LTD Induction Causes Morphological
Changes of Presynaptic Boutons
and Reduces Their Contacts with Spines
Nadine Becker,1Corette J. Wierenga,1,3Rosalina Fonseca,1,2,3Tobias Bonhoeffer,1and U. Valentin Na ¨gerl1,*
1Max Planck Institute of Neurobiology, Am Klopferspitz 18, 82152 Mu ¨nchen-Martinsried, Germany
2Present address: Gulbenkian Institute of Science, Rua da Quinta Grande 6, 2780-156 Oeiras, Portugal
3These authors contributed equally to this work
Activity-dependent changes in the synaptic connec-
tions of the brain are thought to be important for
learning and memory. Imaging techniques have en-
abled the examination of structural rearrangements
during activity-dependent processes at the synapse.
While many studies have examined structural
changes of dendritic spines, little is known about
structural plasticity of presynaptic boutons. We
therefore examined how axonal boutons are affected
during long-term depression (LTD). We used time
lapse two-photon laser scanning microscopy and
extracellular field recordings to monitor simulta-
neously synaptic morphology and activity for up to
five hours in mouse organotypic hippocampal slice
cultures. LTD induction dramatically increased the
turnover of presynaptic boutons, while decreasing
the number of putative synaptic contacts between
Schaffer collateral boutons and spines of CA1 pyra-
tic contribution to activity-dependent morphological
plasticity and provide opportunities for studying the
molecular mechanisms of the structural remodeling
of synaptic circuits.
Experience-dependent changes in synaptic connections in the
brain are thought to play a key role in learning and memory as
ment. A number of imaging studies in recent years have revealed
ity, such aslong-term potentiation (LTP) or long-term depression
(LTD), and structural plasticity at individual synapses (Yuste and
Bonhoeffer, 2001, 2004). For instance, it was shown that the
induction of LTP in hippocampal slices leads to the growth of
new spines (Engert and Bonhoeffer, 1999; Maletic-Savatic
et al., 1999; Toni et al., 1999) and ultimately to the formation of
to cause dendritic spines of potentiated synapses to increase in
size (Lang et al., 2004; Matsuzaki et al., 2004). In addition, the
converse effect was also demonstrated; the induction of LTD
causes spines on CA1 pyramidal neurons in hippocampal slices
to retract and/or shrink (Na ¨gerl et al., 2004; Zhou et al., 2004).
These and other studies support the idea that morphological
changes at the level of dendritic spines provide a potential struc-
tural basis of how transient changes in synaptic strength are
made long lasting.
In contrast to the plasticity of dendritic spines, activity-depen-
dent structural plasticity at the presynaptic site has not been
studied in great detail. While a number of recent studies have
looked at the plasticity of presynaptic boutons in adult and
developing hippocampal and cortical tissue (De Paola et al.,
2003, 2006; Konur and Yuste, 2004; Deng and Dunaevsky,
2005; Umeda et al., 2005; Stettler et al., 2006), most of these
studies have concentrated on plasticity under unstimulated con-
ditions. Therefore, the behavior of presynaptic boutons under
the conditions of classic plasticity paradigms known to induce
synaptic plasticity is still largely unknown.
We examined whether and how presynaptic boutons are
affected by the activity-dependent plasticity paradigm of LTD
induction. In addition, we performed two-color two-photon time
lapse imaging to visualize the structural dynamics of boutons
and associated spines in hippocampal organotypic slice cul-
tures. Since LTD-inducing low-frequency stimulation is known
rons, we examined how boutons of CA3 axons that were in close
association with these spines are affected by this paradigm. Our
data show that the effect of LTD on boutons is, if anything, larger
than on postsynaptic spines. Beyond the well-established role
of dendritic spines in the activity-dependent modifiability of
synaptic connections, our data reveal a significant presynaptic
Imaging Activity-Dependent Structural Plasticity
of Boutons and Spines
We combined two-photon microscopy with electrophysiological
recordings to investigate the structural dynamics of boutons of
CA3 axons subjected to the classic plasticity paradigm of
long-term depression (LTD). At the beginning, we checked the
synaptic nature of the axonal varicosities we set out to study.
590 Neuron 60, 590–597, November 26, 2008 ª2008 Elsevier Inc.
that the vast majority of varicosities was positive for the presyn-
aptic marker protein synapsin (89% ± 3%, n = 191 varicosities in
6 slices), indicating that in our preparation most varicosities
represent functional release sites (Figure 1A). Whether or not a
varicosity was likely to be a release site did not depend on its
size, since also 81% of the smallest varicosities (<0.5 mm3)
were synapsin positive. In addition, immunostainings against the
vesicular glutamate transporter (VGlut1) revealed a high degree
in 7 slices, data not shown), further corroborating the functional
nature of the majority of axonal varicosities.
Second, we looked at the varicosities on the ultrastructural
level by serial section electronmicroscopy (ssEM). To this end,
we loaded individual CA3 pyramidal neurons with a fluorescent
dye and biocytin permitting us to answer the question whether
the varicosities are bona fide synapses. Out of a total of 23 var-
icosities we have examined by ssEM, 21 (91%) varicosities
showed clear ultrastructural signs of synapses (as defined by
the combined presence of  clusters of synaptic vesicles, 
a synaptic cleft, and  a postsynaptic density in the apposing
membrane), while for the remaining two boutons no clear evi-
dence could be observed. Figure 1B shows single EM sections
of two representative varicosities identified by light microscopy,
revealing that they are part of synapses.
Third, besides the structural evidence, we wanted to check
whether the varicosities would also exhibit Ca2+concentration
rises associated with action potential (AP) firing, which is an
important criterion for their functionality. To this end, we filled
individual CA3 pyramidal neurons with the Ca2+indicator dye
Fluo-5F via a patch pipette, which was also used to evoke APs
Figure 1. Axonal Varicosities Mark Functional
Synapses on Schaffer Collaterals
(A) Maximal intensity projection of three subsequent
confocal sections (Dz = 0.5 mm) of a hippocampal slice
culture, immunostained for a-synapsin displayed in
magenta and for eGFP in green (overlapping pixels
are white). Inset shows zoom-in on axonal stretch.
Scale bar, 2 mm.
(B) Correlated light and ssEM, the upper panel shows
the light microscopic image of a stretch of a labeled
axon, the middle panel shows the corresponding
stretch reconstructed by ssEM, and the lower two
panels show EM images of two varicosities (b) that
form asymmetric synapses with postsynaptic spines
(s), the arrowheads marking the postsynaptic densi-
ties. Scale bars, 10 mm, 1 mm and 200 nm.
(C) Ca2+imaging reveals AP-induced Ca2+signals in
boutons. Normalized fluorescence intensity traces;
colors correspond to the regions of interest depicted
in the image. Lines in respective colors indicate the
23 SD threshold.
Scale bars, 10 mm and 2 mm.
by brief current injections. We found that
AP-induced Ca2+transients could readily be
detected in 46 out of 47 boutons (from 15
different cells in 15 slice cultures) that were
examined (Figure 1C). The Ca2+transients
were restricted to the immediate vicinity of the boutons, with
the connecting axonal shafts showing no detectable rises in
Ca2+following the firing of three APs. This observation indicates
that functional voltage-gated Ca2+channels are enriched on the
membrane of the axonal varicosities, supporting the view that
LTD Induction Leads to Structural Plasticity
of Presynaptic Boutons
We used low-frequency stimulation (LFS) to induce LTD at CA3–
LFS resulted in robust LTD evidenced by a significant reduction
of normalized fEPSP slopes compared to baseline (LTD, 66% ±
2%, n = 11; p < 0.001; Figure 2B), which is comparable with the
levels of hippocampal LTD measured by extracellular recordings
previously (Dudek and Bear, 1992).
To study the effects of LTD on presynaptic boutons, we quan-
tified the number of appearing and disappearing boutons on the
axons of CA3 pyramidal neurons under baseline and LTD condi-
tions (Figures 2C–2G). LTD significantly increased the number of
induction compared with control conditions (boutons lost per
100 mm; LTD, 0.9 ± 0.3; n = 26 axons; control, 0.2 ± 0.1; n = 16
axons; p < 0.05; Figure 2G). At the same time, the rate at which
new boutons appeared after LTD was also increased almost
2-fold over control conditions (boutons gained per 100 mm;
LTD, 2.2 ± 0.3; n = 26 axons; control, 1.1 ± 0.2; n = 16 axons;
p < 0.01; Figure 2G). Thus, LTD induction significantly increased
the turnover rate of Schaffer collateral boutons (Table S1).
LTD Remodels Hippocampal Synapses
Neuron 60, 590–597, November 26, 2008 ª2008 Elsevier Inc. 591
To test whether the morphological plasticity of boutons is
specific for the LTD-inducing stimulation (LFS) or is associated
more generally with elevated levels of neuronal activity, we also
investigated slices with matched overall activity that did not in-
duce LTD (see Supplemental Experimental Procedures available
online) as well as unstimulated slices. We determined the turn-
conditions and found them to be statistically indistinguishable
(boutons lost per 100 mm; no-stimulus, 0.4 ± 0.1; n = 14 axons;
p = 0.15 compared to control; matched-activity, 0.4 ± 0.2;
n = 21 axons; p = 0.36 compared to control; control, 0.2 ± 0.1;
n = 16 axons; boutons gained per 100 mm; no-stimulus, 0.5 ±
0.2; n = 14 axons; p = 0.19 compared to control; matched-
activity, 0.7 ± 0.2; n = 21 axons; p = 0.21 compared to control;
control, 1.1 ± 0.2; n = 16 axons; Figure 2G). In addition, we
analyzed two interesting experiments in which LFS failed to
induce LTD, allowing us to ask the question whether the applica-
tion of the LFS alone is sufficient for the structural effects
we observed or whether the successful induction of synaptic
plasticity is required. We found that bouton turnover in those ex-
periments was not significantly different from control experi-
ments (bouton turnover when LFS failed to induce LTD, n = 16
axons, bouton gain, 0.56 ± 0.15; p = 0.05 compared to control;
bouton loss. 0.48 ± 0.14; p = 0.15), indicating that the enhanced
Figure 2. Structural Plasticity of Boutons after LTD Induction
(A) Schematic illustrating the experimental paradigm. Red pipette: extracellular pressure injection of Calcein red-orange AM in the CA3 area. Yellow pipettes:
stimulation (left) and recording (right) electrodes after labeling was finished.
(B)Normalized and averaged fEPSPslopes (mean ±SEM)for all control (gray symbols) and LTD experiments(black symbols). Inset showsraw data tracesbefore
(a) and after (b) induction of LTD.
(C and D) Maximum intensity projections (MIPs) of labeled Schaffer collateral axons.
(E) Higher magnification of the box marked in (D) showing the emergence of a bouton (arrowhead). The graphs next to each example show the corresponding
intensity plots along the parent axon before (gray) and after (black) LTD induction. Arrowheads indicate the intensity peaks of the respective boutons. Eleven
sections were used for the MIPs in (C) and (D), and nine sections in (E) and (F), the Dz spacing was 0.5 mm. All scale bars 2 mm, unless otherwise indicated.
(F) Higher magnification of the box marked in (C) showing the loss of two boutons (arrowheads) after LTD induction.
(G) Summary of bouton turnover for the various experimental conditions (bouton gain in red; bouton loss in blue).
(H) Time course of bouton turnover (mean ± SEM; bouton gain in red; bouton loss in blue).
(I) Volume (mean ± SEM) of stable control boutons (gray bar), lost and gained boutons after LTD (black bars).
(J) Distance to the nearest neighboring bouton (median ± SD) for all (gray bar), simulated lost and gained (sim.; white bars), and measured lost and gained (meas.;
black bars) boutons after LTD. In all panels and following figures: asterisks indicate significant differences: *p < 0.05; **p < 0.01; ***p < 0.001.
the image. Lines in respective colors indicate the 23 SD threshold.
LTD Remodels Hippocampal Synapses
592 Neuron 60, 590–597, November 26, 2008 ª2008 Elsevier Inc.
Under LTD as well as under control conditions there was a net
gain of boutons over the imaging period. Of the paradigms
tested only the LTD-inducing LFS reliably induced presynaptic
morphological plasticity, enhancing both bouton gain and loss.
The time course of the number of boutons gained or lost reveals
that the onset of the increase in bouton turnover was delayed by
about 2 h with respect to the start of the stimulation (Figure 2H).
In order to characterize this presynaptic structural plasticity in
more detail, we measured the volume of the newly gained or lost
boutons and examined their location along the axon. Boutons
that appeared or disappeared after LTD (termed ‘‘plastic’’ bou-
tons) were significantly smaller than stable control boutons
(Figure 2I and Table S2), suggesting a higher stability for larger
neighboring bouton. When we compared these values for plastic
boutons and stable boutons, we found that nearest-neighbor
distances were significantly shorter for plastic boutons (median
± SD; gained, 1.5 ± 0.8 mm, n = 24 axons, p < 0.001; lost, 1.7 ±
1.1 mm, n = 11 axons, p < 0.05; all stable, 2.5 ± 0.7 mm, n = 26
axons; Mann-Whitney U test; Figure 2J). However, if new bou-
tons are inserted randomly between existing boutons, they are
expected to have shorter nearest-neighbor distances than the
existing boutons on a given axon. We therefore tested simulated
random insertion and removal of boutons between the stable
boutons on each axon. Based on the nearest-neighbor distance,
gained boutons could not be distinguished from randomly in-
serted boutons (median ± SD; simulated gained, 1.6 ± 0.3 mm,
n = 2600, p = 0.84). Lost boutons were closer to their neighbors
than expected from simulated random removal (simulated lost,
2.4 ± 0.7 mm, n = 260, p < 0.05). This suggests that boutons
disappeared preferentially if they were close to other boutons.
In order to assess the functionality of newly formed varicosi-
ties, we extended the Ca2+experiments described above by im-
aging Ca2+transients in new varicosities formed after LFS. We
detected AP-induced, spatially restricted Ca2+signals in all
(n = 15) newly formed boutons. Interestingly, in almost all of
them (14) Ca2+transients could be detected as early as the first
timepointtested (Figure2K),whichwas typically within 30minof
their initial appearance. In one of the new varicosities, Ca2+tran-
sients could not be detected 30 min, but only 50 min after the
new varicosity had first become visible. Taken together, the
Ca2+imaging data indicate that voltage-gated Ca2+channels
are rapidly assembled in the membranes of newly formed pre-
synaptic structures, making it likely that they are in the process
of assembling the molecular machinery needed for functional
LTD-Induced Structural Plasticity in Bouton-Spine
Pairs: Elevated Turnover and Net Loss
of Bouton-Spine Pairs
eral boutons that were associated with spines of CA1 pyramidal
neurons. To this end, we labeled axons and dendrites by spec-
trally distinct dyes, allowing us to examine activity-dependent
pre- and postsynaptic structural plasticity simultaneously (see
Experimental Procedures for details on criteria of ‘‘contact’’).
While the majority of bouton-spine pairs was stably associated
pairs (106 of 685) showed structural changes after LTD induction
a 5-fold increase over baseline conditions (Figure 3D). Besides
various pre- and/or postsynaptic structural rearrangements
that caused the breakup or formation of contacts, a substantial
fraction of bouton-spine pairs was lost because the bouton dis-
appeared (33% of lost pairs; Figure 3B). In parallel, new pairs
were gained by the appearance of a new bouton in contact
with a spine, (50% of gained pairs; Figure 3C). The loss or gain
of spines was less frequently responsible for the loss or gain of
bouton-spine pairs (18% and 13%, respectively). Importantly,
we never observed the retraction of a spine where the partner
bouton would be ‘‘pulled’’ along to the dendritic shaft, suggest-
shaft synapses. These results show that LTD induction also
affects boutons that were associated with CA1 spines.
Presynaptic Boutons Are Affected by the Plasticity
of Contacting Spines
Finally, we examined whether boutons or spines are affected if
their associated partner disappears or a new partner appears.
We measured the volumes of boutons and spines whose associ-
ated partner was stable, lost or gained before and after LTD.
Interestingly,the volumeofboutons decreasedsignificantly after
their associated spines were lost (Figures 4A–4C and Table S2),
Figure 3. LTD-Induced Presynaptic Structural Plastic-
ity in Identified Bouton-Spine Pairs
All images are single sections, chosen for best focal plane.
White arrows indicate the bouton-spine pair of interest.
(A) Stable bouton-spine pair under unstimulated control
(B and C) Examples of bouton loss (B) and bouton gain (C) in
bouton-spine pairs after LTD induction.
(D) Percentage of the total number of bouton-spine pairs lost
(blue) or gained (red) under LTD and control conditions
(mean ± SEM).
LTD Remodels Hippocampal Synapses
Neuron 60, 590–597, November 26, 2008 ª2008 Elsevier Inc. 593
while the volumes of boutons contacting stable spines were
unaffected after LTD. Boutons right next to (but not contacting)
disappearing spines had stable volumes, excluding that the
reduced bouton volumes were due to unspecific regional shrink-
age. The converse effect, i.e., a volume increase for boutons
after coming into contact with a new spine, did not quite reach
to be affected at all by the appearance or disappearance of an
associated bouton (Figure 4D and Table S2).
The observation that the retraction of a spine is specifically
associated with a volume reduction in the associated bouton
suggests that those bouton-spine pairs had been functionally
connected. Moreover, these results suggest that boutons are
more affected by plastic changes of their associated partners
Our study aimed at examining activity-dependent structural
plasticity from a presynaptic vantage point. We used dual-label
two-photon time-lapse imaging to monitor Schaffer collateral
in organotypic hippocampal slice cultures. First, we demon-
strated that LTD induction led to a significant remodeling of
studies that focused on postsynaptic structural dynamics.
Second, we showed that LTD induction reduced the number of
boutons that are associated with spines, suggesting that some
CA3–CA1 contacts are lost.
LTD Induces Structural Plasticity
of Presynaptic Boutons
Our experiments show that the induction of LTD led to a pro-
tons. About 11% of boutons imaged during the experiment were
impacted, being either gained or lost. Under control conditions,
may reflect the developmental stage of the tissue with new bou-
tons being continuously added. The observed increased bouton
gain after LTD induction seems to contrast with the net loss of
Figure 4. Characterization of the Volume Changes of Bou-
tons Associated with Plastic Spines
(A) Example of spine loss from a bouton-spine pair.
(B) Changes in bouton volumes for stable pairs (control) and pairs in
which the spine was lost or gained after LTD. Neighboring boutons
that were close but not in contact with the retracting spines did not
undergo any significant change in volume (striped gray bar).
(C) Volume changes shown in (B) plotted for individual boutons
(n = 10).
(D) Changes in spine volumes (mean ± SEM) for stable pairs (control)
and pairs in which the bouton was lost or gained after LTD.
contacts between boutons of CA3 axons and spines on
(De Paola et al., 2006) in the sensory cortex (Knott et al.,
2006) can form synapses, it is likely that at least some of
them ultimately will turn into functional synapses. It will
be important to determine the identity of their postsynaptic part-
ner. For instance, new boutons could form synapses with inhibi-
tory interneurons, which would contribute to an overall decrease
in synaptic transmission from CA3 to CA1. Alternatively, the new
boutons could have split off from existing boutons and may be
part of a dynamic pool of orphan presynaptic constituents that
can be mobilized as a rapid expression mechanism of synaptic
plasticity, as suggested by Krueger et al. (2003). While the pres-
ence of Ca2+transients in the new boutons indicates that they
may be functional, the questions of whether and when they can
The fact that bouton loss takes place in close proximity to ex-
isting boutons is interesting in the light of several recent reports
that neighboring presynaptic terminals can share proteins or
even larger structural complexes (Krueger et al., 2003; Darcy
et al., 2006; Tsuriel et al., 2006) and that this can be done in an
activity-dependent fashion (Waters and Smith, 2002; Vanden
Berghe and Klingauf, 2006). Together with our finding that bou-
ton addition and removal can both occur on the same axons,
this is consistent with a model in which structural elements liber-
ated at one site duringbouton removal can be used atother sites
where new boutons are built.
LTD-Induced Reduction of Contacts between Schaffer
Collaterals and CA1 Pyramidal Neurons
Our experiments show that LTD induction led to a considerable
loss of Schaffer collateral boutons that were closely associated
with CA1 spines. Although a physical contact between a bouton
and a spine is necessary for a synaptic connection it is by no
that a significant fraction of contacts we imaged over time repre-
ing bouton is suggestive of a functional contact. Second, the ob-
servationthatvirtually all bouton-spine pairsmade contactatthe
are typically formed at the head of the spine in a one-to-one rela-
tionshipbetweenCA1 spinesand boutonsof Schaffercollaterals
(Harris and Stevens, 1989; Schikorski and Stevens, 1997). Third,
our EM data indicate that most of the morphological varicosities
LTD Remodels Hippocampal Synapses
594 Neuron 60, 590–597, November 26, 2008 ª2008 Elsevier Inc.
voltage-gated Ca2+entry. This is in line with previous reports
showing that almost all boutons of mature neurons are functional
(Yao et al., 2006). In addition, more than 95% of all spines form
synapses, at least in the neocortex (Arellano et al., 2007).
LTD induction increased the number of newly formed bouton-
spine pairs by 4%, but many more pairs were lost (11%). If these
may represent a structural implementation for the weakening of
CA3–CA1 synaptic transmission. However, as the onset of the
activity-dependent structural dynamics is delayed compared
to expression of LTD, it isunlikely that contact loss playsa signif-
icant role for the early part of LTD. In any case, LTD induction
appears to lead to a reorganization of pre- and postsynaptic
elements, effecting a functional rewiring of the hippocampal net-
work. This view is also supported by a recent study (Bastrikova
et al., 2008) that provides evidence that LFS is associated with
a reduction in bouton-spine contacts.
large as the decrease in the number of contacts, suggesting that
other nonstructural LTD mechanisms, such as weakening of
existing synapses without removing them, are also operational.
Moreover, a discrepancy would be expected in as much as not
all labeled fibers were actually stimulated, diluting the structural
effect. In any case, the relationship between the number of syn-
apses and their combined strength is likely to be highly complex,
and therefore one would not expect to find a linear relationship
between structural plasticity and changes in synaptic transmis-
sion. Therefore, our findings are likely to have physiological
relevance, especially if the structural changes occurred within
defined synaptic pathways, considering that coactivation of just
a few synapses can lead to AP firing.
Our data show that structural plasticity is a feature that is not
specific to dendritic spines. In fact, presynaptic structural
changes may even play a bigger role than spine changes given
that the gain or loss of contacts was frequently due to bouton
changes. Irrespective of its mode, contact loss must entail the
downregulation of cell adhesion that otherwise maintains the
structural integrity of synaptic contacts (Garner et al., 2000). Un-
molecular level will be an important challenge for future studies.
Organotypic Hippocampal Slice Cultures and Recording Solutions
Hippocampal slices (400 mm thick) were prepared from postnatal day 5–7 C57
BL/6 mice, embedded in a plasma clot on glass coverslips, and incubated in a
roller incubator at 35?C, according to the Ga ¨hwiler method (Ga ¨hwiler, 1981).
The slice cultures used in the experiments were maintained 10 to 17 days
in vitro after the preparation. For the experiments, slice cultures were trans-
ferred into a heated recording chamber (35?C), where they were continuously
perfused with carbogenated (95% O2, 5% CO2) artificial cerebrospinal fluid
(ACSF) at pH 7.4 containing 126 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl2,
1.3 mM MgCl2, 10 mM glucose, 1.25 mM NaH2PO4, 26 mM NaHCO3, 1 mM
pyruvate, and 1 mM trolox (Sigma, Munich, Germany).
Labeling and Microscopy
Two-color time-lapse two-photon laser scanning microscopy (TPLSM) was
used to monitor the morphology of CA3 axons and spines of CA1 pyramidal
neurons over time. The red excitation light (l = 790 nm) from a 5 W Mira-Verdi
laser system(Coherent, SantaClara, CA) wasrouted through alaser scanhead
(YanusII, TILL Photonics, Gra ¨felfing,
(CH-700DCXR2638; LOT Oriel, Darmstadt, Germany), and a 403, 1.2 NA
water-immersion objective (Zeiss, Oberkochen, Germany) mounted on an
inverted IX70 microscope (Olympus, Hamburg, Germany). The power of the
excitation light was adjusted to 2 mW after the objective by an acousto-optical
modulator(Polytec, Waldbronn,Germany).The emitted fluorescence wassplit
by a suitable dichroic mirror (HQ572LP; LOT Oriel) into red and green fluores-
cence, filtered by adequate band-pass filters (red channel, HQ590/55; green
channel, HQ525/50; both filters from LOT Oriel), and detected by two external
photomultiplier tubes (R6357, Hamamatsu, Herrsching, Germany). Image
acquisition was performed by custom-programmed software (LabVIEW 8.2,
National Instruments, Austin, TX).
For the structural plasticity experiments, a glass micropipette filled with
0.5 mM Calcein red-orange AM (Invitrogen, Karlsruhe, Germany) diluted in
HEPES-buffered ACSF, connected to a Picospritzer (Parker Hannifin Corpora-
tion, Fairfield, NJ), was placed in the middle of the cell body layer of the CA3
area. The dye was injected into the tissue by applying brief pressure pulses
of 5 to 15 ms every 20 s. After 1 hr, about 30 to 40 CA3 neurons were labeled,
projecting predominantly into the stratum oriens. TPLSM was used to localize
the projection area of the labeled axons of the CA3 neurons in the CA1 region.
Two to four CA1 pyramidal neurons in this area were loaded for 2–3 min via
a patch pipette containing 4 mM Calcein green (Invitrogen), 120 mM K-Gluco-
nate, 10 mM KCl, 20 mM HEPES, 5 mM NaCl, and 12 mM Mg2+-ATP. The field
of view waschosen inthe region withoptimal overlap of theCA3pyramidalcell
axons and the CA1 pyramidal cell dendrites, which was always located within
the stratum oriens. 3D image stacks (spanning 140 mm in x,140 mm in y, and
25–40 mm in z using 1024 3 1024 pixels in xy and 0.5 mm steps in z) were ac-
the duration of the experiment. For the Ca2+imaging experiments, single CA3
cells were patched for up to 8 hr, the pipette containing internal solution (see
above), 100 mm Fluo-5F and 200 mm Alexa Fluor 568. Eighty image frames
of 5 mm 3 5 mm using 64 3 64 pixels were acquired in a single, optimal z plane
of one or two boutons with a frame rate of one image per 15 ms. APs were
evoked after 20 frames. For the EM reconstructions of the axons, single CA3
pyramidal neurons were briefly loaded via a patch pipette with Calcein green
and biocytin to produce an electron-dense label.
Field excitatory postsynaptic potentials (fEPSPs) were recorded from the cell
bodylayer nearthelabeled CA1pyramidal neurons,usingaglassmicropipette
filled with ACSF. CA3 pyramidal neurons were stimulated by brief (0.2 ms)
current pulses from astimulus isolator (WPI, Sarasota,FL) usinga glass micro-
Supplemental Experimental Procedures.
4D (x, y, z, t) image stacks were processed and analyzed using ImageJ (NIH,
Bethesda, MD), Imaris 5.1 (Bitplane, Zu ¨rich, Switzerland), and custom-pro-
grammed MATLAB software (Version 7.1, MathWorks, Natick, MA). Bouton
turnoverwasanalyzed blindly withrespecttocontroland LTD experiments. In-
dividual stacks of the red channel were visualized and filtered with a Gaussian
algorithm in ImageJ. For each experiment, two to four axons were analyzed
over lengths ranging from 55 to 130 mm. The selected axons were displayed
and analyzed as maximum intensity projections. We assessed the presence
of every bouton on the basis of intensity plots, which were calculated as the
average intensity over every line of pixels from a rectangle drawn along short
stretches of axon (Figures 2C and 2D). Boutons were defined by a peak inten-
code to simulate the random insertion and deletion of a single bouton onto
each of the axons we analyzed (100 simulated insertions and 10 simulated re-
movalsperaxon). Spineturnoveranalysiswascarried outonvolume-rendered
images of the green channel, which was spatially filtered by an edge-preserv-
ing algorithm using the Imaris software. In addition, individual sections were
analyzed to confirm if a spine was lost or gained. All spines visible in the field
of view were included in the analysis.
LTD Remodels Hippocampal Synapses
Neuron 60, 590–597, November 26, 2008 ª2008 Elsevier Inc. 595
For the analysis of bouton-spine pairs, contacting boutons and spines were
detected visually in single planes of image stacks with both channels overlaid
and filtered by an edge-preserving algorithm in Imaris. We analyzed bouton-
spine pairs undergoing structural changes, and a random sample of stable
pairs. Only those pairs were analyzed for which both bouton and spine were
centered in the same z section. The volumes of both partners of bouton-spine
pairs were determined by a MATLAB program in the following manner. Stacks
of raw images of both channels were read in as time series, filtered with an
adaptive Wiener algorithm and drift-corrected in all dimensions. Data were
binarized in each channel separately by an intensity threshold. Thresholds
were empirically defined by a semiautomatic two-step procedure. In the first
step, background was distinguished from neuronal structures such as den-
drites and axons (threshold = mean + 33 SD of intensity values of a local
200 3 200 pixel area). In the second step, boutons were distinguished from
axonal shafts (threshold = mean + 1.53 SD of axon pixel intensities). In this
way, thresholds were objective and they were kept constant over all time
points. Outlines of boutons and spines of putative contacts were determined
semiautomatically in the individual z sections of the thresholded images of
the separate channels. From these, the volumes of the individual structures
Data are reported as means ± SEM unless stated otherwise. Statistical
significance of the effects of LTD was calculated using two-tailed, unpaired
t tests. Statistical significance of the effect of changing boutons or spines on
their contact partner was determined using two-tailed, paired t tests. In cases
of multiple comparisons, p values were post hoc Bonferroni corrected. In total,
we analyzed 908 pairs of boutons and spines; of those,223 came from 5 slices
in control experiments and 685 came from 11 slices in LTD experiments.
See Supplemental Experimental Procedures.
Serial Section Electronmicroscopy and 3D Reconstructions
Slices were initially transferred into 35?C 0.1 M phosphate buffer (PB)-based
fixative containing 4% paraformaldehyde, 15% picric acid, and 0.5% glutaral-
dehyde and subsequently stored at 4?C for 4 hr on a shaker. Slices were re-
gradient of PB-based sucrose solutions, and then subjected to a freeze-thaw
cycle using liquid nitrogen. The biocytin label was revealed using a Vectastein
Elite ABC kit (Axxora, Gru ¨nberg, Germany) and 3,3-diaminobenzidenetetrahy-
drochloride (DAB) histology. Briefly, slices were incubated in ABC solution
overnight before the peroxidase reaction end product was revealed by DAB.
tetroxide solution (4%), followed by an uranylacetate (1%) containing ethanol
Switzerland). Sectionswerepreparedforlightmicroscopyandregions ofinter-
est were then observed under transmission EM (see Anderson et al.  for
details). Briefly, serial ultrathin sections were collected at 60 nm thickness on
Pioloform-coated single-slot copper grids (Bio-Rad, Hempstead, England).
ages were processed using the Reconstruct program (John C. Fiala, Boston
University) to outline the labeled structures for 3D reconstructions.
The Supplemental Data include three figures, two tables, and Supplemental
Experimental Procedures and can be found with this article online at http://
The authors acknowledge the collaboration of the BIZ (Ludwig-Maximilians
University Munich) and TILL Photonics GmbH and support from the Boeh-
ringer Ingelheim Fonds (N.B.), the National Research Council of Portugal
(R.F.), the Alexander von Humboldt Stiftung, and a Marie Curie European
fellowship (C.J.W.). We thank C. Huber, N. Sto ¨hr, F. Voss, and M. Braun for
technical assistance, R. Bopp for the electronmicroscopy (in the lab of
K.A.C. Martin, ETH Zu ¨rich, supported by SNF NCCR ‘Neural Plasticity and
Repair’), and T. Keck, T. Mrsic-Flo ¨gel, and V. Scheuss for comments on the
Accepted: September 3, 2008
Published: November 25, 2008
Anderson, J.C., Douglas, R.J., Martin, K.A., and Nelson, J.C. (1994). Map of
the synapses formed with the dendrites of spiny stellate neurons of cat visual
cortex. J. Comp. Neurol. 341, 25–38.
Arellano, J.I., Espinosa, A., Fairen, A., Yuste, R., and Defelipe, J. (2007).
Non-synaptic dendritic spines in neocortex. Neuroscience 145, 464–469.
Bastrikova, N., Gardner, G.A., Reece, J.M., Jeromin, A., and Dudek, S.M.
(2008). Synapse elimination accompanies functional plasticity in hippocampal
neurons. Proc. Natl. Acad. Sci. USA 105, 3123–3127.
Darcy, K.J., Staras, K., Collinson, L.M., and Goda, Y. (2006). Constitutive
sharing of recycling synaptic vesicles between presynaptic boutons. Nat.
Neurosci. 9, 315–321.
De Paola, V., Arber, S., and Caroni, P. (2003). AMPA receptors regulate dy-
namic equilibrium of presynaptic terminals in mature hippocampal networks.
Nat. Neurosci. 6, 491–500.
De Paola, V., Holtmaat, A., Knott, G., Song, S., Wilbrecht, L., Caroni, P., and
Svoboda, K. (2006). Cell type-specific structural plasticity of axonal branches
and boutons in the adult neocortex. Neuron 49, 861–875.
Deng, J., and Dunaevsky, A. (2005). Dynamics of dendritic spines and their
afferent terminals: spines are more motile than presynaptic boutons. Dev.
Biol. 277, 366–377.
Dudek, S.M., and Bear, M.F. (1992). Homosynaptic long-term depression in
area CA1 of hippocampus and effects of N-methyl-D-aspartate receptor
blockade. Proc. Natl. Acad. Sci. USA 89, 4363–4367.
Engert, F., and Bonhoeffer, T. (1999). Dendritic spine changes associated with
hippocampal long-term synaptic plasticity. Nature 399, 66–70.
Ga ¨hwiler, B.H. (1981). Organotypic monolayer cultures of nervous tissue.
J. Neurosci. Methods 4, 329–342.
Garner, C.C., Nash, J., and Huganir, R.L. (2000). PDZ domains in synapse
assembly and signalling. Trends Cell Biol. 10, 274–280.
Harris, K.M., and Stevens, J.K. (1989). Dendritic spines of CA 1 pyramidal cells
in the rat hippocampus: serial electron microscopy with reference to their
biophysical characteristics. J. Neurosci. 9, 2982–2997.
Knott, G.W., Holtmaat, A., Wilbrecht, L., Welker, E., and Svoboda, K. (2006).
Spine growth precedes synapse formation in the adult neocortex in vivo.
Nat. Neurosci. 9, 1117–1124.
Konur, S., and Yuste, R. (2004). Imaging the motility of dendritic protrusions
and axon terminals: roles in axon sampling and synaptic competition. Mol.
Cell. Neurosci. 27, 427–440.
apparatus is functional in the absence of dendritic contact and highly mobile
within isolated axons. Neuron 40, 945–957.
Lang, C., Barco, A., Zablow, L., Kandel, E.R., Siegelbaum, S.A., and Zakhar-
enko, S.S. (2004). Transient expansion of synaptically connected dendritic
spines upon induction of hippocampal long-term potentiation. Proc. Natl.
Acad. Sci. USA 101, 16665–16670.
Maletic-Savatic, M., Malinow, R., and Svoboda, K. (1999). Rapid dendritic
morphogenesis in CA1 hippocampal dendrites induced by synaptic activity.
Science 283, 1923–1927.
Matsuzaki, M., Honkura, N., Ellis-Davies, G.C., and Kasai, H. (2004). Structural
Na ¨gerl, U.V., Eberhorn, N., Cambridge, S.B., and Bonhoeffer, T. (2004). Bidi-
rectional activity-dependent morphological plasticity in hippocampal neurons.
Neuron 44, 759–767.
LTD Remodels Hippocampal Synapses
596 Neuron 60, 590–597, November 26, 2008 ª2008 Elsevier Inc.
Na ¨gerl, U.V., Ko ¨stinger, G., Anderson, J.C., Martin, K.A.C., and Bonhoeffer, T. Download full-text
(2007). Protracted synaptogenesis after activity-dependent spinogenesis in
hippocampal neurons. J. Neurosci. 27, 8149–8156.
Schikorski, T.,and Stevens, C.F. (1997).Quantitative ultrastructural analysis of
hippocampal excitatory synapses. J. Neurosci. 17, 5858–5867.
Stettler, D.D., Yamahachi, H., Li, W., Denk, W., and Gilbert, C.D. (2006). Axons
and synaptic boutons are highly dynamic in adult visual cortex. Neuron 49,
Toni, N., Buchs, P.A., Nikonenko, I., Bron, C.R., and Muller, D. (1999). LTP
promotes formationofmultiple spine synapses between asingle axonterminal
and a dendrite. Nature 402, 421–425.
Tsuriel, S., Geva, R., Zamorano, P., Dresbach, T., Boeckers, T., Gundelfinger,
E.D., Garner, C.C., and Ziv, N.E. (2006). Local sharing as a predominant deter-
minant of synaptic matrix molecular dynamics. PLoS Biol. 4, e271. 10.1371/
Umeda, T., Ebihara, T., and Okabe, S. (2005). Simultaneous observation of
stably associated presynaptic varicosities and postsynaptic spines: morpho-
logical alterations of CA3–CA1 synapses in hippocampal slice cultures. Mol.
Cell. Neurosci. 28, 264–274.
Vanden Berghe, P., and Klingauf, J. (2006). Synaptic vesicles in rat hippocam-
pal boutons recycle to different pools in a use-dependent fashion. J. Physiol.
Waters, J., and Smith, S.J. (2002). Vesicle pool partitioning influences presyn-
aptic diversity and weighting in rat hippocampal synapses. J. Physiol. 541,
Yao, J., Qi, J., and Chen, G. (2006). Actin-dependent activation of presynaptic
silent synapses contributes to long-term synaptic plasticity in developing
hippocampal neurons. J. Neurosci. 26, 8137–8147.
Yuste, R., and Bonhoeffer, T. (2001). Morphological changes in dendritic
spines associated with long-term synaptic plasticity. Annu. Rev. Neurosci.
Yuste, R., and Bonhoeffer, T. (2004). Genesis of dendritic spines: insights from
ultrastructural and imaging studies. Nat. Rev. Neurosci. 5, 24–34.
Zhou, Q., Homma, K.J., and Poo, M.M. (2004). Shrinkage of dendritic spines
associated with long-term depression of hippocampal synapses. Neuron 44,
LTD Remodels Hippocampal Synapses
Neuron 60, 590–597, November 26, 2008 ª2008 Elsevier Inc. 597