Content uploaded by Peter Brennan
Author content
All content in this area was uploaded by Peter Brennan
Content may be subject to copyright.
MHC Class I Peptides as
Chemosensory Signals in the
Vomeronasal Organ
Trese Leinders-Zufall,
1
Peter Brennan,
2
Patricia Widmayer,
3
Prashanth Chandramani S.,
3
Andrea Maul-Pavicic,
4
Martina Ja
¨
ger,
4
Xiao-Hong Li,
1
Heinz Breer,
3
Frank Zufall,
1
*
Thomas Boehm
4
*
The mammalian vomeronasal organ detects social information about gender,
status, and individuality. The molecular cues carrying this information remain
largely unknown. Here, we show that small peptides that serve as ligands for
major histocompatibility complex (MHC) class I molecules function also as
sensory stimuli for a subset of vomeronasal sensory neurons located in the
basal G
a
o- and V2R receptor–expressing zone of the vomeronasal epithelium.
In behaving mice, the same peptides function as individuality signals under-
lying mate recognition in the context of pregnancy block. MHC peptides
constitute a previously unknown family of chemosensory stimuli by which
MHC genotypic diversity can influence social behavior.
The mammalian vomeronasal organ (VNO) is
essential for social recognition. Vomeronasal
sensory neurons (VSNs) detect pheromones
and other chemosignals that carry informa-
tion about gender, sexual and social status,
dominance hierarchies, and individuality, but
it has been very difficult to define the molec-
ular nature of these chemosignals (1–13). The
VNO epithelium is segregated into two dis-
tinct zones, both of which express a unique
set of transduction-related molecules (2, 3,
10–12): (i) an apical (superficial) zone that
expresses the G protein Gai2 as well as mem-
bers of the V1R family of vomeronasal re-
ceptors (È150 genes) and (ii) a basal (deep)
zone that characteristically contains VSNs
that express Gao and members of the V2R
receptor family (9150 genes). The few mol-
ecules that have been identified as sensory
stimuli thus far are all small, urine-derived
volatiles that activate VSN subpopulations in
the apical zone (4, 8, 12, 13). Stimuli for
VSNs in the basal zone have not yet been
found, nor has it been possible to identify any
nonvolatile molecules that are widely assumed
to be detected by the VSNs (1–3, 9, 12).
We hypothesized that peptide ligands of
the major histocompatibility complex (MHC)
class I molecules, in addition to their well-
established role in the immune system (14),
may function as sensory stimuli for VSNs.
MHC peptides are excellent candidates for
social recognition signals that convey infor-
mation about genetic individuality. The poly-
morphisms of MHC molecules directly
translate into structurally diverse peptide-
binding grooves, such that different MHC
molecules bind different peptides (14). Hence,
the structures of peptide ligands mirror the
structures of MHC molecules and thus pro-
vide a unique molecular signature for each
individual. When peptide/MHC complexes
are not retained at the cell surface but are
instead released into the extracellular space
and appear in the urine and other bodily se-
cretions (15), any information contained in
their chemical complexity becomes a property
of the entire individual and potentially can be
used for interindividual communication (16).
By using an intact VNO preparation to
record extracellular field potentials from the
microvillous surface of the sensory epitheli-
um (4), we tested whether known ligands of
MHC class I molecules (14) Etable S1 and
Supporting Online Material (SOM) Text^ elic-
it electrical responses in VSNs of C57BL/6
mice (17). These ligands were chosen to cor-
respond to prototypical representatives for
two disparate H-2 haplotypes, namely
AAPDNRETF (for the H-2
b
haplotype of
C57BL/6 mice) and SYFPEITHI (for the
unrelated H-2
d
haplotype of BALB/c mice)
(14, 18). Both peptides evoked negative field
potentials in a dose-dependent manner in the
VNOs of female C57BL/6 mice carrying the
H-2
b
haplotype, indicating that cognate MHC
class I molecules are not required for this
response. Threshold responses were observed
with concentrations below 10
j11
Mand
10
j12
M, respectively (Fig. 1, A to D).
Control peptides in which the characteristic
anchor residues of the two MHC class I
ligands were replaced by alanines (i.e.,
AAPDARETA and SAFPEITHA, respective-
ly) failed to activate VSNs at all concen-
trations tested (up to 10
j7
M) (Fig. 1, E and
F), indicating that certain structural features
of peptides may be required for VSN ac-
tivation and ruling out an involvement of
trace by-products from peptide synthesis
1
Department of Anatomy and Neurobiology, Univer-
sity of Maryland School of Medicine, Baltimore, MD
21201, USA.
2
Sub-Department of Animal Behaviour,
University of Cambridge, Cambridge CB3 8AA, UK.
3
Institut fu¨r Physiologie, Universita
¨
t Hohenheim, D-
70593 Stuttgart, Germany.
4
Department of Devel-
opmental Immunology, Max-Planck Institute of
Immunobiology, D-79108 Freiburg, Germany.
*To whom correspondence should be addressed.
E-mail: boehm@immunbio.mpg.de (T.B.); fzufa001@
umaryland.edu (F.Z.)
Fig. 1. Class I MHC
ligands induce excitato-
ry electrical responses in
mouse VNO. (A and B)
Examples of negatively
directed field potentials
and their dose depen-
dency registered in in-
tact VNO of female
C57BL/6 mice. Re-
sponses are produced
by 500-ms pulses of
peptides that were fo-
cally ejected from a
multibarrelled stimula-
tion pipette. Responses
are representative of a
total of 17 recordings in
six mice. (C and D)
Dose-response plots of
peak responses from
the two experiments
shownin(A)and(B),
respectively. Smooth
curves are fitted by the
Hill equation, with K
1/2
value and Hill coeffi-
cient of 13.3 pM and 0.9 (AAPDNRETF) and 1.2 pM and 1.0 (SYFPEITHI), respectively. (E and F)Two
control peptides (each at 10
j7
M) failed to elicit an electrical response (n 0 13). (G) The response to
SYFPEITHI and AAPDNRETF (each at 10
j11
M) is reversibly suppressed by 2-APB (50 6M, n 0 3). (H)
Onset kinetics and initial slope of the rising phase of the field potential depend on stimulus
concentration. Responses were scaled to yield the same peak amplitudes.
R EPORTS
www.sciencemag.org SCIENCE VOL 306 5 NOVEMBER 2004
1033
and purification in the registered responses.
Peptide-induced potentials were abolished
by 2-aminoethoxydiphenyl borate (2-APB)
(50 6M) (Fig. 1G), a blocker of Ca
2þ
-
permeable, diacylycerol-gated cation channels
essential for VNO transduction, in keeping
with results obtained for urine stimuli (19).
Next, we investigated the cellular logic
underlying peptide recognition and discrim-
ination in the VNO. We first used freshly dis-
sociated VSNs from a transgenic C57BL/6
mouse strain in which all mature VSNs can
be visualized on the basis of their expres-
sion of green fluorescent protein (GFP)
under the control of the regulatory sequences
of the OMP gene (20). VSNs generate an
increase in intracellular Ca
2þ
concentration
in response to chemostimulation (4, 13,
21). Cells were loaded with a Ca
2þ
in-
dicator dye, fura-2, and cellular responses
were examined optically. Transient somatic
Ca
2þ
elevations to peptide stimulations
were reproducibly detected in a subset of
GFP-positive cells (Fig. 2, A and B). A
total of 27 cells responded to stimulation
with AAPDNRETF or SYFPEITHI pep-
tides. Of these cells, 10 responded only to
the D
b
ligand, 15 only to the K
d
ligand, and
2cellsrespondedtobothpeptides.The
response to structurally different peptides is
thus specific for individual subsets of
VSNs, with only minimal overlap. We then
systematically analyzed the spatial repre-
sentation of peptide responses in large
populations of VSNs in the sensory epithe-
lium by using in situ mapping of neuronal
activity (4). Each peptide produced robust
and reproducible increases in intracellular
Ca
2þ
in specific subsets of VSNs when
tested in such slices (Fig. 2, C to K). In total,
we imaged 5183 VSNs (26 slices from 22
mice) of which 85 cells (1.6%) responded to
peptide ligands. These signals were re-
versibly abolished by 2-APB (50 6M) (Fig.
2E), in accord with the field potential
recordings of Fig. 1G. Analysis of stimulus-
response curves of single VSNs responding
to either AAPDNRETF or SYFPEITHI es-
tablished that these cells are exceptionally
sensitive detectors of MHC peptides, with
activation thresholds near or below 10
j12
M.
VSNs responding to the same peptide ex-
hibited almost identical dose-response curves
(Fig. 2, I and J). The sets of neurons acti-
Fig. 2. MHC peptides are detected by distinct populations of VSNs. (A)
Fluorescence image of freshly dissociated VSNs obtained from an OMP-GFP
mouse. White arrow indicates the cell that was analyzed in (B). Scale bar,
10 6m. (B) Waveform of somatic Ca
2þ
transients evoked by the application
of SYFPEITHI (5 10
j10
M) or KCl (100 mM). This cell did not respond to
AAPDNRETF (5 10
j10
M). The latencies between stimulus onset and
response decrease from left to right because of the design of the perfusion
apparatus. (C and D) Spatial representation of peptide-induced activity in
VNO sensory epithelium. Shown are reconstructed VSN response maps
(%F/F confocal Ca
2þ
images digitally superimposed onto a transmitted light
image of the same slice. F, fluorescence units) for AAPDNRETF (10
j12
M,
green) and SYFPEITHI (10
j12
M, red). Cells responding to both peptides are
color-coded yellow. Black arrows indicate peptide-sensitive VSNs that are
localized at the very base of the epithelium. Black boxes, regions that were
imaged in these experiments. The white box in (C) is shown at higher
magnifications in (F) to (H). Scale bar, 100 6m. (E)Ca
2þ
response to
SYFPEITHI (10
j12
M) is reversibly abolished by 2-APB (50 6M, n 0 7). (F to
H) High-resolution pseudocolor images of the relative increase in peptide-
induced Ca
2þ
fluorescence (ratio between the peak fluorescence before and
after stimulation, %F/F). In this example, AAPDNRETF (10
j12
M, green)
activated three VSNs (cell 2, 3, and 4) and SYFPEITHI (10
j12
M, red)
activated two VSNs (cell 1 and 4). Cell 4 responded to both ligands. Scale
bar, 10 6m. Dose-dependency of stimulus-induced Ca
2þ
peak responses of
12 VSNs that recognized either AAPDNRETF (I) or SYFPEITHI (J). (K)Time
course of peptide-induced Ca
2þ
responsesfromthesamecellsthatare
depicted in (F) to (H).
R EPORTS
5 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org
1034
vated by structurally different peptides were
largely distinct, with only a few cells re-
sponding to both peptides (Fig. 2, C, D, F
to H, and K). Of 2067 imaged VSNs that
were tested with both peptides, 25 (1.2%)
responded only to the D
b
ligand, 20 (1.0%)
only to the K
d
ligand, and 8 cells (0.4%) re-
sponded to both peptides. To support the
physiological relevance of nonvolatile pep-
tides as stimuli of VSNs, we confirmed that
nonvolatile stimuli in urine gain access to the
vomeronasal epithelium in behaving mice
(fig. S1 and SOM Text) and that VSNs ac-
tivated by synthetic peptides respond also to
urine obtained from mice of the relevant
haplotype (fig. S2).
Peptides were recognized by sparse pop-
ulations of VSNs that were widely distrib-
uted in the sensory epithelium (Fig. 2, C and
D). We noted that activated VSNs were
mostly localized to the basal half of the
epithelium. Almost one-third of these cells
(16/53 or 30%) were found at the very base
of the epithelium, close to the basal lamina
(Fig. 2, C and D, black arrows). Do peptide-
detecting neurons thus belong to those of the
basal zone? To address this question, we first
identified peptide-sensitive VSNs by in situ
Ca
2þ
mapping and then immunostained the
tissue with antibodies against Gao Especific
for the basal zone (2, 3, 10–12)^ and phospho-
diesterase PDE4A Especific for the apical
zone (22)^ (fig. S3A). All 18 peptide-sensitive
VSNs were identified as Gao-positive and
PDE4A-negative, irrespective of whether they
were located in deep or more superficial
regions of the epithelium (fig. S3, B to E).
This result was confirmed with dissociated
VSNs by using Gao antibody (23). To
demonstrate that peptide-sensitive VSNs ex-
press V2R receptors, we used an antibody that
recognizes the V2R2 receptor, which is
broadly expressed in the basal VNO layer
(24). Double-label immunohistochemistry
showed that V2R2 is coexpressed in all
Gao-positive VSNs (fig. S3, F to H). Com-
bining in situ Ca
2þ
mapping and V2R2 im-
munolabeling, we showed directly that
peptide-sensitive VSNs express V2Rs (fig.
S3, I to K) (n 0 7).
To determine whether peptide stimula-
tion leads to action potential generation in
single VSNs, we used the loose-patch tech-
nique to register extracellular spike activity
from visually identified VSNs in VNO slices
(4). At 10
j11
M, the MHC peptides elicited
excitatory, sequence-specific responses in a
subset of basal VSNs (Fig. 3A) (n 0 6),
consistent with the Ca
2þ
imaging data. We
also used whole-cell current clamp record-
ings from VSNs in slices (19) to demonstrate
directly that peptide stimulation produced a
membrane depolarization that, in turn,
evoked action potential discharges (Fig. 3B)
(n 0 5).
What are the structural constraints un-
derlying peptide discrimination? We hy-
pothesized that the peptide anchor residues
may substantially contribute to the specific
recognition by VSNs. Indeed, two different
ligands of the K
d
MHC molecule, SYFPEITHI
and SYIPSAEKI (14, 18), which share the
same anchor residues (Y and I) at positions 2
and 9, respectively, but differ substantially in
the other positions, activated the same 6
neurons (out of 1109 cells imaged), and their
stimulus-response curves were nearly identical
(Fig. 4, A to E). VSNs that recognized only
one, but not the other, peptide were not
observed. The recognition mode of such
peptides may thus at least partially resemble
that of MHC molecules.
Given that the anchor residues of peptides
appear to be essential for VSN activation and
that a goldfish V2R-like receptor is known to
recognize free amino acids (25), it was nec-
essary to rule out that the VSN responses
were caused by free amino acids. Isoleucine
(10
j12
M), the C-terminal anchor residue in
the K
d
peptides, failed to generate a Ca
2þ
signal in five of five cells that detected
SYFPEITHI and SYIPSAEKI (Fig. 4D). It
also failed to produce a response in 881 other
VSNs with unknown tuning properties.
Furthermore, in field potential recordings, a
mixture containing all amino acids (in free
form, each at 10
j11
M) that constitute the
SYIPSAEKI peptide failed to induce any re-
sponse (Fig. 4F) (n 0 11). Likewise, a scram-
Fig. 4. Structural features of VSN peptide discrimination. (A to C)Ca
2þ
responses in individual VSNs to two MHC peptides, SYFPEITHI (green) and
SYIPSAEKI (red) (each at 10
j12
M). Both cells responded to both peptides. (D)
Time courses of peptide-evoked Ca
2þ
responses from the two cells shown in
(A) to (C). The free amino acid isoleucine (10
j12
M) failed to elicit a response.
(E) Comparison of the dose-dependency of Ca
2þ
peak responses induced by
SYFPEITHI (black curve, open circles) or SYIPSAEKI (red curve, solid circles).
Each data point represents the mean T SD of at least five independent
measurements. (F) A mixture containing all the amino acids that constitute the SYIPSAEKI peptide (in free form, each at 10
j11
M) fails to elicit a VNO
field potential (11 recordings from three mice). (G) A scrambled version of the AAPDNRETF peptide, ANPRAFDTE (10
j11
M), fails to evoke a field
potential response (14 recordings from five mice).
Fig. 3. MHC peptides induce action potential
generation in individual VSNs. (A) Spontaneous
and stimulus-evoked impulse discharges in a
VSN after successive application of three dif-
ferent peptides (all at 10
j11
M). AAPDNRETF,
but not SYFPEITHI or AAPDARETA, elicited a
transient excitation in this neuron. (B) Whole-
cell current clamp recording of a VSN that
responded to AAPDNRETF (10
j11
M) with a
transient increase in the rate of action potential
firing. Resting potential was –62 mV. Arrows
indicate the time point at which peptide applica-
tion was turned on.
R EPORTS
www.sciencemag.org SCIENCE VOL 306 5 NOVEMBER 2004
1035
bled version of the D
b
ligand AAPDNRETF,
ANPRAFDTE, failed to evoke any response
(Fig. 4G) (n 0 14). Thus, peptides must meet
precise structural specifications for VSN
activation, and peptides of random sequence
are unlikely to function as ligands for the
receptors on VSNs.
Given that MHC peptides activate VSNs
in a sequence-specific manner, they could
potentially function as individuality signals
during social recognition. In mice, selective
pregnancy failure Ethe Bruce effect (26)^ rep-
resents an excellent paradigm to assess this
hypothesis in vivo, because it depends criti-
cally on signaling via the accessory olfactory
system (SOM Text) and requires the capacity
to differentiate between individuality cues
(27). Female mice of the BALB/c inbred
strain (H-2
d
haplotype) were mated with
BALB/c males and then exposed to urine
taken from either a BALB/c male (mating
male urine) or C57BL/6 male (unfamiliar
male urine; H-2
b
haplotype). Application of
the unfamiliar urine, coincident with the
postmating peaks in prolactin levels, reliably
resulted in a high level of pregnancy failure,
whereas the familiar urine did not (Fig. 5,
experiments 1 and 2). When BALB/c fe-
males were mated with C57BL/6 males,
pregnancy block occurred after the applica-
tion of BALB/c urine, but not of C57BL/6
urine (Fig. 5, experiments 3 and 4), estab-
lishing the strain specificity of pregnancy
block (28). To test whether familiar male
urine could be converted to unfamiliar urine,
we added peptides of disparate H-2 haplo-
type specificity. Exposure of BALB/c-mated
BALB/c females to a mixture of H-2
b
class I
peptides in BALB/c mating male urine was
equally effective as C57BL/6 urine (Fig. 5,
experiment 6). The addition of a mixture of
H-2
d
peptides had no effect (Fig. 5, experi-
ment 5). Conversely, exposure of C57BL/6-
mated BALB/c females to BALB/c peptides
in C57BL/6 mating male urine was effective
at blocking pregnancy; here, C57BL/6 pep-
tides were ineffective. Experiments 5 to
8 show that peptides per se do not cause
pregnancy failure and that this function
depends on the previous mating combination.
The overall occurrence of pregnancy block
upon exposure to the strange male H-2
peptides of 64% (n 0 47; combined results
of experiments 6 and 7) was significantly
higher than the 25% (n 0 24; combined re-
sults of experiments 5 and 8) elicited by ex-
posure to mating male H-2 peptides (P 0
0.002, Fisher exact probability test).
Our experiments identified an unexpected
role for MHC class I peptides as chemosen-
sory stimuli. MHC class I ligands are recog-
nized by VSNs in the basal layer of the
VNO. Recognition of peptides by VSNs is
independent of MHC haplotype, and pep-
tides specific for different MHC molecules
(i.e., carrying different anchor residues)
generate unique VSN activation patterns,
providing the basis for the neural represen-
tation of the structural diversity of this new
family of chemosignals.
Our data demonstrate that the VNO can
detect both nonvolatile and volatile stimuli, a
result that is fully compatible with early pre-
dictions (29). Peptide responses were found
exclusively in Gao- and V2R-positive neu-
rons, whereas responses to volatile stimuli
have been mapped to the apical V1R-
expressing zone (4). Whether this functional
segregation is true for all vomeronasal stim-
uli remains to be seen. V2R receptors con-
stitute a large family of orphan receptors
(2, 3, 10–12) that differ from V1Rs and
olfactory receptors by the presence of a large
N-terminal domain. They are coexpressed
with MHC class Ib molecules at the cell sur-
face of VSNs (30, 31). MHC class Ib mol-
ecules can bind peptides, but they lack the
typical peptide-binding groove and specificity
of classical MHC class I molecules (32).
Sequence-specific recognition of peptides may
thus be achieved by the N-terminal domain of
certain V2R receptors (or receptor combina-
tions), whereas the MHC class Ib molecules
may serve as a general presentation device.
Given the limited diversity of amino acid res-
idues occupying the two anchor positions of
mouse MHC class I peptides (14), we estimate
that about 50 different receptors should be suf-
ficient to discriminate ligands from all known
mouse MHC class I molecules.
Considerable work has focused on the
main olfactory system in the detection of
MHC-related odor signals (33, 34). Our
results highlight the role of the VNO in this
process but do not preclude a role of the
main olfactory system in individual recogni-
tion, nor do they preclude a role for other
molecules such as volatile urinary constitu-
ents (34) or polymorphic major urinary pro-
teins (35). MHC peptides may thus form one
class of different signals that may be used in
different behavioral contexts. For a meaning-
ful biological response to occur in MHC-
related behaviors, signals about gender,
reproductive status, and species identity must
be evaluated alongside signals of genetic
individuality, and this may involve remotely
sensed signals as well as signals that are de-
tected during direct contact. A chemosensory
function of MHC peptides provides a direct
link between MHC diversity and MHC-
related behavior, converting a MHC geno-
type into an olfactorily detectable quality.
References and Notes
1. C. J. Wysocki, M. Meredith, in Neurobiology of Taste
and Smell, T. E. Finger, W. L. Silver, Eds. (Wiley, New
York, 1987), pp. 125–150.
2. R. Tirindelli, C. Mucignat-Caretta, N. J. Ryba, Trends
Neurosci. 21, 482 (1998).
3. P. A. Brennan, E. B. Keverne, Curr. Biol. 14, R81
(2004).
4. T. Leinders-Zufall et al., Nature 405, 792 (2000).
5. T. E. Holy, C. Dulac, C. M. Meister, Science 289, 1569
(2000).
6. L.Stowers,T.E.Holy,M.Meister,C.Dulac,G.Koentges,
Science 295, 1493 (2002); published online 31 January
2002 (10.1126/science.1069259).
7. B. G. Leypold et al., Proc. Natl. Acad. Sci. U.S.A. 99,
6376 (2002).
8. K. Del Punta et al., Nature 419, 70 (2002).
9. M. Luo, M. S. Fee, L. C. Katz, Science 299, 1196 (2003).
10. C. I. Bargmann, Cell 90, 585 (1997).
11. C. Dulac, A. T. Torello, Nature Rev. Neurosci. 4,551
(2003).
12. P. Mombaerts, Nature Rev. Neurosci. 5, 263 (2004).
13. C. Boschat et al., Nature Neurosci. 5, 1261 (2002).
14. H. G. Rammensee, J. Bachmann, S. Stefanovic, MHC
Fig. 5. Peptides func-
tion as individuality
signals in the context
of pregnancy block.
Percent pregnancy fail-
ure in female BALB/c
(H-2
d
haplotype) mice
mated with either
BALB/c (H-2
d
haplo-
type) or C57BL/6 (H-
2
b
haplotype) males as
indicated (þ). Exposure
to different male urine
types, which, in some
cases, were supple-
mented with peptides
specific for BALB/c or
C57BL/6 haplotypes, is
indicated. *P G 0.05;
***P G 0.001 (Fisher
exact probability test
with Bonferroni cor-
rection for multiple
comparisons).
R EPORTS
5 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org
1036
Ligands and Peptide Motifs (Landes Bioscience,
Georgetown, TX, 1997).
15. P. B. Singh, R. E. Brown, B. Roser, Nature 327, 161 (1987).
16. T. Boehm, C. C. Bleul, M. Schorpp, Immunol. Rev.
195, 15 (2003).
17. Materials and methods are available as supporting
material on Science Online.
18. Single-letter abbreviations for the amino acid resi-
dues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F,
Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn;
P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp;
and Y, Tyr.
19. P. Lucas, K. Ukhanov, T. Leinders-Zufall, F. Zufall,
Neuron 40, 551 (2003).
20. S. M. Potter et al., J. Neurosci. 21, 9713 (2001).
21. M. Spehr, H. Hatt, C. H. Wetzel, J. Neurosci. 22, 8429
(2002).
22. Y. E. Lau, J. A. Cherry, Neuroreport 11, 27 (2000).
23. T. Leinders-Zufall et al., data not shown.
24. S. Martini, L. Silvotti, A. Shirazi, N. J. Ryba, R. Tirindelli,
J. Neurosci. 21, 843 (2001).
25. D. J. Speca et al., Neuron 23, 487 (1999).
26. H. M. Bruce, Nature 184, 105 (1959).
27. A. Lloyd-Thomas, E. B. Keverne, Neuroscience 7, 907
(1982).
28. The efficiency of pregnancy block by direct applica-
tion of unfamiliar urine is similar to that caused by
the presence of unfamiliar mice (for instance, when
BALB/c females are mated with C57BL/6 males,
presence of BALB/c males causes pregnancy failure
in 9 of 12 mice).
29. R. J. O’Connell, M. Meredith, Behav. Neurosci. 98,
1083 (1984).
30. T. Ishii, J. Hirota, P. Mombaerts, Curr. Biol. 13, 394
(2003).
31. J. Loconto et al., Cell 112, 607 (2003).
32. X. L. He, P. Tabaczewski, J. Ho, I. Stroynowski, K. C.
Garcia, Structure 9, 1213 (2001).
33. D. Penn, W. Potts, Adv. Immunol. 69, 411 (1998).
34. G. K. Beauchamp, K. Yamazaki, Biochem. Soc.
Trans. 31, 147 (2003).
35. J. L. Hurst et al., Nature 414, 631 (2001).
36. We thank R. Escher for peptide synthesis, J. Cherry for
PDE4A antibodies, R. Tirindelli for V2R2 antibodies,
P. Mombaerts for OMP-GFP mice, and K. Kelliher for
the demonstration that nonvolatile chemicals can
reach the VNO in vivo. This work was supported by
the Leibniz program of the Deutsche Forschungsge-
meinschaft (T.B. and H.B.), by NIH/National Institute
on Deafness and other Communication Disorders
(T.L.-Z. and F.Z.), and by the Hochshul- und
Wissenschafts-Programm (P.W.).
Supporting Online Material
www.sciencemag.org/cgi/content/full/306/5698/1033/
DC1
Materials and Methods
SOM Text
Figs. S1 to S4
Table S1
References
15 July 2004; accepted 9 September 2004
Autophagy Defends Cells Against
Invading Group A Streptococcus
Ichiro Nakagawa,
1,3
*
Atsuo Amano,
2,4
Noboru Mizushima,
3,5
Akitsugu Yamamoto,
6
Hitomi Yamaguchi,
7
Takahiro Kamimoto,
7
Atsuki Nara,
6,7
Junko Funao,
1
Masanobu Nakata,
1
Kayoko Tsuda,
7
Shigeyuki Hamada,
1
Tamotsu Yoshimori
4,7
*
We found that the autophagic machinery could effectively eliminate
pathogenic group A Streptococcus (GAS) within nonphagocytic cells. After
escaping from endosomes into the cytoplasm, GAS became enveloped by
autophagosome-like compartments and were killed upon fusion of these
compartments with lysosomes. In autophagy-deficient Atg5
j/j
cells, GAS
survived, multiplied, and were released from the cells. Thus, the autophagic
machinery can act as an innate defense system against invading pathogens.
Autophagy mediates the bulk degradation of
cytoplasmic components in eukaryotic cells
in which a portion of the cytoplasm is
sequestered in an autophagosome and even-
tually degraded upon fusion with lysosomes
(1–3). Streptococcus pyogenes (also known
as group A Streptococcus, GAS) is the etio-
logical agent for a diverse collection of
human diseases (4). GAS invades nonphago-
cytic cells (5, 6), but the destination of GAS
after internalization is not well understood.
To clarify the intracellular fate of GAS,
especially any possible involvement of auto-
phagy, we first investigated whether intra-
cellular GAS colocalizes with LC3, an
autophagosome-specific membrane marker,
following invasion of HeLa cells (7–9).
After infection, GAS strain JRS4 cells colo-
calized with LC3-positive vacuole-like struc-
tures in HeLa cells (Fig. 1A). The size (5 to
10 6m) and morphology of the structures were
distinct from standard starvation-induced
autophagosomes with a diameter of about
1 6m (fig. S1A), so we designated these
structures GAS-containing LC3-positive
autophagosome-like vacuoles (GcAVs). The
number of cells bearing GcAVs, the area of
GcAVs, and ratio of GAS trapped in GcAVs
to total intracellular GAS increased in a time-
dependent manner, reaching a maximum at
3 hours after infection (Fig. 1, B and C;
figs. S1B and S2A). A similar result was ob-
tained in mouse embryonic stem (ES) cells
(figs. S2B and S3A). About 80% of intracel-
lular GAS were eventually trapped by the
compartments (Fig. 1C; fig. S1B). LC3 fre-
quently surrounded GAS, fitting closely around
a GAS chain (Fig. 1, D and E; movie S1).
LC3 exists in two molecular forms. LC3-I
(18 kD) is cytosolic, whereas LC3-II (16 kD)
binds to autophagosomes (7, 8). The amount
of LC3-II, which directly correlates with the
number of autophagosomes (8), increased
after infection (Fig. 1F). Thus, GAS invasion
appears to induce autophagy, specifically
trapping intracellular GAS.
To substantiate this idea, we examined
GcAV formation in Atg5-deficient (Atg5
j/j
)
cells lacking autophagosome formation (7).
In contrast to the wild-type cells (fig. S2, B
and C), no GcAVs were observed in Atg5
j/j
ES cells (J1-2) (Fig. 2A) or in Atg5
j/j
mouse embryonic fibroblasts (MEFs) (fig.
S2C). Thus, GcAV formation requires an
Atg5-mediated mechanism. We also exam-
ined LC3-II formation. During infection with
GAS, Atg5
j/j
cells showed no induction of
LC3-II (Fig. 2B). By electron microscopy, in
wild-type MEF cells infected with GAS, we
observed characteristic cisternae surrounding
GAS in the cytoplasm (Fig. 2C). No GAS
were found surrounded by the membranes in
Atg5
j/j
cells (Fig. 2C). The autophagosome-
like multiple membrane–bound compartment
containing GAS was also found in HeLa cells
(Fig. 2D).
Next, we asked whether the bacteria are
killed or survive after entering the compart-
ments. To address this question, we directly
scored bacterial viability by counting colony-
forming units (CFU viability assay) in wild-type
and Atg5
j/j
MEFs (Fig. 2E). In wild-type
MEFs at 4 hours after infection, intracellular
GAS had been killed (Fig. 2E), whereas the
decrease of GAS viability was suppressed
in the Atg5
j/j
MEFs. Tannic acid is a cell-
impermeable fixative that prevents fusion
between secretory vesicles and the plasma
membrane but does not affect intracellular
membrane trafficking (10). In Atg5
j/j
cells treated with tannic acid to prevent
external escape of GAS, the viable bacteria
increased by 2 hours after infection and
maintained this level at 4 hours after in-
fection (Fig. 2E). In contrast, the numbers
of intracellular GAS decreased rapidly in
tannic acid–treated wild-type cells as well
1
Department of Oral and Molecular Microbiology,
2
Department of Oral Frontier Biology, Osaka Univer-
sity Graduate School of Dentistry, 1-8 Yamadaoka,
Suita-Osaka 565-0871, Japan.
3
PRESTO,
4
CREST,
Japan Science and Technology Agency, Kawaguchi-
Saitama 332-0012, Japan.
5
Department of Bioregu-
lation and Metabolism, The Tokyo Metropolitan
Institute of Medical Science, 3-18-22 Honkomagome,
Bunkyo-ku, Tokyo 113-8613, Japan.
6
Department of
Cell Biology, Faculty of Bio-Science, Nagahama
Institute of Bio-Science and Technology, 1266
Tamura-cho, Nagahama-Shiga 526-0829, Japan.
7
De-
partment of Cell Genetics, National Institute of
Genetics/SOKENDAI, Yata 1111, Mishima-Shizuoka
411-8540, Japan.
*To whom correspondence should be addressed.
E-mail: ichiro@dent.osaka-u.ac.jp and tamyoshi@lab.
nig.ac.jp
R EPORTS
www.sciencemag.org SCIENCE VOL 306 5 NOVEMBER 2004
1037