OLFACTORY CODING AND DECODING BY ENSEMBLES OF NEURONS
, Head, Unit on Sensory Coding and Neural Ensembles
Baranidharan Raman, PhD, Postdoctoral Fellow
Rebecca Vislay-Meltzer, PhD, Postdoctoral Fellow
Iori Ito, PhD, Visiting Fellow
Joby Joseph, PhD, Visiting Fellow
Nobuaki Tanaka, PhD, Visiting Fellow
Stacey Brown Daffron, MS, Technician
Kui Sun, MD, Technician
Rose Chik Ying Ong, MS, Graduate Student
Kelly Chang, College Student
Several mechanisms extract features from natural odor stimuli
Our previous work suggested that neural plasticity is one of the general mechanisms responsible for the several dramatic transformations of information as it moves through brain. To investigate mechanisms underlying the transformations, we delivered repeated, rapid pulses of odors timed to mimic features of natural plumes. At the same time, we monitored, in intact animals, neural activity in several locations: olfactory receptor neurons; ensembles of projection and local first-order interneurons of the antennal lobe (analogous to the olfactory bulb); and second-order Kenyon cells of the mushroom body (analogous to the pyriform cortex). At each location, we sought to understand responses in terms of the interactions of plasticity occurring at earlier sites. We also sought to understand the potential value to the animal of these restructurings.
We found that interneuronal responses to natural forms of odor stimuli are shaped by at least two plastic mechanisms: rapid adaptation in the receptors and relatively enduring facilitation of inhibition within the receptors’ downstream targets. Peripheral adaptation renders the olfactory system relatively insensitive to stimuli that repeat rapidly. Central facilitation of inhibition increases the reliability and sparseness of stimuli that are encountered repeatedly but relatively infrequently. Further, the plastic mechanisms constrain the projection neuron ensemble to provide relatively stable output to its downstream followers, thereby allowing the encoding of information about odor identity and concentration with firing patterns that are not confounded by the timing patterns of the stimulus.
How do the follower neurons decode this time-varying ensemble activity? Intracellular and extracellular recordings from Kenyon cells showed that the cells’ firing rates change dramatically throughout trains of odor pulses in a timing-dependent manner; for brief inter-pulse intervals, the great majority of action potentials fire at the beginning of the train and again at the train’s conclusion. We found that the Kenyon cell firing threshold can be met when projection neurons fire at relatively low rates but are highly synchronized by the oscillatory mechanism of the antennal lobe (as occurs during the onset of the pulse train). On the other hand, the threshold can be met when the instantaneous firing rate of the projection neuron ensemble is high in the absence of pronounced synchronization (as occurs following the offset of the train). Taken together, our work suggests that the nonassociative plasticity elicited by odor plumes leads to responses in the projection neuron ensemble that combine an instantaneous report of sensory input with a record of recent input, allowing the extraction of high-level features.
Brown SL, Joseph J, Stopfer M. Encoding a temporally structured stimulus with a temporally structured neural representation. Nat Neurosci 2005;8:1568-76.
Control of neural oscillation frequency
In many neural systems, oscillations coordinate the activities of populations of neurons. By applying new techniques, we found that a wide variety of odors evoke oscillations in the olfactory system of the moth Manduca sexta, which is a useful model system for such studies. In examining the mechanisms underlying neural synchronous oscillations, we found that odor-elicited action potentials in antennal lobe neurons were tightly phase-locked to the local field potential oscillations recorded in the mushroom body and that the same antennal lobe neurons were phase-locked during oscillations. Further, intracellular recordings from the interneurons showed that subthreshold membrane potential oscillations were highly correlated with the local field potential. We demonstrated that local injection of picrotoxin into the antennal lobe blocked GABAergic inhibition from local neurons onto projection neurons and reversibly abolished oscillations in the local field potential; local saline injections had no such effect. Our results show that, in the moth, odor stimuli lead to the coordinated, oscillatory synchronous firings of projection neurons in the antennal lobe and that this coordinated activity is transmitted to a downstream region—the mushroom body.
In natural settings, as when moths encounter odors while feeding, odor exposure typically endures for several seconds. We found that, over the course of a several-second odor pulse, the frequency of odor-evoked oscillations slowed dramatically from about 40Hz to about 20Hz. We are using a computational model to address how this frequency shift is achieved and what it might mean for understanding oscillatory circuitry in general and for odor coding.
Time-varying neural codes for sensory stimuli
Accumulating evidence suggests that, in many animals from insects to mammals, olfactory information is represented by the temporally structured synchronized firing of a spatially distributed population of neurons in the olfactory system. Using Manduca sexta, which offers experimental advantages for both electrophysiological and behavioral testing, we investigated the information content of time-varying neural codes for odors. We found that moth Kenyon cells responded to odor stimuli with sparse spiking and that spikes were highly phase-locked to local field potential oscillations, indicating that Kenyon cells are influenced by the oscillatory timing machinery of the antennal lobe and favor synchronous inputs from projection neurons. In the moth, odor representations in the antennal lobe are relatively dense and consist of sustained, lengthy, and bursty trains of spikes in projection neurons; the time-varying spiking patterns change with the odorant. In the mushroom body, odor representations have become sparse and consist of a few, rare spikes in Kenyon cells that fire mainly after the onset and offset of the odor. Some Kenyon cells showed both on and off responses to odor pulses, often in an odor-specific fashion, but most showed only on or off responses, and off responses followed long rather than brief stimuli. Thus, the neural representation of the odor stimulus in the mushroom body varied over time.
Our results suggested that different information was available to followers of the Kenyon cells at different time points. However, understanding the information content of these evolving patterns requires an evaluation of behaviors that could be used to assess perception. Thus, using proboscis extension, a natural feeding behavior in the moth, we examined the correlation between time-varying neural responses and perception. We found that time-varying responses in the moth antennal lobe and mushroom body, generated by fast neural oscillatory synchronization and slow, distributed temporal firing patterns, give rise to perceptions that vary with the changing neural representation. Our results have several implications. As suggested by the statistical analyses of neural firing patterns, a single sensory stimulus can give rise to a sequence of evolving neural representations that are not serially redundant; different time points carry distinct information. Our behavioral results show that higher processing levels in the brain have access to separate epochs of the evolving representations and make use of them when preparing actions and forming memories.
Behavioral and physiological development of olfactory processing
How do the sensory capacities of animals develop? Through their innate sensory preferences, animals often demonstrate the existence of inborn information. How is this information encoded, and how does it differ from information acquired through direct experience? We found that newly hatched locusts, literally crawling out of their hatching cups, immediately move toward fresh grass. In a series of behavioral studies with thousands of locusts, we established that the hatchlings choose real grass over visually similar but odorless plastic grass; that they choose paper rubbed with fresh grass over clean paper of the same color; and that they choose paper dabbed with colorless monomolecular odorants that are components of grass odor over paper with other colorless odorants (even when the odorants have been diluted to identical vapor pressures). The results indicate that naive locusts, which had never eaten, touched, or otherwise encountered their natural food source, have a built-in preference for their food’s odor.
This preference might be attributable to a peripheral mechanism; for example, perhaps hatchling locusts have a surplus of odor receptors for grass odors. We tested such a hypothesis by making electroantennograms from hatchlings and found that the antennae respond best to grass odors and less strongly to an assortment of non-plant odors (as do the antennae of adult locusts, as we found). The same held true when we diluted the odors to provide equal vapor pressure. We also found that, in hatchling antennae, sensory adaptation occurs for grass odors with the same timing and extent as for non-plant odors. The results suggest that the behaviorally demonstrated innate preference for grass odors is determined, at least in part, by the prevalence of olfactory receptors for them. The prevalence is in place at hatching and persists into adulthood.
In most respects, odor responses of the hatchling brain resemble those of the adult brain. Recording local field potentials in the mushroom bodies of hatchlings, we found that the oscillatory synchronization mechanism is already intact (although the oscillation frequency is significantly slower than in the adult). Recording intracellularly from antennal lobe neurons, we found odor-specific temporal patterns in the distributed responses of projection neurons. The responses appeared similar to those of adults, although less complex; response patterns were shorter in duration and contained fewer alternating excitatory and inhibitory components. We found no evidence in the antennal lobe for specialized neurons responding only to grass odors. Thus, we concluded that representations of grass and other odors in the hatchling brain, as in the adult brain, are broadly and spatiotemporally distributed.
Odor-evoked oscillations in Drosophila
Transient oscillatory synchronization of ensembles of neurons is a common feature of sensory coding. Drosophila offers several experimental advantages for analyzing neural circuitry underlying complex responses. We found that, as in many other species, odors can evoke neural oscillations in Drosophila. Using new intracellular and extracellular recording techniques as well as several types of genetic manipulation, we are investigating the neural circuitry responsible for generating these oscillations.
We found that, as in locust and moth, odor-elicited oscillations originate in the circuitry of Drosophila’s antennal lobe. Subthreshold oscillations were evident in the membrane potentials of both excitatory projection neurons and two classes of inhibitory local neuron; extracellular local field potential oscillations were detectable in the mushroom bodies, which are projection sites for antennal lobe output cells. Spikes in both projection and local neurons were tightly phase-locked to the field potential oscillations. Bath application of picrotoxin, a blocker of GABAergic transmission, reversibly abolished the oscillations. More specifically, transiently blocking transmission from only one type of local neuron by means of a temperature-sensitive genetic mutation also reversibly blocked the oscillations. Thus, Drosophila appears to make use of a neural coding mechanism similar to that of other insects. The availability of genetic techniques in Drosophila for fluorescently labeling and transiently inactivating specific types of neurons (and even specific neurons) allows unprecedented resolution of the mechanisms underlying the coordination of ensembles of neurons involved in sensory coding.
Contributions of olfactory receptor neuron response dynamics to spatiotemporal sensory coding
Groups of projection neurons in the antennal lobe, like mitral cells of the olfactory bulb, respond to odorants with spatially distributed, temporally complex firing patterns that change with and thus contain information about the odorants that elicit them. We are interested in the neural mechanisms that generate these patterns of early-stage odor representations. Several lines of evidence suggest that interactions between the excitatory projection neurons and inhibitory local neurons are at least partly responsible for the patterns, although computational models incorporating the interactions were unable to reproduce the full complexity of projection neuron firing patterns.
Our recordings from locust olfactory receptor neurons revealed responses to odors that varied significantly in temporal structure. We constructed a computational model of the locust antennal lobe that included complex heterogeneous input from receptors modeled after those we observed in vivo. Compared with a model with homogenous inputs, our model generated more realistic, temporally complex spatiotemporal response patterns, particularly when presented with several rapid stimuli that simulated odor plumes. Thus, heterogeneous, temporally structured input from olfactory receptor neurons appears to contribute significantly to the spatiotemporal complexity of the projection neuron ensemble odor code.
The diverse, time-varying olfactory receptor outputs of the model could be represented as trajectories that showed many of the odor- and concentration-specific characteristics of similar trajectories constructed from projection neuron responses. However, classification analyses revealed that, unlike projection neuron responses, the receptor neuron responses do not decorrelate over time. Thus, decorrelation, which allows for precise identification of similar odor stimuli, appears to be a specific function of the antennal lobe circuitry.
Behavioral and electrophysiological analysis of innate olfactory information
Like many animals, moths rely on innate information to locate their first food sources. We sought to examine the abilities of animals to modify responses to odorants for which they appear to have built-in biases. We tested a set of odorants that included plant volatiles known to attract moths and a set of odorants not found in plants. We found that the plant odors elicited stronger responses from odor receptors. Animals generally learn more readily about strong, salient stimuli versus weak stimuli. To focus our analysis on the role of downstream odor-processing steps in learning, we devised a way to normalize the olfactory input intensities of the odor set by reference to electroantennogram (EAG) responses from the antennae of moths. With the normalized odor set, we trained moths in a classical conditioning paradigm in which they associatively learned that an odor predicts a food reward. We found no significant differences in learning rate among the normalized odors. Further, all normalized odors evoked similar responses recorded in several brain areas. Together, these results suggest that the innate preference for plant odors is encoded as differential expression of odor receptors rather than within the central nervous system. Further, rapid learning about plant volatiles appears to result from the strong signals the volatiles elicit from peripheral odor receptors.
Adaptive regulation of sparseness by feedforward inhibition
Our sensory environment is in near constant flux. Brain sensory systems have evolved the means to adjust their coding properties to adapt to constantly changing signals arising in sensory neurons. Ideally, a coding strategy used by a sensory system should provide efficient representations across the full possible range of stimulation conditions. For the olfactory system, this task involves the encoding of odor intensities, an ability critical for survival in many species. In the mushroom body of insects, odors are represented by very few spikes in a small number of neurons in what is a highly efficient strategy known as sparse coding. Physiological studies of the neurons show that sparseness is maintained across 1,000-fold changes in odor concentration. Using a realistic computational model, we propose that sparseness in the olfactory system is regulated by adaptive feedforward inhibition. When odor concentration changes, feedforward inhibition modulates the duration of the temporal window over which the mushroom body neurons may integrate excitatory presynaptic input. This simple adaptive mechanism can maintain the sparseness of sensory representations across wide ranges of stimulus conditions.
Assisi C, Stopfer M, Laurent G, Bazhenov M. Adaptive regulation of sparseness by feedforward inhibition. Nat Neurosci 2007;10:1176-84.
1 April Chiriboga, BS, former Technician
2 Jacklyn Feldman, former High School Student
3 Jeffrey Tang, former College Student
COLLABORATOR
Collins Assisi, PhD, The Salk Institute for Biological Studies, La Jolla, CA
Maxim Bazhenov, PhD, The Salk Institute for Biological Studies, La Jolla, CA
Kei Ito, PhD, Institute of Molecular and Cellular Biosciences, University of Tokyo, Tokyo, Japan
Gilles Laurent, DVM, PhD, California Institute of Technology, Pasadena, CA
For further information, contact stopferm@mail.nih.gov.
