1
Hebbian spike-timing dependent plasticity at the cerebellar
2
input stage
3
Sgritta M1,3, Locatelli F1, Soda T1,4 Prestori F1, and D’Angelo E1,2
4 5 6
1
Department of Brain and Behavioral Sciences, University of Pavia, 27100 Pavia, Italy
7
2
Brain Connectivity Center, IRCCS C. Mondino, 27100 Pavia, Italy
8
3
European Brain Research Institute “Rita Levi Montalcini”, 00143 Rome, Italy
9
4
Museo Storico della Fisica e Centro Studi e Ricerche Enrico Fermi, 00184 Rome, Italy
10 11 12 13
Correspondence should be addressed to:
14
Prof. Egidio D'Angelo
15
University of Pavia
16
Dept of Brain and Behavioral Sciences
17
via Forlanini 6, 27100 Pavia, Italy
18
[email protected]
19 20
Number of pages: 38
21
Number of figures: 9
22
Number of words: 249-Abstract, 650-Introduction, 1499-Discussion
23 24
Running Title: STDP in the cerebellum
0
25
Acknowledgments. This work was supported by: European Union grant Human Brain Project (HBP-29
26
604102) to ED and Fermi grant [13(14)] to ED and TS. The authors declare no competing financial
27
interests.
28 29
Abstract
30
Spike-timing dependent plasticity (STDP) is a form of long-term synaptic plasticity
31
exploiting the time relationship between postsynaptic action potentials (AP) and EPSPs.
32
Surprisingly enough, very little was known about STDP in the cerebellum, although it is thought to
33
play a critical role for learning appropriate timing of actions. We speculated that low-frequency
34
oscillations observed in the granular layer may provide a reference for repetitive EPSP/AP phase
35
coupling. Here we show that EPSP-spike pairing at 6Hz can optimally induce STDP at the mossy
36
fiber - granule cell synapse in rats. Spike timing-dependent long-term potentiation and depression
37
(st-LTP and st-LTD) were confined to a ±25 ms time-window. Since EPSPs led APs in st-LTP
38
while APs led EPSPs in st-LTD, STDP was Hebbian in nature. STDP occurred at 6-10 Hz but
39
vanished above 50 Hz or below 1 Hz (where only LTP or LTD occurred). STDP disappeared with
40
randomized EPSP/AP pairing or high intracellular Ca2+ buffering and its sign was inverted by
41
GABA-A receptor activation. Both st-LTP and st-LTD required NMDA receptors, but st-LTP also
42
required reinforcing signals mediated by mGluRs and intracellular calcium stores. Importantly, st-
43
LTP and st-LTD were significantly larger than LTP and LTD obtained by modulating the frequency
44
and duration of mossy fiber bursts, probably because STDP expression involved postsynaptic in
45
addition to presynaptic mechanisms. These results thus show that a Hebbian form of STDP occurs
46
at the cerebellum input stage providing the substrate for phase-dependent binding of mossy fiber
47
spikes to repetitive theta-frequency cycles of granule cell activity.
48 49
1
50
Significance statement
51
Long-term synaptic plasticity is a fundamental property of the brain causing persistent
52
modifications of neuronal communication thought to provide the cellular basis of learning and
53
memory. The cerebellum is critical for learning the appropriate timing of sensorimotor behaviors
54
but whether and how appropriate spike patterns could drive long-term synaptic plasticity remained
55
unknown. Here, we show that this can actually occur through a form of spike-timing dependent
56
plasticity (STDP) at the cerebellar inputs stage. Pairing pre- and postsynaptic spikes at 6-10 Hz
57
reliably induced STDP at the mossy fiber - granule cell synapse, with potentiation and depression
58
symmetrically distributed within a ±25 ms time window. Thus, STDP can bind plasticity to the
59
mossy fiber burst phase with high temporal precision.
60 61
2
Introduction
62 63 64
The dualistic function of cerebellum as a "learning and a timing machine" has been early
65
recognized (Eccles, 1967; Marr, 1969). The cerebellum can learn and store information about
66
sensori-motor contingencies and thereby control behavior with millisecond precision (Timmann et
67
al., 1999; Osborne et al., 2007). Nonetheless, if and how timing of incoming signal control the
68
induction of long-term synaptic plasticity remained to be demonstrated.
69
Fundamental elements of spike-timing and long-term synaptic plasticity have been
70
recognized in the cerebellum and the granular layer, located at the cerebellar input stage, has been
71
proposed to exploit these mechanisms to regulate transmission and storage of complex spatio-
72
temporal patterns (D'Angelo and De Zeeuw, 2009; Gao et al., 2012). On the one hand, different
73
forms of long-term synaptic and non-synaptic plasticity have been reported at the mossy fiber -
74
granule cell synapse (D'Angelo, 2014), where LTP induction is favored by long high-frequency
75
bursts and strong membrane depolarization, while LTD induction is favored by short low-frequency
76
bursts and weak membrane depolarization (Sola et al., 2004; Gall et al., 2005; D'Errico et al., 2009).
77
On the other hand, the mossy fiber - granule cell synapse is endowed with specific molecular
78
mechanisms allowing high temporal precision (Silver et al., 1992; Cathala et al., 2005; Kanichay
79
and Silver, 2008) and controlling the delay and persistence of spike emission (Nieus et al., 2006).
80
Now the question is if and how spike-timing can drive synaptic plasticity. Key to understand this
81
relationship is the observation that granule cells show coherent low-frequency oscillations in rats
82
and monkeys in vivo (Pellerin and Lamarre, 1997; Hartmann and Bower, 1998; Courtemanche and
83
Lamarre, 2005; D'Angelo et al., 2009) so that spike timing-dependent plasticity (STDP) may be
84
able to detect the precise phase relationships between postsynaptic action potentials (APs) and
85
EPSPs.
3
86
STDP is a form of long-term synaptic plasticity that exploits the temporal relationship of
87
activity in pre- and postsynaptic neurons to regulate the switch between potentiation and depression
88
(called here st-LTP and st-LTD) (Artola and Singer, 1993; Abbott and Nelson, 2000; Caporale and
89
Dan, 2008; Markram et al., 2011). STDP typically requires repetitive EPSP/AP pairing at low
90
frequency with a “time window” of less than 100 ms, shows a sharp transition between st-LTP and
91
st-LTD, its magnitude is sensitive to the EPSP/AP interval (STDP degrades toward the edges of the
92
time window) and depends on spike back-propagation and membrane depolarization (Markram et
93
al., 1997; Bi and Poo, 1998; Zhang et al., 1998; Stuart and Häusser, 2001; Celikel et al., 2004;
94
Sjöström and Häusser, 2006). STDP is defined Hebbian when potentiation occurs with EPSPs
95
leading APs (st-LTP) and depression occurs with APs leading EPSPs (st-LTD) (Hebb, 1949), vice
96
versa it is called anti-Hebbian (in both cases, spike back-propagation into the dendrites down to the
97
synapses is needed, otherwise STDP in non-Hebbian). Various forms of STDP have been observed
98
in several brain areas including neocortex, striatum, hippocampus and olfactory system (Debanne et
99
al., 1998; Feldman, 2000; Sjöström et al., 2001; Froemke and Dan, 2002; Bender et al., 2006;
100
Pawlak and Kerr, 2008) (Cassenaer and Laurent, 2007). A form of non-Hebbian STDP has been
101
observed at the parallel fiber Purkinje cell synapse of cerebellum (Safo and Regehr, 2008; Piochon
102
et al., 2012) and anti-Hebbian STDP has been reported in the corresponding synaptic relay of
103
cerebellum-like structures [electrosensory lobe of fishes and dorsal cochlear nucleus (Bell et al.,
104
1997; Tzounopoulos et al., 2004; Wu et al., 2015)].
105
In this work, the induction of STDP over a theta-frequency cycle was investigated at the
106
mossy fiber-granule cell synapse of cerebellum. This paradigm reliably led to Hebbian STDP that
107
exploited induction and expression pathways that overlapped in part with those of LTP and LTD
108
induced using duration/frequency-modulated spike bursts (Gall et al., 2005; D'Errico et al., 2009;
109
D'Angelo, 2014). Thus, this study suggests that STDP could have a relevant role in linking spike
110
timing and learning at the input stage of the cerebellar circuit.
111 4
112
Methods
113 114
The experiments reported in this paper have been conducted on 17-to 22-day-old (P0=day of
115
birth) Wistar rats of either sex according to protocols approved by the local ethical committee and
116
by the European Commission (under Human Brain Project). The rats were anesthetized with
117
halothane (Sigma, St.Louis, MO) and killed by decapitation in order to remove the cerebellum for
118
acute slice preparation according to a well-established technique (D'Angelo et al., 1995; Armano et
119
al., 2000; Gall et al., 2005).
120 121
Slices preparation and solutions
122
The cerebellar vermis was isolated and fixed on the vibroslicer’s stage (Leica VT12005)
123
with cyano-acrylic glue. Acute 220 m-thick slices were cut in the parasagittal plane from in cold
124
(2–3ºC) cutting-solution containing (in mM): potassium gluconate 130, KCl 15, ethylene glycol-bis
125
(β-aminoethyl ether) N,N,N’,N’-tetraacetic acid (EGTA) 0.2, N-2-hydroxyethyl piperazine-N-2-
126
ethanesulphonic acid (Hepes) 20, glucose 10, pH 7.4 with NaOH. Slices were incubated for at least
127
1 h at room temperature in oxygenated bicarbonate-buffered (Kreb’s) solution containing (mM):
128
NaCl 120, KCl 2, MgSO4 1.2, NaHCO3 26, KH2PO4 1.2, CaCl2 2, glucose 11 (pH 7.4 when
129
equilibrated with 95%O2–5%CO2), before being transferred to a recording chamber mounted on the
130
stage of an upright microscope (Zeiss, Oberkochen, Germany). The slices were perfused with
131
oxygenated Kreb’s solution and maintained at 32°C with a Peltier feedback device (TC-324B,
132
Warner Instrument Corp., Hamden, CT, USA). For recordings, Kreb’s solution was added with the
133
GABAA receptor antagonist SR95531 (10 M; Tocris Bioscience). Local perfusion with Kreb’s
134
solution and 10 M SR95531 was commenced before seal formation and was maintained until end
135
of recording.
5
136
In a set of experiments SR95531 (gabazine) was omitted from the extracellular solution. Patch
137
pipettes were pulled from borosilicate glass capillaries (Hilgenberg, Malsfeld, Germany) and, when
138
filled with the intracellular solution, had a resistance of 7-9 MΩ before seal formation. The whole-
139
cell recording pipettes were filled with the following solution (in mM): potassium gluconate 126,
140
NaCl 4, HEPES 5, glucose 15, MgSO4 · 7H2O 1, BAPTA-free 0.1, BAPTA-Ca2+ 0.05, Mg2+-ATP
141
3, Na+-GTP 0.1, pH 7.2 adjusted with KOH. This solution maintained resting free- [Ca2+] at 100
142
nM. In a different series of recordings BAPTA was increased to 10 mM.
143 144
Electrophysiological recordings
145
Recordings were performed with Multiclamp 700B [-3dB; cutoff frequency (fc),10 kHz],
146
sampled with Digidata 1550 interface, and analyzed off-line with pClamp10 software (Molecular
147
Devices, CA, USA). Just after obtaining the cell-attached configuration, electrode capacitive
148
transients were carefully cancelled to allow for electronic compensation of pipette charging during
149
subsequent current-clamp recordings (D'Angelo et al., 1995; D'Angelo et al., 1999). Mossy fiber
150
stimulation was performed with a bipolar tungsten electrode (Clark Instruments) via a stimulus
151
isolation unit. The stimulating electrode was placed over the central fiber bundle in the cerebellar
152
lamina to stimulate the mossy fibers, and 200 μs step current pulses were applied at the frequency
153
of 0.33 Hz (in specific experiments, paired-pulse stimulation at 20 ms inter-pulse was used). From
154
comparison with previous data obtained in comparable conditions (Sola et al., 2004; D'Errico et al.,
155
2009), between 1 and 2 mossy fibers were stimulated per granule cells. After evoking excitatory
156
postsynaptic currents (EPSCs) at -70 mV at the test frequency for 10 min (control period), the
157
recording was switched to current clamp. STDP protocols consisted of pairings of presynaptic and
158
postsynaptic stimulations (60 times at either 1 Hz, 6 Hz or 10 Hz) with a time interval (Δt) that
159
could either be fixed (phase-locked) or random. During random stimulation, EPSPs always occurred
160
regularly at 6 Hz but the AP occurred casually between +25 ms and -25 ms. Therefore, the number
161
of EPSPs and APs was the same as in phase-locked STDP experiments. In some experiments we 6
162
varied Δt between −100 and +100 ms to fully explore the time window of plasticity induction. The
163
time interval Δt was defined as the time between the onset of the EPSP and the peak of the AP.
164
Presynaptic activity corresponded to mossy fiber stimulations and postsynaptic activity to an action
165
potential evoked by a brief (5 msec) depolarizing current step (40 pA) in the recorded granule cell.
166
Long-term synaptic efficacy changes were measured after 20 min. After delivering the STDP
167
protocol, re- voltage-clamp at -70 mV was reestablished and stimulation was restarted at the test
168
frequency.
169
The stability of whole-cell recordings can be influenced by modification of Rs. To ensure
170
that Rs remained stable during recordings, passive electrode-cell parameters were monitored
171
throughout the experiments. The granule cell behaves like as a lumped electrotonic compartment
172
and can therefore be treated as a simple resistive - capacitive system, from which relevant
173
parameters can be extracted by analyzing passive current relaxations induced by 10 mV
174
hyperpolarizing steps from a holding potential of -70 mV applied 10 ms before each mossy fiber
175
stimulus (D'Angelo et al., 1995; Silver et al., 1996; D'Angelo et al., 1999). The voltage clamp time
176
constant, τvc, was estimated from bi-exponential fitting to current transients elicited by voltage
177
steps. The 3-dB cut-off frequency of the electrode-cell system was calculated as fvc = (2π • τvc)-1 =
178
2.1 ± 0.1 kHz (n=120). Membrane capacitance Cm = 3.2 ± 0.1 pF (n=120) was measured from the
179
capacitive charge (the area underlying current transient), membrane resistance R m= 1.6 ± 0.1 GΩ
180
(n=120) was obtained from the steady-state current flow, also yielding series resistance Rs = τvc/Cm
181
= 30.3 ± 1.4 MΩ (n=120). In the cells used in this paper, these values did not significantly change
182
after 30 minutes attesting recording stability. Otherwise, variation of series resistance (Rs) >20%
183
leds to the rejection of the experiment.
184 185
Post synaptic currents analysis and data processing
186
All EPSCs were digitally filtered at 1.5 kHz and analyzed off-line. EPSC amplitude was
187
measured as the difference between EPSC peak and the current level just before stimulation. The 7
188
paired-pulse ratio (PPR) between the first and second EPSC in a sequence was PPR=EPSC2/EPSC1.
189
In specific experiments, the acquisition program automatically alternated EPSC with background
190
activity recordings (1 s and 9 s, respectively), from which miniature EPSCs were detected
191
(mEPSC). mEPSC analysis was performed automatically with Clampfit software, setting a proper
192
threshold for event detection. A further visual inspection of detected signals allowed us to reject
193
noisy artifacts. A 10-min period was used to evaluate mean mEPSC amplitude and coefficient of
194
variation (CV).
195 196
Data are reported as mean ± MSE, statistical comparisons are done using unpaired Student’s t-test and differences are considered statistically significant at p (MB/MA) (eq.3). The inequality leads to a topological representation of
218
neurotransmission changes (see Fig. 8) and has been extensively used to interpret the plasticity
219
mechanism (Bekkers and Stevens, 1990; Malinow and Tsien, 1990; Sola et al., 2004; D'Errico et al.,
220
2009). The graph of eq. (3) has the following properties. For an M increase:
221
(i)
when (CVB/CVA)-2 > (MB/MA) both n and p can increase,
222
(ii)
when (CVB/CVA)-2 = (MB/MA) only n can increase,
223
(iii)
when (CVB/CVA)-2 < (MB/MA) neither n nor p can increase implying an increase in q. A
224
pure increase in q will lie on the axis (CVB/CVA)-2 =1.
225
For an M decrease, complementary parameter changes are predicted. A confident interpretation of
226
the plot can be achieved only providing independent estimates of q and cvq, since simultaneous
227
changes of the parameters cannot be excluded by inspection (von Kitzing et al., 1994).
228 229
Calcium imaging
230
Calcium imaging was performed as reported previously in this same preparation, (Gall et al.,
231
2005; D'Errico et al., 2009), by using Oregon green BAPTA-1 (OG1, Molecular Probes). Briefly,
232
the intracellular solution was added with 0.2 mM OG1 in substitution to the BAPTA/Ca2+ buffer.
233
Granule cells were identified with a x63, 0.9 NA water immersion objective (Olympus, Hamburg,
234
Germany). Digital fluorescence images were obtained using an excitation light source from T.I.L.L.
235
Photonics (Planegg, Germany) controlled through Axon Imaging Workbench AIW5.2 (INDEC
236
Systems). Images were acquired with a 50 ms exposure/image at video rate. Acquisition (frame rate
237
= 2 Hz) started after allowing >2 min for dye loading in the neuron. After this time, the resting
238
fluorescence (F0) varied by less than 5% in each analyzed cell region for the entire recording time
239
and the background fluorescence (B0) was also stationary. Each protocol was preceded by 2 s of 9
240
baseline, during which no stimulus was given, to allow subsequent F0 (basal [Ca2+]i) evaluation.
241
[Ca2+]i changes were first elicited during voltage-clamp recordings by depolarizing granule cells
242
from – 70 mV to 0 mV for 400 ms. This depolarization induced [Ca2+]i changes in the dendrites,
243
more evident at the level of terminal digits (Gall et al., 2005) [(ΔF/F0)max = 0.18 ± 0.025; n=6). The
244
[Ca2+]i increase at the somatic level was very low [(ΔF/F0)max = 0.035 ± 0.02, n=6). All stimulation
245
protocols were separated by a minimum of 60 sec in order to allow [Ca2+]i to return to basal level.
246
Cell damage was identified by the following signs: the failure of 200 ms depolarization at 0 mV in
247
voltage clamp to elicit a fluorescence transient or the sudden inability of fluorescence levels to
248
recover to baseline after stimulation. We never observed bleaching of OG1 basal fluorescence
249
during stimulations.
250
Stimulus-induced fluorescence changes were analyzed off-line in the regions of interest
251
(ROIs). For each experiment, regions were drawn by eye defining the ROIs in the first image of a
252
sequence, thus giving a set of two-dimensional arrays of pixels. In addition, background
253
fluorescence was evaluated by defining a background area of similar size close to the cell. For each
254
ROI, a measurement of the relative change in fluorescence during cell stimulation, ΔF/F0 (F0 is the
255
mean resting fluorescence), was obtained as follows: for each consecutive image in the sequence,
256
the fluorescence intensity f(n) was evaluated in the ROI. Background fluorescence was measured
257
simultaneously in the background area, B(n). Care was taken to check that background fluorescence
258
was stationary. The background-subtracted fluorescence F(n)= f(n) - B(n) was then used to evaluate
259
ΔF/F0(n) = (F(n) - F0)/F0, where F0 is the average background-subtracted resting fluorescence over
260
four consecutive images before applying the stimulus. This background subtraction procedure was
261
used to account for slice auto-fluorescence and/or fluorescence arising from outflow of dye from the
262
pipette prior to seal formation. As well as in our previous work (Gall et al., 2005; D'Errico et al.,
263
2009), [Ca2+]i was measured in the granule cell dendritic ending showing the most prominent
264
fluorescence change, in order to exclude non-synaptically activated dendrites, in which voltage-
265
dependent calcium channels open due to electrotonic diffusion of depolarization. 10
266 267
Chemicals
268
The following chemicals were used at the indicated concentrations: D-2-amino-5-
269
phosphono-pentanoic acid (D-APV 50 μM; Tocris Bioscience), SR 95531 hydrobromide (gabazine
270
10 μM; Tocris Bioscience), 7-chlorokynurenic acid sodium salt (50 μM; Tocris Bioscience) were
271
dissolved in the extracellular solution. (RS)-1-Aminoindan-1,5-dicarboxylic acid (AIDA 15 μM;
272
Tocris Bioscience) was perfused locally through a multibarrel pipette. Ruthenium Red (10 μM;
273
Tocris Bioscience), Heparine 1 mg/ml (Tocris Bioscience).
274
Results
275 276
Whole-cell patch-clamp recordings were performed from granule cells in acute cerebellar
277
slices of juvenile rats (P17-P22) using stimulation protocols designed to determine the presence of
278
STDP. The average control EPSC size in the whole experimental data set was -41.5 ± 2.7 pA
279
(n=112). After monitoring baseline EPSCs at low-frequency for 10 min, the recording system was
280
switched to current-clamp for STDP induction, then the EPSCs were monitored again for at least 30
281
min. During induction, mossy fiber stimulation elicited subthreshold EPSPs, which were paired
282
with a single postsynaptic action potential (AP) generated by a 1-ms somatic current pulse (Fig.
283
1A). EPSP-AP pairing was repeated 60 times at 6 Hz using different EPSP-AP phases in different
284
recordings [for similar protocols see (Bell et al., 1997; Feldman, 2000; Froemke and Dan, 2002;
285
Froemke et al., 2005; Fung et al., 2016)].
286 287
Phase-locked pairing determines STDP
288
When EPSP-AP pairing was performed maintaining the same EPSP-AP phase relationship
289
over the 60 repetitions (phase-locking), the sign and degree of EPSC changes depended on the time
290
interval (t) between the onset of the EPSP and the peak of the AP (Fig. 1). When the AP followed 11
291
the EPSP, EPSP-AP pairing with 0
