A non-conducting role of the Ca v 1.4 Ca 2+ channel drives homeostatic plasticity at the cone photoreceptor synapse

In congenital stationary night blindness type 2 (CSNB2), a disorder involving the Cav1.4 (L-type) Ca2+ channel, visual impairment is mild considering that Cav1.4 mediates synaptic release from rod and cone photoreceptors. Here, we addressed this conundrum using a Cav1.4 knockout (KO) mouse and a knock-in (G369i KI) mouse expressing a non-conducting Cav1.4. Surprisingly, Cav3 (T-type) Ca2+ currents were detected in cones of G369i KI mice and Cav1.4 KO mice but not in cones of wild-type mouse, ground squirrel, and macaque retina. Whereas Cav1.4 KO mice are blind, G369i KI mice exhibit normal photopic (i.e., cone-mediated) visual behavior. Cone synapses, which fail to form in Cav1.4 KO mice, are present, albeit enlarged, and with some errors in postsynaptic wiring in G369i KI mice. While Cav1.4 KO mice lack evidence of cone synaptic responses, electrophysiological recordings in G369i KI mice revealed nominal transmission from cones to horizontal cells and bipolar cells. In CSNB2, we propose that Cav3 channels maintain cone synaptic output provided that the nonconducting role of Cav1.4 in cone synaptogenesis remains intact. Our findings reveal an unexpected form of homeostatic plasticity that relies on a non-canonical role of an ion channel.


INTRODUCTION
At the first synapse in the visual pathway, the light-dependent graded electrical signals produced in rod and cone photoreceptors gates the release of glutamate onto postsynaptic neurons.To accomplish this task, photoreceptor synapses are specialized with a ribbon organelle, which helps prime synaptic vesicles 1,2 and postsynaptic dendrites from horizontal and bipolar cells that invaginate deep within the terminal 3 .A variety of proteins interact with the ribbon and synaptic vesicles near release sites (i.e., active zones) 4 .The importance of these proteins for vision is illustrated by the numerous inherited retinal diseases linked to mutations in their encoding genes 5 .
One such gene is CACNA1F, which encodes the voltage-gated Ca 2+ (Cav) channel expressed in retinal photoreceptors, Cav1.4 [6][7][8] .Among the sub-family of Cav1.xL-type channels, Cav1.4 exhibits unusually slow inactivation that is well-matched for supporting the tonic, Ca 2+dependent release of glutamate from photoreceptor synaptic terminals in darkness 9,10 .More than 200 mutations in CACNA1F cause vision disorders including congenital stationary night blindness type 2 (CSNB2) 11,12 .These mutations are broadly categorized as producing a gain of function or loss of function in Cav1.4 13 .How these mutations in CACNA1F lead to the variable clinical phenotypes of CSNB2 is largely unknown.Symptoms may include strabismus, low visual acuity, and in many cases, night blindness 14,15 .The latter suggests a primary defect in rod pathways, which is surprising given that knockout (KO) mice are completely blind and lack any evidence of either rod or cone synaptic responses 6,7,16 .A major caveat is that rod and cone synapses fail to form in Cav1.4 KO mice 7,[16][17][18] .Thus, Cav1.4 KO mice are not suitable for studies of how CACNA1F mutations differentially affect rod and cone pathways or for efforts to uncover how the biophysical properties of Cav1.4 shape photoreceptor synaptic release properties.
Here, we overcome this hurdle with a knock-in mouse strain (G369i KI) expressing a nonconducting mutant form of Cav1.4 19 .We show that cone ribbon synapses in G369i KI mice are largely preserved and that downstream signaling through cone pathways, although greatly impaired, can support visual function.This novel mechanism requires the ability of the Cav1.4 protein, independent of its Ca 2+ conductance, to nucleate the assembly of cone ribbon synapses and involves an aberrant Cav3 (T-type) conductance that appears when Cav1.4 Ca 2+ signals are compromised.

Ca 2+ currents in cones are mediated by Cav3 channels upon Cav1.4 loss-of-function
A prevailing yet unsupported hypothesis regarding the relatively mild visual phenotypes in CSNB2 is that additional Cav subtypes may compensate for Cav1.4 loss of function in cones.If so, then Ca 2+ currents (ICa) mediated by these subtypes should be evident in cones of Cav1.4 KO and G369i KI mice.The G369i mutation is an insertion of a glycine residue in a transmembrane domain, which prevents Ca 2+ permeation through the channel 19 .Rods of G369i KI mice lack any evidence of ICa, despite the normal presynaptic clustering of the mutant channel 19 .To test if this might differ in G369i KI cones, we first analyzed the localization of the mutant G369i Cav1.4 channels in cones by immunofluorescence with antibodies against Cav1.4,as well as cone arrestin (CAR) and CtBP2 to label cone terminals (i.e., pedicles) and ribbons, respectively (Fig. 1).In Cav1.4 KO mouse retinas, cone pedicles were shrunken and retracted into the outer nuclear layer (ONL, Fig. 1a) and had malformed ribbons (Fig. 1b).In contrast, cone pedicles in G369i KI mice were normally localized in the outer plexiform layer (OPL, Fig. 1a) and were populated by multiple ribbons (Fig. 1b).As in WT mice, labeling for Cav1.4 was clustered near elongated ribbons in cones of G369i KI mice (Fig. 1b).Thus, unlike in Cav1.4 KO mice, the mutant Cav1.4     protein is normally localized and supports the integrity of cone pedicles and ribbons in G369i KI mice.
We next compared patch clamp recordings of cones in retinal slices of adult WT, G369i KI, and Cav1.4 KO mice under conditions designed to isolate ICa (Fig. 2).In WT cones, there was a large, sustained ICa that activated around -50 mV and peaked near -20 mV, consistent with the properties of Cav1.4 (Fig. 2a-d).A small-amplitude ICa was detected in cones of G369i KI and Cav1.4 KO mice but activated at significantly more negative voltages than in WT cones (Fig. 2ad; Table 1).Moreover, ICa in the G369i KI and Cav1.4 KO cones activated around -60 mV and peaked near -35 mV.This aberrant ICa was not sustained as in WT cones but inactivated rapidly during 50-ms (Fig. 2b) and 500-ms (Fig. 2e) step depolarizations.Overlay of the conductancevoltage (Fig. 2d) and inactivation curves (Fig. 2f) revealed a sizeable window current.These features of ICa in G369i KI and Cav1.4 KO cones resembled those of Cav3 T-type channels rather than Cav1.4 20 .
Although Cav3 channels were reported in patch clamp recordings of cone pedicles in WT mouse cones 21 , we did not observe a low voltage-activated component of ICa in the I-V curve from our recordings of WT mouse cone somas, which would be indicative of a Cav3 subtype (Fig. 2a,c).Moreover, ICa in WT cones was blunted by the Cav1 antagonist isradipine (Fig. 3a) but not the Cav3 antagonist, ML 218 (Fig. 3b; ML 218 caused a minor suppression of ICa in some WT cones, which could be attributed to weak activity on Cav1.4,Supp.Fig.S1).In contrast, ICa in G369i KI cones was significantly suppressed by ML 218 but not by isradipine (Fig. 3a,b).To further test for a Cav3 contribution to ICa, we analyzed cones of ground squirrel retina, where the large amplitude of ICa facilitates pharmacological analyses.Consistent with its actions on Cav1.4 in WT mouse cones (Fig. 3c) and transfected HEK293T cells (Supp.Fig.S1), ML 218 caused an insignificant inhibition of peak ICa (-7.0 ± 20.8%, n = 6 cones) as well as a negative shift in the voltagedependence of activation in ground squirrel cones (Supp.Fig.S2).As a positive control, we confirmed that ML 218 blocked a prominent Cav3-type current in ground squirrel type 3a OFF cone bipolar cells (Supp.Fig.S2).Application of isradipine followed by ML 218 resulted in a timeand voltage-dependent suppression of the residual ICa, which is a hallmark of Cav1 inhibition by dihyropyridine antagonists such as isradipine (Fig. 3d, Supp.Fig.S2) 22 .Finally, we also performed patch clamp recordings of cone terminals in macaque retina where there was no evidence of a Cav3 current (Supp.Fig.S3).We conclude that Cav3 channels contribute significantly to ICa in cones only when Cav1.4 Ca 2+ signals are absent.

Cone synaptogenesis relies on the Cav1.4 protein but not its Ca 2+ conductance
As shown in previous studies 7, [16][17][18] , CtBP2 labeled stubby, sphere-like structures resembling immature ribbon material in Ca v 1.4 KO mice (Fig. 1).The presence of elongated ribbons in G369i KI mice suggested that, as in rods 19 , the non-conducting mutant Cav1.4 protein may support the molecular assembly of cone synapses.To this end, we analyzed cone synapses by immunofluorescence and confocal microscopy.Along with the major constituents of the ribbon, CtBP2 and RIBEYE 23 , presynaptic proteins such as bassoon and members of the postsynaptic signaling complex in depolarizing (ON) cone bipolar cells (GPR179, mGluR6, and TRPM1 24 ) were enriched near cone ribbons in G369i KI mice as in WT mice (Fig. 4a).Compared to WT mice, the labeled structures occupied a larger volume (Fig. 4b,c), albeit at a much lower density (Fig. 4d), which increased linearly with the volume of the pedicle in G369i KI mice (Fig. 4e-i), perhaps as a homeostatic response to Cav1.4 loss-of-function.
The cone synapse is structurally complex, with the dendritic tips of two horizontal cells and an intervening ON cone bipolar cell invaginating deeply into the pedicle near the ribbon 3 .To test how the switch in Cav subtypes might affect this arrangement of postsynaptic partners, we generated 3D reconstructions of cone synapses by serial block-face scanning electron microscopy (SBFSEM; Fig. 5).As with the enlarged synaptic contacts (Fig. 4), there was some evidence of structural modifications in G369i KI pedicles.Compared to WT pedicles, ribbons appeared disorganized in G369i KI pedicles which extended telodendria laterally rather than apically (Fig. 5a).In addition, a slightly larger fraction of synaptic sites in the G369i KI pedicle (35% vs 14% in WT) formed incorrect postsynaptic partnerships (Fig. 5b-d'; Table 2).
Nevertheless, the number of ribbons is normal in G369i KI cones and about half of the ribbons making invaginating contacts with the appropriate cell types (i.e., both horizontal cells and cone bipolar cells; Table 2).Therefore, while necessary for the structural refinement of the cone synapses, Cav1.4 Ca 2+ signals are largely dispensable for cone synapse assembly.

Cone signaling to a postsynaptic partner is intact in G369i KI mice
The preservation of cone synapses in G369i KI mice allowed the unique opportunity to test whether a Ca v subtype other than Ca v 1.4 could support ribbon-mediated synaptic release.To this end, we compared synaptic transmission between cones and horizontal cells (HCs, Fig. 6) in retinal slices of WT and G369i KI mice.In darkness, glutamate released from cones depolarizes horizontal cells via activation of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic (AMPA)/kainate receptors 25,26 .The resulting excitatory postsynaptic current (EPSC) undergoes a decline in response to light stimuli that hyperpolarize cones 27 .In WT HCs, a 1 s light pulse (=410 nm) inhibited the standing EPSC, which is reflected as an outward (hyperpolarizing "ON") current (Fig. 6a).Upon termination of the light pulse, an inward (depolarizing "OFF") current signaled the resumption of the EPSC (Fig. 6a).Both the ON and OFF components of the EPSCs in WT HCs increased with light intensity, reflecting the impact of luminance on the presynaptic membrane potential of cones and the subsequent change in glutamate release from their terminals (Fig. 6b).The ON and OFF components of EPSCs in G369i KI HCs were measurable, although lower in amplitude than in WT (Fig. 6a,b).The EPSC in darkness and the light response were abolished by the AMPA/kainate receptor antagonist, DNQX, in both WT and G369i KI horizontal cells (Fig. 6c,d).Thus, while greatly impaired, cone synapses can transmit light information to postsynaptic partners in the retina of G369i KI mice.

Light responses of bipolar cells and visual behavior is spared in G369i KI but not Cav1.4 KO mice
While horizontal cells act as inhibitory interneurons that modulate photoreceptor output, the vertical dissemination of visual information from cones to the inner retina is relayed by glutamate to ON and OFF cone bipolar cells (CBCs) which express mGluR6 and AMPA/kainate receptors, respectively.To test whether cone synaptic transmission to CBCs is intact in G369i KI mice, we recorded electroretinograms (ERGs) under light-adapted conditions using Cav1.4KO mice as a negative control (Fig. 7a-d).In these recordings, the light-induced response of photoreceptors and the postsynaptic response of ON CBCs corresponds to the a-and b-waves, respectively.Unlike in Cav1.4 KO mice, the a-waves of G369i KI mice were like those in WT mice, which indicates that cones do not degenerate in this mouse strain.While reduced in amplitude, the b-wave was measurable in G369i KI mice and significantly larger than in Cav1.4 KO mice at the highest light intensities (Fig. 7a,b).We also recorded flicker ERGs using 10 Hz light stimuli that can isolate cone pathways involving both ON and OFF CBCs.In WT mice, flicker responses exhibited two peaks, one at a lower irradiance (-2 log cd•s/m 2 ) and one at higher irradiance (0.5 log cd•s/m 2 ).The peak at the lower irradiance is attributed to responses in both rod and cone pathways, and the peak at higher irradiance is attributed to responses exclusively in cone pathways (Fig. 7c,d) 28,29 .While significantly lower in amplitude than in WT mice, flicker responses in G369i KI mice showed a similar non-monotonic relation.Compared to Cav1.4 KO mice, G369i KI mice showed peak flicker responses that did not differ at -2 log cd•s/m 2 , but were significantly higher at 0.5 log cd•s/m 2 (Fig. 7d).The inverted flicker responses at higher illuminations in G369i KI mice (Fig. 7c) were absent in Cav1.4 KO mice and may result from the hyperpolarizing contribution of cone-to-OFF CBC transmission.These results suggest that cone-to-CBC signaling is intact in G369i KI mice.
To validate these results with respect to vision-guided behavior, we used a swim test that assesses the ability of mice to identify a visible platform 30 .In darkness, G369i KI and Cav1.4 KO mice took significantly longer to find the platform compared to WT mice (Fig. 7e,f).These results are consistent with our flicker response assays (Fig. 7c,d) as well as the absence of ICa in rods and rod-to-rod bipolar cell synaptic transmission in G369i KI mice 19 .In daylight conditions, G369i KI mice but not Cav1.4KO mice performed as well as WT mice (Fig. 7e,f).Thus, G369i KI mice are unique in retaining visual function under photopic but not scotopic conditions.Collectively, our results suggest that Cav3 channels can support cone synaptic responses and visual behavior in G369i KI but not Cav1.4KO mice.

DISCUSSION
The nervous system has remarkable abilities to adapt to pathological perturbations in neuronal activity.In the retina, ablation or degeneration of photoreceptors triggers various postsynaptic mechanisms that maintain some level of visual function in rod or cone pathways.
These include remodeling of bipolar cell dendrites and their synapses [31][32][33] as well as changes in the sensitivity of bipolar cells to photoreceptor input and inhibitory modulation 33,34 .To our knowledge, this study provides the first evidence for a presynaptic form of homeostatic plasticity that originates within photoreceptors.Using Cav1.4 KO and G369i KI mice, we identify the upregulation of a Cav3 conductance as a common response to Cav1.4 loss-of-function in cones.
However, Cav3 channels can only compensate for Cav1.4 loss-of-function when cone synapse structure is maintained.Thus, our results also highlight a crucial, non-conducting role for the Cav1.4 protein that allows cone synapses to function in the absence of Cav1.4 Ca 2+ signals.

A non-canonical role for Cav1.4 in regulating cone synapse assembly
A major finding of our study is that cone synapse formation requires the Cav1.4 protein but not Cav-mediated Ca 2+ influx.As shown for rod synapses 19 , ribbons and other components of the pre-and post-synaptic complex assemble normally at cone synapses in G369i mice (Figs.1,4,5).
Cav3 Ca 2+ signals are dispensable for this process since their presence in Cav1.4 KO cones is not accompanied by any semblance of ribbon synapses (Fig. 1B).An intimate relationship between Cav1.4 and ribbons is supported by the colocalization of Cav1.4 and RIBEYE puncta resembling ribbon precursor spheres in the developing OPL 7 .Moreover, light adaptation, which decreases the size of the ribbon, leads to a reduction in Cav1.4 labeling in mouse retina 2 .While evidence for the binding of RIBEYE to Cav1.4 is lacking, such an interaction could support the oligomerization of RIBEYE A and B domain within a macromolecular complex 35 .Alternatively, Cav1.4 could pioneer sites of ribbon assembly perhaps by serving as a docking or nucleation site for the active zone.Regardless of the mechanism, our findings show that the formation of photoreceptor synaptic complexes does not require Ca 2+ influx through Cav1.4 channels.
Despite a normal molecular organization, G369i KI cone synapses were enlarged and made errors in postsynaptic partner selection.While I-V curves predict similar peak I Ca amplitudes in cones of WT and G369i KI mice near the membrane potential of cones in darkness (-45 to -50 mV 36 , Fig. 1c), the strong inactivation of Cav3 channels predicts greatly diminished Ca 2+ signals in G369i KI pedicles.Paradoxically, reductions in presynaptic Ca 2+ in rod photoreceptors are thought to cause illumination-dependent shrinkage of ribbons in vivo and in vitro 2,37,38 .However, an inverse correlation between presynaptic Ca 2+ influx and ribbon synapse size is seen in sensory hair cells in zebrafish 39 .The mechanism involves decreased mitochondrial Ca 2+ uptake and increase in the redox state of nicotinamide adenine dinucleotide (NAD + /NADH ratio) 40 .If a similar mechanism applied in cones, Cav3 Ca 2+ signals in G369i KI cones may decay too quickly to enable mitochondrial Ca 2+ uptake mechanisms that trim ribbons.The size of ribbons and the postsynaptic specialization increased linearly with pedicle volume in G369i KI but not in WT mice (Fig. 4e-i), which could represent a form of homeostatic synaptic scaling as has been demonstrated at the neuromuscular junction 41,42 .Similarly, the presence of some noninvaginating contacts with incorrect partner pairings in G369i KI mice (Fig. 5, Table 2) could represent a compensatory response to weakened synaptic output.Considering that only a subset of CBC subtypes re-wire correctly following partial ablation of cones in immature mice 32 , the structurally normal cone synapses of G369i KI mice could involve contacts with CBCs that are unusually resilient to loss of presynaptic input.
While it is well-established that Cav1.4 is the main Cav subtype in mouse cones, Cav3 channels, in particular Cav3.2, have been reported to be expressed in mouse cones by electrophysiology 21 and single cell RNA-seq studies 21,43,44 .By pharmacological and other criteria, we found no evidence for a functional contribution of Cav3 in our recordings of cones in WT mice (Figs.2,3), ground squirrels, or macaque (Supp.Figs.S2,S3).A caveat of using dihydropyridine antagonists such as isradipine to isolate the contribution of Cav1 from other Cav subtypes is the relatively low sensitivity of Cav1 channels to these drugs and their strong voltagedependence.At a holding voltage of -90 mV, isradipine at 1 M causes only ~80% inhibition of Ca v 1.4 with greater block at depolarized voltages 22 , which agrees with our recordings of cones in WT mice (Fig. 3a) and ground squirrels (Fig. 3d, Supp.Fig.S2).Moreover, Cav3 blockers such as Z944 45 and ML218 (Supp.Consistent with the diminutive Cav3-mediated ICa, light responses of HCs were evident but smaller in G369i KI than in WT mice.Due to their slow activation and strong inactivation 46 , Cav3 channels may have a reduced ability to fuel the Ca 2+ nanodomains that support fast and sustained components of release, both of which occur only at ribbon sites in cones 1 47 .Cav3 channels may also be located further from the ribbon than Cav1.4,thus lowering the efficiency of coupling to exocytosis.Unfortunately, it was not possible to test this with commercially available antibodies against Cav3.2, which yielded identical patterns of immunofluorescence in the OPL of WT and Cav3.2KO mice (data not shown).A detailed analysis of the Cav3 subtype(s) and their subcellular localization in G369i KI cones is required to unravel the shortcomings of Cav3 channels with respect to cone synaptic release.
Based on extremely heterogeneous clinical presentations, CSNB2 manifests as a spectrum of visual disorders that originate from various mutations in CACNA1F 14,48 .Even though some CSNB2 mutant Cav1.4 channels may traffic normally to the plasma membrane, many of these mutations are expected to produce non-functional, non-conducting Cav1.4 channels. 49,50 the visual phenotypes of CSNB2 patients are not as severe as the complete blindness in Cav1.4 KO mice, which lack any Cav1.4protein expression and exhibit no signs of visual behavior (Fig. 7e,f) 51 .Collectively, our results suggest that G369i KI mice accurately model Ca v 1.4 channelopathies in CSNB2 patients that are characterized by a greater impairment in rod than in cone pathways.This interpretation is supported by our findings that G369i KI mice exhibit horizontal cell responses to bright but not dim illumination (Fig. 6), ERG responses under conditions of light adaptation (Fig. 7a-d) but not dark-adaptation 19 , and visual behavior under photopic but not scotopic conditions (Fig. 7d).Together with the enlargement of synaptic sites, modest levels of synaptic release from cones of G369i KI mice may be sufficient to support nominal transmission of visual information through cone pathways.Furthermore, we acknowledge that G369i KI mice could also exhibit homeostatic alterations in the inner retina, which are known to support visual function when photoreceptor input is severely compromised 32- 34 .Future studies of the retinal circuitry and visual behavior of G369i KI mice could identify compensatory pathways that are recruited upon Cav1.4 loss-of-function and how they might be targeted in novel therapies for CSNB2 and related disorders.

DECLARATION OF INTERESTS
The authors declare no competing interests.(25 th , 75 th quartiles).Vh and k were determined from Boltzmann fits of the G-V and steady-state inactivation curves.Time constant (tau) for activation was obtained from exponential fit of the rising phase of ICa evoked by a 50-ms test pulse to a voltage near the peak of the I-V.Tau deactivation was determined from exponential fit of the decay of the tail current evoked by repolarization to -90 mV from +20 mV.CM, membrane capacitance; RM, input resistance.P-values were determined by Kruskal Wallis test.a , relative to WT. b , relative to G369i KI.The changes produced by ML218 were small but nonetheless significant (sup., n = 7, ISR: p < 0.0001, ML218: p = 0.0047; mid., n = 5, ISR: p < 0.0001, ML218: p = 0.0001; inf., n =5, ISR: p = 0.0019, ML218: p = 0.0460; two-tailed t test).procedures performed at Northwestern University were approved by the Institutional Animal Care and Use Committee.The G369i KI 19 and Cav1.4 KO mouse strains were bred on the C57BL6/J background strain for at least 10 generations.Adult male and female mice were used (6-12 weeks old), and aged-matched C57BL6/J mice were used as the control (WT) animals.

Immunofluorescence
Mice between 6-8 weeks of age were anesthetized using isoflurane and euthanized by cervical dislocation.Eyes were enucleated and hemisected.The eye cups with retina were fixed on ice in 4% paraformaldehyde in 0.1 M phosphate buffer (PB) for 30 min.Fixed eye cups were then washed three times with 0.1 M PB containing 1% glycine followed by infusion of 30% sucrose at 4°C overnight.The eye cups were orientated along their dorsal-ventral axis and frozen in a 1:1  (Table 3).Sections were incubated with primary antibodies for 1 h or overnight and then washed five times with 0.1 M PB.Sections were then incubated with secondary antibodies for 30 min and then washed five times with 0.1 M PB.Trace 0.1 M PB was removed, and sections were then mounted with #1.5H coverslips (ThorLabs) using ProLong Glass Antifade Mountant with or without NucBlue (Thermo Fisher Scientific).
For double labeling with other rabbit polyclonal antibodies, CAR antibodies were conjugated with the CF647 fluorophore (CAR-647) using the Mix-n-Stain antibody labeling kit according to the manufacturer's protocol (Biotium).Sections were processed first with rabbit polyclonal Cav1.4 or EAAT2 antibodies and corresponding secondary antibodies as described above.To prevent CAR-647 from binding to any available sites on the anti-rabbit secondary antibodies previously added to the sections, CaV1.4 or EAAT2 antibodies were readded to the sections and incubated for 30 min.After washing, the sections were incubated with CAR-647 for 1 h, followed by wash steps and cover glass mounting.
Immunofluorescence in labeled retinal sections was visualized using an Olympus FV3000 confocal microscope (Tokyo, Japan) equipped with an UPlanApo 60x oil HR objective (1.5 NA).
Images were captured using the Olympus FLUOVIEW software package.Acquisition settings were optimized using a saturation mask to prevent signal saturation prior to collecting 16-bit.All confocal images presented are maximum z-projections.Images (256 x 256 pixels) used in analyses of synaptic proteins were collected using a 30X optical zoom, 0.6 Airy disk aperture, and voxel size of 0.028  × 0.028  × 0.2  ( ×  × ).Amira segmentation was used for generating 3D binary masks, and Amira 3D label analysis was used for quantification of immunofluorescent and masked images.Non-deconvolved images were used for all analyses.
3D binary masks of individual CAR-labeled pedicles were made by setting the threshold 1 standard deviation above mean fluorescence intensity of each 3D image.3D binary masks of presynaptic labels (CtBP2, CaV1.4,and Bassoon) or postsynaptic labels (GPR179, mGluR6, and TRPM1) within or associated with the pedicle, respectively, were made by setting the threshold 3 standard deviations above the mean fluorescence intensity.To aid in presentation, high-mag images displayed in Figs.3f, 4a, and 6f were deconvolved using cellSens software (Olympus), and any immunofluorescence corresponding to rod synaptic proteins was subtracted from the image.

Patch clamp electrophysiology
Whole-cell voltage clamp recordings of transfected HEK293T cells were performed 48 to 72 hrs after transfection using an EPC-10 amplifier and Patchmaster software (HEKA Elektronik, Lambrecht, Germany).Patch pipette electrodes with a tip resistance between 4 and 6 M were pulled from thin-walled borosilicate glass capillaries (World Precision Instruments, Sarasota, FL) using a P-97 Flaming/Brown Puller (Sutter Instruments, Novato, CA).A reference Ag/AgCl wire was placed into the culture dish mounted on an inverted Olympus IX70 microscope.Recordings were performed at room temperature.A pressurized perfusion pencil multi-barrel manifold controlled with Valve Bank II (AutoMate Scientific, Inc., Berkeley, CA) was used to deliver extracellular solutions.ML218 of different concentrations (0.5, 1, 5, 25 and 100 mM) was added to the extracellular solution the day of experiments.Series resistance was compensated up to 70%, and passive membrane leak subtraction was conducted using a P/−4 protocol.Whole-cell Ca 2+ currents (ICa) of transfected HEK293T cells were evoked for 50 ms with incremental +5 mV steps from -80 mV to +65 mV.Current-Voltage (IV) data were fit with a single Boltzmann equation: where Gmax is the maximal conductance, Vm is the test voltage, Vr is the Ca 2+ reversal potential, Vh is the membrane potential required to activate 50% of Gmax, and k is the slope factor.Data were sampled at 100 kHz, filtered at 3 kHz, and analyzed using custom programs written in IgroPro (WaveMetrics).
Whole-cell voltage clamp recordings of mouse cones, macaque cone terminals, and mouse horizontal cells were performed using an EPC-10 amplifier and Patchmaster software (HEKA).Patch pipette electrodes with a tip resistance between 10 and 14 M for cones and 6-8 for horizontal cells were pulled from thick-walled borosilicate glass (1.5 mm outer diameter; 0.84 mm inner diameter; World Precision Instruments).
To prepare retinal slices, adult mice (6 -8 weeks old) were anesthetized using isoflurane and euthanized by cervical dislocation.Eyes were enucleated, placed into cold Ames' media slicing solution, and hemisected.Following removal of the vitreous, the eye cup was separated into dorsal and ventral halves using a scalpel.Ventral retina was isolated, molded into low-melt agarose and mounted in a Leica VT1200s vibratome (Leica Biosystems).Mouse retina slicing solution was continuously bubbled with 100% O2 and contained the following: Ames' Medium with L-glutamine supplemented with (in mM) 15 NaCl, 10 HEPES, 10 U/mL penicillin, 0.1 mg/mL streptomycin, pH 7.4, osmolarity 300 mOsm/kg.Vertical (~200 µm) or horizontal (~160 µm) retinal slices were anchored in a recording chamber, placed onto a fixed stage, and positioned under an upright Olympus BX51WI microscope equipped with a 60X water-immersion objective (1.0 NA), and superfused with extracellular solution (flow rate of ~1-2 ml/min) at room temperature.Slices were visualized using IR-DIC optics and an IR-2000 (Dage MTI, Michigan City, IN) or SciCam Pro (Scientifica, Uckfield, United Kingdom) CCD camera controlled by the IR-capture software package or µManager, respectively 53,54 .Drugs used in these experiments were added to the mouse extracellular solution the day of experiments at the concentration described in Table 4.A reference Ag/AgCl pellet electrode was placed directly into the recording chamber solution.Data from whole-cell recordings with a series resistance >20 M were discarded.
Cone somas were identified based on their morphology and location (outer ONL cell layer).Cone identity was confirmed by the whole-cell capacitance (~3-4 pF), which is larger than rod whole-cell capacitance (~0.7-1 pF).For cone voltage ramp recordings, cones were held at -90 mV for 200 ms followed by a ramp of +0.5 mV/ms to +40 mV.To determine voltage activation of ICa, whole-cell Ca 2+ currents in cones were evoked for 50 ms with incremental +5 mV steps from -80 mV to +40 mV.The activation voltage of ICa is reported as G/Gmax, where G is the conductance at each test voltage and Gmax is the maximum peak conductance for each cone.
Conductance was calculated using the equation ICa = G(Vm-Vr), where Vr is +60 mV.To determine steady state inactivation of ICa, currents in cones were evoked for 500 ms with incremental +5 mV steps from -90 mV and -30 mV followed by a final step to -30 mV for 50 ms after each test voltage.
The steady state inactivation of ICa is reported as I/Imax, where I is the peak current in the final voltage step to -30 mV and Imax is the maximum peak current for each cone.Data were sampled between 20 and 60 kHz and filtered at 3 kHz.
For horizontal cell light responses, horizontal slices were prepared from central mouse retina.The identity of horizontal cells was determined based on their larger soma diameter (~15 μm) compared to bipolar cell somas (~6 μm).During whole cell patch clamp recordings, horizontal cells were held at -70 mV.Light stimuli (1 s) at 410 nm (630 x 830 μm) were presented onto the retina (at a minimum of 5 s intervals) through the microscope's condenser using a Polygon1000 DMD pattern illuminator (Mightex, Pleasanton, CA, USA) and a custom-built light path.Light intensity (in watts) was measured at the point on the microscope stage where the retina is placed using a power meter (Thorlabs, Newton, New Jersey, USA).Photon flux Фq (photons/s) within the light stimulus area was calculated using the measured light intensities in the formula: where h is Planck's constant (J*s), c is the speed of light (m/s), and λ is the wavelength (m).Light stimulus intensity was increased in Log2 steps from 4.9x10 2 to 2.1x10 5 Фq/μm 2 .For each light intensity step, both ON and OFF current amplitudes were measured from baseline (averaged 5 ms of current prior to light onset) to the maximum positive (after light onset) or maximum negative (after light offset) current, respectively.To confirm the identity of these horizontal cell light responses as AMPA-mediated currents, 1 s light stimuli (410 nm, 1.2x10 5 Фq/μm 2 ) were continuously delivered every 10 s before, during and after bath application of 20 μM DNQX.
For voltage clamp recordings of 13-lined ground squirrel cones, retinal slices were prepared as previously described 55 .The eyecup was divided along the dorsal to ventral axis into superior, middle, and inferior parts.The dorsal area above the line of the optic nerve head was defined as superior, the central region with a width of about 5 mm just ventral to the optic nerve head was middle, and the remaining ventral area was inferior.Isradipine and ML218, alone or in combination, were applied from separate puffer pipettes whose orifices were aimed at the cone synaptic region.Recordings were made with an Axopatch 200B amplifier (Molecular Devices).
Signals were electronically filtered at 5 kHz and digitized at a rate of 10 kHz.Additional Gaussian filtering was added (cutoff frequency of 500 Hz).Tissue was viewed through a 63x water immersion objective on a Zeiss Axioskop FS2 microscope and superfused with extracellular solution at room temperature.Drugs and their concentrations used during these experiments are described in Table 4. Membrane potential was continuously maintained at -85 mV.During ramp stimulation, the membrane potential was depolarized to -85 to +35 mV at a rate of 1 mV/ms.
For rhesus macaque cone terminal recordings, a 12-year-old male was sedated with ketamine (5 mg/kg I.M.) and dexmedetomidine (0.015 mg/kg I.M.).Post-sedation, the animal received buprenorphine (0.02 mg/kg I.M.), atropine (0.02 mg/kg I.M.), and maropitant citrate (1 mg/kg S.Q.).The animal was intubated and maintained with inhaled isoflurane (0.75-2.0%) and propofol (7-8 mg/kg/hr I.V.).Crystalloid fluids (5 mL/kg/hr) were administered I.V., and phenylephrine (5-10 mcg/kg/hr I.V.) was used for blood pressure support.The animal was under anesthesia for approximately 3 hours prior to perfusion.Transcardial perfusion was approached through a midline thoracotomy.Immediately prior to perfusion, 5 mL of Euthasol® (sodium pentobarbital 390 mg/mL/phenytoin sodium 50 mg/mL) was administered I.V.The descending thoracic aorta was clamped, the pericardium was opened, and the right atrium was cut.The apex of the left ventricle was sharply incised, and a large bore cannula (Yankauer suction handle, 5 mm internal diameter) was inserted through the left ventricle until it could be palpated in the ascending aorta.The cannula was clamped in place at the apex of the left ventricle.The animal was perfused with 4000 mL of cold phosphate-buffered saline at ~500 mL/min.Eyes were removed approximately 1 hour following perfusion.Eyes were dissected, and eye cups were allowed to dark adapt for ~30 min and stored in bicarbonate buffered Ames' media at 32C equilibrated with 95% O2/5% CO2 prior to slice preparations.Vertical sections of central retina were prepared from 5 mm retina punches as described for mice.

Serial block-face scanning electron microscopy and 3D reconstructions
Eye cups were prepared from P42 WT and G369i KI littermates and fixed using 4% photoreceptor terminals reconstructed using TrakEM2 (NIH).Postsynaptic partners at cone ribbons were followed to the inner nuclear layer to determine their identity.Amira software was used for 3D visualization of reconstructed profiles.

Electroretinography
Retinal function was assessed using the Celeris system (Diagnosys, Inc.

Visible platform swim test
The visible platform swim test was performed as has been previously described 30 .WT, G369i KI, and Cav1.4 KO male and female mice (6-9 weeks old) were adapted to the procedure room for at least 1 h prior to beginning the experiments.A water-filled, 4-foot diameter galvanized steel tank and a visible white platform with a diameter of 10 cm was used for the swim test.Mice were subjected to 6 swim trials per day.Assays conducted under photopic (55 lux) and scotopic conditions (0 lux) were performed on days 1 and 2, respectively.Light intensity for photopic and scotopic conditions was measured at the platform using an Extech HD450 light meter (FLIR Systems, Nashua, New Hampshire).Mice were given 90 s to find the platform before being removed.After one trial was performed on all mice, the platform was moved to one of three different locations, top left, top center, and top right in relation to the initial site of mouse placement in tank.Each trial was recorded using an infrared camera (Basler AG, Ahrensburg, Germany) and EthoVision XT16 software (Noldus Information Technology).Locating the platform was considered successful when mice contacted the platform with a head-on approach, even if mice failed to escape onto the platform.Mice were allowed to rest on the platform for 15 s at the end of each trial before being returned to a pre-warmed cage.Latency to find the platform was manually recorded and confirmed using recorded videos.Male and female mice were tested separately, and no sex differences in performance were identified post-hoc using two-way RM-ANOVA (Supplementary Table 1).Male and female data were combined into a single group for each genotype.Latency to platform was calculated by averaging the final 3 trials under scotopic and photopic conditions for each mouse and compared using Kruskal  WT (4) G369i KI (5) Ca V 1.4 KO (5) Fig.S2) have additional activity on Cav1 channels at micromolar concentrations including effects on current amplitude and activation voltage.Thus, currents mediated by Cav1.4 that are sensitive to Cav3 blockers (e.g, >5 µM ML218) and spared by dihydropyridines at negative voltages may be mistaken as being mediated by Cav3.In our experiments, the biophysical properties of the isradipine-sensitive ICa in cones of WT mice and ground squirrels resembled only those of Cav1, whereas those of the ML218-sensitive ICa in cones of G369i KI mice resembled only those of Cav3.Therefore, we favor the interpretation that Cav3 contributes to ICa in mouse cones only upon silencing of the Cav1.4Ca 2+ conductance.

Figure 1 .
Figure 1.Cone pedicles and the mutant Cav1.4 channel are normally localized in the OPL

Figure 2 .
Figure 2. A low-voltage activated I Ca is present in cones of G369i KI and Ca V 1.4 KO but absent in WT mice.a,b, Representative traces of ICa evoked by voltage ramps (a) and voltage

Figure 3 .
Figure 3. Pharmacological characterization of ICa in cones of WT and G369i KI mice and

Figure 4 .
Figure 4. Immunofluorescence characterization of cone synapses in WT and G369i KI mice.

Figure 5 .
Figure 5. Cone synapses form incorrect pairings with postsynaptic partners in G369i KI

Figure 6 .
Figure 6.Light responses of horizontal cells are impaired in G369i KI mice.a-b,

Figure 7 .
Figure 7. Photopic vision is reduced but not lost in G369i KI mice.a, Representative traces of photopic ERGs recorded in the presence of background green light (20 cd • s/m 2 ) in WT, G369i wt/vol) mixture of Optimal Cutting Temperature compound and 30% sucrose in a dry ice/isopentane bath.Eye cups were cryosectioned at 20 µm on a Leica CM1850 cryostat (Leica Microsystems), mounted on Superfrost plus Micro Slides (VWR), dried for 5 to 10 min at 42°C, and stored at -20°C until used.Slides with mounted cryosections were warmed to room temperature, washed with 0.1 M PB for 30 min to remove the OCT/sucrose mixture and blocked with dilution solution (DS, 0.1 M PB/10% goat serum/0.5% Triton-X100) for 15 min or overnight at room temperature.All remaining steps were carried out at room temperature.All primary antibodies and appropriate secondary antibodies were diluted in DS at concentrations specified in the Key Resources Table
) paired with the Espion software (Diagnosys, Inc.).Mice were anaesthetized under red light (660 nm) via intraperitoneal (I.P.) injection with ketamine/xylazine mixture (100 mg/kg ketamine, 10 mg/kg xylazine).Tropicamide ophthalmic solution (1 %) and hypromellose lubricant eye gel (0.3 %) were administered topically to both eyes before the mouse was secured to a heated (37 °C) platform to maintain body temperature.Ag/AgCl corneal stimulators were placed on each eye.After collecting data from individual mice, atipamezole (Antisedan, 1-2 mg/kg) was I.P. injection administered to reverse the effects of the ketamine and xylazine.For photopic ERGs, eyes were light adapted using a background green light at 20 cd•s/m 2 for 10 min.Following light adaptation, eight different light intensity pulses (-0.5, 0, 0.5, 1, 1.5, 2, 2.3 and 2.6 log cd•s/m 2 ) were delivered on top of the background green light.ERG a-waves were measured from baseline to the peak of the negative potential.ERG b-waves were measured from the peak of the a-wave to the peak of the positive potential.For flickering ERGs, mice were dark adapted (>12 hrs).White light pulses from -4 to 2.5 log cd•s/m 2 were delivered in 0.5 log unit steps at 10 Hz.Response amplitudes were measured from the trough to the peak of each response at all light intensities.Each intensity stimulus was delivered 10 times, with a 3 s interval between each stimulus, and averaged.The mean response amplitudes recorded in the right and left eye of each mouse are reported for all quantified ERG data.

Table 3 . Key resources used in this study. Primary Antibodies
AnimalsAll mouse and macaque experiments were performed in accordance with guidelines approved by the National Institutes of Health and the Institutional Animal Care and Use Committees at the University of Texas at Austin.All ground squirrel (Ictidomys tridecemlineatus)

Table 4 . Pharmacological drugs and their concentrations used in electrophysiological recordings of cones.
glutaraldehyde in 0.1M cacodylate buffer, pH 7.4, for 4 hours at room temperature followed by additional fixation overnight at 4°C.Glutaraldehyde-fixed eye cups were then washed 3 times in 0.1M cacodylate buffer.Retinas were thereafter isolated and embedded in Durcupan resin after staining, dehydration, and embedding as described previously (Della Santina et al., 2016).A Thermo Scientific VolumeScope serial block face scanning electron micrscope was used to image embedded retinas.Retinal regions comprising a 2x2 montage of 40.96 m tiles were imaged at a resolution of 5nm/pixel and section thickness of 50 nm.Image stacks were aligned, and cone -Wallis one-way ANOVA with Dunn's posthoc multiple comparison test.53.Edelstein, A., Amodaj, N., Hoover, K., Vale, R. & Stuurman, N. Computer control of microscopes using microManager.Curr Protoc Mol Biol Chapter 14, Unit14 20 (2010).54.Edelstein, A.D., et al.Advanced methods of microscope control using muManager software.J Biol Methods 1 (2014).55.DeVries, S.H. & Schwartz, E.A. Kainate receptors mediate synaptic transmission between cones and 'Off' bipolar cells in a mammalian retina.Nature 397, 157-160 (1999).