A genetic variant BDNF polymorphism alters extinction learning in both mouse and human.

Abstract

Mouse models are useful for studying genes involved in behavior, but whether they are relevant to human behavior is unclear. Here, we identified parallel phenotypes in mice and humans resulting from a common single-nucleotide polymorphism in the brain-derived neurotrophic factor (BDNF) gene, which is involved in anxiety-related behavior. An inbred genetic knock-in mouse strain expressing the variant BDNF recapitulated the phenotypic effects of the human polymorphism. Both were impaired in extinguishing a conditioned fear response, which was paralleled by atypical frontoamygdala activity in humans. Thus, this variant BDNF allele may play a role in anxiety disorders showing impaired learning of cues that signal safety versus threat and in the efficacy of treatments that rely on extinction mechanisms, such as exposure therapy.

Full text

Page 1

identical MIS 5e/5a relative sea-level histories
oftectonicallystableBermudaandMallorca.The
very rapid onset and relatively brief nature of the
MIS 5a highstand may have plausibly generated
lags between the timing of sea-level changes and
the timing of coral reef growth, and may provide
a partialexplanation as to why reefs on Barbados
and New Guinea do not record a comparable
eustatic height for this event. This and other
factors that could be part of the apparent
discrepancy are discussed in (9).
The suggestion that MIS 5a sea level was
slightly higher than at present and only slightly
lowerthantheMIS5esealevelimpliesthatmost
of the ice built up during MIS 5b would have
melted during the onset of MIS 5a. The ~84- to
80-ka timing of this highstand closely matches
the June 60°N insolation peak at ~84 ka (Fig.
2E), a pattern that is consistent with the
Milankovitch model. In fact, June insolation at
60°N was higher at ~84 ka than that at 11 ka (27),
and field studies in the Baffin Island region
suggest the complete melting of the Laurentide
Ice Sheet around 80 ka (28). Finally, we find
additional,independentsupportfor anear-modern
eustatic MIS 5a highstand when we consider the
indirectsea-levelestimateof (29) inferred from a
Pacific benthic d18O record, the Vostok atmo-
spheric d18O record, and certain assumptions
about the Dole effect on deep-water temper-
atures (Fig. 2D). The premise of the approach of
(29) is that the deep-sea d18O record does not
capture the true magnitude of eustatic sea-level
change, because the d18O signal is partially
controlled by temperature.
Because of its relation to continental ice vol-
ume, an accurate Quaternary sea-level curve has
been a long-term goal of scientists interested in
ice-age cycles and their causes. Ice-age theory
has long assumed gradual ice buildup and more
rapid ice melting in the generally acceptedmodel
of the ~100-ky cycle of glaciation. Instead, the
emerging body of evidence suggests that both
melting and accumulation can be very rapid dur-
ing discrete intervals of time when specific con-
ditionsprevail.Furthermore,the100-kymodelof
glaciation has always faced the problem that al-
though thedeep-sea d18O record isdominated by
a 100-ky cycle, northern high-latitude summer
insolation has negligible power in this band. Our
data from Mallorca and data from other sites
around the world indicate the possibility that eu-
static sea level was near modern levels at ~80 ka.
If this is true, the 100-ky cycle so universally
accepted as the main rhythm of the Middle and
Late Quaternary glaciations, in fact, applies
rather poorly to ice growth and decay, but much
better to carbon dioxide, methane, and temper-
atures recorded by polar ice (30).
References and Notes
1. P. J. Hearty, Quat. Sci. Rev. 6, 245 (1987).
2. C. D. Gallup, R. L. Edwards, R. G. Johnson, Science 263,
796 (1994).
3. K. Lambeck, J. Chappell, Science 292, 679 (2001).
4. D. R. Muhs, K. R. Simmons, B. Steinke, Quat. Sci. Rev.
21, 1355 (2002).
5. E. K. Potter, K. Lambeck, Earth Planet. Sci. Lett. 217, 171
(2003).
6. J. F. Wehmiller et al., Quaternary Int. 120, 3 (2004).
7. J. Ginés, Endins 20, 71 (1995).
8. P. Tuccimei et al., Earth Surf. Process. Landf. 35, (2010).
9. Materials, methods, and additional discussion are available
as supporting material on Science Online.
10. P. Tuccimei et al., Z. Geomorphol. 50, 1 (2006).
11. R. E. Dodge, R. G. Fairbanks, L. K. Benninger, F. Maurrasse,
Science 219, 1423 (1983).
12. T. M. Esat, M. T. McCulloch, J. Chappell, B. Pillans,
A. Omura, Science 283, 197 (1999).
13. W. X. Li et al., Nature 339, 534 (1989).
14. M. A. Toscano, J. Lundberg, Quat. Sci. Rev. 18, 753 (1999).
15. J. X. Mitrovica, G. A. Milne, Quat. Sci. Rev. 21, 2179
(2002).
16. P. A. Pirazzoli, Quat. Sci. Rev. 24, 1989 (2005).
17. K. Lambeck, M. Azidei, F. Antonioli, A. Benini,
A. Esposito, Earth Planet. Sci. Lett. 224, 563 (2004).
18. W. R. Peltier, Annu. Rev. Earth Planet. Sci. 32, 111 (2004).
19. P. Stocchi, G. Spada, Ann. Geophys. 50, 741 (2007).
20. E.J.Hodge,D.A.Richards,P.L.Smart,A.Ginés,D.P.Mattey,
J. Quat. Sci. 23, 713 (2008).
21. R. L. Edwards et al., Science 260, 962 (1993).
22. P. J. Hearty, Quat. Sci. Rev. 17, 333 (1998).
23. K. R. Ludwig, D. R. Muhs, K. R. Simmons, R. B. Halley,
E. A. Shinn, Geology 24, 211 (1996).
24. H. L. Vacher, P. Hearty, Quat. Sci. Rev. 8, 159 (1989).
25. M.K.Coyne,B.Jones,D.Ford,Quat.Sci.Rev.26,536(2007).
26. D. R. Muhs, J. F. Wehmiller, K. R. Simmons, L. L. York,
in The Quaternary Period in the United States,
A. R. Gillespie, S. C. Porter, B. F. Atwater, Eds.
(Elsevier, Amsterdam, 2004), pp. 147–183.
27. A. Berger, M. F. Loutre, Quat. Sci. Rev. 10, 297 (1991).
28. G. H. Miller et al., Quat. Sci. Rev. 18, 789 (1999).
29. N. J. Shackleton, Science 289, 1897 (2000).
30. J. R. Toggweiler, Paleoceanography 23, PA2211 (2008).
31. This material is based on work supported by NSF
(grant OISE-0826667 to B.P.O. and J.A.D.), the University
of South Florida (grant R058889) to B.P.O., and the
MICINN-FEDER Projects CGL2006-11242-C03-01 and
CGL2009-07392 of the Spanish Government to J.J.F.
We thank L. Vacher, V. Polyak, and E. A. Bettis III for
stimulating discussions, three anonymous reviewers for
insightful criticism that considerably improved the
manuscript, and F. Gràcia and K. Downey for some
of the pictures.
Supporting Online Material
www.sciencemag.org/cgi/content/full/327/5967/860/DC1
Materials and Methods
SOM Text
References
9 September 2009; accepted 7 January 2010
10.1126/science.1181725
A Genetic Variant BDNF Polymorphism
Alters Extinction Learning in Both
Mouse and Human
Fatima Soliman,1,2* Charles E. Glatt,2Kevin G. Bath,2Liat Levita,1,2Rebecca M. Jones,1,2
Siobhan S. Pattwell,2Deqiang Jing,2Nim Tottenham,1,2Dima Amso,1,2Leah H. Somerville,1,2
Henning U. Voss,3Gary Glover,4Douglas J. Ballon,3Conor Liston,1,2Theresa Teslovich,1,2
Tracey Van Kempen,1,2Francis S. Lee,2* B. J. Casey1,2*
Mouse models are useful for studying genes involved in behavior, but whether they are relevant
to human behavior is unclear. Here, we identified parallel phenotypes in mice and humans
resulting from a common single-nucleotide polymorphism in the brain-derived neurotrophic factor
(BDNF) gene, which is involved in anxiety-related behavior. An inbred genetic knock-in mouse
strain expressing the variant BDNF recapitulated the phenotypic effects of the human
polymorphism. Both were impaired in extinguishing a conditioned fear response, which was
paralleled by atypical frontoamygdala activity in humans. Thus, this variant BDNF allele may play a
role in anxiety disorders showing impaired learning of cues that signal safety versus threat and in
the efficacy of treatments that rely on extinction mechanisms, such as exposure therapy.
G
enetically modified mice provide useful
model systems for testing the role of can-
didate genes in behavior. The extent to
which such genetic manipulations in the mouse
andtheresultingphenotypecanbetranslatedacross
species,from mouse to human, is less clear. In this
report, we focused on identifying biologically
valid phenotypes across species. We utilized a
common single-nucleotide polymorphism (SNP)
in the brain-derived neurotrophic factor (BDNF)
gene that leads to a valine (Val) to methionine
(Met) substitution at codon 66 (Val66Met). In an
inbredgeneticknock-inmousestrainthatexpresses
the variant BDNF allele to recapitulate the spe-
cific phenotypic properties of the human poly-
morphism in vivo, we found the BDNF Val66Met
genotype was associated with treatment-resistant
forms of anxiety-like behavior (1). The objective
ofthisstudywastotestiftheVal66Metgenotype
couldaffectextinctionlearninginourmousemod-
el and whether such findings could be generalized
to human populations.
1The Sackler Institute for Developmental Psychobiology, Weill
Cornell Medical College, New York, NY 10065, USA.
2Department of Psychiatry, Weill Cornell Medical College, New
York, NY 10065, USA.3Citigroup Biomedical Imaging Center,
Department of Radiology, Weill Cornell Medical College, New
York, NY 10065, USA.4Lucas Center for Imaging, Department
of Radiology, Stanford University, Stanford, CA 94305, USA.
*To whom correspondence should be addressed. E-mail:
fas2002@med.cornell.edu (F.S.) or fslee@med.cornell.edu
(F.S.L.) or bjc2002@med.cornell.edu (B.J.C.)
www.sciencemag.org
SCIENCE
VOL 32712 FEBRUARY 2010
863
REPORTS

Page 2

BDNFmediatessynapticplasticityassociated
with learning and memory (2, 3), specifically in
fear learning and extinction (4, 5). BDNF-
dependent forms of fear learning have known
biological substrates and lie at the core of a
number of clinical disorders (6, 7) associated
with the variant BDNF (8–10). Fear-learning
paradigms require the ability to recognize and
remember cues that signal safety or threat and to
extinguish these associations when they no
longer exist. These abilities are impaired in
anxiety disorders such as posttraumatic stress
disorder and phobias (11, 12). Behavioral treat-
ments for these disorders such as exposure
therapy rely on basic principles of extinction
learning (13) inwhich an individualis repeatedly
exposed to an event that was previously asso-
ciated with aversive consequences. Understand-
ing the effect of the BDNF Met allele on these
forms of learning can provide insight into the
mechanism of risk for anxiety disorders, can re-
fineexistingtreatments,andmayleadtogenotype-
based personalized medicine.
We examined the impact of the variant BDNF
on classic fear conditioning and extinction
paradigmsthatwereadaptedtobesuitableforeach
species and that are associated with well-known
underlying biological substrates (14, 15, 16). Fear
conditioning consisted of pairing a neutral cue
with an aversive stimulus. With repeated pairings,
the cue itself takes on properties of the aversive
stimulus as it predicts threat of an impending
aversive event. Extinction consisted of presenting
thecuealonefollowingconditioning,wherebythe
association is diminished with repeated exposure
to empty threat.
We tested 68 mice (17 BDNFVal/Val, 33
BDNFVal/Metand 18 BDNFMet/Met) and 72
humans group-matched for age, gender, and
ethnic background (36 Met allele carriers: 31
Val/Met and 5 Met/Met, and 36 non–Met allele
carriers: Val/Val) (table S1). We found no effect
of the BDNF Met allele on fear conditioning in
miceasmeasuredbythepercentageoftimespent
freezing in response to the conditioned stimulus
(F2,65= 1.58, P < 0.22) (fig. S1A) or on general
fear arousal as measured by freezing during the
intertrialinterval(fig.S2).WegroupedhumanMet
allelecarrierstogether(Val/MetandMet/Met)for
analyses, because the rarity of human Met allele
homozygotes prevents enough observations for
meaningful analysis. As we found in the mouse,
there was no effect of the BDNF Met allele on
fear conditioning in humans as measured by skin
conductance response to the cue predicting the
aversivestimulusrelativetoaneutralcue(F1,70=
0.67, P < 0.42) (fig. S1B).
Analysis of extinction trials showed a main
effect of genotype for both mice [(F2,65= 6.55,
P < 0.003); Val/Val, 48.8 T 2.3; Val/Met, 53.2 T
1.8; Met/Met, 61.3 T 2.8] and humans [(F1,70=
4.86, P < 0.03); Val/Val, 0.32 T 0.03; Val/Met,
0.42 T 0.04], such that extinction learning was
impaired in Met allele carriers relative to non–
Met allele carriers. The Met allele carriers showed
slower extinguishing, as indicated by an inter-
actionoftime×genotypeforthe mouse(F2,65=
6.51, P < 0.003) with no differences in freezing
initially,butadoseresponseoftheMetalleleon
the percentage of freezing behavior during late
trials [Val/Val versus Val/Met: t(48) = –2.62, P <
0.01; Val/Val versus Met/Met: t(33) = –4.78, P <
0.0001; Val/Met versus Met/Met: t(49) = –2.90,
P < 0.006] (Fig. 1A). Humans showed a similar
pattern to the mice with no genotypic difference
in the initial human skin conductance response
during early trials of extinction [t(70) = –1.57,
P <0.12],butsignificantdifferencesbylatetrials
[t(70) = –2.43, P < 0.02, corrected for time] (Fig.
1B). These data demonstrate slower or impaired
extinction related to the Met allele in both mouse
and human.
Thelearningparadigmforhumansincludeda
conditioned stimulus paired with the aversive
stimulus and a neutral stimulus that was not
paired with the aversive stimulus. This design
allowed for distinguishing between effects due to
impaired learning versus a general effect of
heightened anxiety, as generalized heightened
anxiety would lead to a similar response to both
the conditioned and neutral cues. Met allele
carriers had an overall heightened response to
both conditioned and neutral cues [main effect of
genotype (F1,70= 7.21, P < 0.009)], but overall,
they differentiated between the conditioned and
neutral cues similarly to the non–Met allele
carriers (fig. S1B). Yet, when we examined these
effects over time, Met allele carriers took longer
to recognize that the neutral cue was not asso-
ciated with the aversive stimulus, as evidenced
by significant genotypic differences during late
trials [t(70) = –3.46, P < 0.001, corrected for
time] but not early trials [t(70) = –1.44, P < 0.16]
(Fig. 2). Thus, the skin conductance response to
theneutralcueduringfearconditioningshoweda
pattern similar to that observed during extinction
trials (16).
The genetic findings for both fear condi-
tioningandextinctionsuggestthatlearningabout
cues that signal threat of an impending aversive
event is intact in Met allele carriers. However,
learning that cues no longer signal threat (e.g.,
extinction) or do not predict threat (cues not
paired with an aversive stimulus) is impaired in
Met allele carriers, which leads to exaggerated
and longer retention of aversive responses where
they are not warranted.
To provide neuroanatomical evidence to
validate our cross-species translation, we used
human functional magnetic resonance imaging
(MRI)todefinetheunderlyingneuralcircuitryof
the behavioral effects of BDNF Val66Met and to
map them to known circuits involved in fear
learning in the rodent (table S2). We targeted
frontoamygdala circuitry that has been demon-
Fig. 1. Altered extinction in mice and humans with BDNF Val66Met. Impaired extinction in Met allele
carriers (Val/Met and Met/Met) as a function of time in 68 mice (A) and 72 humans (B) as indexed by
percentage of time freezing in mice and skin conductance response (SCR) in humans to the conditioned
stimulus when it was no longer paired with the aversive stimulus. All results are presented as means T
SEM. *P < 0.01, Student’s t test. **P < 0.02, Student’s t test. VV, Val/Val; VM, Val/Met; and MM, Met/Met.
Fig. 2. Impairedlearningofneutralcueinhuman
Met allele carriers. Elevated skin conductance
response (SCR) to the cue never paired with the
aversive stimulus during fear conditioning as a
function of time in Met allele carriers (VM) relative
to non–Met allele carriers (VV). All results are
presentedas meansT SEM.*P <0.001,Student’st
test. VV, Val/Val; VM, Val/Met.
12 FEBRUARY 2010 VOL 327
SCIENCE
www.sciencemag.org
864
REPORTS

Page 3

strated to support fear conditioning and extinc-
tion in both rodent (17–20) and human (21–26)
studies. Whereas portions of the amygdala have
been shown to be essential for fear conditioning
(27, 28), ventral prefrontal cortical regions have
been shown to be important for modifying
previously learned associations and extinction
(19, 29). Thus, on the basis of our behavioral
findings in the mouse and human, we hypothe-
sized that ventromedial prefrontal regions, im-
portant in extinction, would be less active in Met
allele carriers relative to non–Met allele carriers
and that amygdala activity may be enhanced.
Totestthishypothesis,weexaminedthemain
effect of genotype on brain activity during
extinctionofthepreviouslyconditionedstimulus.
The analysis directly parallels the observed
behavioralmaineffectsofgenotypeonextinction
asmeasuredbymeanpercentageoftimefreezing
inthe mice (Fig.3A)and mean skin conductance
response in humans (Fig. 3B) (16) with Met
allele carriers showing weaker extinction. The
imaging results showed significantly less ventro-
medial prefrontal cortical (vmPFC) activity
during extinction in Met allele carriers relative
to non–Met allele carriers [t(68) = –3.78, P <
0.05, corrected] (Fig. 3C) (16). In contrast, Met
allele carriers show greater amygdala activity
relative to non–Met allele carriers during extinc-
tion [t(68) = 2.23, P < 0.05, corrected] (Fig. 3D).
These findings indicate that cortical regions previ-
ouslyshowntobeessentialforextinction(vmPFC)
in both rodent and human (19, 26, 30) are
hyporesponsive in Met allele carriers relative to
non–Met allele carriers. Moreover, Met allele
carriers show continued recruitment of the amyg-
dala, a region that should show diminished
activity during the extinction trials of the ex-
periment(26).Thesefindingsaremostlikelydue
to the SNP biasing activity-dependent learning
rather than affecting CNS development per se, as
there was no evidence of genotypic developmen-
tal effects on brain structure in this ethnicity-,
age-,andgender-matchedsamplefromMRI-based
brain morphometry (supporting online text). Fur-
thermore, an association between vmPFC activity
and the strength of fibers connecting frontolimbic
regions is consistent with more effective extinc-
tionlearningasaresultofbettervmPFCmodula-
tionoftheamygdala(supportingonlinetext,figs.
S3 and S4).
These experiments identify a behavioral
phenotype related to BDNF Val66Met across
species providing evidence for translation from
mouse to human. The mouse model provides the
opportunity to test dose-dependent effects of the
BDNFMetalleleinbothacontrolledgeneticand
environmental background not feasible in hu-
mans. These features allow for reliable assign-
ment of behavioral differences to the effects of
the Val66Met polymorphism. The human behav-
ioral and imaging findings provide confidence
thatcross-speciestranslationisbiologicallyvalid,
by defining the underlying neural circuitry of the
behavioral effects of BDNF Val66Met that can
be mapped onto known circuits involved in fear
learning and extinction. The robustness of our
findings across species and paradigms is evi-
denced by work showing slower extinction cou-
pledwithdecreasedneuronaldendriticcomplexity
in vmPFC in the BDNFMet/Metmice in a con-
ditioned taste aversion task compared with wild-
type counterparts (31). Furthermore BDNFMet/Met
mice exhibit a trend toward blunted expression of
c-Fos in the vmPFC as compared with wild-type
miceafterthefearextinctionparadigm(supporting
online text and fig. S5).
Impaired extinction learning has been impli-
cated in anxiety disorders, including phobias and
posttraumatic stress disorder, whereby the indi-
vidual has difficulty recognizing an event as safe
(32). Our neuroimaging findings of diminished
ventromedial prefrontal activity and elevated
amygdala activity during extinction are reminis-
cent of those reported in patients with anxiety
disorders and depression when presented with
an empty threat or aversive stimuli (e.g., fearful
faces) (33, 34). Understanding the effect of the
BDNF Met allele on specific components of a
simple form of learning provides insight into risk
for anxiety disorders and has important impli-
cations for the efficacy of treatments for these
disorders that rely on extinction mechanisms.
One such treatmentisexposure therapy,whereby
anindividualisrepeatedlyexposedtoatraumatic
event in order to diminish the significance of
that event. Our findings suggest that the BDNF
Val66Met SNP may play a key role in the ef-
ficacy of such treatments and may ultimately
guidepersonalizedmedicineforrelatedclinical
disorders.
References and Notes
1. Z. Y. Chen et al., Science 314, 140 (2006).
2. M. F. Egan et al., Cell 112, 257 (2003).
3. A. R. Hariri et al., J. Neurosci. 23, 6690 (2003).
4. J. P. Chhatwal, L. Stanek-Rattiner, M. Davis, K. J. Ressler,
Nat. Neurosci. 9, 870 (2006).
5. L. M. Rattiner, M. Davis, K. J. Ressler, Neuroscientist 11,
323 (2005).
6. D. S. Charney, H. K. Manji, Sci. STKE 2004, re5 (2004).
7. E. J. Nestler et al., Neuron 34, 13 (2002).
8. M. Gratacòs et al., Biol. Psychiatry 61, 911 (2007).
9. J. M. Gatt et al., Biol. Psychol. 79, 275 (2008).
10. X. Jiang et al., Neuropsychopharmacology 30, 1353
(2005).
11. S. L. Rauch, L. M. Shin, E. A. Phelps, Biol. Psychiatry 60,
376 (2006).
12. S. Lissek et al., Behav. Res. Ther. 43, 1391 (2005).
13. B. O. Rothbaum, M. Davis, Ann. N. Y. Acad. Sci. 1008,
112 (2003).
14. M. R. Delgado, A. Olsson, E. A. Phelps, Biol. Psychol. 73,
39 (2006).
15. F. Sotres-Bayon, C. K. Cain, J. E. LeDoux, Biol. Psychiatry
60, 329 (2006).
16. Materials and methods are available as supporting
material on Science Online.
17. K. M. Myers, M. Davis, Neuron 36, 567 (2002).
18. G. J. Quirk, E. Likhtik, J. G. Pelletier, D. Paré, J. Neurosci.
23, 8800 (2003).
19. M. R. Milad, G. J. Quirk, Nature 420, 70 (2002).
20. J. E. LeDoux, Annu. Rev. Neurosci. 23, 155 (2000).
21. K. S. LaBar, J. C. Gatenby, J. C. Gore, J. E. LeDoux,
E. A. Phelps, Neuron 20, 937 (1998).
22. D. Schiller, I. Levy, Y. Niv, J. E. LeDoux, E. A. Phelps,
J. Neurosci. 28, 11517 (2008).
23. M. R. Delgado, K. I. Nearing, J. E. Ledoux, E. A. Phelps,
Neuron 59, 829 (2008).
24. R. Kalisch et al., J. Neurosci. 26, 9503 (2006).
25. J. A. Gottfried, R. J. Dolan, Nat. Neurosci. 7, 1144
(2004).
26. E. A. Phelps, M. R. Delgado, K. I. Nearing, J. E. LeDoux,
Neuron 43, 897 (2004).
27. K. Nader, J. LeDoux, Behav. Neurosci. 113, 152
(1999).
28. M. Kim, M. Davis, Behav. Neurosci. 107, 580 (1993).
29. K. Lebrón, M. R. Milad, G. J. Quirk, Learn. Mem. 11, 544
(2004).
30. M. R. Milad et al., Biol. Psychiatry 62, 446 (2007).
31. H. Yu et al., J. Neurosci. 29, 4056 (2009).
32. D. S. Charney, Am. J. Psychiatry 161, 195 (2004).
33. I. Liberzon et al., Biol. Psychiatry 45, 817 (1999).
34. S. L. Rauch et al., Biol. Psychiatry 47, 769 (2000).
35. We acknowledge two anonymous reviewers for their
thoughtful comments, resources and staff at the
Fig. 3. Neural circuitry of the behavioral effect of BDNF Val66Met during extinction. (A) Average
percentage of time freezing during extinction by genotype in 68 mice. (B) Average skin conductance
response (SCR) during extinction by genotype in 72 humans. (C) Brain activity as indexed by percent
change in magnetic resonance (MR) signal during extinction in the ventromedial prefrontal cortex
(vmPFC) by genotype (x, y, z = –4, 24, 3), with Met allele carriers having significantly less activity than
Val/Valhomozygotes(VM<VVisblue),imagethresholdP<0.05,corrected.(D)Genotypicdifferencesin
left amygdala activity during extinction (x, y, z = –25, 2, –20) in 70 humans, with Met allele carriers
having significantly greater activity than Val/Val homozygotes (VM > VV is orange), image threshold P <
0.05, corrected. *P < 0.05. **MM were included in the analysis with VM, but plotted separately to see the
dose response. All results are presented as means T SEM. VV, Val/Val; VM, Val/Met; MM, Met/Met.
www.sciencemag.org
SCIENCE
VOL 327 12 FEBRUARY 2010
865
REPORTS

Page 4

Biomedical Imaging Core Facility of the Citigroup
Biomedical Imaging Center at Weill Cornell Medical
College, Rafael Oania, a generous gift by the
Dr. Mortimer D. Sackler family and support from
NIH grants MH079513 (B.J.C., F.S.L.), MH060478
(B.J.C.), NS052819 (F.S.L.), HD055177 (B.J.C., S.S.P.),
GM07739 (F.S.), and United Negro College Fund–Merck
Graduate Science Research Dissertation Fellowship
(F.S.), Burroughs Wellcome Foundation (F.S.L.), and
International Mental Health Research Organization
(F.S.L.).
Supporting Online Material
www.sciencemag.org/cgi/content/full/science.1181886/DC1
Materials and Methods
SOM Text
Figs. S1 to S5
Tables S1 and S2
References
14 September 2009; accepted 28 December 2009
Published online 14 January 2009;
10.1126/science.1181886
Include this information when citing this paper.
Vibrio cholerae VpsT Regulates
Matrix Production and Motility
by Directly Sensing Cyclic di-GMP
Petya V. Krasteva,1Jiunn C. N. Fong,2Nicholas J. Shikuma,2Sinem Beyhan,2
Marcos V. A. S. Navarro,1Fitnat H. Yildiz,2* Holger Sondermann1*
Microorganisms can switch from a planktonic, free-swimming life-style to a sessile, colonial state,
called a biofilm, which confers resistance to environmental stress. Conversion between the motile
and biofilm life-styles has been attributed to increased levels of the prokaryotic second messenger
cyclic di-guanosine monophosphate (c-di-GMP), yet the signaling mechanisms mediating such a
global switch are poorly understood. Here we show that the transcriptional regulator VpsT from
Vibrio cholerae directly senses c-di-GMP to inversely control extracellular matrix production and
motility, which identifies VpsT as a master regulator for biofilm formation. Rather than being
regulated by phosphorylation, VpsT undergoes a change in oligomerization on c-di-GMP binding.
I
els of extracellular matrix by means of expres-
sion of Vibrio polysaccharide (vps) genes and
genes encoding matrix proteins. vps expression
is under the control of two positive transcrip-
tional regulators, VpsTand VpsR (5, 6). VpsT is
a member of the FixJ, LuxR, and CsgD family
of prokaryotic response regulators, typically ef-
fectors in two-component signal transduction
systems that use phosphoryl transfer from up-
stream kinases to modulate response-regulator
protein activity (7–9). Although the putative phos-
phorylation site is conserved in VpsT’s receiver
domain, other residues crucial for phosphotransfer-
dependent signaling are not, and no cognate
kinase has been identified to date (fig. S1). Reg-
ulation by VpsT and VpsR has been linked to
signal transduction by using the bacterial second
messenger cyclic di-guanosine monophosphate
(c-di-GMP) (10, 11) (fig. S2), yet little is known
about the direct targets of the nucleotide. A ribo-
switch has been identified as a c-di-GMP target
that regulates gene expression of a small num-
ber of genes, but that is unlikely to account for
the global change in transcriptional profile re-
quired for biofilm formation (12). Neither do PilZ
domain–containing proteins, potential c-di-GMP
effectors, affect rugosity, because a V. cholerae
n Vibrio cholerae, biofilm formation is fa-
cilitated by colonial morphotype variation
(1–4). Rugose variants produce increased lev-
strain lacking all five PilZ domain–containing
proteins retains its colony morphology and abil-
ity to overproduce vps gene products (13).
VpsTconsists of an N-terminal receiver (REC)
domain and a C-terminal helix-turn-helix (HTH)
domain, with the latter mediating DNA binding
(Fig. 1A) (14) (also see supporting online text
for details). Unlike other REC domains, the canon-
ical (a/b)5-fold in VpsT is extended by an addi-
tional helix at its C terminus [helix a6 in (Fig.
1A)]. The HTH domain buttresses an interface
formed by helices 3 and 4 of the N-terminal reg-
ulatory domain. There are two nonoverlapping
dimerization interfaces between noncrystallograph-
ic VpsT protomers [chain A-chain B and chain
A-chain Bsym(symmetrical) in (Fig. 1B)]. The
c-di-GMP–independent interface involves inter-
actions mediated by a methionine residue (M17)
(15) located at the beginning of a1 and a bind-
ing pocket that extends into the putative phos-
phorylation site of the REC domain (fig. S3A).
The second interface involves a6 of the REC
domain, in contrast to canonical response regu-
lators, such as CheY and PhoB, that utilize a
surfaceformedbya4-b5-a5 for dimerization (9).
The binding of two intercalated c-di-GMP mole-
cules to the base of a6 stabilizes VpsT dimers
using this interface (Fig. 1 and fig. S3B).
The binding motif for c-di-GMP in VpsT
consists of a four-residue-long, conserved W[F/
L/M][T/S]R sequence (15) (fig. S1). The side
chains of the tryptophan and arginine form p-
stacking interactions with the purine rings of
the nucleotide (Fig. 1C). While the hydrophobic
residue in the second position plays a structural
role where it is buried in the REC domain, the
threonine residue at position 3 forms a hydrogen
bond with the phosphate moiety of c-di-GMP. A
subclass of VpsT and/or CsgD homologs exists
with a proline substitution in position 3 (W[F/L/
M]PR). Although CsgD is also functionally linked
to c-di-GMP signaling in Escherichia coli and
Salmonella (16, 17), its binding pocket appears
to be distinct from that of VpsT, as it displays a
highly conserved YF[T/S]Q motif that is un-
likely to accommodate c-di-GMP (fig. S3B).
The apparent affinity of VpsT for c-di-GMP,
determined by isothermal titration calorimetry, is
3.2 mM with 1:1 stoichiometry, consistent with a
dimer of c-di-GMP binding to a dimer of VpsT
(fig. S4A). Single point mutations in the con-
served c-di-GMP–binding motif (VpsTR134A,
VpsTW131F, or VpsTT133V) or in the isoleucine
in a6 of the c-di-GMP–stabilized REC dimer-
ization interface (VpsTI141E) abolished c-di-
GMP binding, which indicated that dimeric
REC domains are required for binding (fig. S4B).
Conversely, mutation of a key residue in the
nucleotide-independent interface (VpsTM17D)
had no effect on c-di-GMP recognition. On the
basis of static multiangle light scattering, VpsTM17D
exists as a monomeric species in the absence of
c-di-GMP, whereas intermediate molecular weights
for the wild-type VpsTand the mutants VpsTR134A
and VpsTI141Eindicated fast exchange between
monomers and dimers, presumably through the
c-di-GMP–independent interface (fig. S5 and
table S2). Addition of c-di-GMP increases the
molecular weight of VpsTM17Dand wild-type
VpsT (figs. S5 and S6), whereas the oligomeric
state of VpsTR134Aand VpsTI141Eis insensitive
to the nucleotide.
The role of c-di-GMP recognition and the
relevance of the two dimer interfaces in DNA-
binding and VpsT-regulated gene expression were
assessed by using c-di-GMP binding (R134) and
dimerization (I141or M17) mutants (Fig. 2). In
electromobility shift assays, we used regulatory
sequences upstream of vpsL, a gene under pos-
itive control of VpsT (Fig. 2A) (6). DNA mobil-
ity shifts were observed only for the wild-type
and VpsTM17Dforms, where the effect was pro-
tein specific and c-di-GMP dependent. In ad-
dition, nucleotide-dependent DNA binding of
VpsT was observed to multiple and relatively
remote sites in the regulatory region of vpsL.
To evaluate the functional importance of
VpsT oligomers and c-di-GMP binding in cells,
we measured transcription of vps genes by using
a chromosomal vpsLp-lacZ transcriptional fusion
in the DvpsT strain harboring wild-type vpsT,
vpsT point mutants (vpsTM17D, vpsTR134A, or
1Department of Molecular Medicine, College of Veterinary
Medicine, Cornell University, Ithaca, NY 14853, USA.2Depart-
ment of Microbiology and Environmental Toxicology, Uni-
versity of California, Santa Cruz, CA 95064, USA.
*To whom correspondence should be addressed: yildiz@
metx.ucsc.edu (F.H.Y.); hs293@cornell.edu (H.S.)
12 FEBRUARY 2010VOL 327
SCIENCE
www.sciencemag.org
866
REPORTS