Mechanisms of calcium decay kinetics in hippocampal spines: role of spine calcium pumps and calcium diffusion through the spine neck in biochemical compartmentalization

Abstract

Dendritic spines receive most excitatory inputs in the CNS and compartmentalize calcium. Although the mechanisms of calcium influx into spines have been explored, it is unknown what determines the calcium decay kinetics in spines. With two-photon microscopy we investigate action potential-induced calcium dynamics in spines from rat CA1 pyramidal neurons in slices. The [Ca(2+)](i) in most spines shows two decay kinetics: an initial fast component, during which [Ca(2+)](i) in spines decays to dendritic levels, followed by a slower decay phase in which the spine follows dendritic kinetics. The correlation between [Ca(2+)](i) in spine and dendrite at the breakpoint of the decay kinetics demonstrates diffusional equilibration between spine and dendrite during the slower component. To explain the faster initial decay, we rule out saturation or kinetic effects of endogenous or exogenous buffers and focus instead on (1) active calcium extrusion and (2) buffered diffusion of calcium from spine to dendrite. The presence of an undershoot in most spines indicates that extrusion mechanisms can be intrinsic to the spine. Supporting the two mechanisms, pharmacological blockade of smooth endoplasmic reticulum calcium (SERCA) pumps and the length of the spine neck affect spine decay kinetics. Using a mathematical model, we find that the contribution of calcium pumps and diffusion varies from spine to spine. We conclude that dendritic spines have calcium pumps and that their density and kinetics, together with the morphology of the spine neck, determine the time during which the spine compartmentalizes calcium.

Fig. 1.

Kinetics of calcium decays in spines and adjacent dendrites after an action potential. A, Two-photon image of a living CA1 pyramidal cell filled with 200 μm calcium green through a patch pipette (top leftcorner). Image is a collapsed z-stack of images taken 1 μm apart. Inset shows image of a spine in theboxed region of the dendrite of this cell. The spine image was taken at a resolution of 0.1 μm in z. B,[Ca²⁺]_i kinetics in a spine and parent dendrite after a back-propagating action potential. Immediately after an action potential is fired, [Ca²⁺]_iin the spine (solid line) rises to a level ∼1.5 times higher than that of the dendrite (stippled line). All spine traces in the figures are represented by solid lines, whereas dendritic traces are represented bystippled lines. Thin lines represent exponential fits to the data. The dendrite decays down to baseline levels with a single-exponential time scale (τ = 350 msec). Within the first 200 msec after the action potential [Ca²⁺]_i in the spine decays to near-dendritic levels (τ = 110 msec). After a breakpoint (arrow) it then follows the dendritic time scale down to rest, maintaining a slight undershoot. Shown is an average of 10 trials; data were taken at room temperature. C,[Ca²⁺]_i kinetics in a spine and parent dendrite after a back-propagating action potential at 37°C. [Ca²⁺]_i amplitudes in the spine (solid line) reach a level 2.5 times higher than the dendrite (stippled line). The dendrite decays with a single-exponential time scale (τ = 552 msec), while the spine decays with an initial fast decay (τ_f = 93 msec) to near-dendritic levels and then follows the dendritic time scale to basal levels. During the slow decay the spine maintains [Ca²⁺]_i levels lower than the dendrite (undershoot). Shown is an average of five trials. The traces inB and C are taken at a resolution of 12.5 msec/point and filtered with a seven point smoothing kernel.

Fig. 2.

Dependence of the breakpoint of spine decay kinetics on peak [Ca²⁺]_i.A, [Ca²⁺]_i decay kinetics in spine and dendrite, showing the double- and single-exponential fits to the decay. Note how at ∼150 nm[Ca²⁺]_i the spine switches from a fast to a slow time scale (breakpoint marked with grid lines). The trace was filtered with a three point smoothing kernel; average of six traces. B, Data from a different spine, showing a breakpoint (grid lines) at 90 nm [Ca²⁺]_i. Data were filtered with a three point smoothing kernel; average of eight traces.C, The [Ca²⁺]_i at which the breakpoint occurs can change within a single spine. Shown is [Ca²⁺]_i at breakpoint (taken as [Ca²⁺]_i at three initial fast time constants) plotted against peak [Ca²⁺]_i in a single spine during loading with extrinsic buffer. If intrinsic buffer saturation is responsible for the double-exponential kinetics observed in spines, the [Ca²⁺]_i at the point of the breakpoint should stay constant and correspond to theK_D of the intrinsic buffer. Therefore, these data rule out that the double-exponential decay kinetics observed in spines is attributable to buffer saturation.

Fig. 3.

Slow diffusion of calcium green. A,Measurement of calcium green diffusional recovery time in a spine. The graph shows the fluorescence intensity kinetics in a spine. The high-intensity laser pulse (800 μsec exposure produced with 20 msec increase in intensity in line scan mode) produces a sharp decrease in the fluorescence intensity. Immediately afterward, unbleached fluorophore diffuses in from the dendrite, causing a rise in the fluorescence intensity. The recovery curve of the diffusion of unbleached fluorophore was fit to a single exponential (τ = 671 msec; average of five traces, unfiltered). Inset shows the positioning of the line scan used to bleach the dye in the spine, but not the dendrite. B, Measurement of fura-2 diffusional recovery in a spine. The recovery curve of the diffusion of unbleached fluorophore was fit to a single exponential (τ = 87 msec; average of 10 traces, unfiltered).

Fig. 4.

Correlation between the breakpoint and slow component of spine decay kinetics and dendritic [Ca²⁺]_i kinetics. A,[Ca²⁺]_i in spines plotted against [Ca²⁺]_i from adjacent dendritic shafts at the breakpoint. Note the strong correlation between the two variables (r = 0.83; regression ttest; p < 0.0001). B, Correlation between the slow τ of spine and the dendritic τ (r = 0.92; regression t test;p < 0.001). These data indicate that the breakpoint in the spine represents the point of diffusional equilibrium between spine and dendrite, and the slow decay in the spine after the breakpoint has occurred is driven by diffusion from the dendrite.

Fig. 5.

Correlation between spine neck length and initial spine decay kinetics. A,B, Illustrated are how short-necked spines tend to have faster initial decays than longer-necked spines. A shows a spine with a neck length of ∼0.7 μm and τ of 298 msec (average of six trials), whereasB shows a shorter spine with a neck length of 0.25 μm and τ of 37 msec (average of eight trials). Scale bars, 1 μm. Traces were filtered with a seven point smoothing kernel.C, Correlation between initial τ in spines and spine neck length (n = 22; regression ttest; p < 0.06).

Fig. 6.

SERCA blockers affect the dendritic decay and the initial spine decay phase. A, CPA lengthens the decay kinetics of dendrites. The graph shows the [Ca²⁺]_i decay kinetics in a dendrite before and after the application of CPA (30 μm). Thethin lines represent single-exponential fits to the decay kinetics. The dendritic τ (τ_d) was 700 msec in control conditions and 1100 msec once CPA was applied. After a washing with aCSF, the dendritic decay recovered to 680 msec.B, CPA also lengthens the initial fast decay (τ_f) in spines. Shown is [Ca²⁺]_i decay kinetics in a spine before and after the application of CPA (30 μm). Thethin solid lines represent a double-exponential fit to the initial decay kinetics. This effect was reversible with long (>30 min) washes with normal aCSF. Traces were filtered with a three point smoothing kernel and normalized to account for differences in the amplitudes observed between trials.

Fig. 7.

Model of calcium dynamics in spine and dendrite.A, Drawing of the components of our mathematical model of the spine and adjacent dendrite. Influx occurs in both spine and dendrite through voltage-sensitive calcium channels located on the plasma membrane. Efflux in both spine (γ_s) and dendrite (γ_D) is mediated by plasma membrane pumps and exchangers that remove calcium into extracellular space and pumps that sequester calcium to cytoplasmic organelles. With the assumption that the buffering capacities are independent of calcium concentration, the resultant differential equation yields the exact solution given by Equation 6. Equation 5 was solved numerically, allowing the buffering capacities to vary with calcium concentration, keeping the diffusion across the spine neck constant, and changing the relative rates of extrusion between the spine and dendrite. B, Example of a numerical simulation when the rate of pumping in the spine is greater than in the dendrite. Note how the spine undershoots the dendrite during the slow decay phase. Breakpoints are marked witharrows. C, Simulation when the rate of pumping is greater in the dendrite than in the spine under conditions of equal diffusion as in B. Note how the spine overshoots the dendrite. Parameters used in the model: Dendrite radius, 0.2 μm; dendrite length, 10 μm; spine radius, 0.3 μm;K_D of calcium green, 189 nm;K_D of intrinsic buffer, 140 nm; calcium green concentration, 200 μm; intrinsic buffer concentration, 72.66 μm; clearance rate of dendrite, 0.88 μm/μm⁻² per msec; clearance rate of spine in B, 0.18 μm/μm⁻² per msec; clearance rate of spine in C, 0.06 μm/μm⁻² per msec.

Fig. 8.

Heterogeneity of decay phase calcium kinetics.Top panels show examples of [Ca²⁺]_i decay kinetics in spines with different initial fast τ values. A,[Ca²⁺]_i decay kinetics in a spine with initial τ is fast (20 msec; average of five trials).B, Example of a spine for which the initial τ is 280 msec (average of six trials). The middle panels show that spines also differ in their behaviors during the slow decay phase.C, Shown is a spine that overshoots the dendrite during the slow decay phase, suggesting that pumping rates are faster in the dendrite than in the spine (average of 10 trials). D, The spine maintains levels similar to its parent dendrite, suggesting that diffusion is fast and brings the two compartments into equilibrium (diffuser; average of eight trials). Traces were filtered with a seven point smoothing kernel. The bottom panels show that spines that exhibit anomalous behaviors also have slow decays, which depend on dendritic kinetics. E,[Ca²⁺]_i decay kinetics in a spine for which the adjacent dendrite did not undergo a significant [Ca²⁺]_i increase after an action potential. Note that the spine has a single-exponential decay that equilibrates with dendritic resting [Ca²⁺]_i levels, with a τ of ∼150 msec—similar to the fast τ in double-exponential spines (average of six trials, filtered with a five point smoothing kernel).F, [Ca²⁺]_i decay kinetics in a spine in which the dendritic [Ca²⁺]_i rose after the action potential but did not decay significantly. The spine still undergoes the initial fast decay but then follows dendritic kinetics by maintaining a new resting [Ca²⁺]_i(average of 10 trials, filtered with a five point smoothing kernel).