Calsyntenin-3, an atypical cadherin, suppresses inhibitory basket- and stellate-cell synapses but boosts excitatory parallel-fiber synapses in cerebellum

Cadherins contribute to the organization of nearly all tissues, but the functions of several evolutionarily conserved cadherins, including those of calsyntenins, remain enigmatic. Puzzlingly, two distinct, non-overlapping functions for calsyntenins were proposed: As postsynaptic neurexin ligands in synapse formation, or as presynaptic adaptors for kinesin-mediated vesicular transport. Here, we show that acute CRISPR-mediated deletion of calsyntenin-3 in cerebellar Purkinje cells in vivo causes a large decrease in inhibitory synapses, but a surprisingly robust increase in excitatory parallel-fiber synapses. No changes in the dendritic architecture of Purkinje cells or in climbing-fiber synapses were detected. Thus, by promoting formation of an excitatory type of synapses and decreasing formation of an inhibitory type of synapses in the same neuron, calsyntenin-3 functions as a postsynaptic adhesion molecule that regulates the excitatory/inhibitory balance in Purkinje cells. No similarly opposing function of a synaptic adhesion molecule was previously observed, suggesting a new paradigm of synaptic regulation.


ABSTRACT 23
Cadherins contribute to the organization of nearly all tissues, but the functions of several 24 evolutionarily conserved cadherins, including those of calsyntenins, remain enigmatic. 25 Puzzlingly, two distinct, non-overlapping functions for calsyntenins were proposed: As 26 postsynaptic neurexin ligands in synapse formation, or as presynaptic adaptors for 27 kinesin-mediated vesicular transport. Here, we show that acute CRISPR-mediated deletion 28 of calsyntenin-3 in cerebellar Purkinje cells in vivo causes a large decrease in inhibitory 29 synapses, but a surprisingly robust increase in excitatory parallel-fiber synapses. No Synapse formation, elimination, and remodeling are thought to be organized by synaptic 42 adhesion molecules (SAMs) (Südhof, 2021). Many candidate SAMs have been described, 43 but most SAMs appear to make only partial contributions to the formation and specification 44 of synapses. In particular, few SAMs were consistently found to contribute to the initial 45 formation of synapses. At present, only adhesion-GPCR SAMs, such as latrophilins and 46 BAIs, are known to have a major impact on synapse numbers when tested using rigorous 47  The decrease in inhibitory synapse density raises the question whether inhibitory synaptic 187 transmission is suppressed. To address this question, we recorded miniature inhibitory 188 postsynaptic currents (mIPSCs) from Purkinje cells in the presence of tetrodotoxin ( Figure  189 4). Clstn3 KO produced a large decrease in mIPSC frequency (~60%), without changing 190 the mIPSC amplitude ( Figure 4A-4C). Moreover, the Clstn3 KO increased the rise but not 191 decay times of mIPSCs ( Figure 4D). Measurements of the Purkinje cell capacitance and 192 input resistance showed that the Clstn3 deletion did not produce major changes, 193 demonstrating that it did not globally alter Purkinje cell properties ( Figure S4A). 194 mIPSCs are heterogeneous in Purkinje cells. Smaller mIPSCs are mostly derived from 195 more distant stellate-cell synapses, and larger mIPSCs from more proximal basket-cell 196 synapses (Nakayama et al., 2012). To examine these two types of input synapses 197 separately, we plotted the mIPSC amplitudes in a normal distribution ( Figure 4E). This plot 198 revealed that the majority of mIPSC amplitudes (>90%) are <60 pA. Therefore, we 199 separately analyzed mIPSCs with amplitudes of >60 pA and <60 pA, of which the >60 pA We detected a significant decrease (~40%) in IPSC amplitudes. The decrease in IPSC 210 amplitude is consistent with a loss of inhibitory synapses, but could also be due to a 211 decrease in release probability. However, we detected no major changes in the coefficient 212 of variation, paired-pulse ratio, or kinetics of evoked IPSCs, suggesting that the release 213 probability is normal ( Figure 5B-5F, S4B). These data confirm the morphological results, 214 suggesting that the Clstn3 KO decreases inhibitory synapse numbers. 215

Clstn3 deletion in Purkinje cells increases excitatory parallel-fiber but not 216
climbing-fiber synapse densities. The decrease in inhibitory synapse numbers by the 217 Clstn3 KO is consistent with previous studies suggesting that Clstn3 promotes synapse 218 formation in the hippocampus, but these previous studies primarily identified a decrease in 219 excitatory synapses (Kim et al., 2020;Pettem et al., 2013;Ranneva et al., 2020). We thus 220 tested whether the Clstn3 KO also affects excitatory synapse numbers in cerebellum.  The potential increase in parallel-fiber synapses induced by the Clstn3 KO, suggested by 241 the enhanced vGluT1 staining intensity, is unexpected. This prompted us to examine the 242 levels of GluA1 as an astroglial marker of tripartite parallel-fiber synapses ( Figure 6C;  We next analyzed the density of climbing-fiber synapses by staining cerebellar sections for 247 vGluT2, but detected no significant effect of the Clstn3 KO in Purkinje cells ( Figure 6E). 248 Different from parallel-fiber synapses that contain vGluT1, climbing-fiber synapses are 249 labeled with antibodies to vGluT2 and are readily resolved by confocal microscopy (Figure  250 6E). The number and size of synaptic puncta identified with vGluT2 antibodies were not 251 altered by the Clstn3 KO, although there was a slight trend towards a decrease in 252 climbing-fiber synapse density ( Figure 6F, 6G). These observations suggest that the 253 enhancement of parallel-fiber synapse density by the Clstn3 KO is specific for this type of 254

synapse. 255
The Clstn3 KO increases the spine density of Purkinje cells. It is surprising that the 256 Clstn3 KO in Purkinje cells appears to increase the parallel-fiber synapse density, as one 257 would expect a synaptic adhesion molecule to promote but not to suppress formation of a 258 particular synapse. The parallel-fiber synapse increase is likely not a homeostatic response 259 to the loss of inhibitory synapses because such a response, which would aim to maintain 260 the correct excitatory/inhibitory balance, should involve a decrease, and not an increase, in 261 parallel-fiber synapses. The increase in parallel-fiber synapse numbers is also unexpected 262 given previous results showing that in hippocampal CA1 neurons, the Clstn3 KO decreases confirm this increase, we analyzed the dendritic spine density in Purkinje cells. Since nearly 265 all spines contain parallel-fiber synapses and all parallel-fiber synapses are on spines 266 (Sotelo, 1975), the spine density of Purkinje cells represents a reliable proxy for synapse 267

density. 268
We filled individual Purkinje cells in acute slices with biocytin via a patch-pipette, and 269 analyzed their dendritic structure and spine density by quantitative morphometry ( Figure 7A The Clstn3 KO increases parallel-fiber but not climbing-fiber synaptic transmission. 282 The increase in parallel-fiber synapses could be due to a true enhancement of parallel-fiber 283 synapse formation, or a compensatory reaction to a decrease in parallel-fiber synapse 284 function. Although the latter hypothesis would be consistent with a homeostatic response, it 285 seems unlikely given that in vertebrates, synapses rarely proliferate in response to a 286 functional impairment. To clarify this question, we analyzed parallel-synapse function by 287 electrophysiology, and compared it to climbing-fiber synapse function as an internal control 288 since climbing-fiber synapse numbers are not changed by the Clstn3 KO in Purkinje cells. 289 We first monitored spontaneous miniature synaptic events (mEPSCs) in the presence of 290 tetrodotoxin. We observed an increase in mEPSC amplitudes (~25%) and frequency (~15%) 291 in Clstn3 KO neurons, without a notable change in mEPSC kinetics ( Figure  were monitoring mEPSCs derived from parallel-fiber synapses (whose density is increased 296 morphologically), we analyzed only slow mEPSCs with rise times of >1 ms that are mostly 297 generated by parallel-fiber synapses on distant dendrites (Nakayama et al., 2012;298 Yamasaki et al., 2006). The results were the same as for total mEPSCs, confirming that the 299 Clstn3 KO increases parallel-fiber synaptic activity ( Figure 8E-8H). 300 Finally, we measured evoked parallel-fiber EPSCs, using input-output curves to correct for 301 variations in the placement of the stimulating electrode ( Figure 9A). Consistent with the 302 morphological and mEPSC data, the Clstn3 KO robustly enhanced parallel-fiber synaptic 303 responses (~60% increase) ( Figure 9B-9D). This finding suggests that the Clstn3 KO not 304 only increases the density of parallel-fiber synapses, but also renders these synapses more 305 efficacious. The increased strength of parallel-fiber synaptic transmission was not due to a 306 change in release probability because neither the coefficient of variation nor the 307 paired-pulse ratios of parallel-fiber EPSCs were affected ( Figure 9E-9G). The increase of 308 parallel-fiber EPSCs is consistent with the vGluT1 intensity and mEPSC amplitude changes, 309 providing further evidence that the Clstn3 KO enhances parallel-fiber synapses. 310 In contrast to parallel-fiber EPSCs, climbing-fiber EPSCs exhibited no Clstn3 KO-induced 311 alteration. Specifically, the amplitude, paired-pulse ratio, and kinetics of climbing-fiber 312 EPSCs in control and Clstn3 KO Purkinje cells were indistinguishable ( Figure 9H-9L). 313 These findings are consistent with the lack of a change in vGluT2-positive synaptic puncta 314 analyzed morphologically ( Figure 6E-6G). Viewed together, these data suggest that Clstn3 315 KO produces an increase in excitatory parallel-fiber, but not climbing-fiber, synapses.   Immunohistochemistry on the cerebellar cortex was done as previously reported (Zhang et 490 al., 2015). Mice were anesthetized with isoflurane and sequentially perfused with 491 phosphate buffered saline (PBS) and ice cold 4% paraformaldehyde (PFA). Brains were 492 dissected and post-fixed in 4% PFA overnight, then cryoprotected in 30% sucrose in PBS 493 for 24 h. 40 μm thick sagittal sections of cerebellum were collected using a Leica 494 CM3050-S cryostat (Leica, Germany). Free floating brain sections were incubated with 495 blocking buffer (5% goat serum, 0.3% Triton X-100) for 1 h at room temperature, then 496 treated with primary antibodies diluted in blocking buffer overnight at 4 ℃ (anti-vGluT1, 497 Rabbit, YZ6089, Yenzym, 1:1,000; anti-vGluT2, Rabbit, YZ6097, 1:1,000, Yenzym; 498 anti-vGAT, guinea pig, 131004, Sysy,1:1,000). Sections were washed three times with PBS 499 (15 min each), then treated with secondary antibodies (Alexa goat anti guinea pig 633, 500 A-21105, Invitrogen, 1:1,000; or Alexa goat anti rabbit 647, A-21245, Invitrogen, 1:1,000) 501 for 2 h at room temperature. After washing with PBS 4 times (15 min each), sections were 502 stained with DAPI (D8417, Sigma) and mounted onto Superfrost Plus slides with mounting 503 media. Confocal images were acquired with a Nikon confocal microscope (A1Rsi, Nikon, 504 Japan) with 60x oil objective, at 1024 x 1024 pixels, with z-stack distance of 0.3 μm. All 505 acquisition parameters were kept constant within the same day between control and Clstn3 506 KO groups. Images were taken from cerebellar lobules IV/V. Images were analyzed with 507 Nikon analysis software. During analysis, we divided the cerebellar cortex into different 508 layers to compare Clstn3 KO effects. We defined 0-40% as superficial molecular layer and 509 40-80% as molecular deep layers, 80-100% as PCL, and we analyzed and labeled GCL 510 separately in vGAT staining. Slices were fixed in 4% PFA/PBS solution overnight at 4℃. Slices were then washed 3 X 5 572 mins with PBS, permeabilized, and blocked in 5% goat serum, 0.5% Triton-X100 in PBS at 573 room temperature for 1 h. Then slices were incubated in 1:1,000 diluted Streptavidin 574 Fluor™ 647 conjugate (S21374, Invitrogen) at room temperature for 2 h in 5% goat serum 575 in PBS, washed 5 X 5 min with PBS, and mounted onto Superfrost Plus slides for imaging. 576 Image overviews were obtained with a Nikon confocal microscope (A1Rsi, Nikon, Japan) 577 with a 60x oil objective, at 1024 x 1024 pixels, with z-stack distance of 2 μm. Dendritic tree 578 3D reconstructions were performed using Neurolucida360 software (MBF science, USA) in 579 the Stanford Neuroscience Microscopy Service Center. Note that some somas could not be 580 detected automatically and were manually labeled. Spine images were obtained with a 581 ZESS LSM980 inverted confocal, Airyscan2 for fast super-resolution setup, equipped with 582 an oil-immersion 63X objective. Z-stacks were collected at 0.2 μm intervals at 0.06 μm/pixel 583 resolution with Airyscan2. Spine images were deconvolved using ZEN blue software 584 Accelerating rotarod. Mice were placed on an accelerating rotarod (IITC Life Science). The 590 rod accelerated from 4 to 40 r.p.m. in 5 min. Mice were tested 3 times per day with 1 hour 591 interval and repeated for 3 days. Time stayed on the rod was recorded while the mouse fell 592 off, or hanged on without climbing, or reached 5 min. 593 Three-chamber social interaction. Social interaction was evaluated in a three-chamber box. 594 Mice were placed initially in the central chamber to allow 10 min habituation for all three 595 chambers. For sociability session, a same sex-and age matched stranger mouse 596 (stranger1) was placed inside an upside-down wire pencil cup in one of the side chambers. 597 The other side had the same empty pencil cup. A test mouse was allowed 10 min to 598 investigate the three chambers. For social novelty session, another stranger mouse 599 (stranger2) was placed into the empty pencil cup and test mouse was allowed another 10 600 min to investigate between three chambers. The time mice spent in each chamber was 601 recorded and analyzed using BIOBSERVE III tracking system. 602 603

Data analysis 604
Experiments and data analyses were performed blindly by coding viruses. Unpaired t-test 605 or one-way ANOVA or two-way ANOVA or repeat measures ANOVA were used to analyze 606 slice physiology data or immunohistochemistry data or behavior data as indicated in figure  607 legends. Kolmogorov-Smirnov test was used to analyze the cumulative curves of mEPSCs 608 or mIPSCs. Significance was indicated as *p < 0.05, **p < 0.01, ***p < 0.001. Data are 609 expressed as means ± SEM. synthesis are driven by U6 and CAG promoters, respectively. Control mice were infected 928 with AAVs that lacked sgRNAs but were otherwise identical. 929 (C) Experimental strategy for CRISPR-mediated acute Clstn3 deletions in the cerebellum.

930
AAVs expressing the sgRNAs and tdTomato were stereotactically injected into the 931 The Purkinje cell image is from one of the cells reconstructed during the present study. The 1153 changes summarized on the right were identified in Figures 3-9. 1154