INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Odorant-specific projections to olfactory bulb glomeruli determine the functional units by which olfactory stimuli are encoded (Buck 1996; Mori 1995; Mori et al. 1999). In the rat, each such unit includes 1,000 odorant-specific receptor neurons, 20-25 mitral cells and the glomerular neuropil to which they both project. Inhibition between these functional units, mediated by the arrangement of mitral cells and bulb interneurons, is thought to enhance or tune the code (Wilson and Leon 1987; Yokoi et al. 1995). Mitral cells within a single glomerular unit influence mitral cells of adjacent glomerular units by an extensive system of dendrodendritic synapses that are distributed along mitral cell lateral dendrites. Dendritic propagation of action potentials in mitral cells is necessary to drive transmission at these synapses, thus inhibition depends on action potential backpropagation. Action potentials propagate in apical dendrites without changes in waveform (Bischofberger and Jonas 1997; Chen et al. 1997). In lateral dendrites, attenuation of action potentials has been reported (Lowe 2002; Margrie et al. 2001), although others (Xiong and Chen 2002) have argued that action potentials propagate with full somatic fidelity. Regulation of action potential propagation in lateral dendrites could provide insight into the role of inhibition in odorant coding.

The efficacy of action potential propagation in dendrites varies widely between cell types. In substantia nigra neurons and hippocampal oriens-alveus interneurons, axo-somatically initiated action potentials are conserved throughout the dendritic arbor (Hausser et al. 1995; Martina et al. 2000). However, action potentials decrement along dendrites in cerebellar Purkinje cells, CA1 hippocampal pyramidal cells, and cortical pyramidal cells (Llinas and Sugimori 1980; Spruston et al. 1995; Stuart and Hausser 1994; Stuart and Sakmann 1994). These differences presumably reflect the relative densities of voltage-gated ion channels in dendrites (Hausser et al. 2000). For example, in hippocampal CA1 apical dendrites, the density of TTX-sensitive sodium channels appears uniform (Magee and Johnston 1995), but the density of A-type potassium channels (I_A) steadily increases with distance from the soma (Hoffman et al. 1997). I_A thus can regulate action potential amplitudes in CA1 neurons by shunting voltage-dependent sodium current in distal dendrites (Hoffman et al. 1997). We examined action potential propagation in mitral cell dendrites in olfactory bulb slices. Action potentials were attenuated in lateral dendrites based on somatic and dendritic paired recordings as well as imaging of dendritic calcium transients. The attenuation was prevented by block of A-type potassium channels. These results suggest that the extent of lateral inhibition could be shaped by the incomplete propagation of full amplitude action potentials.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We prepared horizontal slices of main olfactory bulb (300 µM) from Sprague-Dawley rat pups (PND 11-15). Briefly, slices were cut and incubated in an oxygenated solution containing (in mM) 125 NaCl, 25 NaHCO₃, 25 glucose, 2.5 KCl, 1.25 NaH₂PO₄, 2 MgCl₂, and 1 CaCl₂ (pH 7.3). For whole cell current- and voltage-clamp recordings from mitral cells, we used the same solution except MgCl₂ was 1 mM and CaCl₂ was 2 mM. The bath temperature was maintained at 33-35°C.

Infrared/DIC optics were used to identify mitral cell bodies and dendrites. Patch pipettes (soma: 4-5 M; dendrites: 10-12 M) contained (in mM) 125 K-gluconate, 2 MgCl₂, 2 CaCl₂, 10 EGTA, 2 Na-ATP, 0.5 Na-GTP, and 10 HEPES (pH 7.3). For some experiments, synaptic currents were blocked with 100 µM AP5, 10 µM (2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzoquinoxaline-7-sulfonamide) (NBQX), and 10 µM GABAzine (Tocris Cookson, Bristol, UK). To counteract recording errors, we used capacitance neutralization and bridge balance compensation. If V_m was more positive than -50 mV or bridge balance compensation was >15 M (soma) and 40 M (dendrites), we terminated the experiment. Mitral cells were held at a slightly hyperpolarized membrane potential (ca. -60 to -65 mV) to prevent random spiking. Potassium currents were measured in nucleated patches with pipettes (4-5 M) that contained (in mM) 140 KMeSO₄, 7 NaCl, 5 EGTA, 2 Mg-ATP, and 10 HEPES (pH 7.3). TTX (0.5 mM) eliminated sodium conductances. On-line subtraction protocols removed linear leak and capacitance currents; series resistance compensation was used to minimize steady-state voltage errors (>80%). For calcium imaging, mitral cells were filled with the low-affinity (Kd = 20 µM) calcium indicator Oregon Green BAPTA-5N (100 µM, Molecular Probes, Eugene, OR) via the recording pipette. The pipette also contained (in mM) 135 K-gluconate, 6 KCl, 4 MgATP, 0.5 Na-GTP, and 10 HEPES (pH 7.3). After allowing the dye to equilibrate for >20 min, action-potential-evoked changes in fluorescence were recorded in the presence of synaptic receptor blockers (AP5, NBQX, and GABAzine). The fluorescence (F/F) evoked by short train of three action potentials in lateral dendrites (F/F = 113.2 ± 15.5%.) was comparable to three times the fluorescence evoked by the first action potential (F/F = 122.9 ± 21.5%, n = 6), confirming that the dye was not saturated.

Current and voltage signals were recorded with Axopatch 1B and 1D amplifiers (Axon Instruments, Foster City, CA). Signals were filtered with the built-in 4-pole Bessel filter (2-10 kHz) and digitized (20-50 kHz). Data were acquired and analyzed using Axon Instruments software (pClamp 8.2 and AXOGRAPH 4.5). A confocal microscope (Odyssey XL, Noran Instruments, Middleton, WI), equipped with ×40 and ×63 objectives, was used for calcium imaging. Fluorescence was displayed as F/F. We determined statistical significance using standard Student's t-test or repeated-measures ANOVA as appropriate (Microsoft Excel, Redmond, WA).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Action potential backpropagation in mitral cell dendrites

We used paired somatic and dendritic current-clamp recordings to determine the origin and extent of action potential propagation in mitral cell apical and lateral dendrites. Depolarizing current injections (range: 200-1,000 pA, 100 ms) in either the soma or dendrites were used to generate action potentials. Regardless of the current injection location, action potentials were initiated in the soma/axon hillock and backpropagated into the dendrites (Fig. 1A). Consistent with previous reports (Bischofberger and Jonas 1997; Chen et al. 1997; Margrie et al. 2001), apical dendritic and somatic action potential amplitudes were equivalent (dendritic amplitude was 97.0 ± 0.7% of somatic amplitude, n = 9; Fig. 1A). However, we found that action potentials recorded in proximal lateral dendritic segments (90-225 µm) were significantly attenuated (Fig. 1A; 76.4 ± 5.1% of somatic amplitude, n = 12).

View larger version (19K): [in this window] [in a new window]   Fig. 1. Action potential attenuation in mitral cell lateral dendrites. A: paired somatic and dendritic whole cell current-clamp recordings in apical (top) or lateral (bottom) dendrites show that action potentials generated by somatic (left) or dendritic (right) current injection were attenuated in lateral dendrites but not in apical dendrites. The action potentials enclosed with boxes are enlarged below (1-3). B: dendritic action potential amplitudes were plotted as a percentage of the somatic action potential for 27 mitral cell paired recordings. Action potentials were attenuated in lateral dendrites (, >100 µm from soma) but not in apical dendrites ().

The attenuation was not due to granule cell-mediated synaptic inhibition because action potential amplitudes were unchanged when synaptic conductances were blocked with AP5 (100 µM), NBQX (10 µM), and GABAzine (10 µM; Fig. 2, A and B). Furthermore, in control conditions, somatic and dendritic action potential amplitudes were unaffected by repetitive firing (range: 10-80 Hz; Figs. 1A and 2A), activity thought sufficient for recruitment of recurrent inhibition. Xiong and Chen (2002) observed a similar lack of inhibitory control of action potential propagation under physiological conditions. They reasoned that single-cell stimulation was insufficient to overcome magnesium block of N-methyl-D-aspartate (NMDA) receptors in granule cells (Isaacson and Strowbridge 1998; Schoppa et al. 1998). However, shunting of lateral dendritic action potentials has been observed with focal GABA applications (Lowe 2002) and by focally evoked granule cell inhibitory postsynaptic potentials (IPSPs) (Xiong and Chen 2002).

View larger version (25K): [in this window] [in a new window]   Fig. 2. Attenuation was not dependent on synaptic activity. A: bath application of synaptic receptor antagonists (right, 100 µM AP5, 10 µM NBQX, and 10 µM GABAzine) did not affect action potential attenuation in lateral dendrites (top) or in the soma (bottom). The action potential enclosed in boxes are enlarged below (1 and 2). B: action potential amplitudes were plotted as a percentage of the somatic action potential. The amplitude of the 4th action potential in a train was chosen for comparison because feedback inhibition should be activated by repetitive stimuli.

The reduced amplitude of action potentials in proximal lateral dendrites raised the possibility that action potentials might fail in more distal segments. Direct recordings were not possible along the full extent of lateral dendrites (1,000 µm) (Orona et al. 1984) due to pronounced dendritic tapering. Thus we used optical imaging to assess penetration of voltage signals in distal dendritic segments (>200 µm). Mitral cells were loaded individually with the calcium indicator Oregon Green BAPTA-5N (100 µM). Experiments were conducted in the absence of synaptic activity (100 µM AP5, 10 µM NBQX, and 10 µM GABAzine). Single, somatically elicited action potentials (depolarizing current injection: 250-500 pA, 5.0-7.5 ms) evoked rapidly activating calcium transients in dendrites (peak F/F range: 8-48%, time to peak: 2-4 ms, and half-width range: 60-220 ms). The size and shape of the evoked response was stable during 30-60 min of recording, suggesting that the dye concentration was relatively constant. During brief action potential trains, calcium transients added linearly with spike number, indicating that the dye was not saturated (see METHODS).

Dendritic calcium transients could reflect the active propagation of action potentials or the passive propagation of somatic depolarizations. To resolve this question, we used an action potential train waveform (25 Hz, 120 ms) as a somatic voltage-clamp command (Fig. 3A1). Robust calcium transients were generated in the initial dendritic segment (3A2, left) and in distal dendritic segments (Fig. 4A2, right) when TTX-sensitive sodium conductances were left intact. However, calcium transients were diminished in the initial segment and completely absent in distal dendritic segments (3, A2 and B) when sodium-dependent action potential propagation was blocked with TTX (0.5 µM). Thus dendritic calcium transients reflect local calcium entry mediated by action potentials. The TTX block of calcium transients also indicates that mitral cell dendrites lack regenerative calcium spikes (Charpak et al. 2001; Margrie et al. 2001; Xiong and Chen 2002).

View larger version (33K): [in this window] [in a new window]   Fig. 3. Dendritic calcium transients require backpropagating action potentials. A1: we used a 90-ms train of action potentials (33 Hz) recorded in current-clamp as a voltage-clamp command. Dendritic calcium transients evoked by this waveform were recorded in either the initial dendritic segment (0-25 µm) or in a more distal dendritic location (200 ± 25 µm). A2: TTX reduced the calcium transients in the initial dendritic segment (0 µm). In a different cell, TTX eliminated calcium transients at distal dendritic locations (200 µm). B: the nearly complete block of calcium transients in distal segments indicates calcium transients can be used as an indication of backpropagating action potentials. Calcium transients are plotted as F/F.

View larger version (61K): [in this window] [in a new window]   Fig. 4. Action potential-evoked calcium transients in distal lateral dendrites. A: single action potentials evoked a calcium transient (inset) in a lateral dendrite distal segment, 625 µm from the soma. Note, the example image is an ensemble composed of several images acquired at low magnification following the experiment. White box denotes imaging area. B: action-potential-evoked calcium transients had similar amplitudes at 25 and 200 µm in this cell. The calcium transients for each location were stable; compare transients at 25 µm for the beginning (top) and end (bottom) of the recording period. C: data from all cells were binned based on location. The peak amplitudes in each segment were not significantly different.

Action-potential-evoked calcium transients were generated in all dendritic segments monitored (Fig. 4, A and C), suggesting that action potentials penetrate into distal segments. We compared the size of calcium transients evoked in the initial lateral dendritic segment and at distances of 200 µm in the same cell. Calcium transients recorded from both sites were similar in size (F/F = 22.0 ± 3.3 at 0-25 µm; F/F = 24.1 ± 2.4 at 175-225 µm, respectively, n = 4; Fig. 4B). Consistent with this observation, attenuation of calcium transients was not apparent when data were pooled from all sites (Fig. 4C). The lack of calcium transient attenuation does not exclude attenuation of the underlying voltage signal due to differences in the membrane surface-to-volume ratio. Calcium transients in thin dendrites can appear larger than in thick dendrites as demonstrated in neocortical pyramidal cells (Holthoff et al. 2002). The dendritic tapering in mitral cell lateral dendrites (Mori et al. 1983) would tend to underestimate the calcium signal in proximal dendritic sites as compared with distal dendritic sites. In fact, the low surface-to-volume ratio of the soma prevented detection of single action-potential-mediated calcium signals (data not shown).

A-type potassium conductances contribute to action potential attenuation

In hippocampal CA1 apical dendrites, a high density of A-type potassium channels (I_A) underlies action potential attenuation (Hoffman et al. 1997). We examined whether a similar mechanism affects action potential amplitude in mitral cell dendrites. A-currents are present in mitral cells (Wang et al. 1996) as demonstrated by 4-AP (6-10 mM) block of a transient outward current in nucleated outside-out patches (n = 6; Fig. 5A). In paired somatic and dendritic recordings, 4-AP (10 mM) increased the amplitude and half-width of action potentials in the soma, apical, and lateral dendrites (Fig. 5, B and C1-3). The increase in peak amplitude was much greater in lateral dendrites (>100 µm from soma), indicating that A-channels (I_A) contribute to action potential attenuation in lateral dendrites (Fig. 5C3). 4-AP produced similar increases in the action potential half-width in somatic, apical and lateral dendritic sites (Fig. 5C2). In addition, 4-AP decreased the firing rate and the time to initiate the first action potential (Fig. 5B). In contrast to other cell types (Hoffman et al. 1997), 4-AP did not shift the initiation site of action potentials to dendrites with large dendritic current injections (1,000 pA).

View larger version (22K): [in this window] [in a new window]   Fig. 5. 4-Aminopyridine (4-AP) prevented action potential attenuation in lateral dendrites. A: potassium currents were evoked in outside-out nucleated patches. Patches were held at -50 mV, stepped briefly (250 ms) to -110 mV to remove resting inactivation and then stepped to a test potential of +60 mV; sodium channels were blocked TTX (0.5 µM). A2: bath application of 4-AP (6 mM) blocked a rapidly activating and deactivating component, consistent with an A-type potassium current (see subtracted trace, right). B: in a paired recordings of lateral dendrites (top) and soma (bottom), 4-AP (10 mM) preferentially increased the amplitude of the dendritic action potential. Action potentials in boxes are enlarged below (1 and 2, respectively). 4-AP also slowed the firing frequency. C, 1 and 2: 4-AP increased action potential amplitudes in all segments, but the largest increases were in lateral dendrites. Changes in the half-width were not dependent on location. C3: 4-AP markedly reduced action potential attenuation in lateral dendrites.

Lower concentrations of 4-AP (<100 µM) inhibit D-type potassium currents (Storm 1988) that, like A currents, are rapidly activating subthreshold currents (Storm 1988) and thus could contribute to action potential attenuation. However, the specific D-current antagonist -dendrotoxin (1 µM) did not affect action potential attenuation in lateral dendrites. In direct dendritic recordings, -dendrotoxin had no significant effect on the action potential amplitude and increased the half-width marginally (102.0 ± 1.1 and 111.9 ± 5.9% of control, respectively; n = 3). Thus the D-current does not have a role in attenuating action potentials in mitral cell lateral dendrites.

We used optical imaging to assess whether A-current block altered action potential propagation in distal dendritic segments. 4-AP (10 mM) increased the peak calcium transient at all dendritic sites (Fig. 6A). The magnitude of the increase varied with location. The increase at 200 µm on lateral dendrites was greater than increases on the initial dendritic segment or sites on apical dendrites. However, there was no significant difference between 200 µm and sites more distal (Fig. 6B), suggesting that action potentials reach an attenuated plateau potential in control conditions. 4-AP also increased the half-width of calcium transients at all sites (Fig. 6C). Differences between sites were not significant.

View larger version (33K): [in this window] [in a new window]   Fig. 6. 4-AP increased peak calcium transient in lateral dendrites. A: the amplitude of calcium transients at varying distances along their lateral dendrite were compared in the presence and absence of 4-AP (10 mM). The data from each location are from separate cells. Calcium transients evoked by single somatically elicited action potentials were increased by 4-AP. These increases were more prominent in more distal locations (190 and 625 µm). B1: the largest increases in calcium transients occurred in distal segments of the lateral dendrites (200 µm). B2: the effects of 4-AP on the shape of the calcium transient, as measured by half-width, were not location-dependent.

In proximal segments of lateral dendrites, 4-AP increased the action potential amplitude and half-width by 26 and 60%, respectively (see Fig. 5). To determine whether the 4-AP-induced changes in action potential shape could account for changes in the calcium transient, we used action potential waveforms in voltage clamped lateral dendrites (>100 µm; Fig. 7A). The amplitude or half-width of the action potential waveform were modified to match changes induced by 4-AP. Calcium transients were measured in close proximity to the pipette (20 µm) to ensure adequate voltage clamp and in the presence of TTX (0.5 µM) to block active conductances. Action potential waveforms induced robust calcium transients that were comparable to those evoked by unclamped spikes (compare Fig. 7B, 3 and 6). Small increases in the half-width of the calcium transient most likely reflect the slow accumulation and dissipation of calcium in the imaged region due to space clamp issues associated with loss of active conductances. Increasing the action potential waveform amplitude or half-width increased the calcium transient (Fig. 7, B and C). These results indicate that the A-current reduces calcium entry in mitral cell dendrite by controlling the peak and duration of the action potential.

View larger version (57K): [in this window] [in a new window]   Fig. 7. Action potential amplitude and duration can alter the peak calcium transients in lateral dendrites. A, 1 and 2: in the presence of TTX, simulated action potential voltage-clamp commands were applied (right) during direct dendritic recordings, 100 µm from soma. The command protocol was varied to match the effects of 4-AP on action potentials. B: increasing the test action potential amplitude by 126% or half-width by 160% increased the peak calcium transient peak and its half-width. Data are from the same cell. C: histograms summarizing the effects of simulated action potential amplitude and half-width changes on evoked calcium transients.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Attenuation of backpropagating action potentials in lateral dendrites

Our results indicate that block of A-currents in lateral dendrites prevents attenuation of backpropagating action potentials. We used paired somatic and dendritic recordings to measure action potential backpropagation. We found that action potentials were attenuated in proximal lateral dendrites. Lowe (2002) and Margrie et al. (2001), also using paired recordings, reported action potential attenuation in lateral dendrites. Xiong and Chen (2002) using a single dendritic recording pipette, observed smaller action potentials in lateral dendrites, but they attributed it to series resistance errors. Series resistance artifacts cannot explain our results. Our recording criteria was similar to Xiong and Chen (2002); dendritic pipette resistances were 10-12 M and dendritic recordings with a series resistance >40 M were excluded. For series resistance in the range considered acceptable in our experiments, some low-resistance lateral dendritic recordings showed more attenuation than cells with relatively higher series resistance. Furthermore, 4-AP reversed the attenuation and correspondingly increased the peak calcium transients in distal dendritic segments.

Although action potentials were attenuated in lateral dendrites, attenuation did not lead to propagation failure in distal segments. We used single somatically evoked calcium signals as a measure of propagation in distal lateral dendrites. Margrie et al. (2001) and Xiong and Chen (2002) performed similar experiments. Margrie et al. (2001) concluded that action potentials failed to propagate in distal segments, whereas Xiong and Chen (2002) concluded that action potentials traverse the entire dendrite. Our results suggest that action potential amplitudes reach a plateau level in distal dendrites, similar to that described in CA1 (Spruston et al. 1995) and neocortical pyramidal cells (Stuart and Sakmann 1994). This result is particularly important for mitral cells because glutamate released along the lateral dendrites drives dendrodendritic inhibition. The amplitude and duration of the voltage signals will determine the extent of transmitter release. Because calcium transients in lateral dendrites required TTX-sensitive action potentials, regulation of backpropagating action potentials may be an important mechanism in shaping the pool of granule cells activated by lateral dendrites.

A-channels in lateral dendrites

Dendritic excitability depends critically on the density and distribution of voltage-dependent ion channels (Hausser et al. 2000). This is consistent with the role of dendrites in synaptic integration rather than point-to-point transmission of action potentials. Because mitral cell dendrites lack calcium-dependent regenerative potentials (Charpak et al. 2001; Margrie et al. 2001; Xiong and Chen 2002), the mechanisms governing sodium-dependent action potential backpropagation provide functional control of dendritic excitability. In hippocampal CA1 pyramidal cells, a high density of dendritic A-channels underlies action potential attenuation (Hoffman et al. 1997). Block of A-channels also prevented action potential attenuation in our experiments. We did not make an estimate of A-channel current density in lateral dendrites, thus our results could be explained either by a high density of dendritic A-channels or a nonuniform distribution of dendritic sodium channels. It is likely that sodium/potassium channel ratio is lower in mitral cell lateral dendrites than in the soma/axon initial segment because action potentials were always initiated at the somatic electrode even with large current injections in dendrites in control conditions as well as in the presence of 4-AP.

A-channels are good candidates for control of dendritic excitability because they operate at subthreshold voltages and their activity can be modulated (Hoffman and Johnston 1998; Mayer and Sugiyama 1988; Villarroel and Schwarz 1996). For example, decreasing the availability of A-channels due to membrane depolarization could increase dendritic action potential amplitudes. The subthreshold depolarization driven by NMDA autoreceptors (Isaacson 1999; Schoppa and Westbrook 2001) could serve such a role. We did not observe changes in action potential amplitude during repetitive firing (<100 Hz), suggesting that A-channels did not accumulate inactivation. However, this behavior is likely to be strongly influenced by membrane potential. Our results were obtained at resting membrane potentials of approximately equal to 65 mV, whereas Margrie et al. (2001) reported a gradual increase in dendritic action potential amplitudes during spike trains for cells with more depolarized membrane potentials.

Functional implications

The dendritic action potential waveform provides the stimulus necessary for release of glutamate at dendrodendritic synapses. The resulting recurrent and lateral inhibition from synaptically activated granule cells could thus be determined by the shape of the action potential. In our experiments, calcium transients in proximal dendritic segments were sensitive to the amplitude and duration of action potential waveforms. Although voltage-dependent calcium channels are responsible for dendritic glutamate release (Isaacson and Strowbridge 1998; Xiong and Chen 2002), the calcium sensitivity of glutamate release from this unusual synapse is not known. However, it seems likely that modest changes in action potential amplitude might not markedly affect transmitter release, at least for proximal dendritic segments. For example, 4-AP did not increase the size of granule cell excitatory postsynaptic potentials evoked by focal mitral cell stimulation although the action potential was broadened (Schoppa and Westbrook 1999). However, given the small dendritic arbor of granule cells (100 µm) (Mori et al. 1983), these experiments were most likely examining release from proximal lateral dendrites and thus may not accurately reflect the behavior at more distal synapses.

Although the baseline synaptic activity in our experiments did not influence backpropagating action potentials, Lowe (2002) recently demonstrated that exogenous GABA application to lateral dendrites could further attenuate backpropagating action potentials. Focal stimulation of granules cells can also lead to local action potential failure (Xiong and Chen 2002). Thus the extent and pattern of incoming synaptic inhibition could strongly influence the penetration of action potential into distal dendritic segments. Because attenuated spikes in distal segments would presumably be more influenced by such inhibition, it raises the possibility that proximal and distal segments can be independently regulated. One might expect that this would alter the extent of lateral inhibition.