C1QL3 promotes cell-cell adhesion by mediating complex formation between ADGRB3/BAI3 and neuronal pentraxins

Abstract

Synapses are the fundamental structural unit by which neurons communicate. An orchestra of proteins regulates diverse synaptic functions, including synapse formation, maintenance, and elimination – synapse homeostasis. Some proteins of the larger C1q super-family are synaptic organizers involved in crucial neuronal processes in various brain regions. C1Q-like (C1QL) proteins bind to the adhesion G protein-coupled receptor B3 (ADGRB3) and act at synapses in a subset of circuits. To investigate the hypothesis that the secreted C1QL proteins mediate tripartite trans-synaptic adhesion complexes, we conducted an in vivo interactome study and identified new binding candidates. We demonstrate that C1QL3 mediates a novel cell-cell adhesion complex involving ADGRB3 and two neuronal pentraxins, NPTX1 and NPTXR. Analysis of single-cell RNA-Seq data from cerebral cortex shows that C1ql3, Nptx1, and Nptxr are highly co-expressed in the same excitatory neurons. Thus, our results suggest the possibility that in vivo the three co-expressed proteins are pre-synaptically secreted and form a complex capable of binding to post-synaptically localized ADGRB3, thereby creating a novel trans-synaptic adhesion complex. Identifying new binding partners for C1QL proteins and deciphering their underlying molecular principles will accelerate our understanding of their role in synapse organization.

INTRODUCTION

Many well-studied proteins involved in synapse homeostasis are linked to brain disorders, underlining their crucial role (1). Hetero- and homo-trans-synaptic molecular interactions facilitated by synaptic adhesion proteins are essential for the proper function of the synapse (1–4). Synaptic adhesion proteins bind across the synaptic cleft to form a molecular interface between pre- and post-synaptic membranes and initiate intercellular trans-synaptic signaling. These proteins are at the junction of our genes and our experiences and are critical for initiating and stabilizing synaptic changes in a multitude of ways. Synaptic adhesion proteins forming a trans-synaptic complex can involve one (e.g., cadherin), two (e.g., neurexin-neuroligin), or more proteins (e.g., neurexin-CBLN1-GRID2)(5–7). New synaptic adhesion proteins continue to be discovered, reflecting the complexity necessary to govern the exquisite dynamics among diverse synapses.

Proteins of the C1q/tumor necrosis factor (TNF) superfamily have recently entered the spotlight for their function at synapses. These proteins have a structurally similar globular C1q domain (gC1q) named for the founding member, C1q (8). Another member, complement component 1, q subcomponent-like 3 (C1QL3; a.k.a CTRP13) is part of a secreted protein subfamily with four paralogs (C1QL1–4) and is remarkably evolutionarily conserved with only a single amino acid substitution between human and mouse (9). C1ql3 expression was initially thought to be largely brain-specific but subsequently has been shown to be expressed in adipose and pancreatic β-islet cells and functions in energy metabolism, glucose uptake, and insulin secretion (10–13). C1QL proteins are high-affinity ligands for the adhesion G protein-coupled receptor B3 (ADGRB3; a.k.a. BAI3) (14, 15). Similar to C1QL3, ADGRB3 mouse/human orthologs are remarkably evolutionarily conserved and more than 98% identical (16). As ADGRB3 is widely distributed, the synaptic specificities between C1QLs and ADGRB3 lie in the regional expression of C1QLs. ADGRB3 is post-synaptically localized on cerebellar Purkinje cells and interacts with pre-synaptically secreted C1QL1 from the inferior olive climbing fibers; this interaction promotes synapse formation and maintenance (17, 18). We and others have shown that C1QL3 is expressed in a subset of circuits and promotes excitatory synaptogenesis/maintenance and genetically demonstrated its importance in several specific behaviors (19–21).

In the hippocampus, C1QL2 and −3 bind to pre-synaptically localized neurexin 3 and also to post-synaptically localized kainate receptor subunits GRIK2 and GRIK4, forming a complex that alone does not regulate synapse density (22). C1ql3 is expressed in numerous other locations and a biochemical mechanism to explain the influence of C1QL3 on synapse density is lacking (20). The intrinsic biochemical properties of the C1QLs are illuminating as to their role in synaptic organization. The C-terminal gC1q domains of C1QL proteins assemble into obligate trimers, and their N-terminal collagen-like domain allows further multimerization into higher weight assemblies. Crystallographic and biochemical analyses reveal that, despite structural similarities, C1QLs have unique electrostatic surface and metal binding properties compared to other gC1q domain containing proteins, suggesting specific interaction partners and functions for C1QL proteins compared to other members of the C1q/TNF superfamily (23). We hypothesize that C1QL3 is a bidirectional synaptic organizer and that the mechanism of its influence on synapse density involves binding to its post-synaptically localized partner ADGRB3 plus a presently unknown membrane-tethered pre-synaptic partner.

To explore additional binding partners for C1QL3, we used an in vivo interactome approach, which led to the discovery that C1QL3 initiates a tripartite complex between the newly found binding partner of neuronal pentraxin-1 (NPTX1) and ADGRB3. Neuron-specific glycoproteins NPTX1 and its paralog neuronal pentraxin receptor (NPTXR) are found at excitatory synapses and interact with AMPA receptors (24–30). NPTX1 and NPTXR can form an extremely stable heteromeric complex when expressed simultaneously in the same secretory pathway (31). NPTXR was given the name ‘receptor’ because of its transmembrane domain. However, once NPTX1 is secreted, it is unable to interact with NPTXR; therefore, a likely function of NPTXR is to anchor a co-expressed NPTX1 to the cell surface (32).

In this study, we describe a novel C1QL3-NPTX1 interaction, which we suggest is important for a novel trans-synaptic adhesion complex in which C1QL3 bridges the post-synaptically localized ADGRB3 to the pre-synaptically localized NPTX1/NPTXR complex. This is supported by our results herein demonstrating that C1QL3 mediates a quaternary cell-cell adhesion complex in vitro and single-cell RNA sequencing showing that C1ql3, Nptx1, and Nptxr are usually co-expressed in excitatory neurons, suggesting the proteins are secreted as a complex. Our results suggest an additional possibility that C1QL3, by forming a complex with NPTX1 and NPTXR, may influence excitatory synapses through an indirect interaction with AMPA receptors.

MATERIALS AND METHODS

Stereotaxic virus infection

An adeno-associated virus (AAV; DJ serotype) was made to express HA-C1ql3-IRES-eGFP using a CMV promoter (cloning details described below). A control AAV was also used that only expressed GFP. All surgeries were performed on mice 5 weeks old and 0.5 μl (titer = 10¹⁰/ml) purified AAV was injected bilaterally at rate of 0.1 μl/min. Coordinates given as distance from bregma according to The Mouse Brain atlas (33) and units in mm. To target dentate gyrus: −1.83 A/P, −1.8 D/V, ± 1.28 M/L.

Time-controlled crosslinking and preparation of proteins for mass spectrometry analysis

The protocol was adapted from (34). 11 male C57BL/6J mice each were infected with either the C1ql3 or control virus. Mice were perfused one week after infection. Under anesthesia, mice were perfused first with PBS for 1 minute then with 15 ml of 4% paraformaldehyde (PFA) for only 6 minutes. After perfusion, the hippocampal CA3 regions were dissected out and frozen within a maximum of 18 minutes and pooled into either the C1ql3 or control group. Dissected tissue was mixed with homogenization buffer (150 mM NaCl, 50 mM NH₄Cl, 100 mM Tris pH 8, protease inhibitors) in 10:1 ratio of buffer over tissue. Tissue was homogenized as described (34) in equal volume of extraction buffer (150 mM NaCl, 1% deoxycholate, 1% NP-40, 20 mM Tris pH 8), and mixed for 30 minutes at 4°C. To remove insoluble material, the homogenate was centrifuged for 5 min at 1000 g, then the supernatant was centrifuged again for 1 hour at 100,000 g. To immunoprecipitate HA-tagged proteins, supernatant was incubated overnight while mixing at 4°C with monoclonal anti-HA−agarose (A2095 Sigma, USA). The affinity matrix containing the bound proteins was transferred to Screw Cap Spin Columns (Thermo Scientific Pierce, USA), and washed a total of 5 times with 10 column volumes each. First wash was with buffer A (1% NP-40, 150 mM NaCl, 25 mM HEPES, pH 7.5), second with high salt buffer (1% NP-40, 500 mM NaCl, 25 mM HEPES pH 7.5), then again with buffer A, followed by two final washes with PBS. Acid elution was performed with 100 mM glycine, pH 2.5, twice for 5 minutes each. pH was normalized with 1 M HEPES, pH 8.0. Eluates were pooled and analyzed together with the Stanford University Mass Spectrometry core facility.

Protein Digestion and LCMS

All neutralized IP eluates were precipitated with 4X volume −80°C acetone on dry ice for at least 1 hour. Following centrifugation at 12,000 g for 10 minutes at 4°C, the supernatant was removed, and the samples were dried for 5 minutes by vacuum centrifugation. The protein pellet was reconstituted in 35 μL 0.02% of acid labile surfactant protease max (Promega, USA) and 50 mM ammonium bicarbonate. The samples were reduced using DTT (5 mM) at 50°C for 30 minutes followed by alkylation using propionamide (10 mM) for 30 minutes at room temperature. The sample volume was adjusted to 100 μL by the addition of 50 mM ammonium bicarbonate after which 600 ng Trypsin/LysC (Promega, USA) was added. The samples were digested overnight at 37°C, followed by an acid quench using 10 μL 10% formic acid water. Peptides were stage-tip purified on C18 columns packed in-house where the eluent was dried and stored at 4°C. The LC was an Eksigent nano2D run at 600 nL per minute flow rate where the analytical column was packed in house using 3 μM C18 reversed phase particles (Peeke Scientific, USA) at 15 cm in length and the source was a Bruker Captive Spray. The mass spectrometer was an LTQ Orbitrap Velos, set to acquire in DDA fashion where the top 15 most intense precursor ions were selected for fragmentation in the ion trap. The .RAW data were searched using Byonic (Protein Metrics, USA) against the Uniprot mouse database. The tolerances were 16 ppm precursor mass error and 0.25 Da fragment ion error searched in a target decoy approach where all data were filtered to a 1% FDR. The protein assignments were further strengthened by requiring 4 peptides per protein assigned.

Sequences and plasmids

14 new plasmids were created for this manuscript.

NPTX1

• 1To express NPTX1-V5 in HEK-293T cells (Figure 2): pL304 NPTX1-V5. cDNA for human NPTX1 (Uniprot Q15818) purchased from the PlasmID Repository, a service of the DNA Resource Core at Harvard Medical School (Clone ID HsCD00296108 in the vector pDONR221). Invitrogen Gateway recombination cloning was used to move the cDNA into pL304, which added an in-frame V5 tag to the C-terminus. pL304 vector uses a CMV promoter. This vector could be used to make a lentivirus, but was here used for transient transfection to make NPTX1-V5 protein instead.
• 2To express NPTX1-flag in HEK-293T and COS-7 cells (Figures 2B,C and 4B–C): pCMV5 NPTX1-flag. The above-mentioned cDNA used as template for PCR to insert cDNA into pCMV5 vector using HindIII and XbaI R.E. sites. Forward primer 5’-atcgatAAGCTTccaccATGCCGGCCGGCCGCGCC with HindIII 5’ overhang. Reverse primer 5’-tccTCTAGATtACTTGTCGTCATCGTCTTTGTAGTCGTTGATCTGGCGACAGGC with flag tag and XbaI 5’ overhang.
• 3To express NPTX1-flag in FreeStyle 293-F cells for purification or HEK-293T cells (Figures 3D, 4E, 5, and 6): pEB Multi Neo Nptx1-flag-IRES-GFP. Template for PCR was IMAGE clone for mouse Nptx1 (Uniprot Q62443) cDNA (IMAGE ID 8862042, accession number BC139020). The base vector was purchased from Wako Chemicals (Japan): pEB Multi-Neo. The multi-cloning site was modified to contain new R.E. sites: Acc65I-XhoI-BamHI-SalI-BsrGI-NotI. The intermediate vector was further modified by adding a IRES-GFP using the BamHI and BsrgI sites. Insert DNA extracted from pL316. The PCR amplicon of Nptx1 was inserted into the intermediate vector using XhoI and BamHI R.E. sites. Forward primer 5’- gatacaGTCGACccaccATGCTGGCCGGCCGCGCC with SalI 5’ overhang. Reverse primer 5’-tgtatcggatccTtACTTGTCGTCATCGTCTTTGTAGTCtccGTTGATCTGGCGACAAGCCTCGAAT with flag tag and BamHI 5’ overhang.
• 4To express NPTX1-flag-strep in HEK-293T cells and FreeStyle 293-F cells for purification (Figures 2D,3A–C,S1F and S1G): pEB Multi Neo Nptx1-flag-TEV-strep-IRES-GFP. The above construct #3 was modified to introduce a TEV cleavage site and a strep tag after the flag tag. We used the following primers: forward primer 5’- gatacaGTCGACccaccATGCTGGCCGG with a SalI overhang, and a reverse primer 5’-CAACggaGACTACAAAGACGATGACGACAAGGAAAACCTGTATTTTCAGGGTTGGAGCCACCCGCAGTTCGAAAAATGAggatccTCCG ending with a BamHI 5’- overhang.

NPTXR

• 5To express full length NPTXR-flag in HEK-293T and COS-7 cells (Figures 2B,C and 4A): pCMV5 NPTXR-flag. cDNA for human NPTXR (Uniprot Q95502) purchased from the PlasmID Repository (Clone ID HsCD00083063 in the vector pENTR223.1). This was used as template for PCR to insert cDNA into pCMV5 vector using HindIII and XbaI R.E. sites. The start codon of NPTXR had to be switched from CTG to ATG in order to express. Forward primer 5’- atcgatAAGCTTccaccAtgaagttcctggccgtgc with HindIII 5’ overhang. Reverse primer 5’- tccTCTAGAttaCTTGTCGTCATCGTCTTTGTAGTCtgccttggccctcccctt with flag tag and XbaI 5’ overhang.
• 6To express full length untagged NPTXR-flag COS-7 and 293T cells (Figures 4B–C and 6): pCMV5 NPTXR. This vector was created simultaneously as the above pCMV5 NPTXR-flag except using a different reverse primer for PCR: 5’- tccTCTAGAttatgccttggccctccc with XbaI 5’ overhang.
• 7To express a soluble form of NPTXR-flag with no transmembrane domain in HEK-293T cells or FreeStyle 293-F cells for purification (Figures 3D, 4D,5,S1H&S1I): pEB Multi Neo soluble Nptxr-flag-IRES-GFP. The transmembrane domain of NPTXR was replaced with the signal peptide from NPTX1 to produce a soluble secreted protein. Note that NPTXR is a type 2 membrane protein. cDNA for mouse Nptxr (Uniprot Q99J85) was purchased from the Mammalian Gene Collection (MGC; Accession number BC098199, clone ID 5705707). The above-mentioned pEB Multi Neo intermediate vector was used. The MGC cDNA was used as template for PCR of Nptxr and inserted into the intermediate vector using XhoI and BamHI R.E. sites. Forward primer 5’- acaCTCGAGccaccATGCTGGCCGGCCGCGCCGCACGCACCTGTGCGCTGCTCGCCCTCTGCCTCCTGGGCAGTGGGGCCCTGCCCGGCGGCACCGACAA with XhoI 5’ overhang. Reverse primer 5’- tgtatcggatccTtACTTGTCGTCATCGTCTTTGTAGTCtccTGCCTTTGCCCTCCCCTTGCACAC with flag tag and BamHI 5’ overhang.

ADGRB3

• 8To express full length flag-ADGRB3 in COS-7 cells (Figure S1C): pCAGGS flag-BAI3 (Uniprot O60242), described previously (14).
• 9To express full length untagged ADGRB3 in COS-7 cells (Figure 4D,E): pCAGGS BAI3 untagged. From above vector, 1371 bp DNA was removed from 5’ region of cDNA using ClaI and BsrGI R.E. sites to remove the DNA encoding the N-terminal flag tag. Replacement 5’ PCR amplicon inserted using same vector as template. Forward primer 5’- gatacaATCGATATGTCaGCtCTaCTcATaCTtGCaCTaGTaGGtGCaGCtGTaGCaTTTGGATTTAATGCTGCCCAAGACTTCTG with ClaI 5’ overhang. Reverse primer 5’- ATTGGCTGTACATTCAGGGTTATAGCACT with BsrGI 5’ overhang.
• 10To express ADGRB3-V5 extra cellular domain (ECD) from FreeStyle 293-F cells for purification (Figure 4A–C): pEB Multi Neo BAI3-V5-IRES-GFP. Note that this construct was truncated just after the hormone binding domain, excluding the GAIN domain, therefore not the entire ECD was used. C-terminal amino acid (a.a.) sequence reads with V5 tag underlined: …FARCISGKPIPNPLLGLDST*stop*. Protein includes a.a. 20 – 569 of ADGRB3. Note this construct did not express well, for unknown reasons. Template for PCR was the above-described pCAGGS BAI3 untagged. The above-mentioned pEB Multi Neo intermediate vector was used. The PCR amplicon of ADGRB3 was inserted into the intermediate vector using XhoI and BamHI R.E. sites. Forward primer 5’- gatacaCTCGAGccaccATGTCTGCACTTCTGATCCTAGCTC with XhoI 5’ overhang. Reverse primer 5’- tgtatcggatccTtAcgtagaatcgagaccgaggagagggttagggataggcttaccTGATATGCATCTTGCAAAGCTCGG with V5 tag and BamHI 5’ overhang.
• 11To express flag-extended CUB domain of ADGRB3 from FreeStyle 293-F cells for purification (Figure S1B): pEB Multi Neo flag-extended CUB-IRES-GFP. The above-mentioned pEB Multi Neo intermediate vector was used. No PCR was needed. The cDNA for flag-extended CUB domain of ADGRB3 was excised from pCMV5 flag-extended CUB using SalI and BamHI and inserted into pEB Multi Neo using XhoI and BamHI sites. Protein includes a.a. 20 – 290 of ADGRB3.
• 12To express a GPI-anchored CUB domain ADGRB3 with N-terminal flag in COS-7 cells (Figure S1D): pCMV5 flag-CUB GPI anchor IRES mCherry. According to UniProt database, CUB domain corresponds to a.a 30–159 in ADGRB3. This protein includes a.a. 20 – 179 of ADGRB3. First, flag-CUB domain inserted into pCMV5 vector, which was opened with BglII and SalI R.E. Template for PCR was pCAGGS flag BAI3 described in (14). Forward primer 5’-ACAGGATCCTCGAGTCACCATGTCTGCACTTCTGAT with BamHI 5’ overhang and amplicon will include the flag tag just after the signal peptide. Reverse primer 5’-ATCGTCGACGGCGCGCCTTAAGATCTTAAGCAGCTCTCCAACCAAGT with BglII and SalI in 5’ overhang. To add GPI anchor signal sequence: the previous step was opened with BglII and inserted was a cDNA fragment corresponding to the GPI signal sequence of the mouse contactin-1 gene with corresponding sticky ends (sticky ends not included in this sequence): ggctcacgatatataatcacatgggatcacgttgtggcactatccaatgaatctactgtgacaggctacaagatactctatagaccggatggccaacatgacggcaagctgttctcaacccataaacactccatagaagtccccatccccagagacggagagtatgttgtcgaggttcgagcacacagcgacgggggagacggagttgtgtctcaagtcaaaatttcaggcgtgtccacgctgtcttcgagcctcctcagcttgctcctgccctcccttggctttcttgtctactcggaattctga. Once pCMV5 flag-CUB GPI was completed, an IRES and mCherry was added. First, IRES segment excised from pAAV synapsin WGA ΔCre IRES tdTomato described in (20) using AscI and XhoI and inserted into pCMV5 flag-CUB GPI using AscI and SalI sites in the 3’ MCS. Second, mCherry was excised from pL309 +mCherry-IRES C1ql3-HA (20) with BamHI and inserted into pCMV5 flag-CUB GPI IRES with BamHI site in the 3’ MCS.
• 13To express a GPI-anchored extended CUB domain ADGRB3 with N-terminal flag in HEK-293T and COS-7 cells (FigureS1E–I): pCMV5 flag-extended CUB GPI anchor IRES mCherry. The above-mentioned pCMV5 flag-CUB GPI anchor IRES mCherry was used as base vector to insert new cDNA of ADGRB3. The original ADGRB3 cDNA sequence of the CUB domain was excised with SacI and BglII, leaving in place the GPI anchor sequence. A PCR amplicon was inserted in-frame using template from the above-mentioned pEB Multi Neo flag-extended CUB-IRES-GFP. Forward primer 5’- taagcagagctcgtttagtgaacc which includes SacI site. Reverse primer 5’- ccgggatccACCAGTTTGTGCCATAAATTTAGCA with BamHI 5’ overhang. Protein includes a.a. 20 – 290 of ADGRB3.

C1QL3

• 14To over-express HA-C1QL3 via AAV (Figure 1): pAAV HA-C1ql3 IRES eGFP. C1ql3 sequence was murine (Uniprot Q9ESN4). pAAV promoter is CMV. Vector was subcloned from related pAAV2 vectors used in (20). As previously used, the HA tag was added immediately after the signal peptide (14, 20). A PCR product was inserted into pAAV2 using the BamHI and XbaI R.E. sites. Template for PCR was an unpublished vector for making lentiviruses: pL316 HA-C1ql3-IRES-eGFP. PCR amplicon contained the entire HA-C1ql3-IRES-eGFP construct. Forward primer 5’-cggggatccatgGAGACAGACACACTCCTG with BamHI 5’ overhang. Reverse primer 5’- gactctagagcgcagaatccaggtgg with XbaI 5’overhang.
• 15To express HA-C1QL3 (Uniprot Q9ESN4) for transfection in COS-7 cells (Figures 2B,C and S1G&S1I): pDisplay HA-C1ql3, described previously (14). Note that pDisplay vector was modified such that protein is not membrane tethered.
• 16To express HA-C1QL3 (Uniprot Q9ESN4) from FreeStyle 293-F cells for purification (Figures 3 – 6 and S1B–E): pEB Multi Neo HA-C1ql3-IRES-GFP. Template for PCR was pDisplay HA-C1ql3 described in (14). The above-mentioned pEB Multi Neo intermediate vector was used. The PCR amplicon of C1ql3 was inserted into the intermediate vector using XhoI and BamHI R.E. sites. Forward primer 5’- gatacaCTCGAGccaccatgGAGACAGACACACTCCTGCTAT with XhoI 5’ overhang. Reverse primer 5’- tgtatcggatccTCAGTCAGCATAAATAATAAATCCAGAGAATG with BamHI 5’ overhang.
• 17To express the C1q domain of C1QL3 with N-terminal GST tag from bacteria (Figure 2D): pGEX-KG_C1QL3, described previously (14, 23).

Tissue culture cells lines

COS-7 cells were purchased from ATCC (ATCC^® CRL-1651^™). HEK-293T cells were purchased from ATCC (ATCC^® CRL-1573^™). Non-adherent HEK cells purchased from ThermoFisher (USA) (FreeStyle^™ 293-F R79007).

Co-immunoprecipitation

For Figure 2B,C: Expression vectors or empty pCMV5 control vector were transiently transfected into HEK-293T cells with the calcium phosphate method in 6-well dishes. 2 days later, HEK-293T cell media was harvested, pH balanced with 1 M HEPES pH 7.4, and protease inhibitors were added. Media was centrifuged at 4700 rpm for 20 minutes at 4°C to pellet any cell debris then supernatant centrifuged again for 10 min at 14,000 RPM at 4°C. This supernatant was used as ‘media’ in co-IP experiments. Media was incubated overnight at 4°C with the anti-HA affinity matrix as above. For Figure 2C, media from different transfections were mixed for 24 hours prior to adding the anti-HA beads). The following morning, affinity matrix was transferred to Screw Cap Spin Columns (Thermo Scientific Pierce, USA), and washed with wash buffer (50 mM NaCl, 10 mM HEPES pH 7.4, 2 mM CaCl₂). Bound proteins were eluted with 100 mM glycine pH 2.5 and pH-balanced with 1 M HEPES. Eluates were thoroughly denatured with freshly made 50 mM DTT and SDS running buffer for 1 hour at room temperature, then 20 min at 55°C, then 95°C for 5 min. Immunoblotting was with either anti-HA (1:6,000), anti-V5 (1:5,000), or anti-flag antibodies (1:4,000). HRP-conjugated secondary antibody was used at 1:5,000 for ECL detection (GE Healthcare Amersham ECL Western Blotting Detecting Reagents, USA).

For Figure 2D, affinity purified protein samples were incubated in binding buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM CaCl₂) overnight with agitation. To study the complex formation between NPTX1 and gC1QL3, 1.8 μg of dual-tagged NPTX1 (flag-strep tags) and 120 μg of GST-tagged gC1QL3 were mixed overnight. Samples were then incubated for three hours with 40 μL slurry of pre-equilibrated PureCube StrepTactin Hi Cap Magnetic Beads (Cube Biotech, Germany) to allow for the strep-tagged NPTX1 (bait protein) to be immobilized, and to pull-down GST-tagged C1QL3 (prey protein). To remove unbound proteins, the Magnetic Beads were washed twice with 250 μL of binding buffer. Protein(s) immobilized by the strep MagBeads were eluted with 30 μL of with D-Desthiobiotin (1X StrepTactin elution buffer, IBA, Cat. No. 2-1000-025, Germany) supplemented with 2 mM CaCl₂ to maintain calcium level constant throughout the experiment. To analyze whether gC1QL3 formed a complex with NPTX1, SDS-PAGE and corresponding anti-flag and anti-GST Western Blots were used for detection of NPTX1 and gC1QL3, respectively. SIGMAFAST 5-bromo-4-chloro-3-indolyl phosphate (BCIP)/nitro blue tetrazolium (NBT) Alkaline Phosphatase Substrate tablets (B5655, Sigma, USA) was used according to the package directions.

For co-IP of C1QL3 and the CUB domain of ADGRB3 in Figure S1B: Constructs indicated in the figure were transiently transfected into HEK293T cells using the calcium phosphate method in 2 wells of a 6-well plate. All three ADGRB3 vectors had flag tags and C1QL3 had HA tag. After 2 days, conditioned media was collected, sterile filtered, and concentrated to 250 μL using EMD Millipore (USA) Amicon^™ Ultra-0.5 Centrifugal Filter Units with Ultracel-50 membrane (UFC50109). Concentrated proteins were mixed and incubated overnight with anti-HA beads (described above) in Screw Cap Spin Columns (described above).

Media was removed with a spin at 500 RPM for 1 minute. All wash and elution steps were performed at same speed. Proteins were washed with interaction buffer (10 mM HEPES pH 7.4, 50 mM NaCl, 2 mM MgCl₂, 2 mM CaCl₂) 3x then eluted 2x in 30 μL with 100 mM glycine pH 2.5. ECL immunoblot for HA and flag tags was the same as described above.

Antibodies

The following antibodies were used in this study either on cryosections, cultured COS-7 cells, immunoblots, or for SPR: rat anti-HA (clone 3F10, Roche, Switzerland), goat anti-GST-AP (for Figure 2D, AP-1310, Columbia Bioscience, USA), rabbit anti-synaptoporin (for Figure 1, a.k.a. synaptophysin-2, Y941, (35)), rabbit anti-flag (F3165, Sigma, USA), rabbit anti-V5 (for Figure 2, AE071, Abclonal, USA), mouse anti-V5 (for Figure 4, 46–0705, Invitrogen, USA), mouse anti-NPTX1 (for Figure S1, sc-374199, Santa Cruz Biotechnology, USA), and mouse anti-NPTXR (for Figure S1, Abcam, U.K. ab168254). The appropriate secondary Alexa Fluor or HRP-conjugated antibodies (Life Technologies, USA) were used: goat anti-rat Alexa Fluor 633 (A21094), goat anti-rabbit Alexa Fluor 546 (A11035), goat anti-mouse Alexa Fluor 488 (A11029), goat anti-rat HRP (HAF005), goat anti-rabbit HRP (A16104). Goat anti-mouse HRP (ab97023, Abcam U.K.)

Protein expression and purification

For Figure 2D, 3A–C, and S1H,I: NPTX1-flag-strep was produced in HEK-293T. Proteins were expressed and purified from monolayered cultures grown at about 80% confluence in DMEM media supplemented with 10% Fetal Bovine Serum (FBS), on either T300 or cell factories (Corning (USA) CellSTACK Culture chambers, 10-STACK with 6,360 cm² cell growth area). Plasmids were transfected using Polyethylenimine (PEI) (linear, MW ~25,000, Cat# 23966, Polyscience, USA) reagent at a stock concentration of 1 mg/mL and mixed with 1 mg/mL DNA in 1X PBS. Following the addition of the transfection mixture to the cultures, the media was supplemented with Valproic Acid (VPA, prepared from Sodium 2-Propylvalerate, TCI No. S0894) at a final concentration of 4 mM. VPA is a histone deacetylase inhibitor, previously shown to enhance expression of recombinant proteins in mammalian expression systems (36, 37). The media containing the secreted proteins was harvested on the fourth day after transfection for purification. This protocol was adapted from (36). Media was supplemented with 20X PBS to a final concentration of 1X PBS to buffer pH; protease inhibitor cocktail (Roche, Switzerland) was added to prevent protein degradation. To remove cells and cell debris from the media, the supernatant was harvested by centrifugation at 12,100 rpm for 20 minutes at 4 °C. For proteins subsequently purified using the strep-tag, media was supplemented to a final concentration of 1 mM EDTA for optimal binding efficiency to StrepTactin Sepharose High Performance (GE, USA) resin. The clarified supernatant was loaded and passed through a pre-equilibrated StrepTactin Sepharose High Performance (GE) resin connected to a peristaltic pump (PALL). The column was washed with 10X column volumes of 1X PBS (containing 1 mM EDTA). Protein was eluted with D-Desthiobiotin (1X StrepTactin elution buffer, IBA, Germany, Cat. No. 2-1000-025). Protein was further purified with Size Exclusion Chromatography (SEC) by using a Superose 6 Increase 10/300 GL (29091597, GE, USA) column. The proteins were separated in TBS buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl). Characterization of the NPTX1-containing fractions was performed by SDS–PAGE analysis under reducing conditions on a 10 % gel. About 0.3 μg of total protein were analyzed by performing electrophoretic runs on the Mini PROTEAN 3 Cell (BioRad, USA) at 100 V for 60 minutes. Detection of the protein bands was obtained by staining with Coomassie Brilliant Blue G-250 dye or Pierce Silver Stain Kit (Ref.no. 24612), followed by imaging with a Gel Doc XR+ with Image Lab software.

gC1q domain of C1QL3 was produced and purified as described previously (14).

To create purified proteins for Figures 3D, 4 – 6 and S1: Proteins were produced in eukaryotic FreeStyle^™ 293-F Cells (R79007, ThermoFisher, USA). Each protein to be produced was cloned into the pEBMulti-Neo vector (Wako Chemicals, Japan), which harbors EBNA-1 derived from Epstein-Barr Virus. Vectors were transfected into cells using FreeStyle MAX reagent (Life Technologies, USA) and G418 at a 0.40 mg/mL working concentration was used for selection. Proteins were purified from the culture media using methods similar to co-IP protocol described above. Depending on the tag, the following were used to immobilize proteins: monoclonal anti-HA−agarose (A2095 Sigma, USA), anti-flag M2 affinity gel (A2220 Sigma, USA), or V5 epitope tag antibody [agarose immobilized] (NB600–387 Novus Biologicals, USA).

Glycosylation analysis

To determine the level of N-linked glycosylation on NPTX1, highly pure NPTX1 protein was treated with Glycopeptidase F (GPF) (Takara, Code No. 4450, Japan). The sample was treated with GPF under denaturing conditions as per supplier manual. GPF-treated and untreated samples were incubated for 16 hours at 37°C. SDS-PAGE and western blot (anti-flag) under denaturing and reducing conditions were employed to determine whether there was a band-shift, which is indicative of N-glycan removal, and to calculate the percentage of N-glycan relative to the full-length protein.

Dynamic Light Scattering

Prior to DLS measurements, highly pure NPTX1 protein sample was filtered through a PVDF Ultrafree-MC Centrifugal Hydrophilic Filter (UFC30GV25, Millipore, USA) in a refrigerated tabletop centrifuge, by centrifugation at 12,000 for 30 seconds. Dynamic light scattering was performed with 70 μL of the sample using a Malvern Instruments, UK, Zetasizer Nano-S. Data was analyzed using the Zetasizer Software.

Blue Native PAGE analysis

0.3 μg of each purified protein was mixed with Invitrogen NativePAGE 4x Sample Buffer (BN2003, Invitrogen, USA) and loaded onto a NativePAGE 3 to 12%, Bis-Tris gel (BN1001BOX, Invitrogen, USA) using a Mini Gel Tank (A25977, Invitrogen, USA). For molecular weight estimation: NativeMark Unstained Protein Standard (LC0725, Invitrogen, USA). Electrophoresis was performed according to instructions described in http://tools.thermofisher.com/content/sfs/manuals/nativepage_man.pdf with standard anode buffer and light blue cathode buffer containing Coomassie G-250. The protein standard was visualized with routine staining with Coomassie Blue R-250, while rest of gel was transferred to PVDF membrane with a Trans-Blot Turbo (Bio-Rad, USA) and NuPAGE Transfer Buffer (NP0006, Invitrogen, USA), followed by western blot using anti-HA and anti-flag antibodies as described in the section on co-IP.

Cell surface binding assays

1 μg each of expression vector was transiently transfected into COS-7 cells with the calcium phosphate method in 24-well dishes onto glass coverslips. 2 days later, cells washed with binding buffer (DMEM+FBS growth media + 50 mM HEPES pH 7.4 + 2 mM MgCl₂ + 2 mM CaCl₂ + 0.02 % (w/v) NaN₃; azide blocks endocytosis). Purified exogenous HA-C1QL3 applied to cells at 150 nM for 90 min at room temp. Cells were washed with binding buffer twice, then washed with buffer twice (10 mM HEPES pH 7.4, 150 mM NaCl, 2 mM MgCl₂, 2 mM CaCl₂). Cells fixed for 10 min with ice cold 4% PFA made in PBS. Cells washed 3 times with PBS. For testing of ternary interactions, the protein binding step was repeated with an additional purified exogenous protein (ADGRB3-V5 ectodomain, NPTX1-flag, or NPTXR-flag) at 150 nM. Fixing the first soluble protein before applying the second soluble protein improved the signal and interpretability of the images. After second fixation and washes, cells blocked with 5% goat serum in PBS for 1 hour, followed by indirect immunofluorescence using appropriate primary (overnight at 4°C) and secondary antibodies (1 hour at room temperature). Single plane images were acquired on a Nikon (Japan) confocal microscope. To quantify, 12 healthy appearing transfected cells (determined by large and flat morphology) were selected and photographed at random. The magnetic lasso tool was used in Adobe Photoshop to outline the cell’s plasma membrane, then the mean fluorescence intensity value acquired for each color to be quantified (background signal intensity was subtracted by selecting a cell-free region adjacent to the outlined cell). Each measurement was averaged to create n = 1, and the entire experiment was replicated 3 times. For the cell surface binding assays in Figure S1, methods were similar except for: exogenous application of NPTX1-flag and NPTXR-flag which were applied without a prior fixation step, used at concentrations of 1 μM, and images were acquired on a Keyence BZ-X810 widefield microscope.

Surface Plasmon Resonance

Surface Plasmon Resonance data were collected using the ProteOn XPR36 Protein Interaction Array System from Bio-Rad (USA) equipped with a GLC sensor chip. Rat anti-HA antibody was diluted to a concentration of 0.1 mg/mL in a pH 4.5 acetate buffer for protein immobilization through amine coupling. Before immobilization, the GLC sensor chip was activated with NHS/EDC according to standard immobilization procedures. The diluted anti-HA antibody was flowed over vertical flow cells 1 and 2 at a flow rate of 30 μL/min for 450 seconds. Flow cells 3–6 were left blank and not used for data collection. The reaction was quenched by application of ethanolamine and the flow cells were washed with 1 M NaOH. The anti-HA antibody was directly immobilized to allow for the capture of purified HA-tagged C1QL3 to avoid direct immobilization of the C1QL3 protein. Experimental setup and determination of binding affinities were identical for both the C1QL3-NPTX1 and C1QL3-NPTXR interactions. 4.4 μg of purified HA-tagged C1QL3 was diluted in immobilization buffer (10 mM HEPES, 150 mM NaCl, 2 mM CaCl₂, 0.05% Tween, 1 mg/mL BSA, pH 7.5) and flowed over only vertical flow cell 1 at a rate of 30 μL/min for 300 seconds to capture C1QL3 on the sensor surface. Vertical flow cell 2 was left blank to be used as a reference. Before capture of C1QL3 protein, the vertical flow cells were washed with interaction buffer at a flow rate of 100 μL/min for 60 seconds. Interaction buffer was then flowed over the 6 horizontal flow cells at a flow rate of 100 μL/min for 60 seconds. This was followed by simultaneous injection of the respective concentration series (including a blank) for both NPTX1 and NPTXR over all 6 horizontal flow cells at a flow rate of 100 μL/min for 120 seconds. This was followed by a 300 second dissociation in which blank buffer was flowed over the complex. After testing NPTXR, the surface was regenerated by two sequential injections of 100 mM glycine pH 2.5 and 2 M NaCl both at 100 μL/min for 18 seconds. The final sensorgrams were produced through double referencing to control for non-specific binding as well as diffusion of C1QL3 away from the anti-HA antibody. Both data sets were fit best (i.e., had the lowest χ²) with the ‘heterogeneous ligand’ model using the Bio-Rad software (χ² = 17.7 RU for NPTX1 and 3.8 for NPTXR). Assigning Langmuir 1:1 kinetics produced obviously incorrectly fitted curves (χ² = 180.3 RU for NPTX1 and 14.6 for NPTXR).

HEK-293T cell aggregation and flow cytometry

Adherent HEK-293T cells were grown in 6-well plates and transiently transfected using the calcium phosphate method with varying plasmids depending on the condition. In the experimental condition, a plasmid to express ADGRB3 (#8 in the plasmid section above) was co-transfected with a separate plasmid carrying mCherry to mark transfected cells. A NPTX1 and GFP plasmid (#3 in the plasmid section above) was co-transfected with NPTXR (#6 in the plasmid section above) to promote NPTX1 surface tethering. For the various control conditions, an omitted plasmid was replaced with empty pCMV5 vector or a plasmid to only express GFP or mCherry, as required. Co-transfected plasmids were added in equimolar amounts such that 4 μg of total DNA was added to each well. 1 day later, cells were detached with gentle shaking at 37°C with PBS + 1 mM EDTA. Detached cells were removed from wells, placed in a 1.5 mL tube and pelleted at 1000 RPM for 5 minutes. Cell pellets were re-suspended in binding buffer (10 mM HEPES, 50 mM NaCl, 5.5 mM CaCl₂, 2% FBS). An equal amount of each cell population was mixed either in the presence of 1.6 μM purified C1QL3 or with no C1QL3 and incubated for 1 hour at room temperature on a rotisserie. The mixed cell solutions were then analyzed by flow cytometry on an Amnis (USA) ImagestreamX Mark II instrument. 1000 events were collected for both conditions with and without C1QL3. The objective lens was set to allow visualization of cell aggregates (>1 cell). Aggregates were identified by plotting bright-field aspect ratio against bright-field area as a measure of overall cell shape. Aggregates were further analyzed by plotting mCherry intensity against GFP intensity. NPTX1/R-ADGRB3 aggregates were identified as having signal from both GFP and mCherry. Single cells with bleed through into other channels were excluded by focusing only on aggregates as defined above. Gates were determined manually on control cells (cells incubated without C1QL3) and then applied to the experimental condition.

Bioinformatics

We used the comprehensive single-cell RNA sequencing dataset GSE115746 (38) and a bioinformatics pipeline as previously published (39). We quantified Nrxn3 alternative splicing using exon junction reads, defining exon 25a, 25b, and 25c as reads mapping to exon junctions with receptor sites on Chromosome 12 at positions 90,331,484, 90,331,787, and 90,332,108, respectively. As we were only interested in whether a sequence located in exon 25b would be included in the resulting transcript, and while exon 25 has multiple acceptor sites but only one donor site, we merged exons 25a and 25b into a single exon (25ab) for further analysis. Threshold of gene detection was 1 read for all cells and genes analyzed.

Statistical analysis

For statistical analysis of cell surface binding assays and HEK293T cell aggregation assays: GraphPad Prism software was used to perform 1-way ANOVA with Bonferroni’s post-hoc test. Threshold of p ≤ 0.05 indicated statistical significance. (* p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001).

RESULTS

In vivo C1QL3 interactome study reveals new binding partners

We hypothesized the existence of a binding partner for C1QL3 that is tethered to pre-synaptic membranes, thereby allowing the possibility of a ternary trans-synaptic adhesion complex with C1QL3 and the post-synaptic ADGRB3. To identify a novel member that associates with C1QL3, we conducted an in vivo interactome screen utilizing an AAV vector that co-expressed an HA epitope-tagged C1QL3 and a reporter GFP (C1QL3-IRES-GFP); the HA tag did not disrupt C1QL3 function in vivo and the cytoplasmic GFP is not a fusion protein (20). We targeted this AAV or control GFP-only AAV with stereotactic injections to the dentate gyrus granule neurons, because these neurons endogenously express both C1ql2 and C1ql3 (9), their relatively accessible location, and because they send axonal projections to the nearby CA3 region of the hippocampus (Figure 1A). Mice were analyzed one week later. Consistent with the pre-synaptic localization of endogenously expressed C1QL3 that was previously observed in the hippocampus (20, 22), anti-HA immunolocalization was predominantly co-localized with synaptoporin (SYNPR), which marks the mossy fibers emanating from the dentate gyrus (Figure 1B).

Fig. 1: HA-C1QL3 ectopically expressed in dentate gyrus is purified from dissected CA3.

(A) Strategy for purification of C1QL3-associated proteins. AAV encoding HA-C1QL3 and GFP or GFP alone were stereotactically injected into dentate gyrus of 5-week-old mice. 1 week later, mice were perfused/fixed and the CA3 region was dissected for anti-HA immunoprecipitation. The screen consisted of a single cohort of 11 pooled mice for each condition.

(B) Co-immunofluorescence labeling of hippocampus coronal cryosections for HA-C1QL3, GFP, and SYNPR on the mossy fiber projection.

(C) A denaturing immunoblot for the HA tag to reveal the immunoprecipitated C1QL3 in varying degrees of fixation to other monomers and additional binding partners.

To improve the yield of co-immunoprecipitated C1QL3-associated proteins, we adapted a time-controlled transcardial crosslinking protocol (34). The CA3 region of the hippocampus was dissected and homogenized. C1QL3 proteins were immobilized from homogenate, and eluted proteins were immunoblotted for the HA tag, which revealed that exogenously expressed C1QL3 migrated at a variety of sizes (Figure 1C). Some C1QL3 migrated as a monomer and some as a trimer, indicating a desired substantial yet incomplete crosslinking of monomers. The doublet band likely represents differential post-translational modifications (13). Bands of larger molecular weights suggested either trimers crosslinked into higher molecular weight oligomers or C1QL3 crosslinked to binding partners.

NPTX1 co-immunoprecipitates with C1QL3

To identify the co-immunoprecipitated proteins, we performed mass spectroscopy analysis (Table 1). Numerous C1QL2 peptides co-immunoprecipitated with C1QL3, suggesting that C1QL2 and C1QL3 can form a heteromeric complex in vivo. The expected binding partners ADGRB3 and kainate receptor (GRIK2, a.k.a. GluK2) returned only one peptide for each in the HA-C1QL3 sample, compared to no peptides in the control sample, suggesting that our co-immunoprecipitation strategy pertained a bias towards proteins assembled within the secretory pathway of the pre-synaptic neuron. Surprisingly, we found no peptides for neurexin 3, despite our expectations based on a prior study (22). For a subset of the proteins identified with potential to be novel C1QL3 binding partners, corresponding DNA sequences were cloned with epitope tags into eukaryotic expression vectors. To test if these proteins could interact with C1QL3 in vitro, we transfected plasmids encoding each protein (Figure 2A and Table 1) into HEK-293T cells with or without simultaneous co-expression of C1QL3. Two days later, secreted/soluble C1QL3 and associated proteins in the culture media were immunoprecipitated and analyzed by immunoblot using antibody probes to detect any co-precipitating proteins. We found that C1QL3 co-immunoprecipitated NPTX1 (Figure 2B). This was particularly exciting, as NPTX1 has been shown to be released pre-synaptically from neurons (24, 28, 30). C1QL3-NPTX1 interaction was not dependent on co-expression, as co-immunoprecipitation was also observed when each expression vector was transfected into separate populations of cells and the media mixed (Figure 2C).

Table 1: List of proteins identified by mass spectroscopy analysis.

Mice were injected with AAV to either express HA-C1QL3 and GFP, or GFP alone, and dissected brain tissue from each condition was subjected to immunoprecipitation of the HA tag. First left two columns show the number of peptides for each protein identified by mass spectroscopy analysis that co-immunoprecipitated with either the HA tagged C1QL3 or GFP. Hspa5, Cntn1, Nptx1, Nptxr, Lgals3bp, and Nrcam were cloned into expression vectors with epitope tags for additional verification experiments. Last two right columns show the result summary of additional binding assays for each protein by co-immunoprecipitation (co-IP) and a cell surface binding assay. “Variable” for NPTXR assays indicates that observed binding varied with experimental conditions.

# peptides / condition
C1QL3-HA	GFP	Identified Protein	Co-IP?	Cell surface binding?
113	0	C1Q-like 3 (C1QL3)
80	7	78 kDa glucose-regulated protein (HSPA5)	no	no
50	0	C1Q-like 2 (C1QL2)	N/A	N/A
20	4	Contactin-1 (CNTN1)	yes	no
18	0	Neuronal pentraxin-1 (NPTX1)	yes	yes
15	0	Neuronal pentraxin receptor (NPTXR)	variable *	variable #
10	0	Galectin-3-binding protein (LGALS3BP)	no	no
4	0	Neuronal cell adhesion molecule (NRCAM)	no	no
3	0	Neurogenic locus notch homolog protein 1 (NOTCH1)	N/A	N/A
2	0	Cerebellin-4 (CBLN4)	N/A	N/A
1	0	Glutamate receptor, ionotropic kainate 2 (GRIK2)	N/A	N/A
1	0	Adhesion G protein-coupled receptor B3 (ADGRB3)	N/A	N/A

^*Compare Figures 2B to 2C

^#Compare Figures 4A&D to S1G

Fig. 2: NPTX1 co-immunoprecipitates with C1QL3.

(A) Domain structure and types of epitope tags of constructs used in panels B and C. The known cleavage sites of NPTXR are indicated.

(B) The indicated constructs were co-transfected for expression in HEK-293T cells. 2 days later, secreted HA-C1QL3 and associated proteins were immunoprecipitated with an anti-HA affinity matrix from the culture media. (left) Immunoblot on the flow-though (i.e., unbound proteins not immunoprecipitated) to detect either V5 tag or flag tag, as indicated. (right) Immunoblot for eluted/purified proteins.

(C) Identical design as in panel B, except proteins were expressed from separate populations of cells (as indicated) and the secreted proteins mixed prior to anti-HA immunoprecipitation.

(D) Co-immunoprecipitation purified NPTX1-flag-strep and GST-gC1QL3. At top are domain structures and types of epitope tags of constructs used. NPTX1 was detected by western blot (anti-flag) and gC1QL3 was detected by anti-GST antibody. Double negative controls were used ensuring a) GST protein does not directly bind to NPTX1, and b) GST-gC1QL3 does not bind to streptavidin magnetic beads.

Each condition has been replicated at least once, with representative results shown.

It was previously shown that when the type II membrane protein NPTXR is expressed in HEK-293T cells, a fraction of the protein is membrane-tethered but is also cleaved in distinct locations creating two populations of soluble proteins (referred to as “A” and “B”), which we similarly observed here (24) (Figure 2A and 2B). C1QL3 co-immunoprecipitated the larger NPTXR “A” cleavage product when co-expressed in the same cells (Figure 2B), but not when separately expressed (Figure 2C), suggesting a weaker affinity. When NPTX1 and NPTXR are co-expressed in the same cell, they form a tight complex (we refer to as ‘NPTX1/R’) that is either membrane-tethered or released from the surface if NPTXR is cleaved (31). When NPTX1, NPTXR, and C1QL3 were co-expressed, C1QL3 co-precipitated both neuronal pentraxins, suggesting all three can exist in a complex (Figure 2B). To test whether the N-terminal collagen domain of C1QL3 is necessary for NPTX1 binding, we co-immunoprecipitated the gC1q domain of C1QL3 (gC1QL3) alone with NPTX1 and found that the gC1q domain is sufficient to bind NPTX1 (Figure 2D), similar to the gC1q domain binding to ADGRB3 (14), and suggesting that a C1QL3 multimer will provide multiple binding sites. In summary, these results suggest that C1QL3, NPTX1, and NPTXR can form a complex, potentially with a multifarious stoichiometry.

NPTX1 and NPTXR form oligomers larger than previously appreciated

Proteins with a multidomain organization such as C1QL3 and NPTX1 often exhibit complex quaternary structures, especially domains such as the N-terminal collagen-like domain of C1QL3 and the predicted N-terminal coiled-coil domain of NPTX1 (23, 40–42). To determine the oligomeric state and average size of NPTX1, we used size exclusion chromatography (SEC), dynamic light scattering (DLS), and native gel electrophoresis (native PAGE). SEC showed that NPTX1 was eluted in a homogeneous peak that corresponds to a size above 670 kDa of the thyroglobulin standard, suggesting a NPTX1 oligomer to be closer to a size of a pentadecamer (15-mer, 750 kDa) (Figure 3A). To test what fraction of NPTX1 molecular weight is due to N-linked glycosylation, we treated NPTX1 with Glycopeptidase F (GPF) and concluded that NPTX1 molecular weight is ~10% N-linked glycosidic moieties (Figure 3B; similar to previously observed in (29)). Next we analyzed the main SEC peak fraction by DLS, which revealed two populations of NPTX1, 24.6 nm and 186.5 nm in size (Figure 3C). The cumulant size (z-average) was 61 nm with a polydispersity index of 0.5, indicating that NPTX1 is a moderately polydisperse sample. It was previously reported using the technique of sucrose gradient sedimentation that NPTX1 oligomerization is that of a pentamer (~250 kDa) (31). We therefore used a complementary technique, blue native PAGE, to confirm the extent of NPTX1 multimerization suggesting a size of ~750 kDa, consistent with a NPTX1 pentadecamer (Figure 3D). We also show that NPTXR forms an oligomer as well (Figure 3D), though it was previously reported to form at most a dimer if expressed alone (31). The reasons for these differences are unclear. Using native PAGE we also observed purified C1QL3 appearing as three populations, likely corresponding to a hexamer, dodecamer (12-mer), and octadecamer (18-mer) (Figure 3D; assuming each monomer subunit is ~37 kDa). Our results suggest that NPTX1 exists as an obligate multimer likely dependent on its predicted N-terminal coiled-coil domain for assembly. The dodecamer species of C1QL3 appeared most prominently at ~444 kDa, therefore we predict that a typical C1QL3-NPTX1 complex would form an assembly in a mass range of ~1.2 MDa.

Fig. 3: NPTX1 and NPTXR form oligomers larger than previously appreciated.

(A) SEC of NPTX1. Insert identified NPTX1 by coomassie blue (CB) stained and immunoblot (anti-flag) of main peak fractions separated on denaturing PAGE. NPTX1-flag-strep made using plasmid #4 described in methods section and purified using the strep tag.

(B) NPTX1 treatment with Glycopeptidase F (GPF). GPF-treated and non-treated samples were analyzed on denatured PAGE and detected by silver stain and immunoblot (anti-flag).

(C) DLS of NPTX1 particles from the main single peak in SEC. Two unique sample concentrations (red and purple) resulted in one population 24.6 nm in diameter with an average intensity of 30% and a second population 186.5 nm in diameter with an average intensity of 70%.

(D) Purified NPTXR, NPTX1, and C1QL3 separated by blue native PAGE and detected by western blot for epitope tags as indicated. NPTX1-flag made using plasmid #3 described in methods section and purified using the flag tag. The construct to create soluble NPTXR for purification is described in methods section under plasmid #7.

Each condition has been replicated at least once, with representative results shown.

C1QL3 facilitates a quaternary complex with NPTX1-NPTXR-ADGRB3

We next tested if C1QL3 can bind NPTX1 and/or NPTXR on a live cell surface, and if C1QL3 can mediate a complex between NPTX1/R and ADGRB3. The use of live, healthy cells is important as C1QL3 and several homologs have an affinity for lipids exposed on the surfaces of dying cells (43). COS-7 cells were transfected to express NPTXR-flag or to co-express NPTX1-flag and untagged NPTXR. Purified C1QL3 was applied exogenously to the live cells, which were then fixed and immunostained to detect cell surface C1QL3 and flag tag. NPTXR alone showed minimal binding to C1QL3, suggesting a low affinity (Figure 4A). C1QL3 showed robust interaction with cells co-expressing NPTX1/R (Figure 4C). To test if C1QL3 can bind simultaneously to both NPTX1 and the extracellular domain (ECD) of ADGRB3, after unbound C1QL3 was washed off and cells were fixed, we applied purified ECD-ADGRB3. We found that ECD-ADGRB3 was retained on the cell surface only when C1QL3 was first bound to cells expressing NPTX1/R (Figure 4B,C). This was further confirmed in a reciprocal set of binding experiments in which purified exogenous NPTX1 (150 nM) was only retained on the surface of ADGRB3-expressing cells if C1QL3 had been pre-bound to ADGRB3 (Figure 4E). Purified exogenous ectodomain of NPTXR at 150 nM had minimal affinity for C1QL3 (Figure 4D). At a higher application concentration of 1 μM of exogenous NPTXR, binding was more significant (Figure S1G). Taken together, our data demonstrate that C1QL3 can mediate complex formation between neuronal pentraxins and ADGRB3, and that the interaction is greatly enhanced by the co-expression of NPTX1 and NPTXR.

Fig. 4: C1QL3 forms ternary complex with NPTX1/ADGRB3.

(A-C) COS-7 cells were transfected to express either NPTXR alone or NPTX1/R followed by a cell surface binding assay at room temperature for C1QL3 and ECD-ADGRB3 (extracellular domain of ADGRB3). Endocytosis in the live cells was blocked by sodium azide. NPTX1 is secreted/soluble therefore its expression alone causes no flag tag to be retained on the cell surface (not shown). Following final fixation, immunofluorescence without permeabilization for bound proteins was performed for HA-C1QL3, ADGRB3-V5, and NPTXR-flag/NPTX1-flag. In B-C, the NPTXR was not tagged, therefore any flag signal is due to NPTX1 in complex with membrane-tethered NPTXR. Representative pictures of each are shown.

(D,E) The reciprocal design of panels A-C. Untagged ADGRB3 was expressed on cell surface and exogenous/soluble C1QL3 and NPTXR or NPTX1 proteins were applied at 150 nM.

On right: quantifications of immunofluorescence signal intensity. Each marker represents the average of 1 experiment. Statistical comparisons were performed using 1-way ANOVA with Bonferroni’s post-hoc test (* p ≤ 0.05, ** p ≤ 0.01).

ADGRB3 has several extracellular domains (Figure S1A); what domain or domains are minimally required for mediating ternary complex formation with C1QL3 and neuronal pentraxins is not known. Based on the binding between C1QL1 and ADGRB3 (17), we expected a similar minimal binding domain of ADGRB3 for binding to C1QL3 at ADGRB3’s N-terminal CUB domain (a.a. 30 – 159 according to the UniProt database). However, we found that additional C-terminal residues after the CUB domain up to a.a. 290 were required for binding to C1QL3. This ‘extended CUB’ domain was sufficient for binding to C1QL3 by co-IP (Figure S1B). To facilitate investigating if this extended CUB domain was sufficient for ternary complex formation, we fused it to a GPI anchor and expressed this construct in COS-7 cells. This membrane-tethered recombinant protein was sufficient for binding exogenous C1QL3 in a cell surface binding assay (Figure S1E), and formed a ternary complex with neuronal pentraxins (Figure S1G,I). Whether and how the previously reported thrombospondin domains of ADGRB3 affect binding to C1QL3 is unclear (14).

C1QL3 binds to NPTX1 with low micromolar affinity

The observed differences in binding of NPTX1 and NPTXR to C1QL3 in the above experiments led us to quantify their binding affinities using surface plasmon resonance (SPR). We immobilized N-terminally HA-tagged C1QL3 to an anti-HA antibody-coated sensor chip surface and then applied a dilution series of purified NPTX1 or a soluble version of NPTXR lacking a transmembrane domain (Figure 5A). The NPTX1 kinetic data were poorly fit by the standard 1:1 Langmuir binding model (χ² = 180.3 RU), suggesting a more complex binding stoichiometry. Consequently, a ‘heterogeneous ligand’ model was applied (χ² = 17.7 RU), which assumed that C1QL3 presents two or more distinct binding sites, each with a distinct binding affinity. This binding model seems plausible since purified C1QL3 exists in multiple possible oligomeric states (Figure 3D). Moreover, oligomerization likely increases its avidity, and consequently its binding kinetics likely involve multi-step binding and dissociation events. This resulted in at least two distinct observed sets of association and dissociation rates, each with their corresponding affinities.

Fig. 5: Determination of C1QL3-NPTX1 affinity.

(A) Coomassie stain of denatured purified proteins used for SPR.

(B,C) SPR results for (B) C1QL3-NPTX1 interaction and (C) C1QL3-NPTXR. Anti-HA antibody was immobilized on a GLC sensor chip to capture HA-C1QL3. NPTX1 (in A) or NPTXR (in B) was then flowed over the chip surface at the indicated concentrations. Data was double referenced to control for both non-specific binding of NPTX1 or NPTXR to the antibody and for diffusion of C1QL3 away from the antibody. Both affinity measurements were replicated using a distinct batch of proteins, yielding similar results.

Applying this analysis resulted in a K_D1 of 5.49 μM and K_D2 of 1.17 μM for C1QL3-NPTX1 interaction (Figure 5B). Comparable affinities were reported for other synaptic adhesion proteins (17, 22, 44). Consistent with our expectations that binding of NPTXR to C1QL3 would be of lower affinity, the observed SPR traces yielded low and ‘noisy’ response units (Figure 5C). Overall, these data provide further evidence that NPTX1 is the preferred neuronal pentraxin binding partner for C1QL3, suggesting that NPTXR may have an auxiliary function in tethering the complex to the plasma membrane.

C1QL3 promotes cell-cell adhesion on NPTX1/R and ADGRB3 expressing cells

To demonstrate that C1QL3 can create a cell-cell adhesion complex between cells expressing ADGRB3 and NPTX1/R heteromeric complexes, we performed a HEK-293T cell aggregation assay, and quantified this novel interaction by flow cytometry. One population of HEK-293T cells was transfected with plasmids expressing ADGRB3 and mCherry, and a second population was transfected to express NPTX1, NPTXR, and GFP (Figure 6A). Cells were mixed with or without the addition of purified C1QL3 to permit cell aggregation. The mixed cells were analyzed by flow cytometry that measured fluorescence intensity of either single cells or cell aggregates; each measurement is referred to as an “event”. The flow cytometry data were first plotted for mCherry (ADGRB3) fluorescence intensity vs. GFP (NPTX1/R) fluorescence intensity (Figure 6B). Gates were drawn using the control (no C1QL3) data to demarcate events that are ADGRB3-expressing only (purple), NPTX1/R-expressing only (green), or a mixture of ADGRB3/NPTX1/R cells (black). To only consider events that contain more than one cell, the same data were plotted as a function of aspect ratio versus bright-field area (aggregates will inherently take a more irregular shape and cover a larger area than single cells) and a gate was drawn to exclude all events containing a single cell (Figure 6C). Addition of C1QL3 created substantially more cell aggregates.

Fig. 6: C1QL3 promotes cell-cell adhesion between ADGRB3 and NPTX1/R expressing cells.

(A) Experimental design. Separate populations of adherent HEK-293T cells were transfected to express ADGRB3 and mCherry or NPTX1, NPTXR, and GFP. One day post-transfection, cells were detached from the plate and mixed in equal amounts, ± purified C1QL3 (1.6 μM). Fluorescence from cells were then analyzed by flow cytometry.

(B) Representative results including all recorded events plotted as intensities of mCherry vs. GFP. 1000 events recorded for each. Gates were created to represent events that included only mCherry (ADGRB3) expressing events (purple), only GFP (NPTX1/R) expressing events (green), and both mCherry + GFP expressing events (black).

(C) Results including all recorded events plotted as aspect ratio vs. brightfield area. Blue gate represents cell aggregates.

(D) The same data as in panel B but showing aggregates only.

(E) Distribution of cell aggregate (containing both ADGRB3 and NPTX1/R expressing cells) perimeter ± purified C1QL3. Data presented as means ± standard deviation, n = 3.

(F) Quantification of fold increase of cell-cell adhesion. For each indicated transfection condition the number of aggregates containing mCherry and GFP in the plus C1QL3 condition was divided by the corresponding number of aggregates in the minus C1QL3 condition. Each marker represents 1 experiment. Statistical comparisons were performed using 1-way ANOVA with Bonferroni’s post-hoc test (* p ≤ 0.05, ** p ≤ 0.01).

The same data plotted in Figure 6B are presented in Figure 6D, but only showing aggregates, and quantified in Figures 6E,F. The inclusion of C1QL3 caused an increase in both the number and the aggregate perimeter of events indicating increased cell-cell adhesion (Figure 6E). Importantly, C1QL3 caused a ~5-fold increase in aggregates containing both ADGRB3 and NPTX1/R compared to cells mixed without C1QL3 (Figure 6F). This is significantly greater than control transfection conditions that omitted either ADGRB3 or NPTXR. There was a small amount of aggregation between cells expressing ADGRB3 and NPTXR, consistent with above results showing that C1QL3 can bind NPTXR, but distinctly less than compared to NPTX1/R (Figure 6F). These data indicate that C1QL3 can bind to ADGRB3 and NPTX1/R simultaneously, and that this complex stabilizes cell-cell adhesion.

C1ql3, Nptx1, and Nptxr are highly co-expressed in excitatory neurons

To determine if C1ql3, Nptx1, and Nptxr co-express in single brain cells in vivo, we analyzed a comprehensive single-cell RNA sequencing (RNA-seq) dataset of adult mouse cortical cell types (38) using previously published methods (39). We found that C1ql3 is almost exclusively expressed in excitatory neurons (Figure 7A, blue bars). Notably, the co-expression of C1ql3, Nptx1, and Nptxr was particularly high among excitatory neurons. Of the C1ql3-positive excitatory neurons, 89% co-expressed Nptx1, and 88% co-expressed both Nptx1 and Nptxr (Figures 7B, black and green bars; and Figure 7C). These results are consistent with our hypothesis that NPTX1, NPTXR, and C1QL3 form complexes in excitatory neurons. By contrast, of the small subset of inhibitory neurons that expressed C1ql3, only 18% co-expressed Nptx1, suggesting pentraxin binding to C1QL3 is less common in inhibitory neurons (Figure 7A, B, and D). As expected, Adgrb3 was expressed almost uniformly in neurons suggesting that ADGRB3 is available for binding C1QL3 at any synapse where C1QL3 is secreted (Figure 7A) (16, 45).

Fig. 7: C1ql3, Nptx1, and Nptxr are commonly co-expressed in excitatory neurons.

(A) RNA-seq data showing the percent of total cells in each indicated category that express each indicated combination of genes.

(B) The same data as in panel A, but only considering C1ql3-expressing cells and the fraction in each category that co-express the indicated gene combination.

(C) Data plotted as Venn diagram for any excitatory neuron that expressed C1ql3, Nptxr + Nptx1, or Nrxn3+25ab.

(D) Venn diagram as in panel C, but for inhibitory neurons.

Recently it was reported that C1QL3 binds to pre-synaptically localized neurexin 3 (NRXN3) in the hippocampus (22). Nrxn3 is extensively alternatively spliced, and C1QL3 binding occurs if NRXN3 contains the specific alternatively spliced exon of 25b at splice site #5 (22). Since exon 25 has multiple acceptor sites but only one donor site, and both exons 25a and 25b include the 25b sequence, we merged exons 25a and 25b into a single exon (25ab), and analyzed for co-expression of C1ql3 with Nrxn3+25ab. Any Nrxn3 transcripts lacking the critical exon were excluded. We found that some C1ql3-positive excitatory neurons also co-expressed Nrxn3+25ab (33%), although less frequently than co-expression of C1ql3, Nptx1, and Nptxr (Figure 7B, yellow bars). Interestingly, of the smaller subset that co-expressed C1ql3 and Nrxn3+25ab, almost all of those also co-expressed both Nptx1 and Nptxr (Figure 7B, purple bars, and Figure 7C). Together, these findings suggest that a membrane-tethered C1QL3-NPTX1/R complex exists at many synapses in the cerebral cortex, as well as locations where cortical neurons project. Furthermore, these data suggest that a C1QL3-NRXN3 complex likely exists outside the hippocampus where it was initially described (22).

DISCUSSION

C1QL3 mediates a quaternary complex with NPTX1/R-ADGRB3

In our screen for novel C1QL3 binding partners, we identified NPTX proteins as promising candidates. After the screen was conducted, it was shown that ADGRB3 levels are unusually low in the CA3 region of the hippocampus and that C1ql2;C1ql3 double knockout mice have no reduction in synapse density in CA3, revealing that C1QL proteins can have distinct functions depending on available binding partners (22). Fortunately, Nptx1 is expressed strongly in dentate gyrus allowing us to identify it as a binding partner. Here we used five distinct assays to demonstrate that NPTX1 is a novel and bona fide binding partner for C1QL3, and that C1QL3 can simultaneously bind to NPTX1 and ADGRB3. It is likely that the other C1QL paralogs can bind NPTX1 and ADGRB3 simultaneously as well, given their similarities (14, 23), and each complex may function in unique cellular contexts. C1QL3 can bind to NPTXR directly, but has a low affinity, which we suspect is less likely to be physiologically relevant. We suspect that in vivo C1QL3 interacts with NPTXR indirectly, as NPTX1 and NPTXR form a stable heteromeric complex when co-expressed (31). NPTX1 binds taipoxin but NPTXR does not, therefore there is a precedent for non-conserved binding interactions in the NPTX family (31). NPTXR is known to be regulated by the protease TACE that cleaves NPTXR from the membrane (24). We therefore speculate that a similar proteolytic process may regulate the membrane tethering of the NPTX1/R-C1QL3 complex and thus regulate cell-cell adhesion.

Combined with our biochemical results, our data suggest that a trans-synaptic adhesion complex can exist when neurons express C1QL3, NPTX1, and NPTXR pre-synaptically, and ADGRB3 is present at the post-synaptic side. This is further supported by our observation that the presence of all four proteins promotes cell-cell adhesion in HEK-293T cells. The mechanism by which C1QL3 influences synapse density in vivo is unknown; our results propose that C1QL3 could influence excitatory synapse density either by creating or maintaining a trans-synaptic cell-cell adhesion complex or by an indirect interaction with AMPA receptors, which bind to multiple NPTX family members (24–28). Future experiments are needed to demonstrate these points in vivo.

C1QL3 interactions likely facilitate clustering of large pre- and post-synaptic protein complexes

We hypothesize that C1QL3 acts as a “hub” protein at the cell-cell interface, in contrast to more linear interactions between classical homo- or heterodimeric synaptic adhesion proteins such as cadherins and neurexins/neuroligins. Unlike the homolog CBLN1, which also mediates a ternary trans-synaptic adhesion complex, C1QL3 is not restricted to form hexamers, and consequently the C1QL3-NPTX1/R-ADGRB3 complex will likely have a very different stoichiometry compared to the CBLN1-NRXN1-GluD2 complex (46, 47). All proteins participating in this novel complex are multi-domain proteins and at least C1QL3 and NPTX1 alone can form large molecular assemblies. We previously predicted that the collagen domain of C1QL3 will create an octadecamer (~666 kDa), which we observe here, although a dodecamer appears to be more prominent (Figure 3D). We found NPTX1 forms oligomers up to ~750 kDa, which, together with a C1QL3 dodecamer or octadecamer, would result in a complex of a size above 1 MDa that would further increase upon ADGRB3 binding. C1QL3 and NPTX1 may also exist in distinct oligomeric states depending on their in vivo context and/or availability of other binding partners. The gC1q domain of C1QL3 is sufficient for binding to both ADGRB3 and NPTX1 individually in vitro. However, a multimer of C1QL3 trimers may be required to create the cell-cell adhesion complex and to cluster its binding partners at synaptic membranes.

Comparing the predicted coiled-coil region of NPTX1 with the ones of PTX3 reveals that it differs in length, helix position, and numbers as well as distribution of likely important cysteines. We therefore predict that NPTX1 has a unique way to oligomerize via its N-terminal coiled coil domain that is distinct from the proposed PTX3 oligomers (41), and that the pentraxin domain of NPTX1 is sufficient for binding to C1QL3. Our SPR kinetics data analysis provides clues regarding how these macromolecules interact. Of the two calculated C1QL3-NPTX1 affinities (Figure 5B), the second has both slower ‘on’ and ‘off’ rates. We suspect that this slower association is due to steric hindrance; C1QL3 trimer-trimer packing as well as possible NPTX1 oligomerization may influence the dynamics of the second binding event. The second slower “off” rate may be explained by a multivalent C1QL3 multimer allowing NPTX1 molecules to dissociate from one subunit and to immediately bind to a neighboring subunit before being released to the solution. Increasing avidity by multimerization of domains to achieve better binding appears common in extracellular proteins (48). A scheme of multimerization of subunits, which individually have low affinities, has potential evolutionary advantages, allowing for more rapid divergence of function after a gene duplication event (i.e., a fewer number of mutations are likely to be required to alter its binding partner) (49). Future structural and biochemical studies are needed to decipher the exact oligomeric state, structure, and binding modalities of ADGRB3, C1QL3, NPTX1, and NPTXR. We predict that this is a new regulatory pathway specifying how synapse homeostasis is regulated by one set of synaptic organizers that can interact with a variety of receptors. Cracking the molecular code of C1QL3 and its interactions will require more in-depth interdisciplinary and specifically structural research.

C1QL3 likely mediates trans-synaptic adhesion in vivo

Comparing our previous observations of the C1ql3 expression pattern to both previous in situ hybridization analyses of Nptx1/Nptxr and the Allen Brain Atlas reveals very similar expression patterns, especially for C1ql3 and Nptx1 (19, 20, 29, 32). Here, we directly examined gene co-expression in the cerebral cortex, the predominant location where C1ql3 is expressed. We confirm our previous observation that C1ql3 is expressed predominantly in a subset of excitatory neurons, and additionally find that if a neuron expresses C1ql3, it has a very high probability of co-expressing both Nptx1 and Nptxr. Given that all three are known to be targeted pre-synaptically, this strongly suggests that a C1QL3-NPTX1/R ternary complex exists in vivo and is likely located in the synaptic cleft. We also show the near ubiquitous neuronal expression of Adgrb3 in cerebral cortex (Figure 7A). Combined, our findings suggest the possibility that a quaternary protein complex exists with ADGRB3 at excitatory synapses of the cerebral cortex and that C1QL3 completes a trans-synaptic adhesion complex. It is also highly likely that the NPTX1/R-C1QL3 complex exists in the hippocampus and it is possible that it interacts with post-synaptic kainate receptors creating a distinct trans-synaptic adhesion complex in this context.

We investigated the expression of the specific Nrxn3 splice variant that encodes a protein capable of binding to C1QL3 in the hippocampus (22) and its co-expression with other genes in the cerebral cortex. We found a subset of C1ql3-expressing neurons that co-express this Nrxn3 isoform, although co-expression was less frequent than that of C1ql3 and Nptx1. We speculate that if the post-synaptic neuron expresses the appropriate kainate receptor subunits, then it is likely that C1QL3 will mediate a distinct trans-synaptic adhesion complex at these NRXN3-containing synapses. In the subset of neurons that express all C1ql3, Nptxr, Nptx1, and Nrxn3, two unique C1QL3-mediated trans-synaptic adhesion complexes could exist and NRXN3 and NPTX1 might compete for binding to C1QL3.

Nptx1;Nptxr double knockout mice have been created but unlike C1ql3 knockout mice, minor phenotypes were reported (25, 50). The prior genetic analyses were largely done in retinal ganglion cells where C1ql3 likely has minimal expression (19). Nptx1 and Nptxr have yet to be genetically examined at synapses where a NPTX1/R-C1QL3-ADGRB3 complex is likely to be common, such as excitatory synapses in the telencephalon. Of note, both Nptx1 and C1ql3 knockout mice are hyperactive and both transcripts are up-regulated after neuronal activity blockade as part of the homeostatic scaling response (51, 52), suggesting an epistatic genetic interaction is likely.

Conclusion

Here we present biochemical evidence that C1QL3 mediates cell-cell adhesion by simultaneously binding to ADGRB3 and a heteromeric complex containing NPTX1 and NPTXR. In vivo transcriptional analysis suggests that these proteins co-exist at excitatory synapses and likely create a trans-synaptic adhesion complex that may be involved with the presently unknown mechanism by which C1QL3 influences synapse density.

ABBREVIATIONS

a.a.

amino acids

AAV

adeno-associated virus

ADGRB3; a.k.a. BAI3

adhesion G protein-coupled receptor B3

AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

AP

alkaline phosphatase

CBLN1

cerebellin 1

C1ql3; a.k.a Ctrp13

complement component 1, q subcomponent-like 3

CUB

complement protein subcomponents Clr/Cls, urchin embryonic growth factor, and bone morphogenic protein 1

DLS

dynamic light scattering

DTT

dithiothreitol

ECD

extracellular domain

FBS

fetal bovine serum

gC1q

globular C1q domain

GPCR

G protein-coupled receptor

GPF

glycopeptidase F

GPI

glycosylphosphatidylinositol

GRID2

glutamate receptor ionotropic delta 2

GRIK2

glutamate ionotropic receptor kainate type subunit 2

GRIK4

glutamate ionotropic receptor kainate type subunit 4

IP

immunoprecipitation

IRES

internal ribosome entry site

LCMS

liquid chromatography - mass spectrometry

NPTX1, a.k.a. NP1

neuronal pentraxin-1

NPTXR, a.k.a. NPR

neuronal pentraxin receptor

NRXN3

neurexin 3

PAGE

polyacrylamide gel electrophoresis

PEI

polyethylenimine

PFA

paraformaldehyde

SEC

size exclusion chromatography

SPR

surface plasmon resonance

SS

silver stain

SYNPR

synaptoporin

TNF

tumor necrosis factor