SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References
  7. Appendix
  8. Supporting Information

Dendritic spines are the site of most excitatory connections in the hippocampus. We have investigated the diffusibility of a membrane-bound green fluorescent protein (mGFP) within the inner leaflet of the plasma membrane using Fluorescence Recovery After Photobleaching. In dendritic spines the diffusion of mGFP was significantly retarded relative to the dendritic shaft. In parallel, we have assessed the motility of dendritic spines, and found an inverse correlation between spine motility and the rate of diffusion of mGFP. We then tested the influence of glutamate receptor activation or blockade, and the involvement of the actin cytoskeleton (using latrunculin A) on spine motility and mGFP diffusion. These results show that glutamate receptors regulate the mobility of molecules in the inner leaflet of the plasma membrane through an action upon the actin cytoskeleton, suggesting a novel mechanism for the regulation of postsynaptic receptor density and composition.

The motility of mammalian cells is of great importance to the function of many cell types. This is not restricted to overtly motile cells such as macrophages; in the nervous system, it is the ability of growth cones to extend in response to their guidance cues which gives rise to the precise patterns of connectivity essential to development. Recently, it has been shown that the postsynaptic structures known as dendritic spines, once viewed as static ‘receivers’ of information, are themselves highly motile (Fischer et al. 1998, 2000; Dunaevsky et al. 1999). Dendritic spines undergo rapid changes in size and shape even in vivo (Lendvai et al. 2000) and can form and disappear over the course of only a few days (Trachtenberg et al. 2002– but see also Grutzendler et al. 2002), in an experience-dependent manner (Shepherd et al. 2003).

Over the last decade it has become clear from many reports that actin can cause movement. The chief experimental models for this are the membrane ruffles and lamellipodia seen in fibroblasts. These studies have demonstrated that actin polymerization and branching can ‘push’ the plasma membrane forwards (reviewed in Pollard & Borisy, 2003). In some of these studies, it has been proposed that an important element is actin nucleation at lipid rafts in the inner leaflet of the plasma membrane (reviewed in Caroni, 2001). This occurs via the neural Wiskott-Aldrich syndrome protein–Arp 2/3 complex and it is thought that the rapid formation of a dense meshwork of short actin filaments at the membrane gives rise to the protrusive force which moves the membrane forwards, while deeper within the cytoplasm, the actin polymers slowly disassociate to preserve the concentration of free actin monomers (Pollard & Borisy, 2003). Since the membrane is a fluid matrix of diffusible lipids and immobilized rafts, the formation of a physical connection between the plasma membrane and the cytoskeleton will result in changes in the laminar diffusion constant of membrane-bound macromolecules. At its simplest, this can be viewed as a change in the viscosity of the milieu immediately adjacent to the membrane. It is possible to test for such an effect by monitoring the rate of diffusion of a membrane-bound probe.

We have investigated the diffusion of a membrane marker in dendritic spines using fluorescence recovery after photobleaching (FRAP). This technique has proven useful for the study of membrane-limited diffusion (Saxton & Jacobson, 1997), and has been used previously to monitor cytoplasmic coupling between dendritic spines and the dendritic shaft (Svoboda et al. 1996; Majewska et al. 2000). Our chosen probe was an enhanced green fluorescent protein (GFP) linked to the membrane via a short palmitoylated peptide sequence. The use of a surface-bound form of GFP enables us to accurately track the boundaries of individual spines, rather than focusing simply on the deepest areas of cytoplasm (which would be the case if we used cytoplasmically localized GFP). Our membrane-bound GFP is tagged to the membrane via the N-terminal fragment of a myristoylated alanine rich C kinase substrate (MARCKS) mutant where the myristoylation site has been exchanged for a palmitoylation site (De Paola et al. 2003). Use of transgenic mice also enables us to avoid perturbing the system to incorporate our marker. This would ordinarily lead to very high background fluorescence from all the GFP labelled cells, but we have avoided this problem by expressing our membrane GFP in the Thy1.2 expression cassette, which results in only a very small subset of cells being labelled (Caroni, 1997; De Paola et al. 2003). This enables us to visualize a particular cell without interference from high surrounding levels of mGFP fluorescence.

Our analysis of the diffusion of a fluorophore bound to the inner leaflet of the plasma membrane indicated that fluorescence recovery rates differed greatly between the main dendritic shaft, and the heads of dendritic spines, consistent with the bulk of the cytoskeletal contacts being in the spine heads. More interestingly, we found that activation of postsynaptic glutamate receptors induced a change in the rate of fluorescence recovery. This effect directly correlated with the ability of glutamate receptor antagonists and agonists to activate and inhibit spine motility, such that more motile spines exhibited a slower recovery of fluorescence. These data provide the first evidence for a link between cell motility and diffusion within the membrane, and suggest a model in which connections between actin and the plasma membrane are responsible for spine motility; glutamate receptor activation results in a severing of these connections between cytoskeleton and membrane. This glutamatergic regulation of membrane dynamics is likely to have important implications for the clustering of postsynaptic receptors, promoting or inhibiting the exchange of extrasynaptic receptors bound within the post-synaptic density (PSD).

Methods

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References
  7. Appendix
  8. Supporting Information

Transgenic mice

Variegated mice were generated using standard techniques. A construct was generated where the cDNA for enhanced green fluorescent protein was fused to the membrane-anchoring domain (first 41 amino acids) of a double palmitoylated mutant of MARCKS (Laux et al. 2000), to form a membrane-targeted marker (mGFP) under the Thy1 promoter. Twenty-five distinct lines were generated, each with subtly different patterns of expression. Of these, L15 mice were chosen as they had low but consistent levels of mGFP positive cells within area CA1 of the hippocampus.

Slice cultures

Organotypic slice cultures were used for these experiments as they provide the major advantage of exhibiting preserved tissue-specific organization of synaptic connections, in an in vitro preparation suitable for imaging studies. Slices were prepared from the hippocampi of 6-day-old L15 mice killed by decapitation following a protocol approved by the Veterinary Department of the Canton of Zurich. The slices (400 μm) were maintained in roller tubes for 3–6 weeks before use, as previously described for rat (Gahwiler et al. 1997).

Confocal imaging

Slice cultures were transferred to a recording chamber mounted on an upright microscope (Leica DM LFSA, Leica microsystems, Heidelberg, Germany) equipped with a heated (30°C) submersion chamber where slices were continually perfused with a solution comprising (mm): NaCl, 137; KCl, 2.7; CaCl2, 2.5; MgCl, 2; NaHCO3, 11.6; NaH2PO4, 0.4; and glucose, 5.6. The confocal scanhead was a Leica SP2. mGFP was imaged using the 488 nm laser line, with an XY dimension of 26 nm pixel−1, and a Z dimension of  200–250 nm (12–18 images per stack). Stack acquisition took 15–20 s. The objective was a Leica water immersion HCX APO 63× 0.9 NA. objective. Bleaching was minimal for acquisition series of < 30 stacks.

Spine motility

Image stacks (4D) were deconvolved using Huygens Pro (Scientific Volume Imaging B.V., Hilversum, the Netherlands) running on a Silicon Graphics Octane workstation (Mountain View, CA, USA), using a full maximum likelihood extrapolation algorithm. Volume rendering and quantification were carried out using Imaris Surpass software (Bitplane AG, Zurich, Switzerland) running on a Windows 2000 Professional workstation. The same parameters were used for all time points of an experimental series. All spines were tertiary apical dendrites of CA1 pyramidal cells.

Spine motility was assessed by measuring the volume of 3D reconstructed spines over time. Averaging all spines, spine volume did not change over the duration of an experiment. This allowed us to define motility as the mean deviation of spine volume from its average value, over time. Quiescent spines were defined as those whose volume changes did not exceed those seen in latrunculin A-treated preparations (where actin-dependent motility is de facto blocked). In Supplementary material Fig. 1, we show repesentative time series of spines with motility measured according to our volume-based method, and the shape factor method (length/width) of Fischer et al. (1998, 2000). Both analyses show a clear time-dependent effect of AMPA (open bar) on motility.

image

Figure 1. mGFP labels the surface of individual neurones in organotypic slices A, 3D reconstruction of a living CA1 pyramidal neurone expressing mGFP. The cell was reconstructed from 60 optical sections taken at 0.5 μm intervals. Scale bar 5 μm. B, a maximum intensity projection of an individual spine labelled with mGFP. The image stack contained 16 sections at 0.25 μm intervals. The line indicates the position of a 1 μm fluorescence line profile, plotted to the right. Two maxima are seen, corresponding to the plasma membrane on either side of the spine. C, a maximum intensity projection of an individual spine labelled by iontophoretic injection of Lucifer yellow. Image obtained as in B. Note the bright fluorescence within the spine head, and the very dim spine neck. The line indicates the position of a 1 μm fluorescence line profile, plotted to the right.

Download figure to PowerPoint

Fluorescence recovery after photobleaching

Fluorescence recovery after photobleaching (FRAP) experiments were carried out by defining a spot (centred on a single pixel) within a spine of interest, and bleaching that spot with 100% laser intensity for 1 s. Recovery was measured by rapid imaging of a single image plane in letter box mode (64 × 256), with a line scanning frequency of 800 Hz. Experiments were analysed in ImageJ (NIH; http://rsb.info.nih.gov/ij). Two trials were carried out per spine, and averaged. Each curve was then individually fitted either with a single exponential, to derive a value for τ, and to derive D the simple diffusion coefficient assuming 1D diffusion along the dendritic spine shaft, or with the equation:

  • image

which is the approximation of Feder et al. (1996) for anomalous diffusion in bilayers, where F0 is the fluorescence immediately after bleach, F0 is the fluorescence prior to bleach, R is the mobile fraction, and α is the time exponent for anomalous subdiffusion. For comparison of diffusion values between treatments, statistical significance was assessed using Student's t test.

Reagents

AMPA and NBQX were supplied by Tocris Cookson (Bristol, UK), latrunculin A by Molecular Probes (Eugene, OR, USA) and other reagents by Sigma (Bucks, Switzerland).

Results

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References
  7. Appendix
  8. Supporting Information

In any imaging study, the localization of the labelling molecule is crucial. In our case, the use of a GFP bound to the inner surface of the plasma membrane allowed us to follow the movement of membrane rather than redistribution of a cytosolic marker. Figure 1A shows an example of a labelled neurone. The advantages in resolution achieved by cell-surface expression of the fluorophore were further optimized by digital deconvolution using a maximum likelihood extrapolation. A comparison of mGFP and cytoplasmic labelling is shown in Fig. 1B and C. The upper example is of a spine labelled with mGFP; the lower is a spine from a neurone in the same slice, labelled by intracellular injection of Lucifer yellow. To show the difference between the two more clearly, a fluorescence line profile across the spine head is plotted to the right of each spine. While the mGFP signal clearly has maxima corresponding to the membrane bounding the spine head, the Lucifer yellow labelled spine has only a single peak, corresponding to the centre of the spine head. The shape of the two membrane peaks in the line profile gives us an estimate of the XY resolution we have obtained. The half-maximum width of these peaks is ∼270 nm, in good agreement with the theoretical maximum emission/2 = 270 nm). These values reflect the light-limited resolution of a point source.

The motility of dendritic spines is known to be both regulated by glutamate receptors, and dependent on actin polymerization. How does activation of glutamate receptors cause a down-regulation of actin-mediated motility? Cell motility in many cell types involves actin nucleation at the membrane, followed by polymerization, which has the effect of pushing the membrane forwards (Pollard & Borisy, 2003). We speculated that a similar mechanism might be involved in spine motility. Fluorescence recovery after photobleaching (FRAP) is an optical technique that can be used to measure the mobility of fluorescent molecules. If the fluorescent molecules are tethered to the membrane, then FRAP measures the laminar diffusion within the membrane leaflet. We reasoned that if spine motility involves interactions between the spine membrane and the actin cytoskeleton, then changes in motility should produce corresponding changes in the ability of mGFP to diffuse within the membrane. Figure 2 shows an example of FRAP of mGFP in a dendritic spine. The images in Fig. 2A show a spine before, and at intervals after, photobleaching. An example of a FRAP curve is shown in Fig. 2B, where fluorescence from the spine head, the spine shaft, and the dendritic shaft are plotted. Fluorescence in the spine head is greatly reduced immediately after photobleaching, but recovers within 10 s. As might be expected, this time course contrasts with that seen using cytosolic indicators such as fluorescent dyes and GFP, where subsecond recovery curves have been observed (Svoboda et al. 1996; Majewska et al. 2000). We also note that the fluorescence within the dendrite itself did not change as a result of the bleach, indicating that membrane diffusion occurs rapidly within this compartment.

image

Figure 2. Using FRAP to measure the mobility of mGFP in the inner membrane leaflet A, time-lapse series showing the time course of FRAP following a 1 s bleach of the spot indicated by the white square. Time in seconds is indicated in the top right hand corner. B, time course of FRAP. Right hand panel, cartoon indicating the areas measured for comparison of FRAP of the spine head. Left hand panel, following bleach as indicated by the arrow, fluorescence recovered fully within 6–7 s in the spine head; fluorescence recovered faster in the spine neck, and bleach was barely detectable in the dendritic shaft.

Download figure to PowerPoint

It has previously been shown that activation of the (S) amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subtype of glutamate receptors gives rise to spines that round-up and freeze (Fischer et al. 2000). We examined the effect of both AMPA-receptor activation and blockade on the diffusion of mGFP in spines. Figure 3A shows that addition of 0.5 μm AMPA to the perfusion medium caused an increase in the rate of fluorescence recovery (τ= 0.89 ± 0.09 s) compared to controls (τ= 3.4 ± 0.1 s), while addition of 6-nitro-7-sulphamoylbenzo[f]quinoxaline-2,3-dione (NBQX; 10 μm), an AMPA/kainate-receptor antagonist, caused a slowing of fluorescence recovery (τ= 4.38 ± 0.23 s). Since AMPA also blocks the motility of spines, and NBQX enhances it, we wondered if the effects of AMPA and NBQX would correlate with mGFP diffusion in static and motile spines which were otherwise untreated. To investigate this possible link between motility and laminar diffusion of mGFP, we measured the motility of individual spines, and then subjected them to FRAP analysis of the diffusion of mGFP. The motility of dendritic spines was examined by imaging 3D stretches of dendrite at high resolution. We then analysed the motility of spines as the mean deviation of volume for the spines at each time point from the average volume over the entire time course of the experiment. We then picked motile or quiescent spines (see Methods) and carried out FRAP experiments upon them. Figure 3B shows the result when spines are grouped according to their motility; motile spines show a slower time course of FRAP (τ= 4.17 ± 0.33 s) than quiescent spines (τ= 2.18 ± 0.21 s).

image

Figure 3. Spine motility correlates with mGFP mobility in the membrane A, spines were imaged over time and treated with glutamatergic agonist (AMPA) or antagonist (NBQX). AMPA-treated spines (red) showed faster recovery of fluorescence after bleaching than NBQX-treated spines (black). Filled circles represent means, thin lines indicate individual data sets. B, a series of experiments where spines were first imaged over time, classified as ‘static’ or ‘motile’, and then subjected to FRAP analysis. FRAP experiments were carried out at the end of imaging, and the recovery plotted against time. Static spines (red) show a faster recovery (smaller t½) than motile spines (black). C, treatment with latrunculin A (LA) speeds the time course of FRAP. In keeping with an effect downstream of the second messenger cascade (probably intracellular Ca2+) LA exerted this effect irrespective of the presence of NBQX (red indicates LA alone, blue indicates LA + NBQX). NBQX alone (black) is replotted from (A) for purposes of comparison.

Download figure to PowerPoint

We hypothesized that the relationship between spine motility and FRAP reflects interactions between the actin cytoskeleton and the plasma membrane. A test of this hypothesis is to treat slices with latrunculin A, an agent which promotes actin depolymerization by sequestering actin monomers, and thus restricts spine motility by preventing actin polymerization. If the speed of FRAP reflects cytoskeletal binding to the membrane then treatment with latrunculin A should speed FRAP. Furthermore, the effect of latrunculin A should be independent of NBQX, since its site of action is downstream of glutamate receptor activation. The data presented in Fig. 3C confirm that the effect of latrunculin A is dominant over that of NBQX. Latrunculin A was applied at 15 μm for 15 min prior to and during imaging. Latrunculin A-treated spines had a τ of 0.85 ± 0.11 s, and coapplication with NBQX did not significantly change this (τ= 0.89 ± 0.12 s). In Fig. 4A we plot the t½ (which, unlike exponential fits, makes no assumptions as to the shape of the data) of fluorescence recovery of individual spines, plotted against the motility measured for those same spines. A strong linear correlation was observed (r= 0.81).

image

Figure 4. Correlation between half-time of FRAP recovery and spine motility A, scatter plot comparing spine motility with the t½ of FRAP. bsl00000, static spines, bsl00001, motile spines; bsl00085, AMPA treatment; bsl00043, NBQX treatment; bsl00066, latrunculin treatment; bsl00072, spines treated with both latrunculin and NBQX. A strong correlation between spine motility and FRAP t½ is seen. B, data from Fig. 4A are plotted after further analysis. If each FRAP curve is analysed and fitted to a simple model of one-dimensional diffusion, the parameter D, diffusion coefficient can be determined. This is plotted against motility, giving a stronger correlation than simply plotting half-time of fluorescence recovery.

Download figure to PowerPoint

A simple model for the diffusion of mGFP seen in our experiments is a 1-dimensional approximation. This is valid since the dendritic shaft itself represents an infinite source of mGFP (FRAP experiments within the shaft itself recover faster than our resolution; data not shown). In essence, this means that as long as the rate of diffusion in the dendrite itself is faster than in the spine, the rate of fluoresence recovery in the spine is dictated by the rate of diffusion within the spine. As a result, since we measure the recovery of fluorescence within the spine head as a whole, we are measuring the movement of mGFP from the shaft along the spine neck, and into the spine head. This results in a direct relationship between τ, D, the diffusion coefficient, and L, the length of the spine neck, where τ=L2/2D. Clearly this only provides an approximation for the diffusion constant, but the value is useful for comparison between groups. Most importantly, this model allows the length of the spine to be directly taken into account.

In Fig. 4B we plot data from all groups where we measured both the motility (see Methods) and FRAP of individual spines. There is a strong inverse correlation between the diffusion coefficient D and motility (r=−0.91), such that spines which are more motile have a smaller value of D than more quiescent ones. This means that motile spines have more restricted laminar diffusion in the inner leaflet of the plasma membrane. The model is justified, since the data are more tightly grouped than in Fig. 4A, where only the empirical t½ is used. A major factor in this improvement is the spine length – time for diffusion is dependent on distance, whereas D is independent of distance.

Studies of membrane diffusion tend to use either single particle tracking or FRAP measurements, and these tend to give rise to different values for D. This has been explained by the existence of lipid microdomains, which constrain diffusion to limited regions. As a result, diffusion over very short time scales (single particle tracking) tends to be fast relative to the bulk measurements on longer time scales which are measured by FRAP data. This has been formally shown by Feder et al. (1996) in work where they derived an approximation of anomalous diffusion which fitted both FRAP data and single particle tracking data. The retardation in mGFP diffusion observed following AMPA/kainate- receptor blockade could be due to changes in the diffusion coefficient, in the immobile fraction, or in the extent of anomalous diffusion. To answer this, we fitted each individual FRAP curve (see Methods) to determine which parameters changed under each condition. The results are summarized in Table 1. Anomalous diffusion is seen as a slowing of the slow component of diffusion, with little change in the fast component (which lies within the lipid microdomain). Figure 5 shows a graphical representation of some of the data in Table 1. As illustrated in Fig. 5A, there is a modest correlation between the level of fluorescence recovery, and motility, but this disappears when the anomalous diffusion model is used. Anomalous diffusion which is reflected in both the ratio of fast and slow components of diffusion, and the value of α did not appear to change in the conditions examined. Values for α remained largely consistent (see Fig. 5B; α= 0.82 ± 0.11 for AMPA-treated spines compared to 0.72 ± 0.12 for NBQX-treated spines.) As shown in Table 1 and Fig. 5C and D both fast and slow components of diffusion changed in parallel; this implies that anomalous diffusion itself was unchanged by the various treatments. As a result, our data are equally well fitted by the simple model where D is the only value to change. The overall fit to the data is improved, however, as evidenced by the change in R, the mobile fraction (∼80% in simple diffusion fits, compared to ∼96% in the anomalous subdiffusion model).

Table 1.  Comparison of simple and anomalous diffusion models
GroupR: % recovery in simple modelD: diffusion coefficientR: % recovery in complex modelα: coefficient of subdiffusionDf: diffusion coefficient (fast)Ds: diffusion coefficient (slow) n (slices)
  1. Table of fitted values using either a simple model of diffusion (first two columns), or modelling with anomalous subdiffusion (last four columns). R is the mobile fraction, D is the diffusion coefficient for simple diffusion, α the anomalous subdiffusion coefficient, Df diffusion measured for the first 3 s, Ds diffusion over 20–30 s. Last column indicates experimental n, with number of slices in parentheses.

Uncharacterized spines79.38 ± 6.540.24 ± 0.1294.56 ± 5.360.73 ± 0.130.54 ± 0.210.031 ± 0.01412 (6)
AMPA88.26 ± 5.350.81 ± 0.0796.33 ± 4.320.82 ± 0.110.93 ± 0.150.038 ± 0.003 9 (5)
NBQX74.56 ± 6.890.13 ± 0.0995.21 ± 5.480.72 ± 0.120.34 ± 0.110.028 ± 0.009 8 (4)
Static spines81.24 ± 6.110.38 ± 0.1295.86 ± 6.010.77 ± 0.110.68 ± 0.180.034 ± 0.006 8 (4)
Motile spines78.34 ± 6.660.13 ± 0.1293.24 ± 6.540.73 ± 0.120.36 ± 0.120.030 ± 0.008 9 (5)
Latrunculin A87.34 ± 7.220.72 + 0.0997.02 ± 5.330.75 ± 0.100.87 ± 0.110.039 ± 0.004 8 (4)
LA + NBQX85.38 ± 6.970.68 + 0.1196.24 ± 6.010.76 ± 0.140.81 ± 0.140.038 ± 0.006 8 (4)
image

Figure 5. Comparison between simple and anomalous models of diffusion A, the percentage recovery of fluorescence obtained by simple exponential fits (simple diffusion in black bars) and anomalous diffusion (grey bars) is compared. There is a modest correlation between motility and the values for R in the simple model, but this disappears in the anomalous diffusion model, where the additional component of diffusion provides a better fit of the data. B, the value for α, the anomalous diffusion coefficient, is largely unchanged under different conditions. C, the mean value for diffusion over rapid time scales (Df) shows a pronounced relation with motility. Statistical significance as assessd by t test is indicated by * where P < 0.05 and ** where P < 0.01. D, the mean value for diffusion over longer time scales (Ds) shows a modest relation with motility. If anomalous diffusion itself was changing, then Ds would be expected to change with no corresponding change in Df. Since this is not the case, and since α is also unchanged, we can conclude that the slowing of FRAP is due to a direct effect on the diffusion coefficient, presumeably due to the increased viscosity of the environment immediately adjacent to the membrane which results from actin polymerization.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References
  7. Appendix
  8. Supporting Information

We have shown for the first time that the diffusion of a surface marker in a cell membrane can be influenced by receptor activation, and that this occurs within a tissue slice, where cell–cell interactions have not been disrupted. This has been achieved by generation of a transgenic mouse expressing a membrane-bound GFP in a subset of neurones, and analysing membrane diffusion using fluorescence recovery after photobleaching.

The diffusion rates of lipids, lipid-linked proteins and integral membrane proteins have been extensively studied over the last 30 years. (Axelrod et al. 1976; reviewed in Saxton & Jacobson, 1997). These studies, initially dependent on FRAP alone, have been supplemented by single particle fluorescence measurements. Diffusion coefficients measured in these ways have indicated that the rate of diffusion scales according to time and distance from origin (known as anomalous diffusion). Despite this, characteristic values have emerged for rates of diffusion in membranes; lipids and lipid-linked proteins typically have rates of diffusion in the range of ∼1 μm2 s−1, whereas integral membrane proteins such as receptors typically have slower rates of diffusion, in the range 0.01–0.4 μm2 s−1 (Saxton & Jacobsen, 1997). In this study we have characterized the diffusion of a lipid-linked (palmitoylated) GFP in the membrane of dendritic spines, the postsynaptic specializations which are the site of the majority of excitatory connections in the hippocampus.

FRAP analysis of mGFP reveals that fluorescence recovers smoothly following bleach, indicating that mGFP can diffuse in the inner leaflet of the plasma membrane in which it is anchored by the N-terminal double palmitation. Activation or blockade of glutamate receptors leads to a speeding or slowing, respectively, of the rate of fluorescence recovery. This most likely reflects diffusion of unbleached mGFP into the spine from the dendrite.

The influence of microdomains

Microdomains within the plasma membrane have been shown to greatly influence the motility of membrane markers, especially in studies using single particle tracking. These microdomains could be due to recruitment and retention of components to lipid rafts, and also to corralling; retention of markers within regions as a result of cytoskeletal binding to membranes (Sheetz, 2001). To take microdomains into account in our analysis, we used the method proposed by Feder et al. (1996), and looked for evidence of a change in α, the coefficient of anomalous diffusion. Instead, we found changes in other parameters, indicating that the retardation of diffusion we have observed is not due to an increase in the number of microdomains crossed by the diffusing marker. Instead we interpret the slowing of diffusion seen in parallel with spine motility as being due to the increased viscosity of the cytoplasm immediately beneath the inner leaflet of the plasma membrane. The increased viscosity we observe might be due to increased filamentous actin networks, and perhaps actin nucleation at the membrane itself. This is an important point which demonstrates that variations in the effective viscosity of the cytoplasm provide a source of diffusional anomalies in addition to membrane subdomains. In the light of these findings, it is particularly interesting that other results suggest that the spine head itself may constitute a lipid microdomain. Since the entire spine head is bleached within 1 s, the apparent value for D within this area is much faster than the value observed during recovery (trecovery is between 2 and 10 s). Thus potential energy barriers exist which resist the exchange of lipids from the spine head to the dendrite. It would not be startling if spine heads did prove to contain lipid microdomains, since in other preparations they typically have areas of 0.5–1 μm2, which compares well to the surface area of a spine head, ∼1 μm2.

Relating motility to diffusion within the membrane

In this study we relate membrane diffusion, actin dynamics, and dendritic spine motility. Although mGFP diffusion and spine motility are correlated, it is necessary to ask whether this relationship is informative or merely coincidental. How does the actin cytoskeleton cause spines to move? From our data, it seems likely that a connection is established between the actin cytoskeleton and the membrane when motility is up-regulated, as it is following AMPA receptor blockade. Dendritic spines contain F-actin which can be considered to be in two forms (Halpain, 2000): a central core which is extremely resistant to turnover (hence the fact that latrunculin A does not cause spines to disappear), and dynamic actin which appears to be responsible for spine motility (Fischer et al. 1998, 2000; Matus, 2000). Different models have been proposed for actin-mediated motility. In Listeria and Shigella, actin-based motility can be reconstituted using actin together with activated ARP2/3 complex, cofilin and capping protein (Loisel et al. 1999); this is as yet not possible for eukaryote systems, suggesting that a more complex regulation is involved. What appears clear is that a combination of factors, including Wiskott-Aldrych syndrome Protein, Arp2/3, cofilin, profilin and other proteins, all interact to shape the directionality and branching of actin polymerization.

At the plasmalemma, this situation is complicated by the need to manipulate the bilayer itself. Likely intermediaries in this role are phosphatidyl inositol (PI) and lipid rafts (Caroni, 2001). Phosphatidyl inositol 4,5-bisphosphate (PIP2) molecules have been shown to modulate actin dynamics (Martin, 1998), actin nucleation (Hill et al. 2000) and its interaction with lipid bilayers (Takenawa & Itoh, 2001); it is therefore plausible to suggest that it is actin polymerization adjacent to lipid microdomains enriched in PIP2 which drives the membrane movements seen in the motility of dendritic spines. A central role for PIP2 would leave spine motility in line with other forms of motility (Sheetz, 2001), and also endocytosis (Schafer, 2003). An important factor in such interactions is the relative abundance of PIP2 in the membrane; although cytoskeletal elements bind PIP2 with relatively low affinity, the concerted action of thousands of such interactions (there are thought to be 103 or 104 PIP2 molecules per μm2 of membrane) has a marked effect. Consequently, reorganization of the cytoskeleton adjacent to the membrane will exert complementary changes on the behaviour of the membrane.

One model for motility suggests that movement can be regulated through the tension exerted on the membrane by the cytoskeleton (Sheetz, 2001). In this model, based on biophysical observations made on membrane ‘blebs’, loss of cohesion between the cytoskeleton and the membrane results in protrusion of the membrane into the extracellular space (Dai & Sheetz, 1999). A similar phenomenon on a much smaller scale could account for the swelling and shrinkage of spine heads observed during dendritic spine motility.

It is therefore particularly interesting that treatments which impair actin dynamics also affect both normal synaptic transmission (Kim & Lisman, 1999) and synaptic plasticity (Kim & Lisman, 1999; Krucker et al. 2000), and inhibition of PI metabolism produces an impairment in both synaptic plasticity (Lin et al. 2001; Sanna et al. 2002) and associative fear conditioning (Lin et al. 2001). Additional evidence links spine motility to synaptic plasticity: growth of dendritic spines has been associated with long-term potentiation in area CA1 of the hippocampus (Maletic-Savatic et al. 1999; Engert & Bonhoeffer, 1999), and in vivo studies support experience-driven changes in spine morphology (Lendvai et al. 2000; Shepherd et al. 2003).

Implications of modulating membrane diffusion

The finding that the diffusion of macromolecules in the inner leaflet of the plasma membrane can be modulated has important implications for receptor clustering. It has become clear that some receptors, such as a subpopulation of NMDA receptors, are able to freely diffuse in the postsynaptic membrane (Tovar & Westbrook, 2002). The importance of this has been directly shown at the neuromuscular junction; loss of presynaptic activity rapidly leads to dispersal of postsynaptic acetylcholine receptors (Akaaboune et al. 2002). A variation of this phenomenon may play a role in synaptic plasticity at central synapses. The glutamate-dependent lowering of diffusion barriers described in our study, combined with a stronger potential energy trap for receptors, could lead to an increased receptor density. Equally, such a mechanism allows for exchange of receptor subunits. Such a diffusion-mediated shuffling of receptors has advantages over receptor endocytosis since it can very rapidly alter postsynaptic responses (as receptors diffuse from the post synaptic density they will be activated increasingly slowly and weakly as the apparent glutamate concentration drops). A detailed discussion of diffusion-mediated receptor shuffling can be found in Choquet & Triller (2003). This selective trapping of receptors may underlie the recently described ability of stargazin to control the clustering of AMPA receptors (Schnell et al. 2002).

The results presented here demonstrate that the diffusion of a lipid-tethered molecule can be regulated by physiologically relevant stimuli, an important advance in the understanding of how membrane biophysics can affect cellular and physiological function. Our results extend our knowledge of the mechanisms of spine motility, and the role played by the actin cytoskeleton in this process. This may represent a phenomenon specific to these neuronal specializations, or be found in other motile structures. In Fig. 6 we present a summary scheme for the relationship between glutamate receptor activation, actin dynamics, and spine motility. We propose that periods of synaptic inactivity lead to actin nucleation and branching adjacent to, or at, the plasma membrane, leading to an increase in spine motility. Synaptic activity, probably acting through intracellular calcium, reverses this process, leading to the breakdown of branched actin filaments at the membrane, and giving rise to a rounded, static spine.

image

Figure 6. Synaptic activity regulates spine motility through interactions between actin and the plasma membrane A schematic diagram illustrates our model for the regulation of dendritic spine motility by glutamate. In the absence of synaptic activity (for example, during pharmacological blockade) a linkage is formed between the cytoskeleton and the inner leaflet of the plasma membrane; this retards diffusion in the lamina. Actin nucleation then drives changes in spine volume. Activation of glutamate receptors severs the linkage between actin and the membrane, speeding membrane diffusion and rendering spines static.

Download figure to PowerPoint

References

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References
  7. Appendix
  8. Supporting Information

Appendix

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References
  7. Appendix
  8. Supporting Information

Acknowledgements

We thank L. Heeb and L. Rietschin for technical assistance, and Drs J. Bai and C. Richards for advice. This work was supported by the Swiss National Science Foundation (31-61518.00), and the Kanton of Zurich.

Supplementary material

The online version of this paper can be accessed at:

DOI: 10.1113/jphysiol.2004.062091http://jp.physoc.org/cgi/content/full/jphysiol.2004.062091v1/DC1 and contains supplementary material consisting of a figure entitled: The effect of AMPA measured by volumetric and shape factor analyses.

This material is also available at:

Supporting Information

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References
  7. Appendix
  8. Supporting Information

Fig. S1. The effect of AMPA measured by volumetric and shape factor analyses.

FilenameFormatSizeDescription
TJP_351_sm_s1.pdf78KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.