This is a preprint of a chapterpublished in Developmental Psychobiology, Developmental Neurobiology andBehavioral Ecology: Mechanisms and Early Principals. (Blass, E.M., ed.) pp.53-80, Kluwer Academic/Plenum Publishers, New York.

Spatial Coding inthe Olfactory System: The Role of Early Experience

Brett A. Johnsonand Michael Leon

Department ofNeurobiology and Behavior

School ofBiological Sciences

Room 2205 BS II

The University ofCalifornia, Irvine

Irvine, CA92697-4550

Supported by grant DC03545

I. Spatiallyspecific alterations in olfactory bulb function and structure following earlyodor learning

II. Spatialrepresentations of chemical cues

III. A simplemodel of odorant molecular feature processing

IV. Summary

V. References

Spatially specific alterations in olfactory bulb functionand structure following early odor learning

Infant rats are born witha functional olfactory system (Guthrie and Gall, 1999). Within the first days of their lifethey begin to approach the odor of their mother in preference to the odor of avirgin female (Leon and Moltz, 1971). These preferences can be seen when thepups are placed in an apparatus designed to allow them to approach one of twoareas on the basis of odor cues alone. Such a preference also can be inducedwhen the natural situation is mimicked experimentally by pairing a non-maternalodor (such as peppermint extract) with tactile stimulation of the kind that amother might impose on her pups (Coopersmith and Leon, 1984).  These data indicate that pups acquiretheir preference for the mother's odor postnatally, rather than being born withthat ability. We went on tofind that the individuality of the odor of one mother compared to another wasdue to differences in their diet; mothers with identical diets were equallyapproached by their pups (Leon, 1975).

Learning under natural circumstances seemed likely to be thekind of behavioral change that should induce large changes in the relativelysimple brain of developing rats, large enough to visualize through relativelystraightforward analyses. Even small changes in the brain induced by the kindof experience described above would be magnified as the brain developed,allowing us to visualize changes that are difficult to see in the brains ofexperienced adults. We also wantedto capitalize on the fact that the responses of the olfactory system tospecific odorants is highly localized at one level of their processing, therebyallowing us to focus our initial efforts on a very small area in the brain.

Place Figure 1 abouthere

As can be seen in Figure 1, mammalian olfactory codingstarts with the binding of airborne chemicals to receptors located on theolfactory receptor neurons deep within the nasal turbinates.  This induces a cascade of events in theolfactory receptor neuron that eventually causes it to become active.  Thousands of such neurons withidentical chemical receptors converge in the olfactory bulb and the site ofthis convergence is the glomeruli that ring the outer lamina of the bulb(Ressler et al., 1994; Vassar et al., 1994; Mombaerts et al., 1996). Within theglomeruli, the axons of homologous olfactory receptor neurons synapse with thedendrites of the second set of neurons in the olfactory pathway and it iswithin these glomeruli that localization of activity in response to differentodorants occurs and can be observed through contemporary neuroscienceprocesses. Specifically, whendifferent odorants are presented to rats, different olfactory glomeruli areactivated: activation patterns for individual odorants are consistent amonganimals (Stewart et al., 1977; Coopersmith et al., 1986; Royet et al., 1987;Bell et al., 1987; Guthrie et al., 1993). These specific activation patterns were first seen with a technique inwhich rats were injected with tracer amounts of a radiolabelled glucoseanalogue ([¹⁴C] 2-deoxyglucose) before exposure to an odorant.  Unlike glucose, this analogue isincompletely metabolized, leaving its radioactively tagged product in thecell. The assumption is thatincreased neural activity increases the use of glucose in brain cells andactive cells will accumulate increased amount of the radiolabelled product.  The density of the radiolabel insections of the olfactory bulb reveals localized specific responses todifferent odorants and refined image analysis with respect to radioactivestandards allow differences in glomerular activity to be quantified.

In our initial investigation of the neurobiologicalcorrelates of olfactory learning, we wanted to determine whether the uptake of[¹⁴C] 2-deoxyglucose (2-DG) in the glomerular layer of the olfactorybulb occurred differentially in response to learned and control odorants(Coopersmith and Leon, 1984). The volatile components of peppermint extractevoked multiple foci of 2‑DG uptake that were reliably associated withregional glomerular activity even in control rats. After early olfactory preference training, however, thefocal glomerular response to peppermint odor was elevated relative to theresponse of pups that previously received either repeated presentations of odoror tactile stimulation alone, or unpaired presentations of odor, or tactilestimulation alone (Coopersmith and Leon, 1984; Sullivan and Leon, 1986;Sullivan et al., 1989). Only thepairing of odor and tactile stimulation induced both a behavioral preferenceand an enhanced uptake of radiolabelled 2-DG in response to subsequentpresentations of the training odor. Moreover, a variety of stimuli that are associated with the mother, suchas oral infusion of milk, can be paired with the odor to induce both a behaviorpreference and an increase in focal 2‑DG uptake in the olfactory bulbglomerular layer (Sullivan and Leon, 1986; Do et al., 1988; Sullivan et al.,1990). The enhanced 2‑DG uptake induced by early learning persisted intoadulthood (Coopersmith and Leon, 1986).

One possibility for the increase in the glomerular responseis that the pups simply increase their respiration in the presence of an odorthat they have come to like. Theincrease in the resultant amount of the training odor that reaches the nasalepithelium would therefore be expected to increase, perhaps driving an increasein the glomerular response. However, when we monitored the respiration patterns of trained pups inthe presence of the training odorant, we found no difference in such breathingpatterns (Coopersmith and Leon, 1984; Coopersmith et al., 1986), suggestingthat there were no increases in the amount of odorant reaching the olfactorysensory neurons. It was possible that we did not identify a special sniffingpattern that could underlie the differential glomerular response in the absenceof a simple increase in respiration. Therefore, we decided to determine whether the glomerular response woulddiffer between trained and control pups after we imposed the same number ofidentical "sniffs" on them (Sullivan, et al., 1988). Anesthetizedpups were tested for their response to the trained odorant after they had beenplaced on a respirator and the odorant was pulled through their nares withanother respirator. Only thosepups that had been trained to prefer the odor had an increase in 2-DG followingidentical stimulus exposure.

Central changes caused by early preference training were notconfined to the glomeruli. c-Fos expression also specifically increased inperiglomerular cells surrounding the glomerular foci with enhanced 2-DG uptake(Johnson et al., 1995). Mitral ortufted cells in the region of the bulb deep to the foci of high 2‑DGuptake displayed an increased proportion of inhibitory responses, measuredelectrophysiologically, following early preference training (Wilson et al.,1987). Furthermore, learningdecreased the c-Fos response of inhibitory granule cell interneurons locateddeep to the glomerular 2-DG foci (Woo et al., 1996). Granule cells are responsible for lateral inhibition withinthe bulb by way of their reciprocal dendrodendritic connections with excitatorymitral and tufted cells. Together,these data suggest that neurobiological changes localized within the glomerularlayer engage a circuit that that modulates at least one output response of theolfactory bulb.

To investigate possible structural changes that couldunderlie such a long-lasting change in response, we measured the width of theglomerular layer in sections containing 2-DG foci. Trained animals had a wider glomerular layer in focalresponse regions than did control animals (Woo et al., 1987).  This increased width included both anincrease in the diameter of glomeruli and an increase in the number ofperiglomerular cells surrounding the active glomeruli (Woo and Leon,1991). The density of glial cellprocesses within active glomeruli also increased in trained animals (Matsutaniand Leon, 1993). In contrast, thedensity of mitral cells and granule cells in areas deep to the alteredglomeruli did not change with learning (McCollum et al., 1997).

Learning-induced changes in the olfactory bulb appear to belargely confined to the areas showing focal 2-DG uptake. Measurements ofglomerular width in non-focal regions taken systematically with respect tohigh-uptake foci revealed that experience-dependent increases in width werefound only in the regions of high uptake (Woo et al., 1987).  The odorant cyclohexanone (with apowerful, solvent-like odor) evoked 2-DG uptake in posterior, medial glomeruli,and early preference training with cyclohexanone enhanced 2-DG uptake in theseglomeruli (Coopersmith et al., 1986). Glomerular 2-DG uptake did not increase in thesecyclohexanone-responsive regions following training with peppermint odor, andin animals trained with cyclohexanone, 2-DG uptake did not increase in focievoked by peppermint extract (Coopersmith et al., 1986). Similarly, changes inthe activity of mitral cells following odor preference learning occurred in thelateral bulb, where foci of high 2‑DG uptake are observed, but suchchanges were not found in a region of the bulb remote from these foci (Wilsonand Leon, 1988). This specificityimpresses us. It implies that anatomically and functionally distinct regions ofthe bulb have access to motivational and motoric systems that are engaged whenanimals make choices or engage in acts (Holmes, this volume) that reflect theirparticular early olfactory experiences.

Within the portion of the lateral bulb where 2-DG uptake isstimulated by peppermint odor, learning changed the Fos-like response ofperiglomerular cells associated with midlateral 2-DG foci, but learning did notchange the Fos-like response in ventrolateral 2-DG foci (Johnson et al., 1995).These findings of regional heterogeneity in the modification of the bulb byolfactory experience prompted us to explore more intensively the spatialdistribution of experience-dependent changes in 2-DG uptake throughout theglomerular layer. To the extent that the regions that we are exploring arerepresentative of others in the bulb, then the following analyses may serve asa model for future regional studies.

Important to an analysis of the glomerular code is to have atechnique that is capable of resolving activity of individual glomeruli, suchas the uptake of radiolabelled 2-DG. To characterize at this level of resolution the spatial distribution of activity evoked by peppermintextract, we generated maps of 2‑DG uptake across the glomerular layer andaveraged the maps obtained for individual trained and control pups.  The maps of the entire glomerular layerthroughout the bulb reveal the entire pattern of activity that each odorantevokes. Since the maps are essentially data matrices, this technique alsoallows us to average individual maps to obtain a group response that can becompared statistically to the responses of other group responses.  Naive animals increased activity topeppermint extract in four circumscribed regions in the posterior half of thelateral glomerular layer (Johnson and Leon, 1996). Within this general region training did not cause a largechange in the spatial pattern of 2-DG uptake (Johnson and Leon, 1996).  There were circumscribed regionalchanges, however. Three midlateralfields of high uptake in the bulb were markedly enhanced. In contrast, the ventrolateralfield of peppermint-evoked 2-DG uptake was unchanged after the training(Johnson and Leon, 1996). This pattern of midlateral enhancement andventrolateral stability parallels that which we had seen previously forFos-like responses (Johnson et al., 1995). Differences between the learning-dependent changes inmidlateral versus ventrolateral peppermint-activated glomeruli were not limitedto these measures of neural activity. We also found that training decreased ß‑adrenergic receptorbinding in the midlateral glomerular layer, while ventrolateral ß ‑adrenergicreceptors were unaffected (Woo and Leon, 1995b). It may be that the medial glomeruli convey differentinformation regarding specific odorants than lateral responses.  The lateral representation may containadditional information regarding the odorant that includes affective or experientialinformation in addition to the normal information carried by the medialrepresentation. While differentialprojections of medial and lateral bulb have not been well studied, there areseparate lateral and medial olfactory tracts in lower vertebrates that projectto distinct forebrain targets where the different kinds of information may beput to different uses (Eisthen, 1997).

Given that odor learning can cause spatially specificincreases in focal glomerular activity, in local glomerular structure, andleads to specific behavioral change, it is important to understand howindividual glomeruli contribute to odor processing. This knowledge would provide a framework in which tointerpret the changes brought about by experience. It has long been known that different odorants activatedifferent, but overlapping, areas within the glomerular layer of mammalianolfactory bulbs (Stewart et al., 1979; Jourdan et al., 1980; Bell et al., 1987;Royet et al., 1987; Shepherd, 1991; Guthrie et al., 1993).  It now is appreciated that theactivation of specific glomeruli by particular odorants is a phenomenon thatgeneralizes across a number of species (Cinelli et al., 1995; Friedrich andKorsching, 1997; Joerges et al., 1997; Galizia et al., 1999a; Galizia et al.,1999b). Only recently, however,have we begun to understand how spatial distributions of glomerular activitymight be involved in the processing of volatile chemical cues.  We now review recent developments inthe interpretation of glomerular activity patterns. Then, we will address how these developments may relate tothe learning-dependent changes we have observed in the olfactory bulb.

Spatial representations of chemical cues

Understanding the code through which different odors arerepresented in the nervous system is basic to understanding olfactoryneurobiology. Since each of theglomeruli receives input from a set of olfactory receptor neurons with a singletype of receptor protein, the differential activation of glomeruli would be anideal place to use differential activity to try to understand what the systemis coding. Yet the evidence that individual odorants generate characteristicspatial patterns of activity has generally been ignored. Most models ofolfactory coding have favored a low-specificity, broadly tuned system withwidely distributed neural connections. Indeed, there has been little emphasis on specificity at anyanatomical/functional level of the olfactory brain that has been presented inmodels of coding. Unlike othersensory systems, spatial information was considered to contribute little tochemical coding. Rather,olfactory information has been thought to been extracted from a distributedexcitation of the system. Such a system would likely have few receptor types,but those receptors would be expected to respond to a wide variety of chemicalsthat exist in nature. For example,electrophysiological recordings of isolated salamander olfactory receptorneurons indicated that most of these neurons respond to most odorants (Firesteinet al., 1993). Subsequentprocessing of this broadly tuned response would then allow the signalrepresenting the chemicals to be decoded. In addition, all of the "electronic noses"developed to recognize airborne chemicals have utilized electrochemical sensorsthat respond to a broad range of odorants. Subsequent computational processing of the information isused to allow the electronic systems to sense and discriminate odorants  (Dickinson et al., 1998).

Olfactory receptor neurons synapse with olfactory bulb cellswithin the densely synaptic glomeruli. These connections have been reported to be imprecise in nature, withmany of the connections demonstrating no obvious organization (Astic and Saucier,1986; Kauer, 1987). The responseof the glomerular layer to different odorants also seemed to be nonspecific,because different odorants appeared to stimulate much of the glomerular layerin salamander olfactory bulbs (Kauer and Cinelli, 1993; Cinelli et al., 1995).

Moreover, mitral cells, the dominant output neuron emanatingfrom glomeruli and projecting to the olfactory cortex, were reported to have anonspecific response to odorant stimulation. Motokizawa (1996), who recordedfrom mitral cells in rats, found very few differences in their responsepatterns to different odorants. Mitral cells were broadly tuned; they responded to a wide range ofairborne chemicals. Similarconclusions were reached using EEG recordings of the bulb to monitor responsesto odorants in rabbits (Freeman and Skarda, 1985).

Finally, mitral cell projections to the olfactory cortex inrats appeared to be broadly distributed (Haberly and Price, 1977).  Similarly, responses to odorants in theolfactory cortex were difficult to correlate with odorant stimulation, suggestinga random, distributed representation of information regarding olfactory cues inthe olfactory cortex (Haberly and Bower, 1989). In all, the entire system appeared to be broadly tuned, withcoding and decoding accomplished without using spatial arrangements ofprojections. Rather, complexcomputational methods were thought to accomplish odor coding and decoding(Alkasab, et al., 1999).

The lack of spatial specificity of the system was furtheremphasized by complementary lesion studies in which large parts of theolfactory bulb could be removed without affecting the ability of an animal todetect and discriminate a variety of odorants (Slotnick et al., 1987; 1997; Luand Slotnick, 1994; 1998). Evenwhen these lesions targeted specific focal areas of 2-DG uptake evoked bypropionic acid, lesioned animals detected the odorant with high sensitivity(Slotnick et al., 1997) and responded as if the perceived odor of the chemicalhad been unaltered by the lesion (Lu and Slotnick, 1994).  Preservation of function followingremoval of identified 2‑DG foci led to the suggestion that these focalresponses might not be important for odor processing. Together, these multiple approaches to olfactory functionpointed to broadly distributed networks of low specificity.

The discovery of a superfamily of putative olfactoryreceptor genes that are expressed by sensory neurons in the mammalian olfactoryepithelium (Buck and Axel, 1991) began to change the perception of theolfactory system from a low-specificity to a high-specificity system.  Mammals appear to have a singleolfactory receptor type expressed in each olfactory sensory neuron, therebyconferring an unexpected, but exquisite specificity to the system at its outset(Chess et al., 1994; Malnic et al., 1999). Indeed, recent functional studies indicate that responses ofindividual receptor proteins to specific odorants are narrowly tuned with ahigh degree of ligand specificity (Zhao et al., 1998; Krautwurst et al., 1998).

The response of individual rodent olfactory sensory neuronsseems to reflect the specificity of the receptor proteins they express (Malnicet al., 1999). Sato et al. (1994) recorded the responses of a subpopulation ofreceptor neurons in deeply anesthetized rabbits, and found highly specific responsepatterns to low concentrations of chemically related odorants.  Thus each olfactory receptor neuronseems to relay specific odorant information to the olfactory bulb.

Place Figure 2 abouthere

If the olfactory receptor neurons distribute widely to variousparts of the bulb as originally believed, then one would imagine that thisspecificity would be lost at that next level of processing.  Figure 2 illustrates the findings ofMombaerts et al. (1996a), who identified individual olfactory receptor neuronsexpressing the same olfactory receptor gene in mice and found that all of theiraxons converged on as few as two glomeruli in each olfactory bulb, one lateraland one medial. The medial projection area was located more caudally andventrally than the lateral projection (Ressler et al., 1994; Vassar et al.,1994; Sullivan and Dyer, 1996; Mombaerts et al., 1996a), and these identifiedglomeruli are consistent in their location across animals (Ressler et al.,1994; Vassar et al., 1994; Mombaerts et al., 1996a). Accordingly, there appears to be a high degree ofspecificity in the organization of information as it arrives in the olfactorybulb, with respect to both the tuning of individual receptors and theprojections of receptor neurons. The convergence of olfactory information into a small number ofglomeruli constitutes a system that is ideally suited to amplify the olfactorysignal. In such a system, manythousands of receptor neurons carrying the same information stimulate a smallnumber of mitral cells. Thespatial specificity of the organization of the mammalian olfactory system iscarried through to the level of the mitral cells because each mitral cellreceives afferent input only from a primary dendrite that extends into a singleglomerulus. Moreover, the responsepattern of neighboring mitral cells resembles that of each other more closelythan that of distant mitral cells (Buonviso and Chaput, 1990), thereby furtherenhancing the regional efferent signal towards the control of processes influencedby early olfactory experience.

Because the projections of homologous olfactory receptorneurons to the bulb appear to be specific to a small number of glomeruli, itseems possible to form a map of the representation of all possible odorants byassessing with high resolution the responses of glomeruli to differentodorants. We then could seewhether the functional response matched the anatomical specificity in thesystem. The vast number ofodorants to which the system is responsive, however, complicates thispossibility. Estimates of thenumber of possible odorants range from a low of 10,000 to hundreds ofthousands. If the number ofodorants matched the number of putative odorant receptors, then it would bepossible to predict that each receptor could bind one odorant maximally andthen communicate this information to the olfactory cortex.  However, the number of chemicals withdistinct odors is at least ten times the number of odorant receptors.  Therefore, it is unlikely that eachodorant is bound to a single olfactory receptor that recognizes principallythat chemical. It is far morelikely that most olfactory receptor proteins function in the same way as doother receptors, by binding to and detecting specific molecular features(Figure 3; Buck and Axel, 1991; Imamura et al., 1992; Mori et al., 1992; Katohet al., 1993; Sato et al., 1994). The idea that olfactory receptors are molecular feature detectorsfollows from studies of the binding of neurotransmitters and pharmacologicalagents to their receptors, for which specific molecular features determine thebinding affinity (Dean, 1987).

Place Figure 3 about here

The change in the conception of how olfactory systems areorganized, at least in mammals, has been dramatic. The dominant view of itsorganization has changed from a broadly tuned, randomly organized system withwidely distributed activity, to that of a narrowly tuned system with a highdegree of spatial specificity of neural activity playing a key role in thecoding of chemical cues. We willfirst show that the coding at the level of the glomeruli is quite specific andlinked to specific aspects of the airborne molecules that are perceived asodors.

Airborne chemicals are probably represented by a combinationof either shared or unique features of odorant molecules, termed"primitives," "odotopes," or "epitopes"(Shepherd, 1991; 1994; Buck, 1996), that would allow discrimination amongodorants (Figure 3). Sinceseveral odorants would be expected to activate some of the same receptors, eachodor would be coded by the particular combination of receptors activated by thefeatures of that odorant, as illustrated in Figure 4. This concept is the foundation for recent"combinatorial" proposals for the mechanism underlying olfactory coding(Buck, 1996; Friedrich and Korsching, 1997; Johnson et al., 1998; Kauer andCinelli, 1993; Malnic et al., 1999; Ressler et al., 1994; Shepherd, 1991, 1994;Vassar et al., 1994; Vickers and Christensen, 1998).

Place Figure 4 about here

How would a combinatorial code of odor quality be relayedfrom the olfactory receptor proteins to the central nervous system?  Recall that axons from homologousolfactory receptor neurons converge on to a small number of glomeruli in themain olfactory bulb (Ressler et al., 1994; Vassar et al., 1994; Mombaerts etal., 1996a). Given thisinformation, the pattern of glomerular activation should be determined by thespecific olfactory receptor proteins that recognize distinct features of anodorant molecule. Thus, each pureodorant should generate a characteristic spatial pattern of glomerularactivation. Figures 3 and 4 showsthat any two odorants sharing a chemical feature recognized by a singlereceptor should activate the same glomeruli. Odorants with few molecular featuresshould generate simpler patterns of glomerular activation than odorants withmany molecular features, because the latter should bind to additional receptorproteins. Finally, if the specificreceptor proteins encode information in a highly tuned way, the glomerularlayer activity evoked by an odorant feature should be represented bothlaterally and medially, because olfactory sensory neurons expressing a giventype of receptor protein project to both lateral and medial glomeruli in eachbulb. On the other hand, if theolfactory information is widely distributed and the code is represented by verybroad activity patterns in a broadly tuned system, we should not see a focalorganization of activity in the glomerular layer of each bulb.

Place Figure 5 abouthere

Our strategy has been to test the idea that the olfactorysystem functions as a detector of molecular features, and it involvedcharacterizing spatial responses to odorants that with some shared molecularfeatures, while not sharing others (Johnson et al, 1998).  We selected four chemicals for ourinitial study, all with fruit odors: ethyl butyrate, isoamyl butyrate, isoamylacetate, and ethyl acetate (see Figure 5). These chemicals were selected becausethey have a common structure shared by all, and because pairs of the chemicalsshared particular potential molecular features. If the patterns of activity revealed a common representationamong all odorants and the pairs that shared a particular molecular featurealso shared a focal representation, then the data would support the notion of asystem that was sensitive to combinations of molecular features.  We then mapped [¹⁴C]2-deoxyglucose uptake across the entire glomerular layer of the rat olfactorybulb to examine activity patterns evoked by these odorants.

Place Figure 6 abouthere

Each of the four distinctive odorants evoked a distinctspatial pattern of glomerular activation, despite the fact that they differedonly slightly in their molecular features. More importantly, odorants that shared molecular featuresstimulated overlapping areas of activity in the glomerular layer, as can beseen in Figure 6. Both odorantsthat possessed an isoamyl group activated glomerular areas that were not activatedby the two odorants that did not have an isoamyl group (Johnson et al,1998). Similarly, both odorantsthat possessed an ethyl group stimulated areas that were not stimulated by thetwo odorants that did not have an ethyl group. Finally, all four molecules also activated overlapping partsof the bulb, possibly reflecting the common core structure that theyshared. These data are consistentwith a combinatorial mechanism of olfactory coding wherein unitary responses ofolfactory receptors to particular features of a given odorant generate spatialpatterns of bulbar activity that are characteristic of that odorant. Thisimplies that the specific pattern of activity generated by a particular odorantcould be predicted from the activity patterns of other odorants with which itshares nonoverlaping common elements. Another observation that was consistent with such a combinatorialmechanism was that the complexity of the spatial patterns increased withodorant molecular size (Johnson et al, 1998). Increasing molecular size would likely increase the numberof molecular features bound by olfactory receptor proteins.  The increase in the number of molecularfeature that would be processed simultaneously would be expected to increasethe number of focal regions activated in the glomerular layer.

As predicted by the projections of homologous olfactoryreceptor neurons, all maps of activity revealed paired foci, with one focus onthe lateral and one focus on the medial aspect of the bulb.  Moreover, the lateral focus was alwaysmore anterior than the medial focus, again as predicted from the olfactoryneuron projection patterns (Sullivan and Dyer, 1996; Mombaerts et al.,1996a). Given the similarprojection patterns of olfactory receptor neurons expressing individual genes,the spatial relationship between these foci of activity supports the functionalrelevance of the putative olfactory receptor genes.

While the ester odorants that we studied are related, otherodorants are even more closely related. One possibility is that the olfactory system encodes similar odorants inwidely separated parts of the bulb to allow them to be easilydiscriminated. Alternatively, thesystem may cluster the representations of very similar odorants to allow theiridentification as a group and to sharpen the discrimination among similarodorants by lateral inhibition. To distinguish between these two hypotheses, wetested the spatial patterns of activity evoked by odorants that differed by aslittle as one methylene group. Weexposed rats to a series of straight-chained acids varying in length from twoto eight carbons, all with odors characteristic of mammalian bodies.  Then, we mapped [¹⁴C] 2-DGuptake across the glomerular layer. We focused on the question of whetherglomerular responses to these odorants were clustered or dispersed in the bulb(Johnson et al., 1999).

Place Figure 7 abouthere

The acid odorants evoked activity in four discreteglomerular regions that involved two lateral/medial pairs of fields (Figure 7A;Johnson et al., 1999). Odorantsdiffering by a single step in carbon chain length stimulated overlapping butdistinct sets of glomeruli within each of these four fields.  The locations of response within thefields were calculated as centroids of 2-DG uptake (Figure 7B).  For each field, the centroid ofactivity shifted progressively towards more ventral and/or rostral positionswith increasing carbon chain length of the straight-chain acid odorants(Johnson et al., 1999). Thus,within each functional field, glomeruli appear to be chemotopically arranged,such that the nearest neighbors have the most similar odorant specificity.Clustered responses for aldehydes also have been observed in the rat bulb(Rubin and Katz, 1999).

Single-unit recording studies of mitral/tufted cells withina region of the rabbit olfactory bulb that corresponds to a dorsomedial fieldthat we identified in rats reveal responses to various straight chain acids(Mori et al., 1992; Imamura et al., 1992). Furthermore, individual mitral/tufted cells in this regionare tuned to odorants possessing a limited range of carbon-chain lengths (Moriet al., 1992; Imamura et al., 1992). On the basis of their findings, Mori and coworkers proposed thatneighboring glomeruli within the dorsomedial bulb may respond tocarbonyl-containing odorants of different chain lengths, with a preference fora given chain length. Imamura et al. (1992) also suggested that lateralinhibitory interactions between glomeruli and/or mitral/tufted cells couldsharpen the specificity of an individual mitral/tufted cell to a more limitedrange of stimuli by suppressing neighboring responses to closely relatedodorants. Consistent with thishypothesis, they found that a reduction of lateral inhibition among adjacent mitralcells produced a broader response specificity by these cells to differentstraight chain acids (Yokoi et al., 1995). These suggestions are supportedfurther by our data that reveal a chemotopic arrangement of glomerularresponses within this area of the bulb. The use of chemotopic arrangements of glomeruli to achieve odoranttuning provides evidence that spatial distributions of responses may be used bythe olfactory system to encode odor quality.

In our study of ester odorants, we identified high-uptakefoci that responded specifically to isoamyl butyrate (Johnson et al.,1998). These foci were associatedwith very few glomeruli in any given coronal section. Insome cases, the foci seemed to involve only a single glomerulus in asection. Such a finding would bethe predicted consequence of the activation of a single class of olfactoryreceptor protein expressed by olfactory receptor neurons that all project to asingle glomerulus. However, mostother responses to the esters involved foci of 2‑DG uptake that wereassociated with more than one glomerulus in any given coronal section. Our ability to find a closecorrespondence between individual glomeruli and high-uptake foci for theisoamyl butyrate-specific fields suggests that the large number of glomeruliassociated with other foci may be due to the activation of closely relatedreceptor neurons projecting to adjacent sites. The number of glomeruliassociated with a given focus differed for the different esters evoking thefocus (Johnson et al., 1998), which also suggests that neighboring glomerulimight have related, but distinct, specificities.

Clustering of similarly specific glomeruli is furthersupported by data showing that increases in odorant concentration lead toincreases in the area of either focal 2‑DG uptake or c‑fos mRNAhybridization signal (Stewart et al., 1979; Bell et al., 1987; Guthrie andGall, 1995; Johnson et al., 1999). The increased areas at higher odorant concentrations could beinterpreted as a recruitment of nearby glomeruli receiving projections fromolfactory receptors with lower affinity for the odorant, suggesting that justas in all receptor-ligand relationships, a receptor will bind to a range ofmolecular features. Clustering of glomeruli with specificities for chemicallyrelated amino acid odorants also has been observed through Ca2+-sensitive dyerecordings of zebrafish olfactory bulbs, indicating that the use of spatialrelationships to achieve finer odorant tuning might be a phenomenon that can begeneralized across species (Friedrich and Korsching, 1997).

Johnson and Leon (in press) presented two types of odorantsat different concentrations. Oneset of odorants, at least in humans, demonstrates what we may call odorconstancy throughout the range of concentrations. The other set of odorants was selected because humans reportthat these change in quality with increasing concentration.  A combinatorial code would predict thatthere would be distant glomeruli activated at these higher concentrations forthose stimuli that change with concentration. For the odorants with presumed odor constancy, increasedconcentration correlated with increased 2-DG uptake in all glomerular areas ofactivation. Indeed, increasedglomerular activity with increasing odorant concentration has been notedrepeatedly with a variety of techniques (Stewart et al., 1979; Guthrie andGall, 1995; Cinelli et al., 1995; Joerges et al., 1997; Friedrich andKorsching, 1997; Johnson et al., 1999; Rubin and Katz, 1999).  At the same time, those odorantsselected for their ability to change quality evoked activity in parts of thebulb that were not activated at lower concentrations.

Increasingodorant concentrations are thought to recruit additional types of odorantreceptors that have lower affinity for the odorants (Malnic et al, 1999). Theactivation of new glomeruli at higher odorant concentrations probably indicatesthat increasing odorant concentrations recruit new olfactory receptor proteins,since sensory neurons expressing the same receptor gene project to as few asone lateral and one medial glomerulus (Ressler et al., 1994; Vassar et al.,1994; Mombaerts et al., 1996; Wang et al., 1998).

The recruitment of different kinds of olfactory sensoryneurons at higher odor concentrations might not always result in a differentperceived odor. Activation ofdistinct olfactory sensory neurons probably depends on the spatial arrangementsof the glomeruli to which different types of sensory neurons project.  Glomeruli responding maximally toaliphatic acids of a given carbon chain length appear to be located nearglomeruli that respond maximally to acids of slightly different carbon chainlengths, suggesting that sensory neurons expressing receptors of similarselectivity may project to nearby glomeruli (Johnson et al., 1999).  Indeed, olfactory receptor neuronsexpressing similar receptors project to neighboring glomeruli (Tsuboi et al.,1999). Similar specificities of nearby glomeruli also would explain why higherconcentrations of odorants increase the area of focal glomerular responses detectedusing 2-DG uptake (Stewart et al., 1979; Johnson et al., 1999), in situhybridization for c-fos mRNA (Guthrieand Gall, 1995), voltage-dependent dye recording (Cinelli et al., 1995;Friedrich and Korsching, 1997), or optical recording (Rubin and Katz,1999). The increased area ofresponse may reflect the recruitment of neighboring glomeruli that aremaximally responsive to odorants of slightly different chemistries, but thatalso can respond to higher concentrations of suboptimal stimuli (Johnson etal., 1999).

Spatial clustering of glomeruli similar, but not identicalresponse profiles specific has been proposed to stimulate a lateral inhibitorynetwork that tunes projection neurons to specific odorant carbon chain lengths(Yokoi et al., 1995; Friedrich and Korsching, 1997; Johnson et al., 1999).  Since neighboring mitral cells caninhibit one another via granule cell dendrites, the recruitment of aneighboring glomerulus at higher odorant concentrations may not override thegreater inhibition caused by increasing activity of the originally activemitral cells. Thus, mitral cellsassociated with a neighboring glomerulus may not change their activity even iftheir associated glomeruli do. Inthat case, there would not be a change in odor quality.  On the other hand, odorants that evokednew areas of activity far enough from the original glomerular activity to avoidlateral inhibition and their associated mitral cells could well evoke theadditional activity that could underlie an altered perception with increasingconcentration.

In mice, the olfactory receptor proteins expressed bysensory neurons appear to be involved in targeting the axons of the cells tothe appropriate regions of the olfactory bulb (Mombaerts et al., 1996a; Wang etal., 1998). If receptors withsimilar specificity also contain a high degree of amino acid sequence homology,and if similar sequences cause guidance to nearby glomerular locations, thenaxonal guidance by the receptor proteins would be an efficient means to constructclusters of similarly specific glomeruli. The reliability in the locations of uptake across different animalsexposed to the same odorant, which must have been present to obtainstatistically significant results in our analyses, implies that there is aprofoundly rigorous set of parameters used to establish the topography of theprojections of homologous sensory neurons.

The 2-DG maps suggest that the olfactory system is encodingthe information provided by molecular features of chemicals.  If this mechanism is the means by whichthe system codes chemical information, then destruction of such foci shouldprevent the animals from normal recognition of the odorants.  As discussed above, the results ofolfactory bulb lesion studies questioned the importance of focal areas of 2-DGuptake. Prior to our work, therewere numerous reports that propionic acid evoked only a single major focus ofactivity that was located in the dorsomedial bulb (Slotnick et al., 1987; 1989;Mori et al., 1992; Sallaz and Jourdan, 1993). The removal of this focus did notaffect the ability of an animal to detect propionic acid, to distinguishpropionic acid from other chemicals, or to remember the odor of propionic acidlearned prior to the lesion (Slotnick et al., 1987; 1997; Lu and Slotnick,1994; 1998). However, bysystematically mapping the response of the entire glomerular layer, we foundfour fields (i.e., two paired fields) that contained reliable, robust responsesto propionic acid (Johnson et al., 1999). These four fields were distributed in anterior, posterior, medial, andlateral portions of the bulb, including regions not removed in the prior lesionstudies (Johnson et al., 1999). Because important focal areas of 2-DG uptake were not removed in thelesion studies, it remains possible that these areas are critical for odorcoding. It would seem possible tobe able to remove all of the areas that are stimulated by a particular odorantand thereby prevent its normal perception.

Our data on glomerular activity patterns suggested thatthere was considerable localization of responses within the olfactory bulb,with some areas being stimulated far more than others.  Because individual mitral cells in ratsextend their apical dendrites into only a single glomerulus, one would predictthat the mitral cells connected to active glomeruli would also be more activethan mitral cells located elsewhere in the bulb. Such localization of active mitral cells was reported instudies performed on deeply anesthetized rabbits in which recording was deep inthe glomerular layer (Mori et al., 1992; Imamura et al., 1992).  However, other studies indicated thatwidely distributed mitral cells in lightly anesthetized rats respondedsimilarly to many odorants (Motokizawa, 1996). These units may have been secondary or tertiary and not theprimary ones recorded by Mori. Tohelp resolve this issue, we mapped 2‑DG uptake in the external plexiformlayer and in the internal plexiform/superficial granule cell layers of ratsexposed to straight chain acid odorants (Johnson et al., 1999).  Uptake in these layers must reflect theactivation of tufted or mitral cells, because the projections from olfactorysensory neurons do not penetrate further than the glomerular layer.  We found focal areas of response inthese deeper layers that mirrored those detected in the glomerular layer(Johnson et al., 1999). Thesedistinctive responses were located in the same dorsomedial part of the bulbwhere Mori et al. found specific responses to straight chain acids (Mori etal., 1992; Imamura et al., 1992). Furthermore, the uptake in the internal plexiform/superficial granulecell layer differed systematically in location across straight chain acidodorants differing in carbon-chain length, which likely indicates an orderlyspatial arrangement of projection neurons exhibiting different optimalspecificities (Johnson et al., 1999). Our results arose from studies of rats that were not subjected toanesthesia, so that a picture not contaminated by anesthesia has been obtained inrats.

Since homologous olfactory sensory neurons converge intospecific glomeruli within the bulb, glomerular activity should reflect thenarrow tuning of particular olfactory receptor proteins to the specificmolecular features of odorants. This raises the question as to why so many salamander olfactory receptorneurons respond to a given odorant, and why a large number of odorants canstimulate any particular salamander olfactory receptor neuron in vitro(Firestein et al., 1993). Onesimple possibility is that salamanders and rodents have evolved different kindsof tuning mechanisms for their olfactory systems (Sato et al., 1994), apossibility that will be explored in more detail below.  Another possible answer may be similarto that for the low specificity mitral/tufted cell responses detected in theolfactory bulb by Motokizawa (1996). The study showing broad specificity of salamander sensory neurons usedhigh odorant concentrations and did not preselect cells giving the largestresponses. Indeed, in studies thatdid preselect for mouse olfactory receptor neurons located in a portion of theseptum of the epithelium, Sato et al. (1994) found good tuning with respect tothe carbon chain length of straight chain acids and alcohols, especially when theodorants were presented at low concentrations. Similarly, in a study that preselected for mouse sensoryneurons sending axons to glomeruli of the dorsomedial olfactory bulb, whereresponses to straight chain acids are observed, Bozza and Kauer (1998) foundthat individual sensory neurons responded to straight chain acids but not toalcohols of a similar carbon-chain length. Selection for a particular rat olfactory receptor gene alsoresulted in sensory neuron responses that both were finely tuned and highlysensitive (Zhao et al., 1998).

Despite the evidence that the olfactory system is selectivefor a narrow range of chemical structures, crude counts of glomeruli underlyinghigh-uptake foci evoked by high concentrations of valeric acid suggest that upto 5% of all glomeruli were located in the four fields that were activated bythis odorant. High concentrations of caproic acid can stimulate an even greaterproportion of the glomeruli (Johnson et al., 1999), as can high concentrationsof isoamyl butyrate (Johnson et al., 1998). Thus, at high odorant concentrations, the inherentspecificity of the olfactory system may be obscured. Behavioralstudies would help to resolve these issues.

In our study of straight chain acid odorants, both theamount of 2-DG uptake and the number of responsive glomeruli increased withincreasing size of the odorant molecule (Johnson et al., 1999), a findingconsistent with other observations demonstrating elevated responses withincreases in odorant molecular size and/or hydrophobicity.  For example, Sato et al. (1994) foundthat the number of responsive mouse olfactory receptor neurons increases withincreasing carbon number of straight chain acids and alcohols.  Ottoson (1958) found that increasingodorant hydrophobicity correlated with increased electro-olfactograph responseamplitude in frog epithelium. Increased responses of sensory neurons and glomeruli with increasingodorant size may underlie the inverse correlation that exists between thecarbon-chain length of straight chain acids or aldehydes and the concentrationof these odorants required for threshold odor detection in humans(Cometto-Muñiz et al., 1998).

The spatial overlap in responses at the level of theglomerular layer even for odorants that differ greatly in certain molecularfeatures suggests the possibility that even chemically dissimilar odorantsmight be confused on a perceptual level. Humans show surprisingly poor initial performance during odorant identificationtasks, correctly identifying only about half of common, familiar odorants(Cain, 1979; Cain and Potts, 1996). While there are many misidentifications of similar odorants, there alsoare dramatic mistakes involving very different odorants (Cain, 1979; Cain andPotts, 1996) that appear to be related to perceptual confusion rather than toproblems in verbal labeling (Cain and Potts, 1996). Odorant identification becomes virtually perfect, however,with feedback about odorant identification (Cain, 1979), suggesting that humansmay learn to use information that allows them to identify odors. Followingtraining, humans can discriminate between very closely related acids, althoughtheir difficulties increase when these compounds differ by only a single carbonor by changes in branch structure (Laska and Teubner, 1998). Rats can be trained to discriminatebetween propionic acid and acetic acid, which differ by a single step incarbon-chain length, and which have very similar odors to humans (Lu and Slotnick,1994; 1998; Slotnick et al., 1997). However, the possibility that rats would confuse these odorants (or evenmore dissimilar ones) without prior discrimination training has not beenevaluated.

A simple model of odorant molecular feature processing

The spatial patterns of odorant-evoked activity observedwhen 2‑DG uptake is surveyed across the entire glomerular layer areconsistent with a simple model of the processing of odorant molecular featuresin the rat olfactory bulb. Themodel presented below synthesizes aspects of several, previously proposedmodels of odor coding by the olfactory system (Imamura et al., 1992; Kauer andCinelli, 1993; Shepherd, 1994; Axel, 1995; Mori and Yoshihara, 1995; Buck,1996).

In a sense, the odorant molecule is broken up into itscomponent features at the level of the olfactory epithelium and the informationregarding those features is then sent to the olfactory bulb for furtherprocessing. Each receptor proteinexpressed by neurons in the olfactory epithelium should exhibit a preferentialresponse to a given molecular feature that is present in odorantmolecules. The receptor also wouldrespond with lower affinity to variants of that molecular feature.  Each olfactory sensory neuron in ratsappears to expresses only one receptor protein (Chess et al., 1994; Malnic etal., 1999; Mombaerts et al., 1996b) and therefore each sensory neuron shouldrespond preferentially to a single molecular feature.

Olfactory receptor neurons expressing the same kind ofreceptors converge on a limited number of glomeruli (Mombaerts et al.,1996a). The convergence of manyneurons processing the same information regarding a molecular feature shouldincrease the sensitivity of the system because it maximizes the number ofneurons that will stimulate the second-order neurons in the bulb.  Mitral/tufted cells have a singleapical dendrite that arborizes in a single glomerulus (Mori, 1987), therebyassuring that each mitral/tufted cell receives direct projections from sensoryneurons in which activity is related to the binding status of a single receptorprotein.

Glomeruli responding to the same molecular feature appear tobe clustered together, creating a functional field that can be revealed bystudying 2‑DG uptake. Withina given cluster, glomeruli appear to be distributed in such a way that nearestneighbors exhibit the smallest differences in response specificity, perhapsrelated to odorant size or hydrophobicity. This chemotopy is the strongest evidence for spatial codingin the olfactory system. Therigorous spatial arrangement would insure that lateral inhibition by synapticinteractions with interneurons in the glomerular and external plexiform layerswould sharpen mitral and tufted cells to the specific molecular attributes thatwould be difficult to distinguish without tuning.

The effects of tuning would be most pronounced during laterresponses after the lateral inhibition by neighboring mitral cells has beeninitiated. Initial mitral/tuftedcell responses may transmit information about the general class of the odorant(floral), while later, sharpened responses may transmit more specificinformation (rose). Indeed, onecomputational model of coding in the olfactory system predicted a similarhierarchical organization of olfactory coding (Ambros-Ingerson et al., 1990).

In this model, the bulb would be envisioned as an array offunctional fields, in which each field accomplishes tuning of odorant featuresby way of lateral inhibition. Itis not yet clear that there will be any psychophysical meaning associated withthe spatial relationships between distinct fields evoked by structurallydifferent odorants. If the basaldendrites of mitral cells connected to glomeruli at the boundary of afunctional field branch radially in all directions, they may induce inhibitionof neighboring mitral cells connected to glomeruli within a separate functionalfield. Thus, there may beinteractions between parts of adjacent fields that have different molecularfeature specificities. Theseinteractions would lead to the additional prediction that certain odorantfeatures, and thereby certain odorants that evoke activity at the boundaries ofadjacent fields may mask each other more effectively (e.g., in odorantmixtures) than would others that are represented in more distant fields.  On the other hand, it may be thatmitral cells located at the edge of a given functional field have basaldendrites extending only in the direction of neighboring mitral cells withinthe same field, thereby creating an isolated anatomical unit for the processingof chemicals with similar specificities.

At some point in this processing system, the molecularfeatures that had been separately processed in the olfactory bulb must bebrought together to form the perception of a specific odor.  The axon of a mitral cell leaves theolfactory bulb to synapse with cortical and other forebrain targets (Macridesand Davis, 1983). A singlecortical pyramidal cell may receive input from various mitral cells, each ofwhich transmits information about a single feature present in an odorant.  Such an idealized convergence couldprovide coincidence detection to reassemble the features of a pureodorant. The previously reportedlack of topography in mitral cell-to-cortex projections  (Haberly and Price, 1977) could insurethe convergence of most possible sets of molecular features that could bepresent in an odorant or odorant mixture. Recent studies, however, have shown that mitral cells associated with asingle glomerulus may project to only a limited domain within the olfactorycortex (Puche et al., 1998). Activation of one glomerulus thus leads to a modular activation ofcortical pyramidal cells (Puche et al., 1998) Therefore, a hierarchicalreassembly of the information describing distinct molecular features carried bymitral cell axons may occur in the olfactory cortex and the features of otherodors may be rejoined at an even higher level of processing involving neuronassemblies and/or secondary projections of cortical pyramidal cells.

Alternative models of odorant processing

The study of olfaction has employed an extremely wide rangeof animal species, and alternative models of odor quality coding have combinedresults from these numerous species (Kauer and Cinelli, 1993).  However, there are both fundamentalspecies-dependent differences in the anatomy of the olfactory system (Eisthen,1997) and important ethological concerns, that suggest caution in consideringone species as a model for another. There also are considerable gaps in our knowledge concerning some of thespecies that make it difficult to judge the suitability of a given model.

Studies of locusts and honeybees have revealedodorant-specific temporal patterns of responses by individual projectionneurons in the antennal lobe that are phase-linked to each other and towidespread odorant-induced oscillations (Laurent et al., 1996; Wehr andLaurent, 1996; Stopfer et al., 1997). Pharmacological blockade of the oscillations disrupts the learning andmemory of a certain pair of similar chemical odorants detected by honeybees,but not of a pair of more dissimilar chemicals (Stopfer et al., 1997).  Different odorants also activatedistinct subsets of glomeruli in the honeybee (Joerges et al., 1997; Galizia etal., 1999), and the persistence of discrimination between dissimilar odorantsin the absence of oscillations suggests the possibility of spatial coding todistinguish some odorants (Stopfer et al., 1997). Nevertheless, the rigorous correlation between temporalparameters of the projection neuron activity and the identity of the similarodorants raises the possibility that these species also may use a criticalmechanism involving temporal coding to discriminate between similarodorants. On the other hand, thereare alternative interpretations of these data that suggest that the locusts orbees function very much like rats and rabbits in using a highly specificspatial code. The loss ofsynchronized activity in the antennal lobes of these insects is accomplished byblocking GABA receptors, and as noted above, when these receptors are blockedin rabbits, there is a loss of response specificity to odorants, probably dueto the loss of lateral inhibition (Yokoi et al., 1995).  Since the spatial localization of theinitial odorant response in bees and locusts was not identified before andafter drug administration, it is possible that blocking GABA in these insectssimply allows a less specific spatial localization activity, with consequentloss of perceptual specificity. Underthese conditions, one would expect to observe difficulties in discriminatingbetween similar odors and little difficulty discriminating between dissimilarodors.

It has been suggested that these data from insects also mayapply to odor coding in mammals, including rats (Laurent, 1997; Dorries, 1998),despite remarkable anatomical differences between the species.  A given locust projection neuron hasdirect connections to approximately eleven non-contiguous glomeruli (Laurentand Naraghi, 1994), in contrast to the one-to-one relationship between glomeruliand projection neurons in the rat. To our knowledge, olfactory receptor genes in these species have not yetbeen cloned, so the relationship between glomeruli and the projections ofhomologous sensory neurons is unknown. On the other hand, an olfactory receptor gene family has been identifiedin fruit flies, and that species seems to express one receptor per olfactoryreceptor neuron (Vosshall et al., 1999; Clyne et al., 1999).   If there were a one-to-onerelationship between homologous sensory neurons and glomeruli, then a giveninsect projection neuron could receive direct excitatory input relay from abouteleven different olfactory receptors. Such an organization would be expected to increasesensitivity to a broad group of odorants (e.g.: food odors).  However, it could be that all of theglomeruli to which a given locust projection neuron is connected receive axonsonly from homologous sensory neurons, thereby producing the same type ofconvergence present in a rodent. Such an anatomical organization could be expected to increasesensitivity to a specific odorant. If this were the case, rules for the clustering of glomeruli withrelated specificities may be different in the locust, given that the multipleglomeruli influencing a single projection neuron are themselves separated byadditional glomeruli.

There is direct evidence in other invertebrates ofexcitatory and inhibitory actions of different odorants on the same sensoryneuron (Michel and Ache, 1994). This suggests both that a single sensory neuron may express multipleolfactory receptor proteins and that some degree of odor processing may beaccomplished by the first neuron in the pathway. Direct evidence for such processing in mammals has not beenreported, although peripheral interactions among distinct odorants remain apossibility (Bell et al., 1987).

Our data certainly can not rule out the possibility oftemporal representations of odorant quality in the rat, and temporal factorsmay figure heavily in coincidence detection of multiple molecularfeatures. Odorant-evokedoscillations also may be involved in the discrimination between similarodorants that evoke overlapping responses in the rat glomerular layer,especially when these odorants are presented as mixtures.  Indeed, mixtures of propionic andcaproic acids have been reported to induce oscillatory responses in thedorsomedial region of the rabbit olfactory bulb (Mori et al., 1992).

Other studies with salamanders have described broad patternsof activation across bulbar layers, and the patterns evoked by odorants ofdifferent chemistries overlap extensively (Cinelli et al., 1995).  These data have been used to supportdiffusely distributed models of odor processing in mammals.  The broad patterns of voltage-sensitivedye responses to odorants measured in salamanders (Cinelli et al., 1995)contrast greatly with the focal patterns of 2‑DG uptake observed in ratsusing the same odorant, ethyl butyrate (Johnson et al., 1998).  Although these findings may indicate adifferent sensitivity of the two imaging techniques, it also may be indicativeof an important species difference. Like the insect, individual mitral cells of the salamander olfactorybulb have apical dendrites in multiple glomeruli (Nieuwenhuys, 1967; Eisthen, 1997),and the relationships between glomeruli and sensory neurons expressing singletypes of olfactory receptor proteins are not yet known.  As discussed above, it also is possiblethat sensory neurons in salamander (Firestein et al., 1993) are more broadlytuned than are sensory neurons in a rodent (Sato et al., 1994; Zhao et al.,1998; Bozza and Kauer, 1998). Because of the numerous anatomical differences in the olfactory systemsof amphibians and rodents, odor-coding principles also may differ between them.

Fish and rodent olfactory systems are even more different inanatomy than are the olfactory systems of salamanders and rodents (Nieuwenhuys,1967; Eisthen, 1997). Therefore,it is remarkable that clustering of glomeruli of related specificities areobserved in fish (Friedrich and Korsching, 1997) as well as in rodents.  This finding suggests that some tuningprinciples may be present across widely different species.  Indeed, tuning of individual mitralcells to straight chain compounds of a given chain length also has been foundin the frog olfactory bulb (Dôving, 1966).

Obviously, the possible discovery of subtle differences inthe specificities of neighboring glomeruli, as have been observed in fish andrats, requires the presentation of odorants differing systematically inchemical structure. It thereforewould be interesting to determine whether insects and salamanders employ morespatially specific principles when tested with sets of chemically relatedodorants. The behavioral relevanceof the odorants selected for study also should be considered.  For example, the amino acid odorantsthat are represented by clusters of glomeruli with similar specificity inzebrafish (Friedrich and Korsching, 1997) are nonvolatile, water-solublecompounds. These compoundspresumably would not reach the main olfactory epithelium of a land-dwellingmammal, but they represent naturally occurring, important olfactory stimulirelated to food in fish. It islikely that each species is somewhat specialized in terms of which odorants itbest discriminates, and it also is possible that rats and other omnivorousspecies may discriminate a broad range of odorants related to a great varietyof food sources.

Developmental plasticity of the olfactory bulb in lightof the simple model of odorant feature processing

The simple model of odorant feature processing in theolfactory system that we described above provides several interestingpredictions concerning changes in olfactory bulb structure and function causedby early learning about odors.

One specific prediction of the model is that increasedglomerular activity, such as occurs following learning, may lead to increasedbehavioral sensitivity to odors. That the amount of glomerular activity can be correlated with increasedsensitivity of odorant detection follows from our studies of 2-DG uptake inresponse to different odorants. For example, larger odorants are detected at lower concentrations thanare smaller odorants of the same chemical class (Cometto-Muñiz et al., 1998),and larger odorants stimulate more glomerular 2‑DG uptake than do smallerodorants presented at the same vapor phase concentration (Johnson et al., 1998;1999). Increases in sensitivity ofdetection following prior exposure to certain odorants have been described foradult mice (Wang et al., 1993) and humans (Wysocki et al., 1989; Stevens andO’Connell, 1995). In mice, thisincrease in sensitivity is associated with increased activity that can bemeasured in the olfactory nerve, suggesting that the underlying neurobiologicalchanges are likely to involve the olfactory sensory neurons themselves (Wang etal., 1993). Followingdiscrimination learning in adult rats, spatially restricted regions of the olfactoryepithelium increase in activity (Youngentob and Kent, 1995).  Changed activity through an increaseeither in the number of receptors expressed in each sensory neuron, or in thenumber of sensory neurons expressing a given receptor, is likely to causechanged activity in a limited number of glomeruli in the olfactory bulb.

In the simple model of odorant feature processing, glomeruliare envisioned as functional units that serve to segregate responses toindividual molecular features of odorants. Reassembly of these features into odor perceptions is likelyto occur at some higher level of processing. The hedonic value of the odor would likely emerge at stillsome later stage. In other words,the increased glomerular activity during and following learning would notreflect odor preference per se. Consistent with this prediction, increased glomerular uptake of 2-DGalso occurs once young rats have learned an aversion to an odorant (Sullivanand Wilson, 1991).

Different odorants possessing a common molecular featureactivate the same glomeruli. Therefore, after an animal has learned about the significanceone odorant, and thereby has an enhanced response in certain glomeruli, asubsequent exposure to another odorant possessing a molecular feature in commonwith the learned odorant would stimulate some of the same glomeruli, and theseresponses also would be enhanced. This other odorant also would activate additional glomeruli that do notoverlap with those stimulated by the learned odorant. An interesting prediction of this partial overlap is thatthe learning of one odorant could change the spatial pattern of activity evokedby the other odorant by altering the relative contribution of differentglomeruli to the pattern. Becausespatial patterns of activity are likely to be related to odor perception, earlylearning of one odorant then would be predicted to change the perceived odor ofother volatile chemicals. Forodorants that are very closely related in chemistry to a previously learnedodorant, the stimulation of a large number of overlapping, enhanced glomerulimay lead to a greater perceived similarity between these odorants and thelearned one. A potential benefitof this predicted increase in odor generalization is a decreased significance ofminor variations in the compositions of odorant mixtures.  It also would make it possible foranimals to generalize among odorants typical of food classes.  For example, prior preference learningof particular ester odorants associated with one kind of fruit may become morereadily generalized to esters associated with some other fruit.

In a related issue, recall that there are glomeruli in theposterior, ventrolateral bulb that respond to peppermint extract, but that donot increase their 2-DG uptake or Fos-like response following early learning (Johnsonet al., 1995; Johnson and Leon, 1996). One possible explanation for this finding is that control animalspreviously also had experienced the odorant molecular features detected bythese glomeruli. For example, thevolatile components responsible for the odor of peppermint extract areprimarily terpenes related to menthol, and many other plants produce theterpenes that occur in peppermint. Both control and trained animals would have been exposed to volatilecompounds from the hardwood chips used to line their cages, and it is possiblethat these compounds had the same molecular feature responsible for activity inthe ventrolateral glomeruli that is evoked by components of peppermintextract. An alternativeexplanation for the differential effects of learning on midlateral andventrolateral glomeruli is that some glomeruli are more plastic than others(Johnson and Leon, 1996). Differential plasticity would be consistent with the heterogeneousglomerular distribution of proteins that may play a role in plasticity,including the beta-adrenergic receptor (Woo and Leon, 1995a), nitric oxidesynthase, and certain growth factor receptors. It would be interesting to determine with a wide variety ofodorants where in the bulb glomerular responses are plastic in response toearly olfactory learning.

Summary

The olfactory system appears to encode odorants by initiallybreaking them up into their component molecular features.  A large family of odorant receptors ispresent at this first level to bind these features with different levels ofspecificity. The olfactory neuronsbearing a specific kind of receptor all project to a glomerulus either on thelateral or the medial aspect of olfactory bulbs, thereby allowing the signalgenerated by one kind of feature to be amplified. As the signal is transmittedinto the bulb at the level of the glomerular layer, the signal for a specificfeature also can be sharpened by local inhibition of the responses to similarfeatures in the bulb. The signalrepresenting each feature then is transmitted into the olfactory cortex whereit is combined with the information regarding other features to form theperception of the odor. While a representation for odorants is present in the bulb at birth,early learning regarding odors can change the subsequent response in thebulb.

In the coming years, a study of the perceptual consequencesto early alterations in the coding of odorants may reveal the connectionbetween what is represented and what is perceived in the olfactory system.  Another realm of inquiry to be exploredwill be the micro-circuitry of the olfactory bulb and olfactory cortex; muchshould be learned with a careful respect for the spatially specificorganization of the coding process. In addition, the differences among species may be appreciated inrelation to the different olfactory tasks that each is asked to accomplish bythe natural world. Finally, wehave to render a more detailed map of molecular features across the bulbarsurface and then we must correlate those spatial activation patterns with theirpsychophysical properties to gain a full understanding of odor coding.

References

Alkasab,T.K., T.C. Bozza, T.A. Cleland, K.M. Dorries, T.C. Pearce, J. White, J.S.Kauer. (1999) Characterizingcomplex chemosensors: information-theoretic analysis of olfactory systems.Trends Neurosci. 22:102-108.

Ambros-Ingerson,J., R. Granger, and G. Lynch (1990) Simulation of paleocortex performshierarchical clustering. Science 247:1344-1348.

Astic, L. andM. Cattarelli (1982) Metabolic mapping of functional activity in the ratolfactory system after a bilateral transection of the lateral olfactory tract.Brain Res. 245: 17-25.

Axel, R.(1995) The molecular logic of smell. Sci. Am. 273:154-159.

Bell, G.A.,D.G. Laing, and H. Panhuber (1987) Odour mixture suppression: Evidence for aperipheral mechanism in human and rat. Brain Res. 426:8-18.

Bozza, T.C.and J.S. Kauer (1998) Odorant response properties of convergent olfactoryreceptor neurons. J. Neurosci. 18:4560-4569.

Buck, L.B.(1996) Information coding in the vertebrate olfactory system. Ann. Rev.Neurosci. 19:517-544.

Buck, L.B.and R. Axel (1991) A novel multigene family may encode odorant receptors: Amolecular basis for odor recognition. Cell 65:175-187.

Buonviso, N.and M.A. Chaput (1990) Response similarity to odors in olfactory bulb outputcells presumed to be connected to the same glomerulus: Electrophysiologicalstudy using simultaneous single-unit recordings. J. Neurophysiol. 63:447-454.

Cain, W.S.(1979) To know with the nose: Keys to odor identification. Science 203:467-470.

Cain, W.S.and B.C. Potts (1996) Switch and bait: Probing the discriminative basis of odoridentification via recognition memory. Chem. Senses 21:35-44.

Chess, A., I.Simon, H. Cedar, and R. Axel (1994) Allelic inactivation regulates olfactoryreceptor gene expression. Cell 78:823-834.

Cinelli,A.R., K.A. Hamilton, and J.S. Kauer (1995) Salamander olfactory bulb neuronalactivity observed by video rate, voltage-sensitive dye imaging. III. Spatialand temporal properties of responses evoked by odorant stimulation. J.Neurophysiol. 73:2053-2071.

Clyne, P.J.,C.G. Warr, M.R. Freeman, D. Lessing, J. Kim, and J.R. Carlson (1999) A novelfamily of divergent seven-transmembrane proteins: candidate odorant receptorsin Drosophila. Neuron 22:327-338.

Cometto-Muñiz,J.E., W.S. Cain, and M.H. Abraham (1998) Nasal pungency and odor of homologousaldehydes and carboxylic acids. Exp. Brain Res. 118:180-188.

Coopersmith,R. and M. Leon (1984) Enhanced neural response to familiar olfactory cues.Science 225:849-851.

Coopersmith,R. and M. Leon (1986) Enhanced neural response by adult rats to odorsexperienced early in life. Brain Res. 371:400-403.

Coopersmith,R., S.R. Henderson, and M. Leon (1986) Odor specificity of the enhanced neuralresponse following early odor experience in rats. Dev. Brain Res. 27:191-197.

Dean, P.M.(1987) Molecular Foundations of Drug-Receptor Interaction. Cambridge:University Press.

Dickinson,T.A., J. White, J.S. Kauer, and D.R. Walt (1998) Current trends in'artificial-nose' technology. Trends Biotech. 16:250-258.

Do, J.T.,R.M. Sullivan, and M. Leon (1988) Behavioral and neural correlates of postnatalolfactory conditioning: II. Respiration during conditioning. Dev. Psychobiol. 21:591-600.

Dorries, K.M.(1998) Olfactory coding: Time in a model. Neuron 20:7-10.

Dôving, K.B.(1966) An electrophysiological study of odour similarities of homologoussubstances. J. Physiol. 186:97-109.

Eisthen, H.L.(1997) Evolution of vertebrate olfactory systems. Brain Behav. Evol.50:222-233.

Firestein,S., C. Picco, and A. Menini (1993) The relation between stimulus and responsein olfactory receptor cells of the tiger salamander. J. Physiol. 468:1-10.

Freeman W.J.and C.A. Skarda (1985) Spatial EEG patterns, non-linear dynamics andperception: the neo-Sherringtonian view. Brain Res., 357:147-75.

Friedrich,R.W. and S.I. Korsching (1997) Combinatorial and chemotopic odorant coding inthe zebrafish olfactory bulb visualized by optical imaging. Neuron 18:737-752.

Galizia,C.G., S. Sachse, A. Rappert, and R. Menzel (1999a) The glomerular code for odorrepresentation is species specific in the honeybee Apis mellifera. NatureNeurosci. 2; 473-478.

Galizia,C.G., Menzel, R., and Holldobler, B. (1999b) Optical imaging ofodor-evoked glomerular activity patterns in the antennal  lobes of the ant Camponotusrufipes. Naturwissenschaffen,86:533-537.

Guthrie, K.M.and C. Gall (1995) Functional mapping of odor-activated neurons in theolfactory bulb. Chem. Senses 20:271-282.

Guthrie, K.M.and C. Gall (1999) Functional mapping of the developing olfactory bulb.  AchemS, Abstract, 21: 17.

Guthrie,K.M., A.J. Anderson, M. Leon, and C. Gall (1993) Odor-induced increases inc-fos mRNA expression reveal an anatomical unit for odor processing inolfactory bulb. Proc. Natl. Acad. Sci. USA 90:3329-3333.

Haberly, L.B.and J.L. Price (1977) The axonal projection patterns of the mitral and tuftedcells of the olfactory bulb in the rat. Brain Res. 129:152-157.

Haberly L.B.and J.M. Bower (1989) Olfactory cortex: model circuit for study of associativememory? Trends Neurosci.12:258-264.

Imamura, K.,N. Mataga, and K. Mori (1992) Coding of odor molecules by mitral/tufted cellsin rabbit olfactory bulb. I. Aliphatic compounds. J. Neurophysiol.68:1986-2002.

Joerges J,Küttner A, Galizia CG, Menzel R. 1997. Representations ofodours and odour mixtures visualized in the honeybee brain. Nature 387:285-288.

Johnson, B.A.and M. Leon (1996) Spatial distribution of [¹⁴C] 2-deoxyglucoseuptake in the glomerular layer of the rat olfactory bulb following earlyolfactory preference learning. J. Comp. Neurol. 376:557-566.

Johnson, B.A.and M. Leon (in press) Modular glomerular representations of odorants in therat olfactory bulb: The effects of stimulus concentration. J. Comp. Neurol.

Johnson,B.A., C.C. Woo, H. Duong, V. Nguyen, and M. Leon (1995) A learned odor evokesan enhanced Fos-like glomerular response in the olfactory bulb of young rats.Brain Res. 699:192-200.

Johnson, B.A., C.C. Woo, and M. Leon (1998) Spatialcoding of odorant features in the glomerular layer of the rat olfactory bulb.J. Comp. Neurol. 393:457-471.

Johnson,B.A., C.C. Woo, E.E. Hingco, K.L. Pham and M. Leon, (1999) Multidimensionalchemotopic responses to n-aliphatic acid odorants in the rat olfactorybulb. J. Comp. Neurol. 409:529-548.

Jourdan, F.,A. Duveau, L. Astic, and A. Holley (1980) Spatial distribution of [¹⁴C]2-deoxyglucoseuptake in the olfactory bulbs of rats stimulated with two different odours. BrainRes. 188:139-154.

Katoh, K., H.Koshimoto, A. Tani, and K. Mori (1993) Coding of odor molecules bymitral/tufted cells in rabbit olfactory bulb. II. Aromatic compounds. J.Neurophysiol. 70:2161-2175.

Kauer, J.S.(1987) Coding in the olfactory system. In: Neurobiology of Taste and Smell. T.E. Finger and W.S. Silver, Eds. Wiley, NY pp. 205-231.

Kauer, J.S.and A.R. Cinelli (1993) Are there structural and functional modules in thevertebrate olfactory bulb? Micr. Res. Tech. 24:157-167.

KrautwurstD., K.W. Yau and R.R.Reed. (1998)Identification of ligands for olfactory receptors by functional expression of areceptor library. Cell 95:917-26.

Laska, M. andP. Teubner (1998) Odor structure-activity relationships of carboxylic acidscorrespond between squirrel monkeys and humans. Am. J. Physiol.274:R1639-R1645.

Laurent, G.(1997) Olfactory processing: Maps, time and codes. Curr. Opin. Neurobiol.7:547-553.

Laurent, G.and M. Naraghi (1994) Odorant-induced oscillations in the mushroom bodies ofthe locust. J. Neurosci. 14:2993-3004.

Laurent, G.,M. Wehr, and H. Davidowitz (1996) Temporal representations of odors in anolfactory network. J. Neurosci. 16:3837-3847.

Leon, M.(1975) Dietary control of maternal pheromone in the lactating rat. Physiol.Behav. 14:311-319.

Leon, M. andMoltz, H. (1971) Maternal pheromone: discrimination by preweanling albino rats.Physiol. Behav, 7:265-267.

Lu, X.-C. M.and B.M. Slotnick (1994) Recognition of propionic acid vapor after removal ofthe olfactory bulb area associated with high 2-DG uptake. Brain Res. 639:26-32.

Lu, X.-C. M.and B.M. Slotnick (1998) Olfaction in rats with extensive lesions of theolfactory bulbs: Implications for odor coding. Neuroscience 84:849-866.

Macrides, F.and B.J. Davis (1983) The olfactory bulb. In P.C. Emson (ed): ChemicalNeuroanatomy. New York: Raven Press, pp. 391-426.

Malnic, B.,J. Hirono, T. Sato, and L. Buck (1999) Combinatorial receptor codes forodors. Cell, 96:713:723.

Matsutani, S.and M. Leon (1993) Elaboration of glial cell processes in the rat olfactorybulb associated with early learning. Brain Res. 613:317-320.

McCollum,J.F., C.C. Woo, and M. Leon (1997) Granule and mitral cell densities areunchanged following early olfactory preference training. Dev. Brain Res. 99:118-120.

Michel, W.C.and B.W. Ache (1994) Odor-evoked inhibition in primary olfactory receptorneurons. Chem. Senses 19:11-24.

Mombaerts,P., F. Wang, C. Dulac, S.K. Chao, A. Nemes, M. Mendelsohn, J. Edmonson, and R.Axel (1996a) Visualizing an olfactory sensory map. Cell 87:675-686.

Mombaerts,P., F. Wang, C. Dulac, S.K. Chao, A. Nemes, M. Mendelsohn, J. Edmonson, and R.Axel (1996b) The molecular biology of olfactory perception. Cold Spring HarborSymp. Quant. Biol. 61:135-145.

Mori, K.(1987) Membrane and synaptic properties of identified neurons in the olfactorybulb. Progr. Neurobiol. 29:275-320.

Mori, K. andY. Yoshihara (1995) Molecular recognition and olfactory processing in themammalian olfactory system. Progr. Neurobiol. 45:585-619.

Mori, K., N.Mataga, and K. Imamura (1992) Differential specificities of single mitral cellsin rabbit olfactory bulb for a homologous series of fatty acid odor molecules.J. Neurophysiol. 67:786-789.

Motokizawa,F. (1996) Odor representation and discrimination in mitral/tufted cells of therat olfactory bulb. Exp. Brain Res. 112:24-34.

Nieuwenhuys,R. (1967) Comparative anatomy of olfactory centres and tracts. In Y. Zotterman(ed): Progress in Brain Research, New York: Elsevier, pp. 1-64.

Ottoson, D.(1958) Studies on the relationship between olfactory stimulating effectivenessand physicochemical properties of odorant compounds. Acta Physiol. Scand.43:167-181.

Puche, A., V.Aroniadou-Anderjaska, and M. Shipley (1998) Olfactory bulb-olfactory cortexslices in the study of central olfactory CNS circuits.  Soc. Neurosci. Abstr., 34:1885.

Ressler,K.J., S.L. Sullivan, and L.B. Buck (1994) Information coding in the olfactorysystem: Evidence for a stereotyped and highly organized epitope map in theolfactory bulb. Cell 79:1245-1255.

Royet, J.P.,G. Sicard, C. Souchier, and F. Jourdan (1987) Specificity of spatial patternsof glomerular activation in the mouse olfactory bulb: Computer-assisted imageanalysis of 2‑deoxyglucose autoradiograms. Brain Res. 417:1-11.

Rubin, B.D.and L. C. Katz (1999) Optical imaging of odorant representations in themammalian olfactory bulb. Neuron, 23, 499-511.

Sallaz, M.and F. Jourdan (1993) C-fos expression and 2-deoxyglucose uptake in theolfactory bulb of odour-stimulated awake rats. NeuroReport 4:55-58.

Sato, T., J.Hirono, M. Tonoike, and M. Takebayashi (1994) Tuning specificities to aliphaticodorants in mouse olfactory receptor neurons and their local distribution. J.Neurophysiol. 72:2980-2989.

Shepherd,G.M. (1991) Computational structure of the olfactory system. In J. Davis and H.Eichenbaum (eds): Olfaction as a model system for computational neuroscience.Cambridge, MA: MIT Press, pp. 3-41.

Shepherd,G.M. (1994) Discrimination of molecular signals by the olfactory receptorneuron. Neuron 13:771-790.

Slotnick,B.M., S. Graham, D.G. Laing, and G.A. Bell (1987) Detection of propionic acidvapor by rats with lesions of olfactory bulb areas associated with high 2-DGuptake. Brain Res. 417:343-346.

Slotnick,B.M., H. Panhuber, G.A. Bell, and D.G. Laing (1989) Odor-induced metabolicactivity in the olfactory bulb of rats trained to detect propionic acid vapor.Brain Res. 500:161-168.

Slotnick,B.M., G.A. Bell, H. Panhuber, and D.G. Laing (1997) Detection anddiscrimination of propionic acid after removal of its 2-DG identified majorfocus in the olfactory bulb: A psychophysical analysis. Brain Res. 762:89-96.

Somjen, G.(1972) Sensory Coding in the Mammalian Nervous System. New York:Appleton-Century-Crofts.

Stevens, D.A.and R.J. O’Connell (1995) Enhanced sensitivity to androstenone followingregular exposure to pemenone. Chem. Senses 20: 413-419.

Stewart,W.B., J.S. Kauer and G.M. Shepherd (1979) Functional organization of ratolfactory bulb analyzed by the 2-deoxyglucose method. J. Comp. Neurol. 185:715-734.

Stopfer, M.,S. Bhagavan, B.H. Smith, and G. Laurent (1997) Impaired odour discrimination ondesynchronization of odour-encoding neural assemblies. Nature 390:70-74.

Sullivan,R.M. and M. Leon (1986) Early olfactory learning induces an enhanced olfactorybulb response in young rats. Dev. Brain Res. 27:278-282.

Sullivan,R.M., Wilson, D.A., Kim, M.H. and Leon, M. (1988) Behavioral and neural correlates of postnatalolfactory conditioning: I. Effect of respiration on conditioned neuralresponses. Physiol. Behav. 44,85-90.

Sullivan,R.M. and D.A. Wilson (1991) Neural correlates of conditioned odor avoidance ininfant rats. Behav. Neurosci.103:307-312.

Sullivan,R.M., D.A. Wilson, and M. Leon (1989) Norepinephrine and learning-inducedplasticity in infant rat olfactory system. J. Neurosci. 9:3998-4006.

Sullivan,R.M., D.A. Wilson, R. Wong, A. Correa, and M. Leon (1990) Modified behavioraland olfactory bulb responses to maternal odors in preweanling rats. Dev. BrainRes. 53:243-247.

Sullivan,S.L. and L. Dyer (1996) Information processing in mammalian olfactory system.J. Neurobiol. 30:20-36.

Tsuboi A,Yoshihara S, Yamazaki N, Kasai H, Asai-Tsuboi H, Komatsu M, Serizawa S, IshiiT., Y. Matsuda, F. Nagawa, H. Sakano (1999) Olfactory neurons expressing closely linked and homologousodorant receptor genes tend to project their axons to neighboring glomeruli onthe olfactory bulb. J Neurosci 19:8409-8418.

Vassar, R.,S.K. Chao, R. Sitcheran, J.M. Nuñez, L.B. Vosshall, and R. Axel (1994)Topographic organization of sensory projections to the olfactory bulb. Cell79:981-991.

Vickers ,N.J. and T.A. Christensen (1998) A combinatorialmodel of odor discrimination using a small array of contiguous, chemicallydefined glomeruli. Ann. N.Y. Acad.Sci. 855:514-516.

Vosshall,L.B., H. Amrein, P.S. Morozov, A. Rzhetsky, and R.Axel (1999) A spatial map ofolfactory receptor expression in the Drosophila antenna. Cell 96:725-736.

Wang, H.-W.,C.J. Wysocki, and G.H. Gold (1993) Induction of olfactory receptor sensitivityin mice. Science 260: 998-1000.

Wang, F., A.Nemes, M. Mendelsohn, and R. Axel (1998) Odorant receptors govern the formationof a precise topographic map. Cell 93:47-60.

Webster, D.B.and M. Webster (1977) Neonatal sound deprivation affects brain stem auditorynucei. Arch. Otolaryngol. 103: 392-396.

Wehr, M. andG. Laurent (1996) Odour encoding by temporal sequences of firing in oscillatingneural assemblies. Nature 384:162-166.

Wilson, D.A.and M. Leon (1988) Spatial patterns of olfactory bulb single-unit responses tolearned olfactory cues in young rats. J. Neurophysiol. 59: 1770-1782.

Wilson, D.A.,R.M. Sullivan and M. Leon (1987) Single-unit analysis of postnatal olfactorylearning: Modified olfactory bulb output response patterns to learned attractiveodors. J. Neurosci., 7:3154-3162.

Woo, C.C. andM. Leon (1991) Increase in a focal population of juxtaglomerular cells in theolfactory bulb associated with early learning. J. Comp. Neurol. 305:49-56.

Woo, C.C. andM. Leon (1995a) Distribution and development of beta-adrenergic receptors inthe rat olfactory bulb. J. Comp. Neurol. 352: 1-10.

Woo, C.C. andM. Leon (1995b) Early olfactory enrichment and deprivation both decreasefl-adrenergic receptor density in the main olfactory bulb of the rat. J. Comp.Neurol. 360: 634-642.

Woo, C.C., R.Coopersmith, and M. Leon (1987) Localized changes in olfactory bulb morphologyassociated with early olfactory learning. J. Comp. Neurol. 263: 113-125.

Woo, C.C.,M.H. Oshita, and M. Leon (1996) A learned odor decreases the number ofFos-immunopositive granule cells in the olfactory bulb of young rats. Brain Res.716: 149-156.

Wysocki,C.J., K.M. Dorries, and G.K. Beauchamp (1989) Ability to perceive androstenonecan be acquired by ostensibly anosmic people. Proc. Natl. Acad. Sci. USA 86:7976-7978.

Yokoi, M., K.Mori, and S. Nakanishi (1995) Refinement of odor molecule tuning bydendrodendritic synaptic inhibition in the olfactory bulb. Proc. Natl. Acad.Sci. USA 92:3371-3375.

Youngentob,S.L. and P.F. Kent (1995) Enhancement of odorant-induced mucosal activitypatterns in rats trained on an odorant identification task. Brain Res.670:82-88.

Zhao, H., L.Ivic, J.M. Otaki, M. Hashimoto, K. Mikoshiba, and S. Firestein (1998)Functional expression of a mammalian odorant receptor. Science 279:237-242.

Figure Legends

Figure1. The olfactory receptor neuronstransduce chemical stimulation and project into the glomeruli of the olfactorybulb. The axons of the olfactoryreceptor neurons synapse with mitral, tufted and periglomerular cells withinthe glomeruli. Mitral cells arethe dominant output neurons of the bulb, tufted cells project both within thebulbs, as well as to the olfactory cortex and periglomerular cells appear tomediate interglomerular inhibition. The external plexiform layer contains secondary dendrites of mitral andtufted cells as well as tufted cell bodies and granule cell dendrites. The internal plexiform layer containsgranule cell dendrites, as well as mitral and tufted axons.  The granule cells mediate selfinhibition of mitral cells, inhibition among neighboring mitral cells, andinhibition of mitral cells evoked by efferent neurons.  The accessory olfactory bulb (AOB)receives input from a secondary chemical sensing system, which has itsreceptors in the vomeronasal organ

Figure 2. Thisdiagram shows the projection patterns of olfactory sensory neurons that expressthe same putative olfactory receptor gene. The receptor of these homologous olfactory receptor neuronsare distributed across the olfactory epithelium and these neurons project to asingle glomerulus on the medial aspect and a single glomerulus on the lateralaspect of the bulb, albeit in different anterior-posterior planes shown incoronal sections.

Figure 3. Twohypotheses for how odorants would likely be bound to olfactory receptorproteins. The less likely possibility is that each odorant binds to a singlereceptor in its entirety. A morelikely hypothesis is that the molecular features that comprise each odorantbind separately to each type of receptor.

Figure 4. Thisfigure illustrates the combinatorial model of olfactory coding.  The combination of molecular featuresthat describe an odorant constitutes the representation for any odor in theolfactory system.