Dendritic NMDA receptors activate axonal calcium channels

Summary

NMDA receptor (NMDAR) activation can alter synaptic strength by regulating transmitter release from a variety of neurons in the CNS. As NMDARs are permeable to Ca²⁺ and monovalent cations, they could alter release directly by increasing presynaptic Ca²⁺ or indirectly by axonal depolarization sufficient to activate voltage-sensitive Ca²⁺ channels (VSCCs). Using two-photon microscopy to measure Ca²⁺ excursions, we found that somatic depolarization or focal activation of dendritic NMDARs elicited small Ca²⁺ transients in axon varicosities of cerebellar stellate cell interneurons. These axonal transients resulted from Ca²⁺ entry through VSCCs that were opened by the electrotonic spread of the NMDAR-mediated depolarization elicited in the dendrites. In contrast, we were unable to detect direct activation of NMDARs on axons indicating an exclusive somatodendritic expression of functional NMDARs. In cerebellar stellate cells, dendritic NMDAR activation masquerades as a presynaptic phenomenon and may influence Ca²⁺-dependent forms of presynaptic plasticity and release.

Introduction

NMDARs are found throughout the central nervous system and contribute to synaptic excitability and intracellular Ca²⁺ transients. NMDARs were once thought to be expressed exclusively in somatodendritic membranes, concentrated at the postsynaptic density of glutamatergic synapses (Fagg and Matus, 1984; Monaghan and Cotman, 1986). An emerging view suggests that NMDARs are also expressed in axons because NMDAR activation can alter spontaneous and action potential-evoked transmitter release (Berretta and Jones, 1996; Bardoni et al., 2004; Sjostrom et al., 2003; Corlew et al., 2007; Yang et al., 2006; Brasier and Feldman, 2008). Given the Ca²⁺-dependence of neurotransmitter release and modulation (Zucker and Regehr, 2002), these results imply that there is a close spatial association of NMDARs and presynaptic release sites because of the limited intracellular diffusion of NMDAR-mediated Ca²⁺ entry (Mainen et al. 1999; Sabatini et al. 2002). However, results from dentate granule cells and cortical pyramidal cells indicate that subthreshold somatodendritic depolarizations can enhance axonal release by passively depolarizing axonal release sites (Alle and Geiger, 2006; Shu et al., 2006). It is possible, then, that potentiation of release by NMDARs is caused, at least in part, by somatodendritic NMDAR activation and passive propagation of the resulting depolarization to axonal release sites.

Stellate cells are a class of interneuron located in the molecular layer of cerebellar cortex (Palay and Chan-Palay, 1974). They express NMDARs in an atypical pattern in their dendrites in that they surround the postsynaptic density rather than reside within it (Clark and Cull-Candy, 2002). Several studies report that NMDARs are also expressed on the axons of stellate cells. In the absence of action potentials, bath application of NMDA increases the rate of spontaneous exocytotic release of GABA from stellate cells onto Purkinje cells (Glitsch and Marty, 1999; Duguid and Smart, 2004; Huang and Bordey, 2004; Glitsch, 2008) and elevates Ca²⁺ in their axons (Shin and Linden, 2005).However, direct detection of presynaptic NMDA receptors remains elusive (Clark and Cull-Candy, 2002) and the contribution of somatodendritic NMDARs to axonal Ca²⁺ elevation has not been determined.

In this study, we have investigated the distribution of NMDARs in stellate cells using two-photon laser-scanning microscopy and Ca²⁺ imaging. We find that NMDARs are expressed on dendrites but not on axons. However, Ca²⁺ transients are evoked in axons by activation of dendritic NMDARs either by exogenous agonists or synaptic stimulation. NMDAR-mediated Ca²⁺ transients in stellate cell axons result from VSCCs opened by passive spread of the dendritic NMDAR depolarization. Our results suggest that NMDAR depolarization-mediated Ca²⁺ transients in axons will have a profound influence on release.

Results

NMDAR-mediated Ca²⁺ transients in stellate cell dendrites and axons

To probe for NMDAR activity we imaged Ca²⁺ transients evoked by bath applied NMDA (10 µM) in the axons and dendrites of stellate cells in rat cerebellar slices (0 mM Mg²⁺, 0.5 µM TTX). Cells were filled through the patch pipette with the red fluophore Alexa 594 (50 µM) to visualize cell morphology and the green calcium indicator Fluo-5F (200 µM) for measuring Ca²⁺ transients. After loading, dendrites and axons were easily resolved (Figure 1A). Axons were distinguished by a long, complex plexus of collaterals studded with varicosities whereas dendrites were short and not varicose (Palay and Chan-Palay, 1974). Images of dendrites during NMDA application revealed large increases in intracellular Ca²⁺ in current-clamped stellate cells (Figure 1B) consistent with the activation of Ca²⁺–permeable NMDARs (MacDermott et al. 1986). Concurrent with the rise in Ca²⁺, NMDA application also resulted in a significant depolarization (17.7 ± 0.6 mV, n= 28) of the stellate cell as recorded in the soma (Figure 1B). The large magnitude of the NMDAR-mediated response is attributable to the absence of extracellular Mg²⁺ that, when present, blocks NMDARs in a voltage-dependent manner (Mayer et al., 1984;Nowak et al. 1984). For example, with 1 mM extracellular Mg²⁺, 30 µM NMDA, the concentration used in previous reports (Glitsch and Marty, 1999; Huang and Bordey,2004; Glitsch 2008), was required to generate a depolarization of similar size (22.4 ± 1.1 mV, n= 5). It is unlikely that Ca²⁺ influx through VSCCs contributed significantly to the NMDA-evoked Ca²⁺ transient recorded in the dendrite because direct depolarization by somatic current injection resulted in very little Ca²⁺ elevation (Figure 1B).

Figure 1. Bath applied NMDA evokes Ca²⁺ entry in stellate cell dendrites and axons.

(A) Two-photon fluorescence image of a cerebellar stellate cell filled via patch pipette with 50 µM Alexa 594 and 200 µM Fluo-5F. Magnified views show a dendrite (left boxed inset) and an axon studded with varicosities (right boxed inset). The axon segment shown in the right inset is not part of the cell shown in A. (B) A sequence of frame scans show that application of NMDA (10 µM, 0 mM Mg²⁺) increases Ca²⁺ in a segment of dendrite (frame 540s) while somatic depolarization (frame 1460s) resulted in very little Ca²⁺ entry. Plot shows the average Ca²⁺ increase in dendritic segments elicited by NMDA superimposed on the average somatic membrane potential. Images and voltage sweeps taken every 20 seconds. (C₁) Sequence of frame scans from an axon segment showing action potential-evoked Ca²⁺ entry. (C₂) Frame scans from the same axon segment show Ca²⁺ entry induced by bath applied NMDA (10 µM, 0 mM Mg²⁺) and somatic depolarization (frame 540 s and 1420 s, respectively). Plot shows the average change in Ca²⁺ recorded in axon segments and somatic membrane potential. (D) Dependence of axonal Ca²⁺ elevation on distance from the axon hillock (10 µM NMDA, 0 mM Mg²⁺).

We then imaged the bead-like varicosities along axons for NMDAR-mediated Ca²⁺ influx (Shin and Linden, 2005) where the activation of axonal NMDARs is thought to increase the rate of spontaneous inhibitory neurotransmission (Glitsch and Marty,1999; Duguid and Smart, 2004; Huang and Bordey, 2004; Glitsch 2008). Axon varicosities are hot spots of action potential-evoked Ca²⁺ entry (Figure 1C₁) indicative of clustered VSCC activity and therefore are presumed en passant sites of release (Palay and Chan-Palay, 1974; Llano et al., 1997; Forti et al., 2000). We found that bath application of NMDA (10 and 30 µM; 0 and 1 mM Mg²⁺, respectively) also resulted in an elevation of Ca²⁺ in axon varicosities (Figure 1C₂ and 1D, ΔG/R= 0.0219 ± 0.0043, n=30; ΔG/R= 0.0090 ± 0.0005, n= 5, 10 and 30 µM NMDA, respectively, p= 0.24).However, the Ca²⁺ rise in axons was much smaller than that recorded in dendrites (p< 0.05). In contrast to dendrites, matching the NMDA-induced depolarization by somatic current injection resulted in a Ca²⁺ elevation in the axon that was indistinguishable from the transient evoked by bath applied NMDA (n= 4, p= 0.64; Figure 1C₂). This suggests that NMDAR-dependent Ca²⁺ elevation in axon varicosities is controlled by membrane potential and VSCCs rather than direct Ca²⁺ influx through NMDA receptors.

If axonal NMDARs mediate the depolarization underlying Ca²⁺ accumulation in axon varicosities, then bath NMDA-evoked Ca²⁺ transients should be observed throughout the axonal arbor. However, we found that NMDA-evoked Ca²⁺ transients in proximal axon varicosities were often much larger than those evoked in distal varicosities (Figure 1D). Nonlinear regression analysis revealed an inverse correlation (r²= 0.98) between the amplitude of the Ca²⁺ increase in axon varicosities and distance from the soma. A monoexponential fit of binned data (50 µm bins) yielded a decay length constant of 212 µm. We therefore propose that NMDA-evoked Ca²⁺ elevation in axons results from an NMDAR depolarization originating in the somatodendritic compartment that passively propagates into the axon.

Dendritic NMDAR potentials and axonal Ca²⁺ entry

To determine if depolarization resulting from selective activation of dendritic NMDARs is sufficient to control axonal Ca²⁺ entry without the concomitant activation of axonal NMDARs, we used iontophoresis of the NMDAR agonist L-aspartate to locally activate NMDARs on stellate cell dendrites (Figure 2A₁) in the presence of TTX (0.5 µM). In current-clamped cells, dendritic iontophoresis of L-aspartate (−50 to −200 nA; 10–50 msec) elicited brief NMDAR-mediated potentials (29.8 ± 1.7 mV, n=13; Figure 2A₂) that resembled EPSPs (see Figure 7A₂) and evoked Ca²⁺ transients in dendrites near the iontophoretic pipette (Figure 2B and 2C). The NMDAR antagonist (R)-CPP (20 µM) blocked the Ca²⁺ transient evoked by iontophoresis (ΔG/R 5.9 ± 4.3% of control, n= 7;Figure 2C) indicating that the Ca²⁺ transient was NMDAR-mediated. We found that dendritic L-aspartate iontophoresis also evoked Ca²⁺ transients in axon varicosities (Figure 2D). These transients were much smaller in amplitude than action potential-evoked Ca²⁺ transients in axon varicosities or the Ca²⁺ transients evoked by iontophoresis recorded in the dendrite (Figure 2D and 2E).

Figure 2. Dendritic NMDAR-mediated Ca²⁺ entry in axon varicosities.

(A₁) Image of a stellate cell shows the location of the iontophoretic pipette used for dendritic L-aspartate application. A magnified view of the dendrite (inset) shows the location of the line scan (dashed line). (A₂) Dendritic iontophoresis of L-aspartate evoked an NMDAR-mediated depolarization (V_rest= −71 mV). (B) Line scans from the dendrite in response to L-aspartate iontophoretic stimulation. Iontophoretic pulse lasted 15 msec. (C) The iontophoresis-evoked Ca²⁺ transient recorded in the dendrite was blocked by the NMDAR antagonist (R)-CPP (20 µM). (D) Ca²⁺ transients were evoked in an axon varicosity (inset) by an action potential and by dendritic iontophoresis. (E) Average Ca²⁺ transients evoked in dendrites and axon varicosities. (F) Somatic voltage clamp eliminated the Ca²⁺ transient evoked in an axon varicosity by dendritic iontophoresis. (G) Somatic voltage clamp did not alter the iontophoresis-evoked Ca²⁺ transient in a dendrite. (H) Average effect of somatic voltage clamp on Ca²⁺ transients in axons and dendrites evoked by dendritic iontophoresis.

Figure 7. PF-evoked Ca²⁺ transients in axon varicosities.

(A₁) Fluorescence image of a stellate cell shows the dendritic position of the extracellular pipette used to stimulate PFs. Magnified view of an axonal varicosity (boxed inset) used for Ca²⁺ imaging. (A₂) NMDAR-mediated EPSP evoked (V_rest= −70 mV) by focal stimulation of PFs (50 Hz, 80 ms). (B) PF stimulation evoked Ca²⁺ transients in the dendrite and axon varicosity of the cell shown in A₁. The PF-evoked axonal Ca²⁺ transient is shown (bottom) at higher gain. (C₁) Somatic voltage clamp did not alter the dendritic Ca²⁺ transient. (C₂) Somatic voltage clamp eliminated the Ca²⁺ transient in an axon varicosity evoked by dendritic PF stimulation. (D) An axonal Ca²⁺ transient evoked by dendritic PF stimulation. PF stimulation near the axon (middle) failed to evoked a Ca²⁺ in an axon varicosity. The axonal signal with no stimulation (bottom). (E) Summary data show that local stimulation of PFs near axon varicosities failed to evoke a Ca²⁺ transient.

We confirmed that the axonal Ca²⁺ transient depended on the NMDAR-mediated depolarization originating in the dendrite by voltage-clamping the soma. This manipulation prevents the passive propagation of dendritic depolarization into the axon and eliminated iontophoresis-evoked Ca²⁺ entry in axon varicosities (Figure 2F and 2H). This observation is reminiscent of experiments in dentate granule cells and cortical pyramidal cells showing that subthreshold depolarizations originating in the somatodendritic domain can passively propagate into axons (Alle and Geiger, 2006; Shu et al., 2006). Similarly, somatic voltage-clamp dramatically reduced the bath NMDA-evoked Ca²⁺ transient recorded in axon varicosities (ΔG/R 25.3 ± 7.5% of current-clamp,n= 3; p< 0.05) suggesting a similar pathway of Ca²⁺ entry. The Ca²⁺ component remaining in voltage-clamp probably reflects diffusion of Ca²⁺ from the dendrite into the axon during prolonged bath application of NMDA. In contrast, somatic voltage-clamp did not alter the dendritic Ca²⁺ transient evoked by L-aspartate iontophoresis (Figure 2G and 2H) indicating that the dendritic signal is the result of Ca²⁺ influx through local NMDARs.

Cable theory predicts that nonregenerative potentials will passively decay along lengths of dendrite and axon (Rall 1969). Therefore, distance-dependent shaping of depolarizing potentials that control Ca²⁺ entry in axon varicosities should be reflected in the amplitude and shape of the evoked Ca²⁺ transient. As with bath applications of NMDA, the amplitudes of L-aspartate iontophoresis-evoked Ca ²⁺ transients were largest in proximal varicosities and decreased with distance from the soma (ΔG/R= 0.0154 ±0.0038, n=17; and ΔG/R= 0.0019 ± 0.0013, n=7; <150 µm and ≥150µm respectively; p< 0.05) (Figure 3A–3C). In addition, there was a trend for faster Ca²⁺ transients in proximal varicosities, however, small amplitude transients, especially in distal axonal segments, prevented accurate fitting of rise times for all data points (Figure 3B and 3D). Together, these results indicate that NMDAR-mediated potentials originating in the dendrite can control Ca²⁺ entry in axon varicosities.

Figure 3. Distance-dependence of iontophoresis-evoked Ca²⁺ transients in axon varicosities.

(A₁) Stellate cell image shows the axonal recording sites for Ca²⁺ measurement and the dendritic location of the iontophoretic pipette used for NMDAR stimulation. Magnified views of axonal varicosities are shown in the boxed insets with the positions of line scans (dashed lines). (A₂) An NMDAR-mediated depolarization (V_rest= −70 mV) resulted from dendritic iontophoresis of L-aspartate. (B) Ca²⁺ transients recorded in proximal (site 1) and distal (site 2) axon varicosities evoked by dendritic iontophoresis. (C) Amplitudes of Ca²⁺ transients recorded in axon varicosities versus the distance of the recording site from the axon hillock. (D) Single exponential fits of the rising phase of Ca²⁺ transients versus distance from the axon hillock.

Dendritic NMDAR-mediated depolarizations open axonal VSCCs

It is likely that NMDAR-mediated Ca²⁺ entry in axon varicosities occurs through VSCCs because this effect depends on membrane potential and VSCCs are enriched in stellate cell presynaptic specializations (Llano et al., 1997; Forti et al., 2000). Consistent with this notion, a cocktail of VSCC blockers (ω-conotoxin MVIIC 1 µM, SNX 0.3 µM,ω-agatoxin IVA 0.2 µM, nimodipine 20 µM and mibefradil 10 µM) greatly reduced the Ca²⁺ transient evoked in axon varicosities by dendritic-targeted iontophoresis of L-aspartate (Figure 4A and 4D). The unblocked component likely reflects toxin-resistant Ca²⁺ channels (Tottene et al. 2000; Metz et al. 2005; Li et al. 2007) given that action potential-evoked Ca²⁺ transients in axon varicosities were blocked to a similar extent (Figure 4B and 4D). In contrast, iontophoresis-evoked Ca²⁺ transients in dendrites were unaltered by VSCC blockers (Figure 4C and 4D) indicating that most Ca²⁺ entry in this compartment was directly through NMDARs.

Figure 4. VSCCs mediate NMDAR-evoked Ca²⁺ entry in axon varicosities.

(A) Dendritic iontophoresis of L-aspartate evoked a Ca²⁺ transient in an axon varicosity that was inhibited by a cocktail of VSCC blockers (ω-conotoxin MVIIC 1 µM, SNX 0.3 µM, ω-agatoxin IVA 0.2 µM, nimodipine 20 µM and mibefradil 10 µM) (B) Similarly, VSCC blockers inhibited an action potential-evoked Ca²⁺ transient in an axon varicosity.(C) Dendritic Ca²⁺ transient evoked by local iontophoresis was unaffected by VSCC blockers. (D) Inhibition of axonal and dendritic Ca²⁺ transients by VSCC antagonists.

NMDARs are not expressed on axons

Our results indicate that NMDAR-mediated Ca²⁺ entry in axon varicosities does not result from direct influx through axonal NMDARs but rather that NMDAR-mediated depolarization, most likely from dendritic receptors, results in Ca²⁺ influx through VSCCs. If NMDARs were expressed on the axon then axonal L-aspartate iontophoresis would increase axonal Ca²⁺ directly through NMDARs. However, in voltage-clamped cells, L-aspartate iontophoresis failed to evoke Ca²⁺ transients in axon varicosities when focally applied near the axon (ΔG/R= 0.0004 ± 0.0002, n=33; Figure 5A and 5B). Iontophoresis of L-aspartate onto dendrites prior to and following axonal application elicited large Ca²⁺ transients in the dendrites (Figure 5A and 5B) and inward currents (−24.3 ± 3.8 pA, n= 10) in somatic electrical recordings confirming the reliability of iontophoresis within single cell experiments. No current was detected when L-aspartate was applied near the axon (−0.2 ± 0.1 pA, n= 24) similar to the results of Clark and Cull-Candy (2002). Fluorescence measurements obtained from axon varicosities following direct iontophoretic stimulation were unchanged by (R)-CPP (20 µM) (ΔG/R 114.8 ± 29.3% of control, n= 15) confirming that axon varicosities lack functional NMDARs. It is unlikely that low-density NMDAR expression mediated Ca²⁺ transients that were below detection threshold given that two-photon fluorescence imaging has the sensitivity to detect Ca²⁺ entry through single channels (Sabatini and Svoboda, 2000; Nimchinsky et al.2004; see Experimental Procedures).

Figure 5. Axonal iontophoresis of L-aspartate does not evoke NMDAR-mediated Ca²⁺ transients.

(A) Direct iontophoretic application of L-aspartate onto an axon varicosity failed to evoke a Ca²⁺ transient. The amplified record is shown in comparison to that without iontophoresis (each record is an average of several trials). Iontophoresis onto a dendrite of the same cell evoked a Ca²⁺ transient confirming the efficacy of L-aspartate application. (B) Peak Ca²⁺ amplitudes recorded in stellate cell processes versus distance from the axon hillock.

To examine a larger axonal area than afforded by line scans, frame scans were used to image small regions of axons (Figure 6A–D). As with line scans, we were unable to detect elevations of Ca²⁺ in axonal varicosities following iontophoretic application of L-aspartate near the axon of interest (ΔG/R= 0.0006 ± 0.0005, n= 14). Action potential-evoked Ca²⁺ transients confirmed that the selected sites were hot spots of Ca²⁺ entry (Figure 6C). In addition, NMDAR-dependent Ca²⁺ transients were not detected in axon segments between varicosities (ΔG/R= 0.0004 ± 0.0006, n= 9; Figure 6C and 6D). These results suggest that the entire axonal arbor is devoid of functional NMDARs. Therefore, dendritic NMDARs must provide the sole source of depolarization responsible for NMDAR-mediated Ca²⁺ entry in axon varicosities.

Figure 6. Stellate cell axons lack NMDARs.

(A) Axonal (left box) and dendritic (right box) locations of L-aspartate iontophoresis on a stellate cell. (B) Frame-scan sequence shows a dendritic Ca²⁺ transient evoked by dendritic iontophoresis. (C) Action potential stimulation (top) reveals hot spots of Ca²⁺ entry in an axon segment. Frame scans taken during iontophoresis (bottom) onto the same axon segment. (D) Quantification of frame scans of dendritic (B) and axonal segments (C- lower varicosity) during L-aspartate iontophoresis.

NMDAR-mediated EPSPs evoke Ca²⁺ entry in axon varicosities

We next determined whether synaptic stimulation could produce NMDAR-mediated Ca²⁺ transients in axon varicosities similar to those evoked by bath applied NMDA and dendritic L-aspartate iontophoresis. Subthreshold NMDAR-mediated EPSPs (17.6 ± 0.7 mV, n= 23) were evoked in current-clamped cells by extracellular stimulation (50 Hz, 80–140 ms) of parallel fibers (PFs) near the dendritic arbor (Figure 7A₁ and 7A₂). PF stimulation elicited large Ca²⁺ transients localized to dendrites near the stimulating pipette (ΔG/R= 0.1994 ± 0.0338, n=8; Figure 7B; Figure S1A). The spatial extent of Ca²⁺ accumulation was larger than expected from single synapses (Soler-Llavian and Sabatini,2006) owing to the stimulation of multiple PFs, the extrasynaptic localization of NMDARs and glutamate spillover (Carter and Regehr, 2000; Clark and Cull-Candy, 2002). Ca²⁺ transients were also evoked in axon varicosities by dendrite-targeted PF stimulation (Figure 7B). These Ca²⁺ transients were smaller than those evoked by dendritic iontophoresis (ΔG/R= 0.0110 ± 0.0029, n=24; and ΔG/R= 0.0027 ± 0.0006, n=37, iontophoretic and synaptic stimulation, respectively; p< 0.05) most likely because iontophoretically evoked potentials were larger in amplitude (p< 0.05). (R)-CPP (20 µM) blocked PF-evoked Ca²⁺ transients in both dendrites and axon varicosities (ΔG/R 2.8 ± 1.9% of control, n= 3; and 1.0 ± 6.3% of control, n= 5; dendrites and axon varicosities, respectively) confirming that the transients were NMDAR-mediated (Figure S1B and S1C). As with iontophoretic stimulation of dendritic NMDARs, somatic voltage clamp greatly reduced the PF-evoked Ca²⁺ transient recorded in axon varicosities but not in the dendrite (ΔG/R 13.8 ± 9.0%, n= 7; and 95.7 ± 1.8% of control amplitude, n= 5, axon varicosity and dendrite, respectively; p< 0.05; Figure 7C₁ and 7C₂) indicating that the passive propagation of somatodendritic depolarization evoked axonal Ca²⁺ entry.

Positioning the stimulating pipette adjacent to axon varicosities so that local PFs were excited failed to elevate Ca²⁺ (Figure 7D and 7E) again suggesting that there are no functional axonal NMDARs. That axonal Ca²⁺ transients evoked by dendritic PF stimulation rely on passive propagation of somatic depolarizations is further supported by the dependence of their amplitudes on the distance from the soma (Figure S2). Nonlinear regression analysis revealed that the amplitude of these axonal Ca²⁺ transients was inversely correlated (r²= 0.89) with distance from the soma with a monoexponential fit (50 µm bins) yielding a decay length constant of 60 µm (Figure S2C). In contrast, the amplitude of action potential-evoked Ca²⁺ transients in axon varicosities varied irrespective of distance (r²= 0.03).

Discussion

Axonal expression of NMDARs has been reported in many regions of the CNS (Berretta and Jones, 1996; Bardoni et al., 2004; Sjostrom et al., 2003; Corlew et al., 2007; Yang et al., 2006; Brasier and Feldman, 2008) including interneurons in the cerebellar cortex (Glitsch and Marty, 1999; Duguid and Smart, 2004; Huang and Bordey, 2004). However, we find that stellate cell axons are devoid of functional NMDARs. This result does not imply that NMDARs are without influence in the axon. NMDAR-mediated depolarizations originating in the dendrite passively propagated into the axon and opened VSCCs in varicosities where these channels are concentrated. This axonal Ca²⁺ transient is the likely basis for the increase in the strength and frequency of evoked and spontaneous release that have been reported previously (Glitsch and Marty, 1999; Duguid and Smart, 2004; Huang and Bordey, 2004).

Localization of NMDARs

Stellate cell dendrites express both AMPA and NMDA-type glutamate receptors. Whereas unitary PF responses are mediated entirely by AMPARs, NMDARs are recruited only under conditions of dense PF activity suggesting that they are located perisynaptically (Carter and Regehr, 2000; Clark and Cull-Candy, 2002). NMDAR immunoreactivity in the molecular layer of cerebellar tissue is consistent with dendriticand axonal expression in interneurons (Petralia et al., 1994). In cultured stellate cells,axon varicosities are immunolabeled with antibodies raised against NR1 and NR2 subunits (Duguid and Smart, 2004) and patches excised from axon terminals have functional NMDAR channels (Fiszman et al., 2005). However, despite large NMDAR-mediated Ca²⁺ influx in dendrites, our search failed to detect functional NMDARs in axons. In addition, preliminary evidence from basket cells suggests that the axons of this cell type are also devoid of functional NMDARs (unpublished observations), indicating a commonality between the major inhibitory interneurons in the cerebellar molecular layer.

Our method of locating NMDARs depends on the generation of Ca²⁺ influx. Ca²⁺ permeability is diminished when heteromeric NMDA channels incorporate NR3 subunits (Matsuda et al., 2002; Sasaki et al., 2002). It is unlikely that NR3 expression in stellate cells obscured observation of NMDAR-dependent Ca²⁺ entry given that the cerebellum lacks NR3A and NR3B subunits at postnatal day (PND)> 14 (Ciabarra et al., 1995; Sucher et al., 1995; Nishi et al., 2001; Fukaya et al., 2005; but see Wee et al. 2008). If axonal NR3 subunit-containing receptors were activated, then the resulting local depolarization should cause a VSCC-mediated Ca²⁺ transient. However, axonal Ca²⁺ transients were not evoked by local stimulation, but rather, only by manipulations that depolarized the somatodendritic compartment. Therefore, at the very least, NMDAR-dependent dendritic depolarization must be a greater influence on axonal Ca²⁺ entry.

Passive Propagation of Dendritic Depolarization

Excitatory potentials from the somatodendritic compartment passively propagate into the axonal arbor of neurons in the hippocampus and neocortex (Alle and Geiger, 2006; Shu et al., 2006), as they do in frog sciatic nerve (Hodgkin, 1937). Although direct electrical measurements did not reveal VSCC openings in dentate granule cell boutons (Alle and Geiger, 2006), an experimentally derived model suggests that subthreshold waveforms should activate a small population of VSCCs (Li et al., 2007). This discrepancy could result from the difficulty in recording VSCC activity mediated by small depolarizations (Awartramani et al., 2005). Our results indicate that NMDAR-mediated potentials passively propagate from the stellate cell dendrite into the axon and that the resulting voltage waveform is sufficient to open VSCCs in varicosities. Passively propagated NMDAR currents originating in the dendrite were also detected in direct electrical recordings from proximal axon varicosities of cerebellar interneurons in culture (Fiszman et al., 2005).

NMDAR depolarization-evoked Ca²⁺ transients in stellate cell axons were largest in proximal axon varicosities and diminished in size with distance from the soma consistent with passive propagation (Rall 1969). The length constant measured for PF-evoked transients was less than 100 µm, a relatively short distance when compared to the length constants of electrotonic propagation in pyramidal and dentate granule cell axons (τ≈ 400 µm; Alle and Gieger, 2006; Shu et al., 2006). In our measurements, VSCC-mediated Ca²⁺ influx is secondary to the underlying depolarization and most likely not linearly related to depolarization. Likewise, somatic voltage control of action potential-evoked Ca²⁺ transients in mossy fiber boutons decayed with a length constant of τ< 200 µm (Scott and Rusakov, 2006). The effectiveness of this transfer is determined by many factors including the electrotonic structure of the cell (Rall, 1969), number and type of Ca²⁺ channels (Li et al., 2007), and the size and time course of the underlying potential (Shu et al., 2006). For example, the length constant of bath NMDA-evoked Ca²⁺ entry extended beyond 200 µm, probably owing to the prolonged duration of the underlying depolarization. The axonal arbors of stellate cells are relatively compact with the most distal collaterals rarely extending more than 400 µm from the soma (Palay and Chan-Palay, 1974). Therefore, the length constant derived for bath NMDA and PF-evoked Ca²⁺ transients suggests that varicosities throughout a large portion of the stellate cell axonal arbor are influenced by passively propagated depolarizations. In neurons with extensive axon arbors, such as dentate granule cells, length constants of similar distance are unlikely to influence much of the axon (Scott et al., 2008).

Ca²⁺ diffusion is tightly regulated in small neuronal structures by endogenous, low-mobility buffers, pumps and the morphological specialization (Sabatini et al. 2002; Goldberg et al., 2003; Soler-Llavian and Sabatini, 2006; Brenowitz and Regehr, 2007). In CA1 pyramidal cell spines, for example, NMDAR-mediated Ca²⁺ entry is so well compartmentalized that it is insufficient to activate SK channels expressed on the same spines (Bloodgood and Sabatini, 2007). Similarly, Ca²⁺ influx through NMDARs located at olfactory bulb dendrodendritic synapses does not normally trigger release of nearby GABA-filled vesicles (Isaacson, 2001; but see Chen et al. 2000; Halabisky et al. 2000). The effects of Ca²⁺ influx through NMDARs, then, are usually restricted to sub-micron domains. The coupling of VSCC activity to passively propagated NMDAR-mediated potentials provides a mechanism to extend the spatial extent of NMDAR Ca²⁺ signaling beyond simple diffusion.

Functional Implications

Neurotransmission is both triggered and modulated by Ca²⁺ signals in presynaptic elements. Although high concentration Ca²⁺ transients are required to trigger exocytosis,low concentrations precondition vesicles for release via vesicle recruitment, priming and sensitization (Zucker and Regehr, 2002). Short-term elevation of Ca2⁺ occurring, for example, following an action potential leads to facilitation of synaptic transmission to subsequent action potentials (Zucker and Regehr, 2002). Axonal Ca²⁺ may also be elevated directly by ligand-gated channels that are expressed on the bouton that either depolarize the bouton, are permeable to Ca²⁺, or both (MacDermott et al., 1999; Engelman and MacDermott, 2004). Our results indicate that dendritic ligand-gated channels can elevate Ca²⁺ in stellate cell axon varicosities by electrotonic propagation thereby providing an alternative method for elevating Ca²⁺ at presynaptic sites of release.

Although we did not determine the relationship between somatodendritic depolarization-mediated Ca²⁺ entry and release, a facilitatory role is clearly implied from previous investigations in the cerebellum (Glitsch and Marty, 1999; Duguid and Smart,2004; Huang and Bordey, 2004). Recent studies have shown that in neocortical pyramidal cells and dentate granule cells of the hippocampus, passive propagation of somatodendritic depolarization into boutons facilitates release (Alle and Geiger, 2006; Shu et al., 2006). Ca²⁺ elevation was found to be only partly responsible (Alle and Geiger, 2006) suggesting that other mechanisms also contribute to this facilitation (Shu et al., 2006; Kole et al., 2007; Scott et al., 2008). In the calyx of Held, slight depolarization of the presynaptic element by direct current injection results in a small increase in Ca²⁺ and a significant enhancement of synaptic transmission (Awartramani et al., 2005), demonstrating the critical relationship between residual Ca²⁺ and the likelihood of vesicle release. Given the large charge transfer mediated by NMDAR synaptic conductances and the long time constant of the compact stellate cells, dendritic depolarization-mediated Ca²⁺ entry could alter the axonal release probability for significant periods.

Activation of presynaptic VSCCs by dendritic depolarization may be a general feature of neurotransmitter release plasticity. Our measurements of dendritic depolarization-evoked Ca²⁺ transients in axon varicosities were limited to NMDAR-mediated activity and current injection through the patch pipette. It is likely that other ypes of ligand-gated ion channels expressed on the dendrite also control axonal VSCC activation. AMPAR activation elicited by synaptic stimulation or exogenously applied agonists alters the strength and frequency of action potential-evoked and spontaneous release in cerebellar interneurons (Bureau and Mulle, 1998; Satake et al., 2000; Satake et al., 2004; Rusakov et al. 2005; and Liu 2007). Although these effects were ascribed to receptors expressed on axons, VSCCs are at least partly responsible for the observed affect (Bureau and Mulle, 1998). It is clear that multiple mechanisms can alter presynaptic release probability including residual Ca²⁺ from previous action potentials, ligand-gated receptors expressed on axons, and subthreshold activation of axonal VSCCs either by direct axonal depolarization or passive propagation from dendritic sites.

Experimental Procedures

Slice Preparation and Electrophysiology

Following anesthesia, cerebellar vermi from Sprague Dawley rats (PND 15–20) were isolated and cut parasagittally (250–300 µm) on a vibroslicer (Leica Instruments, Nussloch, Germany) in an ice-cold solution containing (in mM) 87 NaCl, 25 NaHO₃, 2.5 KCl, 1.25 NaH₂PO₄, 7 MgCl₂, 0.5 CaCl₂, 10 glucose, and 75 sucrose. Slices were then transferred to a chamber containing artificial cerebrospinal solution (ACSF) containing (in mM) 119 NaCl, 26.2 NaHO₃, 2.5 KCl, 1 NaH₂PO₄, 1.3 MgCl₂, 2 CaCl₂, 11 glucose and incubated at 34°C for 30 min. Thereafter, slices were maintained at 22–25°C. During recordings, slices were superfused with ACSF (22–25°C) altered to facilitate the detection of NMDAR-mediated calcium transients (0 mM MgCl₂, 3 mM CaCl₂). NBQX (20 µM) and picrotoxin (100 µM) were added to block AMPA and GABA_A receptors, respectively, D-serine (10 µM) was added to saturate the NMDAR coagonist-binding site, and cyclopiazonic acid (CPA; 20 µM) was used to deplete internal Ca²⁺ stores. All solutions were bubbled with 95% O₂- 5% CO₂. Animal handling and procedures were performed in accordance with Oregon Health and Science University Institutional Animal Care and Use Committee protocols.

Whole-cell recordings were obtained from stellate cells identified with gradient contrast infrared optics (Dodt et al., 2002) and two-photon fluorescence microscopy based on their location in the molecular layer and distinct morphology (Palay and Chan-Palay, 1974). Pipettes with resistances of 3–4 MΩ were used for patching and contained a solution of (in mM) 142 K-gluconate, 2 KCl, 10 HEPES, 4 MgCl₂, 4 NaATP, 0.5 NaGTP, 0.05 Alexa 594 and 0.2 Fluo-5F (Molecular Probes, Eugene, OR). Electrophysiological potentials and currents were recorded with a Multiclamp 700B amplifier (Molecular Devices, Union City, CA). Electrode series resistance was compensated by bridge balance. The analog signals were filtered at 3–10 kHz and digitized at 10–20 kHz. Data were collected using custom software (J.S. Diamond) written in IgorPro (Wavemetrics, Lake Oswego, OR). Action potentials were evoked by current injection (100–175 pA; 5–10 msec) through the patch pipette. Parallel fibers were stimulated with a pipette filled with ACSF (10–100 V, 30–100 µsec). Iontophoretic pipettes (impedance >100 MΩ) were filled with a 1M solution of L-aspartate (pH 7.0).Pipettes were placed near (5–20 µm) the cellular process of interest. A retention current of ≈−1 nA was applied to prevent passive leakage of L-aspartate while a pulse of 100–200 nA (10–50 msec) was used for ejection. TTX (0.5 µM) was bath applied prior to iontophoretic stimulation in order to prevent the initiation of action potentials.

Pharmacological agents including NBQX, CPA, TTX, and NMDA were obtained from Tocris Cookson (Ellisville, MO). Picrotoxin, nimodipine and mibefradil were from Sigma (St. Louis, MO). Peptides including ω-conotoxin MVIIC, SNX, and ω-agatoxin IVA were purchased from Peptides International (Louisville, KY). CytoC or BSA (Sigma, St. Louis, MO) was added (0.25 mg per ml) to solutions containing peptides to prevent surface adsorption in the perfusion system.

Two-photon Imaging

Fluorescence was monitored with a lab-built two-photon laser-scanning microscope using an Olympus (Melville, NY) upright microscope and objective (60X,0.9 NA) and a Ti:sapphire laser (Coherent, Santa Clara, CA) tuned to 810 nm. Green and red fluorescence was simultaneously collected by photomultiplier tubes (H8224, Hamamatsu, Hamamatsu City, Japan) in both epi- and transfluorescence pathways using a 565 nm dichroic and 525/50 and 620/60 bandpass filters (Chroma, Battleboro, VT). Images were acquired using ScanImage software (Pologruto et al., 2004). Lines scans were obtained at 500 Hz while sequential frame scanning occurred at 7.8 Hz. Fluorescence changes were quantified as increases in green fluorescence normalized to red fluorescence (ΔG/R) (Sabatini et al. 2002). Peak amplitude measurements of evoked Ca²⁺ transients (except for action potentials) reflect the average ΔG/R of a 250 msec epoch centered on the peak. For most axonal line scan measurements, alternating trials in which no stimulation occurred were used to subtract background fluorescence levels.

Sensitivity of fluorescence detection

To estimate the Ca²⁺ signal-detection sensitivity of our microscope, we examined VSCC-mediated Ca²⁺ transients evoked by somatic action potentials in spicules (Figure S3A), the spine-like protrusions located on stellate cell dendrites (Palay and Chan-Palay, 1974). We observed extensive variability in the peak response of evoked transients in individual spicules whereas evoked transients in the attached parent dendrite varied less (≥50 trials) (Figure S3B). As with action-potential evoked Ca²⁺ transients in spines of CA1 pyramidal cells (Sabatini and Svoboda, 2000), we were able to detect obvious failures in several spicules. These failures did not result from action potential propagation failure because Ca²⁺ transients were detected in the parent dendrite in the same trial and the amplitude of the evoked dendritic Ca²⁺ transient was independent of the Ca²⁺ transient in the spicule (Figure S3C₁ and S3C₂). The failure rate in spicules ranged from 0.05–0.28 (average 0.16 ± 0.05; n= 4). Given the probabilistic nature of VSCC opening it is likely that Ca²⁺ transients mediated by single channels were also recorded in some trials (Sabatini and Svoboda, 2000).

Although the Ca²⁺ permeability of NMDA and VSCCs is not dissimilar (Hille 2001), NMDA channels produce a much greater influx of total Ca²⁺ following activation due to their long open time (NMDAR open times ~100 msec, Lester et al. 1990; action potential-mediated open time for VSCCs <1 msec, Bischofberger et al. 2002). In estimates from CA1 pyramidal cell spines where single NMDA channel-mediated Ca²⁺ transients have been measured (Nimchinsky et al. 2004), Ca²⁺ influx through NMDARs activated by synaptic transmission is at least ten-fold greater than VSCC-mediated Ca²⁺ influx evoked by an action potential (Sabatini and Svoboda, 2000; Sabatini et al., 2002). Therefore, if we can detect Ca²⁺ influx mediated by single VSCCs in dendrites, we should be able to detect Ca²⁺ influx through single NMDARs in dendrites or axons.

Data Analysis

Excel (Microsoft) and InStat (GraphPad Software) were used for statistical analysis. ANOVA (Bonferroni post hoc procedures) and t-tests were used as appropriate. A value of p< 0.05 was considered significant. Reported values are mean ± SEM. In figures asterisks denote statistical significance.