|Home | About | Journals | Submit | Contact Us | Français|
Extension of the endoplasmic reticulum (ER) into dendritic spines of Purkinje neurons (PNs) is required for cerebellar synaptic plasticity and is disrupted in animals with null mutations in Myo5a, the gene encoding myosin-Va1–3. Notably, the mechanism ensuring the ER's localization to spines has not been unraveled. While it has been proposed that animal class V myosins localize organelles by tethering them to the actin cytoskeleton4–7, we demonstrate here that myosin-Va acts as a point-to-point organelle transporter to pull ER as cargo into PN spines. Specifically, the myosin accumulates at the ER tip as the organelle moves into spines, and the myosin's ability to hydrolyze ATP is required for spine ER targeting. Moreover, myosin-Va is responsible for the vast majority of spine ER insertional events. Finally, attenuation of the myosin's ability to move along actin filaments reduces the maximum velocity of ER movement into spines, providing direct evidence that myosin-Va drives ER motility. Thus, we establish that an actin-based motor moves ER within animal cells, and we uncover the mechanism that mediates ER localization to PN spines, a prerequisite for synaptic plasticity.
Class V myosins are actin-based motors that mediate the proper intracellular localization of diverse organelles, mRNAs and proteins3. There is convincing evidence that myosin-V drives the motility of organelles along actin filaments in Saccharomyces8 and Dictyostelium9. In metazoans, however, a role for class V myosins as point-to-point cargo transporters has been questioned despite the fact that these myosins possess features that are ideal for driving organelle transport3. Rather, it has been suggested that they function primarily by tethering organelles to the actin cytoskeleton4, 6, 7 or to the plasma membrane5 after delivery by microtubule-based transport.
The heavy chain of mouse myosin-Va is encoded by Dilute (Myo5a), one of three class V myosin genes present in mammals3. Myosin-Va is recruited to melanosomes to mediate their accumulation in the actin-rich periphery of melanocytes3, 7. Therefore, dilute mutations cause pigmentation defects. In addition, mice homozygous for dilute-lethal alleles such as dl20J, a functional null allele of Myo5a, display severe ataxia10. This phenotype, together with the high level of Myo5a expression in cerebellar PNs11, suggest a role for myosin-Va in the cerebellum. Consistently, dilute-lethal PNs in situ display a striking organelle localization defect, as ER is missing specifically from their dendritic spines1, 2.
Dendritic spines are small, actin-rich protrusions on neuronal dendrites that serve as sites of excitatory synaptic input and constitute postsynaptic signaling micro-compartments12. The ER, a dynamic organelle consisting of a continuous network of membrane tubules and cisternae13, normally extends into all PN spines12. This spine ER releases Ca2+ locally via the type 1 inositol 1,4,5-trisphosphate receptor (Itpr1) in response to IP3 produced after metabotropic glutamate receptor (mGluR) activation, thereby facilitating mGluR-dependent long-term depression (LTD)1, 14–16, a form of synaptic plasticity thought to underlie cerebellar motor learning14, 17. Importantly, the mGluR-dependent Ca2+ transient evoked in PN spines by parallel fiber (PF) stimulation is attenuated by ~50% in dilute-lethal PNs and LTD at the PF-PN synapse is abolished in this mutant1.
To confirm that the myosin-Va-dependent localization of ER to spines is essential for the local Ca2+ signal elicited by mGluR-activation, we used two-photon laser glutamate uncaging to stimulate individual PN spines in acute cerebellar slices. PNs were loaded with the Ca2+ indicator Fluo-4 and Alexa-594 to visualize cell volume (Fig. 1a, upper panel). Glutamate was uncaged at single spine heads (Fig. 1a, lower panel) in the presence of the AMPA-receptor inhibitor DNQX, producing a long-latency Ca2+ transient (167 ± 12 msec, range 80 to 330 msec) in the spine but not in the dendritic shaft of control PNs (Fig. 1b; dv/dl20J + DNQX; left: traces from a single spine/dendrite pair; middle: average trace; see Methods for explanation of genotypes). This spine Ca2+ transient requires mGluR1 activation (Fig. 1b; dv/dl20J + DNQX/CPCCOEt) and is entirely absent in the spines of dl20J/dl20J PNs (Fig. 1c; dl20J/dl20J + DNQX; see Fig.1d for a comparison of peak Ca2+ magnitudes). Notably, AMPA-receptor-dependent, fast Ca2+ transients in spines following glutamate uncaging are preserved in dl20J/dl20J PNs (Supplementary Information, Fig. S1). These results verify that the mGluR-dependent Ca2+ transient evoked locally within PN spines depends on myosin-Va and the presence of spine ER.
Given the physiological importance of localizing ER to spines, we next asked how myosin-Va is involved in targeting the ER to these protrusions. There are at least three likely mechanisms for how the myosin might function in this process. First, myosin-Va might function in a non-cell autonomous manner, e.g. in the presynaptic neuron, to confer upon PNs the ability to localize ER to spines (Mechanism 1; Fig. 1e). Second, myosin-Va might function within the PN to mediate the tethering of ER in spines after the organelle has been transported into the spine by a myosin-Va-independent mechanism (Mechanism 2; Fig. 1f). Third, myosin-Va might function as a cargo transporter that associates with ER and moves it along actin filaments into spines (Mechanism 3; Fig. 1g).
To investigate the role of myosin-Va in ER localization, we initially analyzed the dynamic behavior of this organelle in PN spines in dissociated cerebellar cultures from wild type (WT) mice (Fig. 2a–d). For this and all subsequent experiments, the cerebellar cultures were transfected by nucleofection with novel plasmids harboring a PN-specific promotor18 that drives expression of inserted cDNAs specifically in the PNs present within these heterogenous cultures (W.W., Seumas McCroskery, and J.A.H., manuscript in preparation; see also Methods). To visualize the PN's ER and cell volume, we used a plasmid encoding both an mRFP-tagged protein that targets to the ER lumen (mRFPER)19 and a cell volume marker (GFP). Consistent with previous observations of fixed cerebellar sections1, 20, confocal microscopy of live PNs at day in vitro (DIV) 15 shows that almost all spines are fully loaded ER (Fig. 2a; Fig. 5a). Notably, the ER is continuously present within these spines during the time of observation (Fig. 2b; Supplementary Movies 1, 2). In contrast, analysis of the steady state presence of spine ER at 10 DIV (when PN dendrites are starting to grow out; Fig. 2c) shows that ~13% of spine-like protrusions are either empty or only partially filled with ER at any given moment (Supplementary Information, Fig. S2a). Strikingly, events where ER translocates into these protrusions by progressively extending towards the protrusion's tip are common (Fig. 2d; Supplementary Information, Fig. S2b, Movies 1, 2). Retractions of spine ER are also observed at this stage (Supplementary Fig. S2c). Therefore, both ER translocation and ER maintenance potentially contribute to the localization of ER to spines.
To discriminate whether the loss of myosin-Va affects the translocation of ER into spines, or simply the maintenance/tethering of ER within these protrusions, we observed the ER in live dl20J/dl20J PNs (Fig. 2e–h). If only the ER's maintenance within spines is myosin-Va-dependent, then the organelle would be expected to move at least transiently into dl20J/dl20J spines in search of myosin-Va-dependent tethering. Consistent with previous analyses of mutant PNs in situ1, 2, the ER is still present within the dendrites of dl20J/dl20J PNs, but it is almost completely missing from their spines at 15 DIV (Fig. 2e; Fig. 5a) and at 10 DIV (Fig. 2g; Supplementary Information, Fig. S2a). Importantly, time-lapse recording of live 10 DIV cultures shows that the frequency of ER movement into empty/partially-filled spines is reduced ~30-fold in dl20J/dl20J PNs relative to control PNs (Fig. 2h; Supplementary Information, Fig. S2b, Movies 3, 4). Therefore, the movement of ER into PN spines depends critically on myosin-Va. This effectively rules out Mechanism 2 (Fig. 1f), which requires that the movement of ER into spines be driven by a myosin-Va-independent mechanism.
While the ER insertional frequency data argues that the vast majority of these events are driven by myosin-Va, rare myosin-Va-independent insertional events do occur (Supplementary Information, Fig. S2b) and could contribute over time to spine ER targeting. Since microtubules mediate ER motility in animal cells13 and transiently grow into dendritic spines of hippocampal neurons21 and PNs (Supplementary Information, Fig. S3a), we asked whether microtubules contribute to the movement of ER into PN spines. Importantly, while low-dose nocodazole blocks microtubule entry into PN spines, it does not reduce the frequency of ER movement into spines of control dv/dl20J PNs (Supplementary Information, Fig. S3a,e). Moreover, the frequency of microtubule growth into spines is not diminished in dl20J/dl20J PNs (Supplementary Information, Fig. S3b), indicating that the dramatic loss of ER insertional movement seen in the absence of myosin-Va is not caused by failure of microtubules to enter spines. Interestingly, the rare ER insertional events seen in dl20J/dl20J PNs (Supplementary Information, Fig. S2b) are always accompanied by microtubule entry and are abolished by low-dose nocodazole treatment (Supplementary Information, Fig. S3c,e). Taken together, these results show that while microtubule-dependent ER insertion into PN spines can occur, its contribution to the transport of ER into spines in WT PNs must be very minor.
Consistent with previous electron microscopy studies2, 20, control experiments (Supplementary Information, Fig. S4, S5, Movie 5) argue that the ER targeting defect in dl20J/dl20J PNs is not due to a general disruption of spine organization or the targeting of postsynaptic density (PSD) proteins, although the morphology of the dendritic arbor is affected somewhat in cultured dl20J/dl20J PNs at 15 DIV and later (Supplementary Information, Fig. S6). Since cultured PNs receive presynaptic input from granule neurons both in vivo14 and in culture22, and PN development is affected by brain-derived neurotrophic factor secreted by granule neurons23, we considered the possibility that myosin-Va triggers the targeting of ER to PN spines by functioning non-cell autonomously (Mechanism 1; Fig. 1e). To determine if the presence of myosin-Va within PNs is sufficient for localizing the ER to PN spines, cerebellar cultures from dl20J/dl20J mice were transfected with PN-specific plasmids expressing a GFP-tagged version of the brain-spliced isoform of myosin-Va and mRFP-ER. Strikingly, this restores the characteristic tubular ER that protrudes from the dendritic shaft into spines in the dl20J/dl20J PNs (Fig. 3a, b, and top panels in d). Moreover, mGFP-myosin-Va localizes to the tip of the spine ER (see the maximum projection image of a Z-stack in Fig. 3a, the corresponding three-dimensional reconstruction in Supplementary Movie 6, and the examples of single confocal plane images in Fig. 3b), and it remains concentrated there over time, even as these ER tubules grow and shrink, as evidenced by kymograph images (Fig. 3c; see also Supplementary Movies 7–9). To quantify the extent to which ER targeting is rescued, and to confirm that the ER is continuously present in rescued spines as in WT spines at 15 DIV, we expressed a volume marker (GFP) in addition to mRFP-ER and mGFP-myosin-Va in dl20J/dl20J PNs. Analysis of these PNs showed that ER targeting is rescued to the level of WT PNs (Fig. 5a). In addition, time-lapse imaging reveals that the ER is continuously maintained in these spines (Supplementary Movie 10, top panels). Since the exogenous, GFP-tagged myosin-Va is expressed specifically within the PNs present in these mixed cultures (Fig. 3d, top panel), myosin-Va must function within PNs to mediate ER targeting. Consistent with this, ER targeting is not rescued in dl20J/dl20J PNs when co-cultured with cerebellar cells from WT mice (Supplementary Information, Fig. S7). Therefore, despite the fact that myosin-Va could have important functions in other cerebellar cell types, these results argue strongly against Mechanism 1.
To determine whether the myosin's motor activity is necessary for ER targeting, we transfected dl20J/dl20J PNs with plasmids encoding myosin-Va with mutations known (G440A and E442A)24 or expected (R219A)25 to severely impair its steady state ATPase activity. None of the mutant versions of myosin-Va was able to rescue ER targeting (Fig. 3d; Supplementary Movie 9). Thus, ER translocation is dependent on the myosin's ability to hydrolyze ATP. Because the G440A mutation traps myosin-Va in a state with relatively high affinity for actin at physiological ATP levels24, these data also suggest that the myosin's ability to simply link its cargo to F-actin is not sufficient for ER targeting.
If myosin-Va pulls ER as cargo into spines, then the myosin is expected to associate with the spine ER. Several observations indicate that this is the case. First, the striking localization of WT mGFP-myosin-Va at the tip of the spine ER (Fig. 3) does not reflect simply the myosin's association with the PSD26, since the myosin and PSD-proteins (PSD-93 and Homer-3a) localize to different albeit closely-opposed sites (Supplementary Information, Fig. S8, Movie 11). An association of myosin-Va with the spine ER is further supported by previous immuno-electron microscopy data showing endogenous myosin-Va on tubulovesicular ER in PN spines20. Finally, using three-color imaging of live dl20J/dl20J PNs expressing mGFP-myosin-Va, mRFP-ER, and the cell volume marker mCerulean, we find that the myosin co-localizes with the leading tip of the ER tubule as the tubule moves into a preexisting spine-like protrusion (Fig. 4; Supplementary Movie 12).
So far, our findings are completely consistent with Mechanism 3 (Fig. 1g). A prediction from this model is that attenuation of the myosin's ability to move along actin should lead to a reduction in the robustness of ER targeting and in the speed of ER tubule movement towards the tip of spines. To perform this direct test of myosin-Va-dependent ER motility, we analyzed WT PNs and dl20J/dl20J PNs expressing WT myosin-Va or mutant versions of the myosin that display both reduced velocities and run lengths in vitro29–31. These mutants were: (i) myosin-Va with a switch 1 mutation that slows the myosin's ATPase activity (myosin-VaS217A)27, and (ii) two versions of myosin-Va with reduced lever arm length (myosin-Va4IQ and myosin-Va2IQ, which contain four or two of the myosin's six IQ motifs, respectively). Reduction of myosin-Va's lever arm length leads to proportional changes in the myosin's step-size in vitro28, 29, and similar lever arm mutations in Myo2p reduce proportionally the velocities of secretory vesicles, a Myo2p cargo in yeast8.
We initially determined if ER localization to spines is restored in dl20J/dl20J PNs by these mutant myosins at 15 DIV (Fig. 5a). Analysis of dl20J/dl20J PNs expressing GFP-tagged versions of these mutant myosins shows that whereas myosin-Va4IQ rescues ER targeting to the same extent as WT myosin-Va, significantly fewer spines contain ER in the case of both myosin-Va2IQ and myosin-VaS217A (Fig. 5a; Supplementary Movie 10). Similarily, the dendrite morphology defect of dl20J/dl20J PNs is reversed by the expression of WT myosin-Va and myosin-Va4IQ, and partially reversed by the expression of myosin-Va2IQ and myosin-VaS217A (Supplementary Information, Fig. S6).
Strikingly, at 10 DIV, the maximum velocity of ER movement into spines is significantly slower in PNs expressing myosin-Va2IQ or myosin-VaS217A compared to WT PNs and dl20J/dl20J PNs rescued with either WT myosin-Va or myosin-Va4IQ (Fig. 5b, Supplementary Information, Fig. S9). Notably, myosin-Va2IQ and myosin-VaS217A, like WT myosin-Va, are present at the tips of spine ER tubules when expressed as GFP fusions in dl20J/dl20J PNs (Supplementary Information, Fig. S10). We conclude, therefore, that myosin-Va drives the movement of the ER towards the spine tip.
We have shown here that myosin-Va is a point-to-point organelle transporter that translocates ER into PN spines, and that it drives the vast majority of ER insertional movements (although microtubules may make a minor contribution). While previous experiments with cell extracts suggested that myosin-V might mediate ER motility30, 31, our results demonstrate for the first time that an actin-based motor in fact moves ER within animal cells. This function, which may be evolutionarily ancient32, 33, complements the recent demonstration that myosin-Vb moves recycling endosomes into the dendritic spines of hippocampal neurons34. Given that microtubules play a crucial role in ER distribution and movement in neuronal dendrites13, 35, our results also establish a clear example of the dual-filament model of organelle transport36. These findings, together with our confirmation at the single spine level that the mGluR-dependent Ca2+ transient is abolished in dilute-lethal PN spines, establish that myosin-Va-driven ER motility is crucial for at least one physiological function of the ER (local Ca2+ release required for PF-PN LTD). Moreover, we suggest that the myosin-Va-mediated translocation of Itpr1-laden ER may be of general importance for Itpr1-mediated calcium signaling in the nervous system (see e.g.15, 37), since the neurological phenotype of dilute-lethal mice10 is strikingly similar to that of Itpr1 mutant mice16 and much more severe than that of mice merely lacking PF-PN LTD17.
Parasagittal slices (250 μm thick) were cut from the cerebellar vermis of dv/dl20J or dl20J/dl20J mice at postnatal day P15 to P17. Dissections were performed in an ice-cold sucrose solution containing the following (in mM): 75 NaCl, 26 NaHCO3, 75 sucrose, 25 glucose, 2.5 KCl, 1.25 NaH2PO4, 7 MgCl2, and 0.5 CaCl2. Slices were incubated for 30 min at 32°C in sucrose solution, transferred to saline solution (in mM: 125 NaCl, 26 NaCO3, 1.25 NaH2PO4, 2.5 KCl, 1 MgCl2, 2 CaCl2, and 25 glucose), and after 30 min allowed to cool to room temperature. All solutions were bubbled with 95% O2/5% CO2. Whole-cell recordings were obtained from PNs using differential interference contrast microscopy. Glass electrodes (4 to 6 MΩ) were filled with (in mM): 130 KMeSO4, 10 HEPES, 5 NaCl, 1 MgCl2, 4 Mg-ATP, 0.4 Na-GTP, 14 Tris-phosphocreatine, pH 7.3, 290 mOsm. Electrodes also contained Alexa-594 (20 μM) to visualize cell morphology, and fluo-4 (150 μM) to measure Ca2+ signals. Voltage clamp recordings were made with a Multiclamp 700B amplifier (Molecular Devices) filtered at 3 kHz. All recordings were made from slices bathed in the saline solution containing MNI-glutamate (3.75 mM), picrotoxin (20 μM; to block GABA-A receptors) and tetrodotoxin (0.5 μM; to block voltage-gated sodium channels). In some experiments, the AMPA receptor antagonist DNQX (10 μM) and/or the mGluR1 antagonist CPCCOEt (100 μM) were added. Experiments were performed at room temperature. PNs were loaded through the recording pipette for 10 to 20 min before imaging. Combined two-photon laser scanning microscopy and two-photon laser uncaging was performed with a custom-built microscope41. The outputs of two Ti:sapphire lasers (Chameleon, Coherent, Santa Clara, CA) were independently modulated with Pockel's cells (Conoptics). The beams were combined using polarization optics and guided to a single pair of scanning galvanometer (Cambridge Technology). The lasers were tuned to 820 nm and 720 nm for imaging and uncaging, respectively. MNI-glutamate was uncaged with three pulses of 1 msec duration delivered at 100 Hz. Line scans were performed at 500 Hz. Fluorescence signals were collected through the objective (60×, 0.9 NA, Olympus) and an oil condenser (1.4 NA, Olympus) and the signals were summed with a current preamplifier (Stanford Instruments) and digitally sampled at 1.25 MHz (PCI-6110, National Instruments). Fluorescence emission was separated with a dichroic mirror (565DCXR, Chroma) and green and red fluorescence were isolated with bandpass filters (525/50 and 617/73 nm, respectively). Green and red fluorescence was detected with H7422-50 and R9110 photomultiplier tubes, respectively (Hamamatsu). Scanning and signal collection were controlled by custom software written in MATLAB. Laser power for uncaging was adjusted to produce ~40% bleaching of Alexa-594 fluorescence in the spine head42 to provide consistent power delivery for uncaging at different spines.
Timed-pregnant C57BL/6J WT mice were purchased from Charles River Laboratories. C57BL/6J mice heterozygous for the dilute-lethal allele dl20J and the dilute-viral allele dv (i.e. dv/dl20J mice7, 10) were mated to obtain dl20J/dl20J, dv/dl20J and dv/dv progeny. In some cases, the dv/dl20J or dv/dv littermate mice or cultures were used as the normal control. This is valid because (i) the dv allele abolishes Myo5a expression in the skin but not in the brain43, (ii) both dv/dl20J and dv/dv mice do not display the neurological phenotype exhibited by dl20J/dl20J mice10, 43, and (iii) there is no significant difference in the fraction of spines that contain ER, or in the frequency of ER movements into spines, when dv/dl20J PNs are compared with WT PNs (Fig. 5a, Supplementary Information, Fig. S2). Approved protocols were followed for all animal procedures.
Dissociated cerebellar cultures were prepared from E17 or E18 mouse embryos as described44, except that the freshly-dissociated cerebellar cells were transfected with novel L7 (Pcp-2)-based expression plasmids using nucleofection (Amaxa/Lonza) (W.W., Seumas McCroskery, and J.A.H., manuscript in preparation). Embryos were obtained from timed-pregnant WT mice or timed-pregnant dv/dl20J mice (yielding dl20J/dl20J, dv/dl20J and dv/dv littermates) and were treated separately during culture preparation, yielding a single culture per embryo. dl20J/dl20J cultures were used for the analyses of the mutant myosin-Va-null phenotype and for the rescue experiments with full-length myosin-Va cDNA. To distinguish between dl20J/dl20J, dv/dl20J and dv/dv cultures, diagnostic PCR was performed on genomic DNA prepared from embryo brain tissue saved during the dissection procedure, using four primers (3'-CACCATCATCTCATTTCCATCCTGTGTCC-5', 3'-CTCAGGAGGATAATAAATGCACGAGACGC-5', 3'-CTCATCTATACATGGTAATAGCAGGTGGC-5', and 3'-CAGTTAGAGAAGGCTAGAAGTAGCAGAGG-5'). This reaction yields a 334 bp fragment for the dl20J allele, and both a 291 bp fragment and a 251 bp fragment in case of either the dv or the WT allele. For all statistical analyses, the mean values ± SEM are indicated, unless indicated otherwise.
The PN-specific expression plasmid pL7, which will be described in detail elsewhere (W.W., Seumas McCroskery, and J.A.H., manuscript in preparation), is a Bluescript SK+-based vector that contains (i) a promotor sequence derived from the L7 (Pcp-2) gene that, within the cerebellum, drives expression specifically in the PNs18, and (ii) a multiple cloning site inserted into the BamHI site of the L7 (Pcp-2) fragment. This plasmid leads to the specific and efficient expression of inserted cDNAs within PNs present in heterogenous, dissociated cerebellar cultures. Construction of plasmids pL7-mGFP, pL7-mCherry, pL7-mCerulean, pL7-Homer-3a-mGFP, pL7-PSD93-mGFP and pβ-Actin-fGFP will be described elsewhere (W.W., Seumas McCroskery, and J.A.H., manuscript in preparation). To obtain a plasmid that drives the expression of a luminal ER marker under control of the L7(Pcp-2) promotor, a cDNA encoding a fusion protein (mRFP-ER) consisting of the prolactin signal sequence, mRFP and the KDEL ER retention signal19 was inserted into plasmid pL7, creating pL7-mRFP-ER. Similarily, pL7-mRFP-ER-IRES-GFP contains the cDNA encoding mRFP-ER, followed by an internal ribosomal entry side and the EGFP cDNA obtained from pIRES2-EGFP (Clontech). Plasmid pL7- ITPKA-9-52-mGFP was created by inserting the NheI-AgeI ITPKA-9-52 fragment from N9-52-GFP45 into the NheI-AgeI sites of pL7-mGFP, generating a ITPKA-9-52-mGFP fusion identical to that in N9-52-GFP. To create a plasmid for the expression of α2-tubulin under L7-promotor control, a fragment containing the human α2-tubulin cDNA was inserted downstream of GFP in pL7-mGFP. Plasmids pL7-mGFP-myosin-Va and pL7-mCherry-Myosin-Va correspond to pL7-mGFP and pL7-mCherry, respectively, which contain (in frame and downstream of mGFP or mCherry) the cDNA encoding the full-length, WT, brain-spliced isoform43 of mouse myosin-Va. Plasmids pL7-mGFP-myosin-Va-R219A, pL7-mGFP-myosin-Va-G440A, pL7-mGFP-myosin-Va-E442A, and pL7-mGFP-myosin-Va-S217A are derivatives of pL7-mGFP-myosin-Va and encode mutant versions of myosin-Va with the indicated single amino acid changes. To create plasmids pL7-mGFP-myosin-Va-4IQ and pL7-mGFP-myosin-Va-2IQ, an EcoRI-SphI DNA fragment that encompasses myosin-Va's IQ domain was swapped with the respective mutant fragments from mouse myosin-Va constructs with 4 or 2 IQ motifs28. All constructs were confirmed by DNA sequencing.
Cerebellar cultures were imaged using a laser scanning confocal microscope (LSM 510, Carl Zeiss, Inc.) equipped with a 100× objective (1.4 NA). During observation, live cultures were kept at 37 °C and supplied with humidified air containing 5% CO2. To determine the relative enrichment of F-actin in spines versus dendritic shafts, single confocal plane images of dendrites from PNs expressing both IPTKA-9-52-mGFP and mCherry were recorded, avoiding signal saturation. The average signal intensity of IPTKA-9-52-mGFP along a line placed (i) along the length of the spine and (ii) within the shaft was determined. To correct for differences in sample thickness in shaft versus spine (the latter often being thinner than the confocal section), the average signal intensity of mCherry was measured from the same locations as the IPTKA-9-52-mGFP signal, and the product of the IPTKA-9-52-mGFP-spine signal and mCherry-shaft signal was divided by the mCherry-spine signal. Division of this corrected ITPKA-9-52-mGFP-spine signal by the ITPKA-9-52-mGFP-shaft signal yielded the fold-enrichment of ITPKA-9-52-mGFP in spines. For nocodazole experiments, 2 μl of 200 μM nocodazole stock dissolved in DMSO (or, for control, 2 μl DMSO) were directly added to the culture medium (2 ml), before cultures were mounted on the microscope, and movies were recorded during a period of 15 – 70 min after nocodazole (or DMSO) addition. For the co-culture experiment (Supplementary Information, Fig. S7), freshly dissociated WT cerebellar cells were nucleofected with pβ-Actin-fGFP and mixed 1:1 with freshly dissociated dl20J/dl20J cerebellar cells prepared at the same time and nucleofected with pL7-mRFP-ER-IRES-EGFP. The cell mix was plated, cultured and imaged as above.
To determine (1) the frequency of ER movements into spines, (2) the frequency of spine ER retractions, and (3) the frequency at which spines are found filled with ER, partially filled with ER, or devoid of ER (Supplementary Information, Fig. S2, S3), dual-color movies depicting ER and cell volume were acquired from PNs expressing the luminal ER marker mRFP-ER and free GFP as a cell volume marker. The movies were recorded at a rate of 1 fps for 90 – 150 seconds. The following parameters were determined for each individual spine in these movies: the total time of presence of the spine throughout the movie; the time it was entirely, partially, or not filled with ER; the number of ER translocations into spines (partially or all the way; > 0.3 um); the number of ER retractions (> 0.3 um). Analysis of these data yielded the relative time spines were entirely, partially, or not filled with ER. Furthermore, the frequency of ER translocations into spines was determined by dividing the number of total ER insertions by the time (minutes) during which spines were found partially or not filled with ER (i.e. not already fully loaded). Similarily, the frequency of spine ER retractions from spines was determined by dividing the number of total ER retractions by the time (minutes) during which spines that were found filled or partially filled with ER (i.e. where ER retraction is possible). The maximum velocity of ER movement directed towards the distal end of spines (“spine tip”) was determined in six steps, as follows. Step 1: acquisition of dual-color movies depicting ER and cell volume. Movies of PNs expressing the luminal ER marker mRFP-ER, the cell volume marker GFP and, if applicable, a GFP-tagged version of myosin-Va, were recorded at a rate of 1 fps for 120 seconds using a laser scanning confocal microscope. Step 2: identification of spines with ER movement towards the spine tip. ER translocation into spines was not seen in every spine imaged and, if ER translocation into spines occurred, it was a relatively brief event that took place over ~5 – 20 seconds (see examples in Fig. 2d and Supplementary Information, Movies 1 and 2). Therefore, we inspected movies by eye for spines into which ER clearly translocated and subjected these particular spines to kymograph analysis. Step 3: kymograph analysis. Using MetaMorph software (MDS Analytical Technologies), kymographs of cell volume and ER were obtained from spines selected in step 2 by placing a line along the protrusion and precisely in the same direction as the translocating ER. Kymograph analysis was terminated if the spine tilted relative to the kymograph line or moved away from it. Examples of kymographs are shown in Supplementary Information, Figure S9. To facilitate the measurement of the ER's instantaneous velocity (i.e. the distance that the ER traveled within the 1 second-long interval between movie frames), the position of the ER's leading edge and the spine tip were manually determined from the kymograph images at each 1 second interval, and this positional data was recorded in spreadsheets. Graphs generated from these positional data (as well as the kymographs themselves) revealed periods of forward ER movement into spines, but also periods of ER retraction and periods where the ER paused (see examples in Supplementary Information, Figure S5). Step 4: identification of periods of persistent ER movement towards the spine tip. We established specific criteria to select forward ER motility periods from the positional data. These criteria served to limit our measurements to those periods where the ER is undergoing a persistent movement towards the spine tip (the movement that we hypothesized is myosin-Va-driven), and to exclude from our analyses (i) periods where the ER was fully inserted into the spine, and (ii) periods where the ER was undergoing persistent movement away from the spine tip (i.e. periods of retraction). These specific criteria were: (i) the spine ER must make a net forward movement of at least 0.46 μm (i.e. 5 pixels; 1 pixel corresponding to 92 nm) towards the spine tip before or upon reaching the spine tip (i.e. getting as close as 2 pixels, or less, to the volume marker's front), and (ii) the ER must move as defined in (i) without retracting in two or more frames that are either consecutive or spaced by frames without detectable ER movement. In addition, periods of ER motility were further defined to start at the time when the ER was at its lowest (least inserted) position that was closest to the end of the forward motility period. Finally, periods of ER motility ended as soon as the ER reached its outmost (most distal) position for the first time. We note that, according to our criteria, instances where the ER (i) retracts for just a single frame, or (ii) pauses for an unlimited number of consecutive frames (as long as these pauses do not occur at the beginning or end of forward ER motility events), are considered a part of forward ER motility events. The bold portions of the lines in the graphic rendering of the kymographs in Supplementary Information Figure S5 show examples of periods of forward ER motility that were identified based on these criteria. The total numbers of such forward ER motility periods identified and used for further analysis were: 18 for WT; 32 for dl20J/dl20J + MVa-WT; 41 for dl20J/dl20J + MVa-4IQ; 46 for dl20J/dl20J + MVa-2IQ; and 19 for dl20J/dl20J + MVa-S217A. Step 5: measurement of the instantaneous velocities of ER movement. The distance that the ER moved within one second (i.e. its instantaneous velocity) during periods of forward motility, as defined in step 4, was calculated from the positional data in the spreadsheets. As expected, we obtained three classes of instantaneous ER velocities that occurred from one frame to the next (i.e. within one second) for each experimental condition: towards the spine tip (+), away from the spine tip (−), and no movement at all (0). The total numbers of instantaneous ER velocities that were measured and used for further analysis were: 122 for WT; 282 for dl20J/dl20J + MVa-WT; 355 for dl20J/dl20J + MVa-4IQ; 466 for dl20J/dl20J + MVa-2IQ; and 193 for dl20J/dl20J + MVa-S217A. Importantly, for each experimental condition, a similar percentage of the measured instantaneous velocities was (+), i.e. directed towards the spine tip (see below). Step 6: calculation of the maximum velocity of ER movement directed towards the spine tip from the (+) instantaneous velocities of ER movement. Because we wanted to test if myosin-Va specifically drives the forward movement of ER into spines, only those instantaneous ER velocities collected from the motility periods that were directed towards the spine tip (i.e. the instantaneous (+) velocities) were considered for calculating the maximum velocity of ER movement towards the spine tip. The total numbers of instantaneous (+) velocities were: 91 for WT (74.6% of the total instantaneous ER velocities); 207 for dl20J/dl20J + MVa-WT (73.4% of the total instantaneous ER velocities); 264 for dl20J/dl20J + MVa-4IQ (74.4% of the total instantaneous ER velocities); 350 for dl20J/dl20J + MVa-2IQ (75.1% of the total instantaneous ER velocities); and 133 for dl20J/dl20J + MVa-S217A (68.9% of the total instantaneous ER velocities). The maximum velocities shown in the graph in Fig. 5b were obtained by averaging the fastest 10% of these instantaneous (+) velocities.
(a) Fast Ca2+ transients in PN spines are blocked by an AMPA-receptor inhibitor. Ca2+ transients in dv/dl20J PN spines (red traces) and adjacent dendritic shafts (blue traces) evoked by two-photon laser uncaging of MNI-glutamate (3.75 mM) at the spine head at the time indicated (arrowhead), in the presence of CPCCOEt (100 μM; upper panels), and in the additional presence of the AMPA-receptor inhibitor DNQX (10 μM; lower panels). Group data for 13 spines (3 PNs) are shown, with the solid traces indicating the mean and shaded areas indicating SEM. (b) Fast Ca2+ transients are larger in dl20J/dl20J PN spines. Shown are the mean Ca2+ traces (± SEM) from group data for 5 spines (2 PNs), obtained in the presence of CPCCOEt (100 μM; upper panel), and in the additional presence of the AMPA-receptor inhibitor DNQX (10 μM; lower panel). The peaks of the fast Ca2+ transient measured in the presence of CPCCOEt are ΔF/F = 39 ± 5% (dv/dl20J) and 109 ± 32% (dl20J/dl20J).
(a) Illustration of ER motility measurements made from a dl20J/dl20J PN transfected with pL7-mGFP-myosin-Va (MVa-WT) and pL7-mRFP-ER-IRES-EGFP to visualize cell volume (Volume, shown in green) and ER (ER, shown in red). On the left side is an image taken from a time series recorded at a rate of 1 fps. The white box indicates a spine-like protrusion devoid of ER that subsequently received ER. In the center panels, kymograph images obtained from the boxed spine and depicting cell volume (Volume) and ER (ER) are shown. Overlay is also shown (Overlay). The right panel shows a graphic rendering of the corresponding kymograph and indicates the position of the ER's leading edge (red) and the distal end of the spine (green). Events where the spine ER moved persistently for 0.46 μM or more towards the tip of the spine-like protrusion were selected for motility measurements as described in Methods and are shown in bold in the graph. (b) As (a), but illustrating ER motility measurements from a dl20J/dl20J PN transfected with pL7-mGFP-myosin-VaS217A (MVa-S217A) and pL7-mRFP-ER-IRES-EGFP to visualize cell volume (Volume, shown in green) and ER (ER, shown in red).
Time-lapse movies of the dendrites of WT PNs transfected with pL7-mRFP-ER-IRES-EGFP to visualize cell volume (Volume) and ER (ER). The PNs were observed at 15 DIV (upper panels) and 10 DIV (lower panels). The red arrows in the 10 DIV panels indicate the position of the ER-tip during translocations of ER into spine-like protrusions. The movies were recorded at 1 frames per second (fps) and are played back at 10× real time. Size bar, 2 μM.
Time-lapse movies of the dendrites of dl20J/dl20J PNs transfected with pL7-mRFP-ER-IRES-EGFP to visualize cell volume (Volume) and ER (ER), and with either pL7-Myosin-Va-WT (MVa-WT), pL7-Myosin-Va-4IQ (MVa-4IQ), pL7-Myosin-Va-2IQ (MVa-2IQ) or pL7-Myosin-Va-S217A (MVa-S217A). The movies were recorded at 1 fps and are played back at 30× real time. Size bar, 2 μM.
Time-lapse movies of the dendrites of WT PNs transfected with pL7-mCherry-Myosin-Va and PSD-93-mGFP to visualize the myosin (MVa) and the PSD (PSD-93), respectively. Two examples are shown. The PNs were observed at 10 DIV. The movies were recorded at 2 fps and are played back at 30× real time. Size bar, 2 μM.
Time-lapse movie of a spine of a dl20J/dl20J PN transfected with pL7-mCerulean, pL7-mRFP-ER and pL7-mGFP-myosin-Va to visualize cell volume (Volume), myosin-Va (MVa) and ER (ER), respectively. The PN was observed at 8 DIV. The movie was recorded at 2 fps and is played back at 20× real time. Size bar, 1 μM.
Zoom-in time-lapse movies of spines of WT PNs transfected with pL7-mRFP-ER-IRES-EGFP to visualize cell volume (Volume) and ER (ER). The PNs were observed at 15 DIV (upper panels) and 10 DIV (lower panels). Note the translocation of ER into the spine-like protrusions in the 10 DIV panels. The movies were recorded at 1 fps and are played back at 10× real time. Size bar, 1 μM.
Time-lapse movies of the dendrites of dl20J/dl20J PNs transfected with pL7-mRFP-ER-IRES-EGFP to visualize cell volume (Volume) and ER (ER). The PNs were observed at 15 DIV (upper panels) and 10 DIV (lower panels). The movies were recorded at 1 fps and are played back at 10× real time. Size bar, 2 μM.
Zoom-in time-lapse movies of spines of dl20J/dl20J PNs transfected with pL7-mRFP-ER-IRES-EGFP to visualize cell volume (Volume) and ER (ER). The PNs were observed at 15 DIV (upper panels) and 10 DIV (lower panels). The movies were recorded at 1 fps and are played back at 10× real time. Size bar, 1 μM.
Time-lapse movies of the dendrites of dl20J/dl20J and control dv/dv PNs transfected with pL7-mCherry and pL7-ITPKA-9-52-mGFP to visualize cell volume (Volume) and F-actin, respectively. The PNs were observed at 15 DIV. The movies were recorded at 1 fps and are played back at 10× real time. Size bar, 2 μM.
Three-dimensional reconstruction of a dendrite of a dl20J/dl20J PN transfected with pL7-mRFP-ER and pL7-mGFP-myosin-Va to visualize the ER (ER) and the myosin (MVa-WT), respectively. The PN was observed at 15 DIV. Size bar, 2 μM.
Images of live dl20J/dl20J PNs transfected with pL7-mRFP-ER (ER) and either pL7-mGFP-myosin-Va-2IQ (MVa-2IQ) (a) or pL7-mGFP-myosin-Va-S217A (MVa-S217A) (b) show that the mutant versions of myosin-Va localize to the tip of spine ER tubules just like WT myosin-Va. Images depict single confocal planes of PN dendrites (three examples for each mutant myosin are shown) and were taken from time series recorded at 10 DIV. Overlay is also shown (Overlay). Size bar, 2 μM.
Time-lapse movies of the dendrites of dl20J/dl20J PNs transfected with pL7-mRFP-ER and pL7-mGFP-myosin-Va to visualize the ER (ER) and the myosin (MVa-WT), respectively. Three examples are shown. The PNs were observed at 10 DIV. The movies were recorded at 2 fps and are played back at 30× real time. Size bar, 2 μM.
Time-lapse movies of the dendrites of dl20J/dl20J PNs transfected with pL7-mRFP-ER and pL7-mGFP-myosin-Va to visualize the ER (ER) and the myosin (MVa-WT), respectively. Three examples are shown. The PNs were observed at 15 DIV. The movies were recorded at 2 fps and are played back at 30× real time. Size bar, 2 μM.
Time-lapse movies of the dendrites of dl20J/dl20J PNs transfected with pL7-mRFP-ER to visualize the ER (ER) and either pL7-mGFP-myosin-Va (MVa-WT), pL7-mGFP-myosin-Va-R219A (MVa-R219A), pL7-mGFP-myosin-Va-G440A (MVa-G440A), or pL7-mGFP-myosin-Va-E442A (MVa-E442A). The PNs were observed at 13 DIV. The movies were recorded at 2 fps and are played back at 30× real time. Size bar, 2 μM.
To characterize spine ER dynamics, we visualized ER and cell volume by expressing mRFP-ER and free GFP, respectively, in WT, dv/dl20J, and dl20J/dl20J PNs, as well as in dl20J/dl20J PNs transfected with GFP-tagged versions of WT myosin-Va (MVa-WT) or myosin-VaS217A (MVa-S217A). Movies recorded from these PNs at 1 fps at 10 DIV were analyzed. The data were obtained from a set of three or four experiments in the case of WT, dv/dl20J, dl20J/dl20J and MVa-WT, and a set of two experiments in the case of MVa-S217A. On average, 94.5 ± 44.7 (SD) spines were observed per experiment, yielding total times of spine observation of 153.0 ± 70.1 (SD) min per experiment. (a) Steady state presence of spine ER. The relative times that spines are filled with ER, partially filled with ER, or devoid of ER were determined from the recorded movies and are plotted in the graphs as mean ± SEM. At 10 DIV, the spines of WT and dv/dl20J PNs are filled with ER most of the time (dark grey bars), although spines with partially inserted or no ER are also encountered (light grey and white bars, respectively), and together account for ~13% (WT) or ~32% (dv/dl20J). In dl20J/dl20J PNs, the time during which ER is present in spines is dramatically decreased compared to WT and dv/dl20J, and the expression of exogenous WT or S217A versions of myosin-Va in dl20J/dl20J PNs rescues this phenotype fully (MVa-WT) or partially (MVa-S217A). These results confirm and extend those for DIV 15 PNs (Fig. 5a) (b) The frequency of ER movements into spines in dl20J/dl20J PNs is 37-fold lower than in WT PNs and 27-fold lower than in dv/dl20J PNs, and is rescued fully or partially by the expression of exogenous WT myosin-Va or myosin-VaS217A, respectively, in dl20J/dl20J PNs. The graph shows the mean frequency (± SEM) of ER movements towards the spine tip when observing spines that are devoid or only partially filled (i.e. not already fully loaded) with ER. Only motility events where the ER translocated for more than 0.3 μM towards the spine tip were counted. During a two-minute recording, ER insertions take place on average in 14.0% ± 1.5 of all WT spines and 20.9% ± 0.3 of all dv/dl20J spines and, in these spines, happen on average 1.26 ± 0.16 times (WT) or 1.35 ± 0.11 times (dv/dl20J) during this 2 min time interval. This indicates that ER insertions are a common event in these two types of control cells, and that they are not restricted to a small subset of spines. (c) The frequency of spine ER retractions is increased in dl20J/dl20J PNs compared to WT and dv/dl20J PNs, and rescued to control levels by the expression of WT or S217A versions of myosin-Va in dl20J/dl20J PNs. The graph shows the mean frequency (± SEM) of spine ER retractions when observing spines that are fully or partially filled with (i.e. not devoid of) ER. Only motility events where the ER retracted for more than 0.3 μM towards the dendritic shaft were counted. Importantly, myosin-VaS217A is able to maintain the ER within spines as efficiently as WT myosin-Va (compare MVa-WT and MVa-S217A) but is impaired in its abilty to transport the ER into spines (Fig. 5b). Given the ER targeting defect in dl20J/dl20J PNs expressing myosin-VaS217A at 15 DIV (Fig. 5a), these results in aggregate argue that the myosin's ability to maintain the ER in spines is not sufficient to drive ER targeting, but that the transport of ER by myosin-Va is crucial. Together, these results argue strongly against Mechanism 2 and for Mechanism 3 (Fig. 1f, g).
(a) Dynamic microtubules visit the spines of WT and dv/dv PNs at 10 DIV. To determine if microtubules can transiently enter PN spines, we simultaneously visualized microtubules and cell volume by expressing GFP-tagged α2-tubulin and free mCherry, respectively, in WT and dv/dv PNs. The upper left panel is taken from a movie recorded at 1 fps for 2 min from a dv/dv PN and depicts part of a dendrite with a spine that was subjected to kymograph analysis (indicated by the white line). Right panels: Kymographs obtained from this spine for cell volume and microtubules, as well as the overlay image, show that microtubule visits to spines are transient, consistent with previous observations in hippocampal neurons21. The graph (lower left) shows the fraction of WT spines that display microtubule entry during 4 min, as determined from movies recorded at 0.5 fps during a 55 min period that started 15 min after the addition of nocodazole (final concentration: 200 nM; NOCO) or vehicle (DMSO) to the culture medium. The data are the mean (± SEM) obtained from a set of three (DMSO) and four experiments (NOCO). In each experiment, a total of 55 to 118 spines obtained from 4 to 7 different cells were analyzed. The number of spines visited by microtubules is similar with and without DMSO treatment (compare with b). * N/D = not detected. (b) The number of spines that display microtubule entry during 2 min is not reduced in dl20J/dl20J PNs. Movies were recorded as above for 2 min at 1 fps. The graph shows the fraction of spines that display microtubule entry during 2 min. The data are the mean (± SEM) obtained from a set of 23 dv/dv PNs and 15 dl20J/dl20J PNs. An average of 8.7 ± 4.1 (SD) spines were analyzed for each PN. (c) The rare ER movements into spines that occur in dl20J/dl20J PNs take place as microtubules are extending into or are present within that spine. To look for events where the growing end of the microtubule is even with or ahead of the tip of the extending ER tubule (but not behind it), which needs to be the case for either of the two possible mechanisms of microtubule-dependent ER insertion (motor-dependent and TAC-dependent) to work, we simultaneously visualized ER and microtubules by expressing mRFP-ER and GFP-tagged α2-tubulin, respectively, in dv/dl20J and dl20J/dl20J PNs. The upper left panel is taken from a movie recorded at 0.5 fps for 4 min from a dv/dl20J PN and depicts part of a dendrite with a spine that was subjected to kymograph analysis (indicated by the white line). Right panels: Shown are kymographs obtained from this spine during 126 sec for ER and microtubules, as well as the overlay image. The table lists the number of spine ER length extensions that took place in the simultaneous presence (+ MT) or absence (− MT) of a spine microtubule in dv/dl20J and dl20J/dl20J PNs (n = indicates the total number of spine ER extensions analyzed; three experiments were analyzed for both, dv/dl20J and dl20J/dl20J PNs). (d) The steady state presence of spine ER is not affected by nocodazole treatment. To determine if the loss of microtubule insertions into spines affects the steady state presence of spine ER, we visualized ER and cell volume by expressing mRFP-ER and free GFP, respectively, in dv/dl20J, and dl20J/dl20J PNs. At 10 DIV, movies were recorded at 1 fps for 2 min during a 55 min period starting 15 min after the addition of nocodazole (final concentration: 200 nM; NOCO) or vehicle (DMSO) to the culture medium and analyzed as in Supplementary Information, Fig. S2a. The graphs depict the fraction of time that spines are found on average to be filled with ER, partially filled with ER, or devoid of ER. The data are shown as mean (± SEM) and were obtained from a set of three experiments (50 – 186 spines observed per experiment, yielding total times of spine observation of 76 – 346 minutes per experiment), except in case of dl20J/dl20J plus DMSO (indicated by **), where data from a single experiment are shown (174 spines observed, yielding a total time of spine observation of 291 minutes). (e) Blocking microtubule visits to spines abolishes the rare ER movements into dl20J/dl20J spines, but does not affect the frequency of ER movement into control dv/dl20J spines. The frequency of ER movements towards the spine tip in spines that are devoid or only partially filled (i.e. not already fully loaded) with ER was determined from the movie data set described in (d). The graph shows the mean frequency (± SEM), except in case of dl20J/dl20J plus DMSO, where data from a single experiment are shown (indicated by **). Only motility events where the ER translocated for more than 0.3 μM towards the spine tip were counted. The frequency of ER movements into spines is not significantly different between nocodazole-treated, DMSO-treated and untreated dv/dl20J samples (p > 0.4 in all cases, Student t-test; compare with Supplementary Information, Fig. S2b, for the untreated sample). No ER movement into spines was observed in 3 experiments when dl20J/dl20J PNs were treated with 200 nM nocodazole (323 total spines observed, 601 min total time of spine observation). In contrast, the frequency of ER movement into spines of dl20J/dl20J PNs treated with DMSO (0.028/min) was close to the mean of untreated dl20J/dl20J PNs (0.022/min; compare with Supplementary Information, Fig. S2b) in a single experiment where 174 spines were observed for 291 min in total (8 ER insertions were observed). * N/D = not detected (f) Consistent with spine microtubules not playing any major role in spine ER dynamics, blocking microtubule visits to spines does not affect the frequency of ER retractions from control dv/dl20J spines. The frequency of spine ER retractions from spines that are fully or partially filled with (i.e. not devoid of) ER was determined from the movie data set described in (d). The graph shows the mean frequency (± SEM), except in case of dl20J/dl20J plus DMSO, where data from a single experiment are shown (indicated by **). Only motility events where the ER retracted for more than 0.3 μM towards the dendritic shaft were counted. The frequency of ER retractions from spines is not significantly different between nocodazole-treated, DMSO-treated and untreated dv/dl20J samples (p > 0.3 in all cases, Student t-test compare with Supplementary Information, Fig. S2c, for the untreated sample). * N/D = not detected.
(a–f) Shown are confocal images of live PNs from dv/dl20J or dv/dv control mice (a, c, e) and dl20J/dl20J mice (b, d, f) transfected with pL7-mCherry and pL7-ITPKA-9-52-mGFP to visualize cell volume (Volume) and F-actin45, respectively, at 8 DIV (a, b), 10 DIV (c, d) and 15 DIV (e, f). The upper panels show images reconstructed from confocal Z-stacks of PNs. Size bar, 20 μM. The lower panels show a single confocal plane of a PN dendrite (see also Supplementary Movie 5). Size bar, 2 μM. (i) The relative enrichment of F-actin in spines compared to dendritic shafts was measured as described in Methods. The graph shows the mean fold increase (± SEM) of ITPKA-9-52-mGFP fluorescence intensity in spines relative to dendritic shafts (in each case, n = 15 to 20 spine-dendrite pairs). No significant difference between control and dl20J/dl20J PNs was detected (p > 0.473 in all cases; Student t-test), arguing against any gross defect in F-actin accumulation in spines in the absence of myosin-Va.
Shown are live dv/dl20J control PNs (a, c) and dl20J/dl20J PNs (b, d) at 13 DIV, transfected with pL7-mCherry to visualize cell volume (Volume) and either pL7-PSD-93-mGFP (PSD-93) (a, b) or pL7-Homer-3a-mGFP (Homer-3a) (c, d) to visualize the PSD. The upper panels show images reconstructed from confocal Z-stacks of PNs. Size bar, 20 μM. The lower panels show a single confocal plane of a PN dendrite. Size bar, 2 μM.
Shown are representative images reconstructed from confocal Z-stacks covering the entire depth of live dv/dl20J or dl20J/dl20J PNs transfected with pL7-mRFP-ER-IRES-EGFP to visualize cell volume (shown) and ER (not shown). If indicated, dl20J/dl20J PNs were transfected additionally with pL7-mGFP-myosin-Va (MVa-WT), pL7-mGFP-myosin-Va-4IQ (MVa-4IQ), pL7-mGFP-myosin-Va-2IQ (MVa-2IQ) or pL7-mGFP-myosin-Va-S217A (MVa-S217A). Images were recorded at 8 DIV, 10 DIV, 15 DIV and 22 DIV. Size bar, 20 μM. The images indicate qualitatively that the development of the dendritic arbor is somewhat defective in cultured dl20J/dl20J PNs at 15 DIV and later, and that this defect is rescued by the expression of WT myosin-Va and myosin-Va4IQ, and partially rescued by the expression of myosin-Va2IQ and myosin-VaS217A.
To determine whether the ER targeting defect observed in dl20J/dl20J PNs can be rescued by the presence of myosin-Va in other cerebellar cell types present in these heterogenous cultures (i.e. whether or not myosin-Va functions cell autonomously in PNs to mediate ER targeting), freshly dissociated WT cerebellar cells transfected with a β-Actin promotor plasmid expressing farnesylated GFP non-specifically in neurons were mixed 1:1 with freshly dissociated dl20J/dl20J or control dv/dl20J cerebellar cells transfected with an L7-based plasmid expressing both mRFP-ER and free GFP and imaged at 15 DIV. (a) Left panels and inserts: WT cells expressing farnesylated GFP (open arrow heads) and dl20J/dl20J PNs (upper panels) or dv/dl20J PNs (lower panels) expressing ER and volume markers (closed arrow heads) were both present in the cultures. Right panels and inserts: Analysis of ER morphology shows that, in these mixed cultures, spine ER is not present in the dl20J/dl20J PNs, but targeted normally in control dv/dl20J PNs. Size bar, 20 μM (insert: 2 μM). (b) Quantitative analyses show that ER targeting to the spines of dl20J/dl20J PNs is not rescued by the presence of co-cultured, WT cerebellar cells. The graph indicates the fraction of spines (mean ± SEM) that contain ER, as determined from images of PN dendrites recorded at 15 DIV. The numbers of analyzed PNs were 13 (dl20J/dl20J; 19.5 spines/PN analyzed on average) and 9 (dv/dl20J; 24.1 spines/PN analyzed on average). No significant difference was detected between dl20J/dl20J PNs cultured by themselves (Fig. 5a) and dl20J/dl20J PNs co-cultured with WT cells (p = 0.672; Student t-test). 31.
Shown are live WT PNs transfected with pL7-mRFP-ER and pL7-Homer-3a-mGFP (a) or with pL7-mRFP-ER and pL7-PSD-93-mGFP (b) to visualize the ER (ER) and the PSD (Homer-3a or PSD-93). The images show that, unlike myosin-Va, PSD-93 and Homer-3a frequently do not localize at the ER's outermost tip but rather along the spine ER tubule. This observation is consistent with the position of the PSD in PN spines in situ as imaged by electron microscopy46. Images depict a single confocal plane and were recorded at 10 and 15 DIV. Superimposition of the images is also shown (Overlay). Size bar, 2 μM. (c) Images of live WT PNs expressing mCherry-Myosin-Va (MVa) and PSD-93-mGFP (PSD-93) at 10 DIV show that myosin-Va and PSD-93 localize to different, albeit closely opposed, sites within spines. Two examples are shown. Superimposition of the images is also shown (Overlay). The images correspond to single confocal planes and were taken from a time series (see Supplementary Movie 11). Size bar, 2 μM.
We thank Roland Bock and David J. Linden (Johns Hopkins University, Baltimore) for teaching us cerebellar culture preparation, John Oberdick (The Ohio State University, Columbus), Erik L. Snapp and Jennifer Lippincott-Schwartz (NICHD, NIH, Bethesda), Howard D. White (Eastern Virginia Medical School, Norfolk), James R. Sellers, Jose A. Martina and Seumas McCroskery (NHLBI, NIH, Bethesda), Michael J. Schell (Uniformed Services University, Bethesda) and David S. Bredt (University of California, San Francisco) for DNA constructs, Richard E. Cheney (University of North Carolina, Chapel Hill) for advice, and Xufeng Wu (NHLBI, NIH, Bethesda) for microscopy support.
Author Contributions W.W. and J.A.H. designed the project. W.W. carried out the experiments except the Ca2+ imaging, which was carried out by S.D.B.. J.A.H. contributed new reagents and W.W., S.D.B. and J.A.H. analyzed the data and wrote the manuscript.