This is a preprint of an article published in the Journal of Comparative Neurology 503:1-34.

© 2007 Wiley-Liss

Chemotopic Odorant Coding in a Mammalian

Olfactory System

Brett A. Johnson and Michael Leon

Department of Neurobiology and Behavior, University of California, Irvine

Number of text pages: 89

Number of figures: 9 (9 color)

Number of tables: 1

Abbreviated title: Chemotopic odorant coding

Associate editor: Thomas E. Finger

Key words: Sensory coding, rat, 2-deoxyglucose, imaging techniques, mapping

*Correspondence to: Brett A. Johnson, PhD

Dept. of Neurobiology & Behavior

2205 McGaugh Hall

University of California, Irvine

Irvine, CA 92697-4550

telephone: (949)824-7303

fax: (949)824-2447

email: bajohnso@uci.edu

Supported by United States Public Health Service Grants DC03545, DC006391, and DC006516

ABSTRACT

Systematic mapping studies involving 365 odorant chemicals have shown that glomerular responses in the rat olfactory bulb are organized spatially in patterns that are related to the chemistry of the odorant stimuli. This organization involves the spatial clustering of principal responses to numerous odorants that share key aspects of chemistry such as functional groups, hydrocarbon structural elements, and/or overall molecular properties related to water solubility. In several of the clusters, responses shift progressively in position according to odorant carbon chain length. These response domains appear to be constructed from orderly projections of sensory neurons in the olfactory epithelium and may also involve chromatography across the nasal mucosa. The spatial clustering of glomerular responses may serve to “tune” the principal responses of bulbar projection neurons by way of inhibitory interneuronal networks, allowing the projection neurons to respond to a narrower range of stimuli than their associated sensory neurons. When glomerular activity patterns are viewed relative to the overall level of glomerular activation, the patterns accurately predict the perception of odor quality, thereby supporting the notion that spatial patterns of activity are the key factors underlying that aspect of the olfactory code. A critical analysis suggests that alternative coding mechanisms for odor quality, such as those based on temporal patterns of responses, enjoy little experimental support.

Olfactory stimuli

Olfactory stimuli are typically vaporous chemicals that bind to odorant receptors on olfactory sensory neurons in the nasal epithelium (Buck and Axel, 1991; Axel, 1995; Buck, 1996). The odorant molecules are thought to bind and activate receptors through mechanisms similar to those dictating other receptor-ligand interactions (Araneda et al., 2000; Kajiya et al., 2001; Katada et al., 2005). To begin to understand these interactions, it seems reasonable to study pure odorant chemicals of known structure, just as early studies of visual coding used controlled spots of light (Kuffler, 1953; Hubel and Wiesel, 1959) and studies of auditory coding used pure tones of controlled frequency and volume (e.g., Scheich and Zuschratter, 1995).

Because pure chemicals differ from one another incrementally rather than continuously, even small differences in odorant structure can change multiple chemical dimensions that might be relevant to receptor interactions. Many incremental changes can be made to a single molecule, as shown in Figure 1, and each change can result in alterations in associated molecular properties such as length, hydrophobicity, polarity, and flexibility that might affect the ability of different parts of the molecule to associate with any given receptor (Ho et al., 2006b). Therefore, one needs to study a large number of odorant chemicals varying along these different dimensions to be able to understand stimulus coding by the olfactory system.

Anatomical foundation for the early stages of olfactory processing in rodents

There are about 1,000 different odorant receptor genes in rats and mice (Zhang and Firestein, 2002), and most sensory neurons probably express only one of these receptors (Serizawa et al., 2004). Sensory neurons expressing the same receptor are organized in many overlapping but distinct zones stretching anterior to posterior across the olfactory epithelium (Ressler et al., 1993; Vassar et al., 1993; Miyamichi et al., 2005). During passive breathing, an odorized air stream is drawn at low flow rate along the central channel of the nasal epithelium and then across the remaining receptor expression zones, with odorant molecules absorbing differentially into the chemically complex, aqueous mucosa along the way (Mozell, 1964; Hornung and Mozell, 1977; Hornung et al., 1987; Mozell et al., 1987; Figure 2). During active sniffing, such as occurs when rodents are exploring their environment, or performing learned olfactory-guided behaviors, animals alter the airflow dramatically (Youngentob et al., 1987).

Olfactory sensory neurons expressing the same odorant receptor gene converge in their projection into only a few glomeruli in the olfactory bulb, and these glomeruli are consistently located across individuals (Ressler et al., 1994; Vassar et al., 1994; Mombaerts et al., 1996). Sensory neurons located along the nasal septum or in the ventral part of the nasal turbinates project to glomeruli in the medial half of the bulb, whereas homologous sensory neurons located in the lateral turbinates project to lateral glomeruli (Astic and Saucier, 1986; Clancy et al., 1994; Lèvai et al., 2003; Figure 2). Sensory neurons located in the central channel of the nose project to glomeruli in the dorsal half of the bulb, while sensory neurons located in progressively peripheral or ventral parts of the nose project more ventrally (Saucier and Astic, 1986; Schoenfeld et al., 1994; Schoenfeld and Cleland, 2005; Figure 2). These projections form two mirror-image maps of odorant receptor input, one on the lateral aspect and one on the medial aspect of the bulb (Miyamichi et al., 2005; Tsuboi et al., 2006). In addition, a distinct set of sensory neurons expressing characteristic odorant receptor genes is clustered at the tips of certain turbinates rather than along the anterior-posterior expression zones (Strotmann et al., 1992; 1999), and sets of these neurons expressing the same gene project to unpaired glomeruli along the ventral extremity of the bulb (Strotmann et al., 2000). Most evidence suggests that each glomerulus receives input only from sensory neurons homologous with respect to the expression of a single odorant receptor gene (Treloar et al., 2002; Wachowiak et al., 2004).

The convergence of homologous sensory neurons to produce receptor-based maps in the glomerular layer of the main olfactory bulb has provided an obvious target for imaging studies in olfaction. If one could monitor differential odorant-evoked activity across all of the glomeruli in an olfactory bulb, one would obtain a read-out of the differential activation of odorant receptors. By using a large number of systematically chosen odorant stimuli and monitoring the response of the entire glomerular layer, one could determine (1) the features of an odorant stimulus that are responsible for activation of each glomerulus, (2) the aspects of each odorant stimulus that are most relevant to its representation, (3) the features that are ignored by the receptors altogether, and (4) the spatial arrangements of different responses in the glomerular layer that may suggest a specific kind of processing of particular odorant-related information.

We first will review the early work establishing the fact that different odorants and different aspects of odorant chemistry are associated with the activation of neurons in different parts of the olfactory bulb, and then we will discuss the current state of understanding of the relationships between odorant chemistry and activity of various bulbar locations. We will address criticisms of different approaches to study this system, as well as the relevance of the measured spatial patterns to odor perception. We also offer a brief, critical discussion of temporal coding hypotheses that consider spatially distinct activity patterns to be largely irrelevant to olfaction.

“Odotopic” organization of bulbar responses

Adrian concluded from evoked potential studies that different odorants activate different parts of the olfactory bulb, a concept termed “odotopy" (e.g., Adrian, 1950a). A consideration of the anatomy of the olfactory nerve projections led Le Gros Clark (1957) to suggest that different glomeruli might contain converging input from sensory neurons of similar specificity, and that different glomeruli might respond differentially to different odorant chemicals. Indeed, evoked potentials in particular glomeruli then were shown to differ in sensitivity to particular odorants (Levetau and MacLeod, 1966).

Possible relationships between particular odorants and particular bulbar locations were later mapped onto “representative” coronal sections by studying differential degeneration following long-term exposures to an impressively large array of single odorants (e.g., Pinching and Døving, 1974). These degeneration studies also showed a consistency in location across different animals exposed to the same single odorant (Døving and Pinching, 1973; Pinching and Døving, 1974). Despite these early indications of a relationship between particular odorants and particular bulbar locations, the specificity of this relationship was not recognized until the work of Gordon Shepherd and colleagues using the 2DG method (Sharp et al., 1975; 1977; Stewart et al., 1979). They identified foci of 2DG uptake and made two-dimensional activity maps of individual olfactory bulbs. They showed that individual odorant chemicals stimulated a number of segregated areas within the glomerular layer that were consistently located in different individuals, but that were also to some extent overlapping for different odorants (Stewart et al., 1979). These responses were seen in deeper bulbar layers as well (Sharp et al., 1977; Lancet et al., 1982). Other researchers extended the mapping of 2DG uptake to other odorants (Skeen, 1977; Jourdan et al., 1980; Teicher et al., 1980; Coopersmith and Leon, 1984; Coopersmith et al., 1986; Bell et al., 1987; Wilson and Leon, 1988; Sicard et al., 1989; Slotnick et al., 1989), and the collection of 2DG data was systematized and subjected to statistical analysis that showed that spatial patterns of uptake differed significantly for different odorants (Royet et al., 1987).

The odotopic activation of particular parts of the bulb by particular odorants now has been confirmed by many techniques including the mapping of field potentials (Mori et al., 1992), unit recordings of mitral cells (Mori et al., 1992; Imamura et al., 1992; Katoh et al., 1993), up-regulation of immediate early gene products in the glomerular and granule cell layers (Onoda, 1992; Guthrie et al., 1993; 2000; Sallaz and Jourdan, 1993; Schellink et al., 1993; Schaefer et al., 2001a; 2002; Inaki et al., 2002; Montag-Sallaz and Buonviso, 2002; Salcedo et al., 2005), optical imaging of either endogenous responses (Rubin and Katz, 1999; Uchida et al., 2000; Meister and Bonhoeffer, 2001; Takahashi et al., 2004a,b; Igarashi and Mori, 2005) or voltage- and calcium-sensitive dye responses (Wachowiak and Cohen, 2001; 2003; Fried et al., 2002; Spors and Grinvald, 2002; Spors et al., 2006), and functional magnetic resonance imaging (Yang et al., 1998; Xu et al., 2000; 2003; 2005; Schafer et al., 2006).

Odotopy has been extended to other classes of vertebrates (e.g., zebrafish: Friedrich and Korsching, 1997; 1998; Fuss and Korsching, 2001, catfish: Nikonov and Caprio, 2001; Nikonov et al., 2005, salamanders: Cinelli et al., 1995), as well as to insects (e.g., honeybees: Joerges et al., 1997; Galizia et al., 1999a; Sachse et al., 1999, moths: Galizia et al., 2000; Carlsson et al., 2002; Collmann et al., 2004; Skiri et al., 2004; Lei et al., 2004, ants: Galizia et al., 1999b, and Drosophila: Rodrigues, 1988; Wang et al., 2003; Kreher et al., 2005). Given this evidence, it would appear that odorant-specific spatial patterning of glomerular activation is a basic characteristic of olfactory systems.

Odotopy is in fact an inevitable consequence of the fact that sensory neurons expressing different odorant receptors have differential responses to different odorants and project to different locations in the bulb. This fact does not by itself indicate that the brain uses information about the location of the activated glomeruli in olfactory processing, because glomeruli must be located somewhere in space. However, the fact that different individuals exposed to the same odorant have the same pattern of activity would not be predicted by any hypothesis that considers spatial location to be unimportant for olfactory coding.

Chemotopic organization of bulbar responses

Adrian (1950a, 1953) reported that spatial patterns of bulbar activity might be related to odorant chemical features such as functional groups or lipid solubility, but this idea of chemotopy was not pursued until Korsching and coworkers showed that amino acids with similar side chains activated similar sets of glomerular clusters in zebrafish (Friedrich and Korsching, 1997; Fuss and Korsching, 2001). Other odorant classes such as bile acids and pheromones stimulated other parts of the bulb, whereas nucleic acids stimulated characteristic glomerular activity patterns in regions partly overlapping with amino acid-sensitive zones (Friedrich and Korsching, 1998). Segregation of glomeruli responding to some of these same odorant classes also has been clearly demonstrated in catfish (Nikonov and Caprio, 2001), where chemotopic organization extends into the forebrain (Nikonov et al., 2005).

Fish live in an aquatic environment where individual odorant chemicals are water-soluble, often charged molecules of limited diversity; they also have a more limited repertoire of odorant receptors and many fewer glomeruli than do rodents (Alioto and Ngai, 2005), which are exposed to a wider variety of meaningful odorant chemistries through the air. Despite its greater complexity, the rat olfactory system also has a chemotopic organization. We have performed a series of studies focused on systematic sets of odorants organized with respect to their chemical structures (Table 1), and we have mapped 2DG uptake across the entire glomerular layer in response to each odorant (Johnson et al., 1998; 1999; 2002; 2004; 2005a,b; 2006; 2007a,b; Johnson and Leon, 2000a,b; Farahbod et al., 2006; Ho et al., 2006a,b). We average our data across a group of individuals exposed to the same odorant to obtain spatial maps that can be used for statistical analyses to establish where responses are significantly different among odorants. By using the same methods across our different experiments, we have been able to construct an archive of odorant responses that allows relationships between these patterns to be visualized and compared in different orientations and formats (http://leonserver.bio.uci.edu).

Three types of chemotopic organization have emerged from these studies. In the first, a cluster of glomeruli responds similarly to odorants with similar structural features and/or with similar overall molecular properties such as water solubility. In the second type of chemotopy, activated glomeruli within a cluster are arranged systematically in space in relation to a molecular property of the odorant. In the third type, which we have termed “global chemotopy”, the degree of similarity in overall spatial patterns of activity across the glomerular layer is proportional to the degree of similarity in odorant chemistry. We shall discuss each level of chemotopic organization in turn.

Local clusters of glomeruli respond to odorants of similar chemistry

An early piece of evidence that clusters of adjacent glomeruli have similar chemical response specificity came from studies of glomerular responses to increasing concentrations of a single odorant. Imaging methods including 2DG (Stewart et al., 1979; Johnson and Leon, 2000a), in situ hybridization for c-fos mRNA (Guthrie and Gall, 1995), and optical imaging of either voltage-dependent dyes (Cinelli et al., 1995) or intrinsic signals (Meister and Bonhoeffer, 2001), showed that low concentrations of an odorant tend to stimulate very few glomeruli in any given location, consistent with the activation of the highest affinity receptors for the odorant ligand. Increasing concentrations of the same odorant recruit responses in glomeruli located nearby the originally activated ones, with the overall effect of increasing the area of the response roughly in proportion to odorant concentration. The newly activated neighboring glomeruli are likely associated with odorant receptors possessing a lesser affinity for the odorant ligand, perhaps because their best stimuli are closely related odorant chemicals.

The stimulation of a large cluster of neighboring glomeruli at high concentrations of some odorants may be related to the ability of those odorants to assume multiple conformations that satisfy the binding requirements for multiple, related receptors. In general, we have found that flexible, straight-chained aliphatic odorants such as valeric acid and methyl valerate are the ones that tend to activate large clusters of glomeruli at high concentrations (Figure 3A), whereas odorant molecules with less flexibility (fewer rotatable bonds) activate fewer glomeruli (Johnson et al., 1999; 2006). These more rigid odorants include shorter aliphatic molecules such as propionic acid and methyl acetate (Figure 3B), as well as various cyclic odorants such as cyclobutanecarboxylic acid and oxyoctaline formate (Figure 3C). The focal 2DG responses shown in Figure 3 also serve to illustrate that the 2DG method is capable of single-glomerular resolution when it is applied to individual olfactory bulbs.

The best evidence for spatial clustering of glomeruli with similar specificity comes from experiments where responses are mapped to systematic sets of odorants differing by small increments in structure. Typically, different odorants sharing certain aspects of chemistry are found to stimulate overlapping, but distinct sets of glomeruli in the same general area of the bulb. Figure 4 summarizes our understanding of the modular representations of these odorant chemical features across the glomerular layer of the rat olfactory bulb. Some of these glomerular modules have specificities characterized by odorant functional groups, others are associated with elements of odorant hydrocarbon structure, and still others are associated with overall molecular properties of the odorants, independent of their specific structural features. The maps of odorant chemistry appear in duplicate, with one copy on the lateral aspect and one copy on the medial aspect of the bulb (Johnson et al., 1998), an organization that parallels the paired lateral and medial projection of sensory neurons expressing the same odorant receptor gene (Ressler et al., 1994; Vassar et al., 1994; Miyamichi et al., 2005). Our previously published diagrams of odorant response modules (Johnson and Leon, 2000a; Johnson et al., 2002) have obvious relationships to the present summary figure (Figure 4). However, the mapping of hundreds of additional responses predictably has led to adjustments in boundaries, and in some cases adjacent modules that were previously defined as separate entities now have been fused, with the recognition that they actually contain responses to similar compounds.

The specificities of these modules are described in detail in the final section of this review, which both summarizes our own 2DG mapping studies and relates our findings to those from other labs that have used other methods to monitor odorant-evoked glomerular activity. In most cases, our results using 2DG are complementary to observations from optical imaging studies, especially when these are conducted across large portions of the glomerular layer (see Mori et al., 2006).

All possible chemical features are not represented as unique modules

As described above, glomeruli responding to odorants with similar functional groups, hydrocarbon structures, or overall molecular properties are often clustered together in the bulb. However, there are some chemical features that are not represented by their own glomerular modules. For example, although double and triple bonds have unique partial charge distributions and steric configurations that presumably could serve as distinctive binding sites for sets of odorant receptors, odorants sharing such features do not specifically overlap in their stimulation of particular glomeruli (Ho et al., 2006b; Johnson et al., 2007a). Instead, unsaturated bonds seem either to disrupt the recognition of certain odorant ligands by certain receptors or to change the glomeruli that are activated within a module (Ho et al., 2006b; Johnson et al., 2007a). Another example is certain cyclic structures, which would seem to provide enough specific chemical “information” to support recognition by a class of odorant receptors, but which do not activate common glomeruli (Johnson et al., 2006).

More generally, some chemical attributes appear to have greater importance in determining responses than others, just as some odorants seem to evoke more overall activity than others (Johnson and Leon, 2000b; Ho et al., 2006b; Johnson et al., 2006; 2007a). The olfactory system evolved to solve biologically relevant problems under certain constraints, and an all-purpose qualitative analysis of odorant chemistry would not be expected of the system, which is more likely to be focused on sets of odorant chemicals that contribute important information relevant to survival and reproduction. We will revisit this theme in the discussion of chemotopic progressions below.

Chemotopic progressions within glomerular modules

Different odorants activating the same response module typically stimulate overlapping, but distinct, sets of glomeruli within the module. This phenomenon is apparent even after the averaging of patterns of 2DG uptake across both bulbs of several different animals (Johnson et al., 1998; 1999; 2004; Johnson and Leon, 2000b; Farahbod et al., 2006), and we take it as evidence that sensory neurons of distinct, but related specificity project to nearby, quite consistently situated glomeruli in these modules. For three pairs of modules, the arrangement of glomeruli is systematic with respect to odorant chemistry, establishing another of level of chemotopic organization in the system. In all three pairs (shaded orange, green, and light blue in Figure 4), straight-chained, unsaturated, aliphatic odorants of greater carbon number stimulate progressively ventral glomeruli (arrows in Figure 4).

Ventral progressions with increasing carbon number in homologous series have been shown for carboxylic acids and ethyl esters in the anterior domains responding to these compounds (Figure 4, orange shading). These progressions have been seen in studies of 2DG uptake (Johnson et al., 1999; 2004), optical imaging (Uchida et al., 2000), and immediate early gene expression (Inaki et al., 2002). In the lateral member of this pair of domains, the shift also involves a progressive movement rostrally with increasing carbon number, such that aliphatic acids of eight or more carbons stimulate glomeruli located on the ventral part of the rostral pole of the bulb (Johnson et al., 1999; unpublished data). Shifts in the same areas are found in many studies using aldehyde odorants and either 2DG uptake (Johnson et al., 2004) or optical imaging (Rubin and Katz, 1999; Uchida et al., 2000; Meister and Bonhoeffer, 2001), although we interpret these shifts as being related to carboxylic acids present as oxidized contaminants in the aldehyde preparations (Johnson et al., 2004). One study using a recombinant fluorescent marker of synaptic vesicle fusion did not detect chemotopic progressions in this lateral, anterior module in response to a series of aldehydes (Bozza et al., 2004), which may indicate either a variable oxidation state of the odorants used in that study, or a different imaged area.

In any homologous series of straight-chained odorants, a greater carbon number is associated with systematic differences in numerous related chemical properties, including molecular length, molecular volume, and hydrophobicity. Carboxylic acid odorants of different hydrocarbon structures (double bonded, branched, and cyclic) do not show as much covariance in these related properties. We exploited this situation and found that the chemotopic progression of responses in the medial domain was more correlated with molecular length than with the other properties (Johnson and Leon, 2000b).

Increasing carbon number in alcohols and aldehydes is associated with ventral chemotopic progressions in the module preferring these functional groups (green shading in Figure 4), as determined by 2DG uptake (Johnson et al., 2004) and immediate early gene expression (Inaki et al., 2002). The phenomenon is seen for both primary alcohols and secondary alcohols with the hydroxyl group in the 2-position (Johnson et al., 2004). The hydrocarbon chain-related domains (light blue in Figure 4) also show chemotopic progressions for all effective stimuli including esters, acids, alcohols, aldehydes, and alkanes (Johnson et al., 1998; 1999; 2004; Inaki et al., 2002; Igarashi and Mori, 2005; Ho et al., 2006a).

In summary, chemotopic progressions with increasing odorant carbon number have been detected in multiple response modules using multiple odorant series and multiple imaging methods. In many experiments characterizing these progressions, odorants are presented at fixed vapor phase concentrations (Johnson et al., 1998; 1999; 2004) or across a range of concentrations (Meister and Bonhoeffer, 2001), although similar progressions can be found without controlling odorant concentration (Rubin and Katz, 1999; Uchida et al., 2000; Johnson and Leon, 2000b; Inaki et al., 2002; Igarishi and Mori, 2005; Ho et al., 2006a). The suggestion that chemotopic progressions arise as an artifact of different odorant concentrations across a series (Wilson and Mainen, 2006) is therefore unfounded.

Gaps in chemotopic progressions involving homologous series

In our analyses of 2DG uptake, we average data matrices first across the two bulbs of an individual animal and then across different animals. Due to a small amount of biological variance in glomerular location (Royal and Key, 1999; Strotmann et al., 2000; Schaefer et al., 2001b), as well as to experimental variation in dissection and tissue sectioning, the detailed spatial arrangement of activated glomeruli that is evident in original autoradiograms (e.g., Figure 3) often is not apparent in our averaged matrices. Similarly, statistical analyses of the shift in location of responses within a glomerular module are affected by such variance, so that the chemotopic progressions in 2DG uptake often appear to be smooth and continuous (Johnson et al., 1999; 2004). However, when the contrast between differentially active individual glomeruli was enhanced by mathematical filtering following optical imaging studies of a homologous series of straight-chained odorants, it was evident that glomeruli activated by these odorants often are interrupted by glomeruli that do not respond as well to any members of the series (Meister and Bonhoeffer, 2001). What do these gaps indicate about chemotopic progressions?

We have found that odorants of a wide variety of hydrocarbon structure can stimulate modules that also are activated by straight-chained compounds (Johnson and Leon, 2000b). It seems likely that some of these odorants would stimulate individual glomeruli that are distinct from those activated by straight-chained compounds. In the medial acid-preferring domain, the responses to branched, double-bonded and alicyclic acids are laid out chemotopically along with the straight-chained compounds in an arrangement proportional to molecular length (Johnson and Leon, 2000b). Therefore, we hypothesize that some of the less active glomeruli in the optical imaging studies might be activated by related odorants of distinct hydrocarbon structure that were not tested in those optical imaging studies. We cannot exclude the possibility, however, that glomeruli of unrelated specificity might also be present in these domains.

Possible causes of chemotopic progressions

We have considered two explanations for how chemotopic progressions might arise within glomerular domains. The first idea is that the progressions are actually laid out by chromatography of odorants in the olfactory epithelium (Mozell, 1964; Hornung and Mozell, 1977; Hornung et al., 1987; Mozell et al., 1987; Scott et al., 2000), and then are fed forward to the bulb by way of the topographical relationships involved in that projection (Saucier and Astic, 1986; Clancy et al., 1994; Schoenfeld et al., 1994; Schoenfeld and Cleland, 2005). Relative absorption into the olfactory mucosa depends on the air-mucosa partition coefficient. Although some odorants may absorb to the mucosa through interactions with macromolecules (Pelosi, 1996; Johnson et al., 2005b), much of the absorption is thought to be related to water solubility such that more water-soluble members of any given series should absorb earlier in the air path, which projects to more dorsal bulbar locations, while less water soluble odorants can diffuse further through the air to reach parts of the nose projecting to more ventral bulbar locations (Schoenfeld and Cleland, 2005; Zhao et al., 2006). Within any given homologous series of odorants, the odorants with fewer carbons would be more water-soluble and thereby more prone to early absorption and stimulation of more dorsal bulbar locations (Figure 2).

The second explanation is that sensory neurons bearing receptors specific for shorter molecules project to the dorsal-most glomeruli in the domains, while sensory neurons bearing receptors specific for the longer molecules project more ventrally. The molecular processes that cause axons of sensory neurons expressing the same receptor to bundle together to converge onto single glomeruli also might operate to insure that adjacent glomeruli receive projections from sensory neurons of the most similar specificity. The odorant receptor itself is involved in axonal path finding (Wang et al., 1998), and sensory neurons containing receptors of similar sequence indeed project to nearby bulbar areas (Tsuboi et al., 1999; Strotmann et al., 2000). Therefore, if receptors of similar specificity with respect to odorant molecular length also have similar amino acid sequences, then orderly glomerular clustering would be expected. This receptor-based hypothesis for the origin of chemotopic progressions is consistent with our finding that odorant molecular length is more predictive of the location of response within the medial acid-preferring module than is odorant hydrophobicity (Johnson and Leon, 2000b). Similar experiments have not yet been conducted for other odorant-module combinations.

Chromatographic separation of odorants in the olfactory epithelium and orderly, differential bulbar projections of sensory neurons responding to odorants of different length are not mutually exclusive. These two processes could work together to establish chemotopic progressions within glomerular response modules (Scott-Johnson et al., 2000), especially considering that rats probably can optimize the location of epithelial stimulation by modulating various attributes of their respiratory behavior (Youngentob et al., 1987).

Not all modules show chemotopic progressions

The very dorsal, ketone-responsive domain does not show chemotopic progressions of activity proportional to the number of carbons in straight-chained ketone odorants (Johnson et al., 2004); nor have we been able to identify any other molecular feature or property of ketones that is organized chemotopically within that domain (e.g., Johnson et al., 2005a). One possible explanation is that the most dorsal bulbar domains are associated with the central channel (zone 1) of the epithelium (Schoenfeld and Cleland, 2005). Initial airflow through the central channel should be in the anterior-to-posterior direction across zone 1 (Kimbell et al., 1997). Any chromatography of strongly absorbed odorants occurring along this axis would not result in the stimulation of a distinct set of odorant receptors, so that chromatography could not contribute to a chemotopic progression. The domains preferring aliphatic esters (yellow shading in Figure 4) or aromatic hydrocarbons (dark brown shading), both of which also may involve sensory neurons in epithelial zone 1, also do not show evidence for chemotopic progressions (Farahbod et al., 2006).

We also have found no evidence for chemotopic progressions within the ventral domains that respond to bicyclic and camphoraceous odorants (Johnson et al., 2006). These molecules overlap heavily in the areas they stimulate in the ventral domains, and the complexity of the structure of these odorants makes them difficult to classify along any single dimension. The ventral region of the bulb receives projections from sensory neurons expressing receptors in a non-zonal epithelial distribution (Strotmann et al., 1992), which may also explain the absence of chemotopic progression. Finally, we have not seen evidence for chemotopic progressions corresponding to any attribute of the water-soluble odorants stimulating posterior bulbar domains (Figure 4, red shading), although this topic was not the subject of any direct study (Johnson et al., 2007b).

Possible consequences of chemotopic progressions

Mitral and tufted cell projection neurons located in the dorsomedial part of the rabbit olfactory bulb respond to acid and aldehyde odorants possessing a narrow range of carbon number along a homologous series (Mori et al., 1992; Imamura et al., 1992; Yokoi et al., 1995). Yokoi et al. (1995) showed that when inhibition by interneurons was blocked using the GABA receptor antagonist picrotoxin, these mitral cells responded with robust sequences of action potentials to a broader range of straight-chained odorants, suggesting that lateral inhibition among neighboring glomeruli had served to narrow the molecular receptive range of the mitral cells in this region. Such “tuning” by center-surround lateral inhibition should occur preferentially in areas where related odorants directly activate glomeruli organized in a chemotopic progression (Figure 5). Studies of 2DG uptake (Johnson et al., 1999; 2004), optical imaging (Uchida et al., 2000; Meister and Bonhoeffer, 2001; Takahashi et al., 2004a), and expression of immediate-early genes (Inaki et al., 2002) all showed that the area from which Yokoi et al. (1995) recorded is indeed chemotopically organized with respect to the dimension of carbon chain length that was investigated in that study. Therefore, one consequence of chemotopic organization may be to insure that relative responses of nearby glomeruli can be “compared” by way of mutual inhibition to produce a pattern of mitral cell output that is more distinct with respect to these similar odorants (Figure 5).

Given the chemotopic organization of the olfactory bulb, we would not expect mitral cells that are randomly recorded across other parts of the olfactory bulb to show this kind of tuning along any randomly chosen chemical dimension, especially if the definition of “response” were broadened to include inhibition and delayed action potentials. Therefore, the failure of these spatially (and temporally) unconstrained approaches to find evidence for tuning (c.f.: Motokizawa, 1996) cannot be taken as evidence that tuning of odorant responses does not occur.

The rabbits used in these studies of mitral cell tuning (Mori et al., 1992; Imamura et al., 1992; Yokoi et al., 1995) were under urethane anesthesia, and spontaneous mitral cell activity was greatly reduced during the experiments. It has been argued that this condition may have obscured a broader responsiveness of these cells than would have been detected if the animals were differently anesthetized (Motokizawa, 1996) or entirely awake (Bhalla and Bower, 1997; Kay and Laurent, 1999; Rinberg et al., 2006a). Awake animals, such as those we study using 2DG, show complex spontaneous mitral cell activity against which it appears to be easier to detect information about the animal’s behavioral state (e.g., hunger, expectation of odor, alertness, or responses to any odor independent of its identity) than to detect information directly relevant to odor quality (Pager, 1974a,b; Bhalla and Bower, 1997; Kay and Laurent, 1999; Rinberg et al., 2006a), at least when the recordings are made in arbitrary locations in response to arbitrarily chosen odorants. We predict that recording from mitral cells associated with the peak glomerular responses in awake rats might reveal selective, high levels of responsiveness to specific odorants against a lower level of background activity, although these mitral cells are also likely to be highly responsive to top-down control of their activity.

Because mitral cell activity ultimately must be responsible for carrying information about odor identity, downstream processing must somehow extract odorant-specific activity from the background of other “information” that is related to behavioral state. The possibility should be considered that the use of anesthetics simply reveals the most relevant odorant-specific information in mitral cells independent of the influences of behavioral state, information that olfactory cortex would extract in some other way. When we mapped 2DG uptake in the superficial granule cell layer and the external plexiform layer of awake rats, where all activity would be secondary to the activation of mitral cells, we found local areas of increased uptake directly beneath the foci of glomerular uptake, which suggests that a predictable subpopulation of mitral cells is more responsive to a given odorant than are the other mitral cells in the bulb (Johnson et al., 1999). Similar relationships between activity in the glomerular layer and deeper bulbar lamina were noted in earlier 2DG (Sharp et al., 1977) as well as in c-fos studies (Guthrie et al., 1993). Moreover, the largest optically imaged glomerular responses are good predictors of the largest projection neuron responses in both rats (Luo and Katz, 2001) and honeybees (Sachse and Galizia, 2002; 2003). We further found that the granule cell activity shifted progressively in location with increasing carbon chain length, just as it did in the glomerular layer (Johnson et al., 1999), thereby confirming the anatomical foundation for the type of tuning detected in the Yokoi et al. (1995) study.

To date, the Yokoi et al. (1995) study appears to be the only one to have reported direct evidence for tuning by way of lateral inhibition in mammals, although Sachse and Galizia (2002) found that principal neurons in honeybees respond more broadly to odorants in the absence of inhibition by local interneurons, and Luo and Katz (2001) found that rat mitral cells located beneath strongly activated glomeruli show a pattern of excitation while surrounding mitral cells are inhibited, a pattern of response that would be expected if the output of the bulb were subject to lateral inhibition. It would be good to have additional examples of tuning in this system before the phenomenon is completely accepted. One benefit of mapping responses to multiple odorants across the entire bulb is to find areas and candidate odorant series to test such hypotheses, and on the basis of our mapping studies, a reasonable choice would be to use homologous series of carbon chain length and to study any of the six (three pairs of) chemotopically organized domains that responds to the odorants with the chosen functional group (Figure 4).

It also should be noted that simpler olfactory systems with fewer glomeruli might not use center-surround, nearest-neighbor relationships to tune output neurons. For example, inhibitory interneurons can connect nearly all glomeruli in an antennal lobe of a honeybee, and spatial proximity is not as meaningful as similarity of response profile in predicting which glomeruli are functionally involved in inhibitory networks in this species (Linster et al., 2005). Lateral inhibition and spatial arrangements of glomeruli likely would be more important in a larger structure such as a rodent olfactory bulb, where periglomerular and granule cell neurons are expected to exert their inhibitory influences over only a fraction of the total bulbar area (Shepherd, 1972). Differences in anatomical organization may explain why there is more evidence of spatial clustering of responses to similar odorant chemicals in vertebrates than in most invertebrates, although chemotopic organization is apparent in Drosophila larva (Kreher et al., 2005).

Not all continuously varying molecular properties are coded by progressions

A criticism of the hypothesis that center-surround lateral inhibition is used to tune responses in the olfactory system is that odorants vary in a highly dimensional space (Figure 1), whereas center-surround architecture is only two-dimensional (Laurent, 1999; Cleland and Sethupathy, 2006; Wilson and Mainen, 2006). The implication is that any two-dimensional organization would be useful only for a sensory system in which a two-dimensional stimulus space is being represented (e.g., the retina). There are at least four important problems with this criticism. First, as will be discussed in another section, the representation of any odorant usually involves a combination of multiple modules involving independent molecular features of the odorant, and tuning to different features in different modules clearly adds dimensionality to the representation. Second, tuning of some responses does not prevent individual receptors and mitral cells from being represented independently to construct a higher-dimensional combinatorial code. Third, differences along some of the “dimensions” of chemistry that are implied by the criticism, such as the nature of functional groups and certain variations in hydrocarbon structure, actually involve such large changes in the odorant stimulus that incremental differences along these “dimensions” cause the activation of entirely distinct sets of receptors, glomerular modules, and mitral cells, so that no center-surround tuning is required. Fourth, we have evidence from glomerular responses and behavioral analyses that not all chemical dimensions of an odorant stimulus are coded at the same level of detail, reducing the number of dimensions that need to be represented.

We have found only one chemical dimension that is mapped by chemotopic progressions across any given glomerular module, namely carbon number (or a correlated property such as molecular length or hydrophobicity). Other systematic changes in odorant chemistry either do not have much impact on the patterns at all, or are represented by differential activity across different response modules. For example, changes involving the presence, number, position, or stereochemistry of double or triple bonds in hydrocarbons have little impact on 2DG uptake patterns (Ho et al., 2006b). Behavioral studies showed that these differences also did not have significant effects on perception using an assay capable of showing odorant generalization (Ho et al., 2006b). The presence, position and number of methyl group branches in hydrocarbon odorants also do not greatly impact 2DG uptake patterns or perception (Ho et al., 2006b). Functional group position has little impact on patterns evoked by ketones or esters, but completely different glomerular modules respond to alcohols differing in substitution position (Johnson et al., 2005a). Position of substitution has little effect on the representation of aromatic hydrocarbons, whereas aromatic odorants with alkyl substituents activate distinct glomeruli from aromatic odorants with oxygenic substituents (Farahbod et al., 2006).

Data from these individual 2DG mapping studies exemplify how an empirical approach to coding in the olfactory system is more productive than considerations about stimulus dimensionality that are not meaningfully constrained by definitions of those dimensions. The olfactory system is not likely to have evolved into an all-purpose analysis system for identification and relative quantification of each and every chemical that might be synthesized by an organic chemist. Rather, natural selection has more likely led to an olfactory system focused on chemical detection and analysis problems relevant to survival and reproduction. By studying many odorants, we can both discover the factors that are of greatest importance to the animal and identify factors that are not as important to them.

Animals appear to be able to learn to discriminate between virtually any pair of odorant chemicals, even those that do not differ along the dimensions that are encoded by chemotopic progressions or by activity in different modules (Linster et al., 2002). Indeed, optical imaging never fails to show a small difference in the relative activation of individual glomeruli even by odorants of very similar structure (Uchida et al., 2000; Takahashi et al., 2004a), indicating that there may always be sufficient information for any olfactory discrimination if the animal is highly motivated and is trained to identify such differences. However, it also is not clear whether animals in nature would receive the dozens (or hundreds) of reinforced experiences necessary for learning these subtle discriminations (Linster et al., 2002).

In a related issue, it may not be the case that all inhibitory interactions between glomeruli or their underlying mitral cells will have uniform center-surround architecture. Rather, it is possible that certain response units will exert a disproportionate influence on some regions of the bulb compared to others (Willhite et al., 2006).

Global chemotopy

Quantitative mapping allows precise evaluation of the relatedness between different patterns, and it has revealed that overall spatial patterns of activity across the bulb are chemotopically organized. We compare pairs of data matrices of 2DG uptake and express the overall relatedness between odorant-evoked spatial patterns using various indices such as Pearson correlation coefficients and principal components analysis. The similarities and differences found in these comparisons reflect both the relative modular representations of odorant features and the chemotopic progressions within the modules, without the need for a priori definitions of modular boundaries. Overall similarities between patterns then can be used to test hypotheses concerning predicted perceptual similarities between odorants.

We very often find that overall pattern similarity is greater for odorant stimuli that share similarity along a single chemical dimension. For example, patterns are more similar for odorants that have a comparable number of carbons along a homologous series of straight-chain aliphatic odorants (carboxylic acids: Johnson et al., 1999; aldehydes, esters, primary alcohols, secondary alcohols, ethyl esters, and acetates: Johnson et al., 2004; alkanes: Ho et al., 2006a). Overall pattern similarities are proportional to similarities in the molecular length of hydrocarbons that differ in branching and bond saturation (Ho et al., 2006b). Patterns are more similar for odorants with more similar functional group positions in aliphatic alcohols and esters (Johnson et al., 2005a), and patterns are more similar for aromatic hydrocarbons with similar numbers of methyl group substituents (Farahbod et al., 2006). Finally, when viewed across large numbers of odorant pairs, the greatest similarities involve odorants that resemble each other in both functional group and hydrocarbon structure, such as odorant enantiomers and positional isomers of aromatic compounds (Johnson et al., 2002).

Combinatorial coding of odorant molecular features

There are more olfactory perceptions than there are receptors, indicating that the identity of the stimulus must be coded using a combination of responses. Moreover, multiple distinct odorant receptor types and multiple glomeruli are activated by most individual odorants (Polak, 1973; Kauer and Cinelli, 1993; Friedrich and Korsching, 1997; Malnic et al., 1999). While many of the glomeruli activated by a single odorant are located in clusters, there also can be stimulation of glomeruli in domains located in very different parts of the bulb (Stewart et al., 1979; Johnson et al., 1998; 1999; 2002; 2004; 2005b; 2006; Inaki et al., 2002). Such clusters often have a different set of chemical determinants for their activation. For example, aliphatic odorants often activate both glomerular domains directly related to their functional groups and domains related to their hydrocarbon chains (Johnson et al., 1998; 2004; Johnson and Leon, 2000a,b; Ho et al., 2006a).

In our experiments using either simple, straight-chained aliphatic compounds with different functional groups (Johnson and Leon, 2000a; Johnson et al., 2002; 2004) or carboxylic acids and esters differing dramatically in hydrocarbon structure (Johnson et al, 1998; 1999; Johnson and Leon, 2000b), it appeared that responses to functional groups and hydrocarbon elements occurred independently of one another, as if these distinct chemical features were separately detected by subsets of odorant receptors. Encouraged by these results, we formulated a hypothesis involving the combinatorial coding of discrete odorant molecular features by discrete glomerular modules (Johnson et al., 2002; Leon and Johnson, 2003). While this simple model maintains its power to predict accurately the responses to other simple aliphatic odorants with single functional groups, further studies involving more complex odorant structures have shown that there are important interactions between some chemical features, resulting in activity patterns that are not initially predictable from responses to the individual features.

Complex hydrocarbon structural features can prevent modular responses to functional group features that were identified using simple aliphatic compounds. For example, when a benzene ring is substituted with an oxygenic functional group such as a methyl ester, responses are not observed in the anterior domain responding to that functional group in aliphatic odorants. Instead, the aromatic feature trumps that response and activity is confined to the posterior, dorsal domain responding to benzyl odorants with oxygenic substituents (Johnson et al., 2005b; Farahbod et al., 2006). Another example is that glomerular responses related to oxygenic functional groups can be hindered if alicyclic structures (Johnson et al., 2007b) or triple bonds (Johnson et al., 2007a) are located near the functional group in the odorant molecule, perhaps because the hydrocarbon feature interferes with the proper positioning of the odorant functional group at the receptor binding site (Araneda et al., 2000).

When two oxygenic functional groups are present in a single odorant molecule, one rarely sees the modular responses that are related to each of the functional groups separately (Johnson et al., 2007b). Instead, such an odorant stimulates glomeruli in the region responding to highly water-soluble molecules (Johnson et al., 2007b). Responses associated with certain functional groups also are conditional on their position within a molecule (Johnson et al., 2005a).

These observations of interactions between separate odorant structural features in complex molecules indicate that the whole odorant structure must be considered in predicting the response pattern. Odorant receptors do not recognize molecular features independently of one another, and therefore an earlier, simple notion of modular coding of discrete features that still apparently holds for simple odorant structures is, as might have been expected, not applicable to more complex odorants.

The complete data involving isolated odorant chemicals perhaps better fit a model wherein intact odorants are represented by combinations of active glomeruli rather than molecular features being represented by combinations of glomerular modules. In this model, the modules would represent the locations where responses to intact molecules are used to tune or otherwise decorrelate mitral cell responses along different stimulus dimensions. The specificities of both modules and individual glomeruli are still best defined in terms of odorant molecular features given the large number of intact odorants capable of stimulating them. In this newer model, the definition of a molecular feature both for a module and for a glomerulus would be narrower than was envisioned in our earlier model (e.g., instead of the feature “carboxylic acid”, the feature relevant to the anterior-dorsal module might be “aliphatic or alicyclic carboxylic acid, methyl ester, or ethyl ester without additional oxygenic functional groups and without triple bonds or cyclopropyl structures within one carbon atom of the single permissive functional group,” and the feature for a particular glomerulus in the module might be constrained further by a range of overall molecular length.)

Odorant concentration

As measured using various imaging techniques, absolute levels of glomerular activity generally increase with increasing odorant concentration, sometimes approaching plateau values that would be predicted from saturation of odorant receptors across the various sensory neurons projecting to a given glomerulus (Stewart et al., 1979; Cinelli et al., 1995; Guthrie and Gall, 1995; Friedrich and Korsching, 1997; Johnson and Leon, 2000a; Xu et al., 2000; Fried et al., 2001; Meister and Bonhoeffer, 2001; Sachse and Galizia, 2003). In addition, at any given arbitrary level for defining the presence of a “response,” increasing odorant concentrations also are associated with an increased number of responding glomeruli. Typically, the glomeruli recruited at higher concentrations are located near the originally activated glomeruli, a consequence of the chemotopic clustering of glomeruli with similar odorant specificities.

When activity is measured across the entire glomerular layer using the 2DG method, it is possible to express the activity of each glomerulus relative to the activity of all other glomeruli in the bulb. In our experiments, we typically express this relative activity as a z score, where the uptake at each location is calculated as the number of standard deviations above or below the mean glomerular layer uptake (Johnson et al., 1998; 1999). For most odorants, there is little concentration-dependent change in this relative z-score pattern (Figure 6; Johnson and Leon, 2000a; Johnson et al., 2002; 2006), despite clear increases in absolute levels of 2DG uptake (Figure 6; Johnson and Leon, 2000a). Because most odorants are perceived to have the same odor quality at different concentrations, we reasoned that the olfactory system might use some similar relational code to transform bulbar activity patterns into perceptions (Johnson and Leon, 2000a).

Indeed, z-score patterns of uptake were found to be better predictors of odorant discriminations involving different odorants and concentrations than were patterns of absolute uptake (Cleland et al., 2007). Computational modeling further indicated that the olfactory bulb itself has circuitry that is capable of normalizing the input to generate a relational code in a manner similar to how we calculate a z score from the raw data (Cleland et al., 2007). Short-axon cells (Aungst et al., 2003) can relay information about overall levels of glomerular input across the entire olfactory bulb by way of excitatory synapses, and we calculated that they can suppress mitral cells throughout the bulb by way of their connections to inhibitory periglomerular interneurons, thereby insuring that the intensity of output signals are adjusted relative to the overall glomerular activity (Cleland et al., 2007). Indeed, there is evidence that mitral cells do not show monotonic increases in firing rates with increases in odorant concentration (Chalansonnet and Chaput, 1998). These findings suggest that all responses that are recorded in the olfactory system do not necessarily contain the coded odorant information, but may be background responses that are normalized by glomerular-layer networks. Moreover, the ability of these normalized patterns to predict perception accurately (detailed in subsequent sections), raises the possibility that such patterns actually represent the output activity of the bulb, rather than simply the input of the olfactory sensory neurons.

There are a few odorants that evoke different z score-standardized patterns of uptake at different concentrations, and also evoke different odor perceptions at different concentrations (Johnson and Leon, 2000a). For example, at low concentrations, pentanal does not stimulate activity in the anterior parts of the olfactory bulb that prefer carboxylic acid odorants, while at high concentrations anterior activity becomes prominent (Johnson and Leon, 2000a). Others using optical imaging methods also have found that different concentrations of an aldehyde can evoke unique relative activity patterns in the dorsal aspect of the bulb (Fried et al., 2001; Meister and Bonhoeffer, 2001). We now think that a likely cause of this change is the presence of acid contaminants (1-5%) in many preparations of aldehydes (Johnson et al., 2004).

Ketone odorants also evoke different z-score patterns at different concentrations (Johnson and Leon, 2000a; Johnson et al., 2004). Oddly, higher 2-hexanone concentrations cause decreases in the amount of uptake in ventrally located glomeruli while causing new responses in dorsal glomeruli (Johnson and Leon, 2000a). Rats can change the nasal distribution of odorants by adjusting their respiration patterns, and deeper sniffs may be required for full access of odorants to the most ventral and peripheral culs-de-sac of the nasal turbinates (Youngentob et al., 1987; Kimbell et al., 1997; Scott et al., 2006). Therefore, one explanation for the changes in relative patterns evoked by 2-hexanone is that the rats withhold deep sniffs at higher concentrations, a phenomenon known to occur for various other odorants (Alarie, 1973; Youngentob et al., 1987).

Odorant mixtures

Although the use of isolated odorant chemicals of defined structure has been invaluable for our understanding of the relationships between stimulus and response in the olfactory system, the fact remains that natural odor stimuli are not pure chemicals, but rather are mixtures of chemicals emitted from various objects signifying attractive or unattractive food sources, environmental cues, predators, and kin. Any natural selection that may have operated during the formation of the olfactory system would have acted in the context of these natural mixtures as opposed to individual odorant chemicals, raising the possibility that special mechanisms may have arisen either to insure the robust detection and identification of particular biologically relevant odorant mixtures or, more generally, to process information about the kinds of mixtures that often characterize natural odor objects.

Studies of invertebrate olfactory systems have shown that the presence of multiple odorants in a mixture can result in interactions at many levels. In an impressive series of systematic studies reconstructing the responses of the lobster olfactory system to natural food stimuli, individual odorants were found to interact by way of competition for receptor binding sites, synergistic stimulation of distinct excitatory receptors on the same sensory neurons, and inhibitory responses to some odorants in the face of excitatory responses to other odorants by the same sensory neurons (reviewed in Derby, 2000). Similar results for food-related mixtures have been obtained for fish (Kang and Caprio, 1997). Inhibitory interactions between odorants in arbitrary mixtures also were observed during optical imaging of calcium responses in honeybee glomeruli (Joerges et al., 1997). The situation may be somewhat different for mammals, which are generally thought to express only a single odorant receptor gene in a given sensory neuron (Serizawa et al., 2004). However, chemically related odorants indeed compete for binding to individual rodent odorant receptors (Araneda, 2000; Oka et al., 2004; Sanz et al., 2005), and there is some evidence that different odorants can excite and inhibit the same sensory neuron (Sanhueza et al., 2000; DuChamp-Viret et al., 2003), that synergistic responses can occur in single sensory neurons (DuChamp-Viret et al., 2003), that multiple receptor genes can be expressed in certain sensory neurons (Rawson et al., 2000), and that mixtures of chemically unrelated odorants can result in suppression of some responses by way of peripheral mechanisms (Bell et al., 1987). These findings suggest that mammals also may possess a foundation for mixture interactions.

In contrast with the choices of food-related stimuli in experiments on odorant mixtures using aquatic species, studies on mammals typically have used binary mixtures of arbitrary odorant pairs. The typical result has been that the mixture response is well predicted from the responses to the individual odorants (Belluscio and Katz, 2001; Lin et al., 2006). However, if natural selection has shaped mixture interactions, it would be more likely to discover their existence using naturally occurring odorant mixtures rather than arbitrary odorant pairs. Responses to urine, a natural odor mixture, have been detected in remarkably well-confined portions of the ventral part of the mouse bulb through the use of both Fos immunohistochemistry (Schaefer et al., 2001a; 2002) and electrophysiology (Lin et al., 2005). Although a particular chemical detected in urine activates similar bulbar regions and elicits similar behaviors as urine (Lin et al., 2005), the effect of other urine components on the overall mixture-evoked activity has not been investigated.

Limitations of mapping methods

Various methods can be used to measure the differential activity at each anatomical level of the olfactory system. Some of these methods are capable of tracking responses to numerous stimuli in individual animals across small increments of time, but cannot access responses across the entire structure, whereas other methods require between-animal comparisons and long odorant exposure times, but can monitor activity across an entire anatomical level. A complete understanding of olfactory processing probably will require an analysis of every level at all resolutions because each analysis technique captures a different aspect of the neural response, some of which may be more closely related to the information that the brain actually uses to build an olfactory perception. To identify which information is used, behavioral studies are required in order to identify the aspects of the neural response that are critical for perception.

Much of the evidence for chemotopic odorant representations has come from our own studies of 2DG uptake. Louis Sokoloff and coworkers originally developed the 2DG technique to measure real values of glucose utilization in the brain (Kennedy et al., 1975), but now the method is typically used in a semiquantitative manner to compare relative levels of activity in different brain regions. After being taken up through the glucose transporter in proportion to a neuron’s demand for glucose, 2DG becomes phosphorylated. Because the charged product cannot pass back through the cell membrane and because most neurons lack the enzymes capable of further catabolizing the product, the 2DG becomes trapped in the active neuron (Sokoloff et al., 1977). Most evidence suggests that 2DG uptake primarily reflects synaptic rather than somatic activity in numerous brain regions (Schwartz et al., 1979; Nudo and Masterson, 1986), although short-term experiments on radiolabeled glucose uptake indicated the possibility of uptake by both terminals and active cell bodies (Duncan and Stumpf, 1991). In the olfactory bulb, the vast majority of odorant-evoked 2DG uptake occurs in glomerular neuropil, although some labeling is detected in cell bodies (Benson et al., 1985). Because many postsynaptic dendrites in glomeruli are themselves presynaptic to other dendrites (Shepherd, 1972; Wachowiak and Shipley, 2006), the general presumption of presynaptic labeling by 2DG is not definitive with respect to which elements are active. Electron microscopic analysis of labeled glomeruli revealed very small patches of labeling surrounded by larger unlabeled regions, which suggested that olfactory nerve terminals might contribute more to the signal than postsynaptic structures (Benson et al., 1985). As discussed below, uncertainty in the exact nature of the elements being labeled is considered by some to be a disadvantage of the 2DG technique.

Another clear disadvantage of the 2DG method is that only one odorant condition is used for each animal. Because glomeruli associated with the same odorant receptor vary slightly in their position from bulb to bulb (Royal and Key, 1999; Strotmann et al., 2000; Schaefer et al., 2001b), and because of experimental variation in the preparation and sectioning of tissue, the 2DG method does not allow a comparison of the activation of a single glomerulus by multiple odorants. Rather, across-odorant comparisons in 2DG data pertain to the average and variance in activity over an area somewhat broader than a glomerulus. Optical imaging and electrophysiological approaches to measuring odorant-evoked neural activity do not have this same disadvantage. Interestingly, however, the necessity of a statistical approach in studies of 2DG uptake also may be a strength of this technique. The evident details in individual images and electrophysiological time series from single animals and neurons appear to have encouraged an almost anecdotal presentation of data in studies using these techniques, so that it is often not clear to what extent the conclusions are valid for the entire population of animals.

Measurements of 2DG uptake through autoradiography of multiple sections taken throughout the bulb provide access to the entire glomerular layer, allowing the identification of the principal responses to odorants in awake, behaving animals. Many parts of the olfactory system respond in a minor way to many odorants, probably due to the presence of both multiple, low-level contaminants and low-affinity responses to the principal compound. Any small set of cells or glomeruli therefore probably will show some kind of response to every odorant. If there is no knowledge of the presence of much larger responses to those odorants elsewhere in the system, one might come to the conclusion that the system is broadly and not differentially responsive, a very different conclusion than would be made after observing the enormously differential responsiveness that is apparent when one can view the entire bulb (Figure 7). Techniques limited in their spatial scope such as mapping responses only on the dorsal surface or sampling from only several cells randomly located through the bulb risk analyzing only the background responses. Also, if one were to use few odorant stimuli when looking at a limited set of cells or glomeruli, then one might not be aware that those same cells or glomeruli display much larger responses to other odorants, and one might come to incorrect conclusions regarding the breadth of tuning in the system. If, for these technical reasons, there appeared to be no differential spatial responses, one would be forced to look for mechanisms other than a spatial or identity code to represent odorant stimuli.

The problem of defining odorant responses is illustrated in a different way in Figure 8, which shows 2DG uptake in both “absolute” and “relative” scales (Woo et al., 2007). The absolute uptake (Figure 8, upper row) shows responses above background (yellow or warmer colors) to almost all odorants in almost all bulbar locations, which might be taken as “broad tuning” or a “highly distributed response.” Any arbitrary location in the bulb would likely show an above-background response to all odorants, with no clear systematic relationship to either odorant chemistry or perception, similar to results of imaging a small number of glomeruli or a few mitral cells. However, when each data point is expressed relative to the average response in the entire glomerular layer (z-score normalization; Figure 8, lower row), the differential responses of some bulbar locations relative to others are much easier to appreciate.

The fact that glomerular circuitry involving the short-axon cell network may accomplish something very similar to z score standardization (Cleland et al., 2007) suggests that our typical maps of 2DG uptake may in fact be modeling the spatial pattern of input to mitral cells instead of the pattern of input to glomeruli. The mitral cell input pattern would be expected to show a closer correlation with behavior than the glomerular input pattern, which is exactly what is observed across odorant concentrations (Cleland et al., 2007).

Animals in 2DG experiments are not anesthetized, which contrasts with the use of general anesthetics in all optical imaging and functional magnetic resonance imaging studies used to map glomerular activity. In our hands, the use of urethane anesthesia almost entirely abolishes glomerular uptake evoked by the powerful odorant stimuli decanal and methyl benzoate (Johnson, Ong, and Leon, unpublished observations), which suggests the possibility that the responses measured by the other methods may also have been greatly affected by anesthesia. We have not yet determined whether the suppression of 2DG responses by the anesthetic involves direct actions on sensory neurons or whether it arises indirectly from the suppression of odorant inspiration. It is interesting that wakefulness has been reported to result in a sparsening of odorant-evoked mitral cell responses despite an increase in spontaneous activity (Rinberg et al., 2006a).

A criticism that has been leveled against 2DG and certain optical imaging methods is that they do not distinguish between presynaptic and postsynaptic activity, thereby failing to provide information regarding “computations” performed at that level (e.g., Laurent et al., 2001). While it is true that the 2DG method cannot definitively distinguish olfactory nerve terminal activity from other activity in the glomerular neuropil, the strong predictive relationship between 2DG uptake patterns and perception that we will describe later in this review should obviate these concerns. A related, but more vague, criticism of the 2DG technique is that it is “indirect,” with the unstated implication that focal responses somehow arise artifactually from the use of a metabolic marker of activity (e.g., Bhalla and Bower, 1997), but a mechanism by which this artifactual result could arise has not been proposed, or shown to exist. One might reason instead that both the use of glucose for biochemical processes unrelated to activity and the potential equal weighting of secondary excitation and inhibition would cause a metabolic tracking method to show less spatial specificity than actually exists, rather than more.

The 2DG method in its typical form involves exposing animals to an odorant for 45 minutes, whereas odor perception can occur in hundreds of milliseconds (Uchida and Mainen, 2003; Ditzen et al., 2003; Abraham et al., 2004; Rinberg et al., 2006b). Thus, there may be concern that 2DG uptake shows late metabolic events unrelated to perceptual processing. We have found that exposures in 2DG studies can be as short as two minutes and yet result in similar relative patterns of uptake, although the fainter signal against the higher background of unutilized radiolabel is generally undesirable for routine mapping experiments (Woo et al., 2004). Moreover, animals trained to sniff an odorant intermittently for less than a second at a time over the course of a 45-minute period (1-2 minutes total exposure) also show a pattern of 2DG uptake similar to animals exposed continuously to the odorant for 45 minutes (Slotnick et al., 1989). Finally, optical imaging experiments and electrophysiological studies that measure rapid responses in the same areas of the bulb as the 2DG measurements (Mori, et al., 1992; Uchida et al., 2000; Takahashi et al., 2004a,b; Igarashi et al., 2005; Lin, et a., 2005) show patterns of responses that are quite comparable to those from 2DG studies.

Indeed, in addition to the close agreement between 2DG and optical imaging for numerous odorants, almost all methods that map activity across the entire olfactory bulb have come to similar conclusions regarding the locations best stimulated by any given odorant, regardless of the temporal resolution of the techniques. For example, principal responses to carboxylic acids have been localized to the dorsal part of the bulb using 2DG (Bell et al., 1987; Royet et al., 1987; Sallaz and Jourdan, 1993; Johnson et al., 1999; 2004; Johnson and Leon, 2000a,b), immediate-early gene expression (c-fos: Sallaz and Jourdan, 1993; Guthrie et al., 2000; arc: Guthrie et al., 2000; zif/268: Inaki et al., 2002), evoked field potentials (Mori et al., 1992), and recordings of individual mitral cells (Mori et al., 1992). Chemotopic progressions in representations of homologous series are found in the same locations using 2DG (Johnson et al., 1999; 2004), optical imaging, and zif268 expression (Inaki et al., 2002). Posterior and ventral responses to sulfides are found using both mitral cell unit recordings (Lin et al., 2005) and 2DG (Johnson and Leon, unpublished data). Finally, responses to isoamyl acetate are similar for 2DG (Johnson et al., 1998) and fMRI, provided that the larger variance between individual animals in the fMRI method is accounted for by averaging over many rats (Schafer et al., 2006).

In comparing the results from different imaging methods, it also is important to consider the use of different species by different labs. For example, responses to carboxylic acids appear to involve the dorsomedial region of the bulb of rabbits (Mori et al., 1992), rats (Bell et al., 1987), and mice (Royet et al., 1987), suggesting a similarity in spatial areas of activation. However, 2DG experiments show primarily posterior activation by ethyl acetate in rats (Johnson et al., 1998), while c-fos studies show primarily ventral activation by ethyl acetate in mice (Salcedo et al., 2005). To determine if this sort of difference could be due to the use of different species, we have begun an analysis of 2DG uptake in mice using odorants previously characterized in rats, and we are finding that differences in spatial patterns between these species are actually quite common (Johnson, Xu, Ali, and Leon, unpublished data).

The final test of the ability of any method to collect data relevant to odor coding is to test how well olfactory system responses predict odor perceptions. As detailed in the next section, there has been considerable success in relating odor perceptions to spatial activity patterns obtained using the 2DG method, suggesting that any real or imagined shortcomings of the technique have not compromised its usefulness in understanding principles of odor coding.

Correlations between spatial patterns of glomerular activity and behavior

Cross-habituation

Rats initially investigate novel odorants in their environment, but upon repeated presentations, their interest diminishes. If an unrelated odorant then is presented to a rat that has become habituated to the first odorant, the rat will investigate that new odorant. However, if the second odorant is perceived as being the same as the habituated odorant, then there is less investigation. This phenomenon provides a simple assay to determine whether naïve rats, such as are used in our mapping studies, distinguish between odorants. The advantage of using such a technique is that the behavior is spontaneous, does not require hundreds or thousands of rewarded exposures to the odorants and does not raise the possibility of learning-related changes in the system.

The degree of behavioral cross-habituation between odorant pairs has been quantitatively related to the similarity in overall patterns of 2DG uptake in numerous experiments. For example, habituation among members of a homologous series of aliphatic acids or alkanes was proportional to the degree of similarity in their 2DG activity patterns (Cleland et al., 2002; Ho et al., 2006a). Quantitative similarities between 2DG patterns evoked by octane and/or by eight-carbon branched alkanes or eight-carbon alkenes and alkynes also predicted the amount of cross-habituation between the odorant pairs (Ho et al., 2006b). Differences in 2DG uptake patterns evoked by pentadecane obtained from different sources (with a likelihood of different impurities) also correctly predicted that the odorants would be perceived differently in a cross-habituation experiment (Ho et al., 2006a). Finally, the enantiomers D- and L-carvone, which evoked statistically distinct 2DG uptake patterns, were perceived as being different in a cross-habituation assay, while the chemically related enantiomer pairs D- and L-limonene and D- and L-terpinen-4-ol, which were not statistically different in their evoked 2DG patterns, were not distinguished in the cross-habituation assay (Linster et al., 2001).

Rate of acquiring a learned discrimination

Although rats can learn to discriminate between even very similar pairs of odorants with extensive reinforcement, it is sometimes possible to distinguish differences in the rate of acquisition of the behaviors that are used to monitor this kind of learning. Odorant pairs that are perceived to be similar yield slower acquisition than odorants that are perceived to be different. Indeed, rats required additional reinforced trials to learn to discriminate between pairs of odorants giving more similar 2DG patterns along a homologous series of acids, affirming the results of cross-habituation assays (Cleland et al., 2002). In female mice exposed to urine from donor male mice of different genetic backgrounds, relative similarities in bulbar patterns of c-fos in situ hybridization were related to the number of trials required by mice to learn a discrimination between the odors of the urine samples (Schaefer et al., 2002).

Similar to the results from cross-habituation studies, rats required additional reinforced trials to discriminate limonene and terpinen-4-ol enantiomers compared to carvone enantiomers (Linster et al., 2002). Carvone discriminations in these rewarded digging tasks were evident before the tenth trial, whereas discriminations involving the other two enantiomers were evident before the 15^th trial (Linster et al., 2002). Using operant procedures involving water deprivation and water reward, McBride and Slotnick (2006) trained rats to discriminate between 10% dilutions of carvone enantiomers, followed by eight additional steps of dilution, using between 60 and 400 training trials at each concentration. After this extensive experience with discriminating carvone enantiomers, rats were shown to learn to discriminate between terpinen-4-ol enantiomers before the 20th trial (McBride and Slotnick, 2006). Although the authors interpreted these data otherwise, this result is fully consistent with those of the Linster et al. study (2002), where inexperienced rats could discriminate between terpinen-4-ol enantiomers before the 15th trial in a simple digging task. In general, the use of many trials of operant conditioning appears to result in successful discriminations of all odorants when they are tested at detectable concentrations (Slotnick et al., 1987; 1997; Lu and Slotnick, 1994; 1998; Bisulco and Slotnick, 2003; Slotnick and Bisulco, 2003; McBride and Slotnick, 2006), suggesting that the method is poorly suited for the investigation of odor similarities.

Odorant confusion matrix

All of the above behavioral tests of perception rely on the comparison between two odorants, and relationships among a group of odorants can be deduced only from binary comparisons. The correlation between odorant-evoked glomerular responses and odorant-evoked perceptions would be more compelling if one could demonstrate that a quantitative relationship among the patterns evoked by a larger group of unrelated odorants was paralleled by a similar quantitative relationship across their evoked perceptions. To address the comparison of multiple odors in a single experimental protocol, Steven Youngentob developed a five-odorant confusion matrix task in which rats learn to associate each of five different odorants with one of five different tunnels (Youngentob et al., 1990). Rats are given extensive training to identify the correct response tunnels with very few errors. Over thousands of trials, however, enough errors are committed to permit a rigorous statistical analysis. Individual rats tend to commit the same types of errors, revealing that some odorants are more likely to be misidentified than are others (Youngentob et al., 1990).

This experimental design was exploited to test the predictive power of spatial patterns of 2DG uptake, using five odorants of very distinct chemical structure that could not be characterized along any single chemical dimension (Youngentob et al., 2006). Both the 2DG patterns and the behavioral data then were subjected to independent multidimensional scaling analyses. Remarkably, the perceived similarity among the odorants closely matched the pattern of similarity for the 2DG response (Youngentob et al., 2006). Because the peak modular 2DG responses for some pairs of odorants did not overlap, and yet the information regarding their relative similarity was revealed in the behavior, it suggested that the overall activity pattern could be used for odor comparisons under at least some circumstances (Youngentob et al., 2006).

Multidimensional relationships between perception and epithelial activity patterns had been identified previously using these procedures (Kent et al., 1995; 2003), suggesting that bulbar patterns may further develop receptor-dependent information already present in the first neurons to respond to the odorants. Recent work also has shown a similar multidimensional correlation between antennal lobe activity patterns and perception in honeybees (Guerrieri et al., 2005). Behavioral methods that can test the perceived similarities of different odorants, such as the methods mentioned above, as well as certain others (Kay et al., 2006), are indispensable tools in identifying the neural activity that actually carries information relevant to odor perception.

Correlations between rat glomerular 2DG patterns and human odor descriptions

There is a great deal of information about odor quality perception in humans, who can name and verbally describe odors, and who can offer numerous other measures of olfactory performance not easily obtained for experimental animals (Zelano and Sobel, 2005). Given evidence that odor perception might be similar for different species (Laska et al., 1999; Laska and Galizia, 2001), relationships between rat activity patterns and human odor descriptors have been explored. A machine learning technique (support vector machine) was used to extract modules from a database of 172 of our 2DG uptake images on the basis of these modules being able to predict ten human odor descriptors (Yamanaka and Gutierrez-Osuna, 2006). The maps of 2DG uptake were able to predict these odor descriptors at well above chance levels.

As mentioned previously, most odorants evoke the same perceived odor at all concentrations, but certain odorants change in perception with intensity. We found that two odorants, pentanal and 2-hexanone, which are reported by humans to change in odor quality with concentration (Arctander, 1994) also exhibited changes in 2DG uptake patterns with concentration, while three other odorants not reported to change in odor evoked consistent uptake patterns at different concentrations (Johnson and Leon, 2000a).

Although there appears to be at least an overall similarity in olfactory coding between humans and rodents, there also appear to be species-specific differences in the responses of the two olfactory systems to particular classes of odorants. For example, rats exhibit a more focal and intense glomerular response to 2-methylbutyric acid than to other acids, perhaps due to its natural occurrence in rat feces (Johnson and Leon, 2000b). The differential patterns evoked by carboxylic acids of different hydrocarbon structure were not associated with proportionate differences in our own perceptions of these compounds. On the other hand, differences in bond saturation and branching changed our impressions of the odor of hydrocarbons, but did not change either the evoked 2DG pattern or odor perception in rats (Ho et al., 2006b). Heterocyclic compounds containing either oxygen, such as furans, or nitrogen, such as indole, have strong odors to humans, and yet evoked very little 2DG uptake in the rat olfactory bulb (Johnson et al., 2006). Many heterocyclic compounds are formed primarily by way of cyclization of other molecules as a result of heating, such as occurs during the cooking of sugars or amino acids. The odors of these heterocyclic compounds are likely to signify biologically relevant food sources to humans, and therefore, such receptors may have been positively selected. Rodents probably would have received less benefit from any selective detection of these compounds.

Interventive experiments testing the relevance of chemotopic representations

Patterned electrical micro-stimulation of the olfactory bulb

The most definitive interventive tests of bulbar spatial coding hypotheses appear in a series of experiments in which spatial patterns of activity were imposed on the rat olfactory bulb using electrical stimulation (Monod et al., 1981; 1989; Mouly et al., 1985; Mouly and Holley, 1986). Based on observations that humans report sensations of odor when their olfactory bulbs are electrically stimulated, Monod and coworkers (1981) stimulated rat olfactory bulbs and found that the rats began sniffing and orienting their noses in the direction of the incoming air, a typical reaction to the presence of a novel odorant. They further found that such bulbar stimulation could take the place of an odor stimulus in toxicosis conditioning (Monod et al., 1981).

By pairing one site of bulbar stimulation with water that was made palatable with sucrose, and pairing another site with water that was made unpalatable with quinine, rats could be taught to discriminate between different spatial locations of activity as if the different locations were equivalent to different odors (Mouly et al., 1985). The rats also could discriminate between pairs of activity patterns established using sets of four electrodes, with the two sets of electrodes being interdigitated with one another (Mouly et al., 1985). The rats required more training trials to discriminate these patterns as the distance decreased between the positively and negatively associated electrodes (Mouly et al., 1985). This finding is predicted by spatial coding hypotheses and recalls the greater difficulty that rats have in learning to distinguish odorants evoking more similar spatial activity patterns (Linster et al., 2002; Cleland et al., 2002; Schaefer et al., 2002). Similarly, rats could discriminate between a pair of three-site stimulation patterns involving two of the same electrodes, but when a set of four electrodes was stimulated with different intensities at two of the positions, more trials were required to learn this task (Mouly et al., 1985). Again, these data strongly support the existence of an identity code that is related to locations of activity.

On the other hand, rats were not able to learn a discrimination involving stimulation of the same sets of electrodes at different times during the respiration cycle (Monod et al., 1989). Thus, although different spatial locations of stimulation contained information recalling different perceived odors, different temporal aspects of stimulation evoked behavior consistent with a single odor. This observation favors spatial or “identity” coding hypotheses over temporal coding hypotheses, as will be discussed at greater length below.

Effects of odorant enrichment on subsequent habituation

Rats that have undergone daily exposure to individual odorants or odorant pairs spontaneously distinguish novel and familiar odorant pairs that are not discriminated by naïve rats in a cross-habituation assay (Mandairon et al., 2006a,b). The effectiveness of the enrichment procedure was related to the overall similarity between the 2DG pattern evoked by the enrichment odorant and the patterns evoked by the test odorants (Mandairon et al., 2006a,b). For example, enrichment with odorants evoking 2DG patterns that overlapped a great deal with test odorant patterns enhanced spontaneous discrimination, while experience with patterns that did not overlap much did not enhance discrimination (Mandairon et al., 2006a,b).

Perceptual relevance of particular receptors and sensory neurons

There are a number of interventive experiments showing differential contributions of certain sets of glomeruli, receptors, or sensory neurons to the perception of particular odorants. The clearest cases involve invertebrate models such as the special involvement of the V glomerulus in carbon dioxide avoidance in Drosophila (Suh et al., 2004), the special involvement of the odr-10 receptor gene to diacetyl chemotaxis in C. elegans (Sengupta et al., 1996), and the special involvement of a single sensory neuron in chemotactic responses to several other odorants in C. elegans (Wes and Bargmann, 2001). In each of these cases, elimination of particular neurons or genes blocked the ability to perceive specific odors, findings consistent with the presence of an identity code in which specific neurons carry specific odorant information.

Evidence for the perceptual importance of particular odorant receptors is also emerging from studies on rats. Perfusion of the rat nasal cavity with the lectin concanavalin A, which selectively binds to specific carbohydrate moieties on a subset of glycoproteins, impaired the ability of rats to learn a response to low concentrations of the odorant D-carvone, although learning the response to the enantiomer L-carvone was unaffected (Kirner et al., 2003). Intranasal concanavalin A also impaired detection of low concentrations of dimethyl disulfide while not affecting detection of ethyl acetate (Apfelbach et al., 1998). These studies suggest that there are subsets of sensory neurons that are specially involved in the perception of particular odorants. Although animals can detect ethyl acetate even after concanavalin A application, the odor of the chemical apparently changes in quality, because rats do not recognize the compound if they are trained to the odorant in the absence of the lectin and then tested in its presence (Apfelbach, 2004). As measured by Fos-like immunoreactivity, the spatial activity pattern in the olfactory bulb in response to ethyl acetate was correspondingly changed by intranasal concanavalin A, while the pattern evoked by L-carvone was unaffected (Apfelbach, 2004).

In another set of studies, polyclonal antibodies to the rat I7 odorant receptor were applied to the olfactory mucosa with the intention of disrupting interactions between this receptor and one of its preferred odorant ligands, octanal (Araneda et al., 2000; Deutsch and Apfelbach, 2006). After application of the antibody, rats were impaired in their detection of octanal, but not in their detection of another aldehyde, citral (Deutsch and Apfelbach, 2006). The Fos-like immunoreactivity evoked by octanal in the glomerular layer of the olfactory bulb was accordingly reduced by the antibody (Deutsch and Apfelbach, 2006). Although it is possible that the polyclonal antibody to the I7 peptide sequence cross-reacted with a number of related odorant receptors, this result nevertheless appears to show a remarkable degree of importance of specific receptors to perception in rats. The specificity of this relationship is probably related to the focal 2DG response evoked by octanal and the close relationship between the location of this response and the target of the projection of sensory neurons expressing the I7 receptor (Johnson et al., 2004).

The toxin dichlobenil causes the death of sensory neurons in zone 1 of the mouse epithelium, and this observation was exploited to test the relative importance of this zone for the detection of different odorants (Vedin et al., 2004). The toxin increased the threshold of detection for certain odorants, while the detection of other odorants was unaffected, suggesting that different parts of the olfactory epithelium were involved in the perception of different odorants (Vedin et al., 2004). Also, destruction of large portions of the epithelium using methyl bromide gas disrupted odor perception as measured in a confusion matrix task, with differential effects on specific odorants (Youngentob and Schwob, 2006). Given the topography of the epithelium-to-bulb projection (Schoenfeld and Cleland, 2005), these results suggest that different parts of the olfactory bulb are involved in the perception of different odorants.

It is likely that the use of gene-altered mice soon will allow more specific tests of the contribution of particular odorant receptors to the perception of particular odors. Already, m