1: 

2: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

3: \usepackage{graphicx}% Include figure files

4: \usepackage{dcolumn}% Align table columns on decimal point

5: \usepackage{bm}% bold math

6: 

7: \input{psfig.sty}

8: %\textwidth=8in \textheight=7.5in \oddsidemargin=-0.3in

9: %\evensidemargin=-0.3in

10: \topmargin=-0.2in

11: 

12: 

13: %\def\doublespace{\parskip 3pt plus 1.2pt

14:  %    \baselineskip 18pt plus 1pt minus .5pt

15:   %   \lineskip 2pt plus 1pt \lineskiplimit 5pt}

16: 

17: \newcommand{\be}{\begin{equation}}

18: \newcommand{\bea}{\begin{eqnarray}}

19: \newcommand{\ee}{\end{equation}}

20: \newcommand{\eea}{\end{eqnarray}}

21: \def \blank{\mbox{}}

22: \def\d{\mbox{$\partial$}}

23: \def\de{\mbox{$d_E$}}

24: \def\dw{\mbox{$d_W$}}

25: \def\dl{\mbox{$d_L$}}

26: \def\da{\mbox{$d_A$}}

27: \def\hb{\mbox{$\hbar$}}

28: \def\Re{\mbox{\,Re\,}}

29: \def\Im{\mbox{\,Im\,}}

30: \def\dw{\mbox{$d_W$}}

31: \def\nb{\mbox{$N_B$}}

32: \def\A{\mbox{\bf A}}

33: \def\B{\mbox{\bf B}}

34: \def\C{\mbox{\bf C}}

35: \def\M{\mbox{\bf M}}

36: \def\P{\mbox{\bf P}}

37: \def\x{\mbox{{\bf x}}}

38: \def\ol{\mbox{$\bar{\lambda}$}}

39: \def\Q{\mbox{{\bf Q}}}

40: \def\R{\mbox{{\bf R}}}

41: \def\S{\mbox{{\bf S}}}

42: \def\L{\mbox{$L$}}

43: \def\Rd{\mbox{$R^d$}}

44: \def\DF{\mbox{{\bf DF}}}

45: \def\DG{\mbox{{\bf DG}}}

46: \def\DH{\mbox{{\bf DH}}}

47: \def\DFL{\mbox{$\DF^L$}}

48: \def\DGL{\mbox{$\DG^L$}}

49: \def\OSL{\mbox{{\bf OSL}}}

50: \def\T{\mbox{$T$}}

51: \def\Ec{\mbox{{$\cal E$}}}

52: \def\Pc{\mbox{{$\cal P$}}}

53: \def\Ac{\mbox{{$\cal A$}}}

54: \def\l{\mbox{$\lambda$}}

55: \def\g{\mbox{$\gamma$}}

56: \def\oma{\mbox{$\omega_A$}}

57: \def\om{\mbox{$\omega$}}

58: \def\eps{\mbox{$\epsilon$}}

59: \def\Y{\mbox{$\bf{Y}$}}

60: \def\y{\mbox{$\bf{y}$}}

61: \def\v{\mbox{$\bf{y}$}}

62: \def\ynn{\mbox{$\bf{y}^{NN}$}}

63: \def\w{\mbox{$\bf{w}$}}

64: \def\z{\mbox{$\bf{z}$}}

65: \def\e{\mbox{$\bf{e}$}}

66: \def\del{ \mbox{\boldmath{$\Delta$}} }

67: \def\delt{ \mbox{\boldmath{$\delta$}} }

68: \def\domega{ \mbox{$\delta\omega$}}

69: \def\xint{\mbox{$\displaystyle\int d^3x \,$}}

70: \def\f{\mbox{$\bf{f}$}}

71: \def\p{\mbox{$\bf{\rho}$}}

72: \def\F{\mbox{$\bf{F}$}}

73: \def\E{\mbox{$\bf{E}$}}

74: \def\D{\mbox{$\bf{D}$}}

75: \def\H{\mbox{$\bf{H}$}}

76: \def\G{\mbox{$\bf{G}$}}

77: \def\U{\mbox{$\bf{U}$}}

78: \def\J{\mbox{$\bf{J}$}}

79: \def\ca{\mbox{Ca$^{2+}$}}

80: 

81: %\renewcommand{\topfraction}{0.85}

82: %\renewcommand{\textfraction}{0.1}

83: %\renewcommand{\floatpagefraction}{0.75}

84: \begin{document}

85: \preprint{}

86: \title{ Synaptic plasticity of Inhibitory synapse promote synchrony in inhibitory network in presence of

87: heterogeneity and noise}% Force line breaks with \\

88: 

89: \author{Sachin S. Talathi}%

90:  \email{talathi@physics.ucsd.edu}

91: \affiliation{%

92: Department of Physics and Institute for Nonlinear Science \\

93: University of California, San Diego, La Jolla, CA 92093-0402, USA

94: }

95: \date{\today}

96: \begin{abstract}

97: Recently spike timing dependent plasticity was observed in inhibitory synapse in the layer II of entorhinal cortex. The rule provides an interesting zero in the region of $\Delta t=t_{post}-t_{pre}=0$ and in addition the dynamic range of the rule lie in gamma frequency band. We propose a robust mechanism based on this observed synaptic plasticity rule for inhibitory synapses for two mutually coupled

98: interneurons to phase lock in synchrony in the presence of intrisic heterogeneity in firing. We study the stability of the phase locked solution by defining a map for spike times dependent on the phase response curve for  the coupled neurons. Finally we present results on robustness of synchronization in the presence of noise.

99: \end{abstract}

100: \pacs{Valid PACS appear here}

101: \maketitle

102: 

103: It is generally accepted that inhibitory interneurons are

104: important for synchrony in neocortex. Several studies have

105: reported the role for inhibitory interneurons in generating stable

106: synchronous rhythms in neocortex

107: ~\cite{Benardo,Jefferys,Michelson,Whittington,Bragin}. Cortical

108: oscillations in the gamma frequency band (20-80 Hz), are thought

109: to be involved in binding of object properties, a process of great

110: significance for the functioning of the brain. These experimental

111: findings have led to numerous theoretical studies of synchrony

112: among inhibitory interneurons ~\cite{van, Ernst, Traub, Wang}. The

113: principle result of these studies showed that depending on the

114: decay time of the inhibitory synaptic coupling, the mutually

115: coupled inhibitory neurons oscillate in synchrony ( in phase

116: locking) or in antisynchrony (out of phase locking). However much

117: of the above investigations did not explore the effects of

118: heterogeneity in the intrinsic firing rates nor did they take into

119: account noise, which is invariably present in neuronal systems.

120: 

121: In another set of theoretical investigations, ~\cite{White}

122: explored the implications of small heterogeneity for the

123: degradation of synchrony of fast spiking inhibitory neurons and

124: the mechanism by which the degradation occur. They found that

125: introduction of even small amounts of heterogeneity in the

126: external drive, resulted in significant reduction in coherence of

127: neuronal spiking. It is important then to understand what mediates

128: observed in vivo synchrony of neuronal networks under biological

129: realistic conditions of noise induced unreliability and intrinsic

130: heterogeneity in spiking rates of the neuronal ensemble.

131: 

132: In this letter we propose a robust mechanism based on spike timing dependent synaptic plasticity of

133: inhibitory synapses ~\cite{Haas}

134:  by which two coupled interneurons can phase lock in synchrony

135: even under conditions of mild heterogeneity in the firing rates of

136: the coupled neurons and in the presence of noise. Earlier work

137: \cite{Valentin} has explored the function of synaptic plasticity

138: in the excitatory synapse in improving synchronization in

139: unidirectionally coupled neuronal network. We consider a network

140: of two coupled interneurons with self inhibition as shown in

141: Figure \ref{Fig1}b. The self-inhibition is introduced because

142: biological neural networks often have

143:  local inhibitory interneurons which deliver feedback inhibition to the cells activating those

144:  interneurons ~\cite{Sheperd}. The importance of self inhibition that simulates the network effect, is explored in

145:  details in \cite{Talathi_Neurosci}.

146: 

147:  Each neuron in the coupled network is modelled as

148: \begin{eqnarray}

149: \frac{dV_{i}(t)}{dt}&=&I^{i}_{In}+g_{Na}m^{3}(t)h(t)(E_{Na}-V_{i}(t)) \nonumber \\

150: &+&g_{K}n^{4}(t)(E_K-V_{i}(t))+g_{L}(E_L-V_{i}(t))  \nonumber \\

151: &+&I^{ij}_{M}(t)+I^{i}_{S}(t) +\eta \zeta_{i}(t) \nonumber \\

152: \end{eqnarray}

153: where $V_{i}(t)$ (i=A,B) is the membrane potential ,

154: $\zeta_{i}(t)$ is the gaussian synaptic noise of amplitude $\eta$ satisfying $\left<\zeta(t)\right>=0$ and

155: $\left<\zeta_{i}(t)\zeta_{j}(t^{'})\right>=\delta(t-t^{'})\delta_{ij}$.

156: $I^{i}_{In}$, is the external drive, $I^{i}_{S}(t)=g_{s}S(t,V_{i}(t))(E_{I}-V_{i}(t))$, is the synaptic current due to self inhibition and

157: $I^{ij}_{M}(t)=g_{j\to i}S(t,V_{j}(t))(E_{I}-V_{i}(t))$ is the synaptic current from mutual

158: inhibition. $g_{j\to i}=g_{m}G(t)$ is the dynamic synapse, whose strength is determined by the

159: inhibitory synaptic plasticity rule, and $g_{s}$ is the synaptic strength of self inhibition.

160: $E_r$ (r=Na, K, L) are reversal potentials of the sodium and

161: potassium ion channels and the leak channel respectively. $E_{I}$ is the reversal potential of

162: the inhibitory synapse. $S(t,V(t))$ give the fraction of bound receptors and satisfy the first

163:  order kinetic equation, $\dot{S}(t)=\frac{S_{0}(V_{pre}(t))-S(t)}{\hat{\tau}(S_{I}-S_{0}(V_{pre}))}$

164:  where $V_{pre}(t)$ is the presynaptic voltage. It involves two time constants, $\tau_{r}=\hat{\tau}(S_{I}-1)$, the docking time for

165: the neurotransmitter and $\tau_{d}=\hat{\tau}S_{I}$, the undocking time constant for the neurotransmitter binding.

166:  $S_{0}(V)$ is the sigmoidal function given by,

167:  $S_{0}(V)=0.5(1+\tanh(120(V-0.1)))$. The gating

168: variables X(t),(X=m,h,n), satisfy first order kinetic equations,

169:  $\dot{X}(t)=\alpha_{X}(V)(1-X(t))-\beta_{X}(V)X(t)$.

170: We have used standard functions $\alpha_{X}(V)$ and $\beta_{X}(V)$ and parameters for the model \cite{Talathi},

171:  such that the dynamics of the neurons to spiking is through saddle node bifurcation and

172:  the model represents a type I neuron \cite{Ermentrout}.

173:  The model parameters are within physiological range and give high spike rates typical of interneurons.

174: 

175: \begin{figure*}[ht!]

176: \includegraphics[width=5.0in,scale=1,angle=0]{Fig1.eps}

177: \caption{(a) Schematic diagram of the self inhibited neuron

178: considered to determine the different regimes of operation of

179: network of mutually coupled interneurons. (b) Schematic of the

180: mutually coupled interneurons. (c) Ratio of the synaptic decay

181: constant $\tau_{d}$ to firing period T plotted versus $\tau_{d}$.

182: Parameter values for the operation of the neuron in the two

183: regimes, are shown. (d) The ration of firing period of two coupled

184: neurons in phasic and tonic regimes are plotted as function of

185: heterogeneity in external drive.\label{Fig1}}

186: \end{figure*}

187: 

188: We consider two regimes of operation of the network shown in Figure \ref{Fig1}b. The two regimes

189: called the phasic regime and the tonic regime \cite{White} are determined from

190: the firing characteristics of a single self inhibited neuron (Figure \ref{Fig1}a). In the phasic regime

191: the network period depends on the synaptic decay constant ~\cite{White} and in the tonic regime the synaptic dynamics

192: weakly affect the network period. These two regimes of operation are clearly illustrated in

193: the plot of $\tau_{d}/T$, which is the ratio of synaptic decay constant to the firing frequency of the self

194: inhibited neuron versus $\tau_{d}$ in Figure \ref{Fig1}b for various values of $I_{In}$ and $g_{s}$.

195: These two regimes of network oscillations were observed earlier \cite{White}, with the synaptic model

196: for the inhibitory synapse, obeying the first order kinetic equation,

197: $\dot{S}(t)=F(V)(1-S(t))-S(t)/\tau_d$. Changing $\tau_d$ in this situation not only changes the decay time of

198: the synapse but also the rise time given by $\tau_{r}=\frac{\tau_{d}}{\tau_{d}F(V)+1}$ and the

199: saturation level of the synapse $S_{max}=\frac{\tau_{d}F(V)}{\tau_{d}F(V)+1}$. As a result a narrow region of higher harmonic phase locking was observed between the coupled interneurons in the phasic regime.

200:  We have therefore considered the synaptic model

201: presented above where we have control over the decay time for the

202: synapse independent of the rise time and the saturation level of

203: the synapse. Earlier work \cite{Chow} has shown that in presence

204: of mild heterogeneity, coherence in neuronal firing is observed

205: much more for phasic regime than in tonic regime. In the results

206: presented here we therefore

207:  consider the phasic regime of the network operation and study the effect of STDP in inhibitory synapses,

208:   in maintaining coherence in neuronal firing in presence of heterogeneity and noise. Details on the calculations

209:   for the tonic regime will appear elsewhere \cite{Talathi_Neurosci}

210: 

211: In all the subsequent calculations, we fix the parameters of the model in the phasic regime, ($I_{In}=2.5 \mu A/cm^{2}$,

212:  $g_{s}=g_{m}=0.1 mS/cm^{2}$, $\tau_{r}=1.1$ ms, $\tau_{d}=5.0$ ms) and study

213: the effect of the dynamic synapse in maintaining synchrony in presence of

214: heterogeneity and intrinsic noise. In figure \ref{Fig1}d, we plot the ratio of mean firing periods

215:  $\frac{<T_{B}>}{<T_{A}>}$ for the two coupled neurons A and B as function of heterogeneity in the external drive,

216:  defined as

217: $H=100\frac{I^{A}_{In}-I^{B}_{In}}{I^{A}_{In}}$ in absence of

218: noise i.e., $\eta=0$. As can be seen from Figure \ref{Fig1}d, (top

219: panel), the region of 1:1 locking, as function of heterogeneity is

220: much broader in the phasic regime of network operation as compared

221: to the tonic regime, where we set $I^{B}_{In}=5.0 \mu A/cm^{2}$.

222: In addition, higher order synchronization are also present in the

223: phasic regime, as a result coherence between the two neurons is

224: preserved more often in the phasic regime of the network

225: operation.

226: 

227: A spike timing dependent plasticity (STDP) rule for inhibitory synapses has been recently reported in

228: \cite{Haas} and it has the form

229: $\Delta g(\Delta t)=\frac{g_{0}}{g_{norm}}\alpha^{\beta}|\Delta t|\Delta t^{\beta -1}e^{-\alpha|\Delta t|}$,

230:  where $\Delta t=t_{post}-t_{pre}$. $t_{pre}$ is the time of presynaptic spike stimulation and $t_{post}$

231:  is the time of a spike generated by the postsynaptic neuron. $g_{0}$ is the scaling factor accounting for the

232:   amount of change in inhibitory

233: conductance induced by the synaptic plasticity rule and $g_{norm}=\beta e^{-\beta}$ is the normalizing constant.

234:  An empirical fit of the above function to the data gives,

235: $\alpha=1$ and $\beta=10$, giving a window of $\pm 20$ ms  over which the efficacy of synaptic plasticity is non

236: zero.  This implies that

237: in the physiologically important regime of gamma oscillations (25-80 Hz), STDP rule of inhibitory synapses

238: can play a significant role in modulating the firing dynamics of the neuronal network.

239: 

240: \begin{figure*}

241: \includegraphics[width=5.0in,scale=1,angle=0]{Fig2.eps}

242: \caption{(a) Ratio of firing periods of the two coupled

243: interneurons, in the static case, when the inhibitory synaptic

244: strength is constant and dynamic case when the inhibitory synaptic

245: strength is modulated by STDP rule, is plotted as function in

246: heterogeneity H. (b) The evolution of phase difference $\psi^{AB}$

247: is plotted in the static and dynamic case, when the heterogeneity

248: is set at 6 \%. (c) The evolution of the inhibitory synaptic

249: strength between the two coupled interneurons, A and B is plotted

250: as function of time. \label{Fig2}}

251: \end{figure*}

252: 

253: 

254: We now consider two situations in the phasic regime: when the strength of mutual inhibition is static:

255:  $g_{A\to B}=g_{B\to A}=g_{m}$ and when the mutual inhibition strength is dynamic and governed by observed

256:  synaptic plasticity rule, i.e., $g_{A\to B}=g_{m}(1+\frac{1}{g_{m}}\tilde{g}(t))$ and

257: $g_{B\to A}=g_{m}(1-\frac{1}{g_{m}}\tilde{g}(t))$. In order to take into account the effect of multiple spike

258: pairs, we follow \cite{Froemke} and define

259: $\tilde{g}(t)=\sum_{j}\sum_{i}\Delta g(\Delta t_{ij}) \epsilon^{A}_{i}\epsilon^{B}_{j}$, where

260: $\Delta t_{ij}=t^{B}_{j}-t^{A}_{i}$

261: is the difference in spike times of neurons A and B respectively. $\epsilon^{A,B}_{k}$ gives the efficacy of

262: spike in A and B and is defined as $\epsilon^{K}_{i}=e^{-(t^{K}_{i}-t^{K}_{i-1})/\tau_{e}}$.

263: We take $\tau_{e}\approx 55$ms,

264: an average of the efficacy values given in \cite{Froemke} as experimental results on contribution of multiple spike

265: pairing to inhibitory synaptic plasticity are as yet unknown. In Figure \ref{Fig2}a we plot the ratio

266: $<T_{B}>/<T_{A}>$ as function of heterogeneity H in the static and dynamic case.

267: 

268: As shown in Figure \ref{Fig2}a there is considerable increase in 1:1, 1:2 and 2:1 synchronization windows

269: mediated by the dynamic synapse in the phasic regime. This implies an increased probability

270: of observing coherence

271: in the firing pattern of the mutually coupled

272: interneurons even in presence of mild heterogeneity as has been reported in many in vivo experimental data.

273:  In Figure \ref{Fig2}b ,we show the evolution of phase difference

274: $\psi^{AB}_{i}=\left(|t^{B}_{i}-t^{A}_{i}|\mod <T_{A}>\right )$  in the static and dynamic case,

275: with heterogeneity of 6 \%. We see that in the static case the phase difference grows linearly

276: modulo $<T_{A}>$, representing situation of asynchronous firing. However in the dynamic case, after the initial

277: transient is over, the phase difference saturates to a fixed value, (1.4 in this example), representing stable 1:1

278: locking between the two mutually coupled neurons. In Figure \ref{Fig2}c we plot the evolution of the

279: synaptic strength as function of time. We see that the STDP rule results in modulating the synaptic strength

280: so as to phase lock the two mutually coupled neurons.

281: 

282: \begin{figure}[ht!]

283: \includegraphics[width=3.0in,scale=1,angle=0]{Fig3.eps}

284: \caption{Plot of the difference in phase response of the two coupled neurons in the static $\eta^{AB}$ and the dynamic

285: case, $\tilde{\eta}^{AB}$ is plotted as function of the phase difference. Stability analysis on $\eta^{AB}$

286: and $\tilde{\eta}^{AB}$ , determines whether the two coupled neurons will phase lock with each other. \label{Fig3}}

287: \end{figure}

288: 

289: In order to understand the dynamics of phase locking under

290: heterogeneity in the presence of dynamic synapse, we consider the

291: situation initially in the static case. Let the two coupled

292: neurons fire heterogeneously with intrinsic firing period

293: $T^{0}_{A}$ and $T^{0}_{B}$ when they are uncoupled. In the case

294: of 6 \% heterogeneity, we have $I^{A}_{In}>I^{B}_{In}$ so that

295: $T^{0}_{A}< T^{0}_{B}$. Let $\phi(\epsilon)$ be the phase response

296: curve of the neuron model. It is known that for type I neuron

297: models, the phase response curve is positive \cite{Ermentrout},

298:  and as a result every time a spike arrives, the phase of subsequent spike is delayed.  In the

299:  situation considered, with initial phase difference between the two neurons set to zero, the spike times of

300:  individual firing neurons can be written as,

301:   $t^{K}_{j}=t^{K}_{j-1}+T^{0}_{K}+\phi^{K}(\epsilon^{K}_{j-1})$, (K=A,B) with

302:  $\epsilon^{K}_{j-1} >0$,

303:  where $\epsilon^{A}_{j}=t^{B}_{j}-t^{A}_{j}$   and $\epsilon^{B}_{j}=t^{A}_{j+1}-t^{B}_{j}$. $t^{K}_{j}$ is the time

304:  of $j^{th}$ spike of neuron K. The map evolving the

305:  phase difference $\psi^{AB}$  is then,

306: \begin{eqnarray}

307: \psi^{AB}_{j}&=&(T^{0}_{B}-T^{0}_{A})+\psi^{AB}_{j-1}+\phi^{B}(t^{A}_{j}-t^{B}_{j-1}) \nonumber \\

308: &-&\phi^{A}(t^{B}_{j-1}-t^{A}_{j-1}) \nonumber \\

309: &\approx&(T^{0}_{B}-T^{0}_{A})+\psi^{AB}_{j-1}- \eta^{AB}(\psi^{AB}_{j-1})

310: \end{eqnarray}

311:  where $\eta^{AB}(x)=\phi^{A}(x)-\phi^{B}(<T_{A}>-x)$.

312: The fixed point of the map is then given by

313: $$\eta^{AB}(\psi^{AB}_{j})=T^{0}_{B}-T^{0}_{A} $$

314: Numerical solution to above equation gives the fixed points, at $\psi^{AB}_{*s}\approx 3.85$ as can be seen in Figure

315: \ref{Fig3}, where we plot $\eta(\psi^{AB})$ versus $\psi^{AB}$.

316: Stability of the fixed point would require

317: $|1-\frac{d\eta^{AB}(x)}{dx}|_{x=\psi^{AB}_{*s}}<1$ $\Rightarrow$

318: $0< \frac{d\eta^{AB}(x)}{dx}|_{x=\psi^{AB}_{*s}}<2$

319: 

320: For the set of parameters considered however, the numerical value obtained for

321: $\frac{d\eta^{AB}(x)}{dx}|_{x=\psi^{AB}_{*s}}=-2.5$ for the fixed point.

322: 

323: Thus we see that when the synapse is static the phase difference is unstable and the two coupled

324: interneurons fire asynchronously in the

325: situation of

326: 6\% heterogeneity. Now consider the situation in the dynamic synapse case.

327: Again setting the initial phase difference zero, in the presence of dynamic synapse with STDP, we have

328: $t^{K}_{j}=t^{K}_{j-1}+T^{0}_{K}+\Hat{\phi}^{K}(\epsilon^{K}_{j-1},\tilde{g}_{j-1}(t))$, where phase shift given by

329: the phase response curve also depends on the dynamics of the synaptic coupling strength governed by the STDP

330: rule.

331: The map function for evolution of the phase shift $\psi^{AB}$ is then obtained as,

332: \begin{eqnarray}\psi^{AB}_{j}\approx(T^{0}_{B}-T^{0}_{A})+\psi^{AB}_{j-1}-\tilde{\eta}^{AB}(\psi^{AB}_{j-1})\end{eqnarray}

333:  where $\tilde{\eta}^{AB}(x)=\Hat{\phi}^{A}(x)-\Hat{\phi}^{B}(<T_{A}>-x)$.

334:  The fixed point of the map is then given by

335:  $\tilde{\eta}^{AB}(x)=T^{0}_{B}-T^{0}_{A}$.

336: As can be seen from Figure 3b, where we plot $\tilde{\eta}^{AB}(\psi^{AB})$, in the steady state ($t\to

337: \infty$), when $g_{B\to

338: A}=.146$ and $g_{B\to A}=0.054$, we obtain, $\psi^{AB}_{*d}\approx1.4$ and $\psi^{AB}_{*d}\approx 3.3$.

339: Stability of the fixed point given by

340: $0<\frac{d\tilde{\eta}^{AB}(x)}{dx}|_{x=\psi^{AB}_{*}}<2$, implies $\psi^{AB}_{*d}\approx1.4$ is stable as can

341: be also seen from the time evolution of phase in Figure \ref{Fig2}b.

342: Thus we see that STDP of inhibitory synapse, modulates the phase response curve such that the network locks into

343: synchrony even under mild heterogeneity.

344: \begin{figure}[ht!]

345: \includegraphics[width=2.0in,scale=1,angle=-90]{Fig4a.eps}

346: \caption{Ratio of firing periods $<T_{B}>/<T_{A}>$ is plotted as function of heterogeneity H, in presence of

347: synaptic gaussian noise with amplitude a=0.1 \label{Fig4}}

348: \end{figure}

349: In Figure \ref{Fig4} we present results on synchrony in presence of noise. We set the noise amplitude

350: $\eta=0.1$ in equation 1.

351: For mild noise, STDP of inhibitory synapse, is able to maintain synchrony between the two

352: coupled interneurons under conditions of mild heterogeneity in the drive.

353: 

354: We have also tested the dynamics of the network in the tonic regime. STDP of inhibitory synapse, also

355:  significantly increases the window of synchronous oscillations by the same

356: mechanism \cite{Talathi_Neurosci}.

357: 

358: It has been suggested in \cite{Haas} that plasticity of inhibitory synapses may play an important role in

359: balancing the effect of excitatory synapse preventing run away behavior typically observed in

360: epileptogenesis. In this work we present an important function for STDP in inhibitory synapse in maintaining

361: synchrony in networks of coupled interneurons, under biologically realistic situation of mild heterogeneity

362: and noise.

363: 

364: {\bf Acknowledgements}\\

365: This work was partially funded by a grant from the National

366: Science Foundation, NSF PHY0097134. SST was partially supported by

367: the NSF sponsored Center for Theoretical Biological Physics at

368: UCSD. We would like to thank Henry Abarbanel, Julie Haas, Thomas

369: Nowotny and Johnathan Driscoll for instructive feedbacks that

370: improved this work significantly.

371: \begin{thebibliography}{99}

372: \bibitem{Benardo} Benardo L.S., {\em J Neurophysiol}, {\bf 77} 3134-3144, (1997).

373: 

374: \bibitem{Jefferys} Jefferys J.G., Traub R.D., Whittington M.A., {\em Trends Neurosci}, {\bf 19} 202-208, (1996).

375: 

376: \bibitem{Michelson} Michelson H.B., Wong R.K., {\em J Physiol (Lond)}, {\bf 477} 35-45, (1994).

377: 

378: \bibitem{Whittington} Whittington M.A., Traub R.D., Jefferys J.G., {\em Nature}, {\bf 373} 612-615, (1995).

379: 

380: \bibitem{Bragin} Bragin A., Jando G., Nadasdy Z., Hetke J., Wise K., Buzsaki G.,{\em J Neurosci}, {\bf 15} 47-60, (1995).

381: 

382: \bibitem{van} van Vreeswijk C., Abbott L.F., {\em J Comp Neurosci}, {\bf 1} 313-321, (1994).

383: 

384: \bibitem{Ernst} Ernst U., Pawelzik K., Geisel T., {\em Phys Rev Lett}, {\bf 74} 1570-1573, (1995).

385: 

386: \bibitem{Traub} Traub R.D., Jefferys J.G., Whittington M.A., {\em J Comp Neurosci}, {\bf 416} 433-438, (2002).

387: 

388: \bibitem{Wang} Wang X.J., Rinzel J.,  {\em Neural Compute}, {\bf 4} 84-97, (1992).

389: 

390: \bibitem{Froemke} Froemke R.C., Dan Y., {\em Nature}, {\bf 416} 433-438, (2002).

391: 

392: \bibitem{Kempter} Kempter R., Gerstner W., Hemmen van J.L.,  {\em Physical Review E}, {\bf 8} 979-1001, (1996).

393: 

394: \bibitem{Robert} Roberts P.D., {\em J Comp Neurosci}, {\bf 7} 235-246, (1999).

395: 

396:  \bibitem{White} White A.J., Chow C.C., Ritt J., Trevino C.S., Kopell N., {\em J Comp Neurosci}, {\bf 5} 5-16, (1998).

397: 

398:  \bibitem{Chow} Chow C.c., White J.A., Ritt J., Kopell N., {\em J Comp Neurosci}, {\bf 5}, 407-420, (1998).

399:  \bibitem{Haas} Haas  J., Nowotny T., Abarbanel H.D.I.A., {\em J Neurophysiol}, Submitted.

400: 

401: \bibitem{Valentin} Zhigulin V.P., Rabinovich M.I., Huerta R., Abarbanel H.D.I.,{\em Phys Rev E}, {\bf 67} 021901, (2003).

402: 

403: \bibitem{Sheperd} Shepherd G.M., {\em The Synaptic Organization of the Brain}, {\bf 3rd ed}, Oxford University Press, New York, (1990).

404: 

405: \bibitem{Talathi} Abarbanel, H.D. I. and Talathi S.S., {\em Phys Rev Lett}, {\bf 96}, 148104 (2006).

406: 

407: \bibitem{Talathi_Neurosci} Talathi S.S. and Abarbanel H.D.I., Unpublished Results.

408: 

409: \bibitem{Ermentrout} Ermentrout B., {\em Neural Compute} {\bf 8} 979-1001, (1996).

410: 

411: 

412: \end{thebibliography}

413: 

414: \end{document}

415: