Huntingtin-KIF1A-mediated axonal transport of synaptic vesicle precursors influences synaptic transmission and motor skill learning in mice

Neurotransmitters are released at synapses by synaptic vesicles (SVs), which originate from SV precursors (SVPs) that have traveled along the axon. Because each synapse maintains a pool of SVs, only a small fraction of which are released, it is unclear whether axonal transport of SVPs modifies synaptic function. Here, studying the corticostriatal network both in microfluidic devices and in mice, we find that phosphorylation of the Huntingtin protein (HTT) causes it to recruit the kinesin motor KIF1A, which in turn increases axonal transport of SVPs and synaptic glutamate release. In mice, constitutive HTT phosphorylation leads to SV over-accumulation at synapses, increases the probability of SV release, and impairs motor skill learning on the rotating rod. Silencing KIF1A in these mice restored SV transport and motor skill learning to wild-type levels. Axonal SVP transport within the corticostriatal network thus influences synaptic plasticity and motor skill learning.


Introduction 35
Synaptic plasticity underlies our ability to learn. The number of synaptic vesicles (SVs) and release 36 sites at synapses, the probability of the SVs releasing neurotransmitter, and the SV quantal size all affect 37 synaptic strength and thus memory (Katz, 1969). SVs actually begin life as SV precursors (SVPs), which 38 are formed in the cell body and transported along the axon to the presynapse-a distance that can span 39 meters (Guedes-Dias and Holzbaur, 2019; Rizalar et al., 2021; Rizzoli, 2014). It seems intuitively obvious 40 that this long-distance axonal transport influences SV homeostasis, because synapses must be replenished 41 somehow with new vesicles as they release neurotransmitters. Yet the synaptic SV pools contain an 42 average of 400 to 500 vesicles, of which only a few percent participate in synaptic release (Reshetniak 43 and Rizzoli, 2021); these synaptic pools almost certainly serve to ensure that there are always sufficient 44 SVs ready to be released, even with prolonged neuronal stimulation (Denker et al., 2011;Rizzoli, 2014). 45 Moreover, neighboring synapses appear to share localized pools of SVs that circulate between them 46 (Wong et al., 2012). 47 The idea that SVP transport affects synaptic neurotransmitter release derives from studies of 48 mutations that strongly affect SVP transport, neuronal transmission, and behavior in mice, flies, and 49 worms. In Caenorhabditis elegans, null mutants for the kinesin-related gene unc-104 or the vesicle-50 associated protein SAM-4 lead to defects in SVP transport, with a consequent lack of SV at synapses and 51 locomotor deficits (Hall and Hedgecock, 1991;Zheng et al., 2014). In Drosophila, deletion of the imac 52 gene, a kinesin-3 family member, impairs SVP axonal transport and the formation of synaptic boutons 53 SVs at synapses, along with sensorimotor deficits and early postnatal death (Yonekawa et al., 1998). The 56 consequences of completely blocking a molecular motor, however, may not tell us whether enhancement 57 or attenuation of axonal transport influences synapse homeostasis, synaptic transmission, or the function 58 of specific brain circuits. In fact, studies of mice with either constitutively phosphorylated or 59 unphosphorylatable Huntingtin (HTT), a protein that plays a prominent role in axonal transport (Saudou 60 and Humbert, 2016), did not reveal any obvious behavioral phenotypes despite affecting axonal transport 61

HTT-KIF1A-mediated transport regulates the number of SVs at synapses 241
We next investigated whether SVP anterograde transport via the HTT-KIF1A complex regulates 242 the number of vesicles at synapses. We injected lentiviruses encoding either sh-scramble-GFP or sh-243 KIF1A-GFP in layer V of the HTT-SD motor cortex (Fig. 7A), whose neurons project mainly to the DLS 244 (Hunnicutt et al., 2016). We then counted the number of SVs at corticostriatal synapses from sections 245 prepared from WT and HTT-SD brains injected with lentiviral sh-Scr or sh-KIF1A. WT sh-KIF1A 246 presynapses showed significantly fewer SVs than WT sh-Scr presynapses (Fig. 7B). As previously shown 247 (Fig. 1B), there were significantly more SVs at presynapses in HTT-SD, but this number reverted to WT 248 levels in HTT-SD brains treated with sh-KIF1A (Fig. 7C). These data demonstrate that decreasing the To determine whether the modification in anterograde transport via the HTT-KIF1A complex 256 within corticostriatal projecting neurons is responsible for the defect in motor skill learning we observed 257 in HTT-SD mice, we injected lentiviral vectors encoding sh-Scr-GFP and sh-KIF1A-GFP into 3-month-258 old WT and HTT-SD mice. Three weeks after lentiviral injection we performed the same behavioral 259 protocol as in Figure 1 (Fig. 8A). As with the non-injected mice (Figure 2B), HTT-SD mice did not show 260 improvement over 8 days and had a much shorter latency to fall than WT mice (Fig. 8B). Silencing 261 KIF1A improved the performance of the HTT-SD mice, especially over the first 5 days (Fig. 8B, left 262 graph). A careful analysis of behavioral performance on the first and last days revealed that the 263 improvement in motor learning of the HTT-SD mice via sh-KIF1A silencing was significant on the first 264 day of training, but the effect did not last until the eighth day (Fig. 8C) weakness reflects the extra challenge presented by the extremely long axons of the peripheral nervous 310 system, but the cognitive deficits in ADMR type 9 show that disruptions in axonal transport can clearly 311 disturb synaptic transmission and synaptic strength, with obvious consequences for learning and memory 312  Together, these studies indicate that SVP transport is normally fine-tuned to ensure the proper 326 quantity of SVs at the synapse and effective synaptic function. biochemical studies or processed to be imaged by electron microscopy. The number of animals was 379 limited to the minimum number necessary per group in order to have a chance of at least 80% of detecting 380 a significant difference (power 1-β) and a risk of error α of 5%. This number was determined using a 381 statistical test for estimating the optimal sample size using the variances determined in a preliminary 382 study so as to reduce the number of animals used as much as possible while keeping enough to avoid 383 compromising the validity of the experiments that was carried out. 384 For biochemistry and neuronal culture (E15.5), the sex distinction of homozygous or WT mice was 385 not made. Specific ages used for each experiment are indicated in the figure legends. C57BL/6J mice, 386 purchased from Charles River Laboratory, were used for backcrosses to maintain the colony and to obtain 387 WT E15.5 pups. 388 389

Primary neuron culture and transduction 390
Primary cortical and striatal neurons were dissected from E15.5 wild type (C57Bl/6J) or HTT-SA or 391 HTT-SD mouse embryos as previously described (Liot et al., 2013). They underwent a chemical 392 dissociation with papain cysteine solution, DNase (1/100), and FBS (1/10) and were finally mechanically 393 dissociated. They were re-suspended in a growing medium containing a Neurobasal medium, 2% B27, 394 1% Penicillin/streptomycin, and 2 mM glutamax (5 x 10 6 cells in 120 µl). Cortical neurons were plated in 395 the presynaptic chamber coated with poly-D-lysine (0.1 mg/ml) and striatal neurons were plated in the 396 postsynaptic chamber coated with poly-D-lysin and laminin (10 µg/ml) with a final density of ~7000 397 cells/mm 2 . A growing medium was added to the synaptic chamber to equilibrate the flux. Neurons were 398 left in the incubator for 2 hours and then all compartments were gently filled with growing medium. 399 Neurons were cultured at 37°C in a 5% CO2 incubator for 10-12 days.

Stereotaxic injections 416
3-month-old HTT-SD and WT male mice were anesthetized by inhalation of isoflurane associated 417 with a mix of oxygen and room air (3-5% of isoflurane for induction and 1-2% in the mask). The mouse 418 head was then shaved and placed within the stereotaxic frame. The skin was incised, and the skull was 419 bilaterally drilled. The capillary was inserted slowly. We injected bilaterally (position AP:1,54 ML: + or -420 1,6 DV:-0,8) 500 nl of the diluted lentivirus (1/3 dilution in saline solution of the KIF1A shRNA or the 421 Scr shRNA) at 0.5 µl/min speed using a nanoinjector. The capillary was slowly removed one minute after 422 the end of the injection to prevent the leak of the injected solution. The skull was then washed with saline 423 solution, the skin was sutured and 1 ml of NaCl 0.9% was injected subcutaneously. After surgery, mice 424 were put alone in a warmed cage and monitored daily throughout recovery. 425 426

Plasmids 427
The VAMP2-mCherry construct was a kind gift from T. Ryan's laboratory. Vglut1 cDNA sequence 428 was amplified from an adult mouse brain. Its sequence from the 104 th amino acid to the end was cloned 429 after the pHluorin sequence in a Smal site in a superecliptic pHluorin containing vector used in 430 coupled to a spinning-disk confocal system (CSU-W1-T3; Yokogawa) connected to an electron-458 multiplying CCD (charge-coupled device) camera (ProEM+1024, Princeton Instrument) at 37 °C and 5% 459

CO2. 460
For the study of the exocytosis events, images were acquired every 200 ms for 1min on an inverted 461 microscope (Axio Observer, Zeiss) with X63 oil-immersion objective (1.46NA) coupled to a spinning-462 disk confocal system (CSU-W1-T3; Yokogawa) with TIRF microscopy (Nikon/Roper, Eclipse Ti) 463 equipped with a camera Prime 95B sCMOS (Telelyne Photometrics) at 37 °C and 5% CO2. The same 464 three fields per microchambers were acquired before and after a 4AP-bicuculline (respectively 2.5mM 465 and 50µM) stimulation of the presynaptic chamber, four times in total (1 before and three after 466 stimulation). Lysates were then centrifuged (12000 RPM) and the supernatant, considered as the total fraction, is then 500 centrifuged (3000 RPM for 10 minutes). The resulting supernatant was centrifuged (12 000 RCF for 40 501 minutes). The supernatant was then ultracentrifuged (100 000g) to obtain the vesicular fraction (the 502 pellet) and the cytosolic fraction (the supernatant). 503 All types of lysed samples were dosed by a Bradford reagent to quantify the protein concentration

Electron microscopy 513
We anesthetized 3-to 4-month-old animals with 1ml/kg of Doléthal ® and perfused them 514 transcardially with cold PBS followed by 2% paraformaldehyde 2% glutaraldehyde and 0,1M cacodylate 515 cold solution. We removed brains from the skull and fixed them in a 0.1M phosphate buffer pH7.2 with 516 2% of glutaraldehyde and 2% of paraformaldehyde for 48 hours at 4°C before obtaining 2 mm-thick or 517 100µm-thick slices from a mold and a vibratome, respectively. A 1mm square piece of tissue was 518 removed from the dorsolateral striatum; samples were then fixed again with the same solution for 72 519 hours, washed with phosphate buffer, and then post-fixed in a 0.1M phosphate buffer pH 7.2 with 1% 520 Osmium tetroxide for 1 hour at 4°C. After extensive washes with water, samples were then stained with a 521 solution of 1% uranyl acetate pH 4 in water for 1 hour at 4°C. They were further dehydrated through a 522 gradient of ethanol (30%-60%-90% and three at 100%) and infiltrated with a solution of 1/1 epon/alcohol 523 100% for 1 hour and followed by several baths of fresh epon (Fukka) for 3 hours. The resin was then 524 with 95% O2 and 5% CO2. Signals were amplified using EPC10-2 amplifiers (HEKA Elektronik, 545 Lambrecht, Germany). All recordings were performed at 34°C using a temperature control system (Bath- parametric Mann Whitney, one-sample t-tests using the indicated significance threshold (p). 561 562

Mass spectrometry 563
This analysis follows that of (Migazzi et al., 2021). Briefly, vesicular fraction from brains obtained 564 as described earlier was first pre-cleared for an hour at 4°C with protein A Sepharose beads (Sigma 565 Aldrich-P9424) and then immunoprecipitated for 3 hours at 4°C by agarose beads preincubated with 566 rabbit anti-HTT D7F7 antibody (Cell Signaling, Cat#5656). To remove the non-specific binding, the 567 beads were washed three times with the lysis buffer and bound proteins are finally eluted with Laemmli 568 buffer. The HTT corresponding band on the western blot was cut and analyzed. MS was performed with a 569 LTQ Orbitrap XL mass spectrometer (Thermo Scientific), equipped with a nanoESI source (Proxeon). 570 The top eight peaks in the mass spectra (Orbitrap; resolution, 60,000) were selected for fragmentation 571

Exocytosis events 598
The same three fields per microchamber were acquired before and after a 4AP-bicuculline 599 (respectively 2.5mM and 50µM) stimulation of the presynaptic chamber, four times in total (1 before and 600 Where 608 • yi expresses the difference of the number of events after (yi, post) minus before (yi, pre) stimulation in 609 HTT-SD neurons of the field i, 610 • xj expresses the difference of the number of events after minus before stimulation in WT neurons of 611 the field j and n of them have been averaged. 612 • the final value was normalized to 1, i.e. a given HTT-SD neuron field, whose activity after 613 stimulation increased as much as that of the average of the WT neuron fields, will display a value 614 of 1. 615 • The amplitude of the signal from stimulated neurons was normalized by that of the same neuron 616 before stimulation. 617 618

Statistical analysis 619
Statistical calculations were performed using GraphPad Prism 6.0. Statistical parameters 620 (Replication, sample size, SEM, etc…) are reported in the legend figure. For each data set, outliers were 621 identified and removed from the analysis using the ROUT test (Q=1%). Then, the Shapiro-Wilk 622 normality test with the threshold set at a = 0.05 was performed to assess the normality of the data. 623 According to whether the data set followed a normal repartition or not, parametric or non-parametric tests 624 were performed respectively. Then, if two conditions were analyzed, a t-test (or a Mann-Whitney test if 625 nonparametric) was used. If more than two conditions were compared, a one-way ANOVA followed by a 626 Tukey 's post-hoc analysis (or a Kruskal-Wallis test followed by a Dunn's post hoc analysis if 627 nonparametric) was used. If the data set were dependent on another, a two-way ANOVA was used 628 followed by Tukey 's post-hoc analysis if more than two groups are compared or a Sidak's post-hoc 629 analysis if only two groups are analyzed. For the non linear fit, the run test was performed to study 630 whether the curve deviates systematically from the data. Low P value (ns) indicates that the curve poorly 631 describes the data. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant. N number 632 corresponds to biological replicates. 633 634

Data Availability 635
All data generated or analyzed during this study are included in the manuscript and supporting files.