Decoding The Projection Neuron Population Output

Connectivity

Antennal lobe projection neurons project to the mushroom body where they send distributed collateral projections throughout the calyx. In locust, each projection neuron was estimated to make over 600 (possibly as many as 2,000) output synapses (Perez-Orive et al. 2002). [More recent results (Jortner, Farivar, and Laurent, in preparation) indicate that connectivity between projection neurons and Kenyon cells is in fact a lot more widespread, but the reasoning presented below still applies.] Our early estimates come from anatomical analysis of projection neuron axonal projections and from electron microscopy of axonal presynaptic varicosities (Leitch and Laurent 1996). If we suppose that each pro-jectionneuron makes 1200 individual outputs on average, the projection neuron population makes a total of 1200 x 830 output synapses in the calyx. If those synapses are made onto all 50,000 Kenyon cells (i.e., if all Kenyon cells receive olfactory information), each Kenyon cell must be connected to 20 projection neurons on average (Figure 10.4). This estimate may vary across the Kenyon cell population. Our recent estimates, based on paired electrophysiological recordings, are that connectivity between projection neurons and Kenyon cells is, in fact, so dense that any projection neuron has about 50% chance of being connected with any one Kenyon cell (Jortner, Farivar, and Laurent, in preparation). The gain of any such synapse is, however, very small on average, and the firing threshold of a Kenyon cell requires the coactivation of 100-200 presynaptic projection neurons.

In addition, because projection neurons have widespread projection throughout the calyx, any Kenyon cell has access to most (if not all) projection neuron axons' terminals. This suggests that individual Kenyon cells could detect inputs across any possible combination of projection neurons. However, the number of possible combinations of 400 projection neurons among 830 is tremendously large (~1024°). Given that there are only 50,000 Kenyon cells, and if we assume that the distribution of connections is random (we do not know this to be true, but projection neuron axonal projections fail to reveal any gross bias), we conclude that those 50,000 realized combinations must be very different from one another. Herein may reside the key to sparseness and associativity among Kenyon cells. For any one odor, rare will be the Kenyon cells whose inputs

Varicosities Neuron

Figure 10.4 Connectivity between antennal lobe (AL) projection neurons (PNs) and mushroom body (MB) Kenyon cells (KCs). Recent results (Jortner, Farivar and Laurent, in preparation) suggest that each KC receives direct excitatory inputs from about 50% of all PNs on average. Similarly, individual PNs diverge to thousands of KCs. Each KC is thus characterized by the complement of PNs converging onto it and acts as a detector of the coactivation of a fraction of these PNs during each oscillation cycle. The threshold (in terms of number of coactive PNs) for each KC may vary across KCs and across time for each KC. In this illustration, KC3 reaches firing threshold; the others do not. The membrane potentials of all KCs, however, indicate periodic PN input, caused by the subsets of PNs presynaptic to each of them. (Intracellular data from Laurent and Naraghi 1994.) Simple combinatorics show that this distributed pattern of connections between PNs and KCs may in great part explain the KCs' high selectivity and, thus, the sparseness of odor representations in the MBs (see text for details).

Figure 10.4 Connectivity between antennal lobe (AL) projection neurons (PNs) and mushroom body (MB) Kenyon cells (KCs). Recent results (Jortner, Farivar and Laurent, in preparation) suggest that each KC receives direct excitatory inputs from about 50% of all PNs on average. Similarly, individual PNs diverge to thousands of KCs. Each KC is thus characterized by the complement of PNs converging onto it and acts as a detector of the coactivation of a fraction of these PNs during each oscillation cycle. The threshold (in terms of number of coactive PNs) for each KC may vary across KCs and across time for each KC. In this illustration, KC3 reaches firing threshold; the others do not. The membrane potentials of all KCs, however, indicate periodic PN input, caused by the subsets of PNs presynaptic to each of them. (Intracellular data from Laurent and Naraghi 1994.) Simple combinatorics show that this distributed pattern of connections between PNs and KCs may in great part explain the KCs' high selectivity and, thus, the sparseness of odor representations in the MBs (see text for details).

match precisely enough the projection neuron activity vectors occurring at each oscillation cycle. At present, the actual connection matrix is beyond experimental determination, although it is possibly within reach in Drosophila, which has ~ 200 projection neurons and 2500 Kenyon cells on each side of the brain. Figure 10.4 indicates some of the basic features that can explain Kenyon cell odor specificity in locust:

• Each Kenyon cell is connected to specific set of projection neurons (see above).

• Each Kenyon cell is silent at rest (Perez-Orive et al. 2002).

• Each Kenyon cell has a high firing threshold (Perez-Orive et al. 2002).

• Each Kenyon cell has a limited integration time constant (about half an oscillation cycle; Laurent andNaraghi 1994; Perez-Orive et al. 2002). This property is explained in Figure 10.5.

During an odor, each oscillation cycle is caused by a combination of active projection neurons. Each Kenyon cell, by virtue of its particular input set, detects projection neuron coincidence if enough of the projection neurons presynaptic to it are coactive during that cycle. For most Kenyon cells and cycles, too few presynaptic projection neurons are active, leading only to subthreshold responses. Kenyon cells spikes are extremely rare, but subthreshold membrane potential oscillations are always seen. Kenyon cells thus act as binary classifiers of projection neuron activity vectors: When a projection neuron activity pattern matches closely enough the projection neuron connection pattern of one Ken-yon cell, that Kenyon cell can reach threshold and signal the occurrence of that projection neuron combination. This selective detection is repeated at each oscillation cycle.

Integration Window

Critical to the Kenyon cell's specificity, and thus to the sparseness of odor representations in the mushroom bodies, is their short temporal integration window (about one half of each oscillation cycle). We know at least two different mechanisms that ensure the reset of each Kenyon cell before the start of a new oscillation cycle, thus curtailing temporal summation: feedforward inhibition and voltage-gated dendritic properties.

Delayed, Nonspecific, Feedforward Inhibition (Figure 10.5a)

In addition to sending distributed axonal collaterals to the mushroom body, all antennal lobe projection neuron axons terminate in the lateral horn; there, they contact a population of about 60 GABAergic interneurons, whose own axons terminate in the mushroom body calyx, providing feedforward inhibition to Kenyon cells. Because 830 projection neurons converge on a small population of lateral horn interneurons (LHI), LHI responses to odors are oscillatory, precisely locked to projection neuron output, but 180° out of phase with it (Perez-Orive et al. 2002). Because each LHI receives inputs from many projection neurons, LHI responses are poorly odor specific. This has important consequences for individual Kenyon cells: By virtue of its connections to only a fraction of all projection neurons, the depolarization of a Kenyon cell is critically dependent on the coactivation of many of its presynaptic projection neurons.

Was this article helpful?

0 0

Post a comment