Myosin XVI Regulates Actin Cytoskeleton Dynamics in Dendritic Spines of Purkinje Cells and Affects Presynaptic Organization

The actin cytoskeleton is crucial for function and morphology of neuronal synapses. Moreover, altered regulation of the neuronal actin cytoskeleton has been implicated in neuropsychiatric diseases such as autism spectrum disorder (ASD). Myosin XVI is a neuronally expressed unconventional myosin known to bind the WAVE regulatory complex (WRC), a regulator of filamentous actin (F-actin) polymerization. Notably, the gene encoding the myosin’s heavy chain (MYO16) shows genetic association with neuropsychiatric disorders including ASD. Here, we investigated whether myosin XVI plays a role for actin cytoskeleton regulation in the dendritic spines of cerebellar Purkinje cells (PCs), a neuronal cell type crucial for motor learning, social cognition and vocalization. We provide evidence that both myosin XVI and the WRC component WAVE1 localize to PC spines. Fluorescence recovery after photobleaching (FRAP) analysis of GFP-actin in cultured PCs shows that Myo16 knockout as well as PC-specific Myo16 knockdown, lead to faster F-actin turnover in the dendritic spines of PCs. We also detect accelerated F-actin turnover upon interference with the WRC, and upon inhibition of Arp2/3 that drives formation of branched F-actin downstream of the WRC. In contrast, inhibition of formins that are responsible for polymerization of linear actin filaments does not cause faster F-actin turnover. Together, our data establish myosin XVI as a regulator of the postsynaptic actin cytoskeleton and suggest that it is an upstream activator of the WRC-Arp2/3 pathway in PC spines. Furthermore, ultra-structural and electrophysiological analyses of Myo16 knockout cerebellum reveals the presence of reduced numbers of synaptic vesicles at presynaptic terminals in the absence of the myosin. Therefore, we here define myosin XVI as an F-actin regulator important for presynaptic organization in the cerebellum.

The actin cytoskeleton is crucial for function and morphology of neuronal synapses. Moreover, altered regulation of the neuronal actin cytoskeleton has been implicated in neuropsychiatric diseases such as autism spectrum disorder (ASD). Myosin XVI is a neuronally expressed unconventional myosin known to bind the WAVE regulatory complex (WRC), a regulator of filamentous actin (F-actin) polymerization. Notably, the gene encoding the myosin's heavy chain (MYO16) shows genetic association with neuropsychiatric disorders including ASD. Here, we investigated whether myosin XVI plays a role for actin cytoskeleton regulation in the dendritic spines of cerebellar Purkinje cells (PCs), a neuronal cell type crucial for motor learning, social cognition and vocalization. We provide evidence that both myosin XVI and the WRC component WAVE1 localize to PC spines. Fluorescence recovery after photobleaching (FRAP) analysis of GFP-actin in cultured PCs shows that Myo16 knockout as well as PC-specific Myo16 knockdown, lead to faster F-actin turnover in the dendritic spines of PCs. We also detect accelerated F-actin turnover upon interference with the WRC, and upon inhibition of Arp2/3 that drives formation of branched F-actin downstream of the WRC. In contrast, inhibition of formins that are responsible for polymerization of linear actin filaments does not cause faster F-actin turnover. Together, our data establish myosin XVI as a regulator of the postsynaptic actin cytoskeleton and suggest that it is an upstream activator of the WRC-Arp2/3 pathway in PC spines. Furthermore, ultra-structural and electrophysiological analyses of Myo16 knockout cerebellum reveals the presence of reduced numbers of synaptic vesicles at presynaptic terminals in the absence of the myosin. Therefore, we here define myosin XVI as an F-actin regulator important for presynaptic organization in the cerebellum.
Myosin XVI is a neuronally expressed, vertebrate-specific unconventional myosin (Patel et al., 2001;Yokoyama et al., 2011;Cameron et al., 2013). There are indications that the myosin might be involved in actin cytoskeleton regulation. The myosin XVI heavy chain (MYO16, also known as NYAP3) is an F-actin-binding protein with an N-terminal ankyrin repeat domain that binds protein phosphatase 1 catalytic subunits, followed by a myosin motor domain that, in mammals, is likely impaired in its ability to hydrolyze ATP (Patel et al., 2001;Cameron et al., 2007;Kengyel et al., 2015). Via its tail domain, MYO16 binds phosphoinositide 3-kinase (PI3K) and the WRC, an upstream regulator of Arp2/3-dependent actin filament nucleation (Yokoyama et al., 2011). MYO16/NYAP3 and two proteins that resemble the myosin's tail domain (NYAP1, NYAP2) are thought to function redundantly via bridging WRC-PI3K association in order to regulate neurite outgrowth (Yokoyama et al., 2011). Notably, both WRC and Arp2/3 are key factors that drive actin dynamics in hippocampal dendritic spines (Spence and Soderling, 2015;Chazeau and Giannone, 2016).
The WRC constitutes a heteropentameric complex consisting of WAVE1 (or its orthologs WAVE2, WAVE3), CYFIP1 (also known as SRA1; or its ortholog CYFIP2/PIR121), NCKAP1 (also known as NAP1, HEM2; or its ortholog HEM1), ABI1 (or its orthologs ABI2, ABI3) and HSPC300 (also known as BRICK1) (Takenawa and Suetsugu, 2007;Bisi et al., 2013). Activation of the WRC is a multistep process that involves binding to the small GTPase Rac1, and results in exposure of the VCA domain of WAVE (Lebensohn and Kirschner, 2009;Chen et al., 2010). Once exposed, the VCA domain binds and activates Arp2/3. The seven subunit Arp2/3 complex catalyzes the nucleation of new filaments from the side of pre-existing ones, thereby promoting formation of a branched F-actin meshwork (Rotty et al., 2013).
Interestingly, several genes linked to an increased risk of developing ASD encode actin regulators (Joensuu et al., 2018). This includes the genes encoding WRC components CYFIP1 and NCKAP1, two established myosin XVI protein interaction partners (Wang et al., 2009;Yokoyama et al., 2011;Chang et al., 2013;Joensuu et al., 2018). ASD is a complex neuropsychiatric disease characterized by deficits in social interaction and communication, with motor coordination problems as a frequent comorbidity de la Torre-Ubieta et al., 2016). Strikingly, ASD-like phenotypes in mouse models can be reverted by manipulating actin regulators (Dolan et al., 2013;Duffney et al., 2015). Therefore, it has been suggested that alterations in F-actin dynamics are one of the important pathological features in ASD (Spence and Soderling, 2015;Lin et al., 2016;Yan et al., 2016;Borovac et al., 2018;Hlushchenko et al., 2018;Joensuu et al., 2018). Notably, genetic evidence links also MYO16 to an increased risk of developing ASD (Wang et al., 2009;Chang et al., 2013;Liu et al., 2015) and other neuropsychiatric disorders (Rodriguez-Murillo et al., 2014;Kao et al., 2016).
MYO16 occurs in cerebellar PCs (Patel et al., 2001;Cameron et al., 2013), central signal integrators that provide the only output from the cerebellar cortex. Their dendrites project into the cerebellar molecular layer and receive excitatory synaptic input via dendritic spines from axons termed PFs (granule cell axons) and climbing fibers (CFs). PCs are crucial for motor coordination and motor learning (Ito, 2001;Schonewille et al., 2010). However, recent research demonstrates that PCs are also important for social cognition, language and vocalization (Tsai et al., 2012;Fujita-Jimbo and Momoi, 2014;Peter et al., 2016;Sokolov et al., 2017). Interestingly, malfunction of the cerebellum and PCs has been linked to ASD and ASD-like phenotypes in mice de la Torre-Ubieta et al., 2016). For example, PC-specific knockout of ASD genes Shank2 or Tsc1 causes social interaction deficits in mice (Tsai et al., 2012;Peter et al., 2016). Thus, cerebellar PCs appear to be a highly relevant cell type for studying the role of ASD-related genes such as Myo16.
Notably, little is known about F-actin regulation in the dendritic spines of PCs, compared to hippocampal neurons. PCs express WRC components, the formin Daam1, and the Arp2/3formin coordinator MTSS1 (Soderling et al., 2003;Saarikangas et al., 2015;Kawabata Galbraith et al., 2018). However, whether and how the WRC, Arp2/3, and formins affect F-actin turnover in PC spines has not been examined directly. Remarkably, PCs also harbor specific F-actin regulators such as espin and delphilin that are not found in other neurons (Miyagi et al., 2002;Sekerkova et al., 2003). Thus, actin dynamics might be regulated in a unique manner in PC dendritic spines.
The aim of the present study was to obtain insight into the role of myosin XVI in neuronal cells. Given the known interaction of MYO16 with the actin regulator WRC, and considering that several genes associated with ASD encode actin regulators, we hypothesized that myosin XVI is involved in regulating dendritic spine F-actin and, possibly, synaptic structure and function.

Generation and Initial Characterization of Myo16 Knockout Mice
To investigate the role of MYO16 in the brain, we generated two mouse lines that carry constitutive Myo16 knockout alleles (Myo16 em2 , Myo16 em3 ; Figures 1A,B). Western blot analyses confirmed the absence of myosin XVI heavy chain in homozygous Myo16 −/− mice of both lines ( Figure 1C). Since Myo16 is expressed in cerebellum (Patel et al., 2001;Cameron et al., 2007; see also Figure 1C), we examined the anatomical organization of this structure in Myo16 −/− knockout mice. Nissl staining of cerebellar sections did not reveal gross abnormalities regarding foliation and the organization of layers ( Figure 1D). Moreover, immuno-fluorescence labeling for the presynaptic marker VGLUT1 did not expose gross deficits in terms of presence of PF terminals in the molecular layer ( Figure 1E). Given the genetic association of MYO16 with ASD, we also examined the in situ localization of Shank2 and neuroligin-2, two proteins that are strongly linked to ASD and that form postsynaptic clusters at excitatory and inhibitory synapses of PCs, respectively (Zhang et al., 2015;Ha et al., 2016;Peter et al., 2016). Quantification of Shank2 clusters ( Figure 1F) and neuroligin-2 clusters ( Figure 1G) within the cerebellar molecular layer showed that cluster density is unaltered in the absence of myosin XVI. Finally, we characterized protein levels and subcellular distribution of postsynaptic molecules in Myo16 −/− knockout cerebellum (Figures 1H,I). We focused on excitatory synapse proteins (AMPA receptor subunits GluA1 and GluA2, scaffolding molecule PSD-95) and on inhibitory synapse proteins (GABA A receptor subunit α1, neuroligin-2). All of these proteins are present in cerebellum, including postsynaptically in PCs (Briatore et al., 2010;Yamasaki et al., 2011;Zhang et al., 2015). Using differential fractionation of cerebellar extracts, we generated a crude extract (S1), a membrane-enriched fraction (P2) and a fraction enriched for synaptosomal proteins (SYP) (Figure 1H). Quantification of GluA1, GluA2, PSD-95, GABA A α1, and neuroligin-2 amounts showed that the levels of these synaptic proteins in S1, P2, and SYP fractions are not significantly changed in the absence of myosin XVI, compared to littermate control ( Figure 1I). Together, this indicates that gross anatomical organization, as well as expression and localization of selected synaptic proteins, are unaltered in the cerebellum of Myo16 knockout mice.

Myosin XVI Localizes to Purkinje Cell Dendritic Spines
In order to obtain insight into the subcellular locations at which endogenous myosin XVI might act, we determined its distribution using fractionation of cerebellar extracts (Figure 2A). MYO16 was detected in crude extracts (S1), but also in the membrane-enriched and synaptosomal fractions (P2, SYP), suggesting the possibility that the myosin localizes at or close to synapses. Successful enrichment of membranes and synaptosomes in P2 and SYP fractions was verified via PSD-95 enrichment and loss of tubulin subunit alpha-Tubulin-4A (Figure 2A). Since MYO16 is present in cerebellar PCs (Patel et al., 2001;Cameron et al., 2007), we further examined its subcellular location in this cell type. We made use of heterogeneous cerebellar cultures transfected with PC-specific expression plasmids as described (Wagner et al., 2011b) (see section "Materials and Methods"). Observation of live PCs expressing a red-fluorescent cell volume marker and GFPtagged myosin XVI heavy chain (GFP-MYO16) showed that the myosin accumulates in essentially all dendritic spines (Figures 2B,C). Time-lapse movies of PC dendrites revealed that the GFP-MYO16 clusters change their shape over time (Supplementary Movie S1). These dynamic morphology changes were reminiscent of F-actin in the dendritic spines of live PCs (Figures 2D,E; see also Supplementary Movie S2), as visualized via F-tractin (Johnson and Schell, 2009). Indeed, GFP-MYO16 and F-tractin co-localized in spines of live PCs ( Figure 2F; see also Supplementary Movie S3). Thus, the myosin XVI heavy chain localizes to the postsynaptic, F-actin rich spines of cerebellar PCs.
Frontiers in Cellular Neuroscience | www.frontiersin.org FIGURE 2 | MYO16 localizes to the synaptosomal fraction and to Purkinje cell dendritic spines. (A) Cerebellar extracts of 3-4 week old Myo16 +/+ and Myo16 em2 knockout mice (Myo16 −/− ) were subjected to subcellular fractionation. S1, S2, P2, and SYP fractions were analyzed by Western blotting using antibodies against MYO16 (pre-adsorbed), PSD-95 (marker for membrane/synaptosomes), and alpha-Tubulin 4A (TUBA4A; marker for loss of cytosolic components). Note that specific MYO16 signal is present in P2 and SYP fractions (n = 3 experiments). (B) GFP-tagged myosin XVI heavy chain accumulates within the dendritic spines of Purkinje (Continued) Frontiers in Cellular Neuroscience | www.frontiersin.org 5 August 2019 | Volume 13 | Article 330 FIGURE 2 | Continued cells (PCs). Cultured live PC at 14 DIV and co-transfected with L7/Pcp-2 promoter plasmids for PC-specific expression of mGFP-MYO16 (green) and FusionRed (volume marker; magenta). Shown is a maximum projection of a Z-stack of images recorded using spinning disk confocal microscopy. Scale bar, 20 µm. (C) Dendrite branch of a live PC at 14 DIV transfected as in (B). Shown are images of a single Z-plane recorded using spinning disk confocal microscopy. Images correspond to a frame of a time-lapse movie (Supplementary Movie S1). Scale bar, 2 µm. (D) Cultured live PC at 15 DIV and co-transfected with L7/Pcp-2 promoter plasmids encoding the live cell F-actin marker F-tractin (green) and FusionRed (volume marker; magenta). Shown is a maximum projection of a Z-stack of images recorded using spinning disk confocal microscopy. Scale bar, 20 µm. (E) Dendrite branch of a live PC at 15 DIV transfected as in (D). Shown are images of a single Z-plane recorded using spinning disk confocal microscopy. Images correspond to a frame of a time-lapse movie (Supplementary Movie S2). Scale bar, 2 µm. (F) Dendritic spines of a live PC at 15 DIV co-transfected with plasmids encoding mGFP-MYO16 (green) and F-tractin (magenta). Shown are spinning disk confocal images of a single Z-plane that correspond to a frame in a time-lapse movie (Supplementary Movie S3). Scale bar, 5 µm. lying above the Y -axis limit is not shown; p value determined using Mann-Whitney test. * p < 0.05; * * p < 0.01; * * * p < 0.001; * * * * p < 0.0001; n.s., not significant.
assess the turnover of GFP-actin in single PC spines, we used FRAP ( Figure 3A; see also Supplementary Movie S4). To verify that FRAP of GFP-actin reveals the turnover of actin filaments in PC spines (as opposed to diffusion of GFP-actin monomers into spines), cells were treated with the F-actin stabilizing drug jasplakinolide (1 µM) (Bubb et al., 1994;Cramer, 1999). As expected, while GFP-actin fluorescence recovered in spines of vehicle-treated cells following bleaching, recovery was almost entirely blocked upon jasplakinolide treatment ( Figure 3B). This confirms that the observed fluorescence recovery reflects formation of new F-actin in PC spines.
To determine whether myosin XVI is required for F-actin turnover, we performed FRAP analysis of GFP-actin in spines of Myo16 −/− PCs (Myo16 em2 allele; Figure 3C). The plateau of fluorescence recovery indicates the fraction of F-actin that undergoes turnover (referred to as the mobile F-actin pool), while the recovery time constant (τ) is a measure of F-actin turnover rate in spines (Star et al., 2002). Our analyses did not reveal a significant difference regarding the mobile pool of spine F-actin when comparing Myo16 −/− to wild-type littermate PCs (71 vs. 67%, respectively; Figure 3D). However, F-actin turnover rate was significantly faster (i.e., τ was smaller) in Myo16 −/− PC spines compared to control (τ = 23 s vs. τ = 32 s, respectively; Figure 3E). We independently confirmed these results using PCs from the knockout mouse line carrying the Myo16 em3 allele (Figures 3F,G). Therefore, F-actin turnover in PC spines is accelerated upon Myo16 knockout.
Since the actin cytoskeleton is a crucial determinant of spine shape (Hotulainen and Hoogenraad, 2010), we also monitored whether overall morphology and dynamics of spines are changed in Myo16 −/− PCs. Using images of unbleached spines recorded during the GFP-actin FRAP experiments, a spine circularity index was determined as a measure for spine shape ( Figure 3H). Moreover, we monitored spine area (an indirect measure of spine size; Figure 3I), relative spine area size changes over 2.5 min (Figure 3J), and relative changes of actin fluorescence intensity in spines over time ( Figure 3K). None of these parameters were significantly different in Myo16 −/− PCs when compared to wild-type. Thus, overall morphology and dynamics of PC spines appear to be normal in the absence of myosin XVI.

Myosin XVI Acts Within Purkinje Cells to Regulate Spine F-Actin Turnover
Myosin XVI might affect F-actin turnover directly via acting in PC spines. Alternatively, altered F-actin dynamics in Myo16 −/− PCs might be of non-cell autonomous origin such as altered presynaptic input from granule cells. Thus, to determine whether the myosin is required within PCs, we performed PC-specific Myo16 knockdown. First, we identified RNAi sequences that, when embedded in a miR backbone, knockdown Myo16 expression ( Figure 4A). Compared to control levels, the independent knockdown sequences Myo16 KD3 and Myo16 KD5 significantly reduced GFP-MYO16 expression to 33% and 28%, respectively ( Figure 4B). Following a previously used strategy (Alexander and Hammer, 2016), these knockdown sequences as well as a reporter (FusionRed) were placed under control of the L7/Pcp-2 promoter ( Figure 4C) which drives PC-specific expression in cerebellar cultures (Wagner et al., 2011b). Expression of the Myo16 KD3, Myo16 KD5 or scrambled miR knockdown constructs in the analyzed PCs was verified via monitoring FusionRed.
Fluorescence recovery after photobleaching analysis of GFPactin showed that, similar to the situation in Myo16 −/− PCs, the two independent PC-specific Myo16 knockdown constructs both caused a significantly faster turnover of F-actin in spines compared to scrambled control (Myo16 KD3, τ = 21 s; Myo16 KD5, τ = 19 s; Scrambled, τ = 28 s), while the mobile pool of F-actin remained unchanged (Figures 4D-G). Moreover, as in case of Myo16 −/− PCs, no consistent changes in terms of overall morphology and dynamics of spines were observed upon Myo16 knockdown (Figures 4H-K). Therefore, both global Myo16 knockout and PC-specific Myo16 knockdown lead to an identical phenotype, i.e., accelerated F-actin turnover in PC-spines. Since GFP-MYO16 targets to PC spines (Figure 2), this indicates that the myosin functions at the postsynaptic side to attenuate spine F-actin turnover.

Faster F-Actin Turnover in Purkinje Cell Spines Upon WRC Inhibition
The WRC is an activator of Arp2/3-mediated F-actin polymerization (Bisi et al., 2013) and a well-established interaction partner of myosin XVI (Yokoyama et al., 2011). Therefore, we asked whether also the WRC is important for actin cytoskeleton dynamics in dendritic spines of PCs. We first determined whether the WRC-subunit WAVE1 localizes to PC spines in situ. We employed an anti-WAVE1 antibody that detects a band of expected size in cerebellar extracts ( Figure 5A). Immuno-fluorescence labeling of wild-type cerebellum confirmed (Soderling et al., 2003) the presence of WAVE1 in PCs and revealed partial colocalization with PC spines ( Figure 5B). Immuno-electron microscopy showed that WAVE1 is present at pre-and postsynaptic sites within the cerebellar molecular layer ( Figure 5C, Myo16 +/+ panels). In spines, predominantly the postsynaptic density was labeled ( Figure 5C, left Myo16 +/+ panel). In cerebellum of Myo16 −/− mice, no obvious difference in WAVE1 labelling was observed ( Figure 5C, Myo16 −/− panels). Thus, WAVE1 is present in PC spines, and myosin XVI is not essential for WAVE1 targeting to spines.
To interfere with WRC function, we made use of a wellcharacterized dominant-negative version of WAVE1 that lacks the C-terminal VCA domain (WAVE1 VCA) and thus is unable to activate Arp2/3 (Miki et al., 1998;Kim H.J. et al., 2006;Pils et al., 2012). To selectively target the PCs in heterogeneous cerebellar culture, WAVE1 VCA was expressed under control of the L7/Pcp-2 promoter. GFP-actin FRAP analysis revealed that WAVE1 VCA leads to a significantly smaller mobile F-actin pool in PC spines (FRED-WAVE1 VCA: FIGURE 4 | Purkinje cell-specific Myo16 knockdown leads to accelerated F-actin turnover in dendritic spines. (A,B) Identification of Myo16 miR RNAi knockdown constructs. (A) Western blot analysis of HEK293 cells co-transfected with a plasmid encoding mGFP-tagged mouse MYO16 and pcDNA TM 6.2-GW/EmGFP-miR plasmids carrying the indicated knockdown sequences (Myo16 KD1-KD5) or scrambled control (Scr.). For control, untransfected cells (first lane) and cells transfected only with plasmid encoding mGFP-Myo16 were used. Antibodies against GFP and α-Tubulin (α-Tub; loading control) were used. (B) Quantification of GFP-MYO16 protein amount upon co-transfection with Myo16 knockdown constructs KD1-KD5 or scrambled control. Graph depicts GFP-MYO16 signals normalized to tubulin signals and expressed as percentage of control (i.e., cells transfected with GFP-MYO16 plasmid only). Bars indicate mean values (n = 3-4; data points represent experiments) ± SEM (magenta); p values determined using Kruskal-Wallis test (p = 0.0321) followed by Dunn's multiple comparisons test. (C) Schematic representation of plasmids for expressing Myo16 KD3, Myo16 KD5 or scrambled sequence (pre-miRNA) and flanking miR sequences (green) together with FusionRed as a reporter for RNAi expression (magenta) under control of the PC-specific L7/Pcp-2 promoter. (D,E) FRAP analysis of GFP-actin in spines of wild-type PCs transfected with L7/Pcp-2 promoter plasmids carrying Myo16 KD3, Myo16 KD5, or scrambled sequence and FusionRed (reporter for knockdown construct expression). Cells were co-transfected with a plasmid encoding GFP-actin. Graphs depict recovery of GFP-actin fluorescence intensity in spines, data points represent the mean of a representative experiment. For clarity, recovery curves of Myo16 KD3 (D) and Myo16 KD5 (E) are compared to the same scrambled control but shown in separate graphs. (F) GFP-actin FRAP recovery plateau in spines of PCs transfected as described in (D,E). Data are plateau values obtained from independent experiments (n = 7-11; magenta line indicates median); p value determined using Kruskal-Wallis test. (G) GFP-actin FRAP recovery time constant (τ) in spines of PCs transfected as described in (D,E). Data are τ values from independent experiments (n = 7-11; magenta line indicates mean); p values determined using one-way ANOVA (p = 0.0038) followed by Tukey's multiple comparisons test. (H) Circularity index of spines of PCs transfected as described in (D,E). Data points represent single spines, magenta line indicates median; p value determined using Kruskal-Wallis test. (I) Apparent area covered by single spines of PCs transfected as described in (D,E). Data points represent single spines, magenta line indicates median; p value determined using Kruskal-Wallis test. (J) Spine area changes over time of PCs transfected as described in (D,E). Data points represent SD of the relative area change of single spines over 150 s, magenta line indicates median; p values determined using Kruskal-Wallis test (p = 0.0008) followed by Dunn's multiple comparisons test. (K) Change of GFP-actin fluorescence intensity over time in spines of PCs transfected as described in (D,E). Data points represent SD of the relative fluorescence change of single spines monitored over 150 s, magenta line indicates median. For reasons of comparability with the other figures, a single data point of Myo16 KD5 lying above the Y -axis limit is not shown; p value determined using Kruskal-Wallis test. * p < 0.05; * * p < 0.01; * * * p < 0.001; * * * * p < 0.0001; n.s., not significant.
Analyses of PCs expressing GFP-actin showed furthermore that WAVE1 VCA leads to changes in spine morphology. Compared to control, spines of PCs that express WAVE1 VCA were more elongated and larger (Figures 6D,E . Time is indicated in seconds. Scale bar, 2 µm. * p < 0.05; * * p < 0.01; * * * p < 0.001; * * * * p < 0.0001; n.s., not significant. spines over time (Figures 6F,G). This indicates that the WRC promotes circular spine shape and favors spine dynamics in PCs.

Arp2/3 Inhibition, but Not Formin Inhibition, Accelerates F-Actin Turnover in PC Spines
Since Arp2/3 is the downstream target of the WRC (Takenawa and Suetsugu, 2007;Bisi et al., 2013), reduced Arp2/3 activation might be the cause of accelerated F-actin turnover upon WRC inhibition ( Figure 6C). To determine whether Arp2/3 plays a role for F-actin turnover in PC spines, cells were acutely exposed to CK-666, a well-established small molecule inhibitor of Arp2/3 (Nolen et al., 2009), before performing FRAP analysis of GFP-actin ( Figure 7A). Similar as for WRC inhibition, spines of PCs treated with 200 µM CK-666 displayed a significantly smaller mobile F-actin pool (CK-666, plateau at 43%; vehicle control, plateau at 65%; Figures 7B,C). Moreover, like when interfering with the WRC or with myosin XVI, turnover of the mobile F-actin pool was significantly faster upon CK-666 treatment (CK-666, τ = 20 s; vehicle control, τ = 30 s; Figure 7D). Together, our data indicate that the WRC-Arp2/3 pathway determines the relative level of mobile F-actin in PC spines and attenuates its turnover time. We next tested whether also manipulation of actin polymerization factors unrelated to Arp2/3 results in accelerated F-actin turnover in PC spines. For this purpose, cells were acutely treated with SMIFH2 (40 µM), an inhibitor of forminmediated actin filament nucleation and elongation (Rizvi et al., 2009). Similar to interference with WRC-Arp2/3, this led to a significantly smaller mobile F-actin pool compared to vehicle control (SMIFH2, plateau at 40%; Figures 7B,C). Thus, formin activity is required for F-actin dynamics in PC spines. Notably, however, the turnover time of the dynamic pool was not reduced compared to control (SMIFH2, τ = 29 s; Figure 7D). Moreover, the effects of CK-666 and SMIFH2 on mobile pool size of F-actin were additive, with both blockers leading to a significantly reduced plateau compared to CK-666 alone (CK-666+SMIFH2, plateau at 30%; Figure 7C). Therefore, WRC-Arp2/3 and formins appear to act in parallel pathways influencing F-actin dynamics in PC spines, with WRC-Arp2/3 -but not formins -attenuating F-actin turnover rate.
Since myosin XVI attenuates F-actin turnover rate in PC spines (Figures 3, 4) and binds to the WRC (Yokoyama et al., 2011), the myosin might act through the WRC-Arp2/3 pathway. If this is the case, Myo16 knockout may not have an additive effect on actin dynamics upon Arp2/3 inhibition. Indeed, neither the mobile pool of spine F-actin nor its turnover rate was significantly different in spines of CK-666-treated Myo16 −/− PCs when compared to CK-666-treated littermate, wild-type PCs (Figures 7E,F). This argues for myosin XVI regulating F-actin turnover rate in PC spines via Arp2/3. Finally, we examined whether acute exposure to the Arp2/3 or formin inhibitors changes overall spine morphology or dynamics (Figures 7G-O). Similar to WRC inhibition ( Figure 6D) or long-term (24 h) treatment with CK-666 (Kawabata Galbraith et al., 2018), short-term Arp2/3 inhibition led to a significantly more elongated spine shape ( Figure 7G). Moreover, CK-666 treatment resulted in increased spine size ( Figure 7H). Size changes of PC spines or actin intensity changes in spines over 2.5 min were not significantly affected by Arp2/3 inhibition (Figures 7I,J). This indicates that the WRC-Arp2/3 pathway promotes a circular spine shape in PCs, consistent with Arp2/3 driving the formation of a non-uniformly oriented, branched actin meshwork (Chazeau and Giannone, 2016). In contrast, acute formin inhibition affected PC spine morphology and dynamics differently (Figures 7K-O). Unlike CK-666, SMIFH2 did not lead to spine elongation ( Figure 7K) but significantly reduced relative spine size changes and actin fluorescence intensity changes in spines over time (Figures 7M,N). Similar to CK-666 treatment, PC spines adopted a larger area upon exposure to SMIFH2 ( Figure 7L). Therefore, unlike Arp2/3, formin activity is needed to promote fluctuations of PC spine size and actin content.

Altered Synaptic Structure and Transmission in the Cerebellum of Myo16 Knockout Mice
Having established that myosin XVI regulates actin dynamics in cerebellar PCs, we examined in more detail whether synapses formed on PC spines are structurally altered in Myo16 −/− mice in situ. First, using electron microscopy images of the middle third of the molecular layer, we quantified the number of asymmetric synapses per µm 2 ( Figure 8A). No difference in the density of these excitatory synapses that mainly represent PF-PC synapses was detected. Consistent with the finding that spine size of cultured Myo16 −/− PCs is unaltered (Figure 3I), also spine head area of molecular layer synapses measured from electron microscopy images was unaltered (Figures 8B,C).  Moreover, PSD length, PSD area and presynaptic vesicle density were unchanged at these synapses (Figures 8B,C). Strikingly, however, the area of presynaptic axon terminals and the absolute number of synaptic vesicles per terminal were dramatically reduced in the molecular layer of Myo16 −/− cerebellum. Therefore, myosin XVI appears to be important for presynaptic ultrastructure and synaptic vesicle numbers at PF-PC synapses. Both altered postsynaptic actin dynamics and the ultrastructural abnormalities at presynaptic terminals observed . Data points represent analyzed PCs (Myo16 +/+ : n = 17 from seven mice; Myo16 −/− : n = 15 from nine mice). mEPSC amplitude: median ± 95% confidence interval is indicated (magenta); p value determined using Mann-Whitney test. mEPSC inter-event interval: mean ± SEM is indicated (magenta); p value determined using Welch's t-test. (C) PCs filled with biocytin during mEPSC measurement were stained and visualized using confocal microscopy. Images show Z-stack projections. Scale bar, 50 µm. (D) Higher magnification images of PC dendrites with spines obtained as in (C). Scale bar, 2 µm. Graph depicts number of spines per µm dendrite. Data points represent analyzed PCs, mean ± SEM is indicated (magenta); p value determined using Student's t-test. (E) Model of the role of myosin XVI at PF-PC synapses. In PC dendritic spines, myosin XVI is required for the attenuation of F-actin turnover. We propose that the myosin functions by promoting WRC activation which, in turn, activates Arp2/3-mediated branched F-actin formation. In addition to WRC and Arp2/3, also formins promote F-actin mobility in PC spines. However, interference with myosin XVI, WRC, and Arp2/3, but not with formins, accelerates F-actin turnover rate, consistent with formins acting in a separate pathway. At the presynaptic side, loss of myosin XVI leads to reduced numbers of synaptic vesicles and terminals of apparently reduced size. This may be a consequence (i) of a role of the myosin at the presynaptic side (e.g., in F-actin turnover) or (ii) of the postsynaptic changes in Myo16 −/− PCs. * p < 0.05; * * p < 0.01; * * * p < 0.001; * * * * p < 0.0001; n.s., not significant.
in the absence of myosin XVI might affect the function of PF-PC synapses. To test whether Myo16 knockout alters synaptic transmission onto PCs, we measured spontaneous AMPA receptor-mediated mEPSCs from PCs in acute cerebellar slices ( Figure 9A). No difference in the peak amplitude of mEPSCs was observed when comparing Myo16 −/− and wild-type littermate PCs (Figure 9B), suggesting that synaptic AMPA receptor content is unchanged. In contrast, the mEPSC inter-event interval was significantly longer in Myo16 −/− PCs (Figure 9B). Such a reduced frequency of mEPSCs may arise from a reduced number of synapses formed on the PCs. However, no gross abnormalities in dendritic arborization of the analyzed PCs was observed ( Figure 9C). Moreover, spine density on Myo16 −/− PCs was unaltered (Figure 9D), and the density of excitatory molecular layer synapses was unchanged in Myo16 −/− cerebellum ( Figure 8A). Therefore, the increase in mEPSC inter-event interval provides further support for a reduced number In the open field test, both genotypes showed similar locomotor habituation to the paradigm as indicated by the decrease in distance moved (genotype: p = 0.42), decreased mean velocity (genotype: p = 0.88) and an increased mean distance to wall (less thigmotaxis) (genotype: p = 0.47) over a 30 min test duration. (B) Latency to fall from an accelerating rotarod revealed significant increase of duration over trials (p < 0.001) with no effect of genotype (p = 0.77). (C) Both genotypes show similar performance in marble burying (p = 0.61). (D) Time exploring an empty cup vs. a novel mouse, or a familiar vs. a novel mouse during the first 2 min of a social recognition test. Mice spent more exploration time with the novel mouse vs. the empty cup (p < 0.01, no effect of genotype: p = 0.17) and more time investigating the novel mouse vs. the familiar mouse (choice: p < 0.001, no effect of genotype: p = 0.82). The discrimination index showed a significant preference of Myo16 −/− but not the Myo16 +/+ mice for the novel mouse over the empty chamber (p < 0.05; p = 0.372, respectively) and a significant preference of both genotypes for the novel over the familiar mouse (p < 0.05; p < 0.001, respectively). Values represent mean ± SEM (A-D: repeated measures ANOVA, Discrimination index: one-sample t-test, test value: 50). Myo16 −/− and Myo16 +/+ mice: n = 11 each. * p < 0.05; * * p < 0.01; * * * p < 0.001; * * * * p < 0.0001; n.s., not significant.
of presynaptic vesicles at PC synapses in Myo16 knockout cerebellum (Figures 8B,C).

Myo16 Knockout Mice Show Comparable Performance to Controls in Motor-and Social Interaction-Tests
Given the synaptic roles of myosin XVI identified so far (Figure 9E), we examined whether Myo16 −/− mice display behavioral abnormalities. In the open field test (Figure 10A), Myo16 −/− mice were indistinguishable from wild-type littermate controls in terms of locomotor activity (total distance moved, velocity) and regarding a measure of anxiety (mean distance to wall). Since we found that MYO16 is important for F-actin dynamics in cerebellar PCs and that synaptic organization in the cerebellum is altered in Myo16 −/− mice, we next monitored the motor learning ability of the knockout mice. In the accelerating rotarod test, both Myo16 −/− and wild-type control mice improved their performance over the first five test days, without differing significantly between genotypes ( Figure 10B). Therefore, mice lacking myosin XVI perform similar to wildtype controls in this test for motor learning. Given the genetic association of MYO16 and ASD, we also examined Myo16 −/− mice for behaviors altered in mouse models of ASD. A marble burying test did not reveal differences in repetitive behavior of Myo16 −/− mice compared to littermate controls ( Figure 10C). Moreover, a social interaction test did not reveal deficits of Myo16 −/− mice in their ability to discriminate between an empty chamber or a novel mouse, or between a familiar and a novel mouse ( Figure 10D). Therefore, at least this limited set of behavioral assays did not reveal gross abnormalities in mouse behavior upon Myo16 knockout.

DISCUSSION
Here, we define myosin XVI as a novel component of the regulatory machinery of actin turnover in dendritic spines of cerebellar PCs, and we uncover that myosin XVI is important for normal presynaptic organization in the cerebellum.
Our results shed light on the mechanisms that govern F-actin turnover in spines of PCs, a neuronal cell type implicated in motor learning, social behavior and ASD. We show that, similar to other neurons, the WRC-Arp2/3 pathway is crucial for the regulation of postsynaptic actin turnover in PCs. Interference with either WRC or Arp2/3 led to a reduced mobile pool and faster F-actin turnover in spines (FRAP-and PC spine measurement-results are summarized in Table 1). Notably, our observation that acute inhibition of Arp2/3 accelerates turnover rate in dendritic spines is consistent with the finding that CK-666 increases retrograde actin flow rates at the leading edge of neuronal growth cones (Yang et al., 2012). In comparison, diverse outcomes for postsynaptic F-actin turnover upon interfering with Arp2/3 or WRC have been reported previously for other neurons. For example, sparse in vivo knockout of the Arp2/3 subunit ArpC3 in a subset of hippocampal neurons leads to a reduced mobile pool and slower recovery (Kim et al., 2013). Moreover, WAVE1 knockdown was found to reduce the mobile pool of F-actin in cortical spines (Njoo et al., 2015). In contrast, heterozygous knockout of the WRC subunit Cyfip1 increases the mobile pool while leaving the recovery rate of spine F-actin unchanged (Pathania et al., 2014). While Arp2/3 drives the formation of a non-uniformly oriented F-actin meshwork, formins promote the formation of linear actin filaments. We show that formins are also crucial for actin cytoskeleton dynamics in PC dendritic spines, as formin inhibition dramatically reduces the size of the mobile F-actin pool ( Table 1). The effect of SMIFH2 on mobile pool size being additive to CK-666 supports the view that formins promote F-actin dynamics in a pathway parallel to Arp2/3. Moreover, in contrast to Arp2/3, inhibition of formins leaves F-actin turnover rate in PC spines unchanged but dampens spine size fluctuations.
Finally, PC spines become more elongated upon interfering with Arp2/3 or WRC (Table 1), consistent with formin-mediated polymerization of linear F-actin remaining. Along this line, depletion of MTSS1, a positive regulator of Arp2/3-mediated F-actin formation and negative regulator of formin DAAM1, also leads to longer PC spines (Saarikangas et al., 2015;Kawabata Galbraith et al., 2018).
We measured an increased size of PC spines upon dominantnegative inhibition of WRC ( Figure 6E) and upon acute inhibition of Arp2/3 or formins (Figures 7H,L). This is surprising, given that in hippocampal neurons Arp2/3 is essential for the development of bigger, mature spines (Spence et al., 2016). In part, the distinct outcome may be due to different approaches taken (e.g., acute inhibition vs. genetic depletion). However, we propose that differences in the inherent properties of PC spines compared to hippocampal spines are also a major factor. For example, PC spines form independently of presynaptic input following an intrinsic program (Sotelo, 1978). Moreover, unlike hippocampal spines, PC spines do not undergo drastic morphological changes during plasticity (Sdrulla and Linden, 2007). Finally, in addition to WRC and Arp2/3, also specific F-actin regulators exist in PCs that are not found in other neurons (Miyagi et al., 2002;Sekerkova et al., 2003). Increased PC spine size might affect second messenger signaling within these microcompartements, leading to alteration in synaptic plasticity at PF-PC synapses.
Our data suggest a model where myosin XVI activates the WRC-Arp2/3 pathway in PC spines in order to affect F-actin turnover ( Figure 9E). First, Myo16 depletion via knockout or PC-specific knockdown phenocopies interference with the WRC-Arp2/3 pathway in terms of accelerated F-actin turnover rate in PC spines (Table 1). Second, there was no additive effect of Myo16 knockout on F-actin turnover upon CK-666 treatment (Figures 7E,F), indicating that the myosin acts through Arp2/3. Third, the model is also supported by the known physical interaction of myosin XVI with the WRC (Yokoyama et al., 2011) and by the localization of both GFP-MYO16 and the WRC component WAVE1 to PC spines (Figures 2, 5). In contrast to interference with WRC and Arp2/3, however, depletion of myosin XVI did not alter the relative size of the mobile F-actin pool and spine circularity ( Table 1). This is consistent with the existence of alternative, myosin XVI-independent ways of activating WRC-Arp2/3. Similar to certain synaptic scaffolding proteins (e.g., IRSp53, nArgBP2; Takenawa and Suetsugu, 2007;Lee et al., 2016) and transmembrane proteins (e.g., neuroligins, protocadherins; Chen et al., 2014), myosin XVI might recruit the WRC at specific sites and facilitate its local activation. Nevertheless, our data argue against a general requirement of the myosin for localizing the WRC to spines, as WAVE1 is still present in spines of Myo16 knockout cerebellum in situ. Myosin XVI might also function by bringing PP1c and/or PI3K in close proximity to the WRC in PC spines, thereby helping to regulate the local phosphorylation status -and thus activity -of the WRC.
Genetic ablation of Arp2/3 and WRC components leads to deficits in hippocampal synaptic transmission and to behavior deficits (Soderling et al., 2003(Soderling et al., , 2007Kim et al., 2013;Spence et al., 2016). Therefore, also disrupted postsynaptic F-actin turnover caused by Myo16 knockout may result in altered synaptic targeting of neurotransmitter receptors or altered spine morphology. However, our analyses did not reveal changes in PC spine density or spine morphology upon Myo16 knockout. Moreover, subcellular fractionation and measurement of mEPSC amplitudes under basal activity indicate that synaptic AMPA receptor numbers are normal in the absence of myosin XVI. Notably, it has been suggested that the late phase of longterm depression in PCs requires actin polymerization (Smith-Hicks et al., 2010). Thus, while we did not detect alterations in postsynaptic morphology or function in Myo16 −/− cerebellum, it is possible that myosin XVI-dependent actin turnover in PC spines specifically affects synaptic plasticity at PF-PC synapses. Importantly, the absence of a deficit in the rotarod test does not rule out that PC-dependent motor learning impairment is present in Myo16 knockout mice and may be detected using other tests (compare e.g., with Ha et al., 2016).
Both ultrastructural analysis of the cerebellar molecular layer and mEPSC measurements pointed toward a striking presynaptic deficit at Myo16 −/− PF-PC synapses. While synapse density was unchanged, a longer mEPSCs inter-event interval was observed, consistent with less spontaneous fusion events of synaptic vesicles in the absence of myosin XVI. This corroborates the reduction in presynaptic vesicle numbers visualized via electron microscopy. Notably, F-actin is also enriched at presynaptic sites, and presynaptic actin dynamics are thought to play a role for synaptic vesicle fusion (Rust and Maritzen, 2015). Moreover, Wave1 knockout leads to altered morphology of axon terminals (Hazai et al., 2013), and we observed that WAVE1 localizes presynaptically at molecular layer synapses ( Figure 5C). This raises the possibility that myosin XVI regulates actin dynamics at the presynaptic side to promote normal terminal size and synaptic vesicle numbers ( Figure 9E). Future experiments will need to test this possibility.
Interference with the regulation of F-actin dynamics may lead to behavioral alterations. For example, Wave1 knockout mice display motor coordination problems (Soderling et al., 2003). Moreover, an acute requirement of Arp2/3 activity for fear memory formation was demonstrated by microinjecting CK-666 into the lateral amygdala of live rats during fear conditioning (Basu et al., 2016). Since we observed altered F-actin turnover in PC spines and altered synaptic transmission onto PCs upon Myo16 knockout, we carried out an initial behavioral characterization of Myo16 −/− mice. Notably, while common variations at the MYO16 locus are associated with ASD (Wang et al., 2009), it is not known if Myo16 kockout mimics the effect of these alleles. Nevertheless, we included tests for ASDlike behavior. We did not detect phenotypes of Myo16 −/− mice regarding locomotion activity, cerebellum-dependent motor learning, repetitive behavior and social interaction. This suggests that more elaborate future experiments will be needed to determine whether myosin XIV and/or actin turnover play a role for cerebellum-dependent functions such as motor learning or social behavior. Importantly, genetic evidence also indicates links of MYO16 to schizophrenia (Rodriguez-Murillo et al., 2014) and to bipolar disorder (Kao et al., 2016). Moreover, epigenetic changes at the MYO16 locus have been detected in depressed patients (Gross et al., 2017) and following learning tasks in mice (Koberstein et al., 2018). Together, this underscores the significance of determining in the future whether myosin XVI is crucial for the function of neuronal circuitry in vivo, and whether this involves regulation of F-actin dynamics.

Protein Extracts, Fractionation, and Western Blot Analyses
For detection of MYO16 and WAVE1 in cerebellar crude extracts (Figures 1C, 5A), single cerebella were lysed in ice cold lysis buffer (150 mM NaCl, 1% IGEPAL R CA-630 [I8896, Sigma-Aldrich], 50 mM Tris-Cl pH8.0, protease inhibitors [cOmplete TM Protease Inhibitor Cocktail; #04693116001, Roche]) using 12 strokes with Potter S Homogenizer (Sartorius AG). The homogenate was centrifuged at 600 × g for 10 min, 4 • C and the supernatant was mixed with 20% (v/v) loading dye (5% β-mercaptoethanol, 0.02% bromophenol blue, 30% glycerol, 10% sodium dodecyl sulfate [SDS], 250 mM Tris-Cl pH 6.8) and denatured (10 min, 97 • C). Subcellular fractionation of cerebellar extracts (Figures 1H,I) to yield a crude extract (S1), a membrane-enriched fraction (P2), a cytosolic fraction (S2), and a synaptosomal fraction (SYP) was carried out by lysing a complete cerebellum in Sucrose 1 buffer (320 mM sucrose, 1 mM NaHCO 3 , 1 mM MgCl 2 , 500 µM CaCl 2 , 1 µM PMSF) in the presence of EDTA-free protease inhibitors (cOmplete Tablets, EASYpack, 04693132001, Roche) and phosphatase inhibitors (PhosSTOP, EASYpack, 04906837001, Roche), by potterization in 2 ml teflon tubes at 800 rpm and 12 strokes. S1 post-nuclear fraction was obtained by centrifugation of the total lysate at 1,400 × g for 10 min at 4 • C. A fraction of S1 was conserved for further analysis and the rest was centrifugated at 13,800 × g for 10 min at 4 • C to obtain S2 soluble and P2 membranous fractions. S2 fraction was conserved and P2 was resuspended in Sucrose 2 buffer (320 mM sucrose, 1 mM NaHCO 3 ). P2 was shaken for 20-30 min at 4 • C. Sucrose gradient for synaptosomal purification was achieved by sequentially adding three distinct layers of 1.2 M, 1 M, and 0.8 M sucrose in the presence of NaHCO 3 . A part of the P2 fraction was conserved for analysis and 400 µl were placed on top of the gradient and centrifugated at 22,000 rpm using SW40 Ti rotor for 2 h at 4 • C. Synaptosomes were isolated from the interface between 1.2 and 1 M fractions of the gradient using a 1 ml syringe. In order to assess myosin XVI localization in synaptosomes (Figure 2A), an identical protocol was used, however, omitting phosphatase inhibitors and PMSF from sucrose buffer 1 and, occasionally, NaHCO 3 from sucrose gradient. Protein concentration was quantified with Pierce TM BCA Protein Assay kit (#23227, Thermo Fisher Scientific), and identical amounts of proteins were mixed with loading dye and denatured as above.

Live Cell Microscopy
Cultured PCs were observed using a spinning disk confocal microscope (Visitron Systems GmbH) consisting of an inverted microscope (Nikon Ti-E) equipped with a spinning disk (Yokogawa), solid state lasers (405, 488, 561, and 647 nm), a 100× objective, EM-CCD cameras (Hamamatsu Photonics), autofocus system (Nikon TI-ND6-PFS), a multi-point FRAP/photoactivation module (VisiFRAP, Visitron Systems GmbH), and an incubation chamber for controlled environmental conditions (37 • C, 5% CO 2 ). During imaging, cultures were kept in their conditioned growth medium. Images shown in Figures 2B,D were obtained by recording Z-stacks (0.3 µm Z-plane distance) and subjecting them to noise removal (denoise, low pass filter; VisiView R software, Visitron Systems GmbH). Maximum projections were generated and stitched using Fiji/MosaicJ (Thevenaz and Unser, 2007;Schindelin et al., 2012). Images shown in Figures 2C,E,F, as well as Supplementary Movies S1-S3 were obtained recording at a frame rate of 0.5/s, followed by noise removal (de-noise, low pass filter; VisiView R software, Visitron Systems GmbH). Subsequently images were processed in Fiji (Schindelin et al., 2012) using bleach correction (histogram matching) and generation of a three frame rolling average image.

Fluorescence Recovery After Photobleaching
To measure F-actin turnover in PC spines, FRAP experiments were performed with cultured PCs expressing GFP-actin at 14-15 DIV using the spinning disk microscope with a 100× objective. Images were obtained using 488 nm excitation and an exposure time of 500 ms. In each trial, following five frames of starting point recording (frame rate of 1/s), the GFP-actin fluorescence signal of six single spines on a PC was bleached by directing 405 nm laser light to a circular region (12 pixel diameter) placed on the spine (∼2 s total bleach time for six spines). Imaging was resumed immediately after and continued for 145 s at a frame rate of 1/s. If indicated, the following compounds were added to the culture medium prior to carrying out FRAP recordings: jasplakinolide (#420127-50UG, Merck; added to final concentration of 1 µM; FRAP performed 10-60 min after addition), CK-666 (#3950, Tocris Bioscience; added to final concentration of 200 µM; FRAP performed within 90 min after addition), SMIFH2 (S4826, Sigma-Aldrich; added to final concentration of 40 µM; FRAP performed 5-50 min after addition, as with combination of SMIFH2 and CK-666).
GFP-actin fluorescence intensity signal in bleached spines was quantified using Fiji (Schindelin et al., 2012;Schneider et al., 2012). Intensity was measured from a region of constant size encompassing the spine throughout the duration of the experiment. To determine recovery of the bleached fluorescence signal, the intensity measured in each frame was reduced by the intensity remaining in the first frame after bleaching, yielding baseline-corrected intensity values F b . To correct for overall bleaching during the experiment, F b was multiplied with a bleach factor (y) calculated for each trial and time point, yielding bleach-corrected intensity values (F bc ). To obtain y, fluorescence intensity in each frame was measured from part of the cell that included only spines not targeted by FRAP bleaching. After background correction of the measured intensities, bleach factor y was determined for each time point using Excel (Microsoft) by fitting a curve described by the exponential equation y = at b (a and b, fitted variables; t, time) to the values obtained by dividing the initial, background-corrected intensity by the background-corrected intensity at each time point. The mean of F bc (the bleach-corrected intensity values) before FRAP bleaching was normalized to 100% (F 0 ), and F bc values at each time point after bleaching were expressed relative to F 0 (GFP-actin fluorescence,% of initial), yielding the fluorescence recovery curve. For each independent experiment (i.e., for each culture of transfected PCs derived from an individual embryo), a mean recovery curve was calculated from the recovery curves of roughly 50 spines on average, with up to six curves originating from a single FRAP trial. To calculate plateau and time constant τ of GFP-actin fluorescence intensity recovery in spines, the mean recovery curves from each experiment were fitted to the equation y, GFP-actin fluorescence intensity; x, time; P, plateau; k, inverse of τ. In case a mean recovery curve showed an intensity decrease of more than five percentage points at any time during the recovery phase, the values following the pre-decrease maximum value were ignored when fitting the equation. If this led to less than 70 s of recovery time that could be fitted, the data were excluded entirely from analysis. Cells that were immotile and stiff (i.e., apparently dead) during FRAP recording were also excluded from analyses. Spines were excluded if the fluorescence signal was oversaturated, if the spine overlapped with another spine during the 150 s of recording, or if the fluorescence signal was not bleached close to background level.

Spine Morphology and Dynamics of Live PCs
The following parameters were determined using Fiji and the time-lapse movies of GFP-actin recorded during the FRAP experiments. 50 spines (five spines from ten PCs) that were not FRAP-bleached were analyzed for each condition. The area covered by randomly selected, individual spines was determined in each frame by selecting a threshold at the level of dendritic shaft GFP-actin signal intensity, followed by an automatic creation of a smoothened region of interest (ROI) around the above-threshold spine GFP-actin signal in each frame. The "spine area" was determined by averaging the ROI area of the first three frames (1 s recording interval) for each spine. The variation of the spine area during 2.5 min (150 frames of recording), denoted as "SD (area over time), " is the standard deviation of the relative area (% of initial) over 150 frames, with the average area of the first three frames representing 100%. A circularity index (CI) was calculated from the ROI using the formula CI = 4π a p 2 (a, area; p, perimeter) (Rubio et al., 2011) and is given as the average value of the first three frames (1 s recording interval) for each spine. The variation of the GFP-actin signal intensity in spines during 2.5 min (150 frames of recording), denoted as "SD (intensity over time), " is the standard deviation of the relative fluorescence intensity (% of initial) over 150 frames, with the average intensity of the first 20 frames representing 100%. Spine GFP-actin signal intensity was determined for each frame by measuring the integrated intensity within the above determined ROI, followed by background subtraction, bleaching correction and calculation of a 20 frame rolling average intensity value.

Immuno-Histochemistry and Immuno-Electron Microscopy
For WAVE1 immuno-histochemistry (Figure 5B), ∼6 month old adult mice were anesthetized by a mixture of ketanest and rompun and transcardially perfused with 4% paraformaldehyde (PFA) in phosphate-buffer (PB). Brains were post-fixed in PFA/PB (overnight, 4 • C) before 50 µm vibratome sections were cut. Brain sections were washed in PBS, incubated for 10 min in 0.5% (w/v) NaBH 4 in PBS, washed again in PBS, and blocked for 30 min in PBS containing 10% (v/v) horse serum (HS), 0.3% (w/v) BSA and 0.3% (v/v) Triton X-100. Sections were then incubated with primary antibodies diluted in carrier (PBS containing 1% HS, 0.2% BSA and 0.3% Triton X-100) for 24 h at 4 • C with gentle agitation, washed in PBS, and incubated for 2 h with secondary antibodies diluted in carrier. After wash in PBS, sections were mounted on glass slides using ProLong TM Gold Antifade Mountant (P36930; Thermo Fisher Scientific), imaged on an Olympus FV1000 confocal laser scanning microscope (60× objective; 0.8 µm Z-plane thickness).
Similarily, for immuno-histochemistry shown in Figures 1D-G, 19-26 week old mice were perfused as above. Postfixation was carried out for 48 h, followed by cryoprotection of the brains in 30% Sucrose/PBS. Brains were frozen in TissueTek at -80 • C and 40 mm thick sagittal sections were obtained with a cryostat (Cryostar NX70, Thermo Scientific). For staining, sections were washed in PBS and permeabilized 20 min in presence of PBS/0.5% Triton X-100 at RT. Subsequently, sections were washed three times in PBS and blocked with blocking buffer (PBS/10% goat serum/1% BSA) for 1-2 h at RT. Primary antibodies were diluted in Ab incubation buffer (PBS/3% goat serum, 1% BSA/0.05% Triton X-100) and incubated over night in a wet chamber at 4 • C before washing the sections three times in PBS, followed by incubation with the secondary antibodies in Ab incubation buffer (1-2 h at RT). After three washes, sections were mounted using Aqua Poly/Mount (18606-20, Polysciences, Inc.) and observed using confocal laser scanning microscopy as above.
Images were processed and analyzed using Fiji (Schindelin et al., 2012). Shank2 clusters were manually counted in a ∼432 um 2 square area within three sections of the cerebellar molecular layer (identified via calbindin staining) for each animal. NLGN2 clusters were counted in a ∼6000 um 2 square area spanning the whole molecular layer in five sections per animal using the "Analyze Particles" function.
For Nissl staining, sections were delipidized in ethanol with increasing concentration (70, 95, and 100%) followed by submersion in xylen (30 s per step). Rehydration was performed following the opposite order ending in H 2 O for 4 min. After 3-4 min incubation in cresyl violet stain (0.1% cresyl violet acetate, 2.5% glacial acetic acid) sections were rinsed with dH 2 O for 1 min and 6-7 times in 70% ethanol. Subsequently, a differentiation step was carried out by alternating between 95% ethanol and 95% ethanol with 10% acetic acid. Sections were then dehydrated with consecutive 30 s washes in 95% ethanol, 100% ethanol and xylen before covered with entellan. Images were taken with a stereomicroscope (Stemi 2000C, Zeiss, GER).
For ultrastructural electron microscopy (Figure 8), 40 week old anesthetized mice that previously underwent behavioral testing (see below) were transcardially perfused with 4% PFA (postfixed in a mixture of 4%PFA and 1% glutaraldehyde) in 0.1 M PB at pH 7.4. Similarly, for immuno-electron microscopy ( Figure 5C), 38 week old mice were transcardially perfused with a mixture of 4% PFA and 0.1% glutaraldehyde in 0.1 M PB at pH 7.4. Brains were removed and 100 µm thick sagittal sections were cut with a Vibratom (Leica VT 1000S). Thereafter, pre-embedding immuno-electron microscopy was performed as follows: Sections were cryoprotected in 2.3M sucrose and subjected to two cycles of freeze-thaw in liquid nitrogen to aid penetration of immunoreagents into tissue. After rinsing in PBS, sections were incubated with PBS containing 10% HS and 0.2% BSA for 15 min, before being incubated with primary antibody in PBS containing 1% HS and 0.2% BSA overnight. Cells were washed with PBS, then incubated with biotinylated secondary antibody in PBS containing 1% HS and 0.2% BSA for 90 min. After rinsing, they were incubated with ABC (Vector Labs) diluted 1:100 in PBS for 90 min. Sections were washed in PBS and reacted in diaminobenzidine (DAB)-H 2 O 2 solution (Sigma, St. Louis, United States) for 10 min. Thereafter sections were rinsed three times in 0.1 M sodium cacodylate buffer (pH 7.2-7.4) (Sigma-Aldrich, Buchs, Switzerland) and incubated with 1% osmium tetroxide (Science Services, Munich, Germany) in cacodylate buffer for 20 min on ice. The osmication of sections was followed by dehydration through ascending ethyl alcohol concentration steps and rinsed twice in propylene oxide (Sigma-Aldrich, Buchs, Switzerland). Infiltration of the embedding medium was performed by immersing the tissue first in a mixture of 2:1 of propylene oxide and Epon (Carl Roth, Karlsruhe, Germany), then in a 1:1 mixture, and finally in neat Epon and hardened at 60 • C for 48 h. Ultrathin sections (60 nm) were collected and analyzed with an EM902 transmission electron microscope (Zeiss, Germany) equipped with a CCD in lens 2K digital camera and running the ImageSP software (Tröndle, Moorenweis, Germany).
Quantification of synaptic parameters from electron microscopy images was performed using Fiji. For counting molecular layer synapse numbers, 7.086 µm × 7.086 µm images were obtained with 12,000× magnification. 25 consecutive images of the middle third area of the molecular layer were analyzed for each animal, with a total of three animals per genotype. Synapses were counted if they showed a uninterupted outline of the pre-and postsynaptic specialization, ER structure and PSD in spine head and an opposing presynaptic terminal with vesicles. For analyzing spine and presynaptic structural properties, images of the cerebellar molecular layer were taken using 20,000× magnification. Synapses were identified using the same criteria as before.

Mouse Behavior Analyses
All behavior experiments were performed with a cohort of males of the B6-Myo16 em3Hhtg /J line backcrossed to C57BL/6J for at least five generations. Heterozygous mice were bred in house to obtain naïve, age-matched Myo16 +/+ (11 individuals) and Myo16 −/− (11 individuals) mice for the experiments. Weight did not differ significantly between groups. Animals were housed in groups of littermates (3-5 individuals per cage) in an acclimatized animal vivarium (21 ± 1 • C, relative humidity 55 ± 5%) under a 12 h:12 h reversed light/dark cycle and were tested during dark hours. The mice had ad libitum access to food and water. At the beginning of the experiments, animals were between 9 and12 weeks old. All experiments were performed blind to genotype.

Open Field
The open field apparatus consisted of four identical square (50×50×50 cm) arenas made of white polyvinyl foam material. Four lamps were installed that provided even lightning (50 lux) in each arena and a video camera that was mounted directly above the apparatus. The videos were transmitted to a computer running Ethovision tracking software (Version XT8.5, Noldus Technology, Netherlands). Up to 4 mice were tested at the same time, counterbalanced across genotypes but blind to the experimenter. The mice were introduced to one corner of the arena and were allowed to explore undisturbed for 30 min. Total distance moved, velocity and mean distance to the wall were analyzed in 5 min consecutive time bins.

Rotarod
The Ugo Basile Model 47600 (Comerio, VA, Italy) accelerating rotarod for mice was used. The testing area was illuminated diffusely with 30 lux white light. A digital camera was located close to the apparatus to capture videos. The test was performed in two steps. For habituation, up to five subjects were tested simultaneously and placed on the rotating drum at a baseline speed of four rounds per minute (rpm) for up to 180 s. Two trials were performed with an inter-trial interval of 50-60 min. On the same day, three trials were performed for each subject with the speed increasing linearly for 4 min from 4 to 40 rpm. Animals were allowed to stay on the rotating drum for up to 600 s with an inter trial interval of 50-60 min. On the next 5 days each subject was tested in three trials per day of accelerating speed as described above, however, the inter-trial interval was shortened to 10-15 min. The latency to fall was measured. Two capture a second time window, the same animals performed the task again 5 weeks later for seven consecutive days.

Marble Burying Test
Two 26×42 cm cages, filled 6 cm high with fine fresh bedding material were used. Cages were closed with a plastic plate leaving space for air circulation. On the bedding, 20 marbles were placed in a regular pattern, covering the whole area. The apparatus was illuminated by diffuse light of 30 lux. A digital camera was located close to the apparatus to capture videos. Each mouse was introduced into the cage and allowed to explore freely for 30 min. Between subjects, the bedding was stirred thoroughly, pressed down to have a plane surface and marbles were again placed on top. The marbles buried were counted manually in 5 min time bins.

Social Interaction Test
The social interaction test was performed in a white, compartmentalized box made out of polyvinyl foam material. The box had the dimensions of 61 × 37 cm and was divided twice, resulting in three linked compartments: left (22 × 37 cm), center (17 × 37 cm), and right (22 × 37 cm), with doors in between. In the outer compartments, a round chamber (12 cm diameter, 13 cm high) with a heavy lid was placed. The apparatus was illuminated with 30 lux and a digital camera was mounted directly above the setup. Videos were transmitted to a computer running Ethovision tracking software (Version XT8.5, Noldus Technology, Netherlands) equipped with three-point (nose, body center, and tail) detection settings. After 5 min of habituation in the center compartment, doors were opened and mice were allowed to move freely between all three compartments. Active exploration was scored when the nose of the test mouse was detected within a distance of 2 cm to the round chambers. A WT mouse unknown to the test mouse (novel mouse) was present in one chamber, whereas the second chamber was empty. After 10 min a second, unfamiliar WT mouse (novel mouse) was placed in the empty chamber and the test mouse was allowed to explore for another 10 min. In the first part of the test, time spent with the novel mouse compared to the empty chamber was analyzed. In the second part, the time spent with unfamiliar mouse compared to familiar mouse was analyzed. Discrimination index was calculated as time spent with "novel mouse" divided by the time spent with "empty chamber" and "novel mouse" multiplied by 100 (part 1), or time spent with "novel mouse" divided by the time spent with "familiar mouse" and "novel mouse" (part 2) multiplied by 100.

Statistics
Analyses were performed using Prism 7.04 (GraphPad Software, Inc.). All data sets were subjected to Shapiro-Wilk normality test. When comparing two groups fulfilling the normality test, p values were obtained using Student's t-test (if variance in data sets was the same according to F test) or using t-test with Welch's correction (if variance in data sets was different according to F test). When comparing two groups not fulfilling the normality test, exact p values were obtained using Mann-Whitney test. For comparison of more than two groups that all fulfill the normality test, the presence of significant differences was first evaluated using ordinary one-way ANOVA, followed by Dunnett's, Tukey's or Sidak's multiple comparisons test against the control values. For comparison of more than two groups that not all fulfill the normality test, the presence of significant differences was first evaluated using ordinary Kruskal-Wallis test, followed by Dunn's multiple comparisons test against the control values. Significance values are indicated in the figures as * p < 0.05; * * p < 0.01; * * * p < 0.001; * * * * p < 0.0001; n.s., not significant. For behavior analyses, data were analyzed using IBM SPSS Statistics (SPSS Inc., Chicago, IL, United States, Version 22). The statistical tests are indicated in the figure legends of each experiment. For statistical significance a type I error rate of p = 0.05 was defined, all tests were performed two tailed. For analyzing data measured in different time bins, repeated measures ANOVA with genotype as between subject factor and time (bins) as within subject factor was carried out, followed by a Bonferroni post hoc test whenever appropriate.

DATA AVAILABILITY
The datasets generated for this study are available on request to the corresponding author.

ETHICS STATEMENT
This study was carried out in accordance with the recommendations of the European Community Council Directive (2010/63/EU) and the procedures used were approved by the City of Hamburg (Behörde für Gesundheit und Verbraucherschutz, Lebensmittelsicherheit und Veterinärwesen).

AUTHOR CONTRIBUTIONS
WW conceptualized the work and wrote the original draft. All authors contributed to the methodology and wrote, reviewed, and edited the manuscript. MR, FL, SF, MS, and WW contributed to the investigation and formal analysis. IH-B and MK contributed to the resources. MK and WW acquired the funding. SF, JS, and WW supervised the work. projects FOR2419 WA3716/1-1 (to WW), FOR2419 KN556/11-1 and FOR2419 KN556/11-2 (to MK).