Distinct odorant receptor response patterns to aliphatic odorants in freely behaving mice

Abstract

In mammals, odors are encoded by a combinatorial code determined by the pattern of responses across hundreds of odorant receptors expressed monogenically and monoallelically in olfactory sensory neurons. The compositions of these receptor response patterns are largely unknown and overlap between them has yet to be explored. Activity-dependent reporter gene expression in freely behaving S100a5-tauGFP mice allowed capture of activated olfactory sensory neurons and identified 168 receptors responsive to moderate concentrations of 1 or more of 12 aliphatic (5 to 8 carbons) ketones, alcohols, and carboxylic acids. These 12 response patterns are remarkably different, with only 19% of the receptors responding to more than 1 of these odorants. This distinctiveness corresponds with the ease of discrimination of these odorants and may help maintain perceptual constancy in the face of response pattern variability, such as across odorant concentrations. This set of 168 receptors is not specific to aliphatic odorants but instead has 16% overlap with the receptors responsive to 7 odors tested previously in vivo, consistent with a receptor repertoire evolved to produce combinatorial codes. Aliphatic odorant response pattern similarity depends more upon odorant functional group than carbon chain length but the impact of chain length increases with the number of carbons. The response patterns to these aliphatic odorants are mostly composed of unrelated receptors, except some patterns contain minor subsets of closely related receptors. These findings argue that the major selective forces driving OR evolution are expansion of the odorant receptor gene family and the production of distinct response patterns.

1. Introduction

Mammals detect volatile chemicals (odorants) via their physical interactions with protein receptors located in the plasma membrane of cilia elaborated from the dendritic knobs of olfactory sensory neurons (OSNs). Known odorant molecules number in the many thousands, but the number of possible odorants is in the billions (Mayhew et al. 2022). To cope with this vast universe of odorants, the olfactory system evolved hundreds of odorant receptors (ORs). For example, humans have ∼400 intact ORs, mice ∼1,100, and elephants nearly 2,000 (Zhang and Firestein 2002; Liberles and Buck 2006; Niimura et al. 2014). In addition, mammals evolved receptors specialized for detecting odorants containing amines, called trace amine-associated receptors (TAARs). In mice, 14 of the TAARs are expressed in OSNs. Each OSN expresses just 1 allele of 1 OR or TAAR gene (Chess et al. 1994; Mombaerts 2004; Liberles 2015). This allows the response patterns of ORs and TAARs to be unambiguously transmitted to the brain because the axons innervating each glomerular neuropil in the olfactory bulb are solely those of OSNs expressing the same OR or TAAR (Ressler et al. 1994; Vassar et al. 1994; Mombaerts et al. 1996). Receptor response patterns are therefore the foundations of odor recognition and discrimination. Given that odorants and their possible combinations into odor mixtures greatly outnumber ORs, the ability to discriminate odors depends on combinatorial codes that arise from differences in receptor response patterns rather than specific receptors for each odorant or type of odorant. Measures of physiological responses in samples of OSNs or olfactory bulb glomeruli confirm that OSN activity forms combinatorial codes that differ between odors (Duchamp-Viret et al. 1999; Malnic et al. 1999; Oka et al. 2006; Nara et al. 2011). With very few exceptions the identities of the receptors giving rise to combinatorial codes are unknown (McClintock et al. 2014; Jiang et al. 2015; von der Weid et al. 2015; de March et al. 2020; Hu et al. 2020). This impedes understanding how the differential sensitivities of ORs (and to a lesser extent, of TAARs) mesh together to serve as a sensor array capable of producing combinatorial codes for such a vast and structurally discontinuous set of stimuli.

Relevant to the encoding of combinatorial codes by ORs are 2 selective forces that appear to have driven their evolution. First is the need to detect odorants spread across the vast breadth of the universe of possible volatile chemical structures. This favors evolution of large numbers of ORs, each responsive to multiple odorants. Based on the general principle that receptors tend to respond to structurally similar ligands, broad responsiveness should arise mostly from sensitivity to physicochemical features common to the shared odorant agonists. Second is the need for distinctly different combinatorial codes. The more that OR response patterns differ, the more distinct are the combinatorial codes received by the brain and the easier are the tasks of odor discrimination and odor recognition. The first selective force drives gene duplication, which favors the production of sets of related receptors responsive to structurally similar odorants and therefore a bias for overlapping receptor response patterns to structurally similar odorants. However, the overlap in these patterns is likely to decline over time because the second selective force favors mechanisms that diversify OR response patterns. The balance between these 2 selective forces should be apparent in the breadth of OR response patterns, the relatedness of ORs in response patterns, and the degree of OR response pattern overlap between structurally similar odorants.

We measured the receptor response patterns to a set of aliphatic ketones, alcohols, and carboxylic acids in freely behaving mice with an in vivo assay, often called the Kentucky assay. It uses S100a5-tauGFP mice in which the activity-dependent S100a5 gene locus expresses enhanced green fluorescent protein (GFP) in response to brief, intermittent odor stimulation during the natural nocturnal activity period of the mice, allowing capture of recently responsive OSNs by fluorescence-activated cell sorting (McClintock et al. 2014; de March et al. 2020; McClintock et al. 2020a; McClintock et al. 2020b). Transcriptomics of GFP⁺ and GFP⁻ cell samples allow simultaneous measurement of all ORs and TAARs, revealing those receptors specifically enriched in responsive OSNs and thereby identifying them as responsive to the odorant or odor mixture tested. We find OR response patterns identified by this assay to be overlapping but also quite distinct, even between the most structurally similar odorants. The functional groups of the tested odorants—ketone, alcohol, and carboxylic acid—best explain OR response pattern similarity, especially at short carbon chain lengths. However, by 8 carbons instances of similarity across functional group become more common. Intriguingly, ORs responsive to an odorant are mostly unrelated sequences, something that has implications for understanding OR evolution.

2. Materials and methods

2.1. Chemicals

Sigma Aldrich (St. Louis, MO) was the source of the odorants.

2.2. In vivo assay

The in vivo assay procedures have been described in detail previously (McClintock et al. 2014). Briefly, this assay takes advantage of odor-stimulated expression of GFP from the activity-dependent S100A5 gene locus in the S100a5-tauGFP mouse (The Jackson Laboratory, stock number 6709), a mutant strain in which the coding exons of S100a5 have been replaced by a sequence encoding a fusion of tau and GFP (McClintock et al. 2014). The S100a5-tauGFP mice used in this project have been back-crossed for more than 11 generations against C57BL/6J. All procedures with mice were done according to protocols approved by the Institutional Animal Care and Use Committee of the University of Kentucky. After 26 h in individual chambers ventilated at 3.1 L/min with filtered air, heterozygous S100a5-tauGFP mice were stimulated every 20 min during their nocturnal activity period by 10 s flushes of the headspace air from 50 mL vials containing 5 mL of odorant dilutions in dimethyl sulfoxide while controls simultaneously experienced headspace air flushed from vials containing the dimethyl sulfoxide solvent (Table 1). This produced brief, intermittent stimuli that reached the mice within 1 s and tailed off within seconds. Fluorescence-activated cell sorting collected both GFP⁺ and GFP⁻ cells from dissociated olfactory epithelia. We measured the amount of every mRNA species in these samples (Affymetrix Clariom S Arrays, performed by the University of Kentucky Microarray Facility) and determined which OR and TAAR mRNAs were enriched in samples of recently activated OSNs. Because each OSN only expresses a single OR or TAAR gene, the receptor mRNAs enriched in odor-stimulated samples compared to clean air controls must encode the ORs or TAARs that responded to the odorants tested.

Table 1.

Aliphatic alcohols, ketones, and η-carboxylic acids tested with their CAS numbers and the concentration of odorant used in the source dilutions.

Odorant	CAS #	Amount (g/mL)
1-Pentanol	71-41-0	0.412
1-Hexanol	111-27-3	0.205
1-Heptanol	111-70-6	0.410
1-Octanol	111-87-5	0.165
2-Pentanone	107-87-9	0.040
2-Hexanone	591-78-6	0.081
2-Heptanone	110-43-0	0.122
2-Octanone	111-13-7	0.164
η-Pentanoic acid	109-52-4	0.140
η-Hexanoic acid	142-62-1	0.372
η-Heptanoic acid	111-14-8	0.459
η-Octanoic acid	124-07-2	0.364

ORs are labeled by their Olfr designations herein. To cross-reference these names with a recent update in the names of these genes and proteins by the National Center for Biotechnology Information (e.g. Olfr398 is a synonym for Or1r1), see Supplementary Tables S1 and S2.

The microarray data from this project are available in the Gene Expression Omnibus under the accession numbers GSE263063, GSE263065, GSE263066, GSE263068, GSE263070-GSE263075, GSE263077, and GSE263268. Data were initially processed using Affymetrix GeneChip Command Console software to generate globally normalized quantities for each gene transcript cluster. Additional processing to generate GFP⁺/GFP⁻ ratios from the microarray signal intensities was done in Microsoft Excel. These GFP⁺/GFP⁻ enrichment ratios help to normalize effect across the different abundances of OR mRNAs and across differences in constitutive activity of ORs. Data for each gene are reported as the relative response, the GFP⁺/GFP⁻ ratio of signal from odor-stimulated mice divided by the GFP⁺/GFP⁻ ratio of signal from vehicle-stimulated mice.

These experiments were done in a paired design, N = 4 samples, with each sample consisting of a set of 3 mice, and each odor-stimulated set of 3 mice was always paired with another set of 3 control mice simultaneously exposed to clean air. The stability of OR GFP⁺/GFP⁻ ratios makes it possible to screen for responsive receptors using a relatively small number of replications. A Bayesian hierarchical model was used to obtain normalized measures of odorant effect, accounting for 4 sources of variation: (i) basal expression of the receptor, (ii) odorant effect, (iii) nonspecific/batch effect (correlated changes in both odorant and vehicle control in a paired replicate), and (iv) random measurement error. For each odorant effect, the posterior mean divided by the posterior standard deviation provides a measure (Z-statistic) that is approximately normally distributed. A difference of 2 or greater in the mean difference was considered a responsive OR. The responses of the 14 TAARs expressed by OSNs, which are believed to respond only to odorants containing amine groups, were used as an empirical check of this criterion (Liberles 2015). Further analysis and visualization were done using R version 4.0.2.

2.3. In vitro functional assay

Hana3A cells were cultured in Minimum Essential Media supplemented with 10% Fetal Bovine Serum, with penicillin-streptomycin and amphotericin B at 37 °C, saturating humidity, and 5% CO₂. The Dual-Glo luciferase assay (Promega) was used to determine OR activation by monitoring the activity of firefly and Renilla luciferase in Hana3A cells, as previously described (Zhuang and Matsunami 2008). Briefly, firefly luciferase, driven by a cAMP response element promoter (CRE-Luc, Promega), was used to determine OR activation levels and the constitutively produced Renilla luciferase (SV40-RL) was used to normalize the luciferase activity in each well. For each well of a 96-well plate, 5 ng of SV40-RL, 10 ng of CRE-Luc, 5 ng of mouse RTP1S (Wu et al. 2012), 2.5 ng of M3 muscarinic receptor (Li and Matsunami 2011) and 5 ng of Rho-tagged OR plasmid DNA (or empty vector pCI) were transfected 18 to 24 h before odorant stimulations. The stimuli (odorants) were diluted in CD293 media supplemented with copper and glutamine (CD293 stimulation medium) to the desired final concentration and 25 µL of the stimulation solution was injected into each well and incubated at 37 °C, 5% CO₂ for 3 ½ h. Firefly and Renilla luciferase luminescence were then recorded following the manufacturer's protocol on a POLARstar OPTIMA plate reader (BMG Labtech). Data were analyzed using Microsoft Excel and GraphPad Prism. Normalized activity for each well was further calculated as (Luc-400)/(Rluc-400) where Luc, luminescence of firefly luciferase; Rluc, Renilla luminescence and 400 corresponds to the luminescence of an empty well, n = 4 for each receptor. For screening, odorant concentration was set at 100 µM and the luminescence induced by OR activation was normalized and compared to the empty vector. Activation was determined by a 2-way ANOVA and a Sidak's multiple comparisons test between the OR and the empty vector responses. Activation was next verified by a dose–response assay of 7 serial dilutions ranging from 0.316 to 316 µM and CD293 medium (0 µM) for baseline. OR and empty vector responses to odorants were analyzed by fitting a least squares function to the data using GraphPad Prism.

3. Results

To discriminate odorants, especially structurally similar odorants, receptor response patterns must differ. To quantify the degree of receptor response pattern diversity we measured the response patterns to moderate airborne concentrations of 12 aliphatic odorants (Table 1) in freely behaving mice. Structurally, these 12 odorants define a small region of the universe of odorants. We detected responses to at least one of these odorants from a total of 167 ORs and 1 TAAR (Supplementary Table S1). Even though these odorants are similar in structure, the majority of the receptors responded to only one of the tested odorants, with only 32 of the 168 receptors (19%) responding to multiple odorants. Hierarchical clustering of response strength of these 168 receptors reveals substantial diversity across both receptors and odorants (Fig. 1a).

Fig. 1.

Cluster analysis of OR responses to 12 aliphatic odorants. a) Heatmap of response strength of the receptors that exceed the response criterion to at least 1 of the 12 odorants. b) Heatmap of 167 randomly selected nonresponsive ORs. Scale: fold-difference from clean air controls.

The general principle that agonists for a receptor typically share structural similarity predicts that this degree of overlap in receptor response patterns, though small, is nonrandom. As predicted, Fig. 1a shows clusters of subsets of ketones, alcohols, and acids that are not present when the same analysis is done on 167 ORs randomly selected from nonresponsive ORs (Fig. 1b). The most strongly clustered pairs of odorants are hexanone with pentanone, octanol with heptanol, heptanoic acid with hexanoic acid, and octanoic acid with pentanol. The pairing of octanoic acid with pentanol disagrees with expectations about odorant similarity producing response pattern similarity and suggests that the Fig. 1 clusters require refinement. Since nearly every one of the 167 ORs responded to only 1 or a small number of the tested odorants, irrelevant noise from numerous instances of nonresponsive ORs influences the clustering in Fig. 1. To assess odorant response pattern relationships in the absence of these influences, we clustered odorants based on the number of ORs in the overlap between response patterns. This analysis was done in 2 ways: using just the ORs that reach the criterion for response (Fig. 2a) and constraining the analysis so that each response pattern contains the same number of ORs (Fig. 2b).

Fig. 2.

Odorant response pattern similarity. a) Mathews Correlation Coefficient analysis for fold-difference scores of the 167 responsive ORs. Scale: Correlation coefficient. b) Heatmap based on the number of shared ORs between the top 25 OR responses for each odorant. Scale: Number of shared ORs.

Once the effects of nonresponsive ORs are removed, octanoic acid and pentanol show no correlation. Instead, this analysis finds that the strongest correlations are hexanol with pentanol, and hexanone with pentanone (Fig. 2a). Hexanone and hexanoic acid display a robust correlation as well in Fig. 2a, but this is a noisy measurement because the response pattern to hexanoic acid is small. In the analysis that normalizes the number of responsive ORs across odorants, the correlation is much weaker, with just 2 shared responsive ORs, and hexanone no longer clusters with hexanoic acid (Fig. 2b). The most strongly overlapping response patterns are instead hexanone with pentanone, heptanone with hexanone, octanone with heptanone, and heptanone with pentanone (Fig. 2b and Supplementary Table S2). Overlap at more moderate levels also occurs between odorants sharing functional groups (hexanol with pentanol, octanol with pentanol, octanoic acid with heptanoic acid) and odorants that do not share the same functional group (octanone with heptanol, octanone with octanol).

Nearly all of most tightly clustered response patterns in all 3 analyses, 9 of 11 clusters, are comprised solely of odorants sharing the same functional group. In addition, these analyses (Figs. 1a, 2a and b) organize response patterns into broader clusters that also tend to emphasize the relative importance of the functional group over carbon chain length for the tested aliphatic odorants. The ketones tend to cluster together (Figs. 1a, 2a, and b), as do the alcohols (Figs. 1 and 2a) and acids (Fig. 2b). However, some mid-level clusters include odorants differing in functional group. As one would predict, the odorants in these clusters tend to have chain lengths differing only by 1 carbon, for example heptanol, octanol, and octanoic acid (Fig. 2a), as well as octanone, heptanol, and octanol (Fig. 2b).

Overall, the greatest response pattern similarities are among short chain (C5 to C7) odorants having the same functional group, such as pentanone with hexanone (Figs. 1 and 2a, and b) and pentanol with hexanol (Fig. 2a and b). Odorants with 8 carbons (C8), in contrast, show more similarity across functional groups. This includes moderate levels (3 shared responsive ORs) of direct overlap in response patterns—octanone with heptanol; octanone with octanol (Fig. 2b)—and clustering with odorants having different functional groups such as octanol with octanoic acid and heptanol (Fig. 2a and b). These observations suggest the hypothesis that functional group identity is relatively more important than carbon chain length in producing response pattern overlap, especially at short carbon chain lengths, but also begins to suggest that chain length increases in influence with length.

To further assess the importance of factors affecting OR response pattern similarity, especially functional group and carbon chain length, we did principal component analysis (PCA) (Fig. 3). As should be the case, the outliers in this analysis are the ORs with the strongest and broadest responses. In our experiments those are the ketones, so the result of this analysis emphasizes their differences, distributing an arc of ketone-responsive ORs across the PCA plot. The ORs having the strongest association with principal component 2 (PCA2) are the ORs most responsive to short chain ketones (Olfr195 and Olfr531). The OR with the strongest association with principal component 1, Olfr77, responds best to C8 chain length odorants, specifically octanone and octanol. ORs broadly responsive to long-chain ketones such as Olfr398, Olfr1126, Olfr1507, and Olfr796 form a cluster in the arc of ketone-responsive ORs near Olfr77. Another cluster in the ketone arc is the set of ORs responsive to medium chain length (C6 and C7) ketones, including Olfr191, Olfr193, Olfr1480, Olfr532, Olfr921, Olfr923, Olfr45, Olfr122, Olfr176, and Olfr907. The ORs most responsive to alcohols and carboxylic acids, such as Olfr899, Olfr877, Olfr95, Olfr444, Olfr1450, Olfr1208, and Olfr782 cluster separately from the ORs responsive to ketones. One parsimonious interpretation of this analysis is a shift in the relative importance of the 2 major structural features, carbon chain length and functional group, where ORs responsive to aliphatic ketones, alcohols, and acids have the most distinct response patterns at short carbon chains (C5 to C6) and more overlap at long carbon chains (C7 to C8).

Fig. 3.

Principle components analysis plot of OR responses to the 12 aliphatic odorants produces groups according to response sensitivity to carbon chain length and functional group. For clarity only some ORs are identified by their names.

In addition to informing us about the relatedness of OR response patterns to odorants, this data set also provides information about the relatedness of ORs responsive to individual monomolecular odorants. These new, more extensive data strengthen previous evidence that the ORs responsive to an odorant in vivo often are unrelated sequences (McClintock et al. 2014; de March et al. 2020; Hu et al. 2020). We observe that the ORs responsive to each of the 12 aliphatic odorants tend to be spread across the phylogenetic tree of OR sequences (Fig. 4a to l). Clusters of 2 to 5 closely related ORs occur but they represent minor fractions of the number of ORs in these response patterns. This property of OR sequence diversity is not restricted to ORs responsive to only 1 aliphatic odorant. ORs responsive to multiple aliphatic odorants are also spread across the breadth of OR sequences, including both Class I and Class II ORs (Fig. 4m).

Fig. 4.

a) to l) ORs responsive in vivo to aliphatic odorants are typically spread broadly across the diversity of OR sequences. m) ORs responsive to 2 (blue; darker) or 3 (gold; lighter) of the aliphatic odorants tested.

Our previous studies have shown that our in vivo assay accurately identifies ORs responsive to tested odorants. While some ORs do not function when expressed in cultured cells, which can sometimes be attributed to failure to express at the surface of cultured cells, those that do function nearly always provide in vitro confirmation of the in vivo responses (McClintock et al. 2014; de March et al. 2020; Hu et al. 2020). Continuing this practice, we performed in vitro heterologous expression and activation assays on 8 ORs. Four of them show in vitro responses. We confirm the specificity of Olfr95 for octanol (Fig. 5a to c). We confirm that Olfr398, one of the ORs that responded to multiple odorants in vivo, exhibits the strongest response to octanone and responds significantly to other long-chain ketones. Additionally, Olfr398 shows lower but notable responsiveness to octanol, heptanol, and hexanol, aligning with our in vivo findings (Fig. 5d to f). Olfr195 responds to hexanone and hexanol, matching the strongest part of the response profile we observe in vivo (Fig. 5j to l). We also identified a case where the in vivo and in vitro responses were less consistent. Olfr1494 responds to octanol in vitro, which represents only a minor component of its in vivo response profile, yet it does not respond to heptanone, the strongest agonist observed in vivo. (Figure 5g to i). In vivo OR activity arises from a complex cascade of events, including peri-receptor events that can modify odorants before their interaction with ORs at the surface of OSNs (Pelosi 2001; Nagashima and Touhara 2010; Li et al. 2014). In contrast, the in vitro assay enables the exploration of odorant/OR interactions in a controlled setting, free from the influence of accessory proteins or other peri-receptor events. Here, we observe that most of our in vivo findings are validated by the in vitro experiments, indicating that peri-receptor events did not significantly impact OR responses to the majority of odorant/OR pairs.

Fig. 5.

Comparison of in vivo responses (a, d, g, j) with in vitro responses from heterologous expression of individual ORs in cultured cells (n = 4) in screening at 100 µM (b, e, h, k), and dose–response (c, f, i, l). Statistical significance of OR response in comparison to empty vector pCI is assessed by multiple comparisons 2-way ANOVA. Raw and analyzed data are available in Supplementary Tables S3 and S4.

4. Discussion

Our results allow several major conclusions. First, OR response patterns show substantial differences even between the aliphatic odorants that are the sole occupants of a small region of odorant structural space. The magnitudes of these differences appear sufficient to allow substantial variation in response patterns while still maintaining perceptual constancy. Second, the aliphatic odorants' functional groups are more important than carbon chain length for determining response pattern similarity, especially at short carbon chain lengths (C5 to C7). However, the data also begin to suggest that as carbon chain length increases so does its impact on response pattern similarity. Third, the sets of ORs responsive to an odorant are typically spread broadly across the diversity of OR sequences, with only occasional small clusters of related ORs. This observation is informative about the relative importance of forces driving OR evolution. It indicates that expansion and diversification of the OR gene family is the major evolutionary response, presumably due to selection for an OR repertoire capable of producing combinatorial codes sufficient to represent odor environments that are locally and temporally variable subsets of a vast and structurally divergent odor universe. Fourth, OR response patterns to these aliphatic odorants may not be significantly influenced by peri-receptor events. The olfactory mucus contains various proteins, including odorant-binding proteins and enzymes (Pelosi 2001; Nagashima and Touhara 2010; Li et al. 2014), that could potentially modify odorant molecules before they interact with ORs at the surface of OSNs. However, the overall consistency between our in vivo and in vitro OR response patterns in both this project and our previous work (McClintock et al. 2014; de March et al. 2020; Hu et al. 2020) suggests that, for the odorant molecules and ORs studied, these peri-receptor events did not have a strong impact on OR activation patterns.

The OR response patterns we observed differ substantially even when the odorants are structurally similar. These response patterns may consist largely of the most strongly responsive ORs and thereby magnify these differences, but strongly responsive, highly sensitive ORs are likely to be the most important elements of combinatorial codes. Most of the 167 responsive ORs reached significance in response to 1 aliphatic odorant, with only 32 (19%) responding to more than 1 of the 12 aliphatic odorants at the concentrations tested. However, these ORs should not be viewed as selective for aliphatic odorants. For example, 26 ORs (16%) are among the 116 ORs responsive to 7 unrelated odorants or simple mixtures tested in our previous in vivo experiments, including 1-pentanethiol (13 shared ORs), indole (2 ORs), bourgeonal (2 ORs), whiskey lactone (2 ORs), undecanal (none), isoamyl acetate (4 ORs), and citrus accord (6 ORs) (de March et al. 2020; McClintock et al. 2020a; McClintock et al. 2020b). Two of the 26 ORs respond to 2 or more of these 7 odors. Some overlap with the aliphatic odorant response patterns occurs even for odorants that are very different structurally, such as indole and bourgeonal. In contrast, 23% of the ORs responsive to 1-pentanethiol, an aliphatic thiol, also respond to 1 or more aliphatic ketones, alcohols, and acids we tested. This is consistent with the prediction that OR response patterns of structurally similar odorants overlap better than do response patterns of less similar odorants, but as with comparisons between the aliphatic acids, alcohols, and ketones, overlap with the 1-pentanethiol response pattern also involves only a minor fraction of the responsive ORs. By producing response patterns with limited amounts of overlap between the most responsive ORs, even for structurally similar odorants, the OR repertoire would facilitate odor discrimination and recognition. Olfactory bulb recordings agree that response patterns differ even for structurally similar odorants (Uchida et al. 2000; Belluscio and Katz 2001) and behavioral studies confirm that aliphatic odorants, including those tested herein, are easily discriminated (Laska et al. 2008; Can Guven and Laska 2012; Yoder et al. 2014).

The degree of separation we observe at the level of OR response pattern should allow for a substantial amount of variation in the response to an odorant or odor without necessarily forcing loss of perceptual constancy. In fact, perceptual constancy in the face of some variation in the OR response pattern seems certain. For example, concentration-invariant perception of odors is the norm (Gross-Isseroff and Lancet 1988; Krone et al. 2001; Wachowiak and Cohen 2001; Laing et al. 2003; Uchida and Mainen 2007; Homma et al. 2009; Mainland et al. 2014; Wojcik and Sirotin 2014; Sirotin et al. 2015; Wilson et al. 2017) yet published evidence shows that OR response patterns change with concentration (Hu et al. 2020; McClintock et al. 2020b). Not only do less sensitive ORs join the response pattern as concentration increases, some highly sensitive receptors respond less—or even fail to respond—to higher concentrations. Similar effects have been observed in physiological recordings from olfactory bulb glomeruli (Friedrich and Korsching 1997; Rubin and Katz 1999; Meister and Bonhoeffer 2001; Wachowiak and Cohen 2001; Fried et al. 2002; Bozza et al. 2004; Hu et al. 2020; McClintock et al. 2020b). Without a sufficient degree of separation between OR response patterns to structurally similar odorants these variations would threaten odor discrimination and identification.

The substantial differences between the response patterns of these 12 aliphatic odorants prevent strong correlations between these response patterns, but the most similar response patterns, irrespective of the type of analysis, do tend to occur between odorants having the same functional group and differing in chain length by a single carbon. The identity of the functional group tends to be more strongly associated with similarity in OR response pattern than carbon chain length, but this seems most true for shorter carbon chain lengths. Our data suggest the hypothesis that carbon chain length importance relative to functional group grows as chain length increases from 5 carbons to 8 carbons. If true, this shift could have several causes. These include longer carbon chain lengths providing stronger binding to hydrophobic regions of OR binding pockets, or requiring a larger binding pocket involving more extensive differences in OR sequence, or differences in the ability of aliphatic odorants with differing carbon chain lengths to reach OR binding pockets, or combinations of these factors.

The degree of separation between OR response patterns we observed could be an unavoidable consequence of natural selection driving the evolutionary expansion of a set of receptors sufficient to detect the extreme breadth and structural diversity of odorants (Mayhew et al. 2022), or it may instead have selective value itself. The diversity of OR sequences responsive to an odorant argues that selection against multiple ORs responsive to an odorant or purifying selection for ORs highly sensitive to specific odorants is unimportant compared to selection for expansion and diversification of the OR family. As gene duplication over evolutionary time produces new ORs and mutation diversifies these sequences, overlap in odorant sensitivity should repeatedly arise randomly and then diverge, which our data indicate is the case. In particular, the identification of aliphatic-responsive ORs that also respond to unrelated odorants in vivo is consistent with an OR repertoire evolving to function in combinatorial coding. The need for distinct combinatorial codes, however, means that natural selection could also act upon the degree of overlap in agonist selectivity, limiting it to produce distinct patterns for very large numbers of stimuli while still evolving ORs with new odorant sensitivities. An additional factor that may contribute to the evolution of response pattern diversity is concentration-invariant perception. As mentioned above, the need to accommodate some degree of variation, especially concentration-dependent variation, in a combinatorial code may be strong enough to be acted upon by natural selection and help increase the degree of distinctiveness of OR response patterns.

The conclusions we draw suffer from some limitations inherent in the approaches and methods used. We measured all ORs and TAARs in vivo but may not have captured small responses, so the peaks in response patterns are emphasized in our data. We also cannot rule out the possibility that even though the mice experience 40 h of clean air before OR responses are measured the prior history of odor exposure, or odors produced by the mice themselves during experiments, may affect the detection of responses from some ORs due to odorant antagonism or OSN adaptation. These limitations mean the breadth of response patterns is under-represented in our data, but less so than techniques restricted to measuring minor fractions of the OSN response pattern. In addition, by capturing the largest responses, which presumably arise from the most sensitive ORs, we detect the elements of the response pattern thought to matter most for odor discrimination and odor recognition. Response pattern breadth and OR response strength in vivo depend upon concentration (McClintock et al. 2020b) so we could have observed broader response patterns and greater overlap using higher odorant concentrations, but high odorant concentrations raises other problems such as nonmonotonic dose–response relationships observed for some ORs (Hu et al. 2020; McClintock et al. 2020b). Better to test low-to-moderate concentrations and maximize detection of the more sensitive receptors thought to be the key elements for perception (Spors and Grinvald 2002; Wilson et al. 2017; Chong et al. 2020). That these most sensitive receptors respond first, a key element of primacy effect theories, cannot be determined by our in vivo method because it lacks sufficient temporal resolution. Its ability to measure all ORs and TAARs in freely behaving mice is its strength, one that is an important check on the conclusions drawn from the response patterns of subsets of OSNs or olfactory bulb glomeruli. Thus far the general conclusions from these disparate techniques agree. Lastly, the behavior of TAARs in our experiments empirically confirm that the rate of false positive responses is low, near 10%. TAARs are believed to respond only to odorants containing amine groups (Liberles 2015), which were not among the odorants we tested. Only 1 of the 14 TAARs expressed by mouse OSNs, Taar2, exceeded the response criterion and only in response to 1 odorant (pentanoic acid). That Taar2 responds to pentanoic acid remains possible but seems unlikely.

In summary, we confirm that structurally similar odorants evoke overlapping OR response patterns but find that overlap between the stronger elements in OR response patterns typically consist of a minor fraction of these ORs. This distinctiveness of response patterns almost certainly facilitates odor recognition and discrimination. We hypothesize that the substantial degrees of difference we observe between OR response patterns allow for variation in the response pattern to an odor without losing perceptual constancy. Our data also support the hypothesis that response pattern similarity across aliphatic odorants depends most on functional group at short carbon chain lengths but the 2 factors become more equal once chain length reaches 8 carbons. OR response patterns to the tested set of aliphatic odorants consist of ORs showing a spectrum of response specificity. This ranges from broadly responsive ORs, like Olfr398, to ORs responsive to only one of the tested odorants, like Olfr1494. However, only in the context of the set of aliphatic odorants tested would it be appropriate to designate these ORs as broadly and narrowly tuned ORs, respectively. Only a very limited number of odorants and concentrations have been tested across all ORs, making more general conclusions about the response breadth of individual ORs premature. Encouragingly, we continue to find correspondence between OR responses in vivo and in vitro, indicating that odorant metabolism and other peri-receptor events in the olfactory mucus are not responsible for most in vivo OR responses we detect.

Author contributions

CAdM designed experiments, performed experiments, did data analysis and visualization, and helped write the manuscript. PB designed experiments, did data analysis and visualization, and helped write the manuscript. WBT performed experiments. HM designed experiments and helped write the manuscript. TSM designed experiments, performed experiments, and helped write the manuscript.

Supplementary material

Supplementary material is available at Chemical Senses online.

Funding

This work was supported by National Institutes of Health (NIH) grants R01DC014468 (TSM), R01DC020353 (HM), and K99DC018333 (CAdM). HM acknowledges support from National Science Foundation grant 2014217.

Data availability

The data underlying this article are available in the article and in its online Supplementary material, or can be found in the Gene Expression Omnibus under the accession numbers GSE263063, GSE263065, GSE263066, GSE263068, GSE263070-GSE263075, GSE263077, and GSE263268.