INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The calcium-binding protein parvalbumin (PV) is expressed within specific types of neurons throughout the CNS (Celio 1990), where it is implicated in Ca²⁺ homeostasis. Under resting conditions, the Ca²⁺-binding sites of PV are principally occupied by Mg²⁺ ions, which have to be displaced by Ca²⁺, at a rate determined mainly by the dissociation rate of Mg²⁺. This explains why, despite its high affinity for Ca²⁺, the on-rate of Ca²⁺ binding is slow (Schwaller et al. 1999). The intracellular calcium concentration ([Ca²⁺]_i) at rest, or peak [Ca²⁺]_i during a Ca²⁺ transient, is not affected by PV (Lee et al. 2000), but PV accelerates the initial decay of [Ca²⁺]_i (Schwaller et al. 1999).

PV has been implicated in protecting against pathologically high levels of intracellular Ca²⁺ as judged from the high survival rate of PV-containing hippocampal interneurons in temporal lobe epilepsy (Freund et al. 1992; Kamphuis et al. 1989; Sloviter et al. 1991) or after ischemia-induced neuronal degeneration (Freund et al. 1992; but see Hartley et al. 1996). The physiological role of PV at the cellular and network levels is less clear. PV was reported to be concentrated in synaptic terminals of basket type interneurons in the cerebellum (Kosaka et al. 1993), where it is likely to modulate the Ca²⁺-dependent release of GABA. If the presynaptic Ca²⁺ transient evoked by a single spike is insufficient to trigger release, it may prime the terminal for release on subsequent spikes (Kamiya and Zucker 1994). PV may suppress this "residual Ca²⁺"-dependent facilitation if it effectively buffers the presynaptic Ca²⁺ transient. Indeed paired-pulse suppression of IPSCs in Purkinje cells evoked by activating basket cells containing PV in PV+/+ mice turned into facilitation in the cerebellum of PV-deficient mice (Caillard et al. 2000). Paired-pulse suppression of IPSCs could be restored by the addition of the slow Ca²⁺-buffer EGTA (similar Ca²⁺-binding kinetics as PV) into the basket cell via patch pipette.

In the hippocampus, PV is expressed in a subset of interneurons (Aika et al. 1994; Freund and Buzsáki 1996; Maccaferri et al. 2000), most prominently axo-axonic interneurons and a subset of basket type interneurons (Kosaka et al. 1987). PV-containing interneurons selectively innervate the perisomatic region of cells (Gulyás et al. 1993; Ribak et al. 1990) and play a major role in the recurrent inhibitory control of principal cells (Freund and Buzsáki 1996). In addition to building a network by mutual synaptic contacts, PV-containing interneurons form a syncytium throughout the hippocampus by dendro-dendritic gap junctions (Fukuda and Kosaka 2000), which is implicated in mediating inhibition-based coherent gamma rhythms (Tamas et al. 2000). Cortical gamma rhythms are associated with cognition and sensory processing (Engel and Singer 2001).

In the present study, we focus on potentially physiological roles of PV in the hippocampus. We postulate that PV reduces the Ca²⁺-dependent facilitation of GABA release and that, as a consequence, it will affect the inhibition-based gamma rhythms in the hippocampus. To test this hypothesis, we used PV-deficient mice (Schwaller et al. 1999). As a measure of Ca²⁺-dependent GABA release, we monitored the monosynaptic GABA_Aergic IPSCs in CA1 principal neurons. To concentrate on intrinsic modulation of GABA release, we have selected experimental conditions that favor low release probabilities (Lambert and Wilson 1994; Thomson 1997) and minimize presynaptic suppression of GABA release. To assess the functional significance of any changes in amplitudes of rhythmic IPSCs, we measured kainate-induced gamma oscillations in CA3.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Generation of PV-deficient mice

The PV-deficient (PV/) mice used in this study were generated by homologous recombination as previously described (Schwaller et al. 1999). Briefly, targeted embryonic stem cells (E14; derived from 129 Ola Hsd mice) were injected into blastocysts of C57BL/6J mice, and the chimeric offspring mated to C57BL/6J animals. Heterozygous mice (PV+/) were bred to obtain both PV+/+ and PV/, mice, which both have a mixed 129 Ola Hsd × C57BL/6J genetic background. Genotyping was performed using genomic DNA isolated from tail biopsies. This was subjected to PCR using primer pairs that were either specific for exon 3 (deleted in PV/ mice) or for part of the neomycin resistance gene (absent in PV+/+ mice). Mice were housed in groups before use. The genotype of the mice was masked until all experiments and analyses had been carried out. In addition, C57BL/6J mice (Harlan OLAC, Bicester, UK) were used for control experiments. All mice were adult (25-30 g) when used for experiments, which were carried out according to the UK Animals Scientific Procedures Act of 1986 and the European Committee Council Direction of November 24 1986 (86/69/EEC).

Histochemistry and morphometry

Immunohistochemical analysis of cryofixed brain sections was performed as previously described (Celio and Heizmann 1982) with the exception that the bound primary antibody was revealed by the avidin-biotin technique rather than the peroxidase-anti-peroxidase one. For immunostaining, the polyclonal antibody PV4064 (Swant, Bellinzona, Switzerland) was used at a dilution of 1:5,000. To visualize the perineuronal nets normally found around PV-expressing neurons, cryo sections were washed in Tris-buffered saline (pH 7.3) additionally containing (in mM) 0.1 MgCl₂, 0.1 MnCl₂, and 0.1 CaCl₂. Free-floating sections were incubated with the peroxidase-labeled isolectin B4 Vicia villosa agglutinin (VVA), diluted to 20 µg/ml in the Tris-buffered saline described in the preceding text for 3 days at 4°C with gentle agitation. After three washes with the Tris-buffered saline, the lectin-peroxidase complex was visualized by incubating sections with the chromophore 3,3'-diaminobenzidine and hydrogen peroxide. The number of profiles of perineuronal nets surrounding individual somata in the regions CA1, CA2, and CA3 were counted, and the length of the pyramidal layer of the respective areas was estimated applying morphometric techniques. Numbers of perineuronal net-positive cells are given per mm length of pyramidal layer for PV+/+ (n = 3) and PV/ (n = 3) mice.

Electrophysiology

Adult mice, randomly selected from both groups (PV+/+ and PV/) and from controls, were anesthetized by intraperitoneal injection of a ketamine (75 mg/kg)/medetomidine (1 mg/kg) mixture and then killed by cervical dislocation. The brain was quickly removed from the skull and chilled in ice-cold artificial cerebrospinal fluid (ACSF). The composition of the ACSF was (in mM) 125 NaCl, 3 KCl, 26 NaHCO₃, 1.25 NaH₂PO₄, 2 CaCl₂, 1 MgCl₂, 10 D-glucose; pH was equilibrated at 7.4 with a 95% O₂-5% CO₂ gaseous mixture. For the recording of GABA_Aergic responses, the brain was cut into 400-µm-thick transverse slices, using a Vibroslice (Campden Instruments, Sileby, UK). Slices were transferred to a recording chamber (kept at 33°C), wherein they were maintained at the interface between a warm moist gaseous atmosphere (95% O₂-5% CO₂) and ACSF, flowing at a rate of 2 ml/min. The ACSF was supplemented with 20 µM 6-nitro-7-sulfamoylbenz[f]quinoxaline-2,3-dione (NBQX), 25 µM D-2-amino-5-phosphonovaleric acid (APV), 1 µM CGP 55845A, 5 µM atropine sulfate, and 5 µM naloxone hydrochloride to isolate GABA_Aergic responses and minimize presynaptic suppression by GABA_B receptors (Davies and Collingridge 1993; Lambert and Wilson 1994), muscarinic receptors (Hajos et al. 2000), and µ opioid receptors (Lambert et al. 1991), respectively. To promote facilitation (Lambert and Wilson 1994; Thomson 1997), the release probability was reduced by increasing the concentration of MgCl₂ to 3 mM. Brain slices were allowed to equilibrate for 1 h before the onset of recording. Monosynaptic GABA_Aergic responses were evoked by electrical stimulation (0.1-ms square pulse), using a constant voltage stimulus isolator (Digitimer, Welwyn Garden City, UK). The stimulus was applied with a bipolar electrode, constructed from a pair of insulated, and intertwined 50-µm-diam nickel/chromium wires (Advent Research Materials, Halesworth, UK), placed in the pyramidal cell layer of area CA1b, within 0.1 mm from the recording site. Intracellular current-clamp and single-electrode voltage-clamp recordings were taken from neurons within the stratum pyramidale using sharp pipettes filled with 2 M potassium methylsulphate (tip resistance was 50-70 M) connected to an Axoclamp-2A amplifier (Axon Instruments, Burlingham, CA). Impaled cells were first inspected in current-clamp and accepted for recording when the resting membrane potential was at least -55 mV and when the current injection-induced overshooting action potentials. For current-clamp recordings, the resting membrane potential was manually adjusted to -65 mV. Single-electrode switch voltage-clamp recordings were accepted when the switching rate was >4 kHz and voltage-clamp efficiency was >90%, as judged from the difference in the inhibitory postsynaptic potential (IPSP) amplitude between voltage- and current-clamp recordings. The holding potential was -65 mV. Single- and double-pulse stimulations were applied at 15-s intervals, and stimulus trains (10 pulses) were applied at 3-min intervals.

For the recording of network oscillations, 400-µm-thick horizontal slices of the ventral hippocampus were cut and transferred to an interface slice recording chamber where they were allowed to equilibrate for 1 h in standard ACSF at 33°C before the onset of recording. Flow rate was increased to 3-4 ml/min. Extracellular field potential recordings were made with glass pipettes filled with ACSF (tip resistance: 2-4 M) from the pyramidal cell layer of area CA3b/c. Intracellular current-clamp recordings were made from pyramidal neurons in area CA3b/c using methods described in the preceding text.

NBQX, APV, naloxone hydrochloride, and atropine sulfate were obtained from Tocris-Neuramin (Bristol, UK), PV4064 from Swant (Bellinzona, Switzerland), CGP 55845A was a gift from Ciba Geigy (Basel, Switzerland), and all other drugs were purchased from Sigma (Poole, UK).

Signals were low-pass filtered at 3 kHz and sampled at a rate of 10 kHz using a CED 1401 interface and Signal software (Cambridge Electronical Design, Cambridge, UK). Current traces were digitally filtered off-line. The power of the oscillations was measured by performing fast Fourier transformations over five consecutive 10-s traces. Cross-correlation analysis between field potentials was made over five consecutive traces, using Spike2 software (Cambridge Electronic Design).

Data are expressed as means ± SE. Unless otherwise indicated, statistical comparisons were made between experimental groups, using unpaired Student's t-test. The significance criterion was P < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

PV/ mice are not distinguishable from wild-type litter mates

The life span, growth, and breeding of PV/ mice did not differ from those of wild-type (PV+/+) and heterozygous (PV+/) animals (Schwaller et al. 1999); the three genotypes were likewise indistinguishable with respect to behavior and physical activity under standard housing conditions. Light microscopic analysis of hematoxylin/eosin-stained brain sections revealed no histological differences between PV/ and wild-type animals.

Basket cells in PV/ mice manifest no PV immunoreactivity

The B4 lectin from VVA interacts with N-acetylgalactosamine residues alpha-linked to serine or threonine residues in cell-surface glycoproteins. The vast majority of VVA-labeled cells within the hippocampal formation are known to be GABAergic and to express PV (Drake et al. 1991), and in this respect, they resemble previously described VVA-labeled neurons in cerebral cortex (Lüth et al. 1992). Thus VVA staining was used as a means of visualizing, in PV/ mice, the neuronal subpopulation that would normally express PV. Parallel sections were stained either with a PV-specific antiserum or with peroxidase-conjugated VVA. Images taken from the hippocampus of a wild-type mouse are depicted in Fig. 1. These reveal the presence not only of PV-positive cell bodies but also of a diffuse staining within the s. pyramidale, which demonstrates site specificity of the synaptic terminals (Freund and Buzsáki 1996). As expected, sections derived from PV/ mice remained unstained after incubation with the PV antiserum, but distribution and number of VVA-labeled neurons was unchanged. For the quantification, only cells with intact somata localized in the s. pyramidale and adjacent regions (s. oriens, s. radiatum) were counted. The number of perineuronal net-positive cells per millimeter pyramidal layer length was 6.61 ± 0.66 (n = 3) for PV+/+ and 6.29 ± 0.31 (n = 3) for PV/ mice. There was no significant difference between the two groups (P = 0.69). At higher magnifications, no differences in the structure of the perineuronal nets were apparent in PV/ animals (data not shown). Preservation of the perineuronal nets around the cortical neurons of PV-deficient mice has likewise been demonstrated by fluorescence microscopy using the Wisteria floribunda agglutinin (Haunso et al. 2000). In both, PV/ and PV+/+ mice, the perineuronal extracellular matrices were revealed as lattice-like structures around cell soma and dendrites. We conclude that the cell bodies of interneuron types that contain PV in wild-type animals were likewise present and of normal appearance in PV/ mice but lacked this Ca²⁺-binding protein. Given the absence of a band of VVA staining in s. pyramidale in both groups, we cannot make any predictions on the number or distribution of functional synaptic terminals in PV/ mice.

View larger version (97K): [in this window] [in a new window]   Fig. 1. CA1 basket cells are present in PV/ mice but are devoid of parvalbumin (PV). Micrographs of the CA1 area in PV+/+ (left) and PV/ mice (right), immunostained for PV (top) and labeled with VVA (bottom). Basket-type interneurons are present within the hippocampus of PV/ mice (evidenced by the staining characteristics of the extracellular matrix), but do not contain PV.

Single inhibitory postsynaptic responses are not affected by PV deficiency

The most accurate way to investigate the presynaptic role of PV would consist of paired recordings of unitary IPSCs elicited by an interneuron that normally would express PV, like in cerebellar basket cells (Caillard et al. 2000). In the hippocampus, however, it is currently impossible to identify with certainty those interneurons normally expressing PV in wild-type mice, when recording from slices of PV/ mice. Thus we recorded monosynaptic GABA_Aergic IPSCs evoked by stimulation of the near s. pyramidale. Under our experimental conditions, the postsynaptic responses were completely blocked by 20 µM bicuculline methiodide and therefore mediated by GABA_A receptors (data not shown). In current-clamp mode, the maximal IPSP was determined using stimuli of stepwise increasing amplitude. The maximal IPSP peak amplitude did not differ between the groups (-14.5 ± 0.7 mV for 20 cells out of 8 PV+/+ mice; -13.3 ± 0.8 mV for 20 cells out of 7 PV/ mice) and was reached at 30 ± 1 V for PV+/+ and at 33 ± 1 V for PV/ (n.s.). The underlying IPSC was recorded in the same cells at the maximal IPSP stimulus intensity by switching to single-electrode voltage-clamp mode (Fig. 2A). At this stimulus intensity, the GABA_Aergic IPSC showed a late current reversal, probably due to intracellular chloride accumulation and consequent depolarization of the reversal potential of the GABA_Aergic, mixed chloride/bicarbonate conductance (Davies and Collingridge 1993; Kaila 1994). The peak amplitude was not different between the groups (1.2 ± 0.4 nA for PV+/+; 0.8 ± 0.1 nA for PV/, n.s.). The GABAergic neuron population activated by this stimulus intensity is unlikely to be confined to interneurons with perisomal synapses. To improve selectivity of the stimulus, we reduced the stimulus intensity to a level (~16% of the maximal IPSP stimulus intensity) that elicited an IPSP with an amplitude ~30% of the maximal response. Such a stimulus will excite a smaller area of tissue, is better confined to the pyramidal layer, and will recruit a higher proportion of axons from cells that normally contain PV. IPSCs elicited by low-intensity stimulus did not show current reversal (Fig. 2B). The charge carried by the IPSC (measured as area under the curve, excluding the initial artifact) did not differ between the groups (7.7 ± 1.6 pC for 20 cells out of 8 PV+/+ mice; 8.1 ± 1.1 pC for 20 cells out of 7 PV/ mice, n.s.). The monophasic IPSC decay was fitted, starting at 15 ms after the onset of stimulation (t₀; see Fig. 2B), to an exponential function of the form: A(t) = A_max exp((t₀  t)/), where A_max is the amplitude as extrapolated to t₀ and  is the time constant of decay.  did not differ between the groups (23 ± 4 ms for PV+/+ mice; 22 ± 2 ms for PV/ mice, n.s.).

View larger version (17K): [in this window] [in a new window]   Fig. 2. Postsynaptic inhibitory responses. A: single-sweep voltage-clamp recording of an inhibitory postsynaptic current (IPSC) evoked in a cell from a PV+/+ mouse by stimulation of the stratum pyramidale at the intensity required to evoke the maximal inhibitory postsynaptic potential (IPSP) amplitude intensity when recorded in current-clamp. The large amplitude and the current reversal indicate that a large population of interneurons is activated. B: an IPSC recorded in the same cell at a stimulus intensity that evokes an IPSP with an amplitude ~30% of the maximal IPSP amplitude. At this intensity IPSC decay was monophasic and could be fitted with an exponential function of the form: A(t) = Amax exp((t0  t)/), where Amax is the amplitude as extrapolated to t0 and  is the time constant of decay.

Paired-pulse modulation of IPSCs is not affected by PV deficiency

To determine whether the residual Ca²⁺-dependent facilitation (Kamiya and Zucker 1994) of GABA release was affected by the presence of presynaptic PV, pulse pairs at varying inter-pulse intervals were applied (Fig. 3A). Experimental conditions were chosen that favor low release probabilities (Lambert and Wilson 1994; Thomson 1997) and minimize presynaptic suppression of GABA release. For quantification of IPSC modulation by prior activation, the conditioned IPSC was isolated from the conditioning IPSC by digital subtraction of the residual component of the conditioning IPSC remaining at the time of the second stimulus. The charge carried by the conditioned IPSC (normalized to the charge of the conditioning IPSC) was maximally facilitated at 3-5 ms and maximally suppressed at 200-500 ms (Fig. 3B). Paired-pulse modulation of IPSCs did not differ between PV+/+ and PV/ at any interval tested (Fig. 3B). The lack of a change in paired-pulse modulation of IPSCs suggests that the presence of PV does not significantly affect the presynaptic calcium transient induced by a single stimulus. The paired-pulse IPSC suppression at long intervals (Lambert and Wilson 1994; Pearce et al. 1995) may be due to unblocked presynaptic receptors, such as presynaptic metabotropic glutamate receptors (but see Hefft et al. 2002) and cannabinoid receptors (Hajos et al. 2000); but this suppression was unaffected by the lack of PV.

View larger version (21K): [in this window] [in a new window]   Fig. 3. Paired-pulse modulation of IPSCs. A: single-sweep voltage-clamp recordings of double-pulse stimulation at 30% of the maximal IPSP stimulus intensity, at intervals of 3, 25, 100, and 200 ms in the same cell as in Fig. 2. The IPSC is facilitated at short intervals and suppressed at long intervals. B: the conditioned IPSC was isolated by digitally subtracting a single IPSC. The charge carried by the 2nd IPSC (normalized to the charge of the 1st IPSC) is presented as a function of the time interval between paired pulses for 14 cells from 8 PV+/+ mice () and in 14 cells from 7 PV/ mice (). Paired-pulse modulation of the IPSC charge was not different between the groups. Error bars indicate SE.

PV deficiency enhances frequency-dependent facilitation of IPSC trains

To ascertain whether PV had an effect on prolonged or repetitively induced Ca²⁺ transients within the presynaptic terminal, we delivered trains of 10 stimuli. The gradual build-up of the intracellular Ca²⁺ concentration ([Ca²⁺]_i) in the presynaptic terminal during repetitive stimuli is likely to depend on the frequency of stimulation, similar to the situation previously observed in fast-twitch muscles (Schwaller et al. 1999). Thus trains were delivered at different frequencies (0.1, 1, 10, 20, 33, 50, and 100 Hz). At 1 and 10 Hz, IPSC suppression occurred. At frequencies >20 Hz, IPSCs gradually built up with successive stimuli and attained a steady level within 10 stimuli (Fig. 4A). For this cell, the total charge carried by 10 IPSCs at 33 Hz was 188% of that carried by 10 IPSCs at 0.1 Hz. The relative IPSC amplitude measured 10 ms after the stimulus for the nth IPSC is given in Fig. 4B for both groups. At frequencies >20 Hz IPSC facilitation was stronger in seven cells from 6 PV/ mice than in nine cells from 8 PV+/+ mice. The relative difference increased with the number of stimuli and was significant only after four to five stimuli. This difference could not be explained by a slower IPSC decay because the time constant of current decay did not differ between the groups. Figure 4C gives the facilitation/suppression (mean amplitude of the last 5 IPSCs normalized to that of the first) for each group as a function of frequency. Repetitive IPSC facilitation at frequencies >20 Hz was significantly higher in cells from PV/ mice than in those from PV+/+ mice, with the difference at 33 Hz (220%) being relatively greater than that at 100 Hz (46%).

View larger version (22K): [in this window] [in a new window]   Fig. 4. Frequency-dependent IPSC modulation with repetitive stimulation. A: response of a cell in voltage clamp (single sweep) to a train of 10 stimuli at 30% of the maximal IPSP stimulus intensity, delivered at frequencies of 20, 33, 50, and 100 Hz. IPSCs build up gradually to a steady level. B: IPSC amplitude for every nth stimulus (normalized to that of the first), for pulse trains delivered at different frequencies in 9 cells from 8 PV+/+ mice (, - - -) and in 7 cells from 6 PV/ mice (, ). Error bars indicate SE; *, a significant difference at P < 0.05. Lines are smooth fits. C: maximal IPSC amplitude (mean of the last 5 stimuli normalized to that of the first) expressed as a function of pulse-train frequency in the same cells as in B. The greatest relative difference between PV/ (, ) and PV+/+ (, - - -) was found at 33 Hz. Error bars indicate SE; *, a significant difference at P < 0.05. Lines are smooth fits.

Inhibition-based gamma oscillations are facilitated by PV deficiency

In the hippocampus, GABAergic interneurons are involved in coherent network oscillations at frequencies in the gamma band (>30 Hz), which are driven by rhythmic IPSCs (Penttonen et al. 1998; Whittington et al. 1995). The hippocampus-wide network of mutually interconnected PV-containing interneurons (Fukuda and Kosaka 2000) forms a likely substrate for inhibition-based gamma oscillations. Given that the greatest differences in repetitive IPSC facilitation were observed at gamma frequencies, our prediction was that inhibition-based gamma oscillations would be affected in PV/ mice. Assuming that GABAergic terminals from PV-containing interneurons in area CA3 are similar to those in area CA1 (G. Buzsaki, personal communication), we tested this in vitro using the kainate model of inhibition-based gamma oscillations. At submicromolar concentrations, kainate selectively depolarizes interneurons above firing threshold (Cossart et al. 1998), facilitates GABA release between interneurons (Cossart et al. 2001), and induces persistent oscillations that are driven from area CA3 (Hajos et al. 2000; Traub et al. 2000). Extracellular field potential recordings showed that bath-applied kainate (100 nM) caused robust oscillations in area CA3 (Fig. 5A, top). Current-clamp recordings from CA3 pyramidal neurons showed rhythmic IPSPs with amplitudes within the same range as the IPSPs evoked in CA1 neurons by low-intensity stimulation (Fig. 5A, bottom). Simultaneous recordings demonstrated that the extracellular field positivities in s. pyramidale coincide with the on-going phase of the IPSP (Fig. 5B) and that their peak amplitudes correlate with the IPSP amplitudes (Fig. 5C) and therefore most likely reflect the perisomatic population IPSCs (Buhl et al. 1998). Pharmacologically induced increases and decreases of IPSC amplitude resulted in increases and decreases, respectively, of the amplitude of carbachol-induced gamma oscillations in vitro (Stenkamp et al. 2001). The amplitude of the gamma oscillation can therefore be used as a measure of rhythmic IPSC amplitude. Because the maximum power (~35 Hz) was highly variable within animals, power spectra were calculated from 10 slices from both ventral hippocampi for each animal. Maximum power of the kainate-induced oscillation was approximately three times higher in 70 slices from seven PV/ mice than that in 70 slices from seven PV+/+ mice (412 ± 53 vs. 138 ± 26 µV², P < 0.0001). The averaged power spectra from both genotypes revealed the power of PV/ slices to be three to four times larger than that in slices from PV+/+ mice across the full range of the gamma frequency band (Fig. 5D). The dominant frequency was not different (35 ± 1 vs. 36 ± 1 Hz) between genotypes. For slices with a maximum power >10 µV², there was no relationship between dominant frequency and power (R = 0.03).

View larger version (33K): [in this window] [in a new window]   Fig. 5. PV-deficiency facilitates kainate-induced gamma oscillations. A: kainate (100 nM) induces high-frequency network oscillations in the hippocampal CA3 region. Extracellular field potential recording from s. pyramidale of CA3b/c (top) shows rhythmic positivities and intracellular current-clamp recording from a pyramidal neuron near the extracellular electrode (bottom) shows rhythmic IPSPs. Cells fired (if slightly depolarized) on a minority of the cycles. B: field positivities coincide with the downstroke of IPSPs and action potentials precede the field positivity. C: the amplitude of the field positivities correlates with the amplitude of the IPSP (data from a 5 s recording from the cell in B), indicating that the extracellular oscillation reflects population IPSCs. D: average power spectrum of the oscillation induced by 100 nM kainate in 70 slices from 7 PV/ mice (black line) and in 70 slices from 7 PV+/+ mice (gray line). PV-deficiency facilitates gamma (>30 Hz) power by 3-4 times control values. Data are means of pooled slices, dotted lines indicate SE. E: same slices as in C after addition of atropine (5 µM), CGP 55845A (1 µM), naloxone (5 µM), and elevation of MgCl2 to 3 mM. PV-deficiency increases gamma oscillation power ~2- to 3-fold compared with values determined in wild-type mice.

Dual field potentials recordings (0.5 mm apart) in slices with clear gamma oscillations showed that kainate-induced gamma oscillations were tightly phase-locked, although amplitudes could vary for each cycle (Fig. 6A). Despite significant differences in power of the oscillation, cross-correlation analysis showed no differences between the groups (cross-correlation coefficient was 0.65 ± 0.05 at 1.1 ± 0.3-ms phase difference for 9 slices from 7 PV/ mice vs. 0.67 ± 0.04 at 1.0 ± 0.2 ms for 10 slices from 7 PV+/+ mice). Spatial synchronization of kainate-induced gamma oscillations did not depend on the power of the gamma oscillation (Fig. 6B).

View larger version (28K): [in this window] [in a new window]   Fig. 6. PV-deficiency does not affect coherence of gamma oscillations. A: field potentials simultaneously recorded in s. pyramidale of CA3 (0.5 mm apart) show tight phase-locking of events but also a large variability in the amplitude. B: cross-correlation analysis for 9 slices from 7 PV/ mice () and 10 slices from 7 PV+/+ mice () showed no differences in phase difference. The maximum cross-correlation coefficient as a function of maximum power of the gamma oscillation showed no relation with the maximum power in the gamma frequency band.

To reproduce accurately the experimental conditions used to quantify facilitation of IPSCs, we also analyzed the kainate-induced gamma oscillations in the presence of CGP 55845A, atropine, naloxone, and elevated MgCl₂. This treatment shifted the dominant frequency toward lower frequencies (~30 Hz) in both genotypes and maximum power of the kainate-induced oscillation under these conditions was 479 ± 53 µV² for PV/ and 192 ± 47 µV² for PV+/+, P < 0.02 (Fig. 5E). Therefore under both "normal" conditions and low release probability/minimal presynaptic suppression conditions, the kainate-induced inhibition-based oscillations were strongly enhanced in the absence of PV. Small-amplitude gamma oscillations were occasionally observed even in normal ACSF prior to kainate administration. The average power from 30 to 40 Hz was 2.0 ± 0.5 µV² for PV/ and 0.7 ± 0.1 µV² for PV+/+, P < 0.01. Thus even without kainate, gamma oscillations were more prominent in slices from PV/ mice.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The principal results of this study are that PV deficiency in hippocampal interneurons leads to an increased IPSC facilitation with repetitive stimulation, consistent with greater use-dependent build-up of [Ca²⁺]_i in the presynaptic terminals of PV-deficient interneurons, and consequently to stronger inhibition-based gamma rhythms.

The staining with VVA-peroxidase complex demonstrates the presence of neurons that would normally express PV (Drake et al. 1991) in the hippocampus of PV/ mice, whereas staining for PV confirmed that they were devoid of PV. These neurons are likely to make functional synaptic contacts with their normal targets because the IPSC amplitude and kinetics were not different between genotypes; the distribution of the GABA_A receptor subunit 1, a typical marker for hippocampal PV-expressing basket cells (Klausberger et al. 2002), was not different between genotypes (J.-M. Fritschy, personal communication); the distribution of the GABA_A receptor subunit 2, present at synaptic contacts between axo-axonic PV-expressing cells (Nusser et al. 1996) and the initial segment of pyramidal cell axons was not different between genotypes (J.-M. Fritschy); and the functional connectivity between cerebellar basket cells and Purkinje cells was not different between genotypes (Caillard et al. 2000). The theoretical contributions to the effects observed in PV/ mice of either anatomical changes in the terminals of "PV neurons" innervating the recorded pyramidal cells, or of subtle alterations in the presynaptic machinery induced by the PV-deficiency, cannot be excluded and are currently under investigation.

Because it is currently impossible to specifically identify those hippocampal interneurons normally expressing PV in wild-type mice in slices from PV/ mice, we used monosynaptic GABA_Aergic IPSC in CA1 principal cells as a measure of Ca²⁺-dependent GABA release. Mild stimulation of the s. pyramidale activates a mixed population of GABAergic terminals, approximately half of which are PV positive (Ribak et al. 1990). Therefore differences between the genotypes detected using this method, most likely underestimate the PV-deficiency-related changes in PV-expressing synaptic terminals.

Presynaptic PV, as a slow calcium buffer like EGTA, is likely to have little effect on unconditioned transmitter release (Adler et al. 1991). Indeed, unconditioned monosynaptic IPSCs were not different in cells from PV-deficient mice. This is consistent with the lack of an effect of PV on basal [Ca²⁺]_i levels (Lee et al. 2000; Schwaller et al. 1999), Ca²⁺ currents (Chard et al. 1993), and the expression of GABA_A receptors subunits (J.-M. Fritschy).

In normal ACSF, paired activation of putative basket cells and axo-axonic cells results in (presumably presynaptic) paired-pulse suppression of IPSP/Cs in CA1 pyramidal neurons (Ali et al. 1999; Buhl et al. 1995). Under the conditions of reduced release probability (Lambert and Wilson 1994; Thomson 1997) and minimal presynaptic suppression used in this study, the IPSCs exhibited paired-pulse facilitation at short intervals, similar to that found for CA3 neurons (Lambert and Wilson 1994). It is likely that this short-term facilitation results from residual Ca²⁺ in the synaptic terminal (Kamiya and Zucker 1994). As a slow-onset Ca²⁺ buffer, PV does not affect the peak amplitude of the Ca²⁺ transient but accelerates its initial decay (Lee et al. 2000; Schwaller et al. 1999). Given the slow kinetics of the buffer (Schwaller et al. 1999), we predicted therefore that the reduced Ca²⁺ buffering and subsequent accelerated build up of presynaptic [Ca²⁺]_i in the absence of PV (Lee et al. 2000) would favor paired-pulse facilitation of GABA release at intervals between 20 and 300 ms. Indeed, Caillard et al. (2000) demonstrated, for intervals between 30 and 100 ms, increased paired-pulse facilitation of the basket cell triggered IPSC in Purkinje cells from PV/ mice. However, under our conditions, the paired-pulse IPSC facilitation was not enhanced in CA1 neurons from PV/ mice. This suggests that compared with the cerebellum, the presynaptic terminals of hippocampal PV-interneurons have either lower net stimulus-induced Ca²⁺ influx and/or a more efficient Ca²⁺ extrusion. Alternatively, the PV concentration in cerebellar basket cells may be higher than in hippocampal basket cells.

The predicted enhanced facilitation of GABA release in cells from PV/ mice only became apparent in the hippocampus after repetitive stimulation at high frequencies, which will result in a progressive build-up of presynaptic [Ca²⁺]_i if the interval between stimuli is shorter than the [Ca²⁺] relaxation time constant as modeled for the effect of PV in neurons by Lee et al. (2000). An analogous process of accelerated Ca²⁺ build-up occurred in fast-twitch muscles of PV/ mice, which had a faster build-up of tetanic tension when stimulated at frequencies >20 Hz (Schwaller et al. 1999). The slow exogenous Ca²⁺ buffer EGTA has a similar effect on excitatory postsynaptic potentials (EPSPs) (Adler et al. 1991; Hochner et al. 1991). The effect of PV was maximal at ~33 Hz for IPSCs and ~20 Hz for tension build-up in fast-twitch skeletal muscle (Schwaller et al. 1999). The limited role of PV at higher frequencies can be explained by its limited buffer capacity. Once all Ca²⁺-binding sites are occupied, any additional Ca²⁺ influx will increase [Ca²⁺]_i and output (GABA release or force) becomes independent of the presence of PV (Lee et al. 2000; Raymakers et al. 2000).

The maximal effect of PV on repetitive IPSC facilitation in CA1 neurons was observed at frequencies in the gamma frequency band (30-80 Hz). Coherent gamma oscillations are associated with higher cognitive processing (Engel and Singer 2001) and are most probably driven by rhythmic IPSCs (Penttonen et al. 1998). In vitro gamma oscillations induced in the hippocampus by metabotropic glutamate agonists (Whittington et al. 1995) in area CA1 and by carbachol (Fisahn et al. 1998) or kainate (Hajos et al. 2000; Traub et al. 2000) in area CA3, are likewise paced by fast GABA_Aergic IPSCs. The hippocampus-wide network of mutually interconnected PV-containing interneurons (Fukuda and Kosaka 2000) is likely to play a major role in mediating inhibition-based gamma oscillations. Assuming that GABAergic terminals in area CA3 are similar to those in area CA1 (G. Buzsaki, personal communication) and given that rhythmic IPSP amplitudes in CA3 neurons were of comparable amplitude as IPSPs evoked at low stimulus intensity in CA1 neurons, we predicted that the absence of PV would increase the amplitude of the rhythmic IPSCs during gamma oscillations. Indeed, the power of both spontaneous and kainate-induced gamma oscillations was increased in slices from PV/ mice under conditions of low release probability/minimal presynaptic suppression as well as in normal ACSF. This is in line with the increase in gamma power observed in EEG recordings from mice deficient for the Kv3.1 potassium channel highly expressed in PV-containing interneurons and was associated with increased GABA release (Joho et al. 1999). Further support comes from the finding of a decrease in gamma power on suppression of IPSCs after application of cannabinoid receptor agonist (Hajos et al. 2000). The effect of PV deficiency on the gamma oscillation was not dependent on release probability as determined by presynaptic modulation or by extracellular Mg²⁺ but most likely due to a PV-related change in presynaptic calcium homeostasis.

The absence of a direct relationship between gamma power (representing IPSC amplitude) and dominant frequency was surprising, because Whittington et al. (1995) showed that the frequency of IPSC-paced interneuron gamma oscillations induced by metabotropic glutamate receptor activation increased with IPSC amplitude. However, similar uncoupling of maximal power and dominant frequency was described for gamma oscillations induced by carbachol (Hack et al. 2000) or kainate (Hajos et al. 2000). A model study of Traub et al. (2000) suggested that the frequency of kainate-induced oscillations was determined by a build-up of activity in a network of electrotonically coupled pyramidal cell axons after a reset by synchronized IPSCs. Conscious information processing has been associated with changes in coherence of gamma oscillations (Engel and Singer 2001; Miltner et al. 1999). In the present study, coherence of the gamma oscillation was not correlated with the oscillation power and was not affected in PV deficient mice despite gross changes in power. A similar absence of a relation between power and coherence of carbachol-induced gamma oscillations was observed in the isolated guinea pig brain (Dickson et al. 2000).

Functional implications

Under normal conditions, the gradual depression of IPSC amplitude with prolonged high-frequency activation of PV-expressing basket type interneurons and/or axo-axonic interneurons (Maccaferri et al. 2000), leads to a steady inhibitory potential in the pyramidal cell (Ali et al. 1999; Buhl et al. 1995). PV thus maintains the strength of a perisomal synapse near its resting level, allowing integration of EPSPs against a steady background of inhibition; this may increase the sensitivity of the integration process.

The power of gamma oscillations increases with attention and recognition in human subjects (Muller et al. 2000) and with sensory information processing in rats (Penttonen et al. 1998). The increase in gamma power in mice deficient for the Kv3.1 potassium channel was associated with better performance in an active avoidance task (Joho et al. 1999). In contrast, the cannabinoid-induced decrease of gamma power in rats (Hajos et al. 2000) may explain the effects of cannabis on cognitive performance. The presence of PV limits the power of gamma oscillations. On the one hand, considering the role of gamma in cognitive functions, this seems like a hindrance. No obvious behavioral changes have been observed in PV/ mice so far, but more targeted behavioral tasks aimed to specific paradigms (memory, learning) are currently being explored. On the other hand, considering the hyper-synchronous gamma activity at the onset of kainate-induced epileptiform discharges (Medvedev et al. 2000), the limitation of the gamma power by PV might be a safeguard against the onset of epileptic activity. This is in line with the increased susceptibility toward pentylenetetrazol-induced seizures observed in PV/ mice (Tandon et al. 1999; B. Schwaller, unpublished observation).