Action potential propagation in mitral cell lateral dendrites is decremental and controls recurrent and lateral inhibition in the mammalian olfactory bulb

In Vivo and in Vitro Electrophysiology.

Horizontal olfactory bulb slices (300 μm thick) were prepared from anesthetized mice and rats (Ketamine 0.1%/Xylazine 0.1%; P18-P40) and maintained in Ringer's solution containing (in mM): NaCl 125, KCl 2.5, NaHCO₃ 25, NaH₂PO₄ 1.25, MgCl₂ 1, glucose 25, and CaCl₂ 2. In vivo experiments were carried out on freely breathing mice (P21-P35) anesthetized with urethane (1.2 g/kg, i.p.). Access to the olfactory bulb was gained via a hole drilled through the cranium 1 mm anterior to the rhinal fissure and 0.75 mm lateral to the midline. A custom-built chamber was sealed to the cranium to enable superperfusion of solutions (37°C). A custom-built olfactometer (24) was used to present odors to the nares. After the experiments, the brains were fixed, coronally sectioned (250–300 μm), and processed (25) to reveal cell morphology (Fig. 1A). Because of some surface damage from repeated electrode penetrations, the apical tufts of mitral cells sometimes were not recovered. All animal care was performed in accordance with the guidelines of the Max Planck Society.

For both in vitro and in vivo recordings, patch pipettes (3–6 MΩ, 8–15 MΩ for dendritic recordings) were pulled from borosilicate glass tubing and filled with solutions containing (in mM): potassium-gluconate 130, Hepes 10, MgCl₂ 2, MgATP 2, Na₂ATP 2, GTP 0.3, NaCl 4, and biocytin 1–4%, at pH 7.3. In vitro whole-cell recordings were obtained by using standard techniques (26). All data have been corrected for a +7 mV junction potential. Because of the small diameter of the lateral dendrites in mice, dendritic recordings were carried out on mitral cells in rat olfactory bulb slices. Otherwise, all experiments were carried out on mice. To obtain whole-cell recordings in vivo, we used conventional voltage clamp methods (27). Electrodes were advanced into the brain in 2- to 3-μm steps while repeatedly applying voltage steps (6–10 mV, 10 ms duration). Positive pressure (25–50 millibars) was applied to the pipette while it was being advanced. When electrode resistance increased, the positive pressure was removed and, with slight negative pressure, a seal resistance of 1 GΩ or greater was achieved. The whole-cell configuration was achieved by applying a slow ramp of negative pressure. In vivo, mitral cells were identified by their distance from the surface (180–250 μm), input resistance (<160 MΩ), resting membrane potential (−51 to −62 mV), and the appearance of a recurrent IPSP following APs. In vivo and in vitro voltage recordings were amplified with the Axoclamp-2B amplifier in bridge mode (Axon Instruments, Foster City, CA), filtered at 3–10 kHz, and digitized at 5–20 kHz (ITC-16; Instrutech, Mineola, NY) with custom software running under IGOR (WaveMetrics, Lake Oswego, OR) or PULSE (HEKA Electronics, Lambrecht/Pfalz, Germany). Series resistances were from 10–70 MΩ (in vivo), 5–25 MΩ (in vitro somatic), and 15–50 MΩ (in vitro dendritic).

Ca²⁺ Fluorescence Imaging.

Neurons were filled via the recording pipette with either 50–200 μM fura-2 or 50–100 μM calcium orange (Molecular Probes). Fluorescence was imaged with ×60 and ×20 objectives (Olympus, Hamburg) and a fixed-stage microscope (Zeiss). A monochromatic light source was used for fluorescence excitation (TILL Phototonics, Planegg, Germany), and images [full frames and regions of interest (ROI)] were acquired by a back-illuminated frame-transfer charge-coupled device camera (Princeton Instruments, Trenton, NJ). When imaging fura-2 fluorescence, Ca²⁺ concentrations were calculated by using the standard ratioing equations (28, 29). Data acquisition rates ranged from 10 ms per point for ROI to 67 ms per point for full-frame images. Agents used were 2,3-dihydroxy-6-nitro-7-sulfamoylbenzo[f]quinoxaline (NBQX), 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), D-2-amino-5-phosphonovaleric acid (APV), bicuculline, biocytin, n-amyl acetate (AA) (stock: 50 or 100% diluted in mineral oil; Sigma). All data are reported as mean ± SEM.

Odor-Evoked Recurrent Inhibition.

Odor-evoked AP firing was examined by using in vivo whole-cell recordings from mitral cells (Fig. 1A). Odor-evoked responses fell into two general categories (30), with one fraction of cells showing increases in AP rate (Fig. 1B, cells 1 and 2) and other cells showing inhibition by hyperpolarization (Fig. 1C, cells 3–5).

The first type of odor-evoked mitral cell response was characterized by stimulus-locked AP rates up to 250 s⁻¹ (Fig. 1B, cell 2). Mitral cells firing at high rates at the onset of odor presentation often showed a decrease in AP frequency during the stimulus and a hyperpolarization following its termination (Fig. 1B, cells 1 and 2, arrows). We hypothesized that both the decrease in AP rate during stimulus presentation and the after-stimulus hyperpolarization were caused by odor-evoked recurrent IPSPs. To test this view more directly, we examined the pharmacological properties of the odor-evoked IPSP and current injection-evoked IPSPs (i.e., the isolated recurrent IPSP).

Odor-evoked AP patterns were modulated by bicuculline. Fig. 2Ai shows an odor-evoked response that exhibited a prominent hyperpolarization (≈5 mV) after termination of the odor exposure. Superfusion of bicuculline over the MOB reduced the odor-evoked hyperpolarization, resulting in APs that outlasted the odor stimulus (Fig. 2Ai). Bicuculline also abolished the increase in interspike-interval (ISI) that was observed during odor presentation. Under control conditions, the ISI of odor-evoked responses increased ≈2-fold (n = 3) from 77 ± 7 ms (first 15 APs after burst) to 128 ± 16 ms (last 15 APs; Fig. 2B). In bicuculline, the ISI was unchanged during the odor response [49 ± 4 ms (first 15 APs after burst) vs. 39 ± 5 ms (last 15 APs); Fig. 2B].

Depolarizing current injection (100- to 150-ms steps) resulted in trains of APs that were followed by a prolonged hyperpolarization (Fig. 2Aii) comparable to recurrent IPSPs as described (10, 11, 17, 18, 31) in vitro (decay time constant: 459 ± 87 ms, n = 13). This hyperpolarization was blocked also by bicuculline (5.7 ± 0.6 mV vs. 2.2 ± 0.4 mV control and bicuculline, respectively; n = 11, P < 0.001; Fig. 2C), indicating that it represents the isolated, recurrent IPSP.

Therefore, the recurrent IPSP shapes odor-evoked mitral cell AP patterns in two ways. First, during prolonged odor presentation, the IPSP reduces the rate of mitral cell firing. Secondly, after the cessation of a sensory input, the hyperpolarization caused by the recurrent IPSP rapidly and precisely terminates mitral cell firing.

Mechanisms of Recurrent Inhibition.

The in vivo experiments suggest that the amplitude of the recurrent IPSP depends on the rate of mitral cell firing. Consistent with this idea, when isolated recurrent IPSPs were elicited by bursts of APs evoked by somatic current injection, the amplitude of the IPSP depended on the number of APs in the burst [−3.1 ± 0.5 mV (4 AP), −6.4 ± 0.5 mV (10 AP), n = 17; Fig. 3 A and B], suggesting that APs at higher than 10 APs s⁻¹ effectively evoke recurrent IPSPs. A similar dependence of recurrent-IPSP amplitude on AP number was observed in vitro [−1.4 ± 0.4 mV (4 AP), −3.3 ± 0.3 mV (10 AP), n = 30; P < 0.001; Fig. 3B], although the amplitude of the recurrent IPSP in vitro was reduced. In vivo, bursts of spontaneously generated APs, but usually not single APs, also produced prolonged hyperpolarizations (Fig. 3C). Thus, odor-evoked, current-evoked, and spontaneous recurrent IPSPs in vivo increased with AP frequency. The difference in the recurrent IPSP amplitude between the in vivo and in vitro preparations may reflect the elimination of inhibitory synaptic connections during the slicing procedure or a difference in the ionic concentrations in the extracellular milieu and suggest that in vitro experiments underestimate the contribution of the recurrent IPSP in vivo.

Because sampling of natural odors occurs over many seconds, for dendrodendritic inhibition to contribute to odor processing, it must remain effective during long periods of mitral cell activity. Therefore, we next evoked mitral cell activity for periods of 0.1 to 5 s and compared the amplitude of the recurrent IPSP evoked as a function of stimulus duration. The amplitude of the recurrent IPSP was unaffected by increasing the duration of the current step while maintaining a constant AP frequency of ≈50 Hz (100 ms with 1-, 2-, 4-, and 5-s pulses; F_(4,15) = 0.07, P = 0.989, one-way ANOVA; Fig. 4A). This observation indicates that the recurrent IPSP is not attenuated by long periods of high frequency firing and maintains its effectiveness during long periods of sensory stimulation (e.g., Fig. 1B, cell 2 and Fig. 2A).

When the recurrent IPSP was abolished by application of bicuculline (20–40 μM), we found that current steps (100 ms) elicited more APs than in control conditions (e.g., a 200-pA step: 2.3 ± 0.2 APs in control vs. 8.0 ± 1.8 APs in bicuculline; Fig. 4B, n = 8 cells). In the presence of bicuculline, current-evoked APs were followed by a depolarization (2.1 ± 0.9 mV amplitude at peak evoked by 10 AP; n = 5; Fig. 4C Left) that in some cases produced AP firing (Fig. 4C Left). This prolonged depolarization was blocked by glutamate receptor antagonists (APV, 25 μM and NBQX or CNQX, 10 μM), suggesting that AP-evoked release of glutamate from a mitral cell evokes a recurrent excitatory postsynaptic potential (EPSP) mediated by both N-methyl-D-aspartate (NMDA) and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (32–37; Fig. 4C Right). By subtracting the averaged response recorded in the presence of bicuculline from the average control response, we found that the recurrent EPSP in vitro reduces the bicuculline-sensitive component by ≈40% (isolated IPSP = −5.7 ± 1.5 mV, isolated EPSP = 2.4 ± 0.9 mV, for 7 APs, n = 4 cells; Fig. 4C Right). These data show that recurrent inhibition reduces the rate of mitral cell firing and is masked in part by glutamatergic self excitation.

Odor Evoked Lateral Inhibition.

The second type of response of mitral cells to odorants (Fig. 1C) was characterized by a hyperpolarization during the initial 1–2 s of odor presentation (2, 30, 38–39). This hyperpolarization attenuated with time and often was accompanied by an increase in membrane noise (e.g., Fig. 1C, cells 3–5 and Fig. 5 Top). Cells that exhibited this initial hyperpolarization during stimulus presentation under control conditions fired APs in response to the same stimulus after superfusion of bicuculline over the surface of the MOB (n = 3; Fig. 5 Middle). Because the hyperpolarization was not preceded by AP firing, these data show that GABA_A receptor-mediated lateral inhibition prevents odor-evoked firing in mitral cells. Moreover, the attenuation of the hyperpolarization during the odor presentation suggests that the amplitude of lateral inhibition decreases as a function of stimulus duration. Thus, lateral inhibition can prevent or delay firing in mitral cells receiving active ORN input.

Decremental AP Propagation and Ca²⁺ Transients in Lateral Dendrites.

Granule cell-mediated dendrodendritic inhibition underlies recurrent and lateral inhibition between mitral cells and depends on [Ca²⁺]-dependent glutamate release from mitral cell lateral dendrites. We measured AP propagation and the AP-evoked increase of [Ca²⁺] in these dendrites to determine how patterns comparable to odor-evoked mitral cell activity determine the propagation of APs in lateral dendrites and the concomitant transient rise in dendritic [Ca²⁺]. Simultaneous whole-cell voltage recordings from mitral cell somata and lateral dendrites showed that single APs decreased in amplitude by 22% per 100 μm (linear fit to data R² = 0.52, n = 13; Fig. 6 A and B) as they propagated from the soma into lateral dendrites. In contrast, APs in the apical dendrite did not attenuate with distance (<0.1% increase per 100 μm, n = 5; Fig. 6B; refs. 22 and 23). Because odors can evoke mitral cell firing at rates higher than 250 AP s⁻¹, we examined whether propagation in lateral dendrites depended on repetitive APs (40). When a train of APs was elicited at the soma, propagation of APs into distal lateral dendrites increased with AP number. For dendritic recordings further than 75 μm from the soma, attenuation was reduced from 41% for the first AP to 31% by the eighth AP (n = 6; Fig. 6 C and D). This improved propagation was associated with both an increase in AP amplitude in the dendrite and a decrease in AP amplitude in the soma. The larger peak depolarization of the later APs in a train may, therefore, result in more effective activation of voltage-dependent calcium channels in the lateral dendrites and thus in increased lateral inhibition.

Because lateral inhibition is driven by Ca²⁺-dependent glutamate release (17, 19) from the lateral dendrites of mitral cells, we measured the spatial profile of Ca²⁺ transients along these dendrites. Calcium transients observed in response to single somatically evoked APs decreased with distance by 20% per 100 μm from the soma (linear fit to data, R² = 0.47, n = 19; Fig. 7A). When multiple APs were evoked, the amplitude of calcium transients increased nonlinearly with AP number. In the distal lateral dendrites (>250 μm from the soma), calcium transients following five APs were 7-fold larger than those evoked by a single AP (1 AP = 16.0 ± 6.8 nM peak Δ[Ca²⁺] vs. 5 AP = 115.1 ± 40.8 nM peak Δ[Ca²⁺], n = 7 cells; Fig. 7A). In some cells, single APs evoked no measurable calcium transient in distal regions of the lateral dendrite (>300 μm from the soma), whereas trains of APs evoked large calcium elevations in these same regions (Fig. 7 Aii and Biii). Thus, repeated APs act cooperatively to elevate [Ca²⁺] in distal segments of lateral dendrites (Fig. 7 A and B). This observation suggests that, although single APs will fail to evoke glutamate release from the distal half of a lateral dendrite, repetitive APs will progressively recruit release from more distal release sites.

Calcium transients in lateral dendrites evoked by multiple APs increased substantially in the presence of bicuculline (ΔF/F evoked by 5 APs at 50 s⁻¹ was 22 ± 6% in controls vs. 39 ± 8% with bicuculline, n = 5) in dendritic segments >300 μm from the soma. Single AP-evoked calcium transients were unaffected by the addition of bicuculline (ΔF/F evoked by 1 AP was: 3.2 ± 0.4% vs. 3.7 ± 0.6%) in these same dendritic segments. These data suggest that recurrent inhibition regulates the spatial extent of AP propagation evoked by repeated AP firing in mitral cells.