Report

Combinatorial Effects of Odorant Mixes in Olfactory Cortex

See allHide authors and affiliations

Science  10 Mar 2006:
Vol. 311, Issue 5766, pp. 1477-1481
DOI: 10.1126/science.1124755

This article has been retracted. Please see:

Abstract

In mammals, each odorant is detected by a combination of different odorant receptors. Signals from different types of receptors are segregated in the nose and the olfactory bulb, but appear to be combined in individual neurons in the olfactory cortex. Here, we report that binary odorant mixes stimulate cortical neurons that are not stimulated by their individual component odorants. We propose that cortical neurons require combinations of receptor inputs for activation and that merging the receptor codes of two odorants provides novel combinations of receptor inputs that stimulate neurons beyond those activated by the single odorants. These findings may explain why odorant mixtures can elicit novel odor percepts in humans.

Odor detection is mediated by odorant receptors (ORs) (1, 2), which are located on olfactory sensory neurons in the nasal olfactory epithelium (OE). Signals generated in these neurons in response to odorants are transmitted to the olfactory bulb (OB) of the brain, which in turn transmits signals to the olfactory cortex (OC) (35). ORs comprise a diverse family of receptors that number ∼1000 in mice (68). Different odorants are detected by different, but sometimes partially overlapping, combinations of ORs (9, 10). Each sensory neuron expresses a single allele (11) of one OR gene (9). Neurons with the same OR are dispersed in the OE (12, 13). However, their axons synapse in a few OR-specific glomeruli in the OB. The result is a stereotyped map of OR inputs in which signals from different ORs are segregated in different glomeruli and their associated mitral cell relay neurons (1416). In the OC, the axons of OB mitral cells carrying input from a given OR synapse with specific clusters of pyramidal neurons. The result is a stereotyped map of OR inputs distinctly different from that in the OB (17). Inputs from different ORs are mapped onto partially overlapping clusters of pyramidal neurons, and individual neurons appear to receive signals derived from multiple different ORs (17).

The arrangement of OR inputs in the OC raises the intriguing possibility that single cortical neurons can integrate signals from different ORs that detect the same odorant (the odorant's “receptor code”) and, thus, perform an initial step in the reconstruction of an odor image from its deconstructed features. This could derive from a scenario in which the OC neuron not only receives combinatorial OR inputs, but actually requires coincident inputs from more than one OR for its activation. One prediction of this model is that a binary mix of odorants would activate OC neurons beyond those activated by its component odorants. In the simplest version of this model, odorants A and B are each detected by two ORs (1+2 and 3+4, respectively) and the activation of a cortical neuron requires simultaneous inputs from two ORs. Neurons activated by odorant A or B alone receive combined signals from ORs 1+2 or 3+4, respectively. However, a mix of odorants A and B, in addition, stimulates neurons that receive signals from other pairs of the four ORs, such as ORs 1+3 or 2+4.

To test this model, we compared the responses of mouse OC neurons to binary mixtures of odorants versus their individual components. We did this using Arc catFISH (cellular compartment analysis of temporal activity by fluorescence in situ hybridization), a technique that allows visualization of neuronal responses to two different experiences (18). The immediate early gene Arc (arg3.1) is induced by depolarization in hippocampal neurons and appears in visual cortex in response to natural, retina-dependent sensory input (1921). On induction, Arc mRNA is seen first only in the nucleus and later only in the cytoplasm (18). By appropriate temporal spacing of different stimuli, the subcellular location of Arc mRNA can be used to distinguish responses of individual neurons to those stimuli.

In initial experiments, we asked whether exposure to odorants induces Arc mRNA expression in the mouse OC (22). As in hippocampus and visual cortex, Arc expression could be induced in the OC by depolarization and implicates N-methyl-d-aspartate (NMDA) glutamate receptors involved in excitatory synaptic transmission (19, 20). Unanesthetized, unrestrained mice were exposed to the odorant eugenol for 1 min and their brains removed after 5, 15, or 30 min. Tissue sections collected from the brains were subjected to fluorescence in situ hybridization (FISH) with an Arc cRNA probe, and the sections were then counterstained with 4′,6′-diamidino-2-phenylindole (DAPI) to reveal the locations of neuronal nuclei. We focused on one major OC area, the anterior piriform cortex (APC) (5). The hybridized sections showed fluorescent Arc mRNA signal in a subset of APC neurons in layers II and III, the locations of pyramidal neurons that receive direct synaptic input from the OB (5). Similar to observations with hippocampal neurons (18), Arc mRNA signal was seen at one or two bright spots in the nucleus 5 min after odorant exposure, and subsequently appeared in the cytoplasm, where it was seen exclusively in most Arc+ neurons after 30 min (Fig. 1, A to C).

Fig. 1.

Dynamic subcellular localization of Arc mRNA in olfactory cortex. An Arc cRNA probe (red) was hybridized to cortical sections 5 (A), 15 (B), or 30 (C) minutes after exposure to eugenol and cell nuclei were visualized with DAPI (blue). In most labeled neurons, Arc mRNA was first seen at one to two intranuclear spots (A) and then appeared in the cytoplasm (B), where it was found exclusively after 30 min (C). Exposure to two odorants administered first separately (at –30 min), and then as a binary mix (at –5 min), resulted in many neurons with only nuclear Arc signal (arrowheads) in the vicinity of neurons with both nuclear and cytoplasmic Arc signal (arrows) (D). Scale bar, 5 μm.

To further characterize odorant-induced Arc expression in the OC, we analyzed the subcellular patterning of Arc mRNA in coronal sections at 70-μm intervals spanning the APC. Similar to odorant-induced patterns of c-Fos+ neurons in the APC (23, 24), Arc+ neurons were highly distributed along the anterior-posterior length of the APC (Fig. 2A), and the number of Arc+ neurons varied among cortices, with 1431 ± 565 and 666 ± 310 Arc+ neurons detected 5 and 30 min after odorant exposure, respectively (Table 1). However, the percentages of neurons with Arc mRNA in different subcellular compartments were similar among cortices (three cortices from two mice per condition) (Table 1). Five minutes after exposure to eugenol, the majority of Arc+ neurons (81 ± 3%) had Arc mRNA only in the nucleus whereas, after 30 min, most (87 ± 2%) had Arc signal only in the cytoplasm (Table 1).

Fig. 2.

Distribution of Arc+ neurons after odorant exposure. These graphs show the number of neurons with Arc mRNA in the nucleus alone (red), nucleus and cytoplasm (blue), or cytoplasm alone (green) at 70-μm intervals along the anterior-posterior length of the APC in individual mice. (A) Five min after exposure to eugenol (left), most Arc+ neurons contained only nuclear Arc mRNA, but 30 min after exposure (right) most contained only cytoplasmic Arc signal. (B) In mice exposed to a binary odorant mix twice (at –30 and –5 min) (left), neurons with both nuclear and cytoplasmic Arc mRNA predominated, but when two odorants were first given separately and then as a mix (right), many neurons had Arc mRNA in only the nucleus.

Table 1.

Number of Arc+ neurons in olfactory cortex. Average ± SD is indicated.

N only (%)C only (%)N + C (%)TotalNet N/N total, %Net C/C total, %
Eugenol
    5 min
        1 618 (78) 72 (9) 103 (13) 793
        2 1537 (82) 75 (4) 255 (14) 1867
        3 1358 (83) 64 (4) 211 (13) 1633
1171 ± 487 (81 ± 3) 70 ± 6 (6 ± 3) 190 ± 78 (13 ± 1) 1431 ± 565
    30 min
        1 7 (2) 286 (88) 33 (10) 326
        2 60 (8) 628 (85) 52 (7) 740
        3 38 (4) 822 (88) 73 (8) 933
35 ± 27 (5 ± 3) 577 ± 271 (87 ± 2) 53 ± 20 (8 ± 2) 666 ± 310
Odor pair 1
    mix-mix
        1 104 (7) 179 (11) 1289 (82) 1572
        2 160 (11) 161 (11) 1114 (78) 1435
        3 187 (12) 174 (11) 1193 (77) 1554
150 ± 43 (10 ± 3) 171 ± 9 (11 ± 0) 1199 ± 87 (79 ± 3) 1520 ± 74 8 10
    sep-mix
        1 662 (32) 205 (10) 1220 (58) 2087
        2 661 (32) 233 (11) 1174 (57) 2068
        3 599 (35) 184 (11) 913 (54) 1696
641 ± 36 (33 ± 2) 207 ± 25 (11 ± 1) 1102 ± 166 (56 ± 2) 1950 ± 220 36 13
Odor pair 2
    mix-mix
        1 97 (10) 182 (19) 701 (72) 980
        2 113 (8) 130 (10) 1091 (82) 1334
        3 114 (9) 153 (12) 987 (79) 1254
108 ± 9 (9 ± 1) 155 ± 26 (14 ± 5) 910 ± 202 (77 ± 5) 1173 ± 186 10 10
    sep-mix
        1 463 (34) 195 (14) 715 (52) 1373
        2 519 (39) 120 (9) 696 (52) 1335
        3 432 (32) 68 (5) 851 (63) 1351
471 ± 44 (35 ± 4) 128 ± 64 (9 ± 5) 754 ± 85 (56 ± 6) 1353 ± 19 39 9
Odor pair 3
    mix-mix
        1 117 (15) 63 (8) 588 (77) 768
        2 110 (20) 38 (7) 408 (73) 556
        3 68 (9) 75 (10) 603 (81) 746
        4 50 (7) 64 (9) 581 (84) 695
86 ± 32 (13 ± 6) 60 ± 16 (9 ± 1) 545 ± 92 (79 ± 5) 691 ± 95 7 3
    sep-mix
        1 355 (37) 100 (10) 504 (53) 959
        2 261 (37) 72 (10) 368 (52) 701
        3 546 (36) 90 (6) 868 (58) 1504
        4 516 (36) 65 (5) 850 (59) 1431
420 ± 135 (37 ± 1) 82 ± 16 (8 ± 3) 648 ± 251 (56 ± 4) 1149 ± 384 38 5
DMSO
    1 36 (26) 57 (41) 47 (34) 140
    2 55 (39) 53 (37) 34 (24) 142
    3 63 (38) 43 (26) 59 (36) 165
    4 45 (38) 38 (32) 36 (36) 119
50 ± 12 (35 ± 6) 48 ± 9 (34 ± 6) 44 ± 11 (31 ± 5) 142 ± 19
Water
    1 2 (2) 95 (79) 24 (19) 121
    2 4 (4) 81 (84) 11 (11) 96
    3 0 (0) 8 (100) 0 (0) 8
    4 0 (0) 18 (72) 7 (28) 25
2 ± 2 (3 ± 2) 53 ± 44 (84 ± 12) 11 ± 10 (17 ± 12) 66 ± 54

We next used the temporal patterning of Arc mRNA expression to compare the responses of individual OC neurons to binary mixtures of odorants versus their component odorants. At –30 min (30 min before brain removal), mice (n = 6) were exposed sequentially to two different odorants, spaced one minute apart, and then at –5 min, they were exposed to a mixture of the two odorants. Control mice (n = 6) were exposed to the same odorant mixes at –30 and –5 min. We tested three binary mixes of odorants (“odor pairs”) with diverse structures: (i) eugenol (clove) and dimethyl pyrazine (chocolate, nuts), (ii) methenyl methyl ether (citrus) and methylamine (fishy), and (iii) vanillin (vanilla) and ethyl butyrate (apple). A 10-fold increase in odorant concentration can recruit additional ORs into an odor response (9) and increase the number of OC neurons induced to express c-Fos (23). Although the odorants in each binary mix had dissimilar structures, we could not exclude the possibility that some ORs might nonetheless recognize both odorants in a mix. To exclude the possibility that an odorant mix might thereby double the odorant concentration at some ORs, with one exception (odor pair 1, number 1 in Table 1), we doubled the concentration of each odorant when odorants were given separately rather than as a mix.

In mice exposed to an odorant mix twice, the majority of Arc+ APC neurons had Arc mRNA in both the nucleus and cytoplasm, indicating that they had responded to both exposures to the mix (mix-mix, Table 1). These neurons constituted 79 ± 3% of all Arc+ neurons for odor pair 1, 77 ± 5% for odor pair 2, and 79 ± 5% for odor pair 3 (n = 3 to 4 cortices from two mice per odor pair). For each odor pair, a small percentage of labeled neurons had Arc mRNA in only the nucleus (9 to 13%) or cytoplasm (9 to 14%) (Table 1). These percentages were slightly higher than those seen in mice exposed once to eugenol at –30 min (2 to 8% nuclear Arc signal only) or –5 min (4 to 9% cytoplasmic Arc signal only), which suggests that a small percentage of neurons responded to one exposure to the mix, but not the other.

Strikingly different results were obtained when animals were exposed to two odorants separately at –30 min and then as a mix at –5 min (sep-mix, Table 1). In these mice, many neurons contained Arc mRNA in both the nucleus and cytoplasm, which indicated that they had responded to an odorant mix, as well as to one (or both) of its individual components. However, in sharp contrast to when odorants were delivered twice as a mix, there were also many neurons that had Arc mRNA only in the nucleus, which indicated that they had responded to an odorant mix, but not to either of its component odorants alone. In these animals, 33 to 37% of Arc+ neurons had only nuclear Arc signal compared with 9 to 13% in animals given an odorant mix twice. The percentage of APC neurons with Arc mRNA in only the nucleus averaged 33 ± 2% for odor pair 1, 35 ± 4% for odor pair 2, and 37 ± 1% for odor pair 3 (n = three to four cortices from two mice per odor pair). Thus, for each odor pair, about one-third of neurons that responded to the mix failed to respond to either of the single odorants in the mix.

In mice exposed to odorant mixtures by either exposure protocol, Arc+ neurons were highly distributed along the anterior-posterior length of the APC (Fig. 2B). This was the case for neurons with Arc mRNA in both the nucleus and cytoplasm, as well as for neurons with Arc signal only in the nucleus. Most neurons with only nuclear Arc signal were located in the vicinity of neurons with both nuclear and cytoplasmic Arc signal (Fig. 1D). Given that cortical neurons that receive input from the same OR are found in distinct clusters and inputs from different ORs can partially overlap (17), it is possible that neighboring cells with different subcellular patterns of Arc mRNA receive input from different, but partially overlapping sets of ORs.

To more accurately assess the populations of OC neurons responsive to odorants in these studies, we took into account data obtained from control animals exposed twice (at –30 and –5 min) to odorant solvents alone [dimethyl sulfoxide (DMSO) for odor pairs 1 and 3 and water for odor pair 2] (Table 1). We first subtracted the average numbers of neurons with different subcellular patterns of Arc mRNA in the controls from the corresponding numbers in the odorant-exposed animals. We then used the resulting net values for each experiment to calculate the percentage of neurons stimulated by the second odorant exposure, but not the first (net N only/N total), and the percentage stimulated by the first odorant exposure, but not the second (net C only/C total).

In animals given an odorant mix twice, only 7 to 10% of neurons responded to the second exposure, but not the first (Table 1, net N only/N total). In sharp contrast, in animals exposed to odorants first separately and then as a mix, 36 to 39% of neurons that responded to the odorant mix at –5 min did not respond to its component odorants at –30 min. Compared with animals given an odorant mix twice, this percentage increased by 28% (4.5-fold) for odor pair 1, 29% (3.9-fold) for odor pair 2, and 31% (5.4-fold) for odor pair 3. Thus, for each odor pair, about 30% of cortical neurons that responded to the binary odorant mix did not respond to either of the single odorants in the mix.

Interestingly, the converse was not seen. Only a small percentage of neurons responded to the first odorant exposure, but not the second (Table 1, net C only/C total), regardless of whether the first exposure was to odorants delivered separately (5 to 13%) or as a mix (3 to 10%). The percentage changed little when the first odorant exposure was to separate odorants rather than their binary mix (1.3-, 0.9-, and 1.7-fold for odor pairs, 1, 2, and 3, respectively). This indicates that most OC neurons that responded to a single odorant also responded to an odorant mix containing that odorant. Nonetheless, the slight differences seen for odor pairs 1 and 3 in the two exposure protocols suggest that a small percentage of neurons responsive to the individual components of these mixes might have been suppressed when the single components were mixed (mixture suppression). The extent to which this occurs could vary among combinations of odorants. Mixture suppression has previously been reported in the OB (25) and OC (26), as well as the OE (27).

In summary, these studies show that binary odorant mixes stimulate many cortical neurons beyond those that respond to their individual component odorants. About 30% of OC neurons that responded to a binary odorant mix were not stimulated by either of the single odorants in the mix. It was not possible to examine Arc expression in OB mitral cells in these studies, because the mitral cells were intermingled with numerous granule cell interneurons constitutively expressing Arc. However, the present findings in the OC contrast sharply with electrophysiological studies of the mammalian OB, which show that most or all mitral cells responsive to an odorant mix also respond to one component of the mix (25, 28). This is also the case for binary odorant mixes in fish, even though each fish mitral cell is connected to several glomeruli rather than one as in mammals (29, 30). With combinations of large numbers of odorants, more complex interactions may (30) or may not (28) occur in the OB, such as mixture suppression resulting from interneuron-mediated lateral inhibition among mitral cells (30).

The present findings indicate that the OC has an integrative capacity that is lacking from the OB. Signals from different ORs are segregated in different neurons in the nose and bulb, but individual pyramidal neurons in the OC appear to receive combined inputs from different ORs (17). Given that each odorant is detected by a combination of ORs (its receptor code) (9, 10), a straightforward explanation for the present results is that the activation of an OC pyramidal neuron requires input from more than one OR (Fig. 3). In this model, neurons stimulated by an odorant mix, but not its individual components, are those that receive novel combinations of OR inputs that result from merging the receptor codes of two odorants. This would represent a synthetic operation in which the deconstructed features of an odorant, which are carried by different OR inputs, begin to be reconstructed at the level of individual cortical neurons in order to generate a unique odor perception. Synthetic operations could also derive from excitatory association fibers that interlink OC pyramidal neurons and provide a source of disynaptic excitatory input from the OB (31, 32). In this scheme, a neuron might be activated solely by the combined excitatory inputs from other OC neurons responsive to different components of an odorant mix.

Fig. 3.

OR inputs and odor responses in the olfactory system. Inputs from different ORs are segregated in both the olfactory epithelium (OE) and olfactory bulb (OB), but are combined in single neurons in the olfactory cortex (OC). In this simple model, activation of an OC neuron requires input from two ORs. Two odorants are recognized by different pairs of ORs and activate OC neurons that receive input from those OR pairs. Mixing of the two odorants produces additional pairs of OR inputs that stimulate OC neurons that are not activated by either odorant alone.

Interestingly, these findings provide a potential explanation for certain odor mixture effects in humans. Humans have only a limited capacity to detect individual odorants in an odorant mix (33). With binary odorant mixes, individual components may be detected, but lose a described quality, such as “strawberry.” Moreover, mixing two odorants can elicit a novel odor perception. For example, in certain proportions, a binary mix of eugenol (clove) and phenylethyl alcohol (rose) is perceived as “carnation,” a distinctly different scent (34). The present studies suggest that these mixture effects may be due to the novel cortical representations that result from mixing odorants. Given that most natural odors derive from complex blends of odorants, it is quite possible that they emerge from cortical representations that bear only a remote resemblance to those of their component odorants.

Supporting Online Material

www.sciencemag.org/cgi/content/full/311/5766/1477/DC1

Materials and Methods

References

References and Notes

View Abstract

Navigate This Article