Pathway-specific deregulation of striatal excitatory synapses in LRRK2 mutations

LRRK2 is a kinase expressed in striatal spiny projection neurons (SPNs), cells which lose dopaminergic input in Parkinson’s disease (PD). R1441C and G2019S are the most common pathogenic mutations of LRRK2. How these mutations alter the structure and function of individual synapses on direct and indirect pathway SPNs is unknown and may reveal pre-clinical changes in dopamine-recipient neurons that predispose towards disease. Here, R1441C and G2019S knock-in mice enabled thorough evaluation of dendritic spines and synapses on pathway-identified SPNs. Biochemical synaptic preparations and super-resolution imaging revealed increased levels and altered organization of glutamatergic AMPA receptors in LRRK2 mutants. Relatedly, decreased frequency of excitatory post-synaptic currents accompanied changes in dendritic spine nano-architecture, and single-synapse currents, evaluated using 2-photon glutamate uncaging. Overall, LRRK2 mutations reshaped synaptic structure and function, an effect exaggerated in R1441C dSPNs. These data open the possibility of new neuroprotective therapies aimed at SPN synapse function, prior to disease onset.


INTRODUCTION
Gain of function mutations in the leucine rich repeat kinase 2 (LRRK2) gene represent the most common cause of familial Parkinson's disease 1 . Carriers are at risk for late onset PD, which is clinically indistinguishable from sporadic PD, consistent with a possibility of common disease mechanisms 2,3,4,5 . LRRK2 gene product is a large multi-domain protein with two catalytic domains: a GTPase (ROC-COR) domain and a serine/threonine-directed protein kinase domain.
Pathogenic mutations are found predominantly in these two domains, suggesting that LRRK2 enzymatic activities are involved in PD pathogenesis 6,7 . Yet, how LRRK2 mutations in the two distinct functional domains contribute to PD pathogenesis and whether they act through a common mechanism is unknown. Enhanced LRRK2 kinase activity conferred by the G2019S (GS) mutation in the kinase domain is the most extensively studied property of mutant LRRK2 6,8 .
Meanwhile, the R1441C (RC) substitution in the GTPase domain results in impaired GTP hydrolysis, which is thought to indirectly enhance kinase activity through mechanisms that remain to be determined 9,10 .
Despite remaining questions, the last decade marks extensive progress in our understanding of LRRK2 function. This kinase is highly expressed in the spiny projection neurons (SPNs) of the striatum 9,11,12 . LRRK2 expression peaks in a developmental time window of extensive glutamatergic excitatory synapse formation, suggesting that LRRK2 may regulate the development or function of excitatory synaptic networks. Consistently, several lines of evidence suggest that loss of LRRK2 alters striatal circuits during postnatal development 12 , and the GS pathogenic mutation increases glutamatergic activity in cultured cortical neurons 13 as well as in acute striatal slices 14,15 . Recent studies assign a critical role of LRRK2 in presynaptic terminal vesicle function 16 . Here, the GS mutation impairs presynaptic glutamatergic release, suggested to underlie changes in glutamatergic activity of striatal neurons 15 . The potential postsynaptic function of LRRK2 remains less well-characterized. We have previously shown that the RC mutation impedes normal striatal protein kinase A (PKA) signaling, which in turn results in increased GluA1 phosphorylation in developing SPNs, consistent with a postsynaptic mechanism of action 12 . Similarly, glutamate receptor trafficking perturbations were observed in GS knock-in (KI) mice in response to plasticity induction protocols 17 . Furthermore, although an increase in spontaneous excitatory postsynaptic currents has been reported for the dorsomedial striatum, a recent report failed to show this phenotype for the ventral striatum 18 , despite LRRK2 expression in that region 19,20 . These observations suggest that LRRK2 mutations may shape the cortocostriatal synaptic function in a synapse-, cell-and area-specific manner. Overall, despite the links between LRRK2 and glutamatergic synapse dysfunction, the field currently lacks a coherent framework for understanding how the two distinct LRRK2 mutations selectively alter synaptic function in specific cell types.
Given the complementary role of direct and indirect striatal pathways in behavior 21,22,23 the lack of pathway specificity in prior studies limits our understanding of disease related mechanisms associated with LRRK2 mutations. Previous studies did not compare the dysfunction of two LRRK2 mutations which are found in distinct LRRK2 domains and confer divergent biochemical properties to the kinase, choosing instead to focus on the one mutation or the other. Previous studies have mainly focused on the GS pathogenic mutation, and whether the molecular mechanisms underlying pathology across the two most common mutations remain unknown. In addition, the phenotype of LRRK2 mutant models is subtle and corresponds to a moderate susceptibility for PD. Therefore, refined tools are required to distinguish the early pre-pathology functional abnormalities. We have taken the approach of combining molecular, anatomical and electrophysiological approaches that capture global aspects of LRRK2 function in the striatum, but also single synapse specific effects, in case abnormalities are associated with specific synapse subtypes.
Here, we undertook a systematic synapse function and structure analysis in two different mutant LRRK2 mouse lines to evaluate the contribution of this kinase mutations to SPN synapses across direct and indirect pathways. Using a combination of subcellular fractionation biochemistry, superresolution imaging, and 2-photon laser scanning microscopy with whole-cell physiology approaches, we found a critical role for LRRK2 RC, and to a lesser extent GS mutation, in organizing the structure and function of the SPN excitatory synapses, particularly for dSPNs.

LRRK2 +/RC mutation leads to increased synaptic incorporation of GluA1 in the striatum.
In order to resolve LRRK2 mediated synaptic alterations in identified SPNs in a systematic way, we crossed the pathway-specific reporter BAC transgenic mice 24 with two mutant LRRK2 KI mouse lines. Specifically, Drd1-Tomato and Drd2-eGFP reporter mice were crossed with each of the two mutant LRRK2 lines (RC and GS) ( Figure 1A). This platform enabled pathway specific interrogation of each LRRK2 mutation side by side, as well as initial experiments described below that were done on aggregated SPN populations. Our previous findings showed that RC KI neurons displayed altered PKARIIβ localization, as compared to wild-type (+/+) neurons, suggesting altered synaptic PKA activity in SPNs 12 . To further explore the synaptic PKA signaling in both KI mouse lines, we employed differential centrifugation, discontinuous sucrose gradient, and detergent extractions to isolate synaptic subcellular fractions from striatal extracts as shown in Figure 1B 12,25,26 . Consistent with our previous data, PKA activity was found elevated in the P2 crude synaptosomal preparation of +/RC but not +/GS striatal extracts, when phospho-PKA substrate antibody was used to detect phosphorylation of downstream PKA targets. The increased PKA activity in +/RC was further confirmed by an increase in phosphorylation of GluA1, a key downstream target of PKA. It has been previously reported that Rab8A is a physiological target of LRRK2 kinase activity 8,27 , but it is unknown whether this is the case in the striatum. We next investigated the contribution of each mutation to phosphorylation of Rab8A in the P2 striatal fractions. Notably, the RC, compared to the GS mutation, led to a stronger increase in Rab8A phosphorylation ( Figure 1D). This finding stands in accordance with two recent studies in cell lines, showing that the RC and other mutations found in the GTPase domain of LRRK2 result in greater increase kinase activity, compared to the GS mutation 28,29 . We further isolated the post synaptic density (PSD) fraction from the P2 preparation by using a detergent extraction step. The PSD fraction showed enrichment of the postsynaptic marker PSD95, but not presynaptic protein synaptophysin (svp38) (Figure 1C). Given that PKA mediated phosphorylation of S845 in GluA1 has a significant impact on synaptic trafficking of GluA1 12,30 , we examined the levels of GluA1 in the PSD fractions of KI mice. We observed a significant increase in the GluA1 levels in the PSD fraction from +/RC, but not +/GS mice, compared to wild type control (+/RC: 37.7% increase of control; +/GS: 2% increase relative to control, Kruskal-Wallis p=0.0092, post hoc tests as noted, n=4 per group) ( Figure 1E, F), which was paralleled by elevated PKA activity specifically in the PSD fractions in +/RC (+/RC, 24.3 % of control, Kruskal-Wallis p=0.0194, post hoc tests as noted, n=4) but not +/GS mice (+/GS, 0.05% of control, n=4) ( Figure 1E, G). Next, we determined the relative levels of PSD-95 protein in the PSD fraction across genotypes, which was used in the ratios in Figure 1F, G. We run subcellular fractions of all genotypes on the same blot (Supplemental Figure 1A), and normalized PSD95 to another synaptic protein, Homer1. PSD95 band intensities were expressed as fraction of the intensity of wild-type control S1 on the same blot (Supplemental Figure 1B, C). Furthermore, PSD95 band intensities in each of S1, P2, and PSD fractions were expressed as percentage of the PSD95 intensity of the wild type in each fraction. No differences in PSD95 levels in mutant LRRK2 fractions were observed after one way ANOVA. Overall, our data show increased GluA1 levels in the PSD fractions of +/RC striatal extracts, which parallels enhanced synaptic PKA signaling. wild type in each one of the S1, P2, and PSD95 fractions.

Altered nanoscale organization in the dendritic spines of direct and indirect pathway mutant LRRK2 SPNs
To precisely examine the nanoscale organization of GluA1 receptors in the dendritic spines of identified SPNs in mutant LRRK2 mice, we used structured illumination microscopy (SIM). SIM is able to overcome the resolution limits of conventional microscopy, and recently revealed a high degree of organization among scaffold proteins and AMPA receptors, forming nanoscale subsynaptic domains 31,32,33 . Primary cultured striatal neurons from Drd1-dTomato or Drd2-eGFP mice crossed with either +/RC or +/GS mouse lines (Figure 2A), were immunostained for GluA1 and PSD95 ( Figure 2B, and Figure 3A-B, Video S1). We found several lines of evidence demonstrating increased synaptic incorporation of GluA1 in +/RC dSPNs, compared to +/+ ones.
However, the percentage of the total PSD95 area containing GluA1 nanodomains +/GS iSPNs compared to controls was similar (one-way ANOVA, p=0.469) ( Figure 3H). Thus, the increased size of GluA1-PSD95 overlapping domains in the +/GS iSPNs reflects the elevated PSD95 area of these neurons (Supplemental Figure 2C-D), and not synaptic GluA1 incorporation, as found in +/RC. We aimed to further investigate the relationship between overlapping nanodomains size and PSD95 area in the iSPNs dendritic spines. In general, PSD95 area correlated with the overlap GluA1-PSD95 domain size in all genotypes (control, Spearman r=0.7817, p<0.0001, +/RC, r=0.3996, p=0.0003 and +/GS, r=0.6938, p<0.0001) ( Figure 3I-K). Based on Figure 2I, we found that for larger PSD95 domains, mutant LRRK2 dSPNs show a tendency towards larger overlap between GluA1 and PSD95. However, this was not the case for iSPNs of +/RC genotype. In fact, in +/RC iSPNs, some high overlap GluA1-PSD95 domains were characterized by smaller PSD95 areas. This detailed high resolution imaging analysis points to different GluA1 subsynaptic distribution in the iSPNs of +/RC SPNs, compared to +/+ and +/GS. Taken together, these single synapse imaging findings in pathway identified SPNs demonstrate an increased number of GluA1 receptors in the synapses of +/RC dSPNs, pointing towards cell type and mutant specific LRRK2 functions. These observations suggest highly specific and likely regulated nanoscale organization of GluA1 receptors in SPN dendritic spines.

Pathway specific functional alterations in the synapses of R1441C and G2019S SPNs
Since the effects of LRRK2 mutations appear to alter excitatory synapse subunits, next we carried out recordings of pharmacologically isolated miniature excitatory post-synaptic currents (mEPSCs) in acute brain slices from mice of the three genotypes. SPN pathway identity was determined based on the presence or absence of Drd2-eGFP 34

Overall dendritic spine density and morphology is conserved in mutant LRRK2 SPNs
To investigate whether the observed decrease in the mEPSC frequency was associated with global changes in dendritic spine number and morphology, we relied on identifying pathway specific SPNs by injecting a Cre dependent adeno-associated virus (AAV) expressing eGFP virus (AAV8/Flex-GFP) in the striatum of neonatal pups. Drd1-Cre and Adora2a-Cre mice were crossed with +/RC and +/GS mice, as well as controls (Figure 5A, Video S2). We then analyzed dendritic fragments of d-and iSPNs to determine dendritic spine density, spine head width, and spine length. No differences in any of these parameters were observed for all six genotype/cell type combinations (two-way ANOVA, n=9-12 SPNs from 3-4 mice/genotype) (Figure 5B-E). Further dendritic spine classification showed no difference in dendritic spine type in SPNs of either pathway (two way ANOVA, n=9-14 SPNs for 3-4 mice/genotype) ( Figure 5F). Overall, these data suggest that the functional synaptic alterations (Figure 4) observed in +/RC and +/GS mutant SPNs are not reflected in uniform changes in dendritic spine number or morphology.

R1441C mutant SPNs
Our results indicate that the abundance of GluA1 subunits is selectively increased in dendritic spines of +/RC dSPNs, along with synaptic AMPAR presence and the prevalence of larger amplitude mEPSCs. This is consistent with a possibility that a specific fraction of +/RC excitatory synapses are functionally stronger. In order to evaluate the physiology of single dendritic spines in pathway-identified SPNs across three genotypes, we used 2-photon dual laser glutamate uncaging and imaging, combined with whole-cell electrophysiology in voltage clamp mode ( Figure 6A). Drd2-eGFP mice were crossed into G2019S and R1441C KI lines, and maintained on a WT background for control groups. SPNs were filled with a cesium-based internal and Alexa 594 for imaging dendrites and dendritic spines at ~910 nm ( Figure 6B-C); the presence or absence of GFP labeling was used to assign pathway identity. Uncaging evoked EPSCs (uEPSCs) were elicited using 0.5 ms long pulses of 725 nm laser light, directed near dendritic spines (~0.5-1 µm away), in order to drive focal uncaging of MNI-glutamate. Locations of peak responses, sampled for 3 sites near each dendritic spine, were chosen for data acquisition for every synapse. Recordings were carried out in the presence of blockers of GABAARs, NMDARs, muscarinic receptors, and sodium channels, for isolation of single synapse AMPAR responses.

DISCUSSION
In the present study, we characterized synaptic dysfunctions caused by mutant LRRK2 (RC and GS) in pathway-identified SPNs. To our knowledge, this is the first comparative approach focusing on more than one PD related LRRK2 mutation, and employing both global and single synapse approaches for the study of LRRK2 driven striatal remodeling. By studying RC and GS KI mouse lines side by side, we were able to demonstrate that the two most common LRRK2 gain of function pathogenic mutations 35 Figure 6G).
Based on the current knowledge, it remains unclear why the RC mutation is associated with somewhat more pronounced effects on sculpting striatal synapses. The GS mutation, which has been the main focus of the current literature, is found in the kinase domain of LRRK2, whereas the RC mutation is located in the GTPase domain of the protein 36,37 . It has been previously proposed that all pathogenic mutations lead to increased LRRK2 kinase activity, although the mechanism by which RC does this still remains unclear 1,9 . Moreover, it is not known whether aberrant LRRK2 substrate phosphorylation is the predominant pathogenic mechanism in LRRK2 mediated striatal changes. Recent findings have revealed a subset of Rab family of proteins as the long awaited bona fide LRRK2 substrates 8 . Notably, the RC mutation, mainly in the context of heterologous cell lines, leads to higher increase in Rabs phosphorylation compared to the GS mutation, which is localized to the kinase domain 29,28 . By using Rab8A phosphorylation at T72 as a readout for LRRK2 kinase activity, our findings confirmed that RC mutation led to higher phosphorylation in synaptic striatal extracts, compared to GS. Similarly, evidence of increased synaptic PKA activities was observed in +/RC and not +/GS striatal synaptic fractions. This increased PKA activity in the RC synapses is expected to positively correlate with elevations in dSPNs signaling 38 . Specifically, phosphorylation of GluA1 at Ser 845 by PKA facilitates the targeting of the receptors to extrasynaptic membranes and their "priming" for synaptic insertion and this results in elevated GluA1 synaptic incorporation 39 . Indeed, our biochemical and SIM imaging data in RC mice support this notion. In contrast to dSPNs, PKA driven signaling requirements for iSPNs are more complex, due to the involvement of adenosine. This molecule can be released by neurons or glia 40,41 , complicating predictions for PKA involvement in mutant LRRK2 function in iSPNs.
Our previous and current findings showed that LRRK2 contributes to glutamatergic synaptic functions by directing PKA signaling events in SPNs, leading to altered GluA1 synaptic incorporation in the striatum of RC mice 12  EPSCs. This scenario, on top of overall dampening of mEPSC frequency, is expected to bias mutant dSPN responses to be driven by a narrower set of glutamatergic inputs than is normally the case. Overtime, a small bias could be amplified through the recurrent circuitry of basal ganglia loops 23 . While there are some differences in the age of preparation and in recording conditions, our consistent finding of substantial miniature EPSC frequency decreases for both mutations in pathway-identified SPNs are in disagreement with prior recordings of spontaneous events in the GS mutant mice. One recent report showed elevations in spontaneous EPSCs events frequency in GS KI mice at 1-3 months 15 , while another reported changes in spontaneous but not miniature events with the GS mutation in pre-weaning mutants 14 , although recordings were primarily performed in a mixed population of direct and indirect pathway SPNs, and neither study thoroughly compared the two mutations. Such differences in spontaneous event frequency point to a potential circuit-level compensatory mechanism that may be driven by disturbances in a subset of SPN excitatory synapses. Surprisingly, in the presence of clear functional and nanoarchitecture abnormalities, aggregate dendritic spine density and morphology, at least at the ages we evaluated, appeared normal. One possible explanation for this mismatch that requires further investigation is the possibility of higher occurrence of nonfunctional or differently functioning spines in LRRK2 mutant SPNs.
Given the critical role of LRRK2 in excitatory glutamatergic synapses 12,15,14 , a more nuanced understanding of LRRK2 mediated synaptic alterations should facilitate the search for more targeted PD therapies. Here, we focused primarily on interrogating cell type specific LRRK2 based responses. Why does this pathway specificity matter? It is well-established that dopamine loss causes pathway specific morphological and functional changes in SPNs 21,22,44 . Moreover, dSPNs and iSPNs exhibit opposing PKA pathway signaling properties after dopamine receptor activation 45,40 . Given the critical role of LRRK2 in synaptic PKA activities 12,46 , these mutations are poised to have pathway specific effects and present distinct opportunities for pathway-targeted therapies.
Over the past years, a number of transgenic mouse models was generated to study LRRK2 function 47,48 however the results were found to be inconsistent across studies. Gene-targeted mutant LRRK2 KI mice represent the most physiologically relevant model to investigate LRRK2 mediated alterations and unravel disease mechanisms. Here, we attempt to synthesize and further the existing knowledge by employing a comparative approach focusing on two distinct mutations, combined with a powerful suite of techniques allowing for both global and single synapse study of LRRK2 SPNs. Our conclusions provide useful insights on how LRRK2 mutations could confer PD susceptibility for human patients with otherwise sub-clinical alterations in dopamine signaling. This is of particular interest as asymptomatic mutant LRRK2 carriers show early corticostriatal connectivity changes 49,50,51 , in the absence of neuronal degeneration. Therefore, KI LRRK2 mouse models expressing pathogenic mutations represent valuable models that enable investigations of early events that precede neuronal death and may be predictive of future dysfunction.

METHODS
Mouse strains and genotyping. All mouse related experiments followed the guidelines approved by the Northwestern University Animal Care and Use Committee. Young adult male and female mice (postnatal days 30-50) were used in this study. Approximately equal numbers of males and females were used for every experiment. All mice were group-housed, with standard feeding, light-dark cycle, and enrichment procedures. C57BL/6 (wild-type), RC 52 , and GS 53 heterozygous knock in mice were used for subcellular fractionation experiments. For electrophysiological and imaging approaches RC or GS mice were crossed with Drd1α-dTomato and/or Drd2-eGFP mice.
Heterozygotes for RC and GS allele and hemizygotes for Drd1α-dTomato and/or Drd2-eGFP were used in all experiments. For dendritic spine analysis, hemizygous Drd1-Cre, and Adora2a-Cre mice crossed with RC and GS KI mice were used. All animals were backcrossed to C57BL/6 for several generations.

Subcellular fractionation and Western blot analysis
Subcellular fractionation of mouse striatum was performed as previously described 25,12 (Figure   1B). Specifically, mouse striata were dissected (3 striata per experiment were pooled) and rapidly homogenized in four volumes of ice-cold Buffer A (0.32 M sucrose, 5 mM HEPES, pH7.4, 1 mM MgCl2, 0.5 mM CaCl2) supplemented with Halt TM protease and phosphatase inhibitor cocktail (Thermo) using a Teflon homogenizer (12 strokes). Homogenized brain extract was centrifuged at 1,400 g for 10 min. Supernatant (S1) was saved and pellet (P1) was homogenized in buffer A with a Teflon homogenizer (5 strokes). After centrifugation at 700 g for 10 min, the supernatant (S1') was pooled with S1. Pooled S1 and S1′ were centrifuged at 13,800 g for 10 min to the crude synaptosomal pellet (P2) and the supernatant (S2). P2 was resuspended in Buffer B (0.32 M sucrose, 6 mM Tris, pH 8.0) supplemented with protease and phosphatase inhibitors cocktail with Teflon homogenizer (five strokes) and was carefully loaded onto a discontinuous sucrose gradient Rab 8A (Abcam, ab230260) , as well as secondary anti-mouse and anti-rabbit (Thermo Fischer Scientific) Membranes were incubated with Immobilon ® ECL Ultra Western HRP Substrate (Millipore) for 3 minutes prior to image acquisition. Chemiluminescent blots were imaged with iBright TM CL1000 imaging system (Thermo Fisher Scientific). For quantitative analysis, images were analyzed using iBright Analysis Software (Thermo Fisher Scientific).

Primary neuronal cultures
Primary corticostriatal co-cultures were prepared as described previously 12,54 . In brief, striatal cultures were prepared from P0 pups of Drd1-Tomato or Drd2-eGFP mice crossed with mutant KI LRRK2 lines (G2019S and R1441C). Tissues were digested by papain (Worthington Biochemical Corporation) and the striatal and cortical cells were mixed at a ratio of 1:2. The neurons were placed on coveslips with plating medium (Medium I) containing Basal Eagle Medium (Sigma-Aldrich) supplemented with 1 x GlutaMAXTM-I (Gibco TM ), 1 × B27 (Gibco TM ), 1 x N-2 (Gibco TM ), 1 x Antibiotic Antimycotic (Sigma-Aldrich), 5% horse serum (Gibco TM ) and 5% FBS (Gibco TM ) at a density of 4x10 5 for about one hour. After initial plating, medium was changed to Medium I without the horse serum and the FBS supplemented with 2.5 μM arabinosylcytosine (Sigma-Aldrich) (Medium II). Half of the medium was changed with fresh Medium II every 7 days; experiments were conducted 28-30 days after plating.

SIM imaging and analysis
Multichannel SIM images were obtained with a Nikon Structured Illumination super-resolution microscope using a 100x, 1.4 NA objective as previously described 33  GFP/tdTomato signal (Thermo Fischer Scientific, #A10262/#M11217) was used to generate a mask to identify the spine shape, subsequent analysis were done with the nanodomains of GluA1 (NeuroMab #75327) and PSD95 (Thermo Fischer Scientific, #51-6900)within the mask. 3D reconstructions in a single plane used 9 images captured with 2D SIM and reconstructed with 3D SIM utilizes 15 to increase xy and z resolution were generated by Nikon Element and the illumination modulation contrast was set automatically by the software. Nikon Elements software (general analysis) was used for quantification of colocalization of GluA1 and PSD95 proteins. The images were thresholded for each channel and kept constant across experiments. The regions of interest (puncta) were outlined, and total immunofluorescence number and binary area for each region per channel were measured automatically. Regions in one channel were overlayed on the other channel. The minimum distances between puncta (e.g. PSD95 to GluA1) were measured from surface to surface automatically by the software.

Intracranial injections
Intracranial injections were performed as previously described 23 . Briefly, P4-5-day-old pups were placed into a stereotaxic frame under cryoanesthesia. 200 nl of AAV8/Flex-GFP virus (6.2*10 12 used at 1:3 dilution, UNC Vector Core, Chapel Hill, NC) were delivered into the dorsal striatum at a rate of 100 nl min −1 using microprocessor-based controller, Micro4 (WPI). In order to ensure targeting of the dorsal striatum the needle was placed 1 mm anterior to midpoint between ear and eye, 1.5 mm from midline and 1.8 mm ventral to brain surface.

Confocal microscopy and dendritic spine analysis
Confocal images of fixed 80 μm thick brain sections of P30 pups injected with the AAV8/Flex-GFP virus were obtained with the Nikon A1R scope. Fluorescence projection images of dendrites and the corresponding spines were acquired with a 60x oil immersion objective (NA=1.4) at 0.1 μm intervals with 1024x1024 pixel resolution. For each genotype, 2-4 segments per neuron, 9-14 neurons from 3-4 animals were used to generate z-stacks. Fragments between secondary to tertiary dendrites without overlap with other neurons or discontinuous were chosen for analysis. Dendritic spine density and morphology was performed using Imaris 9.21 software (Bitplane, Concord, USA). Images of dendritic fragments were collected from secondary to tertiary dendrites, for tracing of the dendritic fragments the autopath mode of the filament tracer and default settings were selected. The following settings were used for spine detection: 0.5 μm minimal head size, 1.8 μm maximum length, seed point threshold approx. 10, no branched spines were allowed. Spine detection was manually corrected if necessary. Classification of spines into stubby, mushroom-like and filopodia was performed using the Imaris XTension classify spines with following definitions: stubby: spine length < 0.75 μm; mushroom: spine length < 3.5μm, spine head width >0.35 μm and spine head width > spine neck width; filopodia: when it did not fit the criteria mentioned above 55 .

Electrophysiological recordings
Two hundred and fifty µm thick acute coronal brain slices were prepared as described  H7422) with 520/28 nm band pass filters (Semrock), mounted above and below the sample, were used for imaging fluorescence signals. A modified version of ScanImage was used for data acquisition 59,34 .
For n sizes, both the number of trials recorded and the number of animals are provided. All data are expressed as mean + SEM, or individual plots. For two-group comparisons, statistical significance was determined by two-tailed Student's t-tests. For multiple group comparisons, oneway analysis of variance (ANOVA) tests were used for normally distributed data, followed by post hoc analyses. For non-normally distributed data, non-parametric tests for the appropriate group numbers were used, such as Mann-Whitney and Kruskal-Wallis. Spearman correlation was used to detect correlation between two groups of data. p<0.05 was considered statistically significant.