2001 Nature Publishing Group http://neurosci.nature.
com
news and views
erosynaptic nature of kainate-receptor-
mediated LTP in LA/BLA could be a
mechanism for stimulus generalization
that would add significance to stimuli that
are similar to the ones directly associated
with strong emotional content. This
mechanism could also contribute to some
2001 Nature Publishing Group http://neurosci.nature.com
of the more common disorders of emo-
tional memory, including post-traumat-
ic stress disorder.
The implications for the study of
learning and memory are perhaps even
more profound. The most common
strategy for linking synaptic plasticity to
learning has been to manipulate (phar-
macologically or genetically) molecules
that are presumed to be important for
either plasticity or learning and determine
whether the manipulation affects both
equally. This strategy has relied on the
assumption that most synapses that
demonstrate use-dependent plasticity
would use similar mechanisms. Li and
Synapse formation revisited
Graeme W. Davis, Ben Eaton and Suzanne Paradis
A genetic analysis of synapse formation changes the current
understanding of the events that generate acetylcholine
receptor clustering and synapse-specific receptor transcription.
The clustering of neurotransmitter
receptors to a discrete domain of the
postsynaptic membrane is among the
defining events of synapse formation.
For decades, researchers have attempted
to elucidate the sequence of events and
the signaling molecules that lead to this
change. Much of this research has
focused on the events leading to synapse
formation at the cholinergic vertebrate
neuromuscular junction (NMJ). The
model that has emerged from many of
these studies is one of reciprocal induc-
tion. The arrival of the nerve induces
postsynaptic differentiation, which in
turn induces the differentiation of the
presynaptic nerve terminal. The agrin
hypothesis has been proposed as a mol-
The authors are in the Department of
Biochemistry, University of California, San
Francisco, San Francisco, California 94143-
0448, USA.
e-mail: gdavis@biochem.ucsf.edu
558
colleagues have now demonstrated con-
vincingly that this strategy will need to
take into account a wide variety of differ-
ent mechanisms, involving different rules
for synaptic activation and different effec-
tor molecules. Although the prospect of
applying this strategy to complex behav-
iors that may recruit dozens of different
types of synapses spread throughout the
forebrain is daunting, it also provides an
opportunity to dissect the mechanisms of
learning and memory more carefully on
the basis of both anatomical and physio-
logical diversity.
1. Aggleton, J. P. The Amygdala: A Functional
Analysis 2nd edn. (Oxford Univ. Press,
Oxford, 2000).
2. Li, H., Chen, A., Xing, G., Wei, M.-L. &
Rogawski, M. A. Nat. Neurosci. 4, 612620
(2001).
3. Bliss, T. V. P. & Collingridge, G. L. Nature 361,
3139 (1993).
4. Bannerman, D. M., Good, M. A., Butcher, S. P.,
ecular mechanism underlying the induc-
tion of postsynaptic differentiation1.
Agrin is a large secreted glycoprotein
that was purified as an activity that
induces enhanced clustering of acetyl-
choline receptors (AChR) on myotubes
in vitro2. Additional work identified two
muscle proteins that are also critical for
AChR clustering: the transmembrane
tyrosine kinase MuSK and the cytoplas-
mic protein rapsyn3. The agrin hypothe-
sis, in its simplest form, proposes that
agrin is released from the presynaptic
nerve terminal and is deposited in the
synaptic basal lamina. Agrin in the synap-
tic basal lamina then induces the cluster-
ing of AChRs and other molecules at the
site of innervation1. This activity of agrin
requires the activation of postsynaptic
MuSK through an as-yet unidentified
agrin receptor. This model is based, in
part, on the initial characterization of
agrin and MuSK null mutant mice in
which both postsynaptic and subsequent
nature neuroscience volume 4 no 6 june 2001
Ramsay, M. & Morris, R. G. M. Nature 378,
182186 (1995).
5. Tsien, J. Z., Huerta, P. T. & Tonegawa, S. Cell
87, 13271338 (1996).
6. Dudek, S. M. & Bear, M. F. Proc. Natl. Acad.
Sci. USA 89, 43634367 (1992).
7. Chapman, P. F. & Chattarji, S. in The
Amygdala: A Functional Analysis (ed.
Aggleton, J. P.) 117153 (Oxford Univ. Press,
Oxford, 2000).
8. Danober, L., Heinbockel, T., Driesang, R. B. &
Pape, H. C. Neuroreport 11, 25012506
(2000).
9. Li, H., Weiss, S. R., Chuang, D. M., Post, R. M.
& Rogawski, M. A. J. Neurosci. 18, 16621670
(1998).
10. Bortolotto, Z. A. et al. Nature 402, 297301
(1999).
11. Frerking, M., Petersen, C. C. & Nicoll, R. A.
Proc. Natl. Acad. Sci. USA 96, 1291712922
(1999).
12 Frerking, M., Schmitz, D., Zhou, Q.,
Johansen, J. & Nicoll, R. A. J. Neurosci. 21,
29582966 (2001).
13 Husi, H., Ward, M. A., Choudhary, J. S.,
Blackstock, W. P. & Grant, S. G. Nat. Neurosci.
3, 661669 (2000).
presynaptic differentiation are blocked:
AChR clusters are absent or severely
reduced, and the motor nerves fail to
form normal synaptic connections
(instead forming drastically elongated
processes on the muscle fiber surface)4,5.
Two papers, one by Lin and colleagues
in the recent issue of Nature6 and a sec-
ond by Yang and colleagues in Neuron7,
revisit the agrin hypothesis with an exten-
sive genetic analysis of synapse formation.
These new data argue that we need to
revise our view of the early events that ini-
tiate synapse formation at the vertebrate
NMJ (Fig. 1). First, these new data sup-
port previous evidence that AChR clus-
ters are pre-patterned to the eventual site
of muscle innervation (the endplate
band) in the absence of any innervation.
Recent evidence in support of pre-pat-
terning of the muscle membrane came
from analysis of topoisomerase II muta-
tions8. In these mutant mice, the phrenic
nerve navigates to the diaphragm muscle
but fails to innervate the muscle fibers. In
this mutant background, AChRs still form
clusters, which concentrate to the
endplate region of the muscle where the
nerve should have innervated the muscle.
These data suggest that the initial cluster-
ing of AChRs, and the concentration of
these clusters to a site that prefigures
innervation, is a muscle-specific program
(an idea that was initially proposed based
on experiments ablating neurons with
toxins9). There remained the possibility,

Fig. 1. Sequence of signaling events that may
underlie synapse formation at the vertebrate
NMJ. MuSK is required for pre-patterning AChR
clusters in the endplate region. This pre-pattern-
ing is not restricted to AChR clustering, and
includes AChR transcription. Organization of
AChR clustering is a balance between the stabi-
lization of AChR clusters by agrin and the elimi-
2001 Nature Publishing Group http://neurosci.nature.com
nation of aneural AChRs through the activity of a
putative nerve-derived dispersal factor (NDF). A
strong candidate for NDF is the well-studied
activity-dependent downregulation of AChRs.
The distribution of NDF at the endplate is unclear. Synaptogenesis proceeds with further induc-
tive events, the molecular basis of which remains obscure.
however, of transient nerve contact in the
topoisomerase null mutant animals,
because the nerve approached the muscle.
Thus it remains possible, in these mice,
that diffusible signals could have generat-
ed the observed patterning of AChRs.
The papers by Lin and colleagues and
Yang and colleagues now definitively
demonstrate that AChR clusters are pre-
patterned to the endplate region of muscle
in the absence of innervation. Mutant
mice were examined that completely
lacked the motor nerveand AChR clus-
ters still formed and concentrated to the
endplate region where they prefigured the
site of would-be muscle innervation.
Importantly, the average cluster size was
smaller than that observed following
innervation, and the area that the clusters
occupied on the muscle surface was
broader than that observed following
innervation. Thus, innervation further
refined the pre-patterning of the muscle
membrane and enhanced the clustering
of AChRs beneath the nerve terminal.
How is the pre-pattern of AChRs gen-
erated? A remarkable amount of new
genetic data and new analyses of existing
mutations demonstrate that pre-pattern-
ing of AChRs at the muscle endplate is a
MuSK-dependent, agrin-independent
process6,7. At both early (shortly after the
arrival of the nerve at E14.5) and late
(E18.5) stages of synaptic development,
AChR clusters were absent from muscle in
the MuSK null mutant mouse6. Further-
more, in a double-mutant animal, where
both the nerve and MuSK are absent, AchR
clusters were not observed6,7 Thus, pre-pat-
terning of AChR clusters required MuSK.
A similar analysis of agrin mutant
mice, however, reveals a strikingly differ-
ent phenotype. Lin and colleagues gener-
ated a new mutant mouse that eliminated
all isoforms of the agrin gene. In this new
mutant, as well as previous neural agrin
null mutants, AChR clusters still formed
at early stages of synapse development
nature neuroscience volume 4 no 6 june 2001
2001 Nature Publishing Group http://neurosci.nature.com
Pre-pattern Organization
AChR AChR Agrin NDF
(E14.5), demonstrating that agrin was not
necessary for pre-patterning of AChR
clusters6. However, when the same agrin
mutant synapses were analyzed several
days later (E18.5), after prolonged pres-
ence of the nerve, the pre-patterned AChR
clusters were nearly absent 6 . AChRs
appeared to be distributed along the entire
muscle length, rather than being concen-
trated to the endplate, and the remaining
clusters were significantly smaller than at
the earlier stage (E14.5). Furthermore, a
comparison of agrin mutant mice with
double mutants that lack both agrin and
innervation demonstrated that whereas
pre-patterned AChR clusters were present
in the double mutant, these AChR clus-
ters were dissolved and broadly distrib-
uted in the single mutant when
innervation occurred in the absence of
agrin6,7. Thus, in the absence of agrin, the
presence of the nerve disrupted pre-pat-
terned AChR clusters6,7.
Two important insights emerge from
these experiments. First, it seems that
MuSK activation occurs independently
of agrin to induce AChR clustering dur-
ing pre-patterning. Previous data
showed that MuSK can be phosphory-
lated at a low level in the absence of
agrin, suggesting that MuSK could be
autoactivated during synapse forma-
tion1012. Second, because pre-patterned
AChR clusters are disrupted in the
absence of agrin, it seems that agrins
role in vivo might be to stabilize and
enhance pre-existing clusters of AChRs
as well as to induce their formation at
the endplate band. Thus, agrin no longer
seems to be only an inductive cue for
postsynaptic differentiation; it may,
additionally (or instead) consolidate pre-
existing patterning in the muscle.
How is the pre-patterning of AChRs
disrupted in the absence of agrin? Both
recent studies argue that the arrival of the
nerve triggers a program that dissolves the
pre-patterning of AChRs to the endplate
news and views
Induction
Amy Center
antero- retro-
AChR
band6,7. Lin and colleagues suggest the
activity of a nerve-derived dispersal factor
(NDF) that actively disperses or eliminates
non-agrin stabilized receptors. The pro-
posed activity of NDF as a dispersal factor
aptly fits the immunohistochemical images
presented by the authors, in which AChRs
appeared to be dispersed in the muscle
membrane in the presence of the nerve at
agrin-deficient NMJs. There is a wide range
of possible mechanisms, however, by which
this process could be achieved, including
the well-studied activity-dependent down-
regulation of AChRs. NDF could actively
disperse AChRs in the membrane, leading
either directly or indirectly to receptor
internalization and degradation. NDF
might also cause AChR cluster dispersal by
disrupting the MuSK-dependent mecha-
nisms that achieve stable pre-patterning of
AChRs in the middle of the muscle fiber.
It seems likely that this process is related to
the process by which non-synaptic AChRs
are eliminated during the normal process
of synapse formation.
The synapse-specific transcription of
AChRs and the inhibition of AChR tran-
scription at non-synaptic nuclei is anoth-
er important step during synapse
formation at the NMJ. Data presented by
Lin and colleagues and Yang and col-
leagues indicate that subsynaptic AChR
transcription is also pre-patterned. A zone
of AChR transcription is present at the
muscle fiber endplate region in the
absence of innervation. As with AChR
clustering and localization, a genetic
analysis demonstrates that pre-patterning
of AChR transcription in the endplate
region is dependent upon MuSK and is
independent of innervation7. The pre-pat-
terned zone of AChR transcription is
refined to the subsynaptic region upon
innervation. It has yet to be determined,
however, whether the zone of pre-pat-
terned AChR transcription is refined in
the absence of nerve-derived agrin.
Neuregulin (NRG-1), released from the
nerve, is a candidate for the factor that
induces synapse-specific AChR transcrip-
tion13. Although NRG-1 null mutant mice
die before synaptogenesis can be exam-
559

news and views
ined, NRG-1 heterozygous mutant mice
have reduced numbers of AChRs, thus
providing support for this model14. How-
ever, Yang and colleagues analyzed a new
transgenic mouse that inactivates NRG-1
specifically in sensory and motor neurons.
These new data indicate that NRG-1 is not
2001 Nature Publishing Group http://neurosci.nature.com
necessary for the refinement of AChR
expression during synapse formation
(arrival of the nerve), nor is it necessary
for synapse-specific AChR transcription
itself. When NRG-1 was selectively elimi-
nated in sensory/motoneurons, the zone
of AChR transcription at the endplate not
only was normal, but this zone was as nar-
row as that observed at innervated wild-
type muscle. This indicates that the nerve
could still refine the broader pre-patterned
zone of AChR transcription in the absence
of NRG-1 and could also maintain subsy-
naptic AChR transcription. In these exper-
iments, there remains the formal
possibility, although it seems unlikely, that
very low levels of NRG-1 were produced
before gene inactivation occurred and that
these low levels of NRG-1 were sufficient
to induce synapse specific gene expression
of AChRs at the endplate.
Three simple models follow the new
NRG-1 genetic data. One possibility is that
a signal other than neural agrin refines the
pattern of AChR synthesis during synapse
formation. A second possibility is that
neural agrin, requiring MuSK activation,
leads to synapse-specific transcription that
is independent of NRG-1. Finally, it is
Brain at work: play by play
Frank Tong
A functional magnetic resonance imaging (fMRI) study
reveals areas of the human brain that are activated during
the perception of boundaries between events.
In the game of soccer, the players never
stop running. The sport consists of non-
stop action as twelve players battle for
control of the ball. But through the eyes
of the commentator, every dribble, pass,
shot at the goal or block by the goalkeep-
er, is broken downplay by play. From
this breakdown of the game into each
Frank Tong is in the Department of Psychology,
Princeton University, Green Hall, Princeton,
New Jersey 08544, USA.
e-mail: ftong@princeton.edu
560
2001 Nature Publishing Group http://neurosci.nature.com
known that agrin can cluster muscle-
derived NRG-1 (ref. 15). Thus, nerve-
derived agrin may refine the pattern of
AChR expression by stabilizing synaptic
muscle-derived NRG-1 (ref. 7). By exten-
sion, the activity of NDF may not only
contribute to the dissolution of the non-
stabilized, pre-patterned AChR clusters,
but may dissolve all aspects of pre-pat-
terning that are not stabilized by innerva-
tion including aneural AChR expression.
Many new questions are raised con-
cerning the mechanisms by which the
pre-patterning of muscles can be
achieved, maintained and then disrupt-
ed in a precisely regulated manner. Pre-
pattern formation might be initiated
during early myoblast fusion by signal-
ing derived from the muscle founder
cells. Given that a founder cell might ini-
tiate this process, several other questions
arise. How is the pre-pattern maintained
until the time of innervation and, for
that matter, how are even small AChR
clusters formed during this process of
pre-patterning? It is now clear that MuSK
is necessary for the pre-patterning of
AChR synthesis and clustering, but the
precise involvement that MuSK has in
this process remains to be determined.
The regulated disruption of the pre-pat-
terning mechanism will also need inves-
tigation. Does it involve a novel
nerve-derived factorNDFor an
activity-dependent program initiated
upon arrival of the nerve and release of
meaningful event, fans listening to the
radio can translate the commentators
words back into a dynamic mental movie
of the ongoing game.
Many everyday activities, such as mak-
ing a bed or ironing a shirt, also consist of
a continuous stream of actions. Yet what
we tend to perceive is a series of discrete,
meaningful events. How does the human
brain parse the dynamic flow of action in
the world into distinct perceptual events?
A functional brain imaging study by
Zacks and colleagues1 in this issue provides
nature neuroscience volume 4 no 6 june 2001
transmitter? Finally, there are larger ques-
tions regarding the importance of pre-
patterning. Is pre-patterning of the
muscle necessary for synapse formation?
This seems unlikely given that synapse
formation occurs in vitro, where pre-pat-
terning is not observed3. Nevertheless,
pre-patterning of the muscle fiber could
be involved in targeting of axons to the
endplate region of the muscle fiber, if
molecules involved in target recognition
were similarly pre-patterned.
1. McMahan, U. J. Cold Spring Harb. Symp.
Quant. Biol. 55, 40718 (1990).
2. Nitkin, R. M. et al. J. Cell Biol. 105, 24712478
(1987).
3. Sanes, J. R. & Lichtman, J. W. Annu. Rev.
Neurosci. 22, 389442 (1999).
4. Gautam, M. et al. Cell 85, 525535 (1996).
5. DeChiara, T. M. et al. Cell 85, 501512 (1996).
6. Lin, W. et al. Nature 410, 10571064 (2001).
7. Yang, X. et al. Neuron (in press).
8. Yang, X. et al. Science 287, 131134 (2000).
9. Harris A. J. Phil. Trans. Soc. Lond. B Biol. Sci.
293, 287314 (1981).
10. Zhou, H. et al. J. Cell Biol. 146, 11331146
(1999).
11. Jones, G. et al. J. Neurosci. 19, 33763383
(1999).
12. Watty, A. et al. Proc. Natl. Acad. Sci. USA 97,
45854590 (2000).
13. Burden, S. J. Genes Dev. 12, 133148 (1998).
14. Sandrock, A. W. Jr. et al. Science 276, 599603
(1997).
15. Meier, T. et al. J. Cell Biol. 141, 715716 (1998).
important new clues to this question. The
authors first had subjects in an MR scan-
ner passively watch movies of an actor per-
forming an everyday activity such as
ironing a shirt (Fig. 1) or playing the saxo-
phone (Fig. 2). In two later scans, the same
movies were shown again, but this time,
subjects were instructed to press a button
whenever they perceived the end of one
event and the beginning of another. In one
of these later scans, subjects were asked to
divide the movie into the largest units that
seemed natural and meaningful; in the
other, the smallest units. This experimen-
tal design allowed the authors to look at
brain activity during coarse and fine event
boundaries and to test whether similar
activity occurred in the earlier passive
viewing scans when no task was required.
Activation during the passive viewing scans
might indicate that the brain automatical-
ly parses events in the absence of any
explicit instructions to do so.