Myosin Vb Mobilizes Recycling Endosomes and AMPA Receptors for Postsynaptic Plasticity

Summary

Learning-related plasticity at excitatory synapses in the mammalian brain requires the trafficking of AMPA receptors and the growth of dendritic spines. However, the mechanisms that couple plasticity stimuli to the trafficking of postsynaptic cargo are poorly understood. Here we demonstrate that myosin Vb (MyoVb), a Ca²⁺-sensitive motor, conducts spine trafficking during long-term potentiation (LTP) of synaptic strength. Upon activation of NMDA receptors and corresponding Ca²⁺ influx, MyoVb associates with recycling endosomes (REs), triggering rapid spine recruitment of endosomes and local exocytosis in spines. Disruption of MyoVb or its interaction with the RE adaptor Rab11-FIP2 abolishes LTP-induced exocytosis from REs and prevents both AMPA receptor insertion and spine growth. Furthermore, induction of tight binding of MyoVb to actin using an acute chemical genetic strategy eradicates LTP in hippocampal slices. Thus, Ca²⁺-activated MyoVb captures and mobilizes of REs for AMPA receptor insertion and spine growth, providing a mechanistic link between the induction and expression of postsynaptic plasticity.

Introduction

Most models of learning and memory invoke modification of synaptic strength as the underlying mechanism for information storage in the brain. One compelling and intensely studied example is long-term potentiation (LTP) at CA1 synapses in the hippocampus. LTP-inducing stimuli activate synaptic NMDA receptors, leading to increased synaptic AMPA receptors (Kennedy and Ehlers, 2006; Derkach et al., 2007; Shepherd and Huganir, 2007; Newpher and Ehlers, 2008) and rapid alteration of dendritic spine morphology (Engert and Bonhoeffer, 1999; Maletic-Savatic et al., 1999; Alvarez and Sabatini, 2007). The coupling of AMPA receptor insertion and membrane addition to spines suggests the presence of a local intracellular source for these resources. Recent studies demonstrated that recycling endosomes (REs) located within or at the base of spines contain a pool of AMPA receptors to support local receptor cycling (Ehlers, 2000; Cooney et al., 2002; Park et al., 2004). During LTP, REs are rapidly mobilized into spines where their contents are delivered to the plasma membrane (Park et al., 2006). Disrupting RE trafficking blocks not only activity-dependent AMPA receptor insertion but also spine growth and new spine formation (Park et al., 2004; Park et al., 2006), indicating that dendritic REs are the local reservoir of cargo needed for plasticity-induced spine modification.

Although RE trafficking is required for functional and structural changes at synapses during LTP, it is not understood how activation of synaptic NMDA receptors triggers the rapid spine mobilization of REs during LTP. The abundance of actin and the exclusion of microtubules from spines suggest that the trafficking of REs during LTP may involve actin-based myosin motors. The unconventional class V myosins have been widely implicated in vesicle and organelle trafficking (Desnos et al., 2007). Class V myosins consist of two heavy chains, each consisting of an N-terminal motor domain, a neck domain that binds calmodulin and other light chains, a coiled-coil region that mediates dimerization, and a C-terminal globular tail domain (GTD) that associates with cargo (Figure 1A). Of the three class V myosins (Va, Vb, and Vc), myosin Vb (MyoVb) has been shown to associate with REs and regulate the trafficking of a variety of receptors from REs to the plasma membrane in both neuronal and nonneuronal cells (Lapierre et al., 2001; Hales et al., 2002; Volpicelli et al., 2002; Fan et al., 2004; Lise et al., 2006; Nedvetsky et al., 2007; Swiatecka-Urban et al., 2007). The ability of MyoVb to regulate recycling endosome trafficking relies on the interaction of its GTD with the RE-resident GTPase Rab11 and its effector Rab11-family interacting protein 2 (Rab11-FIP2) (Lapierre et al., 2001; Hales et al., 2002). MyoVb is enriched in hippocampus (Zhao et al., 1996), pointing to the possibility that MyoVb could mediate endosomal trafficking for LTP. Experiments utilizing acute inhibition of MyoVb have found that MyoVb tethers endosomes at the cell periphery for local recycling (Provance et al., 2004; Provance et al., 2008).

Figure 1.

MyoVb is Concentrated in Dendritic Spines and Traffics with REs

(A) Schematic diagram of MyoVb and its association with REs via Rab11/Rab11-FIP2.

(B) Cultured hippocampal neurons (DIV19) were fixed and immunolabeled for endogenous MyoVb and PSD-95 (left). Recycling endosomes (REs) were labeled by Alexa 647-transferrin uptake (Alexa-Tf, middle) or TfR-mCherry expression (TfR-mCh, right), followed by MyoVb staining. Colocalization is indicated by yellow arrows. Red arrows indicate REs in dendritic shafts that are not associated with MyoVb. Dashed white lines indicate the dendritic outline determined by a GFP cell fill. Scale bar, 2 μm.

(C-D) Hippocampal neurons expressing TfR-mCh and GFP-MyoVb FL (C) or GFP-MyoVb ΔC (D) and were imaged over time. Red arrows in (C) indicate the coordinated movement of GFP-MyoVb FL and TfR-mCh into and out of the spine head in two examples (C1, C2). Green arrows in (D) indicate the lack of correlated movement between GFP-MyoVb ΔC and TfR-mCh. Time is indicated in min:sec. Scale bars, 2 μm. See Movies S1-S3.

(E) Means ± SEM of the correlation coefficient between GFP-MyoVb FL or GFP-MyoVb-ΔC and TfR-mCh over time. n = 31 spines from 3 neurons for each; p<0.001 for all time points.

Recent biochemical and structural studies indicate that the molecular structure of mammalian MyoVa, the prototype class V myosin, is dynamically regulated by micromolar levels of Ca²⁺. At resting cellular Ca²⁺ levels, MyoVa exists in an inactive folded conformation characterized by a low actin-activated ATPase activity and inhibitory interactions between the N-terminal motor head and the C-terminal GTD. Micromolar Ca²⁺ concentrations lead to the unfolding of MyoVa, a conformational switch that exposes the GTD (Krementsov et al., 2004; Li et al., 2004; Wang et al., 2004; Li et al., 2006; Liu et al., 2006; Thirumurugan et al., 2006). Extrapolated to MyoVb, such a conformation change would be predicted to expose the Rab11/Rab11-FIP2 binding domain enabling association with REs and thereby engaging endosomal transport. Given the well-established requirement for elevated spine Ca²⁺ in LTP (Lynch et al., 1983; Malenka et al., 1988), such studies point to the possibility that MyoVb acts as a Ca²⁺-sensitive trigger and actin motor for RE transport during LTP.

In the present study, we demonstrate that MyoVb mediates Ca²⁺-dependent accumulation of REs and AMPA receptors in spines during synaptic potentiation. MyoVb is highly enriched in spines under basal conditions, and is rapidly recruited to dendritic endosomes upon NMDA receptor activation via a Ca²⁺-dependent conformational switch that enables binding of MyoVb to the Rab11-FIP2 adaptor complex on REs. Disrupting or augmenting this interaction alternately impairs or enhances RE trafficking into spines. Furthermore, loss-of-function and dominant negative experiments show that MyoVb is required for LTP-induced exocytosis from REs, AMPA receptor delivery, and new spine formation. Further, rapidly locking MyoVb on actin in single postsynaptic neurons by a chemical genetic inhibition strategy acutely abrogates LTP in hippocampal slices. Taken together, these results define MyoVb as a Ca²⁺ sensor for postsynaptic membrane trafficking during LTP and demonstrate a mechanistic link between synapse-specific signaling and actin-based transport that couples functional and structural plasticity of glutamatergic synapses. Moreover, these studies identify a novel physiological function for Ca²⁺-dependent activation of class V myosin in cells.

Results

MyoVb is Enriched in Dendritic Spines and Traffics with Spine Endosomes

Endogenous MyoVb in cultured hippocampal neurons was detected immunocytochemically using rabbit polyclonal antibodies generated against rat MyoVb (Figure S1). Synapses were labeled by PSD-95 immunostaining and REs were labeled by internalized Alexa-647 conjugated transferrin (Alexa-Tf) or mCherry-tagged transferrin receptor (TfR-mCh). In 19 day in vitro (DIV) hippocampal neurons, MyoVb was highly enriched in dendritic spines and localized near glutamatergic synapses labeled by PSD-95 (Figure 1B). Consistent with previous findings (Park et al., 2006), REs were often positioned at the base of spines and occasionally within the heads of spines (Figure 1B). REs in spines colocalized with MyoVb, while those in dendritic shafts were rarely decorated with bright MyoVb puncta (Figure 1B).

To gain more insight into the dynamics of MyoVb, the mobility of GFP-tagged MyoVb (GFP-MyoVb) and REs labeled with TfR-mCh were monitored in real time in individual dendritic spines. Consistent with the observed staining pattern for endogenous MyoVb, TfR-mCh labeled REs in spines colocalized with GFP-MyoVb clusters and both showed correlated movement into and out of spines (Figures 1C and 1E; Movies S1 and S2). This co-trafficking was selective for MyoVb, as REs did not show correlated movement with MyoVa (Figure S2). In contrast to full length MyoVb, a truncated MyoVb lacking the cargo-binding GTD (GFP-MyoVb-ΔC) was highly localized to spines and exhibited local movement that was not correlated with REs (Figures 1D and 1E; Movie S3), suggesting that the GTD of MyoVb is important for RE trafficking in spines.

Activation of Synaptic NMDA Receptors Recruits MyoVb to REs

REs move into spines in response to LTP-inducing stimuli (Park et al., 2006). We tested whether activation of synaptic NMDA receptors by glycine stimulation - a protocol used to induce synaptic potentiation in cultured hippocampal neurons (Lu et al., 2001) - alters MyoVb trafficking. Under basal conditions, MyoVb concentrated in spines and rarely colocalized with REs in dendritic shafts. Upon brief glycine application (200 μM, 5 min), MyoVb showed increased overlap with REs at the base of spines (Figure 2A). Quantitative analysis revealed a significant increase in the amount of MyoVb associated with REs (normalized total MyoVb associated with TfR-GFP: unstimulated, 1.00 ± 0.05, n = 25 cells; glycine, 1.61 ± 0.22, n = 21 cells; p<0.01, Student’s t-test; Figure 2B), while the total expression level of myosin Vb was unaltered (data not shown). Moreover, a similar stimulus-dependent increase in colocalization was found between MyoVb and REs labeled by endogenous TfR and internalized HA-GluR1 (Figure S3).

Figure 2.

MyoVb is Recruited to REs that Move Into Spines Upon LTP Stimuli

(A) Hippocampal neurons expressing TfR-GFP (red) to label REs and mCherry as a cell fill were stained for MyoVb (green) either before (Unstim) or after glycine stimulation. White arrows indicate REs co-labeled with MyoVb. Scale bar, 2 μm.

(B) Means ± SEM of the fraction of MyoVb associated with REs before or after a glycine stimulus. * p<0.01 relative to control, t-test.

(C) Hippocampal neurons expressing GFP-MyoVb and TfR-mCh were imaged before and after glycine. Times are indicated in min:sec. Red arrows indicate the movement of GFP-MyoVb and REs into spines in two examples (C1, C2). Scale bars, 2 μm. See Movies S5-S6.

(D) Quantitative analysis of normalized GFP-MyoVb and TfR-mCh intensity in the spines shown in (C) over time. Black lines indicate periods of glycine application and double arrow-headed lines correspond to the time lapse after glycine shown in (C).

(E) Cumulative probability plot of the average intensity of GFP-MyoVb on individual REs before and 5 min after glycine.

(F) Quantitative analysis of GFP-MyoVb and RE movement into spines. The intensity at each data point is the average intensity across a population of spines from the corresponding time periods normalized by the averaged intensity 0-5 min before glycine (Gly) stimulation. *p<0.01 and #p<0.01 relative to GFP-MyoVb and TfR-mCh before glycine respectively.

(G) GFP-MyoVb spine translocation events (events/min/100 μm of dendrite) were measured after glycine (Gly) with or without APV, and normalized to basal conditions. Data represent means ± SEM. **p < 0.001 relative to Gly.

(H) Cumulative probability plot of the average intensity of GFP-MyoVb on individual REs before (Basal) and 5 min after stimulation (5 μM glutamate, 200 μM glycine, 3 min) in the presence (H1) or absence (H2, 0 Ca²⁺) of 2 mM extracellular Ca²⁺.

To investigate the dynamics of NMDA receptor-triggered recruitment of MyoVb to REs, we performed live cell imaging experiments on hippocampal neurons expressing TfR-mCh before, during, and after glycine-induced chemical LTP. Consistent with our immunostaining results (Figure 2A), REs at the base of spines showed little colocalization with GFP-MyoVb puncta under resting conditions (Figure 2C). However, after glycine stimulation, GFP-MyoVb rapidly localized to a population of TfR-positive REs at the base of spines, and subsequently co-trafficked with dendritic REs into the spine head (Figures 2C and 2D; Movies S4 and S5). Quantitative analysis revealed a significant increase in the fractional amount of GFP-MyoVb associated with dendritic REs following glycine stimulation (Figure 2E). Furthermore, the stimulus-induced translocation of REs into spines was accompanied by a parallel increase in the average intensity of GFP-MyoVb in spines (Figure 2F). Blocking NMDA receptors with the antagonist APV (100 μM) prevented the translocation of MyoVb into spines following gly-LTP (Figure 2G). Removing extracellular Ca²⁺ prevented the stimulus-induced recruitment of GFP-MyoVb to REs (Figure 2H), consistent with a requirement for Ca²⁺ entry through activated NMDA receptors. Together, these data show that upon an LTP stimulus, MyoVb is recruited to REs that subsequently move into dendritic spines.

Ca²⁺-Dependent Unfolding and Binding of MyoVb to the RE Adaptor Rab11-FIP2

Due to their high Ca²⁺ permeability, activation of NMDA receptor channels elevates Ca²⁺ in spines to micromolar levels or higher (Yang et al., 1999; Sabatini et al., 2002), which in turn triggers downstream effectors for LTP expression (Malenka and Nicoll, 1999). Interestingly, the actin-activated ATPase activity and conformation of MyoVa, the prototypical class V myosin, are regulated by micromolar Ca²⁺ (Nascimento et al., 1996; Li et al., 2004; Wang et al., 2004; Li et al., 2006). Given the structural similarity between MyoVb and MyoVa, we hypothesized that the conformation of MyoVb may also be regulated by Ca²⁺, allowing it to act as a Ca²⁺ sensor for stimulus-dependent association and spine transport of REs (Figure 3C).

Figure 3.

Binding of MyoVb to Rab11-FIP2 is Induced by a Ca²⁺-Dependent Conformational Switch

(A) Ca²⁺ stimulates the actin-activated ATPase activity of purified MyoVb. Data represent means ± SD; n = 4. The calculated V_max and K_actin are 1.17 ± 0.21 s^-1 motor^-1 and 12.58 ± 6.81 μM in the absence of Ca²⁺, and 12.07 ± 1.21 s^-1motor^-1 and 7.11 ± 1.78 μM in the presence of 100 μM Ca²⁺.

(B) Ultracentrifugation sedimentation analysis of MyoVb under Ca²⁺-free or 100 μM Ca²⁺conditions. The sedimentation coefficient of MyoVb decreased from 14.34 S in the absence of Ca²⁺ to 10.82 S in the presence of 100 μM Ca²⁺.

(C) Model diagram indicating the proposed Ca²⁺-regulated conformational switch and cargo binding by MyoVb. Top panel, wildtype MyoVb is folded into an inactive structure at low Ca²⁺ levels. High Ca²⁺ switches MyoVb to an extended structure that binds to Rab11/Rab11-FIP2, resulting in membrane recruitment. Middle panel, a mutant of MyoVb lacking 15 amino acids required for Rab11-FIP2 binding (MyoVb-ΔRBD) is incapable of binding to REs at low or high Ca²⁺ levels. Bottom panel, deleting part of the coiled-coil region (MyoVb-CCtr) generates a constitutively unfolded MyoVb mutant that binds REs independent of Ca²⁺.

(D) Schematic diagram of MyoVb illustrating its domain architecture. Brackets indicate MyoVb domains that interact with other proteins. CaM, calmodulin. Numbers indicate amino acids.

(E) Brain lysates were subjected to immunoprecipitation (IP) with rabbit Rab11 or MyoVb antibodies and immunoblot (IB) analysis with the indicated antibodies. Pre-immune rabbit serum was used as a negative control.

(F) Co-IP was performed on lysates of 293T cells expressing V5-tagged MyoVb and GFP-tagged Rab11-FIP2 in different concentrations of free Ca²⁺. ms, control mouse serum.

(G) Data represent means ± SEM of the normalized fraction of MyoVb bound to GFP-Rab11-FIP2 at various levels of free Ca²⁺. *p<0.05 relative to 0 mM Ca²⁺.

(H) Co-IP was performed on lysates from 293T cells expressing the indicated constructs at 0 or 1 mM free Ca²⁺. A representative blot from three independent experiments is shown.

To test this hypothesis, we first measured the actin-activated ATPase activity of purified recombinant full length MyoVb under Ca²⁺-free or high (100 μM) Ca²⁺ conditions. We found that high Ca²⁺ enhanced the ATPase activity of MyoVb nearly 10 fold (Figure 3A), similar to the Ca²⁺-dependent regulation of MyoVa. Second, we monitored the conformation of full length MyoVb using analytical ultracentrifugation and sedimentation analysis. In the presence of Ca²⁺, MyoVb transitioned from a species with high sedimentation velocity (14.3S) to a species with lower sedimentation velocity (10.8S) (Figure 3B), consistent with a Ca²⁺-dependent switch from a compact to an elongated conformation.

Next, we examined the binding of MyoVb to Rab11-FIP2. Co-immunoprecipitation (co-IP) revealed that MyoVb forms a complex with Rab11 and Rab11-FIP2 in brain (Figure 3E). Similarly, when expressed in 293T cells, MyoVb co-immunoprecipitated with Rab11-FIP2 (Figure 3F). Notably, the binding between MyoVb and Rab11-FIP2 was enhanced with increasing concentrations of Ca²⁺ (Figures 3F and 3G; % of binding: 19.7 ± 5.4, 44.7 ± 17.1, and 86 ± 25 in 0.01, 0.1 and 1.0 mM free Ca²⁺ respectively; n = 6; p<0.01, t-test). We then sought to determine whether the Ca²⁺-dependent binding of MyoVb to Rab11-FIP2 was due to a conformational switch. We generated a mutant MyoVb with a shortened coiled-coil domain to prevent folding and lock the motor into an extended conformation (MyoVb-CCtr, Δ949-1042; Figures 3C and 3D) (Li et al., 2006), and tested its binding to Rab11-FIP2. Co-IP analysis showed that the MyoVb-CCtr mutant strongly bound to Rab11-FIP2 even under nominally Ca²⁺-free conditions (Figure 3H). As a negative control, a MyoVb mutant defective in Rab11/Rab11-FIP2 binding (MyoVb-ΔRBD, Δ1797-1811; Figures 3C and 3D) bound poorly to Rab11-FIP2 even at 1 mM Ca²⁺ (Figure 3H). These data provide strong evidence that Ca²⁺-dependent unfolding of MyoVb enables its association with cargo.

MyoVb Constitutively Traffics REs in Spines by Interacting with Rab11-FIP2

In hippocampal neurons, Rab11-FIP2 exhibits a punctate distribution in dendrites, axons and neuronal somas (Figure S4). In the soma and dendrites, Rab11-FIP2 localizes to TfR-positive REs (Figure S4). Intriguingly, whereas MyoVb-WT showed only partial overlap with Rab11-FIP2 and TfR-positive REs, the constitutively unfolded MyoVb-CCtr mutant strongly co-localized with Rab11-FIP2 and REs present in both spines and dendritic shafts under basal conditions (Figures 4A-C). In contrast, the Rab11-FIP2 binding-deficient MyoVb-ΔRBD mutant was almost exclusively localized to spines and exhibited much less overlap with Rab11-FIP2 and TfR-positive REs (Figures 4A-C). Quantitative analysis revealed inverse effects of the CCtr and ΔRBD mutations on the co-localization of MyoVb with endogenous Rab11-FIP2 (Figure 4B), and with TfR-positive REs (Figure 4D, normalized MyoVb associated with REs: WT, 1.00 ± 0.05, n = 22 cells; ΔRBD, 0.67 ± 0.04, n = 9 cells; CCtr, 1.63 ± 0.08, n = 17 cells; p<0.001, t-test).

Figure 4.

MyoVb Constitutively Traffics REs in Spines by Interacting with Rab11-FIP2

(A) Hippocampal neurons expressing GFP-MyoVb WT, ΔRBD, or CCtr were stained for endogenous Rab11-FIP2. The dendritic outline was determined by a mCherry cell fill. Colocalization is indicated by white arrows. Scale bar, 2 μm.

(B) Pixel-by-pixel fluorescence intensity correlations between GFP-MyoVb-WT, ΔRBD, or CCtr with endogenous Rab11-FIP2. Correlation coefficients (CC) are indicated. AFU, arbitrary fluorescence units.

(C) Hippocampal neurons expressing TfR-mCh, GFP, and V5-tagged MyoVb WT, ΔRBD or CCtr were stained for V5. The dendritic outline was determined by GFP. Scale bar, 2 μm.

(D) Quantitative analysis of the normalized fraction of MyoVb associated with REs. Data represent means ± SEM. **p <0.001 relative to WT; t-test.

(E) Quantitative analysis of RE mobility in spines over time. Neurons expressing TfR-mCh and GFP-MyoVb WT or mutants were imaged for 5 min at 37°C. The normalized TfR-mCh fluorescence (F/F₀) in 30 dendritic spines for each construct was plotted over time.

(F) Coefficient of variance (CV) of the normalized intensity of TfR-mCh in spines over a 5 min time lapse. Data represent means ± SEM of the CV. *p<0.01 relative to WT; t-test.

We next tested whether the physical association of MyoVb with Rab11-FIP2 is required for ongoing endosome movement in dendritic spines. In cells expressing MyoVb-WT, a plot of TfR-mCh fluorescence intensity over time in spines revealed a mixture of fluctuating and flat lines, representing mobile and stationary REs respectively (Figure 4E). In contrast, in neurons expressing MyoVb-ΔRBD, the fluctuation of TfR-mCh intensity was dampened (Figure 4E), indicating that association with Rab11-FIP2 is required for ongoing RE transport in spines. On the other hand, in neurons expressing MyoVb-CCtr, the fluctuation of TfR-mCh intensity was more pronounced (Figure 4E), indicating that increasing the association of MyoVb with REs increases their shuttling into and out of spines. Introduction of the ΔRBD mutation into MyoVb-CCtr reversed the increased spine transport of REs (Figure 4E). Both ΔRBD and CCtr/ΔRBD mutants impaired the spine motility of REs (Figure 4E), consistent with the ability of MyoVb-ΔRBD to dimerize with MyoVb-WT (data not shown) and act in a dominant negative manner, likely by impairing association with dimeric Rab11 and Rab11-FIP2 (Jagoe et al., 2006) (Figures 4A-B, Figure S5). Quantification of the coefficient of variance (CV) of TfR-mCh intensity confirmed the reduced or augmented spine trafficking of REs in neurons expressing MyoVb-ΔRBD and MyoVb-CCtr, respectively (Figure 4F; TfR-mCh intensity CV in spines: wt, 0.20 ± 0.01, n = 167 spines from 4 cells; ΔRBD, 0.14 ± 0.01, n = 104 spines from 4 cells; CCtr, 0.33 ± 0.05, n = 105 spines from 4 cells; CCtr/ΔRBD, 0.13 ± 0.01, n = 87 spines from 3 cells; p < 0.01 relative to wt).

To further examine the structural requirements for conformation-dependent RE transport by MyoVb, we mutated three conserved residues in the motor domain (D136) and the GTD (K1699, K1772) to alanine residues. These three charged residues are conserved in MyoVa (D136, K1706, K1779) (Figure S6), and ionic interactions between them contribute to the folded structure of MyoVa (Li et al., 2008). Introduction of either D136A or K1699A/K1772A mutations increased the colocalization of MyoVb with REs under basal conditions (data not shown) and both mutants enhanced the mobility of REs in spines to the same extent as MyoVb-CCtr (Figure 4F; TfR-mCh intensity CV in spines: D136A, 0.34 ± 0.04, n = 30 spines from 3 cells; K1699/1772A, 0.35 ± 0.04, n = 40 spines from 4 cells; p < 0.001 relative to wt).

MyoVb Mediates Endosome Translocation into Spines during LTP

Our results above identify a Ca²⁺-dependent conformational change in MyoVb that facilitates its interaction with REs via Rab11-FIP2. To test whether this Ca²⁺-dependent switch mediates LTP-induced mobilization of REs to spines, we performed live cell imaging experiments on hippocampal neurons expressing TfR-mCh and GFP-MyoVb while simultaneously activating synaptic NMDA receptors with glycine. As before (Figure 2C), glycine stimulation induced an abrupt increase in GFP-MyoVb-WT associated with dendritic REs (Figures 5A and 5D, Movie S6) and a simultaneous increase in GFP-MyoVb-WT colocalized with endogenous Rab11-FIP2 in dendrites (Figure S3F-G). Immediately following glycine treatment, MyoVb-laden REs translocated into the spine head, resulting in a 4-fold increase in the intensity of TfR-mCh in spines (Figure 5G; 4.4 ± 1.1 fold increase, n = 22 spines from 3 cells).

Figure 5.

MyoVb Mediates RE Translocation into Spines Following LTP-Inducing Stimuli

(A-C) Time lapse images of TfR-mCh and GFP-MyoVb-WT (A), ΔRBD (B), or CCtr (C) in a dendritic spine. Two time points before and a six minute time lapse after glycine (black arrow) are shown. Red arrows indicate co-transport of GFP-MyoVb WT or CCtr and REs following glycine stimulation. Green arrows indicate the lack of correlated movement between GFP-MyoVb ΔRBD and RE. Time is indicated in min:sec. See Movies S6-S8. Scale bar, 2 μm.

(D-F) Cumulative probability plots of GFP-MyoVb WT (D), ΔRBD (E), and CCtr (F) fluorescence intensity on REs before and 2.5 min after glycine. The ΔRBD mutant blocks, while the CCtr mutant occludes the glycine-induced association of MyoVb with REs.

(G) Data indicate means ± SEM of the fold increase in TfR-mCh fluorescence intensity 10-15 min after glycine in spines initially lacking REs. **p<0.001; t-test.

In contrast to MyoVb-WT, glycine stimulation failed to recruit spine-localized GFP-MyoVb-ΔRBD to REs (Figures 5B and 5E, Movie S7) which did not translocate into spines (Figures 5B and 5G; 1.0 ± 0.2 fold increase, n = 19 spines from 3 cells; p<0.001 relative to WT, t-test). Conversely, GFP-MyoVb-CCtr constitutively localized to REs (Figures 4C, 4D, and 5C), and trafficked into and out of spines both prior to and after glycine stimulation (Figure 5C and 5F, Movie S8). Interestingly, in these neurons, the accumulation of REs in spines was significantly increased following glycine (Figure 5G; 18.9 ± 4.6 fold increase, n = 15 spines from 4 cells; p<0.001 relative to WT, t-test), suggesting that the GTD of MyoVb tethers endosomes within spines following LTP stimuli. Taken together, these results indicate that MyoVb, via its Ca²⁺-regulated interaction with Rab11-FIP2, directs RE translocation into spines during LTP.

MyoVb is Required for LTP-Induced Spine Exocytosis

To test whether MyoVb is required for activity-dependent exocytosis in spines, we used RNA interference (RNAi) for loss of function studies along with expression of RNAi-resistant MyoVb mutants (Figure S7). Hippocampal neurons were infected with lentivirus expressing MyoVb short hairpin (sh) RNA and exocytosis from REs was monitored using the pH-sensitive optical reporter TfR-superecliptic pHluorin (TfR-SEP) (Park et al., 2006). Exocytic events were observed as the sudden appearance of TfR-SEP fluorescence. In neurons expressing an empty vector (Vec) or scrambled shRNA control (Scram), brief glycine application led to extensive exocytosis from REs on neuronal dendrites, most appreciably on dendritic spines (Figures 6A and 6B; Movie S9). This glycine-induced exocytosis was blocked by the NMDA receptor antagonist APV, indicating a requirement for NMDA receptor activation (Figure 6B). In contrast, MyoVb-depleted neurons exhibited little exocytosis from REs following glycine LTP stimulation in either spines or dendritic shafts (Figures 6A and 6B; Movie S10). Importantly, the inhibition of TfR-SEP exocytosis was rescued by expressing an RNAi-resistant MyoVb (Figures 6A and 6B; Movie S11).

Figure 6.

MyoVb is Required for LTP-Induced Exocytosis, AMPA Receptor Insertion, and Spine Growth

(A) Hippocampal neurons expressing MyoVb shRNA were transfected with TfR-SEP either alone (MyoVb RNAi) or together with RNAi-resistant MyoVb-WT (Rescue). Exocytosis from REs was monitored by TfR-SEP fluorescence before and 15 min after glycine. Fluorescence intensity is indicated as a pseudocolor scale. White and red arrows indicated spines with increased and unchanged/decreased TfR-SEP signal after glycine stimulation, respectively. Scale bars, 2 μm and 1 μm. See Movies S9-S11.

(B) Quantitative analysis of surface appearing TfR-SEP in spines. Data represent means ± SEM of the fractional increase in fluorescence intensity (ΔF/F₀). The black bar indicates the period of glycine application. Vec, vector control; Scram, scrambled shRNA.

(C) Data presented as in (B) for neurons expressing MyoVb shRNA along with RNAi-resistant versions of MyoVb-WT, MyoVb-CCtr, or MyoVb-ΔRBD. Note the ordinate scale difference between the graphs in (B) and (C).

(D) Hippocampal neurons expressing MyoVb shRNA were transfected with SEP-GluR1 either alone (MyoVb RNAi) or together with RNAi-resistant MyoVb-WT (Rescue). Stimulus-induced AMPA receptor insertion was monitored by increased SEP-GluR1 fluorescence before and 15 min after glycine stimulation. Yellow and red arrows indicate spines with increased and decreased SEP-GluR1 after glycine respectively. Yellow boxes indicated corresponding magnified regions. Scale bars, 2 μm.

(E) Quantitative analysis of SEP-GluR1 in spines after glycine stimulation. Data represent means ± SEM of the normalized SEP-GluR1 fluorescence intensity in spines (F/F₀). SEP-GluR1 F/F₀ in spines: Vec, 1.87 ± 0.10; Scram, 1.93 ± 0.13; RNAi, 1.07 ± 0.03; WT, 1.97 ± 0.22; CCtr, 1.99 ± 0.14; ΔRBD, 1.22 ± 0.05; Vec + APV (APV), 0.85 ± 0.02. n = 703, 900, 337, 661, 945, and 497 spines from 14, 14, 12, 16, 21, 14 and 10 cells for Vec, RNAi, Scram, WT, CCtr, ΔRBD and APV, respectively; **p<0.001 relative to Vec or Scram; t-test.

(F) Depleting endogenous MyoVb inhibits stimulus-induced spine growth. Yellow and red arrows indicate dendritic protrusions that appeared and disappear after glycine respectively. Scale bars, 2 μm.

(G) Quantitative analysis of spine formation following LTP stimulation. New protrusions per 100 μm of dendrite: Vec, 10.9 ± 2.4; Scram, 10.8 ± 2.1; RNAi, -4.0 ± 2.6; Rescue, 8.9 ± 2.2; Vec + APV, 2.2 ± 1.9. n = 12, 15, 27, 11 and 8 cells for Vec, Scram, RNAi, Rescue, and Vec + APV, respectively. **p <0.001, *p<0.05 relative to Vec or Scram; t-test.

(H) Quantitative analysis of spine growth following LTP stimulation. Normalized spine size S/S₀: Vec, 1.52 ± 0.07; Scram, 1.81 ± 0.15; RNAi, 0.88 ± 0.04; Rescue, 1.91 ± 0.17; Vec + APV, 0.84 ± 0.05. n = 104, 74, 241, 132, and 105 spines from 6, 9, 14, 8 and 5 cells for Vec, Scram, RNAi, Rescue and Vec + APV, respectively. **p <0.001 relative to Vec or Scram; t-test.

To determine whether exocytosis induced by LTP stimuli requires association of MyoVb with the Rab11/Rab11-FIP2 endosomal adaptor complex, we replaced endogenous MyoVb with MyoVb-ΔRBD or MyoVb-CCtr. RNAi-resistant MyoVb-ΔRBD or MyoVb-CCtr cDNAs were expressed in hippocampal neurons previously infected with MyoVb shRNA. In contrast to MyoVb-WT (Figures 6A-C), expression of MyoVb-ΔRBD failed to rescue glycine-induced TfR-SEP exocytosis (Figure 6C). Conversely, expression of MyoVb-CCtr augmented exocytosis above the level observed with MyoVb-WT rescue (Figure 6C). The latter finding is in agreement with the increased recycling endosome localization of MyoVb-CCtr and the enhanced spine trafficking of REs induced by MyoVb-CCtr (Figure 4), and suggests that MyoVb-CCtr increases the pool of REs tethered in spines, priming them for LTP-induced exocytosis. Together these data show that MyoVb is required for LTP-induced exocytosis in dendritic spines, and this exocytosis requires a conformation-dependent association of MyoVb with Rab11-FIP2.

Disrupting MyoVb Prevents Both AMPA Receptor Delivery and Spine Growth Following LTP Stimuli

AMPA receptors are the most celebrated exocytic cargo for glutamatergic synapse plasticity (Derkach et al., 2007; Shepherd and Huganir, 2007; Newpher and Ehlers, 2008). In addition, exocytosis from REs is required for LTP-induced spine growth (Park et al., 2006). To test whether MyoVb mediates requisite exocytic events for postsynaptic plasticity, we monitored GluR1 exocytosis in live hippocampal neurons using SEP-tagged GluR1 (Kopec et al., 2006) together with RNAi knockdown of MyoVb. In control neurons expressing vector alone or a scrambled shRNA, SEP-GluR1 fluorescence in spines increased ∼80% after glycine stimulation, and this increase was blocked by the NMDA receptor antagonist APV (Figure 6D-E). This stimulus-induced increase in SEP-GluR1 fluorescence was abolished in neurons depleted of MyoVb, but was fully rescued by co-expression of RNAi-resistant MyoVb-WT (Figure 6D-E). Moreover, whereas MyoVb-CCtr rescued the effect of MyoVb knock-down, the MyoVb-ΔRBD mutant incapable of binding Rab11-FIP2 did not (Figure 6D-E). Interestingly, MyoVb-CCtr did not over-rescue the insertion of SEP-GluR1 above control as observed for TfR-SEP exocytosis (Figure 6E), suggesting a limiting amount of GluR1 AMPA receptors in REs relative to transferrin receptors. Dominant negative studies similarly showed that the interaction between MyoVb and the RE adaptor Rab11-FIP2 is required for stimulus-induced transport of AMPA receptors to the plasma membrane (Figure S8).

To test whether MyoVb mediates membrane trafficking for spine maintenance, we measured spine number in neurons expressing MyoVb shRNA. Knocking down MyoVb caused a modest decrease in spine number in cultured hippocampal neurons (21.5 ± 1.5% decrease relative to Scram, n = 17 cells, p < 0.05). Next we examined the role of MyoVb in acute spine structural remodeling by assaying glycine-induced growth of dendritic spines in live neurons. In control neurons, glycine LTP was accompanied by an expansion of existing spines and growth of new dendritic spines that was dependent on activation of NMDA receptors (Figures 6F-H). In contrast, glycine-induced spine growth was absent upon MyoVb knock down, and this effect was fully rescued by co-expression of RNAi-resistant MyoVb-WT (Figures 6F-H). These findings demonstrate that MyoVb is required for both AMPA receptor delivery and spine growth during LTP.

Acute Chemical Genetic Inhibition of MyoVb Abolishes LTP in Hippocampal Slices

To rule out effects of chronic disruption of MyoVb and investigate the physiological role of MyoVb in synaptic plasticity, we employed an acute chemical genetic inhibition strategy to rapidly and selectively lock MyoVb on actin and measure its effect on LTP. Transgenic mice expressing MyoVb-Y119G, a mutant MyoVb whose mobility along F-actin can be acutely inhibited by the bulky nonhydrolyzable ADP analog N⁶-2-phenylethyl-ADP (PE-ADP), were generated (Provance et al., 2004). The Y119G mutation enlarges the ATP/ADP binding pocket of MyoVb, rendering it accessible to PE-ADP. In the absence of PE-ADP, MyoVb-Y119G exhibits actin mobility indistinguishable from MyoVb-WT (Provance et al., 2004; Provance et al., 2008). Upon binding the nonhydrolyzable PE-ADP, MyoVb-Y119G tightly associates with F-actin and motor processivity ceases, thereby locking MyoVb in place (Figure 7A). In contrast, wildtype MyoVb is not affected by PE-ADP (Provance et al., 2004).

Figure 7.

Acute Blockade of LTP by Chemical Genetic Inhibition of MyoVb

(A) Schematic diagram illustrating the chemical genetic inhibition strategy using a transgenic mouse line expressing V5-tagged MyoVb-Y119G. See text for details.

(B) Acute inhibition of MyoVb motility disrupts LTP in hippocampal slices. Hippocampal slices were acutely prepared from MyoVb-WT (WT) or MyoVb-Y119G transgenic (Tg) mice, and AMPA receptor-mediated EPSCs were measured as illustrated in Figure S9A. Left panel, LTP was induced using high-frequency stimulation (arrow). PE-ADP or water was included in the recording pipette solution. n = 7, 7 and 12 animals for WT PE-ADP, Tg water and Tg PE-ADP respectively. Error bars indicate SEM. Right panel, representative EPSC traces (averages of six consecutive EPSCs) were evoked from CA1 pyramidal cells in each condition at the times indicated (1 and 2, corresponding to time points in the left panel graph). Scale: 200 pA, 10 msec.

(C) Schematic model for spine mobilization of AMPA receptor-containing recycling endosomes by Ca²⁺-regulated MyoVb during LTP. See text for details.

We used whole-cell patch clamp to measure LTP at the Schaffer collateral-CA1 synapses in hippocampal slices. In control slices from wildtype mice, brief high-frequency afferent stimulation (HFS) of Schaffer collateral inputs (Figure S9A) elicited robust LTP that was unaffected by intracellular perfusion of PE-ADP (Figure 7B). When the ATPase activity of MyoVb was acutely blocked by intracellular perfusion of PE-ADP in CA1 neurons from slices taken from MyoVb-Y119G transgenic mice, LTP was markedly attenuated (Figure 7B). This effect was not due to a general defect in LTP in MyoVb-Y119G mice as robust LTP was elicited upon intracellular perfusion of vehicle control solution (Figure 7B). Moreover, synaptic NMDA receptor currents were unaffected by acute MyoVb inhibition (Figure S9). These results demonstrate that LTP requires the actin-based motility of MyoVb and its AMPA receptor-containing recycling cargo. Moreover, the rapidity of action of PE-ADP in MyoVb-Y199G Tg slices indicates that even a very acute block of MyoVb-dependent trafficking is sufficient to prevent LTP. Thus, Ca²⁺-activated MyoVb mobilizes REs to spines for activity-dependent AMPA receptor insertion and synaptic potentiation.

Discussion

MyoVb as a Multifunctional Motor for Postsynaptic Plasticity

Synaptic NMDA receptor activation causes a fast Ca²⁺ rise in spines and activates downstream effectors to enhance AMPA receptor-mediated synaptic transmission and promote the formation and enlargement of dendritic spines (Alvarez and Sabatini, 2007; Derkach et al., 2007; Shepherd and Huganir, 2007). Dendritic REs containing a local pool of AMPA receptors are rapidly mobilized to the postsynaptic membrane upon an LTP stimulus, providing AMPA receptors, lipid membrane, and presumably additional unknown cargo simultaneously to the spine membrane (Ehlers, 2000; Park et al., 2004; Park et al., 2006). Such findings have implied the existence of one or more “Ca²⁺ sensors” that detects LTP stimuli and converts the signal into enhanced membrane trafficking.

Here we have shown that MyoVb is a Ca²⁺ sensor for actin-based transport of AMPA receptor-containing REs during LTP. LTP stimulation triggers the rapid Ca²⁺-dependent unfolding of MyoVb, allowing binding to Rab11-FIP2 adaptors on REs, and thereby recruiting endosomes into spines to supply AMPA receptors and membrane for functional and structural plasticity (Figure 7C). The interaction between MyoVb and Rab11/Rab11-FIP2 is required for synaptic potentiation. Expression of a MyoVb C-terminal fragment that lacks a region required for Rab11/Rab11-FIP2 binding (Lapierre et al., 2001; Hales et al., 2002; Swiatecka-Urban et al., 2007) had no effect on GluR1 trafficking into spines (Correia et al., 2008), further supporting a requirement for association with Rab11/Rab11-FIP2. Moreover, we have shown that the conformation of MyoVb is regulated by Ca²⁺. Mutations that prevent adoption of the compact folded conformation increase the binding of MyoVb to Rab11-FIP2, likely by preventing intramolecular sequestration of the C-terminal cargo-binding GTD. Under these conditions, endosomal transport and LTP-induced exocytosis in dendritic spines are dramatically increased. In contrast, deletion of the Rab11-FIP2 binding domain renders MyoVb unable to associate with REs and prevents LTP-induced endosomal translocation, AMPA receptor insertion, and synaptic potentiation. By virtue of its localization in dendritic spines, its directional mobility towards the membrane-directed plus end of F-actin (Watanabe et al., 2006), and its Ca²⁺-dependent conformational switch, MyoVb is ideally suited to translate elevations in spine Ca²⁺ into postsynaptic membrane transport.

Organelle Movement and Actin-Based Motors in Synapse Plasticity

The complex geometry of neuronal dendrites and the exquisite synapse-specific regulation of postsynaptic membrane composition has long argued for local mechanisms of regulated membrane trafficking at glutamatergic synapses (Shepherd and Huganir, 2007; Newpher and Ehlers, 2008). The exclusively actin-based cytoskeleton of spines points to the potential involvement of actin-based myosin motors.

Several classes of myosin have been shown to participate in organelle transport in neurons. For example, myosin III mediates light-dependent translocation of visual arrestin in photoreceptor cells (Lee and Montell, 2004). Myosin VI localizes to endocytic vesicles and plays a role in clathrin-mediated endocytosis of AMPA receptors (Osterweil et al., 2005). Myosin Va mediates mRNA translocation into hippocampal neuron dendritic spines following mGluR1 activation (Yoshimura et al., 2006), and is required for the spine localization of smooth ER at cerebellar Purkinje cell synapses (Miyata et al., 2000). Here we have found a unique role for MyoVb in the Ca²⁺-regulated transport of AMPA receptor-containing REs during LTP, the paradigmatic example of regulated membrane trafficking for synapse plasticity. Although MyoVa has been reported to be associated with GluR1 AMPA receptors (Correia et al., 2008), it is not required in vivo for postsynaptic glutamate receptor distribution, excitatory synaptic transmission, short-term plasticity, or LTP (Petralia et al., 2001; Schnell and Nicoll, 2001). Moreover, neither wildtype MyoVa nor the constitutively unfolded D136A mutant MyoVa co-traffic with TfR-positive REs (Figure S2) that contain AMPA receptors (Figure S3C) (Park et al., 2004). Notably, the MyoVa GTD inhibits the ATPase activity of the MyoVb motor domain (Figure S10), likely due to the similar binding interface between the two C-terminal GTDs and the N-terminal motor head domains (Figure S6) (Li et al., 2008). Overexpression of the MyoVa GTD may thus functionally inhibit MyoVb. The very different functions of myosin isoforms at glutamatergic synapses likely reflect the distinct sets of adaptors and cargos that they associate with, and highlight the tight regulation required to tune spine membrane composition and structure.

The Ins and Outs of MyoVb-Mediated Spine Trafficking

The current study raises several questions concerning the details of MyoVb-dependent endosome trafficking in spines. First, upon NMDA receptor-mediated Ca²⁺ influx, how is unfolded MyoVb in spines recruited to REs located predominantly at the base of spines? Biochemical and structural studies of MyoVa show that, in addition to conformational changes, Ca²⁺ also increases MyoVa binding to F-actin and reduces its mobility on F-actin in vitro (Krementsov et al., 2004). Thus, one potential mechanism is that upon Ca²⁺ influx during LTP, increased binding of MyoVb to F-actin combined with actin treadmilling drives MyoVb to the base of spines.

Once associated with Rab11-FIP2 on REs, how are MyoVb-loaded endosomes transported into spines if the motility of MyoVb is decreased by Ca²⁺? Due to the compartmentalized structure of spines, Ca²⁺ influx through NMDA receptors leads to a very local rise in Ca²⁺ concentration (Sabatini et al., 2002), producing a gradient in Ca²⁺ from the postsynaptic membrane to the spine neck entry point off the dendritic shaft. Thus, the motility of MyoVb may be subject to highly local regulation based on Ca²⁺ microdomains, with the reduced Ca²⁺ at the spine entry point sufficient to recover MyoVb processivity.

A third question that emerges from the current study is how exactly cargo from REs is exocytosed at the spine membrane. We have found that the unfolded MyoVb-CCtr mutant increases the mobility of REs in spines but has little effect on the exocytosis rate of REs under basal conditions. Thus we propose that by localizing REs to spines, MyoVb acts to focus or spatially restrict subsequent exocytic transport to the spine plasma membrane, while additional unknown activity-regulated mechanisms are required for the final fusion step from REs during LTP. The mobilization of REs together with localized endocytic cycling adjacent to the PSD (Blanpied et al., 2002; Lu et al., 2007), could both acutely augment and persistently confine new components of a given synapse for enduring synapse modification. More broadly, such a mechanism may provide a general paradigm in diverse cell types for rapid stimulus-dependent regulation of local membrane composition on a micron scale.

Experimental Procedures

DNA constructs, antibodies, cell culture procedures, immunocytochemistry, trafficking assays, and immunoprecipitation methods are included in the Supplemental Experimental Procedures.

Imaging and Image Analysis

Confocal images of fixed samples and live cells were obtained using a Perkin Elmer Ultraview spinning disc confocal microscope with either a 40× 1.3 N.A. objective or a 60× 1.4 N.A. objective. Images were analyzed using Metamorph software (Universal Imaging Corporation). Details are available in the Supplemental Experimental Procedures.

ATPase Activity Assay and Analytical Ultracentrifugation

The ATPase activity of full-length MyoVb was measured as described (Li et al., 2008). Sedimentation velocity analysis was carried out using a Beckman Optimal XL-1 analytical ultracentrifugation as described (Li et al., 2004). Further details are provided in the Supplemental Experimental Procedures.

Transgenic Mice and Electrophysiology

Multiple lines of transgenic mice expressing a V5-tagged sensitized (Y119G) mutant MyoVb were constructed using a SalI digest of the construct previously described in detail (Provance et al., 2004). Whole-cell voltage-clamp recordings were performed on CA1 pyramidal neurons in acute hippocampal slices from 4-5 week old transgenic or wild type FVB mice. See the Supplemental Experimental Procedures for details.

RNA Interference

Four 19-mer shRNA sequence targeting rat MyoVb were examined. The most potent and specific shRNA sequence (CUAACCACAUCUACACUUA) was used to assay the effects of MyoVb depletion on TfR exocytosis, GluR1 insertion and spine growth. Detailed descriptions are available in Supplemental Experimental Procedures.