Encyclopedia of Computational Neuroscience

Living Edition
| Editors: Dieter Jaeger, Ranu Jung

Olfactory Computation in Mitral-Granule Cell Circuits

  • Michele MiglioreEmail author
  • Tom McTavish
Living reference work entry
DOI: https://doi.org/10.1007/978-1-4614-7320-6_615-4



Olfactory computation in mitral-granule cell circuits refers those operations between mitral and granule cells that can directly modulate the mitral cell firing behavior. In vertebrates, they are carried out by the reciprocal synapses established between mitral cells’ lateral dendrites and spines on granule cell dendrites (Shepherd 2004).

Detailed Description

Mitral-granule cell circuits are formed by the continuous and dynamic reconfiguration of their synaptic connections during the entire life of the organism. Although, to date, there is no direct experimental evidence for plasticity of these synaptic connections, their sparse and distributed nature is well established (Willhite et al. 2006) and widely acknowledged as being the main mechanism responsible for the lateral inhibition effects observed in the olfactory bulb (Yokoi et al. 1995). This entry focuses on the circuital aspects of the mitral-granule reciprocal synapses, rather than on how they can be established or destroyed through LTP/LTD or cells’ turnover processes. All mitral-granule circuits are activated by a canonical sequence of steps, in which (a) an odor input facilitates one or more mitral cell somatic action potentials (APs); (b) these APs backpropagate along the lateral dendrites, releasing glutamate thereby activating granule cells along their way; and (c) granule cells locally inhibit the lateral dendrites of the mitral cells to which they are connected. The relative location of the activated granule cell with respect to the soma of a mitral cell can generate different circuits with different computational properties, such as (1) distance-independent lateral inhibition, (2) synchrony, and (3) gating effects.

Distance-Independent Lateral Inhibition

While granule cells connect to dendrodendritic synapses along the full span of the lateral dendrites, the inhibition from any granule cell onto the mitral cell is local. As such, those granule cells near to the mitral cell soma can more easily regulate its firing than those far away. The implication for one mitral cell to strongly laterally inhibit another is that the granule cell shared between the two mitral cells must be near to the soma of the target neuron. While inhibition is local, because of the broad field of granule cells, there is a population of granule cells that will be near to the body of any given mitral cell, and the circuit can ironically obtain distance-independent lateral inhibition. This is schematically shown in Fig. 1, where the synapses relevant to the circuit are highlighted in red (excitatory and inhibitory synapses are represented by closed and open circles, respectively). A somatic action potential elicited in M0 backpropagates (bAP) along the M0 lateral dendrites (in red). It activates both GCs (closed red circles), but the consequent inhibition (open red circles) on M1 (black trace) is mainly determined by GC1, which is close to the soma and independent from the location of M0, provided that the AP can backpropagate along the lateral dendrite.
Fig. 1

Distance-independent lateral inhibition. Inhibition can be independent of the relative distance between mitral cells (M0 and M1) if it is imposed locally by granule cells (GC0 and GC1) activated by a backpropagating action potential (bAP). In this simulation an action potential was generated on M0 (red trace). Its backpropagation activated GC1 (closed red circle on GC1), generating an inhibitory response (open red circle on GC1) close to the soma of M0 (black trace)


When two similar neurons receive a correlated signal, they can synchronize. If two mitral cells are near each other and share granule cells near their somas, they will mutually inhibit each other and synchronize. Similarly, if the two mitral cells in Fig. 1. were activated by an odor, they provide a correlated signal onto their shared granule cells. As such, the granule cells become synchronized and provide a correlated signal onto the mitral cells, which might be widely separated. Such synchrony is reliant on backpropagating action potential travelling the full span of the lateral dendrites. However, this may not always be the case. A typical circuit exhibiting this effect is shown in Fig. 2. The two mitral cells receive the same input, but starting at different times. At t = 500 ms the granule cells are activated and the mitral cells synchronize.
Fig. 2

Mitral cells synchronization. Two mitral cells (M0 and M1) were activated by the same input, but starting with a 20 ms relative latency. At t = 500 ms the granule cells (GC0 and GC1) were activated and the mitral cells synchronized (black and red traces). The reciprocal synapses involved in this circuit are shown in red, with excitation and inhibition represented by closed and open circles, respectively

Gating Effects

Gating effects occur when the backpropagation of action potentials is blocked by the local (dendritic) inhibition activated by other mitral cells. A typical example is illustrated in Fig. 3, with three mitral cells (M0–M2) and three granule cells (GC0–GC2). All mitral cells receive the same odor input with different latency, and only the mitral-granule synapses relevant to this circuit are highlighted in red (excitation and inhibition are represented by closed and open circles, respectively). The plots show membrane potential of M0–M2 soma and lateral dendrite of M0 at 800um from the soma during simulations with GC1 active or inactive. Note that the firing of M2 (blue traces) depends on GC1 local activation through M1 (black trace close to GC1 soma). In this case (top traces), the backpropagation of the APs along the M0 lateral dendrite (black traces) is blocked by GC1 activity, GC2 will not be activated, and M2 can fire APs. With GC1 inactive, APs from M0 can freely backpropagate until they activate GC2, inhibiting M2.
Fig. 3

Gating effects. Three mitral cells (M0M2) receive the same odor input with different latency. The plots show somatic potential of M0M2 (traces above the soma) and the membrane potential of the M0 lateral dendrite (in black) at 800 mm from the soma (traces indicated with dotted lines), during simulations with GC1 active or inactive. The reciprocal synapses involved in this circuit are shown in red, with excitation and inhibition represented by closed and open circles, respectively

Interactive simulations of the circuits illustrated in Figs. 1, 2, and 3 can be downloaded from the ModelDB section of the Senselab database suite (http://senselab.med.yale.edu/, accession number 149415).

Cross-References/Related Terms


  1. Shepherd GM (2004) The olfactory bulb. In: Shepherd GM (ed) The synaptic organization of the brain. Oxford University Press, New YorkCrossRefGoogle Scholar
  2. Willhite DC, Nguyen KT, Masurkar AV, Greer CA, Shepherd GM, Chen WR (2006) Viral tracing identifies distributed columnar organization in the olfactory bulb. Proc Natl Acad Sci USA 103:12592–12597PubMedCentralPubMedCrossRefGoogle Scholar
  3. Yokoi M, Mori K, Nakanishi S (1995) Refinement of odor molecule tuning by dendrodendritic synaptic inhibition in the olfactory bulb. Proc Natl Acad Sci USA 92:3371–3375PubMedCentralPubMedCrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media New York 2013

Authors and Affiliations

  1. 1.Institute of BiophysicsNational Research CouncilPalermoItaly
  2. 2.Department of Cell and Developmental BiologyUniversity of Colorado DenverAuroraUSA