Forward and Backward Propagation of Dendritic Impulses and Their Synaptic Control in Mitral Cells

See allHide authors and affiliations

Science  17 Oct 1997:
Vol. 278, Issue 5337, pp. 463-467
DOI: 10.1126/science.278.5337.463


The site of impulse initiation is crucial for the integrative actions of mammalian central neurons, but this question is currently controversial. Some recent studies support classical evidence that the impulse always arises in the soma-axon hillock region, with back-propagation through excitable dendrites, whereas others indicate that the dendrites are sufficiently excitable to initiate impulses that propagate forward along the dendrite to the soma-axon hillock. This issue has been addressed in the olfactory mitral cell, in which excitatory synaptic input is restricted to the distal tuft of a single primary dendrite. In rat olfactory bulb slices, dual whole cell recordings were made at or near the soma and from distal sites on the primary dendrite. The results show that the impulse can be initiated in either the soma-axon hillock or in the distal primary dendrite, and that the initiation site is controlled physiologically by the excitatory synaptic inputs to the distal tuft and inhibitory synaptic inputs near the soma.

The mitral cell of the mammalian olfactory bulb gives rise to one apical dendrite (termed primary dendrite) and several lateral basal dendrites (termed secondary dendrites) (1). The primary dendrite extends 200 to 350 μm in the rat and enters a single glomerulus, where it ramifies in a distal tuft of branches that receive excitatory synaptic input from terminals of olfactory nerve (ON) (Fig.1A). The secondary dendrites extend horizontally in the external plexiform layer (EPL), mediating recurrent and lateral synaptic inhibition through reciprocal dendrodendritic synapses with inhibitory granule cell interneurons (2).

Figure 1

Increasing dendritic synaptic excitation or hyperpolarizing somatic membrane potential shifted action-potential initiation from near the soma to the distal primary dendrite. (A) Morphology of a biocytin-stained mitral cell simultaneously recorded from the soma and the distal primary dendrite as indicated. Note that the axon originated from the base of the cell body. Some secondary dendrites, which extended horizontally for more than 800 μm, have been truncated. (B) Simultaneous whole cell recordings from proximal (near soma) and distal sites on the primary dendrite (electrode distance: 217 μm). The solid trace shows proximal recording, dashed trace shows distal recording. Stimulus strength delivered to the ON was increased in three steps from 17 μA to 25 μA to 33 μA, producing a shift in site of impulse initiation from near soma site (17 μA and 25 μA, solid tracing first) to distal site (33 μA, dashed tracing first). These magnitudes of ON shocks are relevant only for comparisons within the same preparation; they varied between preparations because of varying distances between the stimulating electrodes and the activated nerve fibers (see also Figs. 2and 3). (C) Another mitral cell showed a shift of impulse initiation site in response to two identical ON stimuli (49 μA, 200 μs) delivered at different somatic membrane potential levels (–58 and –72 mV). The resting membrane potential was –58 mV, and somatic membrane hyperpolarization was obtained by injecting DC (–200 pA) through the somatic electrode. The dendritic recording site was 348 μm from the soma. M, mitral cell soma; a, axon; p, primary dendrite; s, secondary dendrite; Glo, glomerulus; ON, olfactory nerve; G, granule cell.

As inferred by Ramon y Cajal in 1911 and established by later authors (1), the main function of the primary dendrite is clear: It links the synaptic response to olfactory input in the distal tuft to impulse output in the axon. This has served as one of the best examples proving that distal dendritic input is not limited to slow background modulation of soma output, but can mediate rapid information transmission (3). The location of excitatory synapses exclusively on the distal tuft thus furnishes an attractive model for investigating the role of dendrites in linking excitatory synaptic input to impulse output in a mammalian central neuron. In addition, the entirely separate location of granule cell inhibitory input to the soma and secondary dendrites provides a model for investigating the control of excitatory dendritic responses by synaptic inhibition.

Mitral cells were visualized in thick tissue slices of rat olfactory bulb under an infrared differential interference contrast (DIC) microscope (4-6). Primary dendrites were clearly visible and could be traced from the cell bodies all the way into individual glomeruli. Simultaneous whole cell recordings were made from the soma and at various distal sites on the primary dendrite near the glomerular tuft (n = 24 pairs from 22 slices). Dual dendritic recordings from proximal and distal sites (Fig. 1B;n = 5) gave results similar to those from soma and distal sites. In all recordings, fast action potentials were observed in both soma and distal dendrite.

Excitatory postsynaptic potentials (EPSPs) were evoked in mitral cells by stimulation of the olfactory nerve layer. A single shock to the ON evoked an EPSP that was always larger at the distal dendritic recording site, consistent with the synaptic input being confined to the distal glomerular tuft (Fig.1, B and C; n = 17). When the ON stimuli were weak, the action potentials arising from this EPSP occurred first at or near the soma and second at the distal dendritic recording site (Fig. 1B; two impulse pairs on the right; n = 17). This finding is consistent with axosomatic initiation followed by back-propagation of impulses, as observed in neocortical and hippocampal pyramidal cells (7). However, as the stimulus intensity increased, action-potential initiation shifted to the distal primary dendrite, followed by an action potential at the near-soma site (Fig. 1B, leftmost impulse pair; n = 14). Both the somatic and dendritic impulses, when evoked by current pulse injection, were blocked by 1 μM tetrodotoxin in the bath, indicating that these fast action potentials were dependent on the activation of sodium channels [(8); n = 5]. The higher levels of synaptic excitation presumably correspond to stronger odor stimulation, which in vivo elicits large-amplitude slow potentials across the glomeruli, and large amplitude EPSPs giving rise to burst responses in mitral cells (9).

To rule out possible effects of whole cell patch break-in, dual recordings of small amplitude extracellular spikes were made before break-in [(8); n = 8]. These spikes showed the same shift in initiation site with increasing ON synaptic input. After patch break-in, the mitral cells had resting potentials of –55 to –65 mV, and some cells spontaneously fired action potentials, similar to properties seen in vivo (9). Possible initiation of impulses from an axon of dendritic origin [as in substantia nigra neurons (10)] was ruled out by staining the dually recorded mitral cells with fluorescent dyes and biocytin (n = 8); in all cases, the axon was found to emerge from the cell body (Fig.1A), as described previously in a number of morphological studies (1).

In addition to being dependent on excitatory synaptic input, the impulse initiation site was also dependent on inhibitory input. With weak ON stimuli, the somatically initiated impulse could be blocked by hyperpolarizing the somatic membrane, or by an inhibitory postsynaptic potential (IPSP) elicited by a single shock to the EPL (8). With moderate ON stimuli, however, the site of impulse initiation could be shifted from soma to distal dendrite by soma hyperpolarizing current injection (Fig. 1C; n = 10) . The same result could be obtained by an EPL-evoked IPSP. This IPSP was always larger at the soma than at the distal primary dendrite (Fig.2, A and C; n = 9), consistent with passive spread of the IPSP into the distal primary dendrite from dendrodendritic synapses located at the secondary dendrites and soma (1-3). The IPSP could change the initiation site and propagation direction of an impulse evoked either by distal primary dendrite current injection (Fig. 2, A and B;n = 3) or by ON excitatory synaptic input (Fig. 2, C and D; n = 5). Thus, it appeared that both the excitatory synaptic input at the distal primary dendrite and the granule cell inhibition close to the soma can control the relative excitability between soma and distal primary dendrite and thereby change the site of impulse initiation and direction of impulse propagation. Even when impulse initiation would normally occur at the soma, a distal dendritic hot spot can provide a means for a nerve cell to overcome inhibition at the soma.

Figure 2

Inhibitory synaptic input near the soma shifted action-potential initiation from soma to distal dendrite. (A) Effect of an IPSP on impulse initiation evoked by a brief depolarizing current pulse (120 pA, 20 ms) injected into the distal primary dendrite. Upper panel: dendritic recording. lower panel: somatic recording. Two traces are superimposed in each panel: control response to the current pulse and test response preceded by a shock (20 μA, 200 μs) delivered laterally in the deep EPL to evoke a slow, long-latency IPSP in the secondary dendrites of the recorded cell [accordingly, the IPSP was larger at the soma (*) than at the dendritic recording site]. The dendritic recording site was 314 μm from the soma. (B) An expanded view of the evoked action potentials in (A), to show the shift in impulse initiation site. (C) Effect of an IPSP on impulse initiation evoked in another mitral cell by synaptic excitation of the distal dendrite. ON shocks (100 μA, 200 μs) were delivered (see arrowheads) to evoke an EPSP-generated action potential in the absence and presence of an IPSP from a preceding EPL shock (filled circle: 200 μA, 200 μs). The dendritic recording site was 305 μm from the soma. (D) An expanded view of the ON-evoked action potentials in (C), showing the shift in impulse initiation from slightly earlier in the soma (without IPSP) to earlier in the dendrite (with IPSP). For diagram abbreviations, see Fig. 1 caption.

A classic sign of distal dendritic excitability is a fast prepotential recorded from the soma, as first shown in hippocampal pyramidal cells (11). This type of response has been seen in intracellular studies (6, 12) in vitro in the turtle and rat olfactory bulb, as well as in vivo in the rabbit. A fast prepotential was sometimes observed in our paired recordings (n = 7) (Fig. 3, A and B). In the first of two consecutive sweeps (trace 1 in Fig. 3A), the full-size somatic impulse, which lagged the corresponding dendritic impulse, was preceded by a small prepotential (single arrow) that also occurred later than the dendritic impulse (see Fig. 3B, top, for comparison of the somatic and dendritic impulses). In the second sweep (trace 2 in Fig. 3A), in which the soma was slightly hyperpolarized, the full-size somatic impulse failed, yielding an isolated prepotential (double arrow) that appeared to be caused by passive spread of the dendritic impulse to the soma (see Fig. 3B, bottom, for comparison with the dendritic impulse). Occasionally, an EPL-evoked IPSP was strong enough to block forward propagation (Fig. 3, C and D; n = 4), leaving at the soma a small prepotential (trace 4, single arrow) reminiscent of a dendritic action potential (trace 2). This indicates that an action potential can be initiated independently in the distal dendrite, even without pairing with a somatic action potential.

Figure 3

Fast prepotentials recorded at the mitral cell soma were always correlated with a dendritic action potential. (A) Threshold ON stimulation (400 μA, 200 μs) evoked in the mitral cell soma a full-size action potential that was preceded by a small prepotential (single arrow, trace 1). Occasionally, the same stimulation failed to evoke a full-size action potential, yielding an isolated fast prepotential (double arrow, trace 2) at the soma. Membrane potential (with −290 pA DC injection) was −80 mV. (B) Comparisons of dendritic recordings (dashed line) with their corresponding somatic recordings (solid line) shown in traces 1 and 2 of (A), respectively, for the case when the prepotential elicited a soma impulse (top) and when it failed (bottom). The dendritic recording site was 223 μm from the soma. (C) Inhibitory synaptic input near the soma blocked forward propagation of a dendritic impulse, yielding a small somatic prepotential (single arrow, trace 4). Traces 1 and 2 are dendritic recordings, whereas traces 3 and 4 are simultaneous somatic recordings. Traces 1 and 3 are control ON stimulation without prior EPL stimulation; traces 2 and 4 are the EPL-ON sequential activation. EPL stimulation evoked an IPSP that was preceded by an action potential, which was due to direct stimulation of the secondary dendrites of the recorded cell. The IPSP was differentiated from the spike afterhyperpolarization (AHP) by its blockade with 30 μM bicuculline methiodide (8). EPL stimulation: 20 μA, 200 μs. ON stimulation: 100 μA, 200 μs. The dendritic recording site was 286 μm from the soma. (D) An expanded view of the ON-evoked action potentials in (C). Left panel, control traces 1 and 3 without IPSP; right panel, traces 2 and 4 with preceding IPSP.

These results provide a perspective on current questions concerning the site of impulse initiation in mammalian central neurons. It is well known that dendrites in many types of neurons generate Na+impulses (13). In hippocampal and neocortical pyramidal cells, for example, weak excitatory synaptic activation at impulse threshold elicits a somatic impulse which back-propagates into the dendrites (7). There has been evidence that at higher excitation intensities, the impulse trigger zone can shift to the apical dendrites (14). However, it has been suggested that this orthodromically elicited dendritic impulse is a purely local event occurring only during very strong excitation, that any excitable response of the dendrites is not able to propagate effectively to the soma and axon and therefore should be considered an active form of synaptic integration rather than action-potential initiation, and that the axon initial segment is the only site for impulse initiation at all intensities of synaptic input in mammalian central neurons (15).

Our recordings confirm that the classical model applies to the mitral cell at weak levels of synaptic excitation, when the impulse is initiated at the axosomatic region and back-propagates into the dendrites. Both somatic and dendritic impulses are Na+-dependent. With increasing intensity of dendritic depolarization, active responses of the dendrites are initiated, as has also been reported in other neurons (7, 14, 15). Our findings are that in the mitral cell, this distal dendritic activity consists of an impulse that has a large amplitude and rapid time course; it leads directly to impulse generation at the soma, and this link appears to be quite reliable.

The results further show that in a range of EPSP amplitude where the classical model of somatic impulse initiation applies, proximal inhibitory input can shift the impulse origin for the same EPSP to the distal dendrite and change the direction of impulse propagation in the dendrite from backward to forward. These mechanisms likely play critical roles in sensory responses under physiological conditions. It may be postulated that both the large-amplitude EPSPs and large-amplitude IPSPs recorded in vivo in response to strong odor stimulation [(9); see above] function in a coordinated manner to shift impulse initiation to the distal dendrites. This could ensure that the distal tuft continues to respond actively to excitatory input despite the shutdown of impulse output in the axon. Further in vivo studies are needed to test directly for these mechanisms.

How broadly might these properties of mitral cells generalize to other neurons? Although the distal dendritic tuft is specialized for specific sensory input, this arrangement has similarities with pyramidal cells, where the lamination of inputs provides for discrete synaptic sites, both for excitation to the distal apical tuft and inhibition to the axon hillock (16). Distal (layer I) synaptic stimulation in neocortical pyramidal cells can evoke an axosomatic back-propagating impulse (7), as is the case in mitral cells with weak to moderate dendritic excitation. Computer simulations (17) have supported the experimental evidence (13, 14) that forward-propagating fast sodium action potentials can be initiated in the apical dendrites of mature cortical pyramidal neurons. With regard to synaptic influences on impulse propagation in dendrites, it is known that synaptic inhibition can control the extent of dendritic invasion of a back-propagating impulse in cerebellar Purkinje cells and hippocampal pyramidal neurons (18). Further studies are needed to determine whether proximal inhibition in Purkinje cells by basket cells and in pyramidal neurons by chandelier cells could, in addition, shift impulse origin to the dendrites. In these types of neurons that are under strong inhibition at or near the axon hillock, distal dendritic electrogenesis such as we describe in mitral cells provides a means of overcoming that inhibition and thus extending the operating range of the cell. Furthermore, even when impulse output from the axon hillock is inhibited, the dendritic impulse can continue to evoke local Ca2+ influx (18) that enables the distal dendrites to continue to be involved in local processes, such as dendritic release of transmitters or synaptic plasticity.

In summary, mitral cells provide a model for multiple impulse initiation sites within a neuron that explains how they may be controlled by both excitatory and inhibitory synaptic inputs. This model helps to widen the view of how dendritic excitability contributes to processing of information in different types of neurons in the vertebrate brain.


Stay Connected to Science

Navigate This Article