HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) All Versions of this Article: 96/1/175    most recent 01355.2005v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (7) Citing Articles via Google Scholar Google Scholar Articles by Idoux, E. Articles by Moore, L. E. Search for Related Content PubMed PubMed Citation Articles by Idoux, E. Articles by Moore, L. E.

Oscillatory and Intrinsic Membrane Properties of Guinea Pig Nucleus Prepositus Hypoglossi Neurons In Vitro

¹Laboratoire de Neurobiologie des Réseaux Sensorimoteurs, Centre National de la Recherche Scientifique (CNRS)-Université René Descartes (Paris 5) Unité Mixte de Recherche (UMR) 7060, Paris, France; ²Département de Neurosciences Fondamentales, Faculté de Médecine, Université de Genève, CMU, Geneva, Switzerland; and ³Physio-Pathologie des Réseaux Neuronaux du Cycle Veille-Sommeil, CNRS-Université Lyon 1 UMR 5167, Lyon, France

Submitted 22 December 2005; accepted in final form 1 April 2006

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Numerous models of the oculomotor neuronal integrator located in the prepositus hypoglossi nucleus (PHN) involve both highly tuned recurrent networks and intrinsic neuronal properties; however, there is little experimental evidence for the relative role of these two mechanisms. The experiments reported here show that all PHN neurons (PHNn) show marked phasic behavior, which is highly oscillatory in 25% of the population. The behavior of this subset of PHNn, referred to as type D PHNn, is clearly different from that of the medial vestibular nucleus neurons, which transmit the bulk of head velocity-related sensory vestibular inputs without integrating them. We have investigated the firing and biophysical properties of PHNn and developed data-based realistic neuronal models to quantitatively illustrate that their active conductances can produce the oscillatory behavior. Although some individual type D PHNn are able to show some features of mathematical integration, the lack of robustness of this behavior strongly suggests that additional network interactions, likely involving all types of PHNn, are essential for the neuronal integrator. Furthermore, the relationship between the impulse activity and membrane potential of type D PHNn is highly nonlinear and frequency-dependent, even for relatively small-amplitude responses. These results suggest that some of the synaptic input to type D PHNn is likely to evoke oscillatory responses that will be nonlinearly amplified as the spike discharge rate increases. It would appear that the PHNn have specific intrinsic properties that, in conjunction with network interconnections, enhance the persistent neural activity needed for their function.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
The CNS computes and processes both digital and analog signals: some of these operations have been elegantly described as mathematical algorithms, such as multiplication in sound localization (Pena and Konishi 2004), division for certain inhibitory inputs (Borg-Graham et al. 1998), linear and nonlinear addition in the spinal cord or cortex (Shu et al. 2003), and finally integration of dendritic inputs for persistent memory (Polsky et al. 2004) or of an eye-velocity signal to position in the oculomotor system (Robinson 1975). The "oculomotor integrator" is thought to be the same for all oculomotor networks responsible for fixation, orientation, and stabilization (Goldman et al. 2002). It uses eye-velocity-coded saccadic commands and head-velocity vestibular signals to generate eye-position commands (Fukushima and Kaneko 1995; Moschovakis 1997). For the horizontal plane, the prepositus hypoglossi nucleus (PHN) is the principal structure of the oculomotor integrator (Mettens et al. 1994) as shown by comparison among network modeling and pharmacological for lesion experiments (Arnold et al. 1999; Cannon and Robinson 1987; Chéron and Godaux 1987; Robinson 1975; Skavenski and Robinson 1973). The PHN is in the rostral medulla (McCrea and Horn 2005), near the medial vestibular nucleus (MVN), which is one of its major inputs. Theoretical models based on intrinsic cellular properties (Egorov et al. 2002; Loewenstein and Sompolinsky 2003; Marder et al. 1996), the finding of rostrocaudal gradients for neurons showing position versus velocity signals (Delgado-Garcia et al. 1989) and extensive studies related to the oculomotor integrator of goldfish (Aksay et al. 2000, 2001, 2003a,b; Major et al. 2004a,b; Seung et al. 2000) suggest that the stability and robustness of the neuronal integrator (Galiana and Outerbridge 1984; Goldman et al. 2003; Koulakov et al. 2002) require both intrinsic properties and network interconnections (Anastasio 1998; Arnold and Robinson 1991, 1997; Blazquez et al. 2003; Cannon et al. 1983).

PHN neurons (PHNn) have been functionally characterized in vivo (Delgado-Garcia et al. 1989; Lopez-Barneo et al. 1982) and in vitro either morphologically (McCrea and Baker 1985b) or electrophysiologically with intracellular recordings (Bobker 1994; Bobker and Williams 1989–1991, 1995). The present paper links these different classifications and describes a highly oscillatory behavior shown by 25% of the PHNn and not by any MVN neuron (MVNn) (Beraneck et al. 2003; Ris et al. 2001). The whole PHN includes as well a greater percentage of phasic neurons (type B) than the MVN. In addition, it was found that the activation of N-methyl-D-aspartate (NMDA) receptors in type B PHNn leads to discrete oscillatory activity, supporting some theoretical studies showing robust neural integration in neurons having NMDA activated bistable dendrites (Goldman et al. 2003; Koulakov et al. 2002). Some of the results have been presented in abstract form (Idoux et al. 2004; Mühlethaler et al. 1996; Serafin et al. 1996a).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Slice preparation

Experiments were carried out on pigmented guinea pigs of both genders (Animal facility of the Geneva University Medical Center and Elevage de la Garenne, Saint-Pierre d'Exideuil). Compared with rats and mice, guinea pigs are far more developed at birth and do not show as marked developmental differences during the first 5 wk after birth (Dobbing and Sands 1970; Nacher et al. 2000; Staudinger et al. 2005). To assess the potential effect of age in these animals, two groups of guinea pigs were used, one 6–9 days and the other 3–5 wk old. Unless stated otherwise, the sections of the paper referred to as qualitative analysis are from the young animals; the quantitative analysis was done on data from the older animals. The results from these two age groups showed no significant differences; nevertheless some of the results will be presented separately to indicate possible developmental differences. All intracellular recordings were done in vitro at 32°C on guinea pig 500-µm-thick brain stem coronal slices with sharp electrodes (120–150 M). These slices were obtained and maintained using previously described procedures (Beraneck et al. 2003; Serafin et al. 1991; Vibert et al. 1999).

The viability of neurons in the brain stem slice could be improved by replacing the normal artificial cerebrospinal fluid (ACSF) by "sucrose-ACSF", i.e. an ACSF-depleted of sodium chloride and loaded with sucrose during the slice preparation. Statistical comparison of all measured parameters of PHNn between slices obtained with either ACSF showed no difference, as previously reported for lateral vestibular nucleus neurons (Uno et al. 2003). Therefore all recordings from similar age groups were pooled together.

Intracellular current-clamp experiments were done as described in Beraneck et al. (2003). The morphology, the response to neuromodulators and a qualitative determination of the cell types were done on 213 PHNn from guinea pigs <2 wk old (see Classification parameters). The quantitative classification program of Beraneck et al. (2003) was used on the remaining 110 neurons to allow a direct comparison with MVNn. Both methods led to comparable proportions of each cell type, therefore the results of all the 323 PHNn were pooled together to compare the relative distribution of cell types between the PHN and the MVN.

Classification parameters

The quantitative classification of Beraneck et al. (2003, 2004) is based on two quantitative parameters that relate to the qualitative classification used by Serafin et al. (1991): the "dAHP", which estimates the size of a double afterhyperpolarization, and the "IA", which is a measure of the rectification of the potential between spikes and possibly related to an inactivating potassium A-like current. In this paper, the IA will be termed AHPR, afterhyperpolarization rectification, because it is, like dAHP, a less ambiguous descriptive term for the membrane potential profile following the spike. Both the dAHP and the AHPR were evaluated using the first derivative of the variation of the membrane potential versus time. According to these parameters, MVNn were divided into two cell types: type A MVNn and type B MVNn. Type A MVNn were the MVNn with no dAHP (= 0 V/s) and a strong AHPR (0.15 V/s). In contrast, type B MVNn were those with a dAHP and no AHPR or a small one (<0.15 V/s). If a neuron had both a non-null dAHP and a strong AHPR (0.15 V/s), then this cell was said to be an intermediate type C MVNn, but this class was found to be <5% of the total population of MVNn (Table 1 and Fig. 2A). To compare neuronal types within the PHN as well as with the MVN, the following parameters were determined for each neuron: dAHP, AHPR, average resting membrane potential (V_m, mV), spontaneous discharge rate (DR, spikes/s), coefficient of variation of DR (C_v, %), amplitude of the after-hyperpolarization (AHP, mV), width (ms) of the spike and finally, the concavity(Cav, mV) and convexity(Cvx, mV) of the interspike interval (see Fig. 1 in Beraneck et al. 2003).

View larger version (47K): [in this window] [in a new window]   FIG. 2. Anatomical distribution of PHNn within the PHN. A: cells with an AHPR 0.15 V/s and a null dAHP are classified as type A-like neurons (including type A and D PHNn), cells with an AHPR <0.15 V/s are classified as type B-like neurons (including B and B+LTS PHNn). The 3 neurons having an AHPR 0.15 V/s and a non-null dAHP would be type C neurons (see Beraneck et al. 2003). B: photomicrograph of the most rostral 500-µm-thick slice used in the present study, showing the location of the prepositus hypoglossi nucleus. The PHN area contains (arrow) a typical neurobiotin-filled PHNn enlarged in the inset. Notice that the border between the medial vestibular nucleus (MVN) and the PHN is particularly difficult to delineate. C: diagram shows the subdivision of the PHN into 6 areas (LD, LI, LV, MD, MI, and MV, respectively) that were used to map the location of the recorded PHNn throughout the nucleus. D: proportions of the 4 neuronal cell types recorded in the PHN (1 type per column) in each of the 6 areas defined in the preceding text at each rostrocaudal slice level (1 level per line). E: cumulated percentages of the various PHN neuronal types (1 type per column) within the nucleus (in this representation the 3 rostrocaudal levels are merged together, see METHODS). D and E: proportions 70% are written with white font on black background, whereas proportion between 40% and 70% are on gray background. g7, genu of the facial nerve; mlf, medial longitudinal fasciculus; 4V, 4th ventricle; D, dorsal; I, intermediate; L, lateral; M, medial and V, ventral.

View larger version (52K): [in this window] [in a new window]   FIG. 1. Qualitative classification of prepositus hypoglossi nucleus neurons (PHNn) and assessment of their conductances. A, 1–3: type A PHNn displayed a tonic regular activity at rest (A1). These neurons had a medium-broad action potential (A2, inset) followed by a single deep afterhyperpolarization (AHP, A2) leading to an inflection (see ) that is consistent with the presence of an IA current. Application of a depolarizing current step superimposed on a steady-state hyperpolarization (here and in the following figures, , represent the resting membrane potential) was used to indicate the presence of an IA current as an initial short-duration lag, which delayed the firing of the cell (arrow in A3). B,1–3: type B PHNn showed tonic regular activity at rest (B1) and had a short-duration action potential (B2, inset) followed by an early fast and a delayed slower AHP (single and double in B2). These cells showed long-lasting sodium-dependent sub-threshold plateau potentials (B3, *) in response to a short-lasting depolarizing current pulse superimposed on a slight hyperpolarization just sufficient to eliminate their spontaneous firing. The plateau potential was still present in absence of calcium but totally suppressed by TTX at 1 µM (B3, insets). C, 1–3: type B+LTS PHNn showed tonic regular activity at rest (C1); they had a medium-duration action potential (C2, inset) followed by a biphasic afterhyperpolarization potential (AHP; single and double in C2). These cells were mainly characterized by the presence of a potent TTX-insensitive low-threshold calcium spike (LTS) evoked in response to depolarizing current steps applied while the neuron was hyperpolarized by a steady-state current (C3). The LTS persisted in presence of TTX and was totally suppressed by cadmium (C3, insets). D, 1–3: type D PHNn displayed in most cases a slow irregular spontaneous activity at rest (D1). These neurons had a relatively long duration action potential (D2, inset) followed, like type A PHNn, by a single large AHP (D, 2 and 3). They could easily and best be distinguished from all other neuronal types by their capacity to fire in clusters of action potentials ( in D3) intermingled with sub-threshold membrane oscillations (closed squares in D3) when depolarized from rest. In general, such sub-threshold oscillations were also observed at the resting membrane potential (D, 1 and 2). A4–D4: photomicrographs of neurobiotin-injected PHNn of each type.

Ramp stimulations over a 0.3-nA range were done both at rest and during a hyperpolarization of the neuron 10 mV below its resting potential using five slopes, so that 0.3 nA was injected within 5,000, 3,400, 1,800, 600, or 200 ms. Adaptation of the discharge was observed in most neurons and was measured, for the fourth slope, as the difference between the peak instantaneous frequency obtained when the current reached its maximal value and the steady-state value of the instantaneous frequency at the end of the plateau; this difference was called the overshoot (O₆₀₀, spikes/s). The firing threshold of the neuron (FThr, mV) was computed from the lowest slope ramp (t = 5,000 ms), when initiated from a hyperpolarized value. In spontaneously active neurons, FThr is more negative than V_m. Conversely in silent neurons, FThr is positive to V_m.

In contrast to the large signal ramp stimulation, small signal discrete sinusoidal stimulation (0.07–0.09 nA) was done either individually with a set of 16 sine waves at various frequencies (0.2–50 Hz) for 5 s or with 60 sine waves applied simultaneously with random phases as a pseudo-white noise signal. These measurements were done during a hyperpolarization to abolish action potentials, at the resting potential, and at a depolarized potential (10 mV above the resting potential). Both the modulation of V_m (<10 mV) and the instantaneous frequency (IF) of action potentials were divided by the input current and shown as magnitude and phase plots of the impedance (Z, M) and instantaneous frequency transfer functions (IF/I, spikes · s^–1 · nA^–1) (Ris et al. 2001), respectively. Step and sinusoidal membrane potential responses were used to construct a subthreshold electrotonic model of type D PHNn using procedures previously described (Saint-Mleux and Moore 2000a,b). Parameters used to compare the voltage-dependent intrinsic filtering properties of the neurons were obtained for the three levels of polarization: the frequency of stimulation at which the resonance occurs (peak frequency, F_peak, Hz) and the amplitude of this resonance relative to 0.2 Hz (Q_R). These two parameters were calculated using the impedance during a hyperpolarization and the instantaneous frequency transfer function in the other cases. Therefore in the tables and text, a subscript related to the actual state is added, respectively H, R, and D for hyperpolarization, rest, and depolarization. For example, F_peak-D refers to the frequency of stimulation at which the resonance occurs for the depolarized case. Means for the magnitude and phase functions were calculated for all the neurons within one cell type, except for the A PHNn, in which two cells were excluded because their discharge rate was <10 Hz, which could not be adequately modulated.

In addition to the instantaneous frequency transfer function obtained from the time intervals between each successive spike (Moore et al. 2004; Ris et al. 2001), an equivalent spectral method was used, based on the impulse functions derived from the firing times. In that case, spikes were treated as Dirac delta functions where the spike train is of the form

where {tk} at k = 1 to N are spike-threshold times. A spectral analysis of this impulses train provided the Fourier component at the stimulating frequency, from which magnitude and phase were extracted. Indeed, the spike rate transfer function (SRTF) is the ratio of the Fourier transforms of the spike train, S(t) and the injected current, I(t), namely

where T is the stimulus duration and I(f) the Fourier transform of the injected current. This function is less ambiguous than the IF transfer function in its definition of time and will be used in the following text to compare spike rates and membrane potential.

Morphological analysis

At the end of electrophysiological recordings, slices containing neurobiotin-filled neurons (1 per slice side) were immersed in an ice-cold fixative with 3% freshly depolymerized paraformaldehyde in 0.1 M phosphate buffer. Neurobiotin-filled neurons were then visualized using the avidin-biotinylated horseradish peroxydase complex reaction (Vectastain, Elite kit, Vector Laboratories, Burlingame, CA) with 3–3'-diaminobenzidine (Sigma-Aldrich) as a chromogen combined or not with nickel intensification (blue-black vs. brown reaction product). For easier microscopic observation of the morphology of the neurobiotin-filled neurons and to ensure their location within the PHN boundaries, slices were re-sectioned in 50-µm-thick sections using a cryostat. These free-floating sections were then mounted on gelatin-coated slides and counterstained with 1% Neutral Red for the PHN to be better demarcated from the adjacent medial vestibular nucleus and medullary reticular formation. The photomicrographs displayed on Fig. 1 were prepared with Adobe Photoshop 4.0.

To investigate the anatomical distribution of PHNn throughout the nucleus, the latter was first divided into three 500-µm-thick rostrocaudal levels (rostral, intermediate, and caudal levels respectively; an example of the rostral level is shown in Fig. 2B) taking the genu of the facial nerve as a rostral landmark and according to the Gstoettner's atlas of the guinea pig brain (Gstoettner and Burian 1987). At each level, the nucleus was then further subdivided into a medial (M) and a lateral (L) part and dorsoventrally into a dorsal (D), an intermediate (I) and a ventral (V) part (see Fig. 2C). Such a subdivision of the nucleus into 18 areas (6 areas per slice level: LD, LI, LV, MD, MI, and MV, respectively) of equivalent dimensions (around 150 x 150 µm) was used to map the location of 213 PHNn identified unambiguously by their intrinsic membrane properties. At least 10 cells were randomly recorded in each given area allowing the calculation of the relative proportion of each neuronal cell type with respect to the other ones in that area. More recordings were done in the medial areas because type D PHNn were found nearly exclusively in that part of the nucleus. The following is an example of calculation (repeated for each area) done to construct the boxes illustrated in Fig. 2D: in the rostral slice level (1st line in Fig. 2D) a total of 11 cells were recorded in the medioventral (MV) area, including 3 type A (3/11 = 27.3%), 2 type B (2/11 = 18.2%), 1 type B+LTS (1/11 = 9.1%), and 5 type D PHNn (5/11 = 45.4%). The cumulated percentages illustrated in Fig. 2E represent the proportion of a given neuronal cell type in a given area of the nucleus with respect to the other cell types, independently of the rostrocaudal level (in this case the 3 slice levels were merged together). Altogether, a total of 41 cells was recorded in the medioventral (MV) area through the three rostrocaudal slice levels, including 3 type A (3/41 = 7.4%), 14 type B (14/41 = 34.1%), 3 type B+LTS (14/41 = 34.1%), and 10 type D PHNn (10/41 = 24.4%).

Statistical analysis

The whole statistical analysis was done with the Systat 8.0 software (SPSS, Chicago, IL). All results are expressed as means ± SD. Normality of the distribution of each cell parameters' value was tested by a one-sample Kolmogorov-Smirnov test with usual significance threshold (P 0.05). In case of a normal distribution and large-enough sample, parametric tests were done; otherwise the corresponding nonparametric tests were used, namely ANOVA and Kruskal-Wallis ANOVA, respectively, except for two-by-two comparisons, in which cases the Student's t-test and Mann-Whitney U test were preferred. We compared parameters' distribution both between neuronal-types (A, B, B+LTS and D) inside the PHN and between corresponding neuronal-types in MVN and PHN (type A PHNn vs. type A MVNn and type B PHNn vs. type B MVNn). When comparisons concerned the same parameters for different levels of polarization, paired tests were used, either paired parametric (ANOVA followed by paired t-test) or nonparametric tests (Friedman ANOVA followed by signed-rank tests), depending, again, on the normality of the distribution and the size of the sample. Even when the distribution was not normal, the mean and SD are given for consistency.

Neuronal models

The kinetic formulation using x as a generalized kinetic variable (Borg-Graham 1991; Moore and Buchanan 1993; Murphey et al. 1995) is

where I_i is the current in the ith compartment, I_p is the total current due to the different potential-dependent conductances g_p (see following text). I_core is the current between compartments and g_core is the corresponding core conductance, V_i is the membrane potential in the ith compartment, g_soma and I_leak represent a nonspecific leakage conductance and current having V_leak as a reversal potential, A is the total area of the dendritic compartments to the soma, N is the number of compartments, C_soma is the capacitance of the soma. g_p is a generic voltage-dependent ionic conductance with a reversal potential of V_p and the kinetics of which is governed by the unitless variable, x, which has a steady-state value of x. Thus at the half-activation (x = 1/2) voltage, v_x, s_x is the slope of x and t_x is the time constant, _x. Real-time simulations were done with a soma and three dendritic compartments having a distribution of ion channels as described in each simulation. An analytical model was used (Murphey et al. 1995; Rall 1960) for all frequency domain simulations. All simulations were done with Mathematica (Wolfram Research, Champaign, Il). The analytical linearized admittance is given as

where j = –1 and f is the frequency in Hertz. This formulation leads to an active soma and cable (Murphey et al. 1995) having an admittance, Y_a, where L is the electrotonic length

Y_e, the admittance with the electrode properties is

where C_e is the electrode capacitance and R_e is the electrode resistance. Finally, a negative resistance R_ss is added to take into account resistance compensation, giving a total admittance of Y_t = Y_e/(1 + R_ss x Y_e).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Classification of PHNn

The most distinctive features of PHNn in the slice compared with MVNn are their irregularity and oscillatory character. The oscillatory PHNn, referred to as type D, can be easily distinguished from all other neuronal cell types in the PHN and in the MVN by their capacity to fire in clusters of action potentials intermingled with subthreshold membrane oscillations, usually visible at rest or when depolarized (circles for the clusters and squares for the oscillations, respectively in Fig. 1D). In the PHN, the population of recorded neurons was divided into four different neuronal types, namely types A, B (with a subset of B+LTS), and D PHNn. As in the MVNn, type A PHNn (Fig. 1A) were defined qualitatively by broad action potentials and a deep monophasic AHP. Type B PHNn (Fig. 1B) had shorter-duration action potentials and a biphasic AHP. The subset of type B+LTS PHNn (Fig. 1C) were distinguished from type B neurons by bursts of spikes elicited in response to depolarizing steps while maintained hyperpolarized by DC injection and therefore named type B+LTS ("low-threshold spike") PHNn as in the MVN. The proportion of each cell type for both nuclei is summarized in Table 1.

Significantly different distributions of neuronal types within PHNn and MVNn were obtained with a ² test (P < 0.001). As summarized in Table 1, comparison of the proportions of each cell type found in the MVN (Beraneck et al. 2003) and PHN showed that the proportion of type B (including B+LTS neurons) increased while the proportion of type A neurons decreased. Finally, the new type D neurons accounted for 23.5% of PHNn.

As shown by Table 2, type A and B PHNn were not very different for some of their steady-state properties, namely the resting membrane potential (V_m), the spike threshold (FThr), the spontaneous discharge rate (DR) and its coefficient of variation (C_v). As indicated by Table 2, most of these parameters were different for the type D PHNn. The smaller difference between V_m and FThr is consistent with the lower spontaneous spike rate measured for the D cells.

The morphological analysis of neurobiotin-filled PHNn and their local collateralization revealed no striking differences between the neuronal types: all PHNn were variable in shape from triangular to ovoid, had medium- to large-sized somas (15–20 µm in average) with few (generally 2–5) principal dendrites (see Table 1 for values and Fig. 1 for photomicrographs). The axons of the neurons could not always be seen and could never be followed for >50–100 µm. In contrast with the hippocampus, spinal cord, or cerebellum, there are no visible anatomical substructures or cell layers within the PHN (McCrea and Horn 2005). In addition, according to McCrea and Baker (1985a), the initial trajectories of PHNn axons are very variable, and the axons often make loops and turns within the nucleus itself. Hence, it was impossible to know where outside the PHN the axons of the neurobiotin-filled PHNn might project. In a few cases, varicous axonal collaterals were observed within the PHN itself; this confirms the existence of the intra-PHN recurrent network seen by McCrea and Baker (1985a). These neurobiotin-filled neurons were used to determine the distribution of 213 of the recorded neurons as defined in the methods and illustrated in Fig. 2. In summary, type A PHNn (Fig. 2, D and E, 1st column) were slightly more frequently recorded in the medial part. Type B PHNn were present in large proportion everywhere throughout the nucleus intermingled with all other neuronal types, but they were more frequently encountered in its lateral part. Type B+LTS PHNn were rarely recorded in the rostral pole of the nucleus but were more numerous as the recording electrode was moved caudally. A large majority of these neurons were confined in the ventral part of the PHN. Finally at all three slice levels (rostral, intermediate, and caudal), type D PHNn were mainly recorded in the medial part of the nucleus close to the fourth ventricle and increased in number along a ventrodorsal gradient (Fig. 2, D and E, last column). In addition, it is noteworthy that the proportion of type D PHNn also increased in a caudorostral gradient (achieving 60% of the total neuronal population in the more rostromedial part of the nucleus). The different distribution of the four PHNn types through the nucleus (in particular that of B+LTS PHNn and D PHNn) is shown in Fig. 2E where the values in the boxes represent, independently of the rostrocaudal level, the proportion of a given neuronal type in a given area of the nucleus with respect to the other cell types (see METHODS).

Intrinsic and pharmacological properties of PHNn

Although morphologically similar, the different PHNn cell types have different intrinsic membrane properties. After brief low-intensity depolarizing current pulses (10 ms at 0.1 nA) applied to neurons slightly hyperpolarized with a steady-state current to eliminate their spontaneous activity, most of type B PHNn (n = 49/67) responded with long-lasting subthreshold plateau potentials (stars in Fig. 1B3). Such plateau potentials were observed in some type A PHNn (n = 8/18) but only in a few type D (n = 4/19) neurons. The behavior of the plateau potentials is similar to that previously described in the MVN (Serafin et al. 1991). The persistence of such plateau responses in absence of calcium ions and their total suppression following addition of 1 µM TTX (Latoxan, Rosans, France) to the bathing solution (n = 5/5, Fig. 1B3, insets) suggests that they most likely depend on a persistent sodium current similar to that reported in other neurons (Crill 1996).

High levels of 4-aminopyridine (4-AP, 1–5 mM, Sigma-Aldrich, France) suppressed an outward rectification in type A PHNn (n = 4/4, data not shown) similar to that observed in other tissues, which suggests the presence of an I_A-like current in these neurons. The I_A current (McCormick 1991) is a potassium outward current with fast-activating and -inactivating kinetics that is generally suppressed by 4-AP; the IA-like current is important in type A PHNn for both the typical shape of their action potential (Fig. 1A, 1 and 2) and the delay to first spike when depolarized from a hyperpolarized level (Fig. 1A3) by a step of current.

The most striking and distinctive property of type B+LTS PHNn was the presence of a potent TTX-resistant and calcium-dependent (n = 4/4, Fig. 1C3) low-threshold calcium spike evident when the neuron was released from a continuous hyperpolarization, thus removing the inactivation of low-voltage-activated T-type calcium channels (Huguenard 1996).

Type D PHNn subthreshold membrane oscillations were TTX-sensitive (n = 5/5, not shown). The frequencies of the subthreshold membrane potential oscillations and that of the action potentials within the clusters (intracluster frequency) were not fixed but co-varied in a current-dependent manner (Fig. 3B). In contrast, the frequency of occurrence of the clusters (intercluster frequency) was current-independent and remained around 4 Hz. Overall, the frequencies of the subthreshold membrane oscillations at the resting level varied from 20 to 50 Hz, whereas the intracluster frequency co-varied from 15 to 50 Hz. Another distinctive property of type D PHNn was revealed by responses to hyperpolarizing current pulses of increasing amplitude delivered from rest. Although the application of a short-lasting hyperpolarizing small-amplitude current pulse did not interfere with their generally low irregular spontaneous resting activity (Fig. 4A), a stronger hyperpolarization (20 mV) delayed the onset of activity for a long time (double arrowhead in Fig. 4B). Such behavior strongly suggested the removal of the inactivation of a slowly inactivating subthreshold current similar to the potassium I_D current originally described in hippocampal neurons (Bekkers and Delaney 2001; Storm 1988) as well as other CNS neurons (Dekin and Getting 1987; Hammond and Crepel 1992; McCormick 1991; Sah and McLachlan 1992; Spain et al. 1991; Storm 1988; Surmeier et al. 1992)

View larger version (31K): [in this window] [in a new window]   FIG. 3. Analysis of type D PHNn discharge pattern. A, 1–3: voltage dependence of the subthreshold membrane oscillations and clusters evoked by depolarizing current pulses of increasing amplitude at rest in a typical type D PHNn. Depolarizing current steps of increasing amplitude evoke clusters of action potentials () intermingled with subthreshold membrane oscillations (). B: steady-state current dependence (I, nA) of the frequencies (F, Hz) of subthreshold oscillations ( F = 31.4*I +22.5, R2 = 0.79) and intracluster firing ( F = 42.0*I +18.1, R2 = 0.95) in 4 type D PHNn. In contrast, the inter-cluster frequency () appears to be relatively current independent (F 3.9, R2 = 0.00). Note that the horizontal axis is a function of current and not time.

View larger version (28K): [in this window] [in a new window]   FIG. 4. Presence of an ID current in type D PHNn. A and B: hyperpolarizing current steps of increasing amplitude applied at rest (arrowhead, –58 mV) reveal an outward rectification of the membrane potential (double arrowhead) which delays the firing of the cell for several seconds. C, 1–3: application of a 5-s-long depolarizing current step superimposed on a steady-state hyperpolarization shows a delay in the firing (double arrowhead in C1) and then a progressive increase in the spike rate, which are both suppressed by addition of 4-aminopyridine (4-AP) at 50 µm in the bathing solution (C2). C3: increase of the instantaneous frequency (IF) with time (t) (mean values ± SD obtained from 5 traces for successive 500-ms periods) in control solution (squares IF 11.3, R2 = 0.00) and in presence of 50 µM of 4-AP (circles IF = 1.0*t +1.3, R2 = 0.98).

The continuous increase in instantaneous frequency (IF) shown in Fig. 4C3 for type D PHNn was only observed in young animals and was greatly dependent on both the membrane potential and step size. It was clearly not a robust behavior that would be needed if single-cell integration could occur without network interactions. All other properties of the type D neurons were identical in young and old animals. In general, the IF was not constant over time because these neurons typically showed action potential clusters (Fig. 5A), which increased in duration with time after a current step and were interspersed with subthreshold oscillations. It appears that the increase in either cluster duration or instantaneous frequency during a step of current requires type D PHNn to be in a sufficiently hyperpolarized state to allow the I_D conductance to open and then inactivate during a step depolarization.

View larger version (30K): [in this window] [in a new window]   FIG. 5. Membrane potential oscillations in type D PHNn. A: response of a type D PHNn to a depolarizing pulse applied from a hyperpolarized level showing increasing numbers of spikes in clusters. Action potentials have been clipped at –30 mV. The behavior of neurons in older guinea pigs is comparable to that of Figs. 2 and 3 from young animals. O, underlines the oscillations; double arrow, the subthreshold oscillations, which have been zoomed for the offset on the right. B: step responses obtained from the same neuron just below and above threshold showing the voltage dependence of the frequency of inter-spike oscillations. In this instance and in C, the spike frequency decreased during the step pulse. C: step response recorded from the same neuron near resting level showing discrete-like fluctuations as indicated by the plus and minus signs.

Bath-applied NMDA (100 µM, Tocris) depolarized all types of PHNn but induced either oscillatory or plateau like responses only in type B and type D neurons. These oscillatory-like responses were observed in seven of eight type B neurons and two of nine type D neurons. Figure 6 illustrates the effect of NMDA on a type B neuron showing slow transitions with action potential firing in the up state and a short quiescent down state or more oscillatory responses in which spikes occur during the most depolarized phase of the oscillation. In the presence of TTX, the plateau and oscillatory fluctuation responses were clearly independent of voltage-dependent sodium conductances, contrary to those observed without NMDA in type D PHNn.

View larger version (49K): [in this window] [in a new window]   FIG. 6. N-methyl-D-aspartate (NMDA) induced oscillations in a type B PHNn. In every panel, the horizontal arrowhead indicates the resting potential of the cell (–61 mV). All measurements were done in the presence of bath perfused 100 µM NMDA. A1: plateau oscillations of the membrane potential and spike discharge were observed during the injection a hyperpolarizing current of –0.1 nA in the presence of NMDA. Oscillations were not present in absence of an injected current. A2: plateau-like oscillations of the membrane potential are independent of action potential activity because they were observed in presence of 1 µM TTX for same current levels as A1. B1: sinusoidal-like membrane potential oscillations with synchronized spikes at a higher frequency than in A1 were observed with a larger hyperpolarizing current of –0.2 nA in presence of NMDA. B2: oscillatory potential responses in TTX and NMDA show a higher frequency during an increased hyperpolarizing current of –0.2 nA compared with A2. C1: oscillatory or sinusoidal-like membrane potential fluctuations were observed at still a higher frequency in NMDA with –0.3 nA of injected current; however, action potentials were not well synchronized and occurred irregularly after 2 or 3 membrane potential cycles. C2: low-frequency plateau potentials in NMDA and TTX with –0.3 nA of current were found that were interspersed with low-amplitude higher frequency oscillations. D, 1 and 2: fluctuations due to NMDA activation were suppressed by a hyperpolarizing current of –0.4 nA with (D2) or without TTX (D1).

Histamine (100 µM, Sigma-Aldrich) depolarized very clearly all cell types (4/5 type A, 17/18 type B and 7/7 type D). A cholinergic agonist (carbachol 100 µM, Sigma-Aldrich) also depolarized type A and B PHNn (9/9 and 28/29, respectively), whereas its effect on type D PHNn was more complex: 5 cells of 11 were depolarized, whereas the remaining 6 were hyperpolarized.

Responses to large ramp stimuli

Ramp currents having different slopes were applied to evaluate dynamic behavior in the time domain for comparison with frequency domain responses. Both at rest or when starting the ramp of current from a hyperpolarized level, the overshoots of all three different cell types inside the PHN illustrated in Fig. 7 are significant and show no measurable differences. This result is in contrast with those of the MVN neurons where the overshoot of type A is always inferior to that of type B MVNn. Because type A PHNn have a significantly higher overshoot than their MVNn counterparts (Table 3 and Fig. 7), the differences among type A, B, and D PHNn are minimal. Thus the PHNn are similar with regard to their ramp responses, indicating that the active membrane properties determining their phasic behavior is likely to influence the firing behavior of all PHNn.

View larger version (58K): [in this window] [in a new window]   FIG. 7. Response to ramp stimuli. Except for type A MVNn, an overshoot (cf. METHODS) is noticeable for all examples, provided as follows: 1st row, type A (left, PHN; right MVN); 2nd row, type B (same as 1st row); 3rd row left, type D PHNn. Third row right: graph bar (mean ± SD) of the overshoot for the plateau starting after 600 ms (O600-R) for the different types from both nuclei. The value of the O600-R of type A PHNn is significantly higher than found for type A MVNn; however, this is not the case for type B neurons. As a consequence, none of the values of the overshoot are different within a PHNn cell type, leading to a more homogeneous and more phasic nucleus. Left: response of PHNn to ramp stimulation (top: type A PHNn, middle: type B PHNn, and bottom: type D PHNn). Right: response of MVNn to ramp stimulation (top: type A PHNn, middle: type B PHNn). Because no type D neurons exist in the MVN, the last panel shows the values of the overshoot (O600-R, mean ± SD) for each cell type of both nuclei. Note that, unlike type A MVNn, type A PHNn show a pronounced overshoot. The values of sensitivity (kIF, in spike · s–1 · nA–1) for the different slope are significantly different at all slopes between cell types in PHNn, and are, within type B PHNn, significantly increasing with the steepness of the slope (92.03, 94.99, 100.22, 103.02, 112.18 spike · s–1 · nA–1, for ramps with durations of 5,000, 3,400, 1,800, 600, 200 ms, respectively. All of these variables were significantly different with P < 0.01 (n = 43).

View larger version (31K): [in this window] [in a new window]   FIG. 8. Filter properties of PHNn: Bode diagrams of PHNn transfer functions. The Bode diagrams show a log-log plot of the magnitude and log–linear plot of the phase of the transfer functions vs. the stimulation frequency. The transfer function is the ratio of the relevant output (modulation of membrane potential, if no spikes, else modulation of instantaneous firing frequency), and input (current). A: mean magnitude (A1) and phase (A2) of the impedance Z (Z = V/I, membrane potential divided by the applied current) during continuous injection of a hyperpolarizing current to abolish action potentials. #, 0.05 < P < 0.10 (This is not significant difference but shows a strong trend between the response of type D and B); *, P < 0.05 (significant statistical difference between the responses of type D and B). Neither the magnitude component of type A and B nor of type A and D responses nor any of the phase responses showed statistically significant differences. B and C: IF transfer function (instantaneous FM divided by the applied current) at rest and during a depolarization, namely the mean magnitude (B1 and C1) and phase (B2 and C2), respectively. In B1 and C1, * indicates difference between the response of type A () and B () neurons is statistically significant (P < 0.05); type B and D PHNn show no significant statistical difference in B1, whereas in C1, type D PHNn magnitudes are either significantly increased (¤, P < 0.05) or show a trend (#, 0.05<P < 0.10). In B2, <1 Hz, * indicate that type D neurons are significantly more positive compared with other cell types, whereas >1 Hz, the type A neurons are significantly more positive when compared with other cell types. In C2, the phase changes are similar to those of C1; however, the change from type D to A for positive phase shifts occurs at 2 rather than 1 Hz.

Effect of stimulus dynamics on instantaneous frequency: frequency-dependent gate.    The dynamical properties of PHNn were investigated using small-signal sinusoidal current stimulation at frequencies from 0.2 to 50 Hz and measuring both the membrane potential (Fig. 8A) and instantaneous frequency (Fig. 8, B and C). To describe more quantitatively the physiological dynamical behavior, we have analyzed the spike rate and its dependency on the membrane potential using an empirical IF transfer function, namely the modulation of the instantaneous frequency in response to current input (IF/I) at different stimulation frequencies (Fig. 8, B and C). The IF transfer function increases with frequency to a maximal resonance, which often cannot be measured because the maximum modulation frequency is limited by the spontaneous the carrier frequency of the discharge rate. The magnitudes of the type A PHNn IF transfer functions (IF/I) are greater than observed for type D and B PHNn, which are similar at rest but diverge at depolarized potentials (Fig. 8, B and C). When the type A and D PHNn are depolarized (Fig. 8C, 1 and 2), the magnitudes of the IF transfer function increase compared with rest (Fig. 8B, 1 and 2), which is in contrast to the decrease observed in MVN neurons. As suggested in the preceding text, this could be due to the presence of steady-state negative conductances, such as small persistent sodium currents or noninactivating calcium currents, that are likely to be responsible for the oscillatory behavior of type D PHNn.

The instantaneous FM is usually seen as a relatively linear function of small currents over a relatively wide range of stimulation frequencies (du Lac and Lisberger 1995; Nelson et al. 2003; Ris et al. 2001); however, significant nonlinear responses occur in response to large currents (Ermentrout 1998; Kepecs et al. 2002; Ris et al. 2001), which themselves lead to depolarized membrane potentials and activate voltage-dependent conductances.

To estimate the relationship between firing rate and potential of resting and depolarized neurons, it is necessary to extract the sinusoidal membrane potential response in the presence of superimposed action potentials, during sinusoidal current injection. This potential has been referred to as the coarse potential (Carandini 2004) and can be estimated using spectral analysis techniques. The spectral analysis approach was done by determining the Fourier component at the stimulation frequency of the intracellular response in which the action potentials have been clipped at 10 mV above threshold. This procedure reduces the influence of the impulse modulation, which was also measured (see METHODS). In some instances the low frequency components of the after potential seriously distorted the estimation of the coarse potential, which was clearly indicated by phase functions more negative than –90° for frequencies >20 Hz. PHNn showing this behavior were not used in the analysis of the membrane potential. An action potential template subtraction method was also developed to separate the sinusoidal potential response from impulse modulation; however, the results were similar to the spectral methods. These estimates of the membrane potential response were then divided by the injected current to produce a measure of the impedance of the neuron (see Beraneck et al. 2003). These impedance measurements were used to calculate the ratio of instantaneous frequency to membrane potential (Fig. 9) and to estimate parameters for the voltage-dependent conductances responsible for the underlying coarse potential (Fig. 11).

View larger version (12K): [in this window] [in a new window]   FIG. 9. Ratio of spike frequency transfer function and impedance of PHNn. Averaged magnitudes of the transfer functions between spike rate and membrane potential, IF/V, are shown for resting and depolarized types A, B, and D PHNn. Because SRTF = IF/I and Z = V/I, SRTF/Z = IF/V. These ratios were taken from the mean values of the spike rate transfer function (SRTF) as described in the methods and the averaged impedance (Z) values for each cell type as plotted for type D PHNn in Fig. 11B, 1 and 2. The magnitude functions are shown for each cell type at 2 levels of potential (at rest, , and under depolarization, ) *, significant differences with a P value 0.05. For 0.8 and 2 Hz for type D PHNn, the P value is actually 0.10.

View larger version (28K): [in this window] [in a new window]   FIG. 11. Type D neuronal model description of impedance and oscillatory responses. A, 1 and 2: individual type D PHNn impedance magnitude (A1) and phase (A2) plots at rest: the troughs and resonances seen in the individual traces have been averaged in B, 1 and 2. B, 1 and 2: average impedance (membrane potential divided by the injected current) of type D PHNn at 3 levels of polarization. The averaged magnitude (B1) and phase (B2) plots shown for the 3 polarization levels indicate the potential dependence of the ionic conductances. C, 1 and 2: type D neuronal analytical model fit of an individual type D PHNn for 2 polarization levels (–60 and –50 mV). The data are always shown as dashed lines. Also shown is a 3rd impedance simulation (3rd solid line) at –80 mV without data. D, 1 and 2: 3-compartment model fit (bottom solid line in C1) of sag response to a hyperpolarizing current (–0.1 nA, dashed line in C1) and simulations of depolarizing steps (C1: top solid line, 0.1 nA; C2: bottom solid line, 0.05 nA, and top solid line, 0.08 nA). Parameter values for the model fits were as follows. C, 1 and 2, Rall-type analytical model (see METHODS): Csoma = 18 pF; gsoma = 0.3 nS; Vleak = –51 mV; A = 5.7; L = 0.8; Ce = 1.3 pF; Re = 200 M; Rss = –200 M; gshunt = 0.4 nS; Vshunt = –37 mV; gNaP = 0.19 nS; VNaP = 50 mV; vm = –38 mV; sm = 0.08 (mV)-1; tm = 0.5 ms; gK = 0.05 nS; VK = –80 mV; vn = –43.5 mV; sn = 0.067 (mV)-1; tn = 20 ms; vqn = –69 mV; sqn = –0.034 (mV)-1; tqn = 230 ms; gh = 0.34 nS; Vh = –35 mV; vq = –63 mV; sq = –0.1 (mV)-1; tq = 8.1 s. D, 1 and 2, compartmental model: parameter values same except as noted: gsoma = 0.63 nS; gcore = 16 nS; Rss = –105 M; gNaP = 1.5 nS; gK = 100 nS; vn = –29.8 mV; sn = 0.07 (mV)-1; tn = 40 ms; vqn = –45 mV; sqn = –0.06 (mV)-1; tqn = 400 ms. The parameters are as follows: Csoma, the capacitance of the soma; gsoma, the passive conductance of the soma; Vleak, the reversal potential of gsoma; A, the ratio of the total dendritic to soma capacitances; L, the electrotonic length; Ce, the capacitance of the electrode; Re, the electrode resistance in M; Rss, the value of the compensated resistance in M; gshunt, the extra shunt conductance due to the microelectrode, which is present only in the soma; Vshunt, the reversal potential for the gshunt conductance; gNaP, the persistent sodium conductance, VNaP, the reversal potential for gNaP; the activation variables for gNaP were vm, the half-activation potential, sm, the slope of the steady state activation curve at the value of vm and tm, the time constant at vm; gK, the inactivating potassium conductance with its associated vn, sn, tn for activation and vqn, sqn, and tqn for inactivation; gh, the h conductance with its associated vq, sq, and tq. The reversal potentials were, respectively: VK for gK and Vh for gh (see Saint-Mleux and Moore 2000a for further details).

In type B and D PHNn there is a clear increase in the ratio, IF/V at each frequency when the neuron is depolarized; however, for resting and depolarized type A PHNn, the ratios are the same. These data suggest that the frequency-dependent gates of type B and D PHNn are nonlinear with respect to the membrane potential (Beierlein et al. 2002; Kepecs et al. 2002), which in type D PHN show a pronounced peak just above 10 Hz at depolarized levels and appear to be shifted to higher frequencies in type B PHNn. The average magnitude increase of the ratio, IF/V, is significantly greater for type D (254%) than type B (27%) PHNn. Thus the ratio, IF/V, appears to be a more sensitive determinant of differences between the PHNn types than IF/I alone, which shows relatively smaller effects when PHNn were depolarized (Fig. 8, B and C). This nonlinear behavior is in addition to the power law nonlinearity between firing rate and the low frequency or nearly steady-state level of the membrane potential relative to threshold observed in some cortical neurons (Carandini 2004). The power law relationship is a quantitative method to describe the increase in variability that occurs in the conversion of synaptic input to spike rate (Priebe et al. 2004). Thus the more marked nonlinearity of the type D PHNn could play a role in the irregular firing and clustering of action potentials that would even be enhanced as their frequency increases.

Although the filter properties of type D neurons of the PHN provide the most striking differences with those of the MVN, the remaining type A PHNn are also distinctly different from type A MVNn at their resting and depolarized levels, despite their hyperpolarized impedances being the same (data not shown). Interestingly type B PHNn are quite similar to type B MVNn (Fig. 10B, 1–4); however, their proportion including the LTS in the PHN is dominant (65%) and far above that found in the MVN. This gives the PHN a more type B-like or phasic character than observed in the MVN, which is further supported by the findings that instantaneous frequency transfer functions of PHN type A neurons have greater magnitudes and higher resonant frequencies than observed in the MVN (Fig. 10A, 1–4).

<

View larger version (22K): [in this window] [in a new window]   FIG. 10. Bode diagrams and comparison between PHNn and MVNn. A, 1–4: comparison of the mean magnitude and phase for the IF/I of type A MVNn () and type A PHNn (). A, 1 and 2, shows a comparison of the 2 components of the transfer function (A1, magnitude, and A2, phase) at the resting level and A, 3 and 4, during a depolarization (A3, magnitude, and A4, phase). *, IF/I is significantly superior for the type A PHNn than for their MVN counterpart, for a given frequency (P < 0.05). B, 1–4: comparison of the mean magnitude and phase for the IF/I of type B MVNn () and type B PHNn (). B, 1 and 2: comparison of the 2 components of the transfer function (B1 magnitude and B2 phase) at the resting level and B, 3 and 4, during depolarization (B3 magnitude and B4 phase). No significant difference was found at any frequency.