lighting the chandelier: new vistas for axo-axonic cells

INMED/TINS special issue

Lighting the chandelier:new vistas for axo-axonic cellsAllyson Howard1, Gabor Tamas2 and Ivan Soltesz1

1Department of Anatomy and Neurobiology, University of California, Irvine, CA 92697, USA2Department of Comparative Physiology, University of Szeged, Szeged, Hungary H-6726

Chandelier or axo-axonic cells are the most selective of

all cortical GABAergic interneurons, because they exclu-

sively contact axon initial segments of cortical glutama-

tergic neurons. Owing to their privileged location on

initial segments, axo-axonic cells have often been

assumed to have the ultimate control of pyramidal cell

output. Recently, key molecules expressed at the initial-

segment synapses have been identified, and novel

in vitro and in vivo electrophysiological studies have

revealed unexpectedly versatile functional effects

exerted by axo-axonic cells on their postsynaptic

targets. In addition, there is also emerging recognition

of the mechanistic involvement of these unique cells in

several neurological diseases, including epilepsy and

schizophrenia.

Introduction

In his pioneering studies of neuroanatomy, SantiagoRamon y Cajal identified many major interneuronsubtypes, yet never explicitly described the characteristic

**

AIS

(a) (b)

Figure 1. Chandelier cell morphology and axon initial segment (AIS) labeling. (a) Camer

hippocampus, illustrating the many characteristic cartridges that give the chandelier cell

likely to be contacted by this chandelier cell. Scale bar, 50 mm. (b) A Golgi-stained CA1 py

Boutons are represented by small arrows. Scale bar, 10 mm. (c) Light micrographs of the

channel subunit KCNQ2 (green) and ankyrin G (red) to the pyramidal cell AIS. Dark, unstai

lacunosum-moleculare; SO, stratum oriens; SP, stratum pyramidale; SR, stratum radiatu

with permission, from Ref. [4] q (1983) the Society for Neuroscience; (c) adapted, with

Corresponding author: Howard, A. ([email protected]).Available online 26 April 2005

www.sciencedirect.com 0166-2236/$ - see front matter Q 2005 Elsevier Ltd. All rights reserved

cartridges of axo-axonic cells (AACs). Chandelier cells, sonamed by Janos Szentagothai because these cartridges(vertical rows of axonal terminals) resemble rows ofcandles on a chandelier (Figure 1a,b), were discoveredonly 30 years ago [1,2]. Shortly thereafter, in Golgi-stainedtissue from rat cortex, it was recognized that chandeliercells are the source of the previously mysterious axo-axonic boutons [3] and are not dendritically projecting, aspreviously believed. A few years later, the GABAergic, andtherefore inhibitory, nature of these cells was verified [4].This short but distinguished heritage gave rise to a seriesof recent breakthroughs that underlie the unique natureof these elusive interneurons.

Location and basic properties

Curiously, nearly all AACs are located in layered corticalareas – the neocortex or allocortex [3,5–7]. However, AACcartridges have also been found throughout the amygdala,a partly cortical structure that is not laminated [8], andAACs do not exist in the cerebellar cortex, despite its

Review TRENDS in Neurosciences Vol.28 No.6 June 2005

KCNQ2

SR

SLM

SO

SP

SR

SLM

SO

SP

CA3

Ankyrin G

(c)

a lucida drawing of a Golgi-stained chandelier cell from the CA1 region of monkey

its name. The asterisks mark representative pyramidal cell somata whose axons are

ramidal cell, with its AIS contacted by a Golgi-stained axon from an axo-axonic cell.

CA3 pyramidal cell layer of rat hippocampus demonstrating colocalization of the KC

ned circles represent pyramidal cell somata. Additional abbreviations: SLM, stratum

m. Scale bar, 20 mm. Panel (a) adapted, with permission, from Ref. [15]; (b) adapted,

permission, from Ref. [44] q (2004) the Society for Neuroscience.

. doi:10.1016/j.tins.2005.04.004

http://www.sciencedirect.com

Review TRENDS in Neurosciences Vol.28 No.6 June 2005 311

layered nature. Because the home of AACs, the cerebralcortex, is evolutionarily recent, it is perhaps no coinci-dence that AACs have been identified so far only inmammals, including rats [3], guinea pigs [9], cats [10],rabbits [11], mice [12], ferrets [13], monkeys [14,15] andhumans [16].

The first anatomical studies of chandelier cells revealedfusiform somata, and dendrites that are usually clusteredin a bitufted manner, parallel to pyramidal cell apicaldendrites. However, a recent report revealed a potentiallydistinct subset of horizontal AACs in CA1 whose dendritesare confined to the stratum oriens [17]; these AACs mightreceive different inputs to their vertically orientedcounterparts. All AAC axons originate from the soma ora proximal dendrite, receive no synapses on their ownaxon initial segments (AISs) and form a dense network ofterminal cartridges 10–50 mm long, each containing 2–12boutons [14,18,19] (Figure 1a,b). A characteristic, but notfunctionally understood, feature of these cartridges isthat they frequently climb upwards along the AIS – thatis, the proximal bouton targets the distal end of the AIS[10,14,18]. Single AACs have been shown to innervatebetween 250 (in cortex) [14] and 1200 (in hippocampus)[19] pyramidal cells, indicating the potential to synchron-ize many cells.

Identification, please: markers for AACs

As the field of interneuron diversity developed, bio-chemical markers were discovered as distinguishingcharacteristics. The first marker found for AACs was theCa2C-binding protein parvalbumin (PV), expressed innearly every AAC [5,20]. Unfortunately, PV is not specificto AACs – it is also expressed by some basket cells. AACcartridges are reliably stained by the high-affinity GABAtransporter GAT-1, but identification relies on the pres-ence of characteristic cartridges because GAT-1 labelsnearly all GABAergic boutons [21]. Without a specificmarker, studies of AACs must rely on unambiguousmeasurements, such as electron-microscopic verificationof synaptic localization. Even within interneuron sub-types, there is variability that has important effects onpyramidal cell output [22]. AACs are no exception to thisrule of heterogeneity: their axonal and dendritic fields canvary as already described, and some AAC cartridgesexpress additional proteins, such as corticotropin-releasingfactor (CRF) in squirrel monkey neocortex [23,24], poly-sialated neural cell adhesion molecule (PSA-NCAM) inhuman temporal cortex and subiculum [25,26] andcalbindin in human temporal cortex [27].

Picking partners: postsynaptic targets of AACs

Although other cell types show target specificity, AACsoutshine them all with their remarkable precision. Innearly all quantitative studies, regardless of species orbrain area, every chandelier cell terminal synapses withthe AISs of pyramidal cells, granule cells or mossy cells[3,7,10,14,16,28–30]. In the few studies that could notconfirm total specificity, the percentage of synapses ontoAISs was never !90% [31–33]. It could be important thatthese AACs with aberrant targets were all found inrats from a relatively deprived, standard laboratory

www.sciencedirect.com

environment [34], emphasizing the need to investigateeffects of enriched environments on interneuron speci-ficity. AACs have never been demonstrated to target otherinterneurons [14,16] and, although some interneuronsreceive a few synapses on their AIS, the parent cellsremain unidentified [35]. A final layer of discrimination isthat AACs project to preferred subsets of pyramidal cells.In visual cortex, AACs target ten times more cortico-cortical pyramidal cells than corticothalamic cells [36].Most AAC cartridges in human entorhinal cortex arelocated in layers 2–3, the site of projection neurons to thehippocampus [25]. With such location-specificity of theirpostsynaptic targets, their presynaptic inputs might alsobe expected to be precisely tuned. Unfortunately, data onAAC afferents are limited – most studies rely solely onparvalbumin staining and therefore cannot distinguishAACs from basket cells.

Sixteen candles: development of a chandelier

Studying development of specific interneuron subtypes isnot trivial because identification relies on Ca2C-binding-protein expression and neurite arborization patterns,which often appear relatively late in development. Astudy in monkeys of AAC cartridges, visualized usingGolgi staining, PV-immunoreactivity or GAT-1-immunor-eactivity, indicated a biphasic developmental course,increasing from a few stained cartridges at birth to amaximum during adolescence, and then possibly decreas-ing again to a lower level in adulthood [37]. Althoughprecise details of synaptic arborization remain unclear,AACs do not fully develop until well after birth, and theymature later than other interneuron types. For example,morphologically mature AAC cartridges are not observedin cat cortex until postnatal weeks 4–6, well after theappearance of basket cell terminals [38]. This latedevelopmental appearance of AACs could explain whyCajal never encountered these cells, because he often usedtissue from young animals. However, some cells hedescribed as atypical basket cells might be immatureAACs [39]. An electron-microscopic time-course study ofsynapse number on pyramidal cell AISs is needed toquantify AAC synaptic development without relying onthe confounding factors of protein expression.

The other side of the synapse: molecular layout of the

AIS

The AAC cannot be studied without considering its targetthe AIS, and the anatomy of the AIS elucidates its func-tion. This axonal structure is precisely defined and exhibitsunique structural features on electronmicrographs, includ-ing microtubule clusters, a membrane undercoating,clusters of ribosomes, and stacks of Ca2C-storing cisternae[40,41]. The neuronal cytoskeleton, consisting of inter-acting spectrins and actins, forms the neuronal structuralscaffold and is a spatial delimiter for neuronal membraneproteins [42]; the membrane undercoating is a specializedcytoskeletal element, found only in the AIS. The bIVisoform of spectrin (an actin-binding protein) and ankyrinG (a spectrin-binding protein) mutually confine each otherto the AIS [43]. Ankyrin G provides a specific anchor formany AIS-specific proteins, including the KC channel



subunits KCNQ2 and KCNQ3 [44] (Figure 1c); thesechannels can assemble heteromerically, underlie themuscarine-sensitive KC current IM, and might also beresponsible for the slowly activating current Ks. GABAA

receptors are present postsynaptic to AAC boutons, with aparticularly high concentration of the a2 subunit [45,46].AISs of some pyramidal cells in monkey and humantemporal neocortex also express 5-HT1A receptors [47].Molecules localized to the presynaptic terminal includeGAT-1, M2 ACh receptors (in hippocampus) [48] andPSA-NCAM (in human neocortex and subiculum) [25,26](Figure 2, left). Additionally, there are often punctaadherentia, providing more stabilization to the AAC–AISinterface.

Unfortunately, most other studies of protein localiz-ation to the AIS have not looked at cortical pyramidalcells. For instance, ankyrin G is responsible for tetheringthe NaC channel subunit Nav1.6 to the cerebellarPurkinje cell AIS [49] and Nav1.2 specifically localizes tothe AIS in retinal ganglion cells [50]. The Purkinje cell AISalso expresses the cell adhesion molecules NrCAM andneurofascin, where neurofascin binds to ankyrin G andhas an essential role in guiding the basket cell axon to theAIS [51]. Although these data from Purkinje cells andganglion cells are interesting, these cells are not goodmodels for studying the AAC–AIS interface. Purkinje cellsare not only GABAergic but also their AIS is ensheathedwith glia and receives no or at most two synapses fromsurrounding basket cell pinceaux (‘paintbrushes’) [52](Figure 2, right). Retinal ganglion cell AISs are likewisecovered by glia and are completely devoid of synapses.Thus, despite some similarities among all AISs, such asthe presence of ankyrin G and bIV spectrin, there are

AACterminal

Pyramidalcell AIS

Pyramidal cell soma

Purkinjecell AIS

Purkinjecell soma

Gliacell

Gliacel

Figure 2. Axon initial segment (AIS) specializations: the differences between the axo-axon

synapse (right). Some molecules are present only in subsets of AAC boutons, such as P

human temporal neocortex [47] and M2 ACh receptors in rat hippocampus [48]. Imp

neurofascin are only known to be localized at the Purkinje cell AIS [49,51]. The different m

case of the cerebellum are also illustrated. Note that AACs provide multiple synapses e

mostly to Purkinje cell somata, and at most two synapses to the Purkinje cell AIS.


significant anatomical differences, and therefore futurestudies must verify the localization of specific NaC

channel subunits and adhesionmolecules on the pyramidalcell AIS, where AAC synapses actually form.

Chandelier cell function: more than simply inhibitory

Recently, great strides have been made in under-standing physiological roles of AACs. Neocortical andhippocampal AACs, identified by electron microscopy asshown in Figure 3(a), display fast-spiking characteristicsand various degrees of spike frequency accommodation,with the decrement in firing rate varying between2.5% and 81.0% in response to a 500 ms depolarization[13,17,28,30,33,53–55] (Figure 3b). As expected, firing ofAACs evokes inhibitory postsynaptic events in pyramidalcells that are mediated by GABAA receptors and modu-lated by presynaptic GABAB receptors [32]. This strongconnection exhibits a failure rate of only 0–10%, pre-sumably owing to the multiple boutons between each AACand the postsynaptic pyramidal cell [17,53].

AACs also contribute to synchrony and oscillations inthe hippocampus such as the theta rhythm (4–8 Hz),which occurs during environmental exploration and REMsleep, and fast ripples (120–200 Hz), which occur duringslow-wave sleep. In paired recordings, an AAC firing attheta frequency immediately entrains the postsynapticpyramidal cell to its rhythm, causing it to fire antiphase tothe AAC [29]. In vivo, AACs fire consistently at 1858 of thenaturally occurring extracellularly recorded thetarhythm, antiphase to pyramidal cells (Figure 3c).Additionally, the basal firing rate of AACs increasesO100% during theta oscillations [56], suggesting thatAACs have more than a simple inhibitory function.

TRENDS in Neurosciences

GAT-1

Ankyrin G

βIV spectrin

Synaptic vesiclecontaining GABA

GABAA receptor, enriched in α2

Nav 1.6

NrCAM or neurofascin

PSA-NCAM

K+ channel:KCNQ2/3 heteromer

Key

5-HT1A receptor

M2 receptor

Cerebellarbasket

cellpinceau

l

ll

ic cell (AAC)–pyramidal cell synapse (left) and the cerebellar basket cell–Purkinje cell

SA-NCAM in human temporal cortex and subiculum [25,26], 5-HT1A in monkey and

ortantly, the NaC channel Nav1.6 and the cell adhesion molecules NrCAM and

orphologies of the two types of presynaptic terminals and the presence of glia in the

xclusively to pyramidal cell AISs, whereas cerebellar basket cells provide synapses


0.04

0.02

0.00

0.40

0.00

0.05

0.00

0.00

0.10Firi

ng p

roba

bilit

y

–1 0 1Normalized time

0 360 720Theta phase

Fast ripples(a) (b) (i) (c) (i) (d) (i)Theta waves

ExtracellularIntracellular

Pyramidal cells

Axo-axoniccells

(ii)

(ii) (ii)

Figure 3. Chandelier cell ultrastructure and physiology. (a) Electron-micrographs of two of the multiple boutons (labeled b1 and b2,3) from an electrophysiologically recorded,

biocytin-filled axo-axonic cell (AAC) synapsing on the biocytin-filled axon initial segment (ais) of a postsynaptic pyramidal cell. Arrows indicate synaptic contacts. Note the

morphological differences between the boutons. (b) Firing pattern of two AACs from rat hippocampus in response to a 500 ms depolarizing pulse. Note that (i) shows a non-

adapting pattern, whereas (ii) shows a strongly-adapting pattern, demonstrating that AAC physiology is heterogeneous, even within one species and brain area. (c,d) Theta

waves (c) and fast ripples (d) in the stratum pyramidale. Extracellularly recorded theta waves and fast ripples are shown in the upper parts of (c,i) and (d,i), respectively, above

the simultaneous juxtacellular spike rasters of single AACs in each case. (c,ii) Stratum pyramidale extracellular (black) and pyramidal cell intracellular (blue) theta rhythms,

corresponding to histograms of spike firing probability in pyramidal cells (blue) and AACs (red), averaged over several cells. Note that AAC firing is antiphase to pyramidal

cell intracellular theta waves and firing, but only slightly phase-delayed from extracellular theta waves. (d,ii) Extracellular fast ripples, and concomitant spike firing probability

in pyramidal cells (blue) and AACs (red), averaged over several cells. Note that AACs increase firing probability just before the fast ripple, and depress firing after the fast

ripple. Scale bars in (c,i): unit, 0.5 mV; theta, 0.2 mV; time, 300 ms. Scale bars in (d,i): unit, 0.5 mV; fast ripple, 0.1 mV; time, 50 ms. Panel (a) adapted, with permission of

Oxford University Press, from Ref. [30]; (b) adapted, with permission, from Ref. [28]; (c) and (d) adapted, with permission, from Ref. [56].


Chandelier cells fire just before the onset of fast ripplesand then are quiescent during and after this activity [56](Figure 3d). An interesting question is whether AACs areamong the gap-junction-coupled PV-expressing cells [57],because electrical connectivity might add to their oscil-latory effectiveness.

For many years, it was presumed that the role of AACswas to gate pyramidal cell output, shutting down the cellto prevent unnecessary firing. This theory was based onthe assumption that the AIS is the initiation site of actionpotentials, owing to a high density of NaC channels in thisregion [58] and to results showing that the action potentialinitiates at some point in the axon [59]. Along these lines,some data suggest that neocortical AACs dampen exces-sive excitability. In response to whisker stimulationin vivo, AACs do not receive fast sensory input but, rather,a long-latency sequence consisting of an inhibitorypostsynaptic potential (IPSP) followed by an excitatorypostsynaptic potential (EPSP), corresponding to the acti-vation of local cortical circuits. In addition, the spon-taneous firing rate of AACs is much lower than that ofother interneurons in control conditions, but significantlyhigher without GABA-mediated inhibition. Accordingly,AACs could monitor local GABA activity, and only activatewhen this activity is insufficient [54]. This mechanism issupported by the nearly linear summation rules detectedfor interacting axo-axonic inputs, maintaining the impactof the 3–6 individual AACs converging on a single targetpyramidal cell [30].

New results suggest that action potential initiationoccurs distal to the AIS, at least 30 mm from the soma, and


perhaps as far as the first node of Ranvier [60,61]. There-fore, the high density of NaC channels in the AIS couldbackpropagate the action potential into the soma [62]. Inthis case, AACs could effectively decouple suprathresholdoperations between the axon and the soma by regulatingNaC channel activation in the AIS. This mechanismmightbe directionally selective, by blocking axon-to-soma actionpotential propagation efficiently, owing to the character-istic arrangement of proximal boutons on the distal AISalready described. Because precise measurement of anaction potential initiation site can be complex technically,computational modeling could offer a solution. Onecomputational study suggests that action potentialthreshold in the AIS and axon depends on axial resistivity,length, diameter and NaC channel density in this segment[63]. For further predictions from computational models,more precise measurements of NaC channel subunitspecificity and density in the pyramidal cell AIS areneeded.

Chandelier cells in disease states

Given the specificity and functional roles of AACs, adisturbance of these cells would be expected to causemajor problems for normal function. In fact, changes inAACs have been reported in both epilepsy and schizo-phrenia. Many studies of both diseases have shownvarious amounts of decreased PV-immunoreactivity ormRNA expression [64–68], but this measure is not specificfor AACs, nor does it necessarily signal cell death ofPV-positive cells. However, some specific changes suggest



a role for chandelier cell dysregulation in the pathology ofepilepsy and schizophrenia.

Various synaptic rearrangements of AACs have beenobserved in epilepsy, from losses of synapses on AISs toincreases in their number (Figure 4). Synapses can be lostfrom the pyramidal cell AIS in animal models of epilepsy,such as in the somatosensory cortex of monkeys withalumina-gel-induced epilepsy [69] and in CA1 of rats withpilocarpine-induced seizures [70]. However, electron-micro-scopic studies on human tissue from temporal lobe epilepsypatients indicateno change in synapsenumber on survivingCA1 pyramidal cells, and even show an increase in synapsenumber on theAISs of dentate granule cells [71,72]. Finally,human epileptic tissue displays de novo expression ofcalbindin in hippocampal AAC cartridges [26]. Theseseemingly contradictory changes indicate that, althoughreorganization of AACs might contribute to epilepsy, we donot fully understand their precise relevance.

Contr

Epilepsy

Increas

Humandentate G

Decrease

PilocarpineCA1

None

MonkeySS pyramidal

HumanCA1

PV

GAD67

GAT-1

PV

GAD67

GAT-1

Decrease

HippocampusNeocortex

AAC

AAC

Figure 4. Axo-axonic cells (AACs) in disease states. Normal AACs express the follo

decarboxylase (GAD)67 and the high-affinity GABA transporter GAT-1. They also expres

cell) axon initial segment (AIS) can be visualized using GABAA a2-immunoreactivity. In e

brain area, species and cell type. In the neocortex, numbers of synapses onto pyramidal

gel-induced epilepsy [69]. In the hippocampus, numbers of synapses onto CA1 pyramida

but numbers onto CA1 pyramidal cells were decreased in rats with pilocarpine-induced e

with temporal lobe epilepsy [71]. In schizophrenia, there are many reported changes in m

mRNA [65], GAD67 mRNA and GAT-1 mRNA [73,75]. Additionally, the number of GAT-1-

a2-immunoreactive synapses is increased [75], possibly indicating an increase in overa


The prefrontal cortex is essential to normal workingmemory, which is disrupted in schizophrenia. Schizo-phrenia is a neurodevelopmental disorder that presents inlate adolescence, when the density of AAC cartridges ischanging. Loss of AAC function in schizophrenic prefron-tal cortex would be significant because no pyramidal cellloss occurs [73]. No counts have been done of AIS synapsenumber in schizophrenic tissue, to measure AAC cartridgeloss directly. However, there are many protein and mRNAexpression changes that are consistent with AAC dys-function (Figure 4). Specifically, the number of GAT-1-immunoreactive cartridges decreases [74], and thenumber of GABAA-a2-receptor-positive AISs increases[75]. Finally, levels of mRNA for PV, GAT-1 and theGABA synthetic enzyme GAD67 have been reported todecrease [73,75]. Functionally, this could correspond todecreased GABA production, which spurs a decrease inpresynaptic GABA reuptake to allow GABA to remain in

TRENDS in Neurosciences

Schizophrenia

Postsynapticcell

GAT-1 GABAA α2

ol

Number of GAT-1-IR cartridges

e

C

GAD67mRNA

PVmRNA

GAT-1mRNA

Number of GABAA

α2 AIS

AAC

wing mRNA and proteins in their cell bodies: parvalbumin (PV), glutamic acid

s GAT-1 protein in their cartridges. The normal postsynaptic (granule or pyramidal

pilepsy, three different fates have been observed for AAC cartridges, depending on

cell AISs were reduced in the somatosensory cortex (SS) of monkeys with alumina-

l cells were unchanged in CA1 tissue from humans with temporal lobe epilepsy [72],

pilepsy [70] and were increased in dentate granule cells (CG) in tissue from humans

RNA expression at the cellular level of the AAC, such as reduced expression of PV

immunoreactive (GAT-1-IR) cartridges is decreased [74], and the number of GABAA

ll expression of this receptor.



the synapse, and an increase in postsynaptic GABAreceptors so that the lower level of GABA can be detectedby the pyramidal cell [75].

Future directions

A wealth of recent information about anatomy andphysiology of AACs and the AIS indicates that theseassociated structures play an important part in corticalnetwork activity. However, more work is needed tounderstand fully the role of AACs. For example, synapticinnervation of the pyramidal cell AIS needs to be studiedusing electron microscopy over the course of postnataldevelopment, and in both control and schizophrenictissues, to verify changes seen in protein expressionstudies. Given the inherent limitations of the Purkinjecell AIS model system, the pyramidal cell AIS will need tobe examined further for the presence of specific NaC

channel isoforms and cell adhesion molecules, and for itsrole in action potential generation or backpropagation.Detailed neurochemical and electrophysiological analysisof intracortical and subcortical afferents will be essentialfor better understanding of AAC function. Finally, pairedrecordings combined with imaging will be crucial indetermining the effect of AACs on distinct subcellulardomains of pyramidal cells, and on populations ofpyramidal cells in a variety of brain states and areas.

Acknowledgements

This work was supported by the NIH (grant NS38580 to I.S.).

References

1 Szentagothai, J. and Arbib, M.A. (1974) Conceptual models of neuralorganization. Neurosci. Res. Program Bull. 12, 305–510

2 Jones, E.G. (1975) Varieties and distribution of non-pyramidal cells inthe somatic sensory cortex of the squirrel monkey. J. Comp. Neurol.160, 205–267

3 Somogyi, P. (1977) A specific ‘axo-axonal’ interneuron in the visualcortex of the rat. Brain Res. 136, 345–350

4 Somogyi, P. et al. (1983) Glutamate decarboxylase immunoreactivityin the hippocampus of the cat: distribution of immunoreactivesynaptic terminals with special reference to the axon initial segmentof pyramidal neurons. J. Neurosci. 3, 1450–1468

5 Schmidt, S. et al. (1993) Parvalbumin-immunoreactive structures ofthe adult human entorhinal and transentorhinal region. Hippo-campus 3, 459–470

6 Kosaka, T. (1983) Axon initial segments of the granule cell in the ratdentate gyrus: synaptic contacts on bundles of axon initial segments.Brain Res. 274, 129–134

7 Soriano, E. and Frotscher, M. (1989) A GABAergic axo-axonic cell inthe fascia dentata controls the main excitatory hippocampal pathway.Brain Res. 503, 170–174

8 Kemppainen, S. and Pitkanen, A. (2000) Distribution of parvalbumin,calretinin, and calbindin-D(28k) immunoreactivity in the rat amyg-daloid complex and colocalization with gamma-aminobutyric acid.J. Comp. Neurol. 426, 441–467

9 Gulyas, A.I. et al. (1993) Precision and variability in postsynaptictarget selection of inhibitory cells in the hippocampal CA3 region. Eur.J. Neurosci. 5, 1729–1751

10 Fairen, A. and Valverde, F. (1980) A specialized type of neuron in thevisual cortex of cat: a Golgi and electron microscope study ofchandelier cells. J. Comp. Neurol. 194, 761–779

11 Muller-Paschinger, I.B. et al. (1983) Chandelier neurons within therabbits’ cerebral cortex. A Golgi study. Anat. Embryol. (Berl.) 166,149–154

12 Chiu, C.S. et al. (2002) Number, density, and surface/cytoplasmic


distribution of GABA transporters at presynaptic structures of knock-in mice carrying GABA transporter subtype 1-green fluorescentprotein fusions. J. Neurosci. 22, 10251–10266

13 Krimer, L.S. and Goldman-Rakic, P.S. (2001) Prefrontal microcircuits:membrane properties and excitatory input of local, medium, and widearbor interneurons. J. Neurosci. 21, 3788–3796

14 Somogyi, P. et al. (1982) The axo-axonic interneuron in the cerebralcortex of the rat, cat and monkey. Neuroscience 7, 2577–2607

15 Somogyi, P. et al. (1983) A new type of specific interneuron in themonkey hippocampus forming synapses exclusively with the axoninitial segments of pyramidal cells. Brain Res. 259, 137–142

16 Kisvarday, Z.F. et al. (1986) Synaptic connections of axo-axonic(chandelier) cells in human epileptic temporal cortex. Neuroscience19, 1179–1186

17 Ganter, P. et al. (2004) Properties of horizontal axo-axonic cells instratum oriens of the hippocampal CA1 area of rats in vitro.Hippocampus 14, 232–243

18 Peters, A. et al. (1982) Chandelier cells in rat visual cortex. J. Comp.Neurol. 206, 397–416

19 Li, X.G. et al. (1992) Axonal and dendritic arborization of anintracellularly labeled chandelier cell in the CA1 region of rathippocampus. Exp. Brain Res. 90, 519–525

20 DeFelipe, J. et al. (1989) Visualization of chandelier cell axons byparvalbumin immunoreactivity in monkey cerebral cortex. Proc. Natl.Acad. Sci. U. S. A. 86, 2093–2097

21 DeFelipe, J. and Gonzalez-Albo, M.C. (1998) Chandelier cell axons areimmunoreactive for GAT-1 in the human neocortex. NeuroReport 9,467–470

22 Aradi, I. and Soltesz, I. (2002) Modulation of network behaviourby changes in variance in interneuronal properties. J. Physiol.

538, 227–25123 Lewis, D.A. and Lund, J.S. (1990) Heterogeneity of chandelier

neurons in monkey neocortex: corticotropin-releasing factor- andparvalbumin-immunoreactive populations. J. Comp. Neurol. 293,599–615

24 Lewis, D.A. et al. (1989) Corticotropin-releasing factor immunoreac-tivity in monkey neocortex: an immunohistochemical analysis.J. Comp. Neurol. 290, 599–613

25 Arellano, J.I. et al. (2002) PSA-NCAM immunoreactivity in chandeliercell axon terminals of the human temporal cortex. Cereb. Cortex 12,617–624

26 Arellano, J.I. et al. (2004) Histopathology and reorganization ofchandelier cells in the human epileptic sclerotic hippocampus. Brain127, 45–64

27 del Rio, M.R. and DeFelipe, J. (1997) Colocalization of parvalbuminand calbindin D-28k in neurons including chandelier cells of thehuman temporal neocortex. J. Chem. Neuroanat. 12, 165–173

28 Buhl, E.H. et al. (1994) Physiological properties of anatomicallyidentified axo-axonic cells in the rat hippocampus. J. Neurophysiol.71, 1289–1307

29 Cobb, S.R. et al. (1995) Synchronization of neuronal activity in hippo-campus by individual GABAergic interneurons. Nature 378, 75–78

30 Tamas, G. andSzabadics, J. (2004) Summation of unitary IPSPs elicitedby identified axo-axonic interneurons. Cereb. Cortex 14, 823–826

31 Soriano, E. et al. (1990) Axo-axonic chandelier cells in the rat fasciadentata: Golgi-electron microscopy and immunocytochemical studies.J. Comp. Neurol. 293, 1–25

32 Buhl, E.H. et al. (1994) Diverse sources of hippocampal unitaryinhibitory postsynaptic potentials and the number of synaptic releasesites. Nature 368, 823–828

33 Kawaguchi, Y. and Kubota, Y. (1998) Neurochemical features andsynaptic connections of large physiologically-identified GABAergiccells in the rat frontal cortex. Neuroscience 85, 677–701

34 Somogyi, P.P. and Klausberger, T. (2005) Defined types of corticalinterneuron structure space and spike timing in the hippocampus.J. Physiol. 562, 9–26

35 Matyas, F. et al. (2004) Convergence of excitatory and inhibitoryinputs onto CCK-containing basket cells in the CA1 area of the rathippocampus. Eur. J. Neurosci. 19, 1243–1256

36 Farinas, I. and DeFelipe, J. (1991) Patterns of synaptic input oncorticocortical and corticothalamic cells in the cat visual cortex. II. Theaxon initial segment. J. Comp. Neurol. 304, 70–77



37 Cruz, D.A. et al. (2003) Postnatal development of pre- and postsyn-aptic GABA markers at chandelier cell connections with pyramidalneurons in monkey prefrontal cortex. J. Comp. Neurol. 465, 385–400

38 Alcantara, S. and Ferrer, I. (1994) Postnatal development ofparvalbumin immunoreactivity in the cerebral cortex of the cat.J. Comp. Neurol. 348, 133–149

39 DeFelipe, J. and Jones, E.G., eds (1988) Cajal on the Cerebral Cortex:an Annotated Translation of the Complete Writings, Oxford UniversityPress

40 Palay, S.L. et al. (1968) The axon hillock and the initial segment.J. Cell Biol. 38, 193–201

41 Benedeczky, I. et al. (1994) The cisternal organelle as a Ca2C-storingcompartment associated with GABAergic synapses in the axon initialsegment of hippocampal pyramidal neurones. Exp. Brain Res. 101,216–230

42 Winckler, B. et al. (1999) A diffusion barrier maintains distribution ofmembrane proteins in polarized neurons. Nature 397, 698–701

43 Komada, M. and Soriano, P. (2002) bIV-spectrin regulates sodiumchannel clustering through ankyrin-G at axon initial segments andnodes of Ranvier. J. Cell Biol. 156, 337–348

44 Devaux, J.J. et al. (2004) KCNQ2 is a nodal KC channel. J. Neurosci.24, 1236–1244

45 Nusser, Z. et al. (1996) Differential synaptic localization of two majorg-aminobutyric acid type A receptor a subunits on hippocampalpyramidal cells. Proc. Natl. Acad. Sci. U. S. A. 93, 11939–11944

46 Nyiri, G. et al. (2001) Input-dependent synaptic targeting of a2-subunit-containingGABAAreceptors in synapses ofhippocampalpyramidal cellsof the rat. Eur. J. Neurosci. 13, 428–442

47 DeFelipe, J. et al. (2001) Pyramidal cell axons show a localspecialization for GABA and 5-HT inputs in monkey and humancerebral cortex. J. Comp. Neurol. 433, 148–155

48 Hajos, N. et al. (1998) Distinct interneuron types express M2muscarinic receptor immunoreactivity on their dendrites or axonterminals in the hippocampus. Neuroscience 82, 355–376

49 Jenkins, S.M. and Bennett, V. (2001) Ankyrin-G coordinates assemblyof the spectrin-based membrane skeleton, voltage-gated sodiumchannels, and L1 CAMs at Purkinje neuron initial segments. J. CellBiol. 155, 739–746

50 Boiko, T. et al. (2003) Functional specialization of the axon initialsegment by isoform-specific sodium channel targeting. J. Neurosci. 23,2306–2313

51 Ango, F. et al. (2004) Ankyrin-based subcellular gradient of neuro-fascin, an immunoglobulin family protein, directs GABAergic inner-vation at purkinje axon initial segment. Cell 119, 257–272

52 Palay, S.L. and Palay, V. (1974) Cerebellar Cortex: Cytology andOrganization, Springer

53 Gonzalez-Burgos, G. et al. (2005) Functional properties of fast spikinginterneurons and their synaptic connections with pyramidal cells inprimate dorsolateral prefrontal cortex. J. Neurophysiol. 93, 942–953

54 Zhu, Y. et al. (2004) Chandelier cells control excessive corticalexcitation: characteristics of whisker-evoked synaptic responses oflayer 2/3 nonpyramidal and pyramidal neurons. J. Neurosci. 24,5101–5108

55 Pawelzik, H. et al. (2002) Physiological and morphological diversity of

Reproduction of material

Interested in reproducing part or all of an article published by Elsevier,

Department with details of how and where the requested material w

http://www.elsevier.com/wps/find/obtainpermi

Alternatively, ple

Elsevier Global Rig

Phone: (+44) 18

permissions@e


immunocytochemically defined parvalbumin- and cholecystokinin-positive interneurones in CA1 of the adult rat hippocampus. J. Comp.Neurol. 443, 346–367

56 Klausberger, T. et al. (2003) Brain-state- and cell-type-specific firing ofhippocampal interneurons in vivo. Nature 421, 844–848

57 Galarreta, M. and Hestrin, S. (1999) A network of fast-spiking cells inthe neocortex connected by electrical synapses. Nature 402, 72–75

58 Catterall, W.A. (1981) Localization of sodium channels in culturedneural cells. J. Neurosci. 1, 777–783

59 Stuart, G.J. and Sakmann, B. (1994) Active propagation of somaticaction potentials into neocortical pyramidal cell dendrites. Nature367, 69–72

60 Stuart, G. et al. (1997) Action potential initiation and backpropagationin neurons of the mammalian CNS. Trends Neurosci. 20, 125–131

61 Clark, B.A. et al. (2005) The site of action potential initiation incerebellar Purkinje neurons. Nat. Neurosci. 8, 137–139

62 Colbert, C.M. and Pan, E. (2002) Ion channel properties underlyingaxonal action potential initiation in pyramidal neurons. Nat.Neurosci. 5, 533–538

63 Mainen, Z.F. et al. (1995) A model of spike initiation in neocorticalpyramidal neurons. Neuron 15, 1427–1439

64 Beasley, C.L. et al. (2002) Selective deficits in prefrontal corticalGABAergic neurons in schizophrenia defined by the presence ofcalcium-binding proteins. Biol. Psychiatry 52, 708–715

65 Hashimoto, T. et al. (2003) Gene expression deficits in a subclass ofGABA neurons in the prefrontal cortex of subjects with schizophrenia.J. Neurosci. 23, 6315–6326

66 Marco, P. et al. (1997) Loss of inhibitory synapses on the soma andaxon initial segment of pyramidal cells in human epileptic peritu-moural neocortex: implications for epilepsy. Brain Res. Bull. 44, 47–66

67 Ferrer, I. et al. (1994) Parvalbumin and calbindin-D28k immunocyto-chemistry in human neocortical epileptic foci. J. Neurol. Sci. 123, 18–25

68 DeFelipe, J. (1999) Chandelier cells and epilepsy.Brain 122, 1807–182269 Ribak, C.E. (1985) Axon terminals of GABAergic chandelier cells are

lost at epileptic foci. Brain Res. 326, 251–26070 Dinocourt, C. et al. (2003) Loss of interneurons innervating pyramidal

cell dendrites and axon initial segments in the CA1 region of thehippocampus following pilocarpine-induced seizures. J. Comp. Neurol.459, 407–425

71 Wittner, L. et al. (2001) Preservation of perisomatic inhibitory input ofgranule cells in the epileptic human dentate gyrus. Neuroscience 108,587–600

72 Wittner, L. et al. (2005) Surviving CA1 pyramidal cells receive intactperisomatic inhibitory input in the human epileptic hippocampus.Brain 128, 138–152

73 Akbarian, S. et al. (1995) Gene expression for glutamic aciddecarboxylase is reduced without loss of neurons in prefrontal cortexof schizophrenics. Arch. Gen. Psychiatry 52, 258–266

74 Woo, T.U. et al. (1998) A subclass of prefrontal gamma-aminobutyricacid axon terminals are selectively altered in schizophrenia. Proc.Natl. Acad. Sci. U. S. A. 95, 5341–5346

75 Lewis, D.A. et al. (2004) Selective alterations in prefrontal corticalGABA neurotransmission in schizophrenia: a novel target for thetreatment of working memory dysfunction. Psychopharmacology(Berl.) 174, 143–150

from Elsevier articles

or one of our article figures? If so, please contact ourGlobal Rights

ill be used. To submit a permission request on-line, please visit:

ssionform.cws_home/obtainpermissionform

ase contact:

hts Department

65-843830

lsevier.com


lighting the chandelier: new vistas for axo-axonic cells

Documents