SPARCL1 Promotes Excitatory but Not Inhibitory Synapse Formation and Function Independent of Neurexins and Neuroligins

Home , Excitatory synapse

Research Articles: Cellular/Molecular SPARCL1 promotes excitatory but not inhibitory synapse formation and function independent of neurexins and neuroligins https://doi.org/10.1523/JNEUROSCI.0454-20.2020

Cite as: J. Neurosci 2020; 10.1523/JNEUROSCI.0454-20.2020 Received: 25 February 2020 Revised: 17 June 2020 Accepted: 12 July 2020

This Early Release article has been peer-reviewed and accepted, but has not been through the composition and copyediting processes. The final version may differ slightly in style or formatting and will contain links to any extended data.

Alerts: Sign up at www.jneurosci.org/alerts to receive customized email alerts when the fully formatted version of this article is published.

4 SPARCL1 promotes excitatory but not inhibitory synapse formation and 5 function independent of neurexins and neuroligins

8 Kathlyn J. Gan1,# and Thomas C. Südhof1,2,#

10 1Department of Molecular and Cellular Physiology, 2Howard Hughes Medical Institute,

11 Stanford University School of Medicine, 265 Campus Drive, Stanford, CA 94305, USA

13 Short title: Mechanism of SPARCL1-induced synaptogenesis

15 Keywords: SPARC-like protein 1 (SPARCL1); synapse; synapse formation; NMDA 16 receptor; neurexin; neuroligin

18 Number of figures: 8 19 Number of tables: 1 20 Extended data: 8 Excel files 21 Number of words: Abstract = 228; Introduction = 681; Discussion = 1287 22

23 #Corresponding authors: [email protected] and [email protected]

24 25 The authors declare no competing financial interests.

26 ABSTRACT 27 Emerging evidence supports roles for secreted extracellular matrix proteins in boosting 28 synaptogenesis, synaptic transmission, and synaptic plasticity. SPARCL1 (a.k.a. Hevin), 29 a secreted non-neuronal protein, was reported to increase synaptogenesis by 30 simultaneously binding to presynaptic neurexin-1α and to postsynaptic neuroligin-1B, 31 thereby catalyzing formation of trans-synaptic neurexin/neuroligin complexes. However, 32 neurexins and neuroligins do not themselves mediate synaptogenesis, raising the 33 question of how SPARCL1 enhances synapse formation by binding to these molecules. 34 Moreover, it remained unclear whether SPARCL1 acts on all synapses containing 35 neurexins and neuroligins or only on a subset of synapses, and whether it enhances 36 synaptic transmission in addition to boosting synaptogenesis or induces silent synapses. 37 To explore these questions, we examined the synaptic effects of SPARCL1 and their 38 dependence on neurexins and neuroligins. Using mixed neuronal and glial cultures from 39 neonatal mouse cortex of both sexes, we show that SPARCL1 selectively increases 40 excitatory but not inhibitory synapse numbers, enhances excitatory but not inhibitory 41 synaptic transmission, and augments NMDA-receptor-mediated synaptic responses 42 more than AMPAR-mediated synaptic responses. None of these effects were mediated 43 by SPARCL1-binding to neurexins or neuroligins. Neurons from triple neurexin-1/2/3 or 44 from quadruple neuroligin-1/2/3/4 conditional knockout mice that lacked all neurexins or 45 all neuroligins were fully responsive to SPARCL1. Taken together, our results reveal that 46 SPARCL1 selectively boosts excitatory but not inhibitory synaptogenesis and synaptic 47 transmission by a novel mechanism that is independent of neurexins and neuroligins.

48 49 50 51 52 53 54 55 56

57 SIGNIFICANCE STATEMENT 58 Emerging evidence supports roles for extracellular matrix proteins in boosting synapse 59 formation and function. Previous studies demonstrated that SPARCL1, a secreted non- 60 neuronal protein, promotes synapse formation in rodent and human neurons. However, 61 it remained unclear whether SPARCL1 acts on all or on only a subset of synapses, 62 induces functional or largely inactive synapses, and generates synapses by bridging 63 presynaptic neurexins and postsynaptic neuroligins. Here, we report that SPARCL1 64 selectively induces excitatory synapses, increases their efficacy, and enhances their 65 NMDA receptor content. Moreover, using rigorous genetic manipulations, we show that 66 SPARCL1 does not require neurexins and neuroligins for its activity. Thus, SPARCL1 67 selectively boosts excitatory synaptogenesis and synaptic transmission by a novel 68 mechanism that is independent of neurexins and neuroligins. 69

70 INTRODUCTION

71 During development and throughout life, synapses are formed by a multi-component 72 process that is controlled by synaptic organizer molecules. Synapse formation requires 73 the initial establishment of contacts between an axon and a target cell, assembly of the 74 presynaptic release machinery and the postsynaptic receptor apparatus, and activity- 75 dependent specification of synapse properties (Südhof, 2018). Synapse formation is 76 guided by two types of organizer molecules: trans-synaptic adhesion molecules and 77 secreted factors (Ferrer-Ferrer and Dityatev, 2018; Yuzaki, 2018). Among adhesion 78 molecules, neurexins and their ligands, including neuroligins, orchestrate synapse 79 specification by mediating bidirectional, trans-cellular signaling (Südhof, 2017; Cao and 80 Tabuchi, 2017; Katzman and Alberini, 2018; Wang et al., 2019). Secreted factors act by 81 several mechanisms, for example by enhancing the recruitment of presynaptic active 82 zone proteins and postsynaptic scaffolding proteins (Wnts; Sahores et al., 2010; Cerpa 83 et al., 2011) or by connecting pre- and postsynaptic adhesion molecules (cerebellins; 84 Yuzaki, 2018). Extracellular matrix (ECM) proteins, reportedly secreted by glia (such as 85 thrombospondins and glypicans), may bridge pre- and postsynaptic structures and cluster 86 ionotropic glutamate receptors at postsynaptic sites (Christopherson et al., 2005; Allen et 87 al, 2012). Although an important role in synaptogenesis is emerging for several proteins 88 traditionally associated with the ECM, the underlying molecular mechanisms remain 89 unknown.

90 The present study examines SPARCL1 (‘secreted protein acidic and rich in cysteine-like 91 1’, a.k.a. Hevin), an ECM protein that induces synapse formation in rodent and human 92 neurons (Kucukdereli et al., 2011; Singh et al., 2016; Gan and Südhof, 2019). SPARCL1 93 belongs to the SPARC family of extracellular glycoproteins (Brekken and Sage, 2001). 94 SPARCL1 contains an N-terminal acidic domain, a central cysteine-rich follistatin domain, 95 and a C-terminal calcium-binding domain (Brekken and Sage, 2000). Little is known about 96 how SPARCL1 is secreted and processed. Like other ECM proteins, SPARCL1 is more 97 abundant during embryonic development than during adulthood, but its expression is 98 upregulated in response to trauma or disease (Lively and Brown, 2008 a,b).

99 In brain, SPARCL1 is localized to postsynaptic membranes and perisynaptic glial 100 processes in the cerebral cortex and the cerebellum (Lively et al., 2007), raising the 101 possibility that SPARCL1 regulates synapse formation. Indeed, Kucukdereli et al. (2011) 102 showed that SPARCL1 induced synapses in mouse retinal ganglion cells. Surprisingly, 103 however, these synapses were morphologically normal but did not appear to be 104 functional. In contrast, we found that SPARCL1 promotes formation of active synapses in 105 human excitatory neurons derived from embryonic stem cells (Gan and Südhof, 2019). 106 Moreover, SPARCL1 reportedly enhanced synapse formation by simultaneously binding 107 to presynaptic neurexin-1α (Nrxn1α) and postsynaptic neuroligin-1B (Nlgn1B) to form a 108 trans-synaptic complex (Singh et al., 2016). This finding was puzzling because Nrxn1α 109 forms tight complexes with all neuroligins in the absence of other factors (Boucard et al., 110 2005; Chih et al., 2006), and because biophysical studies failed to detect binding of 111 SPARCL1 to Nlgn1 (Elegheert et al., 2017). Moreover, deletion of all neuroligins 112 (Varoqueaux et al., 2006; Jiang et al., 2017; Chandra et al., 2017) or all neurexins (Missler 113 et al., 2003; Chen et al., 2017) did not decrease excitatory synapses numbers but 114 impaired synaptic transmission. Given that neurexins and neuroligins are apparently not 115 required for establishing initial synaptic contacts, in remained unclear how SPARCL1 116 could form non-functional excitatory synapses by binding to these adhesion molecules.

117 In the present study, we examined how SPARCL1 potently boosts synaptogenesis, 118 whether it acts selectively on subsets of synapses, and what role neurexins and/or 119 neuroligins play in its activity. Because key prior studies of SPARCL1-induced 120 synaptogenesis were conducted in mouse neurons (Kucukdereli et al., 2011; Singh et al., 121 2016), we performed our experiments in mixed cultures of mouse cortical neurons and 122 glia. We confirm that SPARCL1 induces excitatory synapse formation; however, in 123 contrast with previous studies, we show that SPARCL1 also enhances neurotransmission 124 by differentially elevating AMPA-receptor- (AMPAR-) and NMDA-receptor- (NMDAR-) 125 mediated synaptic responses. SPARCL1 exerts a much larger effect on NMDARs, 126 increasing the NMDAR/AMPAR ratio. Importantly, using cultures from conditional 127 knockout mice, we determined that neurexins and neuroligins do not mediate any of these 128 effects. Thus, SPARCL1 selectively activates excitatory synapse formation and signaling 129 to boost synaptic connectivity by a neurexin- and neuroligin-independent mechanism.

130 MATERIALS AND METHODS

131 Experimental design and statistical analyses. Male and female newborn (P0) mice 132 were used for all experiments. Nrxn123 cKO and Nlgn1234 conditional knockout (cKO) 133 mice were described previously (Chen et al., 2017; Wu et al., 2019, respectively). Mice 134 were housed in the SIM1 Animal Facility at Stanford University on a 12:12 hour light-dark 135 cycle with ad libitum access to food and water. All procedures conformed to institutional 136 guidelines and were approved by the Administrative Panel on Laboratory Animal Care 137 associated with the Stanford University School of Medicine.

138 Statistical analyses were performed using Prism 8 software (GraphPad Software) and are 139 summarized in Table 1 (see Extended Data for full statistical analyses). In the figures, 140 quantitative data shown are means ± SEMs. All experiments were independently 141 repeated at least 3 times. All statistical analyses were performed using unpaired t-tests 142 or one-way ANOVAs with Tukey’s post-hoc tests, comparing control and treated 143 conditions within the same experiments. P-values for all t-tests and Tukey’s post-hoc 144 tests for multiple comparisons are provided as Extended Data.

145 Primary cultures. Cortical neurons were cultured from male and female P0 mice 146 essentially as described (Maximov et al., 2007). Dissected cortices were dissociated by 147 papain digestion for 20 min at 37o C, filtered through a 70 μm cell strainer, plated on 148 borosilicate glass coverslips (Carolina; 0.09-0.12 mm thickness, 15 mm diameter) 149 precoated with Matrigel (Corning), and placed in 24-well plates. Plating medium contained 150 10% fetal bovine serum (Atlanta), Gem21 NeuroPlex supplement (Gemini), 0.4% D- 151 glucose and 2 mM glutamine in Minimal Essential Medium (MEM; Life Technologies). 152 Two hours after plating, the medium was completely exchanged to growth medium 153 comprised of 5% fetal bovine serum, Gem21 NeuroPlex supplement and 2 mM glutamine 154 in Neurobasal A (Life Technologies). At 3-4 days in vitro (DIV3-4), 50% of the growth 155 medium was exchanged with serum-free Neurobasal A supplemented with Gem21 156 NeuroPlex, 2 mM glutamine, and 4 μM cytosine arabinofuranoside (Sigma) to prevent 157 glial overgrowth. From DIV7 onwards, 30% of the medium was exchanged every 3 days 158 until analysis at DIV14-16. To prepare pure glial cultures, cortices from P0 mice were 159 dissected and digested as described above and then harshly triturated. Dissociated cells

160 from 4 cortices were plated in a T75 flask in Dulbecco’s Modified Eagle Medium (DMEM) 161 supplemented with 10% fetal bovine serum (FBS). Upon reaching confluence, glial cells 162 were trypsinized and re-plated at a lower density in 24-well plates. Glia were analyzed at 163 DIV14-16.

164 Lentivirus preparation and infection of cortical neurons. Lentiviruses for conditional 165 deletion of neurexins and neuroligins in cortical neurons were generated from the 166 following expression plasmids: Syn-Cre-EGFP (active Cre recombinase) and Syn-ΔCre- 167 EGFP (inactive Cre recombinase used as a control; Kaeser et al., 2011). Lentiviruses 168 were produced in HEK293T cells essentially as described (Zhang et al., 2013). Using 169 calcium phosphate, HEK293T cells were cotransfected with the three helper plasmids 170 pRSV-REV (3.9 μg), pMDLg/pRRE (8.1 μg), and VSV-G (6.0 μg) and with either Cre or 171 ΔCre lentiviral plasmid (12 μg). Lentiviruses in the culture medium were harvested 48 h 172 after transfection, and virus particles were pelleted by ultracentrifugation at 19,000 rpm 173 for 1.5 h. Pellets were resuspended in 100 μL DMEM, aliquoted, and stored at -80oC. 174 Primary cortical neurons were infected at DIV3 by diluting the concentrated virus 1:1000 175 into the culture medium in each well.

176 Sparse transfection of cortical neurons. To assess dendritic arborization in isolated 177 cortical neurons, cultures were sparsely transfected at DIV10 with pmU-β-actin-eBFP 178 (Gary Banker, Oregon Health and Sciences University) using a calcium phosphate 179 method (CalPhos Mammalian Transfection Kit, Takara; manufacturer’s protocol modified 180 as described here). A DNA/calcium phosphate precipitate was prepared by mixing the

181 following (per 24-well reaction): 1 μg DNA, 3.1 μL 2 M CaCl2, and ddH2O to a volume of 182 25 μL. The DNA mixture was added dropwise under low-powered vortex pulses to an 183 equal volume of 2X HEPES-buffered saline. The precipitate formed for 20 min prior to 184 addition to the cultures. Cultured neurons were transferred into fresh serum-free growth 185 medium that was conditioned overnight by pure glia (0.5 mL per 24-well). DNA/calcium 186 phosphate precipitate was added dropwise to each well, and the cultures were incubated o 187 at 5% CO2, 37 C. After 1.5 h, the precipitate was dissolved using growth medium that

188 was pre-acidified in a 10% CO2 incubator to increase the viability of transfected neurons. 189 Following complete dissolution of the precipitate, neurons were returned to their “home”

o 190 plate containing their original conditioned medium and maintained at 5% CO2, 37 C until 191 morphological analysis at DIV14.

192 Recombinant expression of SPARCL1. Expression constructs encoding the full-length 193 mouse cDNA clone of SPARCL1 (Dharmacon) or mClover as a control (pEB-Multi-Neo- 194 SPARCL1, pEB-Multi-Neo-mClover; Gan and Südhof, 2019) were transfected into 195 HEK293T cells using calcium phosphate and incubated in serum-free medium. After 72 196 h, the culture medium was collected, passed through a 0.45 μm syringe filter to remove 197 cell debris, and concentrated by centrifugal ultrafiltration through an Amicon regenerated 198 cellulose membrane with a nominal molecular weight limit of 3 kDa (Millipore). Samples 199 were centrifuged at 14,000 x g for 30 min at 4o C, and the protein concentrates were 200 subsequently recovered from the membranes with a reverse spin at 1,000 x g for 2 min 201 at room temperature. Concentrated supernatants were diluted 1:100 in serum-free 202 neuronal growth medium and added to cultured neurons at DIV13. Neurons were assayed 203 within 1-3 days after treatment.

204 Immunocytochemistry. All immunocytochemistry experiments were performed 205 essentially as described (Kwinter et al., 2005). Briefly, cortical neurons cultured at low 206 density were fixed with 4% paraformaldehyde for 15 min at 37oC, permeabilized with 0.1% 207 Triton-X for 10 min at room temperature and blocked with 0.5% fish skin gelatin for 1 h at 208 37oC. To quantify excitatory and inhibitory synapses localized to dendritic segments, 209 neurons were stained with anti-VGLUT1 guinea pig (Millipore; 1:1000) or anti-VGAT 210 rabbit (Millipore; 1:1000) and counterstained with anti-MAP2 mouse (1:1000; Sigma). All 211 primary antibodies were diluted in blocking solution and incubated overnight at 4oC with 212 gentle agitation. Neurons were subsequently incubated with compatible secondary 213 antibodies conjugated to Alexa-488, Alexa-546 and Alexa-647 fluorophores (1:500; Life 214 Technologies) for 1.5 h at 37oC, washed in PBS, and mounted in DAPI Fluoromount-G 215 (Southern Biotech).

216 Image acquisition and analysis. Images of neurons with pyramidal morphology were 217 acquired using a Nikon A1RSi confocal microscope with constant laser gain and offset 218 settings, scanning speed, and pinhole size. These settings yielded images in which the 219 brightest pixels were not oversaturated. Each Z-stack was comprised of 10 serial images

220 acquired in 0.5-μm steps. Maximum intensity projections of the Z-stacks were generated 221 for quantification purposes. Synaptic puncta were counted along well isolated primary 222 dendrites (5 x 100-μm dendritic segments per cell) using the “Count Nuclei” application 223 in Metamorph (Molecular Devices). To measure dendritic length and branching, field 224 images of low-density neuronal cultures expressing eBFP and counterstained with anti- 225 MAP2 mouse and anti-NeuN (Millipore; rabbit polyclonal, 1:1000) were analyzed using 226 the “Neurite Outgrowth” application in Metamorph. Constant threshold settings to exclude 227 background signals were maintained for all experimental conditions.

228 Electrophysiology. Electrophysiological recordings were performed in the whole cell 229 configuration essentially as described (Maximov et al., 2007; Sando et al., 2019). Patch 230 pipettes were pulled from borosilicate glass capillary tubes (Warner Instruments) using a 231 PC-10 pipette puller (Narishige). The resistance of pipettes filled with intracellular solution 232 varied between 3.0-4.0 MΩ. The bath solution in all experiments contained (in mM): 140

233 NaCl, 5 KCl, 2 CaCl2, 2 MgCl2, 10 HEPES, and 10 glucose (pH adjusted to 7.4 with NaOH, 234 300 mOsm/L). Excitatory miniature and evoked synaptic responses were recorded in 235 voltage-clamp mode with a pipette solution containing (in mM) 135 Cs-methanesulfonate,

236 8 CsCl, 10 HEPES, 0.25 EGTA, 0.3 Na2GTP, 2 MgATP, 7 phosphocreatine, and 10 QX- 237 314 (pH adjusted to 7.3 with CsOH; 303 mOsm/L). Inhibitory miniature and evoked 238 synaptic responses were recorded in voltage-clamp mode with a pipette solution

239 containing (in mM) 146 CsCl, 10 HEPES, 0.25 EGTA, 2 MgATP, 0.3 Na2GTP, 7 240 phosphocreatine and 10 QX-314 (pH adjusted to 7.3 with CsOH; 303 mOsm/L). Evoked 241 synaptic responses were triggered by 0.5-ms current (100 μA) injection through a local 242 extracellular electrode (FHC concentric bipolar electrode) placed 100-150 μm from the 243 soma of neurons recorded. The frequency, duration and magnitude of the extracellular 244 stimulus were controlled with a Model 2100 Isolated Pulse Stimulator (A-M Systems, Inc.) 245 synchronized with the Clampex 10.1 data acquisition software (Molecular Devices). 246 Miniature excitatory and inhibitory postsynaptic currents (mEPSCs and mIPSCs) were 247 monitored in the presence of tetrodotoxin (1 μM). Evoked AMPAR- and NMDAR-EPSCs 248 were pharmacologically isolated with picrotoxin (50 μM) and recorded at -70 mV or +40 249 mV holding potentials, respectively. GABAR IPSCs were pharmacologically isolated with 250 D-AP5 (50 μM) and CNQX (10 μM). Data were digitized at 10 kHz with a 2 kHz low-pass

251 filter using a Multiclamp 700B Amplifier (Molecular Devices). Data were analyzed using 252 Clampfit 10.5 (Molecular Devices) software.

253 Semiquantitative RT-PCR. Total RNA was isolated from cortical and glial cultures using 254 the PureLink RNA Mini Kit (Thermo Fisher) according to the manufacturer’s protocol. RNA 255 concentrations were measured using a ND-1000 Nanodrop spectrophotometer (Thermo 256 Fisher). Intron spanning primers were designed to distinguish between gDNA and cDNA 257 amplification. The following primer sequences for SPARCL1 and β-actin were used: 258 SPARCL1-F, 5’-AGACGCTACAGTCCCCATTG-3’; SPARCL1-R, 5’- 259 GCCTGCACCATGCTTAGAGT-3’; β-actin-F, 5’- CCCCTGAACCCTAAGGCCA-3’; β- 260 actin-R, 5’- CGGACTCATCGTACTCCTGC-3’. Reverse transcription and subsequent 261 cDNA amplification were performed using the SuperScript IV One-Step RT-PCR System 262 (Thermo Fisher). Reaction components were assembled as follows: 25 μL of 2X Platinum 263 SuperFi RT-PCR Master Mix, 2.5 μL of gene-specific forward primer (10 μM), 2.5 μL of 264 gene-specific reverse primer (10 μM), 0.5 μL of SuperScript IV reverse transcriptase mix,

265 500 ng of template RNA, and nuclease-free ddH2O to 50 μL. The following cycling 266 parameters were used: reverse transcription, 50 oC for 10 min; reverse transcriptase 267 inactivation/initial denaturation, 98 oC for 10 s; cDNA amplification (25 cycles), 98 oC for 268 10 s, 60 oC for 10 s, 72 oC for 30 s; final extension, 72 oC for 5 min. Amplicons were 269 analyzed by electrophoresis on a 0.8% agarose gel. Quantitative analysis was performed 270 using a Molecular Imager GelDoc XR+ system and Image Lab software (Bio-Rad).

271 Immunoblotting. Conditioned media from cortical and pure glial cultures were collected 272 and precipitated with 1/5 volume of trichloroacetic acid (TCA) at 4 oC for 30 min. 273 Precipitates were centrifuged at 14,000 x g for 15 min, aspirated, and washed twice with 274 ice-cold acetone. Dried protein pellets were resuspended in sample buffer containing 200 275 mM Tris-Cl pH 6.8, 8% SDS, 0.4% bromophenol blue and 40% glycerol and boiled in the 276 presence of 1% β-mercaptoethanol for 10 min. Cortical neurons and pure glia in 24-well 277 plates were lysed in RIPA buffer containing 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 278 25 mM Tris-HCl pH 7.4, and Complete EDTA-Free Protease Inhibitor Cocktail (Sigma). 279 Lysates were incubated on ice for 30 min and clarified by centrifugation at 14,000 x g for 280 30 min at 4 oC. Lysates were boiled in sample buffer containing 1% β-mercaptoethanol 281 for 10 min. Proteins were analyzed by SDS-PAGE using 4-20% Mini-Protean TGX

282 precast gels (Bio-Rad). Proteins were transferred onto nitrocellulose membranes for 10- 283 12 min at 25 V using the Trans-Blot Turbo transfer system (Bio-Rad). Membranes were 284 blocked in 5% milk diluted in TBST for 1 h at room temperature. Membranes were then 285 incubated overnight at 4 oC with the following primary antibodies diluted in blocking 286 solution: anti-SPARCL1 goat (R&D Biosystems; 1:5000), anti-pan-Nrxn rabbit (G394, 287 Südhof Lab, 1:500), anti-Nlgn1 mouse (SySy, 1:1000), anti-Nlgn2 mouse (SySy, 1:1000), 288 anti-Nlgn3 rabbit (639B, Südhof Lab, 1:500), anti-β-actin mouse (A1976, Sigma, 1:2000). 289 Membranes were subsequently incubated for 1 h at room temperature with the following 290 compatible secondary antibodies (LI-COR), diluted 1:10,000 in blocking solution: IRDye 291 680LT donkey anti-goat; IRDye 800CW donkey anti-rabbit. Quantitative analysis was 292 performed by a dual-channel infrared imaging system, an Odyssey Infrared Imager CLX 293 and Image Studio 5.2.5 software (LI-COR).

294 295 RESULTS 296 Endogenous SPARCL1 protein is secreted at low levels by cultured mouse glia but 297 is abundant in mouse serum. SPARCL1 is expressed by astrocytes (Saunders et al., 298 2018; The Tabula Muris Consortium, 2018; Zeisel et al., 2018). Nevertheless, addition of 299 SPARCL1 to human neurons co-cultured with glia increases synapse formation and 300 activity, suggesting that endogenous SPARCL1 secreted by glia does not fully stimulate 301 neurons (Gan and Südhof, 2019). To examine how much SPARCL1 is expressed by glia, 302 we developed a semiquantitative RT-PCR assay (Figure 1A, 1B). Intron-spanning primers 303 for the SPARCL1 and β-actin (control) genes were designed to distinguish between 304 mRNA and genomic DNA amplification (Figure 1A). We performed RT-PCR on total RNA 305 isolated from glia cultured alone or from glia co-cultured with cortical neurons. SPARCL1 306 and β-actin transcripts from both sources were amplified either individually or in 307 multiplexed format to estimate the relative abundance of SPARCL1 by normalization to 308 β-actin (Figure 1B). Analysis by electrophoresis revealed specific amplification of 309 SPARCL1 and β-actin mRNA without genomic DNA contamination (Figure 1B). 310 Quantification of multiplexed amplicons showed that SPARCL1 mRNA is expressed by 311 glia at similar levels when cultured alone or when co-cultured with neurons (Figure 1C). 312 These results confirm that most cortical SPARCL1 is produced by glia.

313 Next, we examined SPARCL1 protein expression in conditioned media and lysates from 314 glia cultured alone or co-cultured with neurons (Figure 1C). We also compared glial 315 SPARCL1 protein expression to SPARCL1 protein present in serum (Figure 1C). Using 316 immunoblotting with an antibody directed to the N-terminal acidic domain of SPARCL1 317 (previously validated in Gan and Südhof, 2019), we detected full-length SPARCL1 protein 318 (~130 kDa) in both the conditioned medium and lysates of glia cultured alone or co- 319 cultured with neurons (Figure 1C). Similar SPARCL1 levels were observed in both of 320 these culture systems, implying that most cortical SPARCL1 is secreted by glia. Note that 321 SPARCL1 migrates on SDS-PAGE gels at an apparent molecular weight of ~130 kDa, at 322 nearly twice its predicted size of 75 kDa. This abnormally high apparent molecular weight 323 is due to the high content of charged residues in SPARCL1 (Girard and Springer, 1995 324 and 1996). As previously reported (Gan and Südhof, 2019), we also observed SPARCL1 325 in mouse serum (Figure 1C). Serum SPARCL1 exhibited an apparent molecular weight 326 of ~50 kDa, consistent with earlier findings that SPARCL1 is physiologically processed 327 by matrix metalloproteases (Weaver et al., 2010). In order to detect SPARCL1 in 328 conditioned media from pure or co-cultured glia, we needed to concentrate the media 10- 329 fold, whereas no such concentration was required to detect SPARCL1 in serum. 330 Furthermore, quantitative immunoblotting revealed that the amounts of SPARCL1 protein 331 synthesized and secreted by glia are 30% (lysates) and 3% (media; taking into account 332 the 10-fold concentration of the media) of the amount present in serum (Figure 1C). 333 These observations suggest that SPARCL1 is secreted from glia at low levels.

334 Recombinant SPARCL1 potently increases excitatory but not inhibitory synapse 335 numbers in mixed cortical cultures of neurons and glia. SPARCL1 increases synapse 336 numbers when added to excitatory human neurons co-cultured with mouse glia (Gan and 337 Südhof, 2019). Does SPARCL1 also act on mouse neurons co-cultured with glia, and is 338 SPARCL1 a general synaptotrophic factor, or is it specific for subsets of synapses? To 339 answer these questions, we expressed native SPARCL1 in HEK293T cells, collected the 340 supernatant containing the secreted recombinant protein, and diluted it into the growth 341 medium of neuron-glia cultures from the cortex of newborn mice (hereafter termed “mixed 342 cortical cultures”). As a negative control, we used supernatant from HEK293T cells 343 transfected with mClover. As we demonstrated previously (Gan and Südhof, 2019), the

344 low expression levels of native recombinant SPARCL1 made it difficult to measure its 345 concentration in the HEK cell supernatant (Figure 1D). However, we compared 346 recombinant and endogenous SPARCL1 levels by immunoblotting (Figure 1D). 347 Quantitation using fluorescently labeled secondary antibodies revealed that the amount 348 of recombinant SPARCL1 used in our experiments was approximately equivalent to the 349 amount of endogenous SPARCL1 secreted by glia in cortical cultures (Figure 1D).

350 Mixed cortical cultures were treated with the SPARCL1 or control supernatants at 13 days 351 in vitro (DIV13), and the neurons were analyzed by immunocytochemistry at DIV14-16 352 (Figure 2A). Immunocytochemistry for the excitatory pre- and postsynaptic markers 353 vGluT1 and Homer1, respectively, showed that SPARCL1 elevated the density of 354 excitatory synapses (>50% increase) without changing their apparent size (Figure 2B, 355 2C). However, immunocytochemistry for the inhibitory pre- and postsynaptic markers 356 vGAT and gephyrin, respectively, detected no significant difference in inhibitory synapse 357 density between control and SPARCL1-treated neurons (Figure 2D, 2E). Thus, SPARCL1 358 appears to be specific for excitatory synapses.

359 A possible explanation for the large effect of SPARCL1 on excitatory synapse density is 360 that SPARCL1 enhances the development of excitatory neurons. To investigate this 361 possibility, we sparsely transfected mixed cortical cultures at DIV10 with enhanced blue 362 fluorescent protein (eBFP) to visualize dendritic arbors. We treated the cultures with 363 recombinant SPARCL1 at DIV13 and measured the neuronal cell density, number of 364 primary dendrites, number of dendritic branches, and size of the neuronal cell body at 365 DIV14-16. SPARCL1 did not alter any of these parameters, suggesting that it does not 366 affect overall neuronal development (Figure 2F-2H). In addition, we used whole-cell patch 367 clamping to measure the capacitance and input resistance of the control- and SPARCL1- 368 treated neurons, and again detected no difference (Figure 2I, 2J). Thus, SPARCL1 369 selectively enhances excitatory synapse numbers in mixed cortical cultures without 370 affecting the development of other neuronal features.

371 Cre-recombinase expression substantially reduces neurexin and neuroligin 372 protein expression in conditional knockout (cKO) neurons. To determine whether 373 the cellular effects of SPARCL1 are dependent on neurexins and neuroligins as

374 suggested (Singh et al., 2016), we generated mixed cortical cultures from triple neurexin 375 cKO mice (Nrxn123 cKO, Chen et al., 2017) and quadruple Nlgn1234 cKO mice (Wu et 376 al., 2018). At DIV3, we infected the cells with lentiviruses expressing inactive mutant Cre- 377 recombinase (ΔCre) to retain wildtype neurexin or neuroligin expression, or active Cre- 378 recombinase (Cre) to delete all neurexin isoforms except for Nrxn1J, a minor variant 379 (Sterky et al., 2017) or all neuroligin isoforms (Figure 3). Lentiviral expression of ΔCre 380 and Cre was driven by the neuron-specific synapsin promoter, and both proteins 381 contained nuclear localization signals and a fused eGFP-moiety for visualization of 382 expression (Kaeser et al., 2011). The lentiviruses infected all neurons (Figure 3A). 383 Previous studies documented that lentiviral expression of Cre causes complete 384 recombination of neurexins and neuroligins in cultured neurons (Khajal et al., 2020; 385 Chanda et al., 2017). To confirm this conclusion in the present study, we performed 386 immunoblots on lysates from Nrxn123 cKO and Nlgn1234 cKO cortical cultures infected 387 with lentiviral ΔCre or Cre (Figure 3B, 3C). Cre expression caused a substantial loss of 388 all neurexin isoforms (~80%; Figure 3B) and of neuroligin-1 to -3 (~80-83%; Figure 3C). 389 Note that neuroligin-4 is not significantly expressed in this culture system (Chanda et al., 390 2017; Varoqueaux et al., 2006; Hoon et al., 2011). Our results validate the use of Nrxn123 391 and Nlgn1234 cKO neurons for characterizing the cellular effects of SPARCL1.

392 Deletion of neurexins fails to block excitatory synaptogenesis induced by 393 recombinant SPARCL1. To determine whether SPARCL1 requires neurexins for 394 synaptogenesis, we generated mixed cortical cultures from triple neurexin cKO mice 395 (Nrxn123 cKO, Chen et al., 2017). We observed that the neurexin deletion had no effect 396 on excitatory synapse numbers (Figure 4A-4B; note that pre-synaptic VGluT1 is used as 397 a proxy for excitatory synapses) and caused a small, statistically insignificant decrease in 398 inhibitory synapse numbers (Figure 4A, 4C; note that pre-synaptic VGAT is used as a 399 proxy for inhibitory synapses). Furthermore, the neurexin deletion failed to block the 400 selective induction of excitatory synapses by SPARCL1 (Figure 4A-4C). Thus, neurexins 401 are not required for the synaptogenic activity of SPARCL1.

402 SPARCL1 increases spontaneous excitatory but not inhibitory synaptic 403 transmission independent of neurexins. Does SPARCL1 induce functional synapses,

404 and do neurexins influence the effect of SPARCL1 on synaptic transmission, even if they 405 do not regulate synapse numbers? To address these questions, we treated mixed cortical 406 cultures from Nrxn123 cKO mice with SPARCL1 and recorded miniature excitatory 407 postsynaptic currents (mEPSCs) in the presence of tetrodotoxin and picrotoxin. In 408 wildtype (ΔCre) neurons, SPARCL1 increased the mEPSC frequency (~300%) but did 409 not significantly change the mEPSC amplitude (Figure 5A). Notably, SPARCL1 increased 410 the mEPSC frequency more robustly than the excitatory synapse density (Figure 4A-C), 411 suggesting that SPARCL1-induced synapses are fully functional. In Nrxn123 cKO (Cre) 412 neurons, SPARCL1 potentiated the mEPSC frequency similar to wildtype neurons, 413 showing that SPARCL1 acts independent of neurexins (Figure 5A). Deletion of neurexins 414 in itself significantly decreased the mEPSC frequency (~25%), again without altering the 415 mEPSC amplitude (Figure 5A).

416 We next recorded miniature inhibitory postsynaptic currents (mIPSCs) in the presence of 417 tetrodotoxin, CNQX and D-AP5 (Figure 4B). SPARCL1 had no effect on mIPSC frequency 418 and amplitude in the presence (ΔCre) or absence (Cre) of neurexins. The neurexin 419 deletion in itself decreased the mIPSC frequency (~50%) without significantly altering the 420 mIPSC amplitude (Figure 5B).

421 Together, these results suggest that, in addition to promoting excitatory synapse 422 formation, SPARCL1 selectively and potently enhances excitatory synaptic transmission. 423 Importantly, neurexins do not mediate these effects. Rather, neurexins are primarily 424 required for normal synaptic transmission because their deletion decreases excitatory 425 and inhibitory spontaneous activity without changing synapse numbers, but the neurexin 426 and SPARCL1 effects are completely independent of each other.

427 Deletion of neuroligins also fails to block excitatory synaptogenesis induced by 428 recombinant SPARCL1. Even though SPARCL1 does not require neurexins for 429 synaptogenesis, it is still possible that it might act by binding to neuroligins, which likely 430 bind to other adhesion molecules in addition to neurexins (Connor et al., 2019). Thus, we 431 asked whether SPARCL1 acts via neuroligins. We infected mixed cortical cultures from 432 quadruple Nlgn1234 cKO mice (Wu et al., 2018) with lentiviruses encoding 'Cre to retain 433 wildtype neuroligin expression or Cre to induce the pan-neuroligin deletion (Figure 2A;

434 Figure 3). We then analysed the neurons by immunocytochemistry (Figure 6) and by 435 electrophysiological recordings of spontaneous and evoked synaptic transmission 436 (mEPSCs and mIPSCs, Figure 7; EPSCs and IPSCs, Figure 8).

437 As described above, quantifications of synaptic puncta confirmed that SPARCL1 438 significantly increased the density of VGluT1-positive excitatory synapses (~50%) without 439 affecting the density of VGAT-positive inhibitory synapses (Figure 6). Again, the size and 440 staining intensity of excitatory and inhibitory synapses were unaffected by SPARCL1 441 treatment. The pan-neuroligin deletion had no effect on excitatory and inhibitory synapse 442 density, size, or staining intensity (Figure 6). Notably, the pan-neuroligin deletion did not 443 prevent SPARCL1 from selectively increasing excitatory synapse density (Figure 6).

444 Measurements of mEPSCs and mIPSCs from wildtype ('Cre) and neuroligin-deficient 445 (Cre) neurons confirmed that SPARCL1 potently but selectively increased the mEPSC 446 frequency (Figure 7A). We also observed that SPARCL1 significantly enhanced the 447 mEPSC amplitude in wildtype neurons. As before, SPARCL1 had no effect on the mIPSC 448 frequency or amplitude (Figure 7B). Strikingly, the pan-neuroligin deletion also did not 449 block SPARCL1 from enhancing excitatory synaptic activity, although the pan-neuroligin 450 deletion in itself produced major impairments in both excitatory and inhibitory synaptic 451 transmission. Specifically, the pan-neuroligin deletion decreased the frequencies of both 452 mEPSCs and mIPSCs (~50%) and reduced the amplitudes of mEPSCs (~30%) and 453 mIPSCs (~45%). Taken together, these results show that neuroligins are not required for 454 SPARCL1 to enhance excitatory synapse density and synaptic transmission, but that 455 neuroligins act as major synaptic organizers whose deletion severely impairs excitatory 456 and inhibitory synapse function without changing synapse numbers.

457 SPARCL1 enhances excitatory but not inhibitory evoked neurotransmission and 458 increases the synaptic NMDAR content independent of neuroligins. Previously, we 459 showed in human neurons that SPARCL1 not only increases spontaneous synaptic 460 transmission but also boosts NMDAR-mediated synaptic responses (Gan and Südhof, 461 2019). To determine if SPARCL1 also exhibits this specific activity in mouse neurons, we 462 measured synaptic responses evoked by single action potentials in mixed cortical 463 neurons as a function of SPARCL1 treatment. In the same experiments, we also tested

464 the possibility that neuroligins are selectively required for SPARCL1 to enhance NMDAR- 465 mediated responses because neuroligin-1 has been linked to NMDARs in other studies 466 (Chubykin et al., 2007; Jiang et al., 2017; Wu et al., 2019).

467 We treated wildtype ('Cre) and neuroligin-deficient (Cre) neurons with control or 468 SPARCL1 medium (as described above) and monitored evoked EPSCs and IPSCs. We 469 examined AMPAR- and NMDAR-mediated EPSCs separately by recording them at 470 holding potentials of -70 mV and +40 mV, respectively. We quantified AMPAR-EPSCs at 471 peak amplitudes and NMDAR-EPSCs at 50 ms after the stimulus (Figure 8A). Consistent 472 with the elevations in synapse density and mEPSC frequency, SPARCL1 significantly 473 increased the amplitude of AMPAR-EPSCs (~50%) in both control and neuroligin- 474 deficient neurons (Figure 8B). Strikingly, SPARCL1 increased the amplitude of NMDAR- 475 EPSCs to an even larger extent (>200%) in both wildtype and neuroligin-deficient neurons 476 (Figure 8B). In line with this observation, SPARCL1 also enhanced the NMDAR/AMPAR 477 ratio (~100%; Figure 8B). These data show that SPARCL1 not only promotes synapse 478 formation, but also changes the receptor composition of synapses by increasing their 479 NMDAR responses independent of neuroligins. Notably, the deletion of neuroligins 480 produced a significant decrease both in AMPAR-EPSCs (~40%) and in NMDAR-EPSCs 481 (~50%); this decrease was not affected by SPARCL1.

482 Finally, we recorded evoked IPSCs as a function of SPARCL1 treatments and neuroligin 483 expression (Figure 8C, 8D). SPARCL1 did not alter the amplitude of IPSCs, consistent 484 with its lack of effect on inhibitory synapse numbers and spontaneous activity as 485 described above. However, the neuroligin deletion suppressed inhibitory synaptic 486 transmission (~50%), consistent with a major function of neuroligins in both excitatory and 487 inhibitory synapses (Figure 8C, 8D). Taken together, these results show that neuroligins 488 are not required for SPARCL1 to selectively enhance evoked neurotransmission at 489 excitatory synapses.

490 491 DISCUSSION 492 Emerging evidence supports roles for secreted ECM proteins in synapse formation, 493 synapse maturation, and synaptic plasticity (Ferrer-Ferrer and Dityatev, 2018; Yuzaki,

494 2018). SPARCL1 is a secreted, ubiquitously expressed protein that is highly synaptogenic 495 when added to cultured neurons, inducing the formation of synapses that have been 496 described as largely inactive (Kucukdereli et al., 2011) or active (Gan and Südhof, 2019). 497 Moreover, SPARCL1 is abundant in the blood of young mice but declines precipitously in 498 blood with age, suggesting a possible role as a systemic aging factor (Gan and Südhof, 499 2019). In the present study, we aimed to explore the mechanism by which SPARCL1 500 boosts synapse formation, guided by the exciting discovery that SPARCL1 may catalyze 501 the formation of trans-synaptic neurexin/neuroligin complexes (Singh et al., 2016). 502 Specifically, three questions arose from previous studies: (1) Does SPARCL1 act on all 503 synapses, or is it selective for a subset of synapses? (2) Does SPARCL1 induce 504 functional or largely inactive synapses? (3) Does SPARCL1 boost synapse formation by 505 binding to neurexins and/or neuroligins, or is a different, presently unknown receptor 506 responsible for its powerful synaptogenic effect?

507 We addressed these questions using mixed cultures of glia and neurons prepared from 508 the cortex of newborn mice, and we employed rigorous genetic manipulations to examine 509 the role of neurexins and neuroligins in SPARCL1-induced synaptogenesis. Our 510 observations enable the following conclusions. First, SPARCL1 specifically boosts the 511 formation of excitatory but not inhibitory synapses without changing neuronal size and 512 dendritic arborization (Figures 2, 4, and 6); thus, SPARCL1 acts on a subset of synapses 513 without affecting general neuronal maturation. Second, SPARCL1 strengthens excitatory 514 synapses more than it elevates excitatory synapse density; thus, SPARCL1 either 515 induces new, highly efficacious synapses or strengthens existing synapses (Figures 5, 7, 516 and 8). Third, at excitatory synapses, a brief SPARCL1 treatment increases the 517 magnitude of NMDAR-mediated responses to a larger extent than AMPAR-mediated 518 responses, thereby elevating the NMDAR/AMPAR EPSC ratio by ~50% (Figure 8). 519 Fourth, SPARCL1 is secreted from glia at only low levels, such that endogenous 520 SPARCL1 is insufficient to stimulate synapse formation, and that exogenous – possibly 521 blood-derived, systemic SPARCL1 (Gan and Südhof, 2019) –boosts synapse formation. 522 Fifth, deletion of neurexins (except for the minor variant Nrxn1J) or of neuroligins does 523 not alter any of the SPARCL1-induced effects on neurons (Figures 2,4-7), and thus 524 SPARCL1 does not act by binding to neurexins and neuroligins as hypothesized (Singh

525 et al., 2016). Sixth, deletion of neurexins or of neuroligins has no effect on either excitatory 526 or inhibitory synapse numbers in cultured neurons, and thus neither neurexins nor 527 neuroligins are essential for synapse formation or maintenance (Figures 4 and 6). 528 Seventh, deletion of neurexins decreases the frequencies of mEPSC and mIPSCs 529 (~50%) without affecting their amplitudes (Figure 5), consistent with previous reports that 530 the neurexin deletion impairs synapse organization (Missler et al., 2001; Chen et al., 531 2017; Luo et al., 2020). Eighth, finally, deletion of neuroligins also suppressed the 532 frequencies of mEPSCs and mIPSCs (~50%), but additionally decreased their amplitudes 533 (~30% and ~45%, respectively; Figure 7). Thus, the neuroligin deletion impaired synaptic 534 transmission more severely than did the neurexin deletion, suggesting that neuroligins 535 may bind to other trans-synaptic ligands besides neurexins.

536 Our experiments generate a new, overall picture of SPARCL1 function that answers the 537 three questions posed above: SPARCL1 not only selectively induces functional excitatory 538 synapses but increases their efficacy and enhances their NMDAR content more than their 539 AMPAR content. Moreover, SPARCL1 clearly acts independent of neurexins and 540 neuroligins. The specificity of SPARCL1 for excitatory synapse function is unexpected. 541 SPARCL1 likely acts by binding to a high-affinity receptor – as yet unidentified – that 542 induces an intracellular signal for synapse formation and synaptic organization, which is 543 specific for excitatory neurons.

544 Our results are consistent with earlier results documenting the potent synaptogenic action 545 of SPARCL1 (Kucukdereli et al., 2011; Gan and Südhof, 2019). However, our results are 546 at odds with two previous conclusions about SPARCL1, namely that SPARCL1 induces 547 non-functional synapses, and that SPARCL1 enhances synapse formation by bridging 548 presynaptic Nrxn1α and postsynaptic Nlgn1B (Kucukdereli et al., 2011; Singh et al., 549 2016). In assessing these apparent contradictions, it should be recalled that the previous 550 studies were rather preliminary. These studies suggested exciting hypotheses, but 551 performed few direct tests of synapse function, did not directly demonstrate binding of 552 SPARCL1 to neurexins or neuroligins, and did not show that deletions of neurexins or 553 neuroligins ablate SPARCL1 activity. Here, we demonstrate using rigorous 554 electrophysiological analyses that SPARCL1 is even more effective than envisioned 555 earlier in promoting synapse function since it not only boosted synapse numbers, but also

556 enhanced synaptic strength and increased NMDAR signaling. Thus, SPARCL1 promotes 557 formation of highly functional synapses. Moreover, we find that SPARCL1 is poorly 558 secreted from glia, whereas exogenous SPARCL1 potently enhances synapses in the 559 presence of glia. This observation suggests that glial SPARCL1 does not contribute 560 significantly to this effect. Finally, our conclusion that SPARCL1 acts independent of 561 neurexins and neuroligins is based on solid genetic manipulations and is consistent with 562 several previous studies. Specifically, biophysical assays failed to detect SPARCL1 563 binding to neuroligins (Eleghart et al., 2017). Furthermore, neurexins and neuroligins are 564 thought to function in both excitatory and inhibitory synapses (Missler et al., 2001; Chen 565 et al., 2017; Varoqueaux et al., 2006; Chubykin et al., 2007), yet we found that SPARCL1 566 acts only on excitatory synapses. The use of transient shRNA knockdowns in Singh et 567 al., 2016 versus conditional knockouts in the present study may account for these 568 discrepancies.

569 Our results also contradict some previous studies on neurexins and neuroligins, but again 570 are consistent with others. Specifically, many studies using shRNAs (or microRNAs 571 operating by the same mechanism) often detect massive decreases in synapse numbers 572 when a neurexin or neuroligin is targeted (e.g., see Chih et al., 2004 Science; Li et al., 573 2017; Roppongi et al., 2020; Kwon et al., 2012; Bemben et al., 2014). Yet, genetic 574 deletions failed to reveal a decrease in excitatory synapse numbers upon knockout of 575 neurexins or neuroligins, even when all neurexins or neuroligins were targeted (e.g., 576 Missler et al., 2003; Chen et al., 2017; Varoqueaux et al., 2006; Chanda et al., 2017). Our 577 current results confirm these latter reports. The unchanged synapse numbers observed 578 upon genetic deletions are not due to unspecified compensatory effects. Like RNAi- 579 mediated deletions, conditional mutations also operate on an acute time frame but with 580 higher specificity, suggesting that the phenotypes of RNAi-mediated neurexin or 581 neuroligin mutations reflect off-target effects.

582 How might neuronal transmembrane proteins other than neurexins and neuroligins 583 function as SPARCL1 receptors that mediate excitatory synapse formation and NMDAR 584 recruitment? In one scenario, SPARCL1 might stabilize interactions between diffusible 585 secreted factors and their synaptic receptors. SPARCL1 is reported to activate Wnt 586 signaling by direct interaction with various Frizzled (Fz) receptors and lipoprotein-related

587 proteins, thereby stabilizing Wnt-receptor complexes (Zhao et al., 2018). Wnt7a and its 588 receptor Fz5 promote synapse formation in hippocampal neurons by stimulating 589 recruitment of synaptic vesicles and active zone proteins (Sahores et al., 2010). 590 Furthermore, Wnt5a binds to multiple Fz receptors to specifically upregulate synaptic 591 NMDAR currents in hippocampal slices (Cerpa et al., 2011). Thus, SPARCL1-induced 592 activation of Wnt signaling may underlie excitatory synapse formation and NMDAR 593 recruitment. Another possibility is that SPARCL1 might bind and activate extracellular 594 matrix receptors such as integrins. The follistatin-like domain of SPARC, which is also 595 present in SPARCL1, interacts with β-integrins at developing synapses (Jones et al., 596 2011). Activated β-integrins trigger rapid phosphorylation of Src family kinases, which in 597 turn increases tyrosine phosphorylation of GluN2A and GluN2B NMDAR subunits and 598 enhances NMDAR activity (Park and Goda, 2016). Thus, SPARCL1 could potentially 599 enhance synaptic NMDAR responses via binding to β-integrins. Additional studies will be 600 required to test novel synaptic receptors for SPARCL1 and elucidate downstream 601 signaling mechanisms that regulate synapse formation and specification.

602 603 ACKNOWLEDGEMENTS 604 K.J.G. was funded by a postdoctoral research fellowship from the Canadian Institutes of 605 Health Research (#MFE-141209). This study was supported by grants from the NIMH 606 (MH052804 and MH092931 to T.C.S.).

607 608 AUTHOR CONTRIBUTIONS 609 K.J.G. and T.C.S. designed the study, K.J.G. performed all experiments, and K.J.G. and 610 T.C.S. analyzed the data and wrote the manuscript. 611

612 REFERENCES 613 Allen NJ, Bennett ML, Foo LC, Wang GX, Chakroborty C, Smith SJ, Barres BA (2012) Astrocyte 614 glypicans 4 and 6 promote formation of excitatory synapses via GluA1 AMPA receptors. Nature 615 486:410-4. 616 Bemben MA, Shipman SL, Hirai T, Herring BE, Li Y, Badger JD 2nd, Nicoll RA, Diamond JS, 617 Roche KW (2014) CaMKII phosphorylation of neuroligin-1 regulates excitatory synapses. Nat 618 Neurosci 17:56-64. 619 Boucard AA, Chubykin AA, Comoletti D, Taylor P, Südhof TC (2005) A splice code for trans- 620 synaptic cell adhesion mediated by binding of neuroligin 1 to alpha- and beta-neurexins. Neuron 621 48:229-36. 622 Brekken RA, Sage EH (2001) SPARC, a matricellular protein: at the crossroads of cell-matrix 623 communication. Matrix Biol 19:816-27. 624 Brekken RA, Sage EH (2000) SPARC, a matricellular protein: at the crossroads of cell-matrix. 625 Matrix Biol 19:569-80. 626 Cao X, Tabuchi K (2017) Functions of synapse adhesion molecules neurexin/neuroligins and 627 neurodevelopmental disorders. Neurosci Res 116:3-9. 628 Cerpa W, Gambrill A, Inestrosa NC, Barria A (2011) Regulation of NMDA-receptor synaptic 629 transmission by Wnt signaling. J Neurosci 31:9466-71. 630 Chanda S, Hale WD, Zhang B, Wernig M, Südhof TC (2017) Unique versus redundant functions 631 of neuroligin genes in shaping excitatory and inhibitory synapse properties. J Neurosci 37:6816- 632 6836. 633 Chen LY, Jiang M, Zhang B, Gokce O, Südhof TC (2017) Conditional Deletion of All Neurexins 634 Defines Diversity of Essential Synaptic Organizer Functions for Neurexins. Neuron 94:611-625. 635 Chih B, Gollan L, Scheiffele P (2006) Alternative splicing controls selective trans-synaptic 636 interactions of the neuroligin-neurexin complex. Neuron 51:171-8. 637 Christopherson KS, Ullian EM, Stokes CC, Mullowney CE, Hell JW, Agah A, Lawler J, Mosher 638 DF, Bornstein P, Barres BA (2005) Thrombospondins are astrocyte-secreted proteins that 639 promote CNS synaptogenesis. Cell 120:421-33. 640 Chubykin AA, Atasoy D, Etherton MR, Brose N, Kavalali ET, Gibson, JR, and Südhof, TC (2007) 641 Activity-Dependent Validation of Excitatory vs. Inhibitory Synapses by Neuroligin-1 vs. Neuroligin- 642 2. Neuron 54:919-931. 643 Connor SA, Elegheert J, Xie Y, Craig AM (2019) Pumping the brakes: suppression of synapse 644 development by MDGA-neuroligin interactions. Curr Opin Neurobiol 57:71-80. 645 Elegheert J, Cvetkovska V, Clayton AJ, Heroven C, Vennekens KM, Smukowski SN, Regan MC, 646 Jia W, Smith AC, Furukawa H, Savas JN, de Wit J, Begbie J, Craig AM, Aricescu AR (2017) 647 Structural Mechanism for Modulation of Synaptic Neuroligin-Neurexin Signaling by MDGA 648 Proteins. 95:896-913. 649 Ferrer-Ferrer M, Dityatev A (2018) Shaping Synapses by the Neural Extracellular Matrix. Front 650 Neuroanat 12:40. 651 Gan KJ and Südhof, TC (2019) Specific factors in blood from young but not old mice directly 652 promote synapse formation and NMDA-receptor recruitment. PNAS 116:12524-12533.

653 Girard JP, Springer TA (1995) Cloning from purified high endothelial venule cells of hevin, a close 654 relative of the antiadhesive extracellular matrix protein SPARC. Immunity 2:113-23. 655 Girard JP, Springer TA (1996) Modulation of endothelial cell adhesion by Hevin, an acidic protein 656 associated with high endothelial venules. J Biol Chem 271:4511-7. 657 Hoon M, Soykan T, Falkenburger B, Hammer M, Patrizi A, Schmidt K, Sassoe-Pognetto M, Lowel 658 S, Moser T, Taschenberger H, Brose N, Varoqueaux F (2011) Neuroligin-4 is localized to 659 glycinergic postsynapses and regulates inhibition in the retina. PNAS 108:3053-3058. 660 Jiang M, Polepalli J, Chen LY, Zhang B, Südhof TC, Malenka RC (2017) Conditional ablation of 661 neuroligin-1 in CA1 pyramidal neurons blocks LTP by a cell-autonomous NMDA receptor- 662 independent mechanism. Mol Psychiatry 22:375-383. 663 Jones EV, Bernardinelli Y, Tse YC, Chierzi S, Wong TP, Murai KK (2011) Astrocytes control 664 glutamate receptor levels at developing synapses through SPARC-beta-integrin interactions. J 665 Neurosci 31:4154-65. 666 Kaeser PS, Deng L, Wang Y, Dulubova I, Liu X, Rizo J, Südhof TC (2011) RIM proteins tether 667 Ca2+ channels to presynaptic active zones via a direct PDZ-domain interaction. Cell 144:282-95. 668 Katzman A, Alberini CM (2018) NLGN1 and NLGN2 in the prefrontal cortex: their role in memory 669 consolidation and strengthening. Curr Opin Neurobiol 48:122-130. 670 Khajal AJ, Sterky FH, Sclip A, Schwenk J, Brunger AT, Fakler B, Sudhof TC (2020) Deorphanizing 671 FAM19A proteins as pan-neurexin ligands with an unusual biosynthetic binding mechanism. J 672 Cell Biol, in press. 673 Kucukdereli H, Allen NJ, Lee AT, Feng A, Ozlu MI, Conatser LM, Chakraborty C, Workman G, 674 Weaver M, Sage EH, Barres BA, Eroglu C (2011) Control of excitatory CNS synaptogenesis by 675 astrocyte-secreted proteins Hevin and SPARC. PNAS 108:E440-9. 676 Kwinter DM, Lo K, Mafi P, Silverman MA (2009) Dynactin regulates bidirectional transport of 677 dense-core vesicles in the axon and dendrites of cultured hippocampal neurons. Neuroscience 678 162:1001-10. 679 Kwon HB, Kozorovitskiy Y, Oh WJ, Peixoto RT, Akhtar N, Saulnier JL, Gu C, Sabatini BL (2012) 680 Neuroligin-1-dependent competition regulates cortical synaptogenesis and synapse number. Nat 681 Neurosci 15:1667-74. 682 Li J, Han W, Pelkey KA, Duan J, Mao X, Wang YX, Craig MT, Dong L, Petralia 683 RS, McBain CJ, Lu W (2017) Molecular Dissection of Neuroligin 2 and Slitrk3 Reveals an 684 Essential Framework for GABAergic Synapse Development. Neuron 96:808-826. 685 Lively S, Brown IR (2008a) The extracellular matrix protein SC1/hevin localizes to excitatory 686 synapses following status epilepticus in the rat lithium-pilocarpine seizure model. J Neurosci Res 687 86:2895-905. 688 Lively S, Brown IR (2008b) Localization of the extracellular matrix protein SC1 coincides with 689 synaptogenesis during rat postnatal development. Neurochem Res 33:1692-700. 690 Lively S, Ringuette MJ, Brown IR (2007) Localization of the extracellular matrix protein SC1 to 691 synapses in the adult rat brain. Neurochem Res 32:65-71. 692 Luo F, Sclip A, Jiang M, and Südhof TC (2020) Neurexins Cluster Ca2+ Channels within 693 presynaptic Active Zone. EMBO J. in press. 694 Maximov A, Pang ZP, Tervo DG, Südhof TC (2007) Monitoring synaptic transmission in primary 695 neuronal cultures using local extracellular stimulation. J Neurosci Methods 161:75-87.

696 Missler M, Zhang W, Rohlmann A, Kattenstroth G, Hammer RE, Gottmann K, Südhof TC (2003) 697 Alpha-neurexins couple Ca2+ channels to synaptic vesicle exocytosis. Nature 423:939-48. 698 Park YK, Goda Y (2016) Integrins in synapse regulation. Nat Rev Neurosci 17:745-756. 699 Roppongi RT, Dhume SH, Padmanabhan N, Silwal P, Zahra N, Karimi B, Bomkamp C, Patil 700 CS, Champagne-Jorgensen K, Twilley RE, Zhang P, Jackson MF, Siddiqui TJ (2020) LRRTMs 701 Organize Synapses through Differential Engagement of Neurexin and PTPσ. Neuron pii: S0896- 702 6273(20)30003-9. 703 Sahores M, Gibb A, Salinas PC (2010) Frizzled-5, a receptor for the synaptic organizer Wnt7a, 704 regulates activity-mediated synaptogenesis. Development 137:2215-25. 705 Sando R, Jiang X, Südhof TC (2019) Latrophilin GPCRs direct synapse specificity by coincident 706 binding of FLRTs and teneurins. Science 363(6429). 707 Saunders A, Macosko E, Wysoker A, Goldman M, Krienen F, Rivera H. de, Bien E, Baum M, 708 Wang S, Goeva A, et al. (2018) A Single-Cell Atlas of Cell Types, States, and Other 709 Transcriptional Patterns from Nine Regions of the Adult Mouse Brain. Cell 174:1015-1030. 710 Singh SK, et al. (2016) Astrocytes Assemble Thalamocortical Synapses by Bridging NRX1α and 711 NL1 via Hevin. Cell 164:183-196. 712 Sterky FH, Trotter JH, Lee SJ, Recktenwald CV, Du X, Zhou B, Zhou P, Schwenk J, Fakler B, 713 Südhof, TC (2017) Carbonic anhydrase-related protein CA10 is an evolutionarily conserved pan- 714 neurexin ligand. PNAS 114:E1253-1262. 715 Südhof TC (2018) Towards an Understanding of Synapse Formation. Neuron 100:276-293. 716 Südhof TC (2017) Synaptic Neurexin Complexes: A Molecular Code for the Logic of Neural 717 Circuits. Cell 171:745-769. 718 The Tabula Muris Consortium (2018) Single-cell transcriptomics of 20 mouse organs creates a 719 Tabula Muris. Nature 562, 367–372. 720 Varoqueaux F, Aramuni G, Rawson RL, Mohrmann R, Missler M, Gottmann K, Zhang W, Südhof 721 TC, Brose N (2006) Neuroligins determine synapse maturation and function. Neuron 51:741-54. 722 Wang R, Dong JX, Wang L, Dong XY, Anenberg E, Jiang PF, Zeng LH, Xie YC (2019) A negative 723 regulator of synaptic development: MDGA and its links to neurodevelopmental disorders. World 724 J Pediatr 15:415-421. 725 Weaver MS, Workman G, Cardo-Vila M, Arap W, Pasqualini R, Sage EH (2010) Processing of 726 the matricellular protein hevin in mouse brain is dependent on ADAMTS4. J Biol Chem 285:5868- 727 77. 728 Wu X, Morishita WK, Riley AM, Hale WD, Südhof TC, Malenka RC (2019) Neuroligin-1 Signaling 729 Controls LTP and NMDA Receptors by Distinct Molecular Pathways. Neuron 102:621-635. 730 Yuzaki M (2018) Two Classes of Secreted Synaptic Organizers in the Central Nervous System. 731 Annu Rev Physiol 80:243-262. 732 Zeisel, A., Hochgerner, H., Lönnerberg, P., Johnsson, A., Memic, F., van der Zwan, J., Häring, 733 M., Braun, E., Borm, L.E., La Manno, G., et al. (2018) Molecular Architecture of the Mouse 734 Nervous System. Cell 174, 999–1014. 735 Zhang Y, Pak C, Han Y, Ahlenius H, Zhang Z, Chanda S, Marro S, Patzke C, Acuna C, Covery 736 J, Xu W, Yang N, Danko T, Chen L, Wernig M, Südhof TC (2013) Rapid single-step induction of 737 functional neurons from human pluripotent stem cells. Neuron 78:785-98.

738 Zhao SJ, et al. (2018) SPARCL1 suppresses osteosarcoma metastasis and recruits macrophages 739 by activation of canonical WNT/β-catenin signaling through stabilization of the WNT-receptor 740 complex. Oncogene 37:1049-1061.

741 TABLE 1. Summary of statistical analyses 742 Sample Degree of freedom Figure Test used size and p-value 1B Unpaired t-test 3 T(4) = 0.2138, p = 0.8412 1C One-way ANOVA 3 F(4,10) = 829.3, p < 0.0001 1D One-way ANOVA 3 F(2,6) = 99.09, p < 0.0001 2C (VGlut density) Unpaired t-test 29 T(56) = 5.467, p < 0.0001 2C (Homer density) Unpaired t-test 29 T(56) = 5.778, p < 0.0001 2C (VGlut size) Unpaired t-test 29 T(56) = 0.4105, p = 0.683 2C (Homer size) Unpaired t-test 29 T(56) = 0.6777, p = 0.508 2E (VGAT density) Unpaired t-test 29 T(56) = 0.03323, p = 0.9736 2E (Gephyrin density) Unpaired t-test 29 T(56) = 0.2890, p = 0.7737 2E (VGAT size) Unpaired t-test 29 T(56) = 0.3999, p = 0.2708 2E (Gephyrin size) Unpaired t-test 29 T(56) = 0.0330, p = 0.1702 2G Unpaired t-test 21 T(40) = 0.3137, p = 0.7553 2H (Primary dendrites) Unpaired t-test 21 T(40) = 0.1163, p = 0.908 2H (Dendritic branching) Unpaired t-test 21 T(40) = 0.09528, p = 0.9246 2I Unpaired t-test 21 T(40) = 0.06135, p = 0.9514 2J (Capacitance) Unpaired t-test 21 T(40) = 0.3649, p = 0.7171 2J (Input resistance) Unpaired t-test 21 T(40) = 0.7329 , p = 0.45 3B Unpaired t-test 3 T(4) = 12.46, p = 0.0002 3C (Nlgn1) Unpaired t-test 3 T(4) = 15.68, p < 0.0001 3C (Nlgn2) Unpaired t-test 3 T(4) = 44.02, p < 0.0001 3C (Nlgn3) Unpaired t-test 3 T(4) = 51.86, p<0.0001 4B (Density) One-way ANOVA 24 F(3,92) = 5.122, p = 0.0025 4B (Size) One-way ANOVA 24 F(3,92) = 3.167, p = 0.0281 4B (Intensity) One-way ANOVA 24 F(3,92) = 2.005, p = 0.1188 4C (Density) One-way ANOVA 24 F(3,92) = 3.777, p = 0.0132 4C (Size) One-way ANOVA 24 F(3,92) = 0.1854, p = 0.9061 4C (Intensity) One-way ANOVA 24 F(3,92) = 0.9585, p = 0.4159 5A (Frequency) One-way ANOVA 18 F(3,68) = 31.66, p < 0.0001 5A (Amplitude) One-way ANOVA 18 F(3,68) = 3.063, p = 0.0338 5B (Frequency) One-way ANOVA 17 F(3,64) = 15.74, p<0.0001 5B (Amplitude) One-way ANOVA 17 F(3.64) = 1.865, p = 0.1443 6B (Density) One-way ANOVA 24 F(3,92) = 9.735, p < 0.0001 6B (Size) One-way ANOVA 24 F(3,92) = 5.245, p = 0.0022 6B (Intensity) One-way ANOVA 24 F(3,92) = 5.799, p = 0.0011 6C (Density) One-way ANOVA 24 F(3,92) = 1.185, p = 0.32 6C (Size) One-way ANOVA 24 F(3,92) = 3.829, p = 0.0124 6C (Intensity) One-way ANOVA 24 F(3,92) = 3.692, p = 0.0147 7A (Frequency) One-way ANOVA 18 F(3,68) = 57.57, p < 0.0001 7A (Amplitude) One-way ANOVA 18 F(3,68) = 21.08, p < 0.0001 7B (Frequency) One-way ANOVA 18 F(3,68) = 17.53, p <0.0001 7B (Amplitude) One-way ANOVA 18 F(3, 68) = 16.44, p < 0.0001 8B (AMPAR amplitude) One-way ANOVA 24 F(3,92) = 38.35, p < 0.0001 8B (NMDAR amplitude) One-way ANOVA 24 F(3,92) = 119, p < 0.0001 8B (NMDAR/AMPAR) One-way ANOVA 24 F(3,92) = 24.68, p < 0.0001 8D One-way ANOVA 24 F(3,92) = 10.82 , p < 0.0001

743 744 This table summarizes the statistical tests, degrees of freedom and significance for 745 each figure. Tukey’s test of multiple comparisons was performed post-hoc for all one- 746 way ANOVAs (see Extended Data, Figures 1-1 to 8-1). 747 748

27 749 FIGURES and FIGURE LEGENDS

750 751 Figure 1: Endogenous SPARCL1 is secreted at low levels by cultured mouse glia 752 but is abundant in mouse serum.

753 (A) Design of intron-spanning primers for semiquantitative RT-PCR. Primers against SPARCL1 754 and β-actin (control) genes were situated in exons to distinguish between gDNA and mRNA 755 amplification (left). Expected amplicon sizes are indicated (right). 756 (B) Most cortical SPARCL1 mRNA is expressed by glia. RT-PCR was performed on total RNA 757 isolated from glia cultured alone or from glia cocultured with cortical neurons at 14 DIV. 758 Amplification of SPARCL1 and β-actin transcripts was optimized to 25 PCR cycles to ensure that 759 quantification was performed prior to saturation. SPARCL1 and β-actin transcripts from both 760 sources were amplified either individually or in multiplexed format to estimate the relative 761 abundance of SPARCL1 cDNA by normalization against β-actin. Analysis by electrophoresis 762 shows specific amplification of SPARCL1 and β-actin cDNA without gDNA contamination. 763 SPARCL1 mRNA is expressed at similar levels by pure glia and by neurons cocultured with glia, 764 indicating that most cortical SPARCL1 is expressed by glia. SPARCL1 band intensities are 765 normalized to β-actin band intensities within multiplexed RT-PCR reactions. 766 (C) Endogenous SPARCL1 protein is secreted at low levels in primary cortical cultures. Cortical 767 glia were cultured alone (G) or with cortical neurons (N+G) for 14 DIV (left: Coomassie stain 768 showing the total protein composition of conditioned media and cell lysates taken from both types 769 of cultures and from mouse serum; right: Immunoblot of SPARCL1 (green) from conditioned 770 media and cell lysates). Full-length SPARCL1 is secreted and present intracellularly (green 771 arrow). Secreted SPARCL1 is proteolyzed into an array of fragments (light green arrows). 772 Truncated SPARCL1 is present in serum (dark green arrow) but is undetectable in primary cortical 773 cultures. 774 (D) Comparison of recombinant and endogenous SPARCL1 protein levels. Equal volumes of 775 HEK293T cell supernatant containing native recombinant SPARCL1 and of conditioned medium 776 harvested from primary cortical cultures before SPARCL1 treatment (13 DIV) and 24 h after 777 SPARCL1 treatment were analyzed by SDS-PAGE and immunoblotting. Left: Coomassie stain 778 showing the total protein composition of the HEK cell supernatant and of the conditioned medium 779 before (“untreated”) and after (“treated”) SPARCL1 treatment. Right: immunoblot of recombinant 780 SPARCL1 protein secreted by HEK cells and of SPARCL1 present in the conditioned medium 781 before and after treatment. Full-length and proteolyzed SPARCL1 are indicated by the dark grey 782 and light grey arrows, respectively. 783 Bar graphs show means ± SEM; 3 independent cultures were analyzed. Statistical significance 784 was evaluated by a Student’s T-test. Non-significant relations (n.s.) are indicated. See Figure 1- 785 1 (Extended Data) for complete statistical analyses. 786

787 788 Figure 2: Recombinant SPARCL1 selectively boosts excitatory but not inhibitory 789 synapse numbers in primary cultures of cortical neurons and glia 790 (A) Experimental strategy. Cortical neurons and glia were cultured from P0 wildtype, triple 791 neurexin (Nrxn123), or quadruple neuroligin (Nlgn1234) conditional knockout mice. At DIV3, 792 neurons were infected with lentiviruses expressing nuclear Cre-recombinase fused to EGFP and

793 driven by the synapsin promoter (Syn-Cre-EGFP), or nonfunctional mutant Cre (Syn-ΔCre-EGFP) 794 as a negative control. At DIV10, neurons were transfected with β-actin-eBFP (eBFP) for 795 morphological analyses. Neurons were treated with recombinant SPARCL1 at DIV13 and 796 analyzed by electrophysiology, immunocytochemistry, and immunoblotting at DIV14-16. In the 797 experiments described in the present figure analyzing wild-type neurons, cultures expressing 798 ΔCre were examined. In all subsequent figures describing experiments on conditional neurexin 799 and neuroligin mutants, cultures expressing ΔCre and Cre were investigated. 800 (B) Representative images showing that recombinant SPARCL1 treatment increases the 801 excitatory synapse density in wild-type neurons. Images depict dendritic segments of wildtype 802 neurons treated at DIV13 with the supernatants of HEK cells expressing either mClover (‘Control’) 803 or recombinant SPARCL1 (‘SPARCL1’). Neurons were immunostained for vGluT1 (excitatory 804 presynaptic marker), Homer (excitatory postsynaptic marker), and MAP2 (dendritic marker) at 805 DIV14. 806 (C) Quantifications showing that recombinant SPARCL1 increases the density (left), but not the 807 size of excitatory synapses (right). 808 (D) Representative images showing that recombinant SPARCL1 treatment does not alter 809 inhibitory synapse in wild-type neurons. Images depict dendritic segments of wildtype neurons 810 treated as described in B, and immunostained for vGAT (inhibitory presynaptic marker), gephyrin 811 (inhibitory postsynaptic marker), and MAP2 (dendritic marker) at DIV14. 812 (E) Quantifications showing that recombinant SPARCL1 does not alter the density (left) nor size 813 (right) of inhibitory synapses. 814 (F) Representative images showing that recombinant SPARCL1 treatment does not alter the 815 survival or dendritic morphology of wildtype neurons. Images depict neurons treated as described 816 in B, but additionally sparsely transfected with eBFP to visualize dendritic arborizations of 817 neurons. Neurons were counterstained with MAP2. 818 (G-I) Quantifications showing that recombinant SPARCL1 does not alter the neuronal cell density 819 (G), number of primary dendrites, and dendritic branches (H), or soma size of neurons (I). 820 (J) Electrophysiological measurements of the capacitance (left) and input resistance (right) of 821 wild-type neurons treated with SPARCL1 as described in B show that SPARCL1 has no significant 822 effect on these passive electrical membrane properties. 823 All bar graphs display means ± SEM; numbers of cells/independent cultures analyzed are shown 824 within bars. Statistical significance was evaluated by Student’s t-test; non-significant relations are 825 not indicated. See Figure 2-1 (Extended Data) for complete statistical analyses. 826

31 827 828 Figure 3: Cre-recombinase expression substantially reduces neurexin and 829 neuroligin protein expression in conditional knockout neurons 830 (A) Lentiviral infection of primary cortical neurons. Representative low-magnification images show 831 14 DIV neurons expressing ΔCre-EGFP or Cre-EGFP. Neurons were counterstained for MAP2 832 to visualize neuronal dendrites, respectively. All neuronal nuclei were positive for ΔCre-EGFP or 833 Cre-EGFP expression, yielding an infection efficiency of 100%. 834

835 (B) Analysis of neurexin protein expression in primary cortical neurons from Nrxn123 cKO mice. 836 Immunoblots were performed on lysates from cortical cultures infected with lentiviral ΔCre or Cre. 837 Left: representative blots; right: summary graph of neurexin protein levels. The pan-neurexin 838 antibody detects full-length (black arrow) and truncated neurexin. A non-specific band is also 839 observed (red arrow). The neurexin signals were normalized to corresponding β-actin signals for 840 quantification. 841 842 (C) Analysis of neuroligin protein expression in primary cortical neurons from Nlgn1234 cKO mice. 843 Immunoblots were performed on lysates from cortical cultures infected with lentiviral ΔCre or Cre. 844 Left: representative blots of Nlgn1, Nlgn2, and Nlgn3; right: summary graph of neurexin protein 845 levels. The Nlgn3 antibody detects Nlgn3 (black arrow) and two non-specific bands (red arrows). 846 The neuroligin signals were normalized to corresponding β-actin signals for quantification. 847 848 All bar graphs display means ± SEM; numbers of cells/independent cultures analyzed are shown 849 within bars. Statistical significance was evaluated by Student’s t-test; non-significant relations are 850 not indicated. See Figure 3-1 (Extended Data) for complete statistical analyses. 851 852

33 853 854 Figure 4: Deletion of neurexins in dissociated cultures of mouse neurons and glia 855 neither decreases excitatory or inhibitory synapse numbers nor impairs the 856 increase in excitatory synapse numbers produced by recombinant SPARCL1 857 (A) Deletion of all neurexins does not significantly decrease synapse numbers and does not block 858 excitatory synaptogenesis induced by recombinant SPARCL1. Representative images show triple 859 neurexin (Nrxn123) control (ΔCre) and conditional knockout (Cre) neurons treated at 13 DIV with 860 the supernatants of HEK cells expressing either mClover (‘Control’ medium) or recombinant

861 SPARCL1 (‘SPARCL1’). Neurons were immunostained for VGLUT1 (excitatory puncta), VGAT 862 (inhibitory puncta) and MAP2 at 14 DIV. Higher-magnification images (right) were taken from the 863 boxed areas shown in the corresponding lower-magnification images (left). 864 (B-C) Quantifications showing that recombinant SPARCL1 increases the density of excitatory 865 synapses (top summary graphs) but not inhibitory synapses (bottom summary graphs) in control 866 and Nrxn123 cKO neurons. SPARCL1 did not affect the size and staining intensity of excitatory 867 and inhibitory synapses. 868 All bar graphs display means ± SEM; numbers of cells/independent cultures analyzed are shown 869 within bars. Statistical significance (*p<0.05) was evaluated by one-way ANOVA with Tukey’s 870 post-hoc comparisons; non-significant relations are not indicated. See Figure 4-1 (Extended 871 Data) for complete statistical analyses.

872 873 874 Figure 5: Recombinant SPARCL1 increases spontaneous excitatory but not 875 inhibitory synaptic transmission independent of neurexins 876 (A) SPARCL1 increases mEPSC frequency but not amplitude independent of neurexins. Triple 877 neurexin (Nrxn123) control (ΔCre) and conditional knockout (Cre) neurons were treated at 13 DIV

878 with control medium or recombinant SPARCL1. Spontaneous excitatory synaptic activity was 879 assessed at 14 DIV by mEPSC recordings in the presence of TTX and PTX. Above: 880 representative traces; below: summary graphs of the mEPSC frequency and mEPSC amplitude. 881 (B) SPARCL1 does not affect spontaneous inhibitory synaptic activity. Nrxn123 control and cKO 882 neurons were treated at 13 DIV with control medium or recombinant SPARCL1. Spontaneous 883 inhibitory synaptic activity was assessed at 14 DIV by mIPSC recordings in the presence of TTX, 884 CNQX and D-AP5. Above: representative traces; below: summary graphs of the mIPSC 885 frequency and mIPSC amplitude. 886 All bar graphs are means ± SEM; numbers of cells/independent cultures analyzed are shown 887 beside the bars. Statistical significance (***p<0.001) was evaluated by one-way ANOVA with 888 Tukey’s post-hoc comparisons; non-significant relations are not indicated. See Figure 5-1 889 (Extended Data) for complete statistical analyses. 890 891 892 893 894

895 896 Figure 6: Deletion of neuroligins does not significantly decrease synapse numbers 897 and does not block excitatory synaptogenesis induced by recombinant SPARCL1 898 (A) Deletion of all neuroligins does not significantly decrease synapse numbers and does not 899 block excitatory synaptogenesis induced by recombinant SPARCL1. Representative images 900 show quadruple neuroligin (Nlgn1234) control (ΔCre) and conditional knockout (Cre) neurons 901 treated at 13 DIV with the supernatants of HEK cells expressing either mClover (‘Control’ medium) 902 or recombinant SPARCL1 (‘SPARCL1’). Neurons were immunostained for VGLUT1 (excitatory

903 puncta), VGAT (inhibitory puncta) and MAP2 at 14 DIV. Higher-magnification images (right) were 904 taken from the boxed areas shown in the corresponding lower-magnification images (left). 905 (B-C) Quantifications showing that recombinant SPARCL1 increases the density of excitatory 906 synapses (top summary graphs) but not inhibitory synapses (bottom summary graphs) in control 907 and Nlgn1234 cKO neurons. SPARCL1 did not substantially affect the size and staining intensity 908 of excitatory and inhibitory synapses. 909 All bar graphs display means ± SEM; numbers of cells/independent cultures analyzed are shown 910 within bars. Statistical significance (*p<0.05) was evaluated by one-way ANOVA with Tukey’s 911 post-hoc comparisons; non-significant relations are not indicated. See Figure 6-1 (Extended 912 Data) for complete statistical analyses.

913 914 Figure 7: Recombinant SPARCL1 increases spontaneous excitatory but not 915 inhibitory synaptic transmission independent of neuroligins. 916 (A) Deletion of all neuroligins significantly reduces mEPSC frequency and amplitude but does not 917 prevent SPARCL1 from increasing mEPSC frequency. Quadruple neuroligin (Nlgn1234) control 918 (ΔCre) and conditional knockout (Cre) neurons were treated at 13 DIV with control medium or 919 recombinant SPARCL1. Spontaneous excitatory synaptic activity was assessed at 14 DIV by

920 mEPSC recordings in the presence of TTX and PTX. Above: representative traces; below: 921 summary graphs of the mEPSC frequency and mEPSC amplitude. 922 (B) Deletion of all neuroligins significantly reduces mIPSC frequency and amplitude, which are 923 unaltered by SPARCL1. Nlgn1234 control and cKO neurons were treated at 13 DIV with control 924 medium or recombinant SPARCL1. Spontaneous inhibitory synaptic activity was assessed at 14 925 DIV by mIPSC recordings in the presence of TTX, CNQX and D-AP5. Above: representative 926 traces; below: summary graphs of the mIPSC frequency and mIPSC amplitude. 927 All bar graphs are means ± SEM; numbers of cells/independent cultures analyzed are shown 928 beside the bars. Statistical significance (***p<0.001) was evaluated by one-way ANOVA with 929 Tukey’s post-hoc comparisons; non-significant relations are not indicated. See Figure 7-1 930 (Extended Data) for complete statistical analyses. 931

932 933 Figure 8: SPARCL1 selectively enhances excitatory evoked neurotransmission and 934 increases NMDAR-mediated synaptic responses independent of neuroligins 935 (A) Deletion of all neuroligins reduces the amplitudes of evoked AMPAR- and NMDAR-EPSCs. 936 Treatment with recombinant SPARCL1 increases these parameters and particularly enhances 937 the NMDAR/AMPAR ratio independent of neuroligins. Nlgn1234 control and cKO neurons were 938 treated at 13 DIV with control medium or recombinant SPARCL1. Neurons were analyzed at 14- 939 16 DIV by recording EPSCs evoked by extracellular stimulation. AMPAR-EPSCs and NMDAR- 940 EPSCs were pharmacologically isolated with PTX and monitored at -70 mV and +40 mV holding 941 potentials, respectively. Representative traces of evoked EPSCs are shown for all conditions, 942 (B) Summary graphs of evoked AMPAR-EPSC amplitude, NMDAR-EPSC amplitude, and 943 NMDAR/AMPAR ratio. NMDAR current amplitudes were measured 50 ms after stimulation.

944 (C) Deletion of all neuroligins reduces the amplitudes of evoked GABAR-IPSCs, whereas addition 945 of SPARCL1 does not alter these currents. Nlgn1234 control and cKO neurons were treated at 946 13 DIV with control medium or recombinant SPARCL1. Neurons were analyzed at 14-16 DIV by 947 recording IPSCs evoked by extracellular stimulation. GABAR-IPSCs were pharmacologically 948 isolated with D-AP5 and CNQX and monitored at -70 mV. Representative traces of evoked IPSCs 949 are shown for all conditions. 950 (D) Summary graph of evoked GABAR-IPSC amplitude. 951 All bar and line graphs show means ± SEM; numbers of cells/independent cultures analyzed are 952 shown within the bars. Statistical significance (*p<0.05; **p<0.01; ***p<0.001) was evaluated by 953 one-way ANOVAs and Tukey’s post-hoc comparisons. Non-significant relations are not indicated. 954 See Figure 8-1 (Extended Data) for complete statistical analyses. 955 956 957 958 959 960 961 962

963 EXTENDED DATA FIGURE LEGENDS 964 Figure 1-1: Complete statistical analyses for “Figure 1: Endogenous SPARCL1 is 965 secreted at low levels by cultured mouse glia but is abundant in mouse serum” 966 (1B) Most cortical SPARCL1 mRNA is expressed by glia. RT-PCR was performed on total RNA 967 isolated from glia cultured alone or from glia cocultured with cortical neurons at 14 DIV. SPARCL1 968 band intensities are normalized to β-actin band intensities within multiplexed RT-PCR reactions. 969 Primary data comprise means ± SEM; 3 independent cultures were analyzed. Statistical 970 significance was evaluated by an unpaired T-test. 971 (1C) Endogenous SPARCL1 protein is secreted at low levels in primary cortical cultures. Cortical 972 glia were cultured alone (G) or with cortical neurons (N+G) for 14 DIV. Full-length SPARCL1 973 protein expression was detected by immunoblot and quantified. Primary data comprise means ± 974 SEM; 3 independent cultures were analyzed. Statistical significance was evaluated by a one-way 975 ANOVA with Tukey’s post-hoc test of multiple comparisons. 976 (1D) Comparison of recombinant and endogenous SPARCL1 protein levels. Full-length 977 SPARCL1 protein expression was detected by immunoblot and quantified. Primary data comprise 978 means ± SEM; 3 independent cultures were analyzed. Statistical significance was evaluated by 979 a one-way ANOVA with Tukey’s post-hoc test of multiple comparisons. 980 Test parameters highlighted in blue were used to determine statistical significance and are 981 summarized in Table 1. P-values and significance levels for individual comparisons between 982 specific groups are highlighted in yellow. 983 984 Figure 2-1: Complete statistical analyses for “Figure 2: Recombinant SPARCL1 985 selectively boosts excitatory but not inhibitory synapse numbers in primary 986 cultures of cortical neurons and glia” 987 (2C): Quantifications showing that recombinant SPARCL1 increases the density but not the size 988 of excitatory synapses. Primary data comprise means ± SEM; 29 cells from 3 independent 989 cultures were analyzed. Statistical significance was evaluated by an unpaired T-test. 990 (2E): Quantifications showing that recombinant SPARCL1 does not alter the density nor size of 991 inhibitory synapses. Primary data comprise means ± SEM; 29 cells from 3 independent cultures 992 were analyzed. Statistical significance was evaluated by an unpaired T-test. 993 (2G-I) Quantifications showing that recombinant SPARCL1 does not alter the neuronal cell 994 density (G), number of primary dendrites, and dendritic branches (H), or soma size of neurons (I). 995 Primary data comprise means ± SEM; 21 cells from 3 independent cultures were analyzed. 996 Statistical significance was evaluated by an unpaired T-test. 997 (2J) SPARCL1 has no significant effect on capacitance and input resistance. Primary data 998 comprise means ± SEM; 21 cells from 3 independent cultures were analyzed. Statistical 999 significance was evaluated by an unpaired T-test.

1000 Test parameters highlighted in blue were used to determine statistical significance and are 1001 summarized in Table 1. P-values and significance levels for individual comparisons between 1002 specific groups are highlighted in yellow. 1003 1004 Figure 3-1: Complete statistical analyses for “Figure 3: Cre-recombinase 1005 expression substantially reduces neurexin and neuroligin protein expression in 1006 conditional knockout neurons” 1007 (3B) Analysis of neurexin protein expression in primary cortical neurons from Nrxn123 cKO mice. 1008 Immunoblots were performed on lysates from cortical cultures infected with lentiviral ΔCre or Cre. 1009 The neurexin signals were normalized to corresponding β-actin signals for quantification. Primary 1010 data comprise means ± SEM; 3 independent cultures were analyzed. Statistical significance was 1011 evaluated by an unpaired T-test. 1012 1013 (3C) Analysis of neuroligin protein expression in primary cortical neurons from Nlgn1234 cKO 1014 mice. Immunoblots were performed on lysates from cortical cultures infected with lentiviral ΔCre 1015 or Cre. The neuroligin signals were normalized to corresponding β-actin signals for quantification. 1016 Primary data comprise means ± SEM; 3 independent cultures were analyzed. Statistical 1017 significance was evaluated by an unpaired T-test. 1018 1019 Test parameters highlighted in blue were used to determine statistical significance and are 1020 summarized in Table 1. P-values and significance levels for individual comparisons between 1021 specific groups are highlighted in yellow. 1022 1023 Figure 4-1: Complete statistical analyses for “Figure 4: Deletion of neurexins in 1024 dissociated cultures of mouse neurons and glia neither decreases excitatory or 1025 inhibitory synapse numbers nor impairs the increase in excitatory synapse 1026 numbers produced by recombinant SPARCL1” 1027 (4B-C) Quantifications showing that recombinant SPARCL1 increases the density of excitatory 1028 synapses (B) but not inhibitory synapses (C) in control and Nrxn123 cKO neurons. SPARCL1 did 1029 not affect the size and staining intensity of excitatory and inhibitory synapses. Primary data 1030 comprise means ± SEM; 24 cells from 3 independent cultures were analyzed. Statistical 1031 significances were evaluated by one-way ANOVAs with Tukey’s post-hoc tests of multiple 1032 comparisons. 1033 Test parameters highlighted in blue were used to determine statistical significance and are 1034 summarized in Table 1. P-values and significance levels for individual comparisons between 1035 specific groups are highlighted in yellow. 1036

1037 Figure 5-1: Complete statistical analyses for “Figure 5: Recombinant SPARCL1 1038 increases spontaneous excitatory but not inhibitory synaptic transmission 1039 independent of neurexins” 1040 (5A) SPARCL1 increases mEPSC frequency but not amplitude independent of neurexins. Triple 1041 neurexin (Nrxn123) control (ΔCre) and conditional knockout (Cre) neurons were treated at 13 DIV 1042 with control medium or recombinant SPARCL1. Spontaneous excitatory synaptic activity was 1043 assessed at 14 DIV by mEPSC recordings in the presence of TTX and PTX. Primary data 1044 comprise means ± SEM; 18 cells from 3 independent cultures were analyzed. Statistical 1045 significance was evaluated by a one-way ANOVA with Tukey’s post-hoc test of multiple 1046 comparisons. 1047 (5B) SPARCL1 does not affect spontaneous inhibitory synaptic activity. Nrxn123 control and cKO 1048 neurons were treated at 13 DIV with control medium or recombinant SPARCL1. Spontaneous 1049 inhibitory synaptic activity was assessed at 14 DIV by mIPSC recordings in the presence of TTX, 1050 CNQX and D-AP5. Primary data comprise means ± SEM; 17 cells from 3 independent cultures 1051 were analyzed. Statistical significance was evaluated by a one-way ANOVA with Tukey’s post- 1052 hoc test of multiple comparisons. 1053 Test parameters highlighted in blue were used to determine statistical significance and are 1054 summarized in Table 1. P-values and significance levels for individual comparisons between 1055 specific groups are highlighted in yellow. 1056 1057 Figure 6-1: Complete statistical analyses for “Figure 6: Deletion of neuroligins does 1058 not significantly decrease synapse numbers and does not block excitatory 1059 synaptogenesis induced by recombinant SPARCL1” 1060 (6B-C) Quantifications showing that recombinant SPARCL1 increases the density of excitatory 1061 synapses but not inhibitory synapses in control and Nlgn1234 cKO neurons. SPARCL1 did not 1062 substantially affect the size and staining intensity of excitatory and inhibitory synapses. Primary 1063 data comprise means ± SEM; 24 cells from 3 independent cultures were analyzed. Statistical 1064 significances were evaluated by one-way ANOVAs with Tukey’s post-hoc tests of multiple 1065 comparisons. 1066 Test parameters highlighted in blue were used to determine statistical significance and are 1067 summarized in Table 1. P-values and significance levels for individual comparisons between 1068 specific groups are highlighted in yellow. 1069 1070 Figure 7-1: Complete statistical analyses for “Figure 7: Recombinant SPARCL1 1071 increases spontaneous excitatory but not inhibitory synaptic transmission 1072 independent of neuroligins” 1073 (7A) Deletion of all neuroligins significantly reduces mEPSC frequency and amplitude but does 1074 not prevent SPARCL1 from increasing mEPSC frequency. Quadruple neuroligin (Nlgn1234) 1075 control (ΔCre) and conditional knockout (Cre) neurons were treated at 13 DIV with control medium

1076 or recombinant SPARCL1. Spontaneous excitatory synaptic activity was assessed at 14 DIV by 1077 mEPSC recordings in the presence of TTX and PTX. Primary data comprise means ± SEM; 18 1078 cells from 3 independent cultures were analyzed. Statistical significance was evaluated by a one- 1079 way ANOVA with Tukey’s post-hoc test of multiple comparisons. 1080 (7B) Deletion of all neuroligins significantly reduces mIPSC frequency and amplitude, which are 1081 unaltered by SPARCL1. Nlgn1234 control and cKO neurons were treated at 13 DIV with control 1082 medium or recombinant SPARCL1. Spontaneous inhibitory synaptic activity was assessed at 14 1083 DIV by mIPSC recordings in the presence of TTX, CNQX and D-AP5. Primary data comprise 1084 means ± SEM; 18 cells from 3 independent cultures were analyzed. Statistical significance was 1085 evaluated by a one-way ANOVA with Tukey’s post-hoc test of multiple comparisons. 1086 Test parameters highlighted in blue were used to determine statistical significance and are 1087 summarized in Table 1. P-values and significance levels for individual comparisons between 1088 specific groups are highlighted in yellow. 1089 1090 Figure 8-1: Complete statistical analyses for “Figure 8: SPARCL1 selectively 1091 enhances evoked neurotransmission and increases NMDAR-mediated synaptic 1092 responses independent of neuroligins” 1093 (8B) Deletion of all neuroligins reduces the amplitudes of evoked AMPAR- and NMDAR-EPSCs. 1094 Treatment with recombinant SPARCL1 increases these parameters and particularly enhances 1095 the NMDAR/AMPAR ratio independent of neuroligins. Quantifications of evoked AMPAR-EPSC 1096 amplitude, NMDAR-EPSC amplitude, and NMDAR/AMPAR ratio. NMDAR current amplitudes 1097 were measured 50 ms after stimulation. Primary data comprise means ± SEM; 24 cells from 3 1098 independent cultures were analyzed. Statistical significance was evaluated by a one-way ANOVA 1099 with Tukey’s post-hoc test of multiple comparisons. 1100 (8C) Deletion of all neuroligins reduces the amplitudes of evoked GABAR-IPSCs, whereas 1101 addition of SPARCL1 does not alter these currents. Quantifications of evoked GABAR-IPSC 1102 amplitude. Primary data comprise means ± SEM; 27 cells from 3 independent cultures were 1103 analyzed. Statistical significance was evaluated by a one-way ANOVA with Tukey’s post-hoc test 1104 of multiple comparisons. 1105 Test parameters highlighted in blue were used to determine statistical significance and are 1106 summarized in Table 1. P-values and significance levels for individual comparisons between 1107 specific groups are highlighted in yellow.