IO

La

1b

c

v

AtiorifBtanbCcescttssP

Ki

CA(mCefaSsd

*g8EAcnmed

Neuroscience 170 (2010) 881–892

0d

NTEGRATION OF CO2 AND ODORANT SIGNALS IN THE MOUSE

LFACTORY BULB

rasmsmpgsSsoWtdAGc

CodHgitsObvcpa2oHciACotoCC

cqr

. GAO,a,b J. HU,b C. ZHONGb AND M. LUOb,c*

Institute of Biophysics, Chinese Academy of Sciences, Beijing,00101 PR China

National Institute of Biological Sciences, Beijing, 102206 PR China

Department of Biological Sciences and Biotechnology, Tsinghua Uni-ersity, 100084 PR China

bstract—Carbon dioxide (CO2) is an important environmen-al cue for many animal species. In both vertebrates andnvertebrates, CO2 is detected by a specialized subset oflfactory sensory neurons (OSNs) and mediates several ste-eotypical behaviors. It remains unknown how CO2 cues arentegrated with other olfactory signals in the mammalian ol-actory bulb, the first stage of central olfactory processing.y recording from the mouse olfactory bulb in vivo, we found

hat CO2-activating neurons also respond selectively to odor-nts, many of which are putative mouse pheromones andatural odorants. In addition, many odorant-responsive bul-ar neurons are inhibited by CO2. For a substantial number ofO2-activating neurons, binary mixtures of CO2 and a spe-ific odorant produce responses that are distinct from thosevoked by either CO2 or the odorant alone. In addition, for aubstantial number of CO2-inhibiting neurons, CO2 additionan completely block the action potential firing of the cells tohe odorants. These results indicate strong interaction be-ween CO2 signals and odorant signals in the olfactory bulb,uggesting important roles for the integration of these twoignals in CO2-mediated behavioral responses. © 2010 IBRO.ublished by Elsevier Ltd. All rights reserved.

ey words: mitral/tufted cells, pheromone, mixture, lateralnhibition, olfactory coding.

O2 is one of the major byproducts of cellular metabolism.lthough the basal CO2 level in the air is relatively low

0.038%), local atmospheric CO2 level can fluctuate dra-atically with the metabolic activity of organisms. AmbientO2 level can signal the presence of food, predators, ornvironmental stress and mediate stereotypical behaviorsor animals across phyla (Stange and Stowe, 1999; Suh etl., 2004; Thom et al., 2004; Jones et al., 2007; Hallem andternberg, 2008; Luo et al., 2009). Recent studies havehown that, for both vertebrates and invertebrates, CO2 isetected by a specialized subset of olfactory sensory neu-

Correspondence to: M. Luo, NIBS, #7 Science Park Road, Zhong-uancun Life Science Park, Beijing, 102206, PR China. Tel: �86-10-0726688-8320; fax: �86-10-80723342.-mail address: luominmin@nibs.ac.cn (M. Luo).bbreviations: CAII, carbonic anhydrase II; cGMP, guanosine 3=,5=-yclic monophosphate; CNG channel, cyclic nucleotide-gated chan-el; GC-D, guanylyl cyclase-D; MOB, main olfactory bulb; M/T cell,itral/tufted cell; OSN, olfactory sensory neuron; PDE2A, phosphodi-

tsterase 2A; PSTHs, peri-stimulus time histograms; 2,5-DMP, 2,5-imethyl pyrazine.

306-4522/10 $ - see front matter © 2010 IBRO. Published by Elsevier Ltd. All rightoi:10.1016/j.neuroscience.2010.08.006

881

ons (OSNs) (Suh et al., 2004; Thom et al., 2004; Luo etl., 2009). Although CO2 is odorless to humans, it is sen-itively detected by a specialized olfactory subsystem inice (Hu et al., 2007; Sun et al., 2009). The CO2-respon-

ive OSNs in mammals uniquely express several signalingolecules, such as guanylyl cyclase-D (GC-D) and phos-hodiesterase 2A (PDE2A) that suggest their use ofuanosine 3=,5=-cyclic monophosphate (cGMP) as theecond messenger (Fulle et al., 1995; Juilfs et al., 1997;un et al., 2009). In addition, they project to a set of theo-called necklace glomeruli at the caudal end of the mainlfactory bulb (MOB) (Juilfs et al., 1997; Hu et al., 2007;alz et al., 2007). Because GC-D� OSNs are not known

o express signaling components in the canonical trans-uction pathway within the typical OSNs, such as Golf andCIII (Juilfs et al., 1997; Meyer et al., 2000), it is likely thatC-D� neurons represent a specialized channel dedi-ated to detecting CO2 but not typical olfactory signals.

A key question that remains unresolved is whetherO2-activating bulbar neurons can also respond to otherdorants. CO2 activates bulbar neurons that extend theirendrites into the necklace glomeruli (Hu et al., 2007).owever, necklace glomeruli are intermingled with otherlomeruli in the main olfactory bulb, suggesting potential

nteraction between CO2 signals and olfactory signals inhe MOB. A recent tracing study suggests that at leastome necklace glomeruli receive inputs from GC-D�

SNs (Cockerham et al., 2009). Thus, the CO2-responsiveulbar neurons may also be excited by odorants that acti-ate these GC-D� OSNs. Although mitral/tufted (M/T)ells—the projection neurons of the bulb—extend theirrimary dendrite into only one glomerulus, they also inter-ct extensively with their basal dendrites (Shepherd et al.,004). Inhibitory lateral connections are believed to shapedorant responses (Mori et al., 1999; Tan et al., 2010).owever, it remains unclear whether CO2-activating cellsan be inhibited by odorants and conversely whether typ-

cal odorant-responsive M/T cells can be inhibited by CO2.n interesting feature of CO2 signaling lies in the fact thatO2 levels are correlated with the metabolic activity ofrganisms including plants and animals. It is conceivablehat CO2 cues are often simultaneously released with otherdorants in natural environment. It remains untested howO2-responsive cells in the bulb respond to the mixtures ofO2 and odorants.

In this study, we carried out in vivo physiological re-ordings from the mouse olfactory bulb to address theseuestions. We find that a majority of CO2-responsive neu-ons are selectively activated by other odorants. Many of

hese odorants are putative mouse pheromones and nat-s reserved.

FaosigRCTRswl

L. Gao et al. / Neuroscience 170 (2010) 881–892882

A B

CBenz-

aldehydecinnam-aldehydeα-iononebenzyl

etherstyrene(-)-α-pinenetoluenebenzene

allyl cinnamate

benzyl acetate

ethyl benzoate

ethyl butyrate

propyl butyrate

octyl acetate

hexyl acetate

butyl acetate

aceto-phenone

4-heptanone

2-octanone

cyclo-hexanone

2-sec-butylcyclo-hexanone

menthone(-) carvone(-) fenchone

3-decanone

acetyl butyryl

2-pentanone

trans-anetholephenetoleanisolecreosolguaiacol

2-phenyl-ethanol

( ) 1-phenyl-ethanol

(-)carveolbenzyl alcoholcontrol1-octanol1-hexanoln-butanol

octanalheptanalpentanalpropanaloctanoic acid

hexanoic acid

butyric acid

propionic acid

n-pentyl acetate

n-butyl-amine

3-heptanonefarneseneisovaleric

acidgeraniollinaloolcitral

C57BL/6 female urine

C57BL/6 male urine

2,5-dimethyl-pyrazine

2-heptanone

ethyl vanillin

2-butanone

ethyl acetate

ethyl propion-

ate

2,5-DMPCO2

1 m

VPutative

Pheromone

HO n

O

O n

HO n

Plant constituent

On n

n

O

On

O

Aromatic or Cyclic

CO2

fresh air

odorsfiltered air

soda limeCO2 Control

recording electrode

soda lime active charcoal

active charcoal

active charcoal

3-way valves (x4)

CO2

monitor

flowmeters

filtered air

CO2

vacuumrobotic

olfactometer

0

20

60

40

0 1 2 3 4 5 6

Nor

mal

ized

resp

onse

str

engt

h

-1.520 40 600

Odorant

-0.5

0

0.5

1

-1

0

20

60

40

0 1 2 3 4 5 6

Farnesene

-0.4

-0.2

Nor

mal

ized

resp

onse

str

engt

h

Odorant20 40 60

0

0.2

0.4

0.6

0.8

1

0

D E

0 1 2 3 4 5 60

20

60

40

Time (sec)

Res

pons

e st

reng

th

(spi

kes/

s)

Time (sec) Time (sec)

ig. 1. CO2-activating neurons in the olfactory bulb respond selectively to typical odorants. (A) A schematic diagram illustrates our method of odorantnd CO2 applications. Sixty-three odorants were stored in a panel containing 8�8 headspace vials and delivered with a custom-made roboticlfactometer. Purified room air from one headspace vial was used as the control for flowrate. CO2 was delivered separately by a dedicated line andolenoid valves. Purified room air with CO2 eliminated by soda lime was used as the control for CO2 pulses. (B) A diverse set of odorants was used

n this study. Pink indicates putative pheromone; green, plant leaf constitutive; red, acids; yellow, aldehyde; light green, alcohol; orange, ether; darkreen, ketone; blue, acetate; purple, aromatic ring; white, control. Some odorants have multiple functional groups, although only one is indicated. (C)epresentative physiological traces (upper) and PSTHs (lower) show the responses of a mitral cell to CO2 and odorants. This cell was excited by 1%O2 and 2,5-dimethyl pyrazine (2,5-DMP, 1% saturated vapor) and inhibited by farnesene (1%). Horizontal bars indicate 2-s CO2 or odorant pulse.races below the firing patterns indicate simultaneously recorded respiratory rhythms. (D) The olfactory tuning curve of the same cell as shown in (C).esponse intensities were normalized to the maximal response of the optimal odorant. Odorant identities were aligned to produce a smooth andymmetrical curve, with the maximal excitatory response in the middle. Dashed line, zero response. (E) Mean olfactory tuning curve of nine cells thatas completed with the testing of 63 odorants. Same conventions as in (D). Error bars, s.e.m. For interpretation of the references to color in this figure

egend, the reader is referred to the Web version of this article.

uraCb

A

AtdCoiu

E

At(dtnws(snCpSr

uaw

C

TbwoCsafiwHTa

6rh(uMrtn2tp(Fst

Trto

O

11223ABCCCEEEF*GHIL

*

L. Gao et al. / Neuroscience 170 (2010) 881–892 883

ral odorants. In addition, many odorant-responsive neu-ons are inhibited by CO2 and their odorant responses areffected by the addition of CO2. These results indicate thatO2 and odorant signals are integrated in the olfactoryulb with diverse patterns.

EXPERIMENTAL PROCEDURES

nimals and anesthesia

nimal care and use conformed to the institutional guidelines ofhe National Institute of Biological Science and were in accor-ance with the policies of the US National Institute of Health. Male57/BL6 mice (8–12 weeks) were obtained from VitalRiver Lab-ratory Animal Inc. (Beijing, China). For anesthesia, mice were

njected with atropine (50 mg/kg, i.p., Sigma-Aldrich) and thenrethane (1.64 g/kg, i.p., Sinopharm, Shanghai, China).

lectrophysiology

fter anesthesia, a mouse was placed on a custom-made stereo-axic apparatus (BrainKing Biotech, Beijing) and kept warm39 °C) with an electrical heating pad. The animal anesthesiaepth and health state were monitored by recording the respira-

ion rate with a strain gage affixed to chest skin. CO2-responsiveeurons were recorded from the caudal end of the olfactory bulb,here the necklace glomeruli are located (Hu et al., 2007). Boro-ilicate microelectrodes were pulled on a Sutter horizontal pullerP-97, Sutter Instrument Co., Novota, CA, USA) and filled with 1 Modium chloride (resistance 10�20 M�). Electrophysiological sig-als were amplified by AxoClamp 2B (Axon Instruments, Unionity, CA, USA), band-pass filtered at 0.5–3 kHz and further am-lified with a second amplifier (Brownlee 440, Brownlee Precision,an Jose, CA, USA), and then digitized at 6 kHz. Respiratory

hythm was recorded simultaneously for offline analysis. Single-

able 1. The identity of odorants that elicited either excitatory or inhibitoow consists of the identity of a specific odorant, the number of cells ahat inhibited the CO2-activating cells, the number of CO2-activating cdorant activated the CO2-inhibiting cells, those that inhibited the CO

dorant CO2-activating cells

Excitatory responseto odorant (numberand percentage)

Inhibitory responsto odorant (numband percentage)

-hexanol 5 (29.4%) 1 (5.9%)-octanol 3 (17.6%) 3 (17.6%),5-DMP 4 (21.1%) 2 (10.5%)-heptanone 5 (26.3%) 1 (5.3%)-heptanone 4 (22.2%) 0 (0%)nisole 3 (21.4%) 3 (21.4%)utyric acid 3 (16.7%) 2 (11.1%)itral 7 (35%) 0 (0%)innamaldehyde 7 (58.3%) 1 (8.3%)reosol 4 (26.7%) 4 (26.7%)thyl acetate 4 (20%) 4 (20%)thyl propionate 10 (40%) 2 (8%)thyl vanillin 5 (26.3%) 2 (10.5%)arnesene 4 (22.2%) 3 (16.7%)Female mouse urine 5 (26.3%) 2 (10.5%)eraniol 6 (30%) 2 (10%)exanoic acid 6 (33.3%) 2 (11.1%)

sovaleric acid 5 (27.8%) 3 (16.7%)inalool 9 (42.9%) 0 (0%)

Female urine was obtained from adult C57BL/6 female mice.

nit recordings were confirmed by the inter-spike interval histogramsnd autocorrelograms. Data acquisition and experimental controlere implemented by custom-written Visual Basic programs.

O2 and odorant delivery

he methods for delivering CO2 and odorant pulses are illustratedy a schematic diagram in Fig. 1A. CO2-free clean air was filteredith soda lime and active charcoal and then presented continu-usly before the mouse nostrils at a constant flow rate of 4 L/min.O2 was delivered into the constant air flow with a channel-pecific solenoid valve controlled by Master-8 (A.M.P.I., Israel)nd a custom-made electronic circuit. CO2 was diluted with puri-ed room-air, and final CO2 concentrations at the mouse nostrilsere monitored with two infrared-based analyzers (GXH-3010D,uayun Instruments, Beijing and CO2WFA, CardioPulmonaryechnologies). CO2-free controls were provided by filtering roomir through soda lime and then active charcoal.

Odorants were stored in headspace vials and delivered with a4-channel robotic olfactometer equipped with a liquid dispensingobot (model 7200, I & J Fisnar, Fair Lawn, NJ, USA). This roboteld an interchangeable 8�8 rack containing 64 sampler vialsFig. 1B; 63 odorants plus one blank control). The odorant stim-lus set consisted of 61 pure odorants (Sigma-Aldrich, St. Louis,O, USA) and two samples of mouse urine. To ensure the accu-

acy of the timing of odorant delivery, odorants were diverted intohe vacuum for the first 1.5 s and then released to the mouseostrils by switching open a three-way solenoid valve for 2 or.5 s. An additional air pulse (100 ml/min) was delivered into thehree-way solenoid valve for 2.5 s after the termination of odorantulses to eliminate the odorant residual from the previous trialFig. 1A). Teflon tubing was used to reduce cross contamination.or mixture delivery, CO2 and odorant pulses were deliveredimultaneously through independent Teflon tubing. Intervals be-ween trials of stimulus delivery were at least 15 s to reduce

ses from either CO2-activating or CO2-inhibiting bulbar neurons. Eachod (%) in which the odorant activated the CO2-activating cells, those

d for this odorant, the number of cells and likelihood (%) in which theg cells, and the number of CO2-inhibiting cells tested for this odorant

CO2-inhibiting cells

mbercellsted

Excitatory responseto odorant (numberand percentage)

Inhibitory responseto odorant (numberand percentage)

Numberof cellstested

2 (22.2%) 2 (22.2%) 93 (33.3%) 2 (22.2%) 94 (28.6%) 2 (14.3%) 144 (36.4%) 2 (18.2%) 115 (41.7%) 1 (8.3%) 123 (33.3%) 3 (33.3%) 94 (40%) 3 (30%) 104 (36.4%) 2 (18.2%) 113 (42.9%) 4 (57.1%) 71 (11.1%) 3 (33.3%) 92 (16.7%) 1 (8.3%) 123 (23.1%) 5 (38.5%) 133 (27.3%) 2 (18.2%) 112 (20%) 1 (10%) 104 (36.4%) 3 (27.3%) 114 (33.3%) 3 (25%) 122 (20%) 3 (30%) 102 (20%) 2 (20%) 105 (45.5%) 1 (9.1%) 11

ry responnd likelihoells teste

2-inhibitin

eer

Nuoftes

17171919181418201215202519181920181821

ph

D

DMcqcoaoseilisHPmaretr

fn

ftTawwwpnsnnoossw

Cs

Ta

Ficcrr

L. Gao et al. / Neuroscience 170 (2010) 881–892884

otential response habituation (Tan et al., 2010). No obviousabituation was observed for any odorant tested.

ata analysis

ata analysis was carried out with custom-written programs inatlab (Mathworks). Peri-stimulus time histograms (PSTHs) were

omputed by counting the number of action potentials within se-uential time bins (200 ms). The response strength was firstalculated by subtracting the spontaneous firing rate before thedorant application from the firing rate during the time when wepplied odorant stimuli within each trial. Because control stimuliccasionally evoked weak responses from some cells, we furtherubtracted the control response strength from the responsesvoked by stimuli. Thus, positive values of response strength

ndicate excitation whereas negative values indicate inhibition. Fol-owing the conventional approaches of recordings from free-breath-ng mice in the field, in this study we did not synchronize the onset oftimulus delivery to a specific phase of the animal respiratory cycle.owever, three to eight trials were applied for each stimulus and theSTHs were further averaged across trials. This averaging mini-ized the variability resulted from difference in stimulus onset timingmong trials. The respiratory rates of different animal individualsanged from 2.5 to 5 cycles/s. However, the respiratory frequency ofach animal remained stable during the recording session. Becausehe responses evoked by a stimulus was compared to the controlesponse in the same recording session, the variability in respiratory

A

B

1% CO2

1 m

VR

espo

nse

str

engt

h (s

pike

s/s)

0

Time (sec)1 52 3 4

20

40

0

20

40

0

20

40

0

0.1% CO2

0.5% CO2

1% CO2

ig. 2. CO2 strongly inhibits odorant-responsive bulbar cells in a dosenhibited by a 2-s 1% CO2 pulse (bar). Bottom traces indicate respioncentrations of 0.1, 0.5 and 1% (horizontal bars). (C) Plot of theoncentrations. (D) Concentration curves of 15 different CO2-inhibiting

eduction of firing rates to the spontaneous firing rates of the cell. Thick red curveferences to color in this figure legend, the reader is referred to the Web vers

requency among animals does not affect our conclusion on theeuronal responsiveness to a given stimulus.

Dose-response curves were fitted to a sigmoid function:�x��1 ⁄ �l�exp��a��x�b���, in which x represents CO2 concen-rations, a, slope, and b, half maximal inhibitory response (EC50).o plot the odorant tuning curves, we sorted the response value ofll test odorants so that the most prominent excitatory responsesere in the middle and the most prominent inhibitory responsesere at both ends. The response selectivity to a set of odorantsas quantified by calculating the lifetime sparseness using non-arametric statistics (Vinje and Gallant, 2000). Lifetime sparse-ess S��1���ri ⁄ N�2 ⁄ ��ri

2 ⁄ N�� ⁄ �1��1 ⁄N��, where ri is the re-ponse strength to odor i and N is the total number of the stimuli. Anyegative value of ri was set to zero before computing lifetime sparse-ess. S measures the peakedness of a response distribution. Valuesf S near 0 indicate a flat distribution of the responses to variousdorants, and values near one indicate a sparse distribution of re-ponses to the stimulus set. Student T-tests were performed to testignificant difference between responses evoked by different stimuli,ith P�0.05 considered statistically significant.

RESULTSO2-activating neurons in the olfactory bulb respondelectively to odorants

o test whether bulbar neurons can respond to both CO2

nd odorants, we first searched for CO2-activating neurons

C

D

0 0.2 0.4 0.6 0.8 1-16

-12

-8

-4

0

CO2 concentration (%)

Res

pons

est

reng

th (s

pike

s/s)

Nor

mal

ized

resp

onse

str

engt

h

0 0.5 1 1.5

CO2 concentration (%)

-1.0

-0.8

-0.6

-0.4

-0.2

0

0.2

ent manner. (A) Physiological trace shows that a cell was profoundlythms. (B) PSTHs of the same cell as shown in (A) for CO2 at the

ry response strengths of the cell shown in (A) to CO2 at different(black lines). Inhibitory strengths were calculated by normalizing the

6

-dependratory rhy

inhibitoneurons

e represents fitting with a sigmoidal function. For interpretation of theion of this article.

bo(ifrpnuas1carom

mFasaFhs

soTc1lmtmw

sAw1��v

rmiaweWretapeat

C

Mitt2Ogegoe

Fcsc

L. Gao et al. / Neuroscience 170 (2010) 881–892 885

y targeting recording electrodes into the caudal end of thelfactory bulb where the necklace glomeruli are locatedHu et al., 2007; Walz et al., 2007). Putative M/T cells weredentified based on the fact that these cells were recordedrom a thin layer with clear spontaneous activity coupled toespiratory rhythms. CO2-responsiveness was tested byuffing 2-s CO2 pulses (0.5–1%), which usually saturateeuronal responses (Hu et al., 2007). About 10% of single-nits in this region were activated by CO2. After a CO2-ctivating M/T cell was identified, we examined its re-ponses to odorants by using a robotic olfactometer (Fig.A). The odorant panel consisted of 63 different odorants,overing a broad spectrum of functional groups includingldehydes, acid, acetate, alcohol, ketone and aromaticings (Fig. 1B). Some odorants were putative mouse pher-mones, including complex stimuli such as male and fe-ale mouse urines.

Out of a total of 33 CO2-activating cells tested, a vastajority (n�30 cells) responded to at least one odorant.or example, the cell shown in Fig. 1C was vigorouslyctivated by CO2 (Fig. 1C left-hand panel; responsetrength�13.8�1.2 spikes/s, mean�SEM). This cell waslso strongly activated by 2,5-dimethyl pyrazine (2,5-DMP;ig. 1C middle panel; 28.1�3.2 spikes/s) and clearly in-ibited by farnesene (Fig. 1C right-hand panel; �6.1�2.0pikes/s).

CO2-activating M/T neurons responded to odorantselectively. Typically, a cell was excited and inhibited bynly a few odorants out of a total of 63 odorants (Fig. 1D).his selective tuning was evident from the mean tuningurve for all cells that were tested with 63 odorants (Fig.E, n�9 cells). We quantified the olfactory response se-

ectivity with lifetime sparseness, which has been used toeasure the response selectivity for several sensory sys-

ems (Vinje and Gallant, 2000; Wilson et al., 2004). Theean lifetime sparseness for these CO2-activating cellsas 0.65�0.08, which was comparable to that of randomly

ig. 3. Bulbar neurons that are inhibited by CO2 can be excited and inhell (0.65% CO2, top panel) was excited by 1% 3-heptanone (middle p

how the odorant responses of another CO2-inhibiting cell. This cell was exciteurve of five of CO2-inhibiting neurons, illustrating selective responses of these

ampled M/T cells in the dorsal MOB (Tan et al., 2010).fter we pooled data from 33 cells (each cell was testedith at least 20 odorants, including the nine cells in Fig.E), we collected a total of �1000 cell-odorant pairs. Only20% of responses were significantly excitatory and10% were inhibitory. This again indicated that CO2-acti-

ating neurons respond selectively to odorants.Interestingly, many of the odorant stimuli that evoked

esponses from CO2-activating M/T cells were putativeouse pheromones or chemical compounds richly present

n plants. For example, the cell shown in Fig. 1C–D wasctivated by 2-heptanone, 2,5-DMP and mouse urine. Itas inhibited by farnesene. These odorants were consid-red to be mouse pheromones (Dulac and Torello, 2003).e identified 19 odorants that most frequently elicited

esponses from CO2-responsive neurons (Table 1). Sev-ral putative mouse pheromones had a likelihood of 20%o excite the CO2-activating neurons. Our odorant panellso contained several chemicals that are richly present inlants, including linalool, citral, geraniol, ethyl acetate,thyl vanillin, ethyl propionate and cinnamaldehyde. CO2-ctivating neurons could also be excited or inhibited byhese odorants with high likelihood (Table 1).

O2 inhibits odorant-responsive neurons in the MOB

/T cells in the MOB are far outnumbered by GABAergicnterneurons, which are believed to enable a given M/T cello modulate the olfactory responses of other M/T cellshrough lateral inhibition (Mori et al., 1999; Shepherd et al.,004; Tan et al., 2010). Because CO2 activates GC-D�

SNs and bulbar neurons associated with the necklacelomeruli, we asked whether CO2 signaling is limited toxcite bulbar neurons directly associated with the necklacelomeruli or it can inhibit other M/T cells as well. Webserved clear inhibition of many bulbar neurons by CO2,specially in the caudal part of the bulb (n�49 cells).

different odorants. (A) Physiological traces show that a CO2-inhibitinginhibited by 1% amyl acetate (bottom panel). (B) Physiological traces

ibited byanel) and

d by 1% citral and inhibited by 1% toluene. (C) Mean olfactory tuningcells.

Fdssbobw

L. Gao et al. / Neuroscience 170 (2010) 881–892886

Mixture response (spikes/s)

E

CO2 maleurine

mixture

5

10

15

25R

espo

nse

stre

ngth

(spi

kes/

s)

0

20

male urine mixtureCO2

1 m

V

0 1 2 3 4 5 6

20

40

60

0

Time (sec)0 1 2 3 4 5 6

20

40

60

0

Time (sec)0 1 2 3 4 5 6

20

40

60

0

Time (sec)

Res

pons

e st

reng

th(s

pike

s/s)

0

4

8

12

16

Res

pons

e st

reng

th (s

pike

s/s)

CO2 linalool mixture

linalool mixtureCO2

1 m

V

10

20

30

00 1 2 3 4 5 6

Time (sec)0 1 2 3 4 5 6

10

20

30

0

Time (sec)0 1 2 3 4 5 6

10

20

30

0

Time (sec)

Res

pons

e st

reng

th(s

pike

s/s)

F

G H

B

-2 -1 0 1 2 3 4 5 60

1

2

3

4

5

6

Mixture/CO2

Mix

ture

/odo

rant

D

Res

pons

e st

reng

th (s

pike

s/s)

10

20

30

0

-10CO2 odor mixture

-10 0 10 20 30 40 50 600

10

20

30

40

50

60

Sum

mat

ion

of C

O2

and

odor

ant r

espo

nse

(spi

kes/

s)A

C

Res

pons

e st

reng

th (s

pike

s/s)

10

20

30

40

50

60

-10CO2 odor mixture

0

ig. 4. For some bulbar neurons that are excited by both CO2 and odorants, their responses by the mixture of CO2 and odorants are significantlyifferent from those evoked by CO2 alone. (A) Plot of the linear summation of CO2 and odorant response strengths against the recorded responsetrengths to binary mixtures of CO2 and odorants. All tests were performed on CO2-activating cells. Red dots indicate mixture responses that wereignificantly different from those by either CO2 or odorant alone; green, mixture responses similar to those by odorants but different from those by CO2;lue, mixture responses different from those by odorants but similar to those by CO2; black, mixture responses similar to the responses evoked bydorants and CO2. (B) Plot of the mixture/odorant response ratio against mixture/CO2 response ratio. Dashed lines indicate values of 1 (domination

y either CO2 or the odorant in the mixture response). Colors correspond to dots of the same colors in (A). (C) Plot showing that responses to mixturesere significantly different from the responses evoked by CO2 and odorants separately. Red lines correspond to red dots in (A). (D) Plot showing that

Snif

CahtC1crcobin

viotsi(6otiepa�msoic

Po

StiA2cbc

HsiP(eImCe

fsltTmrumca

rCaialsd

C

r(um(((r

Tattaot

D

SSS

L. Gao et al. / Neuroscience 170 (2010) 881–892 887

imilar to the inhibition evoked by odorants on bulbareurons, the inhibitory response to CO2 became apparent

mmediately after CO2 pulse onset and the cell reboundedollowing the termination of the pulse (Fig. 2A).

The strength of the CO2-evoked inhibition depends onO2 concentration. Typically, the inhibition was moderatet the concentration of 0.1% CO2, became more potent atigher concentrations, and saturated by CO2 at concen-rations�1% (Fig. 2B, C). The concentration curves of 15O2-inhibiting neurons are shown in Fig. 2D. Fitting of the5 curves to a sigmoidal function indicates that the con-entration for EC50 is about 0.4% CO2 and that for satu-ated inhibition is about 1.5% CO2 (Fig. 2D, thick redolor). These values are higher than the respective valuesf excitatory responses evoked by CO2 in the olfactoryulb (Hu et al., 2007), possibly because the inhibition was

ndirectly produced by lateral inhibition from CO2-activatingeurons associated with the necklace glomeruli.

The CO2-inhibiting neurons respond selectively to a di-erse set of odorants. For example, one cell that was potentlynhibited by CO2 (strength��9.0�0.5 spikes/s) was vigor-usly activated by 3-heptanone and clearly inhibited by men-hone (Fig. 3A; response strength to 3-heptanone�54.9�1.7pikes/s; to menthone��17.5�1.5 spikes/s). Another CO2-nhibiting cell was excited by citral and inhibited by tolueneFig. 3B). For five CO2-inhibiting cells that were tested with3 odorants, their lifetime sparseness value of excitatorydorant responses was 0.68�0.10. In addition, the meanuning curve of these cells resembled that of CO2-activat-ng cells (Fig. 3C), suggesting similar olfactory tuning prop-rties. As the same with CO2-activating neurons, weooled all the cells that were tested with at least 20 odor-nts and collected a total of �600 cell-odorant pairs. Only20% of them were significantly excitatory and approxi-ately equal number of the responses was inhibitory, further

uggesting that CO2-inhibiting neurons were also selective todorants. Interestingly, a substantial number of CO2-inhibit-

ng cells were excited by plant-derived compounds, such asitral, cinnamaldehyde and linalool (Table 1).

atterns of responses to the mixtures of CO2 anddorants by CO2-activating neurons

ince CO2-activating neurons respond to odorants, wehen examined whether CO2 and olfactory signals can bentegrated by testing neuronal responses to their mixtures.

total of 13 CO2-activating neurons were presented with1 different binary mixtures consisting of CO2 and a spe-ific odorant. Earlier studies have shown that responses toinary odorant mixtures are typically dominated by oneomponent (Giraudet et al., 2002; Tabor et al., 2004).

esponses to mixtures were significantly different from those to CO2 alupper panel) and PSTHs (lower panel) show the excitatory responserine at the same concentrations. (F) Bar plot shows that the responsouse urine for the cell shown in (C) (** P�0.01; t-test for CO2 vs. mix

lower panel) show the excitatory responses of a cell to CO2 (0.9%), liH) Bar plot shows that the response intensity to the mixture was s

** P�0.01; t-test for CO2 vs. mixture. P�0.84 for linalool vs. mixture). For intereferred to the Web version of this article.

owever, the overall strength of mixture-evoked re-ponses was close to the linear sum of the responses to

ndividual components (Fig. 4A; slope�0.81, P�0.05;earson’s correlation). In a substantial number of tests

n�6/21 tests), mixture responses were significantly differ-nt from those by CO2 or odorants alone (Fig. 4B; Table 2).n the rest of tests (n�15/21 tests), the responses toixtures resembled the responses elicited by odorants orO2 (Fig. 4B; Table 2), suggesting dominancy by onelement within the binary mixtures.

When mixtures produced responses that were differentrom those evoked by both CO2 and odorants, the re-ponse strengths to mixtures were more often significantly

arger than the responses of either component (n�4 out 6ests; Fig. 4C, E). One example is illustrated in Fig. 4E, F.his cell was activated by CO2 and male mouse urine. Theixture of CO2 and male mouse urine evoked excitatory

esponse that was significantly stronger than either CO2 orrine at the same concentrations. For the other two tests,ixtures resulted in response intensities that were signifi-

antly lower than those evoked by either CO2 or the odor-nt alone (Fig. 4B, C).

In some cases, mixtures of CO2 and an odorant elicitedesponses that were significantly different from those byO2 but on the other hand similar to those by odorantslone (n�5/21 tests; green in Fig. 4A, B, D). One example

s shown in Fig. 4G, H. This cell was mildly excited by CO2

nd vigorously activated by linalool. Mixture of CO2 andinalool produced a strong excitatory response with inten-ity significantly larger than that by CO2 alone while notifferent from that by linalool (Fig. 4G, H).

In another five tests, neuronal responses to the mixture ofO2 and an odorant largely followed the pattern and strength

imilar to those by odorants (green dots in A). (E) Physiological tracesl to CO2 (1%), male mouse urine (0.6%), and the mixture of CO2 andy to the mixture was significantly higher than those to CO2 and maleurine vs. mixture). (G) Physiological traces (upper panel) and PSTHs9%), and the mixture of CO2 and linalool at the same concentrations.y higher than that to CO2 and but not different from that to linalool

able 2. The comparison between responses evoked by mixtures of CO2

nd an odorant and those by their components. We classified the rela-ionship to four patterns: (1) mixtures evoked responses different fromhose by either CO2 or the odorant; (2) similar to those by the odorantlone; (3) similar to those by CO2 alone; (4) similar to both of CO2 and thedorant. Each row shows the number of tests for each pattern out of 21ests for CO2-activating cells and 20 tests for CO2-inhibiting cells

CO2-activatingcells (N�21 tests)

CO2-inhibitingcells (N�20 tests)

ifferent from both CO2

and the odorant6 1

imilar to the odorant 5 10imilar to CO2 5 7imilar to both CO2 andodorant

5 2

one but ss of a cele intensitture and

nalool (1.ignificantl

pretation of the references to color in this figure legend, the reader is

oeblrCm

wn

Po

B

Fsrc(htaat

L. Gao et al. / Neuroscience 170 (2010) 881–892888

f responses evoked by CO2 alone (Fig. 5A). One suchxample is shown in Fig. 5C, D. This cell was strongly excitedy CO2 and linalool. Its response to the mixture of CO2 and

inalool was significantly higher than that by linalool alone butesembled that elicited by CO2, suggesting domination byO2 stimulus. In the rest of cases (n�5/21 tests), binary

C

A

0 1 2 3 4 5 6

10

40

60

0

20

30

50

citralCO2

1 m

V

Time (sec)0 1 2 3 4 5 6

10

40

60

0

20

30

50

Time (sec)

Res

pons

e st

reng

th(s

pike

s/s)

0 1 2 3 4 5 6

15

30

45

0

Time (sec)

CO2 linalool

1 m

V

0 1 2 3 4 5 6

15

30

45

0

Time (sec)

Res

pons

e st

reng

th(s

pike

s/s)

E

35

0

5

10

15

20

25

30R

espo

nse

stre

ngth

(spi

kes/

s)

CO2 Odor Mixture

ig. 5. For some bulbar neurons that are excited by both CO2 and odohowing that responses to mixtures were significantly different from thesponses to mixtures were not significantly different from those eitherolors in (Fig. 4A). (C) Physiological traces (upper panel) and PSTHs0.9%), and the mixture of CO2 and linalool at the same concentrationsigher than that to linalool but not different from that that to CO2 (** P�0races (upper panel) and PSTHs (lower panel) show the excitatory ret the same concentrations. (F) Bar plot shows that the response intenspplied separately (P�0.19 for CO2 vs. mixture; P�0.84 for citral vs. m

he reader is referred to the Web version of this article.

ixtures of CO2 and a given odorant elicited responses that e

ere not significantly different from either of the subcompo-ents (Fig. 5B). One such example is shown in Fig. 5E, F.

atterns of responses to the mixtures of CO2 anddorants by CO2-inhibiting neurons

ecause CO can inhibit bulbar neurons (Figs. 2 and 3), we

CO2 citral mixture0

10

20

30

40

Res

pons

e st

reng

th (s

pike

s/s)

2 3 4 5 6

mixture

Time (sec)

0

5

10

15

20

25

30

Res

pons

e st

reng

th (s

pike

s/s)

CO2 linalool mixture

mixture

2 3 4 5 6Time (sec)

D

F

35

0

5

10

15

20

25

30

Res

pons

e st

reng

th (s

pike

s/s)

CO2 Odor Mixture

eir responses to their mixture largely follow those to odorants. (A) Plotdorants alone but similar to those by CO2. (B) Plot showing that ther odorants. In both (A) and (B), colors correspond to dots of the sameanel) show the excitatory responses of a cell to CO2 (0.5%), linaloolplot shows that the response intensity to the mixture was significantly

t for linalool vs. mixture; P�0.96 for CO2 vs. mixture). (E) Physiologicalof a cell to CO2 (0.6%), citral (1%), and the mixture of CO2 and citralmixture was not significantly different from those to CO2 or citral whenests). For interpretation of the references to color in this figure legend,

0 1

10

40

60

0

20

30

50

0 1

15

30

45

0

B

rants, those by oby CO2 o(lower p

. (D) Bar.01; t-tessponsesity to theixture; t-t

2

xamined whether it could modulate odorant responses of

Cwoilolm(

s6Leapcp

t

tFw2rpts

IpbfrtHf

FlwabrcsPtsc

L. Gao et al. / Neuroscience 170 (2010) 881–892 889

O2-inhibiting neurons. A total of 13 CO2-inhibiting neuronsere tested with 20 groups of stimuli consisting of CO2, andorant, and their binary mixtures. Although mixtures elic-

ted responses with intensities largely predicted by theinear summation of the individual responses to CO2 anddorants (Fig. 6A; slope�0.77, P�0.05; Pearson’s corre-

ation), in a majority of tests the responses to the binaryixtures followed the responses to either CO2 or odorants

Fig. 6B; Table 2).Strikingly, CO2 completely blocked the excitatory re-

ponses of odorants in a substantial number of tests (Fig.C; n�7/20 tests). Fig. 6D shows one such example.ikely because of a low spontaneous firing rate, this cellxhibited mild reduction of action potential firing from basalctivity in response to a CO2 pulse and a more obviousost-inhibitory rebound. Linalool vigorously activated theell. However, the excitatory response to linalool was com-letely blocked after the addition of CO2 (Fig. 6D, E).

More frequently, the responses of CO2-inhibiting neuronso the mixtures of CO2 and a given odorant largely followed

D

B

-5

-15 -10 -5 0 5 10 15 20Mixture response (spikes/s)

0

-10

5

-15

10

15

20

Sum

mat

ion

of C

O2

and

odor

ant r

espo

nse

(spi

kes/

s)

A

linalool

1 m

V

CO2

0 1 2 3 4 5 60

10

20

30

Time (sec)0 1 2 3 4 5 6

0

10

20

30

Time (sec)

1

2

3

Res

pons

e st

reng

th(s

pike

s/s)

-8 -6 -4

M

-4

-3

-2

-1

0

1

2

Mix

ture

/odo

rant

ig. 6. For some bulbar neurons that are inhibited by CO2, CO2 additinear summation of CO2 and odorant response strengths against theere performed on CO2-inhibiting cells. Red dots indicate the mixturelone; green, mixture responses similar to those by odorants but differeut similar to those by CO2; black, mixture responses similar to both oatio against mixture/CO2 response ratio. Dashed lines indicate valueorrespond to dots of the same colors in (A). (C) Plot showing that resimilar to those evoked by CO2 alone. Green lines correspond to greSTHs (lower panel) show the inhibitory response of a cell to CO2 (0

he mixture of CO2 and linalool at the same concentrations. (E) Bar

ignificantly lower than that to linalool (** P�0.01; t-test for linalool vs. mixtureolor in this figure legend, the reader is referred to the Web version of this arti

he responses evoked by the odorant alone (n�10/20 tests;ig. 7A). Fig. 7B, C illustrates one such example. This cellas inhibited by CO2 and activated by 2,5-DMP. Resembling,5-DMP, mixture of CO2 and 2,5-DMP evoked excitatoryesponses. Although there was a trend in change of responseattern and a reduction of response strength of the mixture,

he response difference of mixture and 2,5-DMP was nottatistically significant (Fig. 7C).

DISCUSSION

n mice, CO2 activates a specialized subset of OSNs thatroject to the necklace glomeruli in the caudal olfactoryulb (Hu et al., 2007; Sun et al., 2009). Our recordingsrom this study revealed that CO2-activating bulbar neu-ons also respond to odorants, suggesting strong interac-ion between CO2 and odorant signals in the olfactory bulb.ere we will discuss the implications of our data on the

unction of CO2 as a behavioral cue as well as the possible

0

4

-4

8

12

16

CO2 linalool mixtureRes

pons

e st

reng

th (s

pike

s/s)

mixture

mixture

3 4 5 6

me (sec)

E

2 4

2

C

-10

-5

0

5

10

15

Res

pons

e st

reng

th (s

pike

s/s)

CO2 odor

letely abolishes their excitatory responses to odorants. (A) Plot of theresponse strengths to binary mixtures of CO2 and odorants. All testss that were significantly different from those by either CO2 or odorantose by CO2; blue, mixture responses different from those by odorants

voked by odorants and CO2. (B) Plot of the mixture/odorant responseomination by either CO2 or odorant in the mixture response). Colors

mixture were significantly different from those by odorants alone butf the same colors in (A). (D) Physiological traces (upper panel) andng excitatory response to linalool (2.5%), and inhibitory response tos that the response intensity to the mixture was similar to CO2 but

0 1 20

0

0

0

Ti

-2 0

ixture/CO

ion comprecordedresponsent from thf those es of 1 (dponses toen dots o.9%), stroplot show

; P�0.21 for CO2 vs. mixture). For interpretation of the references tocle.

ca

F

Sc(aodOnoafstaTwatn

b

rCtpscobost

taamapcslafe

Fr6trPa

L. Gao et al. / Neuroscience 170 (2010) 881–892890

ircuitry mechanisms underlying the integration of CO2

nd odorant signals.

unctional implications of our results

tudies from several groups have revealed multiple spe-ialized subsystems within the main olfactory systemJuilfs et al., 1997; Ma et al., 2003; Luo, 2008; Fleischer etl., 2009). For example, the GC-D� OSNs within the mainlfactory epithelium project to the necklace glomeruli andetect CO2 (Hu et al., 2007; Sun et al., 2009). In addition,SNs within the Grüneberg ganglion at the rostral tip of theose project to a distinct subset of glomeruli in the caudallfactory bulb and detect alarm pheromones (Brechbuhl etl., 2008). Consistent with the view of specialized channelsor detecting CO2, in insects CO2 is also detected by apecialized set of olfactory receptor neurons, which projecto a single glomerulus in the antennal lobe, the anatomicalnalog of the vertebrate olfactory bulb (Suh et al., 2004;hom et al., 2004; Jones et al., 2007). It remains unclearhether the signals detected by these specialized OSNsre processed by dedicated and independent channels orhey interact with other olfactory signals in the centralervous system.

By single-unit recordings from the mouse olfactory

B

20

-5

0

5

10

15

Res

pons

e st

reng

th (s

pike

s/s)

CO2 odor mixtu

A

2,5-DMPCO2

1 m

V

0

5

10

15

20

0 1 2 3 4 5 6

Time (sec)

0

5

10

15

20

0 1 2 3 4 5 6

Time (sec)

Res

pons

e st

reng

th(s

pike

s/s)

ig. 7. In many cases when cells are inhibited by CO2, CO2 addition pesponses to mixtures were significantly different from those by CO2 aA). (B) Physiological traces (upper panels) and PSTHs (lower panelso 2,5-DMP (1.5%), and similar excitatory responses to the mixture oesponse intensity to the mixture was significantly higher than that to C�0.61 for 2,5-DMP vs. mixture). For interpretation of the referencesrticle.

ulb, we find that a vast majority of CO2-activating neu- o

ons are activated or inhibited by odorants. Interestingly,O2-activating neurons respond to a rather limited spec-

rum of odorant stimuli, which include putative mouseheromones and odorant compounds richly present inome plants. In addition, neurons in the olfactory bulban be inhibited by CO2. Finally, mixtures of CO2 anddorants produce a diverse response patterns that cane dramatically distinct from those evoked by CO2 ordorants alone. These results strongly suggest that CO2

ignals are integrated with odorant signals at the level ofhe olfactory bulb.

In natural environments, ambient CO2 levels can fluc-uate for a variety of reasons, such as the respiratoryctivity of other conspecific individuals or predators. Inddition, local CO2 levels can change because of theetabolic activity of plants. Since CO2-activating neuronslso respond to pheromones and chemicals derived fromlants, it is thus possible that these odorant cues may beombined with CO2 signals in the olfactory bulb to triggerpecific behavioral changes that are mediated by CO2. In

aboratory tests, mice avoid CO2 at concentrations as lows 0.2% (Hu et al., 2007). Future studies on the behavioralunction of CO2 signals in natural environment will helplucidate the exact behavioral significance of CO2 and

mixture

-2

4

6

8

10

Res

pons

eSt

reng

th (s

pike

s/s)

2

0

-4

-6

-8

CO2 2,5-DMP

mixture

2 3 4 5 6

ime (sec)

C

o clear effect on their responses to odorants. (A) Line plots show thatsimilar to those by odorants. Lines correspond to green dots in (Fig.

he inhibitory responses of a cell to CO2 (1.2%), excitatory responsesd 2,5-DMP at the same concentrations. (C) Bar plot shows that theilar to that evoked by 2,5-DMP (** P�0.01; t-test for CO2 vs. mixture;

in this figure legend, the reader is referred to the Web version of this

re

0 10

5

10

15

20

T

roduces nlone but) show tf CO2 anO2 but simto color

dorant signal integration.

rmcmtpoTo2siisiacimo

CC

Tpwrctab23su(t1kpcde

ad2CpnonGCmuai

sbiGitilbee

oisuiplnaiotibaiCC

oebCroaaaoocbC

At(C

B

C

L. Gao et al. / Neuroscience 170 (2010) 881–892 891

Consistent with earlier bulbar recordings on mixtureesponses, we observed several cases of non-linear sum-ation of CO2 and odorant responses in CO2-activating

ells. For a majority of tests, however, the responses toixtures are often dominated by one element in the mix-

ure, either CO2 or odorant. The results are consistent withrevious physiological recordings of bulbar neurons todorant mixtures (Giraudet et al., 2002; Tabor et al., 2004).hey are also in line with behavioral studies showing thevershadowing effect in mixture perceptions (Kay et al.,005; Riffell et al., 2009). Nevertheless, synergistic oruppressive interactions between CO2 and odorants occurn some mixture tests (Fig. 4A–F; Table 2). When addednto a mixture, an odorant can modulate the CO2 re-ponses of CO2-activating neurons (Fig. 4G, H), suggest-ng a modulatory effect on CO2 responses by odorants. Inddition, CO2 exerts profound inhibitory effects on manyells and can block their excitatory responses to odorantsn the mixture (Fig. 6D, E). Thus, the presence of CO2 may

ask the physiological and behavioral responses to somedorants.

ircuit mechanisms underlying the integration ofO2 and odorant signals

he OSNs that detect CO2 express several signaling com-onents that are not found in the canonical signaling path-ay in the typical olfactory sensory neurons. These neu-

ons selectively express carbonic anhydrase II (CA II) thatatalyzes the conversion of CO2 to bicarbonates and pro-ons. They also express GC-D (Fulle et al., 1995; Juilfs etl., 1997; Hu et al., 2007), which is stimulated by bicar-onate ions to produce cGMP (Sun et al., 2009; Guo et al.,009). Unlike typical olfactory neurons that use adenosine=,5=-cyclic monophosphate (cAMP) as the second mes-enger, these CO2-responsive OSNs express cGMP-stim-lated PDE2A and cGMP-sensitive cyclic nucleotide-gatedCNG) channels, further suggesting their use of cGMPs ashe second messenger (Fulle et al., 1995; Juilfs et al.,997; Meyer et al., 2000). Finally, these neurons are notnown to express typical odorant receptors that are Grotein-coupled receptors or their downstream signalingomponents such as Golf. It thus appears that CO2 isetected by a dedicated subset of OSNs in the olfactorypithelium.

How can the CO2-responsive bulbar neurons also bectivated by odorants? Because CAII-mutant mice cannotetect CO2 at behaviorally-relevant concentrations (Hu et al.,007), GC-D� neurons likely represent the sole source ofO2-evoked excitatory signals into the olfactory bulb. Oneossible mechanism of odorant-evoked responses of CO2

eurons in the bulb is that GC-D� OSNs are activated bydorants through a non-conventional signaling mecha-ism. However, imaging and recordings indicate thatC-D� OSNs are not stimulated by odorants that activateO2-activating bulbar neurons, such as 2,5-DMP andouse urine (Leinders-Zufall et al., 2007; Han and Luo,npublished observations). Thus, it is more likely that odor-nt-responsiveness of the CO -activating bulbar neurons

2

s imparted by the intrabulbar circuits. Recently, it has been

hown that necklace glomeruli receive synaptic input fromoth GC-D� and GC-D- OSNs (Cockerham et al., 2009). It

s thus plausible that odorants may activate these non-C-D OSNs that project to the necklace glomeruli, which

n turn excite the CO2-activating M/T cells associated withhese glomeruli. Alternatively, there may exist excitatorynteractions between M/T cells associated with the neck-ace glomeruli and typical odorant glomeruli in the olfactoryulb, although physiological studies suggest that the lat-ral connections between M/T cells associated with differ-nt glomeruli are predominantly inhibitory in nature.

What mechanism underlies the inhibitory responses todorants by CO2-activating neurons? Our previous imag-

ng and electrophysiological recordings indicate that es-entially all GC-D� OSNs are activated by CO2. It remainsnclear whether these CO2-activating OSNs can also be

nhibited by odorants and whether this inhibition can beropagated onto bulbar neurons associated with the neck-

ace glomeruli. A recent study reveals that CO2-activatingeurons in the fly antenna are inhibited by specific odor-nts (Turner and Ray, 2009), suggesting that similar direct

nhibition may also apply to GC-D� OSNs in mice. On thether hand, odorants may inhibit CO2-activating neurons

hrough lateral connections in the olfactory bulb. M/T cellsn the olfactory bulb receive mainly inhibitory lateral inputy forming synapses with GABAergic interneurons (Mori etl., 1999; Shepherd et al., 2004; Tan et al., 2010). The

nhibitory responses of odorant-responsive neurons toO2 may be mediated by lateral inhibition originated fromO2-activating neurons in the bulb.

Finally, the odorant-responsiveness of CO2 cells in thelfactory bulb raises the question on how the CO2 signal isncoded in the bulb. If a CO2-activating cell is also excitedy an odorant, how can the animal identify the stimulus asO2 or an odorant? This question may be answered by a

elatively large number (20) of necklace glomeruli in thelfactory bulb (Juilfs et al., 1997; Walz et al., 2007; Luo etl., 2009). While CO2 is likely encoded by the activation ofll necklace glomeruli, an odorant may activate only one orfew necklace glomeruli plus typical glomeruli in the main

lfactory bulb. Consequently, the mixture of CO2 and andorant may be encoded by simultaneous activation of aombinatory set of glomeruli, with a few necklace glomerulieing more strongly activated due to the presence of bothO2 and the given odorant.

cknowledgments—We thank A.L. Person for critical reading ofhe manuscript. This study was supported by a Project 973 grant2010CB833902) and a Project 863 grant (2008AA022312) fromhina Ministry of Science and Technology to ML.

REFERENCES

rechbuhl J, Klaey M, Broillet MC (2008) Grueneberg ganglion cellsmediate alarm pheromone detection in mice. Science 321:1092–1095.

ockerham RE, Puche AC, Munger SD (2009) Heterogeneous sen-sory innervation and extensive intrabulbar connections of olfactory

necklace glomeruli. PLoS One 4:e4657.

D

F

F

G

G

H

H

J

J

K

L

L

L

M

M

M

R

S

S

S

S

T

T

T

T

V

W

W

L. Gao et al. / Neuroscience 170 (2010) 881–892892

ulac C, Torello AT (2003) Molecular detection of pheromone signalsin mammals: from genes to behaviour. Nat Rev Neurosci 4:551–562.

leischer J, Mamasuew K, Breer H (2009) Expression of cGMP sig-naling elements in the Grueneberg ganglion. Histochem Cell Biol131:75–88.

ulle HJ, Vassar R, Foster DC, Yang RB, Axel R, Garbers DL (1995)A receptor guanylyl cyclase expressed specifically in olfactorysensory neurons. Proc Natl Acad Sci U S A 92:3571–3575.

iraudet P, Berthommier F, Chaput M (2002) Mitral cell temporalresponse patterns evoked by odor mixtures in the rat olfactorybulb. J Neurophysiol 88:829–838.

uo D, Zhang JJ, Huang XY (2009) Stimulation of guanylyl cyclase-Dby bicarbonate. Biochemistry 48:4417–4422.

allem EA, Sternberg PW (2008) Acute carbon dioxide avoidance inCaenorhabditis elegans. Proc Natl Acad Sci U S A 105:8038–8043.

u J, Zhong C, Ding C, Chi Q, Walz A, Mombaerts P, Matsunami H,Luo M (2007) Detection of near-atmospheric concentrations ofCO2 by an olfactory subsystem in the mouse. Science 317:953–957.

ones WD, Cayirlioglu P, Kadow IG, Vosshall LB (2007) Two chemo-sensory receptors together mediate carbon dioxide detection inDrosophila. Nature 445:86–90.

uilfs DM, Fulle HJ, Zhao AZ, Houslay MD, Garbers DL, Beavo JA(1997) A subset of olfactory neurons that selectively expresscGMP-stimulated phosphodiesterase (PDE2) and guanylyl cy-clase-D define a unique olfactory signal transduction pathway.Proc Natl Acad Sci U S A 94:3388–3395.

ay LM, Crk T, Thorngate J (2005) A redefinition of odor mixturequality. Behav Neurosci 119:726–733.

einders-Zufall T, Cockerham RE, Michalakis S, Biel M, Garbers DL,Reed RR, Zufall F, Munger SD (2007) Contribution of the receptorguanylyl cyclase GC-D to chemosensory function in the olfactoryepithelium. Proc Natl Acad Sci U S A 104:14507–14512.

uo M (2008) The necklace olfactory system in mammals. J Neuro-genet 22:229–238.

uo M, Sun L, Hu J (2009) Neural detection of gases—carbon dioxide,oxygen—in vertebrates and invertebrates. Curr Opin Neurobiol

19:354–361.

a M, Grosmaitre X, Iwema CL, Baker H, Greer CA, Shepherd GM(2003) Olfactory signal transduction in the mouse septal organ.J Neurosci 23:317–324.

eyer MR, Angele A, Kremmer E, Kaupp UB, Muller F (2000) AcGMP-signaling pathway in a subset of olfactory sensory neurons.Proc Natl Acad Sci U S A 97:10595–10600.

ori K, Nagao H, Yoshihara Y (1999) The olfactory bulb: coding andprocessing of odor molecule information. Science 286:711–715.

iffell JA, Lei H, Christensen TA, Hildebrand JG (2009) Characteriza-tion and coding of behaviorally significant odor mixtures. Curr Biol19:335–340.

hepherd GM, Chen WR, Greer CA (2004) Olfactory bulb. In: Thesynaptic organization of the brain, (Shepherd GM, ed), 165–217.New York, NY: Oxford University Press.

tange G, Stowe S (1999) Carbon-dioxide sensing structures in ter-restrial arthropods. Microsc Res Tech 47:416–427.

uh GS, Wong AM, Hergarden AC, Wang JW, Simon AF, Benzer S,Axel R, Anderson DJ (2004) A single population of olfactory sen-sory neurons mediates an innate avoidance behaviour in Drosoph-ila. Nature 431:854–859.

un L, Wang H, Hu J, Han J, Matsunami H, Luo M (2009) Guanylylcyclase-D in the olfactory CO2 neurons is activated by bicarbon-ate. Proc Natl Acad Sci U S A 106:2041–2046.

abor R, Yaksi E, Weislogel JM, Friedrich RW (2004) Processing ofodor mixtures in the zebra fish olfactory bulb. J Neurosci 24:6611–6620.

an J, Savigner A, Ma M, Luo M (2010) Odor information processingby the olfactory bulb analyzed in gene-targeted mice. Neuron65:912–926.

hom C, Guerenstein PG, Mechaber WL, Hildebrand JG (2004) FloralCO2 reveals flower profitability to moths. J Chem Ecol 30:1285–1288.

urner SL, Ray A (2009) Modification of CO2 avoidance behaviour inDrosophila by inhibitory odorants. Nature 461:277–281.

inje WE, Gallant JL (2000) Sparse coding and decorrelation in pri-mary visual cortex during natural vision. Science 287:1273–1276.

alz A, Feinstein P, Khan M, Mombaerts P (2007) Axonal wiring ofguanylate cyclase-D-expressing olfactory neurons is dependent onneuropilin 2 and semaphorin 3F. Development 134:4063–4072.

ilson RI, Turner GC, Laurent G (2004) Transformation of olfactoryrepresentations in the Drosophila antennal lobe. Science 303:

366–370.

(Accepted 3 August 2010)(Available online 6 August 2010)