Homeostatic plasticity scales dendritic spine volumes and ... - bioRxiv

bioRxiv preprint first posted online Apr. 26, 2018; doi: http://dx.doi.org/10.1101/308965. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission.

1 2

Homeostatic plasticity scales dendritic spine volumes and changes the threshold and specificity of Hebbian plasticity

3 4 5 6 7 8 9 10 11 12 13 14 15 16

Authors:

17

Anna F. Hobbiss1, Yazmin Ramiro Cortés1,2, Inbal Israely1,3

18 19 20

1. Champalimaud Centre for the Unknown, 1400-038 Lisbon, Portugal.

21

2. Instituto de Fisiología Celular, Universidad Nacional Autónoma de México, Ciudad Universitaria,

22 23

Circuito exterior s/n, Ciudad de México 04510, México. 3. Department of Pathology and Cell Biology in the Taub Institute for Research on Alzheimer's

24

Disease and the Aging Brain, Department of Neuroscience, College of Physicians & Surgeons,

25

Columbia University, New York, New York, 10032, USA.

26 27

Corresponding author:

28

Inbal Israely. Electronic address: [email protected]

29

1


30

Abstract

31

Information is encoded within neural networks through synaptic weight changes. Synaptic

32

learning rules involve a combination of rapid Hebbian plasticity with slower homeostatic

33

synaptic plasticity (HSP) that regulates neuronal activity through global synaptic scaling. While

34

Hebbian plasticity has been extensively investigated, much less is known about HSP. Here we

35

investigate the structural and functional consequences of HSP at dendritic spines of mouse

36

hippocampal neurons. We find that prolonged activity blockade induces spine growth,

37

paralleling synaptic strength increases. Following activity blockade, glutamate uncaging-

38

mediated long-term potentiation at single spines leads to size-dependent structural plasticity:

39

smaller spines undergo robust growth, while larger spines remain unchanged. Moreover, we

40

find that neighboring spines in the vicinity of the stimulated spine exhibit volume changes

41

following HSP, indicating that plasticity has spread across a group of synapses. Overall, these

42

findings demonstrate that Hebbian and homeostatic plasticity shape neural connectivity

43

through coordinated structural plasticity of clustered inputs.

44

Introduction

45

Hebbian synaptic plasticity, widely regarded as the leading biological mechanism for

46

information storage, involves activity-dependent changes in synaptic connectivity (Malenka

47

and Bear, 2004). Notably, these changes have a physical component and excitatory

48

postsynaptic current is highly correlated with dendritic spine volume (Matsuzaki et al., 2001).

49

However, left unchecked activity would lead to a positive feedback loop in which changes in

50

synaptic weight are further reinforced by future events. This has been proposed to result in

51

loss of functionality of the system, by either driving synapses towards saturation, or through

52

silencing the population (Miller and MacKay, 1994). How neurons maintain stability in the

53

face of destabilizing cellular and circuit events has been a longstanding question (LeMasson

54

et al., 1993). One method neurons employ to resolve this is Homeostatic Synaptic Plasticity

55

(HSP) (Turrigiano et al., 1998), a feedback mechanism through which a population of synapses

56

can be maintained to function within optimal bounds. During HSP, a decrease in global activity

57

drives a counteracting scalar increase in synaptic strengths to bring the network into an

58

optimal functioning range, and conversely, network activity increases will promote synaptic

59

weakening (Turrigiano, 2011). This scaling is instantiated in part by AMPA receptor trafficking

2


60

to and from the post-synaptic density, as well as through presynaptic changes which modify

61

neurotransmitter release, leading to concomitant changes in synaptic strength (Murthy et al.,

62

2001; O’Brien et al., 1998). HSP has been observed both in vitro and in vivo (Desai et al., 2002),

63

supporting a fundamental role for this mechanism in the proper functioning of a neural

64

system.

65

It is well established that HSP modulates synaptic function; however, it is currently unknown

66

how this process also correlates with spine structural changes. Moreover, how HSP affects

67

the induction and longevity of subsequent forms of Hebbian plasticity at individual inputs

68

remains to be determined. In this study, we examine how the induction of HSP affects the

69

structure and function of hippocampal pyramidal neuron synapses, and what are the

70

consequences of these changes for future Hebbian plasticity. We find that the induction of

71

HSP through activity blockade leads to an overall increase in the size of spines, corresponding

72

to a structural scaling that matches the functional scaling of synapses. Through precise 2-

73

photon mediated glutamate uncaging, we further investigate how these homeostatic

74

functional and structural plasticity changes impact the ability of individual inputs to undergo

75

subsequent plasticity. We demonstrate that after HSP, spines express increased longevity of

76

LTP, an increased growth rate after stimulation, and a reduced plasticity threshold. We find

77

that HSP enhances the magnitude of synaptic potentiation through the preferential

78

modulation of small spines and by promoting structural plasticity at clustered inputs following

79

Hebbian activity at single synapses. Together, these changes provide a mechanism by which

80

homeostatic plasticity can modulate synaptic efficacy and enhance future learning without

81

compromising previously stored information.

82 83

Results

84

Structural correlates of homeostatic plasticity

85

Hebbian plasticity at an individual input is linearly correlated with volume changes in the

86

corresponding spine (Govindarajan et al., 2011; Harvey and Svoboda, 2007; Matsuzaki et al.,

87

2004). Since HSP is known to change the functional properties of the spine, we reasoned that

88

homeostatic modifications of efficacy would be accompanied by structural plasticity of inputs.

89

We induced HSP using prolonged activity blockade through Tetrodotoxin (TTX) inhibition of

3


90

sodium channels, for either 0 h, 24 h, 48 h or 72 h (Figure 1a), as this has been shown to

91

induce scaling of synaptic strengths in a variety of systems (Karmarkar and Buonomano, 2006;

92

Turrigiano et al., 1998). We chose to conduct our experiments in mouse hippocampal

93

organotypic slice cultures, which maintain a physiologically relevant tissue architecture and

94

are amenable to genetic manipulation. To verify that functional synaptic scaling occurred, we

95

recorded spontaneous miniature excitatory post-synaptic currents (mEPSCs) from both TTX

96

treated and Control CA1 hippocampal pyramidal neurons at 48 h after the beginning of the

97

HSP induction period (Figure 1b-d). All recordings were performed in the presence of acute

98

TTX to block action potentials. As expected, we found a significant increase in mEPSC

99

amplitude following 48 h of activity blockade (Figure 1b,c; Control = 18.6 ± 0.25 pA, TTX =

100

20.72 ± 0.28 pA, p = 1.55e-8 Mann Whitney). The distribution of the TTX mEPSCs scaled

101

linearly to overlay with the control distribution (Figure 1d), in accordance with the synaptic

102

scaling theory. We next investigated whether homeostatic plasticity leads to structural

103

modification of synapses by live, 2-photon imaging of GFP-labelled CA1 dendrites (Figure 1e).

104

The volumes of all visible spines within the image were measured using SpineS, a custom in-

105

house developed Matlab toolbox (Erdil et al., 2012). We found that spines from chronically

106

TTX treated cells were significantly bigger than those from control cells beginning at 48 h

107

(Figure 1e,f; Control = 0.136 ± 0.0062 µm³, TTX = 0.211 ± 0.0123 µm³). Surprisingly, as opposed

108

to the linear functional scaling of mEPSCs seen in response to HSP, we found that structural

109

scaling of spine volumes at 48 h are best fit to controls with a second order equation (Figure

110

1g). This superlinear scaling resulted from a preponderance of large spines after TTX

111

treatment, for which correspondingly large mEPSCs were not observed.

112 113

After 72 h of activity blockade, synapses maintained the enhancement of mEPSC size (Figure

114

1h; Mean ± SEM: control 72 h = 20.35 ± 0.27 pA, TTX 72 h= 22.18 ± 0.239 pA, p = 1.88e-15,

115

Mann-Whitney) and spine volume increases (Figure 1i; Mean ± SEM: control 72 h= 0.16 ±

116

0.0073 µm³, TTX = 0.19 ± 0.0093 µm³, p = 0.0008 Mann-Whitney). However, at this time,

117

scaling was instantiated in a linear fashion at both the level of mEPSC amplitude and spine

118

volumes (Figure 1j,k). This suggests that the expression of functional and structural scaling

119

evolves dynamically over time in response to prolonged activity blockade.

120 121

4


122

Reversibility of homeostatic plasticity mediated structural changes

123

Synaptic strength modifications that occur in response to activity blockade are reversible

124

upon re-exposure of a circuit to activity (Desai et al., 2002; Wallace and Bear, 2004). We

125

tested whether the structural changes that resulted following activity blockade were also

126

reversible when activity is restored. After 48 h of homeostatic plasticity induction, by which

127

time significant structural and functional scaling have occurred (Figure 1), we removed TTX

128

allowing slices to resume activity and measured spontaneous firing using whole-cell patch

129

clamp recordings (Figure 2a,b). Activity blockade for 48 h led to higher firing rates compared

130

to untreated cells (Figure 2c; Control (median) = 1.33, IQ range = 0.11:18.83 spikes/min; TTX

131

(median) = 13.33, IQ range = 11.5:21.06 spikes/min. p=0.03, Kruskal-Wallis post-hoc Dunn).

132

To exclude the possibility that the higher firing rates we observed were due to a rebound after

133

the acute withdrawal of TTX from the system, a set of control neurons were briefly incubated

134

in TTX (2-4h, named “Acute TTX condition”). This manipulation did not significantly alter firing

135

rates, which were similar to untreated controls (Figure 2c; Acute-TTX: med = 4.44, IQ range =

136

2.0:29.33 spikes/min vs Control no-TTX: Median = 1.33, IQ range = 0.11:18.83 spikes/min,

137

p>0.9 Kruskal-Wallis post-hoc Dunn). Thus, the firing rate changes we observed were the

138

result of HSP and did not arise from the acute withdrawal of TTX. We then investigated

139

whether the circuit would return to its original level of activity once neurons were released

140

from blockade. Indeed, after having removed TTX for 48 h, firing rates were indistinguishable

141

from controls (Figure 2c; Control (median) = 1.33, IQ range = 0.11:15.56. TTX reversal

142

(median) = 1.67, IQ range = 0.194:11.76. p>0.05, Kruskal-Wallace post-hoc Dunn). We then

143

examined whether structural modifications accompanied these firing rate changes, and

144

observed a similar reversal in spine volumes (Figure 2d,e). TTX treated spines significantly

145

reduced in size 48 h after TTX removal (Figure 2e; mean volume ± SEM: TTX 48 h = 0.21 ±

146

0.0063 µm³; TTX 48 h reversal = 0.16 ± 0.0064 µm³; p=4.1852e-07). These findings support

147

the idea that functional plasticity is matched by structural plasticity, and that spines sizes

148

reversibly and accurately reflect the activity landscape.

149 150

Enhanced Hebbian potentiation of small spines after HSP

151

Functional plasticity can be enhanced upon homeostatic modulation in hippocampal slices

152

(Arendt et al., 2013; Félix-Oliveira et al., 2014), but occlusion of LTP may also occur (Soares et

153

al., 2017). Given our finding that robust growth of spines occurs during homeostatic plasticity,

5


154

we wanted to determine whether individual inputs were able to undergo further activity-

155

dependent structural plasticity. We therefore induced HSP in hippocampal slices for 48 h and

156

followed this with synaptic potentiation at visually identified dendritic spines through 2-

157

photon mediated glutamate uncaging (Figure 3a,b). This paradigm elicits potentiation and

158

growth of only the stimulated input (Govindarajan et al., 2011; Harvey and Svoboda, 2007).

159

We followed the growth of spines relative to their average baseline volumes, in both the TTX

160

and control conditions (Figure 3c,d). During the initial 45 minutes post-stimulation, both TTX-

161

treated and control spines grew to a similar extent, indicating that homeostatic structural

162

scaling does not occlude activity-dependent structural plasticity (mean ± SEM; Control

163

stimulated 45’= 1.45 ± 0.10; Control neighbors 45’= 1.08 ± 0.04; TTX stimulated 45’= 1.59 ±

164

0.14; TTX neighbors 45’= 1.14 ± 0.04; p > 0.99) (Figure 3d). However, after 120 minutes, only

165

the stimulated inputs in the TTX-treated neurons remain significantly larger than their

166

neighbors (Figure 3d; mean ± SEM: TTX stimulated 120’= 1.53 ± 0.13; TTX neighbors 120’=

167

1.16 ± 0.04), as control spines begin to return to their initial volume (Figure 3d; Control

168

stimulated 120’= 1.30 ± 0.01; Control neighbors 120’= 1.14 ± 0.04). Although the TTX and

169

control spines are not significantly different at 120’ (p = 0.1345, Mann Whitney), they begin

170

to diverge, as seen relative to their respective neighbors. Thus, at scaled synapses, activity

171

that would otherwise lead to short lived structural plasticity now elicits long lasting structural

172

plasticity.

173 174

Structural plasticity at individual inputs varies with activity and spine size. While weak

175

stimulations preferentially affect small spines (Matsuzaki et al., 2004), strong stimuli can elicit

176

structural plasticity across a variety of spine sizes (Govindarajan et al., 2011; Oh et al., 2013;

177

Ramiro-Cortés and Israely, 2013). Our stimulated spines spanned a range of sizes, allowing us

178

to test whether size interacted with prior expression of HSP when expressing synaptic

179

potentiation and further spine growth (Figure 3e). We examined the amount of structural

180

plasticity expressed by spines of different initial sizes after activity blockade. We found a

181

negative correlation between a spine’s initial average value and its normalized final volume

182

when the spine population had first undergone homeostatic plasticity, but not in the control

183

population, indicating that spine volume is an important modulator of the potential for

184

plasticity after HSP (Figure 3ef; r2 = 0.48 for TTX, p < 0.01, r2 = 0.06 for control, p > 0.3). To

185

further examine how HSP modulates the size dependence of structural plasticity, we classified

6


186

spines as ‘large’ if their initial volume was more than 150% of the median initial volume of all

187

spines, and the remainder were placed in the ‘small’ category. Among the small spines, we

188

observed a significant increase in the magnitude of LTP in the TTX condition compared to the

189

controls (Figure 3g; p=0.046, 2-way repeated measures ANOVA). On the other hand, large

190

spines expressed a short-lasting growth that quickly decayed to baseline and was not

191

dependent on whether they had been subjected to activity blockade (Figure 3h; p=0.70 2-way

192

repeated measures ANOVA). Therefore, small spines preferentially underwent long lasting

193

structural plasticity after HSP. We further observed that small spines, which had undergone

194

HSP, tended to show a greater degree of initial growth in response to the induction of

195

plasticity (Figure 3g, time point 0). To quantify the dynamic growth of these spines, we

196

analyzed high-speed images of the spine head taken throughout the stimulation period

197

(approximately 20 Hz) and did not observe a significant difference in the growth curves

198

between the control and TTX treated spines (Figure 4a,b). When we examined the rate of

199

growth that each spine expressed in the first minute after the stimulation however, we found

200

that TTX treated small spines grew more than control small spines (Figure 4c; p = 0.018, 1-

201

way ANOVA with post-hoc Tukey’s), while large spines showed no difference between the

202

two conditions. Therefore, glutamate stimulation leads to a sustained growth of

203

homeostatically modified small spines, which may reflect prolonged signaling at these

204

synapses. Together, these data indicate that HSP facilitates Hebbian plasticity, and that this

205

is accomplished at the individual spine level through preferential structural plasticity of small

206

inputs.

207 208

Homeostatic plasticity facilitates the induction of Hebbian structural plasticity

209

Having observed that small spines showed enhanced responses to stimulation after

210

homeostatic plasticity, as reflected by a higher rate of growth and longer lasting structural

211

plasticity (Figure 3g and 4c), we reasoned that these inputs may respond more robustly to a

212

weaker stimulation. This is also supported by previous findings that loss of activity at

213

individual inputs lowers their threshold for subsequent plasticity (Lee et al., 2010). To test this

214

possibility, we applied a sub-threshold stimulation that utilizes a shorter laser pulse of 1 ms

215

(instead of 4 ms) during the glutamate uncaging protocol, since this does not induce structural

216

plasticity at spines (Govindarajan et al., 2011; Harvey and Svoboda, 2007). Indeed, control

217

spines that underwent this sub-threshold stimulation showed only a transient post-stimulus

7


218

growth that lasted 5 min, resulting in no long term plasticity (Figure 4d; Control spines, 1 ms

219

stimulated vs neighbors: p > 0.05, 2-way repeated measures ANOVA, post hoc Bonferroni).

220

These results were significantly different from what we had observed in response to the 4ms

221

stimulation (Figure 4d; 1 ms stimulated vs 4 ms stimulated; p = 0.04, 2-way repeated

222

measures ANOVA; compared to data from Figure 3c). In contrast to this, HSP modified spines

223

grew significantly in response to the subthreshold stimulation and remained larger than their

224

neighbors (Figure 4e; TTX spines, 1 ms stimulated vs neighbors p = 0.01 2-way repeated

225

measures ANOVA, post hoc Bonferroni). Following TTX treatment, both weak and strong

226

stimulations elicited similar levels of long lasting growth of spines (Figure 4e; 1ms stimulated

227

vs 4 ms stimulated: p = 0.27, 2-way repeated measures ANOVA). Thus, homeostatic plasticity

228

facilitates structural plasticity by lowering the threshold for stimulation, without affecting the

229

magnitude of plasticity. In this way, HSP acts locally, in conjunction with activity, to increase

230

the likelihood that an individual synapse will participate in the encoding of information.

231 232

HSP influences structural plasticity of neighboring spines

233

In addition to modifying synaptic inputs through scaling, homeostatic plasticity can also

234

influence cellular firing rates by modulating the intrinsic excitability of neuronal membranes

235

(Desai et al., 1999). We reasoned that such alterations could facilitate the expression of

236

structural plasticity not only at stimulated spines but also at nearby inputs within the dendritic

237

branch. To test this idea, we examined whether homeostatic plasticity altered neighboring

238

spine volume dynamics following the potentiation of single inputs. We ranked neighboring

239

spines according to their distance from the stimulated spine (with 1 representing the closest

240

neighbor to a stimulated spine) and plotted their growth dynamics for two hours following

241

the stimulation (Figure 5a). We found that spines located in close proximity to the stimulated

242

spine increased in volume only in neurons that had first undergone homeostatic plasticity

243

(compare the first 20 spines, from 0 to 60 minutes after stimulation, Figure 5a). We classified

244

the unstimulated neighbors as “near” or “far” - located either within or beyond 5 µm of the

245

stimulated spine respectively - and quantified spine volume changes over time (Figure 5b-d).

246

As expected, none of the neighbors of stimulated spines changed significantly from their

247

original size in untreated neurons (Figure 5b,c). However, after HSP, neighbors that were

248

within 5 µm of the target spine exhibited significant growth in the 5 minutes following the

249

stimulation (Figure 5b) and remained significantly larger than more distant spines (Figure

8


250

5b,d; p = 0.001, 2-way repeated measures ANOVA). Interestingly, farther neighbors (located

251

up to 20 µm away from the site of stimulation) tended to decrease in size for several minutes

252

following activity, although this was not significantly different from baseline (Figure 5d). We

253

next investigated whether there was a correlation between the structural dynamics of each

254

stimulated spine and its neighbors. We calculated correlation coefficients between the

255

volume change of the stimulated spine and those of its neighbors across the time-course of

256

the experiment. While fluctuations in the volumes of neighboring spines in control conditions

257

were independent of the stimulated input (Figure 5e; r2 = 0.00263, p = 0.61), spine volumes

258

in TTX treated neurons changed in accordance with the stimulated spine (Figure 5f; r2 = 0.156,

259

p < 0.001). Specifically, near TTX neighbors had a positive correlation with the stimulated

260

input, reflecting growth, which eventually became a negative correlation at further distances

261

from the stimulated spine. These findings indicate that HSP reduces the input specificity of

262

activity in a distance dependent manner that is at once both cooperative with near, and

263

competitive with far, inputs. This allows for homeostatic modulation to effect structural

264

plasticity within a dendritic region through activity at a single spine, which favors clustered

265

plasticity of synaptic inputs.

266 267

Discussion

268

Synaptic networks must balance the need to enhance efficacy during learning without

269

saturating their capacity for further changes. Homeostatic plasticity can achieve this

270

regulatory role by effecting a gain modulation on neurons in order to maintain synaptic

271

activity within an optimal target range. It is unclear what are the consequences of

272

implementing HSP for the network, and specifically, how its interaction with Hebbian

273

plasticity impacts the subsequent encoding of information at the level of single inputs. Due

274

to the linear relationship between synaptic structure and function, we used live 2-photon

275

imaging and glutamate uncaging to probe the physical consequences of inducing homeostatic

276

plasticity on a neuron, and to determine how this impacts the ability of single synapses to

277

undergo further changes in efficacy. We show that homeostatic synaptic plasticity, induced

278

by 48 hours of activity blockade, leads to the reversible growth of dendritic spines. We

279

demonstrate that this form of modulation facilitates subsequent structural plasticity at single

280

synapses and that this effect preferentially occurs at smaller inputs. After HSP, activity elicits

9


281

a faster growth rate at these spines, and converts an otherwise subthreshold stimulation into

282

one that is now capable of eliciting structural plasticity. Interestingly, we find that the

283

induction of Hebbian plasticity on a background of homeostatic plasticity leads to

284

compromised input specificity, as neighboring spines grow when they are located in close

285

proximity to a stimulated synapse. Taken together, our results show that homeostatic

286

plasticity can modulate a neuron’s response to activity by facilitating the sensitivity of smaller

287

inputs and inducing structural plasticity at neighboring synapses.

288

The structural scaling that we observe after 48 hours of activity blockade results in a

289

non-linear upscaling of spines and an increased number of large spines (Figure 1g). By 72

290

hours, this structural scaling becomes linear (Figure 1k), suggesting that the initial changes

291

may represent a physical overshooting of the target size. This has also been observed on short

292

timescales following glutamate stimulation of single spines, where an initial large volume

293

change is followed by stabilization of the spine at a more modest size (Matsuzaki et al., 2004).

294

Upon the reinstatement of activity, we find that spines return to their original size as firing

295

patterns normalize by 48 hours (Figure 2c-e), indicating that homeostatic structural changes

296

are plastic and reversible, similarly to homeostatic responses to fluctuating activity levels.

297

We demonstrate that activity blockade increases spine sizes across the population,

298

raising the question of whether further enhancement of potentiation and spine growth could

299

be achieved across inputs of different sizes. Both enhancement and occlusion of plasticity

300

have been reported following HSP at the population level (Arendt et al., 2013; Félix-Oliveira

301

et al., 2014; Soares et al., 2017), but it is unclear how individual inputs are modulated. Further,

302

while larger synapses have been shown to undergo functional homeostatic modulation

303

(Thiagarajan et al., 2005), inputs show a size dependent variation in their ability to undergo

304

Hebbian structural plasticity, large spines being more stable and requiring stronger

305

stimulation in order to be potentiated (Govindarajan et al., 2011; Matsuzaki et al., 2004). We

306

probed plasticity with glutamate uncaging and found that HSP leads to structural plasticity

307

lasting over two hours preferentially at small spines, while control spines decay to baseline

308

after one hour (Figure 3g,h). This suggests that a long lasting, protein synthesis dependent

309

form of Hebbian plasticity was induced at these inputs (Govindarajan et al., 2011), supported

310

by recent findings that TTX-mediated activity blockade leads to the specific production of

311

plasticity related proteins (Schanzenbacher et al., 2018). Closer examination of the structural

312

plasticity we induced revealed that the population of small spines express the majority of the

10


313

growth, while large spines remain stable (Figure 3g,h). This result implies that plasticity at

314

large spines is saturated after homeostatic modifications, and that their threshold for

315

structural plasticity is not altered by HSP.

316

Upon stimulation, we did not observe a significant difference in the magnitude of the

317

potentiation that spines expressed, nor in their immediate response to the stimulation itself,

318

but rather we found that spines post-HSP express a significantly faster growth rate compared

319

to their counterparts in the first minute after stimulation (Figure 4a-c). This enhanced

320

response may be due to amplified signaling cascades shared between these forms of plasticity

321

(Fernandes and Carvalho, 2016), which could facilitate the induction of long lasting structural

322

plasticity. It was therefore not surprising to find that a subthreshold stimulation elicits long

323

lasting growth only at inputs that have undergone homeostatic plasticity (Figure 4 d,e),

324

without affecting the initial magnitude of the response between the HSP modified synapses

325

and the control populations. Changes to NMDA receptor composition that occur following

326

activity blockade could provide a means by which to enhance the calcium permeability of

327

synapses and thus would allow for prolonged signaling responses (Barria and Malinow, 2005;

328

Lee et al., 2010; Paoletti et al., 2013). Taken together, these data show that the effects of

329

homeostatic plasticity preferentially impact small spines during Hebbian regulation of

330

synaptic strength by predisposing them to undergo long lasting structural changes. Such

331

modulation may result in the conversion of weaker stimuli into more salient forms, and

332

proposes a more global role for HSP in information encoding beyond the optimization of

333

neuronal activity.

334

A hallmark of activity dependent changes during Hebbian plasticity is input specificity

335

(Barrionuevo and Brown, 1983). In contrast, homeostatic plasticity is generally expressed over

336

a wider synaptic range, and how this modulation impacts Hebbian learning rules at individual

337

inputs is unknown (Vitureira and Goda, 2013). Computational studies have postulated that

338

homeostatic plasticity can influence previous Hebbian events at a synapse (Rabinowitch and

339

Segev, 2008). Additionally, several lines of evidence indicate that activity can lead to local

340

homeostatic changes within dendrites (Branco et al., 2008; Ju et al., 2004; Liu, 2004; Sutton

341

et al., 2006). We therefore considered the possibility that HSP may function to influence

342

subsequent Hebbian plasticity. Indeed, we find that HSP induces the spillover of Hebbian

343

structural plasticity within the dendrite, as seen by the growth of neighboring unstimulated

344

spines following the activation of one input (Figure 5). Thus, homeostatic plasticity drives the

11


345

local expression of Hebbian plasticity and reduces input specificity. While cooperative

346

interactions have been shown to occur between inputs when multiple synapses are

347

stimulated (Govindarajan et al., 2011; Harvey and Svoboda, 2007), in the absence of such co-

348

activation, structural changes at neighbors have thus far served to counterbalance the

349

direction of plasticity (Oh et al., 2015). Our observations that HSP promotes activity-driven

350

growth of a group of synapses within a five micron area is in alignment with the observed

351

distance over which plasticity related proteins can spread between co-active synapses

352

(Harvey et al., 2008). Thus, the interaction between homeostatic and Hebbian plasticity may

353

coordinately delimit a region of clustered plasticity, which has been proposed to increase the

354

memory capacity of a neural circuit (Chklovskii et al., 2004; Govindarajan et al., 2011, 2006;

355

Ramiro-Cortés et al., 2014). Although clustered activity and its structural correlates have

356

begun to emerge (Fu et al., 2012; Kleindienst et al., 2011; Makino and Malinow, 2011; Yadav

357

et al., 2012), the mechanisms which drive the physical organization of inputs are unknown.

358

HSP may provide a basis for such organization by reducing the threshold for cooperativity

359

between inputs, allowing primed synapses located within several microns of an active spine

360

to be more easily potentiated. This may serve as a first step in the physical arrangement of

361

synapses into clusters. The broader consequence of this could be the binding together of

362

information of varying saliencies, within a delimited region, into the same engram.

363

Our finding that synaptic threshold modulation after HSP is implemented at small,

364

rather than at large spines, and in a reversible manner, suggests that synaptic scaling

365

mechanisms are separable from plasticity mechanisms. However, by promoting long-lasting

366

plasticity at select spines, and potentiating unstimulated neighboring spines within a

367

delimited dendritic region, homeostatic plasticity may interact with and reduce the input

368

specific nature of Hebbian plasticity, enhance the clustering of synaptic inputs, and shape the

369

long-term organization of neural circuits. In this way, despite the global nature of homeostatic

370

plasticity, inputs may be locally modulated in an activity-dependent manner.

12


371

Materials and Methods

372

1. Slice culture preparation and biolistic transfection

373

Mouse hippocampal organotypic slice cultures were prepared using p7-10 C57BL/6J mice as

374

previously described (Govindarajan et al., 2011). Briefly, hippocampi were dissected and 350

375

µm slices were cut with a tissue chopper in ice-cold artificial cerebral spinal fluid (aCSF)

376

containing 2.5 mM KCl, 26 mM NaHCO3, 1.15 mM NaH2PO4, 11 mM D-glucose, 24 mM

377

sucrose, 1 mM CaCl2 and 5 mM MgCl2. The slices were cultured on membranes (Millipore),

378

and maintained at an interface with the following media: 1x MEM (Invitrogen), 20% horse

379

serum (Invitrogen), 1 mM GlutaMAX (Invitrogen), 27 mM D-glucose, 30 mM HEPES, 6 mM

380

NaHCO3, 2mM CaCl2, 2mM MgSO4, 1.2% ascorbic acid, 1 µg/ml insulin. The pH was adjusted

381

to 7.3, and the osmolarity adjusted to 300–310 mOsm. All chemicals were from Sigma unless

382

otherwise indicated. Media was changed every 2-3 days.

383

Biolistic transfection of slice cultures was accomplished using a Helios gene gun (Bio-Rad)

384

after 4–7 days in vitro (DIV). Gold beads (10 mg, 1.6 µm diameter, Bio-Rad) were coated with

385

100 mg AFP-plasmid DNA (a GFP-expressing plasmid driven by the β-actin promoter (Inouye

386

et al., 1997) according to the manufacturer’s protocol and delivered biolistically to the slices,

387

using a pressure of 160-180 psi.

388

2. Homeostatic plasticity induction by activity block

389

TTX (1 µM) was added to the culture media at 7-9 DIV. The day of application was then

390

designated day 0 for experiments. Control experiments were maintained in normal culture

391

media, and were age- and animal-matched to treated slices for experiments.

392

3. Patch Clamp Electrophysiology

393

Hippocampal slice cultures were perfused continuously with aCSF (as above, with the addition

394

of 0.5 µM TTX for all mEPSC recordings) for a pre-incubation period of 15 to 30 min. Whole

395

cell voltage-clamp recordings were performed in CA1 pyramidal neurons, using 7–8 MΩ

396

electrodes. For mEPSC recordings, the internal solution contained 135 mM Cs-

397

methanesulfonate, 10 mM CsCl, 10 mM HEPES, 5 mM EGTA, 2 mM MgCl2, 4 mM Na-ATP and

398

0.1 mM Na-GTP, with the pH adjusted to 7.2 with KOH, at 290-295 mOsm. Cells were voltage

399

clamped at -65 mV. Cellular recordings in which series resistance was higher than 25 mV were

13


400

discarded. Stability was assessed throughout the experiment, with cells whose series

401

resistance changed more than 30% being discarded. mEPSCs recordings were started 3

402

minutes after break-in and continued for 10 minutes. Signals were acquired using a

403

Multiclamp 700B amplifier (Molecular Devices), and data was digitized with a Digidata 1440

404

at 3 kHz. mEPSC events were detected off-line using Mini-Analysis Program (Synaptosoft).

405

Events smaller than 15 pA fell within the range of noise in the system and were not included

406

in the analysis. For spontaneous activity recordings, slices were perfused continuously with

407

aCSF without the addition of TTX for a pre-incubation period of 5 to 10 min. The internal

408

solution for the electrodes contained 136.5 mM K-Gluconate, 9 mM NaCl, 17.5 mM KCl, 10

409

mM HEPES, 0.2 mM EGTA, and 0.025 mM Alexa 594, with the pH adjusted to 7.2 with KOH,

410

at 280-290 mOsm. In current clamp with no external current applied, an IV curve was first

411

recorded to check for spike frequency accommodation in order to validate the identity of

412

pyramidal neurons. Spiking events were then recorded for a period of 6-9 minutes. The

413

addition of Alexa-594 allowed cells to be imaged post-recording.

414

4. Two-photon Imaging

415

Two-photon imaging was performed on a BX61WI Olympus microscope, using a

416

galvanometer-based scanning system (Prairie Technologies /Bruker) with a Ti:sapphire laser

417

(910 nm for imaging AFP; Coherent), controlled by PrairieView software (Prairie

418

Technologies). Slices were perfused with oxygenated aCSF containing 127 mM NaCl, 2.5 mM

419

KCl, 25 mM NaHCO3, 1.25 mM NaH2PO4, 25 mM D-glucose, 2 mM CaCl2, 1 mM MgCl2 and 0.5

420

µM TTX (equilibrated with O2 95%/CO2 5%) at room temperature, at a rate of 1.5 ml/min.

421

Secondary or tertiary apical dendrites of CA1 neurons (where the apical trunk is counted as

422

the primary branch) were imaged using a water immersion objective (60x, 1.0 NA, Olympus

423

LUMPlan FLN) with a digital zoom of 8x. For each neuron, 2-3 dendrites were imaged. Z-stacks

424

(0.5 µm per section) were collected at a resolution of 1024 x 1024 pixels, resulting in a field

425

of view of 25.35 x 25.35 µm. 3 images were taken per dendrite at 5-minute intervals, with the

426

reported spine volume being the average of the 3 images. Images were taken at the highest

427

possible fluorescence intensity without image saturation in order to accurately quantify spine

428

volumes (see below).

14


429

5. Glutamate uncaging

430

Uncaging experiments (Pettit et al., 1997) and caged glutamate calibration were carried out

431

as previously described (Govindarajan et al., 2011), and briefly as follows. MNI-caged-L-

432

glutamate (MNI-Glu) (Tocris) was reconstituted in the dark in aCSF lacking MgCl2 or CaCl2 to

433

make a 10 mM stock solution. Individual aliquots were diluted to the working concentration

434

of 2.5 mM MNI-Glu in uncaging aCSF (see below), in 3 ml volumes. We tested each batch of

435

reconstituted MNI-Glu as previously described (Govindarajan et al., 2011). Briefly, five

436

uncaging test pulses of 1 ms were delivered to single spines and evoked EPSCs were measured

437

by whole cell patch clamp recordings. We compared these to spontaneous mEPSCs, and

438

determined the power needed (60mW at the back aperture) to produce an uncaging EPSC of

439

comparable size to that of an average spontaneous mEPSC. During plasticity experiments,

440

slices were incubated for 30-45 minutes in normal aCSF as described above. Uncaging was

441

performed using a Ti:sapphire laser (720 nm; Coherent), controlled by PrairieView software

442

(Prairie Technologies, Bruker). All stimuli were carried out in uncaging-aCSF containing: 2.5

443

mM MNI-Glu, 0 mM MgCl2, and 4 mM CaCl2. The uncaging train consisted of 30 pulses at 0.5

444

Hz, with a pulse width of either 4 ms (standard protocol) or 1 ms (sub-threshold protocol),

445

using 30mW power as measured at the back aperture (half the power as determined in the

446

calibration step, as previously described for an uncaging LTP protocol (Govindarajan et al.,

447

2011)). The uncaging point was positioned 0.6 µm from the end of the spine head, away from

448

the parent dendrite. Each experiment began with a sham stimulation in uncaging-aCSF lacking

449

MNI-Glu on a control dendrite of the experimental neuron, to rule out the possibility of non-

450

specific structural changes due to phototoxicity or poor neuronal health. Subsequently,

451

experimental spines on new dendrites were identified, uncaging-aCSF containing 2.5mM

452

MNI-glutamate was recirculated for 5 minutes, and the uncaging protocol was delivered. Fast

453

imaging (approximately 20 Hz) of a region of interest (ROI) of the stimulated spine head and

454

neck was performed throughout the 60 second stimulation, to record the growth dynamics

455

for this time period. After the stimulation, the perfusion was returned to normal aCSF for the

456

remainder of the experiment (1-2 hours). The first image was taken immediately after

457

returning to normal aCSF and is designated as time = 0 post stimulation.

15


458

6. Spine Volume Determination

459

To quantify spine volume, we used the custom built Matlab plug-in SpineS, which uses semi-

460

automatic detection, automatic alignment and segmentation of spine heads (Erdil et al., 2012;

461

Ghani et al., 2017). Spine volume was calculated using the summed fluorescence intensity of

462

the spine, normalized to the median fluorescence intensity of the parent dendrite, in order

463

to correct for changes in overall fluorescence levels within a cell. Fluorescence intensity was

464

converted to real volumes by taking Full Width Half Max (FWHM) measurements of spines

465

and calculating a conversion factor between the two measures (Matsuzaki et al., 2004). For

466

the homeostatic plasticity analysis (Figures 1 & 2), all spines with a discernible head within

467

the field of view were included in the analysis, unless obstructed by other structures. For the

468

uncaging analysis (Figures 3, 4 & 5), between 2 and 8 neighboring spines were analyzed from

469

the same dendrite, within the whole field of view. Normalization of spine volumes throughout

470

the

experiment

was

performed

using

the

mean

of

three

baseline

images.

471 472

7. Statistical analyses

473

All statistical analyses were carried out in Graphpad Prism. Stars (*) represent degrees of

474

significance as follows: (*) = p < 0.05; (**) = p < 0.01; (***) = p < 0.001).

475

476

Acknowledgements

477

We thank Ali Ozgur Argunsah for his advice and help with data analysis. We thank Steven

478

Kushner, Rui Costa, and members of the Israely lab for their support and critical reading of

479

the manuscript. This work was supported by grants from the Bial Foundation (to II), Fundação

480

para a Ciência e a Technologia (to II and AH), and CONACYT 254878 (to YRC).

481 482

Author Contributions:

483

AFH, YRC and II designed the experiments. AFH performed the experiments and analyzed

484

the data. AFH, YRC and II wrote the manuscript.

16


485

486

Competing interests

487

The authors declare no competing interests

17


488 489 490

Bibliography Arendt, K.L., Sarti, F., Chen, L., 2013. Chronic inactivation of a neural circuit enhances LTP by

491

inducing silent synapse formation. J. Neurosci. 33, 2087–2096.

492

https://doi.org/10.1523/JNEUROSCI.3880-12.2013

493

Barria, A., Malinow, R., 2005. NMDA receptor subunit composition controls synaptic

494

plasticity by regulating binding to CaMKII. Neuron 48, 289–301.

495

https://doi.org/10.1016/j.neuron.2005.08.034

496 497 498

Barrionuevo, G., Brown, T.H., 1983. Associative long-term potentiation in hippocampal slices. Proc. Natl. Acad. Sci. 80, 7347–7351. https://doi.org/10.1073/pnas.80.23.7347 Branco, T., Staras, K., Darcy, K.J., Goda, Y., 2008. Local Dendritic Activity Sets Release

499

Probability at Hippocampal Synapses. Neuron 59, 475–485.

500


501 502 503

Chklovskii, D.B., Mel, B.W., Svoboda, K., 2004. Cortical rewiring and information storage. Nature 431, 782–788. https://doi.org/10.1038/nature03012 Desai, N.S., Cudmore, R.H., Nelson, S.B., Turrigiano, G.G., 2002. Critical periods for

504

experience-dependent synaptic scaling in visual cortex. Nat. Neurosci. 5, 783–789.

505

https://doi.org/10.1038/nn878

506

Desai, N.S., Rutherford, L.C., Turrigiano, G.G., 1999. Plasticity in the intrinsic excitability of

507

cortical pyramidal neurons. Nat. Neurosci. 2, 515–520. https://doi.org/10.1038/9165

508

Erdil, E., Yagci, A.M., Argunsah, A.O., Ramiro-Cortes, Y., Hobbiss, A.F., Israely, I., Unay, D.,

509

2012. A Tool for Automatic Dendritic Spine Detection and Analysis . Part I : Dendritic

510

Spine Detection Using. Proc. Int. Conf. Image Process. Theory, Tools {&} Appl.

511

Félix-Oliveira, A., Dias, R.B., Colino-Oliveira, M., Rombo, D.M., Sebastião, A.M., 2014.

512

Homeostatic plasticity induced by brief activity deprivation enhances long-term

513

potentiation in the mature rat hippocampus. J. Neurophysiol. 112, 3012–3022.

514

https://doi.org/10.1152/jn.00058.2014

515 516

Fernandes, D., Carvalho, A.L., 2016. Mechanisms of homeostatic plasticity in the excitatory synapse. J. Neurochem. 139, 973–996. https://doi.org/10.1111/jnc.13687

18


517

Fu, M., Yu, X., Lu, J., Zuo, Y., 2012. Repetitive motor learning induces coordinated formation

518

of clustered dendritic spines in vivo. Nature 483, 92–95.

519

https://doi.org/10.1038/nature10844

520

Ghani, M.U., Mesadi, F., Kanik, S.D., Argunsah, A.O., Hobbiss, A.F., Israely, I., Unay, D.,

521

Tasdizen, T., Cetin, M., 2017. Dendritic spine classification using shape and appearance

522

features based on two-photon microscopy. J. Neurosci. Methods 279, 13–21.

523

https://doi.org/https://doi.org/10.1016/j.jneumeth.2016.12.006

524

Govindarajan, A., Israely, I., Huang, S.-Y., Tonegawa, S., 2011. The dendritic branch is the

525

preferred integrative unit for protein synthesis-dependent LTP. Neuron 69, 132–46.

526


527

Govindarajan, A., Kelleher, R.J., Tonegawa, S., 2006. A clustered plasticity model of long-

528

term memory engrams. Nat. Rev. Neurosci. 7, 575–583.

529

https://doi.org/10.1038/nrn1937

530 531 532

Harvey, C.D., Svoboda, K., 2007. Locally dynamic synaptic learning rules in pyramidal neuron dendrites. Nature 450, 1195–1200. https://doi.org/10.1038/nature06416 Harvey, C.D., Yasuda, R., Zhong, H., Svoboda, K., 2008. The spread of Ras activity triggered

533

by activation of a single dendritic spine. Science 321, 136–140.

534

https://doi.org/10.1126/science.1159675

535

Inouye, S., Ogawa, H., Yasuda, K., Umesono, K., Tsuji, F.I., 1997. A bacterial cloning vector

536

using a mutated Aequorea green fluorescent protein as an indicator. Gene 189, 159–

537

162. https://doi.org/10.1016/S0378-1119(96)00753-6

538

Ju, W., Morishita, W., Tsui, J., Gaietta, G., Deerinck, T.J., Adams, S.R., Garner, C.C., Tsien,

539

R.Y., Ellisman, M.H., Malenka, R.C., 2004. Activity-dependent regulation of dendritic

540

synthesis and trafficking of AMPA receptors. Nat. Neurosci. 7, 244–53.

541


542

Karmarkar, U.R., Buonomano, D. V, 2006. Different forms of homeostatic plasticity are

543

engaged with distinct temporal profiles. Eur. J. Neurosci. 23, 1575–84.

544

https://doi.org/10.1111/j.1460-9568.2006.04692.x

545

Kleindienst, T., Winnubst, J., Roth-Alpermann, C., Bonhoeffer, T., Lohmann, C., 2011.

19


546

Activity-Dependent Clustering of Functional Synaptic Inputs on Developing

547

Hippocampal Dendrites. Neuron 72, 1012–1024.

548


549 550 551 552 553

Lee, M.-C., Yasuda, R., Ehlers, M.D., 2010. Metaplasticity at single glutamatergic synapses. Neuron 66, 859–870. https://doi.org/10.1016/j.neuron.2010.05.015 LeMasson, G., Marder, E., Abbott, L.F., 1993. Activity-dependent regulation of conductances in model neurons. Science 259, 1915–1917. https://doi.org/10.1126/science.8456317 Liu, G., 2004. Local structural balance and functional interaction of excitatory and inhibitory

554

synapses in hippocampal dendrites. Nat. Neurosci. 7, 373.

555


556

Makino, H., Malinow, R., 2011. Compartmentalized versus global synaptic plasticity on

557

dendrites controlled by experience. Neuron 72, 1001–1011.

558


559 560 561

Malenka, R.C., Bear, M.F., 2004. LTP and LTD: an embarrassment of riches. Neuron 44, 5–21. https://doi.org/10.1016/j.neuron.2004.09.012 Matsuzaki, M., Ellis-Davies, G.C.R., Nemoto, T., Miyashita, Y., Iino, M., Kasai, H., 2001.

562

Dendritic spine geometry is critical for AMPA receptor expression in hippocampal CA1

563

pyramidal neurons. Nat. Neurosci. 4, 1086–1092. https://doi.org/10.1038/nn736

564

Matsuzaki, M., Honkura, N., Ellis-Davies, G.C.R., Kasai, H., 2004. Structural basis of long-term

565

potentiation in single dendritic spines. Nature 429, 761–766.

566

https://doi.org/10.1038/nature02594.1.

567

Miller, K.D., MacKay, D.J.C., 1994. The Role of Constraints in Hebbian learning. Neural

568

Comput. 6, 100–126. https://doi.org/https://doi.org/10.1162/neco.1994.6.1.100

569

Murthy, V.N., Schikorski, T., Stevens, C.F., Zhu, Y., 2001. Inactivity produces increases in

570

neurotransmitter release and synapse size. Neuron 32, 673–682.

571

https://doi.org/10.1016/S0896-6273(01)00500-1

572

O’Brien, R.J., Kamboj, S., Ehlers, M.D., Rosen, K.R., Fischbach, G.D., Huganir, R.L., 1998.

573

Activity-dependent modulation of synaptic AMPA receptor accumulation. Neuron 21,

574

1067–1078. https://doi.org/https://doi.org/10.1016/S0896-6273(00)80624-8

20


575

Oh, W.C., Hill, T.C., Zito, K., 2013. Synapse-specific and size-dependent mechanisms of spine

576

structural plasticity accompanying synaptic weakening. Proc. Natl. Acad. Sci. U. S. A.

577

110, E305-12. https://doi.org/10.1073/pnas.1214705110

578

Oh, W.C.C., Parajuli, L.K.K., Zito, K., 2015. Heterosynaptic structural plasticity on local

579

dendritic segments of hippocampal CA1 neurons. Cell Rep. 10, 162–169.

580

https://doi.org/10.1016/j.celrep.2014.12.016

581

Paoletti, P., Bellone, C., Zhou, Q., 2013. NMDA receptor subunit diversity: impact on

582

receptor properties, synaptic plasticity and disease. Nat. Rev. Neurosci. 14, 383–400.

583

https://doi.org/10.1038/nrn3504

584

Pettit, D.L., Wang, S.S.H., Gee, K.R., Augustine, G.J., 1997. Chemical two-photon uncaging: A

585

novel approach to mapping glutamate receptors. Neuron 19, 465–471.

586

https://doi.org/10.1016/S0896-6273(00)80361-X

587

Ramiro-Cortés, Y., Hobbiss, A.F., Israely, I., 2014. Synaptic competition in structural plasticity

588

and cognitive function. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 369, 20130157.

589

https://doi.org/10.1098/rstb.2013.0157

590 591 592

Ramiro-Cortés, Y., Israely, I., 2013. Long Lasting Protein Synthesis- and Activity-Dependent Spine Shrinkage and Elimination after Synaptic Depression. PLoS One 8. Schanzenbacher, C.T., Langer, J.D., Schuman, E.M., 2018. Time- and polarity-dependent

593

proteomic changes associated with homeostatic scaling at central synapses. Elife 1–20.

594

https://doi.org/10.7554/eLife.33322

595

Soares, C., Lee, K.F.H., Béïque, J.-C., 2017. Metaplasticity at CA1 Synapses by Homeostatic

596

Control of Presynaptic Release Dynamics. Cell Rep. 21, 1293–1303.

597

https://doi.org/10.1016/j.celrep.2017.10.025

598

Sutton, M. a, Ito, H.T., Cressy, P., Kempf, C., Woo, J.C., Schuman, E.M., 2006. Miniature

599

neurotransmission stabilizes synaptic function via tonic suppression of local dendritic

600

protein synthesis. Cell 125, 785–799. https://doi.org/10.1016/j.cell.2006.03.040

601

Thiagarajan, T.C., Lindskog, M., Tsien, R.W., 2005. Adaptation to synaptic inactivity in

602

hippocampal neurons. Neuron 47, 725–737.

603


21


604

Turrigiano, G., 2011. Too Many Cooks? Intrinsic and Synaptic Homeostatic Mechanisms in

605

Cortical Circuit Refinement. Annu. Rev. Neurosci. 34, 89–103.

606

https://doi.org/10.1146/annurev-neuro-060909-153238

607

Turrigiano, G.G., Leslie, K.R., Desai, N.S., Rutherford, L.C., Nelson, S.B., 1998. Activity-

608

dependent scaling of quantal amplitude in neocortical neurons. Nature 391, 892–896.

609

https://doi.org/10.1038/36103

610

Vitureira, N., Goda, Y., 2013. The interplay between hebbian and homeostatic synaptic

611

plasticity. J. Cell Biol. 203, 175–186. https://doi.org/10.1083/jcb.201306030

612

Wallace, W., Bear, M.F., 2004. A Morphological Correlate of Synaptic Scaling in Visual

613 614

Cortex. J. Neurosci. 24, 6928–6938. https://doi.org/10.1523/JNEUROSCI.1110-04.2004 Yadav, A., Gao, Y.Z., Rodriguez, A., Dickstein, D.L., Wearne, S.L., Luebke, J.I., Hof, P.R.,

615

Weaver, C.M., 2012. Morphologic evidence for spatially clustered spines in apical

616

dendrites of monkey neocortical pyramidal cells. J Comp Neurol 520, 2888–2902.

617

https://doi.org/10.1002/cne.23070

618

22


619

Figure 1

23


620

Figure 1. Spines undergo structural scaling following activity blockade

621

a) Experimental protocol to induce HSP in hippocampal organotypic slices. b) Example mEPSC

622

recording traces from control and TTX-treated conditions at the 48 h time point. Bottom:

623

representative individual mEPSCs. c) Quantification of mEPSC amplitudes after 48 h of activity

624

blockade, represented as mean amplitudes ± SEM. ***p = 1.553e-08, Mann-Whitney. For all

625

data presented, total number of spines or mEPSCs are represented by the first number (n)

626

and followed by the number of independent neurons (N = in parentheses) from which they

627

were collected. Number of mEPSCs (n) from independent neurons (N): Control = 552 (8), TTX

628

= 742(8). d) Ranked control mEPSC amplitudes plotted against TTX condition at 48 h. Best fit

629

line to the data: TTX = Control x 1.45 – 6.06. e) Representative images of dendrites at either

630

0 or 48 h after TTX treatment or in control media. Asterisks indicate all the spines that were

631

analyzed. Scale bar = 5µm. f) Quantification of spine volumes for the indicated conditions.

632

***p = 0.91e-10, Two-way ANOVA, post-hoc Bonferroni. Number of spines (n) from

633

independent neurons (N) for the 0 h, 24 h and 48 h timepoints: Control = 103(7), 245(8), 285

634

(7). TTX = 172 (7), 232 (5), 208 (6). g) Ranked control spine volumes plotted against TTX

635

condition at 48 h. Best fit line to the data: TTX = Control2 x 5.44 + Control x 0.25 + 0.55. h)

636

Quantification of mEPSC amplitudes after 72 h in TTX. ***p = 1.883e-15, Mann-Whitney.

637

Number of mEPSCs (n) from independent neurons (N): Control = 1164 (10), TTX = 1302 (12).

638

i) Quantification of spine volume after 72 h in TTX. ***p = 0.0008, Mann-Whitney. Number of

639

spines (n) from independent neurons (N): Control = 258 (6), TTX = 214 (6). j) Ranked control

640

72 h mEPSC amplitudes plotted against TTX mEPSC amplitudes. Best fit line to the data: TTX

641

= Control x 1.10 – 0.01. k) Ranked control 72 h spine volumes plotted against TTX 72 h spine

642

volumes. Best fit line to the data: TTX = Control x 1.26 – 0.004).

643

24


644

Figure 2

645

25


646

Figure 2 Structural and functional changes after HSP are reversible

647

a) Experimental timeline for reversal experiments. b) Example traces showing cell firing

648

following 48 h of activity block. Spikes are demarcated with black circles. c) Quantification of

649

spikes per minute throughout experiment. *p48 control – 48 TTX = 0.03, Kruskal-Wallis post hoc

650

Dunn. **p48 TTX – 48 post TTX = 0.010, Kruskal-Wallis post hoc Dunn. Number of neurons: 48h TTX

651

condition: 48 h Control = 17, 48 h Acute TTX = 20, 48 h TTX = 26; 48h ‘post-TTX’ Control = 15,

652

48 h post-TTX Reversal = 18. d) Representative images from dendrites after 48 h of HSP

653

induction and at 48 h post TTX. Scale bar = 5µm. e) Quantification of spine sizes throughout

654

the experiment. Graph is plotted as mean ± SEM. All p values calculated using Kruskal-Wallis

655

test, post-hoc Dunn’s. ***p48 TTX – 48 post TTX = 4.1852e-07. Number of spines (n = first number)

656

of Neurons (N = in parentheses): 48 h Control = 285 (7), 48 h TTX = 208 (6), 48 h ‘post-TTX’

657

Control = 514 (14), 48 h post-TTX = 430 (15).

658

26


659 660 661

Figure 3

27


662

Figure 3. Potentiation of single inputs is enhanced at smaller spines following HSP

663

a) Timeline for uncaging experiments. b) Representative images of structural plasticity

664

following single spine glutamate uncaging. Arrowheads and dots indicate stimulated and

665

neighboring non-stimulated spines respectively. Scale bar = 5µm (main image), 1 µm (inset).

666

c) Normalized volume of stimulated and neighboring spines for control and TTX-treated

667

conditions over 2 hours of LTP induction. Volumes were calculated from Z-stacks taken in 5

668

minute bins from time point 0 (thus the stack represented as 0 was collected between 0-5

669

minutes, for example). For all data, the total number of spines are represented by the first

670

number (n), and followed by the number of independent neurons (N = in parentheses) from

671

which they were collected. Control Stimulated = 17 (17), TTX Stimulated = 18 (18), Control

672

neighbors = 85 (17), TTX neighbors = 102 (18). d) All analyzed stimulated and neighboring

673

spine volumes at time points 45 minutes and 120 minutes post stimulation. Significance was

674

calculated using one-way ANOVA post-hoc Bonferroni. Control condition: **pstimulated45

675

neighbors45 = 8.4246e-04. pstimulated120 – neighbors120 = 0.6881. TTX condition: ***pstimulated45 – neighbors45

676

= 2.9383e-04. **pstimulated120 – neighbors120 = 0.0046. (n same as in panel c). e) Initial absolute spine

677

size vs. normalized spine volume at 120 minutes. TTX spines linear regression: r2 = 0.48, **p

678

= 0.002. Control linear regression: r2 = 0.06, p = 0.346. f) Distribution of all initial volumes of

679

stimulated spines, with example spine images. The cut-off for defining ‘large’ spines was the

680

median volume off all spines multiplied by 1.5. Scale bar = 1 µm. g) Time-course of the

681

structural LTP experiment plotting only the small spines. *p = 0.0458, 2-way repeated

682

measures ANOVA). Number of spines (n) from neurons (N = in parentheses): Control = 13 (13),

683

TTX = 11 (11). h) Time-course of structural LTP plotting only large spines. p = 0.699, 2-way

684

repeated measures ANOVA. Number of spines (n) from neurons (N = in parentheses): Control

685

= 4 (4), TTX = 17 (17).

–

686

28


687

Figure 4

29


688

Figure 4. HSP increases the efficacy and reduces the threshold for induction of LTP at

689

individual spines

690

a) Representative image of a single spine during the course of uncaging. Scale bar = 5µm (main

691

image), 1µm (inset). b) Individual traces of spine growth during glutamate uncaging

692

stimulation. Control spine growth vs TTX spine growth: p = 0.198, 2-way repeated measures

693

ANOVA. Number of spines (n): Control = 16, TTX = 18. c) Growth rates of stimulated spines

694

for 1 minute after the end of stimulation. *psmall control–small TTX = 0.018, 1-way ANOVA with post-

695

hoc Tukey’s. plarge control–large TTX < 0.99, 1-way ANOVA with post-hoc Tukey’s. Number of spines

696

(n): Control small = 13, TTX small = 11, Control large = 4, TTX large = 7. d) Quantification of

697

normalized spine volume with different stimulation pulse durations in the control condition.

698

p1ms

699

Number of spines (n) from Neurons from which they were collected (N = in parentheses):

700

stimulated 1ms = 11 (11), neighbors 1ms = 52 (11), stimulated 4ms = 17 (17). *p1ms stimulated–

701

4ms stimulated

702

Quantification of normalized spine volume with different stimulation pulse durations in the

703

TTX condition. **p1ms stimulated–1ms neighbors = 0.007, 2-way rm ANOVA. p1ms stimulated– 4ms stimulated =

704

0.27, 2-way rm ANOVA). Number of spines (n) from Neurons (N): stimulated 1ms = 11 (11),

705

neighbors 1ms = 52 (11), stimulated 4ms = 18 (18). Note 4ms data is the same as in Figure 3.

stimulated–1ms neighbors

= 0.292, 2-way repeated measures ANOVA post-hoc Bonferroni.

= 0.039 2-way rm ANOVA. Note 4ms data is the same as shown in Figure 3. e)

30


706

Figure 5

31


707

Figure 5. Stimulation of single inputs leads to plasticity of clustered neighbors after HSP

708

a) Heat maps of growth dynamics of neighboring spines over time. Neighbors from all

709

experiments were pooled and ranked in order distance from the stimulated spine. Index 1 is

710

the spine nearest to its stimulated spine, and 80 is the furthest away. Heat map = normalized

711

spine volume. Number of spines (n) from Neurons from which they were collected (N = in

712

parentheses): Control = 80 (17), TTX = 80 (18). b) Mean volumes for the first 10 minutes

713

following stimulation for the different neighboring conditions. ***pTTX near – TTX far = 1.0674e-

714

05, Kruskal-Wallis. Number of spines (n) from Neurons from which they were collected (N =

715

in parentheses): Control near = 22 (17), control far = 63 (17), TTX near = 32 (18), TTX far = 70

716

(18). c) Normalized volumes of near and far neighboring spines in the control condition. p =

717

0.987, 2-way rm ANOVA. Number of spines – same as above. d) Normalized volumes of

718

neighboring spines in the TTX condition. *p = 0.0124, 2-way repeated measures ANOVA.

719

Number of spines – same as above. e) & f) Correlation coefficients of volume change of

720

stimulated spines vs volume change of neighboring spine, plotted against the distance

721

between the neighboring spine and the stimulated spine. e) Linear regression control: r2 =

722

0.003, p = 0.641. f) Linear regression TTX: r2 = 0.156, ***p = 3.91e-05).

32