INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The classical conditioning of eyelid/nictitating membrane responses is a widely used experimental procedure for studying the neuronal mechanisms underlying associative learning and memory storage (Clark and Squire 1998; Gormezano et al. 1983; Thompson and Krupa 1994; Woody 1986). Indeed, lid movements are especially suitable for this experimental approach, particularly when recorded with the search coil technique, which allows a precise quantification of evoked reflex and conditioned responses (CRs) (Evinger et al. 1991; Fuchs and Robinson 1966; Gruart et al. 1995). The kinematics, frequency-domain properties, and movement topography of spontaneous, reflex, passive, and conditioned eyelid responses have been described and quantified in different species, including humans (Domingo et al. 1997; Evinger et al. 1991; Gormezano et al. 1983; Gruart et al. 1995; Welsh 1992). Neuronal centers involved in lid responses have been determined, and the electrical activity of representative neuronal types have been recorded and analyzed in both in vivo and in vitro studies (Berger et al. 1983; Gruart and Delgado-García 1994; McEchron and Disterhoft 1997; Sanchez-Andres and Alkon 1991; Trigo et al. 1999).

Nevertheless, the neuronal centers where classically conditioned eyelid responses are learned and/or stored have been a matter of continuous debate for the past four decades (Aou et al. 1992; Bliss and Collingridge 1993; Bloedel 1992; Gormezano et al. 1983; Thompson and Krupa 1994; Woody 1986). In particular, the hippocampus has been implicated in a wide variety of learning and memory experimental paradigms and clinical conditions (Berger et al. 1983; Hoehler and Thompson 1980; McEchron and Disterhoft 1997; Sanchez-Andres and Alkon 1991). It has been shown in both experimental animals and humans that bilateral hippocampal lesions impair the acquisition, but not the retention, of trace eyeblink conditioning, whereas they do not alter delay conditioning (Moyer et al. 1990; Thompson 1988). A proposed explanation is that trace conditioning requires a conscious knowledge (Clark and Squire 1998) and/or declarative or explicit memory (Eichenbaum 1999) of relevant relationships between stimuli used as conditioned (CS) and unconditioned (US) stimuli, a fact apparently not required for delay conditioning to occur. However, multiunitary recordings in awake rabbits during the classical conditioning of the nictitating membrane response have shown that pyramidal and other hippocampal cell types fire indistinctly, to CS-US presentations, for both conditioning paradigms (Berger et al. 1983; McEchron and Disterhoft 1997). The evoked neuronal firing was reported to precede the beginning of the nictitating membrane CR (Berger et al. 1983). These contradictory results have been interpreted as an indication of a putative role of hippocampus in the signaling of the temporal relationships between environmental sensory cues (Eichenbaum 1999), and a less-defined role in the generation of CR profiles (Berger et al. 1983).

The aim of the experiments presented here was to determine whether hippocampal pyramidal cell responses were related to CS sensory modality, CR latency and topography, or CS predictive value. The unitary activity of identified pyramidal cells was recorded in alert behaving cats during different (2 trace and 1 delay) conditioning paradigms. Since it has been described that different CSs are able to evoke CRs different in latency, amplitude, and profile (Rescorla 1988), two distinct sensory modalities (tone and air puff) were used as CS in the two trace conditioning paradigms. The US used in the three conditioning paradigms was always a long, strong air puff applied to the cornea. Eyelid movements were recorded across conditioning with the search coil technique (Gruart et al. 1995). It has been described that the output of hippocampus is facilitated during fast beta and gamma oscillatory activity and blocked during theta rhythmic states (Herreras et al. 1987; Stang Leung 1998; Traub et al. 1999; Vanderwolf 1969; Weiss et al. 1996), and that scopolamine impairs both fast oscillatory activity and information processing in the hippocampus (Múnera et al. 2000). Moreover, the presence of beta oscillation has been related to the perception of sensory cues of particular relevance (Traub et al. 1999). Thus extracellular field potentials were also recorded during conditioning trials, and analyzed with a Morlet wavelet transform to determine the dominant oscillatory activity in hippocampal structures during selected time windows before, during, and after CS-US paired presentation. A short report of these results has appeared in abstract form (Múnera et al. 1999).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects and preexperimental surgical procedures

Eight cats (Iffa-Credo, Lyon, France) were used. Experiments were carried out according to guidelines of the European Union (86/609/EU) and current Spanish legislation (BOE 67/8509-12 1988) for the use of laboratory animals in chronic experiments, and following the approval of the local Ethical Committee. Under controlled anesthesia (pentobarbital sodium, 35 mg/kg ip and atropine sulfate, 0.5 mg/kg im), a Teflon-coated, stainless steel, five-turn coil (3 mm diam) was implanted into the center of the upper left lid of each animal. A bipolar stimulating electrode (200 µm diam, enamel-coated silver wire) was implanted in the right fornix following stereotaxic coordinates (Reinoso-Suárez 1961). To allow transcortical access to right posterior and dorsal hippocampus, a recording window of 8 × 8 mm was drilled in the occipital bone. The dura mater was removed, and an acrylic recording chamber was constructed around the window. The recording chamber was maintained covered and in sterile conditions between recording sessions. Finally, a head-holding system consisting of three bolts cemented to the skull perpendicular to the stereotaxic plane was implanted. Wire terminals from recording and stimulating electrodes were soldered to a socket cemented to the holding system (Gruart and Delgado-García 1994; Gruart et al. 1995; Trigo et al. 1999).

General conditions of recording sessions

Recording sessions started 2 wk after surgery and lasted for <3 h per day for a maximum of 16 days. For experiments, the animal was lightly restrained, placed on the recording table, and its head immobilized in the head-holding system. The recording chamber was opened and a microelectrode advanced toward the hippocampus. Following two sessions dedicated to finding the proper recording site, each animal was assigned randomly to one of the conditioning or pseudoconditioning paradigms described below. Conditioning sessions lasted 10 days and were preceded by two habituation sessions and followed by two of extinction.

Recording techniques

Neuronal electrical activity was recorded with glass micropipettes filled with 2 M NaCl (3-6 M of resistance) and filtered in a bandwidth of 1 Hz to 10 kHz. The recording micropipette was always removed at the end of each recording session. Field potentials were recorded with low-resistance electrodes (1-3 M) or low-pass filtered (>100 Hz) from unitary recordings. The recording area was approached with the help of stereotaxic coordinates (Reinoso-Suárez 1961), and field potentials were evoked by electrical stimulation of the fornix (Berger et al. 1983). The recording electrode tip was tilted 30° anteriorly and moved in the sagittal and coronal planes in 0.1-mm steps. Electrical stimulation of the fornix consisted of single cathodal 50-µs square pulses with current intensities <0.2 mA. Criteria to determine whether the recorded and the activated neuron were the same and to discriminate somatic versus axonic recording were systematically followed (see Gruart and Delgado-García 1994 for details). Eyelid movements were recorded and quantified with the search coil in a magnetic field technique (Gruart et al. 1995).

Classical conditioning paradigms

The classical conditioning of eyelid responses was achieved by the use of two trace and a delay paradigms. For trace air puff-Air Puff conditioning, animals (n = 2) were presented with a short (20-ms), weak (1-kg/cm²) air puff directed at the left cornea as CS. The CS was followed 500 ms after its end by a 100-ms, 3-kg/cm² air puff directed at the same eye as US. Trace tone-Air Puff conditioning was achieved (n = 2 animals) with a 20-ms, 600-Hz, 90-dB tone followed 500 ms after its end by a similar US applied to the left eye. For delay Tone-Air Puff conditioning (n = 2 animals), a 600-ms, 600-Hz, 90-dB tone was used as CS. The tone was followed 500 ms from its onset by a 100-ms, 3-kg/cm² air puff directed at the left cornea as US. Thus the tone and the air puff terminated simultaneously.

Each conditioning session consisted of 12 blocks separated by a variable (4-6 min) interval. Each block consisted of 10 trials separated at random by intervals ranging from 20 to 40 s. The CS was presented alone in the first trial of each block. A complete conditioning session lasted 2 h. During habituation and extinction sessions, the CS was presented alone for the same number of blocks/session and trials/block and with similar random interblock and intertrial distribution (Gruart et al. 1995). Although conditioned criterion (90% of CRs/session) was reached in conditioned animals by the 4th to 6th conditioning session, conditioning was maintained for 10 sessions to obtain the maximum number of recorded neurons. Two additional animals were pseudoconditioned with the unpaired presentation of tones (20 ms, 600 Hz, 90 dB), air puffs (20 ms, 1 kg/cm²), and Air Puff (100 ms, 3 kg/cm²) used here for trace (tone-Air Puff and air puff-Air Puff) conditioning (Domingo et al. 1997; Gruart et al. 1995).

Histology

At the end of recording sessions, animals were deeply anesthetized (pentobarbital sodium, 50 mg/kg ip). Electrolytic marks were placed at selected recording sites with a tungsten microelectrode and at stimulating sites with the help of the implanted electrode (1 mA for 10 s). Animals were then perfused transcardially with saline and 10% phosphate-buffered Formalin. Serial 50-µm brain sections were mounted on glass slides and stained with toluidine blue for analysis.

Data collection and analysis

Eyelid position, neuronal activity, and rectangular pulses corresponding to blink-evoking stimuli or to CS and US presentations were stored digitally on a videotape recording system at a sampling frequency of 22 kHz for biopotentials and 11 kHz for the other signals. Data were transferred through an analog/digital converter (CED 1401 Plus, Cambridge, UK) to a computer for off-line analysis. Most data were sampled at 1-4 kHz with an amplitude resolution of 12 bits, but selected unitary records were sampled at 22 kHz for representational purposes. In addition, action potentials were fed into a window discriminator, and the resulting Schmidt trigger pulses were stored on the computer using the same A/D conversion card.

Commercial computer programs (Spike 2 and SIGAVG from CED) were modified, and new programs were developed to display single, overlapping, averaged, and raster representations of eyelid position, velocity and acceleration, and neuronal and field potential activities. Color rasters were made with the help of a representation program written in Java language by one of us (R. Fernández-Mas). Velocity and acceleration profiles were computed digitally as the first and second derivative of lid position records after low-pass filtering of the data (3 dB cutoff at 50 Hz and a 0 gain at 100 Hz). The instantaneous firing rate was calculated as the inverse of the interspike intervals (Trigo et al. 1999).

The same computer programs also allowed the quantification of lid position and neuronal parameters. Data were processed for statistical analysis with the SPSS (Chicago, IL) for Windows package, for two-tailed tests with a statistical significance level of P = 0.05. Putative relationships between neuronal firing rate and lid position, velocity, and acceleration were checked by linear regression analysis. For this, eyelid position, velocity, or acceleration was plotted versus the instantaneous firing rate, in 1-ms point-to-point process. A quantitative study of neuronal firing rate, field potential, and lid acceleration evolution across conditioning trials was also carried out. The power of the spectral density function (i.e., the power spectrum) of selected 0.5- to 2.5-s segments from field potentials and eyelid acceleration recordings was calculated using a Morlet wavelet transform to define the relative strength of the different frequencies present in field potentials and eyelid acceleration responses (Clarençon et al. 1996; Daubechies 1988). The implementation of this fast wavelet transformation algorithm allowed us to analyze field potential and eyelid acceleration recordings, using a 32-point moving window relative to 0.25-Hz discrete steps (Schiff et al. 1994). Significance of power spectrum peaks was tested with the ²-distributed test for spectral density functions. Statistical differences of mean values were determined with the help of ANOVA, followed by a contrast analysis when needed. Polynomial contrast was used to assess parameter evolution across conditioning (Domingo et al. 1997; Trigo et al. 1999).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

General properties of recorded cells

The recording area, located in posterior and dorsal hippocampus, was delimited with the help of unitary recordings and field potential profiles, during the first two recording sessions, while blink-evoking stimuli were randomly presented (Fig. 1, A-C). Across the whole experiment, a total of 253 neurons were recorded in this specific area. Neurons were identified by their antidromic activation from the fornix, but only those with an antidromic latency 3 ms and a spike duration 0.5 ms were considered as pyramidal cells and further recorded and analyzed (n = 220). The spontaneous firing rate of recorded cells ranged from 2 to 10 spikes/s. All recorded neurons presented complex spikes, both spontaneously and during evoked responses.

View larger version (26K): [in this window] [in a new window]   Fig. 1. Experimental design and firing properties of recorded pyramidal cells. A: diagram illustrating the location of eyelid search coil and stimuli used to evoke reflex and conditioned eyelid responses. B: location of the stimulating electrode in fornix (Fx St) and the recording sites (Rec) in posterior and dorsal hippocampus (Hi). C: an example of the collision test used for unit identification (asterisk). D-F: firing rate of 3 representative pyramidal cells during reflex and conditioned eyelid responses. Recordings were carried out during the presentation of blink-evoking stimuli (a 20-ms, 600-Hz, 90-dB tone in D, and a 100-ms, 3-kg/cm2 air puff in E) and during the paired presentation of the same stimuli across the 6th conditioning session of a trace tone-Air Puff conditioning in F. Lid position and velocity are also illustrated. All records were averaged (n = 30 for D and E, and n = 100 for F). G and H: histograms of the interspike interval distribution (G), and the autocorrelation histogram (H) during spontaneous neuronal activity (skyline) and during the CS-US interval (filled bars). I: averaged (n = 500) field potential triggered by the recorded spike of a representative pyramidal cell. CS and US, conditioned and unconditioned stimuli, respectively; d, down; u, up.

Each neuron was recorded to a maximum of 2 h, i.e., the duration of the conditioning session. A minimum of one and a maximum of four pyramidal cells were recorded per habituation, conditioning, extinction, or pseudoconditioning session. A minimum of 29 and a maximum of 41 neurons were recorded per conditioned animal across the successive habituation (n = 2), conditioning (n = 10), and extinction (n = 2) sessions.

According to histological reconstruction and analysis, recorded cells were located mainly (n = 194, i.e., 88.2%) in the CA1, and the others (n = 26, i.e., 11.8%) in the CA3 area (Fig. 2, A-C). No significant difference between the firing properties of CA1- and CA3-located neurons could be established, because of the low number of pyramidal cells recorded in the CA3 area in comparison with the different paradigms used in the present study. Accordingly, present results should be considered as a description of CA1 pyramidal cell firing properties.

View larger version (88K): [in this window] [in a new window]   Fig. 2. Location of recording sites. A: a horizontal reconstruction of recording sites (plus signs). The dot indicates the location of the electrolytic mark shown in B and C. B: coronal section at the level indicated in A (a-a'). C: an enlargement of section illustrated in B. Calibration bar for A-C: 2 mm. B and C correspond to the coronal plane A4 according to the atlas of Reinoso-Suárez (1961). DG, dentate gyrus; DH, dorsal hippocampus; Fi, fimbria; SUB, subiculum; VH, ventral hippocampus; C, D, L, M, R, V, caudal, dorsal, lateral, medial, rostral, ventral, respectively.

A total of 95% (209 of 220) of identified pyramidal cells fired to tone and air puff presentations with a brief (134 ± 20 ms, mean ± SD), weak (3.3 ± 2.4 spikes; range, 1-11) burst of activity (Fig. 1, D and E). This firing diminished following the repeated unpaired presentation of the stimuli, when presented at a high-frequency (1/s). However, from the first session during the classical conditioning of eyelid blinks, the paired presentation of the same stimuli increased the response to the stimulus (tone or short, weak air puff) used as CS in >60% of identified cells, particularly when compared with the weaker response evoked by the air puff used as US (Fig. 1F), i.e., for paired CS-US presentations, pyramidal cells fired much more to the CS than to the US.

The interspike interval distribution of pyramidal cells was always right-skewed, leptokurtic, and unimodal (modal interval 3.4-7.1 ms, n = 220 neurons), both during spontaneous firing and in the CS-US interval (Fig. 1G). Most (85%) of the recorded cells presented a theta rhythmicity (3-9 Hz) in autocorrelation histograms; however, this rhythmic activity disappeared during the CS-US interval (Fig. 1H). The profiles of spike-triggered field potential averages indicated that these pyramidal cells fired preferentially during a conspicuous fast beta (16-20 Hz) extracellular oscillatory activity preceded and followed by a fast theta rhythm of 5-9 Hz (Fig. 1I).

General dynamics of pyramidal cell firing during classical conditioning of eyelid responses

Figure 3 illustrates the firing rate of an antidromically identified pyramidal cell recorded during the ninth conditioning session of a trace tone-Air Puff conditioning. The neuron fired a short burst of spikes starting 75.2 ± 17.3 (n = 120 trials) ms following CS presentation. Outside the CS-US time window, the illustrated pyramidal cell fired irregularly in brief bursts. The evoked firing response of the cell to CS presentation was consistently maintained throughout the whole session (120 trials). Apart from the CS-US time interval, no pyramidal modified significantly its spontaneous firing rate across conditioning, for the three different conditioning paradigms used here.

View larger version (61K): [in this window] [in a new window]   Fig. 3. Firing rate of a pyramidal cell, field potential activity, and lid conditioned responses (CRs) evoked across the 9th session of a trace tone-Air Puff conditioning. A: illustrations from top to bottom are as follows: 1) a diagram showing the conditioning paradigm; 2) the firing activity of a pyramidal cell recorded across the 2-h session. The top trace shows a firing profile (in spikes/s) of the cell during a single trial. The bottom raster display shows the firing rate of the same cell, averaged each 5 trials, during the 120 trials of the session. Note the repetitive cell response to CS presentation across the session. 3) The field potential recorded with the same microelectrode, and lid position and acceleration records obtained during the same single trial. The bottom raster display shows lid acceleration recordings averaged each 5 trials, throughout the conditioning session. Vertical dashed lines indicate the 4 2.5-s epochs used for spectral analysis. B, top panels: on a trial-by-trial basis, the power spectra of the field potential, computed for the 4 2.5-s epochs, are indicated by the dashed lines. Note the appearance of a significant (P < 0.001) beta (16-20 Hz) band, outlasting by 2 s the CS-US time window. The bottom panels show the power spectra of lid acceleration during the 120 session trials. Note the significant (P < 0.01) peak at 20 Hz during eyelid CRs. d, down; u, up.

Moreover, during the 2.5 s following CS presentation, a fast beta (16-20 Hz, P < 0.01; Fig. 3B, top set of records) extracellular field potential oscillation was noticed, superimposed on theta activity, in single records and confirmed by spectral analysis. The unitary pyramidal firing to CS presentation typically was noticed in coincidence with this fast beta oscillation recorded extracellularly.

The beta activity observed to be evoked by CS presentation was not related to lid oscillation during CRs (Domingo et al. 1997), as confirmed by cross-correlation analyses. The power spectra of lid acceleration records showed a different dominant peak at 20 Hz (P < 0.001), over a broader band of at least 18-22 Hz (Fig. 3B, bottom set of records).

The fact that US presentations evoked weaker neuronal firing responses than did CS (Figs. 1F and 3A) was even more evident during trace air puff-Air Puff conditioning, when a short, weak air puff was used as CS, preceding by 500 ms the presentation of a long, strong air puff directed toward the same eye as US (Fig. 4B). A total of 35 of 54 (65%) neurons recorded during the 8th to 10th conditioning sessions for the 3 conditioning paradigms used here evoked larger peak firing rate responses (125%, P < 0.0001, n = 120 trials/neuron) to CS presentations than to US (Fig. 4, A-C).

View larger version (36K): [in this window] [in a new window]   Fig. 4. Comparison of response latency for pyramidal cell firing, averaged field potential and conditioned responses (CRs) across conditioning. A-C: the 3 conditioning paradigms were trace tone-Air Puff (t-AP, in A), trace air puff-Air Puff (ap-AP, in B), and delay Tone-Air Puff (T-AP, in C). From top to bottom are illustrated the conditioning paradigm, the averaged (n  80 records) firing rate of 3 representative pyramidal cells, the averaged (n  80) field potentials recorded in the same site, and the averaged (n = 120) lid CRs. All records were collected from the 10th conditioning session. The bottom plots illustrate the best linear fit of latency evolution of neuronal firing bursts (squares and continuous line), evoked field potentials (triangles and dashed line), and eyelid CRs (circles and dotted line) to CS presentation across the 10 conditioning sessions. Each square, triangle, and circle represents the mean value for 10 recordings. The coefficients of correlation for CR latency evolution across conditioning trials were r = 0.63 (A), 0.94 (B), and 0.81 (C), with P  0.001. Coefficient of correlation values of regression lines for neuronal activation (A-C) and field potential (A-C) latencies were r  0.38 (P  0.01). d, down; u, up.

Temporal relationships between CS-evoked pyramidal cell firing and eyelid CRs

The activation latency of recorded pyramidal cells during trace tone-Air Puff conditioning (n = 60) ranged between 45 and 110 ms (mean 73.6 ± 15.8 ms) after CS presentation, with a statistically nonsignificant (P  0.54) slope across conditioning. Contrarily, the latency of evoked CRs decreased from 350 ± 20 ms, during the first conditioning session, to 112.5 ± 13 ms, during the 10th session (Fig. 4A).

To check whether pyramidal cell responses were causally related to eyelid CRs, a trace air puff-Air Puff conditioning was used. It has been described that this conditioning paradigm in cats evokes short-latency CRs (Gruart et al. 1995). As illustrated in Fig. 4B, this conditioning paradigm did effectively produce short-latency CRs decreasing from 27 ± 4 ms during the 1st conditioning session to 14.5 ± 2.1 ms during the 10th. However, in this case too, the latency of recorded pyramidal neurons (n = 83) ranged from 49 to 105 ms (mean 68.1 ± 16.2 ms), being nonsignificantly different (P = 0.71) from values obtained during trace tone-Air Puff conditioning.

The effects of a delay Tone-Air Puff conditioning on the discharge rate of pyramidal cells were also checked. Here again, the activation latency (69.2 ± 17.6 ms; range, 55-99 ms) of recorded neurons (n = 32) and evoked field potentials did not change across conditioning, while CR latencies decreased from 150 ± 12 ms during the 1st conditioning session to 60 ± 8 ms during the 10th.

For the three conditioning paradigms used here, the ANOVA of repeated measurements of mean CR latencies, session by session, across conditioning, showed significant (P < 0.01) differences. A contrast analysis revealed that CR latencies computed during trace tone-Air Puff conditioning were greater than the ones computed during delay Tone-Air Puff (240%; P < 0.001) and trace air puff-Air Puff (775%; P < 0.001) conditionings (Fig. 4, A and B). Moreover, the latter presented shorter (24.2%) CR latency values (P < 0.01) than the ones obtained using delay Tone-Air Puff conditioning (Fig. 4, B and C). The same analysis applied for mean neuronal and field potential activation latencies across conditioning for the three paradigms used here did not show any difference (P  0.55). Finally, correlation analyses also showed significant negative slopes for CR latencies (trace tone-Air Puff = 0.23 ms/trial; trace air puff-Air Puff = 0.004 ms/trial; delay Tone-Air Puff = 0.08 ms/trial) across conditioning, but not for neuronal unitary and field potential response latencies (Fig. 4, A-C). Polynomial analysis revealed a statistically significant negative linear tendency across conditioning (P < 0.001) for CR latency, but not for the neuronal and field potential activation latencies (P  0.33, Fig. 4, A-C). According to these results, pyramidal cells seem to fire preferentially to CS presentation, regardless of the sensory modality of the presented stimulus (a tone or a short, weak air puff), of the conditioning paradigm (trace or delay), and of the timing and profile of evoked CRs.

Evolution of CS-evoked pyramidal cell responses across conditioning

Figures 5 and 6 illustrate the enhancement of the hippocampal neural response and the CR across conditioning in two selected animals during trace tone-Air Puff (Fig. 5A) and air puff-Air Puff (Fig. 6A) conditioning. Despite the conspicuous differences in the profile evolution of lid CRs evoked by the two different CSs, pyramidal cell firing rate responses were not significantly different in latencies (P = 0.63) and profiles. During habituation, there was a small pyramidal cell response. In successive conditioning sessions, pyramidal cell responses to CS presentation increased, while the neuronal responses evoked by US presentation remained unchanged (P  0.56). Finally, during extinction, even though the lid response had already disappeared, a cellular response slightly stronger (47% larger for trace tone-Air Puff conditioning, P < 0.05, and 36% for trace air puff-Air Puff conditioning, P < 0.05; see Figs. 5 and 6) than that evoked during habituation was still present. It is noteworthy to indicate that changes in firing rate responses were often evident sessions in advance of the appearance of eyelid CRs (arrowheads in Fig. 5A).

View larger version (27K): [in this window] [in a new window]   Fig. 5. Temporal evolution of pyramidal cell firing rate and lid position averages throughout 10 successive conditioning and pseudoconditioning sessions. A: the conditioning was carried out with a trace tone-Air Puff paradigm. From left to right, panels represent recordings obtained during the 1st habituation session (H1, left), the 10 conditioning sessions (C1-C10, middle), and the 2nd extinction session (E2, right). The beginning of CS and US presentations are indicated by arrows. Each neuronal recording represents the mean firing rate recorded across the conditioning session for a single, identified pyramidal cell (n  120 records). Lid position recordings represent the average response across the whole session (n = 120 trials). Note the increase in neuron firing across conditioning to CS presentation. Arrowheads in lid position profiles indicate the appearance of an eyelid CR. B: temporal evolution of pyramidal cell firing rate and lid position averages throughout 10 successive pseudoconditioning sessions. The animal was pseudoconditioned by the unpaired presentation of the tone and Air Puff used in A. Pyramidal cell firing and lid position were recorded and averaged as in A. Note the absence of any noticeable evolution in neuron firing across pseudoconditioning. Although the tone did not evoke any noticeable lid response, while the strong Air Puff evoked an evident blink reflex, the response of the 10 neurons (1 per session) was similar to both stimuli. Note that the illustrated mean firing rate (n  120 trials) was about 50% smaller than that evoked by the same CS when paired. Calibrations in A are also for B.

View larger version (27K): [in this window] [in a new window]   Fig. 6. Temporal evolution of pyramidal cell firing rate and lid position averages throughout 10 successive conditioning and pseudoconditioning sessions. A: the conditioning was carried out with a trace air puff-Air Puff paradigm. From left to right, panels represent recordings obtained during the 1st habituation session (H1, left), the 10 conditioning sessions (C1-C10, middle), and the 2nd extinction session (E2, right). The beginning of CS and US presentations are indicated by arrows. Each neuronal recording represents the mean firing rate recorded across the conditioning session for a single, identified pyramidal cell (n  120 records). Lid position recordings represent the average response across the whole session (n = 120 trials). Note the increase in neuron firing across conditioning to CS presentation. B: temporal evolution of pyramidal cell firing rate and lid position averages throughout 10 successive pseudoconditioning sessions. The animal was pseudoconditioned by the unpaired presentation of the air puff and Air Puff used in A. Pyramidal cell firing and lid position were recorded and averaged as in A. Note the absence of any noticeable evolution in neuron firing across pseudoconditioning. Although the weak air puff only evoked a brief lid response, while the strong Air Puff evoked an evident blink reflex, the response of the 10 neurons (1 per session) was similar to both stimuli. Note that the illustrated mean firing rate (n  120 trials) was about 50% smaller than that evoked by the same CS when paired. Calibrations in A are also for B.

To check the effect of pseudoconditioning on hippocampal pyramidal cell firing, two selected naïve animals were pseudoconditioned with the unpaired presentation of the tone and Air Puff used for trace tone-Air Puff conditioning (Fig. 5B) or with the same air puff and Air Puff used for trace air puff-Air Puff conditioning (Fig. 6B). Pseudoconditioning procedures did not noticeably modify the firing response of pyramidal cell to tone (Fig. 5B) or to air puff (Fig. 6B) across the successive pseudoconditioning sessions.

In all of the animals, from the 6th to the 10th sessions (n = 30), the mean firing rate of pyramidal cells to CS presentations was 79% larger (P < 0.001) than that evoked by the same CS when presented unpaired during pseudoconditioning sessions (n = 10). These results stress the increase in CS predictive value (or in its associative strength) when presented in advance to the US (see Figs. 5 and 6).

Although the repeated presentation of the tone during the pseudoconditioning session (Fig. 5B, left set of records) did not evoke any noticeable lid response, while the Air Puff evoked a definite blink response (Fig. 5B, right set of records), the response of neurons recorded across pseudoconditioning sessions appeared similar to both (tone and Air Puff) stimuli (see also Fig. 1, D and E). In the same way, the unpaired presentation of weak air puffs and strong Air Puffs during the pseudoconditioning sessions illustrated in Fig. 6B evoked a small and a large blink response, respectively. Nevertheless, the firing rate of pyramidal cells recorded during those pseudoconditioned sessions was similar to air puff and Air Puff stimulations.

A further analysis of the recorded data were carried out to determine whether the increased pyramidal cell discharge provided a basis for predicting that the CS would be followed temporally by another stimulus (the US), i.e., that neural firing was related to the CS predictive value. Figure 7 illustrates the firing rate of 15 pyramidal neurons recorded in the same animal throughout a trace tone-Air Puff conditioning. The firing of this population of pyramidal cells in response to tone increased at a rate of 0.035 spikes/s/trial (r = 0.68, P < 0.001) across conditioning (Fig. 7B). Similar increases in the slope of pyramidal cell firing recorded throughout conditioning sessions were observed during trace air puff-Air Puff (0.042 spikes/s/trial; r = 0.72, P < 0.001) and delay Tone-Air Puff (0.027 spikes/s/trial; r = 0.71, P < 0.001) conditionings. As illustrated in Fig. 7A, the increase in neuronal firing started from the first conditioning session, while the CR was clearly noticeable only by the 4th to 7th session, depending on the conditioning paradigm. Finally, the CS-alone presentation induced a decrease in neuronal firing during habituation and extinction sessions; that is, when this CS was not signaling US presentation (Fig. 7B).

View larger version (52K): [in this window] [in a new window]   Fig. 7. Evolution of pyramidal cell firing rate, and lid conditioned responses across recording sessions of a trace tone-Air Puff conditioning. A: from left to right, panels represent raster displays of pyramidal cell firing rate, lid acceleration, and field potential and lid acceleration power spectral density (PSD) functions. The displays are segmented in 3 rows, corresponding to 2 habituation (2 neurons; 240 trials), 10 conditioning (15 neurons; 1,200 trials), and 2 extinction (4 neurons; 240 trials) sessions. Each segment is organized, on a trial-by-trial basis, from bottom to top, as indicated by the long arrow. The short arrows indicate the presentation of conditioned (CS) and unconditioned (US) stimuli. Note the synchronization of cell firing responses following CS presentation across conditioning. Also, note lid acceleration bands radiating from the US toward the CS. A noticeable shift of field potential dominant-frequency components toward the beta band was evident during the CS-US time interval. The evolution across conditioning of these 4 parameters at the time (a-d) and frequencies (e-g) indicated by dashed lines and letters (a-g) are illustrated in B-E. B: firing rate evolution across habituation (Hab.), conditioning, and extinction (Ext.) sessions measured 150 ms before (a, red circles and lines) and 130 ms after (b, black dots and lines) CS presentation. C: evolution of lid acceleration computed 150 ms before (c, red circles and lines) and 475 ms after (d, black dots and lines) CS presentation. D and E: evolution of peak power spectra of field potential (D) and eyelid acceleration (E) PSDs across conditioning. Power spectrum values were computed at 5 Hz (e, red circles and lines) and 16 Hz (f, black dots and lines) for field potential (D), and at 20 Hz (g, black dots and lines) for eyelid acceleration (E) power spectra, respectively. Each red circle, and black dot, represents the mean value for 10 recordings. Regression lines labeled b, d, and e-g reached coefficient of correlation values ranging between r = 0.52 and 0.97 (P < 0.001). Note that neuronal firing rate, lid (downward) acceleration, and lid (downward) acceleration power spectra increased across conditioning [slopes: 0.035 spikes/ s/trial, 0.879 (deg/s2)/trial, 1.5 × 106 (deg/s2)2/trial, respectively], while field potential power spectra at both theta and beta bands decreased across it (slopes: 0.011 and 0.050 mV2/trial).

While lid 20-Hz oscillation evidenced during the CR increased (P < 0.001) in power across conditioning (Fig. 7E), the power spectra of field potentials showed a decrease in both theta and beta components (P  0.001) present during the first recording session for the three conditioning paradigms (Fig. 7D).

Although checked systematically with linear regression analysis, the firing rate of pyramidal cells did not seem to encode eyelid position, velocity, or acceleration for either reflex or conditioned eyelid blinks (n = 10 neurons for reflex responses and n = 15 neurons/conditioning paradigm, r  0.07, P  0.05).

Pseudoconditioned animals did not show any significant CR. The firing rate of recorded pyramidal neurons to CSs (tone, n = 20; air puff, n = 25) and USs (Air Puff, n = 45) during pseudoconditioning sessions was similar to that observed during single-stimulus presentations.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Role for hippocampus in associative learning

According to present results, the hippocampal output, represented here by the discharge activities of antidromically identified CA1 pyramidal neurons, is related to the CS-US associative strength, or to the predictive value of stimuli used as a CS (Eichenbaum 1999; Rescorla 1988). Thus the unitary activity of pyramidal neurons evoked by the CS-US paired presentation was not dependent on the different conditioning paradigms (trace, delay), CS sensory modalities (acoustic, tactile), or CR latency, kinematics, and frequency-domain properties. Neuronal firing increased in response to CS presentation across conditioning, whereas neuronal responses did not change following US presentation throughout successive conditioning trials. The slow building up of neuronal responses across conditioning (i.e., a mean increase of 0.035 spikes/s/trial) suggests a weak, although progressive, molecular process possibly involving mechanisms such as cooperativity, input specificity, and associability (Bliss and Collingridge 1993).

It should be stressed that pyramidal cells' firing was exclusively modified during the short time window represented by the CS-US interval, as their spontaneous firing was not affected by the paired CS-US presentation. Thus hippocampal pyramidal evoked firing appeared restricted to a narrow time window closely related to the time of CS-US presentation. It might be proposed that, in the same way that hippocampus seems to be involved in the generation of an ensemble code for animal location in space (Wilson and McNaughton 1993), it could also generate a sort of brief temporal permissive window during which relevant relationships between external sensory cues are allowed to reach long-lasting memory storage mechanisms (Eichenbaum 1999). This temporal permissive window would be time related to the presence of beta oscillations or, even, higher frequency activity in hippocampal structures, a phenomenon apparently triggered by the paired CS-US presentation. This beta oscillation of local field potentials could be the result of the high-frequency synaptic activity from nearby neuronal elements on pyramidal cell distal dendrites (Traub et al. 1999). Moreover, this extracellular field potential feature could be related with a facilitating mechanism in the consolidation process of the learned association.

It has recently been proposed that hippocampal activity is necessary during classical conditioning paradigms in which conscious knowledge is required (Clark and Squire 1998). Other authors have proposed that the hippocampus is required for explicit memory processes and/or for the transfer of declarative knowledge to other cortical structures (Bechara et al. 1995; Bontempi et al. 1999; Eichenbaum 1999). A parsimonious interpretation of the present results is that hippocampus is mostly involved in the determination of CS-US associative strength or CS predictive value over many other simultaneous environmental cues, but not in the generation of the CR itself. Such interpretation is reinforced by the slow building up of CS-related neuronal response, and by reported findings, in both humans and experimental animals (Bechara et al. 1995; Clark and Squire 1998; Moyer et al. 1990; Thompson 1988), indicating that hippocampus is not needed for the performance of CRs generated with delay conditioning paradigms.

As already described in unitary in vivo recordings (McEchron and Disterhoft 1997), hippocampal pyramidal cell firing to CS presentation increased sessions in advance of behavioral conditioning. However, contrary to previous reports (Berger et al. 1983; Schmajuk 1990), neuronal firing within individual trials did not always precede the beginning of the CR. Since it has been described that different CSs are able to evoke CRs different in latency, amplitude, and profile (Rescorla 1988), two distinct sensory modalities (tone and air puff) were used here as CS in the two trace conditioning paradigms. In addition, a delay paradigm was also carried out for comparative purposes regarding CR latency and profile. In fact, pyramidal firing either preceded or followed CRs according to the paradigm (trace, delay) or CS sensory modality (tone, air puff) used for conditioning. Consequently, pyramidal cell discharge responses were not correlated to any measured parameter of eyelid CRs, such as latency, peak amplitude, velocity, or acceleration.

Neural activity of hippocampal cells in alert behaving animals during classical conditioning of nictitating/eyelid responses

The present results are in some contradiction with previous reports on hippocampal activity of rabbits during classical conditioning of the nictitating membrane response (Berger et al. 1983; McEchron and Disterhoft 1997, 1999; Moyer et al. 1990, 1996; Thompson 1988; Thompson and Krupa 1994). Nevertheless, the same authors have also reported important differences to CS presentation between rabbits and cats (Patterson et al. 1979). There are some considerations that might bear on the observed differences.

Although cats and rabbits are both mammals, they belong to different orders. Since the former are predators and the latter are prey, their cognitive strategies to cope with specific tasks could be different. Moreover, there are noticeable differences in the control of lid movements in cats and rabbits. For example, whereas eye retraction in rabbits is the key component of the blink, in cats the active closure of the lid by the contraction of the orbicularis oculi muscle is more crucial (Gruart et al. 1995, 2000). Importantly, this is very evident for learned blink responses (Trigo et al. 1999). In addition, there is an anatomical peculiarity of the rabbit's hippocampus that has not been identified in other species (Gloor 1997), i.e., a recurrent projection from the subiculum to the CA1 and CA2 regions (Berger et al. 1980). These differences in motor and neural control of lid movements are not restricted to the facial motor system, because important differences have also been reported for the oculomotor system of both species (Graf and Simpson 1981).

Moreover, we have used here a delayed paradigm with a 500-ms interstimulus interval. Although such an interval has been occasionally used, a 250-ms one has been more extensively used. This strategy was selected to dissect neuronal responses to the conditioned and unconditioned stimuli within a larger time window. In the present experiments, the trace conditioned stimuli lasted only 20 ms, in contrast with the usual 100-500 ms. This feature made the task more cognitively demanding and allowed a large time window for both unitary and field potential recordings in the absence of any experimentally evoked sensory stimulus.

The usual technique for the blink conditioning in rabbits involves a restriction of lid movements and the recording of the nictitating membrane displacement. In contrast, in the present experiments we allowed freedom of movement of the lid and recorded its position using the search coil in a magnetic field technique. As already shown by our group in rabbits (Gruart et al. 2000), there are important differences between lid and nictitating membrane kinematics. Mainly, lid responses are the result of the direct action of the orbicularis oculi muscle and not a passive displacement following eyeball retraction into the orbit. As a consequence, lid displacement has a shorter latency and a less damped profile. These differences are extremely important for the establishment of quantitative relationships between movement profiles and neural firing rates.

Here we used single-unit recording with glass micropipettes directed toward the dorsal hippocampus through a recording chamber of 8 × 8 mm. This chamber allowed an actual recording area of about 18 mm², comprising predominantly the CA1 region, and a narrow band of the CA3 region. Isolated units were identified as pyramidal neurons using not only their antidromic activation from the fornix but also the firing profile criteria used by others (Berger et al. 1983; Fox and Rank 1981; McEchron and Disterhoft 1997). Previous studies of hippocampal activity in awake rabbits during classical conditioning of nictitating membrane responses were, in general, based on multiunit activity records obtained using fixed or adjustable arrays of metallic electrodes. Multiunit activity was either processed as a whole or discriminated by means of computer-assisted devices. The latter, however, are prone to mistakes, due to the variability in shape of the extracellularly recorded action potential of pyramidal and nonpyramidal cells. Moreover, the activity of nearby, different subtypes of hippocampal interneurons could shadow the activity of actual pyramidal cells (Parra et al. 1998). This fact is evidenced in Fig. 1I, in which a single spike of an identified pyramidal cell is surrounded by electrical activity from nearby cells. Notwithstanding, the computer-based discrimination technique has demonstrated a wider range of response profiles of the hippocampal neurons during the classical conditioning of the nictitating membrane response. Nevertheless, Berger et al. (1983), using single-unit recording with tungsten electrodes, reported a population of hippocampal neurons that displayed, as reported here, strong CS-evoked and weak US-evoked responses (see their Fig. 10B). Those neurons were not antidromically or orthodromically activated, but their firing properties were similar to hippocampal pyramidal neurons reported in the present work. Also, in coincidence with the present results, those authors reported similar alterations in CA1 and CA3 pyramidal neurons during conditioning (Berger et al. 1983).

In the present experiments, CS-triggered averages of instantaneous firing rate profiles across successive trials were used to summarize pyramidal cell responses during conditioning trials. In contrast, other groups used either peristimulus time histograms or standard firing scores to analyze multiunit responses in the CA1 pyramidal layer. However, when we built up peristimulus time histograms from our data, the resulting profiles were almost identical to the instantaneous firing rate average. Therefore it is unlikely that the difference in the results was due to the analytic procedure, but instead was due to the points indicated above.

In any case, since hippocampal pyramidal cells firing is characterized by brief bursts, the claim that their firing forms a temporal model of the CR is difficult to be explained. There are only two conditions in which such a model could be generated by a single pyramidal neuron: first, by changing its firing pattern from phasic to tonic; second, as an averaging artifact if the neuron fires bursts randomly around the time of US presentation. Previous studies about hippocampal pyramidal cells activity during classical conditioning of blink responses did not report any shift in their firing pattern, which discards the first possibility. Even, if the second one turns out to be right, it is difficult to accept that such a random firing could have any functional significance for the control of CR execution.

Oscillatory activity in hippocampus during classical conditioning

It has long been known that hippocampal field potentials oscillate at gamma and/or beta frequencies, or present an irregular activity, during automatic movements such as licking, face-washing, or blinking, while theta activity is present during patterned voluntary behavior such as walking or jumping (Klimesh 1999; Stang Leung 1998; Vanderwolf 1969). In addition, consummatory behavior seems to be accompanied by beta oscillations in the hippocampus (Elazar and Adey 1967), and desynchronization of cortical structures is a well-known companion of attentive states (Eliade 1996). Recently, it has been proposed that beta oscillation indicates the presence of percepts of a particular interest or novelty (Traub et al. 1999). On the other hand, it has been described that the output of hippocampus is facilitated during fast beta and gamma oscillatory activity and blocked during theta rhythmic states (Bennet 1973; Herreras et al. 1987; Stang Leung 1998; Traub et al. 1999; Vanderwolf 1969; Weiss et al. 1996), and the presence of beta oscillation has been related to the perception of sensory cues of particular relevance (Traub et al. 1999).

As reported here, a fast oscillatory activity in the hippocampus was time locked to CS-US paired presentation and to pyramidal cell firing. The amplitude of the population spike evoked in the CA1 region by the electrical stimulation of the perforant pathway reaches a maximum when the hippocampus displays irregular field potential activity (Herreras et al. 1987). Lesion, pharmacological, and in vitro studies indicate that the activation of CA1 pyramidal cells is facilitated by cholinergic inputs of septal origin (Auerbach and Segal 1996; Múnera et al. 2000; Ridley et al. 1996; Tsubokawa and Ross 1996) and by fast electrical stimulation (Bliss and Collingridge 1993), in both cases in coincidence with fast or desynchronized activity at pyramidal dendritic levels. Some increased excitability is still evident when CA1 pyramidal cells are recorded in hippocampal slices following training (Moyer et al. 1996; Power et al. 1997; Sanchez-Andres and Alkon 1991). The appearance, reported here, of beta oscillation during CS-US presentation could represent a sort of permissive window to facilitate pyramidal firing and hippocampal output to higher levels of neuronal processing.

In summary, the present results and the above considerations suggest that the hippocampus is involved in the abstract analysis of CS-US associative strength or, perhaps, CS predictive value, but not in the genesis and/or performance of acquired CRs.