In vivo vomeronasal stimulation reveals sensoryencoding of conspecific and allospecific cues by themouse accessory olfactory bulbY. Ben-Shaula, L. C. Katzb,1, R. Mooneyb, and C. Dulaca,2

aHoward Hughes Medical Institute, Department of Molecular and Cellular Biology, Harvard University, Cambridge, MA 02138; and bDepartment ofNeurobiology, Duke University School of Medicine, Durham, NC 27710

Communicated by Cornelia Bargmann, Rockefeller University, New York, NY, January 13, 2010 (received for review November 17, 2009)

The rodent vomeronasal system plays a critical role in mediatingpheromone-evoked social and sexual behaviors. Recent studies ofthe anatomical and molecular architecture of the vomeronasalorgan (VNO) and of its synaptic target, the accessory olfactorybulb (AOB), have suggested that unique features underlie vomer-onasal sensory processing. However, the neuronal representationof pheromonal information leading to specific behavioral andendocrine responses has remained largely unexplored due to theexperimental difficulty of precise stimulus delivery to the VNO. Todetermine the basic rules of information processing in thevomeronasal system, we developed a unique preparation thatallows controlled and repeated stimulus delivery to the VNO andcombined this approach with multisite recordings of neuronalactivity in the AOB. We found that urine, a well-characterizedpheromone source in mammals, as well as saliva, activates AOBneurons in a manner that reliably encodes the donor animal’s sex-ual and genetic status. We also identified a significant fraction ofAOB neurons that respond robustly and selectively to predatorcues, suggesting an expanded role for the vomeronasal systemin both conspecific and interspecific recognition. Further analysisreveals that mixed stimuli from distinct sources evoke synergisticresponses in AOB neurons, thereby supporting the notion of inte-grative processing of chemosensory information.

pheromones | sensory processing | vomeronasal | accessory olfactorybulb | mouse

In most animal species, the detection of pheromonal cues isessential to trigger and modulate sexual and social inter-

actions between conspecifics. Genetic and surgical manipu-lations in the mouse have demonstrated the essential role of thevomeronasal system in this process (1–4). Moreover, tracingstudies revealed dense projections from the vomeronasal systemto hypothalamic nuclei involved in behavioral and endocrinecontrol (5). The vomeronasal system thus offers a uniqueopportunity to explore principles of information processingunderlying animal–animal communication and species-specificsocial and sexual interactions.Although the rodent vomeronasal system is clearly involved in

pheromone sensing, the range of stimuli it can detect is largelyunknown (5). At one extreme, it may be exclusively dedicated toprocessing pheromonal (i.e., conspecific) information. Con-sistent with this view, several classes of conspecific cues wereshown to directly activate isolated vomeronasal organ (VNO)preparations (see ref. 6 for a recent review). Moreover, geneticas well as surgical VNO silencing significantly impairs behavioralresponses to conspecific cues (1, 2). Alternatively, the vomer-onasal system may resemble the main olfactory system in which awide variety of stimuli, including both pheromonal and non-pheromonal signals, are processed. This idea is supported by thefinding that in vitro VNO preparations can be activated by largesets of diverse chemicals (7, 8).The type and specificity of information extracted by the acces-

sory olfactory bulb (AOB) represent another unresolved issue.

AOBneurons can reliably detect conspecific sex from urinary cues(9) and strain information from as yet unidentified cues (10). Onehint that the vomeronasal system can provide finer discriminationamong individuals emerges from its role in the Bruce effect (4), inwhich a gestating female terminates pregnancywhen exposed to anunfamiliar male. However, how sex, strain, and individual identityare encoded by AOB neurons remains poorly understood.Exactly which stimulus features are encoded by the vomeronasal

system depends ultimately on the nature and extent of informationtransmitted from the primary sensory neurons to the AOB. Unlikemitral cells in the main olfactory bulb (MOB), in which responsesare shaped by input from one or a few receptor cell classes (11, 12),AOB output neurons receive convergent inputs from multipleglomeruli (13, 14). This architecture could provide a platform forintegrative information processing that significantly differs fromthat observed in the main olfactory system. However, direct phys-iological evidence for integrative processing by AOB neurons isstill lacking.A major challenge to resolving these issues is that stimulus

delivery to the VNO requires active pumping that is triggered onlyduring exploratory behaviors (15). The relatively low throughput ofchronic recording methods and, more fundamentally, the lack ofcontrol over stimulus uptake in freely behaving mice are not wellsuited for the systematic investigation of information processing inthe AOB. Recently (9, 16), direct stimulus perfusion into the VNOof the anesthetized opossum and mouse has been used to show theimportance of the AOB in distinguishing conspecific sex throughurinary cues. However, delivery by perfusion requires relativelylarge samples, typically pooled from multiple individuals, hencelimiting the ability to systematically address the processing ofpheromonal stimuli with limited availability. Furthermore, stimulusdelivery by perfusion effectively bypasses VNO pump dynamics, aphysiological parameter that may play a significant role in thefaithful transductionof stimulus information todownstreamtargets,much as sniffing does in the main olfactory system (17).To overcome these limitations, we developed a unique exper-

imental approach that enables repeatable and naturalistic deliveryof small and physiologically relevant stimulus volumes to the intactVNO. By combining this approach with multisite electrophysio-logical recordings, we explored how AOB neurons respond tofeatures in conspecific urine and saliva. These experiments re-vealed that AOB neurons respond to these complex stimuli in amanner that reliably encodes information about the strain and sex,as well as finer aspects that might represent the physiological state

Author contributions: Y.B.-S., L.C.K., and C.D. designed research; Y.B.-S. performed re-search; Y.B.-S., R.M., and C.D. analyzed data; and Y.B.-S., R.M., and C.D. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1Deceased November 26, 2005.2To whom correspondence should be addressed. E-mail: dulac@fas.harvard.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0915147107/DCSupplemental.

5172–5177 | PNAS | March 16, 2010 | vol. 107 | no. 11 www.pnas.org/cgi/doi/10.1073/pnas.0915147107

Dow

nloa

ded

by g

uest

on

Nov

embe

r 23

, 202

0

and the individual identity of conspecifics. This approach also ledus to the surprising finding that cues present in a variety of pred-ator urines can serve as potent and selective activators of AOBneurons. These results offer unique insights into how the vomer-onasal system functions to enable individual recognition impor-tant to social and sexual behavior, while also suggesting anexpanded role for this system in the recognition of other species.

ResultsA Naturalistic Preparation for Recording Neural Activity in theVomeronasal System. To activate vomeronasal pumping in anes-thetized animals, we implanted custom-built cuff electrodes overthe cervical region of the sympathetic nerve trunk using the carotidartery as a scaffold (Fig. 1A). Initial experiments established thatbrief trains of biphasic current pulses applied to the sympatheticnerve (amplitude ±100 μA, frequency 33 Hz, duration 1.6 s)inducedefficient uptakeoffluorescent dye into theVNO(Fig. 1A).To flush the nasal cavity andVNO lumen and thus enablemultiplecycles of stimulus presentation, we induced a flow of Ringer’ssolution from the nostril through the nasopalatine duct and to theoral cavity (18) while repeatedly activating the VNO pump (Fig.1A). A planar array of 32 electrodes (19) was inserted into theexternal cellular layer (ECL) of the AOB that contains the cellbodies of mitral and tufted cells (20). Subsequent confirmation ofelectrode placement in the ECL was performed by histologicalinspection of the fluorescent dye-painted (DiI) electrode tracks(Fig. 1A). Neuronal activity recorded in the AOB when a smallvolume of diluted urine (2 μL) was presented to the nasal cavityeither alone or with sympathetic nerve stimulation is shown in Fig.1B. These recordings reveal that both stimulus application andsubsequent sympathetic stimulation are typically required to evokea neuronal response in the AOB and show that repeated sym-pathetic stimulation during flushing can be used to efficiently“clean” the VNO (Fig. 1B).

Basic Characteristics of Vomeronasal Responses.Using this approach,we analyzed responses of 711 single and 2,332AOBmultiunits to avariety of stimuli. Unless otherwise indicated, all analyses refer tosingle-unit data. Evoked responses of two single units recordedsimultaneously following six trials of interleaved presentations ofmale and female urine are shown in Fig. 2A. In addition to dem-onstrating the repeatability and reliability of our experimentalprocedure, these data highlight the known (9) specificity of AOBunits to male and female stimuli. The response specificity to sex-specific stimuli is further documented in SI Text.Baseline firing patterns of single units were low (0.94± 0.05 Hz,

mean ± SEM), similar although slightly lower than values pre-viously recorded in the anesthetized and behaving mouse (9, 10).These lower baseline values could reflect amore effective removalof residual stimuli with our cleaning procedure, our nonbiasedapproach to finding single units, or the anesthetic state. Rateincreases started 4.34 ± 0.03 s following nerve stimulation andpeaked at 10.9 ± 0.10 s, with a half time of 12.8 ± 0.06 s (Fig. 2B).Similar quantification of the temporal profiles of rate decreaseswas impractical due to the low baseline rates. The averageresponse delay (4.34 s) was also similar to that measured in freelybehaving mice (3.6 ± 0.7 s) (10). Because the vomeronasal sig-naling cascade has been estimated to require <0.5 s (21), theresponse latency observed here could reflect the combined delaysof sympathetic stimulation, pump activation, and stimulus suction.Comparison of the temporal response profiles observed here withdirect recordings of VNO pump activity in anesthetized (22) andawake animals (15) suggests that each sympathetic stimulation inour preparation triggers a single suction event. Thus, the basicvomeronasal response is significantly slower than the unitarysensory events in the main olfactory system, which occurs at atypical frequency of 4–12 Hz (23).

Stimulus-Induced Responses Require Functional TRPC2 Signaling. Incontrast to wild-type mice, recordings from TRPC2−/−mice failedto reveal any responses when stimulus application to the nasalcavity was paired with sympathetic nerve stimulation. Specifically,although baseline activity from AOB neurons of adult TRPC2−/−

mice could be recorded (n = 221 units including multi-unitactivity, eight recording sites from four males and two females),the distribution of stimulus-evoked response strengths was clearlydifferent from that of WT mice (Fig. 3A) and, more importantly,the rate of significant responses (P < 0.01) was at chance levels(Fig. 3B). This finding indicates that the AOB responses evokedby pairing stimulus presentation with sympathetic nerve stim-ulation require an intact TRPC2 transduction pathway in theVNO. Thus, in contrast to recent reports that postulate residualTRPC2-independent vomeronasal activity (24), the present re-sults employing direct measurement of electrical activity in theAOB reveal that the elimination of vomeronasal signaling iseffectively complete following genetic ablation of TRPC2.

AOB Readout of Sex and Strain. Prior studies involving chronicrecordings from behaving mice revealed that AOB neurons candistinguish strain-specific cues (10), but the pheromonal source ofthis information was not identified. Interestingly, these recordingsrevealed strong and selective bursts of activity correlated with theanimal’s investigation of both the anogenital and the facial regionof conspecifics (10), suggesting that at least two bodily sourcescontain strain-specific information. Moreover, exocrine glandsecreted peptides (ESPs) found in tears, mucosa, or saliva in a sex-and strain-specific manner have been shown to be effective vom-eronasal stimuli (25, 26). We explored this issue by testing theresponses ofAOBneurons tourine and saliva samples frommiceofdifferent sex and strain combinations. Of 107 single units testedwith urinary stimuli, 40 (37%) showed a significant response to oneor more stimuli (P < 0.01, nonparametric ANOVA) with a preva-lence of responses to only one stimulus (60%, 24 of 40; SI Text). Of

LD

L M

B

A

Stimulus application

bloodvessel

VNOlumen

Sympathetic stimulation

Time (seconds)

Flushing

0 10 20 0 10 20 0-10 -10 -10 10 2030 30 30

10H

z

AOB

VNO

Nasopalatineduct sympathetic

nerve trunk

GlECL

Gr

D

VA P

BV

Fig. 1. Experimental setup. (A) Mouse nasal cavity and relevant structures.(Left Inset) Coronal section of the VNO showing uptake of DiI ipsilateral tonerve stimulation. BV, blood vessel; L, VNO lumen. (Right Inset) Sagittal AOBsection showing a DiI-painted probe within the external cell layer (ECL). GL,glomerular layer; Gr, granule cell layer. (B) Sequence of events in a single trialshowing the serial application of mouse urine followed by sympathetic stim-ulation and flushing. Each panel includes a cartoon of the presumed state ofVNO, a raw electrode signal (stimulation artifacts clipped), and firing rates.

Ben-Shaul et al. PNAS | March 16, 2010 | vol. 107 | no. 11 | 5173

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Nov

embe

r 23

, 202

0

23 single units tested with salivary stimuli, 13 (57%) showed a sig-nificant response (P< 0.01) to one ormore stimuli (Fig. 4B andE).As seen with urinary stimuli (Fig. 4A), AOB neurons displayresponses triggered exclusively by saliva from one or a few strains,by all animals of a given sex, or in more complex patterns (Fig. 4Band SI Text). For both urine and saliva, units responding exclusivelytourineof both sexes fromthe same strainwerenot observed.Thus,both urinary and salivary stimuli can elicit highly specific responsesin single units in a manner that could be used by the mouse tounambiguously detect the strain and sex of the stimulus animal.Are the sex-specific responses observed here and elsewhere

merely a by-product of the finer discrimination of strain and sexselectivity? To test this issue we analyzed the population responsesto urinary and salivary stimuli using hierarchical clustering. Thisanalysis revealed a perfect segregation of responses to male vs.female urine and an almost perfect segregation for salivary cues(Fig. 4 C and D). Thus, although individual AOB neurons areoften tuned to specific strain/sex combinations, sex neverthelessemerges as a primary parameter distinguishing different classes ofresponses in the AOB.To explore whether AOB neuron responses to individual

samples reflect the genetic background of the donor animals, wepresented three distinct urine samples from males of each ofthree distinct strains (CBA, BalbC, or C57Bl6). For each strain,two samples were collected from one individual and one wascollected from a different mouse. Of 51 single units tested, 16showed a significant response to at least one stimulus. Pop-

ulation analysis of these responses showed that urinary cues fromthe same strain and, for two of three strains, even from the sameindividual within a strain clustered together (Fig. 4F). However,in addition to neurons that responded to all samples of a givenstrain, we observed specific responses to samples from an indi-vidual mouse and even to individual samples from a given mouse(Fig. 4C). The subset of neurons that differentiated samplesfrom different mice and even different samples of the samemouse could provide the basis for detection of more subtlefeatures, which might reflect the individual’s physiological stateand identity. One should stress that these responses are not likelyto reflect random fluctuations in the neuronal responses, as theyare consistent across multiple presentations of the same stimuli.

Responses to Allospecific Stimuli. To investigate the potential roleof the mouse vomeronasal system in detecting allospecific stim-uli, we measured the responses of AOB single units to a mix ofurinary cues from three mouse predators: bobcat, fox, and rat.We found that the mix of predator urines could evoke remark-ably robust responses from AOB neurons. Of 186 single unitstested with both mouse and predator stimuli, an equal fraction(19%) showed a significant (P < 0.01) response to mouse (maleor female) or to predator urine. Importantly, most neurons(60%, 31 of the 51 single units that showed any response)responded specifically only to predator or mouse urine (Fig. 5 Aand C). This specificity and the similar distribution of responsemagnitudes to mouse and to predator stimuli (Fig. 5B) suggestthat mouse AOB neurons respond to predator urine because itcontains specific predator-related cues and not simply because itcontains cues also common to mouse urine.To further explore the specificity of these responses, we tested

the responsiveness of individual AOB units to the urine of indi-vidual predators. Of 35 units tested, 12 showed a response to atleast one of these stimuli (P < 0.01). Of these 12, 5 responded toonly one predator stimulus in addition to the mixed predator cues(Fig. 5D), 2 responded to all stimuli, and the rest responded to onlyone of the stimuli (Fig. 5B and SI Text). The robustness of predatorresponses and the specificity of discrimination betweenmouse andpredator cues show that the vomeronasal system could participatein the detection and discrimination of allospecifics.

Responses to Conflicting Stimuli. The detection of conspecificfemale, male, or predator cues triggers dramatically differentbehavioral responses. It is therefore intriguing that although∼60%of the single units (31 of the 51 responding units) are specifically

0

10

400

10

0

5

0

5

40time (s)

11

neuron female urine male urine

2

mea

n re

spon

ses

neuron 2

neuron 1

neuron 2

Hz

trial #

Hz

time (s)

12 213 214 215 216 2

A

neuron 1

Time (seconds)

Num

ber o

f cas

es

1000start time

500peak time

0 5 10 15 20 25 300

500

B

half time

Fig. 2. Sex specificity of responses and temporal response profiles. (A) Sympathetic stimulation-induced spike times in interleaved presentations of dilutemale or female mouse urine from two simultaneously recorded neurons (orange and green). (Lower) Averaged responses (mean ± SEM). Time 0: sympatheticstimulation. (B) Histograms of start time (latency), peak time, and half time of individual responses relative to the onset of sympathetic stimulation (time 0).

0WT

5

10

15

% s

igni

fican

t res

pons

es (P

<0.0

1)

20

25

TRPC-/-0 -2 -1 0 1 2 3 4response (Hz)

5 6 7 8 9 10

WT (1195)

n=221n=1195

TRPC-/- (221)

15

30

45

% re

spon

se s

treng

th d

istri

butio

n

60

75

90 BA

Fig. 3. Dependence of AOB responses on functional TRPC2 signaling. (A) Per-centagesof stimulus-induced rate changesof individual units to themost effectivestimulus (stimuli were male, female, and predator urine). (B) Percentage of sig-nificant responses (P<0.01) tomouseurine inwild-type (WT)andTRPC2−/−mutantmice. Data in this analysis includemulti-unit activity. (n = 1,195,WT; 221, TRPC−/−).

5174 | www.pnas.org/cgi/doi/10.1073/pnas.0915147107 Ben-Shaul et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 23

, 202

0

activated by only one stimulus class (P< 0.01) and thus can conveyunambiguous information to downstream targets, a substantialfraction (the remaining 40%) respond to multiple stimulus classeswith conflicting significance (Fig. 5 A and E).To explore how conflicting cues are processed by the AOB, we

measured responses to male and female mouse urinary cues, topredator urine, and to their mixtures [n = 61 single units, ofwhich 36 (59%) showed significant response (P < 0.05) to at leastone stimulus]. The mixtures were prepared as averages of theelemental stimuli (SI Methods). Specifically, we asked whether amix of conflicting cues generates an intermediate or a novelresponse and whether the response to one cue can overrideothers. At the population level, similarity relationships (given ascorrelation distances; SI Text) reveal that female and male micestimuli are considerably closer to each other than each is topredator urine (0.46 vs. 0.78 and 0.82, normalized distances).

This result indicates that AOB activity readily allows differ-entiation of conspecific from nonconspecific urine stimuli.Moreover, mixes of predator with female or with male urineyielded intermediate representations between mouse and pred-ator stimuli. Thus, the predator stimulus does not override orinhibit the response to conspecific cues. These relationships aredepicted in Fig. 6A using multidimensional scaling (MDS) toapproximate these distances in two-dimensional space (Meth-ods). For comparison, also shown are distance relationships forsimulated cases in which the predator response entirely inhibitsor, alternatively, combines linearly with the mouse urine re-sponse. Comparison of the actual data with the two simulatedcases supports the idea that responses to mixed stimuli resembleintermediate responses to the elemental stimuli.At the single-unit level, a substantial proportion of responses to

the combined predator and mouse stimuli (38%, 20 of 53 cases)

0

10

0

5

0

5

0 20 40

18

0 20 40 0 20 40

0

5

0 10 20 300

7

0 10 20 30 0 10 20 30

0

6

0 10 20 300

10

0 10 20 30 0 10 20 300 10 20 30

0

16

0 10 20 30 0 10 20 30 0 10 20 30

bobcat fox rat

0

A C

D

-1 0 1 2 3 4-1

0

1

2

3

4

mouse response (Hz)

Pre

dato

r Res

pons

e (H

z)

Response Strength

31.4%

7.8%

13.7% 7.8%

17.7%

9.8%

11.8%

B

E

time (seconds)

time (seconds)time (seconds)

Response Frequencies

Hz

Hz

Hz

all

Fig. 5. Responses to predator urine. (A) Percen-tages of units with significant responses (P < 0.01)to each of the stimulus combinations (n = 51single units responding to at least one stimulus).Stimulus categories are male mouse urine, femalemouse urine, and predator urine. Mouse urinewas pooled from strains BalbC, C57Black6, andCBA. (B) Response magnitudes of these singleunits to mouse and predator urine. (C) Threesingle units with selective responses to femalemouse, male mouse, or combined predator urine.(D) Single units responding selectively to differ-ent predators. The icons designate, from left toright, bobcat, fox, rat, and combined urine fromthese sources. (E) Single units responding acrossstimulus categories.

10

0 40 0 400 40 0 400 40 0 40

10H

z

0

4

15

0

10

4

C57 BC FVB C57 BC FVB

0 40 0 400 40 0 400 40 0 40

3

4

5

BC (M)

C57 (M)

FVB (M)

BC (F)FVB (F)

C57 (F)

CBA (M)

CBA (F)

C57(M)

C57 (F)

BC (M)

BC (F)

4

5

6

dist

ance

C57 BC CBA C57 BC CBA

Female urine Male urine

03

0

6

0 4008

0 40 0 40 0 40 0 400 40 0 400 40 0 40

BC-1 BC-1B C57-1 C57-1B C57-2CBA-1B CBA-2CBA-1 BC-2

0 40 0 40 0 40 0 40 0 400 40 0 400 40 0 40

0 20 0 20 0 20 0 20 0 200 20 0 200 20 0 20

CBA-1BCBA-1CBA-2BC-1BBC-1BC-2C57-1BC57-1C57-2

1

2

3

4

Hz

Hz

Individual Male SamplesA

B FED

C

M/F Urine M/F Saliva Male UrineFemale saliva Male saliva

time (s)

time (s)

Fig. 4. Responses to urine and saliva. (A) Responses of three different single units to urine from mice of distinct sex and strain. Significant responses (P < 0.01,either positive or negative) are shown on a gray background. (BC, BalbC; C57, C57/Black6). (B) Same as A for saliva stimuli. (C) Responses to distinct urinesamples. Samples ending with -1 and -1B denote different samples from the same individual whereas the sample ending with -2 is from a different individual.(D–F) Hierarchical clustering of population responses to the stimuli in A (n = 23), B (n = 13), and C (n = 16). A–C show responses of distinct units.

Ben-Shaul et al. PNAS | March 16, 2010 | vol. 107 | no. 11 | 5175

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Nov

embe

r 23

, 202

0

display intermediate magnitudes to that of the individual stimuli,consistent with the population data shown in Fig. 6A. Surprisingly,the data also reveal a prevalence of synergistic interactionsbetween predator and mouse stimuli (42%) (Fig. 6B). In com-parison, mixes of urine from distinct samples of mouse urine eli-cited fewer and weaker synergistic interactions (SI Text). Becausemouse and predator stimuli are likely to contain a large number ofnonoverlapping components, as compared to two conspecificstimuli, this result suggests that a specific combination of compo-nents is sometimes required to elicit a robust response, thus pro-viding direct support to the notion that AOB units integrateinformation from distinct components.

DiscussionThe importance of the mouse vomeronasal system in processingchemical cues required for sex- and species-specific social andreproductive behaviors has been documented in various studies(1, 2, 4, 27). However, many fundamental aspects of vomeronasalfunction have not been addressed yet: Is the vomeronasal systemdedicated to pheromone detection? How are the representationand computation of social parameters achieved? The challengesposed by the unique mode of stimulus delivery to the vomer-onasal organ, primarily the sympathetic activation of a vascularpump, have so far hampered in vivo physiological investigation ofvomeronasal function, leaving the range of chemical stimulidetected by the vomeronasal system and the basic rules ofvomeronasal information processing largely unknown.Herewedescribe a unique experimental preparation for studying

vomeronasal function that is noninvasive, engages the naturalpumping mechanism, requires small volumes of test stimuli, andallows many cycles of stimulus presentation. By combining thismethod with multielectrode recording in the AOB of the anes-thetized mouse, we confirmed the high selectivity of vomeronasalactivation by male and female mouse urinary and salivary cues andshowed that when challenged with increasingly more subtle dis-criminations, AOB neurons can achieve multilevel distinctions,namely, among animals from different strains, among individualsfrom the same strain, and even among distinct samples from anindividual mouse.The wiring diagram of the AOB identified by anatomical and

genetic tools has suggested that, in contrast to the MOB, inte-gration of information by output neurons may involve multiplechemosensory receptors (20, 28). Whereas the VNO appears toreliably convey information about sex (21, 29), an earlier studysuggested that a population code is required to derive informa-tion about strain (30). In our AOB recordings we identifiedmostly units with selectivity to sex/strain combinations, a fewsingle units with generalized sex responses, and no units with

specific responses to strain regardless of sex (Fig. 4). In contrastto VNO recordings, we also observed single units that general-ized across all male samples of the same strain or across distinctsamples from the same individual (Fig. 4).We found here that many AOB units could respond to cues

present in predator urine. These findings suggest that in additionto serving an important role in individual recognition of con-specifics, the vomeronasal system of the mouse is also likely toserve an important role in recognizing animals of other species,perhaps especially predators. Responses of AOB neurons topredator urines were robust in magnitude and frequency andappear highly specific, even among distinct predator species.Intriguingly, reproductive processes can be affected by thepresence of predator stimuli (31). Because the behavioral (1) andphysiological (4, 32) involvement of the vomeronasal system inreproductive processes is well established, predator odors mayact via the VNO to affect such functions. However, predator-related responses have been recently detected in the mainolfactory system (33), raising the question about the respectiverole of each chemosensory system in this process. The slowtiming of AOB activation after stimulus detection, in the order ofseveral seconds compared to a fraction of seconds in the mainolfactory system, together with the need for contact for VNOstimulus detection, suggests that the vomeronasal system mayprovide a means to survey the presence of predators within theecological niche. This, in turn, could provide long-term modu-lation of the mouse’s reproductive physiology, whereas olfactorydetection may provide immediate avoidance of predators. Pre-vious work showed the importance of the vomeronasal system forpredator and prey detection in reptiles (34, 35), and more recentstudies implicated the rodent vomeronasal system in detectingpredator cues as well (35–37). However, our findings reveal thatsuch odorants elicit robust, frequent, and specific electricalactivity in the mouse AOB.Cues from conspecifics and predators provide ecologically

conflicting information that could lead to strikingly differentbehavioral/physiological outputs, the former by enticing socialand reproductive interactions, the latter by arousing fear anddelaying reproduction. How would animals resolve the simulta-neous detection of conflicting stimuli? In one scenario, infor-mation about conflicting classes of stimuli is already processed atearly stages by largely independent units that in turn convey theprospect for friendly or harmful encounters to distinct effectornuclei in the amygdala and hypothalamus. In addition, somecross-inhibition in the response to both stimuli presented simul-taneously may exist, such that the anticipation of a mating partner,for example, can be dampened by the detection of a possiblethreat. We indeed identified AOB neurons that displayed specific

suppression

n=11 n=20 n=22

intermediateresponse synergy

Responses to combined stimulus sets

cases

0

2

4

6

8

resp

onse

(Hz)

A B

1 Actual data

Pur

e st

imul

i

23 Linear combination (sim)

Predator inhibition (sim)

3

3

32

2

2

1

11

Mix

ed s

timul

i

projection 1

pro

ject

ion

2

LEGEND

Fig. 6. Responses to conflicting stimuli. (A) Rela-tionship of population responses to the elementalstimuli and to their combinations using multi-dimensional scaling (MDS). Each circle representsthe population response to one stimulus (see colorlegend).The “predator inhibition” simulationmodels the response were predator urine toentirely inhibit the response to mouse urine. (B)Responses of individual units to mixed stimuli.Each line represents the response of one singleunit to one mix. The edges of the line are definedby the responses of a given unit to the two ele-mental stimuli (male or female mouse urine andpredator urine) and the + indicates the responseof the same unit to their mixture. Responses arecategorized into those showing suppression orintermediate or synergistic interactions followingmixing. n = 53 cases. Male and female mouseurine was pooled as described in Fig. 5.

5176 | www.pnas.org/cgi/doi/10.1073/pnas.0915147107 Ben-Shaul et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 23

, 202

0

responses to either female or male mice or to predators. How-ever, we also detected surprisingly large cohorts of neurons thatresponded to both mouse and predator urine, which is hard toreconcile with a highly segregated opponent mechanism. What isthe functional significance of this dual specificity? One possibilityis that features used to distinguish between two classes of con-specific stimuli may also be present in predator urine. Alter-natively, the feature space defined by AOB neurons may not beexplicitly related to the ecological significance of the stimuli.Instead, AOB processing might maximize the efficiency ofinformation transfer to downstream targets regardless of theirethological value. Under this scenario, the significance of thevarious response patterns is relegated to subsequent processingstages in the amygdala and hypothalamus. Finally, the synergisticresponses to mouse/predator odors uncovered in our study mayprovide a sensory alert for the presence of conflicting cues.According to this hypothesis, AOB processing would not onlyencode chemical features, but also attribute salience to a specificcombination of cues. Regardless of their significance, theobservation of these synergistic responses provides direct phys-iological support to the notion that some AOB mitral cellseffectively integrate information from multiple glomeruli.Summing the frequencies of responses to each of the stimuli

tested indicates that, on average, single units respond to more thanone type of stimulus. Indeed, some units display responses acrossdistinct stimulus classes (i.e., conspecific and predator urine; Fig. 5).This result suggests that single units cannot unambiguously encodethe identity of the stimulus. Instead, it seems that activity across apopulation is required to extract information about stimuli and therelationships between them. This is supported by our populationlevel analyses, which show that small ensembles can reliably encodethe relationships between stimuli (e.g., Fig. 4 D–F).

Finally, we have described a powerful experimental platformthat, in addition to providing insights about AOB function,should be further instrumental in exploring the processing ofvomeronasal as well as olfactory chemosensory information inamygdala and hypothalamic behavioral control centers to gen-erate physiologically relevant responses.

MethodsSurgical Procedures. Experiments were performed under National Institutesof Health, Duke University, and Harvard University guidelines. See SI Text fordetails on surgery, stimulus delivery, and electrophysiology.

Data Analysis. Responses were quantified as the rate change followingstimulation (30 s) relative to the preceding 10 s. Response significance for agiven stimulus was determined with a one-way nonparametric ANOVA withthe set of poststimulation rates (over a 30-s period) compared to the set of allprestimulation rates. All data analyses were performed with custom writtenor built-in MATLAB code. See SI Text for further details.

Stimuli. Fresh urine samples were frozen at −80 °C. For stimulation, urine wasdiluted in water or in Ringer’s solution (1/100). Saliva was collected from theoral cavity using amicropipette. Salivation rate was increased by pilocarpine–HCL injections (10 mg/kg, i.p.). Bobcat and fox urine was kindly provided byPredatorPee. Rat urine was collected similarly to mouse urine. Artificial urineand Ringer’s solution were prepared as described in ref. 21. See SI Text forfurther details.

ACKNOWLEDGMENTS. We thank G. Buzsaki for help with the recordingsetup; D. Purves and G. Feng for assistance with the surgical procedure;R. Irving and D. Kloetzer for technical support; A. Adani, A. Reinhart, L. Pont-Lezica, and D. Lin for stimuli; R. Hellmiss for artwork; and B. Arenkiel,I. Davison, S. Shea, the Dulac laboratory, and two anonymous reviewers forhelpful comments and suggestions.

1. Kimchi T, Xu J, Dulac C (2007) A functional circuit underlying male sexual behaviour inthe female mouse brain. Nature 448:1009–1014.

2. Stowers L, Holy TE, Meister M, Dulac C, Koentges G (2002) Loss of sex discriminationand male-male aggression in mice deficient for TRP2. Science 295:1493–1500.

3. Del Punta K, et al. (2002) Deficient pheromone responses in mice lacking a cluster ofvomeronasal receptor genes. Nature 419:70–74.

4. Bruce HM (1959) An exteroceptive block to pregnancy in the mouse. Nature 184:105.5. Halpern M, Martínez-Marcos A (2003) Structure and function of the vomeronasal

system: An update. Prog Neurobiol 70:245–318.6. Tirindelli R, Dibattista M, Pifferi S, Menini A (2009) From pheromones to behavior.

Physiol Rev 89:921–956.7. Trinh K, Storm DR (2003) Vomeronasal organ detects odorants in absence of signaling

through main olfactory epithelium. Nat Neurosci 6:519–525.8. SamM, et al. (2001) Neuropharmacology. Odorants may arouse instinctive behaviours.

Nature 412:142.9. Hendrickson RC, Krauthamer S, Essenberg JM, Holy TE (2008) Inhibition shapes sex

selectivity in the mouse accessory olfactory bulb. J Neurosci 28:12523–12534.10. Luo M, Fee MS, Katz LC (2003) Encoding pheromonal signals in the accessory olfactory

bulb of behaving mice. Science 299:1196–1201.11. Luo M, Katz LC (2004) Encoding pheromonal signals in the mammalian vomeronasal

system. Curr Opin Neurobiol 14:428–434.12. Fantana AL, Soucy ER, Meister M (2008) Rat olfactory bulb mitral cells receive sparse

glomerular inputs. Neuron 59:802–814.13. Wagner S, Gresser AL, Torello AT, Dulac C (2006) A multireceptor genetic approach

uncovers an ordered integration of VNO sensory inputs in the accessory olfactorybulb. Neuron 50:697–709.

14. Takami S, Graziadei PP (1991) Light microscopic Golgi study of mitral/tufted cells inthe accessory olfactory bulb of the adult rat. J Comp Neurol 311:65–83.

15. Meredith M (1994) Chronic recording of vomeronasal pump activation in awakebehaving hamsters. Physiol Behav 56:345–354.

16. Holekamp TF, Turaga D, Holy TE (2008) Fast three-dimensional fluorescence imagingof activity in neural populations by objective-coupled planar illumination microscopy.Neuron 57:661–672.

17. Verhagen JV, Wesson DW, Netoff TI, White JA, Wachowiak M (2007) Sniffing controlsan adaptive filter of sensory input to the olfactory bulb. Nat Neurosci 10:631–639.

18. Wöhrmann-Repenning A (1984) [Comparative anatomical studies of the vomeronasalcomplex and the rostral palate of various mammals. I]. Gegenbaurs Morphol Jahrb130:501–530.

19. Csicsvari J, et al. (2003) Massively parallel recording of unit and local field potentialswith silicon-based electrodes. J Neurophysiol 90:1314–1323.

20. Larriva-Sahd J (2008) The accessory olfactory bulb in the adult rat: A cytological studyof its cell types, neuropil, neuronal modules, and interactions with the main olfactorysystem. J Comp Neurol 510:309–350.

21. Holy TE, Dulac C, Meister M (2000) Responses of vomeronasal neurons to naturalstimuli. Science 289:1569–1572.

22. Meredith M, O’Connell RJ (1979) Efferent control of stimulus access to the hamstervomeronasal organ. J Physiol 286:301–316.

23. Kepecs A, Uchida N, Mainen ZF (2006) The sniff as a unit of olfactory processing.Chem Senses 31:167–179.

24. Kelliher KR, Spehr M, Li XH, Zufall F, Leinders-Zufall T (2006) Pheromonal recognitionmemory induced by TRPC2-independent vomeronasal sensing. Eur J Neurosci 23:3385–3390.

25. Kimoto H, et al. (2007) Sex- and strain-specific expression and vomeronasal activity ofmouse ESP family peptides. Curr Biol 17:1879–1884.

26. Kimoto H, Haga S, Sato K, Touhara K (2005) Sex-specific peptides from exocrineglands stimulate mouse vomeronasal sensory neurons. Nature 437:898–901.

27. Leypold BG, et al. (2002) Altered sexual and social behaviors in trp2 mutant mice. ProcNatl Acad Sci USA 99:6376–6381.

28. Dulac C, Wagner S (2006) Genetic analysis of brain circuits underlying pheromonesignaling. Annu Rev Genet 40:449–467.

29. Leinders-Zufall T, et al. (2000) Ultrasensitive pheromone detection by mammalianvomeronasal neurons. Nature 405:792–796.

30. He J, Ma L, Kim S, Nakai J, Yu CR (2008) Encoding gender and individual informationin the mouse vomeronasal organ. Science 320:535–538.

31. Apfelbach R, Blanchard CD, Blanchard RJ, Hayes RA, McGregor IS (2005) The effects ofpredator odors in mammalian prey species: A review of field and laboratory studies.Neurosci Biobehav Rev 29:1123–1144.

32. Lomas DE, Keverne EB (1982) Role of the vomeronasal organ and prolactin in theacceleration of puberty in female mice. J Reprod Fertil 66:101–107.

33. Kobayakawa K, et al. (2007) Innate versus learned odour processing in the mouseolfactory bulb. Nature 450:503–508.

34. Halpern M, Frumin N (1979) Roles of the vomeronasal and olfactory systems in preyattack and feeding in adult garter snakes. Physiol Behav 22:1183–1189.

35. Miller LR, GutzkeWH (1999) The role of the vomeronasal organ of crotalines (Reptilia:Serpentes: Viperidae) in predator detection. Anim Behav 58:53–57.

36. McGregor IS, Hargreaves GA, Apfelbach R, Hunt GE (2004) Neural correlates of catodor-induced anxiety in rats: Region-specific effects of the benzodiazepinemidazolam. J Neurosci 24:4134–4144.

37. Masini CV, et al. (2010) Accessory and main olfactory systems influences on predatorodor-induced behavioral and endocrine stress responses in rats. Behav Brain Res 207:70–77.

Ben-Shaul et al. PNAS | March 16, 2010 | vol. 107 | no. 11 | 5177

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Nov

embe

r 23

, 202

0