Abstract
At the glutamatergic synapses between rod photoreceptors and ON-type bipolar cells (BCs), neurotransmitter is detected by the postsynaptic metabotropic glutamate receptor mGluR6. This receptor forms trans-synaptic interactions with ELFN1, a presynaptic cell adhesion molecule expressed in rods, and ELFN1 is important for mGluR6 localization at BC dendritic tips. Here, we show that in mice of either sex lacking mGluR6, the presynaptic localization of ELFN1 is disrupted. In rods of mGluR6 null mice, ELFN1 is still restricted to the axon terminal spherules but is only partially colocalized with synapses. The ELFN1 localization defect is rescued by expressing mGluR6-EGFP in ON-BCs. In vitro binding experiments demonstrated that the leucine-rich repeat (LRR) and LRR C-terminal cap (LRRCT) regions of the ELFN1 extracellular domain (ECD) are necessary and sufficient for binding to all of the Group 3 mGluRs, including mGluR6. ELFN1-flag expressed in rods of wild-type mice is correctly localized at presynapses, colocalizing with the postsynaptic marker TRPM1 in the outer plexiform layer. Deletion of the LRRCT domain abolished trafficking of ELFN1-flag to rod spherules, whereas deletion of other parts of the ELFN1 ECD did not prevent axonal trafficking or correct presynaptic localization. Our results demonstrate bidirectional mutual regulation of presynaptic enrichment of ELFN1 and postsynaptic enrichment of mGluR6 at photoreceptor synapses.
Significance Statement
Metabotropic glutamate receptors (mGluRs) play important roles at synapses throughout the central nervous system. The Group 3 mGluRs participate in trans-synaptic interactions with ELFN synaptic adhesion molecules, which regulate synapse formation, mGluR recruitment, and mGluR function. In the retina, mGluR6 detects neurotransmitter at synapses between photoreceptors and depolarizing bipolar cells. Unlike conventional synapses, where postsynaptic ELFN1 interacts with presynaptic mGluRs, ELFN1 is located at presynapses in photoreceptors and interacts with postsynaptic mGluR6. ELFN1 knock-out leads to mGluR6 mislocalization. Here, we show that loss of mGluR6 also disrupts ELFN1 localization. These results demonstrate a bidirectional role of the ELFN1–mGluR6 complex in mediating synaptic enrichment of both parties, which may have broad implications for formation and function of excitatory synapses.
Introduction
In the retina outer plexiform layer (OPL), rod photoreceptors transmit information to rod bipolar cells (BCs) at ribbon synapses. The neurotransmitter glutamate, released in the dark, activates the postsynaptic metabotropic glutamate receptor, mGluR6 (Nakajima et al., 1993; Nomura et al., 1994; Masu et al., 1995; Vardi et al., 2000). Subsequent activation of Go leads to inhibition of the TRPM1 transduction channel (Nawy, 1999; Dhingra et al., 2000; Tian and Kammermeier, 2006; Morgans et al., 2009; Shen et al., 2009). Reduced glutamate release in the light relieves the inhibition, causing channel opening and BC depolarization. Cone photoreceptors relay information to ON-type cone BCs using a very similar mGluR6-dependent mechanism. Additionally, cones form sign-preserving synapses with OFF-type BCs that make use of ionotropic glutamate receptors (reviewed in Hoon et al., 2014).
ELFN1 and ELFN2 are paralogous single-pass transmembrane proteins containing extracellular leucine-rich repeat (LRR) and fibronectin Type 3 domains (Dolan et al., 2007). They are widely expressed at postsynapses throughout the brain, with functions in recruiting presynapses and regulating presynaptic release (Sylwestrak and Ghosh, 2012; Tomioka et al., 2014; Dunn et al., 2019; Stachniak et al., 2019; Matsunaga and Aruga, 2021; Zucca et al., 2025). ELFN1 has been implicated in several neurological diseases and neurodevelopmental disorders (Tomioka et al., 2014; Dursun et al., 2021; Girgenti et al., 2021; Rasheed et al., 2021; Liu et al., 2025). Rod photoreceptors express ELFN1 at presynapses, where it interacts “in trans” with postsynaptic mGluR6 (Cao et al., 2015). In Elfn1 knock-out mice, rod synaptic ultrastructure is disturbed, and mGluR6 enrichment at rod BC dendritic tips is abolished (Cao et al., 2015). At cone synapses, ELFN1 and the related protein ELFN2 cooperate during development, and loss of both proteins also results in loss of mGluR6 at cone ON-BC dendritic tips (Cao et al., 2020). These results have led to a model in which presynaptic ELFN1 is required for mGluR6 postsynaptic localization. Consistent with this, we showed that mGluR6 congenital stationary night blindness (CSNB) mutants that fail to bind ELFN1 also fail to localize to rod BC dendritic tips (Pindwarawala et al., 2024).
While mGluR6 expression is restricted to the retina (Nakajima et al., 1993), the other Group 3 mGluR family members (mGluR4, mGluR7, and mGluR8) are expressed elsewhere in the central nervous system and can also interact with ELFN1 and ELFN2 (Tomioka et al., 2014; Dunn et al., 2018, 2019; Stachniak et al., 2019). However, unlike mGluR6, which is postsynaptic and the primary neurotransmitter receptor, other Group 3 mGluRs are presynaptic and interact with postsynaptic ELFN1 (Nomura et al., 1994; Masu et al., 1995; Sylwestrak and Ghosh, 2012; Tomioka et al., 2014). ELFN proteins appear to play an instigating role in recruiting presynaptic mGluRs. For instance, ELFN1 mediates recruitment of mGluR7-positive presynapses to somatostatin interneurons in the hippocampus (Tomioka et al., 2014) and stabilizes mGluR4 at excitatory presynapses in the nucleus accumbens (Zucca et al., 2025). In cell-based assays, provision of ELFN2 “in trans” increases mGluR4 plasma membrane localization (Dunn et al., 2019).
In this study, we show that presynaptic ELFN1 localization is regulated by postsynaptic mGluR6, in a novel trans-synaptic mutual recruitment arrangement at rod synapses. This interaction is mediated by the LRR and LRR C-terminal domains of ELFN1, which are necessary and sufficient for interaction with mGluR6 as well as other Group 3 mGluRs.
Materials and Methods
DNA constructs
Mouse ELFN1 cDNA was from Dharmacon (Clone ID: 6811341, NP_001371115.1), and mouse ELFN2 cDNA was from Horizon Discovery (Clone ID: 5706857, NP_001345621.1). Full-length mouse ELFN1-flag, ELFN1-Myc, and ELFN2-flag were cloned into pCDNA3.1 for expression in HEK cells and pMOPS for expression in photoreceptors. The pMOPS-EGFP vector was first constructed from pCAG-GFP (Addgene plasmid #11150, a gift from Connie Cepko; Matsuda and Cepko, 2004) by replacing the SpeI/EcoRI fragment (containing CMV enhancer, CAG promoter, and introns) with the minimal murine opsin promoter consisting of nucleotides −218 to +17 of rhodopsin (Pawlyk et al., 2005; Khani et al., 2007), amplified from pTR-MOPS-GFP (Flannery et al., 1997). The EGFP was then replaced by ELFN1-flag using the AgeI/NotI sites. Deletion mutants of ELFN1-flag and ELFN2-flag were constructed using overlap extension PCR and then cloned into pCDNA3.1; ELFN1 mutants were also cloned in pMOPS. Full-length ss-Myc-ELFN1 in pCDNA3.1 was constructed by inserting a Myc tag after the signal peptide predicted using SignalP-5.0 (Almagro Armenteros et al., 2019), between a.a. 25 and 26, using overlap extension PCR with WT and deletion mutant templates. The WT mouse ELFN1 and ELFN2 ECD-Fc constructs, as well as a negative control Fc (HA signal sequence-flag-Fc), were previously described (Miller et al., 2024). Deletion and truncation mutant Fc constructs were made by amplifying from full-length WT or mutant templates and cloning into the same Fc vector to make Fc fusions. To make the ELFN1(1–255-link) and ELFN2 (1–249-link) constructs, we added the flexible linker (GGGSGGGSGGGSGGGGG) in the PCR primer. This linker replaced the GG in the GGAAA sequence between the ELFN1 ECD and the BamHI-XbaI human Fc fragment from Addgene plasmid #59313 (Scheiffele et al., 2000) in our standard ECD-Fc constructs. The ELFN1 ECD-flag construct was made in pCDNA3.1 with WT mouse ELFN1(1–418) followed immediately by a flag tag.
pCDNA3.1 with untagged mouse mGluR6 (NP_775548.2) was described previously (Kang et al., 2014; Agosto et al., 2021). Mouse mGluR4 (clone ID: 6402528, NP_001013403.1), mouse mGluR7 (Clone ID: 6853901, NP_796302.2), and human mGluR8 (clone ID: 7939570, NP_000836.2) were obtained from Horizon Discovery. mGluR4, 7, and 8 were cloned with a C-terminal Myc tag in pCDNA3.1. Untagged WT and N445Q mouse mGluR6 fused to C-terminal EGFP via a flexible linker (GGGSGGG) and cloned into pGrm6P (Agosto et al., 2018a) for specific expression in ON-BCs using the Grm6 promoter (Kim et al., 2008) were previously described (Agosto et al., 2021; Miller et al., 2024).
Cell culture and transfections
HEK293T cells were maintained in DMEM with 10% fetal bovine serum (Corning) and no antibiotics, in a 5% CO2 humidified incubator. For transfection, cells were seeded in 24-well plates or on poly-d-lysine–coated coverslips in 24-well plates. Unless indicated otherwise, cells were transfected with 0.6 µg DNA and 1 µl Lipofectamine 2000 (Invitrogen) per well according to the manufacturer's instructions. For expression of Fc fusion proteins for pull-down experiments, media were replaced with DMEM containing 10% ultra-low IgG fetal bovine serum (Avantor Seradigm) before transfection. For pull-down experiments, cells in normal media were transfected with mGluRs the following day.
ELFN binding assays
Assays were performed as previously described (Miller et al., 2024). Briefly, media were harvested from wells transfected with ELFN ECD-Fc or negative control Fc and supplemented with 1/20 volumes of 1 M Tris pH 7.4 and PMSF. Media were incubated with Protein G Plus agarose beads (Calbiochem) or Protein G Sepharose Fast Flow beads (Cytiva) at room temperature with mixing and then washed in PBS. Cells were harvested from wells transfected with mGluRs, washed in PBS, resuspended in ice-cold lysis buffer [PBS supplemented with 50 mM NaCl, 1% Triton X-100, and 1× complete EDTA-free protease inhibitors (Roche)], incubated on ice for 15 min, and centrifuged at 11,200 × g for 10 min at 4°C. Supernatant was added to the washed Fc-bound beads and incubated for ∼90 min with mixing at 4°C and then washed four times with wash buffer (PBS supplemented with 50 mM NaCl, 1% Triton X-100, and a dash of PMSF). For heteromerization assays, cells were cotransfected with 0.4 µg WT or mutant ELFN1 ECD-Fc and 0.4 µg WT ELFN1 ECD-flag, and cell culture media were harvested and incubated with protein G beads as described above, and then beads were washed in wash buffer.
For pull-down assays with endogenous retinal mGluR6, 8–10 retinas from WT CD1 mice were washed five times in cold PBS and homogenized in PBS with protease inhibitors (Roche) by ∼100 passes through a 23 G needle, followed by ∼40 passes through a 26 G needle. Homogenates were supplemented such that the final composition was the same as lysis buffer, incubated on ice for 30 min with frequent mixing and additional homogenization after 15 min, and centrifuged at 11,200 × g for 10 min at 4°C. Supernatant was added to washed beads bound to ELFN1 ECD-Fc proteins as described above. Binding reactions were incubated for ∼2 h with mixing at 4°C and then washed as above. Both male and female mice were used.
Input, flow-through, and bead samples were mixed with reducing and denaturing sample buffer and loaded on SDS-PAGE gels without heating. Samples were transferred to nitrocellulose, and membranes were blocked in 5% milk in TBST, blotted with mGluR6 mAb 312 (Agosto et al., 2021; 1 µg/ml), anti-Myc (Proteintech #16286-1-AP, rabbit polyclonal; 1 µg/ml), or anti-Flag (Proteintech #66008-4-Ig, mouse monoclonal; 0.5 µg/ml), overnight, and then horseradish peroxidase-conjugated anti-mouse (Jackson ImmunoResearch Laboratories; 0.08 µg/ml) or anti-rabbit (Proteintech; 0.02 µg/ml). For experiments with retina lysates, anti-mouse light-chain–specific secondary antibody (Jackson ImmunoResearch Laboratories; 0.16 µg/ml) was used to avoid detection of endogenous mouse IgG that binds to the protein G beads. Fc fusion proteins were detected with DyLight 680-conjugated anti-human (Thermo Fisher Scientific #SA510138; 0.05–0.5 µg/ml) incubated overnight as a primary antibody. Blots were imaged using SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific) and an Azure 500 digital imager.
HEK cell IF and surface expression measurements
Cells on coverslips were fixed for 10 min in 2% paraformaldehyde (PFA) in PBS, then washed in PBS, and blocked in either PBSA (PBS with 1% BSA) for labeling nonpermeabilized cells or in PBSAT (PBS with 1% BSA and 0.1% Triton X-100) for permeabilizing. Cells were incubated with primary antibody diluted in blocking buffer for 1 h—Myc 9E10 hybridomas obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, and purified previously (Agosto et al., 2021; 2 µg/ml), Myc rabbit polyclonal (Proteintech #16286-1-AP, 1–2 µg/ml), ELFN1 rabbit polyclonal (Synaptic Systems #448003, 5 µg/ml), or anti-Flag mouse mAb (Proteintech #66008-4-Ig, 1 µg/ml)—washed in PBS, incubated with Alexa Fluor dye-conjugated secondary antibodies (Invitrogen) diluted to 2 µg/ml in blocking buffer for 30 min, washed in PBS, and mounted with Prolong Diamond (Life Technologies). Surface expression was quantified from samples labeled in nonpermeabilizing conditions, using total intensity of images after thresholding at 0.03 in Mathematica v13 (Wolfram) as described (Pindwarawala et al., 2024), from 3–6 images per coverslip.
Animals
Protocols were approved by the Dalhousie University Committee on Laboratory Animals, and all procedures were performed in accordance with regulations established by the Canadian Council on Animal Care. CD1 albino mice were from Charles River Laboratories (CD1-Elite #482). The mGluR6 null nob3 mice (Grm6nob3/nob3; Maddox et al., 2008) were originally obtained from Jackson Laboratory (#016883) and back-crossed to CD1 (#022; Charles River Laboratories) five times (Agosto et al., 2021). Age and sex of animals used are shown in Figure S2C.
Subretinal injection and electroporation
Injection and electroporation were performed essentially as described (Matsuda and Cepko, 2004; Agosto and Wensel, 2021). DNA was prepared using a Qiagen maxiprep kit and dissolved in water. P0 mice were anesthetized on ice, the future edge of the eyelid was opened, a pilot hole was made with a 30 G beveled needle, and a 33 G blunt needle connected to a UMP3 microinjection syringe pump and MICRO2T controller (World Precision Instruments) was used to deliver ∼450 nl of injection solution [PBS, 0.1% Fast Green FCF dye (Thermo Fisher Scientific), 2 mg/ml pGrm6P-mGluR6-EGFP plasmid, and 1 mg/ml pGrm6P-DsRed] into the subretinal space. For electroporation of ELFN1 constructs, 2 mg/ml of pMOPS-ELFN1-flag (WT or deletions in full-length ELFN1) and 1 mg/ml pMOPS-EGFP were used. Five 50 ms 80 V pulses, separated by 950 ms intervals, were delivered using tweezer-type electrodes placed over the eyes and an ECM 830 square wave electroporator (BTX Harvard Apparatus). Injected eyes were fixed 4–6 weeks later.
Retina section IF and image analysis
Intact eyes were fixed in 2% PFA in PBS for 20 min, washed extensively in PBS, and cryoprotected in 30% sucrose in PBS at 4°C overnight. Eyecups with lenses were embedded in OCT, and 16 µm cryosections were collected on SuperFrost Plus slides and stored at −20°C. For immunostaining, sections were blocked in PBS with 10% donkey serum, 5% BSA, and 0.2% Triton X-100 for at least 2 h at RT and then incubated with primary antibody diluted in blocking buffer at 4°C overnight. Primary antibodies were mGluR6 mAb 366 (Agosto et al., 2021) 5 µg/ml; TRPM1 mAb 545H5 or 274G7 (Agosto et al., 2018b) 5 µg/ml; PSD95 mAb K28/43 (Rasband and Trimmer, 2001; #75-028 cell culture supernatant obtained from the Developmental Studies Hybridoma Bank, The University of Iowa) diluted 1:40; ELFN1 rabbit polyclonal (Synaptic Systems #448003) 4–5 µg/ml; and anti-Flag rabbit monoclonal (Cell Signaling Technology #14793) 1:500. After washing in PBS, sections were incubated with Alexa Fluor dye-conjugated secondary antibodies, diluted to 8 µg/ml in blocking buffer, for 2 h at RT. For labeling of mGluR6 and TRPM1 on the same slide, clones 366 (IgG1) and 545H5 (IgG2b) were detected using isotype-specific secondary antibodies. Similarly, PSD95 (IgG2a) was used together with TRPM1 274G7 (IgG1). Labeled slides were mounted with Prolong Diamond (Life Technologies).
Images were acquired with a Zeiss LSM880 confocal microscope and 63× oil immersion objective. Unless indicated otherwise, 1,024 × 512 pixel images of single optical sections, showing mainly the OPL and INL, were acquired at zoom 2.6 (50.7 nm/pixel). Total intensity and OPL puncta intensity measurements were obtained using Mathematica v14 (Wolfram) essentially as described (Miller et al., 2024). Briefly, images were rescaled to the average of the 100 lowest and 100 highest pixel values, and background correction was applied by subtracting the mean pixel value of the first row, located in the ONL. Images were masked by manually removing the INL, and ROIs corresponding to OPL puncta were derived from scaled and background-corrected masked images using MorphologicalComponents with method Convex Hull, threshold value of 0.3, and maximum size of <250 pixels. The resulting ROIs were applied to unscaled background-corrected images to obtain the intensity within OPL puncta, which was then divided by either the total intensity of the unscaled unmasked image (TRPM1 channel) or the unscaled masked image (ELFN1 channel). For rescue experiments, the intensity of ELFN1 in OPL puncta was reported without normalization. For rescue experiments with the N445Q mutant, images were cropped to 402 × 512 pixels before analysis, to mitigate the low transfection efficiency observed in these animals.
Experimental design and statistical analysis
GraphPad Prism was used for statistical tests. For comparison of two samples immunostained together on the same slide, paired t tests were used. For comparison of multiple mutants where each mutant was normalized to WT, mutants were compared with the value 1 using one-sample t tests, and p values were multiplied by the number of mutants to correct for multiple comparisons. Tests are described in the figure legends. For experiments with animals, the number of animals used is indicated by individual points in graphs, unless indicated otherwise in the figure legend. Animals used for immunostaining endogenous ELFN1 are described in Figure S2C. Because subretinal injections were performed at P0, the distribution of sex for different plasmid injections was random.
Results
ELFN1 is mislocalized in mGluR6 null retina
Presynaptic ELFN1 in rod photoreceptors is required for normal rod synapses, including postsynaptic enrichment of mGluR6 (Cao et al., 2015, 2020). To determine whether mGluR6 reciprocally affects the presynaptic enrichment of ELFN1, we labeled WT and nob3 retina sections with ELFN1 antibody. The nob3 line is a spontaneous mutant of the Grm6 gene leading to a splicing defect and introduction of an early stop codon, in which no mGluR6 protein is detected in the retina (Maddox et al., 2008; Cao et al., 2009; Agosto et al., 2021; Fig. 1A,B). TRPM1 dendritic tip enrichment was reduced in the nob3 retina, as expected (Cao et al., 2011; Xu et al., 2012; Agosto et al., 2021; Fig. 1B,C). In WT CD1 mice, ELFN1 immunostaining was primarily in OPL puncta, which at the resolution of conventional confocal microscopy, largely colocalized with TRPM1 and mGluR6 at rod BC dendritic tips (Fig. 1A), as reported previously (Cao et al., 2015, 2020).
ELFN1 is mislocalized in the OPL of nob3 retina. WT (A) and nob3 (B) retina sections were labeled with antibodies for mGluR6, ELFN1, and TRPM1. The white box indicates the magnified region shown at the bottom right. Arrowheads in B point to residual colocalization of ELFN1 and TRPM1 in OPL puncta in the nob3 retina. Images were obtained from four biological replicates, and representative images are shown. WT and nob3 images were acquired and processed identically; except where indicated (“brighter”), levels were turned up to highlight localization. C, Quantification of TRPM1 total intensity in the field and ratio of signal in OPL puncta to total. D, Quantification of ELFN1 total intensity in the OPL and ratio of signal in OPL puncta to total OPL. Points are means of at least three images each from different animals, and paired values represent WT and nob3 sections labeled together on the same slide. Error bars indicate mean ± SD of biological replicates. WT and nob3 samples were compared by paired t test, and p values are shown.
In nob3 retinas, ELFN1 was detected in the OPL but at significantly reduced levels (Fig. 1B,D). Furthermore, while partial colocalization with rod synapses was observed (Fig. 1B, arrowheads), ELFN1 immunostaining appeared hazy, with a significant reduction in the fraction of OPL staining present in puncta (Fig. 1B,D). Nevertheless, ELFN1 immunostaining in photoreceptors was largely contained within the PSD95 staining of spherules (Fig. 2A,B). Line profiles through rod spherules show a wide and apparently disorganized distribution of ELFN1, which partially colocalized with TRPM1 (Fig. 2A,B, right; Fig. S1), with a width at 30% of maximum that was significantly larger for nob3 than WT. These results indicate that ELFN1 is successfully trafficked to the rod spherule but is not appropriately concentrated at presynapses.
ELFN1 is mislocalized in the OPL of the nob3 retina. WT (A) and nob3 (B) retina sections were labeled with antibodies for ELFN1, TRPM1, and PSD95. Images were obtained from three animals of each genotype, and representative images are shown. The ELFN1 channel is shown at increased brightness in the nob3 image to better show localization. Example intensity profiles of lines drawn through OPL puncta and spanning the spherule width are shown at the right, and the corresponding lines are shown superimposed on the TRPM1 images. Profiles are shown normalized to minimum and maximum values. C, Quantification of profile width at 30% of maximum. Each point represents a spherule, and 9–12 spherules from 2–3 images each from three animals were combined for analysis. WT and nob3 values were compared by one-way ANOVA and Tukey's posttest: #p < 0.001; n.s., not significant. All line profiles are shown in Figure S1.
Immunostaining with the ELFN1 antibody was also observed in the inner plexiform layer, with no obvious differences between WT and nob3 mice (Fig. S2A,B). No ELFN1 immunostaining was observed elsewhere in photoreceptors for either strain (Fig. S2D). To test whether the ELFN1 antibody cross-reacts with ELFN2, we transfected HEK cells with ELFN1-Myc or ELFN2-flag and labeled both coverslips with ELFN1 and flag antibodies (Fig. S2E). Some cross-reactivity of the ELFN1 antibody with ELFN2 was observed, which may explain the apparent ELFN1 immunostaining at cone synapses (Fig. 1A) and possibly also in the IPL. In adult WT mice, ELFN1 is expected to be present at presynapses in rods, but not cones, while ELFN2 is present only in cones (Cao et al., 2015, 2020). We confirmed that the ELFN1 antibody does not cross-react with mGluR6 in transfected HEK cells (Fig. S2F).
Electroporation of mGluR6-EGFP in the nob3 retina rescues ELFN1 synaptic enrichment
To determine whether introduction of mGluR6 in BCs of postnatal nob3 mice could rescue presynaptic ELFN1 enrichment, we performed in vivo electroporation of plasmid DNA expressing mGluR6-EGFP under control of the ON-BC specific Grm6 promoter (Matsuda and Cepko, 2004; Kim et al., 2008). As we reported previously (Agosto and Wensel, 2021; Agosto et al., 2021), WT mGluR6-EGFP was predominantly located in OPL puncta and rescued the TRPM1 localization defect in nob3 mice (Fig. 3A,B). We also observed a striking rescue of ELFN1 enrichment at OPL synapses containing mGluR6-EGFP (Fig. 3A; Figs. S3, S4). Comparing images containing transfected BCs with images of untransfected regions from the same retina sections, both ELFN1 OPL total intensity and the amount of OPL immunostaining in puncta were significantly increased (Fig. 3C; Figs. S4, S5). Among technical replicate images for each animal, the ELFN1 signal in OPL puncta was positively correlated with the mGluR6-EGFP signal in OPL puncta (Fig. S5C,D). We also observed some degree of ELFN1 rescue with the mGluR6 mutant N445Q, in which the N-linked glycosylation site at N445 is mutated (Fig. S4). We previously showed that this mutant has reduced ELFN1 binding and reduced dendritic tip localization in BCs (Miller et al., 2024). However, the ELFN1 rescue suggests that the mutant can localize to the dendritic tip plasma membrane and still interact productively with ELFN1 in vivo.
Expression of mGluR6-EGFP in nob3 BCs rescues ELFN1 localization. Sections from nob3 retinas electroporated with mGluR6-EGFP using an ON-BC–specific promoter were labeled with antibodies for TRPM1 and ELFN1. A, Example images of fields with and without transfected cells from the same retina labeled on the same slide are shown. Examples of images from two other animals are shown in Figure S3. Control experiments showing that in the absence of antibody labeling no signal is detected in the ELFN1 channel, and that a similar rescue was observed when ELFN1 was detected in a different channel, are shown in Figure S4. B, C, Quantification of TRPM1 signal in puncta, and ELFN1 total and puncta intensity. Points show means of at least three images from different animals, and paired values represent images with (Txf) and without (Untxf) transfected cells from the same animal. Error bars indicate mean ± SD of biological replicates. Txf and Untxf values were compared by paired t test, and p values are shown. Quantification of technical replicate images is shown in Figure S5.
ELFN1 LRR + LRRCT domains are necessary and sufficient for mGluR6 binding
The ELFN1 extracellular domain (ECD; a.a.1–418) contains an LRR domain, LRR C-terminal cap (LRRCT), proline-rich region (PRR), and a fibronectin Type 3 (FN3) domain (Dolan et al., 2007). To identify the domain(s) of the ECD responsible for the trans-synaptic interaction with mGluR6, we constructed deletion mutants of the mouse ELFN1 ECD (Fig. 4A): Δ26–188 (LRR deleted), Δ190–253 (LRRCT deleted), and Δ256–399 (PRR and FN3 deleted). Binding of mGluR6 from transfected HEK cell lysate to WT and deletion mutant ECDs, fused to Fc, was tested with a protein G bead pull-down assay (Fig. 4B–D). The Δ256–399 mutant retained the ability to precipitate mGluR6, while Δ26–188 (ΔLRR) and Δ190–253 (ΔLRRCT) did not, suggesting that LRR and LRRCT domains are required for binding. To identify domain(s) sufficient for binding, we initially made truncation mutants 1–188 (LRR), 1–253 (LRR + LRRCT), and ss-189–253 (signal sequence + LRRCT). Surprisingly, all three of these constructs exhibited very little or no binding. While a faint mGluR6 band could sometimes be observed for 1–253, it was not significantly different from the negative control Fc. The differences between 1–253, which did not bind, and Δ256–399, which was indistinguishable from WT, are the presence of a.a. 254–255 after the LRRCT and a.a. 400–418 at the C-terminal end of the ECD (Fig. 4A). A 1–255 construct behaved identically to 1–253, ruling out the role of the former. We next hypothesized that the reduced binding of 1–255 compared with Δ256–399 was due to proximity of the Fc domain and/or the protein G beads, causing steric hindrance preventing mGluR6 binding. To test this, we constructed 1–255-link, which has an additional 17 a.a. flexible linker, consisting of glycine and serine, inserted between the ELFN1 fragment and the Fc. This construct exhibited robust mGluR6 binding that was about half that of WT. No binding was observed with the negative control Fc, and all mutants were expressed at levels similar to or greater than WT (Fig. 4C). Line profiles of mGluR6 in bead samples showed that whenever binding was detected, it was restricted to the upper band of mGluR6 (Fig. 4D), consistent with our previous finding that ELFN1 only binds the complex N-glycosylated upper band (Miller et al., 2024). Together, these results demonstrate that the LRR + LRRCT region is necessary and sufficient for mGluR6 binding.
The ELFN1 LRR + LRRCT domains are necessary and sufficient for mGluR6 binding. A, Diagram of WT full-length ELFN1, ELFN1 ECD, and mutants. The hashed line indicates the flexible synthetic linker. B, Example pull-down experiments with ELFN1 ECD fused to Fc, and lysate from HEK cells transfected with mGluR6. Input lysate, ∼5× equivalent amount of bead samples, and flow-through (unbound) samples were blotted with mGluR6 mAb 312. The blot with bead samples was also labeled with anti-human for detection of Fc in a different channel. C, Quantification of mGluR6 and Fc fusion proteins in bead samples as shown in B. For mGluR6, dimer bands (between ∼200 and 250 kDa) were included. D, Line profiles of mGluR6 dimer bands drawn as shown by the arrow in B. The left-hand peak represents the upper band. E, Example pull-down experiments with ELFN1 ECD-Fc and mouse retina lysate. Input lysate, ∼4× equivalent amount of bead samples, and flow-through (unbound) samples were blotted as in B. F, Quantification of mGluR6 and Fc fusion proteins in bead samples as shown in E. For mGluR6, the dimer band at ∼250 kDa was used. In C and F, values were normalized to WT on the same blot; each point is from an independent experiment, and error bars indicate means ± SEM. Black symbols above bars: mutants were compared with 1 using one-sample t tests, and p values were multiplied by the number of mutants to correct for multiple comparisons; white symbols superimposed on bars: mutants were compared with the negative control Fc by ANOVA and Dunnett's posttest; *p < 0.05; **p < 0.01; #p < 0.001.
To determine whether the ELFN1 binding determinants are the same for endogenous mGluR6, we performed similar experiments with recombinant ELFN1 ECD-Fc proteins bound to beads and then incubated with retina lysate from WT CD1 mice (Fig. 4E,F). The Δ256–399 and 1–255 link mutants both robustly precipitated retinal mGluR6, depleting it from the flow-through, while the Δ26–188 (ΔLRR) and Δ190–253 (ΔLRRCT) mutants did not. These results indicate that, like with mGluR6 from HEK lysates, the LRR + LRRCT region is necessary and sufficient for binding endogenous mGluR6 from the retina.
Similar results were observed with ELFN2 ECD deletion mutants: Δ250–379 (FN3 deleted) pulled-down mGluR6, but Δ23–183 (LRR deleted) and Δ185–247 (LRRCT deleted) did not (Fig. 5). However, the Δ23–183 mutant of ELFN2 was poorly expressed (Fig. 5B,C), making this result less meaningful. The ELFN2 fragment 1–249 with the 17 a.a. flexible linker described above was sufficient for interaction with mGluR6. As shown for ELFN1 (Fig. 4D), the WT and mutant ELFN2 that bind mGluR6 all pull down predominantly the upper band (Fig. 5D).
The ELFN2 LRR + LRRCT domains are necessary and sufficient for mGluR6 binding. A, Diagram of WT full-length ELFN2, ELFN2 ECD, and mutants. The hashed line indicates the flexible synthetic linker. B, Example pull-down experiments with ELFN2 ECD fused to Fc. Input mGluR6, ∼5× equivalent amount of bead samples, and flow-through (unbound) samples were blotted with mGluR6 mAb 312. The blot with bead samples was also labeled with anti-human for detection of Fc in a different channel. C, Quantification of mGluR6 and Fc fusion proteins in bead samples. For mGluR6, dimer bands (between ∼200 and 250 kDa) were included. Values were normalized to WT on the same blot; each point is from an independent experiment, and error bars indicate means ± SEM. Black symbols above bars: mutants were compared with 1 using one-sample t tests, and p values were multiplied by 5 to correct for multiple comparisons; white symbols superimposed on bars: mutants were compared with the negative control Fc by ANOVA and Dunnett's posttest; *p < 0.05; #p < 0.001. D, Line profiles of mGluR6 dimer bands drawn as shown by the arrow in B. The left-hand peak represents the upper band.
ELFN1 interaction is conserved in Group 3 mGluRs
Like mGluR6, pull-down assays with mGluR4-Myc, mGluR7-Myc, and mGluR8-Myc revealed that the LRR + LRRCT region is necessary and sufficient for binding (Fig. 6). With mGluR7, we tested 1–255 in addition to 1–255-link and observed very similar behavior to that of mGluR6. mGluR7 had barely detectable binding to the 1–255 construct, but adding the 17 a.a. linker in 1–255-link returned mGluR7 binding to normal levels, providing further evidence that proximity of the LRRCT to the Fc or the beads inhibits mGluR binding. Our results demonstrate a conserved protein–protein interaction between ELFN1 and all of the group III mGluRs. However, because different glycosylated forms of mGluR4, mGluR7, or mGluR8 were not resolved in the SDS-PAGE gel, we did not determine whether the requirement for complex glycosylation of mGluR6 is also conserved among the Group 3 mGluRs.
The ELFN1 LRR + LRRCT domains are necessary and sufficient for Group 3 mGluR binding. A, Example pull-down experiments with ELFN1 ECD and myc-tagged mGluR4, mGluR7, and mGluR8. In some panels, irrelevant lanes were cropped out as shown by black outlines. B, Quantification of mGluR (dimer bands) and Fc proteins in bead samples. Values were normalized to WT on the same blot; each point is from an independent experiment, and error bars indicate means ± SEM. Black symbols above bars: mutants were compared with 1 using one-sample t tests, and p values were multiplied by 9 (mGluR7) or 5 (mGluR4 and mGluR8) to correct for multiple comparisons; white symbols superimposed on bars: mutants were compared with the negative control Fc by ANOVA and Dunnett's posttest; **p < 0.01; #p < 0.001.
Deletion of LRRCT, but not LRR, abolishes synaptic localization in WT photoreceptors
We first determined that the deletion mutants, in the context of full-length ELFN1 protein, are expressed and trafficked to the plasma membrane in HEK cells. N-terminally Myc-tagged ELFN1 mutants were labeled in nonpermeabilizing conditions to detect surface protein (Fig. S6A,B). Though surface expression of the deletion mutants was generally reduced compared with WT, all three had easily detectable surface protein. Total expression of all ELFN1-flag constructs expressed in HEK cells was observed in Western blot (Fig. S6C).
To determine which domains of the ELFN1 ECD are required for synaptic localization in photoreceptors, we expressed full-length WT ELFN1-flag and deletion mutants, also in the context of full-length ELFN1, by subretinal injection and electroporation of plasmid DNA. A minimal murine opsin promoter was used to restrict expression to photoreceptors (Pawlyk et al., 2005; Khani et al., 2007). WT ELFN1-flag expressed by electroporation was mostly localized at presynaptic sites in photoreceptors, colocalizing with postsynaptic TRPM1 puncta in the OPL (Fig. 7A; Fig. S7A). The Δ256–399 mutant (PRR and FN3 deleted) was trafficked to the OPL and also appeared to be correctly localized (Fig. 7D; Fig. S7D), though both WT and Δ256–399 ELFN1-flag did appear to be mislocalized in some spherules (Fig. S7A,D). The Δ26–188 mutant (LRR deleted), though detectable in few cells, also appeared to be correctly localized (Fig. 7B; Fig. S7B). Pull-down experiments with WT ELFN1 ECD-flag and mutant ECD-Fc suggest that all of the deletion mutants are able to heteromerize with WT ELFN1 (Fig. S6D,E), which may explain the synaptic localization of Δ26–188. In contrast to the other mutants, Δ190–253 (LRRCT deleted) was severely mislocalized in photoreceptors. However, unlike in the nob3 mice, where endogenous ELFN1 was still trafficked to the OPL (Fig. 2), this mutant was located in the inner segments and somas (Fig. 7C), suggesting that the LRRCT region has other roles in trafficking. We did not observe any examples of EGFP-containing spherules in which there was detectable Δ190–253 protein colocalized with TRPM1 puncta (Fig. 7C; Fig. S7C).
Localization of ELFN1 deletion mutants in rod photoreceptors. A–D, WT mice were electroporated with WT or mutant full-length ELFN1-flag constructs using a photoreceptor-specific promoter, along with an EGFP-expressing plasmid to identify transfected cells. Sections were labeled with antibodies for TRPM1 and flag. Left, z-projections. Right, Gallery of individual spherules of transfected cells cropped from high-magnification images of single optical planes. Images were processed independently to show localization, and intensities should not be compared. Additional spherules are shown in Figure S7.
Discussion
In this study, we show that mGluR6 in rod BC dendritic tips has a role in recruiting or stabilizing presynaptic ELFN1. It was previously reported that in retinas lacking mGluR6, ELFN1 was still targeted to presynapses, albeit at reduced levels, suggesting ELFN1 localization is independent of postsyaptic components (Cao et al., 2022). Consistent with that, we found ELFN1 present at reduced levels in the OPL of nob3 mice. However, we further observed mislocalization of ELFN1 within the OPL, with significantly reduced enrichment at presynapses (Figs. 1, 2). These results suggest that mGluR6 plays a role in recruiting or stabilizing presynaptic ELFN1. This conclusion is further supported by the ability of electroporated mGluR6-EGFP to rescue ELFN1 localization in nob3 mice (Fig. 3). The nob3 mice used in this study were only back-crossed five times to the CD1 background, and it is possible that there are strain differences that contribute to ELFN1 localization. However, the rescue of ELFN1 localization upon expression of mGluR6 in nob3 mice mitigates this concern, since comparison of fields with and without mGluR6 within the same retina confirmed the effect of mGluR6 on ELFN1 localization. Together with previously reported evidence that ELFN1 recruits postsynaptic mGluR6 (Cao et al., 2015, 2022), our results reveal that the trans-synaptic interaction between mGluR6 and ELFN1 is mutually beneficial in stabilizing both partners at the synapse.
Although ELFN1 colocalization with synapses was reduced in the nob3 retinas, some synaptic localization was still observed, likely reflecting an interaction with another synaptic protein. A possible candidate could be the presynaptic Cav auxiliary subunit α2δ4, which has been reported to interact with ELFN1 and appears normally localized in nob3 mice (Wang et al., 2017). However, α2δ4 colocalizes with ribeye at the synaptic ribbons (Lee et al., 2015; Wang et al., 2017; Kerov et al., 2018), which is inconsistent with colocalization with ELFN1 at synaptic puncta, though this does not preclude an ELFN1-α2δ4 interaction during trafficking and/or development. Roles of the ELFN1 cytoplasmic domain at photoreceptor presynapses have yet to be identified. However, the cytoplasmic domain was found to be important for postsynaptic localization in primary cortical neurons (Dunn et al., 2025), suggesting it mediates cytoplasmic interactions that have yet to be identified. Furthermore, mutations associated with epilepsy and other neurological disorders have been identified in the cytoplasmic domain of ELFN1 (Tomioka et al., 2014), pointing to possible functional interactions in the cytoplasm.
ELFN1 may also have additional trans-synaptic interactions that contribute to synaptic enrichment in photoreceptors. For example, it is possible that TRPM1 contributes to the presynaptic enrichment of ELFN1. Like mGluR6, the dendritic tip localization of TRPM1 is reduced in nob3 mice and rescued by mGluR6-EGFP electroporation (Agosto et al., 2021). However, it was reported that ELFN1 localization was normal in Trpm1−/− mice (Cao et al., 2015), and experiments demonstrating binding and allosteric modulation using transfected heterologous cells support a direct interaction between ELFN1 and mGluR6 (Cao et al., 2015; Dunn et al., 2018), consistent with the conclusion that the rescue observed in nob3 mice is mediated by mGluR6. Regardless, our results indicate retrograde recruitment and/or stabilization of presynaptic ELFN1.
The mechanism of trans-synaptic regulation of ELFN1 localization could be structural, preventing ELFN1 diffusion in the plasma membrane and/or preventing ELFN1 internalization. The ELFN1 immunostaining within the spherules of nob3 photoreceptors suggests the presence of an intracellular pool, supporting the latter function. Furthermore, the presence of ELFN1 exclusively in the spherules suggests that there is a mechanism mediating anterograde axonal trafficking that is distinct from actual synaptic enrichment. The ability of electroporated mGluR6-EGFP to rescue ELFN1 localization in trans confirms plasma membrane insertion of the exogenous mGluR6 at BC dendritic tips. The ELFN1 rescue could therefore be used as an in vivo assay for surface expression of mGluR6 mutants that are competent for ELFN1 binding. For example, we previously showed that a C-terminal truncation mutant of mGluR6, in which the entire cytoplasmic domain was deleted, appeared to have normal dendritic tip localization, but we were not able to confirm surface expression at the synapse (Agosto et al., 2021).
Because the transfection efficiency using electroporation is very low, we did not attempt to measure functional rescue of rod synaptic transmission in the nob3 mice. However, it was previously reported that expression of mGluR6 in a mGluR6 null mouse line using adeno-associated virus did not rescue synaptic transmission to ON-BCs measured by electroretinography (Varin et al., 2021). Possibly the transduction efficiency was still too low, or alternatively it is possible that by P15, when the virus was injected, some other necessary component or synapse structure may have been irreversibly lost. In Elfn1 knock-out mice, restoration of ELFN1 in 2-month-old animals rescued mGluR6 localization and synaptic function, indicating considerable plasticity at this synapse, at least on the postsynaptic side (Cao et al., 2022). However, the window during which endogenous ELFN1 can be rescued by restoring mGluR6 in mGluR6-deficient retina is unknown. Our results suggest that in CSNB cases with mGluR6 variants that are mislocalized and unable to bind ELFN1 (Pindwarawala et al., 2024), presynaptic ELFN1 enrichment is likely also affected. Determining if and when synapse function can be rescued in mGluR6 null or loss-of-function retina will be important for understanding the development of the synapse and possible window for treatment of CSNB. Conversely, in cases of photoreceptor degeneration, the remodeling of BCs is severe at late stages, including loss of mGluR6 and loss of BC dendrites (Jones et al., 2012). The participation of postsynaptic mGluR6 in presynapse integrity has implications for attempts to form functional connections with new photoreceptors introduced by transplantation (Gasparini et al., 2019).
Our results raise the possibility that a similar bidirectional mechanism could occur at trans-synaptic interactions between ELFN1 and other Group 3 mGluRs. This would provide an elegant general mechanism by which both presynaptic and postsynaptic cells could modulate synapse function by regulating expression or trafficking of their part of the trans-synaptic complex. At rod synapses, ELFN1 and mGluR6 have unusual localizations at pre- and postsynapses, respectively. Determining whether presynaptic mGluR4, mGluR7, and mGluR8 have similar roles in postsynaptic ELFN1 localization at other central synapses is an important future direction.
In vitro binding experiments demonstrated that the LRR + LRRCT domains of ELFN1 are necessary and sufficient for binding to Group 3 mGluRs. These experiments were done with deletion and truncation mutants of the ELFN1 soluble ECD fused to Fc, for consistency with previous literature and ease of experimentation. Our results conflict with the result from a previous study that found that the FN3 domain of ELFN1 is required for binding mGluR6, based on a truncated ECD-Fc construct lacking this domain (Dunn et al., 2018). The lack of mGluR6 interaction with this mutant may have been due to proximity of the Fc fusion, consistent with the lack of binding to our 1–255 construct, which was rescued by addition of a synthetic linker separating it from the Fc domain (Fig. 4). Similarly, it is possible that the negative results we observed with the 1–188 and 189–253 fragments could be due to proximity or steric hindrance of the Fc; these fragments should also be tested with the synthetic linker to rule that out. Binding determinants may also be different in the context of the physiological trans-synaptic interaction. In particular, the constraints imposed on both proteins by being anchored in different lipid bilayers could affect the binding interaction.
We observed that the ELFN1-flag Δ26–189 mutant (LRR deleted in the full-length protein) appears to be correctly localized at PR presynapses, while the Δ190–253 mutant (LRRCT deleted) is not (Fig. 7). Possible interpretations of this finding are that the LRR domain is not required for the trans-synaptic interaction of full-length ELFN1 with mGluR6 in vivo, or that interaction mediated by this domain may not be required for synaptic targeting of ELFN1 in vivo. Alternatively, synaptic localization of Δ26–189 may be due to dimerization with endogenous WT ELFN1 or interaction with another synaptic protein. The observation that the deletion mutants can heteromerize with WT, at least in the context of isolated ECDs, supports this idea. The Δ190–253 mutant (LRRCT deleted) was located in the inner segments and outer nuclear layer, unlike endogenous ELFN1 in nob3 mice, which was mislocalized but still within the confines of the spherules in the OPL. This suggests that deletion of the LRRCT results in a trafficking defect in photoreceptors, such that the mutant may not encounter mGluR6.
This study highlights the multifunctional role of mGluR6 at photoreceptor-BC synapses and the importance of the trans-synaptic interaction with ELFN1. We demonstrate that this interaction affects the rod presynapse, in addition to its known importance for the postsynapse, revealing bidirectional regulation of synaptic components.
Footnotes
This work was supported by grants from the Natural Sciences and Engineering Research Council of Canada (RGPIN-2022-02982 to M.A.A.) and Alcon Research Institute (to M.A.A.), by the Canada Foundation for Innovation John R. Evans Leaders Fund, and by Research Nova Scotia. R.V.M. was partially supported by a Dr. R Evatt and Rita Mathers Trainee Scholarship.
The authors declare no competing financial interests.
This paper contains supplemental material available at: https://doi.org/10.1523/JNEUROSCI.0785-25.2025
- Correspondence should be addressed to Melina Agosto at melina.agosto{at}dal.ca.
