A molecular and bioinformatics search for genes expressed differentially in neuronal subpopulations of the mouse MOE and VNO led to the identification of an uncharacterized H2B histone variant (symbol: Hist2h2be; named for its position (e) among H2B-encoding genes within mouse histone cluster 2 (Marzluff et al., 2002) and referred to herein as H2be). Microarray and in situ hybridization (ISH) analyses revealed high expression of H2be in the MOE and apical VNO neuroepithelium (Figure 1A,B), but no expression in the olfactory bulb (OB) or brain. The Genepaint and Genomics Institute of the Novartis Research Foundation (GNF) GeneAtlas (Su et al., 2004) databases, which contain transcriptional information for embryo sections (Figure 1C) and 61 mouse tissues (Figure 1D), respectively, confirmed the remarkable specificity of H2be expression in the MOE and VNO.

Figure 1

Download asset Open asset

The H2be mRNA possesses a long 3′-untranslated region (UTR) and a poly-A tail. This contrasts with typical histone transcripts that lack poly-A tails but contain short 3′-stem loop UTRs, which facilitate coordination of histone expression with the cell cycle. These features, together with the observed presence of its mRNA in post-mitotic neurons, suggest that H2be encodes a replication-independent replacement histone. H2BE displays only five amino-acid differences with the canonical mouse H2B protein (Figure 1E). Potential human (Collart et al., 1992), rat, and bovine orthologs to H2be exist, although their expression is uncharacterized.

To facilitate the characterization and unambiguous identification of H2BE from canonical H2B, we fused a FLAG tag to the N-terminus of H2BE, an approach used successfully for several other H2B proteins (Kao and Osley, 2003), and constructed a bacterial artificial chromosome (BAC) transgenic mouse line expressing the tagged protein (Figure 2A). Control experiments confirmed insertion of the transgene into the genome as a single copy and recapitulation of endogenous H2be expression (Figure 2B–E). We refer to the H2be:Flag-H2be mouse line as Flag-H2be.

Figure 2

Download asset Open asset

Due to the abundance of available molecular and genetic tools, we focused our study on the prospective role of H2be in the MOE. Initial expression analyses revealed that H2be mRNA and protein levels are not uniform among olfactory neurons. Rather, neurons with variable expression levels appear intermingled, and apically-located neurons generally display the highest levels of expression (Figures 1A and 2B). The heterogeneous expression of H2be among MOE neurons could potentially reflect differences in the identity of the co-expressed OR, in neuronal maturity, or in some unknown biological feature. To examine the relationship between H2be and OR expression, we used fluorescent ISH (FISH) to identify mature neurons expressing each of 42 specific OR genes. We then used quantitative fluorescence microscopy (Waters, 2009) (see ‘Materials and methods’) to assess the level of FLAG-H2BE within nuclei of the identified neurons relative to their surrounding neuronal field, imaged in tissue sections spanning the length of the MOE, and across multiple animals (n ≥ 2 per OR). Strikingly, each OR tested appears consistently and reproducibly associated with a stereotyped level of H2BE in mature neurons (Figure 2F,G). Indeed, for 73% of the tested ORs, neurons expressing the same OR display significantly less variation with each other in their H2BE level than is observed within the entire MOE neuronal population (p<0.05 after false-discovery rate [FDR] correction; 1-tailed F-test). These results indicate that H2BE expression in a given neuron is tightly correlated with the identity of the co-expressed OR, and not with a specific stage of neuronal development or maturity.

To further investigate the function of H2be, we generated an H2be-null line by targeted replacement of the H2be coding sequence with a membrane-localized mCherry reporter (Gap43-mCherry; Figure 3A). We refer to this line as H2be-KO. Analysis of GAP43-mCherry fluorescence in the MOE of H2be-KO mice revealed a variable intensity of the fluorescent reporter similar to that observed for the endogenous gene (Figure 3B). Analysis of GAP43-mCherry in the olfactory bulb revealed fluorescence limited to sensory axon termini in the glomerular layer. Fluorescence intensities appeared to vary widely among glomeruli, which are each innervated by neurons expressing a given OR, a result consistent with our observations of variable but stereotyped H2BE levels in neurons expressing specific ORs (Figure 3C).

Figure 3

Download asset Open asset

H2be-KO mice appear healthy and fertile. Training of H2be-KO mice and littermate controls to discriminate between odors for a water reward revealed that H2be-KO mice learn the task more slowly and perform less efficiently than the controls, suggesting a defect in olfactory function (Figure 3D). To investigate a potential defect in olfactory signal transduction, we performed electro-olfactograms on MOEs from H2be-KO and heterozygous littermates (Figure 3E) but found no obvious anomalies, indicating that the observed phenotype is not due to gross defects in odor-evoked signaling.

Because histones are central components of chromatin, we suspected that loss of H2be might affect olfactory gene expression. Microarray analysis of whole MOE tissue in 6-month old H2be-KO and WT mice revealed that approximately 6% of transcripts are differentially expressed (p<0.05; Supplementary file 1A; Figure 4—source data 1), a value that may underestimate the effects of H2be loss in the fraction of MOE cells that express H2be. Qualitatively, the effects of H2be loss appear to depend on the normal expression level of a given gene: highly transcribed genes tend to be up-regulated, while moderately transcribed genes appear up- or down-regulated in H2be-KO mice (not shown). Although the differentially-expressed genes do not pass the p<0.05 threshold after FDR adjustment for the 28,064 probe sets tested, likely due to their moderate fold changes, they do show statistically significant enrichment in several gene ontology categories (FDR-adjusted p<0.05; Figure 4A). The categories enriched most strongly among up-regulated genes include ‘olfactory detection’ (consisting of OR genes) and ‘RNA processing’, and among down-regulated genes include ‘developmental process’ and ‘positive regulation of transcription’.

Figure 4

Download asset Open asset

Figure 4—source data 1

Effects of H2be loss of function on gene expression in the main olfactory epithelium of 6-month old mice.

https://doi.org/10.7554/eLife.00070.007

Download elife-00070-fig4-data1-v1.xlsx

The highly significant enrichment of the ‘olfactory detection’ gene category (FDR-adjusted p=2 × 10⁻¹⁷) reveals that a large number of ORs may be up-regulated in the olfactory epithelia of H2be-KO mice, which we indeed confirmed by quantitative PCR on a subset of genes (Figure 4B). Interestingly, a comparison of gene expression at 6 months with an earlier age of 5 weeks revealed that OR expression differences between WT and H2be-KO mice increase dramatically with age (Figure 4C), and display a significant bias towards up-regulation in the KO (p<10⁻³⁰, 1-tailed paired t-test). To determine if defects in OR mRNA levels originate from differences in the frequency of OR expression in the MOE, in the cellular level of OR transcripts in individual neurons, or in both, we analyzed the expression of specific ORs by RNA FISH. For each of four ORs quantified, expression differences in H2be-KO and WT mice strongly correlate with differences in OR frequencies (Figure 4D,E) and not cellular levels (not shown). Remarkably, expression differences observed for various ORs in H2be-KO and WT mice were also found to correlate with the level of H2BE with which they are normally co-expressed: ORs with increased expression in H2be-KO mice are associated with high levels of H2BE, while ORs with decreased or unchanged expression appear associated with low H2BE levels (Figure 4F).

Changes in the representation of neurons expressing specific ORs upon loss of H2be led us to predict that ectopic over-expression of H2be might also affect OR frequencies. We generated an Omp:Flag-H2be transgenic mouse line (referred to as H2be-GF) in which Flag-H2be is expressed under the control of the promoter for olfactory marker protein (Omp; Figure 5A), a gene highly expressed in mature olfactory neurons (Danciger et al., 1989). An H2be-GF founder line was identified by quantitative PCR as expressing the transgene at a approximately 10-fold higher level than the endogenous H2be gene (not shown). This line displayed a high level of FLAG-H2BE in all mature olfactory neurons except for a subset in zone 2 (Figure 5B). We reasoned that this serendipitous mosaic expression provided an internal control for the effects of H2be over-expression. Comparison of overall gene expression in MOE tissue from H2be-GF and WT mice revealed widespread differences (Supplementary file 1B; Figure 5—source data 1). Strikingly, both up- and down-regulated genes are dominated by ORs, exhibiting FDR-adjusted p-values of 7 × 10⁻²⁹ and 3 × 10⁻¹⁸, respectively, for enrichment in the ‘olfactory detection’ category (Figure 5C). Using an FDR-adjusted p-value cutoff of 0.05, we found that approximately 7% of ORs display significant expression differences in H2be-GF mice (Figure 5D). As observed for H2be-KO mice, OR expression changes in H2be-GF mice are strongly correlated with H2BE expression (Figure 5E,F,H) and reflect changes in OR expression frequency, such that ORs that are consistently co-expressed with the transgene are reduced in frequency, while ORs that escape co-expression with the transgene within zone 2 are increased (Figure 5F–I).

Figure 5

Download asset Open asset

Figure 5—source data 1

Effects of the ectopic over-expression of H2be (expressed from an Omp-promoter-driven transgene and tagged with a FLAG epitope) on gene expression in the main olfactory epithelium of 5-week old mice.

https://doi.org/10.7554/eLife.00070.009

Download elife-00070-fig5-data1-v1.xlsx

Altered OR expression frequencies in H2be-KO and H2be-GF mice could result from changes in OR gene choice or in the life span of neurons expressing a given receptor. To investigate the temporal onset of H2BE expression relative to OR gene choice, we took advantage of the basal to apical gradient of neuronal development in the MOE. Analysis of the onset of H2BE relative to ORs normally associated with high levels of H2BE revealed that H2BE expression is initiated well after OR expression (Figure 6A). We next compared the onset of H2be expression to that of Neurod1, Gap43, and Omp, which are expressed prior to, concurrently with, and subsequent to OR gene choice, respectively (Cau et al., 2002; Iwema and Schwob, 2003; Kolterud et al., 2004). We found that the onset of H2be expression follows the window of Neurod1 expression, as observed at both the protein and RNA levels in MOE tissue from Flag-H2be and WT mice, respectively (Figure 6B,C). Moreover, H2BE expression begins after GAP43 (Figure 6D) and before OMP (Figure 6E), but overlaps extensively with both. Together, these results confirm that H2BE is present prior to full neuronal maturity, but subsequent to OR choice. Thus H2BE is unlikely to participate in the process of OR choice.

Figure 6

Download asset Open asset

Next, we further analyzed H2be-KO mice for evidence of defects in choice, or in stabilization of OR gene choice. We found no defects in OR co-expression frequencies (Figure 7A) or in the position or number of glomeruli (Figure 7B) in H2be-KO mice, which are phenotypes expected to be associated with the aberrant expression of multiple OR genes or the abnormal switching between distinct ORs (Shykind et al., 2004), suggesting that H2BE is not likely involved in the maintenance of OR gene choice. Finally, as described previously, gene expression analyses indicate that OR expression defects in H2be-KO mice are markedly more severe at 6 months of age than at 5 weeks (Figure 4C). Since disruption in OR gene choice would be expected to cause defects in OR expression that are independent of age, the delayed appearance of OR expression defects in H2be-KO mice is inconsistent with a role for H2BE in this process.

Figure 7

Download asset Open asset

We next investigated the alternative hypothesis that altered OR frequencies in H2be-KO and H2be-GF mice may result from changes in neuronal longevity. Analysis of active-CASP3-labeled mature olfactory neurons in H2be-KO and littermate controls revealed a 45% lower frequency of apoptosis in the mutant epithelium (Figure 7C). Moreover, analysis of the life span of 5-bromo-2′-deoxyuridine (BrdU)-labeled neurons revealed a significantly higher survival rate in H2be-KO mice compared to controls (Figure 7D). These results indicate that the loss of H2be increases the life span of mature neurons. Accordingly, similar analyses in H2be-GF mice revealed a twofold higher rate of apoptosis among mature neurons (Figure 7E) and a significantly reduced survival rate compared to Flag-H2be mice (Figure 7F). Remarkably, the initial BrdU labeling frequency of H2be-GF neurons is approximately 2.9-fold higher than that observed in control neurons, reflecting an elevated rate of neurogenesis, presumably a consequence of the increased rate of apoptosis (Figure 7G). To ensure that the observed increase in apoptosis in transgenic epithelia is not due to a non-specific toxic effect of the overexpressed histone, we expressed H2be or Flag-H2be at high levels in cultured fibroblast (NIH-3T3) or embryonic kidney (HEK-293T) cells. Transgene-expressing cells showed no increase in cell death compared to control cells overexpressing the canonical H2b or containing no transgene (not shown), suggesting that H2BE's action in affecting cellular longevity requires an olfactory-specific cellular context. Taken together, these data indicate that the expression of H2be in mature olfactory neurons directly affects neuronal longevity, such that neurons with low expression of H2BE tend to have longer life spans, while neurons with elevated H2BE tend to be relatively short-lived.

We next sought to identify the source of heterogeneity in H2BE level among neurons expressing different ORs in WT mice. To investigate a possible link with neuronal activity, we performed unilateral naris occlusion (UNO) by surgically blocking airflow into one nostril of Flag-H2be mice for 10 days. Upon analysis of H2BE in the MOE of these mice, we observed dramatically higher H2BE expression on the closed side of the epithelium relative to the open side. This increase can be observed at the cellular level in neurons expressing a specific OR (Figure 8A,B), and in the GAP43-mCherry fluorescence within the ipsilateral and contralateral olfactory bulbs of H2be-KO^(+/−) mice (Figure 8C). Together, these results indicate that olfactory deprivation causes an up-regulation of H2BE.

Figure 8 with 1 supplement see all

Download asset Open asset

To test whether neuronal stimulation results in reduced H2BE levels, we exposed mice to a mixture of odorants or, as a negative control, to mineral oil. The odorant mixture consisted of four ligands corresponding to known ORs: heptanal/octanal for Olfr2 (Bozza et al., 2002), lyral for Olfr16 (Touhara et al., 1999), and eugenol for Olfr73 (Oka et al., 2004) and Olfr958 (Oka et al., 2006). After 3 weeks of exposure to the odor mixture or mineral oil, the four stimulated neuronal subtypes showed significantly reduced levels of H2BE compared to the same subtypes in control mice (Figure 8D–F; Figure 8—figure supplement 1). In contrast, neurons expressing two randomly picked orphan ORs, Olfr167 and Olfr653, showed no statistically significant differences between odor-exposed and control mice. Thus, olfactory stimulation of specific neurons results in a relative reduction in their nuclear H2BE levels.

To further corroborate the activity-dependence of H2BE, we used GAP43-mCherry fluorescence as a reporter of H2be gene expression in H2be-KO^(+/−) mice and measured its intensity relative to that of tyrosine hydroxylase, a marker of neuronal activity in the glomerular layer of the olfactory bulb (Cho et al., 1996). These analyses revealed a significant negative correlation between the two markers, further supporting the view that H2be gene expression is inversely related to neuronal activity (Figure 8G,H).

To investigate the possibility that H2BE expression is regulated by cAMP or Ca²⁺, two important signaling components and modulators of gene expression in the MOE (Mori and Sakano, 2011), we looked for evidence of changes in H2BE levels in Adcy3-null and Cnga2-null mice, respectively. We found that olfactory neurons that contain undetectable levels of H2BE in WT mice frequently show extremely high levels in their Adcy3^(−/−) counterparts (Figure 9A). However, using GAP43-mCherry as a reporter of H2be expression in Cnga2^(+/−) mosaic females, we observed no elevation in GAP43-mCherry fluorescence in Cnga2⁻ glomeruli relative to Cnga2⁺ glomeruli (Figure 9B). These results suggest that the activity-dependent down-regulation of H2BE is cAMP- but not Ca²⁺-mediated. The precise mechanism of H2BE regulation by cAMP is unclear. One possibility is that H2be expression is suppressed by a factor such as ICER, a cAMP-dependent repressor form of the cyclic-AMP responsive element binding protein (CREB) family member CREM (Lyons and West, 2011). Interestingly, we identified two CRE half-sites within the H2be coding region, although their function, if any, is unknown.

Figure 9

Download asset Open asset

To investigate a potential role for H2be in mediating activity-dependent changes in gene expression, we performed UNO on H2be-KO and WT mice and compared gene expression differences for each genotype on the open and closed sides of the MOE after 3 weeks of nostril closure. Consistent with a recent study (Coppola and Waggener, 2011), we identified widespread differences in gene expression after activity deprivation in WT mice, with ‘olfactory detection’ (ORs) identified as the most highly enriched gene ontology category among both up- and down-regulated genes (Figure 10A; Figure 10—source data 1). Using an FDR-adjusted p-value cutoff of 0.05, approximately 11.5% of ORs display significant expression differences after olfactory deprivation in WT mice (Figure 10B). FISH analyses revealed that these differences reflect altered OR expression frequencies within the MOE (Figure 10C,D).

Figure 10

Download asset Open asset

Figure 10—source data 1

Effects of H2be loss of function on gene expression changes in the main olfactory epithelium (MOE) as a result of activity deprivation through unilateral naris occlusion (UNO) in 5-week old mice.

https://doi.org/10.7554/eLife.00070.016

Download elife-00070-fig10-data1-v1.xlsx

We then investigated a possible relationship between H2BE levels and changes in OR expression after UNO in WT mice. Remarkably, this analysis revealed that ORs that are down-regulated following UNO are normally co-expressed with high levels of H2BE, while up-regulated ORs are normally co-expressed with low H2BE levels (Figure 10E,F). Thus, if H2BE level is taken as a measure of neuronal inactivity, these data suggest that olfactory deprivation causes normally inactive neurons to decrease in relative abundance, while leading to a relative increase in neurons that are highly active.

Comparison of MOE expression data obtained for WT and H2be-KO mice revealed that for a subset of UNO-altered genes (Figure 11A; Figure 10—source data 1), including a large fraction of ORs (Figure 11B,C), UNO-mediated changes are significantly attenuated in the MOE of the KO compared to the WT. These results suggest that H2BE participates in activity-dependent modulation of olfactory gene expression, though it is clearly not the only mediator of the observed changes.

Figure 11

Download asset Open asset

Our findings support a model in which the level of olfactory activity of a neuron determines its level of H2BE, which in turn affects its life span (Figure 12). According to this model, inactive neurons express high levels of H2BE, which leads to reduced longevity, while active neurons maintain low levels of H2BE and are relatively long-lived. This model predicts that OR expression frequencies should change with age and experience and lead to the gradual enrichment of active olfactory neurons at the expense of inactive neurons in the MOE. Consistent with this prediction and previous reports of age-dependent changes in OR expression (Lee et al., 2009; Rimbault et al., 2009; Rodriguez-Gil et al., 2010), we found that 65% of all ORs are significantly altered in their expression from 5 weeks to 6 months of age in WT mice (FDR-adjusted p<0.05; not shown). Future experiments will determine the physiological significance of the ORs displaying age-dependent altered expression.

Figure 12

Download asset Open asset

To further investigate H2BE's function at the molecular level, we first sought confirmation that transgenic FLAG-H2BE protein, a tool critical for these studies, is normally integrated into chromatin. Fractionation of unfixed MOE cell nuclei from Flag-H2be and H2be-GF transgenic mice showed that FLAG-H2BE is present at barely detectable levels in soluble nucleoplasm and is instead almost entirely chromatin-bound (Figure 13A). Moreover, mass spectrometric analysis of proteins associated with FLAG-H2BE mononucleosomes from Flag-H2be MOE tissue identified numerous chromatin-associated proteins, including several canonical and variant histones (Figure 13B). Finally, chromatin immunoprecipitation (ChIP) using anti-FLAG antibodies yielded large quantities of DNA from Flag-H2be MOE tissue (Figure 13C). Together, these results provide strong evidence that FLAG-H2BE is readily integrated into the chromatin of olfactory neurons and support the use of Flag-H2be and H2be-GF mice for investigating the molecular function of H2BE.

Figure 13

Download asset Open asset

Using FLAG-H2BE as a molecular surrogate for H2BE, we examined the histone variant's localization on genomic DNA. High-resolution confocal imaging of neuronal nuclei in the MOE of Flag-H2be mice revealed that H2BE is not confined to nuclear puncta, but rather is widely distributed throughout the nucleus (Figure 13D). Analysis of H2BE localization by genome-wide chromatin immunoprecipitation (ChIP) in MOE neurons confirmed its wide distribution and revealed a slight enrichment near gene promoters, especially in genes highly expressed in mature olfactory neurons (Figure 13E). Interestingly, these analyses and subsequent qPCR experiments revealed that the protein-coding regions of histone genes are particularly enriched for H2BE, while those of OR and vomeronasal receptor (VR) genes are relatively devoid of the variant (Figure 13F–H). Because the ChIP input was obtained from a heterogeneous mixture of MOE nuclei and approximately 99.9% of OR alleles and all VR alleles are silent in a given olfactory neuron, the latter results may reflect the lower accessibility of silent OR and VR loci for replacement of canonical H2B by H2BE, perhaps due to the highly compacted nature of these loci (Magklara et al., 2011). Together, our results support a model in which H2BE replacement of canonical H2B within olfactory chromatin is widespread, but most extensive within transcriptionally active loci.

Four of the five amino acid variant positions within H2BE are located near post-translational modification (PTM) sites of canonical H2B (Figure 1E), raising the question of whether PTMs may differ between the two proteins. We analyzed three relatively well-characterized H2B PTMs: mono-methylation and acetylation of lysine 5 (Lys5-Me and Lys5-Ac), which lies close to the H2BE variant residue P3L, and ubiquitination of lysine 120 (Lys120-Ub), near the variant residue S124A. Both H2B-Lys5 PTMs have been shown to be positively correlated with transcriptional activity, with Lys5-Me and Lys5-Ac enriched within transcribed and promoter regions, respectively (Barski et al., 2007; Wang et al., 2008). H2B-Lys120 ubiquitination, which has also been linked with transcription, appears to play a complex regulatory role (Chandrasekharan et al., 2010).

Analysis of H2BE and H2B-Lys5-Me levels in Flag-H2be MOEs revealed a striking inverse correlation between the two staining patterns, such that neurons expressing a high level of H2BE display conspicuously low H2B-Lys5-Me immunoreactivity, and vice versa (Figure 14A,B). Using an ELISA assay, we ruled out the possibility that the polyclonal antibody against H2B-Lys5-Me is incompatible with the H2BE sequence (Figure 14C). To determine if there is a causal link between the seemingly mutually exclusive expression of H2BE and H2B-Lys5-Me, we examined the prevalence of H2B-Lys5-Me in H2be-KO and H2be-GF mice. Remarkably, loss of H2be leads to widespread H2B-Lys5-Me immunoreactivity in H2be-KO MOEs (Figure 14D), while over-expression of H2BE further reduces H2B-Lys5-Me staining in H2BE-positive nuclei (Figure 14E). Western analysis of MOE lysates from Flag-H2be mice provided a biochemical confirmation that the Lys5-Me modification is absent from the tagged H2BE (Figure 14F). Together, our results suggest that, unlike canonical H2B, H2BE does not undergo detectable mono-methylation at Lys5 and that, consequently, replacement of H2B by H2BE causes a direct reduction of the Lys5-Me modification in H2BE expressing cells. Interestingly, we observe an increase in the number of H2BE expressing neurons, and in H2BE level in individual cells in the MOEs of older animals (Figure 14G–I). This is supported by western analysis, which revealed that the overall level of H2BE in the MOE doubles from 7 weeks to 45 weeks (approximately 10 months) of age (Figure 14J). In addition, we observe that the mutual exclusion between H2BE and H2B-Lys5-Me dramatically increases with age (Figure 14G–I), indicating the gradual replacement of canonical H2B by H2BE, which proceeds to near-completion in a fraction of neurons by 36 weeks (approximately 8.5 months) of age. Further, it suggests that the accumulation of H2BE through replacement of canonical H2B proceeds at a faster pace than it's loss through neuronal turnover.

Figure 14

Download asset Open asset

Further analyses revealed that the presence of H2BE causes a similar reduction in the level of acetylation at Lys5, indicating that the variant also receives less of this modification compared to canonical H2B (not shown). In contrast, H2BE appears to undergo higher levels of ubiquitination at Lys120, though analyses of the modification in H2be-KO and H2be-GF mice indicate that the association is correlative but not causative (not shown).

The Flag-H2be transgenic mouse line (Figure 2A), which expresses FLAG-H2BE under control of the H2be promoter, was generated based on a described protocol (Yang et al., 1997). Briefly, a FLAG-encoding DNA sequence was inserted through homologous-recombination immediately upstream of the H2be CDS within BAC RP23-16G3, which contains a 200-kb region of mouse genomic sequence surrounding the H2be gene. The modified BAC was amplified in E. coli, confirmed by sequencing, and injected (Harvard Genome Modification Facility) into fertilized mouse zygotes. Transgenic founders were crossed to C57Bl/6 mice to establish the Flag-H2be line. Heterozygous Flag-H2be mice contain a single genomic copy of the transgene that is expressed in a pattern and at a level indistinguishable from that of the endogenous gene (see Figure 2B–E).

The H2be-KO mouse line (Figure 3A), in which the endogenous H2be CDS is replaced with a sequence encoding GAP43-mCherry (an N-terminal fusion of the first 20 amino acids of GAP43 to mCherry), was generated through homologous recombination of the endogenous H2be locus in mouse embryonic stem cells (ESCs) using standard methods. Following selection, ESCs were screened for the desired recombination events, confirmed by sequencing, and injected (Harvard Genome Modification Facility) into mouse blastocysts. Founders were crossed to C57Bl/6 mice to establish the H2be-KO line, in which Gap43-mCherry is expressed in a pattern indistinguishable from that of H2be (see Figure 3B).

The H2be-GF transgenic mouse line (Figure 5A), which expresses Flag-H2be under control of the olfactory marker protein (Omp) promoter, was generated by complete replacement of the Omp CDS in plasmid pJOMP (Danciger et al., 1989) with a sequence encoding FLAG-H2BE, followed by pronuclear injection (Harvard Genome Modification Facility) of the linearized construct into fertilized mouse zygotes. Transgenic founders were crossed to C57Bl/6 mice to establish the H2be-GF line. Heterozygous mice contain approximately 12 genomic copies of the transgene, which are expressed in mature olfactory neurons throughout the MOE with the exception of a band of neurons near zone 2 (see Figure 5B), a pattern that was reproducibly observed in all individuals examined (n = 8).

The P2-IRES-Tau-LacZ mouse line (Mombaerts et al., 1996), which was used to identify possible axon guidance defects in H2be-KO mice (Figure 7B), and the Cnga2-null/Tau-LacZ (Zhao and Reed, 2001) and Adcy3-null (Wong et al., 2000) mouse lines, which were used to identify second messengers affecting H2be expression (Figure 9) were described previously.

• Active-CASP3 (rabbit polyclonal; Promega, Madison, WI, USA; G7481)

• β-GAL (rabbit polyclonal; MP Biomedicals, Solon, OH, USA; 55976)

• BrdU (mouse monoclonal Bu20a; Dako North America, Carpinteria, CA, USA; M0744)

• FLAG (mouse monoclonal M2; Sigma-Aldrich, St. Louis, MO, USA; F1804)

• FLAG (rabbit polyclonal; Sigma-Aldrich; F7425)

• GAP43 (rabbit polyclonal; Novus Biologicals, Littleton, CO, USA; NB300-143A4)

• H2B (rabbit polyclonal; EMD Millipore, Billerica, MA, USA; 07-371)

• H2B-Lys5-Me (rabbit polyclonal; Abcam, Cambridge, MA, USA; ab12929)

• H2B-Lys5-Ac (rabbit monoclonal; Abcam; ab40886)

• H2B-Lys120-Ub (mouse monoclonal NRO3; Medimabs, Montréal, Québec, Canada; MM-0029)

• Histone H3 (rabbit polyclonal; Abcam; ab1791)

• NEUROD1 (goat polyclonal; Santa Cruz Biotechnology, Santa Cruz, CA, USA; sc1084)

• OMP (goat polyclonal; Wako Chemicals USA, Richmond, VA, USA; 344-10001)

• Tyrosine hydroxylase (rabbit polyclonal; Millipore; AB152)

All histological OR gene expression analyses were performed using fluorescent in situ hybridization (ISH). With the exception of H2be mRNA analyses (Figure 1A,B), which were performed using chromogenic ISH, all other histological analyses were performed using immunofluorescence (IF) or immunohistochemistry (IHC; for Active-CASP3, Figure 7C,E). Unless noted, all images are of coronal tissue sections.

ISH target sequences were amplified by PCR and inserted into the pCRII-TOPO vector (Life Technologies, Grand Island, NY, USA). OR antisense probes were designed to span 500–1000 base pairs, to target CDS or UTR gene regions, and to have <70% identity to any other sequence in the mouse genome. Probes were generated from 1 μg of linearized plasmid template using T7 or Sp6 polymerases (Promega) and digoxigenin or fluorescein RNA labeling mixes (Roche Applied Science, Indianapolis, IN, USA), treated with DNaseI (Promega) and ethanol precipitation, and dissolved in a 30-μL volume of water.

Whole tissues were carefully dissected from surrounding bones, frozen immediately in OCT compound (Sakura Finetek USA, Torrance, CA, USA) on dry ice, and stored at −80°C. Tissue blocks were cut into 12-μm thick cryo-sections, placed onto slides, and stored at −80°C. Chromogenic ISH experiments were performed essentially as described (Schaeren-Wiemers and Gerfin-Moser, 1993).

Fluorescent ISH experiments were performed using a modified version of the chromogenic ISH method. Briefly, slide-mounted sections were warmed (37°C, 10 min), equilibrated in phosphate-buffered saline (PBS; pH 7.2; 5 min, room temperature [RT]), fixed in paraformaldehyde (PFA; 4% in PBS; 10 min, RT), washed in PBS (3 min, RT), permeabilized with Triton-X-100 (0.5% in PBS; 10 min, RT) followed by sodium dodecyl sulfate (1% in PBS; 5 min, RT), washed in PBS (3 × 3 min, RT), incubated in acetylation solution (triethanolamine [0.1 M; pH 7.5], acetic anhydride [0.25%]; 10 min, RT), washed in PBS (3 × 3 min, RT), incubated in hybridization solution (formamide [50%], SSC [2×], Denhardts [5×], yeast tRNA [250 μg/mL], herring sperm DNA [200 μg/mL], EDTA [1 mM], sodium phosphate [0.05 M; pH 7]; 30 min, RT), hybridized with a digoxigenin-labeled antisense RNA probe (1:1000 in hybridization solution; 16 hr, 42°C), washed with SSC (2×; 5 min, 42°C), washed with SSC (0.2×; 3 × 30 min, 42°C), incubated in H₂O₂ (3% in TN [Tris–HCl (0.1 M; pH 7.5), 0.15 M NaCl]; 30 min, RT), washed in TNT (Tween-20 [0.05%] in TN; 3 × 3 min, RT), incubated in TNB (Blocking Reagent [Perkin Elmer, Waltham, MA, USA; 0.05% in TN]; 30 min, RT), incubated with anti-digoxigenin-POD antibody (Roche; 1:1000 in TNB; 12 hr, 4°C), and washed in TNT (3 × 20 min, RT). Fluorescent signals were generated using the Tyramide Signal Amplification (TSA) Plus Fluorescein Kit (Perkin Elmer) according to the manufacturer's instructions. Slides were mounted using Vectashield (Vector Laboratories, Burlingame, CA, USA) containing DAPI (5 μg/mL).

Two-color ISH was performed as described for one-color ISH, with the following modifications: Tissue sections were simultaneously hybridized with both digoxigenin- and fluorescein- or dinitrophenyl-labeled antisense RNA probes (1:1000 each in hybridization solution). Following incubation in TNB (30 min, RT), sections were incubated with anti-fluorescein-POD antibody (Roche; 1:1000 in TNB; 12 hr at 4°C) or anti-dinitrophenyl-HRP antibody (Perkin Elmer; 1:350 in TNB; 3 hr at 25°C) and washed in TNT (3 × 20 min, RT). Fluorescent signals corresponding to the fluorescein- or dinitrophenyl-labeled probes were generated using the TSA Plus Fluorescein Kit, after which sections were washed in TNT (2 × 3 min, RT), incubated in H₂O₂ (3% in TN; 1 hr, RT), washed in TNT (3 × 3 min, RT), incubated with anti-digoxigenin-POD antibody (1:1000 in TNB; 12 hr, 4°C), and washed in TNT (3 × 20 min, RT). Fluorescent signals corresponding to the digoxigenin-labeled probe were generated using the TSA Plus Cyanine5 Kit (Perkin Elmer) according to the manufacturer's instructions. Slides were mounted using Vectashield containing DAPI (5 μg/mL).

Combined ISH and IF experiments were performed as described for one-color ISH, with the following modifications: Acetylation, which dramatically reduces detection of the FLAG epitope, was omitted. Following incubation in TNB (30 min, RT), sections were incubated with a mixture of anti-digoxigenin-POD and mouse anti-FLAG antibodies (each 1:1000 in TNB; 12 hr, 4°C) and washed in TNT (3 × 20 min; RT). Fluorescent signals corresponding to the digoxigenin-labeled RNA probe were generated using the TSA Plus Fluorescein Kit, after which sections were washed in TNT (2 × 3 min, RT), incubated with anti-mouse-Alexa647 antibody (Invitrogen; 1:1000 in TNB; 12 hr, 4°C), and washed in TNT (3 × 20 min, RT). Slides were mounted using Vectashield containing DAPI (5 μg/mL).

Animals were anesthetized with ketamine and perfused transcardially on ice with ice-cold PBS (25 mL) followed by ice cold PFA (4% in PBS; 25 mL). Whole tissues were carefully dissected from surrounding bones and immersed in ice-cold PFA (4% in PBS; 1 hr [OB] or overnight [MOE]). MOE tissue was decalcified in EDTA (250 mM in PBS, pH 8.5; 2 days, 4°C), and all tissues were cryoprotected in sucrose (10, 20, and 30% in PBS; 2 hr, 2 hr, and overnight, respectively). Tissues were frozen in OCT on dry ice and stored at −80°C. Tissue blocks were cut into 12-μm thick cryo-sections, placed onto slides, and stored at −80°C.

IF experiments were performed as follows: briefly, slide-mounted sections were warmed (37°C, 10 min), equilibrated in PBS (5 min, RT), fixed in PFA (4% in PBS; 10 min, RT), washed in PBS (3 min, RT), permeabilized with Triton X-100 (0.5% in PBS; 10 min, RT) followed by SDS (1% in PBS; 5 min, RT; omitted if preservation of intrinsic mCherry was necessary), washed in TNT (3 × 5 min, RT), blocked in fetal bovine serum (FBS; 10% in TN; 30 min, RT), incubated with primary antibodies (typically diluted 1:500–1:1000 in 10% FBS; 12 hr, 4°C), washed in TNT (3 × 5 min, RT), incubated with secondary antibodies (typically, Alexa488-labeled [or Alexa488- and Alexa647-labeled, for two-color IF]; Invitrogen; 1:1000 in 10% FBS; 12 hr, 4°C), and washed in TNT (3 × 15 min, RT). Slides were mounted using Vectashield containing DAPI (5 μg/mL).

Mice were perfused and the MOE tissue processed as described for IF with the following modifications: After fixation in PFA and PBS washes, slide-mounted sections were permeabilized with Triton X-100 (0.5% in PBS; 30 min, RT), washed with PBS (3 × 3 min, RT), incubated in H₂O₂ (3% in TN buffer; 30 min, RT), washed with TNT (3 × 3 min, RT), blocked in TNB (30 min, RT), incubated with a mixture of anti-active-CASP3 and anti-OMP antibodies (each 1:300 in TNB; 12 hr, 4°C), washed with TNT (3 × 3 min, RT), and incubated with anti-rabbit-HRP (Jackson; 1:500 in TNB) for 12 hr, 4°C. Fluorescent signals corresponding to active-CASP3 were generated using the TSA Plus Fluorescein Kit, after which sections were washed in TNT (3 × 5 min, RT), incubated with anti-goat-Cy5 (Jackson ImmunoResearch Laboratories, West Grove, PA, USA; 1:500 in TNB; 12 hr, 4°C), and washed in TNT (3 × 15 min, RT). Slides were mounted using Vectashield containing DAPI (5 μg/mL).

Mice were injected intraperitoneally with BrdU (3 × 50 mg/kg in PBS; injections spaced 30 min apart) and sacrificed at the indicated timepoints (Figure 7D,F). The T = 0 timepoint, defined as 15 days post-injection, was chosen to avoid analysis of immature neurons, a large fraction of which are known to die prior to maturity (Kondo et al., 2010).

Mice were perfused and the MOE tissue processed as described for IF with the following modifications: After permeabilization with Triton X-100 and SDS, slide-mounted sections were washed with PBS (3 × 3 min, RT) and water (3 min at RT), incubated in HCl (2 N; 1 hr, 37°C), and washed with TNT (3 × 3 min, RT). Sections were blocked and further processed as described. Primary antibodies were used at concentrations of 1:50 (BrdU) and 1:300 (OMP).

Images were obtained using LSM710 and AxioImager Z2 (Carl Zeiss, Oberkochen, Germany) microscopes. Confocal images (1–5-μm thick optical sections) were used to quantify fluorescence signals, with care taken to ensure that exposures not exceed the instrument's dynamic range. Intensities were quantified using Zen software (Zeiss). For quantification of nuclear fluorescence in the MOE, circular regions encompassing individual nuclei were defined by DAPI fluorescence. Within each quantified image, a region surrounding the neuronal population (excluding immature and sustentacular cells) was defined to allow normalization to average neuronal nuclear fluorescence. For quantification of fluorescence in the OB, circular regions encompassing individual glomeruli were defined based on surrounding periglomerular cells, which were identified by morphology and DAPI fluorescence. p-Values corresponding to relative H2BE expression variances associated with specific ORs were calculated using a one-tailed F-test with FDR correction for multiple comparisons (Benjamini and Hochberg, 1995).

Fluorescent olfactory neuron counts corresponding to MOE tissue from an individual mouse were determined from a series of 10–12 stained coronal sections located approximately 400 μm apart and spanning the anterior–posterior length of the organ. Fluorescent cell counting was performed using Velocity software (Perkin Elmer) or, when necessary due to difficulties in resolving individual olfactory neurons, manually. Epithelial volumes were calculated from areas determined using Velocity software, based on OMP and DAPI signals.

Coding sequences for H2BE, FLAG-H2BE, and consensus H2B were inserted into the pLNCX2 vector (Clontech Laboratories, Mountain View, CA, USA). Retroviruses carrying the resulting clones were generated and used according to the Retroviral Gene Transfer and Expression User Manual (Clontech). NIH-3T3 and HEK-293 cells were transduced by retroviral infection and cell lines stably expressing high levels of each transgene were selected using Geneticin (500 µg/mL; Invitrogen). Expression of transgenes in selected lines was verified by quantitative RT-PCR. Transgenic and non-transgenic cell lines were analyzed for cell viability using a Vi-Cell XR Cell Viability Analyzer (Beckman Coulter, Brea, CA, USA).

For experiments leading to the initial identification of H2be, RNA was obtained by laser-capture micro-dissection (LCM) of apical and basal neurons in the VNO. Briefly, whole VNOs from 8-week old CD1 male mice were carefully dissected from surrounding bones and immediately frozen in OCT. Tissue blocks were cut into 12-μm thick cryo-sections, placed alternately onto Superfrost and Superfrost plus slides (VWR, Radnor, PA, USA). Sections on Superfrost plus slides were stained by ISH for the Gnai2 and Gnao genes, which mark the apical and basal zones, respectively, and used as guides for LCM. Sections on Superfrost slides were stained with toluidine blue and used for LCM of approximately 100 apical and 100 basal neurons per section using a PixCell II LCM system (Arcturus, now Life Technologies, Grand Island, NY, USA). Samples were immediately frozen and pooled into groups of 10 (approximately 1000 cells per group), from which the RNA was extracted using the Arcturus Picopure Kit (Life Technologies, Grand Island, NY, USA). LCM RNA samples were amplified in parallel with whole VNO RNA samples using the Arcturus RiboAmp OA RNA Amplification Kit (Applied Biosystems) and analyzed using Mouse Genome 430 2.0 Arrays (Affymetrix, Santa Clara, CA, USA) according to the manufacturer's instructions.

For all other data reported in this study, RNA was prepared from whole MOE tissue or, in the case of UNO expression analyses, from MOE halves that had been carefully removed from the medial bone. RNA was isolated using Trizol Reagent (Invitrogen) and purified using an RNeasy Miniprep Kit (Qiagen, Valencia, CA, USA). Experiments were performed using 3–6 biological replicates per condition or genotype and MOE tissue from 2–4 individual mice per replicate. Samples were processed and applied to Affymetrix Mouse Gene 1.0 ST Arrays according to the manufacturer's instructions.

Probe cell intensity files (CEL) were analyzed for potential outliers using the Bioconductor software package arrayQualityMetrics (Kauffmann et al., 2009). CEL files were processed with the Affymetrix Expression Console software to generate probe level summarization files (CHP) using the iterative Probe Logarithmic Intensity Error Estimation (IterPLIER) and sketch-quantile normalization algorithms. Statistical analyses of differential expression between groups were carried out using the Bioconductor limma package (Smyth, 2004) implemented through the Bioconductor affylmGUI software (Wettenhall et al., 2006) to generate unadjusted p-values and false discovery rate (FDR) corrections for multiple comparisons. For analyses of gene expression defects in H2be-KO MOE tissue, FDR corrections were made based on all represented genes with log₂ expression levels above five. For analyses of OR expression defects in H2be-GF mice and expression differences following UNO, FDR corrections were made based on all represented OR genes with log₂ expression levels above seven or five, respectively.

Gene ontology analyses were performed using the GOrilla software (http://cbl-gorilla.cs.technion.ac.il/) (Eden et al., 2007; Eden et al., 2009). All analyses were carried out using the ‘single ranked list of genes’ mode, with the exception of the analysis summarized in Figure 11A, which was performed using the ‘two lists of genes’ mode. Reported gene ontology summaries represent a subset of biological process terms selected based on redundancy using the REViGO software (http://revigo.irb.hr/; Supek et al., 2011). All reported enrichment p-values are FDR-adjusted using the Benjamini–Hochberg method (Benjamini and Hochberg, 1995).

cDNA samples for qPCR analysis were prepared using the QuantiTect Reverse Transcription Kit (Qiagen) starting from whole MOE RNA prepared using Trizol Reagent and purified using an RNeasy Miniprep Kit. Single-plex experiments (Figure 2D) were performed using the QuantiTect SYBR Green PCR Kit (Qiagen) with an MJ Opticon 2 instrument (Bio-Rad Laboratories, Hercules, CA, USA). Multiplex experiments (Figure 4B) were performed using the QuantiTect Multiplex PCR Kit (Qiagen) with an ABI 7900HT instrument (Applied Biosystems). Primer pairs (Integrated DNA Technologies, Coralville, IA, USA) and fluorophore/quencher-labeled probes (Eurofins MWG Operon, Huntsville, AL, USA; Integrated DNA Technologies) were designed using the Primer-BLAST tool (NCBI). Primer efficiencies were assessed using standard curves and pairs exhibiting efficiencies of >99% were used for analysis.

14-day old mice were administered Buprenorphine (0.05 mg/kg), anesthetized using isoflurane (confirmed through a tail pinch), and subjected immediately to electrocautery for approximately 5 s on the right nostril under a dissecting microscope (with care taken to avoid contact of the electrocautery unit with any non-superficial tissues). Mice were administered additional doses of Buprenorphine 12 and 24 hr after the procedure and examined on a daily basis to ensure complete blockage of the right nostril through scar formation (typically approximately 3–5 days after the procedure) and normal mouse development and activity.

Odors were presented as mixtures in mineral oil (octanal [0.31%], heptanal [0.068%], eugenol [2.65%]) or propylene glycol (lyral [10%]) continuously for 21 days at concentrations designed to achieve a vapor pressure of approximately 1 Pa per odorant. Odorants were presented in 100-µL volumes of each solution applied to a cotton pad and inserted into a metal mesh tea ball, which was suspended in the middle of the mouse cage by a metal chain. Odorants were exchanged every 24 hr.

Preparation of soluble nuclear proteins, chromatin, and mononucleosomes from MOE tissue and immunoprecipitation of FLAG-H2BE-containing mononucleosomes were performed as described (Okada and Fukagawa, 2006; Okada et al., 2006), with modifications. Briefly, MOE tissue from four 8-week old Flag-H2be and four 8-week old WT mice were dissected and immediately minced and mechanically homogenized. Nuclei were filtered through a cell strainer, pelleted, washed, and disrupted by sonication. Chromatin was separated from soluble nucleoplasm by centrifugation, washed, and digested with micrococcal nuclease to mononucleosomes. Mononucleosomes were solubilized in 350 mM KCl buffer and affinity-purified using Anti-FLAG M2 affinity gel (Sigma-Aldrich). Proteins were eluted with FLAG peptide (Sigma-Aldrich), TCA-precipitated, run on a 15% SDS-PAGE gel, and separated into the following molecular weight fractions: >100, 35–100, 18–35, and <18 kDa. Gel fractions were submitted for MS/MS analysis and protein identification within the Harvard University Microchemistry and Proteomics Analysis Facility.

Chromatin immunoprecipitation (ChIP) and ligation-mediated PCR amplification of the purified DNA were performed essentially as described (Lee et al., 2006). Input chromatin was prepared from whole MOE tissue dissected from 5-week old male Flag-H2be mice using mouse anti-FLAG and rabbit anti-histone H3 antibodies. Amplified DNA samples from the FLAG and H3 ChIPs were separately fragmented and analyzed in duplicate (2 mice per replicate) using GeneChip Mouse Promoter 1.0R Arrays (Affymetrix), which tile 10 kb of DNA surrounding the transcript start site of approximately 25,500 mouse genes. CEL files were processed using Tiling Analysis Software (Affymetrix) to generate BAR files.

ChIP signals were analyzed at each position in the genome as a ratio of FLAG to H3 and assigned to a specific gene promoter region using Galaxy (http://main.g2.bx.psu.edu/). Transcripts were grouped based on their expression level in OMP⁺ olfactory neurons, using expression values downloaded from (Sammeta et al., 2007), or based on their gene family. Within each group, signals at each position relative to the transcript or CDS start site were averaged and plotted to obtain signal profiles for each gene group.

Quantitative PCR analyses of relative FLAG-H2BE levels in the protein-coding (CDS) regions of histone and OR genes were performed on ligation-mediated PCR-amplified DNA from FLAG or H3 ChIP samples. Experiments were performed in triplicate for each input (FLAG or H3) and primer pair combination starting from 0.2 ng of input DNA per reaction. Reactions were performed using the QuantiTect SYBR Green PCR Kit (Qiagen) and an MJ Opticon 2 instrument (Bio-Rad). Primer pairs (Integrated DNA Technologies) were designed using the Primer-BLAST tool (NCBI). Primer efficiencies were assessed using standard curves and pairs exhibiting efficiencies of >99% were used for analysis.

Protein samples were separated by electrophoresis on a TRIS-glycine-SDS polyacrylamide gel (10–20% gradient; Bio-Rad), and transferred to nitrocellulose membranes. Membranes were washed in TN (5 min, RT), blocked in Blotto (Santa Cruz Biotechnology; 5% in TN; 1 hr, RT), incubated with primary antibodies (each diluted 1:2000 in 5% Blotto [in TNT]; 12 hr, 4°C), washed in TNT (3 × 5 min, RT), incubated with a mixture of Alexa488-conjugated anti-rabbit and Alexa647-conjugated anti-mouse secondary antibodies (Invitrogen; 1:2000 in 5% Blotto [in TNT]; 1–2 hr, RT), and washed in TNT (3 × 10 min, RT). Blots were scanned using a Typhoon Trio Imager and analyzed using ImageQuant software (GE Healthcare Biosciences, Pittsburgh, PA, USA).

Eight-well Streptavidin High Binding Capacity strips (Thermo-Fisher Scientific, Waltham, MA, USA) were washed (Tris [25 mM, pH 7.5], NaCl [150 mM], BSA [1%], and Tween-20 [0.05%]). Custom-synthesized C-terminally-biotinylated peptides (Abgent, San Diego, CA, USA) corresponding to the N-terminal 13 amino acids of H2BE or the mouse consensus H2B and containing a methyl-modified lysine residue at position five were immobilized within the well strips (1 μg per well in block buffer [0.2 µg mouse IgG in wash buffer]; 3 hr, RT), washed, incubated with anti-H2B-Lys5-Me antibody (1:5000–1:5,000,000 in block buffer; 12 hr; 4°C), washed, incubated with anti-rabbit-HRP (1:2000 in block buffer; 1 hr, RT), and washed. ELISA signals were developed using the 1-Step Ultra TMB-ELISA reagent (Thermo Scientific) according to the manufacturer's instructions.

Ten 4-month old H2be-KO and heterozygous littermates were subjected to odor discrimination training using water restriction for motivation, under a behavioral paradigm similar to that described (Uchida and Mainen, 2003), but adapted for mice. In an initial experiment, mice were challenged to discriminate between hexanol and hexanoic acid, and in a second experiment, were challenged with the two stereoisomers of carvone. Prior to the initial experiment, mice were trained to obtain water from two ports unconditionally (day 1), after first poking an odor port (days 2–5), after poking an odor port presenting isoamyl-acetate odor (days 6–13), and after only a single poke of an odor port presenting isoamyl-acetate odor (days 14–20). Odor discrimination training began on day 21. Odors streams, controlled by an olfactometer, were generated from air passed through a filter containing 10% of the concentrated odorant in mineral oil and diluted 1:20 into a stream of clean air.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank N. Uchida and M. Uchida for assistance with odor-discrimination experiments, J. Bergan for assistance with electro-olfactogram experiments, N. Rubinstein for assistance with statistical analyses, D. Storm for Adcy3-KO mice, and D. Reinberg, N. Francis, and members of the Dulac lab for helpful discussions.

Animal experimentation: This study was performed within the facilities of the Harvard University Faculty of Arts and Sciences (HU/FAS) in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All animals were handled according to a protocol approved by the Harvard University Institutional Animal Care and Use Committee (IACUC; protocol #97-03). The HU/FAS animal care and use program maintains full AAALAC accreditation, is assured with OLAW (A3593-01), and is currently registered with the USDA. Every effort was made to minimize animal suffering during this study.

© 2012, Santoro and Dulac

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.