An Epigenetic Trap Stabilizes Singular
Olfactory Receptor Expression
David B. Lyons,1William E. Allen,3Tracie Goh,2Lulu Tsai,3Gilad Barnea,3and Stavros Lomvardas1,2,*
2Department of Anatomy
University of California, San Francisco, San Francisco, CA 94158, USA
3Department of Neuroscience, Brown University, Providence, RI 02912, USA
The molecular mechanisms regulating olfactory
receptor (OR) expression in the mammalian nose
are not yet understood. Here, we identify the tran-
sient expression of histone demethylase LSD1 and
the OR-dependent expression of adenylyl cyclase 3
(Adcy3) as requirements for initiation and stabiliza-
tion of OR expression. As a transcriptional coactiva-
tor, LSD1 is necessary for desilencing and initiating
OR transcription, but as a transcriptional core-
pressor, it is incompatible with maintenance of OR
expression, and its downregulation is imperative for
stable OR choice. Adcy3, a sensor of OR expression
and a transmitter of an OR-elicited feedback, medi-
ates the downregulation of LSD1 and promotes the
differentiation of olfactory sensory neurons (OSNs).
This novel, three-node signaling cascade locks the
epigenetic state of the chosen OR, stabilizes its sin-
gular expression, and prevents the transcriptional
activation of additional OR alleles for the life of the
Olfactory receptors (ORs) are G protein-coupled receptors that
detect odors and regulate the projection of olfactory sensory
neurons (OSNs) to the brain (Buck and Axel, 1991). In the mouse,
ORs are encoded by ?1,400 genes (Young et al., 2002) orga-
nized in gene clusters found on most chromosomes (Sullivan
2011; Michaloski et al., 2006) and are expressed in the main
olfactory epithelium (MOE) in a monogenic and monoallelic
fashion (Chess et al., 1994). An unusual, MOE-specific epige-
netic signature of OR loci, characterized by enrichment for
H3K9me3 and H4K20me3, likely governs what is a seemingly
stochastic OR expression pattern (Magklara et al., 2011). This
epigenetic silencing is reinforced by the aggregation of silenced
OR genes in a few heterochromatic foci that preserve the
expression of only one OR allele in each OSN (Clowney et al.,
2012). The active OR allele in each OSN escapes from the OR
aggregates and relocates to a euchromatic territory where it
frequently interacts with a distant OR enhancer, the H enhancer
(Clowney et al., 2012; Lomvardas et al., 2006). The singularity of
OR expression is essential for olfactory perception because ORs
are localized in the OSN dendrites and axons (Barnea et al.,
2004) and participate in both odorant detection and axon target-
ing (Wang et al., 1998). Neurons expressing the same OR
converge their axons in distinct glomeruli of the olfactory bulb
(Mombaerts et al., 1996), by a process that relies on the identity
of the OR protein and its basal activity levels (Mori and Sakano,
2011). Thus, maintaining the stable and singular expression of
the same OR throughout the life of the neuron is necessary for
the integrity of the topographic map in the olfactory bulb, such
that coherent OR expression may be required for proper odor
decoding in the brain.
Although the molecular mechanisms that stabilize OR expres-
sion are not known, it is established that the expression of
transgenic ORs elicits a negative feedback that prevents the
expression of endogenous OR genes (Lewcock and Reed,
2004; Nguyen et al., 2007; Serizawa et al., 2003). Moreover, line-
age-tracing experiments monitoring the expression of OR alleles
from their endogenous loci showed that OR expression elicits a
positive feedback signal that stabilizes its own choice and pre-
vents OR gene switching in most OSNs (Shykind et al., 2004).
Together, these observations raise questions regarding possible
mechanisms that could stabilize the transcription of one OR
allele, while simultaneously preventing the expression of all
the other OR genes. A simple model that could account for the
existence of an OR-elicited feedback signal emerged from the
able OR transcription, and that OR choice coincides with a
switch from the repressive histone H3K9 methylation to the acti-
vating H3K4 methylation (Magklara et al., 2011). Based on these
observations, singular OR expression could become permanent
if the choice of an intact OR allele suppresses H3K9 and H3K4
demethylases, so that one active OR allele and ?1,000 silent
OR genes preserve their distinct epigenetic states for the life of
Demethylation of H3K9me3 and H3K4me3 are both stepwise
processes that require removal of one methyl group first,
creating H3K9me2 and H3K4me2, respectively, which are then
Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc. 325
demethylase 1, Kdm1a), an amine oxidase, is the only protein
with the enzymatic ability to catalyze lysine demethylation
reactions for both intermediates, H3K9me2 and H3K4me2, and
therefore act as transcriptional coactivator or corepressor,
respectively (Metzger et al., 2005; Shi et al., 2004). The exact
mode of action of LSD1 is influenced by the context of the tran-
scription factor that recruits it to a specific locus and the nature
of the local histone modifications. For example, LSD1 recruit-
ment by androgen receptor to specific loci results in H3K9me2
demethylation and transcriptional activation, whereas LSD1 de-
methylates H3K4me2 and represses transcription as a compo-
nent of the CoREST complex (Metzger et al., 2005; Shi et al.,
2005; Wang et al., 2007; Wissmann et al., 2007).
Here, we show that LSD1, which is transiently expressed dur-
ing OSN differentiation, is involved in both OR gene activation
and post-choice gene switching and, thus, plays a dual role in
OR regulation. Genetic ablation of LSD1 activity prior to OR
choice results in widespread loss of OR expression and failure
of the OSNs to mature and to project axons to the brain. Deletion
of LSD1 immediately after OR activation has no detectable con-
sequences in OR expression and OSN targeting, suggesting that
LSD1 activity is needed only during the initial derepression of the
selected OR.OR expressioninducesthe subsequent expression
of adenylyl cyclase 3 (Adcy3), which promotes OSN maturation
and LSD1 downregulation. Lineage-tracing experiments reveal
increased OR gene switching in Adcy3 knockout (KO) mice, sug-
gesting a requirement for timely LSD1 downregulation for the
stabilization of OR expression. Ectopic expression of transgenic
LSD1 in mature OSNs also perturbs the stability of OR choice,
suggesting that Adcy3 stabilizes OR transcription by downregu-
lating LSD1. Thus, our data connect OR choice with the terminal
differentiation of olfactory neurons and uncover the molecular
underpinnings of a feedback loop that preserves the epigenetic
state of active and silent ORs for the life of the neuron.
Dynamic LSD1 Expression Is Required for Initiation but
Not Maintenance of OR Expression
We hypothesized that initiation of OR transcription requires
H3K9 demethylation, whereas stabilization of singular OR tran-
scription requires suppression of H3K9 and H3K4 demethylases
involved in OR regulation. Because LSD1 has both H3K9 and
H3K4 demethylase activities, we asked whether its expression
pattern is compatible with such a role in OR regulation. RNA-
sequencing (RNA-seq) analysis (Magklara et al., 2011) from
mature OSNs (mOSNs, olfactory marker protein [OMP]-positive)
and progenitor/immature neuronal populations (Neurogenin-1
[Ngn1]-positive) shows that LSD1 is expressed in the Ngn1-
positive cells but reduced by 3.6-fold during differentiation to
the OMP-positive stage (Figure 1A). Immunofluorescence (IF)
reveals high levels of LSD1 protein in the nuclei of Ngn1-positive
cells and their immediate progeny and significant reduction of
LSD1 in more apical, mOSN populations of the adult MOE (Fig-
ure 1B). Two-color RNA fluorescence in situ hybridization
(FISH) experiments demonstrate the mutually exclusive expres-
sion patterns of LSD1 and OMP, verifying the transcriptional
downregulation of LSD1 in OR-expressing mOSNs (Figure 1C).
The dynamic pattern of LSD1 expression, together with recent
microarray analyses showing that injury-induced neurogenesis
in the MOE coincides with LSD1 upregulation (Krolewski et al.,
2013), supports a role for this protein in OR regulation.
To functionally test the role of LSD1 in initiation and mainte-
nance of OR gene expression, we used a conditional LSD1 KO
(Wang et al., 2007) that we deleted at three distinct develop-
mental time points: prior, during, and after OR gene activation.
This was accomplished by using Foxg1-Cre (He ´bert and
McConnell, 2000), MOR28-IRES-Cre (Shykind et al., 2004),
LSD1 deletion before OR expression results in widespread loss
of OR expression, based on both ISH with a pool of OR RNA
probes and IF with OR antibodies as well as a general targeting
deficit of the OSN axons (Figure 1E and Figures S1A and S1B
available online). This analysis was performed in embryonic
day (E) 18.5 MOE sections due to perinatal lethality. In contrast
to the early LSD1 KO, IF for MOR28 shows that LSD1 deletion
immediately after MOR28 activation has no measurable effects
on OR expression or OSN targeting (Figures 1F–1H and S1C).
Similarly, RNA ISH and IF as above show that LSD1 deletion in
mOSNs has no detectable effects on OR expression (Figures
1G and S1D). These data suggest that LSD1 activity is neces-
sary for OR desilencing and initiation of OR transcription but
dispensable for OSN function following OR choice, at least
within the kinetic restrictions imposed by available genetic
To determine the extent of the transcriptional effects on OR
expression, we performed RNA-seq analysis with complemen-
tary DNA (cDNA) libraries prepared from the MOEs of control
and LSD1 KO mice at E18.5. This approach also shows signifi-
cant reduction of OR transcription, both regarding the total num-
ber of OR messenger RNA (mRNA)reads and the numberof ORs
that can be detected in the LSD1 KO mice (Figure 2A). In control
mice, we detect 662 ORs, and in LSD1 KO mice, we only detect
212 ORs with ?4-fold fewer reads in those ORs that are
expressed in LSD1 KO MOEs. This analysis also revealed that
transcription factors known or suspected to activate OR tran-
scription, such as Emx2 and Lhx2 (Hirota and Mombaerts,
2004; McIntyre et al., 2008), are still expressed in the LSD1 KO
mice (Figure 2B), supporting a direct role of LSD1 in OR regula-
tion. Developmental markers of progenitor cells, such as Neu-
rod1, are not affected by LSD1 deletion (Figure 2C). Importantly,
developmental markers that are postmitotic and synchronous
to OR expression, such as GAP43 and NCAM1 or Stmn1,
Dpysl5, Marcksl1, and Ablim1 (Iwema and Schwob, 2003; Kro-
lewski et al., 2013; Nickell et al., 2012), are also only moderately
affected by LSD1 deletion, based on our RNA-seq and ISH
experiments (Figures 2C and S2, respectively). This suggests
that the loss of LSD1 activity and not some downstream devel-
opmental deficit is the cause of OR downregulation in the
LSD1 KO MOE. In contrast, mOSN markers are markedly down-
regulated in the LSD1 KO MOEs (Figures 2F and 2G). The loss of
the mOSN layer in the LSD1 KO likely explains the thinner
expression pattern of GAP43 and NCAM1 and the ?2-fold
reduction observed by RNA-seq (Figures 2C and S2), given
that at this developmental stage, there is some overlap of these
immature markers with mOSN markers.
326 Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc.
ORs Induce Adcy3 Expression
Thisdevelopmental arrestinvitesthehypothesis thatOR expres-
sion might be a prerequisite for OSN maturation. Consistent with
this, an ‘‘empty’’ OSN that does not express OR protein, like
olfactory receptor neurons generated in Drosophila (Hallem
and Carlson, 2006), has yet to be described in the mouse, where
neurons with detectable pseudogene OR expression are imma-
ture and OMP negative (Shykind et al., 2004). Thus, we sought to
rescue the LSD1 KO phenotype by ectopic OR expression (Fig-
ure 3A). We crossed a tetO-MOR28-IRES-LacZ transgene
(Clowney et al., 2012) to two tTA drivers: Gg8-tTa (Nguyen
et al., 2007), which is not affected by the LSD1 deletion since
its expression initiates in immature OSNs (Ryba and Tirindelli,
1995), and OMP-IRES-tTA (Yu et al., 2004), which might
Figure 1. Transient LSD1 Expression Is
Required for OR Expression
(A) mRNA-seq reads per million mapped per
thousand base pairs of exon model (RPKM) for
LSD1 and LSD2 in the mature versus immature/
globose basal cells (Ngn1+).
(B) LSD1 immunofluorescence (IF, red) in the
Ngn1-GFP+MOE at P30.
(C) LSD1 and OMP two-color RNA ISH at P5. DAPI
nuclear stain is shown in blue.
three different MOE-specific Cre recombinase
(E) OR ISH probe pool for 8 class II OR genes in
Foxg1-Cre; LSD1flox/+ and Foxg1-Cre; flox/flox
(class I OR ISH is shown in Figure S1).
(green) in MOE with MOR28 immunofluorescence
(red); coexpressing cells are stably expressing
MOR28 in the absence of LSD1.
(G) Class II OR ISH in OMP-IRES-Cre; LSD1 flox/+
and flox/flox MOEs at P1.
(H) Olfactory bulbs of MOR28-IRES-Cre; LSD1+/+
and MOR28-IRES-Cre; LSD1flox/flox animals
at P30 with a two-color membrane-bound Cre-
reporter: mT before Cre; mG after Cre (mT/mG;
Muzumdar et al., 2007).
See also Figure S1.
preserve the expression of transgenic
MOR28 in mOSNs. These three alleles
were put into the Foxg1-Cre; LSD1 KO
background (Figure 3B). Embryos were
collected at E18.5 and subjected to
whole-mount X-gal staining (Figure S3).
Many LacZ-positive neurons are de-
tected in the MOEs of these mice, and
the X-gal-stained cells have dendrites
that reach the lumen of the olfactory
epithelium (Figures 3Band S3).Strikingly,
IF shows that ectopic expression of
MOR28 restores Adcy3 immunoreac-
tivity in the LSD1 KO mice (Figure 3B),
showing that OR expression controls
The fact that Adcy3 constitutes a faithful marker for OR
expression, in addition to being a marker of OSN maturation,
allowed us to also test whether the levels of LSD1 affect the ki-
netics of OR choice and OSN maturation. We detect a significant
reduction in the number of Adcy3-positive cells in heterozygote
LSD1 KO mice compared to wild-type (WT) littermates (Fig-
ure 3C). This suggests that the levels of LSD1 are not saturating
during OR choice, which may result in a slow and inefficient pro-
Adcy3 Promotes OR Choice Stabilization and OSN
Differentiation via LSD1 Downregulation
The intriguing observation that Adcy3 expression is mutually
exclusive with LSD1 expression and depends upon OR
Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc. 327
expression prompted us to test whether this protein plays a role
Adcy3 is the main adenylyl cyclase in OSNs, and previous
reports have shown that Adcy3 KO OSNs have severe targeting
defects (Chesler et al., 2007; Col et al., 2007; Zou et al., 2007). A
role of Adcy3 in stabilization of OR expression could account for
these targeting deficits, together with activity-dependent pro-
cesses that regulate axon guidance. Given that gene switching
requires the repression of the previously chosen OR and the
desilencing of a new OR allele, both of which could be accom-
plished in part by the dual enzymatic activities of LSD1, we
examined whether Adcy3 deletion affects LSD1 expression.
In agreement with an instructive role for Adcy3 in LSD1 gene
regulation, we find that Adcy3 deletion causes a dramatic exten-
sion of LSD1 immunoreactivity toward the mOSN layers at
postnatal day 21 (P21) (Figures 4A and S4A). Similarly, GAP43
expression is also apically expanded, showing that Adcy3 dele-
tion delays the terminal differentiation of these neurons (Figures
4B and S4B). RNA ISH for OMP shows that the mOSN layer is
reduced and restricted only to the most apical OSN layer,
Figure 2. Early Deletion of LSD1 with
Foxg1-Cre Causes Massive Reduction in
OR Gene Expression and Developmental
Arrest at a Differentiation Stage Synchro-
nous to the Onset of OR Transcription
(A) mRNA-seq RPKM for each Refseq OR in
mouse genome from E18.5 MOE sample. Each
spoke of a given color is the expression value for
that OR in the MOE of that genotype (red: Foxg1-
Cre; LSD1flox/+; green: Foxg1-Cre; Lsd1flox/
OR genes, not actual length of chromosome.
Summary boxplot is shown within Circos plot;
p value < 0.001, unpaired Student’s t test.
(B) RPKM values of the two known transcriptional
activators of OR genes in the LSD1 heterozygote
(bottom panels), respectively: Neurod1, GAP43,
NCAM1, and OMP.
(G) IF for Adcy3 at same embryonic stage; DAPI
nuclear stain is shown in blue.
See also Figure S2.
residing below the sustentacular cells
(Figure 4C). Similarly, IF for b-galactosi-
dase, which is expressed instead of
Adcy3 in this KO strain, shows that tran-
scription of the Adcy3 locus is also
restricted to the most apical OSN layer,
positive-feedback loop, whereby stable
Adcy3 expression is contingent on OR
and Adcy3 proteins (Figures 4D and
S4C). Notably, the Adcy3 KO MOEs
have a thin mOSN layer at this age, sug-
gesting that eventually a small proportion
of neurons settle to a stable and robust
choice of a single OR, as previously reported (Zou et al., 2007),
possibly via low-level, paralogous adenylyl cyclase activity.
This is also evident by the OR expression pattern because at
P21, OR IF shows robust OR expression in the thin LSD1-nega-
tive apical layer of the Adcy3 KO MOE (Figure 4E).
Consistent with a delay in the stabilization of OR expression,
RNA-seq analysis of WT and Adcy3 KO MOEs at P21 shows
that overall OR mRNA levels are moderately reduced in a statis-
tically significant manner (Figure 4F, left panel). Because Adcy3
is expressed after OR choice and in an OR-dependent manner,
these effects likely do not reflect a developmental delay in the
initiation of OR expression but, rather, post-OR choice insta-
bility. Indeed, our RNA-seq analysis shows that overall expres-
sion of pseudogene ORs is not decreased but rather slightly
increased, both at absolute and relative levels (Figure 4F, right
panel and Figure S4D, respectively), supporting the notion that
LSD1+OSNs continue to search ORs, even after the choice of
an intact OR. Because OR pseudogenes can be chosen at the
same frequency as intact ORs (Shykind et al., 2004), but their
expression is less stable, a general deficit in stabilization of
328 Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc.
intact OR expression would favor the representation of pseudo-
To directly test the role of Adcy3 in the stability of OR expres-
sion, we performed lineage-tracing experiments (Shykind et al.,
2004) in control and Adcy3 KO mice by crossing MOR28-
IRES-Cre mice to a Cre-inducible GFP reporter (Muzumdar
et al., 2007). If intact ORs switch in the absence of Adcy3, then
a fraction of GFP-positive neurons should stop expressing the
MOR28-IRES-Cre allele that recombined and activated the
reporter, generating GFP+/Cre?neurons. Moreover, if switching
is rapid, then a fraction of Cre-expressing neurons should
be GFP negative because Cre-mediated recombination takes
6–24 hr (Hayashi and McMahon, 2002; Nakamura et al., 2006).
We performed this analysis at P2 because the Adcy3 KO mice
fail to thrive, and mice with all four alleles did not survive beyond
Figure 3. Ectopic Expression of Transgenic
MOR28 in the LSD1 KO MOE Can Rescue
the Loss of Adcy3 Expression
(A)Model summary of
expression study. Using two tTA drivers, one
active in the immature neuron (Gg8-tTa), and
one active in the mature neuron (OMPitTA), it is
possible to express high levels of MOR28 in the
LSD1 KO MOE in a sporadic fashion. We find that
OR expression is followed by the onset of Adcy3
(B) X-gal staining in sections of LSD1 KO MOE
shows infrequent transgenic MOR28 expression
under the control of two tTA drivers. Whole-mount
image is shown in Figure S3.
(C) Foxg1-Cre; tetO-MOR28-lacZ MOE at E18.5
with either Lsd1 flox/flox; OMPitTA (left panels) or
Lsd1 flox/flox; Gg8-tTa; OMPitTA (right panels).
Adcy3 IF (green); Beta-galactosidase IF (red); and
(D) LSD1 dosage positively correlates with Adcy3
immunoreactivity. Adcy3+cells in E18.5 MOEs
were quantified per unit area in ImageJ. y axis
units are Adcy3+cells per micron of MOE area
considered.Barplot shows mean of two quantified
regions of MOE from one experiment. Error bars
are mean ± standard error of the mean (SEM).
this age. At this age, the majority of the
OSNs are immature, yet we detect OR-
expressing OSNs in Adcy3 KO and an
apical expansion of LSD1 expression,
mice (Figures 4G, S4E, and S4F). Lineage
tracing, however, shows that Adcy3 KO
mice have an ?2-fold increase (Student’s
neurons (Figures 4H and 4I), supporting
the frequent switching phenotype sug-
gested by the increase of pseudogene
The rapid OR switching phenotype
observed in Adcy3 KO pups suggests
that the ectopic LSD1 expression in the Adcy3 KO is the cause
of post-choice OR downregulation, and that sustained LSD1
expression is incompatible with OR transcription. To directly
test the post-choice effects of LSD1 expression, we generated
transgenic tetO-LSD1, which we crossed to OMPitTA mice to
drive expression of LSD1 in mOSNs (Figures 5A and 5B).
OSNs of these mice retain high levels of LSD1 even after OR
choice, causing significant downregulation of OR expression
by IF (Figure 5B and quantification in Figure S5A) and RNA ISH
(Figure 5C). Feeding these mice doxycycline for 3 weeks shuts
off tTA-driven LSD1 expression in mOSNs and restores robust
OR expression (Figures 5B and S5A).
We also brought the P2-IRES-taulacZ reporter (Mombaerts
et al., 1996) into the LSD1-overexpressing background. Consis-
tent with the aforementioned decrease in OR expression, there
Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc. 329
was a dramatic reduction of this OR reporter gene (Figure 5D).
Moreover, olfactory-bulb targeting becomes perturbed in the
LSD1-overexpressing MOE, with regional targeting, by and
large, unaffected but the total number of targeted glomeruli
increasing from 1–2 in the control to roughly a dozen in the
OMPitTA; tetO-LSD1 (Figure 5E and data not shown), further
supporting that LSD1-overexpressing OSNs are unable to settle
on a single OR and switch frequently to other OR alleles from the
LSD1 Triggers Guanine Oxidation of the Active OR DNA
The likely transient interaction of LSD1 with a chosen OR makes
it technically impossible to detect by chromatin immunoprecipi-
tation (ChIP) the binding of LSD1 on an active OR. This technical
ture’’ for the presence of LSD1 at the active OR allele. Such a
mark could be generated by hydrogen peroxide, which, in addi-
Figure 4. Adcy3 Removal Triggers Upregu-
lation of LSD1 Protein Levels and Increase
in OR Gene Switching
(A) P21 sections with IF for LSD1 in Adcy3+/?(top)
and?/?(bottom), respectively. See Figure S4 for
(B and C) Fluorescent RNA ISH for immature
(GAP43) and mature (OMP) neurons in Adcy3+/?
(top) and?/?(bottom), respectively.
(D) Beta-gal (green) and LSD1 (red) IF in Adcy3+/?
porter is knocked into the Adcy3 locus. See the
Extended Experimental Procedures for details.
(E) P21 sections from Adcy3+/?(top) and
(bottom), stained with OR (green, MOR28 and
M50) and LSD1 (red) antibodies. ORs are detected
in a more basal layer in Adcy3+/?but not in
(F) Boxplot of RNA-seq RPKM values for all de-
tected Refseq OR genes (left, n = 1072; p value <
0.01; unpaired Student’s t test) and OR pseudo-
genes (right, n = 48, not significant) in corre-
sponding genotype from P21 MOE. RPKMs of
unexpressed intact and pseudogene ORs were
excluded. See also Figure S4.
(G)P3 RNA ISHforM50 and MOR28 as in(E).IF for
ORs and LSD1 is shown in Figure S4.
(H) P2 MOE from MOR28-IRES-Cre; Cre-reporter
mice in Adcy3 WT or KO background. Cre IF
(magenta) and mT/mG reporter (green) are shown.
sections of P2 WT and Adcy3 KO mice were
countedand plotted as aratio of single- to double-
positive cells. Single animal, ten sections quanti-
fied. P value = 0.007, Student’s unpaired t test.
?/?(bottom), respectively. A lacZ re-
ical byproduct of LSD1-mediated lysine
demethylation (Anand and Marmorstein,
2007). Importantly, other histone deme-
thylases do not involve the generation of
reactive oxygen species like FAD-depen-
dent LSD1 (Hou and Yu, 2010). Hydrogen
peroxide tends to selectively oxidize guanosine to 8-oxoguano-
sine (8-oxodG), thus, we reasoned that the extensive demethy-
lation of an OR locus would generate enough hydrogen peroxide
to locally modify guanosines of the chosen allele, as has
been implied by the recruitment of OGG1, a DNA repair protein
that binds to 8-oxodG at LSD1-regulated promoters (Perillo
et al., 2008).
To detect 8-oxodG on a genomic locus, we developed a DNA
immunoprecipitation (DIP) assay with an antibody specific for
this modified base. We performed preliminary titration experi-
ments with a synthetic DNA template derived from the Cre
sequence. This analysis showed that a commercially available
antibody could immunoprecipitate this PCR-synthesized Cre
terpart (Figure S6A; see Experimental Procedures). Thus, a DIP-
based strategy is sensitive and specific enough for the detection
of this modified base on an active OR allele. DIP-quantitative
330 Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc.
PCR (qPCR) analysis with DNA prepared from E18.5 MOEs from
WT, heterozygote, and homozygote LSD1 KO mice shows
LSD1-dependent 8-oxodG enrichment on two OR loci tested,
P2 and MOR28 (Figure 6A). The enrichment levels for 8-oxodG
are low in this experiment and comparable with the enrichment
of a control locus, likely because we performed DIP in whole
MOE populations in which the two OR alleles are expressed in
a very low fraction of cells.
To test whether 8-oxodG enrichment stems from transcrip-
tionally active OR alleles, we fluorescence-activated cell-sorted
(FACS) GFP+neurons expressing olfactory receptor P2 from
P2-IRES-GFP knockin mice. Using DIP-qPCR analysis, we
Figure 5. Ectopic Expression of Trans-
genic LSD1 in the Mature Neuron Layer
Causes Reversible Destabilization of OR
(A) Model summarizing results in adult MOE
regarding the expression pattern of ORs and
LSD1 under different genetic manipulations: OR-
expressing OSNs are prevalent in LSD1-negative
layer regardless of genotype. Weakly OR-ex-
pressing OSNs are present in OMPitTA; tetO-
LSD1 mice before dox, but robust expression
returns following dox and the reduction of LSD1
(B) Adult OMP-tTA; tetO-LSD1 mice were raised
until 3 weeks and either placed on doxycycline for
3 weeks to shut off tTA activation or maintained on
dox-free food. Control littermate mice (OMPitTA
only)were also placed ondoxycycline for 3weeks.
Six-week-old MOEs were harvested, and IF was
performed for LSD1 (red top panel) or Olfr49 (C6)
(red two bottom panels with or without DAPI). See
also Figure S5.
(C) Misexpressing LSD1 in the MOE with OMPitTA
reduced OR expression in the MOE. Chromogenic
ISH OR pool (15 OR probes total) in OMPitTA (left)
and OMPitTA; tetO-LSD1(right).
(D) P2-lacZ (top left) and P2-lacZ;OMPitTA; tetO-
LSD1 (bottom left) MOE and bulb following whole-
mount X-gal staining in P25 mice. Olfactory bulb
with beta-galactosidase IF (green). Despite the
low levels of beta-gal at the cell bodies due to
switching, the protein appears stable at the axons
(Clowney et al., 2012), which allows the visualiza-
tion of additional glomeruli.
quantified the relative enrichment of
8-oxodG on the active and inactive OR
alleles. We detect an ?3-fold higher
enrichment of 8-oxodG on the P2 allele,
compared to the enrichment of this base
on the inactive OR genes tested (Fig-
ure 6B). Because the majority of the
sorted P2 neurons are mature, their
transcription was initiated days or weeks
before, and thus they have long ago
downregulated LSD1. The enrichment
levels we obtained imply that 8-oxodG
is stable on OR DNA, due to a probably inefficient DNA repair
process stemming from the low expression levels of OGG1
and NEIL1 (Klungland and Bjelland, 2007), as shown by our
RNA-seq analysis (data not shown). To test this, we used a
Cre-OR knockin line whereby Cre replaces the coding sequence
of MOR28. MOR28-delete-Cre-expressing OSNs treat this allele
like a pseudogene OR and switch from it in order to express a
functional OR, often the functional MOR28 allele (Shykind
et al., 2004). We crossed this delete-Cre line to the membrane-
GFP (mT/mG) Cre reporter (Muzumdar et al., 2007) and isolated
the GFP-positive neurons with FACS. Although transcription
from the deleted MOR28 allele has ceased in most GFP-positive
Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc. 331
neurons (Figure 6C), we detect significant enrichment for
8-oxodG on the Cre locus in the GFP-positive cells (Figure 6D).
Interestingly, we also detect enrichment for this modified base
on the WT MOR28 allele, which is explained by the high fre-
quency by which these OSNs switch to this allele.
To test the possibility that 8-oxodG is a reflection of unpro-
tected DNA due to transcription and not a direct consequence
of LSD1-dependent demethylation, we performed Illumina
sequencing in DIP from the whole MOE. DIP-seq analysis shows
that the enrichment for 8-oxodG does not correlate with levels of
transcription. We sorted the mouse genes into four quartiles of
transcription levels, and we plotted 8-oxodG levels for each
quartile. The mean 8-oxodG reads per kilobase per million map-
ped reads (RPKMs) are essentially identical between the four
expression quartiles, suggesting that the enrichment of this
base on the chosen OR allele is not a byproduct of the unusual
transcription rates of OR alleles, but rather it is indicative of
LSD1’s proximity to that OR locus (Figure 6E).
Figure 6. LSD1 Generates Stable 8-oxodG
at Active OR Genes
(A) 8-oxodG DIP was performed on sonicated
genomic DNA (gDNA) from LSD1 WT, heterozy-
gote, and KO MOEs at E18.5.
(B) 8-oxodG-DIP-qPCRs from gDNA of P2-GFP-
sorted cells from P30 mice; values shown are
mean of qPCR performed in duplicate from one
representative experiment. Error bars represent
(C) P30 MOE of MOR28-del-Cre; Cre-reporter
mouse, with Cre IF (magenta, arrowhead) and
mT/mG Cre-reporter (green are cells that have
expressed Cre to levels sufficient to recombine
reporter locus). DAPI nuclear stain is blue.
(D) 8-oxodG-DIP-qPCRs from Cre-reporter-posi-
tive neuron gDNA at P30, as shown in (C). Values
shown are mean of qPCR performed in duplicate
from one representative experiment. Error bars
represent the SEM.
(E) DIP-seq analysis of an E18.5 WT 8-oxodG
generated from Figure 1 mRNA-seq. Boxplots
show mean 8-oxodG RPKM for each expression
quartile, where mean is demarcated by red bar.
(F and G) 8-oxodG-DIP-qPCRs from gDNA from
whole MOE of LSD1-overexpressing mice and
Adcy3 KO mice, respectively. Relative enrichment
is ± SEM as above.
See also Figure S6 for control experiments.
Finally, we measured the enrichment
levels of 8-oxodG in Adcy3 KO and
OMPitTA; tetO-LSD1 MOEs, which have
abnormally high levels of LSD1 protein.
We find significant increases of 8-oxodG
levels in both the Adcy3 KO and the
LSD1-overexpressing mice (Figures 6F
and 6G). Notably, in WT adults, the base-
line 8-oxodG levels are lower than in em-
bryos, which is probably explained by the
significantly higher proportion of LSD1-
expressing cells in embryonic than in adult MOEs. Interestingly,
under these overexpression conditions, there is an expected
loss of specificity at the LSD1-dependent DNA oxidation. More-
over, these results also show that in the OMPitTA; tetO-LSD1
MOE, only a small fraction of OR genes become ectopically
demethylated. A 10- to 20-fold increase of 8-oxodG levels in
DIPs from mixed MOE populations suggests that each OR allele
is H3K9 free in 1%–2% of the cells instead of 0.1% that is calcu-
lated in WT MOEs. In agreement with the notion that ORs remain
epigenetically silenced in >95% of the cells, and that in each
ChIP-qPCR for H3K9me3 shows similar enrichment levels
between WT and LSD1-overexpressing MOEs (Figure S6C).
Our understanding of OR regulation changed by the realization
that OR expression elicits a feedback signal that prevents the
332 Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc.
expression of additional ORs and/or stabilizes the expression of
the chosen one (Lewcock and Reed, 2004; Nguyen et al., 2007;
Serizawa et al., 2003; Shykind et al., 2004). However, neither
the mechanism of OR gene activation nor the pathway that sta-
bilizes this activation were previously understood. We recently
showed that ORs undergo heterochromatic silencing in the
MOE, at a stage prior to OR expression, and that OR choice co-
incides with an epigenetic switch from H3K9me3 to H3K4me3
at the chosen allele (Magklara et al., 2011). The data presented
here demonstrate that the epigenetic signature of active and
silent ORs affords the deployment of a feedback mechanism
that prevents the activation of additional ORs, while at the
same time stabilizing the expression of the chosen allele. This
epigenetic switch, combined with the dual function of LSD1
as H3K9 and H3K4 demethylase, not only renders the chosen
OR immune to the feedback signal but also makes the subse-
quent downregulation of LSD1 imperative for the stabilization
of OR choice.
These observations pose a significant question: why does
LSD1 activate OR transcription before an OR is chosen but
repress OR transcriptions after OR choice? Before OR choice,
ORs are marked only by H3K9 methylation, thus LSD1 can
However, after OR choice, the chosen OR switches from H3K9
to H3K4 methylation (Magklara et al., 2011). Therefore, at this
stage, if LSD1 is still present and recruited to the chosen OR, it
can only demethylate H3K4, resulting in the repression of this
OR. Thus, the same molecule before OR choice is by default
an activator but after choice is a repressor for an already acti-
vated OR and a potential activator for the remaining silenced
OR genes. Therefore, local epigenetic context determines the
exact action of LSD1 and makes the chosen OR susceptible to
LSD1-mediated repression.Alternatively,iftheOSN cannotsup-
port OR transcription from two different loci simultaneously, it is
Figure 7. A Three-Node Signaling Cascade
Combined with a Feedback Signal that
Generates Epigenetic Memory
(A) Stabilization of OR expression is achieved by
an Adcy3-dependent ‘‘trap’’ such that the func-
tional chosen OR cannot be turned off once LSD1
is downregulated by its induction of Adcy3. This
trap is caused by removing LSD1 from the
signaling circuit, which allows stable transcription
to ensue (represented by dashed line that reflects
the indirect OR stabilization by Adcy3).
(B) Pseudogene ORs (ORJ) are unable to activate
Adcy3, and thus OSNs that have chosen these
ORs maintain the ability to re-choose and use
LSD1 to transcriptionally silence the ORJ.
(C) Alternatively, LSD1 may not directly silence
the previously chosen OR but instead cause its
repression by activating an additional OR allele.
possible that the downregulation of the
chosen OR is not a direct consequence
of H3K4 demethylation by LSD1 but an
indirect effect of the fact that an addi-
tional OR has been activated. Finally, as
in every genetic manipulation, indirect effects from either the
deletion or the overexpression of LSD1 could contribute to the
It is worth emphasizing that the genetic manipulations pre-
sented here affected only the stability of OR choice and not the
singularity of expression, unlike when we disrupted nuclear OR
aggregation by ectopic LBR expression (Clowney et al., 2012).
The fact that ORs aggregate in large nuclear foci makes the
majority of ORs inaccessible to LSD1, explaining why most
ORs remain heterochromatic after LSD1 overexpression. Thus,
the mechanism that affords the selection of only one out of thou-
sands of OR alleles is different than the mechanism that makes
this selection permanent.
Feedback Signal versus Feedforward Loop
Our data, together with the established requirement of intact
OR protein for the generation of the feedback signal, lead to
the following regulatory model: LSD1, in complex with an as-
yet-unidentified H3K9me3 demethylase, desilence a previously
heterochromatinized OR allele, allowing H3K4 trimethylation
and transcriptional activation. If this allele encodes an intact
OR, then it will induce Adcy3 expression, LSD1 downregulation,
and OSN maturation, generating an ‘‘epigenetic’’ trap that will
preserve OR expression, cellular identity, and targeting speci-
ficity, as long as the underlying transcription factor milieu
remains unaltered (Figure 7A). In contrast, if the initially chosen
allele is a pseudogene and does not produce OR protein, then
this particular choice cannot induce Adcy3 expression, and
LSD1 will not be turned off. With LSD1 activity still present, an
additional OR can become desilenced via H3K9 demethylation,
but also the previously chosen OR allele can become demethy-
lated at H3K4 and turned off. Thus, failure to terminate LSD1
activity results in OR gene switching, and this process will
continue until an intact OR is expressed (Figure 7B).
Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc. 333
Different versions of feedforward developmental loops in tran-
scription factor regulation were recently described in the differ-
entiation of Drosophila photoreceptor neurons (Johnston et al.,
2011) and in various examples of mammalian differentiation
(Neph et al., 2012). In these systems, as in a plethora of cases
where a network of interactions has been mapped (Alon,
2007), establishment of cellular identity did not require downre-
gulation of the activator that initiates a specific differentiation
program, as is the case with LSD1 here. A major difference in
OR regulation is the existence of an extraordinary number of
similar promoters and a strict requirement for singularity in OR
expression, which likely makes a feedforward circuit ineffective.
Instead, the three-node signaling cascade described here,
which locks the epigenetic states of the chosen allele and of
the silent ORs, can assure both singularity and robustness.
Using LSD1, which has coactivator and corepressor activities,
as an initiator of this cascade provides the additional advantage
of autocorrection, through the post-choice repression of
Adcy3 Has Multiple Functions in the MOE
The finding that Adcy3, which requires the LSD1-dependent OR
gene activation to be expressed, induces rapid LSD1 downre-
gulation makes this protein both a sensor for a productive OR
choice and a transmitter of the feedback signal that stabilizes
its own expression and the expression of the chosen OR, while
promoting OSN differentiation. It is intriguing that Adcy3
induces LSD1 downregulation because the role of OR activity
in the feedback signal is unclear. Although odorant-induced
activity is required for singular OR expression in the septal
organ (Tian and Ma, 2008), deletion of various components of
OR signaling, such as Golf(a) and Cnga2, do not affect OR
expression (Belluscio et al., 1998; Brunet et al., 1996), unlike
in Drosophila photoreceptor neurons (Vasiliauskas et al.,
2011). Furthermore, mutation of the DRY motif, which prevents
OR interaction with G(a) proteins, has no impact on the singu-
larity of OR expression (Imai et al., 2006). Although the effects
of the DRY mutation in the stability of OR expression were not
addressed, it is possible that there are additional, G(a)-indepen-
dent mechanisms by which an OR activates Adcy3 signaling, or
that LSD1 dowregulation requires only OR-dependent Adcy3
expression and not OR-dependent Adcy3 activation. In either
scenario, only low levels of cAMP, generated by basal Adcy3
activity, might be sufficient to induce LSD1 downregulation
and OSN maturation, as in the Adcy3 KO mice, some OSNs
eventually mature and turn off LSD1. Low expression of other
adenylyl cyclases may eventually generate enough cAMP to
elicit a feedback signal in the Adcy3 KO OSNs. In any case,
Adcy3 occupies a critical checkpoint role in the development
of the peripheral olfactory system: it regulates the stability of
OR choice, the targeting of OSN axons, and the longevity of
olfactory neurons (Santoro and Dulac, 2012).
A Transcriptional Regulator with Mutagenic Potential
The detection of 8-oxodG on the chosen OR allele is sugges-
tive of extensive LSD1-mediated demethylation activity in
close proximity to the chosen OR allele. Given that ORs are
embedded in continuous blocks of methylated H3K9 (Magklara
et al., 2011), demethylation of this lysine residue during OR
activation is a plausible explanation for the local production
of hydrogen peroxide and the accumulation of 8-oxodG at
the chosen OR. Guanosine oxidation, by default, would not
have consequences in the genomic stability of OSNs, as they
are postmitotic and relatively short-lived. However, were an
LSD1-mediated demethylation responsible for OR activation
during spermatogenesis (Fukuda et al., 2004), this could pro-
vide a mechanistic explanation for the high AT-rich content of
OR genes and the extreme intra- and interspecies polymor-
phisms observed in this gene family, as 8-oxodG frequently
pairs with adenosine rather than cytosine during DNA replica-
tion (Grollman and Moriya, 1993). Thus, LSD1-mediated dere-
pression in the germline could explain both the drift toward
high AT content (Glusman et al., 2001) and the evolutionary
plasticity of olfactory receptor genes(Clowney et al., 2011; Nii-
mura and Nei, 2007). Moreover, the observation that deletion of
Adcy3 results in substantial increase of DNA oxidation in the
OSN nuclei invites speculation regarding the role of neuronal
activity pathways in protecting central nervous system (CNS)
neurons from DNA oxidation and its deleterious long-term
In summary, ORs provide an unusual example in biology,
whereby a transmembrane receptor protein specialized in
odorant detection functions also as a molecular organizer of
the sensory neuron. The finding that Adcy3 expression and
OSN differentiation depend upon OR expression suggests that
there are no temporal restrictions or developmental windows
ses a functional OR, allowing a slow, inefficient, and stochastic
process for the choice of only one out of thousands of available
alleles. The pleiotropic function of ORs in odor detection, OSN
maturation, axonal wiring, and OSN longevity makes the periph-
eral olfactory system ‘‘self-organizing’’ and centered solely
around the identity of the OR, which may have facilitated the
rapid expansion of this gene family during tetrapod evolution
(Niimura and Nei, 2007). To accommodate adaptation in novel
andvariable ecologicalniches, olfaction hasremainedextremely
plastic, both at the level of the genomic integrity of the chemore-
ceptors and at the transmission and interpretation of odorant
information in piriform cortex (Choi et al., 2011). For a sensory
system that lacks ‘‘labeled lines’’ and where polymorphisms
appear constantly, ascribing such a central developmental role
to the receptor protein itself prevents pseudogene ORs from
compromising the sensitivity and discriminatory power of the
olfactory system (Shykind et al., 2004). The initial screening for
OR quality is further fortified by a secondary, activity-dependent
screen that gradually eliminates OSNs that are seldom used,
affording individualized adaptation to an extremely plastic
system (Santoro and Dulac, 2012).
Mice and Strains Used
All mice were housed in standard conditions with a 12 hr light/dark cycle and
access to food and water ad libitum and in accordance with the University of
California IACUC guidelines. All strains were maintained on a mixed genetic
background. Detailed information on the various mouse strains used is pro-
vided in the Extended Experimental Procedures.
334 Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc.
In Situ Hybridization and Immunofluorescence
IF and ISH was performed as previously described (Clowney et al., 2012).
Information on the riboprobes and antibodies used can be found in the
Extended Experimental Procedures. Confocal images were collected with
the Zeiss LSM 700, and brightfield images were collected on the Zeiss Axio-
skop Plus. All image processing was carried out with ImageJ (NIH).
ChIP was performed as described in Magklara et al. (2011).
DNA Preparation and Immunoprecipitation
Genomic DNA was purified either following FACS or total MOE dissection
using DNeasy genomic DNA isolation kit (QIAGEN). Purified genomic DNA
wassonicatedinPBSwith0.5% Tween-20 toapeak around400bpfragments
using the Bioruptor (Diagenode). For sorted cells, fragmentation of DNA was
assumed to be complete following 15 to 30 min of sonication using medium-
tohigh-power outputwithsamplesinicewater. 8-oxodG monoclonalantibody
(Trevingen) was incubated with DNA rotating overnight at 4?C. Immunoprecip-
itation and washes were carried out in PBS-Tween 0.05%, and DNA elution
buffer consisted of 0.1M NaAOc and 1% SDS in TE (pH 8).
DNA Deep Sequencing
Oligonucleotide reads were generated for Lsd1 and Adcy3 mutant and control
mRNA libraries as well as 8-oxodG libraries using the Genome Analyzer IIx or
HiSeq2000 (Illumina). Sequencing libraries were prepared with standard
methods (Magklara et al., 2011), but in the case of the mRNA, the ScriptSeq
kit (Epicenter) was used. Detailed information can be found in the Extended
Supplemental Information includes Extended Experimental Procedures and
six figures and can be found with this article online at http://dx.doi.org/10.
Excellent technical support was provided by Zoe Evans. We would like to
thank Dr. Geoff Rosenfeld for the LSD1 conditional KO mice and Dr. Nicholas
Ryba for the Gg8 tTA transgenic mice. We would also like to thank Drs. Shah
and Ngai, as well as members of the Lomvardas lab, for critical reading of the
manuscript. This project was funded by the Roadmap for Epigenomics grant
#5R01DA030320-02 and a EUREKA grant #5R01MH091661-02, as well as
the Mcknight Foundation.
Received: February 14, 2013
Revised: May 6, 2013
Accepted: June 20, 2013
Published: July 18, 2013
Alon, U. (2007). Network motifs: theory and experimental approaches. Nat.
Rev. Genet. 8, 450–461.
Anand, R., and Marmorstein, R. (2007). Structure and mechanism of lysine-
specific demethylase enzymes. J. Biol. Chem. 282, 35425–35429.
Barnea, G., O’Donnell, S., Mancia, F., Sun, X., Nemes, A., Mendelsohn, M.,
and Axel, R. (2004). Odorant receptors on axon termini in the brain. Science
are anosmic. Neuron 20, 69–81.
Brunet, L.J., Gold, G.H., and Ngai, J. (1996). General anosmia caused by a
targeted disruption of the mouse olfactory cyclic nucleotide-gated cation
channel. Neuron 17, 681–693.
Buck, L., and Axel, R. (1991). A novel multigene family may encode odorant
receptors: a molecular basis for odor recognition. Cell 65, 175–187.
Miller, M.C., and Firestein, S. (2007). A G protein/cAMP signal cascade is
required for axonal convergence into olfactory glomeruli. Proc. Natl. Acad.
Sci. USA 104, 1039–1044.
Chess, A., Simon, I., Cedar, H., and Axel, R. (1994). Allelic inactivation
regulates olfactory receptor gene expression. Cell 78, 823–834.
Choi, G.B., Stettler, D.D., Kallman, B.R., Bhaskar, S.T., Fleischmann, A., and
Axel, R. (2011). Driving opposing behaviors with ensembles of piriform
neurons. Cell 146, 1004–1015.
Clowney, E.J., Magklara, A., Colquitt, B.M., Pathak, N., Lane, R.P., and Lom-
vardas, S. (2011). High-throughput mapping of the promoters of the mouse
olfactory receptor genes reveals a new type of mammalian promoter and
provides insight into olfactory receptor gene regulation. Genome Res. 21,
Clowney,E.J.,LeGros, M.A.,Mosley,C.P., Clowney,F.G.,Markenskoff-Papa-
dimitriou, E.C., Myllys, M., Barnea, G., Larabell, C.A., and Lomvardas, S.
(2012). Nuclear aggregation of olfactory receptor genes governs their mono-
genic expression. Cell 151, 724–737.
Col, J.A., Matsuo, T., Storm, D.R., and Rodriguez, I. (2007). Adenylyl cyclase-
dependent axonal targeting in the olfactory system. Development 134, 2481–
and Jaenisch, R. (2004). Mice cloned from olfactory sensory neurons. Nature
acterizationof amousetesticular olfactory receptorand its role in chemosens-
ing and in regulation of sperm motility. J. Cell Sci. 117, 5835–5845.
Glusman, G., Yanai, I., Rubin, I., and Lancet, D. (2001). The complete human
olfactory subgenome. Genome Res. 11, 685–702.
Grollman, A.P., and Moriya, M. (1993). Mutagenesis by 8-oxoguanine: an
enemy within. Trends Genet. 9, 246–249.
Hallem, E.A., and Carlson, J.R. (2006). Coding of odors by a receptor reper-
toire. Cell 125, 143–160.
Hayashi, S., and McMahon, A.P. (2002). Efficient recombination in diverse tis-
sues by atamoxifen-inducibleform of Cre: atool for temporally regulated gene
activation/inactivation in the mouse. Dev. Biol. 244, 305–318.
He ´bert, J.M., and McConnell, S.K. (2000). Targeting of cre to the Foxg1 (BF-1)
locus mediates loxP recombination in the telencephalon and other developing
head structures. Dev. Biol. 222, 296–306.
Hirota, J., and Mombaerts, P. (2004). The LIM-homeodomain protein Lhx2 is
required for complete development of mouse olfactory sensory neurons.
Proc. Natl. Acad. Sci. USA 101, 8751–8755.
Hou, H., and Yu, H. (2010). Structural insights into histone lysine demethyla-
tion. Curr. Opin. Struct. Biol. 20, 739–748.
Imai, T., Suzuki, M., and Sakano, H. (2006). Odorant receptor-derived cAMP
signals direct axonal targeting. Science 314, 657–661.
Iwema, C.L., and Schwob, J.E. (2003). Odorant receptor expression as a
function of neuronal maturity in the adult rodent olfactory system. J. Comp.
Neurol. 459, 209–222.
Johnston, R.J., Jr., Otake, Y., Sood, P., Vogt, N., Behnia, R., Vasiliauskas, D.,
McDonald, E., Xie, B., Koenig, S., Wolf, R., et al. (2011). Interlocked feedfor-
ward loops control cell-type-specific Rhodopsin expression in the Drosophila
eye. Cell 145, 956–968.
Klungland, A., and Bjelland, S. (2007). Oxidative damage to purines in DNA:
role of mammalian Ogg1. DNA Repair (Amst.) 6, 481–488.
Krolewski, R.C., Packard, A., and Schwob, J.E. (2013). Global expression
profiling of globose basal cellsand neurogenicprogressionwithin theolfactory
epithelium. J. Comp. Neurol. 521, 833–859.
Lewcock, J.W., and Reed, R.R. (2004). A feedback mechanism regulates
monoallelic odorant receptor expression. Proc. Natl. Acad. Sci. USA 101,
Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc. 335
Lomvardas, S., Barnea, G., Pisapia, D.J., Mendelsohn, M., Kirkland, J., and
Axel, R. (2006). Interchromosomal interactions and olfactory receptor choice.
Cell 126, 403–413.
Magklara, A., Yen, A., Colquitt, B.M., Clowney, E.J., Allen, W., Markenscoff-
Papadimitriou, E., Evans, Z.A., Kheradpour, P., Mountoufaris, G., Carey, C.,
et al. (2011).An epigenetic signature for monoallelic olfactory receptor expres-
sion. Cell 145, 555–570.
McIntyre, J.C., Bose, S.C., Stromberg, A.J., and McClintock, T.S. (2008).
Emx2 stimulates odorant receptor gene expression. Chem. Senses 33,
Metzger, E., Wissmann, M., Yin, N., Mu ¨ller, J.M., Schneider, R., Peters, A.H.,
Gu ¨nther, T., Buettner, R., and Schu ¨le, R. (2005). LSD1 demethylates repres-
sive histone marks to promote androgen-receptor-dependent transcription.
Nature 437, 436–439.
Michaloski,J.S., Galante, P.A.,and Malnic,B.(2006). Identification of potential
regulatory motifs in odorant receptor genes by analysis of promoter
sequences. Genome Res. 16, 1091–1098.
Mombaerts, P., Wang, F., Dulac, C., Chao, S.K., Nemes, A., Mendelsohn, M.,
Edmondson, J., and Axel, R. (1996). Visualizing an olfactory sensory map. Cell
Mori, K., and Sakano, H. (2011). How is the olfactory map formed and inter-
preted in the mammalian brain? Annu. Rev. Neurosci. 34, 467–499.
Muzumdar, M.D., Tasic, B., Miyamichi, K., Li, L., and Luo, L. (2007). A global
double-fluorescent Cre reporter mouse. Genesis 45, 593–605.
Nakamura, E., Nguyen, M.T., and Mackem, S. (2006). Kinetics of tamoxifen-
regulated Cre activity in mice using a cartilage-specific CreER(T) to assay
temporal activity windows along the proximodistal limb skeleton. Dev. Dyn.
Neph, S., Stergachis, A.B., Reynolds, A., Sandstrom, R., Borenstein, E., and
Stamatoyannopoulos, J.A. (2012). Circuitry and dynamics of human transcrip-
tion factor regulatory networks. Cell 150, 1274–1286.
Nguyen, M.Q., Zhou, Z., Marks, C.A., Ryba, N.J., and Belluscio, L. (2007).
Prominent roles for odorant receptor coding sequences in allelic exclusion.
Cell 131, 1009–1017.
Nickell, M.D., Breheny, P., Stromberg, A.J., and McClintock, T.S. (2012).
Genomics of mature and immature olfactory sensory neurons. J. Comp. Neu-
rol. 520, 2608–2629.
Niimura, Y., and Nei, M. (2007). Extensive gains and losses of olfactory recep-
tor genes in mammalian evolution. PLoS ONE 2, e708.
Perillo, B., Ombra, M.N., Bertoni, A., Cuozzo, C., Sacchetti, S., Sasso, A.,
Chiariotti, L., Malorni, A., Abbondanza, C., and Avvedimento, E.V. (2008).
DNA oxidation as triggered by H3K9me2 demethylation drives estrogen-
induced gene expression. Science 319, 202–206.
Ryba, N.J., and Tirindelli, R. (1995). A novel GTP-binding protein gamma-
subunit, G gamma 8, is expressed during neurogenesis in the olfactory and
vomeronasal neuroepithelia. J. Biol. Chem. 270, 6757–6767.
Santoro, S.W., and Dulac, C. (2012). The activity-dependent histone variant
H2BE modulates the life span of olfactory neurons. Elife 1, e00070.
Serizawa, S., Miyamichi, K., Nakatani, H., Suzuki, M., Saito, M., Yoshihara, Y.,
and Sakano, H. (2003). Negative feedback regulation ensures the one recep-
tor-one olfactory neuron rule in mouse. Science 302, 2088–2094.
Shi, Y., Lan, F., Matson, C., Mulligan, P., Whetstine, J.R., Cole, P.A., Casero,
R.A., and Shi, Y. (2004). Histone demethylation mediated by the nuclear amine
oxidase homolog LSD1. Cell 119, 941–953.
of LSD1 histone demethylase activity by its associated factors. Mol. Cell 19,
Shykind, B.M., Rohani, S.C., O’Donnell, S., Nemes, A., Mendelsohn, M., Sun,
Y., Axel, R., and Barnea, G. (2004). Gene switching and the stability of odorant
receptor gene choice. Cell 117, 801–815.
Sullivan, S.L., Adamson, M.C., Ressler, K.J., Kozak, C.A., and Buck, L.B.
(1996). The chromosomal distribution of mouse odorant receptor genes.
Proc. Natl. Acad. Sci. USA 93, 884–888.
neurons expressing multiple odorant receptors in the mouse septal organ.
Mol. Cell. Neurosci. 38, 484–488.
Jr., Lidder, P., Vogt, N., Celik, A., and Desplan, C. (2011). Feedback from
rhodopsin controls rhodopsin exclusion in Drosophila photoreceptors. Nature
Wang, F., Nemes, A., Mendelsohn, M., and Axel, R. (1998). Odorant receptors
govern the formation of a precise topographic map. Cell 93, 47–60.
Wang, J., Scully, K., Zhu, X., Cai, L., Zhang, J., Prefontaine, G.G., Krones, A.,
Ohgi, K.A., Zhu, P., Garcia-Bassets, I., et al. (2007). Opposing LSD1 com-
Nature 446, 882–887.
Wissmann, M., Yin, N., Mu ¨ller, J.M., Greschik, H., Fodor, B.D., Jenuwein, T.,
Vogler, C., Schneider, R., Gu ¨nther, T., Buettner, R., et al. (2007). Cooperative
gene expression. Nat. Cell Biol. 9, 347–353.
Young, J.M., Friedman, C., Williams, E.M., Ross, J.A., Tonnes-Priddy, L., and
Trask, B.J. (2002). Different evolutionary processes shaped the mouse and
human olfactory receptor gene families. Hum. Mol. Genet. 11, 535–546.
Yu, C.R., Power, J., Barnea, G., O’Donnell, S., Brown, H.E., Osborne, J., Axel,
R., and Gogos, J.A. (2004). Spontaneous neural activity is required for the
establishment and maintenance of the olfactory sensory map. Neuron 42,
Zhang, X., and Firestein, S. (2002). The olfactory receptor gene superfamily of
the mouse. Nat. Neurosci. 5, 124–133.
Zou, D.J., Chesler, A.T., Le Pichon, C.E., Kuznetsov, A., Pei, X., Hwang, E.L.,
and Firestein, S. (2007). Absence of adenylyl cyclase 3 perturbs peripheral
olfactory projections in mice. J. Neurosci. 27, 6675–6683.
336 Cell 154, 325–336, July 18, 2013 ª2013 Elsevier Inc.