Dynamic Changes in Neural Circuit Topology Following Mild MechanicalInjury In Vitro

TAPAN P. PATEL,1 SCOTT C. VENTRE,1 and DAVID F. MEANEY1,2

1Department of Bioengineering, University of Pennsylvania, 220 S 33rd St, Philadelphia, PA 19104, USA; and 2Department ofNeurosurgery, University of Pennsylvania, Philadelphia, PA 19104, USA

(Received 23 July 2011; accepted 24 August 2011; published online 13 October 2011)

Associate Editor Stefan M. Duma oversaw the review of this article.

Abstract—Despite its enormous incidence, mild traumaticbrain injury is not well understood. One aspect that needsmore definition is how the mechanical energy during injuryaffects neural circuit function. Recent developments incellular imaging probes provide an opportunity to assessthe dynamic state of neural networks with single-cell reso-lution. In this article, we developed imaging methods toassess the state of dissociated cortical networks exposed tomild injury. We estimated the imaging conditions needed toachieve accurate measures of network properties, and appliedthese methodologies to evaluate if mild mechanical injury tocortical neurons produces graded changes to either sponta-neous network activity or altered network topology. Wefound that modest injury produced a transient increase incalcium activity that dissipated within 1 h after injury.Alternatively, moderate mechanical injury produced imme-diate disruption in network synchrony, loss in excitatorytone, and increased modular topology. A calcium-activatedneutral protease (calpain) was a key intermediary in thesechanges; blocking calpain activation restored the networknearly completely to its pre-injury state. Together, thesefindings show a more complex change in neural circuitbehavior than previously reported for mild mechanicalinjury, and highlight at least one important early mechanismresponsible for these changes.

Keywords—Traumatic brain injury, Concussions, Calcium

indicator dyes, Stochastic integrate-and-fire model, Neural

network, Synchrony, Cluster, Excitatory tone, Calpain.

INTRODUCTION

Concussions are a form of mild traumatic braininjury that represents at least 80% of the traumaticbrain injuries occurring each year in the UnitedStates.24 Although often associated with sports, the

majorities of concussions occur outside of sportingactivities and include motor vehicle accidents, falls,and any situation involving sudden acceleration ordeceleration of the head. Long considered an injurythat was ‘inevitable’ with little recourse, the growingawareness of the long-term effects of concussion led toa renewed focus on understanding, preventing, andtreating concussions. New technology to assess themechanical circumstances of concussions is leading toa new perspective on effective protective equipment.Alongside the advances in basic science understandingof concussions, new diagnostic imaging methods arealso becoming available to better evaluate and assistclinicians in the care of patients with a concussion.14

The cellular substrate of concussion is largelyunknown.36 One theory on the mechanical origin ofconcussions proposes that the cortical strains appear-ing in the gray matter at the moment of injury are themost likely reason for the impairments that occur in aconcussion.46 Alternatively, others pose that defor-mation of the brainstem structures are key in produc-ing the brief periods of unconsciousness associatedwith severe concussion.6,55 A common theme amongthese concussion theories is that the mechanical energyduring rapid head motions will deform brain tissueand, ultimately, the cellular ensembles that form thesetissues. However, at the multicellular scale, it is notclear if these transient mechanical events will causepermanent changes in the neuronal morphology or,alternatively, the wiring of neural circuits within brainregions associated with concussion. These changes inboth the neuronal connections and, alternatively, thewiring of neural networks can play a key role in therecovery of the brain following a concussion.

Several in vitro models of injury have been designedto investigate the response of the brain parenchyma tomechanical stimulus in isolation, without confounders

Address correspondence to David F. Meaney, Department ofBioengineering, University of Pennsylvania, 220 S 33rd St, Phila-delphia, PA 19104, USA. Electronic mail: dmeaney@seas.upenn.edu

Annals of Biomedical Engineering, Vol. 40, No. 1, January 2012 (! 2011) pp. 23–36DOI: 10.1007/s10439-011-0390-6

0090-6964/12/0100-0023/0 ! 2011 Biomedical Engineering Society

23

such as vascular damage, systemic inflammation,fluctuations in pH, temperature, oxygenation level orintracranial pressure.38,40 Examples of such modelsinclude the transection model,41,42 the compression orweight-drop model,10,53 the hydrostatic pressure43 orhydrodynamic model,28 shear strain models,4,25 andseveral stretch injury models.34,39,49 Of these, thestretch injury or substrate deformation model capturesessential features of the biomechanics of an in vivobrain deformation in concussions.

Along with advances in animal and in vitromodels ofinjury, there has been a concerted effort to start map-ping the connections among neurons within the brainfor both small and large organisms. In parallel, a sub-stantial new set of tools appeared for formulatingprinciples of information flow among neural net-works, measuring their relative connectivity map, andunderstanding key points of control for complex net-works.7,9,18,33,50,57 Key in this progress is a newemerging set of fluorescent reagents that provide stable,long-term optically based recording of dissociatedneurons, ensembles of neurons within acute brain slices,and the living cortex.15,37,60 Collectively, these newtools allow for the direct visualization of the nervoussystem at an unprecedented scale—it allows one toestimate the input/output connectivity maps amongneurons, assess the relative strength of connectionsamong neurons in a network, and to define the networkscaling principles used in the design of the network.

In this study, we critically analyze fluorescencebased imaging methods for mapping neural networkproperties, and use this analysis to design experimentsfor measuring changes in network behavior followingmild and moderate stretch injury to cortical neuronalnetworks in vitro. Unlike higher injury levels (>80%stretch) that will cause neuronal death within 24 hpost-injury, we find that mild levels of mechanicalinjury (35% stretch) causes widespread changes innetwork topology, synchrony, and excitatory trans-mission. The network that was once highly intercon-nected prior to injury breaks off into multiple, smallerclusters of locally connected modules. Long-rangesynaptic transmission is impaired. Functionally, thereis a significant drop in network synchrony. At thesestretch levels, there is delayed proteolysis of voltagegated sodium channels, mediated by the calcium sen-sitive protease, calpain.62 We show that pharmaco-logical blockade of calpain improves networksynchrony. Interestingly, calcium influx through theNMDA receptor is critical in mediating some, but notall, of these changes. Alternatively, even mildermechanical injury (10% stretch) produced very littlechanges in network topology or synchrony, however,there was a brief period of heightened spontaneousactivity following injury.

MATERIALS AND METHODS

In Silico Simulation of Network Behavior

We used a stochastic integrate-and-fire model toevaluate network dynamics, predict alterations in cyto-solic calcium, and use these estimates of cytosolic calciumto predict measurements derived from fluorescent indi-cator dyes and genetically encoded calcium indicators(see Simulating CalciumDynamics). Each neuron withinthe model is designated as either an inhibitory (I) orexcitatory (E) neuron. We base this model on previousmodels fornetworkbehavior,which containneurons thatshow either a regular spiking (RS) or intrinsic bursting(IB) behavior.17 When modeled as a regular spikingneuron, the membrane potential of a single neuron willreceive a spiking input, followed by a refractory perioddefined by a monoexponential decay. To monitor chan-ges in membrane potential v(t), we also include a leakcurrent, a calcium sensitive hyperpolarizing current forpotassium channels (IK(Ca)), a noise component, and ageneralized formulation of the synaptic current:

Cdv

dt! Ileak " Ispike " IK#Ca$ " Inoise " Isyn;

where C is the overall capacitance of neuronal mem-brane. The synaptic current has contributions fromboth the AMPA and NMDA receptors,

Isyn ! dj IAMPA " INMDA# $;

where dj is a synaptic depression factor that accountsfor the loss of vesicles from the j pre-synaptic termi-nals. AMPA and NMDA currents are computed fromprevious Smoldyn simulations that give a temporalprofile of the number of open receptors (NAMPA andNNMDA), following synaptic vesicular release54:

IAMPA ! gAMPA % Vm & E# $ % NAMPA;

INMDA ! gNMDA % Vm & E# $ % NNMDA;

where gAMPA = 12 pS, gNMDA = 45 pS, and thereversal potential E = 0. The remainder of modelparameters are consistent with.17

For an intrinsic bursting neuron, we follow Frenchet al. and add a low threshold calcium channel toaugment the existing components incorporated for abursting neuron. By incorporating the stochastic nat-ure of AMPA and NMDA receptor opening, we en-hance the previous model and account for the observedvariability in activity patterns.

Defining the Connectivity of the In Silico Network

The connectivity of our in silico model was pre-scribed to match in vitro neural network development.

PATEL et al.24

Past work using dissociated cortical neurons indicatethat the input and output degree distribution is Pois-son.1,61 In selecting targets for neuron i’s outputs, weutilized a weighted random sampling function, whereneurons closest to, and in the direction of i’s axonalgrowth cone are preferentially connected. The number,or degree, of inputs (Ki) and outputs (Ko), and therelative distribution of the degree among an ensembleof neurons, defines the overall connectivity in theneural circuit. Models of cortical neural networks weredeveloped and executed using a custom MATLABscript (Mathworks, Inc.). We simulated the activity of100 neurons for 10 s, with 2-ms timesteps, and averageKi = Ko = 20, where each connection was furtherdivided into 20–50 synapses for a total of ~10,000synapses. Synapses were populated with 80 AMPARsand 20 NMDARs. Simulations ran for 3 h in simula-tion time to correspond with 10 s of imaging data.

Simulating Calcium Dynamics

Calcium influx is simulated by incorporating a stepin free cytosolic calcium with each action potential,followed by an exponential decay. Calcium traces wereused to predict changes in the fluorescence of twocommonly used calcium indicator dyes, fura-2 andfluo-4. Predicted fluorescence is dependent on a com-bination of the bu!ering property of the indicator (seeTable 1), the acquisition rate of the imaging camera(0.4–20 frames per second; see methods below), and thesensitivity of the detector. The fluorescence, R, wasdetermined from:

'Ca2"( ! R& Rmin

Rmax & R% Kd

and then downsampled to match experimental imagingspeeds. Rmin was determined by imaging neurons in acalcium free saline solution and Rmax was determinedby imaging neurons in normal saline solution, exposedto 5 lM ionomycin (Sigma-Aldrich). The fraction ofaction potentials that was resolved by the calciumindicator dye was used to compare different imagingconditions.

Cell Culture

Animal procedures were performed in accordancewith the Institutional Animal Care and Use Committeeat the University of Pennsylvania. We used previousmethods developed in our group to generate mixedcultures of neurons/glia from embryonic day 18 rats(E18; Charles River, Wilmington, MA).56 Cell sus-pensions isolated from embryonic rats were plateddirectly onto treated silicone substrates (Sylgard184 + 186 mix, Dow Corning, Midland, MI). Neu-rons were cultured for 18–21 days in Neurobasalmedia (Invitrogen) with B27 and Glutamax (Gibco) ina humidified incubator at 37 "C, 5% CO2. The lengthof time for culturing was used to allow for the fullexpression of glutamate receptors.20

Drug Treatments

All compounds were solubilized in controlled salinesolution (CSS; in mM: 120 NaCl, 5.4 KCl, 1 MgCl2,1.8 CaCl2, 25 HEPES, 15 glucose, pH 7.3), added tothe cells 5 min prior to injury, and remained on thecultures for the duration of the experiment unlessotherwise noted. Cortical neurons were incubated inbicuculline methobromide 50 lM (Tocris), APV25 lM (Tocris), or MDL28170 50 lM (Calbiochem)as needed.

Mechanical Injury of Cortical Neurons

Cortical neurons at 18–21 days in vitro (DIV) onflexible membranes were placed on a stainless steel plateand covered with a top plate to form a sealed chamber.Increasing the chamber pressure to its peak level in15 ms caused the compliant silicone membrane tostretch, in turn applying a stretch to the cultured neu-rons. This pressurization phase was immediately fol-lowed by a release of the pressure, typically within25 ms of achieving the peak pressure. Although cellswere cultured over a circular area (23 mm diameter), wedesigned the supporting stainless steel plate to exposeonly cells in a defined region (a 3 mm 9 18 mm rect-angular region; approximately 1/3 of the culture surfacearea) to a stretch, designing the system to provide auniform membrane stretch over 95% of the surfacewithin that region.34 The membrane in the injuredregion was stretched uniaxially, where the width wasextended 10% (mild) or 35% (moderate) beyond itsinitial width before returning to its original dimension.We did not observe any obvious detachment of the cellsfrom the membrane following stretch injury, nor did weobserve any gross morphological changes in the cul-tures. This stretch level does not cause a significantincrease in cell death compared to unstretched

TABLE 1. Properties of two commonly used calciumindicator dyes.

Fura-2 Fluo-4

Kd 145 nM 345 nMRmin 0.3982 121Rmax 2.7021 549Frame rate 0.4 Hz 0.4–20 Hz

Neural Network Activity After Mild Mechanical Injury 25

controls.35 Naıve unstretched neurons served as refer-ence conditions for statistical comparison.

Calcium Imaging

Three distinct calcium indicators were used in ourexperiments. The ratiometric dye Fura-2AM (Molec-ular Probes, Eugene, OR; 5 lM) was incubated witheach culture for 45 min at 37 "C in CSS, supplementedwith NaHCO3 26 mM. In separate experiments, thesingle emission indicator Fluo-4 (2 lM) was loadedinto cells at 37 "C for 30 min. For imaging the long-term changes in network activity following mechanicalinjury, cells were transduced with a virus expressingthe genetically encoded calcium indicator GCaMP3(AAV2/1-hSynapsin1-GCaMP3, UPenn vector core,Addgene #22692) at DIV 7, and the virus was allowedto express for 10 days before testing.60

Immediately prior to imaging, cultures were rinsedin bu!ered saline solution to remove any residualindicator dye and placed on a spinning disk micro-scope (CSU-10b, Solamere Technologies), configuredon a Nikon TE2000 microscope (Optical Apparatus,Ardmore, PA). A CCD camera (Photometric Cool-Snap HQ2, BioVision, Exton PA) was used to captureimages at a relatively high framing rate for bothGCaMP3 and Fluo-4 labeled cells. Spontaneous cal-cium activity from ~200 neurons in the field of viewwas recorded for 5 min, at 20 Hz acquisition speedusing a 488-nm excitation laser and Nikon 109/0.45Plan Apo objective.

A subset of cultures comparing Fluo-4 and Fura-2indicator dyes was tested on a TE300 microscope. ForFura-2 imaging, cells were alternately excited at 340and 380 nm using an excitation shutter filter wheel(Sutter Instruments, Novato, CA) and the corre-sponding emission images (510 nm) were collectedusing a 14-bit Hamamatsu camera (model c4742-98;Optical Apparatus, Ardmore, PA) at a rate of one ratioimage approximately every 2 s. Fluo-4 images were alsocollected at the same acquisition rate (0.4 Hz), andhigher rates (1 and 10 Hz) for comparison purposes.

Imaging Analysis

Custom-coded MATLAB scripts were used to ana-lyze the image stacks. Individual neurons in the field ofview were identified and the mean fluorescence withineach soma was computed across the image stack. Highfrequency noise was eliminated with a wavelet-basedalgorithm,48 and peak detection algorithm identifiedthe timestamps of calcium transient onset.44 Thetimestamps were used to generate a raster plot, whichprovides a compact visual representation of the pop-ulation activity over time.

Measurements of Synchrony and Network Topology

Synchronization of the network was evaluated usinga previously published set of analysis methods, modi-fied to include non-linear interactions.30,51 Each neu-ron j has a discrete sequence of spike times, tj(n). Weassigned a time-varying instantaneous phase to eachneuron j within the nth inter-spike interval, given by:

uj t; n# $ ! 2pn" 2pt& tj n# $

tj n" 1# $ & tj n# $ :

The circular variance of the phase di!erencebetween pair-wise neurons j and k, given by:

Sjk ! ei#uj t# $&uk t# $!! !!

was used to generate a phase synchronization matrix.An eigenvector based algorithm identified clusters oflocally synchronized neurons and yielded a globalsynchronization index, which ranges from 0 (random,non-coordinated activity) to 1 (perfectly synchronousactivity).51 Functional connectivity was determined byusing surrogate time-series to test for significantinteractions between all pairs of neurons.31 The mod-ularity of the resulting network, which ranges from 0(no modules or community structure) to 1 (non-over-lapping modules) was determined to describe changesin network topology.45,50

Excitatory tone in the network was determinedusing a modified approach of disintegrating the giantcomponent.5 Rather than using field stimulus andexposing neurons to large currents, we applied bicu-culline and relied on the native excitatory circuitryto drive synchronous oscillations. By slowly addingincreasing concentrations of AMPA receptor antago-nist NBQX, we were able to stop synchronous oscil-lations and arrive at a critical value of [NBQX]/Kd thatdisintegrates the excitatory network13 (Kd = 47 nm).

Statistics

Data were analyzed by ANOVA with Tukey’s posthoc test. At least n = 10 cultures, from three differentisolations were tested for each experimental condition.

RESULTS

In Silico Neural Network Yields Estimates of OptimalImaging Parameters Necessary to Achieve Accurate

Spike Detection

We first developed an in silico model of neuralnetwork activity, conducting a parametric analysis tounderstand advantages and drawbacks of choices thatare commonly used in optical imaging of neural

PATEL et al.26

activity. Our network model used a weighted pseudo-randomized network connection topology, and indi-vidual synapses, populated with AMPA and NMDAreceptors, were assumed to follow a stochastic activa-tion pattern per a recent report54 (Fig. 1a). Integrationof current from individual synaptic release events,which could include a random vesicular release event(i.e., synaptic noise), showed development of sponta-neous network activity that would vary temporally(Fig. 1b). Within individual neurons, these periodicbursts of activity would slightly precede or followneighboring neurons, would show some modestdiversity in the spontaneous timing of bursts. Repeatsimulations of the same network topology yieldedslightly different patterns of activity, owing to the sto-chastic synaptic activation scheme. Inhibitory neurons

in the network provided modulation of the activitypattern, as reducing the relative strength of the inhib-itory neuronal tone led to a progressive synchroniza-tion of network activity (Fig. 1c), resembling acommon technique to synchronize the electrical activ-ity of cortical neuron cultures in vitro.

Using estimates of calcium binding for di!erentcalcium indicator dyes, we investigated how well fura-2and fluo-4 capture underlying electrical activity. Underthe most ideal conditions, without hardware noise, wefound neither Fura-2 nor Fluo-4 provided accurateestimates of precise timing of action potential firingwhen sampled at relatively low rates. Although Fura-2o!ers approximately double the accuracy in detectingaction potentials within networks under very lowframing rate conditions (0.4 frames per second), the

FIGURE 1. In silico model of neural network activity. (a) Pyramidal excitatory neuronal network was constructed, where eachneuron connected to 20 other neurons on average, chosen from a Poisson distribution. Synaptic connections occurred ondendritic spines, populated with AMPA and NMDA receptors. (b, c) Stochastic integrate-and-fire model reproduced spontaneousactivity patterns, composed of flickering events, intermixed with network-wide bursts. Reducing the strength of the inhibitorynetwork increased synchronous oscillations. (d) Calcium transients predicted from underlying action potentials were converted tofluorescence traces for fura-2 and fluo-4, taking into account their unique buffering properties. The accuracy of resolving actionpotentials using fluorescent dye is greater for fluo-4 and increases with imaging acquisition speed.

Neural Network Activity After Mild Mechanical Injury 27

predicted accuracy was still less than 10% (Fig. 1d).Fluo-4 offers the ability, with its single emissionwavelength, to capture fluorescence imaging data at amuch faster rate limited only by the imaging hard-ware.59 Across typical imaging rates available incommercial interline CCD and EMCCD camera, wefind that image acquisition rate will substantiallyimprove the fidelity between increases in calciumindicator fluorescence changes and action potentialactivity (Fig. 1d). Optimal image acquisition rates withthe Fluo-4 indicator exceed 80% accuracy in capturingaction potentials (85 ± 7.8%) at reasonable framingrates, and approach 60% (58 ± 8.52) even at 10 framesper second.

Optical Imaging of Network Activity In Vitro RevealsNetwork Modules and Minimum Acquisition

Requirements

We tested these in silico predictions directly on ourimaging system, assessing both the relative perfor-mance of different indicator dyes at low framing ratesand the performance of the Fluo-4 indicator at higherframing rates. Several algorithms are available todetect spike patterns from optical imaging data, andour wavelet-based algorithm yielded consistent, repeat-able measures of activity on a single neuron basis thatwas not complicated by altering or changing thresholdconditions and baseline shifts that are often challeng-ing experimental factors (Figs. 2a, 2b). Applying thisdetection algorithm to neuronal network cultures(n = 10), we found no discernable difference betweenFluo-4 and Fura-2 at framing rates typically used intime-lapse imaging applications (0.4 Hz) (Figs. 2c, 2d).In the same culture, we progressively increased framingacquisition rates using the Fluo-4 indicator dye.Acquiring images at 1 and 20 Hz, rather than 0.4 Hz,revealed a much more dynamic activity pattern, pick-ing up many short lived calcium transients (Figs. 2d–2f). At higher imaging rates, we could no longer detectactivity patterns clearly evident at lower framing rates,presumably due to the aggregate detection limits of theobjective, filters, camera, and other optical elements inthe light path. We observed no significant difference inthe activity patterns measured with either Fluo-4 orGCaMP3, a recently introduced genetically engineeredcalcium sensor (data not shown).60

After establishing the necessary acquisition speedsto accurately detect activity, we next considered boththe number of neurons and length of imaging timenecessary to accurately estimate network synchroni-zation, a key parameter often used to determine net-work modules.3 Across multiple cultures (n = 8),we found that the phase difference in the calciumtransients was not spatially sensitive, suggesting a

randomized network topology often associated withdissociated cortical networks.23 The periodic burstingbehavior observed with neural networks, whenacquired at high framing rates (20 Hz), yielded a syn-chronization index for the entire culture and alsoproduced measures of smaller subclusters of neuronswith highly coordinated firing activity (Figs. 3b, 3c).To obtain accurate measures of synchronization, weneeded imaging data from at least 40 individual neu-rons to achieve less than 5% error (Fig. 3e). Similarly,we found that at least 2.5 min of imaging spontaneouscalcium activity in the dissociated cultures was neces-sary to achieve reasonably stable estimates of thesynchronization index (Fig. 3f).

A final parameter that we investigated experimen-tally is the excitatory tone of the network. Networkdisinhibition with bicuculline caused synchronizedcalcium oscillations (Fig. 4a) whose frequency wassystematically reduced with increasing dose of AMPAreceptor antagonist NBQX (Figs. 4b, 4c), until oscil-lations abruptly stopped at certain threshold [NBQX](Fig. 4d). The [NBQX]threshold/Kd was recorded as theexcitatory tone. Similar to network synchronization,sampling from a large neuronal population may notbe necessary to achieve accurate estimates of thisparameter. Using a progressive inhibition of AMPAreceptors in dissociated culture, we found that themeasured excitatory tone would require recordingactivity from at least 30 neurons to achieve a stablemeasure with less than 5% variation. Using this as aguide, we verified that excitatory tone was modulatedacross development time in the dissociated cultures.Several studies indicate a progressive increase inAMPA receptor number indicated with immunoblot-ting and immunocytochemistry,32 and we found asimilar increase in the excitatory tone that progressedand stabilized at later DIV (Fig. 4e).

Mechanical Injury to Cortical Neurons ChangesNetwork Dynamics and Functional Connectivity

Upon establishing the necessary imaging conditionsto capture accurate measures of network behavior, wenext studied the response of cortical neurons to twodistinct levels of mechanical injury. Neither level ofmechanical injury causes neuronal death 24 h follow-ing injury.35 One level of mechanical injury (35 ± 5%)will cause a delayed proteolysis of the voltage gatedsodium channel, as measured by immunoblotting.62

For this injury level, we observe a consistent andimmediate decrease in spontaneous activity, consistentwith a recent report that excitatory transmission isaltered following mechanical injury in cultured neu-rons21 (Figs. 5a, 5b). Our detailed analysis of calcium

PATEL et al.28

FIGURE 2. Comparison of in vitro calcium imaging with fura-2, fluo-4 and GCaMP3 at varying frame rates. (a) Time-lapse image ofGCaMP-3 transduced cortical neurons captured during a network-wide burst event (20 frames per second acquisition speed). (b)Fluorescence trace of two ROIs across 5 min acquisition. Onsets of calcium transients were automatically detected using acontinuous-wavelet transform with >98% accuracy compared to manual detection. (c, d) Representative examples of spontaneousactivity of a mature culture (DIV 18), captured with fura-2 and fluo-4 at the same acquisition speed. The same fluo-4 loaded culturewas imaged at higher speeds (e, 1 Hz) and (f, 20 Hz), revealing richer network activity patterns.

Neural Network Activity After Mild Mechanical Injury 29

transients on an individual neuron basis revealed asignificant decrease in network synchronization(Fig. 5c). These changes in spontaneous activity andsynchronization persisted for 6 h. Functional connec-tivity was also reduced, and the network topologyshowed signs of damage or rewiring as the previouslyintegrated network, consisting of a single intercon-nected module, broke off into multiple sub-modulessuggesting a loss in long-range connectivity. Finally,excitatory tone was decreased relative to uninjuredcontrol cultures, consistent with a broad loss in activityand synchronization (Fig. 5c).

Broadly, these data show that mechanical injury tocortical neurons will incur immediate and broad dis-ruptions in network behavior. However, finite elementmodels suggest a wide range of in vivo brain defor-mations occur during concussion. As such, we alsotested a less severe injury level (10 ± 4%). Distinctfrom studies at the moderate injury level, we observedno change in the network synchronization or excit-atory tone (Figs. 5d–5f). Interestingly, the frequency ofnetwork-wide spontaneous oscillations significantlyincreased acutely after injury and returned to pre-stretch levels by 1 h.

Blocking Calpain Activation After Mechanical Injury toCortical Neurons Restores Synchrony

Given the divergence in network activity changesthat occur following mechanical injury in vitro, wefocused our efforts on the mechanical injury level thatproduced a broader disruption in synchrony andfunctional connectivity. At this injury level, paststudies show a calcium-activated neutral protease(calpain) is capable of cleaving the a-subunit of thevoltage gated sodium channel.52,62 Residing principallyat the axon initial segment, cleavage of the a-subunit iscapable of reducing the ability of the sodium channelto inactivate or, alternatively, increases the thresholdfor firing an action potential. Indeed, both the delayedinactivation and/or the increased firing threshold in thein silico model are capable of generating broad dis-ruptions in network synchrony (Figs. 6a, 6b). Experi-mentally, preventing the activation of calpain withMDL28170 significantly improved network synchrony(p< 0.01, Fig. 6c). Interestingly, blocking calciuminflux with APV pre-treatment greatly attenuated theloss in synchronization, but it did not rescue it com-pletely (p = 0.0423). This maybe due to the partial loss

FIGURE 3. Synchronization analysis for the calcium activity in Fig 2f. The instantaneous phase for each neuron’s calcium activity(a) was used to compute pair-wise phase synchronization index (b). An eigen-vector based algorithm identified 3 clusters ofsynchronously oscillating neurons (c, neuron IDs rearranged by cluster assignment, global synchronization index: 0.4304). (d) Theraster plot of activity from Fig. 2f is rearranged and color-coded by the 3 identified clusters. e Repeated sampling of fraction ofneurons in the field of view show that at least 40 neurons need to be imaged to achieve <5% error in synchronization measurement.(f) Similarly, at least 2.5 min imaging duration yields <5% error.

PATEL et al.30

of NMDAR magnesium block and pathological acti-vation of different NMDAR subtypes that then con-tribute to uncoordinated calcium transients.54

DISCUSSION

In this study, we assess fluorescence based imagingmethods for mapping neural network properties, anduse our analysis to design experiments for measuringchanges in network behavior following mild mechani-cal injury to cortical neural networks in vitro. Unlikehigher injury levels that will cause neuronal deathwithin 24 h post-injury, we find that mild mechanicalinjury causes two different types of changes at thenetwork level. At moderate in vitro injury levels (35%stretch), we find an acute loss in network synchrony,changes in the network topology, and reduction in theexcitatory tone that is not recovered up to 6 h fol-lowing injury. At mild levels of mechanical injury, wesee a different response—the network shows anenhanced frequency in firing that returns to baselinewithin 1 h post-stretch and does not show any change

in synchronization or excitatory tone. The more sig-nificant changes in network behavior at moderatestretch are influenced by NMDA receptor activationand subsequent proteolytic changes in the neuronalpopulations, as these changes would be attenuatedsignificantly with NMDAR antagonists and, sepa-rately, a calpain inhibitor.

A number of recent e!orts developed and charac-terized the ability of existing and new molecular probesto function as a surrogate indicator of single-cellelectrical activity. These e!orts are now positioned toprobe network dynamics and complement traditionalmicroelectrode array technologies. Our predictions ofprobe sensitivity, accuracy, and minimum detectionthresholds for network topology are supported bymeasurements in parallel on dissociated cortical neu-ron networks. Similar to past work,16,58 we find therelatively high speed imaging requirements restricts theuse of probes to those with high efficiency and singleemission capability—notably, Fluo-4 and someGCaMP variants. New contributions from our workinclude estimates for the number of neurons and theacquisition time period necessary to accurately capture

FIGURE 4. Excitatory tone measurement. Forced oscillations were induced by network disinhibition with bicuculline (a). Pro-gressive application of increasing [NBQX] decreased the frequency of oscillations until a critical threshold was reached whensynchronous activity abruptly stopped (b–d). The excitatory tone increases as neurons mature in culture (e).

Neural Network Activity After Mild Mechanical Injury 31

descriptors of the network. These minimal imagingrequirements will help set the framework for experi-ments where it may be useful to map the response of thenetwork to any given stimulus. It is important to pointout that our analysis assumed stability of the networktopology during the imaging period. Certainly, anyin vitro neural circuit exhibits the potential to changethe strength, number, and degree of connections overtime. The relatively short imaging periods needed toestimate network connectivity makes this approachideal for studying changes in neural circuits that occurduring development, in response to prolonged exposure

to chemical treatment, or mechanical stimulus. Lessideal are conditions where the network changes inseconds; to our knowledge, there is no optical imagingtechnology available to perform such a rapid exami-nation of the network and microelectrode arrays maybe a more appropriate approach.

We used these optimized imaging algorithms tostudy changes in dissociated cortical neurons followingmechanical injury. An emerging set of studies showimmediate disruptions in the network activity follow-ing mechanical trauma. Cullen et al. show that a rap-idly applied fluid shear stress was capable of disrupting

FIGURE 5. Changes in network activity following 35 and 10% mechanical stretch. The spontaneous activity of a mature culture isgreatly reduced and de-synchronized 1 h after 35% stretch (a, b). There is a significant drop in global synchronization andexcitatory tone (p< 0.01) and an increase in modularity of the network (c). Alternatively, network activity is unaltered after 10%stretch and shows a transient period of heightened activity (d, e). Synchronization, modularity and excitatory tone remainunchanged after 10% stretch (f).

PATEL et al.32

spontaneous action potential firing within selectedneurons,27 and similar changes were observed whenthree-dimensional constructs were injured in pureshear.12 Using organotypic hippocampal slice cultures,Morrison et al. demonstrate that sublethal biaxialstretch causes electrophysiological dysfunction in theCA1, CA3, and dentate gyrus and results in changes inshort-term plasticity.63 Similarly, a suppression inbroad periodic calcium oscillations appeared followinginjury to dissociated cortical neurons, and this sup-pression in activity persisted 2 days following the ini-tial injury and was accompanied with a change in thecomposition of synaptic glutamate receptors.21 Therecent results from Goforth et al. are most directlycomparable to our tests using a moderate level ofmechanical injury, where our results also show animmediate reduction in spontaneous activity. More-over, we observed a similar reduction in the excit-atory tone of the network post-injury, using a methoddistinct from the electrophysiological techniquesemployed by Goforth et al. By acquiring imaginginformation at higher imaging rates and performingindividual cell-based analysis, our data also reveal newfeatures of this response that included an emergence ofnumerous local modules within the network, reductionin the functional connectivity, and a broad loss insynchronization.

These new data suggest that mechanical trauma canimmediately change the topology of a network. Less

clear, though, is the origin or functional consequenceof this change in network wiring. Individual neuronswith numerous connections to neighboring neurons areoften critical in setting the dynamics of the network,even though they may represent a small fraction of thenetwork.33 We do not know if mechanical injuryselectively impairs these ‘driver nodes’ in the network,therefore dramatically reducing the network excitabil-ity or controllability, or if the mechanical injury affectsindividual neurons more uniformly. Answering thisquestion may be a key next step, since approaches torestore network dynamics may differ if the focus is ona small subpopulation of neurons (i.e., the driverneurons) or the entire network. Similarly, the degra-dation of the network into smaller modules requires acloser examination because it may provide a mecha-nism for subpopulations to restore connections amongits neighboring neurons before re-integrating into thelarger network. Alternatively, the increased modularitymay represent an adaptive response to facilitate rapidand robust response to different environmental cues,2 afeature that may be useful to counter subsequentinjuries or aid in the recovery process after injury.

Given past work showing that the NMDA receptoris a key glutamate mechanical sensor during injury,we were not surprised that blocking this mechanicalsensor would help alleviate the e!ects of mechanicalinjury on the network. At this level of injury, we donot observe significant changes in the membrane

FIGURE 6. Rescuing synchronization. Using our in silico model, we show that axonal pathology can significantly reduce syn-chronization ((a) pre-simulated pathology, (b) post-simulated pathology). Blocking calcium entry with APV or calpain activationwith MDL28170 significantly attenuate the loss in synchronization, following 35% stretch (*p< 0.01, #p< 0.05) (c).

Neural Network Activity After Mild Mechanical Injury 33

permeability, often termed mechanoporation.11,19,26 Assuch, the NMDAR emerges as the most likely con-tributor to neuronal dynamics following mechanicalinjury. Indeed, the nearly complete restoration ofnetwork synchrony with NMDAR inhibition suggeststhe NMDA receptor is a primary mechanistic pathwaythat can influence subsequent changes in the neuralcircuit behavior. The small reduction in synchrony,even with NMDAR blockade, may be due to calciumactivity from partial loss of magnesium block.64 Wealso found that a downstream consequence ofNMDAR activation—the activation of calpain—alsoappears critical in mediating the functional impairmentobserved in neural circuits after TBI. Past work showsthat protease activation following mechanical injury issufficient to cause breakdown of sodium channel sub-units in axons within minutes following mechanicalinjury,22 and our data now suggest the effect of calpaincan be even more rapid. The immediate influence ofcalpain activation suggests a critical role for this pro-tease in the earliest phase after injury, where main-taining network synchrony may be important topromote or enhance pro-survival signaling throughsynaptic NMDARs. The impact of calpain disruptingthe timing of neural circuitry may adversely affect anyenhancement of prosurvival signaling through theNMDARs, as this signaling is strongly activated inhighly synchronized networks.29,47 This persisting lossin synchrony likely extends well beyond the initialphase of calcium influx and calpain activation, asturnover of sodium channels and insertion of newsodium channels is slow.8 Therefore, defining thewindow of altered synchrony will be a key step inassessing recoverability of the network to mechanicalinjury. For example, persisting alterations in syn-chrony mediated by even transient calpain activationmay also make it more difficult to establish networkactivity patterns that can lead to either synapticstrengthening or depotentiation, two key steps in cel-lular models of learning and memory. Alternatively, aloss in synchrony may limit the overexcitatory inputthat may appear in sustained, aberrant bursts of thenetwork following trauma.

Taken together, these data point to new opportu-nities for merging information from di!erent perspec-tives (in silico and in vitro) and length scales to studythe dynamic state of neural circuits following trau-matic mechanical injury. Although the emerging pic-ture on mild mechanical injury to neural circuitsremains to be fully developed, our studies suggest abroad continuum of responses in the network behav-ior. In addition, the ability to measure single neuronalproperties from this network response will lead to anew understanding of the initiating injury mechanismsthat influence the consequence of mechanical injury to

in vitro networks, leading potentially to a betterunderstanding of circuit dysfunction in mild traumaticbrain injury.

ACKNOWLEDGMENTS

We thank Dr. Pallab Singh (University of Penn-sylvania) for providing NMDAR and AMPAR acti-vation profiles and for helpful discussions that led tothe development of our stochastic integrate-and-firemodel. Funding for this work was provided from theNational Institutes of Health and the Department ofDefense (W911NF-10-1-0526).

CONFLICT OF INTEREST

The authors do not have any conflict of interest.

REFERENCES

1Amit, D. J., and N. Brunel. Model of global spontaneousactivity and local structured activity during delay periodsin the cerebral cortex. Cereb. Cortex 7:237–252, 1997.2Babu, M. M., N. M. Luscombe, L. Aravind, M. Gerstein,and S. A. Teichmann. Structure and evolution of tran-scriptional regulatory networks. Curr. Opin. Struct. Biol.14:283–291, 2004.3Bialonski, S., and K. Lehnertz. Identifying phase syn-chronization clusters in spatially extended dynamical sys-tems. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 74:051909, 2006.4Bottlang, M., M. B. Sommers, T. A. Lusardi, J. J. Miesch,R. P. Simon, and Z. G. Xiong. Modeling neural injury inorganotypic cultures by application of inertia-driven shearstrain. J. Neurotrauma 24:1068–1077, 2007.5Breskin, I., J. Soriano, E. Moses, and T. Tlusty. Percola-tion in living neural networks. Phys. Rev. Lett. 97(18):188102, 2006.6Browne, K. D., X. H. Chen, D. Meaney, and D. H. Smith.Mild traumatic brain injury and diffuse axonal injury inswine. J. Neurotrauma 2011. doi:10.1089/neu.2011.1913.7Bullmore, E., and O. Sporns. Complex brain networks:graph theoretical analysis of structural and functionalsystems. Nat. Rev. Neurosci. 10:186–198, 2009.8Cantrell, A. R., and W. A. Catterall. Neuromodulation ofNa+ channels: an unexpected form of cellular plasticity.Nat. Rev. Neurosci. 2:397–407, 2001.9Cao, C., and S. Slobounov. Alteration of cortical func-tional connectivity as a result of traumatic brain injuryrevealed by graph theory, ICA, and sLORETA analyses ofEEG signals. IEEE Trans. Neural Syst. Rehabil. Eng.18:11–19, 2010.

10Church, A. J., and R. D. Andrew. Spreading depressionexpands traumatic injury in neocortical brain slices. J.Neurotrauma 22:277–290, 2005.

11Cullen, D. K., and M. C. LaPlaca. Neuronal response tohigh rate shear deformation depends on heterogeneity ofthe local strain field. J. Neurotrauma 23:1304–1319, 2006.

PATEL et al.34

12Cullen, D. K., M. C. Lessing, and M. C. LaPlaca. Colla-gen-dependent neurite outgrowth and response to dynamicdeformation in three-dimensional neuronal cultures. Ann.Biomed. Eng. 35:835–846, 2007.

13Dev, K. K., V. Petersen, T. Honore, and J. M. Henley.Pharmacology and regional distribution of the binding of6-[3H]nitro-7-sulphamoylbenzo[f]-quinoxaline-2,3-dione torat brain. J. Neurochem. 67:2609–2612, 1996.

14Difiori, J. P., and C. C. Giza. New techniques in concussionimaging. Curr. Sports Med. Rep. 9:35–39, 2010.

15Dombeck, D. A., C. D. Harvey, L. Tian, L. L. Looger, andD. W. Tank. Functional imaging of hippocampal placecells at cellular resolution during virtual navigation. Nat.Neurosci. 13:1433–1440, 2010.

16Fetcho, J. R., K. J. Cox, and D. M. O’Malley. Monitoringactivity in neuronal populations with single-cell resolutionin a behaving vertebrate. Histochem. J. 30:153–167, 1998.

17French, D. A., and E. I. Gruenstein. An integrate-and-firemodel for synchronized bursting in a network of culturedcortical neurons. J. Comput. Neurosci. 21:227–241, 2006.

18Garofalo, M., T. Nieus, P. Massobrio, and S. Martinoia.Evaluation of the performance of information theory-basedmethods and cross-correlation to estimate the functionalconnectivity in cortical networks. PLoS ONE 4(8):e6482,2009.

19Geddes-Klein, D. M., K. B. Schiffman, and D. F. Meaney.Mechanisms and consequences of neuronal stretch injuryin vitro differ with the model of trauma. J. Neurotrauma23:193–204, 2006.

20Geddes-Klein, D. M., G. Serbest, M. N. Mesfin, A. S.Cohen, and D. F. Meaney. Pharmacologically inducedcalcium oscillations protect neurons from increases incytosolic calcium after trauma. J. Neurochem. 97:462–474,2006.

21Goforth, P. B., J. Ren, B. S. Schwartz, and L. S. Satin.Excitatory synaptic transmission and network activity aredepressed following mechanical injury in cortical neurons.J. Neurophysiol. 105:2350–2363, 2011.

22Iwata, A., P. K. Stys, J. A. Wolf, X. H. Chen, A. G. Taylor,D. F. Meaney, and D. H. Smith. Traumatic axonal injuryinduces proteolytic cleavage of the voltage-gated sodiumchannels modulated by tetrodotoxin and protease inhibi-tors. J. Neurosci. 24:4605–4613, 2004.

23Kamioka, H., E. Maeda, Y. Jimbo, H. P. Robinson, and A.Kawana. Spontaneous periodic synchronized burstingduring formation of mature patterns of connections incortical cultures. Neurosci. Lett. 206:109–112, 1996.

24Kraus, J. F., and P. Nourjah. The epidemiology of mild,uncomplicated brain injury. J. Trauma 28:1637–1643, 1988.

25LaPlaca, M. C., D. K. Cullen, J. J. McLoughlin, and R. S.Cargill. High rate shear strain of three-dimensional neuralcell cultures: a new in vitro traumatic brain injury model.J. Biomech. 38:1093–1105, 2005.

26Laplaca, M. C., and G. R. Prado. Neural mechanobiologyand neuronal vulnerability to traumatic loading. J. Bio-mech. 43:71–78, 2010.

27LaPlaca, M. C., G. R. Prado, D. K. Cullen, and H. R.Irons. High rate shear insult delivered to cortical neuronsproduces heterogeneous membrane permeability altera-tions. Conf. Proc. IEEE Eng. Med. Biol. Soc. 1:2384–2387,2006.

28LaPlaca, M. C., and L. E. Thibault. An in vitro traumaticinjury model to examine the response of neurons to ahydrodynamically-induced deformation. Ann. Biomed. Eng.25:665–677, 1997.

29Leveille, F., S. Papadia, M. Fricker, K. F. S. Bell, F. X.Soriano, M. A. Martel, C. Puddifoot, M. Habel, D. J.Wyllie, C. Ikonomidou, A. M. Tolkovsky, and G. E.Hardingham. Suppression of the intrinsic apoptosis path-way by synaptic activity. J. Neurosci. 30:2623–2635, 2010.

30Li, X. L., D. Cui, P. Jiruska, J. E. Fox, X. Yao, and J. G.R. Jefferys. Synchronization measurement of multipleneuronal populations. J. Neurophysiol. 98:3341–3348, 2007.

31Li, X., G. Ouyang, A. Usami, Y. Ikegaya, and A. Sik.Scale-free topology of the CA3 hippocampal network: anovel method to analyze functional neuronal assemblies.Biophys. J. 98:1733–1741, 2010.

32Lin, Y. C., Z. H. Huang, I. S. Jan, C. C. Yeh, H. J. Wu,Y. C. Chou, and Y. C. Chang. Development of excitatorysynapses in cultured neurons dissociated from the corticesof rat embryos and rat pups at birth. J. Neurosci. Res.67:484–493, 2002.

33Liu, Y. Y., J. J. Slotine, and A. L. Barabasi. Controllabilityof complex networks. Nature 473:167–173, 2011.

34Lusardi, T. A., J. Rangan, D. Sun, D. H. Smith, and D. F.Meaney. A device to study the initiation and propagationof calcium transients in cultured neurons after mechanicalstretch. Ann. Biomed. Eng. 32:1546–1558, 2004.

35Lusardi, T. A., J. A. Wolf, M. E. Putt, D. H. Smith, andD. F. Meaney. Effect of acute calcium influx aftermechanical stretch injury in vitro on the viability of hip-pocampal neurons. J. Neurotrauma 21:61–72, 2004.

36Meaney, D. F., and D. H. Smith. Biomechanics of con-cussion. Clin. Sports Med. 30:19–31, vii, 2011.

37Mittmann, W., D. J. Wallace, U. Czubayko, J. T. Herb,A. T. Schaefer, L. L. Looger, W. Denk, and J. N. Kerr.Two-photon calcium imaging of evoked activity from L5somatosensory neurons in vivo. Nat. Neurosci. 14(8):1089–1093, 2011.

38Morrison, B., B. S. Elkin, J. P. Dolle, and M. L. Yarmush.In vitro models of traumatic brain injury. Annu. Rev. Bio-med. Eng. 13:91–126, 2011.

39Morrison, B., 3rd, D. F. Meaney, and T. K. McIntosh.Mechanical characterization of an in vitro device designedto quantitatively injure living brain tissue. Ann. Biomed.Eng. 26:381–390, 1998.

40Morrison, B., 3rd, K. E. Saatman, D. F. Meaney, and T. K.McIntosh. In vitro central nervous system models ofmechanically induced trauma: a review. J. Neurotrauma15:911–928, 1998.

41Mukhin, A. G., S. A. Ivanova, J. W. Allen, and A. I.Faden. Mechanical injury to neuronal/glial cultures in mi-croplates: role of NMDA receptors and pH in secondaryneuronal cell death. J. Neurosci. Res. 51:748–758, 1998.

42Mukhin, A. G., S. A. Ivanova, S. M. Knoblach, and A. I.Faden. New in vitro model of traumatic neuronal injury:evaluation of secondary injury and glutamate receptor-mediated neurotoxicity. J. Neurotrauma 14:651–663, 1997.

43Murphy, E. J., and L. A. Horrocks. A model for com-pression trauma: pressure-induced injury in cell cultures.J. Neurotrauma 10:431–444, 1993.

44Nenadic, Z., and J. W. Burdick. Spike detection using thecontinuous wavelet transform. IEEE Trans. Biomed. Eng.52:74–87, 2005.

45Newman, M. E. Modularity and community structure innetworks. Proc. Natl Acad. Sci. USA 103:8577–8582, 2006.

46Ommaya, A. K., and T. A. Gennarelli. Cerebral concussionand traumatic unconsciousness. Correlation of experimen-tal and clinical observations of blunt head injuries. Brain97:633–654, 1974.

Neural Network Activity After Mild Mechanical Injury 35

47Papadia, S., F. X. Soriano, F. Leveille, M.-A. Martel, K. A.Dakin, H. H. Hansen, A. Kaindl, M. Sifringer, J. Fowler,V. Stefovska, G. McKenzie, M. Craigon, R. Corriveau,P. Ghazal, K. Horsburgh, B. A. Yankner, D. J. A. Wyllie,C. Ikonomidou, and G. E. Hardingham. Synaptic NMDAreceptor activity boosts intrinsic antioxidant defenses. Nat.Neurosci. 11:476–487, 2008.

48Pardo, E., J. L. San Emeterio, M. A. Rodriguez, andA. Ramos. Noise reduction in ultrasonic NDT usingundecimated wavelet transforms. Ultrasonics 44(Suppl 1):e1063–e1067, 2006.

49Pfister, B. J., T. P. Weihs, M. Betenbaugh, and G. Bao. Anin vitro uniaxial stretch model for axonal injury. Ann.Biomed. Eng. 31:589–598, 2003.

50Rubinov, M., and O. Sporns. Complex network measuresof brain connectivity: uses and interpretations. NeuroImage52:1059–1069, 2010.

51Rummel, C., M. Muller, and K. Schindler. Data-drivenestimates of the number of clusters in multivariate timeseries. Phys. Rev. E Stat. Nonlin. Soft Matter Phys.78:066703, 2008.

52Saatman, K. E., B. Abai, A. Grosvenor, C. K. Vorwerk,D. H. Smith, and D. F. Meaney. Traumatic axonal injuryresults in biphasic calpain activation and retrogradetransport impairment in mice. J. Cereb. Blood Flow Metab.23:34–42, 2003.

53Sieg, F., P. Wahle, and H. C. Pape. Cellular reactivity tomechanical axonal injury in an organotypic in vitro modelof neurotrauma. J. Neurotrauma 16:1197–1213, 1999.

54Singh, P., A. J. Hockenberry, V. R. Tiruvadi, and D. F.Meaney. Computational investigation of the changingpatterns of subtype specific NMDA receptor activationduring physiological glutamatergic neurotransmission.PLoS Comput. Biol. 7:e1002106, 2011.

55Smith, D. H., M. Nonaka, R. Miller, M. Leoni, X. H.Chen, D. Alsop, and D. F. Meaney. Immediate coma

following inertial brain injury dependent on axonal damagein the brainstem. J. Neurosurg. 93:315–322, 2000.

56Spaethling, J. M., D.M. Klein, P. Singh, and D. F. Meaney.Calcium-permeable AMPA receptors appear in corticalneurons after traumatic mechanical injury and contribute toneuronal fate. J. Neurotrauma 25:1207–1216, 2008.

57Stam, C. J., and J. C. Reijneveld. Graph theoretical anal-ysis of complex networks in the brain. Nonlinear Biomed.Phys. 1:3, 2007.

58Stosiek, C., O. Garaschuk, K. Holthoff, and A. Konnerth.In vivo two-photon calcium imaging of neuronal networks.Proc. Natl Acad. Sci. USA 100:7319–7324, 2003.

59Takahashi, N., T. Sasaki, A. Usami, N. Matsuki, andY. Ikegaya. Watching neuronal circuit dynamics throughfunctional multineuron calcium imaging (fMCI). Neurosci.Res. 58:219–225, 2007.

60Tian, L., S. A. Hires, T. Mao, D. Huber, M. E. Chiappe,S.H.Chalasani, L. Petreanu, J.Akerboom, S.A.McKinney,E. R. Schreiter, C. I. Bargmann, V. Jayaraman, K. Svoboda,and L. L. Looger. Imaging neural activity in worms, flies andmice with improved GCaMP calcium indicators. Nat.Methods 6:875–881, 2009.

61van Vreeswijk, C., and H. Sompolinsky. Chaos in neuronalnetworks with balanced excitatory and inhibitory activity.Science 274:1724–1726, 1996.

62von Reyn, C. R., J. M. Spaethling, M. N. Mesfin, M. Ma,R. W. Neumar, D. H. Smith, R. Siman, and D. F. Meaney.Calpain mediates proteolysis of the voltage-gated sodiumchannel alpha-subunit. J. Neurosci. 29:10350–10356, 2009.

63Yu, Z., and B. Morrison, 3rd. Experimental mild traumaticbrain injury induces functional alteration of the developinghippocampus. J. Neurophysiol. 103:499–510, 2010.

64Zhang, L., B. A. Rzigalinski, E. F. Ellis, and L. S. Satin.Reduction of voltage-dependent Mg2+ blockade ofNMDA current in mechanically injured neurons. Science274:1921–1923, 1996.

PATEL et al.36