identiﬁcation of basolateral amygdala projection cells and...

Identification of Basolateral Amygdala Projection Cells and InterneuronsUsing Extracellular Recordings

Ekaterina Likhtik,1 Joe Guillaume Pelletier,2 Andrei T. Popescu,1 and Denis Pare1

1Center for Molecular and Behavioral Neuroscience, Rutgers, The State University of New Jersey, Newark, New Jersey; and 2Departementde Physiologie, Universite de Montreal, Montreal, Quebec, Canada

Submitted 2 June 2006; accepted in final form 10 July 2006

Likhtik, Ekaterina, Joe Guillaume Pelletier, Andrei T. Popescu, andDenis Pare. Identification of basolateral amygdala projection cells andinterneurons using extracellular recordings. J Neurophysiol 96:3257–3265, 2006; doi:10.1152/jn.00577.2006. This study tested whetherfiring rate and spike shape could be used to distinguish projection cellsfrom interneurons in extracellular recordings of basolateral amygdala(BLA) neurons. To this end, we recorded BLA neurons in isoflurane-anesthetized animals with tungsten microelectrodes. Projection cells wereidentified by antidromic activation from cortical projection sites of theBLA. Although most projection cells fired spontaneously at low rates(�1 Hz), an important subset fired at higher rates (up to 6.8 Hz). In fact,the distribution of firing rates in projection cells and unidentified BLAneurons overlapped extensively, even though the latter cell group pre-sumably contains a higher proportion of interneurons. The only differ-ence between the two distributions was a small subset (5.1%) of uniden-tified neurons with unusually high firing rates (9–16 Hz). Similarly,distributions of spike durations in both cell groups were indistinguishable,although most of the fast-firing neurons had spike durations at the lowend of the distribution. However, we observed that spike durationsdepended on the exact position of the electrode with respect to therecorded cell, varying by as much as 0.7 ms. Thus neither firing rate norspike waveform allowed for unequivocal separation of projection cellsfrom interneurons. Nevertheless, we propose the use of two firing ratecutoffs to obtain relatively pure samples of projection cells and interneu-rons: �1 Hz for projection cells and �7 Hz for fast-spiking interneurons.Supplemented with spike-duration cutoffs of �0.7 ms for projection cellsand �0.5 ms for interneurons, this approach should keep instances ofmisclassifications to a minimum.

I N T R O D U C T I O N

The role of the basolateral amygdala (BLA) in the formationof appetitive and aversive memories continues to attract muchinterest (see Baxter and Murray 2002; Everitt et al. 2003;Fanselow and Poulos 2005; LeDoux 2000; McGaugh 2004;Phelps 2006; Walker and Davis 2002). One of the techniquesoften used to investigate these questions involves extracellularneuronal recordings (reviewed in Holland and Gallagher 2004;Maren and Quirk 2004; Pelletier et al. 2005). However, theBLA contains multiple cell types, including a prevalent groupof projection cells and multiple subtypes of local-circuitGABAergic neurons (McDonald 1992), complicating the in-terpretation of extracellular data. This problem is compoundedby the fact that, in contrast with the hippocampus whereneurons are segregated in a laminar fashion, different cell typesare intermingled in the BLA. As a result, the position of BLAcells cannot be used to infer their identity. Thus it would be of

great practical interest to define a set of criteria that could beused to distinguish projection cells from interneurons on thebasis of their extracellularly recorded activity.

Intracellular and patch-clamp studies of morphologicallyidentified BLA neurons offer clues as to what properties mightbe useful to identify BLA cells. Studies performed in vivo(Chen and Lang 2003; Lang and Pare 1997a,b, 1998; Pare et al.1995) and in vitro (Faber and Sah 2002, 2003; Faber et al.2001; Faulkner and Brown 1999; Mahanty and Sah 1998;Martina et al. 2001; Pape et al. 1998; Rainnie et al. 1993;Szinyei et al. 2000; Washburn and Moises 1992) indicate thatmost principal cells have a comparatively hyperpolarizedmembrane potential, show little spontaneous firing at rest, andgenerate trains of long-duration action potentials that exhibitvarious degrees of frequency adaptation. In contrast, a subtypeof BLA interneurons (fast-spiking cells), thought to expressparvalbumin and accounting for roughly 50% of BLA inter-neurons (Mascagni and McDonald 2003), have a compara-tively more depolarized membrane potential, are often sponta-neously active at rest, and generate trains of short-durationspikes that generally display little or no spike frequency adap-tation. The latter property is thought to result from the expres-sion of K� channels with fast activation kinetics (McDonaldand Mascagni 2006).

In addition to their distinct intrinsic membrane properties,synaptic factors also contribute to make interneurons moreactive than projection cells. Indeed, intracellular studies in vivo(Lang and Pare 1997a,b) revealed that the spontaneous activityof projection cells is dominated by large-amplitude hyperpo-larizing potentials generated by a combination of synaptic�-aminobutyric acid types A and B (GABAA and GABAB,respectively) conductances (Rainnie et al. 1991; Washburn andMoises 1992) and by a synaptically activated Ca2�-dependentK� current (Chen and Lang 2003; Danober and Pape 1998;Faber et al. 2005; Lang and Pare 1997b). In contrast, fast-spiking interneurons lack GABAB receptors (Huttman et al.2006; Martina et al. 2001) and their GABAA reversal potentialis �15 mV positive to that of projection cells, a differencereflecting the presence of cotransporters that accumulate chlo-ride in fast-spiking interneurons and extrude chloride in pro-jection cells (Martina et al. 2001).

Together, these contrasting patterns of GABA responsive-ness and intrinsic properties should conspire to make projec-tion cells less active than fast-spiking interneurons. Thus dif-ferences in firing rates and action potential duration might

Address for reprint requests and other correspondence: D. Pare, CMBN,Aidekman Research Center, Rutgers, The State University of New Jersey, 197University Avenue, Newark, NJ 07102 (E-mail: [email protected]).

The costs of publication of this article were defrayed in part by the paymentof page charges. The article must therefore be hereby marked “advertisement”in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

J Neurophysiol 96: 3257–3265, 2006;doi:10.1152/jn.00577.2006.

32570022-3077/06 $8.00 Copyright © 2006 The American Physiological Societywww.jn.org

constitute useful criteria to distinguish these two BLA celltypes. The present study tested this possibility using extracel-lular recordings of BLA cells, some of which could be unam-biguously identified as projection cells using antidromic acti-vation from known cortical projection sites of the BLA.

M E T H O D S

All procedures were approved by the Institutional Animal Care andUse committee of Rutgers State University, in compliance with theGuide for the Care and Use of Laboratory Animals (Department ofHealth and Human Services). Adult male cats (2.5–3.5 kg) werepreanesthetized with a mixture of ketamine and xylazine [15 and 2mg/kg, administered intramuscularly (im)] and artificially ventilatedwith a mixture of ambient air, oxygen, and isoflurane. Atropine (0.05mg/kg, im) was administered to prevent secretions. The end-tidalconcentration in CO2 was kept at 3.7 � 0.2% and the body temper-ature was maintained at 37–38°C with a heating pad. The level ofanesthesia was assessed by continuously monitoring the electroen-cephalogram and electrocardiogram.

The bone overlying the regions of interest was removed and thedura mater opened. High-impedance (10- to 12-M�) tungsten micro-electrodes (FHC, Bowdoin, ME) were lowered to the basolateralnucleus (BL) using a micromanipulator. To physiologically identifyBL projection cells, concentric stimulating electrodes were stereotaxi-cally inserted in several cortical projection sites of the cat BL nucleusincluding the perirhinal cortex, infralimbic cortex, and insula (Krettekand Price 1977a,b; Pare et al. 1995; Room and Groenewegen 1986;Smith and Pare 1994). To be classified as antidromic, evoked unitresponses had to meet at least two of the following three criteria(Lipski 1981): 1) stable latency (�0.3-ms jitter), 2) collision withorthodromically evoked or spontaneously occurring spikes, and 3)ability to respond faithfully to high-frequency stimuli (300 Hz). Thespontaneous and evoked activity of responsive neurons was observedon a digital oscilloscope, digitized at 20 kHz, and stored on disk foroff-line analysis. We considered only neurons generating spikes witha high signal-to-noise ratio (�3).

To determine whether the duration of action potentials constituteda fixed property of each neuron or whether it depended on the exactposition of the microelectrode tip with respect to the recorded cell, we

performed a “distance test.” In this test, we first located a well-isolated, spontaneously active neuron and then moved the electrodeback in 5-�m steps over a distance of �110 �m from the recordingsite where the highest signal-to-noise ratio was observed.

At the end of experiments, the animals were administered anoverdose of sodium pentobarbital (50 mg/kg, administered intrave-nously) and extracellular recording sites were marked with electro-lytic lesions (0.6 mA for 10 s). The brains were then taken out, fixedin 2% paraformaldehyde and 1% glutaraldehyde, sectioned on avibrating microtome (at 100 �m), and stained with neutral red orcresyl violet to reveal electrode placements. The microelectrode trackswere reconstructed by combining the micrometer readings and histol-ogy. To be included in the analysis, cells had to be histologicallyconfirmed as being located in the BL nucleus.

Analyses were performed off-line with commercial software (IGOR,WaveMetrics, Lake Oswego, OR; Matlab, The MathWorks, Natick, MA)and custom-designed software running on personal computers. Spikeswere detected using a window discriminator after digital filtering of theraw waves. All values are expressed as means � SE.

R E S U L T S

Spontaneous firing rates of BL neurons

To minimize the risk of missing BL neurons with lowspontaneous firing rates, microelectrodes were lowered gradu-ally while applying electrical stimuli to the various stimulationsites (0.3 Hz, 0.1–0.5 mA, 100–300 �s). When an ortho- (Fig.1A) or antidromically (Fig. 1B) responsive and/or spontane-ously active BL neuron with a high signal-to-noise ratio wasencountered, a 3-min period of spontaneous activity was re-corded. Then, its responsiveness to perirhinal, insular, andinfralimbic stimuli was tested. In this case, our goal was todetermine whether antidromic discharges could be elicitedfrom one or more of these stimulation sites, to formallyestablish whether this cell was a projection neuron. It should bestressed that this technique is prone to false negatives becausefailure to antidromically activate a cell might not mean that therecorded cell is a local-circuit neuron but a projection cell

FIG. 1. Categorization criteria for our sample.A: example of BL neuron that was orthodromicallyactivated by insula (INS) stimuli but could not beantidromically activated from any of the testedcortical stimulation sites. This cell was thereforeclassified as “unidentified.” A1: 23 superimposedtraces. A2: peristimulus histogram of responsesevoked from the insula, but based on a largernumber of trials. B: BL neuron that could be anti-dromically activated from the infralimbic (IL, B1)and perirhinal (PRH, B2) cortex. This cell wastherefore classified as a projection cell. Note fixedlatency of antidromic spikes (B1 and B2, 10 super-imposed sweeps in both cases) and collision withspontaneously occurring spikes (B3). Arrowheadsin A and B mark the stimulus artifacts. C: frequencydistribution of firing rates in our entire sample (C1),among physiologically identified projection cells(C2) and in the unidentified neurons (C3). Firingrates of projection cells and unidentified neuronsoverlap extensively, with the exception of a fewunidentified cells that exhibited much higher firingrates (C3).

3258 LIKHTIK, PELLETIER, POPESCU, AND PARE

J Neurophysiol • VOL 96 • DECEMBER 2006 • www.jn.org

whose axon does not end (or course) in close proximity to thestimulating electrodes. Nonetheless, a positive finding, that isthe presence of electrically evoked antidromic discharges, canbe interpreted as unambiguous evidence that the recorded cellis a projection neuron.

Figure 1B shows an example of a BL neuron that could bepositively identified as such because electrical stimulation ofthe infralimbic (Fig. 1B1) or perirhinal (Fig. 1B2) corticesevoked antidromic spikes characterized by a fixed latency andcollision with spontaneously occurring action potentials (Fig.1B3). Overall, 52.6% of recorded BL neurons (or 41 of 78)were identified as projection cells using this technique. Most ofthese cells could be antidromically activated from only one ofthe tested sites (insula, 11.5%; infralimbic, 26.9%; perirhinal,8.9%), but 5.1% projected to at least two of the tested sites, asin the example of Fig. 1B.

Next, we compared the spontaneous firing rate of physio-logically identified projection cells to that of unidentifiedneurons. We reasoned that because this second group of BLcells should contain a higher proportion of interneurons, dif-ferences in the distribution of firing rates between the twogroups might allow us to derive criteria that could be used todistinguish projection cells from interneurons. Figure 1Cshows the distribution of firing rates in the entire sample (Fig.1C1), in the subset of physiologically identified projection cells(Fig. 1C2), and in the unidentified neurons (Fig. 1C3).

The spontaneous firing rates of BL cells ranged from 0 to 16Hz (average 2.1 � 0.4 Hz; Fig. 1C1). Although nearly 51.2%of projection cells fired �1 Hz (Fig. 1C2), we encounteredprojection cells with spontaneous firing rates as high as 6.8 Hz(Fig. 2). Overall, projection cells had a slightly lower averagefiring rate (1.3 � 0.3 Hz, range, 0–6.8 Hz) than that ofunidentified neurons (3.1 � 0.7 Hz; Mann–Whitney U test,

P � 0.05). Yet, distributions of firing rates in the two groupsoverlapped extensively (Fig. 1, C2 and C3), with the exceptionof a few unidentified neurons (5.1% of entire sample) withunusually high rates of spontaneous firing (9.9–16 Hz). Thusthese results suggest that a cutoff firing frequency of 7 Hzmight be used to distinguish projection cells from fast-spikinginterneurons. However, because counts of parvalbumin immu-noreactive cells predict that random samples of BL neuronswill include 10% of fast-spiking cells (as estimated in rats;Mascagni and McDonald 2003), and not 5.1% as in oursample, a 7-Hz cutoff is likely to lead to the erroneousclassification of nearly half the fast-spiking cells as projectionneurons. Moreover, considering that extracellular recordingsare biased toward neurons with high firing rates, the latterfigure likely underestimates the incidence of misclassifications.

Spike durations in BL neurons

Because the above analyses suggest that firing rate is acriterion of limited sensitivity to distinguish projection cellsfrom interneurons, we next turned our attention to spike dura-tion. As reviewed in the INTRODUCTION, intracellular and patch-clamp studies of morphologically identified BL neurons indi-cate that spike duration (typically measured at half-amplitude)is significantly longer in projection cells (1.0–1.5 ms) com-pared with fast-spiking interneurons (0.5–0.7 ms). However,these recordings are typically obtained from the soma of BLcells. This differs from extracellular recordings where themicroelectrode tip (�1 �m) can be positioned at any pointaround the axonal, somatic, or dendritic compartments of BLcells.

In patch-clamp studies of cortical pyramidal cells, spikedurations were shown to vary depending on the cellular com-

FIG. 2. Some physiologically identifiedBL projection cells exhibit high spontaneousfiring rates. A: example of BL neuron thatwas physiologically identified as a projectioncell on the basis of its antidromic responsesto stimulation of the insula. Note fixed la-tency of antidromic spikes (A1, 10 sweepsoverlaid) and collision with spontaneousspikes (A2). B: spontaneous activity of thesame neuron. B1: sample of spontaneousactivity. This cell had a firing rate of 6.8 Hz.B2 and B3: interspike interval histogramswith bins of 10 ms (B2) and 2 ms (B3). Inthis cell, spike duration was 0.82 ms whenmeasured between the negative and positivepeaks with 0.1- to 10-kHz filtering. Abbre-viations: X, average interspike-interval; N,number of intervals in the depicted range.

3259EXTRACELLULAR IDENTIFICATION OF BLA CELLS


partment from which the recording was obtained. In particular,dendritic spikes generally last longer (sometimes twice asmuch) than somatic spikes (e.g., see Gulledge and Stuart 2003;Larkum et al. 2001). In addition, extracellularly recordedaxonal spikes were reported to be briefer (0.4–0.7 ms) thansomatodendritic spikes in multiple cell types (reviewed inLemon 1984).

Thus there is reason to suspect that the spikes generated bysingle cells can have different durations depending on thecellular compartment that is sampled by extracellular micro-electrodes, a contention supported by a recent modeling study(Gold et al. 2006). As a result, for interneurons and projectioncells to be reliably distinguished on this basis, within-cellvariations in spike durations arising from electrode locationmust be smaller that the variations between cell types. To testwhether this is the case in the BL nucleus, we thus performedexperiments designed to estimate these two variables.

To assess the impact of microelectrode location on spikeduration, we performed a “distance test.” In this test, spikesgenerated by spontaneously active BL cells (n � 11) wererecorded as the electrode was moved in 5-�m steps over adistance of �110 �m from the recording site where the highestsignal-to-noise ratio was observed. To minimize the risk ofconfusing the action potentials generated by different cells, weconsidered recordings where only one spontaneously activeneuron was present and where the signal-to-noise ratio was�5. Last, we limited our attention to cells generating primarilynegative spikes followed by a slower positive potential, themost frequently encountered spike shape in extracellular re-cordings. Spike duration was defined as the interval betweenthe negative and positive peaks because this method proveduseful to distinguish principal cells from interneurons in theneocortex (Bartho et al. 2004) and hippocampus (Csicsvari et

al. 1999) where, in simultaneous extra- and intracellular re-cordings, a close correspondence was found between thisparameter and the duration of intracellularly recorded actionpotentials.

Figure 3 shows examples of spikes recorded at seven dif-ferent locations around a representative cell with a slow (Fig.3A) and a fast time base (Fig. 3B). Throughout the distancetest, the spike amplitude remained well above noise and firingrates remained stable. The left inset of Fig. 3B plots spikeduration (left axis) and amplitude (right axis) as a function ofdistance from the original recording site. In this cell, spikedurations changed by as much as 45% from the first (blackline) to the last recording site (red line). Overall, depending onthe electrode position relative to recorded cells, spike durationsvaried by as much as 45.5 � 4.5% (range 17–66%) of thelongest measured value (1.5 � 0.67 ms). Thus these observa-tions imply that depending on the position of the recordingelectrode, one can expect maximal within-cell variations ofspike durations of about 0.70 ms.

Next, we compared within-cell variations in spike dura-tions arising from electrode position to the variability seenbetween different neurons. With respect to firing rates, wecompared the subset of neurons that were physiologicallyidentified as projection cells to the group of unidentifiedneurons that presumably included a higher proportion oflocal-circuit cells. However, measuring spike durations inextracellular recordings is not trivial because spike shapesvary. Although most cells generated primarily negative-going action potentials (72.2%, as in Fig. 3), some generatedprimarily positive-going spikes (18.1%), and others posi-tive-negative ones (9.7%). Importantly, the particular spikeshape was not a specific property of each neuron because it

0.2

mV

0 µm 10 µm 40 µm 80 µm 90 µm 100 µm 110 µm

0.1

mV

0.2

mV

A

B

IS

0.5 ms

0.5 ms

0 µm-110 µmS

pike

dur

atio

n (m

s)

12080400Distance (µm)

Spike am

plitude (mV

) 0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

0.6

FIG. 3. Spike duration varies depending on themicroelectrode position with respect to the recordedcell. A: examples of action potentials recorded atvarious positions (numbers) around the same cell.Initial recording is marked 0 �m. Electrode waswithdrawn gradually �110 �m from the original site.It should be noted that, even though the electrode wasmoved 110 �m, the total travel distance between theelectrode tip and recorded cell was probably lowerbecause the movement of the electrode likely causedthe tissue to move with it. B: graph plotting spikeduration (left axis, filled circles) and amplitude (rightaxis, empty squares) as a function of electrode dis-tance from original site (x-axis). Changes are illus-trated with an expanded superimposition of the spikesrecorded at these various positions. Values markedby the red and black circles in the graph on the left,respectively, correspond to the spikes drawn with redand thick black traces. Bottom right: scheme showinghypothesized trajectory of the electrode in relation tothe recorded cell. Because spikes recorded toward theend of the trajectory (red circle) were briefer andexhibited a more obvious break between the initialsegment (IS) and somatodendritic (SD) componentsthan those recorded at the original site (black circle),we presume that the thinner spikes were recorded inproximity to the initial axonal segment, whereas thelongest one was recorded in the dendrites. That initialsegment spikes are generally of lower amplitude thansomatodendritic spikes was established in experi-ments where simultaneous extra- and intracellularrecordings of action potentials were performed (Ter-zuolo and Araki 1961).



could change from one type to another when the electrodewas moved around the recorded cell (Fig. 4A).

This variability in spike shapes led us to measure spikedurations in two ways. In the first approach, spike durationswere measured as in the distance test (interval between thenegative and positive peaks for neurons generating primarilynegative action potentials, Fig. 4B1; interval between positiveand negative peaks for cells generating primarily positivespikes, Fig. 4B2). In the second approach, spike duration wasdefined as the interval between the initial change from baselineto return to baseline (Fig. 4C), including all components of thespike shape, as in previous studies on BL neurons (Rosenkranzand Grace 1999, 2001). It should be noted that because thisapproach includes the slower potential associated with spikeafterhyperpolarizations, it sometimes yields long spike dura-tions (�4 ms).

Figure 4, D and E compares the distribution of spike dura-tions as measured with the two methods in the entire sample ofBL neurons (Fig. 4, D1 and E1), in the subset of identifiedprojection cells (Fig. 4, D2 and E2), and in the unidentified

neurons (Fig. 4, D3 and E3). The two methods yielded differ-ent ranges of values with a correlation coefficient of r � 0.65.However, in both cases the distributions ranged widely, withconsiderable overlap between projection cells and unidentifiedneurons. The difference in average spike durations (listed inthe top right corner of each graph) between projection cellsand unidentified neurons did not reach significance with eithermethod (t-test, P � 0.05) and qualitatively identical resultswere obtained with different filtering settings (1 Hz to 3 kHzvs. 100 Hz to 10 kHz). Importantly, no correlation was ob-served between spike duration and firing rate with eithermethod (Fig. 4F). However, most of the cells with extremelyhigh firing rates did have spike durations at the lower end of thedistributions (empty circles in Fig. 4F).

To test whether taking into account the entire spike shaperather than a particular attribute (such as duration) would betterdifferentiate between cell types, we performed principal-com-ponent analysis (PCA) on our sample. Previously, PCA wasused to search for better classification criteria than the imme-diately observable data points, solving the problem of reducing

0.5 mV

1 ms

A

1 ms

B1

C1

B2

C2

2 ms

12

10

8

6

4

2

0

Cel

l cou

nts

2..52..01..51..00..50..0Spike duration (ms)

All Cells

12

10

8

6

4

2

0

Cel

l cou

nts


Unidentified Cells

12

10

8

6

4

2

0

Cel

l cou

nts


Projection Cells

D1

D2

D314

12

10

8

6

4

2

0

Cel

l cou

nts

543210Spike duration (ms)

14

12

10

8

6

4

2

0

Cel

l cou

nts


Unidentified Cells

All Cells

14

12

10

8

6

4

2

0

Cel

l cou

nts


Projection Cells

E1

E2

E3

F1

4..0

3..5

3..0

2..5

2..0

1..5

1..0

1614121086420

2..0

1..5

1..0

0..5

Spi

ke d

urat

ion

(ms)

1614121086420

Firing rate (Hz)

Spi

ke d

urat

ion

(ms)

Firing rate (Hz

X=0.91

X=0.85

X=1.9

X=1.85

F2

Projection cell

Unidentified Cell

Presumed ITN

FIG. 4. Spike shape and duration are unre-liable indicators of BL cell identity with high-impedance extracellular electrodes. A: spikepolarity, amplitude, and duration vary depend-ing on the position of the microelectrode withrespect to the recorded cell. Overlaid examplesof spikes generated by the same cell, but re-corded at different positions. B and C: 2 meth-ods used to measure spike durations. B: in thefirst method, spike duration was defined as theinterval between trough and peak for preva-lently negative spikes (B1) or peak to troughfor prevalently positive spikes (B2). C: in thesecond method, spike duration was defined asthe interval between the initial change frombaseline to return to baseline for both spikeshapes. D and E: frequency distribution ofspike durations as estimated with the first (D)and second (E) method in our entire sample(1), in the subset of physiologically identifiedprojection cells (2), and among the unidenti-fied neurons (3). F: plots of spike duration(y-axis) vs. firing rate (x-axis) using the first(1) and second (2) method for measuring spikeduration. Different symbols are used depend-ing on the identity of the cells: red circles forphysiologically identified projection cells,black circles for unidentified cells, empty cir-cles for cells presumed to be interneuronsbecause of their high spontaneous firing rates(�7 Hz). Abbreviation: X, average spikeduration.



data dimensionality (Harris et al. 2000). All spikes werecentered on their peak and the amplitudes normalized. Featurevectors were extracted based on PCA and used to project theoriginal data (Fig. 5). The first 10 components (explaining 91%of the variability) were used to cluster the spikes in two groups,using the K-means method. In general, the first componentcontained most of the information used to generate the twogroups. A �2 analysis revealed that the incidence of projectioncells and presumed interneurons did not differ between the twogroups (�2 � 0.68, P � 0.41). Similarly, the ratio of projectioncells to presumed interneurons in the two groups did not differfrom that expected by chance on the basis of the incidence ofthe two cell types reported in previous studies of GABAimmunoreactivity (group 1, �2 � 0.51, P � 0.48; group 2,�2 � 0.01, P � 0.77). Similar results were obtained whenusing the first two to ten components. Overall, these observa-tions are in keeping with the fact that spike shapes can changedepending on electrode position around recorded cells and thuscannot serve as a reliable criterion to differentiate projectioncells from interneurons (Figs. 3 and 4A).

D I S C U S S I O N

The present study was undertaken to test whether projectioncells and interneurons of the BL nucleus can be distinguishedreliably in extracellular recordings with high-impedance elec-trodes using criteria such as firing rate and spike waveform.Given that a complete understanding of emotional learning willrequire an analysis of the interaction between projection cellsand interneurons, developing reliable criteria for the extracel-lular identification of different BL cells types is of greatimportance. In the following account, we will first considerearlier attempts at classifying cell types on the basis of extra-cellular data and then consider the applicability and reliabilityof such criteria in the BL nucleus.

In some structures, extracellular data can be used todistinguish different cell types

In a number of cortical areas and subcortical structures, itwas reported that different cell types have easily distinguish-able firing properties. In the thalamus, for instance, corticallyprojecting and reticular neurons can be distinguished by theirbursting behavior. Thalamocortical cells generate bursts withinterspike intervals of progressively increasing durations,whereas reticular thalamic neurons produce bursts where suc-cessive interspike intervals initially show a progressive de-crease and then a gradual increase in duration (Domich et al.1986). Unfortunately, this criterion cannot be used in theamygdala because bursting projection cells account for a mi-nority of projection neurons (Pare and Gaudreau 1996) andlocal-circuit cells with an indistinguishable burst-firing patternexist in the BL nucleus (Rainnie et al. 1993).

Other classification schemes have been based on the differ-ential responsiveness of neurons to particular drugs. In thesubstantia nigra, for instance, physiologically identified nigro-striatal neurons were reported to be inhibited by D2 receptoragonists, whereas other types of nigral neurons were unrespon-sive (Hainsworth et al. 1991). This differential responsivenesscoincided with distinct electrophysiological signatures such aslong-duration action potentials (�2.5 ms) and low firing ratesin nigrostriatal cells compared with spikes of short-durationand elevated basal discharge rates in presumed interneurons orthalamically projecting cells (Deniau et al. 1978; Grace andBunney 1983; Guyenet and Aghajanian 1978; Hainsworth etal. 1991; Paladini and Tepper 1999; Tepper et al. 1987).However, to the best of our knowledge, there is no well-established pattern of pharmacological responsiveness thatcould be used to distinguish BL projection cells and interneu-rons. Besides, routinely conducting such tests would be im-practical.

In the hippocampus, where the position of recorded neuronswith respect to the pyramidal layer is already a strong indica-tion of their identity, it was reported that a combination ofextracellular properties including spike duration, firing rate,and pattern could be used to reliably distinguish pyramidalcells from interneurons (e.g., see Csicsvari et al. 1999; Fox andRanck 1975). However, this task is more difficult in theamygdala because interneurons and projection cells are inter-mingled and their dendrites are randomly oriented.

A possible solution to this problem was proposed by Buzsakiand colleagues (Bartho et al. 2004). By performing high-density neuronal recordings with silicon probes in layer 5 ofthe somatosensory cortex, this group examined the functionalinteractions between simultaneously recorded neurons andidentified a subset of cells whose activity was negativelycorrelated to that of simultaneously recorded neurons. Suchcells were regarded as putative GABAergic interneurons,whereas cells with positively correlated activity were classifiedas pyramidal neurons. The spike shapes of these presumedinterneurons and projection cells were then used as templatesto classify all recorded cells. Spike duration (defined as theinterval between the negative and positive peaks, as in thepresent study) measured from unfiltered data (1 Hz to 5 kHz)was found to be the most reliable criterion to distinguish thetwo cell types.

0 0.5 1 1.5 2 2.5

0

0.5

1

1.5

10

4070

100

P I U

Principal component #1

Prin

cipa

l com

pone

nt #

2

Projection CellsInterneurons

Unidentified Cell Type

Group 1Group 2

-0.5

-1.5 -1 -0.5-1

FIG. 5. Clustering of BL cells based on principal-component analysis ofspike shapes. Cells were separated in two groups (red symbols, group 1; blacksymbols, group 2) based on the first 10 principal components (representingmost of the variability in spike shape), using the K-means technique (see text).Graph plots first (x-axis) vs. second (y-axis) component for all recordedneurons using different symbols depending on their identity (see legend onbottom right). Inset: percentage of projection cells (P), presumed interneurons(I, based on the 7-Hz cutoff in firing rate), and unidentified (U) cells in eachof the 2 groups (red bars, group 1; black bars, group 2).



Previous attempts to distinguish BL projection cells versusinterneurons with extracellular data

Although many investigators have conducted extracellularrecordings in the BL nucleus, to the best of our knowledge,only two groups have proposed criteria to distinguish projec-tion cells from interneurons (Pare and Gaudreau 1996; Rosen-kranz and Grace 1999, 2001). Both classifications were basedon firing rate. In the classification of Rosenkranz and Grace(1999, 2001), neurons with firing rates �0.5 Hz were classifiedas projection cells, whereas neurons with higher dischargerates where classified as interneurons. This bipartite classifica-tion of firing rate was reported to coincide with a bimodaldistribution of spike durations (Rosenkranz and Grace 1999,2001) where projection cells generated spikes of longer dura-tion than interneurons. In contrast, Pare and Gaudreau (1996)proposed a cutoff firing rate of 10 Hz but did not examine therelation between firing rate and spike durations.

Applied to the present sample of BL neurons, the criteriaproposed by Rosenkranz and Grace would have led to theerroneous classification of many projection cells as interneu-rons. Indeed, although the present study and a previous one inunanesthetized animals (Pare and Gaudreau 1996) reported thatmany projection cells fire at extremely low rates, more thanhalf of them had firing rates �0.5 Hz, with some firing at ratesas high as 6.8 Hz. Also, spike durations were not distributedbimodally in our sample and several physiologically identifiedprojection cells had relatively short spike durations. Consistentwith this, the sample of BL cells described by Rosenkranz andGrace (1999) contained a high proportion of presumed inter-neurons, roughly threefold higher than that expected on thebasis of counts of GABA immunoreactive neurons in the ratBL nucleus (Mascagni and McDonald 2003).

However, methodological differences might account forsome of the discrepancies between these studies. First, Rosen-kranz and Grace (1999, 2001) used a different anesthetic(chloral hydrate) that might have reduced the excitability ofprojection cells. Second, our study was conducted in cats,whereas Rosenkranz and Grace (1999, 2001) used rats. Giventhat the rat BL nucleus was reported to contain a higherproportion of principal cells with rapid spike frequency adap-tation (Sah et al. 2003), this fact might also explain the higherproportion of principal cells with low firing rates in the studiesreported by Rosenkranz and Grace (1999, 2001). In contrast,there is no evidence in the literature to explain the extremelyhigh proportion of fast-spiking cells they observed. Althoughthe percentage of GABAergic interneurons in the cat amygdalais unknown, observations in primates (McDonald and Augus-tine 1993) suggest it increases with phylogeny, as was ob-served in the thalamus (Arcelli et al. 1997; Winer and Larue1996) and cortex (Fairen et al. 1984; Jones 1993; Rakic 1975).

Misclassification using firing rates and spike durations

In the present study, comparing the distributions of firingrates in physiologically identified versus unidentified BL neu-rons revealed only one nonoverlapping region: neurons withfiring rates �7 Hz (seen only in the unidentified neurons). It istherefore probable that cells with such high firing rates areinterneurons. This contention is supported by the fact that mostof these cells had spike durations at the low end of thedistributions.

Similarly, there was much overlap between the distributionof spike durations seen in physiologically identified projectioncells and unidentified neurons. As mentioned earlier, neuronswith high firing rates did tend to generate spikes of shortduration, but so did many projection cells. Although this seemsto contradict earlier intracellular and patch-clamp studies (seereferences in INTRODUCTION), we believe the discrepancy re-flects the fact that high-impedance extracellular electrodes canbe located at any point along the soma, dendrites, or axons ofrecorded cells, whereas intracellular and patch-clamp measure-ments are typically obtained from the soma. Consistent withthis, we observed that spike shapes and durations were not afixed property of each cell but that they varied depending onthe electrode position with respect to the recorded cell.

Thus our results suggest that the only conclusive way toidentify projection cells remains antidromic activation from BLprojection sites. By reason of the fact that projection cellsaccount for most BL neurons, a majority of unidentified neu-rons, especially those with low firing rates, are likely projectioncells. In contrast, instances of misclassifications will likely behigher for interneurons. Using a cutoff of 7 Hz on our dataimplies that only about 5% of recorded cells were interneurons.This figure is roughly half the incidence expected on the basisof previous immunohistochemical studies, even ignoring thefact that extracellular recordings are biased toward cells withhigh firing rates. These considerations lead us to conclude thatwith a cutoff firing rate of 7 Hz, the rate of false positiveidentification of fast-spiking neurons is probably low, but thatthe rate of false negatives is probably very high, around 50%.This high rate of false negatives might be explained by the factthat there are multiple physiological subtypes of parvalbumin-positive interneurons in the BL nucleus. In support of thiscontention, Woodruff and Sah (2005) reported that �20% ofparvalbumin-positive BL interneurons have a regular firingpattern that is indistinguishable from that of projection cells,

FIG. 6. Estimated probability that a BL cell is a projection cell or inter-neuron depending on its firing rate. Exponential curves were fitted to thefrequency distribution of firing rates among projection cells and divided by thenumber of recorded cells firing at each frequency. Thus the black linerepresents the probability that a cell at a particular firing rate is a projectioncell. Dashed line is the inverse of the solid line and thus an estimate that cellswith a particular firing rate are interneurons.



except for the short duration of the action potentials theygenerate.

In an attempt to describe the probability that a cell with aparticular firing rate is a projection cell or an interneuron, wefitted exponential curves to the distribution of firing rates forboth identified projection cells and the whole neuronal popu-lation in our data. Figure 6 (continuous line) shows the ratio ofknown projection cells to the entire population at all encoun-tered frequencies. The one’s complement of this probabilitycurve yields an estimate (dashed line) of the likelihood thatneurons with particular firing rates are interneurons. On thisbasis, we propose that the optimal approach to obtain puresamples of projection cells and interneurons is to use two firingcutoffs: �1 Hz for projection cells and �7–8 Hz for fast-spiking interneurons. Indeed, the probability functions of Fig.6 suggest that, at a cutoff frequency of 7.4 Hz, the probabilityof incorrectly identifying an interneuron as a projection cell is�5%. The incidence of misclassifications should be very lowwith this approach, especially if used in combination with twospike duration cutoffs (�0.7 ms for projection cells and �0.5ms for fast-spiking interneurons). The drawback of this ap-proach is that it forces investigators to ignore many of therecorded cells. On the other hand, one can be confident aboutthe identity of recorded neurons. A challenge for future studieswill be to test the validity of this approach by directly identi-fying the morphology of presumed interneurons using juxta-cellular labeling (Pinault 1996) of BLA neurons with neurobi-otin.

A C K N O W L E D G M E N T S

We thank J. Tepper and members of the Pare lab for comments on an earlierversion of this manuscript.

G R A N T S

This work was supported by National Science Foundation Grant 0208712and by National Institute of Mental Health (NIMH) Grants R01MH-073610-01to D. Pare and NIMH National Research Service Award Grant 1 F31 MH-076415-01 to E. Likhtik.

R E F E R E N C E S

Arcelli P, Frassoni C, Regondi MC, De Biasi S, and Spreafico R. GABAer-gic neurons in mammalian thalamus: a marker of thalamic complexity?Brain Res Bull 42: 27–37, 1997.

Bartho P, Hirase H, Monconduit L, Zugaro M, Harris KD, and Buzsaki G.Characterization of neocortical principal cells and interneurons by networkinteractions and extracellular features. J Neurophysiol 92: 600–608, 2004.

Baxter MG and Murray EA. The amygdala and reward. Nat Rev Neurosci 3:563–573, 2002.

Chen JC and Lang EJ. Inhibitory control of rat lateral amygdaloid projectioncells. Neuroscience 121: 155–166, 2003.

Csicsvari J, Hirase H, Czurko A, Mamiya A, and Buzsaki G. Oscillatorycoupling of hippocampal pyramidal cells and interneurons in the behavingrat. J Neurosci 19: 274–287, 1999.

Danober L and Pape HC. Mechanisms and functional significance of a slowinhibitory potential in neurons of the lateral amygdala. Eur J Neurosci 10:853–867, 1998.

Deniau JM, Hammond C, Riszk A, and Feger J. Electrophysiologicalproperties of identified output neurons of the rat substantia nigra (parscompacta and pars reticulata): evidences for the existence of branchedneurons. Exp Brain Res 32: 409–422, 1978.

Domich L, Oakson G, and Steriade M. Thalamic burst patterns in thenaturally sleeping cat: a comparison between cortically projecting andreticularis neurones. J Physiol 379: 429–449, 1986.

Everitt BJ, Cardinal RN, Parkinson JA, and Robbins TW. Appetitivebehavior: impact of amygdala-dependent mechanisms of emotional learn-ing. Ann NY Acad Sci 985: 233–250, 2003.

Faber ES, Callister RJ, and Sah P. Morphological and electrophysiologicalproperties of principal neurons in the rat lateral amygdala in vitro. J Neu-rophysiol 85: 714–723, 2001.

Faber ES, Delaney AJ, and Sah P. SK channels regulate excitatory synaptictransmission and plasticity in the lateral amygdala. Nat Neurosci 8: 635–641, 2005.

Faber ES and Sah P. Physiological role of calcium-activated potassiumcurrents in the rat lateral amygdala. J Neurosci 22: 1618–1628, 2002.

Faber ES and Sah P. Ca2�-activated K� (BK) channel inactivation contrib-utes to spike broadening during repetitive firing in the rat lateral amygdala.J Physiol 552: 483–497, 2003.

Fairen A, DeFelipe J, and Regidor J. Nonpyramidal neurons; generalaccount. In: Cellular Components of the Cerebral Cortex, edited by PetersA and Jones EG. New York: Plenum, 1984, p. 201–253.

Fanselow MS and Poulos AM. The neuroscience of mammalian associativelearning. Annu Rev Psychol 56: 207–234, 2005.

Faulkner B and Brown TH. Morphology and physiology of neurons in the ratperirhinal-lateral amygdala area. J Comp Neurol 411: 613–642, 1999.

Fox SE and Ranck JB Jr. Localization and anatomical identification of thetaand complex spike cells in dorsal hippocampal formation of rats. Exp Neurol49: 299–313, 1975.

Gold C, Henze DA, Koch C, and Buzsaki G. On the origin of the extracel-lular action potential waveform: a modeling study. J Neurophysiol 95:3113–3128, 2006.

Grace AA and Bunney BS. Intracellular and extracellular electrophysiologyof nigral dopaminergic neurons—1. Identification and characterization.Neuroscience 10: 301–315, 1983.

Gulledge AT and Stuart GJ. Excitatory actions of GABA in the cortex.Neuron 37: 299–309, 2003.

Guyenet PG and Aghajanian GK. Antidromic identification of dopaminergicand other output neurons of the rat substantia nigra. Brain Res 150: 69–84,1978.

Hainsworth AH, Roper J, Kapoor R, and Ashcroft FM. Identification andelectrophysiology of isolated pars compacta neurons from guinea-pig sub-stantia nigra. Neuroscience 43: 81–93, 1991.

Harris KD, Henze DA, Csicsvari J, Hirase H, and Buzsaki G. Accuracy oftetrode spike separation as determined by simultaneous intracellular andextracellular measurements. J Neurophysiol 84: 401–414, 2000.

Holland PC and Gallagher M. Amygdala-frontal interactions and rewardexpectancy. Curr Opin Neurobiol 14: 148–155, 2004.

Huttmann K, Yilmazer-Hanke D, Seifert G, Schramm J, Pape HC, andSteinhauser C. Molecular and functional properties of neurons in thehuman lateral amygdala. Mol Cell Neurosci 31: 210–217, 2006.

Jones EG. GABAergic neurons and their role in cortical plasticity in primates.Cereb Cortex 3: 361–372, 1993.

Krettek JE and Price JL. Projections from the amygdaloid complex to thecerebral cortex and thalamus in the rat and cat. J Comp Neurol 172:687–722, 1977a.

Krettek JE and Price JL. Projections from the amygdaloid complex andadjacent olfactory structures to the entorhinal cortex and to the subiculum inthe rat and cat. J Comp Neurol 172: 723–752, 1977b.

Lang EJ and Pare D. Similar inhibitory processes dominate the responses ofcat lateral amygdaloid projection neurons to their various afferents. J Neu-rophysiol 77: 341–352, 1997a.

Lang EJ and Pare D. Synaptic and synaptically activated intrinsic conduc-tances underlie inhibitory potentials in cat lateral amygdaloid projectionneurons in vivo. J Neurophysiol 77: 353–363, 1997b.

Larkum ME, Zhu JJ, and Sakmann B. Dendritic mechanisms underlying thecoupling of the dendritic with the axonal action potential initiation zone ofadult rat layer 5 pyramidal neurons. J Physiol 533: 447–466, 2001.

LeDoux JE. Emotion circuits in the brain. Annu Rev Neurosci 23: 155–184,2000.

Lemon R. Methods for Neuronal Recordings in Conscious Animals. Chich-ester, UK: Wiley, 1984.

Lipski J. Antidromic activation of neurones as an analytic tool in the study ofthe central nervous system. J Neurosci Methods 4: 1–32, 1981.

Mahanty NK and Sah P. Calcium-permeable AMPA receptors mediatelong-term potentiation in interneurons in the amygdala. Nature 394: 683–687, 1998.

Maren S and Quirk GJ. Neuronal signaling of fear memory. Nat RevNeurosci 5: 844–852, 2004.

Martina M, Royer S, and Pare D. Cell-type-specific GABA responses andchloride homeostasis in the cortex and amygdala. J Neurophysiol 86:2887–2895, 2001.



Mascagni F and McDonald AJ. Immunohistochemical characterization ofcholecystokinin containing neurons in the rat basolateral amygdala. BrainRes 976: 171–184, 2003.

McDonald AJ. Projection neurons of the basolateral amygdala: a correlativeGolgi and retrograde tract tracing study. Brain Res Bull 28: 179–185, 1992.

McDonald AJ and Augustine JR. Localization of GABA-like immunoreac-tivity in the monkey amygdala. Neuroscience 52: 281–294, 1993.

McDonald AJ and Mascagni F. Differential expression of Kv3.1b and Kv3.2potassium channel subunits in interneurons of the basolateral amygdala.Neuroscience 138: 537–547, 2006.

McGaugh JL. The amygdala modulates the consolidation of memories ofemotionally arousing experiences. Annu Rev Neurosci 27: 1–28, 2004.

Paladini CA and Tepper JM. GABA(A) and GABA(B) antagonists differ-entially affect the firing pattern of substantia nigra dopaminergic neurons invivo. Synapse 32: 165–176, 1999.

Pape HC, Pare D, and Driesang RB. Two types of intrinsic oscillations inneurons of the lateral and basolateral nuclei of the amygdala. J Neurophysiol79: 205–216, 1998.

Pare D and Gaudreau H. Projection cells and interneurons of the lateral andbasolateral amygdala: distinct firing patterns and differential relation to thetaand delta rhythms in conscious cats. J Neurosci 16: 3334–3350, 1996.

Pare D, Pape HC, and Dong J. Bursting and oscillating neurons of the catbasolateral amygdaloid complex in vivo: electrophysiological properties andmorphological features. J Neurophysiol 74: 1179–1191, 1995.

Pelletier JG, Likhtik E, Filali M, and Pare D. Lasting increases in basolat-eral amygdala activity after emotional arousal: implications for facilitatedconsolidation of emotional memories. Learn Mem 12: 96–102, 2005.

Phelps EA. Emotion and cognition: insights from studies of the humanamygdala. Annu Rev Psychol 57: 27–53, 2006.

Pinault D. A novel single-cell staining procedure performed in vivo underelectrophysiological control: morpho-functional features of juxtacellularlylabeled thalamic cells and other central neurons with biocytin or neurobiotin.J Neurosci Methods 65: 113–136, 1996.

Rainnie DG, Asprodini EK, and Shinnick-Gallagher P. Excitatory trans-mission in the basolateral amygdala. J Neurophysiol 66: 986–998, 1991.

Rainnie DG, Asprodini EK, and Shinnick-Gallagher P. Intracellular record-ings from morphologically identified neurons of the basolateral amygdala.J Neurophysiol 69: 1350–1362, 1993.

Rakic P. Local circuit neurons. Neurosci Res Prog Bull 13: 295–416, 1975.Room P and Groenewegen HJ. Connections of the parahippocampal cortex

in the cat. II. Subcortical afferents. J Comp Neurol 251: 451–473, 1986.Rosenkranz JA and Grace AA. Modulation of basolateral amygdala neuronal

firing and afferent drive by dopamine receptor activation in vivo. J Neurosci19: 11027–11039, 1999.

Rosenkranz JA and Grace AA. Dopamine attenuates prefrontal corticalsuppression of sensory inputs to the basolateral amygdala of rats. J Neurosci21: 4090–4103, 2001.

Sah P, Faber ES, Lopez De Armentia M, and Power J. The amygdaloidcomplex: anatomy and physiology. Physiol Rev 83: 803–834, 2003.

Smith Y and Pare D. Intra-amygdaloid projections of the lateral nucleus in thecat: PHA-L anterograde labeling combined with postembedding GABA andglutamate immunocytochemistry. J Comp Neurol 342: 232–248, 1994.

Szinyei C, Heinbockel T, Montagne J, and Pape HC. Putative corticaland thalamic inputs elicit convergent excitation in a population of GABA-ergic interneurons of the lateral amygdala. J Neurosci 20: 8909–8915,2000.

Tepper JM, Sawyer SF, and Groves PM. Electrophysiologically identifiednigral dopaminergic neurons intracellularly labeled with HRP: light-micro-scopic analysis. J Neurosci 7: 2794–2806, 1987.

Terzuolo CA and Araki T. An analysis of intra- versus extra-cellularpotential changes associated with activity of single spinal motoneurons. AnnNY Acad Sci 94: 547–558, 1961.

Walker DL and Davis M. The role of amygdala glutamate receptors in fearlearning, fear-potentiated startle, and extinction. Pharmacol Biochem Behav71: 379–392, 2002.

Washburn MS and Moises HC. Electrophysiological and morphologicalproperties of rat basolateral amygdaloid neurons in vitro. J Neurosci 12:4066–4079, 1992.

Winer JA and Larue DT. Evolution of GABAergic circuitry in themammalian medial geniculate body. Proc Natl Acad Sci USA 93:3083–3087, 1996.

Woodruff AR and Sah P. Four distinct interneuronal networks in thebasolateral amygdala. Program No. 271.11. 2005 Abstract Viewer/ItineraryPlanner. Washington, DC: Society for Neuroscience, 2005, Online.



identiﬁcation of basolateral amygdala projection cells and...

Documents