Intact synapse structure and function after combined knockout of PTPδ, PTPσ, and LAR

It has long been proposed that leukocyte common antigen-related receptor protein tyrosine phosphatases (LAR-RPTPs) are cell-adhesion proteins that control synapse assembly. Their synaptic nanoscale localization, however, is not established, and synapse fine structure after knockout of the three vertebrate LAR-RPTPs (PTPδ, PTPσ, and LAR) has not been tested. Here, superresolution microscopy reveals that PTPδ localizes to the synaptic cleft precisely apposed to postsynaptic scaffolds of excitatory and inhibitory synapses. We next assessed synapse structure in newly generated triple-conditional-knockout mice for PTPδ, PTPσ, and LAR, complementing a recent independent study of synapse function after LAR-RPTP ablation (Sclip and Südhof, 2020). While mild effects on synaptic vesicle clustering and active zone architecture were detected, synapse numbers and their overall structure were unaffected, membrane anchoring of the active zone persisted, and vesicle docking and release were normal. Hence, despite their localization at synaptic appositions, LAR-RPTPs are dispensable for presynapse structure and function.


Introduction
Presynaptic nerve terminals are packed with neurotransmitter-laden vesicles that fuse at the active zone, membrane-attached protein machinery that forms vesicular release sites. Work over the past two decades has established that the unique synaptic architecture with nanoscale apposition of these secretory hotspots with receptors on postsynaptic cells allows for robust signal transmission (Biederer et al., 2017;Südhof, 2012). The assembly mechanisms of these transcellular molecular machines, however, remain largely obscure Rizalar et al., 2021;Südhof, 2018).
We first aimed at resolving the subsynaptic localization of LAR-RPTPs using STED microscopy. PTPd antibody specificity was established using cTKO RPTP neurons as negative controls, while antibodies suitable for superresolution analyses of PTPs or LAR could not be identified. To determine PTPd localization, we selected side-view synapses with bar-like postsynaptic receptor scaffolds (PSD-95 and Gephyrin for excitatory and inhibitory synapses, respectively) on one side of a Synaptophysin-labeled nerve terminal (Figure 1-figure supplement 2, Held et al., 2020;Wong et al., 2018). PTPd, detected with antibodies against the extracellular fibronectin domains (Shishikura et al., 2016), was concentrated apposed to PSD-95 and Gephyrin, respectively, at distances of 24 ± 17 nm (PSD-95) and 28 ± 11 nm (Gephyrin) ( Figure 1D-I). Only background signal typical for quantification of raw images Held et al., 2020;Wang et al., 2016;Wong et al., 2018) remained in cTKO RPTP neurons in STED ( Figure 1D-I) and confocal (Figure 1-figure supplement 3) microscopy. This establishes that the extracellular portion of PTPd localizes to the synaptic cleft. Given the presynaptic roles of LAR-RPTPs at invertebrate synapses and in synapse formation assays (Ackley et al., 2005;Kaufmann et al., 2002), the interactions with the active zone protein Liprin-a (Pulido et al., 1995;Serra-Pagès et al., 1998;Wong et al., 2018), and the asymmetry of the average STED side-view profile of PTPd with a bias toward the presynapse ( Figure 1F,I), it is likely that most PTPd is presynaptic and localized at the active zone, but postsynaptic components cannot be excluded. Furthermore, most synapses contain PTPd, as 88% of excitatory and 92% of inhibitory synapses had PTPd peak intensities higher than three standard deviations above the average of the cTKO RPTP signal.
The subsynaptic PTPd localization and its presence at most synapses is consistent with general roles of LAR-RPTPs in synapse organization. However, the synapse density, measured as Synaptophysin puncta, was unchanged in cTKO RPTP neurons ( Figure 1L-O). Small increases in puncta intensity and area were detected ( Figure 1L-O), consistent with enlargements observed in invertebrates (Ackley et al., 2005;Kaufmann et al., 2002). A recent independent study that also ablated LAR-RPTPs using mouse genetics found normal synapse densities as well (Sclip and Südhof, 2020). Together, these data challenge the model that LAR-RPTPs are master synapse organizers (Dunah et al., 2005;Fukai and Yoshida, 2020;Han et al., 2018;Han et al., 2020a;Han et al., 2020b;Kwon et al., 2010;Takahashi and Craig, 2013;Um and Ko, 2013;Yim et al., 2013). It  remains possible that LAR-RPTPs control assembly of a specific subset of synapses, which may explain why PTPd ablation causes modest layer-specific impairments of synaptic strength (Park et al., 2020).
We next examined whether LAR-RPTPs have specific roles in presynaptic nanoscale structure. Electron microscopy of high-pressure frozen neurons (Figure 2A-E) revealed that synaptic vesicles were efficiently clustered at cTKO RPTP synapses. A~15% increase in the total synaptic vesicle number per synapse profile was detected, matching the modestly increased Synaptophysin signals ( Figure 1) and the enhanced presence of vesicular markers in C. elegans mutants (Ackley et al., 2005). Notably, no differences in vesicle docking (defined as vesicles for which the electron dense membrane merges with the density of the target membrane) were observed. Synapse width, measured as the distance over which the pre-and postsynapse are apposed to one another and separated by a synaptic cleft, was increased by~30%, again matching invertebrate phenotypes (Kaufmann et al., 2002). These data establish that LAR-RPTP ablation does not strongly impair synapse ultrastructure. LAR-RPTPs may shape aspects of the synaptic cleft, consistent with their localization and transsynaptic interactions and possibly similar to other synaptic cell-adhesion proteins, for example SynCAMs (Perez de Arce et al., 2015).
We assessed whether active zone proteins, which are present at normal levels in western blots after LAR-RPTP ablation (Sclip and Südhof, 2020), are anchored at the presynaptic membrane by LAR-RPTPs. STED microscopy was used to measure localization and peak levels of active zone proteins at excitatory ( Figure 2F-I) and inhibitory ( Figure 2J-M) synapses. RIM, Munc13-1, Ca V 2.1, and Liprin-a3 were localized within 30-60 nm of the postsynaptic scaffolds in control RPTP and cTKO RPTP synapses, as expected for these proteins (Held et al., 2020;Wong et al., 2018). Overall, there were no strong changes in their levels, but small increases in RIM and small decreases in Liprin-a3 were detected in both types of cTKO RPTP synapses either by STED ( Figure 2F-M) or confocal (Figure 2figure supplement 1) microscopy. While binding between Liprin-a and LAR-RPTPs (Pulido et al., 1995;Serra-Pagès et al., 1998) may explain Liprin-a3 reductions, these data establish that other pathways are sufficient to recruit most Liprin-a3 to active zones. The higher levels of RIM may be compensatory to reductions in Liprin-a3 and could be related to the liquid-liquid phase separation properties of both proteins McDonald et al., 2020;Wu et al., 2019). Overall, we conclude that the active zone remains assembled and anchored to the target membrane in the absence of LAR-RPTPs.
A previous study found that LAR-RPTP ablation induced no strong defects in glutamate release, but regulated NMDARs through a transsynaptic mechanism (Sclip and Südhof, 2020). These findings are consistent with the near-normal synaptic ultrastructure and active zone assembly ( Figure 2). We complemented this recent study by whole-cell recordings of inhibitory postsynaptic currents (IPSCs, Figure 3). Release evoked by single action potentials was similar between control RPTP and cTKO RPTP neurons, and IPSC kinetics were unaffected. The IPSC ratio of two consecutive stimuli (paired pulse ratio), which is inversely proportional to vesicular release probability (Zucker and Figure 1 continued structure of LAR-RPTPs and the antibody recognition site for PTPd (antibodies were generated using a peptide containing fibronectin domains 2, 3 and 8 Shishikura et al., 2016). (D-F) STED images (D), quantification of the peak intensity of PTPd (E, STED) and average intensity profiles for PTPd (STED) and PSD-95 (F, STED) at excitatory synapses. Side-view synapses were identified by the presence of bar-like PSD-95 signals (STED) at the edge of the vesicle cloud marked by Synaptophysin (confocal). Intensity profiles of shaded areas in the overlap images were used to determine the peak intensity of the protein of interest, and are shown on the right of the corresponding image. N (control RPTP ) = 50 synapses/3 cultures, N (cTKO RPTP ) = 54/3. (G-I) Same as D-F, but for inhibitory synapses identified by Gephyrin. N (control RPTP ) = 58/3 cultures, N (cTKO RPTP ) = 59/3. (J-M) Confocal images of cultured neurons stained with anti-Synaptophysin antibodies (J) and quantification of Synaptophysin puncta density (K), intensity (L) and size (M) detected using automatic two-dimensional segmentation. N (control RPTP ) = 20 images/3 independent cultures; N (cTKO RPTP ) = 21/3. The Synaptophysin confocal data are from the experiments shown in D-I. Data are plotted as mean ± SEM and were analyzed using two-way ANOVA tests (F, I, genotype *** for PTPd), t-tests (E, L, M) or Mann-Whitney rank sum tests (H, K). **p<0.01, ***p<0.001. The online version of this article includes the following figure supplement(s) for figure 1:   Gephyrin (au)   Regehr, 2002), was also unaffected. We conclude that synaptic vesicle exocytosis, here monitored via IPSCs, is not impaired by LAR-RPTP knockout.

Discussion
Overall, we demonstrate that ablation of LAR-RPTPs from hippocampal neurons does not alter synapse density, synaptic vesicle docking, membrane anchoring of active zones, and synaptic vesicle release. This aligns with a parallel study that reported no loss of synaptic puncta and efficient release at excitatory synapses in cultured hippocampal neurons and in acute hippocampal brain slices (Sclip and Südhof, 2020) upon LAR-RPTP knockout, but contrasts RNAi-based studies that led to models in which these RPTPs are major synapse organizers (Dunah et al., 2005;Fukai and Yoshida, 2020;Han et al., 2018;Han et al., 2020a;Han et al., 2020b;Kwon et al., 2010;Takahashi and Craig, 2013;Um and Ko, 2013;Yim et al., 2013). LAR-RPTPs belong to the superfamily of RPTPs (Johnson and Van Vactor, 2003), and it is possible that other RPTPs compensate for their loss. We note, however, that the time course of deletion in our knockout experiments is similar to the time course that is used in most RNAi-knockdown studies, and is hence unlikely to explain the differences. Other contributing factors could be different experimental preparations and off-target effects of knockdowns, which may generate artifacts in synapse formation experiments (Südhof, 2018). Altogether, we conclude that, while biochemical and synapse formation assays support synaptogenic activities for these proteins, synapses persist upon LAR-RPTP ablation, and their structure and function do not necessitate these proteins.
Our study establishes specific localization of PTPd extracellular domains to the synaptic cleft. Hence, PTPd is correctly positioned to locally execute synaptic functions, for example for shaping cleft geometry, to modulate presynaptic plasticity, or to control postsynaptic receptors (Biederer et al., 2017;Sclip and Südhof, 2020;Uetani et al., 2000). Such functions would not be at odds with the at most mild structural and functional effects after LAR-RPTP ablation, nor with upstream functions in neurite outgrowth and axon targeting (Ackley et al., 2005;Clandinin et al., 2001;Desai et al., 1997;Krueger et al., 1996;Prakash et al., 2009;Shishikura et al., 2016). Mechanisms of active zone anchoring to the target membrane, however, remain unresolved. Deletion of the major candidates, Ca V 2 channels (Held et al., 2020), Neurexins (Chen et al., 2017), and now LAR-PTPs (Figures 1 and 2), produces no major structural defects, indicating that active zones are most likely anchored to the plasma membrane through multiple parallel pathways that may or may not include these proteins (Emperador-Melero and Kaeser, 2020). Synaptic cell-adhesion proteins that contribute to synapse formation and function, for example SynCAMs, EphBs, Cadherins, Teneurins, and FLRTs, are plausible candidates that could act on their own or in concert with other proteins, including LAR-RPTPs, to contribute to active zone membrane anchoring (Biederer et al., 2017;Südhof, 2018

Mouse lines
PTPd (Ptprd) mice were acquired as frozen embryos from the Welcome Trust Sanger Institute (Ptprd tm2a(KOMP)Wtsi ; clone EPD0581_9_D04, MGI:4458607, RRID:IMSR_EM:11805), and the same mutant allele was described in previous studies (Farhy-Tselnicker et al., 2017;Sclip and Südhof, 2020). PTPs (Ptprs) mice were obtained as frozen sperm from the Canadian Mouse Mutant Repository at the Hospital for Sick Children (C57BL/6N-Ptprs tm1a(KOMP)Mbp /Tcp; clone DEPD00535_1_D11; MGI:4840831, RRID:IMSR_CMMR:ABCA) and were also used previously (Bunin et al., 2015;Sclip and Südhof, 2020). Embryonic stem cells containing the LAR (Ptprf) mutant allele were obtained from the Helmholtz Zentrum Mü nchen (Ptprf tm1a(EUCOMM)Wtsi ; clone EPD0697_1_D03; MGI:4887720). Mutant alleles were originally generated using homologous recombination by the international knockout consortium (Bradley et al., 2012;Skarnes et al., 2011). Frozen embryos (PTPd), frozen sperm (PTPs), or embryonic stem cells (LAR) were used to establish the respective mouse lines through the Transgenic Mouse Core (DF/HCC) at Harvard Medical School. For generation of the LAR mutant mice, the embryonic stem cells were expanded, the genotype was confirmed by PCR and sequencing, and injection into C57BL/6 blastocysts was used to generate chimeric founders. After germline transmission, the mice were crossed to Flp-expressing mice (Dymecki, 1996) to remove the LacZ and Neomycin cassettes to generate the conditional allele. The same crossing was performed with the cryo-recovered PTPd and PTPs mice. This strategy generated conditional 'floxed' alleles for each gene, in which exon 23 for Ptprd, exon 4 for Ptprs, and exons 8, 9, and 10 for Ptprf were flanked by loxP sites. Survival of each individual floxed allele was analyzed in offsprings of heterozygote matings through comparison of obtained genotypes of offsprings on or after P14 to expected genotypes for Mendelian inheritance. The three floxed lines were intercrossed and maintained as triple-homozygote mice. The conditional PTPd, PTPs, and LAR alleles were genotyped using the oligonucleotide primers CAGAGGTGGCTCATGTGC and GCCCAACCC TCAATTGTCAGAC (PTPd, 465 and 287 bp bands for the floxed and wild-type alleles, respectively), GAGTCCTCAAACCAGGCCCTG and GGTGAGACCAGGGTGGGTTC (PTPs, 522 and 345 bp bands for the floxed and wild-type alleles, respectively), and GATGGTCCCTCTGGAGAC and GCCAAGCCCATGCTCAGAG (LAR, 498 and 289 bp bands for the floxed and wild-type alleles, respectively). All animal experiments were approved by the Harvard University Animal Care and Use Committee.

Neuronal cultures and production of lentiviruses
Primary hippocampal cultures were prepared as described Held et al., 2020;Wong et al., 2018). Briefly, hippocampi of newborn (postnatal days P0 or P1) pups were digested in papain, and neurons were plated onto glass coverslips in plating medium composed of mimimum essential medium (MEM) supplemented with 0.5% glucose, 0.02% NaHCO 3 , 0.1 mg/ml transferrin, 10% fetal select bovine serum, 2 mM L-glutamine, and 25 mg/ml insulin. After 24 hr, plating medium was exchanged with growth medium composed of MEM with 0.5% glucose, 0.02% NaHCO 3 , 0.1 mg/ml transferrin, 5% fetal select bovine serum (Atlas Biologicals), 2% B-27 supplement, and 0.5 mM L-glutamine. At DIV2-3, cytosine b-D-arabinofuranoside (AraC) was added to a final concentration of 1-2 mM. Cultures were kept in a 37˚C incubator for a total of 14-16 d before analyses proceeded. Lentiviruses were produced in HEK293T cells maintained in DMEM supplemented with 10% bovine serum and 1% penicillin/streptomycin. HEK293T cells were transfected using the calcium phosphate method with a combination of three lentiviral packaging plasmids (REV, RRE, and VSV-G) and a separate plasmid encoding either Cre recombinase or inactive Cre, at a molar ratio of 1:1:1:1. Twenty-four hours after transfection, the medium was changed to neuronal growth medium, and 18-30 hr later, the supernatant was used for viral transduction. Neuronal cultures were infected 6 d after plating with lentiviruses expressing EGFP-tagged Cre recombinase (pHN131014) or an inactive variant (pHN131015) expressed under the human Synapsin promotor (Liu et al., 2014), and infection rates were assessed via nuclear EGFP fluorescence. Only cultures in which no non-infected neurons could be detected were used for analyses.

Immunofluorescence staining of neurons
Neurons grown on #1.5 glass coverslips were fixed at DIV15 in 4% paraformaldehyde (PFA) for 10 min (except for staining with anti-Ca V 2.1 and PTPd antibodies, for which 2% PFA was used), followed by blocking and permeabilization in PBS containing 3% BSA/0.1% Triton X-100/PBS for 1 hr at room temperature. Incubation with primary and secondary antibodies was performed overnight at 4˚C and for 1 hr at room temperature, respectively. Samples were post-fixed in 4% PFA for 10 min and mounted onto glass slides using ProLong diamond mounting medium. Antibodies were diluted in blocking solution. Three 5 min washes with PBS were performed between steps. Primary antibodies used were as follows: rabbit anti-Liprin-a3 (

STED and confocal imaging
All images were acquired as described Held et al., 2020;Wong et al., 2018) using a Leica SP8 Confocal/STED 3Â microscope equipped with an oil-immersion 100 Â 1.44 N.A objective, white lasers, gated detectors, and 592 nm and 660 and 770 nm depletion lasers. For every region of interest (ROI), quadruple color sequential confocal scans for Synaptophysin, PSD-95, Gephyrin, and a protein of interest (RIM, Munc13-1, PTPd, Liprin-a or Ca V 2.1) were followed by triple-color sequential STED scans for PSD-95, Gephyrin, and the protein of interest. Synaptophysin was only imaged in confocal mode because of depletion laser limitations. Identical settings were applied to all samples within an experiment. For analyses of synapse density, Synaptophysin signals were used to generate ROIs using automatic detection with a size filter of 0.4-2 mm 2 (code available at https://github.com/kaeserlab/3DSIM_Analysis_CL and https://github. com/hmslcl/3D_SIM_analysis_HMS_Kaeser-lab_CL) and as described before Held et al., 2020;Liu et al., 2018). To measure synaptic levels of PTPd, RIM, Munc13-1, Liprin-a3, and Ca V 2.1 in confocal mode, a mask was generated in ImageJ using an automatic threshold in the Synaptophysin or the PSD-95 channel, and the levels were measured within that mask. For STED quantification, side-view synapses were selected while blind to the protein of interest. They were defined as synapses that contained a vesicle cluster (imaged in confocal mode) with a single bar-like Gephyrin or PSD-95 structure (imaged by STED) along the edge of the vesicle cluster. A 1 mm long, 250 nm wide profile was selected perpendicular to the postsynaptic density marker and across its center. The peak levels of the protein of interest were then measured as the maximum intensity of the line profile within 100 nm of the postsynaptic density marker peaks (estimated area based on Wong et al., 2018) after applying a 5-pixel rolled average. For side-view plots, line scans from individual side-view synapses were aligned to the peak of PSD-95 or Gephyrin after the 5-pixel rolling average was applied, and averaged across images. Only for representative images, a smooth filter was added, brightness and contrast were linearly adjusted, and images were interpolated to match publication standards. These adjustments were made identically for images within an experiment. All quantitative analyses were performed on original images without any processing, and all data were acquired and analyzed by an experimenter blind to genotype. For PTPd STED analyses, synapses were considered PTPd positive if the peak intensity was higher than three standard deviations above the average of the cTKO RPTP signal, assessed separately in each individual culture.

High-pressure freezing and electron microscopy
Electron microscopy was performed as previously described (Held et al., 2020;Wang et al., 2016).
Briefly, DIV15 neurons grown on 6 mm sapphire cover slips were frozen with a Leica EM ICE highpressure freezer in extracellular solution containing 140 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 10 mM glucose, 10 mM Hepes, 20 mM CNQX, 50 mM AP5, and 50 mM picrotoxin (pH 7.4, 310 mOsm). Freeze substitution was done in acetone containing 1% osmium tetroxide, 1% glutaraldehyde, and 1% H 2 O as follows: À90˚C for 5 hr, 5˚C per hour to À20˚C, À20˚C for 12 hr, and 10˚C per hour to 20˚C. Samples were then infiltrated in epoxy resin and baked at 60˚C for 48 hr followed by 80˚C overnight. Next, sapphire coverslips were removed from the resin block by heat shock, and samples were sectioned at 50 nm with a Leica EM UC7 ultramicrotome. Sections were mounted on a nickel slot grid with a carbon-coated formvar support film and counterstained by incubation with 2% lead acetate solution for 10 s, followed by rinsing with distilled water. Samples were imaged with a JEOL 1200EX transmission electron microscope equipped with an AMT 2 k CCD camera. Images were analyzed using SynapseEM, a MATLAB macro provided by Drs. M. Verhage and J. Broeke. Bouton area was measured by outlining the perimeter of each synapse profile. Docked vesicles were defined as vesicles touching the presynaptic plasma membrane opposed to the PSD, with the electron density of the vesicular membrane merging with that of the target membrane. Synapse width was measured as the area between synaptically apposed cells in which an evenly spaced cleft was present and associated with pre-and postsynaptic densities. All data were acquired and analyzed by an experimenter blind to the genotype.

Electrophysiology
Electrophysiological recordings were performed as described before Held et al., 2020;Wang et al., 2016). Neurons were recorded at DIV15-16 in whole-cell patch-clamp configuration at room temperature in extracellular solution containing (in mM) 140 NaCl, 5 KCl, 1.5 CaCl 2 , 2 MgCl 2 , 10 HEPES (pH 7.4), and 10 Glucose, supplemented with 20 mM CNQX and 50 mM D-AP5 to block AMPA and NMDA receptors, respectively. Glass pipettes were pulled at 2.5-4 MW and filled with intracellular solution containing (in mM) 40 CsCl, 90 K-gluconate, 1.8 NaCl, 1.7 MgCl 2 , 3.5 KCl, 0.05 EGTA, 10 HEPES, 2 MgATP, 0.4 Na 2 -GTP, 10 phosphocreatine, CsOH, and 4 mM QX314-Cl (pH 7.4). Neurons were clamped at À70 mV, and series resistance was compensated to 4-5 MW, and recordings in which the uncompensated series resistance was >15 MW at any time during the experiment were discarded. Electrical stimulation was applied using a custom bipolar electrode made from Nichrome wire. A Multiclamp 700B amplifier and a Digidata 1550 digitizer were used for data acquisition, sampling at 10 kHz and filtering at 2 kHz. Data were analyzed using pClamp. The experimenter was blind during data acquisition and analyses.

Statistics
Summary data are shown as mean ± SEM. Unless noted otherwise, significance was assessed using t-tests or Mann-Whitney U tests depending on whether assumptions of normality and homogeneity of variances were met (assessed using Shapiro or Levene's tests, respectively). Two-way ANOVA tests on a 200 nm wide window centered around the PSD-95 peak were used for line profile analyses of STED data, and chi-square tests were used to assess mouse survival ratios. All data were analyzed by an experimenter blind to the genotype. For all quantifications, the specific tests used are stated in the corresponding figure legends.