Logo of jphysiolThe Journal of Physiology SiteMembershipSubmissionJ Physiol
J Physiol. 2003 Jan 15; 546(Pt 2): 363–374.
Published online 2002 Nov 29. doi:  10.1113/jphysiol.2002.031245
PMCID: PMC2342519

Theta oscillation coupled spike latencies yield computational vigour in a mammalian sensory system


Theoretical work carried out almost a decade ago proposed that subthreshold oscillations in membrane potential could be used to convert synaptic current strength into a code reliant on action potential (AP) latencies. Using whole-cell recordings we present experimental evidence for the occurrence of prominent network-driven subthreshold theta oscillations in mitral cells of the mouse olfactory bulb. Activity induced by both injected current and sensory input was accurately reflected in initial AP latency from the beginning of each oscillation cycle. In a network model we found that an AP latency code rather than AP number or instantaneous firing rate provided computational speed and high resolution, and was easily implemented. This coding strategy was also found to be invariant to the total input current as long as the relative input intensities to glomeruli remained constant. However, it was highly sensitive to changes in the ratios of the input currents and improved by lateral inhibitory mechanisms. Since the AP latency-based coding scheme was dependent on the subthreshold oscillation we conclude that the theta rhythm serves a functional role in temporally reformatting the strengths and patterns of synaptic input in this sensory system.

Anatomical (Ramón y Cajal, 2001), molecular (Buck & Axel, 1991; Mombaerts et al. 1996) and functional imaging studies (Kauer, 1988; Joerges et al. 1997; Friedrich & Korsching, 1997; Rubin & Katz, 1999; Uchida et al. 2000; Meister & Bonhoeffer, 2001; Wachowiak & Cohen, 2001) provide convincing evidence that different odours activate spatially distributed combinations of glomeruli across the olfactory bulb (OB). However, due to the comparatively slow dynamics of odour stimuli it has also been suggested that time could be used as a computational variable for odour representation (Laurent, 1999; Friedrich & Laurent, 2001). In this way action potential (AP) timing or rate could code non-temporal stimulus features such as quality or intensity. Recent studies in olfactory bulb slices have shown that mitral cells receiving synaptic input within a common glomerulus laterally excite one another through chemical and/or electrical transmission (Isaacson, 1999; Carlson et al. 2000; Urban & Sakmann, 2002; Schoppa & Westbrook, 2002). In response to extracellular stimulation of the olfactory receptor neuron (ORN) input, this lateral excitation preferentially synchronizes the activity of intraglomerularly coupled mitral cells at approximately 2 Hz (Schoppa & Westbrook, 2001; Urban & Sakmann, 2002). Thus the early olfactory network provides an intrinsic amplification mechanism that could preferentially boost activity in the theta range.

In mammals the respiration cycle provides a mechanism for repetitive and periodic sampling of odour stimuli and is the source of much of the theta band activity (Macrides & Chorover, 1972; Bressler, 1987; Gray & Skinner, 1988). During nasal inhalation, in response to either odorous room air or controlled odour stimuli, electro-olfactogram recordings from ORNs show changes in potential that are coupled to respiration (Ottoson, 1959; Chaput, 2000). This rhythmic ORN activity propagates to the main OB and is reflected in the patterning of mitral cell discharge (Adrian, 1950; Walsh, 1956; Gault & Coustan, 1965; Macrides & Chorover, 1972; Onoda & Mori, 1980; Chaput & Holley, 1985; Doving, 1987; Ravel & Pager, 1990; Sobel & Tank, 1993; Philpot et al. 1997; Chalansonnet & Chaput, 1998; Lestienne et al. 1999; Charpak et al. 2001; Luo & Katz, 2001; Belluscio et al. 2002). The respiration-coupled theta rhythm displays spatiotemporal dynamics in response to odour stimuli (Spors & Grinvald, 2002) and is present in downstream piriform cortex (Bressler, 1987; Wilson, 1998a,b) indicative of a functional role in the representation of odour stimuli. Here we have therefore examined the possibility that individual mitral cells could use the theta rhythm as a coding device to temporally reformat the strength of synaptic input.


In vivo electrophysiology

Whole-cell recordings were carried out on mitral and granular cells of the olfactory bulb of freely breathing mice (C57BL6, P21-P35) anaesthetized with urethane (1.2 g kg−1i.p.) or with a ketamine (50 mg kg−1)/xylazine (5 mg kg−1) mixture as described previously in accordance with the ethical guidelines of the Max-Planck-Gesellschaft (Margrie et al. 2001, 2002). Patch pipettes (3-6 MΩ) were filled with (mm) potassium gluconate 130, sodium gluconate 10, Hepes 10, phosphocreatine 10, Mg2+-ATP 4, GTP 0.3, NaCl 4, biocytin 0.3-1 %, pH 7.2. The respiration cycle was recorded with a plethysmograph constructed from a piezo-electric strap fitted around the thorax to measure the distension signal (Kent Scientific Corporation, Litchfield, CT, USA). Mitral cells were identified by their membrane potential, input resistance, firing profile and the appearance of a recurrent IPSP (Margrie et al. 2001). Granule cells were identified by their distance from the surface of the bulb (> 300 μm), their hyper-polarized membrane potential (-71 ± 2 mV, n = 41), input resistance (429 ± 21 MΩ, n = 41) and in some cases their recovered morphology. A custom built olfactometer was used to present odour to the nares. For odour presentation positive pressure pulses (50 mbar; usually 5 s in duration) of filtered room air (the carrier stream) were propelled through a glass chamber containing the odour solution into the stimulus air line. A series of Teflon valves allowed for rapid redirection of the pulsed air to alternative odorant chambers. For current step injection-evoked AP experiments (Fig. 3B), current steps were of 100 ms duration (40-450 pA). Naris occlusion was achieved under visual guidance. With a micromanipulator a small piece of silicon gel was firmly placed against and/or marginally inside the cavity. All pharma-cology experiments were carried out on cells that exhibited the oscillation regularly for at least 10 min. Data are means ± s.e.m. unless stated otherwise. ANOVA and unpaired t tests were used to determine significance values. Agents used were TTX, 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo(F)quinoxaline (NBQX), 2-amino-5-phosphonovalerate (D-APV) and n-amyl acetate (1 %), ethyl butyrate (1 %), 2-methylbutyric acid (1 %) (Sigma); lavender and cherry (50 or 100 % of stock) was obtained from a local retailer. All odorants were diluted in 100 % mineral oil (Sigma).

Figure 3
Synaptically evoked AP firing rates and latencies in individual mitral cells

Network model

Simulations were performed in Matlab 6.0 (The MathWorks, Natick, MA, USA). The network consisted of an interconnected mitral cell layer each receiving independent input within an individual glomerulus. Thus each mitral cell reflected the activity of all cells associated with a glomerulus (Schoppa & Westbrook, 2001; Urban & Sakmann, 2002). Mitral cells were modelled as leaky integrate-and-fire neurons with an AP threshold of 15 mV, input resistance of 100 MΩ and τ = 10 ms. Subthreshold oscillation was mimicked by a 4 Hz sinusoidal current injection with a peak amplitude of 125 pA. If threshold was reached at t = t0, an AP was elicited (Vmem(t0) = 100 mV), the membrane potential (Vmem) was then reset to 12 mV (80 % of threshold) and hyperpolarizing current was injected to mimic an after-hyperpolarization (AHP) (linearly decreasing from an initial amplitude of 200 pA within 50 ms). Incorporating heterogeneity in membrane potentials (changes by up to ± 5 mV) did not alter the results. AP-evoked lateral inhibition in all other mitral cells (White et al. 1992) was delayed by 25 ms and followed a bi-exponential time course i(t)= 200 pA (1 – et/30ms)et/100ms as extracted from Urban & Sakmann (2002). Stimulation was mimicked by a 1 s step current injection starting at t = 1000 ms locked to the beginning of the oscillations. With these parameters current injections of 30-200 pA evoked 1-5 APs per oscillation cycle. Using theta locked current injection in addition to the step current did not affect the described computational properties (not shown). To analyse to what extent oscillation-mediated AP latencies can indeed readily represent stimulus features, an explicit mechanism was designed to extract stimulus quality (the pattern of mitral cell activation) from action potential latencies. To obtain this, ‘analysis cells’ were introduced (Fig. 4C) that received input with varying delays from all mitral cells, such that a mitral cell AP resulted in a postsynaptic potential in the analysis cell (VPSP(t) = ± (1 – et/2 ms)e−t/5 ms, Fig. 4C and D). The number of analysis cells with non-overlapping receptive fields (as defined by the fraction of stimulus space that excites the analysis cell maximally) indicates the number of different stimulus patterns that can be uniquely represented and detected by the coding scheme.

Figure 4
A network model with firing properties based on experimental data

Connectivity details were as follows. Connections to analysis cell n from mitral cells that responded with action potentials to stimulus pattern n were set excitatory (+ sign in the above time course), others were inhibitory. Delays were chosen according to the recorded simulated AP latencies (range, 1-54 ms; mean ± s.d., 17.5 ± 14.6 ms for excitatory connections) or randomly selected from a uniform distribution between 0 and 65 ms for inhibitory connections. While the excitatory delays are well within the expected range for mono- and di-synaptic excitation in piriform cortex (Ketchum & Haberly, 1993b), what is critical for the model is the relative latencies of the APs from different mitral cells projecting to the same AC. The range for inhibition was chosen to ensure that the decay was complete within the time frame revealed by current-density source analysis in the piriform cortex (Ketchum & Haberly, 1993a).

To assess the computational power of this oscillation/AP latency-based scheme it was compared to (i) a spatial scheme relying only on the identity of the activated glomeruli and (ii) a spatial scheme relying on the number of APs per cycle in the mitral cells. Comparison was made in a network with four glomerular units using a template matching procedure: A stimulus pattern S0 was selected and presented to the mitral cells. The resulting (i) glomerular activation pattern, (ii) number of action potentials and (iii) depolarization in ‘analysis cell’ AC0 were taken as templates for the respective coding strategies. A stimulus pattern S1 was assumed to be identical to S0 if (i) glomerular activation or (ii) action potential numbers matched the respective template or (iii) the resulting depolarization in AC0 was larger than or equal to the template depolarization. This was done for all S1 fulfilling the following conditions: (1) two of the four glomeruli receive input and (2) input is 5, 10, … or 200 pA. This resulted in a number of erroneous matches that could be expressed as a fraction of the entire stimulus space (with six possible combinations of active glomeruli, stimulus space consisted of 40 × 40 × 6 stimuli). This was repeated for 14 different S0 (150 or 200 pA; 50, 75, 100,…. or 200 pA). To study robustness, Gaussian noise (σ = 50 pA) was added point by point independently of the 1 kHz-sampled signal, resulting in an amplitude similar to the experimentally observed intrinsic noise. Control simulations in a compartmental model of a mitral cell (Shen et al. 1999) were performed in NEURON 4.1.2 (http://www.neuron.yale.edu) and yielded similar results.


Consistent with previous studies in the mammalian olfactory bulb (Adrian, 1942; Ottoson, 1959; Gault & Leaton, 1963; Macrides & Chorover, 1972; Onoda & Mori, 1980; Bressler & Freeman, 1980; Chaput & Holley, 1985; Gray & Skinner, 1988; Charpak et al. 2001; Luo & Katz, 2001) whole-cell recordings from mitral cells show a dominating rhythm synchronized to the respiration cycle (n = 23, Fig. 1A and B). This locking of membrane potential to the respiration cycle was observed for periods ranging from tens of seconds up to for as long as the recording persisted (> 1 h) and could be evoked by room air or by overt odour stimuli (n = 23; Fig. 1A-C) consistent with recent voltage-sensitive dye measurements in rodents (Spors & Grinvald, 2002). In single cells, the subthreshold oscillation had an amplitude (peak to trough) of 6.9 ± 1.3 mV, an average frequency of 3.7 ± 0.2 Hz (n = 23; Fig. 1D) and was abolished following superfusion of the sodium channel blocker TTX or NBQX and D-APV over the surface of the olfactory bulb (control 6.3 ± 0.9 mV, n = 11; TTX 2-20 μm, 0.2 ± 0.3 mV n = 5; NBQX (50-200 μm) plus D-APV (200-400 μm), 0.9 ± 0.4 mV (n = 5); all P values were < 0.05, Fig. 1E). Closure of the ipsilateral naris also diminished the subthreshold theta rhythm in mitral cells (control 6.3 ± 0.9 mV, n = 11; 0.6 ± 0.4 mV, n = 3; P < 0.05; Fig. 1E). Fast Fourier transform (FFT) analysis of membrane potential in mitral cells revealed that the respiration-locked activity occurs in the theta bandwidth (Fig. 1D). The LFP measured in the upper third of the external plexiform layer exhibited identical pharmacological properties, power spectrum and correlation structure (n = 7; not shown). This theta rhythm is therefore dependent on nasal airflow, requires AP generation/propagation and glutamatergic transmission within the olfactory bulb. Furthermore the general kinetics of sensory input-evoked synaptic current in mitral cells is defined by the inspiration-exhalation cycle (Fig. 1C).

Figure 1
Odour-evoked theta oscillations in mitral cells depend on synaptic transmission in the olfactory bulb

The influence of subthreshold membrane potential oscillations on AP timing was then investigated in detail by direct current input via the patch pipette (Fig. 2A). Progressively more positive constant current resulted in a decreased latency to the first AP and in an increase in AP number (Fig. 2B). Since current intensity and AP number covaried, AP number can be used as an indicator of input strength. Figure 2C shows that odour-evoked and direct current injection-evoked APs exhibit the same relationship between initial AP latency and AP number per oscillation cycle. In the case of synaptically evoked APs, the majority of APs occurred towards the end of the inhalation phase of the respiration (Fig. 3A 1). In 95 % of the odour-evoked suprathreshold oscillation cycles five or less APs were evoked (Fig. 3A 2).

Figure 2
Direct current injection into mitral cells produces a current amplitude-dependent increase in AP number and decrease in the initial AP latency

When coding of information by individual neurons is considered, there exist a number of features of individual spike trains that may convey information about a stimulus. Within each oscillation cycle these may include inter-spike interval (ISI), AP timing or latencies, AP number and higher-order features such as changes in sustained firing rate (Laurent, 1999; Laurent et al. 2001). Figure 3B shows that the average ISI for mitral cells receiving this oscillatory drive remains constant throughout the range from two to five APs both for odour- and current injection-evoked suprathreshold activity (F(3,221) = 1.8, P > 0.1; 39 ± 3.9(2 APs) vs. 42.7 ± 5.8 Hz (5 APs); n = 4 for odour-evoked APs; 46.7 ± 4.9 Hz (2 APs) vs. 51.1 ± 5.1 Hz (5 APs) for current injection-evoked APs; n = 5). As this held true for non-oscillating cells (56.4 ± 4.2 Hz (2 APs) vs. 61.6 ± 2.7 Hz (5 APs); n = 7), the observed constant firing rate is likely to be due to intrinsic (Desmaisons et al. 1999) or possible feed-forward inhibitory mechanisms, rather than the consequence of the oscillatory drive. Since the instantaneous firing rate remains constant for less than six APs per cycle (which accounts for over 95 % of all observed cycles) firing rate cannot represent the strength of network-driven input during the vast majority of respiration cycles (Fig. 3A 2 and B). Thus, we investigated whether the distribution of spikes across each cycle efficiently conveys information about input strength.

If AP latencies were to play any role in sensory representation in the olfactory system, then computation in downstream neuronal elements would rely on the differential latencies of incoming APs. To assess this possibility experimentally it is therefore necessary to use an external reference that is insensitive to sensory input and olfactory bulb activity. We found that when the mitral cell membrane potential was aligned to the beginning of the respiration cycle (Fig. 3C 1) the latency to the initial spike decreased as the total number of spikes per cycle increased (1 AP (104.2 ± 3 ms) vs. 5 APs (46.9 ± 4 ms), P < 0.0001; Fig. 3C 2). This covariance between AP number and latency to the initial spike was highly significant (latency to initial AP for the range of 1-8 APs; F(7,179) = 30.3, P < 0.0001) which was not the case for the average AP latency per cycle (103 ± 2 ms (2 APs) vs. 108 ± 5 ms, (6 APs); F(7,940) = 1.7, P > 0.1). Due to the activity-dependent distribution of spikes (Fig. 3D) the initial AP latency may provide a reliable and rapid means of representing input strength.

To assess the potential computational role of AP latency and the subthreshold oscillatory drive (Hopfield, 1995), we constructed a network consisting of leaky integrate-and-fire models of mitral cells receiving independent input that were connected via lateral inhibition (Fig. 4A). In this model, stimulation was mimicked by a 1 s step on top of a sinusoidal subthreshold current injection (Fig. 4A). Stimulus space consists of all combinations of current injection amplitudes into mitral cells that result in up to six APs. Since experimental work suggests that only a subset of glomeruli are activated by an individual stimulus (Kauer, 1988; Joerges et al. 1997; Friedrich & Korsching, 1997; Rubin & Katz, 1999; Uchida et al. 2000; Meister & Bonhoeffer, 2001; Wachowiak & Cohen, 2001), the fraction of simultaneously activated glomeruli in the model was chosen to be 50 % (Hopfield, 1999; see discussion).

The intrinsic cellular properties of model mitral cells including AHPs and firing rates were comparable to the experimental data (Fig. 4B 1 and B3). Importantly, the initial AP latency decreased logarithmically as a function of AP number while the mean latency remained constant as in the experimental data (Fig. 4B 2). We then investigated the possibility that AP latencies could be used by downstream circuits to extract stimulus-related information about the net strength of synaptic input to mitral cells. Therefore, we constructed a second layer of ‘analysis cells’ that receive convergent input from mitral cells with varying delays (Fig. 4C) and should each respond optimally to a unique stimulus S (see Methods).

Successive presentation of 20 stimulus patterns to a network of 10 mitral cells resulted in analysis cell n mainly depolarizing following presentation of stimulus pattern n (diagonal in Fig. 4D). This indicates that 20 stimulus patterns can be distinguished unequivocally. To further assess, quantify and compare the computational vigour of the proposed AP latency coding strategy, we used a smaller network of only four mitral cells so the entire stimulus space could be examined. We chose to compare three coding strategies, namely (1) spatial pattern of activated glomeruli (2) pattern of AP numbers and (3) pattern of AP latency. Template matching showed that the pure spatial strategy wrongly matched 9.08 ± 0.67 % of stimulus space whereas the second and third coding strategies matched 1.28 ± 0.14 and 0.73 ± 0.18 % respectively (Fig. 5A). Addition of 50 pA noise resulted in 9.47 ± 0.77, 1.33 ± 0.14 and 0.95 ± 0.2 % erroneous matches, thus retaining computational supremacy of the AP latency coding scheme (Fig. 5A).

Figure 5
AP latency coding provides computational advantage and enables scale invariant pattern detection dependent on subthreshold oscillations

Erroneous matches represent a (usually undesired) generalization from the stimulus template to a larger region of stimulus space. To further assess the nature of the generalization we plotted the entire stimulus space for one analysis cell. From Fig. 5B it is apparent that only sensory input into glomeruli 1 and 2 resulted in erroneous matches in any coding strategy. We next examined in greater detail the subset of stimulus space (1/6 of the total) corresponding to the activation of the two correct glomeruli (Fig. 5B). We find that strategies 1 and 2 simply generalize approximately radially from the template. The AP latency strategy, however, detects not only the template stimulus but also linearly scaled versions of the template. Thus the ratio rather than the total synaptic inputs onto mitral cells is reliably encoded by the subthreshold oscillation-driven AP latencies (Hopfield, 1995). This is consistent with an approximately logarithmic relation between initial AP latency and input strength as seen both in simulation and in the experimental data (Fig. 2C and Fig. 4B). Decreasing the amplitude of the subthreshold oscillation resulted in an approximately 10-fold increase in the number of erroneous matches (Fig. 5C), indicating that indeed the theta oscillation is necessary to retain high fidelity stimulus pattern encoding.

It has been proposed that lateral inhibition provides the olfactory bulb with a mechanism for amplifying differences in mitral cell activity patterns (Shepherd & Greer, 1998). Consistent with the notion that the theta rhythm-locked AP firing in mitral cells evokes lateral inhibition across the bulb we observed EPSPs in granular cells (amplitude = 12.5 ± 2.2 mV, n = 18, Fig. 6A and A). In the limited number of cases when granular cells reached threshold, AP firing was also locked to the respiration cycle. Consistent with very low overall firing rates (Margrie et al. 2002) in both mitral and in particular granule cells (< 0.01 Hz, n = 18), we observed respiration-coupled inhibitory postsynaptic potentials in 3/23 mitral cells (Fig. 6B). In line with previous in vitro studies on recurrent inhibition in mitral cells (Isaacson & Strowbridge, 1998) the respiration-locked inhibition was mediated by fast putatively unitary events (Fig. 6B). The majority of these events occurred between 20 and 40 % of the respiration cycle length (0.32 ± 0.09, mean ± s.d., n = 162; Fig. 6B2) and thus on the early phase of the distribution of mitral cell firing.

Figure 6
Theta oscillation synchronous lateral inhibition increases AP latency-dependent coding capacity

Removing lateral inhibition in the model (amplitude = 0) resulted overall in a 2-fold increase in the number of erroneous matches (Fig. 6C). This was not significantly altered if a tonic hyperpolarizing current was added to account for the overall loss of inhibition. As noted above, the AP latency scheme predominantly generalizes from a stimulus to linearly scaled versions of the stimulus, detecting the ratio between the input currents rather than the total amount of current input to the glomeruli (Fig. 5B). In Fig. 6D1 the set of linearly scaled versions of the stimulus is indicated by the black line through the origin and the correct stimulus. To assess how lateral inhibition contributes to shaping this profile, we determined the distance of an erroneous match to this ‘constant-ratio’ set of stimuli (Fig. 6D1, dotted line). From Fig. 6D2 it becomes apparent that lateral inhibition does not only result in an overall reduction of erroneous matches but predominantly restricts the erroneous matches to a narrow region in stimulus space. This demonstrates that for the AP latency coding scheme, lateral inhibition is not crucial but does indeed enhance pattern separation. In particular, it contributes to restricting the ‘erroneous’ matches to a well-defined region in stimulus space.


Studies on various sensory modalities report both beta and gamma frequency activity in response to sensory stimuli (Singer & Gray, 1995; Laurent et al. 2001). Here we report experimental and theoretical evidence that a respiration-coupled subthreshold theta oscillation in mitral cells can provide a temporal platform for reformatting net synaptic input intensities into AP latencies. This subthreshold oscillation in mitral cells was dependent on activation of AMPA and NMDA receptors indicating that excitatory transmission in the olfactory bulb is essential for the generation of the respiration coupling. Since excitatory synaptic input to mitral cells is confined to the glomerulus it is likely that ORN input drives the oscillatory activity (Ottoson, 1959; Chaput, 2000). Intra-glomerular excitation between the tufts of mitral cells may then serve as a mechanism for amplification and synchronization (Schoppa & Westbrook, 2001; Urban & Sakmann, 2002).

In this study we used the respiration cycle as a reference point from which to measure AP latency. Since the respiration cycle length can be reliably recorded and varies only marginally across the recording period it is a useful external reference to determine the relative AP latency across each theta cycle and different cells in different preparations. We found that the latency to the first AP co-varied with AP number but not ISI to reflect input intensity. The logarithmic relationship between AP latency and input intensity (Fig. 2C and Fig. 4B) might allow efficient encoding of relative input intensities (Hopfield, 1995) as described by our network model.

To reduce the number of free parameters in the model neither short-term dynamics nor variability in synaptic kinetics or efficacy were utilized. Such features can in addition increase computational power of a latency-based scheme. For computational reasons we chose a network with only four glomerular units, two of which were activated by a given stimulus. However, increasing the total number of glomerular units did not significantly change our results while any increase in the number of activated glomeruli actually substantially enhanced the computational supremacy of the latency-based coding scheme (not shown). Analysis cells were introduced in our model predominantly as a direct test of this coding strategy. However, the simplicity of the detection mechanism used in our model (delays and postsynaptic summation) and that simple learning rules can be applied to build the described connectivity make it a biophysically realistic implementation. In addition, the parameters used in the model are consistent with the functional architecture of piriform cortex (Wilson & Shepherd, 1998; Zou et al. 2001). In particular the pyramidal cells of the piriform cortex receive input (both monosynaptic extrinsic and polysynaptic intrinsic) with varying delays up to several hundred milliseconds following afferent activation (Ketchum & Haberly, 1993a). In a previous study (White et al. 1998) a similar delay scheme was used to facilitate odour detection in an artificial nose system demonstrating a direct application of this computational parameter. In our model the analysis cells (ACs) are tuned to particular glomerular activity patterns. As in experimental data from second order olfactory neurons in mammals and insects, our ACs are sparsely activated in a highly stimulus-specific manner (Wilson, 2001; Perez-Orive et al. 2002).

Compared to unit recordings under the same anaesthesia conditions whole-cell recordings from mitral cells and granule cells reveal very low rates of spontaneous activity (Margrie et al. 2002). Consistent with this we observed respiration-coupled IPSPs in a limited set of mitral cells. The distribution of these IPSPs was found to coincide with the early phase of the distribution of APs in mitral cells. This is in agreement with recent in vitro experiments revealing that early action potentials in mitral cells are most efficient at evoking lateral IPSPs (Urban & Sakmann, 2002). This lateral inhibition will contribute to AP timing making it very likely that increasing odour concentration would result in a complex interaction between excitatory ORN input and lateral inhibition. For example, in cells where an increase in lateral inhibition dominates, the initial AP latency will increase rather than decrease with increasing concentration (Kauer, 1974; Yokoi et al. 1995; Margrie et al. 2001). In support of a role of lateral inhibition in tuning mitral cell responses (Yokoi et al. 1995) our data show that lateral inhibition can indeed refine an AP latency-based coding scheme in the early olfactory system. Although this study has not determined directly what features of an odour stimulus might be represented by the initial AP latency we can predict that via lateral inhibition, increasing odour concentration will have a heterogeneous effect on the initial AP latency in individual mitral cells. Further studies ideally presenting a large number of odours, varying odour quality and recording from multiple mitral cells or downstream elements are likely to yield more insight into the aspects of olfactory stimuli that are encoded in AP latencies.

There exist a number of advantages for using the initial AP latencies to represent the strength of synaptic input. By developing a scheme to detect stimulus patterns using AP latencies we have shown that such a strategy is not simply faster but also more efficient than, for example, codes reliant on AP number. In addition AP latencies will define distinct classes of stimuli where ratios of individual components are kept constant. Removing lateral inhibition resulted in additional erroneous matches specifically located distant from the axis of the linearly scaled versions of the stimulus. Thus in our model lateral inhibition enhanced the sharpness of the representation projected by AP latencies. Nevertheless, inhibition seems not to be crucial for discrimination of glomerular excitation patterns as overall the absence of lateral inhibition increased the number of erroneous matches on average only 2- to 3-fold. However, the oscillatory drive described here provides strongly activated cells with a reduced initial AP latency which will enable successful lateral inhibition of the firing onset of cells receiving weaker net synaptic input (Margrie et al. 2001). Taken together, these data suggest that the respiration cycle in mammals might serve a functional role in reformatting spatially organized olfactory input.


This work was supported by The Alexander von Humboldt Foundation, The National Health and Medical Research Council of Australia (T.M.), The Boehringer Ingelheim Fonds and the Heidelberger Akademie der Wissenschaften (A.S.). We thank J. Bekkers, M. Brecht, A. Brichta, R. Friedrich, J. Isaacson and N. Urban for comments on early versions of the manuscript and thank B. Sakmann for his support and encouragement.


  • Adrian ED. Olfactory reactions in the brain of the hedgehog. J Physiol. 1942;100:459–473. [PMC free article] [PubMed]
  • Adrian ED. The electrical activity of the mammalian olfactory bulb. Electroencephalogr Clin Neurophysiol. 1950;2:377–388. [PubMed]
  • Belluscio L, Lodovichi C, Feinstein P, Mombaerts P, Katz LC. Odorant receptors instruct functional circuitry in the mouse olfactory bulb. Nature. 2002;419:296–300. [PubMed]
  • Bressler SL. Relation of olfactory bulb and cortex. I. Spatial variation of bulbocortical interdependence. Brain Res. 1987;409:285–293. [PubMed]
  • Bressler SL, Freeman WJ. Frequency analysis of olfactory system EEG in cat, rabbit, and rat. Electroencephalogr Clin Neurophysiol. 1980;50:19–24. [PubMed]
  • Buck L, Axel R. A novel multigene family may encode odorant receptors: a molecular basis for odor recognition. Cell. 1991;65:175–187. [PubMed]
  • Carlson GC, Shipley MT, Keller A. Long-lasting depolarizations in mitral cells of the rat olfactory bulb. J Neurosci. 2000;20:2011–2021. [PubMed]
  • Chalansonnet M, Chaput MA. Olfactory bulb output cell temporal response patterns to increasing odor concentrations in freely breathing rats. Chem Senses. 1998;23:1–9. [PubMed]
  • Chaput MA. EOG responses in anesthetized freely breathing rats. Chem Senses. 2000;25:695–701. [PubMed]
  • Chaput MA, Holley A. Responses of olfactory bulb neurons to repeated odor stimulations in awake freely breathing rabbits. Physiol Behav. 1985;34:249–258. [PubMed]
  • Charpak S, Mertz J, Beaurepaire E, Moreaux L, Delaney K. Odor-evoked clacium signals in dendrites of rat mitral cells. Proc Natl Acad Sci U S A. 2001;98:1230–1234. [PMC free article] [PubMed]
  • Desmaisons D, Vincent JD, Lledo PM. Control of action potential timing by intrinsic subthreshold oscillations in olfactory bulb output neurons. J Neurosci. 1999;19:10727–10737. [PubMed]
  • Doving KB. Response properties of neurones in the rat olfactory bulb to various parameters of odour stimulation. Acta Physiol Scand. 1987;130:285–298. [PubMed]
  • Friedrich RW, Korsching SI. Combinational and chemotopic odorant coding in the zebrafish olfactory bulb visualized by optical imaging. Neuron. 1997;18:737–752. [PubMed]
  • Friedrich RW, Laurent G. Dynamic optimization of odor representations by slow temporal patterning of mitral cell activity. Science. 2001;291:889–894. [PubMed]
  • Gault FP, Coustan DR. Nasal air flow and rhinencephalic activity. Electroencephalogr Clin Neurophysiol. 1965;18:617–624. [PubMed]
  • Gault FP, Leaton RN. Electrical activity of the olfactory system. Electroencephalogr Clin Neurophysiol. 1963;15:299–304. [PubMed]
  • Gray CM, Skinner JE. Centrifugal regulation of neuronal activity in the olfactory bulb of the waking rabbit as revealed by reversible cryogenic blockade. Exp Brain Res. 1988;69:378–386. [PubMed]
  • Hopfield JJ. Pattern recognition computation using action potential timing for stimulus representation. Nature. 1995;376:33–36. [PubMed]
  • Hopfield JJ. Odor space and olfactory processing: collective algorithms and neural implementation. Proc Natl Acad Sci U S A. 1999;96:12506–12511. [PMC free article] [PubMed]
  • Isaacson JS. Glutamate spillover mediates excitatory transmission in the rat olfactory bulb. Neuron. 1999;23:377–384. [PubMed]
  • Isaacson JS, Strowbridge BW. Olfactory reciprocal synapses: dendritic signaling in the CNS. Neuron. 1998;20:749–761. [PubMed]
  • Joerges J, Kuttner A, Galizia GC, Menzel R. Representations of odors and odor mixtures visualized in the honeybee brain. Nature. 1997;387:285–288.
  • Kauer JS. Response patterns of amphibian olfactory bulb neurones to odour stimulation. J Physiol. 1974;243:695–715. [PMC free article] [PubMed]
  • Kauer JS. Real-time imaging of evoked activity in local circuits of the salamander olfactory bulb. Nature. 1988;331:166–168. [PubMed]
  • Ketchum KL, Haberly LB. Membrane currents evoked by afferent fiber stimulation in rat piriform cortex. I. Current source-density analysis. J Neurophysiol. 1993a;69:248–260. [PubMed]
  • Ketchum KL, Haberly LB. Membrane currents evoked by afferent fiber stimulation in rat piriform cortex. II. Analysis with a system model. J Neurophysiol. 1993b;69:261–281. [PubMed]
  • Laurent G. A systems perspective on early olfactory coding. Science. 1999;286:723–728. [PubMed]
  • Laurent G, Stopfer M, Friedrich RW, Rabinovich MI, Volkovskii A, Abarbanel HDI. Odor encoding as an active, dynamical process: experiments, computation and theory. Annu Rev Neurosci. 2001;24:263–297. [PubMed]
  • Lestienne R, Tuckwell HC, Chalansonnet M, Chaput M. Repeating triplets of spikes and oscillations in the mitral cell discharges of freely breathing rats. Eur J Neurosci. 1999;11:3185–3193. [PubMed]
  • Luo M, Katz LC. Response correlation maps of neurons in the mammalian olfactory bulb. Neuron. 2001;32:1165–1179. [PubMed]
  • Macrides F, Chorover SL. Olfactory bulb units: activity correlated with inhalation cycles and odor quality. Science. 1972;175:84–87. [PubMed]
  • Margrie TW, Brecht M, Sakmann B. In vivo, low-resistance, whole-cell recordings from neurons in the anaesthetized and awake mammalian brain. Pflügers Arch. 2002;444:491–498. [PubMed]
  • Margrie TW, Sakmann B, Urban NN. Action potential propagation in mitral cell lateral dendrites is decremental and controls recurrent and lateral inhibition in the mammalian olfactory bulb. Proc Natl Acad Sci U S A. 2001;98:319–324. [PMC free article] [PubMed]
  • Meister M, Bonhoeffer T. Tuning and topography in an odor map on the rat olfactory bulb. J Neurosci. 2001;21:1351–1360. [PubMed]
  • Mombaerts P, Wang F, Dulac C, Chao SK, Nemes A, Mendelsohn M, Edmondson J, Axel R. Visualizing an olfactory sensory map. Cell. 1996;87:675–686. [PubMed]
  • Onoda N, Mori K. Depth distribution of temporal firing patterns in olfactory bulb related to air-intake cycles. J Neurophysiol. 1980;44:29–39. [PubMed]
  • Ottoson D. Studies on the slow potentials in the rabbit's olfactory bulb and nasal mucosa. Acta Physiol Scand. 1959;47:136–148. [PubMed]
  • Perez-Orive J, Mazor O, Turner GC, Cassenaer S, Wilson RI, Laurent G. Oscillations and sparsening of odor representations in the mushroom body. Science. 2002;297:359–365. [PubMed]
  • Philpot BD, Foster TC, Brunjes PC. Mitral/tufted cell activity is attenuated and becomes uncoupled from respiration following naris closure. J Neurobiol. 1997;33:374–386. [PubMed]
  • Ramón y Cajal S. Histology of the Nervous System of Man and Vertebrates. New York: Oxford University Press; 2001.
  • Ravel N, Pager J. Respiratory patterning of the rat olfactory bulb unit activity: nasal versus tracheal breathing. Neurosci Lett. 1990;115:213–218. [PubMed]
  • Rubin BD, Katz LC. Optical imaging of odorant representations in the mammalian olfactory bulb. Neuron. 1999;23:499–511. [PubMed]
  • Schoppa NE, Westbrook GL. Glomerulus-specific synchronization of mitral cells in the olfactory bulb. Neuron. 2001;31:639–651. [PubMed]
  • Schoppa NE, Westbrook GL. AMPA autoreceptors drive correlated spiking in olfactory bulb glomeruli. Nat Neurosci. 2002;5:1194–1202. [PubMed]
  • Shen GY, Chen WR, Midtgaard J, Shepherd GM, Hines ML. Computational analysis of action potential initiation in mitral cell soma and dendrites based on dual patch recordings. J Neurophysiol. 1999;82:3006–3020. [PubMed]
  • Shepherd GM, Greer CA. Olfactory bulb. In: Shepherd GM, editor. Synaptic Organization of the Brain. New York: Oxford University Press; 1998. pp. 159–203.
  • Singer W, Gray CM. Visual feature integration and the temporal correlation hypothesis. Annu Rev Neurosci. 1995;18:555–586. [PubMed]
  • Sobel EC, Tank DW. Timing of odor stimulation does not alter patterning of olfactory bulb unit activity in freely breathing rats. J Neurophysiol. 1993;69:1331–1337. [PubMed]
  • Spors H, Grinvald A. Spatio-temporal dynamics of odor representations in the mammalian olfactory bulb. Neuron. 2002;34:301–315. [PubMed]
  • Uchida N, Takahashi YK, Tanifuji M, Mori K. Odor maps in the mammalian olfactory bulb: domain organization and odorant structural features. Nat Neurosci. 2000;3:1035–1043. [PubMed]
  • Urban NN, Sakmann B. Reciprocal intraglomerular excitation and intra- and interglomerular lateral inhibition between mouse olfactory bulb mitral cells. J Physiol. 2002;542:355–367. [PMC free article] [PubMed]
  • Wachowiak M, Cohen LB. Representation of odorants by receptor neuron input to the mouse olfactory bulb. Neuron. 2001;32:723–735. [PubMed]
  • Walsh RR. Single spike activity in the olfactory bulb. Am J Physiol. 1956;186:255–257. [PubMed]
  • White J, Dickinson TA, Walt DR, Kauer JS. An olfactory neuronal network for vapor recognition in an artificial nose. Biol Cybern. 1998;78:245–251. [PubMed]
  • White J, Hamilton KA, Neff SR, Kauer JS. Emergent properties of odor information coding in a representational model of the salamander olfactory bulb. J Neurosci. 1992;12:1772–1780. [PubMed]
  • Wilson DA. Habituation of odor responses in the rat anterior piriform cortex. J Neurophysiol. 1998a;79:1425–1440. [PubMed]
  • Wilson DA. Synaptic correlates of odor habituation in the rat anterior piriform cortex. J Neurophysiol. 1998b;80:998–1001. [PubMed]
  • Wilson DA. Receptive fields in the rat piriform cortex. Chem Senses. 2001;26:577–584. [PubMed]
  • Wilson M, Shepherd GM. Olfactory cortex. In: Arbib MA, editor. Handbook of Brain Theory and Neural Networks. London, UK: MIT Press; 1998. pp. 669–673.
  • Yokoi M, Mori K, Nakanishi S. Refinement of odor molecule tuning by dendrodendritic synaptic inhibition in the olfactory bulb. Proc Natl Acad Sci U S A. 1995;92:3371–3375. [PMC free article] [PubMed]
  • Zou Z, Horowitz LF, Montmayeur JP, Snapper S, Buck LB. Genetic tracing reveals a stereotyped sensory map in the olfactory cortex. Nature. 2001;414:173–179. [PubMed]

Articles from The Journal of Physiology are provided here courtesy of The Physiological Society
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • Cited in Books
    Cited in Books
    PubMed Central articles cited in books
  • PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...