anticonvulsant drugs are neuronal network-modifying agents...

Anticonvulsant Drugs Are Neuronal Network-Modifying Agents (NMAs)CL Faingold, Southern Illinois University School of Medicine, Springfield, IL, United States

� 2017 Elsevier Inc. All rights reserved.

Introduction 1Background 1Approaches to Understanding CNS Drug Mechanisms 1Network Perspective on Brain Function 2Networks in Epilepsy 2Current Classification of Anticonvulsant NMA Mechanisms 3Audiogenic Seizure Models 3Neuronal Network for Audiogenic Seizures 4

Approaches to Seizure Networks 4AGS Network Findings 4Anticonvulsant NMA Action on the AGS Network 5Networks and CNS Pathophysiology 8

Anticonvulsants in Epilepsy and Pain Networks 9Future Directions 11Acknowledgements 12References 12

Introduction

Despite decades of research, it is still not really understood precisely how anticonvulsant drugs act to control the epilepsies, whichare chronic neurological disorders. The acute actions of anticonvulsants have been evaluated on isolated neurons in vitro, whichis the standard approach to defining mechanisms of action. These acutely determinedmechanisms may help explain the ability tostop ongoing seizures, which likely contribute to the ability of chronically-administered anticonvulsants to prevent seizures.However, translating these findings from the dish (in vitro) to the intact brain is problematic. A major reason for this quandaryis the complexity of the human brain, which contains billions of neurons and trillions of synapses. A true understanding of therelevant actions of these drugs requires studying neurons in the intact brain that are located in brain regions that are important tothe generation of the seizures using therapeutic doses of the anticonvulsants. Neuronal networks play important roles in allseizures, including those once considered to be strictly localized (focal) in nature. It is becoming clear that neuronal networksin the intact brain have the capability to exhibit “additional” properties that arise from neuronal interactions between nucleiwithin the network that may not be present in isolated neurons even from the same brain structure (Faingold, 2004; Faingoldand Blumenfeld, 2015). These “new” properties that arise from the intact network are called “emergent properties,” and recentfindings indicate that anticonvulsant drugs can act selectively on these properties to exert their therapeutic effects (Faingold,2014b; Faingold and Blumenfeld, 2015). Therefore, agents that are currently classified as anticonvulsants are more accuratelytermed neuronal network-modifying agents (NMAs) (Faingold, 2009). A major reason for using this drug classification is thatmany anticonvulsant NMAs are also effective in chronic pain disorders as well as a number of neuropsychiatric disorders,such as anxiety, obesity, alcoholism, migraine, and mania. Each of these disorders is mediated by a neuronal network in the brain(Faingold and Blumenfeld, 2014a).

Background

Approaches to Understanding CNS Drug Mechanisms

Early in the development of CNS drugs, investigators proposed that most agents exert therapeutic effects by selective actions onneurons in specific brain sites rather than by acting globally on all neurons. An example of this earlier neuropharmacology para-digm is the theory that sedative-hypnotic drugs produce sleep by selective actions on the brainstem reticular formation, a wake-fulness center, which was a key element in the “centrencephalon,” an early example of what we now call a neuronal network(Penfield and Jasper, 1954). This theory was never disproven and has actually been refined in recent studies (Sukhotinsky andDevor, 2014). However, for a number of years this approach was superseded by another paradigm, based on the fact thatmost CNS drugs affect neurotransmitter receptors or ion channels, mechanisms that had been identified by studying isolatedneurons in vitro. Subsequent work has shown that the selective effects seen in in vitro studies are not always seen in the sameneurons in the intact animal in vivo, as discussed below. Such inconsistencies have emphasized that the neuronal networkapproach may yield a more accurate and complete understanding of CNS drug action, especially since it has become well-established that neurological and neuropsychiatric diseases are mediated by neuronal networks rather than single CNS neurons

Reference Module in Neuroscience and Biobehavioral Psychology http://dx.doi.org/10.1016/B978-0-12-809324-5.00009-2 1

http://dx.doi.org/10.1016/B978-0-12-809324-5.00009-2

in one locus within the brain (Faingold and Blumenfeld, 2014a). The inconsistencies between actions of drugs in vivo versusin vitro can be explained by the fact that neurons within networks may express additional emergent properties at which drugscan potentially act. These emergent properties may be absent from the same neurons in vitro, because important influenceson these neurons are lost when the neurons are isolated (see Fig. 5).

Network Perspective on Brain Function

The many complex functions that the brain performs involve many different distributed neuronal networks. The term neuronalnetwork has been used in a wide variety of ways and in different contexts (Stam, 2014; Faingold and Blumenfeld, 2014a). Thisreview will concentrate on relatively large-scale networks, each of which consists of several distinct brain regions. These networkswere defined using direct recordings of neuronal action potentials within specific brain sites (network hubs) that were initially iden-tified, using neuroanatomically-based techniques, such as neuroimaging (Faingold and Blumenfeld, 2015). For example, a compre-hensive network has been identified that controls normal locomotion, and another distinct widespread network mediates hearing(Jordan and S1awi�nska, 2014; Brozoski and Bauer, 2014). Neuronal network connections can be strengthened and, in unusual cases,new ones can even be formed, which modify brain function, both positively, as seen in learning (Schafe, 2014) and negatively, asseen in epilepsy (N’Gouemo et al., 2014). Many other brain disorders involve disrupted neuronal network functions, as seen indegenerative diseases, such as Parkinson’s disease (Lakraj et al., 2014). Other brain disorders are due to the formation of abnormalconnectivity within neuronal networks due to neuroplasticity, for example fiber sprouting and other network expansion mecha-nisms that occur in several chronic forms of epilepsy (N’Gouemo et al., 2014).

Although neuroanatomical studies have identified essential details about the connections between brain neurons, a func-tional neuronal network can greatly differ from the neuroanatomy of that pathway. Functional studies indicate that a significantproportion of anatomical connections, when activated physiologically, only result in subthreshold responses, limiting the acti-vation of many neurons in the network. Such subthreshold responsiveness is seen to an extensive degree in non-primarysensory and motor brain regions in areas of the brain termed “conditional multi-responsive” (CMR) regions (Faingold,2008; Faingold et al., 2014b). CMR brain regions are particularly important in the ability of the brain networks to undergofunctional changes. These subthreshold responses, which are so predominant in CMR brain regions, can reach threshold underpathophysiological conditions, resulting in the activation of a specific disease network. Such findings have led to the idea thatCNS disorders are largely governed by what could be termed the “Law of Conservation of Networks” (Faingold, 2004). That is,the pathophysiology of CNS disorders is mediated largely by normal neuronal networks that are interacting abnormally withother normal networks. For example, an epilepsy model called audiogenic seizures (AGS), described in detail below, involvesan aberrant interaction between the normal auditory network and the normal network for locomotion to produce the bilater-ally symmetrical motor (convulsive) seizure in response to intense acoustic stimuli. Thus, the convulsion occurs due to exces-sive stimulation of the auditory pathway, which extensively activates the locomotor network that is only transiently activatedby acoustic stimuli under normal conditions, as seen in the acoustic startle response (Faingold and Tupal, 2014). Multiplenetwork interactions can also occur, such as the interaction of the auditory and locomotor network seen in AGS, which subse-quently interacts importantly with the respiratory network in certain AGS models. This additional network interaction mediatesseizure-induced respiratory arrest, leading to death that is seen in DBA/1 and DBA/2 mouse models of sudden unexpecteddeath in epilepsy (SUDEP), which is a major problem in human epilepsy (Faingold and Tupal, 2014; Faingold et al., 2015;Thurman et al., 2014). Many mechanisms are known to control network function (Table 1), including volume transmissionof neuroactive agents, such as CNS drugs, mediated by diffusion from blood vessels and via the cerebrospinal and extracellularfluid (Agnati et al., 2014). Volume transmission is especially important for all systemically-administered CNS drugs, such asanticonvulsant NMAs.

Networks in Epilepsy

As noted above, all forms of epilepsy involve a distributed epileptic network rather than just abnormal neurons in one brain loca-tion (focus). This view is based on studies of small networks in vitro, animal models of temporal lobe epilepsy, human intracranialEEG, and human and animal neuroimaging. Neuroimaging is one of the currently most useful approaches to networks. However,the minimum time scale that is achievable with neuroimaging is too great to observe neuronal firing changes that occur duringa seizure, which is usually relatively brief. Neuroimaging often does not indicate the nature (inhibition or excitation) of neuronalfiring changes, which are the critical events underlying elaboration of the seizure (Faingold and Blumenfeld, 2014c). Although neu-roimaging can be very useful for correlating anatomical and physiological changes associated with neurological or psychiatric disor-ders, neuroimaging data cannot determine if the brain regions that exhibit changes are requisite, ancillary, or compensatory to thedisease process (Faingold and Blumenfeld, 2015). Direct observation of neuronal firing changes recorded with extracellular micro-electrodes in behaving organisms that occur before and during the seizure or other behavioral manifestation of the disorder allowsobservation at a millisecond time scale of events. This provides details about the occurrence of excitation or inhibition during themanifestation as well as after the event, such as during the post-ictal depression period seen after seizures. The neuronal recordingapproach allows observation of the dynamic functional changes that occur in each network nucleus. The effects of anticonvulsantNMAs on neuronal firing within the requisite network nuclei have further illuminated the mechanisms of network operation inepilepsy and provided important insights on the actions of these agents, as discussed in detail below.

2 Anticonvulsant Drugs Are Neuronal Network-Modifying Agents (NMAs)

Current Classification of Anticonvulsant NMA Mechanisms

The current mechanisms of anticonvulsant NMA action, which were largely determined in vitro, include, for the most part,effects on neurotransmitter actions or directly on ion channels (Bialer et al., 2015). Enhancement of the inhibitory actionof gamma-aminobutyric acid (GABA) at GABAA receptors is thought to contribute importantly to the anticonvulsant effectsof barbiturates, benzodiazepines, vigabatrin, and tiagabine, and this action may contribute to the anticonvulsant effects of val-proic acid and gabapentin. Use-dependent block of sodium channels is another important anticonvulsant drug mechanism forphenytoin, carbamazepine, oxcarbazepine, rufinamide, lacosamide, and lamotrigine. Opening of potassium channels by ezo-gabine is a newer anticonvulsant drug mechanism. Antagonism of specific types of calcium channels is another mechanism ofanticonvulsant NMA action, including ethosuximide, gabapentin, and pregabalin. Blockade of glutamate receptors has alsobeen observed with perampanel, barbiturates, felbamate, and lamotrigine. Many anticonvulsant NMAs exert multiple actionswhen examined at the single cell level in vitro. However, the actions of these anticonvulsant NMAs have received minimal eval-uation in seizure networks, and the limited seizure network research that has been done has shown differential effects onneurons within specific nuclei of the networks, as discussed below. Such differential effects may be particularly relevant to ther-apeutic mechanisms when the doses given to the intact animal do not greatly exceed those required to produce anticonvulsanteffects in the same animal. By limiting the doses the non-therapeutic (adverse) effects that are exerted by all anticonvulsants inexcessive doses can be reduced or eliminated, as discussed below. Many anticonvulsant NMAs exert multiple mechanisms ofaction in vitro, but one of these mechanisms may actually predominate in vivo at therapeutic doses. Interestingly, even withinthe same brain structure anticonvulsant drugs that block sodium channels only inhibit neurons in the excitatory subnetworkand do not reduce the activity of neurons in the inhibitory subnetwork of the same structure at the same doses (Pothmannet al., 2014). These findings suggest that these excitatory neurons possess an emergent property that is not present on the inhib-itory neurons (Faingold and Blumenfeld, 2015), despite the importance of sodium channels in generation of action potentialsin both types of neurons.

Audiogenic Seizure Models

Seizure models in animals have played a key role in the discovery of all antiepileptic drugs. In vitro testing cannot replace animalmodels, because it cannot model the entire spectrum of pharmacodynamic actions required for seizure blockade, does not provideinformation about the pharmacokinetics that occur in intact organisms, and obscures potentially important toxicological effectsthat may occur in the intact animal.

Only a few epilepsy models have well-established neuronal networks (Faingold and Blumenfeld, 2014c), and this review willemphasize a well-defined network in an inherited epilepsy model, since genetics is well-known to play an important role in a signif-icant percentage of human epilepsies. Genetically epilepsy-prone rats (GEPRs), which model generalized convulsive seizures, are

Table 1 Elements that control network function and interaction

A. Neuronal properties1. Burst firing2. Gap junctions3. Electrical field effects (ephaptic interactions)

B. Neuroactive agent (endogenous and exogenous, e.g., CNS drugs)4. Excitatory or inhibitory5. Endogenous or exogenous6. Synaptic or volume transmission7. Tonic or phasic

C. Neuronal milieu8. Extracellular ions and gases (O2)9. Temperature

10. Buffering capacity (pH)D. Network connections

11. Interneuron activity12. Astrocytic integration13. External stimuli14. Multiplicity of synaptic inputs

E. Neuronal “life” cycle- (development, experience, degeneration, aging and repair)15. Brain state (circadian, sleep, coma, etc.)16. Synaptic plasticity (strength changes, synaptogenesis)17. Neurogenesis, neurodegeneration

Reproduced from Faingold, C.L., Blumenfeld, H., 2014. Introduction to neuronal networks of thebrain. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disor-ders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 1–10 with permission.

Anticonvulsant Drugs Are Neuronal Network-Modifying Agents (NMAs) 3

abnormally sensitive to many seizure induction methods, including intense acoustic stimuli, which induce audiogenic seizures(AGS) (Jobe et al., 1992; Faingold and Naritoku, 1992). There are numerous naturally-occurring rodent forms of AGS, and a numberof gene knockout mouse strains also exhibit AGS susceptibility. AGS can also be readily induced in rodents by several different treat-ments (Faingold et al., 2014a). In all forms of AGS, high intensity acoustic stimulation results in ictal behaviors that begin with wildrunning and progress to generalized (all-limb) clonus or tonus. Under conditions of seizure repetition the tonus can be followed bypost-tonic generalized clonus, closely mimicking human tonic-clonic seizures, as discussed below. The terminal convulsivebehavior is followed by post-ictal depression of behavior, including loss of the righting reflex, like that seen in human generalizedtonic-clonic seizures in humans. In human epileptic patients, generalized clonus and tonus occur, but these convulsions are rarelyevoked by acoustic stimuli, although they may be evoked by visual stimuli. This sensory difference between humans and rodentsmay be due to the fact that rodents, unlike humans, are nocturnal and rely less than humans do on the visual network and rely moreon other senses, especially the auditory network. Many anticonvulsant NMAs have been developed using AGSmodels, and the anti-convulsant effect of levetiracetam was initially observed in an AGS model but not in the models that were standard for drugscreening at the time.

Neuronal Network for Audiogenic Seizures

Inherited AGS models provide specific experimental advantages for network determination and functional studies. No invasivetechniques are needed to induce the seizure, and there is the ability to precisely control the seizure-inducing stimulus, which allowsexternal control of seizure induction and the ability to examine responses to stimuli at intensities that are below as well as above theseizure threshold. Research on the neuroanatomy and neurophysiology of the auditory system is also well-developed (Brozoski andBauer, 2014). Once it was realized that the locomotion network was also involved in the AGS network, this well-developed neuro-science knowledge base provided further important information (Jordan and S1awi�nska, 2014). This established neuroscienceknowledge base is readily applicable to AGS, including potential candidate sites within the auditory and locomotor pathways toevaluate for involvement in AGS.

Approaches to Seizure Networks

The use of chronic stereotaxically implanted cannulae and neuronal recording electrodes to experimentally probe the AGS networkin unanesthetized and behaving rats allows determination of the role of specific brain structures in the epilepsy network (Faingoldand Blumenfeld, 2014b). Based on the ability to block AGS by inhibition of specific putative network sites, the requisite structures inthe seizure network were determined. In subsequent experiments extracellular neuronal responses to acoustic stimuli were recordedinitially at intensities below that which will initiate AGS and then at a seizure-inducing intensity. Neuronal firing and behavior wererecorded on video (split-screen) simultaneously during AGS. The neuronal firing data were analyzed using post stimulus-time histo-grams, and quantified neuronal firing changes were statistically compared (Faingold, 2012).

AGS Network Findings

Using these approaches the inferior colliculus (IC) was established as the consensus seizure initiation site. The major defect thatleads to AGS susceptibility in GEPRs is the reduced effectiveness of GABA-mediated acoustically-evoked inhibition in ICneurons that occurs at higher stimulus intensities in normal animals (Faingold, 2002; Faingold et al., 2014a). This reducedinhibition allows high intensity acoustic stimuli to induce excessive firing of IC neurons, which in turn results in abnormallyintense activation of neurons in midbrain locomotor network sites to which the IC projects [deep layers of superior colliculus(DLSC), periaqueductal gray (PAG), pontine reticular formation (PRF)], and substantia nigra reticulata (SNr) (Fig. 1). Thus,excessive activity in the normal auditory network abnormally activates the locomotor network, which projects to the spinal cordand produces the bilaterally symmetrical convulsive behaviors.

The neuronal responsiveness to acoustic stimuli in each of the AGS network sites in GEPRs is significantly greater than in normalanimals, even in response to stimuli below the seizure threshold (Faingold et al., 2014a). The firing patterns of neurons in the AGSnetwork during the seizure change dramatically in each site. However, the patterns are consistently different in each site, as theseizure progresses, depending on the temporal sequence of convulsive behaviors that is occurring. Neuronal recording studies indi-cate that the seizure network operates hierarchically during AGS; that is, intense activation of each network nucleus occurs ina consistent and specific order, which precedes and likely initiates each convulsive behavior, as they appear sequentially (Fig. 2).Thus, IC neurons fire most intensely just prior to and during AGS initiation, while neurons in DLSC fire most intensely as thewild running begins. PAG and PRF neurons fire most intensely just prior to and during tonic convulsions, PRF and SNr neuronsfire most intensely just prior to and during tonic hind limb extension. During post-ictal depression only neurons in the PRF andSNr remain active until recovery of the righting reflex, when neurons in all sites become active again. These findings suggest thatthe network hierarchy begins with IC dominance (greatest change from pre-seizural firing), but the DLSC becomes dominant duringwild running. Then the PAG, PRF, and SNr become dominant during tonic behavior, while the PRF and SNr dominate during tonichind limb extension.


Anticonvulsant NMA Action on the AGS Network

As noted above, abnormal acoustically-evoked neuronal firing prior to seizure occurs in all AGS network nuclei of the GEPR, and theeffects of several anticonvulsant drugs on these elevated neuronal responses were evaluated in these seizure nuclei network inawake, behaving GEPRs. Several of the anticonvulsant NMAs examined (at doses that effectively reduce AGS) exerted differentialeffects on firing of neurons in each network nucleus (see Fig. 3). Since this network operates hierarchically, this suggests that thesite in the network at or closest to the seizure initiation site in IC is likely to be a critical site for drug action when it is administeredat the lowest effective dose and acts as the critical therapeutic target for each specific anticonvulsant NMA.

For example, competitive N-methyl-D-aspartate (NMDA) receptor antagonists are effective anticonvulsants against many formsof experimental seizures, including AGS in GEPRs. Systemic administration of these agents at anticonvulsant doses, prior to seizureinduction, results in significant reductions of IC neuronal responses to acoustic stimuli at all stimulus intensities, suggesting thatcompetitive NMDA receptor antagonists act in the AGS network at or afferent to the IC. However, an uncompetitive NMDAreceptor antagonist, MK-801 (dizocilpine), does not alter the responses of IC neurons, despite the fact that this agent blocksAGS in very low doses (Faingold, 2014b).

Tiagabine blocks the GABA transporter, prolonging its action. Systemically administered tiagabine blocked AGS in GEPRs, butunlike the competitive NMDA receptor antagonists, significant IC neuronal firing reduction prior to seizure was seen only at highacoustic intensities. The time course of the reduced IC neuronal firing paralleled that of AGS suppression with tiagabine, suggestingthat tiagabine enhanced that form of GABAA receptor-mediated inhibition in IC neurons that is prominent at high acoustic inten-sities (Faingold, 2002). However, other anticonvulsant drugs that block AGS, including phenytoin, do not affect IC neuronalresponses to acoustic stimuli, indicating that their actions are exerted at network sites efferent to the IC.

The PAG and PRF are implicated in generation of the tonic convulsive behaviors. AGS in GEPRs that culminate in tonic hindlimb extension, and elevated acoustically-evoked rapid tonic and/or burst neuronal firing, respectively, immediately preceding

CN

SOC

ICDLSC

PAG

SNR BRF

56

7

8

AcousticStimulus

Wild Running(DLSC)

Tonic Flexion(PAG)

Tonic extension(SNR, BRF)

AUDIOGENIC SEIZURE

9

1

2

3

410

11

Figure 1 Diagram of the neuronal network for audiogenic seizures (AGS). The network is organized as a hierarchy beginning with the acousticstimulus (1) input into the auditory input pathway (2–4), including neurons in these nuclei (up to the level of the inferior colliculus (IC) (5), which isthe consensus seizure-initiating site. The IC projects to the brainstem locomotion network nuclei, including the deep layers of superior colliculus(DLSC) (6), projecting to the periaqueductal gray (PAG) (7) and substantia nigra reticulata (SNR) (8) and brainstem reticular formation (BRF) (9)which project to the spinal cord (10). The hierarchical activation of each requisite network nucleus produces the sequential behaviors of AGS (wildrunning followed by tonic flexion and tonic extension) (11). (The critical structure that initiates each behavior is shown below the behavior.) Neuronsin the BRF and SNR are the only regions that are active during the post-ictal behavioral depression that follows the tonic extension behavior. (CN,cochlear nucleus; SOC, superior olivary complex, which are auditory structures important for input to the IC). Brain regions in the auditory pathwayrostral to the IC are not requisite for seizure induction, and, in fact, no regions rostral to the level of the midbrain are requisite to these seizures. Itshould be noted neurons in the lateral amygdala are affected by these seizures, but blockade of this region does not block seizures, rendering theamygdala as an ancillary site. With seizure repetition (AGS kindling) the network expands to include forebrain structures, including the amygdala (seeN’Gouemo et al., 2014). Modified from Faingold, C.L., Blumenfeld, H., July 6, 2015. Targeting neuronal networks with combined drug and stimula-tion paradigms guided by neuroimaging to treat brain disorders. Neuroscientist. pii:1073858415592377 (Epub ahead of print) with permission.


tonic convulsions, have been observed in these brain sites. Phenytoin can block AGS completely, but lower doses of this agent canblock the tonic phase of AGS selectively. Anticonvulsant doses of phenytoin induce consistent changes in PAG and PRF neuronalfiring and behavior in GEPRs (Faingold, 2014b). Phenytoin in doses that selectively suppressed tonic convulsions did not consis-tently alter PAG neuronal responses to acoustic stimuli prior to seizure, but the same doses of phenytoin resulted in significantsuppression of PRF pre-seizural acoustically-evoked neuronal firing and suppressed firing during seizure as well. Higher doses ofphenytoin which completely blocked AGS, significantly reduced PAG acoustically-evoked neuronal firing and more greatlydecreased PRF firing. These results suggest a critical role for PRF, but not PAG neurons, in generation of tonic convulsive behaviorsof AGS. The suppression of PAG and PRF neuronal firing induced by phenytoin, associated with complete seizure blockade, isconsistent with vital roles for both structures in the seizure network. The differential suppressive effect of phenytoin on both pre-seizural and seizural neuronal firing in PRF as compared to the lack of effect on PAG firing indicate that this experimental approachidentified the most sensitive therapeutic target for the action of this anticonvulsant NMA. For an anticonvulsant NMA to completelyblock seizures, however, actions on multiple network sites, including the PRF and PAG in the GEPR, may be needed to significantlyaffect the emergent properties of the seizure network. Interestingly, as discussed in detail below, the PAG is also implicated in thepain network, and during the post-ictal period, it has been observed that analgesia lasting for hours occurs after seizures, includingAGS in GEPRs. The PAG also plays a major role in this seizure-induced reduction in pain sensitivity, since focal blockade of thisstructure will reverse the post-ictal analgesia (Samineni et al., 2011).

Another agent with anticonvulsant actions, the uncompetitive NMDA antagonist, MK-801, did not consistently affect pre-seizural acoustically-evoked neuronal firing in IC, DLSC, PRF, or PAG, despite blocking AGS in GEPRs. However, MK-801 induced

Figure 2 Composite behavioral activity in GEPR-9s and concurrent typical examples of neuronal firing in the major network nuclei for AGS seenduring each seizure behavior. The structures include (A) the inferior colliculus central (ICc) and (B) inferior colliculus external (ICx) nuclei, (C)deep layers of superior colliculus (DLSC), (D) periaqueductal gray (PAG), (E) pontine brainstem reticular formation (PRF), and (F) substantia nigrareticulata (SNr). Thus, ICc and ICx firing is greatest during AGS initiation, while the greatest increase in DLSC neuronal firing occurs at the onsetof wild running. The greatest increase in PAG, SNr, and BRF neuronal firing occur during the tonic and clonic phase(s) of AGS. The BRF and SNrare the only areas active during post-ictal depression of consciousness. Note: Recovery of SNr firing occurred but was not recorded in theexample in F. (Acoustic stimulus parameters: 12 kHz tone burst, 100 ms duration, 5 ms rise-fall, 100 dB SPL, 2–4 Hz repetition rate). Each row isfrom the same rat. Drawings courtesy of Naritoku, D.K., Mecozzi, L.B., Aiello, M.T., Faingold, C.L., 1992. Repetition of audiogenic seizures ingenetically epilepsy-prone rats induces cortical epileptiform activity and additional seizure behaviors. Exp. Neurol. 115 (3), 317–324 and Faingold,C.L., 2012. Brainstem networks: reticulo-cortical synchronization in generalized convulsive seizures. In: Noebels, J.L., Avoli, M., Rogawski, M.A.,Olsen, R.W., Delgado-Escueta, A.V. (Eds.), Jasper’s Basic Mechanisms of the Epilepsies. fourth ed. pp. 257–271 by permission of Oxford Univer-sity Press, New York, NY, USA.


a significant increase of neuronal firing in SNr neurons concomitant with suppression of AGS (Fig. 4), indicating that SNr neuronsare an important therapeutic target for this agent. In contrast, MK-801 induced minimal effect on neuronal firing when perfuseddirectly onto SNr neurons in brain slices in response to a broad range of electrical stimuli and drug concentrations. These findingssuggest that the significantly increased firing induced by systemically administered MK-801 is an effect on the emergent properties ofthe AGS network, critically involving SNr neurons (Faingold, 2004, 2014b). How can a drug produce a major effect on specificneurons in vivo, but not produce this effect on the same neurons in vitro? As noted above and detailed in the example in Fig. 5,there are a number of influences that control neuronal function. In the intact unanesthetized brain in vivo, all of these mechanismsare present. However, many of these mechanisms are absent and others may be greatly altered in the same neurons isolated in vitro(Faingold, 2014a).

If we examine the influences exerted on neurons in a specific brain site, such as the IC, we can see that there are numerous influ-ences, some of which are shown in Fig. 5. When IC neurons are studied in in vitro brain slices, many of these influences are elim-inated (eg, ascending and descending projections) or greatly modified (eg, ionic and O2 levels) (Fig. 5). Also, achieving anappropriate concentration of the drug in vitro is highly problematic. This is exemplified by the “amplification” phenomenon

CN

SOC

ICDLSC

PAG

SNr BRF

56

7

AcousticStimulus

Wild Running(DLSC)

Tonic Flexion(PAG)

Tonic extension(SNr, BRF)

AUDIOGENIC SEIZURE

8

1

2

3

4

911

uc-NMDAantagonists

VT

ST

GABA Agonists

c-NMDAantagonists

Phenytoin

Gabapen n

10

Figure 3 Diagram of the neuronal network for audiogenic seizures (AGS) with emergent properties (indicated by cylinders) in each requisite site atwhich systemically-administered drugs that block these seizures may act at therapeutic doses. The network is organized as a hierarchy beginningwith the acoustic stimulus (1) input into the auditory pathway (2–4), including neurons in these nuclei (up to the level of the inferior colliculus (IC)(5), which is the consensus seizure-initiating site. The IC projects to the brainstem locomotion network nuclei, including the deep layers of superiorcolliculus (DLSC) (6), projecting to the periaqueductal gray (PAG) (7) and substantia nigra reticulata (SNr) (8) and brainstem reticular formation(BRF) (9) which project to the spinal cord (10). The hierarchical activation of each requisite network nucleus produces the sequential behaviors ofAGSz (wild running followed by tonic flexion and tonic extension) (11). (The critical structure that initiates each behavior is shown below thebehavior.) Neurons in the BRF and SNr are the only regions that are active during the post-ictal behavioral depression that follows the tonic extensionbehavior. Therapeutic doses of several drugs with anticonvulsant properties act to inhibit neurons in the IC, including competitive (c-) NMDA antago-nists, such as 2-amino-7-phosphonoheptanoate and GABA uptake inhibitor, tiagabine, as well as ethanol. Therapeutic doses of other drugs that areeffectively anticonvulsant exert no effect on IC neurons, including an uncompetitive (uc-) NMDA antagonist (MK-801), gabapentin, lamotrigine, andphenytoin. Other effective anticonvulsant NMAs, such as gabapentin, act to reduce PAG neuronal firing. Therapeutic doses of phenytoin selectivelyact to inhibit neurons in the BRF of the pons. The SNr is the target of MK-801, but the effect is to increase neuronal firing. The emergent property ofeach nucleus is seen as a confluence of influences onto the neurons in each nucleus, including neuroactive substances (red squares) released ontothe neurons via synaptic transmission (ST) and from the blood vessels, cerebrospinal and extracellular fluids via volume transmission (VT), as shownin the expanded diagram of an emergent property on the left of the network. Systemically administered drugs reach each site via VT. Direct stimula-tion of any site within the network, either chemically or electrically, can affect seizure susceptibility and may modify the emergent properties in theaffected site. (CN, cochlear nucleus; SOC, superior olivary complex, which are auditory structures important for input to the IC). Reproduced fromFaingold, C.L., Blumenfeld, H., July 6, 2015. Targeting neuronal networks with combined drug and stimulation paradigms guided by neuroimaging totreat brain disorders. Neuroscientist. pii:1073858415592377 (Epub ahead of print) with permission.


that occurs in vivo, so that much smaller concentrations of drugs affecting seizures are needed in vivo as compared to in vitro toobtain the same effect (Narahashi et al., 2007).

The realization that functioning brain networks often exhibit emergent properties (Faingold, 2014a) has led to the hypothesisthat these properties themselves are potentially critical targets for the action of CNS drugs (Faingold, 2004). When neurons are iso-lated (in vitro) from their network, regional selectivity of drug action can be lost. In this situation, a neuron in vivo may be affectedby a given drug, but the same drug may not affect the same neuron in vitro, as seen with MK-801. Conversely, the emergent propertytheory also suggests that a neuron may not be affected by a given drug in vivo, even if the same drug has clear effects on the sameneuron in vitro, which has been observed with ethanol. Thus, the network can actually alter the sensitivity of neurons to a drug, asseen in data comparing the regional effects of ethanol in vivo versus in vitro (see Faingold, 2004).

Neuronal networks may be responsible for conferring the selectivity of drug action, especially anticonvulsant NMAs. Most CNSdrugs at therapeutic concentrations exert selective actions on neurons in a specific site within the brain. However, this selectivity isnot due solely to the intrinsic properties of those neurons, because this selectivity can change when the neuron is isolated from thenetwork. The human brain is estimated to contain 50–100 billion neurons and trillions of synapses. Several nuclei have been shownto become organized into complex interconnected systems or networks that perform complex functions in health and disease, asdiscussed above (Faingold and Blumenfeld, 2014a). Network function is subject to “complexity theory” - ie, when the multitude ofelements in a complex system interact, new and unexpected properties can emerge due, in part, to self-organization (Faingold,2014a). Complexity theory can help explain why a network can confer emergent properties on a neuron or set of neurons, enablingthose properties to be a selective target of CNS drug action.

Networks and CNS Pathophysiology

As noted above, CNS neuronal networks often express emergent properties. Specific elements within these networks may be a criticaltherapeutic target on which CNS drugs, including anticonvulsant NMAs, exert their pharmacological effect. These emergent

0

100

200

300

% o

f Con

trol

*

0

50

100

Control MK-801

(A)

(B)

(C)

(D)

% o

f Con

trol

in vivo

in vitro

Figure 4 An uncompetitive NMDA antagonist (dizocilpine, MK-801) blocked audiogenic seizures (AGS) in genetically epilepsy-prone rats (GEPR-9s)and in ethanol withdrawn (ETX) rats and induced significantly increased evoked (acoustic) neuronal responses in substantia nigra reticulata (SNr)in vivo, as shown in A, but this effect on SNr neurons was lost in vitro (brain slices), as shown in C. In B an example of the effect of systemicadministration of MK-801 (0.05 mg kg�1) on acoustically-evoked SNr neuronal responses in a behaving epileptic rat (A) is shown. In control this SNrneuron exhibited a consistent response, as shown by the example in the left panel. However, 30 min after MK-801 (0.05 mg kg�1, i.p.) the firing wasmore than doubled as seen in the right panel, and AGS was blocked. Recovery from the effect occurred by 24 h. As shown in D, no change wasinduced in SNr neurons (N ¼ 6) in vitro. The effect of MK-801 in concentrations up to 200 mM on the mean firing of SNr neurons when depolarizedwas 12.8 � 2.6 (SEM) during MK-801 perfusion as compared to 14.2 � 1.3 action potentials before MK-801 was not significant. Post-stimulus timehistograms (PSTHs) in B, and examples of digital oscilloscope tracings of the extracellular action potentials are shown in C, and the increased firingof the SNr neuron is seen in the right panel. (Amplitude of action potential is 430 mV, 50 ms scan length) [PSTH parameters, peak amplitude (rightPSTH) axis-32 action potentials/bin, 200 ms scan length, 1 ms bin width, 50 stimulus presentations, N, # of action potentials/PSTH]. Reproducedfrom Faingold, C.L., 2014b. Neuronal network effects of drug therapies for CNS disorders. In: Faingold, C.L., Blumenfeld, H. (Eds.), NeuronalNetworks in Brain Function, CNS Disorders, and Therapeutics, Academic Press/Elsevier, San Diego, pp. 443–465 with permission.


properties of network neurons can result from the intensification of a specific receptor or channel or result from a critical interactionof certain of the many influences affecting these neurons in that specific site (Faingold, 2014a), causing these cells in this site to bea selective target for the therapeutic action of the specific anticonvulsant NMA. However, this emergent property may be absent inthe same neurons when they are isolated in a brain slice or culture in vitro (Faingold, 2004, 2014a). The net drug effect in an intactorganism is the sum of all the actions exerted on various levels of these networks by the drug, and may vary by dose, particularlywith doses in the toxic range. It is well-established that neurological and psychiatric disorders are also mediated by networks ofbrain nuclei (Faingold and Blumenfeld, 2014a). A drug’s effect may be exerted on multiple networks or common elements of thesenetworks, as discussed in detail below for the anticonvulsant NMA, gabapentin.

Anticonvulsant NMAs successfully treat seizure disorders, but many of these same drugs are also effective in treating chronic paindisorders and psychiatric diseases, such as anxiety disorders and bipolar disorder, as mentioned above. Each of these disorders isthought to be subserved by specific brain networks (Faingold and Blumenfeld, 2014a). The gate control theory of pain is an exampleof a prototypical neuronal network in the spinal cord (Melzack and Wall, 1965; Faingold and Tupal, 2014), and several techniques,including neuroimaging, have identified a neuromatrix that is activated in chronic pain syndromes (Roy et al., 2014; Sukhotinskyand Devor, 2014). Neuroimaging differences are seen between patients in pain as compared to pain-free individuals, and many ofthese differences are reversible with effective drug therapy. As noted above, many anticonvulsant NMAs effectively treat chronic painsyndromes. These agents may act by affecting emergent properties in specific sites within the neuronal network involved in the path-ophysiology of pain.

Anticonvulsants in Epilepsy and Pain Networks

Anticonvulsant NMAs, including gabapentin and pregabalin, are also effective in the treatment of neuropathic pain syndromes(Bialer, 2012; Mehta et al., 2014). Epilepsy and pain are mediated by separate neuronal networks in the brain, but these networksalso contain common brain regions (Linnman et al., 2012). Thus, human and animal neuroimaging studies have shown thata number of brain areas, including the PAG, are strongly implicated in pain networks, and functional changes in these pathwaysare induced by several drugs that possess analgesic properties (Wager et al., 2013; Becerra et al., 2013). The brainstem and the

(midline)

Volume transmission (B)

Capi

llary

Synap�c Transmission (B)

*

Syna

p�c

Tran

smiss

ion

Syna

p�c

Tran

smiss

ion

+

+ -

-.interneuron

astrocyte

+ -(A)(A)

(D)(D)

(D)

(D)

(B)

(B)

(C)

(E)

Figure 5 Diagram of a number of the types of influences that control network function. In this example two neurons of a bilateral structure areshown, which connect across the midline. The influences on these neurons include (A) neurophysiological mechanisms, such as ion channels, whichmediate specific neuronal firing patterns, such as burst firing. The function of a network can also be controlled by (B) the action of neuroactiveagents, which are active both via synaptic transmission as well as volume transmission. Endogenous neuroactive agents include monoamines, whichcan activate inactive networks. Exogenous neuroactive agents, including CNS drugs, such as anesthetics and stimulants, can also exert profoundeffects on network function. (C) The neuronal milieu, including levels of oxygen and temperature, can also alter network function significantly.(D) Network connections, including interneurons, astrocytes, external inputs, and the multiplicity of synaptic inputs. Other elements (E), includingneuronal “life cycle” events, such as circadian rhythms and sleep state, and also synaptic plasticity, and neurogenesis associated with learning andaging, as well as neuronal degeneration of certain CNS disorders. Reproduced from Faingold, C.L., Blumenfeld, H., 2014b. Introduction to neuronalnetworks of the brain. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders, and Therapeutics. AcademicPress/Elsevier, San Diego, pp. 1–10.


PAG, in particular, are among the areas that show significant neuroimaging changes when gabapentin and other drugs that exertanalgesic properties are administered in humans and animals (Iannetti et al., 2005; Governo et al., 2008; Takemura et al., 2011).

In chronic CNS disorders, such as chronic pain syndromes and progressive epilepsies, additional emergent properties of neuronsmay develop that become the critical targets for the action of anticonvulsant NMAs, such as gabapentin. Recent evidence suggeststhat the PAGmay be a major nexus for both epileptic and chronic pain neuronal networks that is the target of anticonvulsant NMAs,such as gabapentin, and explain why these drugs are effective in these different classes of CNS disorders. Gabapentin exerts effects onPAG neurons in AGS in a somewhat different manner from that of phenytoin described above. Repeated, periodic induction of AGSresults in AGS kindling, which increases seizure duration and, in GEPRs (substrain 9), induces an additional generalized clonusphase that follows the tonic extension seizure [post-tonic clonus (PTC)] (Naritoku et al., 1992; N’Gouemo et al., 2014). Systemicadministration of gabapentin in AGS-kindled GEPRs blocks this PTC behavior, and the seizures temporarily revert to the unkindledpattern of behaviors that end in tonic hind limb extension (Tupal and Faingold, 2012). In AGS kindling the brainstem networkexpands to include the amygdala (N’Gouemo et al., 2014), and focal blockade of the amygdala will also cause the AGS-kindledseizure to revert temporarily to the unkindled pattern (Feng et al., 2001). The pathway between the amygdala and PAG isimplicated in production of PTC, and AGS kindling-induced changes in this pathway were evaluated by recording PAG neuronalresponses evoked by electrical stimulation of the amygdala (Tupal and Faingold, 2012). Electrical stimuli in the amygdalaevoked intensity-dependent PAG neuronal firing that was enhanced significantly above that seen prior to AGS kindling (seeN’Gouemo et al., 2014). Gabapentin blocked PTC in AGS-kindled GEPRs, but the other convulsive behaviors (wild running andtonic hind limb extension) seen during the seizures before AGS kindling were not affected by this same drug dose.Simultaneous with the block of PTC, gabapentin significantly reduced PAG neuronal responses to amygdala stimulation(Fig. 6). These neuronal response patterns returned to levels similar to those seen prior to AGS kindling (Tupal and Faingold,2012). These data suggest that the amygdala to PAG pathway may be critical in mediating the emergence of PTC during AGSkindling. The ability of gabapentin to suppress this pathway to the PAG may be important for its anticonvulsant effects inAGS-kindled GEPRs. In addition, an effect on PAG neurons may also contribute to gabapentin’s effectiveness in anxietydisorders and chronic pain, since the networks that mediate these CNS disorders also involve the pathway between theamygdala and PAG, which is supported by recent data (Samineni et al., 2012), as discussed below.

Chronic pain syndromes often occur as an adverse effect of cancer chemotherapeutic drugs, such as paclitaxel (Yared andTkaczuk, 2012). Neuroimaging data in pain suggest that gabapentin may act on the PAG, as noted above, and the effect of this agenton PAG neuronal firing in response to noxious stimuli of PAG neurons was evaluated in awake, behaving rats. Gabapentin wasadministered to rats that had previously been treated with a paclitaxel administration paradigm that induces a chronic painsyndrome (Flatters and Bennett, 2004), and the changes in pain threshold and PAG neuronal firing were examined (Fig. 6). Thenociceptive response in this chronic pain model was significantly reduced by gabapentin in the same dose that blocked PTC inkindled GEPRs. However, gabapentin in the same dose did not affect the nociceptive response or PAG neuronal firing in normal

0102030405060708090

100

% o

f Con

trol

Neu

rona

l Firi

ng

KindledAGS

ChronicPain

AcutePain

**

Control

Control

Control

Gabapentin

Gabapentin

Gabapentin

Figure 6 An anticonvulsant network-modifying drug (gabapentin, 50 mg kg�1 i.p.) significantly reduced periaqueductal gray (PAG) neuronal firingand seizure severity in a repetitive epilepsy model [kindled audiogenic seizures (AGS)] in GEPR-9s 60 min after administration. The same dose of thisdrug significantly reduced PAG neuronal firing and thermal pain in a chronic pain model (paclitaxel), but this dose had no effect on PAG neuronalfiring in acute (thermal) pain. In the left pair of bars a significantly reduced PAG (ventrolateral) neuronal responsiveness to amygdala (central) elec-trical stimulation was induced by gabapentin at all stimulus intensities tested [repeated measures ANOVA and post-hoc paired t-test (*p < 0.01)].The middle set of bars shows thermal-evoked PAG neuronal responses in rats treated with a chronic pain-inducing protocol (paclitaxel), and a signifi-cant reduction in PAG neuronal responses was also observed 60 min after gabapentin. However, in an acute pain protocol (radiant heat presented tothe paw at 53 �C) the same dose of gabapentin induced no significant change in PAG neuronal firing. Data in the left pair of bars are from awake,behaving GEPR-9s subjected to AGS kindling and in the middle and right pairs of bars are from normal rats. *Significance at p < 0.01 (Repeatedmeasure ANOVA). Reproduced from Faingold, C.L., 2014b. Neuronal network effects of drug therapies for CNS disorders. In: Faingold, C.L.,Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders, and Therapeutics, Academic Press/Elsevier, San Diego, pp. 443–465with permission.


rats prior to paclitaxel treatment. Thus, the same dose of gabapentin significantly reduced the firing of PAG neurons to nociceptivestimuli in the chronic pain model. However, this dose of gabapentin did not affect PAG neuronal firing or responses to the noci-ceptive stimuli in the rats that were not subject to chronic pain (Samineni et al., 2012). These data, along with the effects of gaba-pentin in the AGS kindling paradigm discussed above, suggest that a similar change in responsiveness of PAG neurons appears toemerge in both of these chronic conditions. This change causes PAG neurons to become a critical target for therapeutic doses ofgabapentin in both pain and epilepsy, which is consistent with the important role of the PAG in the neuronal networks forboth disorders. In both cases the therapeutic doses are the same, and in each case this effect is seen only after chronic expansionof the network is induced. Based on the concept of emergent properties (Faingold, 2014a), an additional (emergent) propertyappears to develop in the PAG after either chronic protocol that causes PAG neurons to become highly sensitive to gabapentin,but this sensitivity is absent prior to these network expansion-inducing experiences (N’Gouemo et al., 2014). Gabapentin is notthe only anticonvulsant NMA that is effective in epilepsy and pain. Phenytoin has long been used in chronic pain syndromes(Cheshire, 2007), and, as noted above, this anticonvulsant NMA also inhibits PAG neuronal firing in GEPRs, further supportingthe concept that PAG neurons may be important in both pain and epilepsy networks.

Gabapentin, which is proposed to act primarily by binding to alpha(2) delta Ca2þ channels (N-type) in vitro (Rogawski andBazil, 2008; Geisler et al., 2015), may act on the PAG, because the network influences occurring during chronic pain or chronicepilepsy induce an increased number of channels or channel affinity for this drug. Alternatively, gabapentin has also beenreported to increase the concentration of GABA in the brains of experimental animals (Richerson and Wu, 2004), and humanimaging studies also indicate significant increases in GABA after gabapentin administration (Cai et al., 2012). This GABAergicmechanism may become more sensitive to gabapentin in the intact network due to the chronic conditions. The emergentproperty may also develop because both of these mechanisms are activated. It is also possible that the mechanisms responsiblefor the effectiveness of this anticonvulsant NMA in the chronic conditions may involve secondary events triggered by the diseaseprocess. Thus, the receptor subtypes or subunits may undergo switching due to the chronicity of the conditions, and thesealtered subunits may become a drug target because of increased sensitivity of the particular subunit (Ueda, 2006; Matta et al.,2011). Another possible effect of chronic CNS disorders is that intracellular mechanisms, such as NMDA receptor-mediatedincreases in Ca2þ entry into neurons, cause altered network function and expansion due to chronic network activations causedby seizure repetition or chronic pain (Grabenstatter et al., 2012; N’Gouemo et al., 2014).

Several other mechanisms of action have been proposed for gabapentin (Zhang et al., 2013; Kumar et al., 2012). The changesinduced by chronic activation of a network may cause these “minor”mechanisms to become expressed in specific network nuclei toa greater extent. Resolving these possibilities will require extensive further investigation. Thus, these emergent properties may beselective for the specific network sites in which they are observed, or the same emergent property may also occur in networks inrelated disorders, such as different forms of epilepsy, as well as different classes of CNS disorders, as seen with gabapentin in chronicpain and epilepsy discussed above (Faingold, 2014b).

Neuroimaging techniques have also identified pathophysiological networks in psychiatric disorders, such as anxiety, that expressmajor differences between anxious patients and non-anxious individuals (see Schafe, 2014), and many of these differences are alsoreversible with successful therapy. Evidence suggests that certain anticonvulsant NMAs (eg, valproate, lamotrigine, pregabalin, andgabapentin) may be effective in treating anxiety disorders. It may be worthwhile noting that the PAG is also a key site in theneuronal network that mediates anxiety (Linnman et al., 2012), suggesting a commonality of network structures (PAG) in anxiety,pain and epilepsy, and all these neurological disorders are treatable with the anticonvulsant NMA, gabapentin.

Future Directions

This neuronal network approach adds another dimension (emergent property alteration) to the understanding of how anticonvul-sants block seizures. This approach also sheds light on the other therapeutic effects exerted by these agents, including usefulness inchronic pain syndromes and anxiety disorders, leading to the idea that anticonvulsants are more properly classified as neuronalnetworkmodifying agents (NMAs). Additional experimentation is required to further develop the value of this approach. Systematicexplorations are needed of the variety of emergent properties that exist in other epileptic neuronal networks, as well as in the otherneurological disorders, such as chronic pain, in which these drugs are effective. There is a hierarchy of levels within the nervoussystem at which emergent properties may occur. The concept that an emergent property is a critical target for CNS drug actionwas developed in the GEPR, and additional anticonvulsant NMAs need to be evaluated in this model. More comparisons betweenin vitro versus in vivo effects of the same agent are needed. The generality of this approach also needs to be tested in closely relatedseizure models (Faingold et al., 2014a). The mechanisms need to be examined in other seizure models to identify whether differ-ences and similarities occur. This methodology can generate a compendium of emergent targets for drug action. This approach ispotentially important for the development of improved drug treatment for epilepsy, and a similar approach may be applicable toother CNS disorders, such as anxiety and pain, to determine if there are common actions, as suggested by the findings with gaba-pentin discussed above.

Seizure networks are subject to plasticity, which may result in the expression of additional emergent properties. For example,partial seizures initially involve a relatively circumscribed neuronal network, but if the seizure secondarily generalizes, the seizurenetwork expands and additional properties emerge. These properties may be additive with the original emergent properties.Blockade of these separable emergent properties may require drugs with different actions. Alternatively, effective anticonvulsants


may act at multiple levels. As noted above, gabapentin, for example, acts on isolated alpha(2) delta Ca2þ channels, which can beconsidered the “ground level” in the hierarchy of emergent properties affected by this agent. However, gabapentin also acts on emer-gent properties at higher organizational levels in vivo, including specific network nuclei, as seen in the studies cited above where itselectively targets PAG neurons within the AGS network and in a chronic pain network in non-epileptic animals. Identification ofhow the anticonvulsant NMAs affect the emergent properties subserving the pathophysiology of epilepsy and of chronic pain is vitalin order to identify drugs that more selectively target these properties. The studies described above apply this approach by investi-gating network mechanisms for seizure and chronic pain control and evaluating the anticonvulsant NMAs on neurons in the nucleiof these networks. These studies are yielding a detailed understanding of the way these networks function and how anticonvulsantNMAs modify these mechanisms. The results of initial studies indicate that specific sites within the intact network are most sensitiveto therapeutic doses of a given drug; this site sensitivity may be different with each specific anticonvulsant examined, since evendrugs with similar mechanisms of action can act at different brain sites, as discussed for the NMDA receptor antagonists. However,the findings with the anticonvulsant NMA, gabapentin, indicate that a specific brain nucleus, the PAG, may play a common role intreatment of both seizure and chronic pain, suggesting the possibility that this may also occur with other anticonvulsant NMAs thatare effective in multiple CNS disorders.

Acknowledgements

The author wishes to thank Gayle Stauffer for editorial assistance. Supported by CURE, the Epilepsy Foundation, and Southern Illinois UniversitySchool of Medicine.

References

Agnati, L.F., Genedani, S., Spano, P.F., Guidolin, D., Fuxe, K., 2014. Volume transmission and the Russian-doll organization of brain cell networks: aspects of their integrativeactions. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 103–119.

Becerra, L., Upadhyay, J., Chang, P.C., et al., 2013. Parallel buprenorphine phMRI responses in conscious rodents and healthy human subjects. J. Pharmacol. Exp. Ther. 345 (1),41–51.

Bialer, M., 2012. Why are antiepileptic drugs used for nonepileptic conditions? Epilepsia 53 (Suppl. 7), 26–33.Bialer, M., Johannessen, S.I., Levy, R.H., et al., 2015. Progress report on new antiepileptic drugs: a summary of the Twelfth Eilat Conference (EILAT XII). Epilepsy Res. 111, 85–141.Brozoski, T.J., Bauer, C.A., 2014. Auditory neuronal networks and chronic tinnitus. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders,

and Therapeutics. Academic Press/Elsevier, San Diego, pp. 261–275.Cai, K., Nanga, R.P., Lamprou, L., et al., 2012. The impact of gabapentin administration on brain GABA and glutamate concentrations: a 7T (1)H-MRS study.

Neuropsychopharmacology 37 (13), 2764–2771.Cheshire, W.P., 2007. Trigeminal neuralgia: for one nerve a multitude of treatments. Expert Rev. Neurother. 7 (11), 1565–1579.Faingold, C.L., 2002. Role of GABA abnormalities in the inferior colliculus pathophysiology - audiogenic seizures. Hear. Res. 168 (1–2), 223–237.Faingold, C.L., 2004. Emergent properties of CNS neuronal networks as targets for pharmacology: application to anticonvulsant drug action. Prog. Neurobiol. 72, 55–85.Faingold, C.L., 2008. Electrical stimulation therapies for CNS disorders and pain are mediated by competition between different neuronal networks in the brain. Med. Hypotheses 71,

668–681.Faingold, C.L., 2009. Anticonvulsant drugs as neuronal network-modifying agents. In: Schwartzkroin, P.A. (Ed.), Encyclopedia of Basic Epilepsy Research, vol. 1. Academic Press/

Elsevier, San Diego, pp. 50–58.Faingold, C.L., 2012. Brainstem networks: reticulo-cortical synchronization in generalized convulsive seizures. In: Noebels, J.L., Avoli, M., Rogawski, M.A., Olsen, R.W.,

Delgado-Escueta, A.V. (Eds.), Jasper’s Basic Mechanisms of the Epilepsies, fourth ed. Oxford University Press, New York, NY, pp. 257–271.Faingold, C.L., 2014a. Emergent properties of neuronal networks. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders, and Therapeutics.

Academic Press/Elsevier, San Diego, pp. 419–428.Faingold, C.L., 2014b. Neuronal network effects of drug therapies for CNS disorders. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders,

and Therapeutics. Academic Press/Elsevier, San Diego, pp. 443–465.Faingold, C.L., Blumenfeld, H., 2014a. In: Neuronal Networks in Brain Function, CNS Disorders, and Therapeutics. Academic Press/Elsevier, San Diego.Faingold, C.L., Blumenfeld, H., 2014b. Introduction to neuronal networks of the brain. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders,

and Therapeutics. Academic Press/Elsevier, San Diego, pp. 1–10.Faingold, C.L., Blumenfeld, H., 2014c. Future trends in neuronal networksdselective and combined targeting of network hubs. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal

Networks in Brain Function, CNS Disorders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 467–485.Faingold, C.L., Blumenfeld, H., July 6, 2015. Targeting neuronal networks with combined drug and stimulation paradigms guided by neuroimaging to treat brain disorders.

Neuroscientist pii:1073858415592377 (Epub ahead of print).Faingold, C.L., Naritoku, D.K., 1992. The genetically epilepsy-prone rat: neuronal networks and actions of amino acid neurotransmitters. In: Faingold, C.L., Fromm, G.H. (Eds.),

Drugs for Control of Epilepsy: Actions on Neuronal Networks Involved in Seizure Disorders. CRC Press, Boca Raton, pp. 277–308.Faingold, C.L., Tupal, S., 2014. Neuronal network interactions in the startle reflex, learning mechanisms, and CNS disorders, including sudden unexpected death in epilepsy. In:

Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 407–418.Faingold, C.L., Raisinghani, M., N’Gouemo, P., 2014a. Neuronal networks in epilepsy: comparative audiogenic seizure networks. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal

Networks in Brain Function, CNS Disorders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 349–373.Faingold, C.L., Riaz, A., Stittsworth Jr., J.D., 2014b. Neuronal network plasticity and network interactions are critically dependent on conditional multireceptive (CMR) brain regions.

In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 387–406.Faingold, C.L., Randall, M., Long, X., et al., 2015. Neurotransmitters implicated in control of sudden unexpected death in epilepsy in animal models. In: Lathers, C.M.,

Schraeder, P.L., Leestma, J.E., Wannamaker, B.B., Verrier, R.L., Schachter, S.S. (Eds.), Sudden Unexpected Death in Epilepsy: Mechanisms and New Methods for AnalyzingRisks. CRC Press/Taylor and Francis, Boca Raton, pp. 251–267.

Feng, H.J., Naritoku, D.K., Randall, M.E., Faingold, C.L., 2001. Modulation of audiogenically kindled seizures by gamma-aminobutyric acid-related mechanisms in the amygdala.Exp. Neurol. 172 (2), 477–481.

Flatters, S.J., Bennett, G.J., 2004. Ethosuximide reverses paclitaxel- and vincristine-induced painful peripheral neuropathy. Pain 109 (1–2), 150–161.


Geisler, S., Schopf, C.L., Obermair, G.J., 2015. Emerging evidence for specific neuronal functions of auxiliary calcium channel alpha(2)delta subunits. General Physiol. Biophys. 34(2), 105–118.

Governo, R.J., Morris, P.G., Marsden, C.A., Chapman, V., 2008. Gabapentin evoked changes in functional activity in nociceptive regions in the brain of the anaesthetized rat: an fMRIstudy. Br. J. Pharmacol. 153 (7), 1558–1567.

Grabenstatter, H.L., Russek, S.J., Brooks-Kayal, A.R., 2012. Molecular pathways controlling inhibitory receptor expression. Epilepsia 53 (Suppl. 9), 71–78.Iannetti, G.D., Zambreanu, L., Wise, R.G., et al., 2005. Pharmacological modulation of pain-related brain activity during normal and central sensitization states in humans. Proc. Natl.

Acad. Sci. U.S.A. 102 (50), 18195–18200.Jobe, P.C., Mishra, P.K., Dailey, J.W., 1992. Genetically epilepsy-prone rats: actions of antiepileptic drugs and monoaminergic neurotransmitters. In: Faingold, C.L., Fromm, G.H.

(Eds.), Drugs for Control of Epilepsy: Actions on Neuronal Networks Involved in Seizure Disorders. CRC Press, Boca Raton, FL, pp. 253–275.Jordan, L.M., S1awi�nska, U., 2014. The brain and spinal cord networks controlling locomotion. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS

Disorders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 215–233.Kumar, P., Kalonia, H., Kumar, A., 2012. Possible GABAergic mechanism in the neuroprotective effect of gabapentin and lamotrigine against 3-nitropropionic acid induced

neurotoxicity. Eur. J. Pharmacol. 674 (2–3), 265–274.Lakraj, A.A.D., Jabbari, B., Machado, D.G., 2014. Neuronal networks and therapeutics in neurodegenerative disorders. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks

in Brain Function, CNS Disorders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 335–348.Linnman, C., Moulton, E.A., Barmettler, G., Becerra, L., Borsook, D., 2012. Neuroimaging of the periaqueductal gray: state of the field. Neuroimage 60 (1), 505–522.Matta, J.A., Ashby, M.C., Sanz-Clemente, A., Roche, K.W., Isaac, J.T., 2011. mGluR5 and NMDA receptors drive the experience- and activity-dependent NMDA receptor NR2B to

NR2A subunit switch. Neuron 70 (2), 339–351.Mehta, S., McIntyre, A., Dijkers, M., Loh, E., Teasell, R.W., 2014. Gabapentinoids are effective in decreasing neuropathic pain and other secondary outcomes after spinal cord injury:

a meta-analysis. Archiv. Phys. Med. Rehab. 95 (11), 2180–2186.Melzack, R., Wall, P.D., 1965. Pain mechanisms: a new theory. Science 150 (3699), 971–979.Narahashi, T., Zhao, X., Ikeda, T., Nagata, K., Yeh, J.Z., 2007. Differential actions of insecticides on target sites: basis for selective toxicity. Hum. Exp. Toxicol. 26 (4), 361–366.N’Gouemo, P., Garcia-Cairasco, N., Faingold, C.L., 2014. Physiological and pathophysiological expansion of neuronal networks. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal

Networks in Brain Function, CNS Disorders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 375–385.Naritoku, D.K., Mecozzi, L.B., Aiello, M.T., Faingold, C.L., 1992. Repetition of audiogenic seizures in genetically epilepsy-prone rats induces cortical epileptiform activity and

additional seizure behaviors. Exp. Neurol. 115 (3), 317–324.Penfield, W., Jasper, H.H., 1954. Epilepsy and the Functional Anatomy of the Human Brain. Little, Brown & Co., Oxford.Pothmann, L., Muller, C., Averkin, R.G., Bellistri, E., Miklitz, C., et al., 2014. Function of inhibitory micronetworks is spared by Naþ channel-acting anticonvulsant drugs. J. Neurosci.

34 (29), 9720–9735.Richerson, G.B., Wu, Y., 2004. Role of the GABA transporter in epilepsy. Adv. Exp. Med. Biol. 548, 76–91.Rogawski, M.A., Bazil, C.W., 2008. New molecular targets for antiepileptic drugs: alpha(2)delta, SV2A, and K(v)7/KCNQ/M potassium channels. Curr. Neurol. Neurosci. Rep. 8 (4),

345–352.Roy, M., Shohamy, D., Daw, N., Jepma, M., Wimmer, G.E., et al., 2014. Representation of aversive prediction errors in the human periaqueductal gray. Nat. Neurosci. 17 (11),

1607–1612.Samineni, V., Premkumar, L., Faingold, C.L., 2012. Chronic pain induces neuroplastic changes in periaqueductal gray neuronal firing: blockade by gabapentin. Soc. Neurosci.

Abstract 178.11/GG2.Samineni, V.K., Premkumar, L.S., Faingold, C.L., 2011. Post-ictal analgesia in genetically epilepsy-prone rats is induced by audiogenic seizures and involves cannabinoid receptors

in the periaqueductal gray. Brain Res. 1389, 177–182.Schafe, G.E., 2014. The fear memory network. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS Disorders, and Therapeutics. Academic Press/

Elsevier, San Diego, pp. 167–178.Stam, C.J., 2014. Modern network science of neurological disorders. Nat. Rev. Neurosci. 15 (10), 683–695.Sukhotinsky, I., Devor, M., 2014. Networks for the modulation of acute and chronic pain. In: Faingold, C.L., Blumenfeld, H. (Eds.), Neuronal Networks in Brain Function, CNS

Disorders, and Therapeutics. Academic Press/Elsevier, San Diego, pp. 311–326.Takemura, Y., Yamashita, A., Horiuchi, H., Furuya, M., Yanase, M., et al., 2011. Effects of gabapentin on brain hyperactivity related to pain and sleep disturbance under

a neuropathic pain-like state using fMRI and brain wave analysis. Synapse 65 (7), 668–676.Thurman, D.J., Hesdorffer, D.C., French, J.A., 2014. Sudden unexpected death in epilepsy: assessing the public health burden. Epilepsia 55 (10), 1479–1485.Tupal, S., Faingold, C.L., 2012. The amygdala to periaqueductal gray pathway: plastic changes induced by audiogenic kindling and reversal by gabapentin. Brain Res. 1475, 71–79.Ueda, H., 2006. Molecular mechanisms of neuropathic pain-phenotypic switch and initiation mechanisms. Pharmacol. Ther. 109 (1–2), 57–77.Wager, T.D., Atlas, L.Y., Lindquist, M.A., Roy, M., Woo, C.W., et al., 2013. An fMRI-based neurologic signature of physical pain. N. Engl. J. Med. 368 (15), 1388–1397.Yared, J.A., Tkaczuk, K.H., 2012. Update on taxane development: new analogs and new formulations. Drug Des. Dev. Ther. 6, 371–384.Zhang, J.L., Yang, J.P., Zhang, J.R., Li, R.Q., Wang, J., et al., 2013. Gabapentin reduces allodynia and hyperalgesia in painful diabetic neuropathy rats by decreasing expression

level of Nav1.7 and p-ERK1/2 in DRG neurons. Brain Res. 1493, 13–18.


anticonvulsant drugs are neuronal network-modifying agents...

Documents