Alternative splicing of latrophilin-3 controls synapse formation – Nature

Mouse handling

C57BL/6 (JAX, 000664) mice were used for tissue RT–PCR experiments, and CAG-Cas9 mice (JAX, 024858) were used for all of the other experiments. Mice were weaned at P21 and housed in groups of maximum 5 under a 12 h–12 h light–dark cycle with food and water ad libitum, in the Stanford Veterinary Service Center. All of the procedures conformed to National Institutes of Health Guidelines for the Care and Use of Laboratory Mice and were approved by the Stanford University Administrative Panel on Laboratory Animal Care.

Plasmids

Plasmids for the TRUPATH assay were from the Roth lab (Addgene, 1000000163). Pink flamindo 2, GFP and PDE7b constructs were obtained from previous study⁸. Mouse Lphn3 of specified splicing variants (all have the following splicing configuration: E6⁻E9⁺E15⁺E28⁻E29⁻) were cloned into the pCMV vector using In-Fusion HD assembly. For manipulating Lphn3 KO, Lphn3 E31 KO or Lphn3 E31(ΔPBM), gRNAs were cloned into lentiCRISPR v2 (Addgene, 52961), followed by human SYN1-promoter driven eGFP (for calcium imaging and RNA-seq) or mScarlet-I (for synapse puncta staining) or Cre recombinase (for monosynaptic rabies tracing), using In-Fusion HD assembly. jRGECO 1a³⁸ was cloned into the FSW lentiviral vector. Gkap, Homer3, Psd95, Shank3 coding regions¹⁵ containing N-terminal 6×His and 3C protease cleavage site were cloned into the pCT10 vector. N-terminal Flag-tagged LPHN3 E31 and LPHN3 E32 (with full splicing combination: E31: E6⁻E9⁺E15⁺E24⁺E28⁻E29⁻E30⁻E31⁺; E32: E6⁻E9⁺E15⁺E24⁺E28⁻E29⁻E30⁻E32⁺) were cloned into the lenti_CMV-TetO2 vector. Sequences of all constructs were confirmed by Sanger sequencing at Elim Biopharm or by long-read sequencing at Primordium.

Genetic CRISPR manipulations

Four gRNAs were designed in this study. The control gRNA (5′-CCGGAAGAGCGAGCTCTACT-3′) was designed to have no target in the mouse genome. The Lphn3 KO gRNA (5′-GCCCGGACAACGGAGCTCAA-3′) targets the constitutive E7 to induce a frameshift. The Lphn3 E31 gRNA (5′-TCTTGTAATCTTTTTCAGAG-3′) targets the splicing acceptor site immediately upstream of E31, to disrupt the inclusion of E31. The Lphn3 E31(ΔPBM) gRNA (5′-AGACTAGTGACCAAGTGCGC-3′) targets the PBM of Lphn3. Potential off-target effects were assessed using Cas-OFFinder⁴¹ to ensure specificity.

Generation of the reference exon list

The exon coordinates of Lphn3 were extracted from GFF annotation of mouse genome GRCm38/mm10. Non-overlapping exons were named numerically in ascending order from 5′ to 3′ of the transcript. For exons with overlapping regions (mostly due to alternative splicing donor/acceptor site), they were named with the same number but with different letters. For exons at the 5′ and 3′ untranslated region, only the longest annotated exons were used, as the current study focuses on the coding region. This generated the draft of exon list. As the annotated exon list may contain exons that never translate to proteins (mostly due to incomplete splicing/incorrect annotation), the draft exon list was used to map the reads of Lphn3 from Ribotaq sequencing dataset¹⁷, which is highly enriched for translating mRNAs, during which only exons detected in this dataset were preserved to produce the final reference exon list for this study.

Analysis of high-throughput sequencing data

This study analysed six datasets from published studies and two datasets generated from this study:

1. (1)

   Reads from PacBio long-read mRNA sequencing data¹⁶ were aligned to reference genome (GRCm38/mm10) using gmaps. Reads belonging to Lphn3 from above five tissue samples (four developmental stages for the retina and P35 for the cortex) were combined to increase the read depth, for analysing the abundance of full-length transcripts.

2. (2)

   Cell-type-specific RiboTag sequencing data¹⁸ reads were aligned to reference genome (GRCm38/mm10) using STAR. Reads belonging to Lphn3 were used for calculating the PSI of exons.

3. (3)

   Neuronal activity regulated bulk transcriptome (KCl and kainate treated)³⁴ and picrotoxin treated³⁵ datasets were downloaded from the Gene Expression Omnibus (GEO) under accession numbers GSE175965 and GSE104802. Reads were aligned as described above using STAR. Reads belonging to Lphn3 were used for calculating the PSI of exons.

4. (4)

   Analysis of native neuronal activity from the mPFC region used a scRNA-seq dataset³⁷. Smartseq reads were mapped to a custom genome, and individual Lphn3 exons were counted individually. Cells were unbiasedly clustered on the basis of their transcriptomes. Immediate early genes were identified by ranking each genes’ correlation to Fos expression. An IEG score was calculated by combining the expression of Fos, Ier2, Egr1, Junb and Dusp1, and this score was used to categorize the activation status of each cell. Single neurons with at least one count for E31 or E32 were used for splicing analysis.

5. (5)

   Reads of Lphn3-KO and Lphn3-E31-KO studies of this work were aligned as above. HTSeq was used to count reads. Only genes with more than 350 reads were used for DEseq2 analysis.

6. (6)

   Reads of PBM deletion from the amplicon sequencing dataset were aligned as described above. Paired-read sequences near the edited site were extracted and the length of each read was calculated. Insertions/deletions that caused frameshifts or mutations within the PBM region were classified as PBM-KO events.

7. (7)

   From the scRNA-seq analysis in the primary visual cortex and anterior lateral motor cortex dataset from the Allen Institute⁴², processed read densities are publicly available.

Calculation of exon PSI

PSI is the percentage of reads containing the target exon among all reads at the target region. For alternative exons containing both 5′ and 3′ flanking exons (for example, E24 and E30b), only the reads spanning the target exon–exon junctions were used for the calculation. For alternative exons at the 3′ termini (E31 and E32), all reads containing the target exon were used for calculation as the 3′ termini of Lphn3 ends with either E31 or E32. E31 and E32 reads were normalized to exon length before calculating the PSI.

Sample preparation of Lphn3 KO and Lphn3-E31 KO neurons for next-generation sequencing

Primary hippocampal culture neurons were infected with lentiviruses expressing gRNAs (control, Lphn3 KO, Lphn3-E31-only KO) at day 3 in vitro (DIV3) and maintained until DIV14. One coverslip of culture was resuspended with 200 µl TRIzol on ice, mixed with 50 µl chloroform and incubated at room temperature for 2 min. The samples were centrifuged at 12,000g for 15 min at 4 °C in an Eppendorf 5417C centrifuge. The aqueous layers were added to 100 µl ice-cold isopropanol for thorough mixing before incubation at −80 °C for 1 h. The samples were thawed on ice and centrifuged at 20,817g for 20 min at 4 °C in the Eppendorf 5417C centrifuge. Pellets were washed with 0.5 ml ice cold 75% ethanol before being centrifuged at 20,817g for 10 min at 4 °C, and subsequently resuspended by 40 µl double-distilled H₂O containing 0.2 U µl⁻¹ SUPERase·In RNase inhibitor. Total RNA samples were converted to a library using Illumina Stranded mRNA kit and sequenced in the NovaSeq (paired-end 150 bp) system with 40 million paired reads at Medgenome.

Sample preparation of ΔPBM neurons for amplicon sequencing

Primary hippocampal culture neurons were infected with lentiviruses expressing gRNAs (control or ΔPBM) at DIV3 and maintained until DIV14. Total RNA samples were extracted as above, and converted to cDNA using the PrimeScript RT-PCR Kit. The PBM region of Lphn3 was amplified using the primers 5′-AAACCTGGGCTCCAGAAACC-3′ and 5′-GGAAAGATTGGGGCACAGGA-3′, and converted to a library using Nextera XT adaptor/indexes, and sequenced in the MiSeq system at the Stanford Functional Genomics Facility.

TRUPATH G-protein-coupling assay

HEK293T cells were obtained from ATCC and maintained, passaged and transfected in DMEM medium containing 10% FBS, 100 U ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin (Gibco-ThermoFisher) in a humidified atmosphere at 37 °C and 5% CO₂. After transfection, cells were plated in DMEM containing 1% dialysed FBS, 100 U ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin for BRET assays. Constitutive activity of LPHNs was accomplished by using the previously optimized Gα–Rluc8, β-subunit and N-terminally tagged γ-GFP2 subunit pairs described previously¹⁹. HEK cells were plated in a 12-well plate at a density of 0.3–0.4 × 10⁶ cells per well with DMEM containing 10% FBS, 100 U ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin. Then, 6 h later, cells were transfected with a 1:1:1 ratio of optimized Gα:β:γ pairings at 100 ng and various amounts of receptor (25 ng, 50 ng, 100 ng, 200 ng, 300 ng) using the TransIt-2020 (Mirius Bio) reagent. To establish a baseline for the cells, pcDNA was used at 100 ng and referred to as 0 ng. The next day, cells were removed from the 12-well plate with trypsin and seeded into a 96-well white clear-bottomed plate (Greiner Bio-One) with DMEM containing 1% dialysed FBS at a cell density of 30,000–35,000 cells per well. Cells were incubated overnight to allow for attachment and growth. The next day, the medium was aspirated from the wells. A solution of assay buffer (20 mM HEPES, Hank’s balanced salt solution, pH 7.4) and 5 μM of coelentrazine 400a (Nanolight Technology) was prepared and added to each well. Cells were allowed to equilibrate with the coelentrazine 400a in the dark for 10 min. Corresponding BRET data were collected using a Pherastar FSX Microplate Reading with luminescence emission filers of 395 nm (RLuc8-coelentrzine 400a) and 510 nm (GFP2) and an integration time of 1 s per well. BRET ratios were calculated as the ratio of the GFP2:RLuc8 emission. The constitutive coupling (0 ng) was used as the baseline to subtract NET BRET of the experimental conditions for each receptor. Three independent cultures with seven technical replicates in each culture were used in total.

cAMP reporter assay

HEK293T cells were maintained in DMEM + 10% FBS at 37 °C 5% CO₂, and seeded onto a 24-well plate. During calcium transfection of each well, eGFP (0.23 µg), pink flamindo 2 (0.23 µg), Gα_s (0.16 µg), Gβ (0.16 µg) and Gγ (0.16 µg) were used for all conditions. When indicated, additional constructs were co-transfected including PDE7b (0.23 µg) and six isoforms of Lphn3 (E24⁺E30b⁻E31⁺E32⁻, E24⁺E30b⁺E31⁺E32⁻, E24⁺E30b⁻E31⁻E32⁺, E24⁻E30b⁻E31⁺E32⁻, E24⁻E30b⁺E31⁺E32⁻, E24⁻E30b⁻E31⁻E32⁺, 0.23 µg each). Then, 16 h after transfection, the medium was replaced with 0.5 ml DMEM + 10% FBS. Then, 36–48 h after transfection, medium of all cultures was replaced with 0.5 ml imaging buffer (20 mM Na-HEPES pH 7.4, 1× HBSS (Gibco, 14065056)) and incubated at room temperature for 30 min. When indicated, 2.5 µM forskolin and 5 µM IBMX were added to the culture for 5 min. Imaging was performed under Nikon confocal microscopy under a ×10 objective.

Primary hippocampal neuron culture

Neonatal P0 mice pups of CAG-Cas9 mice (JAX, 024858) were dissected in ice cold HBS to obtain hippocampi, which were digested in 1% (v/v) papain suspension (Worthington) and 0.1 U µl⁻¹ DNase I (Worthington) for 15 min at 37 °C. Hippocampi from two pups were washed with calcium-free 1× HBS (pH 7.3) and dissociated using gentle pipetting in plating medium (MEM containing 5% FBS, 0.6% glucose, 2% Gem21 NeuroPlex Supplement, 2 mM GlutaMAX), filtered through 70 µm cell strainer, and seeded onto Corning Matrigel-coated 12 mm cover glasses in one 24-well plate, and maintained at 37 °C under 5% CO₂. Then, 16 h after seeding (DIV1), 90% of the medium was replaced with maintenance medium (Neurobasal A with 2% Gem21 NeuroPlex Supplement, 2 mM GlutaMAX). At DIV3, 50% of the medium was replaced with fresh maintenance medium supplemented 4 µM Ara-C (cytosine β-d-arabinofuranoside hydrochloride), and lentivirus expressing gRNA. When indicated, lentiviruses expressing jRGECO1a were added at DIV7. At DIV7, 10 and 13, 30% of the medium was replaced with fresh maintenance medium, before analysis at DIV14.

Virus preparation

Lentiviruses were produced in HEK293T cells using the second-generation packaging system. Per 150 cm² of cells, 186 µl 2 M CaCl₂ containing 5.8 µg of lentivirus shuttle vector, 2.5 µg pVSVG (Addgene, 12259) and 4.2 µg Gag-Pol-Rev-Tat (Addgene, 12260) at a total volume of 1.5 ml was added dropwise to an equal volume of 2× HBS (280 mM NaCl, 10 mM KCl, 1.5 mM Na₂HPO₄, 12 mM glucose and 50 mM HEPES, pH 7.11) under constant mixing, incubated for 15 min at room temperature and added dropwise to the cells. Then, 8–12 h after transfection, the medium was replaced with DMEM with 10% FBS. Then, 48 h after transfection, the cell medium was cleared by centrifuging in table-top centrifuge at 2,000g for 3 min, and filtered through a 0.45 µm PES membrane. The viral supernatant was loaded onto a 2 ml 30% sucrose cushion in PBS and centrifuged in the Thermo Fisher Scientific SureSpin 630 rotor at 19,000 rpm for 2 h. The viral pellet was resuspended in 30 µl MEM and flash-frozen in liquid nitrogen. AAVs (CAG-DIO-RG and CAG-DIO-TCB-mCherry) in capsid 2.5 and pseudotyped rabies virus RbV-CVS-N2c-deltaG-GFP (EnvA)⁴³ were prepared at Janelia Farm Viral core facility.

Monosynaptic retrograde rabies tracing

P0 neonatal mouse pups were anaesthetized on ice for 4 min and were head-fixed using ear bars and a 3D-printed mould. Then, 0.35 µl 1 × 10⁹ IU ml⁻¹ lentiviruses (SYN1-gRNA-NLS-cre) was injected unilaterally to CA1 at the coordinates anteroposterior (AP) +0.95 mm, mediolateral (ML) −0.92 mm, dorsoventral (DV) −1.30 mm (zeroed at Lambda). At P21, mice were anaesthetized by avertin (250 mg per kg) and head-fixed on a stereotaxic injection rig, and 0.2 µl AAVs (CAG-DIO-RG, 3.6 × 10¹² genome copies per ml; and CAG-DIO-TCB-mCherry, 6.35 × 10¹² genome copies per ml; 1:1 volume mix) were co-injected to CA1 at coordinates AP −1.80 mm, ML −1.35 mm, DV −1.30 mm (zeroed at bregma). At P35, the same CA1 site was injected with 0.15 µl EnvA-pseudotyped rabies virus RbV-CVS-N2c-deltaG-GFP at 2 × 10⁸ IU ml⁻¹. After the surgery, the incisions of P21 and P35 mice were closed by suture and 3M Vetbond tissue adhesive (1469SB). After all of the injections, the mice were allowed to recover on a heating pad before returning to their home cage. At P42, the mouse brains that had been perfused were fixed in 4% PFA (Electron Microscopy Sciences, EM grade, 15714) in PBS for 4 h at room temperature, subsequently incubated in 30% (w/v) sucrose in PBS at 4 °C overnight and cryopreserved in Tissue-Tek O.C.T. compound (Sakura) on dry ice. Frozen tissue blocks were cut into 20 µm coronal sections on a cryostat and collected on glass slides (Globe Scientific, 1358W). The sections were air dried, stained in 1 µg ml⁻¹ DAPI for 10 min, washed once with PBS and sealed in Fluoromount-G (Southern Biotech, 0100-01). The sections were imaged on the Olympus VS200 slide scanner at ×10. A total of 5–6 mice was used per condition for the study.

RT–PCR analysis of Lphn3 alternative exons in tissues

C57BL/6 mice at P4, P9, P14, P21 and P35 were euthanized and the brains were dissected to isolate the olfactory bulb, cerebellum, hippocampus, prefrontal cortex, striatum and retina. Tissues were grinded with 500 µl TRIzol on ice, mixed with 125 µl chloroform and incubated at room temperature for 2 min. The samples were centrifuged at 12,000g for 15 min at 4 °C in the Eppendorf 5417C centrifuge. The aqueous layers were added to 250 µl ice-cold isopropanol for thorough mixing before incubation on ice for 5 min. The samples were centrifuged at 20,817g for 20 min at 4 °C in the Eppendorf 5417C centrifuge. Pellets were washed with 0.5 ml ice cold 75% ethanol before being centrifuged at 20,817g for 10 min at 4 °C, and subsequently resuspended in 40 µl double-distilled H₂O containing 0.2 U µl⁻¹ SUPERase•In RNase inhibitor. A total of 100 ng of total RNA was used for cDNA conversion using the PrimeScript RT-PCR Kit using random 6-mers. In total, 1 µl cDNA was used for PCR targeting exon–exon junction regions using Ex Taq DNA Polymerase. The following primers were used: Actb (5′-TCTACAATGAGCTGCGTGT-3′, 5′-CGAAGTCTAGAGCAACATAG-3′), Lphn3 E6 (5′-CCACAGCTACTCATCCTCAC-3′, 5′-GCTCTCGATCATGATGACGT-3′), Lphn3 E15 (5′-GGGGACATCACCTACTCTGT-3′, 5′-TCAGGTCTCTCCAGGCATTC-3′), Lphn3 E24 (5′-CCTGAATCAGGCTGTCTTGA-3′, 5′-AAATGGTGAAGAGATACGCC-3′), Lphn3 E31 (5′-TCCAGGACGGTACTCCACA-3′, 5′-GGCATTGTTCAGAAGCCCCT-3′), Lphn3 E32 (5′-TCCAGGACGGTACTCCACA-3′, 5′-TCCTGTGTCCTGTTTCGGGA-3′). The PCR program was as follows: 94 °C for 1 min; then 31 cycles of 94 °C for 30 s, 55 °C for 30 s, 72 °C for 1 min. PCR products were separated on 2% agarose gel in 1× TAE buffer and imaged using the BioRad Gel Imaging system.

RT–qPCR analysis of Lphn3-KO and Lphn3-E31-KO neurons

Total RNA (80 ng) for each culture was used for converting to cDNA using the PrimeScript RT-PCR Kit using random 6-mers. A total of 1 µl cDNA was used for qPCR experiments with the TaqMan Fast Virus 1-Step Master Mix using PrimeTime Std qPCR designed primer-probe sets: Actb: 5′-GACTCATCGTACTCCTGCTTG-3′, 5′-GATTACTGCTCTGGCTCCTAG-3′, /56-FAM/CTGGCCTCA/ZEN/CTGTCCACCTTCC/3IABkFQ/; Lphn3 E27–E31 junction: 5′-CCTTCATCACCGGAGACATAAA-3′, 5′-GTGGTAGAGTATCCATGACACTTG-3′, /56-FAM/CA GCTCAGC/Zen/ATCGCTCAACAGAGA/3IABkFQ/; Lphn3 E27–E32 junction: 5′-CAGTCAGAGTCGTCCTTCATC-3′, 5′-GTCAGTCTCAGGTCCATAAGTC-3′, /56-FAM/AACAGCTCA/Zen/GCATCGCTCAACAGA/3IABkFQ/. The PCR program was as follows: 95 °C for 20 s; then 41 cycles of 95 °C for 3 s, 60 °C for 30 s. C_t values for the Lphn3 E31 and E32 sample were subtracted by that of Actb from the same sample to get ΔC_t. All ΔC_t values were normalized to the control gRNA. A total of eight cultures was used.

RT–PCR analysis of Lphn3-KO, Lphn3-E31-KO and ΔPBM in neurons

Total RNA (80 ng) for each culture was used for converting to cDNA using the PrimeScript RT-PCR Kit with random 6-mers. A total of 1 µl cDNA was used for PCR targeting exon–exon junction regions using Ex Taq DNA Polymerase. The following primers were used: Actb: 5′-TCTACAATGAGCTGCGTGT-3′, 5′-CGAAGTCTAGAGCAACATAG-3′; Lphn3 E31: 5′-GTCAGAGTCGTCCTTCATCAC-3′, 5′-AGTTGTTCACCAGTTTGTTCATC-3′; Lphn3 E32: 5′-CGGATTCGGAGAATGTGGAA-3′, 5′-CCACAGATAACGTGTGTGGT-3′. The expression level of E31 and E32 were normalized to Actb.

Immunoblotting analyses

One well of neuron culture from a 24-well plate was lysed in 50 µl lysis buffer (20 mM Tris pH 7.5, 500 mM NaCl, 1% Triton X-100, 0.1% SDS 1× Roche EDTA-free protease inhibitor) at room temperature for 5 min. A total of 20 µl 5× SDS loading buffer was added and the samples were analysed using SDS–PAGE. Gels were transferred to a 0.2 µm nitrocellulose membrane in the Trans-Blot Turbo Transfer system (Bio-Rad) and blocked by western blocking buffer (5% BSA in 1× TBST) at room temperature for 30 min. Mouse anti-LPHN3 (Santa Cruz Biotech, sc-393576, 1:1,000) and mouse anti-actin (Sigma-Aldrich, A1978, 1:3,000) antibodies in western blocking buffer were added and incubated at 4 °C for overnight. The membranes were washed in western blocking buffer three times for 10 min each, and IRDye 800CW donkey anti-mouse IgG secondary antibodies (Li-cor, 926-32212, 1:20,000) in western blocking buffer were added to the membrane, which was incubated at room temperature for 1 h and washed in 1× TBST three times for 10 min each. The samples were imaged using the Odyssey Imager (Li-Cor). Quantifications of LPHN3 level were normalized to β-actin.

Calcium imaging

Primary culture neurons were maintained as described above, except they were infected with lentiviruses expressing SYN1-gRNA-EGFP at DIV3, and lentiviruses expressing SYN1-jRGECO1a at DIV7. At DIV14, the coverslips containing neurons were washed once with 37 °C warmed Tyrode buffer (25 mM Na-HEPES pH 7.4, 129 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 15 mM glucose and transferred to 12-well glass plate (Cellvis, P12-1.5H-N) in Tyrode buffer. After 30 min of incubation in Tyrode buffer at 37 °C under 5% CO₂, the cultures were imaged under the Leica microscope at 37 °C under 5% CO₂, with 50 ms exposure, 85 ms interval for 1 min for each field of view (FOV). A total of 6–8 fields of view was recorded for each coverglass of culture. For each condition from one batch of culture, 3–5 cover glasses of cultures were imaged. Three batches of culture were used in total.

Immunohistochemistry and synapse puncta imaging

Primary hippocampal neurons were washed in Tyrode buffer (25 mM Na-HEPES pH 7.4, 129 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 15 mM glucose) and fixed in 4% PFA and 4% sucrose in 1× DPBS at 37 °C for 15 min. Neurons were next washed three times with 1× DPBS for 5 min each, and permeabilized in 0.1% Triton X-100 in 1× DPBS for 10 min at room temperature without shaking. After blocking with 0.5% fish skin gelatin in 1× DPBS at 37 °C for 1 h, the culture was stained with chicken anti-MAP2 (Encor, CPCA-MAP2, 1:1,000), guinea pig anti-vGluT1 (Milipore, AB5905, 1:1,000), and rabbit anti-HOMER (Milipore, ABN37, 1:1,000) antibodies in blocking buffer at 4 °C overnight. The samples were washed three time with 1× DPBS for 8 min each, and incubated with secondary antibodies (anti-chicken Alexa 405, anti-guinea pig Alexa 647 and anti-rabbit Alexa 488) in blocking buffer at 37 °C for 1 h. Next, the culture coverslips were washed three times with 1× DPBS for 8 min each, once with double-distilled H₂O briefly, before being loaded onto glass slides (Globe Scientific, 1358W) in Fluoromount-G (Southern Biotech, 0100-01) and sealed in nail polish (Amazon, B000WQ9VNO). The samples were imaged under the Nikon confocal microscope at ×60, with a 0.35 µm step size and 4–6 z stacks. For each coverglass of the culture, about 20 neurons containing well-isolated dendrites were imaged. For each condition of one batch of culture, two cover glasses of the culture were imaged. Three batches of the culture were used in total.

Image analyses

Five types of image analyses were performed.

1. (1)

   Quantification of excitatory synapse puncta density. Maximum-intensity files were produced from z-stacked images. The background was subtracted and the 5–10 well-isolated secondary dendrites were cropped from each neuron in Fiji (v.2.9.0) for processing. Excitatory synapses, especially in mature mushroom spines, are localized ~0.5–1 µm away from the dendrite due to the long neck of the spine³. Our confocal images have an interpixel unit of 0.20714 µm per pixel. Thus, in our analyses, we include vGluT1/HOMER signals within 5 pixels away from dendrite. The cropped files were converted to binary images using the same threshold for the same channel, for the same batch of experiment. For calculating excitatory synapse puncta, the overlapped region of vGluT1 and HOMER binary images were generated, and the overlapped regions containing more than two neighbouring pixels were considered to be puncta, and were searched and quantified using the scikit-image (v.0.20.0) package⁴⁴. To calculate dendrite length, binary MAP2 channel images were skeletonized by scikit-image to a 1 pixel representation of which the length was measured using FilFinder (v 1.7.3) package⁴⁵. For each cropped file, the puncta number divided by dendrite length produced the puncta density. All of the imaged regions from one batch of the experiment were averaged to calculate the puncta density for one condition. Three batches of data were plotted in total.

2. (2)

   Calcium imaging. Time-lapsed videos of calcium imaging files were processed using the CaImAn package⁴⁶ to search for spiking somas and generate corresponding fluorescence intensity (ΔF/F) over time. The key parameters were: decay_time=0.4, p=1, gnb=2, merge_thr=0.85, rf=60, stride_cnmf=6, K=10, gSig=[40,40], method_init=‘greedy_roi’, ssub=1, tsub=1, min_SNR = 200, rval_thr=0.85, cnn_thr=0.99, cnn_lowest=0.1. ΔF/F traces of all detected spiking somas from one field of view were averaged to produce one synchronized firing trace. SciPy (v1.10.1)⁴⁷ algorithm “find_peaks” (height=0.15, width = (2,20), distance=20) was used to detect the spiking number and signal strength (ΔF/F) for each synchronized firing trace. The synchronizing firing rate was calculated by dividing spiking number against total time for each trace. To plot the firing rate (or ΔF/F) for each condition, the median of the firing rate (or ΔF/F) from all traces of one batch was used. Three batches of culture were plotted in total.

3. (3)

   Rabies tracing. Coronal sections corresponding to bregma −1.55 to −2.03 mm⁴⁸ for hippocampal formation and Bregma −3.8 to −4.1 mm⁴⁸ for the LEC were processed in Fiji by background subtraction. Regions of the ipsilateral CA1, ipsilateral CA3, contralateral CA3 and ipisilateral LEC were cropped in Fiji for processing in scikit-image (v.0.20.0). The cropped regions were converted to binary images using the same threshold for the same channel, for the same batch of experiment. Binary regions containing more than 80 neighbouring pixels (red channel for CA1) and 150 neighbouring pixels (green channel for CA3 and LEC) were considered to be neuron soma, and were counted using the scikit-image⁴⁴ functions measure.label and measure.regionprops. All counts from one mouse were used to calculate the connectivity strength of ipsilateral CA3–CA1, contralateral CA3–CA1, and LEC–CA1.

4. (4)

   cAMP imaging using pink flamindo 2. After background subtraction, the 488 and 546 nm channel signals from one field of view was used to calculate pink flamindo 2/GFP. In total, 3–10 fields of view were imaged per condition per batch of culture. Three batches of cultures were used in total.

5. (5)

   Phase-transitioned droplet. After background subtraction, signals from the indicated channels were used for analysis. A 12.86 µm linear region across the diameter of the droplets was used to plot the signal from the edge to the centre of the droplets. To calculate the number of droplets per cluster, contacting droplets were counted as one cluster. The scikit-image (v.0.20.0) package⁴⁴ was used to count the size of droplets. Three independent replicates were used for each experiment.

Protein purification

We used truncated GKAP and SHANK3 to retain essential interaction modules and obtain soluble proteins. 6×His-tagged GKAP, SHANK3, HOMER3 and PSD95 were purified as described previously¹⁵ with slight modifications. Constructs were transformed into BL21 (DE3) pLysS, which were induced at an optical density at 600 nm of 0.6 with 0.25 mM IPTG at 16 °C for 18 h. Cells were lysed in Ni-buffer A (20 mM Tris pH 8, 500 mM NaCl, 5% glycerol, 4 mM BME, 20 mM imidazole, 1× Roche EDTA-free protease inhibitor, 100 U ml⁻¹ benzonase) and cleared at SS34 rotor at 14,000 rpm for 30 min at 4 °C. Proteins were loaded onto the Ni-NTA column, washed in Ni-buffer A and eluted in Ni-buffer B (20 mM Tris pH 8, 250 mM NaCl, 5% glycerol, 4 mM BME, 400 mM imidazole). His-tags were removed by 3C protease. Finally, the proteins were purified in a size-exclusion column (SD75 10/300 for GKAP, and SD200 10/300 for others) in SEC buffer (20 mM Tris pH 8, 300 mM NaCl, 2 mM DTT). Lentiviruses containing CMV-TetO₂-Flag-Lphn3 E31, E32 and E31(ΔPBM) were used to express proteins in FreeStyle 293-F cells at 37 °C under 8% CO₂. Cells were collected 60 h after 5 µg ml⁻¹ doxycycline induction, and lysed in lysis buffer (20 mM Na-HEPES pH 7.4, 500 mM NaCl, 1% DDM, 0.1% CHS, 30% glycerol, 1× Roche EDTA-free protease inhibitor cocktail, 100 U ml⁻¹ benzonase). The lysate was incubated with 2 mg ml⁻¹ iodoacetamide, cleared by centrifugation at SS34 rotor at 16,000 rpm for 30 min. The supernatant was loaded onto anti-Flag M1 Agarose Affinity Gel in wash buffer 1 (20 mM Na-HEPES pH 7.4, 500 mM NaCl, 0.01% LMNG, 0.001% CHS, 2 mM CaCl₂, 20 µM leupeptin). Bound protein was washed with wash buffer 1 and wash buffer 2 (20 mM Na-HEPES pH 7.4, 150 mM NaCl, 0.01% LMNG, 0.001% CHS, 2 mM CaCl₂), and eluted in elution buffer (20 mM Na-HEPES pH 7.4, 150 mM NaCl, 0.01% LMNG, 0.001% CHS, 5 mM EGTA, 0.2 mg ml⁻¹ Flag peptide), and further purified on the SD200 10/300 column in SECL buffer (20 mM Na-HEPES pH 7.4, 150 mM NaCl, 0.002% LMNG, 0.0002% CHS). 6×His-tagged TENM2 and FLRT3 were cloned into the pCMV vector and expressed in Expi293F cells. Then, 4 days after transfection, the medium was collected and loaded onto the Ni-NTA column, washed in Ni-buffer C (20 mM HEPES pH 7.4, 150 mM NaCl, 20 mM imidazole pH 7.6) and eluted in Ni-buffer D (20 mM HEPES pH 7.4, 150 mM NaCl, 250 mM imidazole pH 7.6). His-tags were removed by 3C protease, and the proteins were purified in a size-exclusion column (SD200 10/300) in SEC buffer (20 mM HEPES pH 7.4, 150 mM NaCl).

Fluorescence labelling of proteins

For HOMER, PSD95 and SHANK, proteins were buffer-exchanged to labelling buffer 1 (100 mM NaHCO₃, pH 8.2, 100 mM NaCl) at a final protein concentration of 2–20 µM. NHS-dyes (AAT iFluor NHS-405, AAT iFluor NHS-546, Invitrogen Alexa NHS-647) were added to the protein at 1:1 molar ratio, and the labelling proceeded at room temperature for 1 h. The reaction was quenched by 100 mM Tris pH 8.2. Free dyes were removed using the PD10 desalting column (Cytia). The labelling efficiency was 50–100%. Labelled proteins were mixed with unlabelled proteins so that the labelling efficiency was about 2%, and the sample was concentrated to 200–1,600 µM. The samples were cleared at 14,000 rpm in the Eppendorf 5417C centrifuge for 10 min before freezing in liquid nitrogen. Purified LPHN3 proteins were directly labelled using AAT iFluor NHS-488 as described above.

Phase-transition imaging and sedimentation assay

Unless otherwise indicated, proteins were added to a final concentration of 10 µM GKAP, 10 µM HOMER3, 10 µM PSD95, 10 µM SHANK3, 6 µM LPHN3 E31, 6 µM LPHN3 E32, 6 µM LPHN3 E31(ΔPBM), 10 µM TEN2, 10 µM FLRT3 in assay buffer (20 mM Na-HEPES pH 7.4, 150 mM NaCl, 0.002% LMNG, 0.0002% CHS). Protein mixtures were incubated at room temperature for 10–20 min. For imaging experiments, 5 µl of sample was loaded onto a channelled slide (ibidi, 80666), which was designed with a cover to minimize evaporation of small-volume samples. For the pelleting experiments, 10 µl of sample was centrifuged at 5,000 rpm in the Eppendorf 5417C centrifuge for 5 min. The supernatant was immediately removed and the pellet was resuspended in 2× SDS loading buffer. All of the samples were analysed using SDS–PAGE and stained in Coomassie G-250 blue. We quantified the LPHN3 pellet percentage using the N-terminal domain, which has the same sequence for all three Lphn3 constructs.

FRAP analysis

After phase separation was completed, strong-excitation laser intensities were used to bleach all of the channels of a small area for approximately 10 seconds, after which the fluorescence of the photobleached spot was recorded for 6 min. Recovery traces were fitted with the exponential equation y = ae^−bx + c to extrapolate the t_1/2 = (ln2)/b. Only the FRAP recovery kinetics were interpreted because the recovery percentage is highly sensitive to the duration of photobleaching, which was not precisely controlled in this experiment.

Statistics and reproducibility

Most statistical tests were performed using two-sided t-tests, as indicated. To control for family-wise error during multiple comparisons, two-sided Tukey’s tests were used in parallel and the adjusted P values are summarized in Supplementary Tables 1 and 2, and do not change the conclusions drawn from t-tests in this work. Gene counts from the high-throughput sequencing dataset were analysed using two-sided Wald test of DESeq2 for bulk RNA-seq datasets, and two-sided Wilcoxon rank-sum tests for the scRNA-seq dataset. For all box plots: the lowest datapoint shows the minimum value; the highest datapoint shows the maximum value; the centre line shows the median; and the box limits show the interquartile range (25th to 75th percentile). Representative experiments were repeated independently the following number of times: Fig. 2b, n = 6; Fig. 2e, n = 3; Fig. 3a, n = 3; Fig. 3d,e, n = 5 for control/Lphn3 KO and 6 for E31 KO; Fig. 4b, n = 4; Fig. 4d,f, n = 3; Extended Data Fig. 3a, n = 3; Extended Data Fig. 5a, n = 3; Extended Data Fig. 5e, n = 3 (rows 1 and 2), n = 5 (rows 3 and 4) and n = 6 (row 5); Extended Data Fig. 6a–i, n = 2; Extended Data Fig. 7a, n = 3; Extended Data Fig. 7c, n = 3; Extended Data Fig. 7f, n = 3; Extended Data Fig. 7h, n = 3; Extended Data Fig. 8d,e, n = 3; Extended Data Fig. 8f,g, n = 6; Extended Data Fig. 8h, n = 10; Extended Data Fig. 9d, n = 3; Extended Data Fig. 9f, n = 3.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.