings were performed as previously described (Ding et al., 2008;Smeal et al., 2008;Blundell et al., 2010). Briefly, 2-3 week old WT and KO mice were anesthetized with ketamine/xylazine (20%/10%, respectively) in 0.85% (w/v) saline cocktail, perfused intracardially with 10-20 ml of artificial cerebrospinal fluid (aCSF), and decapitated. The brain was quickly isolated and chilled in dissecting solution. The dissecting solution contained the following (in mM): 75 sucrose, 81 NaCl, 2.5 KCl, 1.0 NaH 2 PO 4 , 0.1 CaCl 2 , 4.9 MgCl 2 , 26.2 NaHCO 3 , and 1.5 glucose (300-305 mOsm). Slices 360 µm thick were cut using a model 300 Vibratome. Slices were incubated in the bathing solution at 32 ± 1 • C for 30-40 min. Slices were then stored at room temperature until transferred to a submersion-type recording chamber after at least 80 min of recovery following slicing. The bathing solution (regular aCSF) contained (in mM): 124 NaCl, 3 KCl, 1.25 NaH 2 PO 4 , 2 CaCl 2 , 1 MgSO 4 , 26 NaHCO 3 , and 1 glucose (saturated with 95% O 2 /5% CO 2 ). Whole-cell patch-clamp recordings from striatal tdTomato-or from eGFP-expressing MSNs were performed using micropipettes with a resistance of 3-5 M made from 1.1/1.5 mm borosilicate glass (Sutter Instruments). Recording pipettes were filled with the following solution (in mM): 117 Cs-methanesulphonate, 2.8 NaCl, 5 TEA-Cl, 2 QX-314 [N-(2,6dimethylphenylcarbamoylmethyl)triethylammonium chloride], 0.4 EGTA, 2 ATP-Mg, 0.25 [275][276][277][278][279][280][281][282][283][284][285]. A junction potential of 12 mV between the internal and the external solution was calculated theoretically (using the corresponding function in Clampfit), corroborated empirically, and used to correct voltages online. Access resistance was frequently checked to be <30 M and stable (less than 20% variability). Protocols were designed to study input/output curves, NMDA/AMPA ratio, miniature EPSCs and, paired pulse ratio (PPR; see below). At all times, at least five breeding pairs were kept as resource for experimental mice. For each genotype/striatal pathway/protocol, useful data sets were obtained from six or more mice from three or more litters, where 1-3 cells per slice/pathway and 1-5 cells per mouse/pathway were recorded. Results from a total of 185 WT EPSCs were evoked (eEPSCs) by stimulating the brain at the boundary of the corpus callosum and the dorsolateral striatum. Stimuli were elicited for 0.2 ms using bipolar electrodes and a model A365 battery-driven stimulus isolator (WPI). For input/output (I/O) experiments, the lowest stimulation intensity to elicit events in at least three out of five stimulations was considered the threshold for the experiment. The intensity of the stimulus was then scaled 1.5× the previous intensity and in some experiments, up to 25.6× threshold (labeled as ''Threshold Fold'' in Figure 1C). I/O linearity was usually lost in stimulation intensities around 17.4× threshold; therefore, I/O analysis was done on stimulation intensities that ranged from threshold to 11.4× threshold. Evoked NMDA/AMPA ratios were determined using standard, published methods (Myme et al., 2003). In the presence of picrotoxin, AMPAR currents were measured at the peak and at a voltage of −80 mV, at which most NMDAR currents are blocked by Mg 2+ . Stimulation intensities were adjusted so the peak amplitude of the AMPA current was maintained within the 200-600 pA (mostly within 300-500 pA) range. NBQX (10 µM), an AMPA receptor antagonist, was added to the bath after recording AMPAR currents. In the same cell, NMDAR currents were measured in a 2 ms window 48 ms after stimulus onset at a voltage of +40 mV. In a subset of these experiments Ro 25-6891 (0.5-1 µM), an antagonist of GluN2Bcontaining NMDARs, was added to the bath. To determine the current contribution of GluN2B-containing NMDARs vs. the current contributed by putative GluN2A-containing NMDARs, the remaining current after application of Ro 25-6891 (the putative GluN2A component) was subtracted off-line from the total NMDAR current to obtain the GluN2B-component. PPR was done at two inter-event intervals (IEIs) that corresponded to the maximal facilitation (40 ms) or the maximal depression (400 ms; Akopian and Walsh, 2007;Ding et al., 2008). Stimulation intensities were adjusted so the peak amplitude of the first pulse was maintained within the 150-300 pA range. Even though stimulus intensities were intended to induce a homogenous conditioning stimulus across groups, the resulting peak current amplitudes for P1 differed among them (not shown). Therefore, to cancel this artificial difference, all events within a group were normalized to the average current amplitude for P1. These normalized data was then used for statistical analysis to compare P1 vs. P2 between genotypes and genders and presented in Table 2 as normalized event amplitude 1 of short inter-pulse interval (NE1S) vs. NE2S or as normalized event amplitude 1 of long inter-pulse interval (NE1L) vs. NE2L. For recording of mEPSCs, the bathing solution was supplemented with Tetrodotoxin (0.75-1 µM) and Picrotoxin (50-100 µM). After breaking in, the internal solution was dialyzed for at least 8 min, then cells were held at −70 mV and spontaneous mEPSC transmission was recorded for 2-5 min or until at least 200 events were obtained.

Statistical Analysis
All statistical analyses were conducted using Statistica software (version 5.5; Statsof, Tulsa, OK, USA) or Prism software (version 6.02; GraphPad, La Jolla, CA, USA). Statistica was used either for two-way ANOVAs or for three-way repeated measures ANOVA (rmANOVA). Sex and genotype were the two factors in two-way ANOVAs. In addition, for the three-way ANOVA, either the stimulus threshold fold (STF; for I/O curves), or the pulse number (for PPR) were used. If a main effect of sex was found, the Tukey Honest Significant Difference with unequal N (the Spjotvoll-Stoline) post hoc test was applied. Both male and female mice were used for the experiments. For the purpose of these studies, where the effects of constitutively deleting the NL1 protein are central to our investigations, it was required that the Spjotvoll-Stoline test rendered a significant difference when comparing genders within the same genotype. Otherwise, any statistically significant ''main effect of sex'' obtained in the analysis was considered either spurious or of little value to understand the importance of NL1-deletion in striatal pathophysiology. Importantly, not a single parameter fulfilled the above requirement (see Tables 1, 2), therefore, genders were pooled throughout this work. When appropriate, Graphpad Prism software was used for linear regression analysis where slope = r 2 and probability are reported. For ease of flow of data presentation, only rmANOVAs corresponding to the analysis of I/O curves are presented as part of the main text. For the rest of the data, average ± SEM are presented, and P values correspond to those from the ANOVA analysis. P values between 0.1005 and 0.0504 were considered of potential interest to readers and were included in the figures. For further reference, a complete display of the statistical results for ANOVAs and rmANOVAs is presented in Tables 1, 2.

Strength of Glutamatergic Inputs onto MSNs of the Direct and of the Indirect Pathways
Mice constitutively deficient in NL1 show alterations in glutamatergic synaptic function (Chubykin et al., 2007;Blundell et al., 2010). These mice, however, demonstrate increased repetitive grooming behaviors that are potentially mediated by the striatum along with a decrease in NMDA/AMPA ratio at excitatory inputs onto striatal MSNs (MSNs; Blundell et al., 2010). This suggests either a decrease in NMDARmediated current, an increase in AMPAR-mediated current or some combination of the two. Measuring NMDA/AMPA ratio is also not useful for determining baseline synaptic strength. Thus, to compare synapse strength between genotypes, we examined input/output (I/O) curves of evoked excitatory synaptic transmission onto both DR1 (direct pathway) and DR2 (indirect pathway) MSNs of NL1 KO mice compared to WT littermate controls ( Figure 1A). Synaptic currents were evoked by stimulation in the nearby corpus callosum. Recordings were done at a membrane potential of −80 mV, where the slowly activating NMDA receptors are blocked by extracellular Mg 2+ ions. Due to this and because current amplitudes were determined at the peak of the eEPSCs, this experimental protocol mainly determines the strength of AMPA receptor signaling. Typically, stimulation thresholds to elicit a recognizable eEPSC response in three out of five stimuli were between the 15-30 µA. To reduce the likelihood of evoking polysynaptic events due to stimulation of clusters of fibers distant to the stimulating electrode, experiments with thresholds above 60 µA were not used in I/O analysis. Using these parameters, maximum stimulation intensities never exceeded 700 µA. No main effect of gender or of genotype was found in either pathway (Figures 1B-D; I/O three way rmANOVA. DR1 MSNs: main effect of sex, F (1,30) = 0.21, P = 0.652; main effect of genotype, F (1,30) = 0.01, P = 0.937; main effect of STF, F (6,180) = 60.11, P < 0.0001; sex × genotype interaction, F (1,30) = 3.46, P = 0.073; sex × STF interaction, F (6,180) = 1.24, P = 0.289; genotype × STF interaction, F (6,180) = 0.07, p = 0.999; sex × genotype × STF interaction; F (6,180) = 2.24, P = 0.041; DR2 MSNs: main effect of sex, F (1,33) = 0.63, P = 0.433; main effect of genotype, F (1,33) 0.09, P = 0.768; main effect of STF, F (6,198) = 43.24, P < 0.0001; sex × genotype interaction, F (1,33) = 0.41, P = 0.529; sex × STF interaction, F (6,189) = 0.34, P = 0.917; genotype × STF interaction, F (6,189) = 0.05, P = 0.999; sex × genotype × STF interaction; F (6,189) = 0.66, P = 0.686). In line with previous reports in other brain regions on NL1 KO mice (Chubykin et al., 2007;Jung et al., 2010;Soler-Llavina et al., 2011), these experiments suggest that baseline, largely AMPAR-mediated glutamatergic evoked synaptic responses are not altered in the striatum of NL1 KO mice.

NMDA/AMPA Ratio is Selectively Reduced in the Direct Pathway
Both intracellular and extracellular domains of NL1 are thought to modulate clustering of NMDARs at the synapse (Barrow et al., 2009;Budreck et al., 2013). This is in line with the strong correlation between NMDA/AMPA ratio (recorded in cells in culture as well as in acute slices) and NL1 expression levels manipulated by diverse strategies. In general, overexpression increases the NMDA/AMPA ratio whereas knockdown or knockout strategies result in decreases in NMDA/AMPA ratio (Chubykin et al., 2007;Futai et al., 2007;Kim et al., 2008b;Jung et al., 2010;Soler-Llavina et al., 2011;Kwon et al., 2012;Budreck et al., 2013). Because NL1 KO mice exhibit reduced NMDA/AMPA ratio in MSNs in the striatum (Blundell et al., 2010), we further examined NMDA/AMPA ratio specifically in DR1 or in DR2 MSNs of WT and NL1 KO mice.
To investigate if this could be happening in our recording conditions, we performed regression analysis comparing the total NMDAR Q among all experiments within the same genotype and pathway to the Q fraction of their corresponding GluN2A or GluN2B components. This analysis indicated no correlation between normalized total NMDAR and the Q fraction driven by either GluN2A-or GluN2B-containing NMDARs, indicating that significant differences in the relative amount of neurotransmitter released among these experiments (partially reflected in the size of the total NMDAR Q) do not exist, and, therefore, do not drive our main observation ( Figure

The Frequency of Spontaneous mEPSC Neurotransmission is Reduced in the Indirect Pathway of NL1KO Mice
In vitro and in vivo experimental procedures that affect NL1 expression levels can impact spontaneous miniature excitatory neurotransmission (Prange et al., 2004;Nam and Chen, 2005;Chen et al., 2010;Mondin et al., 2011;Burton et al., 2012;Kwon et al., 2012). Furthermore, studies in cortex and hippocampus have shown that NL1 levels modulate synaptic vesicle accumulation and neurotransmission (Wittenmayer et al., 2009;Stan et al., 2010), and synaptic release probability (Futai et al., 2007). Thus, we reasoned that constitutive deletion of NL1 may affect spontaneous miniature excitatory neurotransmission in the striatum. To investigate this, we compared the properties of miniature excitatory synaptic transmission in both DR1 and DR2 MSN pathways between NL1 KO and WT mice. NL1 KO mice showed no difference in mEPSC neurotransmission in the direct pathway (WT amplitude, 10.3 ± 0.34 pA, n = 18 vs. KO amplitude, 11 ± 0.48 pA, n = 18, P = 0.850; WT frequency, 6.1 ± 0.70 Hz, n = 18 vs. KO frequency, 5.5 ± 0.76 Hz, n = 20, P = 0.755). The lack of mEPSC amplitude difference is consistent with our conclusion that the direct pathway alterations in NMDA/AMPA ratio are mediated by altered NMDAR-mediated currents as no change in the largely AMPAR-mediated mEPSC was identified. However, mEPSC frequency was reduced in the indirect pathway (Figures 4A,B; WT frequency, 5.6 ± 0.92 Hz, n = 20; KO frequency, 3.1 ± 0.40 Hz, n = 20, P = 0.023). A small change was also observed in the amplitude in the indirect pathway, but it did not reach statistical significance (WT amplitude, 10.7 ± 0.53 pA, n = 20 vs. KO amplitude, 9.6 ± 0.33 pA, n = 20, P = 0.090).

Marginal Differences in Paired Pulse Facilitation in NL1 KO Mice
A reduction in miniature frequency may be related to a decrease in release probability of synaptic vesicles. PPR is frequently used as an initial approach to investigate changes in the probability of neurotransmitter release (Salin et al., 1996;Kaplan et al., 2003). PPR is studied by pairing two stimuli of the same inputs with specific, brief inter-stimulus intervals (inter-event interval, IEI). Our unpublished data and those reported by other groups (Akopian and Walsh, 2007;Ding et al., 2008) indicate that stimulation of glutamatergic afferents to the striatum at short for AMPARs and NMDARs or for the NMDA/AMPA ratio (right). NL1 KO mice showed a deficit in NMDAR currents, as well as a reduction in the NMDA/AMPA ratio, but only in the direct pathway. **p < 0.01, ***p < 0.001. (C) Regression analysis of normalized AMPAR currents vs. the NMDA/AMPA ratio. AMPAR currents of NL1 KO mice were normalized to the maximal peak current for their WT control group within the direct or indirect pathway. The NMDA/AMPA ratio showed no significant correlation to the normalized amplitude of the AMPAR current.
IEIs (i.e., 30-70 ms) induce a mild facilitation whereas depression is recorded at IEIs above 150 ms. Within the same PPR protocol, here we used IEI 400 ms (in odd sweeps) and IEI 40 ms (in even sweeps) because they elicit the maximum levels of depression and of facilitation, respectively.
To determine if PP facilitation (PPF) or PP depression (PPD) existed within a group, we first compared currents elicited by pulse 1 and 2 for each IEI, followed by comparison of PPRs between groups. Both, DR1 and DR2 pathways showed significant levels of depression at the IEI of 400 ms (DR1 WT-P1, −197 ± 6.9 pA, vs. DR1 WT-P2, −153 ± 6.3 pA, n = 20, P < 0.001; DR1 KO-P1, −222 ± 7.1 pA vs. DR1 KO P2, −166 ± 4.8, P < 0.001, n = 20; DR2 WT-P1, −210 ± 6.1 pA vs. DR2 FIGURE 3 | Reduction in NMDAR currents is mainly due to a decrease in GluN2A currents. (A) AMPA receptor currents were blocked with the antagonist 5 µM NBQX (left). Afterwards, baseline NMDAR currents were recorded at +40mV followed by the application of 0.5-1 µM Ro25-6891 (a GluN2B antagonist, right). The remaining current (blue trace) is considered to be driven by NMDARs containing GluN2A subunits. To estimate the contribution of GluN2B subunits to the total NMDAR current, the GluN2A component was subtracted from the total. The difference (red trace) was considered to be driven by GluN2B. (B) Left panel. Integration of total NMDAR current confirmed the selective reduction in this conductance in the direct pathway. Middle and right panels. In the direct pathway, GluN2A was significantly reduced whereas GluN2B shows a trend towards reduction. Instead, in the indirect pathway neither conductance showed significant changes. **p < 0.01. (C) Regression analysis of NMDAR currents (normalized) vs. the fraction of GluN2A (left) or the fraction GluN2B (right). NMDAR currents were normalized to the maximal peak current for their WT control group within the direct and indirect pathways. No significant correlation was found between the amplitude of the normalized NMDAR current and the fraction contributed by GluN2A or GluN2B currents in either pathway.
WT-P2, −161 ± 5.8 pA, n = 20, P < 0.001; DR2 KO-P1, −196 ± 17.6 pA vs. DR2 K0-P2, −153 ± 13.3, n = 20, P < 0.001). When considering the PPR at this IEI, no difference between genotypes was observed (Figures 4C,D; DR1 WT, −0.78 ± 0.02 pA, n = 20 vs. DR1 KO, −0.75 ± 0.01 pA, n = 20, P = 0.592; DR2 WT, −0.77 ± 0.03 pA, n = 19, vs. DR2 KO, −0.80 ± 0.02 pA, n = 21, P = 0.485). Again, at the short IEI of 40 ms, both genotypes showed similar results (i.e., facilitation) in the DR2 pathway (WT P1, −207 ± 5.7 pA vs. WT P2, −225 ± 10.5 pA, n = 19, P = 0.011; KO P1, −224 ± 7.6 pA vs. KO P2 −231 ± 8.8, n = 20, P < 0.001), with no difference in PPR between genotypes (WT, −1.09 ± 0.05 pA, n = 20, vs. KO, 1.17 ± 0.04 pA, n = 21, P = 0.157). On the contrary, in the DR1 pathway, WT mice showed significant facilitation at 40 ms IEI, but not NL1 KO (WT P1, −197 ± 6.9 pA vs. WT P2 −223 ± 10.5 pA, FIGURE 4 | Characteristics of spontaneous miniature neurotransmission and short-term plasticity in the direct and indirect pathway MSNs of NL1 KO mice. (A) In the presence of picrotoxin (75-100 µM) and of Tetrodotoxin (0.75-1 µM), spontaneous miniature neurotransmission was recorded. After the membrane patch was broken by suction, the pipette intracellular solution was dialyzed for 8-10 min and recordings were taken for 2-5 min in order to observe at least 200 events per recorded neuron. The upper 2 traces and the lower 2 traces correspond to a 2 s segment recorded in the direct pathway or in the indirect pathway, respectively. The size and number of events look similar between genotypes in DR1 neurons, however, DR2 neurons show fewer events in NL1 KO mice. (B) DR2 neurons of NL1 KO mice show increased IEI (bottom right panel) reflected as an approximately 2-fold reduction in the frequency of mini events (inset). Cumulative distributions for amplitudes and for inter-event intervals (IEIs) were built with exactly 200 events per experiment to avoid distribution bias by experiments with higher number of events. (C) paired pulse ratio (PPR) average traces at IEIs of 40 ms or 400 ms. Downward deflections for stimulation # 1 for both IEIs are superimposed, therefore, only one event is apparent. The current relaxation after the peak for stimulus # 1 (P1) at 400 ms was used to determine the baseline level (BL) for stimulation # 2 (P2) at both IEI 40 ms and IEI 400 ms (P2-40 and P2-400, respectively). Peak amplitude of P2-40 was calculated by subtracting the amplitude of the BL at the time of the peak of P2 (BL 40) from its absolute amplitude. BL400 was taken as the average current amplitude of a 2 ms window just before P2-400 was triggered. Similarly, peak amplitude for P2-400 was calculated by subtracting BL400 from the absolute peak value for the corresponding event. The dashed-dotted line at the top corresponds to the steady-state baseline. Dotted lines in both P2s correspond to their recalculated amplitudes after baseline subtraction. (D) No significant changes in PPR were identified between genotypes in either pathway, though a trend is apparent in DR1 neurons at IEI of 40 ms. *P < 0.05. P = 0.002, n = 20; KO P1, −224 ± 7.6 pA vs. KO P2, −231 ± 8.8, n = 20, P = 0.355). However, when PPRs were compared between WT and KO at this short interval, no significance (although a trend) was reached in the DR1 pathway (WT, −1.14 ± 0.05 pA, n = 20, vs. DR1 KO, 1.04 ± 0.03 pA, n = 21, P <= 0.065). These findings suggest no alterations in presynaptic release probability to explain the altered mEPSC frequency in the indirect pathway.

Discussion
Our findings suggest a decrease in NMDAR-mediated synaptic transmission in NL1 KO mice that is selective for DR1 (direct pathway) MSNs. More specifically, the reduction of NMDAR currents in DR1 neurons is mainly due to decreased currents driven by NMDARs containing the GluN2A subunit, although a trend in reduced currents from GluN2B containing NMDARs is also present. These findings confirm and extend our previous report on an overall decrease in NMDA/AMPA ratio in the total pool of striatal MSNs (Blundell et al., 2010). Additionally, we also show that mEPSC frequency is reduced in the indirect pathway with no change in PPR.
While it is clear that NMDAR-mediated currents are decreased in the direct pathway of NL1 KO mice and the mechanism of this decrease is due to alterations in GluN2A containing receptors, the molecular mechanisms involved in the reduction of NMDAR/GluN2A currents in the direct pathway in NL1 KO mice remain uncertain. That said, several lines of evidence suggest that NL1 modulates NMDAR function at excitatory synapses. First, NL1 is preferentially, if not uniquely, localized to excitatory synapses (Song et al., 1999;Chubykin et al., 2007;Dahlhaus et al., 2010). Second, NL1 has been shown to have indirect (through PSD-95; Irie et al., 1997;Prange et al., 2004;Gerrow et al., 2006;Barrow et al., 2009) and, direct (through an extracellular domain) physical interaction with NMDARs (Budreck et al., 2013). These interactions have been shown to be important to induce and stabilize clusters of NMDARs at the synapse, both during development and during adulthood, and may favor the lateral movement of NMDARs from extrasynaptic to synaptic locations upon use-dependent blockage of synaptic NMDARs (Tovar and Westbrook, 2002;Budreck et al., 2013). Third, manipulations of NL1 expression levels do indeed demonstrate a strong positive correlation with the amplitude of NMDAR currents and with the NMDA/AMPA ratio in other brain regions and neurons (Chubykin et al., 2007;Futai et al., 2007;Kim et al., 2008b;Jung et al., 2010;Soler-Llavina et al., 2011;Kwon et al., 2012;Budreck et al., 2013). These data strongly suggest that the levels of NMDAR currents are in part controlled by NL1 without a concomitant alteration of AMPAR currents. However, when NL1 is overexpressed, changes in NMDAR currents do not always correlate with similar changes in NMDA/AMPA ratio because AMPAR currents are also increased (Futai et al., 2007;Shipman et al., 2011;Schnell et al., 2012;Shipman and Nicoll, 2012). Also, in some of these reports, the changes in both AMPA and NMDARmediated currents were in line with changes in the number of synapses (Shipman et al., 2011;Schnell et al., 2012;Shipman and Nicoll, 2012), suggesting that the changes in synaptic currents were in part a reflection of the changes in synapse number.
In our study, we identified decreased NMDAR currents only in D1 MSNs of the direct pathway but no alterations in mEPSC frequency in the direct pathway to suggest altered synapse number, and no change in mEPSC amplitude to suggest alterations in AMPAR-mediated synaptic responses. Likewise, we identified no alterations in D1 MSN input/output curves to suggest decreased synapse number. This is in line with our previous report regarding synapse numbers in hippocampus (Blundell et al., 2010) and from another group in cortex (Kwon et al., 2012), where constitutive loss of NL1 did not induce a decrease in synapse number. Thus, it appears that NL1 deletion can selectively alter NMDAR-mediated currents in MSNs of the direct pathway. This effect is largely mediated by alterations in NMDARs containing GluN2A subunits. The mechanism of this effect as well as its selectivity for direct pathway MSNs and for GluN2A-containing NMDARs is of great interest for future studies.
There are numerous potential explanations for the synapse specificity of the effects of NL1 KO. In NL1 KO mice, the effect of constitutively ablating NL1 expression could differ among glutamatergic synapses onto the direct pathway vs. glutamatergic synapses onto the indirect pathway, perhaps due to differences among these synapses in the expression of specific presynaptic neurexins, of leucine-rich repeat transmembrane protein isoforms (LRRTMs), or of other neuroligins that compete with NL1 for neurexin binding (Ko et al., 2009a;Linhoff et al., 2009;Shipman et al., 2011;Soler-Llavina et al., 2011). Furthermore, differential expression of NL1 and its partners/competitors among different synapses could also explain the selective effects that manipulating NL1 expression levels have on plastic properties observed in glutamatergic inputs to the amygdala from cortex vs. those from the thalamus (Jung et al., 2010), as well as the differences in levels of glutamatergic currents and on plastic properties between CA1 and dentate gyrus neurons in the adult hippocampus (Shipman and Nicoll, 2012). In line with the idea that expression levels of some synaptic proteins vary between direct and indirect pathways, differential mRNA expression levels between striatonigral MSNs and striatopallidal MSNs has been shown for Nxn1 (Lobo et al., 2006) as well as for NL3 (Rothwell et al., 2014).
Striatal synaptic NMDAR currents and/or the NMDA/AMPA ratio are also affected in other animal models of ASD and obsessive compulsive disorder (OCD; Welch et al., 2007;Peça et al., 2011;Kouser et al., 2013). This reduction in NMDAR currents can correlate with a reduction in NMDAR subunit protein levels in total brain extracts (Peça et al., 2011). Similarly, a generalized increase in the levels of GluN1 or GluN2 subunits is induced upon NL1 overexpression (Budreck et al., 2013;Hoy et al., 2013), supporting the idea that the size of synaptic NMDAR currents can correlate with a generalized alteration of NMDAR expression levels. In stark contrast, no reduction in the levels of NMDAR subunits due to constitutive loss of NL1 in vivo has been demonstrated to our knowledge. For example, in spite of a reduction of NMDAR currents in NL1 KO mice (Blundell et al., 2010;Budreck et al., 2013), no changes in levels of these subunits were demonstrated either in total brain homogenates or in biotinylated surface proteins. Together, this evidence suggests that mechanisms other than protein expression levels are involved in the reduction of NMDAR currents in NL1 KO mice. Budreck et al. (2013) indicate that NL1 and NMDARs directly interact through extracellular domains. This interaction seems to be important for the recruitment of extrasynaptic NMDARs into the synapse, stabilizing clusters of synaptic NMDARs without significantly affecting total NMDAR expression levels or synapse numbers (Budreck et al., 2013). Additionally, constitutive loss of NL1 may induce other type of posttranslational modifications that impact NMDAR channel function that could contribute to reduce NMDAR currents without affecting NMDAR expression levels.
We and others have demonstrated altered mEPSC synaptic transmission in response to experimental manipulations that modify the expression levels of NL1. In general, NL1 overexpression increases mEPSC frequency, whereas NL1 ablation, knockdown, or expression of dominant-negative NL1, decreases it (Prange et al., 2004;Nam and Chen, 2005;Chen et al., 2010;Mondin et al., 2011;Burton et al., 2012;Kwon et al., 2012;Schnell et al., 2012). Even though NL1 is a postsynaptic protein, it has been shown to modulate synaptic vesicle clustering through direct or indirect interactions with presynaptic proteins. Indeed, overexpression of NL1 increases the number of vesicles at the synapse (Dahlhaus et al., 2010), an effect that depends on the presence of functional scaffolding proteins S-SCAM, on the postsynaptic side, and N-cadherin, presynaptically (Stan et al., 2010). In addition, synaptic vesicles can also be modulated by the association of NL1 with β-neurexin (Dean et al., 2003). In this regard, we previously reported that the NL1 KO brain shows a significant reduction in several presynaptic proteins including Nxn-1 and mammalian uncoordinated-18 (Munc18; Blundell et al., 2010). These alterations alone could be affecting neurotransmitter release because Munc18 is essential for synaptic vesicle priming (Deák et al., 2009), whereas neurexins, besides modulating synaptic clustering, they can also link presynaptic Ca 2+ channels to the synaptic vesicle release machinery (Missler et al., 2003;Kattenstroth et al., 2004;Zhang et al., 2005). Of related interest are the findings of Lobo et al describing differential mRNA expression between direct and indirect pathway MSNs, including a 2-fold difference in neurexin-1 expression (Lobo et al., 2006). However, NL1 KO mice showed no difference in PPR, even at various Ca 2+ concentrations, making the latter possibility of low likelihood. Importantly, multiple lines of evidence indicate that evoked and spontaneous neurotransmission may differ in the modulation of synaptic vesicle release properties (Smith et al., 2012) which may even utilize different pools of synaptic vesicles (Kavalali, 2015). This varies among brain regions and synapse types and could be due to local synapse-specific expression of synaptic proteins that modulate both types of neurotransmission. Likewise, differences in the baggage of synaptic proteins that comprise local release machineries could be involved in the distinct effects between striatal pathways reported here. If this were the case, the constitutive absence of NL1 could lead mainly to postsynaptic alterations in the direct pathway vs. mainly presynaptic alterations in the indirect pathway. Our present results leave open this interesting and complex line of inquiry for future studies.
Similar to several other mouse models of ASD or of OCD (Greer and Capecchi, 2002;Welch et al., 2007;Etherton et al., 2009;Shmelkov et al., 2010;Peça et al., 2011), NL1-null mice show an ∼2-fold increase in repetitive grooming (Blundell et al., 2010). We do not presently know what the specific impact of the changes in striatal neurotransmission reported here may have on grooming behavior of NL1 KO mice, however, some testable hypotheses may be drawn from the literature. Given that activation of the direct pathway using DR1 agonists increased grooming behavior in rodents (Van Wimersma Greidanus et al., 1989;Berridge and Aldridge, 2000;Taylor et al., 2010), whereas facilitation of the indirect pathway by pretreatment with DR2 antagonists prevented the effect of DR1 agonists on grooming (Molloy and Waddington, 1987;Taylor et al., 2010), the emerging model is that dopamine facilitates the activation of the direct pathway while, at the same time, it dampens the inhibitory effects of the indirect pathway on motor program completion. Importantly, antagonism of Glun2A containing-NMDARs, favors dopaminergic release (Schotanus and Chergui, 2008) and, therefore, could have a stimulating impact on grooming. This is in agreement with the rescue of grooming behavior that systemic injections of D-cycloserine have in NL1 KO mice. However, it will remain to be determined in future studies if reductions in NMDAR currents are intertwined with the dopaminergic modulation of grooming in this mouse.
In summary, we have demonstrated reduced NMDAR currents selectively in the direct pathway MSNs of dorsal striatum in NL1 KO mice. This is due to a selective decrease of GluN2A-containing NMDAR currents at excitatory synapses onto direct pathway MSNs. Furthermore, action potential-independent spontaneous neurotransmission is selectively reduced in the indirect pathway of NL1 KO mice. These findings correlate altered striatal synaptic function to the increased grooming phenotype of NL1 KO mice. Indeed they lead to testable hypotheses of how altered striatal synaptic function may result in increased repetitive grooming.