Brain mechanisms of behavior (PSY 445)Lecture for Day 19

Required reading:Enhanced partner preference in a promiscuous species by manipulating the expression of a single gene. Miranda Lim, Zuoxin Wang, Daniel Olazábal, Xianghui Ren, Ernest Terwilliger, and Larry Young. http://www.neuro.uoregon.edu/wehr/coursepapers/Lim-Young-2004.pdf

Further readings: Love & Sex: The Vole Story.Kate Egan, 1998.http://whsc.emory.edu/_pubs/em/1998summer/vole.html

How could we ever understand the neural, genetic, and molecular mechanisms of a be-havior as complex as love or pair bonding? Psychologists have learned a great deal about the importance of bonding — the formation of social attachments— in social de-velopment. But surprisingly it was field biologists, back in the 1980’s, who laid the foun-dations for the astounding discoveries in the following decades that have started to re-veal the neural and molecular mechanisms underlying bonding. The field biologists dis-covered that several species of voles — small mouse-like burrowing rodents — appear very similar on the outside, but form strikingly different social structures in the wild. Mon-tane voles and meadow voles, like 95% of all mammals, are polygamous. Meadow voles are loners: they are solitary, and mate promiscuously — a different partner each time. The males don’t take part in raising offspring, and even the females abandon their offspring soon after birth. Prairie voles could not be more different. They are a model of family values. They naturally form life-long monogamous pair bonds. Once they have mated, they vary rarely mate with a different partner for the rest of their life. After bond-ing, the males react aggressively towards other males. Both males and females prefer the company of their mate to others, build a nest together, and provide extensive and prolonged bi-parental care. By discovering this naturally occurring difference in pair-bonding in such closely related species in the wild, the field biologists opened the door for neuroscientists, chiefly Tom Insel and his colleagues at NIMH, to try to bring these animals into the lab, where they could quantify these behavioral differences, and start to decipher the neural mechanisms underlying pair bond formation.

Quantifying partner preferenceIn order to study pair bonding quantitatively, Insel built a wide variety of mazes and chambers of different shapes and sizes. Each apparatus was designed to test a differ-ent aspect of bonding behaviors. Here we will focus on partner preference. The basic setup for testing preferences is to use a 3-part apparatus. In this case, after mating has occurred and bonding has been established, we want to test the preference of a vole for their partner, compared to a stranger. We tether the partner in one chamber, and the stranger in another. Then we place the test animal in the neutral chamber in the middle, and record the amount of time they spend in each chamber. Prairie voles spend the

Neural mechanisms of pair bonding in voles 1

most time in the partner chamber. Montane voles, on the other hand, spend the most time alone in the neutral chamber.

Neural mechanisms of pair bonding in voles 2

Prairie voles and meadow voles are very similar in appearance. Prairie voles are monogamous (pair bonding, biparental care) while meadow voles are solitary and polygamous.

132 | FEBRUARY 2001 | VOLUME 2 www.nature.com/reviews/neuro

P E R S P E C T I V E S

changes have now been observed with olfac-tory conditioning in mice49. In sheep, thesechanges seem to be under the influence ofgonadal steroids50, but it is otherwise not yetclear how the process in sheep is distinctfrom olfactory learning that does not involvean enduring, selective attachment to theyoung. Specifically, we do not know how, insheep, the ‘imprinting’ of the lamb’s odourpermanently opens a sensory gate that signalsacceptance to the ewe, whereas rejectionremains her response to all other lambs. Thepermanent experience-dependent changes inthe post-partum ewe offer an excellentopportunity for identifying the cellularmechanisms for long-term memory in eitherthe olfactory bulb or the higher order stagesof olfactory processing.

Studies in sheep might not only yieldimportant clues for the cellular mechanismsof long-term memory, they might also holdgreat promise for revealing the neural mecha-nisms of attachment. The key will be to linkthe process of olfactory learning to the moti-vation for maternal care. Somehow theprocess of birth (or VCS) leads to willingnessto interact with a lamb and then within a fewhours, a permanent change in the behaviour-al,physiological and neurochemical responsesto this and no other lamb. The investigationof molecular and cellular changes in the ewe’sbrain during this initial period of interactionwith the lamb, like the process of imprintingin the chick, should reveal critical clues to theneurobiology of attachment.

Adult–adult pair bond formationAbout 5% of mammals are monogamous andbiparental51,52.With the development of quan-titative, operational definitions of variousbehavioural aspects of pair bonding, the voles(microtine rodents) have proven to be excel-lent model species for molecular and cellularstudies of complex social behaviours53. TwoNorth American species have been comparedextensively for neural differences:prairie volesthat are monogamous and montane voles thatfail to form social bonds.

Oxytocin and vasopressin. As oxytocin hasbeen implicated in maternal behaviour in ratsand sheep, it seems plausible that oxytocin orits close relative vasopressin might also beinvolved in adult–adult attachment.Prairievoles normally form pair bonds aftermating53. As copulation (or VCS) releases oxy-tocin and vasopressin54, one possibility is thatthese neuropeptides are involved in theprocess of pair bond formation after mating.Indeed, all of the major behavioural aspects ofmonogamy can be facilitated in the prairie

and act through the main olfactory bulb34.The main event seems to be at the granulecell–mitral cell synapses in the bulb (BOX 1).

The ewe learns the identity of her lambthrough a nitric-oxide-dependent processthat occurs within a few hours of birth,resulting in an increase in the ratio of excita-tory to inhibitory tone in the granulecell–mitral cell synapses in the olfactorybulb48. But this process might not be uniqueto the post-partum ewe, as analogous

lamb odours before birth or in response tounfamiliar lambs post-partum. These changesare reminiscent of the enhancement ofamino-acid release that is observed afterimprinting in chicks.

How does this reorganization of the olfac-tory bulb take place? We know that, as is thecase for olfactory learning in rat pups, affer-ent projections from noradrenaline cells inthe brainstem are critical. We also know thatthe signals from the lamb are largely volatile

PLC

NAcc

PrairieMontanea

PrairieMontane

NAccCombined

CSF

PLC

OTA

CP

Partner

chamber

Neutral

chamber

Stranger

chamber

0

40

80

120

100

80

60

40

20

0

Tiim

e s

pent

in c

ham

ber

(min

s)

Co

nta

ct

tim

e (m

ins)

b

c

*

*

Figure 2 | Oxytocin and social attachment in the monogamous prairie vole female. a | Prairie and

montane voles have different distributions of oxytocin receptors in the brain, particularly in the nucleus

accumbens (NAcc) and the prelimbic cortex (PLC; arrow). b | In the laboratory, pair bonding is assayed

using a partner-preference test. After mating or cohabitation with a ‘partner,’ the experimental vole is

place in a three-chambered arena in which the partner is tethered in one chamber (green) and an

equivalent non-familiar vole or ‘stranger’ is tethered in another chamber (yellow). The experimental animal

has free access to both chambers. After mating, female prairie voles spend more time in the partner’s

chamber (green bar) than in the neutral or stranger’s chamber (yellow bar), indicating a partner

preference. By contrast, mated montane voles show no preference for either the partner or the stranger,

indicating the absence of a mating-induced social attachment. c | Partner preference formation is

blocked in the mating female prairie vole by infusions of oxytocin receptor antagonist (OTA) into the NAcc

as well as the PLC, whereas cerebrospinal fluid (CSF) into either area or OTA infused into the caudate

putamen (CP) had no effect on partner preference formation. OTA had no effect on mating. (Panel c

modified from REF. 65).

© 2001 Macmillan Magazines Ltd

Montane voles are solitary, prairie voles show partner preference.

The 3-chamber setup being used by Zuoxin Wang in the Insel lab.

Oxytocin and vasopressin:Oxytocin and vasopressin are ancient neuropeptides that are at least 700 milllion years old. They (or close analogs) are found in such widely disparate species such as hydra, worms, insects, and vertebrates. What is a hydra, you might ask? Hydra is a genus of freshwater multicellular organisms that are typically about 1 mm in length. To give you a sense of how long ago 700 million years is, recall that dinosaurs roamed the earth 60 million years ago. 700 million years ago (that’s pre-Cambrian!), there was only a single supercontinent, the earth was covered by glacial ice sheets even at the equator, and the first multicellular organisms are thought to have emerged. That’s pretty ancient!

Oxytocin and vasopressin are both 9 amino acids in length, and differ by only 2 amino acides. Vasopressin influences male sociosexual behaviors: erection, ejaculation, ag-gression, territoriality, and pair bonding, from rodents to humans. Oxytocin influences female sociosexual behaviors: intercourse, childbirth, lactation, maternal bonding, pair bonding. At the end of pregnancy, a surge of oxytocin starts labor contractions. You may have heard of Pitocin, the trade name for oxytocin. I was introduced to Pitocin when the doctors gave my wife an I.V. drip of it to induce labor. It’s very effective at this! Oxytocin is also released during suckling, and acts to stimulate lactation (the “let-down” reflex). Injecting oxytocin into the spinal cord of male rats causes spontaneous erections. Clearly, these versatile hormones have a wide variety of sociosexual actions. We now know that they are also involved in pair bonding in voles (and perhaps us as well).

Neural mechanisms of pair bonding in voles 3

Oxytocin and vasopressin are ancient neuropeptides that are at least 700 million years old and are found in hydra, worms, insects, and vertebrates.

Effects of oxytocin and vaso-pressin in pair-bonding:

In the 1990’s, a series of experiments from the Insel lab showed that oxyto-cin and vasopressin are involved in the formation of pair bonding in prai-rie voles. Insel injected vasopressin into the ventral forebrain of male prai-rie voles, and then housed the male with an unreceptive female. Normally, since the animals do not mate when the female is unreceptive, no bonding would occur. But the vasopressin in-duced the formation of a pair bond, which could be measured with the partner preference test. After vaso-pressin, males spent much more time with the partner than with a stranger. Control injections of cerebrospinal fluid (CSF) did not produce any bond-ing, and neither did oxytocin: after these control injections, males spent the same amount of time with either

the partner or the stranger.

Neural mechanisms of pair bonding in voles 4

Injecting vasopressin into male brains induces partner preference.

Injecting oxytocin into female brains induces partner preference.

What happens if you inject oxytocin into female vole brains? Although oxytocin had no effect on males, it caused pair bonding in females. Before mating, when normally no bonding has occurred, oxytocin caused a huge increase in partner preference. Then they took a different set of females, and let them mate with males in order to form nor-mal pair bonds. In some of these females, they injected an oxytocin antagonist that blocks the effect of oxytocin naturally released during mating. The antagonist prevented the formation of pair bonding, and these females showed no partner preference, spend-ing equal amounts of time with the partner and stranger.

Insel’s group also demonstrated that injecting males with a vasopressin antagonist pre-vents pair bonding in males. So what do these experiments demonstrate about the roles of oxytocin and vasopressin in pair bonding? Are the hormones necessary? Are they sufficient?

Ventral forebrain:Where are the receptors for these powerful hormones? We’ll focus on va-sopressin, since that’s the one manipu-lated by Lim et al. Insel’s group has demonstrated that of the several vaso-pressin receptors, only one (V1aR) ap-pears to be involved in pair bonding. The others mediate a huge range of other actions, from regulating urine ex-cretion in the kidney to causing the re-lease of stress hormones from the ad-renal glands. V1aR vasopressin recep-tor expression is very strong in the ven-tral forebrain in prairie voles, but not in meadow voles.

Can we get any more specific than “ventral forebrain,” which is pretty vague? It appears that the relevant area is the ventral pallidum (VP in the figures).

The ventral pallidum is one of the nu-clei of the basal ganglia, which are part of the limbic system. The dorsal part of this structure is called the globus palli-dus, which is latin for “pale globe,” al-though nobody knows who named it or understands how it got that name, since it isn’t shaped like a globe at all. More recently, people have proposed simply calling it the pallidum. The ventral part of this structure is therefore referred to as ventral pallidum. As if this terminology weren’t arbitrary enough,

Neural mechanisms of pair bonding in voles 5

the ventral pallidum lies within a structure with my all-time favorite name, the substantia innominata (which is Latin for un-named substance).

Now we have all the pieces in place to understand what Lim et al. did. (By the way, these experiments were done by Larry Young’s group at Emory University in Atlanta. Young used to be a postdoc in Insel’s lab). They used one of the molecular techniques we learned about earlier in the course, and built a virus (adeno-associated virus, or AAV) that infects cells and delivers a genetic payload encoding V1aR. By injecting the virus through a needle carefully placed at the coordinates of the ventral pallidum, they tried to cause expression of V1aR where it is not normally expressed in meadow voles (but where it is expressed in prairie voles). In other words, they tried to turn meadow voles into prairie voles. Of course, some of their injections missed the target. But the beauty of this experiment is that afterwards, they could look to see where the receptors were expressed. They used a clever trick to visualize the receptors: they took advan-tage of the fact that the receptors are very, very good at selectively binding to vaso-pressin. So they applied radioactively-tagged vasopressin to brain slices, and then pressed the slices against unexposed film. Where the radiolabelled ligand is bound to the receptors, the radioactivity exposes the film, and you get a nice image of your ex-pression pattern of receptors. In this way they could identify animals for which the injec-tion missed the target. This is a nice control, because it demonstrates just how precise the V1aR expression pattern needs to be. They also used a control virus that didn’t en-code V1aR. Of course, they probably would have preferred to have injections that hit the target every time. But this is a great example of taking an experimental bug and turning it into a feature.

Neural mechanisms of pair bonding in voles 6

Student’s t-test), but not significantly more time with the strangerthan the control animals (P . 0.05, Student’s t-test), indicating thattheir partner preference was not a result of a generalized, non-selective increase in affiliative behaviour.

Although several control animals did seem to display selectiveaffiliation with the mate, none of them were placed within the topfive strongest pair bonds, as defined by the total time spent huddlingwith the partner. However, this illustrates that individual variationdoes exist within the meadow vole population, indicating that thepotential for pair bond formation might be present in promiscuousvole species. Meadow voles do not in fact completely lack V1aRs inthe ventral pallidum, and thus the potential to engage affiliativeneural circuits is in place. One recent study found that a subpopu-lation of meadow voles formed selective affiliations under certainlaboratory conditions, although this phenomenon has not beenreported in other laboratories or in nature15. Our results show thatoverexpression of V1aRs to a degree beyond naturally occurringlevels in meadow voles can shift every individual in the species toform pair bonds.

To determine whether similar proximate neural mechanismsunderlie pair bond formation in V1aR-vpmeadow voles and prairievoles, we next examined the role of dopaminergic neurotrans-mission. Before pairing with the same receptive female for 24 h,all animals were pretreated with 50mg kg21 intraperitoneal eticlo-pride, a D2-receptor antagonist. This dose of eticlopride blockspartner preference formation in prairie voles16. Subsequent partnerpreference testing showed that V1aR-vp animals, like controlanimals, did not significantly prefer the partner over the stranger(P . 0.05, Student’s t-test) (Fig. 4a). In addition, there were nosignificant group differences in the total time spent huddling withthe partner (P . 0.05, Student’s t-test). Analysis of the distributionof individuals within each group indicates that the distribution ofcontrol animals was unaffected by pretreatment with eticlopride,whereas the distribution of V1aR-vp animals was shifted back to the

normal continuum of control meadow voles, with no significantpreference for the partner over that expected from random chance(P . 0.05, x 2 analysis) (Fig. 4b). This shows that pair bondformation in the transgenic meadow voles, as in prairie voles,depends on dopaminergic neurotransmission.Because monogamous-typical behaviours typically include

biparental care in addition to pair bond formation and increasedsociality, we proposed that V1aR overexpression in the ventralpallidum might result in increased paternal care by the males.There were no significant differences between the groups in ‘classic’paternal behaviours, such as the latency to retrieval or the total timespent in physical contact (namely licking, grooming or huddling)with the pups. However, V1aR-vp animals did display increasedsocial approach behaviours towards the pups, including a shorterlatency to approach (P , 0.05, Student’s t-test), more time near thepups (P , 0.05, Student’s t-test), and less time self-grooming(P , 0.005, Student’s t-test) than control animals, which is con-sistent with the finding of increased total social contact observedduring the partner preference test. These results indicate that,although both behaviours are regulated by central V1aR, theunderlying mechanisms of paternal care and pair bond formationprobably use different neural circuits. In prairie voles, the admin-istration of V1aR antagonist into the ventral pallidum selectivelyblocks partner preference4 but not paternal care; similarly, V1aRantagonist in the medial amygdala blocks paternal care but notpartner preference (M.M.L. and L.J.Y., unpublished data). A similarbehavioural dissociation between pair bond formation and paternalcare has been reported in wild subpopulations of prairie voles:Illinois and Kansas prairie voles both form partner preferences, butonly Illinois voles display high levels of paternal care17. Thus, thesame V1aR gene that regulates the suite of monogamous-typicalbehaviours can have different effects on behaviours, depending onthe specific neural circuits that are engaged.What proximate mechanisms might underlie pair bond for-

Figure 1 Comparison of brain neurochemistry and behaviour in prairie and meadowvoles. a, b, Although prairie voles and meadow voles are similar in physical appearance,

prairie voles are highly affiliative as depicted here in ‘huddling’ side by side (a), whereasmeadow voles are solitary (b). c, d, Partner preference test. After mating and cohabitatingwith a female, a male prairie vole tended to spend significantly more time in contact with

the partner (filled columns) than the stranger (open columns) (P , 0.05, Student’s t-test)

(c), whereas meadow voles do not form partner preferences and spent relatively little time

huddling with either female (d). Error bars, standard error. e, f, Autoradiograms of theventral forebrain illustrating the typical prairie vole (e) and meadow vole (f) expressionpattern of V1aR as shown by V1aR autoradiography. g, h, Despite considerable speciesdifferences in V1aR pattern, D2 receptor distribution was broadly similar in prairie

voles (g) and meadow voles (h), as shown by D2 receptor binding in the nucleusaccumbens (NAcc) and the olfactory tubercle (ot). Scale bar, 1 mm.

letters to nature

NATURE |VOL 429 | 17 JUNE 2004 | www.nature.com/nature 755© 2004 Nature Publishing Group

Strong ventral forebrain vasopressin receptor (V1aR) expression in prairie voles but not meadow voles. Dopamine (D2) expression patterns are the same in both species.

prairie vole

meadow vole

V1aR D2

V1aR D2

prairie vole

meadow vole

Neural mechanisms of pair bonding in voles 7

mation? V1aRs are also crucially involved in social memory for-mation. V1aR knockout mice display deficits in individual dis-crimination, whereas overexpression of V1aR in the rat brain byviral vector gene transfer increases the duration of social mem-ory18,19. We propose that, during pair bond formation, the con-current activation of individual recognition and reward pathwaysresults in convergent V1aR and D2 receptor activation in the ventralforebrain, leading to an association between the rewarding nature ofsex and the olfactory signature of the partner, and thus thedevelopment of a conditioned partner preference. Similarly, thelack of V1aR in the ventral forebrain in promiscuous species might

lead to a lack of this selective association of reward with the partner’solfactory signature. In prairie voles, D1 receptors are upregulatedafter mating; the administration of a D1 receptor agonist inhibitspair bond formation, providing a potential mechanism forpreventing the formation of subsequent pair bonds with novelindividuals20.

Although D2 receptor activation is required for pair bondformation in prairie voles, we did not observe any obvious speciesdifferences in D2 receptor patterns in the ventral forebrain (Fig.1g, h). Thus, monogamous social organization might be the resultof the insertion of the V1aR system into this ancient pre-existingreward circuit. Because our manipulation was performed on adultanimals, this suggests that the pre-existing genetic complement andneural circuits underlying behaviour are fundamentally similaramong vole species, and thus a simple alteration in the expressionof a single gene can have a profound impact on behaviour.

The hypothesis that the vast majority of organismal complexityarises from differences in gene regulation, rather than frommutations in the coding regions themselves, has received muchattention recently21. In prairie and pine voles, the 5 0 regulatoryregion of V1aR contains a microsatellite region that is virtuallyabsent from montane and meadow voles9. This species-specificmicrosatellite polymorphism modulates gene expression in a cell-type-dependent manner22. Microsatellite DNA can rapidly contractand expand in length, providing a molecular mechanism for thegeneration of individual variability in brain V1aR patterns withinthe prairie vole species and between vole species23,24. V1aR geneexpression is exceptional in its phylogenetic plasticity in that no twospecies studied have identical V1aR distributions23,25. Thus, theexceptional plasticity of V1aR patterns could result in an alteredbehavioural response to vasopressin, creating diversity in complexsocial behaviours and enabling adaptation to changing socioecolo-gical factors. It is noteworthy that similar genetic polymorphisms inthe human V1aR promoter region exist, and one such allelic varianthas been linked to autism26, raising the possibility that humanV1aRpolymorphismsmight also contribute to human variability in socialbehaviour.

We have shown that changes in the regional expression of a singlegene can have a profound effect on the social behaviour ofindividuals within a species. Given a population of voles withvariability in V1aR patterns and selection pressure for pair bondformation, one can see howmonogamous social organization could

Figure 2 V1aR autoradiography at the level of the ventral pallidum. a, Meadow vole

overexpressing the V1aR gene in the ventral pallidum by AAV-mediated gene transfer

(V1aR-vp). b, Meadow vole infused with the AAV control vector expressing the lacZ gene

into the ventral pallidum (Ctrl-vp). c, A stereotactic injection inadvertently placed too

rostral to the ventral pallidum, in this case located just ventral to the nucleus accumbens.

d, A stereotactic injection placed too caudal to the ventral pallidum, in this case just

ventral to the fornix. Arrows depict ectopic AAV-mediated V1aR expression in c and d.Animals with AAV vector placement outside the ventral pallidum were placed in a second

control group (Ctrl-other). Scale bar, 1 mm.

Figure 3 Partner preference test. a, V1aR-vp meadow voles spent significantly more time

huddling with the partner (filled column) than the stranger (open column), whereas control

animals (Ctrl-vp) and stereotactic misses (Ctrl-other) did not (P , 0.01, Student’s t-test).

Error bars, standard error. b, A plot of the percentage of time spent with the partner for

each subject indicates a shift from randomly distributed preferences in the control groups

to 100% of animals preferring the partner in the V1aR-vp group (P , 0.001, x 2

analysis). The y axis was calculated as the time spent huddling with the partner divided by

the total time spent huddling with the partner and stranger, multiplied by 100.

Figure 4 Partner preference test after eticlopride pretreatment. a, Eticlopride blockedpartner preference in the V1aR-vp group. V1aR-vp, Ctrl-vp and Ctrl-other animals did not

spend significantly more time with either the partner (filled columns) or the stranger (open

columns) (P . 0.05, Student’s t-test). Error bars, standard error. b, A plot of the

percentage of time spent with the partner for each subject shows a similar distribution

between control and V1aR-vp groups, with no significant difference from that expected by

random chance (P . 0.05, x 2 analysis).

letters to nature

NATURE | VOL 429 | 17 JUNE 2004 | www.nature.com/nature756 © 2004 Nature Publishing Group

Using a virus in meadow voles to produce V1aR expression that mimics that in prai-rie voles.

hit (good injection)

control (injection miss)

control virus without V1aR

control (injection miss)

mation? V1aRs are also crucially involved in social memory for-mation. V1aR knockout mice display deficits in individual dis-crimination, whereas overexpression of V1aR in the rat brain byviral vector gene transfer increases the duration of social mem-ory18,19. We propose that, during pair bond formation, the con-current activation of individual recognition and reward pathwaysresults in convergent V1aR and D2 receptor activation in the ventralforebrain, leading to an association between the rewarding nature ofsex and the olfactory signature of the partner, and thus thedevelopment of a conditioned partner preference. Similarly, thelack of V1aR in the ventral forebrain in promiscuous species might

lead to a lack of this selective association of reward with the partner’solfactory signature. In prairie voles, D1 receptors are upregulatedafter mating; the administration of a D1 receptor agonist inhibitspair bond formation, providing a potential mechanism forpreventing the formation of subsequent pair bonds with novelindividuals20.

Although D2 receptor activation is required for pair bondformation in prairie voles, we did not observe any obvious speciesdifferences in D2 receptor patterns in the ventral forebrain (Fig.1g, h). Thus, monogamous social organization might be the resultof the insertion of the V1aR system into this ancient pre-existingreward circuit. Because our manipulation was performed on adultanimals, this suggests that the pre-existing genetic complement andneural circuits underlying behaviour are fundamentally similaramong vole species, and thus a simple alteration in the expressionof a single gene can have a profound impact on behaviour.

The hypothesis that the vast majority of organismal complexityarises from differences in gene regulation, rather than frommutations in the coding regions themselves, has received muchattention recently21. In prairie and pine voles, the 5 0 regulatoryregion of V1aR contains a microsatellite region that is virtuallyabsent from montane and meadow voles9. This species-specificmicrosatellite polymorphism modulates gene expression in a cell-type-dependent manner22. Microsatellite DNA can rapidly contractand expand in length, providing a molecular mechanism for thegeneration of individual variability in brain V1aR patterns withinthe prairie vole species and between vole species23,24. V1aR geneexpression is exceptional in its phylogenetic plasticity in that no twospecies studied have identical V1aR distributions23,25. Thus, theexceptional plasticity of V1aR patterns could result in an alteredbehavioural response to vasopressin, creating diversity in complexsocial behaviours and enabling adaptation to changing socioecolo-gical factors. It is noteworthy that similar genetic polymorphisms inthe human V1aR promoter region exist, and one such allelic varianthas been linked to autism26, raising the possibility that humanV1aRpolymorphismsmight also contribute to human variability in socialbehaviour.

We have shown that changes in the regional expression of a singlegene can have a profound effect on the social behaviour ofindividuals within a species. Given a population of voles withvariability in V1aR patterns and selection pressure for pair bondformation, one can see howmonogamous social organization could

Figure 2 V1aR autoradiography at the level of the ventral pallidum. a, Meadow vole

overexpressing the V1aR gene in the ventral pallidum by AAV-mediated gene transfer

(V1aR-vp). b, Meadow vole infused with the AAV control vector expressing the lacZ gene

into the ventral pallidum (Ctrl-vp). c, A stereotactic injection inadvertently placed too

rostral to the ventral pallidum, in this case located just ventral to the nucleus accumbens.

d, A stereotactic injection placed too caudal to the ventral pallidum, in this case just

ventral to the fornix. Arrows depict ectopic AAV-mediated V1aR expression in c and d.Animals with AAV vector placement outside the ventral pallidum were placed in a second

control group (Ctrl-other). Scale bar, 1 mm.

Figure 3 Partner preference test. a, V1aR-vp meadow voles spent significantly more time

huddling with the partner (filled column) than the stranger (open column), whereas control

animals (Ctrl-vp) and stereotactic misses (Ctrl-other) did not (P , 0.01, Student’s t-test).

Error bars, standard error. b, A plot of the percentage of time spent with the partner for

each subject indicates a shift from randomly distributed preferences in the control groups

to 100% of animals preferring the partner in the V1aR-vp group (P , 0.001, x 2

analysis). The y axis was calculated as the time spent huddling with the partner divided by

the total time spent huddling with the partner and stranger, multiplied by 100.

Figure 4 Partner preference test after eticlopride pretreatment. a, Eticlopride blockedpartner preference in the V1aR-vp group. V1aR-vp, Ctrl-vp and Ctrl-other animals did not

spend significantly more time with either the partner (filled columns) or the stranger (open

columns) (P . 0.05, Student’s t-test). Error bars, standard error. b, A plot of the

percentage of time spent with the partner for each subject shows a similar distribution

between control and V1aR-vp groups, with no significant difference from that expected by

random chance (P . 0.05, x 2 analysis).

letters to nature

NATURE | VOL 429 | 17 JUNE 2004 | www.nature.com/nature756 © 2004 Nature Publishing Group

Student’s t-test), but not significantly more time with the strangerthan the control animals (P . 0.05, Student’s t-test), indicating thattheir partner preference was not a result of a generalized, non-selective increase in affiliative behaviour.

Although several control animals did seem to display selectiveaffiliation with the mate, none of them were placed within the topfive strongest pair bonds, as defined by the total time spent huddlingwith the partner. However, this illustrates that individual variationdoes exist within the meadow vole population, indicating that thepotential for pair bond formation might be present in promiscuousvole species. Meadow voles do not in fact completely lack V1aRs inthe ventral pallidum, and thus the potential to engage affiliativeneural circuits is in place. One recent study found that a subpopu-lation of meadow voles formed selective affiliations under certainlaboratory conditions, although this phenomenon has not beenreported in other laboratories or in nature15. Our results show thatoverexpression of V1aRs to a degree beyond naturally occurringlevels in meadow voles can shift every individual in the species toform pair bonds.

To determine whether similar proximate neural mechanismsunderlie pair bond formation in V1aR-vpmeadow voles and prairievoles, we next examined the role of dopaminergic neurotrans-mission. Before pairing with the same receptive female for 24 h,all animals were pretreated with 50mg kg21 intraperitoneal eticlo-pride, a D2-receptor antagonist. This dose of eticlopride blockspartner preference formation in prairie voles16. Subsequent partnerpreference testing showed that V1aR-vp animals, like controlanimals, did not significantly prefer the partner over the stranger(P . 0.05, Student’s t-test) (Fig. 4a). In addition, there were nosignificant group differences in the total time spent huddling withthe partner (P . 0.05, Student’s t-test). Analysis of the distributionof individuals within each group indicates that the distribution ofcontrol animals was unaffected by pretreatment with eticlopride,whereas the distribution of V1aR-vp animals was shifted back to the

normal continuum of control meadow voles, with no significantpreference for the partner over that expected from random chance(P . 0.05, x 2 analysis) (Fig. 4b). This shows that pair bondformation in the transgenic meadow voles, as in prairie voles,depends on dopaminergic neurotransmission.Because monogamous-typical behaviours typically include

biparental care in addition to pair bond formation and increasedsociality, we proposed that V1aR overexpression in the ventralpallidum might result in increased paternal care by the males.There were no significant differences between the groups in ‘classic’paternal behaviours, such as the latency to retrieval or the total timespent in physical contact (namely licking, grooming or huddling)with the pups. However, V1aR-vp animals did display increasedsocial approach behaviours towards the pups, including a shorterlatency to approach (P , 0.05, Student’s t-test), more time near thepups (P , 0.05, Student’s t-test), and less time self-grooming(P , 0.005, Student’s t-test) than control animals, which is con-sistent with the finding of increased total social contact observedduring the partner preference test. These results indicate that,although both behaviours are regulated by central V1aR, theunderlying mechanisms of paternal care and pair bond formationprobably use different neural circuits. In prairie voles, the admin-istration of V1aR antagonist into the ventral pallidum selectivelyblocks partner preference4 but not paternal care; similarly, V1aRantagonist in the medial amygdala blocks paternal care but notpartner preference (M.M.L. and L.J.Y., unpublished data). A similarbehavioural dissociation between pair bond formation and paternalcare has been reported in wild subpopulations of prairie voles:Illinois and Kansas prairie voles both form partner preferences, butonly Illinois voles display high levels of paternal care17. Thus, thesame V1aR gene that regulates the suite of monogamous-typicalbehaviours can have different effects on behaviours, depending onthe specific neural circuits that are engaged.What proximate mechanisms might underlie pair bond for-

Figure 1 Comparison of brain neurochemistry and behaviour in prairie and meadowvoles. a, b, Although prairie voles and meadow voles are similar in physical appearance,

prairie voles are highly affiliative as depicted here in ‘huddling’ side by side (a), whereasmeadow voles are solitary (b). c, d, Partner preference test. After mating and cohabitatingwith a female, a male prairie vole tended to spend significantly more time in contact with

the partner (filled columns) than the stranger (open columns) (P , 0.05, Student’s t-test)

(c), whereas meadow voles do not form partner preferences and spent relatively little time

huddling with either female (d). Error bars, standard error. e, f, Autoradiograms of theventral forebrain illustrating the typical prairie vole (e) and meadow vole (f) expressionpattern of V1aR as shown by V1aR autoradiography. g, h, Despite considerable speciesdifferences in V1aR pattern, D2 receptor distribution was broadly similar in prairie

voles (g) and meadow voles (h), as shown by D2 receptor binding in the nucleusaccumbens (NAcc) and the olfactory tubercle (ot). Scale bar, 1 mm.

letters to nature

NATURE |VOL 429 | 17 JUNE 2004 | www.nature.com/nature 755© 2004 Nature Publishing Group

Are you convinced?

prairie vole

meadow vole

meadow vole after virus injection

Neural mechanisms of pair bonding in voles 8

V1aR-vp voles spend more time huddling with partner than with stranger. Controls did not.

mation? V1aRs are also crucially involved in social memory for-mation. V1aR knockout mice display deficits in individual dis-crimination, whereas overexpression of V1aR in the rat brain byviral vector gene transfer increases the duration of social mem-ory18,19. We propose that, during pair bond formation, the con-current activation of individual recognition and reward pathwaysresults in convergent V1aR and D2 receptor activation in the ventralforebrain, leading to an association between the rewarding nature ofsex and the olfactory signature of the partner, and thus thedevelopment of a conditioned partner preference. Similarly, thelack of V1aR in the ventral forebrain in promiscuous species might

lead to a lack of this selective association of reward with the partner’solfactory signature. In prairie voles, D1 receptors are upregulatedafter mating; the administration of a D1 receptor agonist inhibitspair bond formation, providing a potential mechanism forpreventing the formation of subsequent pair bonds with novelindividuals20.

Although D2 receptor activation is required for pair bondformation in prairie voles, we did not observe any obvious speciesdifferences in D2 receptor patterns in the ventral forebrain (Fig.1g, h). Thus, monogamous social organization might be the resultof the insertion of the V1aR system into this ancient pre-existingreward circuit. Because our manipulation was performed on adultanimals, this suggests that the pre-existing genetic complement andneural circuits underlying behaviour are fundamentally similaramong vole species, and thus a simple alteration in the expressionof a single gene can have a profound impact on behaviour.

The hypothesis that the vast majority of organismal complexityarises from differences in gene regulation, rather than frommutations in the coding regions themselves, has received muchattention recently21. In prairie and pine voles, the 5 0 regulatoryregion of V1aR contains a microsatellite region that is virtuallyabsent from montane and meadow voles9. This species-specificmicrosatellite polymorphism modulates gene expression in a cell-type-dependent manner22. Microsatellite DNA can rapidly contractand expand in length, providing a molecular mechanism for thegeneration of individual variability in brain V1aR patterns withinthe prairie vole species and between vole species23,24. V1aR geneexpression is exceptional in its phylogenetic plasticity in that no twospecies studied have identical V1aR distributions23,25. Thus, theexceptional plasticity of V1aR patterns could result in an alteredbehavioural response to vasopressin, creating diversity in complexsocial behaviours and enabling adaptation to changing socioecolo-gical factors. It is noteworthy that similar genetic polymorphisms inthe human V1aR promoter region exist, and one such allelic varianthas been linked to autism26, raising the possibility that humanV1aRpolymorphismsmight also contribute to human variability in socialbehaviour.

We have shown that changes in the regional expression of a singlegene can have a profound effect on the social behaviour ofindividuals within a species. Given a population of voles withvariability in V1aR patterns and selection pressure for pair bondformation, one can see howmonogamous social organization could

Figure 2 V1aR autoradiography at the level of the ventral pallidum. a, Meadow vole

overexpressing the V1aR gene in the ventral pallidum by AAV-mediated gene transfer

(V1aR-vp). b, Meadow vole infused with the AAV control vector expressing the lacZ gene

into the ventral pallidum (Ctrl-vp). c, A stereotactic injection inadvertently placed too

rostral to the ventral pallidum, in this case located just ventral to the nucleus accumbens.

d, A stereotactic injection placed too caudal to the ventral pallidum, in this case just

ventral to the fornix. Arrows depict ectopic AAV-mediated V1aR expression in c and d.Animals with AAV vector placement outside the ventral pallidum were placed in a second

control group (Ctrl-other). Scale bar, 1 mm.

Figure 3 Partner preference test. a, V1aR-vp meadow voles spent significantly more time

huddling with the partner (filled column) than the stranger (open column), whereas control

animals (Ctrl-vp) and stereotactic misses (Ctrl-other) did not (P , 0.01, Student’s t-test).

Error bars, standard error. b, A plot of the percentage of time spent with the partner for

each subject indicates a shift from randomly distributed preferences in the control groups

to 100% of animals preferring the partner in the V1aR-vp group (P , 0.001, x 2

analysis). The y axis was calculated as the time spent huddling with the partner divided by

the total time spent huddling with the partner and stranger, multiplied by 100.

Figure 4 Partner preference test after eticlopride pretreatment. a, Eticlopride blockedpartner preference in the V1aR-vp group. V1aR-vp, Ctrl-vp and Ctrl-other animals did not

spend significantly more time with either the partner (filled columns) or the stranger (open

columns) (P . 0.05, Student’s t-test). Error bars, standard error. b, A plot of the

percentage of time spent with the partner for each subject shows a similar distribution

between control and V1aR-vp groups, with no significant difference from that expected by

random chance (P . 0.05, x 2 analysis).

letters to nature

NATURE | VOL 429 | 17 JUNE 2004 | www.nature.com/nature756 © 2004 Nature Publishing Group

100% of V1aR-vp animals spent more time with partner, but control animal preferences were randomly distributed. This is another way of looking at the same data. It provides a fuller picture than the bar graphs.

So what did they find? After they overexpressed V1aR in the ventral forebrain of male meadow voles, they paired each male with a receptive female for 24 hours. Afterwards they ran the partner preference test, to compare how much time the males spent with the partner or the stranger. Normal meadow voles, and those with control injections, spent equal time huddling with either the partner and the stranger, and spent more time alone in the neutral chamber. But the meadow voles with the injections acted like prairie voles. They spent much more time with the partner. This is strong evidence that they had formed a pair bond.

In prairie voles, pair bonding can be blocked by pretreating them with a dopamine an-tagonist before pairing. This indicates that dopaminergic transmission is necessary for pair bond formation (although, since the antagonist was delivered systemically, we don’t know where the critical dopamine transmission is taking place). With the genetically al-tered meadow voles, pair bonding is also susceptible to dopaminergic blockers. This suggests that pair bonding in both prairie voles and transgenic meadow voles depends on dopaminergic signalling. We read last time about the role of dopamine as an error signal in reinforcement learning, which causes animals to place a high value on the states of the world that lead up to reward. The fact that pair bonding depends on dopa-mine suggests that reward pathways may act to increase partner preference by means of reinforcement learning. Perhaps love, in addition to being mediated by vasopressin V1aR receptors in the ventral pallidum, can also be expressed as a highly valued state that is the end product of temporal difference reinforcement learning.

Neural mechanisms of pair bonding in voles 9

mation? V1aRs are also crucially involved in social memory for-mation. V1aR knockout mice display deficits in individual dis-crimination, whereas overexpression of V1aR in the rat brain byviral vector gene transfer increases the duration of social mem-ory18,19. We propose that, during pair bond formation, the con-current activation of individual recognition and reward pathwaysresults in convergent V1aR and D2 receptor activation in the ventralforebrain, leading to an association between the rewarding nature ofsex and the olfactory signature of the partner, and thus thedevelopment of a conditioned partner preference. Similarly, thelack of V1aR in the ventral forebrain in promiscuous species might

lead to a lack of this selective association of reward with the partner’solfactory signature. In prairie voles, D1 receptors are upregulatedafter mating; the administration of a D1 receptor agonist inhibitspair bond formation, providing a potential mechanism forpreventing the formation of subsequent pair bonds with novelindividuals20.

Although D2 receptor activation is required for pair bondformation in prairie voles, we did not observe any obvious speciesdifferences in D2 receptor patterns in the ventral forebrain (Fig.1g, h). Thus, monogamous social organization might be the resultof the insertion of the V1aR system into this ancient pre-existingreward circuit. Because our manipulation was performed on adultanimals, this suggests that the pre-existing genetic complement andneural circuits underlying behaviour are fundamentally similaramong vole species, and thus a simple alteration in the expressionof a single gene can have a profound impact on behaviour.

The hypothesis that the vast majority of organismal complexityarises from differences in gene regulation, rather than frommutations in the coding regions themselves, has received muchattention recently21. In prairie and pine voles, the 5 0 regulatoryregion of V1aR contains a microsatellite region that is virtuallyabsent from montane and meadow voles9. This species-specificmicrosatellite polymorphism modulates gene expression in a cell-type-dependent manner22. Microsatellite DNA can rapidly contractand expand in length, providing a molecular mechanism for thegeneration of individual variability in brain V1aR patterns withinthe prairie vole species and between vole species23,24. V1aR geneexpression is exceptional in its phylogenetic plasticity in that no twospecies studied have identical V1aR distributions23,25. Thus, theexceptional plasticity of V1aR patterns could result in an alteredbehavioural response to vasopressin, creating diversity in complexsocial behaviours and enabling adaptation to changing socioecolo-gical factors. It is noteworthy that similar genetic polymorphisms inthe human V1aR promoter region exist, and one such allelic varianthas been linked to autism26, raising the possibility that humanV1aRpolymorphismsmight also contribute to human variability in socialbehaviour.

We have shown that changes in the regional expression of a singlegene can have a profound effect on the social behaviour ofindividuals within a species. Given a population of voles withvariability in V1aR patterns and selection pressure for pair bondformation, one can see howmonogamous social organization could

Figure 2 V1aR autoradiography at the level of the ventral pallidum. a, Meadow vole

overexpressing the V1aR gene in the ventral pallidum by AAV-mediated gene transfer

(V1aR-vp). b, Meadow vole infused with the AAV control vector expressing the lacZ gene

into the ventral pallidum (Ctrl-vp). c, A stereotactic injection inadvertently placed too

rostral to the ventral pallidum, in this case located just ventral to the nucleus accumbens.

d, A stereotactic injection placed too caudal to the ventral pallidum, in this case just

ventral to the fornix. Arrows depict ectopic AAV-mediated V1aR expression in c and d.Animals with AAV vector placement outside the ventral pallidum were placed in a second

control group (Ctrl-other). Scale bar, 1 mm.

Figure 3 Partner preference test. a, V1aR-vp meadow voles spent significantly more time

huddling with the partner (filled column) than the stranger (open column), whereas control

animals (Ctrl-vp) and stereotactic misses (Ctrl-other) did not (P , 0.01, Student’s t-test).

Error bars, standard error. b, A plot of the percentage of time spent with the partner for

each subject indicates a shift from randomly distributed preferences in the control groups

to 100% of animals preferring the partner in the V1aR-vp group (P , 0.001, x 2

analysis). The y axis was calculated as the time spent huddling with the partner divided by

the total time spent huddling with the partner and stranger, multiplied by 100.

Figure 4 Partner preference test after eticlopride pretreatment. a, Eticlopride blockedpartner preference in the V1aR-vp group. V1aR-vp, Ctrl-vp and Ctrl-other animals did not

spend significantly more time with either the partner (filled columns) or the stranger (open

columns) (P . 0.05, Student’s t-test). Error bars, standard error. b, A plot of the

percentage of time spent with the partner for each subject shows a similar distribution

between control and V1aR-vp groups, with no significant difference from that expected by

random chance (P . 0.05, x 2 analysis).

letters to nature

NATURE | VOL 429 | 17 JUNE 2004 | www.nature.com/nature756 © 2004 Nature Publishing Group

Pretreating animals with a D2 dopamine antagonist prevents pair bonding (forma-tion of partner preference). What does this mean about the role of dopamine in pair bonding?