The purpose of this study was to examine whether variability in the shape of dendritic spines affects protein movement within the plasma membrane. Using a combination of confocal microscopy and the fluorescence loss in photobleaching technique in living hippocampal CA1 pyramidal neurons expressing membrane-linked GFP, we observed a clear correlation between spine shape parameters and the diffusion and compartmentalization of membrane-associated proteins. The kinetics of membrane-linked GFP exchange between the dendritic shaft and the spine head compartment were slower in dendritic spines with long necks and/or large heads than in those with short necks and/or small heads. Furthermore, when the spine area was reduced by eliciting epileptiform activity, the kinetics of protein exchange between the spine compartments exhibited a concomitant decrease. As synaptic plasticity is considered to involve the dynamic flux by lateral diffusion of membrane-bound proteins into and out of the synapse, our data suggest that spine shape represents an important parameter in the susceptibility of synapses to undergo plastic change.
Dendritic spines are small protrusions consisting of a head bearing the postsynaptic components of excitatory synapses and a neck connecting the spine head to the dendritic shaft. The structure of heads and necks varies considerably among different spines within the cortex and hippocampus (Harris and Kater 1994) and over time (Fischer et al. 1998), suggesting that changes of structure entail functional consequences. It has been previously demonstrated that spine shape influences the compartmentalization of electrical signals (Tsay and Yuste 2004), calcium dynamics (Hayashi and Majewska 2005; Holcman et al. 2005; Noguchi et al. 2005) and protein synthesis (Steward and Schuman 2001) in many different neuronal systems.
The diffusion of soluble cytoplasmic proteins through the dendritic spine neck has been shown to be regulated by neuronal activity, a compartmentalization that was independent of spine shape (Bloodgood and Sabatini 2005). In cultured hippocampal neurons, a 20 times faster exchange of surface α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptors was observed in stubby spines compared with mushroom spines (Ashby et al. 2006). However, morphological parameters such as surface or length of spines, known to vary considerably among different spines (Harris and Kater 1994), were not directly deducible with the pHluorin-tagged receptor approach employed in this study (Ashby et al. 2006). It remains therefore unknown whether spine morphological parameters influence the lateral diffusion of membrane-associated proteins, a question that is of great importance with respect to synaptic membrane-associated protein trafficking during synaptic plasticity. To address this point, we studied the exchange of membrane associated proteins from dendritic shafts to spine heads with a combination of confocal microscopy, allowing the 3D reconstruction of dendritic spines, and fluorescence loss in photobleaching (FLIP) to observe membrane associated protein trafficking, in living hippocampal CA1 pyramidal neurons expressing membrane-linked green fluorescent protein (mGFP).
Materials and Methods
Variegated mice were generated by using standard techniques. A construct was generated where the cDNA for enhanced GFP was fused to the membrane-anchoring domain (first 41 aa) of a palmitoylated mutant of MARCKS29 under the Thy1 promoter. Twenty-five distinct lines were produced, each with subtly different patterns of expression (De Paola et al. 2003). Of these, L15 mice were chosen because they had a low, but consistent, number of mGFP-labeled cells within the CA1 area of the hippocampus.
Organotypic slice cultures were used for these experiments because they provide the major advantage of exhibiting preserved tissue-specific organization of synaptic connections in an in vitro preparation suitable for imaging studies. Slices (400 μm) were prepared from the hippocampi of 6-day-old L15 mice and maintained in roller tubes for 2–4 weeks before use, as previously described (McKinney et al. 1999).
Slice cultures were transferred to a recording chamber mounted on an upright microscope (DMLFSA, Leica Microsystems, Heidelberg, Germany) equipped with a heated (32 °C) submersion chamber where slices were continuously perfused with a solution comprising 137 mM NaCl, 2.7 mM KCl, 2.5 mM CaCl2, 2 mM MgCl2, 11.6 mM NaHCO3, 0.4 mM NaH2PO4, and 5.6 mM glucose. The confocal scanhead was a Leica SP2. mGFP was imaged by using the 488 nm argon laser line, with voxel dimensions of 9 × 9 x 100–150 nm. Tertiary and some secondary dendrites of mGFP-labeled pyramidal cells were imaged using a 63× water immersion long-working distance lens (HCX APOL 63× 0.9 w U-F-I, Leica Microsystems, Heildberg, Mannheim, Germany). Additional optical sections were taken above and below the structure of interest to register changes over time. Once the images were captured, the area of interest was cropped and further processed.
Image stacks (3D) were deconvolved with Huygens Pro software (Scientific Volume Imaging, Hilversum, The Netherlands, supplied by Bitplane, Zurich, Switzerland) running on a Silicon Graphics Octane workstation (Mountain View, CA) using a full maximum likelihood extrapolation algorithm. Volume rendering and quantification was carried out with Imaris Surpass software (Bitplane, Zurich, Switzerland) running on a Windows 2000 workstation (Professional version, Microsoft, Redmond, WA). The same parameters were used for all time points of an experimental series.
Fluorescence Loss in Photobleaching
FLIP experiments were carried out by repeatedly bleaching the dendritic shaft with 100% of the 488-nm laser line during image frame acquisition at a frequency of 2.49 Hz during 280.7 s (line scanning frequency of 1000 Hz). Average fluorescence was measured in the spine head. Each FLIP session was preceded by acquisition of 50 frames (at 2.49 Hz) to measure the intrinsic (i.e., loss of fluorescence induced by image frame acquisition) bleaching and the maximal fluorescence. Intrinsic bleaching measured in the spine head after these 50 first frames was fitted by the following monoexponential function: y = 0.99 + 0.048 · exp(−t/3.44); (n = 54 spines in 16 cultures) and was always less than 5% after 400 s (n = 3; Supplementary Fig. 1). For image frame acquisition, the pinhole was completely opened (600 μm) to minimize the impact of spine movements on fluorescence measurements. The image pixel dimension was 20–30 × 20–30 nm. Loss in fluorescence during photobleaching was measured subsequently in the whole spine head with Leica LCS Lite software, fitted and statistically analyzed using Origin 5.0 software (Microcal software, Leica Microsystems, Heidelberg, Mannheim, Germany). Numbers are given as mean ± SEM.
Epileptiform Activity Protocol
Epileptiform activity was induced by blocking fast inhibitory transmission with the GABAA (gamma-aminobutyric acid) receptor antagonist bicuculline methochloride (Sigma, St. Louis, MO, 20 μM) and by extracellular electrical stimulation (every 15 s) applied via a monopolar glass stimulating electrode placed in the stratum radiatum of the CA3 hippocampal region. The induction of epileptiform activity was confirmed by recording the population burst discharge (field potentials) with extracellular glass electrodes containing 3 M NaCl placed close to the imaged cell in the stratum radiatum of the CA1 hippocampal region (Supplementary Fig. 2).
FLIP Kinetic Parameters Vary between Different Spines
We used organotypic hippocampal slice cultures from variegated transgenic mice expressing enhanced GFP linked to the membrane (mGFP) in a small fraction of CA1 pyramidal neurons, allowing the visualization of their neuronal membranes (Fig. 1). In all experiments, we first imaged ≈10 μm portions of second–third order dendrites using confocal microscopy, and subsequently reconstructed the spines in 3 dimensions (Fig. 1). To compare the membrane-associated protein flux between dendritic shafts and spine heads, we used a FLIP approach, which allowed us to assess the exchange rate of fluorescent molecules between these 2 compartments.
Repetitive illumination of dendritic shafts resulted in a progressive loss of fluorescence in neighboring dendritic spines (Fig. 2). After 700 illuminations at 2.49 Hz, the fluorescence of neighboring dendritic spine heads was reduced by 70.0 ± 1.4% (n = 54 spines in 16 cultures). This decrease in spine head fluorescence is reproducible after 15 min (Supplementary Fig. 1), and best fitted with a biexponential function: y = y0 + Afast · exp(−x/τfast) + Aslow · exp(−x/τslow), where in average y0 = 30.0 ± 1.0%, corresponding to the immobile fraction (i.e., the fluorescence that cannot be removed from the spine head compartment, whereas bleaching the parent dendritic shaft), Afast = 44.3 ± 2.3%, Aslow = 25.3 ± 1.6%, τfast = 10.4 ± 0.64 s and τslow = 101.4 ± 13.2 s (all n = 54 spines in 16 cultures). The immobile fraction is significantly different from 0 (t-test, P < 0.01), indicating that a fraction of the dendritic spine membrane is not exchanged with the dendritic shaft in the time frame of the experiment. The rate of decrease in fluorescence described by the time constants (τfast and τslow) of these exponential functions corresponds to the kinetics of mGFP exchange between the dendritic shaft and the spine head compartment. FLIP kinetic parameters varied considerably between different spine shapes (Fig. 2), revealing differences in mGFP exchange between dendritic shaft and spine head compartments.
FLIP Kinetics Correlate with Dendritic Spine Shape Parameters
To assess the role of spine shape in membrane-associated protein exchange, we checked for correlations between FLIP kinetics and spine topological parameters. No correlation was detected between the immobile fraction and any of the other FLIP kinetic parameters, indicating that the size of immovable membrane pools (mostly intraspine membranes and to a small extent constrained synaptic membrane) is in this time frame not directly linked to the loss of fluorescence kinetics (R < 0.3 and/or P > 0.05, n = 54 spines).
Neither Afast/slow nor τslow were correlated with any of the measured spine topological parameters (R < 0.3 and/or P > 0.05, n = 54 spines for length measurements and 47 for surface measurements, in 16 cultures). Interestingly, τfast, the time constant of the fast exponential describing the decrease of 63.3% of fluorescence loss, is linearly correlated with total spine length and spine neck length (R = 0.64 and R = 0.51 respectively, P = 1· 10−4, n = 54 spines in 16 cultures, Fig. 3). Furthermore, τfast is also linearly correlated with the total spine surface and the spine head surface (R = 0.65 and R = 0.57 respectively; P = 1· 10−4, n = 47 spines in 16 cultures, Fig. 3). These data demonstrate the involvement of spine topological characteristics, particularly the length of spine necks and the surface of spine heads, in the exchange of mGFP from dendritic shafts to spine heads.
Epileptiform Activity-Induced Reduction in Spine Length and Surface Area is Associated with Accelerated FLIP Kinetics
On the basis of our results, we hypothesized that a reduction in length and/or volume of a given dendritic spine will be followed by an acceleration of mGFP exchange between the dendritic shaft and the spine head (and consequently by a reduction of τfast). To test this hypothesis, we took advantage of the reduction in spine length and surface occurring within minutes after induction of epileptiform activity (Supplementary Fig. 2). Epileptiform activity reduced dendritic spine length by 15 ± 6% (1.6 ± 0.1 μm vs. 1.4 ± 0.1 μm, n = 15 spines in 2 cultures, paired t-test, P = 9· 10−3) and dendritic spine surface by 21 ± 6% (3.1 ± 0.6 μm2 vs. 2.4 ± 0.5 μm2, n = 12 spines in 2 cultures, paired t-test P = 6· 10−3) (Fig. 4). As FLIP kinetics remained unchanged during successive bleaching experiments with 20 min intervals (Supplementary Fig. 1), we were able to measure FLIP kinetics before and after 30 min of epileptiform activity. The reductions in dendritic spine length and surface were accompanied by a 23 ± 7% reduction of τfast (τfast control = 12.5 ± 1.4 s vs. τfast epileptiform = 9.6 ± 1.4 s, n = 12 in 2 cultures, paired t-test, P = 0.02, Fig. 4). The concomitant change of spine length/surface and τfast confirms the involvement of spine topological parameters in the control of lateral diffusion of membrane-associated molecules from dendritic shafts to spine heads.
The present data show that neck length and head surface influence the kinetics of mGFP exchange between dendritic shaft and spine head compartments. Dendritic spines with long necks and/or voluminous heads exhibit slower membrane kinetics than those with short necks and/or small heads. To rule out the possibility that the slower kinetics in short spines are due to a difference in spine development, the spine head/dendritic shaft mGFP exchange was studied in spines that exhibited reduced length and volume as a result of epileptiform increased activity. Under these conditions, the reduction in spine neck length and head surface was clearly associated with a decrease in τfast, thus confirming the role of spine shape in the exchange of membrane-bound molecules.
As suggested by previous modeling experiments, a consequence of a spine's topology is the reduced effective surface diffusivity of membrane-bound molecules (Biess et al. 2007; Bressloff and Earnshaw 2007); this diffusion trap due to spine geometry is sufficient to account for our data.
The measured decrease in spine head fluorescence corresponds to the exchange of spine head mGFP with dendritic shaft bleached mGFP. This loss is most presumably due to lateral diffusion within the plasma membrane, as the described local recycling occurs within the dendritic spine (Brown et al. 2007) and does not involve the bleached area, and therefore cannot reduce global fluorescence of the spine head. Nevertheless, a contribution of spine head/dendritic shaft vesicle exchange in the FLIP kinetics cannot be excluded. Nevertheless, only vesicle transfer from spine head toward the shaft would result in a change in FLIP kinetics, as recycling endosome vesicles are transferred from the shaft (Brown et al. 2007) and are therefore bleached and cannot contribute to the loss of fluorescence. Such a hypothetical loss in fluorescence due to vesicle transfer from spine heads to dendritic shafts would be likely to undergo slower kinetics than lateral diffusion within the plasma membrane, and both phenomena would therefore be resolved as having distinct time constants. τslow, which is not correlated with the spine geometry, might correspond to this vesicle transfer.
We have used a GFP that is tagged to the membrane by fusion with a plasmalemmal targeting sequence modified to include palmytoylation sites (De Paola et al. 2003). At our spatial resolution, mGFP appears to uniformly stain the plasma membrane, with no detectable signal in microdomains (Supplementary Fig. 3 and also Galimberti et al. 2006). Although an artificial construct, this lipid-tethered molecule is likely to diffuse in a manner similar to endogenous freely diffusing membrane-associated molecules. This feature is particularly relevant for neurotransmitter receptors subjected to nonconfined diffusion in nonsynaptic areas (Borgdorff and Choquet 2002; Dahan et al. 2003). Despite a faster diffusion of mGFP compared with AMPA receptors, both membrane-associated proteins seem to be subjected to the same constraints in dendritic spine/shaft exchange (Ashby et al. 2006). Synapses are known to be dynamic structures where membrane-associated proteins are continuously entering and leaving the synaptic scaffolding structures through lateral diffusion in the membrane leaflets (Borgdorff and Choquet 2002; Dahan et al. 2003; Ashby et al. 2004; Triller and Choquet 2005). The exchange of membrane-associated proteins between spine heads, where most excitatory postsynaptic elements are localized, and dendritic shafts, where freely diffusible neurotransmitter receptors are present at low density (Borgdorff and Choquet 2002; Dahan et al. 2003; Ashby et al. 2004; Triller and Choquet 2005; Scott et al. 2006), is therefore of fundamental importance for activity-dependent synaptic plasticity. Consequently, synaptic plastic events can be interpreted as changes in the rate of exchange of synaptic proteins such as AMPA receptors and postsynaptic density proteins such as PSD-95 between synaptic and nonsynaptic areas.
The involvement of receptor trafficking in synaptic plasticity is suggested by experiments showing that the removal of AMPA receptors from synapses during chemically induced LTD is preceded by a transient endocytosis specifically of extrasynaptic AMPA receptors (Ashby et al. 2004). Conversely, incorporation of AMPA receptors into the synapse has been demonstrated during chemically induced long-term potentiation (LTP; Malinow and Malenka 2002), and trafficking of GluR1 from the nonsynaptic dendritic surface to the synaptic spine surface is considered to play a significant role in LTP (Bredt and Nicoll 2003). The recent demonstration of LTP-induced incorporation of AMPA receptors into synapses (Kopec et al. 2006) does not contradict this mechanism as the exact locus of exocytosis remains unknown and the diffusion of synaptic receptors subsequent to dendritic shaft exocytosis cannot be excluded.
A similar diffusion trap mechanism has also been implicated in the activity-dependent spine head recruitment of dopamine 1 receptors from dendritic shaft membrane stores (Scott et al. 2006), and the “spine geometry” diffusion trap suggested by our data might act along with the diffusion trap involving interaction with synapse scaffolding proteins (Ehlers et al. 2007) to sequester AMPA receptors in synaptic areas.
The effect of dendritic spine topology on the exchange of membrane-associated molecules between synaptic and nonsynaptic areas is likely to be a critical factor controlling the probability that a spine undergoes synaptic plasticity, as suggested by modeling studies (Biess et al. 2007). Thus, we suggest that spine shape reflects the susceptibility for plasticity rather than the state of plasticity of the synapses it bears. Our data offer an explanation for the association between plasticity and spine shape that has been observed in the hippocampus (Matsuzaki et al. 2004) and in the amygdala (Humeau et al. 2005) and for the 20 times faster turnover of AMPA receptors in short stubby spines vs. mushroom spines in cultured hippocampal neurons (Ashby et al. 2006).
Dendritic spine morphology not only varies among different spine types but also over time (Fischer et al. 1998). To assess the impact of rapid changes in spine shape, we took advantage of an epileptiform activity paradigm reducing spine length and surface within minutes. Although induced by non physiological activity, the rapid spine shape remodeling is associated with changes in the kinetics of membrane-associated exchange of molecules between spine heads/dendritic shafts, consistent with our results.
In addition to spine shape, other factors such as lipid rafts (Hering et al. 2003) and protein scaffolds can alter the lateral mobility of membrane-associated proteins. Moreover, glutamate receptor activation regulates the mobility of membrane-associated molecules in spine heads through an action on the actin cytoskeleton (Richards et al. 2004) and the trafficking of transmitter receptors (Groc et al. 2004). Together with these other mechanisms modifying lateral mobility of membrane-associated proteins between synaptic and nonsynaptic areas, rapid changes in dendritic spine shape (Fischer et al. 1998) will likely contribute to the fine-tuning of the susceptibility for plasticity of synapses located on dendritic spines.
Swiss National Science Foundation Grant (31-61518.00); Canadian Institutes of Health Research operating grant to R.A.M.; Canadian Foundation of Innovation operating grant to R.A.M.; and National Sciences and Engineering Research Council of Canada operating grant to R.A.M.
We thank L. Rietschin for excellent technical assistance and Prof. U Gerber for detailed comments and critical reading of the manuscript. Conflict of Interest: None declared.