Encoding Network States by Striatal Cell Assemblies

Luis Carrillo-Reid, Fatuel Tecuapetla, Dagoberto Tapia, Arturo Hernndez-Cruz, Elvira Galarraga, Ren Drucker-Colin and Jos Bargas
You might find this additional info useful... This article cites 56 articles, 22 of which can be accessed free at: http://jn.physiology.org/content/99/3/1435.full.html#ref-list-1 This article has been cited by 12 other HighWire hosted articles, the first 5 are: Spike-Train Communities: Finding Groups of Similar Spike Trains Mark D. Humphries J. Neurosci., February 9, 2011; 31 (6): 2321-2336. [Abstract] [Full Text] [PDF] Functional Connectome of the Striatal Medium Spiny Neuron Nao Chuhma, Kenji F. Tanaka, Ren Hen and Stephen Rayport J. Neurosci., January 26, 2011; 31 (4): 1183-1192. [Abstract] [Full Text] [PDF]
Downloaded from jn.physiology.org on April 29, 2011

J Neurophysiol 99:1435-1450, 2008. First published 9 January 2008; doi:10.1152/jn.01131.2007

Dynamics of the Parkinsonian Striatal Microcircuit: Entrainment into a Dominant Network State Omar Jidar, Luis Carrillo-Reid, Adn Hernndez, Ren Drucker-Coln, Jos Bargas and Arturo Hernndez-Cruz J. Neurosci., August 25, 2010; 30 (34): 11326-11336. [Abstract] [Full Text] [PDF] Sequentially Switching Cell Assemblies in Random Inhibitory Networks of Spiking Neurons in the Striatum Adam Ponzi and Jeff Wickens J. Neurosci., April 28, 2010; 30 (17): 5894-5911. [Abstract] [Full Text] [PDF] Muscarinic Enhancement of Persistent Sodium Current Synchronizes Striatal Medium Spiny Neurons Luis Carrillo-Reid, Fatuel Tecuapetla, Nicolas Vautrelle, Adn Hernndez, Ramiro Vergara, Elvira Galarraga and Jos Bargas J Neurophysiol 2009; 102 (2): 682-690. [Abstract] [Full Text] [PDF] Updated information and services including high resolution figures, can be found at: http://jn.physiology.org/content/99/3/1435.full.html Additional material and information about Journal of Neurophysiology can be found at: http://www.the-aps.org/publications/jn

This infomation is current as of April 29, 2011.

Journal of Neurophysiology publishes original articles on the function of the nervous system. It is published 12 times a year (monthly) by the American Physiological Society, 9650 Rockville Pike, Bethesda MD 20814-3991. Copyright 2008 by the American Physiological Society. ISSN: 0022-3077, ESSN: 1522-1598. Visit our website at http://www.the-aps.org/.

J Neurophysiol 99: 14351450, 2008. First published January 9, 2008; doi:10.1152/jn.01131.2007.

Encoding Network States by Striatal Cell Assemblies

Luis Carrillo-Reid,1 Fatuel Tecuapetla,1 Dagoberto Tapia,1 Arturo Herna ndez-Cruz,1 Elvira Galarraga,1 2 1 Rene Drucker-Colin, and Jose Bargas
Departamentos de Biof sica and 2Neurociencias, Instituto de Fisiolog a Celular, Universidad Nacional Auto noma de Me xico, Mexico City, Mexico
Submitted 12 October 2007; accepted in nal form 8 January 2008
1

Carrillo-Reid L, Tecuapetla F, Tapia D, Hernandez-Cruz A, Galarraga E, Drucker-Colin R, Bargas J. Encoding network states by striatal cell assemblies. J Neurophysiol 99: 14351450, 2008. First published January 9, 2008; doi:10.1152/jn.01131.2007. Correlated activity in cortico-basal ganglia circuits plays a key role in the encoding of movement, associative learning and procedural memory. How correlated activity is assembled by striatal microcircuits is not understood. Calcium imaging of striatal neuronal populations, with single-cell resolution, reveals sporadic and asynchronous activity under control conditions. However, N-methyl-D-aspartate (NMDA) application induces bistability and correlated activity in striatal neurons. Widespread neurons within the eld of observation present burst ring. Sets of neurons exhibit episodes of recurrent and synchronized bursting. Dimensionality reduction of network dynamics reveals functional states dened by cell assemblies that alternate their activity and display spatiotemporal pattern generation. Recurrent synchronous activity travels from one cell assembly to the other often returning to the original assembly; suggesting a robust structure. An initial search into the factors that sustain correlated activity of neuronal assemblies showed a critical dependence on both intrinsic and synaptic mechanisms: blockage of fast glutamatergic transmission annihilates all correlated ring, whereas blockage of GABAergic transmission locked the network into a single dominant state that eliminates assembly diversity. Reduction of L-type Ca2-current restrains synchronization. Each cell assembly comprised different cells, but a small set of neurons was shared by different assemblies. A great proportion of the shared neurons was local interneurons with pacemaking properties. The network dynamics set into action by NMDA in the striatal network may reveal important properties of striatal microcircuits under normal and pathological conditions.
INTRODUCTION

programs (Barnes et al. 2005; Graybiel 1995; Grillner et al. 2005a,b; Takakusaki et al. 2004a). A major component of the BG is the striatum, which receives a widespread input from the cerebral cortex and thalamus. Striatal circuits process corticothalamic inputs to produce specic outputs consisting of bursts of action potentials during the execution of motor tasks, probably following voltage transitions to depolarized up-states (Hikosaka et al. 2000; Kasanetz et al. 2006; Mahon et al. 2006; Romo et al. 1992; Schultz et al. 1993; Wilson 1993). As in other isolated nervous tissue preparations known to contain CPGs (e.g., Guertin and Hounsgaard 1998), addition of NMDA to neostriatal circuits in vitro (Vergara et al. 2003) and in vivo (Herrling et al. 1983) induces recurrent bursting and pattern generation in single neurons. Moreover, intrastriatal application of NMDA in vivo generates turning behavior when administered unilaterally in freely moving animals (Ossowska and Wolfarth 1995); demonstrating that motor behaviors arise from the striatal processing of enhanced excitatory drives. This evidence suggests that the striatum posses the connectivity and intrinsic mechanisms to orchestrate pattern generation (Grillner 2006). To test this hypothesis, we used calcium imaging of neuronal populations in a corticostriatal slice preparation to monitor, with single-cell resolution, dozens of cells simultaneously and then discern pattern generation produced at the microcircuit level by the activity of cell assemblies. Simultaneous electrophysiological recordings from striatal neurons demonstrated that calcium transients result from neurons bursting on top of suprathreshold up-states (Kerr and Plenz 2002).
METHODS

Downloaded from jn.physiology.org on April 29, 2011

A central pattern generator (CPG) produces specic activity patterns in the absence of sensory inputs (Grillner et al. 2005b) and can transform afferent inputs into detailed spatiotemporal outputs (Grillner 2006; Yuste et al. 2005). In vitro experiments with circuits containing CPGs demonstrate that tonic excitation produced by the glutamate agonist N-methyl-D-aspartate (NMDA) can activate the stereotyped electrical behavior that neuronal networks exhibit in more intact preparations (such as ctive locomotion). This is observed as recurrent bursting activity in single neurons, while simultaneous unitary and population recordings demonstrate synchronicity and alternation of cell assemblies activity during bursting (e.g., Gordon and Whelan 2006; Grillner et al. 1981; Guertin and Hounsgaard 1998; Hsiao et al. 1998; Kiehn 2006; Takakusaki et al. 2004b). Basal ganglia (BG) contain CPGs that activate innate behavioral routines, procedural memories, and learned motor
Address for reprint requests and other correspondence: J. Bargas, Instituto de Fisiolog a Celular UNAM, PO Box 70-253, Mexico City, DF 04510 Mexico (E-mail: jbargas@ifc.unam.mx). www.jn.org

Slice preparation
Transverse corticostriatal slices (300 m thickness) were obtained from PD14-29 Wistar rats as previously described (Kawaguchi et al. 1989; Vergara et al. 2003). All procedures conformed to the guidelines of the UNAMs Animals Scientic Procedures Committee. Slices were obtained with ice-cold saline (4C) containing in mM: 123 NaCl, 3.5 KCl, 1 MgCl2, 1 CaCl2, 26 NaHCO3, and 11 glucose (25C; saturated with 95% O2-5% CO2; pH 7.4; 298 mosM/l). Slices were then transferred to saline at room temperature (2125C) where they remained for 1 h before recording. The cationic concentration of this saline favors the appearance of up states in vitro (Vergara et al. 2003). Although the present results were performed mainly in young animals (PD14-21), basically the same network behavior was observed in older animals (PD25-29) (J. Bargas, unpublished data and see RESULTS).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1435

0022-3077/08 $8.00 Copyright 2008 The American Physiological Society

1436

CARRILLO-REID ET AL.

Calcium imaging
Slices were incubated at room temperature in the dark for 20 30 min in the presence of 10 20 M uo 4-AM (Tef Labs, Austin, TX) in 0.1% dimethylsulphoxide (35C), equilibrated with 95% O25% CO2. Slices were perfused with control saline (see preceding text) in a perfusion chamber on the stage of an upright microscope equipped with a 10 water-immersion objective (Eclipse E600FN; Nikon, Melville, NY). Excitation at 488 nm was performed with a Lambda LS illuminator (Sutter instruments, Novato CA). Experiments were performed at room temperature. Images were acquired with a cooled digital camera (SenSys 1401E, Roper Scientic, Tucson, AZ) at 250 500 ms/frame. Imaging software used was RS Image (Photometrics; Roper Scientic). The imaged eld was 800 600 m in size. Short movies (100 250 s, 50to 100-ms exposure, 2 4 image/s) were taken at time intervals of 520 min during 1 h. The number of uo 4-loaded neurons in the eld was determined at the end of the experiment with a 5-s puff of 50 mM KCl. This maneuver disclosed all uo-4-labeled neurons (either active or silent during the experiment). Cells active during the experiment were analyzed, and the ratio of active/silent cells was obtained. Spontaneous or evoked calcium transients together with voltage responses were recorded electrophysiologically in some cells, both in control saline and during the application of 512 M NMDA (Sigma-Aldrich-RBI, St. Louis, MO). In some experiments, cortical sensory motor areas were stimulated with a concentric bipolar electrode (12 m; FHC, Bowdoinham, ME). Stimuli consisted of different trains of 510 stimuli at 20 Hz. Each stimulus was 100 200 s and 50 120 A. In some experiments, we used the minimal stimulus intensity necessary to evoke peaks of synchrony with amplitudes above chance (P 0.05). This allowed us to study changes of electrical evoked synchrony under different pharmacological conditions.

An Axoclamp 2B amplier (Axon Instruments, Foster City, CA) was used to perform whole cell current- and voltage-clamp recordings. Signals were ltered at 13 kHz and digitized at 39 kHz with an AT-MIO16E4 board (National Instruments, Austin, TX) in a PC computer. Data acquisition used a software designed in the LabView environment (Lemus-Aguilar et al. 2006). Patch pipettes (3 6 M) were lled with (in mM) 115 KH2PO4, 2 MgCl2, 10 HEPES, 0.5 EGTA, 0.2 Na2ATP, and 0.2 Na3GTP. In some experiments, biocytin 0.5%, and uo-4 salt (20 30 M) were added to the recording pipettes.

Image analysis
Image processing was carried out with Image J (v.1.36, National Institutes of Health), Multicell 2.0 (kindly supplied by Robert Froemke), and custom-made programs written in IDL (Cossart et al. 2003; Mao et al. 2001; Schwartz et al. 1998) or MATLAB (The Math-Works, Natick, MA). All active neurons in a eld were semi-automatically identied, and their mean uorescence was measured as a function of time. Singlepixel noise was discarded using a 5-pixel ratio mean lter. Calciumdependent uorescence signals were computed as (Fi Fo)/Fo, where Fi: uorescence intensity at any frame and Fo: resting uorescence, i.e., average uorescence of the rst four frames of the movie. Calcium signals elicited by action potentials were detected based on a threshold value given by their rst time derivative (2.5 times the SD of the noise value). Thus we obtained a C F binary matrix; were C represents the number of active cells and F the number of frames for each movie. Spike onsets were signaled by ones in the matrix representing transitions to the up states. Recordings were inspected manually to remove artifacts and slow calcium transients which are likely to correspond to glial cells (Ikegaya et al. 2005; Sasaki et al. 2007).

Downloaded from jn.physiology.org on April 29, 2011

Statistical methods
To determine if calcium transients recorded from different cells were correlated, the numbers of simultaneous activations per trial (onset of signals occurring within 3 frame windows) were detected. To determine the P value of simultaneous transients occurring by chance, the distribution under the null hypothesis of independent transients using Monte Carlo simulations with 1,000 replications were computed (Mao et al. 2001). The degree of correlation between active cells was calculated with the Jaccard correlation coefcient. Nevertheless the magnitude of the correlations is difcult to discern when many lines are superimposed. Thus we constructed cross-correlation maps of Jaccard correlation coefcients to show the magnitude of the correlations between all cells pairs. In addition, we identied the sets of cells activated simultaneously over time. To identify peaks of synchronous activity (i.e., that included more cells than those expected by chance), Monte Carlo simulations were also used to estimate the signicance of their ring together. The threshold corresponded to a signicance level of P 0.05. We then examined peaks that were signicant at P 0.05. Peaks of synchronous and recurrent bursting activity that remained signicant during the experiment were selected for further analysis. To analyze the dynamics of cell assemblies over time (network dynamics), D N matrices were constructed, where D represents the number of active neurons in a set of experiments, and N denotes the ring of cells during 250-ms to 1-s time bins (NMDA-induced up states last between 0.5 and 5 s) (Vergara et al. 2003). Peaks of recurrent synchronous activity were vectorized so that bursting over time was associated with different neurons. Each vector element is formed by the sum of the number of calcium spikes displayed by a single neuron during the time bin, where peak derivatives denote the time onset of electrophysiologically recorded up states (see preceding text and RESULTS). Therefore the set of these vectors denote network activity as a function of time (Brown et al. 2005; Sasaki et al. 2007).
www.jn.org

Immunohistochemistry
Sections were processed to x uo-4 uorescence or cells active during experimentswith N-(3-dimethylaminopropyl)-N-ethylcarbodiimide hydrochloride (EDAC), and to perform conventional immunocytochemistry in uorescent cells to demonstrate either substance P (SP) or enkephalin (ENK) on uo-4-labeled cells using commercially available antisera (Peninsula Labs, San Carlos, CA) conjugated to CY3 or CY5. Slices were not processed for both antisera, but one was chosen in each case. Thus in each trial, either SP or ENK positive and negative neurons could be observed. Briey, sections were rinsed in PBS and incubated for 18 24 h at 4C with primary rabbit antibody against ENK or SP (diluted 1:200). Sections were mounted in an anti-quenching media (Vectashield, Vector Laboratories) and examined under a confocal microscope (MRC1024; Bio-Rad, Natford, UK) equipped with a kryptonargon mixed-gas laser. Immunostained cells were studied in either single confocal images or reconstructed sections made by projecting z-series of 10 40 consecutive confocal images 1 m apart collected throughout the thickness of the section. The background noise was reduced averaging three to six images. Digitized images were transferred to a personal computer (Confocal Assistant, T. C. Brelje). More than 80% of uo-4-loaded cells were medium spiny neurons.

Drugs
Stock solutions were prepared before each experiment and added to the perfusion solution in the nal concentration indicated. NMDA, APV, nicardipine, CNQX, biocytin, and bicuculline methiodide or hydrochloride were obtained from Sigma (St. Louis, MO).

Electrophysiology
Calcium imaging and simultaneous electrophysiological recordings were obtained from areas of the dorsal striatum previously shown as receiving numerous cortical bers (Vergara et al. 2003).
J Neurophysiol VOL

99 MARCH 2008

STRIATAL NETWORK DYNAMICS

1437

To measure the similarity index between the network vectors, we computed the normalized dot product of all possible vector pairs, which is equivalent to the cosine of the angle between the vectors (Sasaki et al. 2006; Schreiber et al. 2003). Then we plotted the similarity indexes as a pseudocolor matrix in which functional states sustained by signicantly correlated or synchronized bursting cell assemblies appear as cluster-like structures (Sasaki et al. 2006). To reduce the dimensionality of the network vectors, locally linear embedding (LLE) was chosen. LLE is an unsupervised learning algorithm that discloses nonlinear structures from multi-dimensional data (Roweis and Saul 2000). Multidimensional scaling (MDS) (Systat, Richmond, CA) gave results quantitatively similar from those obtained with LLE. Nevertheless LLE depicted more effectively the trajectories of the functional states in the network. The trajectories of our high-dimensional data were not well described by linear dimensionality reduction methods (see Brown et al. 2005). To choose the optimal number of states depicted from the LLE reduction, we used hard and fuzzy clustering algorithms taking the Dunns index as a validity function (Bezdek et al. 1997; Sasaki et al. 2007). To determine the number of neurons in each state, hierarchical cluster analysis was computed using Euclidean distances and the nearest neighbor single linkage method (Systat Software, San Jose, CA).

RESULTS

Optical imaging from populations of striatal neurons To study striatal microcircuits in vitro, we used Ca2 imaging (Fig. 1) to measure electrical activity in many cells simultaneously with single-cell resolution. Activity was measured indirectly, as changes in uorescence. Fields from the dorsal striatum were imaged in slices loaded with the Ca2 indicator uo-4 AM in 174 experiments performed in 86 corticostriatal slices. Figure 1A shows all dye-loaded neurons (see METHODS). Contours of cells both active and inactive during the experiment are depicted in Fig. 1B (lled circles and empty contours, respectively). Only neurons active during an experiment were analyzed. Under control conditions (with no drugs added), only a few cells were active, and their ring was asynchronous (lled circles in Fig. 1B). Records of Ca2 transients from three spontaneously active neurons are shown (Fig. 1C). These experiments conrm that the striatal circuitry is mostly quiescent (n 40 slices) under control conditions. Simultaneous voltage recordings and Ca2 imaging in medium spiny neurons (Fig. 1, D and E) demonstrate the corre-

Downloaded from jn.physiology.org on April 29, 2011

FIG. 1. Optical recording in striatal neurons. A: neurons in a striatal slice loaded with uo-4 AM. Picture is the result of averaging 200 consecutive frames and background subtraction (see METHODS). Scale bar: 100 m. B: automatic contour detection of 376 cells from A. Dark circles indicate neurons exhibiting spontaneous calcium transients under control conditions with no drugs added (14/376 or 3.7%). Under control conditions, most striatal cells loaded with uo-4 remain silent. C: recurrent calcium transients recorded from 3 of the active cells shown in B (280 ms per frame). D: a uo-4-loaded cell targeted for electrophysiological recording (1). Fluo-4 salt was also administered through the recording pipette (bottom). Scale bar: 10 m. Voltage responses (top) to current steps (bottom) recorded from the neuron shown in D1 (2). Inward rectication and long latency to 1st spike are characteristics of medium spiny neurons. Steady-state current-voltage relationship measured in current-clamp mode from traces shown in D2 (3). Most calcium transients and electrophysiological recordings shown in the next gures are from this class of neurons. E: simultaneous recordings of voltage transitions (1) and calcium transients (2) from the cell shown in D1 induced by the presence of N-methyl-D-aspartate (NMDA) in the bath. The duration of rst derivatives of the calcium transients [dashed line indicates 2.5 times the SD of the noise; 3; peaks in d(F/F)/dt gray stripes] match the duration of electrophysiological up states. Dots indicate events where d(F/F)/dt 2.5 times SD; used to build raster plots (see following text). Histogram showing a bimodal distribution of membrane potential (4); taken from electrophysiological recordings in 1. Current-voltage relationship measured in voltage-clamp conguration in the presence of NMDA (5). Note 3 crossing points in the voltage axis and a negative slope conductance region.

J Neurophysiol VOL

99 MARCH 2008

www.jn.org

1438

CARRILLO-REID ET AL.

spondence of suprathreshold activity and somatic calcium transients (n 15 cells). As it has been observed in many other central circuits (e.g., Gordon and Whelan 2006; Grillner et al. 1981; Guertin and Hounsgaard 1998; Hsiao et al. 1998; Kiehn 2006; Takakusaki et al. 2004b), we were able to generate bistability in striatal neurons after bath application of NMDA (Vergara et al. 2003), a transmitter known to induce motor behavior when administered in the striatum (Fig. 1E) (Ossowska and Wolfarth 1995). This treatment induces persistent bursting behavior in neostriatal neurons (1 h) without the need of electrical stimulation (spontaneous). Most active cells during a given experiment had the electrophysiological characteristics of medium spiny neurons (Fig. 1D). Ca2 transients (F/F) corresponding to up states, had time derivatives [d(F/F)/dt] that matched bursts duration (Cossart et al. 2003; Kerr and Plenz 2002). Clearly membrane potential distribution is bimodal in bursting cells (Fig. 1E4). The current-voltage relationship (I-V plot) measured in voltage clamp at the end of 400to 500-ms commands shows a negative slope conductance region (NSCR), indicating bistability, in NMDA-treated cells (Fig. 1E5) (e.g., Hsiao et al. 1998; Izhikevich 2007; Vergara et al. 2003). Ca2 imaging and simultaneous electrophysiological recordings also show that only bursts with two or more action potentials produce detectable Ca2 transients in striatal neurons (n 10 neurons; Fig. 2). These experiments conrmed that the striatal neurons can be activated in vitro by NMDA bath application. Therefore the next step was to ask if this activity was correlated and synchronous throughout the network, in which case, it may correspond to a network dynamics capable of producing pattern generation. Cortical stimulation synchronizes widespread striatal neurons In corticostriatal slices, electrical stimulation of the cortex evokes long-lasting depolarizations with overriding spikes in medium spiny neurons (Bargas et al. 1991; Vergara et al. 2003). Trains of cortical stimuli (Fig. 3A) also result in pro20 mV

longed synaptic depolarizations with action potentials (Fig. 3B). Figure 3B illustrates an experiment where whole cell current-clamp recordings (Fig. 3B1) and calcium imaging (Fig. 3B2) were performed simultaneously from a medium spiny neuron (stimulus frequency: 0.1 Hz; stimuli are signaled with arrows at the bottom). A calcium transient accompanies each orthodromic response to the cortical stimulus (Fig. 3B2, arrows at the bottom). The time derivatives of the calcium responses are shown in the third row (Fig. 3B3); note similar duration of derivative positive peaks and voltage responses. Accordingly, each peak from the differentiated calcium signal generates a dot used to build raster plots as that illustrated in Fig. 3D. Each row in the raster plot represents an active neuron (lled circles in Fig. 3C). Filled circles in Fig. 3C indicate uo-4-loaded neurons within the eld of observation that responded to cortical stimulation in one representative experiment. Empty circles indicate loaded neurons that did not respond to cortical stimulation (see METHODS). Responsive cells are a minority of the total number of neurons which are scattered throughout the observational eld among many unresponsive neurons (Fig. 3C). Cortical stimulation activated and synchronized 10 20% of uo-4-loaded cells. Only a few neurons were active before and after the stimulus train, and these cells present a lack of correlation (Fig. 3, D and E). At each trial, 90% of the responsive neurons followed the cortical stimulus (Fig. 3D, bottom). Some neurons spontaneously active before the stimulus did not follow the electrical stimulation (blue lines). Targeted electrophysiological recordings of cells responsive to cortical stimulation revealed that most active neurons were medium spiny neurons (Fig. 3B). Network dynamics set into action by NMDA Ca2 imaging allowed us to observe a population of striatal cells activated by cortical stimulation in control conditions (Fig. 3) and in the bath presence of NMDA (512 M; Fig. 4). Simultaneous voltage recordings of imaged neurons (Fig. 4A)

Downloaded from jn.physiology.org on April 29, 2011

A
-75 mV

200 ms
FIG. 2. Action potentials necessary to produce detectable calcium transients. A: example of action potentials evoked by brief intracellular depolarizing current steps (not shown) in a medium spiny neuron. Arrows point to the same records shown below in a slower time base. B: actions potentials were evoked repetitively (as in Fig. 1A) in successive series (delimited by dashed boxes) of 1 4 action potentials. C: simultaneous recording of calcium transients accompanying the action potentials shown in B. Notice that signicant calcium transients follow voltage responses with 2 action potentials. D: 1st derivative of calcium records shown in C. Detection threshold (dashed line) was set at 2.5 SD of the noise. Threshold is reached when stimulus evokes 2 action potentials.

5% F/F 15 s

C
10%/s d(F/F)/dt

J Neurophysiol VOL

99 MARCH 2008

www.jn.org

STRIATAL NETWORK DYNAMICS

1439

Downloaded from jn.physiology.org on April 29, 2011

FIG. 3. Widespread distribution of striatal neurons synchronized by cortical stimulation. A: scheme of the experimental arrangement: eld stimulation was delivered to corticostriatal afferents (electrode in cortex) while neurons from a dorsal striatum area (square) were recorded with simultaneous calcium imaging and whole cell, patch-clamp recording techniques. B: synaptic response from a medium spiny neuron to a train of synaptic potentials (10 eld stimuli at 20 Hz; arrows indicate train stimuli); note ring of action potentials on top of synaptic response. A sequence of synaptic responses such as that illustrated at left (1). Calcium transients recorded from the spiny neuron (280 ms/frame) responding to cortical stimulation in 1 (2). Note that each synaptic response has a corresponding calcium transient. First derivative obtained from calcium recordings (3). Dots indicate the onset of the responses used to build the raster plots shown in D. C: mapping of all neurons present in the eld of view (circles). Cells triggered by cortical stimulation: 33/252 (13%) are indicated with red lled circles. Cells unresponsive to cortical stimulus but excitable, as shown after high K application, are indicated by empty circles in this and other gures. Scale bar: 50 m. D: population raster plot: each row represents an active neuron. Some cells present spontaneous activity before and after the stimulus. Note synchronized responses from neurons several hundreds of microns apart during cortical stimulation (red). Histogram at the bottom represents the percentage of responding cells triggered by each cortical stimulus. E: cross-correlation map including all active cells (P 0.05; see METHODS). One spontaneously active cell never responded to cortical stimulation (navy blue lines).

show that voltage transitions (Vergara et al. 2003) were accompanied by Ca2 transients during NMDA activation (Kerr and Plenz 2002, 2004). Cell activity in the circuit was followed by recording their Ca2 transients and plotting them as raster plots (Fig. 4B). Similar NMDA-induced electrical activity has been observed in neurons that are part of a CPG (e.g., Grillner et al. 1981; Guertin and Hounsgaard 1998; Hsiao et al. 1998; Takakusaki et al. 2004b). Interestingly, the raster plot (Fig. 4B, top) shows that several neurons are involved in the activity. Moreover, the time histogram (Fig. 4B, bottom; asterisks, red lines) clearly shows periods of increased synchrony and correlated ring that occur spontaneously in different sets of neurons (Cossart et al. 2003; Ikegaya et al. 2004). Statistically signicant (threshold P 0.05) spontaneous peaks of synchrony occurred in sets of neurons from n 38/45 slices in these conditions. On average, there were 2.5 0.2 peaks of synchronous bursting per 201 10-s time epoch. Peaks of recurrent synchronous activity showed no apparent periodicity, and occurred with a mean interval of 46 10 s. Notably, neurons ring synchronously during NMDA treatment could
J Neurophysiol VOL

be hundreds of microns apart intermingled with silent cells (Fig. 4C, lled circles). Spatial correlation maps (Fig. 4D) show the pairs of neurons that exhibit statistically signicant (P 0.05) correlated activity (line thickness is proportional to the degree of correlation). On average, 77 4% of active cells had correlated activity (n 38 slices) at any given moment. Cells active during synchrony peaks represent most of the correlations between cells. Correlation plots of this activity (Fig. 4E; P 0.05) show that the degree of synchronization among active neurons is heterogeneous, thus the network activity is not due to chance correlation between any pair of neurons but to the spatiotemporal dynamics of several cells. Furthermore, application of NMDA to neostriatal circuits in vitro (Vergara et al. 2003) and in vivo (Herrling et al. 1983) induces bursting activity and generates turning behavior in freely moving animals (Ossowska and Wolfarth 1995), suggesting a general mechanism preserved in a broad range of ages. Structured network dynamics with the same characteristics have been shown in the cortex of young mice in vitro (PD13-22) and adult cats in vivo, demonstrating that network
www.jn.org

99 MARCH 2008

1440

CARRILLO-REID ET AL.

A
-75 mV

B
50

Active cell # % coactive cells

40 30 20 10

80 mV

5 sec

10 5

**

P<0.05

20% F/F

50 s
Downloaded from jn.physiology.org on April 29, 2011

D
41 47 45 43 44 42 40 39 38 37 36 35 34 33 32 31 29 28 30 27 26 25 24 23 22 19 20 21 18 1716 14 15 13 12 10 11 9 8 7 6 5 4 3 12 46

E
40 0.75 jaccard correlation

neuron i

30 20 10 10 20 30 40 0

neuron i

FIG. 4. NMDA-induced network activity. A: simultaneous electrophysiological and calcium imaging recordings from a spiny neuron in the presence of 8 M NMDA. Cortical stimulus (arrows) produces plateau potentials with detectable calcium transients. Recurrent membrane potential transitions, with bursting and corresponding calcium transients, are visible several minutes (up to 1h) after interrupting cortical stimulation. B: raster plot of NMDA-induced activity (no cortical stimulation). Each row represents an active cell. Activity can be followed in several neurons simultaneously with calcium imaging. Histogram at the bottom shows the percentage of co-active neurons as a function of time. Peaks of spontaneous synchrony (red), indicated with asterisks, show that recurrent bursting is shared by sets of neurons in the circuit, suggesting the emergence of a spatiotemporal pattern (250 ms/frame). C: mapping of neurons in the eld of observation that exhibited recurrent bursting (lled red and black circles), after NMDA administration (without cortical stimulation; 47/291 cells; 16%). Scale bar: 100 m. Red circles indicate cells active during peaks of synchronous activity. D: spatiotemporal correlation map of NMDA-induced activity. Lines connect neurons the ring of which was signicantly correlated (P 0.05): n 43/47 (91%). Red circles indicate active neurons involved in the synchrony peaks. Note the widespread distribution of cells ring together. E: cross-correlation map of all possible pairs of neurons. Note the heterogeneous distribution of the correlation coefcients. (For better visualization, cross-correlation of the same cells was removed from the map; black line).

activity, intrinsic to specic nuclei, is preserved in slices (Ikegaya et al. 2004). But to address this issue in the striatal microcircuit, we divided our data into age groups to discern possible maturational variables. Figure 5 shows the age independency of NMDA-induced network dynamics. In spite of the previously described reduction in the number of loaded cells as a function of age (PD14-29; Fig. 5, AC) (Froemke et al. 2002; Peterlin et al. 2000), the number of peaks of synchronous bursting per time epoch was maintained across PD14-29 (Fig. 5D), conrming a robust mechanism, across this age range, for the network behavior described here (Ikegaya et al. 2004). These experiments demonstrated the following. 1) NMDA treatment induces pattern generation in the neostriatum in vitro. 2) Activity is distributed throughout the circuit, and it is shared by a signicantly higher percentage of neurons than those active under control conditions. 3) Activity may be synchronous and correlated ring is seen in sets of neurons: 4) these phenomena continue for an extended period of time without electrical
J Neurophysiol VOL

stimulation once the network becomes active. 5) The network dynamics observed occur in a wide range of ages (PD14-29). Accordingly, we hypothesized that the network dynamics set into action by NMDA should reveal sets of related neurons (cell assemblies) that alternate their activity to generate spatiotemporal patterns of synchronization. Visualizing functional states during network dynamics We next investigated network dynamics in the striatum over long periods of time (n 15 slices). Brief movies (100- to 250-s epochs) were taken at different intervals for up to an hour (e.g., Fig. 6C, 15, separated by lines). To analyze network dynamics, we binned the raster plots of all the neurons involved in the peaks of synchrony (Fig. 6). Time segments represent vectors coding coactive burst activity of individual neurons (see METHODS). Network dynamics consisted of a set of N vectors (bins) in D dimensions (active neurons). Vectorizawww.jn.org

99 MARCH 2008

STRIATAL NETWORK DYNAMICS

1441

PD 14

PD 29

C
400

D
5 350 300 250 200 14 18 22 26 30 14 18 22 26 30

Loaded cells

# Peaks/epoch

4 3 2 1

Downloaded from jn.physiology.org on April 29, 2011

Postnatal Day
FIG.

Postnatal Day

5. Age independency of network dynamics. Neurons in a striatal slice loaded with uo-4 AM, PD 14 (A) and PD 29 (B). Pictures are the result of averaging 720 consecutive frames (background is not subtracted). Scale bar: 100 m. C: number of loaded cells as a function of postnatal day. Each point represents one brain slice. The line represents an exponential t. D: number of peaks per 201 10 s time epoch as a function of postnatal day. F, 1 brain slice. , best linear t. Note that although the number of loaded cells decreases with age, the network dynamics remains constant.

tion allowed us to compare different states of network activity rigorously over time (Brown et al. 2005; Sasaki et al. 2007). The normalized inner product (see METHODS) of all possible vector pairs provides a measure of the similarity among states (Sasaki et al. 2006; Schreiber et al. 2003). To evaluate this similarity along network activity, we plotted all vector pairs as a matrix (Fig. 6A). Notably, abrupt transitions in the similarity index showed the presence of cluster-like structures (Sasaki et al. 2006). To compare network responses induced either by cortical stimulation or by NMDA bath application, we reduced the dimensionality of the vectors using LLE (see METHODS), a technique for nonlinear dimensionality reduction (Brown et al. 2005; Roweis and Saul 2000; Stopfer et al. 2003). The new vectors were projected in two dimensions (Fig. 6B). Clusters of points formed trajectories representing the responses over time to specic stimuli. Trajectories illustrate different subgroups of neurons coactive within each functional state (Brown et al. 2005; Stopfer et al. 2003). Thus changes in the functional state of the network can easily be followed. In Fig. 6B, state 1 represents network response to cortical stimuli in control conditions. State 2 depicts NMDA-induced network activity without electrical stimulation. State 3 represents network response to the same cortical stimuli (given in the state 1) in the bath presence of NMDA. Figure 6C, top, shows the raster plot used to reconstruct network states, whereas bottom shows histograms with the percentage of co-active neurons along time. Cortical stimulation in the presence of NMDA recruited many
J Neurophysiol VOL

more striatal neurons (Fig. 6C3, blue dots and peaks) than in control conditions (Fig. 6C1, green dots and peaks) and produced peaks of synchrony much larger than those produced by NMDA only (Fig. 6C, 2, 4, and 5, red dots and peaks). Spontaneous peaks of synchrony induced by bath NMDA were 20% the size of those induced by cortical stimulation (cf. Fig. 6C, 15, bottom). Nonetheless the general behavior of the active network after a cortical stimulus (Fig. 6C, 4 and 5) remained essentially comparable to the activity before the stimulus (Fig. 6C2 and see following text). In fact, time histograms showed similar numbers of spontaneous peaks of synchronous activity (Fig. 6C, bottom; P 0.05) before and after the cortical stimulus. Moreover, cortical stimulation (Fig. 6C3) elicited a trajectory looping back to the NMDA-induced dynamics (Fig. 6B), showing recurrent activation of the same assembly after a perturbation (Stopfer et al. 2003). Figure 6D shows the spatial distribution of the neurons involved in the different states. Hierarchical cluster analysis revealed the existence of different neuronal subsets (assemblies) that underlie network states (Fig. 6F). Each state was basically sustained by a different neuronal assembly. A state transition implies a change of neuronal assembly (alternation). However, some elements are shared by different states (core neurons). Thus using a simple algorithm we could distinguish recursive peaks of synchronization reecting functional states that involve different sets of neurons. We conclude that spatiotemporal ring patterns codied in multidimensional vectors are sufcient to reconstruct the dynamics of a given network.
www.jn.org

99 MARCH 2008

1442

F A
% coactive cells Active cell #
20 40 60 80 100 120 140 0 10 5

state 3

state 2

state 1

vector i (t)
20 40 60 0

state 1
20

vector i (t)

P<0.05
40 60

state 2
* *

Downloaded from jn.physiology.org on April 29, 2011

J Neurophysiol VOL

similarity index

3
B

CARRILLO-REID ET AL.

Active cell #
state 3
* * *

99 MARCH 2008

LLE 2
0 1 2 -3 -2 -2 -1

www.jn.org

state 2

-1

4
0

state 1

LLE 1

% coactive cells
*

10
*

20

state 3

1&2 1&3 2&3

# states 1 min

137 135 126 123 120 115 112 109 104 92 88 82 77 61 43 37 6 1 3 34 40 60 66 80 87 91 102 105 110 114 116 122 125 127 136 70 51 10 4 20 69 134 128 103 98 95 85 73 63 53 49 46 42 39 35 30 24 16 8 11 22 28 32 36 41 44 47 50 58 72 76 94 96 99 106 140 143 129 81 64 52 33 27 48 54 74 84 138 55 5 29 132

STRIATAL NETWORK DYNAMICS

1443

Pattern generation denoted by functional states sustained by cell assemblies The preceding analyzed states included some with imposed synchronization, i.e.: with electrical cortical stimulation. If network dynamics induced by NMDA are mediated by cell assemblies, then one may expect the appearance of distinguishable functional states evoked by the excitatory tonic drive brought about by NMDA only (without electrical stimulation). To address this issue, we investigated the relationships between the peaks of synchronous activity induced only by NMDA (similar results were obtained in n 14/15 slices that presented NMDA-induced peaks of synchrony). To observe alternation between functional states, we investigated the NMDA-induced network dynamics in the striatum over long periods of time (Fig. 7, n 15 slices). Brief image series (100 250 s in duration) were obtained at different intervals for 1 h (Figs. 7C, 1-,3 top, representative image series obtained at different times during the experiment are shown separated by vertical lines). A similarity correlation matrix (similarity index, Fig. 7A) between these vectors clearly disclosed abrupt transitions indicating the presence of statistically signicant cluster-like structures (Sasaki et al. 2006). We projected these vectors into two dimensions (Fig. 7B, colored circles) by further reducing vectors dimensionality with LLE (Brown et al. 2005; Roweis and Saul 2000; Stopfer et al. 2003). This projection allowed us to observe a variety of functional states in the network as clusters that follow a series of trajectories in sequence. We found a different set of neurons (cell assemblies) with correlated synchronous ring for each functional state (Brown et al. 2005; Stopfer et al. 2003) (Fig. 7D; blue, red, green). Therefore the method allows us to follow the evolution in time of the functional states within the network. Experiments similar to that illustrated in Fig. 7B demonstrated the existence of robust nonrandom cell assemblies displaying co-active neurons along time with recurrent and alternating activity (Figs. 7, B and C) (Sasaki et al. 2006). States were continuously revisited with no state having a preference or a signicantly higher probability of recurrence (Fig. 7B; see percentages of trajectories out of each state including recurrences), demonstrating the existence of various semi-stable network attractors. Time histograms with several synchrony peaks (asterisks; Fig. 7C, bottom) show the alternation of activity between different cell assemblies thus demonstrating multistable dynamics. Figure 7D shows the spatial distribution of neurons generating the patterns and producing the different functional states. Network states share some elements but most of the neurons are dissimilar (Fig. 7E). Finally, hierarchical cluster analysis conrmed the recruitment of different cell assemblies during

specic states (Fig. 7F). Nevertheless there is a small core of cells shared by all the network states (8% of neurons involved in the synchrony peaks). We concluded that functional states, denoted by spontaneous peaks of synchrony (Barnes et al. 2005), are generated by the coordinated participation of cell assemblies, as it is the case of unit CPGs (Grillner 2006). The transitions from one state to the other display long-term network dynamics. The return of the same cell assemblies after net traveling through different trajectories shows that the elements of these modules are robustly associated, allowing alternating participation. Also, a core of neurons is shared by all the states. Mechanisms underlying network dynamics for pattern generation For a preliminary investigation into the synaptic and intrinsic requirements for multistable network dynamics, we challenged network activity with both synaptic and intrinsic ion channel antagonists. Fast synaptic inhibition within the striatal circuit (Czubayko and Plenz 2002; Koos et al. 2004; Tepper et al. 2004; Tunstall et al. 2002) was tested with the GABAA receptor antagonist, bicuculline, which was applied once the network became active. Figure 8 shows that on exposure to 10 M bicuculline, the number of state transitions is drastically reduced in the NMDA-treated slices (n 6 slices). The network becomes locked in a preferred state, which is recurrently revisited. Similarity indexes representing network dynamics suggest the absence of diverse cell assemblies (Fig. 8A). Only occasional transitions to another state occurred (Fig. 8B). Paradoxically, the raster plot of network activity showed an increased frequency of recurrent synchrony peaks: from 2.7 0.2 peaks in the NMDA condition to 8.3 2 peaks in the presence of bicuculline (P 0.01). There was also an increase in the number of coactive cells supporting the preferred state (Fig. 8C). Most synchrony peaks, however, correspond to the same dominant state. Neurons involved in network sates are shown in Figs. 8, D and E. Hierarchical cluster analysis did not reveal cluster-like structures in spite of synchronization (Fig. 8F), meaning that all active neurons were embedded into the same state. Cortical stimulation in the presence of bicuculline recruited even more striatal neurons than those obtained in control conditions or in the presence of NMDA alone (data not shown). Thus fast GABAergic transmission is a necessary requirement for the network to exhibit multistable dynamics and different functional states. We then tested the role of fast AMPA glutamatergic transmission in the orchestration of cell assemblies. Note that

Downloaded from jn.physiology.org on April 29, 2011

FIG. 6. Visualizing network states. A: similarity indices of all vectors representing network dynamics as a time function (see METHODS). Note cluster-like structures in the pseudocolored matrix (Sasaki et al. 2006). B: multidimensional reduction of vectors using locally linear embedding (LLE). Each point represents a vector at a given time (see METHODS). Consecutive time points form different trajectories representing specic experimental conditions, each producing a different network state. State 1: cortical stimulation under control conditions. State 2: NMDA-induced network activity. State 3: cortical stimulation in the presence of NMDA in the bath. C: raster plot used to reconstruct the network states (370 ms/frame; top). Rows represent activity of individual cells. Vertical lines delimit time series C1C5 (186 s/image series; C1: t 10 min, C2: t 35 min, C3: t 40 min, C4: t 45 min, C5: t 50 min). Histogram (bottom) represents percentage of coactive cells over time in the same experiment. Peaks of synchrony without electrical stimulation (asterisks) were present both before and after cortical stimuli (perturbation). Cortical stimulation (imposed synchrony or perturbation) apparently did not change the NMDA-induced network dynamics (spontaneous). Note that patterns of active cells change over time. The heights of synchrony peaks during cortical stimulation appear truncated (1 and 3). D: spatial distribution of neurons involved in the different states. Scale bar 100 m. E: percentage of coactive cells during different functional states. Note the participation of very few cells in different states. F: hierarchical cluster analysis of cells participating in the network states (n 92 cells). colored boxes indicate the different states. Note that different cell assemblies sustain each recurrent state (imposed or spontaneous) although some cells are shared by different states.

J Neurophysiol VOL

99 MARCH 2008

www.jn.org

1444

CARRILLO-REID ET AL.

A
30 similarity index

B
vector i (t)
1
1

33% state 1

67%

LLE 2

20

0 -1 -2 50%

state 2 state 3 50% -2 -1 0

33%

10

0 10 20 30

67% 1

vector i (t)

LLE 1

C
90 80 70 60 50 40 30 20 10 0

Active cell #

Downloaded from jn.physiology.org on April 29, 2011

% coactive cells

10

P<0.05 * *

1 min

% coactive cells

state 1

state 2

state 3

E
20

10

0 1&2 1&3 2&3

F
state 1 state 2 state 3

# states

FIG. 7. Network states and striatal cell assemblies. A: similarity indices of all possible vector pairs of representative NMDA-induced activitywithout cortical stimulation. Note cluster-like structure of distribution indices (Sasaki et al. 2006). B: different states depicted by LLE. Note jumps from 1 state to the other, as well as recurrent revisiting of the same state along time. Percentages over trajectories signal probability of leaving a given state. Probability of jumping from 1 state to the other is high. C: raster plot and time histogram of network dynamics. Recurrent peaks of synchronous bursting activity are marked with asterisks. Vertical lines separate different series of images (186 s each; C1: t 35 min, C2: t 45 min, C3: t 50 min). Colors denote different states. Note transitions between states. D: spatial maps of cell assemblies sustaining the different states. Scale bar 100 m. E: percentage of cells coactive in 2 states. Note overlap of some cells active in different states. F: hierarchical clustering of cells involved in the peaks of synchrony revealed the participation of different cell assemblies in different functional states. Note there is also a core of neurons active in all functional states.

88 78 71 68 64 54 27 10 1 5 25 37 58 67 69 75 79 89 91 39 48 76 62 57 52 47 23 19 29 49 55 60 73 80 38 87 32 4 43 93 33 83 81 65 46 34 24 20 3 8 21 31 41 50 72 82 90

Active cell #

core ensemble

J Neurophysiol VOL

99 MARCH 2008

www.jn.org

STRIATAL NETWORK DYNAMICS

1445

A
vector i (t)

14 12 10 8 6 4 2 2 4 6 8 10 12 14

B
similarity index 1
2.0 12%

LLE 2

1.0 0.0 -1.0 100% -2.0 -3.0 -2.0 -1.0

state 1 state 2

88%

0.0

1.0

vector i (t)

LLE 1 + bicuculline

C
80

NMDA Active cell #

Downloaded from jn.physiology.org on April 29, 2011

60 40 20

% coactive cells

20 P<0.05 10

* ** * ** * * * *

1 min % coactive cells

D
state 1 state 2

15 10 5 0 1&2

# states

F
state 1 state 2 53 12 3 1 8 13 80 71 68 65 61 57 55 48 45 40 37 31 29 27 22 19 15 11 9 14 16 20 26 28 30 36 38 43 47 51 56 60 64 66 69 75 81

Active cell #
FIG. 8. Fast GABAA inhibition is necessary for state transitions. A: similarity index matrix of the NMDA-induced dynamics in the bath presence of bicuculline. Note the heterogeneous distribution of the indices showing the absence of cluster-like structures. B: in the bath presence 10 M bicuculline, the LLE algorithm reveals the recurrence of a preferred state with occasionally jumps to another state. C, top: the gure shows the raster plots of 2 different series of images (220 s each): in 8 M NMDA (35 cells; t 25 min; left) and in 10 M bicuculline (56 cells; t 40 min; right), in the continuous presence of NMDA. In this experiment, 8 cells were active in both series of images (440 ms/frame). Rows represent activity of individual cells. The vertical line delimits series of images 1 and 2. Note overall synchronization of NMDA-induced activity in the presence of bicuculline. C: histogram representing percentage of coactive cells over time in the same experiment (bottom). Peaks of synchrony (asterisks) increased in frequency and amplitude after bicuculline application (in the presence of NMDA). D: neurons involved in the 2 states. Scale bar: 100 m. E: percentage of coactive neurons in the 2 states. Note that all the cells of the state 2 also participate in state 1. F: hierarchical cluster analysis of cells active in the synchrony peaks shows only 1 group of active cells.

J Neurophysiol VOL

99 MARCH 2008

www.jn.org

1446

CARRILLO-REID ET AL.

A
Active cell #

1
50 40 30 20 10

B
NMDA + CNQX

1
60

NMDA

+ nicardipine

Active cell #
* * *

40

20

% coactive cells

15 10 5

P<0.05

% coactive cells

10 5

P<0.05

1 min

1 min

NMDA

+ CNQX

NMDA

+ nicardipine

Downloaded from jn.physiology.org on April 29, 2011

28 15 11 15 10 20

22

34 16

FIG. 9. Blockade of AMPA transmission or L-type calcium channels disrupts NMDA-induced network dynamics. A1: raster plot shows activity during the bath presence of 8 M NMDA (left, 36 cells; t 20 min) and 10 M 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX, right, 19 cells; t 35 min, top). In this experiment, 4 cells had activity in both conditions. The gure shows 2 series of images (180 s each; 250 ms/frame), 1 for each condition. Rows represent activity of individual cells. The vertical line delimits series of images 1 and 2. Histogram represents percentage of coactive cells over time in the same experiment (bottom). Peaks of synchrony (asterisks) were abolished after CNQX application (in the presence of NMDA) and overall activity was drastically reduced. A2: spatial maps of cells involved in the peaks of synchronous activity. Scale bar: 100 m. Gray circles represent cells involved in the synchrony peaks that continued active in the presence of CNQX in the bath (n 3 neurons). B1: the raster plot illustrates activity from 2 series of images (220 s each, top). Active cells in the bath presence of 8 M NMDA (left, 41 cells; t 25 min) and after 5 M nicardipine (right, 35 cells; t 40 min). In this experiment, 10 cells were active in both series of images (440 ms/frame). Rows represent activity of individual cells. The vertical line separates series of images 1 and 2. Histogram represents percentage of coactive cells over time in the same experiment (bottom). Peaks of synchrony (asterisks) were abolished after nicardipine application (in the continuous presence of NMDA). B2: spatial maps of active cells in the peaks of synchrony. Gray circles represent cells active in the presence of nicardipine in the bath (n 6 cells). Scale bar: 100 m.

NMDA blockers cannot be used: either the withdrawal of NMDA or the inhibition of NMDA receptors stops multistate dynamics in slices from this quiescent circuit under control conditions. Active networks were challenged with the AMPA receptor antagonist CNQX (10 M) (Fig. 9A). As shown in the raster plot of Fig. 9A1, CNQX drastically reduced the number of active cells (cf., left and right) and disrupted the appearance of synchrony peaks in all NMDA-treated slices (Fig. 9A, bottom; n 6 slices). Interestingly, a few neurons that were active during synchrony peaks maintained spontaneous activity in the absence of fast AMPA transmission (Fig. 9A, 1 and 2, gray circles) (Vergara et al. 2003). Also in the presence of CNQX, cortical stimulation was unable to recruit striatal neurons (data not shown) and all correlated activity was suppressed (Fig. 9A). Intrinsic conductances are known to participate in pattern generation in single cells. In particular, in spiny neurons during activity set into action by a tonic excitatory drive, the activity of voltage-gated L-type calcium channels has been shown to be essential to produce bistability and I-V plots with a negative slope conductance region (see Fig. 1E5) (see also Vergara et al. 2003). L-type calcium channels promote the generation of plateau potentials capable of sustaining repetitive ring that outlasts input duration. We therefore tested the effects of blocking L-type calcium channels with the antagonist nicardipine (5 M). As shown in the raster plot of Fig. 9B1, nicardiJ Neurophysiol VOL

pine only slightly reduces the number of active neurons (n 5 slices; cf., left and right panels; n 41 vs. 35 neurons), it nevertheless reduces the number of peaks with spontaneous synchronous activity. Many cells dening previous functional states continued spontaneously active in the presence of nicardipine (Fig. 9B, 1 and 2, gray circles), but they were rarely synchronized to produce functional states. We conclude that plateau potentials resulting from the activity of voltage-gated L-type calcium channels are required for network synchronization (Yuste et al. 2005) and pattern generation produced by cell assemblies in NMDA-treated slices. Shared core of neurons In most slices presenting multistable network dynamics (n 20/25 slices), we found a core of neurons, which were shared by all states. So we aimed at recording electrophysiologically some of these neurons after their identication as a part of a core. These neurons were 8 2% of all active cells in the network states (n 20 slices). Some of these core neurons exhibited periodic calcium transients during NMDA treatment or CNQX. Bicuculline disrupted the periodic ring of these cells. Cellattached (Fig. 10A1) or whole cell current-clamp recordings (Fig. 10, C1 and D1) were performed on some of these cells, and we found that 40% of the core assembly (n 4/10 cells) showed ring properties characteristic of GABAergic interwww.jn.org

99 MARCH 2008

STRIATAL NETWORK DYNAMICS

1447

Downloaded from jn.physiology.org on April 29, 2011

FIG. 10. GABAergic interneurons are part of the core assembly. A: cell-attached (top) and calcium imaging recordings (bottom) from a neuron with regular bursting activity in the presence of 8 M NMDA that belongs to a core of shared neurons (1). Records show high-frequency regular bursts with their correspondent calcium transients. Interspike interval histogram of intraburst activity. Note a peak value around 11 ms. B: uorescence image of the cell recorded in A. Scale bar 15 m. C: voltage responses (top) to current steps (bottom) recorded from another neuron showing regular (pacemaker) bursting activity (1). The same neuron lled with biocytin, showing that it corresponds to a cell with aspiny varicose dendrites. Scale bar 15 m (2). D: cell in C exhibits regular bursting during NMDA (8 M). Note the periodicity of the up states (1). Histogram (from 1) shows the bimodal distribution over time of membrane potential (2). D3: power spectrum from D1. Note activity with a period of 2.18 s.

neurons (Plenz and Aertsen 1996; Tepper et al. 2004), such as high-frequency bursts (Fig. 10A) and periodic pacemaking activity (Fig. 10D, 13). However, by xing uorescence of cells active during a given experiment (see METHODS), Fig. 11 shows that most active cells in the different assemblies were immunoreactive to either substance P or enkephalines.
DISCUSSION

We demonstrate that an isolated striatal slice has the necessary circuitry to transform a tonic excitatory drive into sequential correlated activity of several cell assemblies that exhibit recurrent, alternating, and synchronized spatiotemporal activJ Neurophysiol VOL

ity patterns. Although most striatal cells are silent under control conditions, either cortical stimulation or bath application of NMDA, induce synchronized burst ring among assemblies of striatal neurons. Cell assemblies dened functional states during a given experiment and alternate their activity as though belonging to unit CPGs (Grillner 2006). Alternation of activity and frequent recurrences of the same states suggest attractor network dynamics. Ca2 imaging of neuronal populations (Mao et al. 2001; Sasaki et al. 2006; Schwartz et al. 1998; Stopfer et al. 2003) and simultaneous whole cell recordings show that calcium transients correspond to burst ring in single cells, the synchronization and recurrence of which generates the acwww.jn.org

99 MARCH 2008

1448

CARRILLO-REID ET AL.

Fluo4
A

ENK
B

Merged
C

Fluo4
D

SP
E

Merged
F
Downloaded from jn.physiology.org on April 29, 2011

FIG. 11. Most active cells in cell assemblies are medium spiny projection neurons. A: confocal image of a eld of view of the dorsal striatum showing cells active during the experiment: loaded with uo-4 AM. Fluorescence with EDAC, Scale bar 10 m. B: confocal image of the same eld showing that several of these neurons were immunoreactive for ENK, a specic marker of spiny neurons. C: superimposition of A and B. D: confocal image of a eld of view with some cells loaded with uo-4 AM in the dorsal striatum. Fluorescence with EDAC, scale bar 10 m. E: confocal image of the same eld showing that several of these neurons were immunoreactive for SP, another specic marker of spiny neurons. F: superimposition of D and E.

tivity patterns. Pattern generation during network states depend on synaptic and intrinsic properties as well as the activity of a core of active neurons (Berke et al. 2004). In summary, this work shows how pattern generation and correlated activity is generated in the striatal microcircuitry in vitro identifying the neuronal elements involved. The connections and modulation of these elements deserve further study because they may reveal basic general properties of striatal microcircuitry and may further suggest in vivo experiments. Voltage transitions in striatal neurons Striatal neurons, both in vivo and in vitro, exhibit spontaneous voltage transitions that sustain burst ring. Transitions occur between a quiescent down state and a recurrent bursting up state (Herrling et al. 1983; Kerr and Plenz 2002, 2004; Mahon et al. 2006; Vergara et al. 2003; Wilson 1993). It is largely unknown how these transitions reect network activity and how they propagate through the network. It is probable that different classes of bursts exist, for example, during different functional states (e.g., sleep vs. movement) (N. Vautrelle, personal communication). Here we took advantage of the tonic drive provided by NMDA to induce the transitions because in the striatum (Ossowska 1995), as in other circuits (Gordon and Whelan 2006; Grillner et al. 1981; Guertin and Hounsgaard 1998; Hsiao et al. 1998; Kiehn 2006; Takakusaki et al. 2004b),
J Neurophysiol VOL

NMDA sets into action patterned activity that generates movement. Accordingly, it was demonstrated that NMDA induces correlated activity that evolves in time. In contrast to control conditions in which the striatum has only a few active cells with no correlated activity, both cortical stimulation and the tonic excitatory drive provided by NMDA exposure (Vergara et al. 2003) recruit neurons spread over a wide area; which, in the case of NMDA, spontaneously express particular spatiotemporal dynamics, indicating that the striatal microcircuit in vitro preserves a set of unit CPGs (Grillner 2006), that is, even if most connections present in more intact preparations are severed, remaining sets of neurons, probably belonging to larger CPGs or modules in vivo, preserve some of the properties of these larger modules. In support of this inference, independent evidence in the cortex suggests that cell assemblies activity is preserved along different spatial scales (Plenz and Thiagarajan 2007). Mechanisms of network dynamics The conditions that generate circuit dynamics reside in both the synaptic and intrinsic properties of striatal neurons. NMDA induces synchronous peaks of neuronal activity emerging from the co-activation of robust and recurrent cell assemblies that alternate in their patterns displaying a sequence of trajectories. These results suggest the presence of robust mechanisms that maintain and stabilize the cells participating in the assemblies.
www.jn.org

99 MARCH 2008

STRIATAL NETWORK DYNAMICS

1449

In particular, AMPA transmission appears to be necessary for turning on network dynamics. In fact, the number of active neurons drastically falls when AMPA transmission is blocked. Because all glutamatergic synapses in the striatum originate from cortical or thalamic afferents, cortico-thalamic afferents are likely to be the source of tonic excitatory drive underlying synchronous activity. NMDA presumably generates an increase in the tonic excitatory drive conveyed by corticothalamic afferents (Grillner 2006; Grillner et al. 1981), and although some neurons maintain spontaneous activity in the presence of CNQX or after dissecting away the cerebral cortex (Vergara et al. 2003)the analysis shows that this behavior is asynchronous so that the emergence of striatal cell assemblies appears to require fast AMPA glutamatergic transmission, presumably originating from cortical and thalamic neurons (Kasanetz et al. 2006; Yuste et al. 2005). The most parsimonious interpretation of these results is that corticothalamic afferents are essential for the appearance of multistate network dynamics. That is, network dynamics are commanded by the cortex and/or the thalamus (Kasanetz et al. 2006; Magill et al. 2001). However, the actual relationship between cortical or thalamic activation and cell assembly activity in the striatum needs further investigation. It is known that activation of NMDA receptors is enough to induce bistability in striatal neurons, amplifying the excitatory drive conveyed by cortical or thalamic inputs (Grillner 2006; Grillner et al. 1981) so that activity (e.g., from the cortex or thalamus) can be relied on. And although cell assembly activity has been recorded in the cortex using NMDA plus dopamine, it has not been recorded after NMDA only (Tseng and ODonnell 2005). Finally, striatal assemblies receive the convergence of widespread cortical areas, and slices cannot contain all these areas (Parthasarathy and Graybiel 1997). However, cortical cell assemblies driving striatal cell assemblies are not excluded and may be possible in several conditions (e.g., with added dopamine receptor agonists). Therefore striatal assemblies may represent the coordinated activity between cortical areas, that once formed by development and learning, only need a cortical trigger or a subset of the original stimuli for activation. In addition, our data show that the striatum does not simply follow cortical commands. The role of the striatal circuit proper in the management of cortical drives is revealed by the result of inhibiting GABAA receptors. Blockade of GABAA inhibition not only increases the peaks of synchronous activity and recruits more cells during synchrony peaks but also shifts the network toward a preferred state that recruits most active neurons. Without inhibition, the striatal circuit gets tied up in a preferred state, from which alternation and selection among different cell assemblies becomes unlikely. Therefore we conclude that inhibitory circuits in the striatum are responsible for transforming cortical commands into a sequential activity of striatal cell assemblies (Ossowska 1995; Wickens and Oorschot 2000). In addition, the blockade of L-type calcium channels disrupts the peaks of synchrony. This result conrms that intrinsic conductances, such as voltage-gated calcium channels, by generating plateau potentials may work as synchronization enablers of neuronal networks (Kiehn 2006; Yuste et al. 2005). Thus NMDA-induced network dynamics in the striatum requires the participation of both synaptic and intrinsic mechanisms.
J Neurophysiol VOL

Finally, electrophysiological recordings of neurons exhibiting Ca2 transients showed that most of the active neurons in cell assemblies were medium spiny projection neurons. This was conrmed by immunocitochemistry (Fig. 11). Nevertheless, a subpopulation of neurons shared by all functional states (see Parthasarathy and Graybiel 1997) exhibited periodic Ca2 transients and the ring properties of pacemaking GABAergic interneurons (e.g., Berke et al. 2004; Tepper et al. 2004). The fact that striatal GABAergic interneurons comprise 25% of striatal cells and yet make up 40% of the neurons active in multiple assemblies suggests that striatal interneurons participate in the orchestration of multistate dynamics (Berke et al. 2004). Activity of pacemaker neurons is a frequent nding in CPGs (Grillner 2006; Yuste et al. 2005). Functional implications and perspectives Recurrent and alternating bursting is characteristic of cell assemblies included in CPGs in vivo and in vitro (Barnes et al. 2005; Grillner 2006; Ikegaya et al. 2004). Synchronized activity of these modules exhibit attractor dynamics (Cossart et al. 2003), providing a unied description for circuits encoding the storage and retrieval of long-term and working memory. Attractors are seen as memory traces retrieved through excitatory tonic driving or partial cues useful for executing motor programs. The persistence of cell assemblies along time is due to recurrent connectivity (Barnes et al. 2005; Tsodyks 2005). The alternation of their activity is under neuromodulatory control (Yuste et al. 2005). Apparently, the properties of these microcircuits are at the interface between small networks and global brain functions (Plenz and Thiagarajan 2007). Their disturbance provokes abnormal processes of synchrony, associated with different disorders such as schizophrenia and Parkinson Disease (Schnitzler and Gross 2005; Uhlhaas and Singer 2006). We show that an isolated circuit set into action, in vitro, can exhibit synchronized states in specic cell assemblies emerging and returning during relatively long periods of time.
ACKNOWLEDGMENTS

Downloaded from jn.physiology.org on April 29, 2011

We thank Drs. Ranulfo Romo, Roman Vidaltamayo, and Nicolas Vautrelle for critically reading the present manuscript. We thank R. Vela zquez for programming a part of the software for analysis, O. Jaidar and A. Herna ndez for some experiments in older animals, and also A. Laville, C. V. Rivera, T. Fiordelisio, and N. Jime nez, for technical support and advice.
GRANTS

This work was supported by grants from a Project Program grant IMPULSA 03 to J. Bargas, A. Hernandez-Cruz, E. Galarraga, and R. Drucker-Colin, by Consejo Nacional de Ciencia y Tecnolog a (Me xico) Grants 42636 to E. Galarraga and 49484 to J. Bargas, and by Direccio n General de Asuntos del Personal Acade mico, Universidad Nacional Auto noma de Me xico Grants IN201607 to J. Bargas and IN201507 to E. Galarraga.
REFERENCES

Bargas J, Galarraga E, Aceves J. Dendritic activity on neostriatal neurons as inferred from somatic intracellular recordings. Brain Res 539: 159 163, 1991. Barnes TD, Kubota Y, Hu D, Jin DZ, Graybiel AM. Activity of striatal neurons reects dynamic encoding and recoding of procedural memories. Nature 437: 1158 1161, 2005. Berke JD, Okatan M, Skurski J, Eichenbaum HB. Oscillatory entrainment of striatal neurons in freely moving rats. Neuron 43: 883 896, 2004. Bezdek JC, Li WQ, Attikiouzel Y, Windham M. A geometric approach to cluster validity for normal mixtures. Soft Comput 1: 166 179, 1997. www.jn.org

99 MARCH 2008

1450

CARRILLO-REID ET AL. Mao BQ, Hamzei-Sichani F, Aronov D, Froemke RC, Yuste R. Dynamics of spontaneous activity in neocortical slices. Neuron 32: 883 898, 2001. Ossowska K. Interaction between striatal excitatory amino acid and gammaaminobutyric acid (GABA) receptors in the turning behaviour of rats. Neurosci Lett 202: 57 60, 1995. Ossowska K, Wolfarth S. Stimulation of glutamate receptors in the intermediate/caudal striatum induces contralateral turning. Eur J Pharmacol 273: 89 97, 1995. Parthasarathy HB, Graybiel AM. Cortically driven immediate-early gene expression reects modular inuence of sensorimotor cortex on identied striatal neurons in the squirrel monkey. J Neurosci 17: 24772491, 1997. Peterlin ZA, Kozloski J, Mao BQ, Tsiola A, Yuste R. Optical probing of neuronal circuits with calcium indicators. Proc Natl Acad Sci USA 97: 3619 3624, 2000. Plenz D, Aertsen A. Neural dynamics in cortex-striatum co-cultures. II. Spatiotemporal characteristics of neuronal activity. Neuroscience 70: 893 924, 1996. Plenz D, Thiagarajan TC. The organizing principles of neuronal avalanches: cell assemblies in the cortex? Trends Neurosci 30: 101110, 2007. Romo R, Scarnati E, Schultz W. Role of primate basal ganglia and frontal cortex in the internal generation of movements. II. Movement-related activity in the anterior striatum. Exp Brain Res 91: 385395, 1992. Roweis S, Saul LK. Nonlinear dimensionality reduction by locally linear embedding. Science 290: 23232326, 2000. Sasaki T, Kimura R, Tsukamoto M, Matsuki N, Ikegaya Y. Integrative spike dynamics of rat CA1 neurons: a multineuronal imaging study. J Physiol 574: 195208, 2006. Sasaki T, Matsuki N, Ikegaya Y. Metastability of active CA3 networks. J Neurosci 27: 517528, 2007. Schnitzler A, Gross J. Normal and pathological oscillatory communication in the brain. Nat Rev Neurosci 6: 285296, 2005. Schreiber S, Fellous JM, Whitmer D, Tiesinga P, Sejnowski TJ. A new correlation-based measure of spike timing reliability. Neurocomputing 52 54: 925931, 2003. Schultz W, Apicella P, Ljungberg T, Romo R, Scarnati E. Reward-related activity in the monkey striatum and substantia nigra. Prog Brain Res 99: 227235, 1993. Schwartz TH, Rabinowitz D, Unni V, Kumar VS, Smetters DK, Tsiola A, Yuste R. Networks of coactive neurons in developing layer 1. Neuron 20: 541552, 1998. Stopfer M, Jayaraman V, Laurent G. Intensity versus identity coding in an olfactory system. Neuron 39: 9911004, 2003. Takakusaki K, Oohinata-Sugimoto J, Saitoh K, Habaguchi T. Role of basal ganglia-brainstem systems in the control of postural muscle tone and locomotion. Prog Brain Res 143: 231237, 2004a. Takakusaki K, Saitoh K, Harada H, Kashiwayanagi M. Role of basal ganglia-brain stem pathways in the control of motor behaviors. Neurosci Res 50: 137151, 2004b. Tepper JM, Koos T, Wilson CJ. GABAergic microcircuits in the neostriatum. Trends Neurosci 27: 662 669, 2004. Tsodyks M. Attractor neural networks and spatial maps in hippocampus. Neuron 48: 168 169, 2005. Tunstall MJ, Oorschot DE, Kean A, Wickens JR. Inhibitory interactions between spiny projection neurons in the rat striatum. J Neurophysiol 88: 12631269, 2002. Uhlhaas PJ, Singer W. Neural synchrony in brain disorders: relevance for cognitive dysfunctions and pathophysiology. Neuron 52: 155168, 2006. Vergara R, Rick C, Hernandez-Lopez S, Laville JA, Guzman JN, Galarraga E, Surmeier DJ, Bargas J. Spontaneous voltage oscillations in striatal projection neurons in a rat corticostriatal slice. J Physiol 553: 169 182, 2003. Tseng KY, ODonnell P. Post-pubertal emergence of prefrontal cortical up states induced by D1-NMDA co-activation. Cereb Cortex 15: 49 57, 2005. Wickens JR, Oorschot DE. Neural dynamics and surround inhibition in the neostriatum: a possible connection. In: Brain Dynamics and the Striatal Complex, edited by Miller R, Wickens JR. Australia: Harwood Acad, 2000, p. 141149. Wilson CJ. The generation of natural ring patterns in neostriatal neurons. Prog Brain Res 99: 277297, 1993. Yuste R, MacLean JN, Smith J, Lansner A. The cortex as a central pattern generator. Nature Reviews 6: 477 483, 2005.

Brown SL, Joseph J, Stopfer M. Encoding a temporally structured stimulus with a temporally structured neural representation. Nat Neurosci 8: 1568 1576, 2005. Cossart R, Aronov D, Yuste R. Attractor dynamics of network UP states in the neocortex. Nature 423: 283288, 2003. Czubayko U, Plenz D. Fast synaptic transmission between striatal spiny projection neurons. Proc Natl Acad Sci USA 99: 15764 15769, 2002. Froemke RC, Kumar VS, Czkwianianc P, Yuste R. Analysis of multineuronal activation patterns from calcium-imaging experiments in brain slices. Trends Cardiovasc Med 12: 24752, 2002. Gordon IT, Whelan PJ. Deciphering the organization and modulation of spinal locomotor central pattern generators. J Exp Biol 209: 20072014, 2006. Graybiel AM. Building action repertoires: memory and learning functions of the basal ganglia. Curr Opin Neurobiol 5: 733741, 1995. Grillner S. Biological pattern generation: the cellular and computational logic of networks in motion. Neuron 52: 751766, 2006. Grillner S, Hellgren J, Menard A, Saitoh K, Wikstrom MA. Mechanisms for selection of basic motor programsroles for the striatum and pallidum. Trends Neurosci 28: 364 370, 2005a. Grillner S, Markram H, De Schutter E, Silberberg G, LeBeau FE. Microcircuits in action-from CPGs to neocortex. Trends Neurosci 28: 525533, 2005b. Grillner S, McClellan A, Sigvardt K, Wallen P, Wilen M. Activation of NMDA-receptors elicits ctive locomotion in lamprey spinal cord in vitro. Acta Physiol Scand 113: 549 551, 1981. Guertin PA, Hounsgaard J. NMDA-induced intrinsic voltage oscillations depend on L-type calcium channels in spinal motoneurons of adult turtles. J Neurophysiol 80: 3380 3382, 1998. Herrling PL, Morris R, Salt TE. Effects of excitatory amino acids and their antagonists on membrane and action potentials of cat caudate neurons. J Physiol 339: 207222, 1983. Hikosaka O, Takikawa Y, Kawagoe R. Role of the basal ganglia in the control of purposive saccadic eye movements. Physiol Revs 80: 953978, 2000. Hsiao C, del Negro CA, Trueblood PR, Chandler SH. Ionic basis for serotonin-induced bistable membrane properties in guinea pig trigeminal motoneurons. J Neurophysiol 79: 28472856, 1998. Hutchinson WD, Dostrovsky JO, Walters JR, Courtemanche R, Boraud T, Goldberg J, Brown P. Neuronal oscillations in the basal ganglia and movement disorders: evidence from whole animal and human recordings. J Neurosci 24: 9240 9243, 2004. Ikegaya Y, Le Bon-Jego M, Yuste R. Large-scale imaging of cortical network activity with calcium indicators. Neurosci Res 52: 132138, 2005. Ikegaya Y, Aaron G, Cossart R, Aronov D, Lampl I, Ferster D, Yuste R. Synre chains and cortical songs: temporal modules of cortical activity. Science 304: 559 564, 2004. Izhikevich EM. Dynamical Systems In Neuroscience. Cambridge, MA: MIT Press, 2007. Kasanetz F, Riquelme LA, ODonnell P, Murer MG. Turning off cortical ensembles stops striatal up states and elicits phase perturbations in cortical and striatal slow oscillations in rat in vivo. J Physiol 577: 97113, 2006. Kawaguchi Y, Wilson CJ, Emson PC. Intracellular recording of identied neostriatal patch and matrix spiny cells in a slice preparation preserving cortical inputs. J Neurophysiol 62: 10521068, 1989. Kerr JN, Plenz D. Dendritic calcium encodes striatal neuron output during up-states. J Neurosci 22: 1499 1512, 2002. Kerr JN, Plenz D. Action potential timing determines dendritic calcium during striatal up-states. J Neurosci 24: 877 885, 2004. Kiehn O. Locomotor circuits in the mammalian spinal cord. Annu Rev Neurosci 29: 279 306, 2006. Koos T, Tepper JM, Wilson CJ. Comparison of IPSCs evoked by spiny and fast-spiking neurons in the neostriatum. J Neurosci 24: 7916 7922, 2004. Lemus-Aguilar I, Bargas J, Tecuapetla F, Galarraga E, Carrillo-Reid L. Disen o modular de instrumentacio n virtual para la manipulacio n y el ana lisis de sen ales eletrotisiolo gicas. Rev Mex Ing Biomed 27: 8292, 2006. Magill PJ, Bolam JP, Bevan MD. Dopamine regulates the impact of the cerebral cortex on the subthalamic nucleus-globus pallidus network. Neuroscience 106: 313330, 2001. Mahon S, Vautrelle N, Pezard L, Slaght SJ, Deniau JM, Chouvet G, Charpier S. Distinct patterns of striatal medium spiny neuron activity during the natural sleep-wake cycle. J Neurosci 26: 1258712595, 2006.

Downloaded from jn.physiology.org on April 29, 2011

J Neurophysiol VOL

99 MARCH 2008

www.jn.org