Studies in dopamine-depleted rats indicate that the external globus pallidus (GPe) contains two main types of GABAergic projection cell; so-called “prototypic” and “arkypallidal” neurons. Here, we used correlative anatomical and electrophysiological approaches in rats to determine whether and how this dichotomous organization applies to the dopamine-intact GPe. Prototypic neurons coexpressed the transcription factors Nkx2-1 and Lhx6, comprised approximately two-thirds of all GPe neurons, and were the major GPe cell type innervating the subthalamic nucleus (STN). In contrast, arkypallidal neurons expressed the transcription factor FoxP2, constituted just over one-fourth of GPe neurons, and innervated the striatum but not STN. In anesthetized dopamine-intact rats, molecularly identified prototypic neurons fired at relatively high rates and with high regularity, regardless of brain state (slow-wave activity or spontaneous activation). On average, arkypallidal neurons fired at lower rates and regularities than prototypic neurons, and the two cell types could be further distinguished by the temporal coupling of their firing to ongoing cortical oscillations. Complementing the activity differences observed in vivo, the autonomous firing of identified arkypallidal neurons in vitro was slower and more variable than that of prototypic neurons, which tallied with arkypallidal neurons displaying lower amplitudes of a “persistent” sodium current important for such pacemaking. Arkypallidal neurons also exhibited weaker driven and rebound firing compared with prototypic neurons. In conclusion, our data support the concept that a dichotomous functional organization, as actioned by arkypallidal and prototypic neurons with specialized molecular, structural, and physiological properties, is fundamental to the operations of the dopamine-intact GPe.
Most circuit-level schemes of basal ganglia (BG) organization embody the external globus pallidus (GPe) as a homogeneous entity (Albin et al., 1989; Mink, 1996; Wichmann and DeLong, 1996; Redgrave et al., 1999; Nambu et al., 2002). For example, the influential “direct/indirect pathways” model (DeLong, 1990; Smith et al., 1998) posits that GPe is populated by a single GABAergic cell type that is conceptually “downstream” of striatum and “upstream” of the subthalamic nucleus (STN). Although such schemes are necessarily idealized, this notion of functional homogeneity is nevertheless contradicted by a host of evidence of cellular diversity and functional specialization within GPe. Indeed, electrophysiological diversity has long been observed in GPe, with some of the first recordings in awake primates distinguishing two major groups of GPe neurons based on their firing rates and patterns (DeLong, 1971). Moreover, it was evident from early studies of rodent pallidum that GPe neurons have diverse intrinsic membrane properties (Kita and Kitai, 1991; Nambu and Llinaś, 1994). The molecular expression profiles of GPe neurons also vary, which partly stems from their diverse developmental origins (Kita, 2007; Flandin et al., 2010; Nóbrega-Pereira et al., 2010). Neurons in GPe likewise exhibit diverse somatodendritic and axonal structure (Kita, 2007); although most neurons innervate STN, some also target one or more or all of the other BG nuclei (Bevan et al., 1998; Smith et al., 1998; Sato et al., 2000). It is thus apparent that defining the nature and necessity of GPe neuron heterogeneity is imperative for understanding BG function and dysfunction.
To progress toward comprehensive definitions of GPe cell types, diverse and highly dynamic electrophysiological phenotypes require framing in the context of neuronal properties that are fixed over longer timescales, such as molecular profiles and structure. Studies in dopamine-depleted parkinsonian rats suggest that GPe has a dichotomous functional organization that arises from the existence of two major GABAergic projection cell types, “prototypic” neurons and “arkypallidal” neurons (Mallet et al., 2012). Prototypic GPe neurons conform to conventional views of GPe cell functions because they fire tonically at high rates in vivo, often express parvalbumin (PV), and always innervate STN. In contrast, arkypallidal neurons fire differently, express preproenkephalin (PPE), and innervate only striatum (Mallet et al., 2012). Therefore, in the dopamine-depleted state at least, two populations of GPe neuron can orchestrate activities across the entire BG in a cell-type-specific manner. However, it is unknown whether and to what extent the core cellular determinants of this dichotomous GPe organization extend to the healthy, dopamine-intact brain.
To address these key issues, we used stereological analyses of immunofluorescently labeled neurons in dopamine-intact GPe to elucidate the definitive molecular architecture of prototypic and arkypallidal neurons, and to generate unbiased estimates of their population sizes. Exploiting the unambiguous molecular signatures of these two cell types, we then defined the brain state-dependent firing of identified neurons in vivo as well as the autonomous firing and intrinsic membrane properties of identified neurons in vitro. Our data provide unique insights into how these dichotomous cell types could subserve a division of labor in GPe.
Materials and Methods
Experimental procedures were performed on male Sprague Dawley rats (Charles River) and were conducted either in Oxford in accordance with the Animals (Scientific Procedures) Act, 1986 (United Kingdom), or in Bordeaux according to institutional guidelines and the European Communities Council Directive 86/609/EEC and its successor 2010/63/EU. All experimental work adhered to Society for Neuroscience Policies on the Use of Animals in Neuroscience Research.
Quantification of molecular marker expression in populations of GPe neurons.
Fifteen adult rats (3–4 months old; 290–430 g) and 20 juvenile rats (postnatal days [P] 28–35) were killed with pentobarbital (1.5 g/kg, i.p.; Ayrton Saunders) and transcardially perfused with 0.05 m PBS, pH 7.4 (PBS), followed by 4% w/v PFA in 0.1 m phosphate buffer, pH 7.4 (PB). Adult rat brains were left overnight in fixative at 4°C and then stored in PBS at 4°C before being cut into 50 μm coronal sections on a vibrating microtome (VT1000S; Leica Microsystems). Juvenile rat brains were left overnight in fixative at 4°C, immersed in PBS containing 20% w/v sucrose for 24 h, and then stored in this solution at −80°C before being cut into 50 μm coronal sections on a cryostat (CM3000; Leica Microsystems). Free-floating tissue sections were collected in series, washed in PBS, and those containing “rostral,” “central,” and “caudal” GPe were selected (see Fig. 1A; in adults, these sections, respectively, correspond to approximate distances of 0.6, 1.0, and 1.4 mm posterior to Bregma) (Paxinos and Watson, 2007) and then processed for indirect immunofluorescence to reveal molecular markers (Mallet et al., 2012; Sharott et al., 2012; Nakamura et al., 2014). Briefly, after 1–2 h of incubation in “Triton PBS” (PBS with 0.3% v/v Triton X-100 and 0.02% w/v sodium azide; Sigma) containing 10% v/v normal donkey serum (Jackson ImmunoResearch Laboratories), sections were incubated overnight at room temperature in Triton PBS containing 1% v/v normal donkey serum and a mixture of between two and four of the following primary antibodies: goat anti-choline acetyltransferase (ChAT; 1:500 dilution; Millipore, AB144P) (Stacy et al., 2005); mouse anti-ChAT (1:1000; gift from C. Cozzari) (von Engelhardt et al., 2007); rabbit anti-ER81 (also known as ETS variant gene 1 protein; 1:5000; gift from S. Arber) (Arber et al., 2000); goat anti-forkhead box protein P2 (FoxP2; 1:500; Santa Cruz Biotechnology, sc-21069); rabbit anti-FoxP2 (1:1000; Sigma, HPA000382) (Reimers-Kipping et al., 2011); mouse anti-human neuronal protein HuC/HuD (HuCD; 1:200; Invitrogen, A-21271, clone 16A11) that specifically binds to an epitope present in gene products of several Hu genes, most notably of HuC and HuD (Marusich et al., 1994; Pascale et al., 2004); mouse anti-LIM homeobox protein 6 (Lhx6; 1:2000; Santa Cruz Biotechnology, sc-271433); this monoclonal antibody recognizes a single protein band of the predicted molecular mass of Lhx6 on Western blots of cell lysates (manufacturer's data); rabbit anti-meis homeobox 2 protein (Meis2; 1:2000; gift from A. Buchberg) (Swift et al., 1998; Agoston and Schulte, 2009); mouse anti-neuronal nuclei protein (NeuN, also known as hexaribonucleotide-binding protein 3; 1:500; Millipore, MAB377, clone A60); rabbit anti-neuronal PAS domain protein 1 (Npas1; 1:500; gift from S.L. McKnight) (Erbel-Sieler et al., 2004); rabbit anti-NK2 homeobox 1 protein (Nkx2-1, also known as thyroid transcription factor 1; 1:500; Santa Cruz Biotechnology, sc-13040) (Magno et al., 2011), or mouse anti-Nkx2-1 (1:50; Leica Biosystems, NCL-TTF-1), which were raised against different Nkx2-1 peptide antigens but resulted in a 99.5% overlap in immunoreactive neurons; guinea pig anti-PV (1:1000; Synaptic Systems, 195004) (Massi et al., 2012); and rabbit anti-PPE (1:5000; LifeSpan Biosciences, LS-C23084); in control experiments, the use of this PPE antiserum resulted in immunolabeling that was indistinguishable from that achieved with an affinity-purified antibody raised against PPE in guinea pigs (Lee et al., 1997). To optimize immunolabeling of Meis2, Npas1, and PPE, we used a heat pretreatment (2, 2, and 6 h, respectively) as a means of antigen retrieval (Mallet et al., 2012). After exposure to primary antibodies, sections were washed in PBS and incubated overnight at room temperature in Triton PBS containing a mixture of secondary antibodies (all raised in donkey) that were conjugated to the following fluorophores: AMCA (1:250 dilution; Jackson ImmunoResearch Laboratories); AlexaFluor-488 (1:500; Invitrogen); Cy3 (1:1000; Jackson ImmunoResearch Laboratories); or DyLight 649 (1:500; Jackson ImmunoResearch Laboratories). All secondary antibodies were highly cross-adsorbed by the manufacturers to reduce cross-species reactivity. After washing in PBS, sections were mounted in Vectashield (Vector Laboratories) and imaged on an epifluorescence microscope (Carl Zeiss, AxioImager.M2) running Axiovision software (Carl Zeiss) and equipped with a StereoInvestigator system (MBF Bioscience). Appropriate sets of filter cubes were used to image the fluorescence channels: AMCA (excitation 299–392 nm, beamsplitter 395 nm, emission 420–470 nm); AlexaFluor-488 (excitation 450–490 nm, beamsplitter 495 nm, emission 500–550 nm); Cy3 (excitation 532–558 nm, beamsplitter 570 nm, emission 570–640 nm); and DyLight 649 (excitation 625–655 nm, beamsplitter 660 nm, emission 665–715 nm). Images of each of the channels were taken sequentially and separately to negate possible crosstalk of signal across channels.
Using a series of partly overlapping, complementary immunofluorescence protocols, GPe neurons were tested for their combinatorial expression of all the molecular markers described above. ROIs (i.e., the borders of rostral, central, and caudal GPe) were first defined using a 5× 0.16 NA objective lens. The medial and lateral borders of GPe were readily delineated according to the internal capsule and dorsal striatum, respectively (see Fig. 1A). The ventral border of GPe was conservatively assigned such that it was set well apart from the poorly defined boundary between GPe and the functionally distinct ventral pallidum (see Fig. 1A, red lines). For the purposes of this study, it was not necessary to determine absolute numbers or densities of GPe neurons expressing these markers. We thus used a version of design-based stereology, the “modified optical fractionator” (West, 1999, 2012), to generate unbiased cell counts and map cell distributions and determine the proportions of a given population of neurons that express certain combinations of markers. Stereological sampling was performed over the entire GPe in both hemispheres, as sectioned at the designated rostral, central, and caudal levels. For each immunofluorescence protocol, a series of completely tessellated, z-stacked images were acquired using a 40× 1.3 NA oil-immersion objective lens and 1.0 μm steps (“optical sections”) at depths of 2–12 μm from the upper surface of each section at the level of GPe. To minimize confounds arising from surface irregularities, neuropil within a 2-μm-thick “guard zone” at the upper surface was not imaged. This sampling strategy thus defined a 10-μm-thick “optical disector” that was used with abutting, unbiased 2D counting frames (200 × 150 μm; consisting of two perpendicular exclusion lines and two inclusion lines) to generate all cell counts and marker expression profiles (West, 1999, 2012; Glaser et al., 2007). A neuron was only counted once through the series of optical sections when its nucleus came into sharp focus within the disector; neurons with nuclei already in focus in the top optical section of the disector were ignored (West, 1999, 2012; Glaser et al., 2007). The use of stereology, and this optical disector probe in particular, ensured that we could generate robust and unbiased cell counts in a timely manner. For a given molecular marker, X, we designate positive immunoreactivity (confirmed expression) as X+, and undetectable immunoreactivity (no expression) as X−. In this and related immunofluorescence analyses, a neuron was classified as not expressing the tested molecular marker only when positive immunoreactivity could be observed in other cells on the same optical section as the tested neuron. Immunoreactivity was scored (either positive or undetectable) by experienced investigators; the results of most immunofluorescence protocols were scored by at least two individuals, and the variability in the normalized scores of multiple investigators was low (<5%). Each immunofluorescence protocol was repeated (and thus, each neuronal expression profile was quantified) at the three levels along the rostrocaudal axis of GPe in a minimum of three adult rats. Data from rostral, central, and caudal GPe of both hemispheres were exported to Excel (Microsoft) and, unless stated otherwise, were pooled for further analyses. In the bar graphs shown in figures, gray bars represent grand averages from all rats tested, with each superimposed black dot representing the average from an individual rat to give an indication of cohort variance.
Viral vector-mediated retrograde tract-tracing.
To elucidate the axonal projections of populations of GPe neurons, we performed retrograde tract-tracing using a recombinant, pseudotyped lentiviral vector (RV-G-MND-EGFP-WPRE) injected into either the STN or dorsal striatum. The RV-G-MND-EGFP-WPRE vector is based on HIV Type 1, is pseudotyped with rabies virus glycoprotein, and contains a gene encoding enhanced green fluorescent protein (EGFP). Intracerebral injection of this vector results in the delivery of the EGFP transgene to neurons through retrograde transport from their axon terminal fields; indeed, its utility for retrogradely labeling several different cell types that provide synaptic inputs to striatum has been demonstrated in rodents (Kato et al., 2007). Eleven adult rats (290–310 g) were anesthetized with isoflurane (3% and 1.5% v/v in O2 for induction and maintenance, respectively), and then placed in a stereotaxic frame (Kopf). Wound margins were infiltrated with local anesthetic (xylocaine), and body temperature was maintained at 37 ± 0.5°C by a homeothermic heating device (Harvard Apparatus). In five rats, discrete bilateral injections of RV-G-MND-EGFP-WPRE vector (0.2 μl per site, titer of 3.4 × 109 infectious particles/ml) were made into the STN (3.8 mm posterior and 2.5 mm lateral of Bregma, 7.5 mm from brain surface) (Paxinos and Watson, 2007). In another six rats, discrete bilateral injections of the same vector (0.4 μl per site) were made into a central region of dorsal striatum (1.8 mm anterior and 2.5 mm lateral of Bregma, 5.0 mm from brain surface). Allowing 17 d for neuronal transduction and labeling with EGFP, animals were killed and transcardially perfused with 100 ml of PBS, followed by 300 ml of 4% w/v PFA in PB. Brains were left overnight in fixative at 4°C and then stored for 3–7 d in PBS at 4°C before sectioning. Coronal sections (50 μm) were cut from each brain using a vibrating microtome, collected in series, washed in PBS, and processed for indirect immunofluorescence using the antibodies and protocols described above. Transduced neurons were revealed after sequential incubation with rat anti-green fluorescent protein (1:500; Nacalai Tesque, 04404-84) and donkey anti-rat conjugated to AlexaFluor-488 (1:500; Invitrogen). The borders of the STN were readily delineated by immunoreactivity for FoxP2 (Campbell et al., 2009). To minimize the possibility of GPe neurons being transduced through their local axon collaterals during vector injections into striatum, we ensured that the outer limits of the striatal zones containing many EGFP+ cells, which presumably indicated sites of vector injection, were >400 μm away from the rostral border of GPe. Neurons in the GPe-expressing EGFP were tested for coexpression of FoxP2, Nkx2-1, PV, and/or Npas1. After binding of primary and secondary antibodies, and final washing in PBS, sections were mounted in Vectashield and imaged on an epifluorescence microscope as described above, with the exception that GPe neurons expressing EGFP were imaged and counted throughout the whole depth of the tissue (excluding upper and lower guard zones). For the immunofluorescence images shown in Figures 123–4, digital images generated in this and preceding anatomical experiments were acquired as z-stacks (optical section thickness: 0.8–1.5 μm) using an epifluorescence or confocal microscope (Carl Zeiss LSM 710; see below for laser and filter settings). After selecting the thinnest z-stack necessary to convey the essential information (3–10 μm stack for Figs. 12–3; 18–28 μm for Fig. 4), raw images were cropped to ROIs, with brightness and contrast adjusted when necessary, using Photoshop software (Creative Suite 3; Adobe Systems).
In vivo electrophysiological recording and juxtacellular labeling of individual GPe neurons.
Recording and labeling experiments were performed in 20 anesthetized dopamine-intact rats (3–4 months old, 305–400 g) and 16 anesthetized 6-hydroxydopamine (6-OHDA)-lesioned rats (3–5 months old and 295–470 g at the time of recording), as previously described (Mallet et al., 2008a, b, 2012). Briefly, anesthesia was induced with 4% v/v isoflurane in O2 and maintained with urethane (1.3 g/kg, i.p.; ethyl carbamate, Sigma), and supplemental doses of ketamine (30 mg/kg, i.p.; Willows Francis) and xylazine (3 mg/kg, i.p.; Bayer). Wound margins were infiltrated with local anesthetic (0.5% w/v bupivacaine; Astra). Animals were then placed in a stereotaxic frame (Kopf). Body temperature was maintained at 37 ± 0.5°C by a homeothermic heating device (Harvard Apparatus). Electrocorticograms (ECoGs) and respiration rate were monitored constantly to ensure the animals' well-being. The epidural ECoG was recorded above the frontal (somatic sensory-motor) cortex (4.0 mm anterior and 2.0 mm lateral of bregma) (Paxinos and Watson, 2007) and was referenced against the ipsilateral cerebellar hemisphere (Mallet et al., 2012). Raw ECoG was bandpass filtered (0.3–1500 Hz, − 3 dB limits) and amplified (2000×; DPA-2FS filter/amplifier; NPI Electronic Instruments) before acquisition. Extracellular recordings of single-unit activity, that is, the action potentials (“spikes”) fired by individual neurons in the GPe, were made using standard-wall borosilicate glass electrodes (10–30 MΩ in situ; tip diameter ∼1.2 μm) containing 0.5 m NaCl solution and neurobiotin (1.5% w/v; Vector). Electrodes were lowered into the brain under stereotaxic guidance and using a computer-controlled stepper motor (IVM-1000; Scientifica), which allowed electrode placements to be made with submicron precision. Electrode signals were amplified (10×) through the bridge circuitry of an Axoprobe-1A amplifier (Molecular Devices), AC-coupled, amplified another 100×, and filtered at 300–5000 Hz (DPA-2FS filter/amplifier). The ECoG and single-unit activity were each sampled at 17.9 kHz using a Power1401 Analog-Digital converter and a PC running Spike2 acquisition and analysis software (Cambridge Electronic Design). As described previously (Mallet et al., 2008a, 2012), single-unit activity in GPe was recorded during cortical slow-wave activity (SWA), which is similar to activity observed during natural sleep, and/or during episodes of spontaneous “cortical activation,” which contain patterns of activity that are more analogous to those observed during the awake, behaving state (Steriade, 2000). It is important to note that the neuronal activity patterns present under this anesthetic regimen may only be qualitatively similar to those present in the unanesthetized brain. Nevertheless, the urethane-anesthetized animal still serves as a useful model for assessing the impact of extremes of brain state on functional connectivity within and between the BG and cortex in dopamine-intact and parkinsonian animals (Magill et al., 2006; Mallet et al., 2008a, b; Sharott et al., 2012). Hindpaw withdrawal reflexes were not present during episodes of prolonged, spontaneous cortical activation, thus indicating that anesthesia was adequate throughout recordings. Following electrophysiological recordings, single neurons were juxtacellularly labeled with neurobiotin (Magill et al., 2001; Sadek et al., 2007; Baufreton et al., 2009; Mallet et al., 2012). Briefly, positive current pulses (2–10 nA, 200 ms, 50% duty cycle) were applied until the single-unit activity became robustly entrained by the pulses. Single-unit entrainment resulted in just one neuron being labeled with neurobiotin. Two to six hours after labeling, animals were killed and transcardially perfused with 100 ml of PBS, followed by 300 ml of 4% w/v PFA in PB. Brains were left overnight in fixative at 4°C and then stored for 1–3 d in PBS at 4°C before sectioning.
Molecular characterization of recorded and juxtacellularly labeled neurons.
Parasagittal sections (50 μm) were cut from each brain using a vibrating microtome, collected in series, and washed in PBS. Free-floating sections were then incubated overnight at room temperature in Triton PBS containing Cy3-conjugated streptavidin (1:1000 dilution; Invitrogen). Sections containing neurobiotin-labeled neuronal somata (those marked with Cy3) were then isolated for molecular characterization by indirect immunofluorescence using the antibodies and protocols described above. Recorded and identified GPe neurons were tested for expression of FoxP2, Nkx2-1, PPE, and/or PV. After binding of primary and secondary antibodies, and final washing in PBS, sections were mounted in Vectashield and imaged on confocal fluorescence microscopes (Carl Zeiss, LSM 510 or LSM 710), as previously described (Mallet et al., 2012; Sharott et al., 2012; Nakamura et al., 2014). Briefly, verification of molecular identity was performed by assessing single-plane confocal images acquired using a constant 1.0-μm-thick optical section (obtained by varying the pin hole size) and a 40× 1.3 NA or a 63× 1.4 NA oil-immersion objective lens. Each fluorescence channel was imaged in a separate track. Cy3 fluorescence signal was imaged with excitation from a helium/neon 543 nm laser, with emission restricted on the LSM 510 by a LP 560 filter (for the shorter wavelengths) and a NFT 635 VIS 2° dichroic (for longer wavelengths). Emission on the LSM 710 was restricted by selecting an emission range of 552–639 nm. AlexaFluor-488 fluorescence was imaged with excitation from an argon 488 nm laser, with emission restricted by either a BP 510–530 filter (on the LSM 510) or by setting emission at 493–542 nm (on the LSM 710). AMCA fluorescence was imaged on the LSM 710 with excitation from a diode 405 nm laser, and selecting emission at 409–485 nm. DyLight 649 fluorescence was imaged with excitation from a helium/neon 633 nm laser, with emission restricted by either a LP 650 filter (LSM 510) or by setting emission at 637–757 nm (LSM 710). A juxtacellularly labeled neuron was classified as not expressing the tested molecular marker only when positive immunoreactivity could be observed in other cells on the same focal plane as the tested neuron.
6-OHDA lesions of dopamine neurons.
Unilateral 6-OHDA lesions were induced in 210–280 g rats, as detailed previously (Mallet et al., 2008a, b, 2012). The neurotoxin 6-OHDA (hydrochloride salt; Sigma) was dissolved in 0.9% w/v NaCl solution containing 0.02% w/v ascorbate to a final concentration of 4 mg/ml. Approximately 25 min before the injection of 6-OHDA, all animals received desipramine (25 mg/kg, i.p.; Sigma) to minimize the uptake of 6-OHDA by noradrenergic neurons (Schwarting and Huston, 1996b). Anesthesia was induced and maintained with isoflurane (1.5%–3% v/v in O2), and 3 μl of 6-OHDA solution was injected near the medial forebrain bundle (4.1 mm posterior and 1.2 mm lateral of Bregma, and 7.9 mm ventral to the dura) (Paxinos and Watson, 2007). Lesions were assessed 14 or 15 d after 6-OHDA injection by challenge with apomorphine (Schwarting and Huston, 1996a) (0.05 mg/kg, s.c.; Sigma) and were considered successful when animals made ≥80 net contraversive rotations in 20 min. Electrophysiological recordings were performed in the GPe ipsilateral to 6-OHDA lesions in anesthetized rats 21–35 d after surgery.
Analyses of in vivo electrophysiological data.
Data from the recording session were visually inspected, and epochs of robust cortical SWA or cortical activation were selected according to the previously described characteristics of these brain states (Magill et al., 2006; Mallet et al., 2008a, b; Sharott et al., 2012). ECoGs were assumed to be realizations of stationary, zero-mean time series; their power spectra during SWA were calculated with a frequency resolution of 0.25 Hz using the MATLAB (MathWorks) toolbox Neurospec 2.0 for multivariate Fourier analyses (www.neurospec.org). A portion of the spike train recorded during each defined brain state was isolated and used for statistical analyses (average epoch duration of 119 ± 7 s, mean ± SEM). We assumed spike trains to be realizations of stationary, stochastic point processes. Putative single-unit activity was isolated with standard “spike sorting” procedures (Mallet et al., 2008a), including template matching, principal component analysis, and supervised clustering (Spike2). Isolation of a single unit was verified by the presence of a distinct refractory period in the interspike interval (ISI) histogram. Only neurons in which <1% of all ISIs were <2 ms were analyzed in this study. For further analysis, single-unit activity was converted so that each spike was represented by a single digital event (Spike2). Mean firing rate (spikes per second; spk/s) was calculated from the total number of spikes per data epoch. The coefficient of variation of the ISIs (CVISI), a value used widely as an indicator of regularity in point processes (Johnson, 1996), and the SD of the ISIs (SDISI) were also calculated. To be included in the CVISI and SDISI analyses, a GPe neuron had to fire at least eight spikes during the in vivo recording.
To investigate how the spike firing of individual GPe neurons varied in time with respect to ongoing cortical slow oscillations during SWA, we used the Hilbert transform to analyze the instantaneous phase relationships between GPe spike times and cortical oscillations in a frequency band of 0.4–1.6 Hz (Sharott et al., 2012; Nakamura et al., 2014). Signal conditioning and analyses were performed using MATLAB (version 7.3; MathWorks). ECoG signals containing robust SWA were first filtered at 0.4–1.6 Hz using a second-order Butterworth filter. Subsequently, the instantaneous phase and power of the ECoG were calculated from the analytic signal obtained via the Hilbert transform (Lachaux et al., 1999; Le Van Quyen et al., 2001). In this formalism, peaks in the ECoG oscillations correspond to a phase of 0° and troughs to a phase of 180°. Linear phase histograms, circular phase plots, and circular statistical measures were calculated using the instantaneous phase values for each spike. Descriptive and inferential circular statistics were then calculated using the CircStat toolbox (Berens, 2009) for MATLAB. For the calculation of vector lengths and statistical comparisons, we included only those GPe neurons that fired ≥40 spikes during the entire recording during SWA. These neurons were then tested for significantly phase-locked firing (defined as having p < 0.05 in Rayleigh's Uniformity Test). The null hypothesis for Rayleigh's Uniformity Test was that the spike data were distributed in a uniform manner. We and others have previously remarked that the nonsinusoidal nature of some field potential oscillations, such as the cortical slow oscillation, can confound standard circular statistics, especially Rayleigh's Uniformity Test (Siapas et al., 2005; Mallet et al., 2008a; Sharott et al., 2012; Nakamura et al., 2014). Thus, for analysis of GPe neuron firing relationships with cortical slow oscillations, Rayleigh's Uniformity Tests were only performed after any phase nonuniformities of the slow oscillations were corrected with the empirical cumulative distribution function of MATLAB (Siapas et al., 2005; Nakamura et al., 2014). For each of the neurons that were significantly phase-locked, the mean phase angle was calculated. Differences in the mean phase angles of groups of neurons were tested by the Watson–Williams F test (p < 0.05 for significance). The mean resultant vector length (referred to hereafter as simply “vector length”) of the phase distribution, bound between 0 and 1 (the closer to 1, the more concentrated the angles), was used to quantify the level of phase locking around the mean angle for individual neurons (computed using the angles of each spike) and for populations of neurons (computed using the mean angle for each neuron). Differences between the vector lengths of two neuron populations were tested for using the Mann–Whitney U test, and differences between those of three populations with Kruskal–Wallis ANOVA on Ranks. Where data are displayed in circular plots (see Figs. 6 and 10), lines radiating from the center are the vectors of the preferred phases of firing (with the center and perimeter of the circle representing vector lengths of 0 and 1, respectively); thin lines indicate preferred firing of individual neurons, whereas thick black lines indicate population vectors. The small open circles on the perimeter represent the preferred phases of each neuron.
In vitro electrophysiological recording of individual GPe neurons.
Visualized recordings of GPe neurons were performed in brain slices acutely prepared from 52 juvenile (P20-P35) dopamine-intact rats, as previously described (Baufreton et al., 2005, 2009; Miguelez et al., 2012). Briefly, animals were anesthetized with a mixture of ketamine and xylazine (75 mg/kg and 10 mg/kg, respectively, i.p.) and transcardially perfused with ice-cold modified “artificial CSF” (ACSF) that was equilibrated with 95% O2 and 5% CO2 and contained the following (in mm): 230 sucrose, 26 NaHCO3, 2.5 KCl, 1.25 Na2HPO4, 0.5 CaCl2, 10 MgSO4, and 10 glucose. The brain was then quickly removed from the skull, blocked in the parasagittal plane, glued to the stage of a vibrating microtome (VT1200S; Leica Microsystems), and submerged in ice-cold modified ACSF. Slices containing the GPe (350 μm thick) were then cut and transferred to a holding chamber at room temperature in standard ACSF that was equilibrated with 95% O2 and 5% CO2, and contained the following (in mm): 126 NaCl, 26 NaHCO3, 2.5 KCl, 1.25 Na2HPO4, 2 CaCl2, 2 MgSO4, and 10 glucose. Single slices were then transferred to a recording chamber, perfused continuously with oxygenated ACSF at 35°C–37°C, and visualized using infrared gradient contrast video microscopy (Eclipse workstation; Nikon) and a 60× water-immersion objective (Fluor 60×/1.00 W; Nikon). Recordings from individual GPe neurons were made in a cell-attached or whole-cell configuration (voltage-clamp mode) or in a perforated-patch configuration (current clamp). In experiments designed to define the autonomous and driven firing of GPe neurons, all cell-attached and perforated-patch recordings were made in the presence of potent and selective blockers of glutamatergic synaptic transmission (50 μm APV and 20 μm DNQX; Abcam) and GABAergic synaptic transmission (20 μm SR-95531/gabazine and 1 μm CGP55845; Tocris Bioscience). Cell-attached recordings were made using standard-wall borosilicate glass electrodes filled with ACSF and were performed at a holding potential of 0 mV after the stable establishment of the gigaseal. Somatic perforated-patch recordings were made using pipettes prepared from standard-wall borosilicate glass capillaries and were front-filled with the following (in mm): 110 KMeSO4, 25 KCl, 3.6 NaCl, 1 MgCl2 · 6H2O, 10 HEPES, 0.1 Na4EGTA, 0.4 Na3GTP, and 2 Mg1.5ATP. The pH and osmolarity of the pipette solution were 7.3 and 290 mOsm, respectively. Patch pipettes were then back-filled with the same pipette solution containing gramicidin (∼15 μg/ml; Sigma). Gramicidin channels are only permeable to monovalent cations and small neutral molecules, allowing better preservation of the intrinsic physiological properties of GPe neurons, particularly their autonomous firing (Surmeier et al., 2005; Miguelez et al., 2012). For perforated-patch recordings, electrode capacitance was compensated online. Series resistance was regularly monitored, but not compensated, during recordings. On average, the series resistances during perforated-patch recordings of identified prototypic and arkypallidal neurons were similar (69.4 ± 4.9 and 57.9 ± 8.0 MΩ, respectively). Voltage errors arising from series resistance were corrected off-line. Deliberate or accidental establishment of the whole-cell configuration was recognized as a sudden drop in series resistance and an offset (∼5 mV) in membrane potential. The value of this offset was smaller than the experimentally measured, and the empirically calculated, junction potential between the electrode solution and the external media of 9 mV (Baufreton et al., 2005). The recorded membrane potential was therefore ∼4 mV more depolarized than the true membrane potential and was corrected off-line. Somatic patch-clamp recordings were obtained using a Multiclamp 700B amplifier and Digitdata 1320A digitizer controlled by Clampex 9.0 software (Molecular Devices). Electrode signals were low-pass filtered at 4 kHz and sampled at 20 kHz. Following electrophysiological recording in either a cell-attached or perforated-patch configuration, the electrode was carefully withdrawn and the soma of the same neuron was repatched (Oren et al., 2009) using new glass pipettes filled with a solution containing 1.5% w/v biocytin (Sigma) and (in mm): 110 KMeSO4, 25 KCl, 3.6 NaCl, 1 MgCl2 · 6H2O, 10 HEPES, 0.1 Na4EGTA, 0.4 Na3GTP, and 2 Mg1.5ATP. The pH and osmolarity of this pipette solution were 7.2 and 290 mOsm, respectively. Subsequent establishment of a whole-cell recording configuration resulted in the rapid dialysis and anterograde filling of the GPe neuron with biocytin. This repatching procedure was used to increase the yield of biocytin-filled neurons after cell-attached or perforated-patch recordings of long duration; whole-cell dialysis with a pipette solution containing gramicidin can lead to widespread perforation of intracellular and plasma membranes, potentially killing the recorded neuron and, thence, precluding its biocytin labeling and post hoc verification. For measurements of “persistent” sodium channel currents (INaP), neurons were recorded using a whole-cell voltage-clamp configuration in the presence of receptor blockers (as above), and held at −95 mV before and after the application of a slowly depolarizing ramp voltage command (from −80 to −15 mV). The ramp command was set at 13.3 mV/s, which reliably inactivates transient voltage-gated sodium channels, leaving the inward sodium current identified as INaP (Park et al., 2013). Ramps were applied 3 times per neuron, and responses (peak amplitudes of INaP) were then averaged. Series resistance was regularly monitored, but not compensated, during recordings; if there was >20% change in series resistance, neurons were excluded from further analyses. On average, the series resistances during whole-cell recordings of identified prototypic and arkypallidal neurons were similar (15.1 ± 1.3 and 15.6 ± 0.7 MΩ, respectively).
One to two hours after cell filling with biocytin, slices were immersed in 4% w/v PFA in PB for 1–7 d at 4°C and then cyroprotected in 20% w/v sucrose in PBS, embedded in Tissue-Tek (Fisher Scientific), frozen in isopentane (Sigma) at −45°C, and stored at −80°C for >2 d before resectioning.
Molecular characterization of GPe neurons recorded in vitro.
Frozen slices were resectioned in the parasagittal plane at 50 μm using a cryostat (CM3000; Leica Microsystems), collected in series, and washed in PBS. Endogenous peroxidases were inactivated by immersion of the free-floating sections in 3% v/v H2O2 in PBS for 30 min. Subsequently, sections were incubated in Triton PBS containing 1% w/v BSA for 30 min and then incubated overnight at room temperature in Triton PBS containing AlexaFluor-568-conjugated streptavidin (1:500 dilution; Invitrogen). Biocytin-labeled neuronal somata were then isolated for molecular characterization by indirect immunofluorescence using the antibodies and protocols described above. Specifically, in vitro recorded and identified GPe neurons were tested for expression of FoxP2 and PV. After binding of primary and secondary antibodies, and final washing in PBS, sections were mounted in Vectashield and imaged on a confocal fluorescence microscope (TCS SP8; Leica Microsystems) equipped of a white light laser 2 with freely tunable excitation from 470 to 670 nm.
Analyses of in vitro electrophysiological data.
Data were analyzed using Origin 7.0 (Microcal Software). Mean firing rates, CVISI, and SDISI were calculated as for recordings of GPe neurons in vivo. To quantify the autonomous firing of neurons recorded in the perforated-patch configuration, contiguous 60 s epochs of activity with zero current injection were analyzed. The input resistance of each neuron was calculated from the difference between the minimum membrane voltage reached during injection of a −100 pA, 500 ms current pulse, and the average subthreshold membrane potential during the 10 s preceding the current injection (Deister et al., 2013). As previously described (Baufreton et al., 2005), action potential threshold was determined using a custom algorithm that detected the first point of sustained positive “acceleration” of membrane voltage [(dV/dt)/dt] that was >2 SDs of baseline voltage before action potential threshold. Action potential half-width was calculated by measuring the action potential width at the midpoint between threshold and peak (Bugaysen et al., 2010). Frequency-current (F-I) response curves were obtained by injecting incremental current pulses from −50 pA to 200 pA (2 s duration; 25 pA steps). An index of firing rate adaptation was calculated as the ratio between the instantaneous firing frequencies of the first two spikes and last two spikes fired in response to a 2 s current pulse of 100 pA. “Rebound” burst firing was defined as spikes occurring at an instantaneous rate in excess of mean spontaneous firing + 3 SDs in response to the injection of a 500 ms current pulse of −150 pA. In voltage-clamp recordings, persistent sodium channel currents were transformed into conductance (G) using Ohm's law in the form: G = I/(V − ENa), where V is the test potential and ENa is the sodium equilibrium potential calculated using the Nernst equation. Conductance was normalized, plotted against V, and fitted with a Boltzmann function of the form: G/Gmax = 1/(1 + exp[(V − V1/2)/k]), where V1/2 is the half-activation voltage and k is the slope factor.
Further statistical testing.
The Shapiro–Wilk test was used to judge whether noncircular datasets were normally distributed (p ≤ 0.05 to reject). Because some datasets were not normally distributed, we used nonparametric statistical testing throughout (SigmaPlot 12; Systat Software). The Mann–Whitney U test was used for comparisons of unpaired data, whereas the Wilcoxon Signed Rank Test was used to compare paired datasets. For multiple group comparisons, we performed a Kruskal–Wallis ANOVA on Ranks, with Dunn's Test for further post hoc definition of comparisons. Significance for all statistical tests was set at p < 0.05 unless noted otherwise. Data are represented as group mean ± SEM unless stated otherwise.
Most GABAergic GPe neurons in dopamine-intact rats express either PV or PPE
In adult rats rendered parkinsonian by 6-OHDA lesions, most prototypic GPe neurons express the calcium-binding protein PV, whereas arkypallidal neurons instead express the neuropeptide precursor PPE (Mallet et al., 2012). We took these differences in the neurochemistry of prototypic and arkypallidal neurons in dopamine-depleted animals as a starting point for our investigation of GPe cell-type identity in dopamine-intact adult rats. We thus performed stereological analyses of immunofluorescence signals in rostral, central, and caudal GPe (Fig. 1A) to generate unbiased estimates of the proportions of neurons that express PV, PPE, both markers, or neither marker. In accordance with previous quantitative estimates of PV expression in adult rat GPe (Kita and Kitai, 1994; Kita and Kita, 2001; Hoover and Marshall, 2002), PV+ neurons constituted 59% of all GPe neurons, the latter being identified here by immunoreactivity for the pan-neuronal marker HuCD (Fig. 1B,C). On average, PPE+ neurons were less numerous than PV+ neurons but still constituted over one-fourth (28%) of all GPe neurons (Fig. 1B). Importantly, coexpression of PV and PPE by individual GPe neurons was negligible (<0.2% of PV+ neurons expressed PPE and vice versa; Fig. 1C–E). However, a substantial proportion of all GPe neurons (10%) expressed neither PV nor PPE (PV−/PPE− neurons; Fig. 1B,C). Cholinergic GPe neurons, identified by immunoreactivity for ChAT, comprised 4% of all GPe neurons (Fig. 1B). Coexpression of ChAT and PV was not observed (Figs. 1D and 2H). A small proportion (13%) of ChAT+ neurons coexpressed PPE (Fig. 2H), but, because cholinergic GPe neurons are rare, this meant that <1% of PPE+ neurons coexpressed ChAT (Fig. 1E). Moreover, the scarcity of cholinergic neurons indicates that they could only account for a minority of PV−/PPE− neurons in the GPe regions that were analyzed. PV-expressing neurons, PPE+ neurons, and PV−/PPE− neurons were distributed throughout GPe and were intermingled with each other (Fig. 1C,F). On average, the proportions of all GPe neurons that were PV+, PPE+, or PV−/PPE− were relatively constant across the rostral, central, and caudal levels of GPe (Fig. 1G). Together, these data show that most GPe neurons express either PV or PPE. However, assuming that all noncholinergic GPe neurons in rodents and primates use GABA as their primary neurotransmitter (Smith et al., 1987; Delfs et al., 1995), our data also show that PV and PPE do not collectively capture all GABAergic GPe neurons.
The two major populations of GABAergic GPe neurons in dopamine-intact rats are delineated by their selective expression of transcription factors
A specific molecular marker for prototypic GPe neurons has not yet been determined (Mallet et al., 2012). Moreover, PPE is the only marker known to distinguish arkypallidal neurons from prototypic neurons (Mallet et al., 2012). There is thus a pressing need to identify other molecular markers that more comprehensively define these two cell types in the dopamine-intact GPe. With this in mind, we interrogated the published literature and other online resources (Allen Brain Atlas; http://www.brain-map.org/) and judged six transcription factors (FoxP2, Meis2, ER81, Nkx2-1, Lhx6, and Npas1) to be especially promising candidate markers. We then used the same stereological sampling approach to quantify their (co)expression in PPE+, PV+, and PV−/PPE− neuronal populations. Neurons expressing PPE, FoxP2, or Meis2 were similarly abundant in GPe (Fig. 1B), and critically, there was an almost perfect overlap in the expression of these three markers. Indeed, >99% of PPE+ neurons expressed FoxP2 and vice versa (Figs. 1E and 2A,D), and >99% of FoxP2+ neurons expressed Meis2 and vice versa (Fig. 2B,D,E). In stark contrast, coexpression of FoxP2 (or Meis2) and PV was extremely rare, such that <1% of FoxP2+/Meis2+ neurons expressed PV and vice versa (Figs. 1D and 2A,B,D,E). Coexpression of ChAT by FoxP2+ neurons was similarly rare (Fig. 2D). These data suggest that arkypallidal neurons are not only defined by their expression of PPE but also by expression of FoxP2 and/or Meis2. However, it follows that these two transcription factors do not mark the PV−/PPE− GABAergic neuron population in GPe.
Studies of the embryonic and early postnatal development of mouse ventral telencephalon (subpallium) show that the molecular diversity of GPe neurons extends to the selective expression of other transcription factors that are critical for the genesis, migration, specification, and maturation of neurons (Flandin et al., 2010; Nóbrega-Pereira et al., 2010). For example, expression of the ETS transcription factor ER81 informs on the progenitor domains of the embryonic subpallium that give rise to GPe neurons (Flames et al., 2007; Flandin et al., 2010; Nóbrega-Pereira et al., 2010). We determined that, on average, ER81+ neurons constituted 56% of all GPe neurons in dopamine-intact adult rats (Fig. 1B). Coexpression of ER81 and PV in individual GPe neurons was common; almost all ER81+ neurons (97%) coexpressed PV, and 90% of PV+ neurons coexpressed ER81 (Figs. 1D and 2C,F). However, ER81+ GPe neurons did not express FoxP2 (Fig. 2C,D,F), a proxy marker for PPE+ GPe neurons (see above). Thus, although the selective expression of ER81 further resolves the molecular phenotype of adult rat GPe neurons, it is largely comparable with that of PV.
Because many neurons in normal mouse GPe express the homeobox transcription factor Nkx2-1 (Magno et al., 2009; Nóbrega-Pereira et al., 2010) and because mutant mice with disrupted Nkx2-1 gene function lack a readily discernible GPe (Sussel et al., 1999; Flandin et al., 2010), we next quantified the expression of Nkx2-1 in the GPe of dopamine-intact adult rats. On average, Nkx2-1+ neurons constituted approximately two-thirds (67%) of all GPe neurons (Fig. 1B). Although a small fraction (5%) of these Nkx2-1+ neurons coexpressed ChAT (Figs. 2G and 3E), the majority of Nkx2-1+ neurons (86%) coexpressed PV instead (Fig. 3A,E). However, almost all (98%) PV+ neurons coexpressed Nkx2-1 (Figs. 1D and 3A). Thus, although there is a large overlap in the expression of Nkx2-1 and PV, these data further highlight that GABAergic Nkx2-1+ GPe neurons are more abundant than GABAergic PV+ GPe neurons. Importantly, Nkx2-1+ GPe neurons did not express FoxP2 (Figs. 2D and 3A,E).
We also analyzed the expression of the transcription factor Lhx6, which acts downstream of Nkx2-1 to direct the migration and specification of several telencephalic cell types, including some GPe neurons (Sussel et al., 1999; Liodis et al., 2007; Du et al., 2008; Flandin et al., 2010). In line with the requirement of Nkx2-1 for the expression of Lhx6 (Sussel et al., 1999; Butt et al., 2008; Flandin et al., 2010), coexpression of Nkx2-1 and Lhx6 in individual GPe neurons was very common. Indeed, 96% of Nkx2-1+ neurons also expressed Lhx6, and 97% of Lhx6+ neurons also expressed Nkx2-1 (Fig. 3B,E,F). Virtually all (99%) PV+ neurons coexpressed Lhx6, whereas only 86% of Lhx6+ neurons coexpressed PV (Figs. 1D and 3B,F). Coexpression of Lhx6 and ChAT in GPe neurons was exceedingly rare (Figs. 2H and 3F). As could be expected from the near-equivalent expression profiles of Lhx6 and Nkx2-1 in GABAergic GPe neurons, coexpression of Lhx6 and FoxP2 was not observed (Figs. 2D and 3B,F). Because Nkx2-1+/Lhx6+ neurons and PPE+/FoxP2+ neurons accounted for ∼68% and 30% of GABAergic GPe neurons, respectively, the expression of Nkx2-1 and/or Lhx6 ostensibly captures the remaining PV−/PPE− GABAergic neuron population. These data collectively suggest that, within a framework of dichotomous cell types in adult GPe, Nkx2-1 and Lhx6 can be used interchangeably as both comprehensive and highly selective markers for GABAergic prototypic neurons.
The basic helix-loop-helix transcription factor Npas1 has been recently shown to mark a novel population of GABAergic cells in mouse GPe, although reports disagree about the prevalence and extended molecular profile of these Npas1+ neurons (Flandin et al., 2010; Nóbrega-Pereira et al., 2010). We found that, on average, Npas1+ neurons constituted 41% of all GPe neurons in adult rats (Fig. 1B). The majority (71%) of Npas1+ GPe neurons coexpressed FoxP2 (Fig. 3C,D,G), whereas virtually all (>99%) FoxP2+ neurons coexpressed Npas1 (Fig. 2D). Most Npas1+/FoxP2− neurons coexpressed Lhx6 and Nkx2-1, and approximately one-half of these neurons additionally expressed PV (Fig. 3D,G). Npas1-expressing neurons comprised 15% of Lhx6+ neurons (Fig. 3F), 13% of Nkx2-1+ neurons (Fig. 3E), and 8% of PV+ neurons (Fig. 1D). Coexpression of Npas1 and ChAT was not observed (Figs. 2H and 3G). These data show that Npas1 is expressed by at least three molecularly distinct GABAergic cell types in GPe and that Npas1 expression cannot be used to neatly discriminate FoxP2+ arkypallidal neurons from Nkx2-1+/Lhx6+ prototypic neurons per se.
Pallidosubthalamic neurons in dopamine-intact rats express Nkx2-1 and PV but not FoxP2
In dopamine-depleted rats, prototypic GPe neurons can be further distinguished from arkypallidal neurons on the basis of their structure and, particularly, by their long-range axonal projections. As a working definition, arising from a modest sample of individually recorded, labeled, and reconstructed neurons (Mallet et al., 2012), all prototypic neurons innervate at least one of the “downstream” BG nuclei, such as the STN, whereas arkypallidal neurons innervate only the striatum. When considering these structural differences in light of the molecular expression profiles we detail above for large GPe cell populations, one might expect that, in dopamine-intact animals, Nkx2-1+/Lhx6+ neurons constitute the majority of GPe neurons innervating the STN, whereas FoxP2+ GPe neurons innervate the striatum but not the STN. To test these two predictions, we retrogradely transduced GPe neurons innervating either STN or striatum with a lentiviral vector expressing EGFP (Kato et al., 2007) and then quantified the proportions of EGFP+ neurons expressing Nkx2-1, PV, and/or FoxP2. Injections of the lentiviral vector into the STN (Fig. 4A) resulted in robust somatodendritic expression of EGFP by GPe neurons (Fig. 4C). The vast majority (99%) of sampled EGFP+ GPe neurons innervating STN (n = 377 neurons from 5 rats) coexpressed Nkx2-1 (Fig. 4C,F). Most (93%) of these EGFP+/Nkx2-1+ neurons additionally expressed PV (Fig. 4C,F). However, EGFP+ pallidosubthalamic neurons did not express FoxP2 (Fig. 4C,F). These data confirm that Nkx2-1+(PV+) neurons are indeed the major GPe cell population innervating STN, thus further ascribing the prototypic cell-type identity to them, and also show that FoxP2+ GPe neurons do not innervate STN. Injections of the lentiviral vector into the dorsal striatum of 6 other rats (Fig. 4B) also resulted in EGFP+ GPe neurons (Fig. 4D), but their predominant molecular phenotypes differed from those innervating STN. Indeed, on average, approximately one-half (48%) of sampled EGFP+ GPe neurons innervating striatum (n = 390 neurons from 6 rats) coexpressed FoxP2, but not Nkx2-1 or PV (Fig. 4D,G). Almost all (96%) of the remaining EGFP+/FoxP2− pallidostriatal neurons coexpressed Nkx2-1; and, of these, approximately one-half also expressed PV (Fig. 4D,G). These data collectively show that FoxP2+ GPe neurons innervate striatum but not STN, thus further ascribing the arkypallidal identity to them. Furthermore, FoxP2+ arkypallidal neurons are the major GPe cell type innervating the central regions of dorsal striatum, with smaller cohorts of Nkx2-1+/PV− neurons and Nkx2-1+/PV+ neurons making up the remainder of the pallidostriatal cell population.
In accordance with previous work (Bevan et al., 1998; Mallet et al., 2012), our data show that both arkypallidal neurons and a minor population of prototypic neurons contribute to the pallidostriatal projection as a whole. Thus, pallidostriatal identity does not tally with a molecular dichotomy defined on the basis of a mutually exclusive expression of FoxP2 or Nkx2-1/Lhx6. Given that Npas1 was not only expressed by virtually all FoxP2+ neurons but also by a minority of Nkx2-1+/Lhx6+ neurons (Figs. 2D and 3C–F), we next quantified the relationship between Npas1 expression and pallidostriatal identity. Of the EGFP+ GPe neurons innervating striatum that were tested for Npas1, FoxP2, and PV (n = 177 neurons), the vast majority (93%) coexpressed Npas1 (Fig. 4E,H). Of these EGFP+/Npas1+ neurons, 48% were FoxP2+ and 22% were instead PV+ (Fig. 4E,H). It is likely that most of the remaining Npas1+/FoxP2−/PV− pallidostriatal neurons were Nkx2-1+/Lhx6+ (Fig. 3G, and compare Fig. 4G,H). These data show that Npas1 is a highly selective molecular marker for GABAergic pallidostriatal neurons, regardless of arkypallidal or prototypic cell-type identities.
Molecular signatures and projection targets of prototypic neurons and arkypallidal neurons in the dopamine-intact rat GPe
Data from our cell counting and retrograde tract-tracing experiments converge to show that prototypic and arkypallidal neurons are optimally and comprehensively delineated by their expression of different sets of transcription factors (Fig. 5). Prototypic neurons are defined by their expression of Nkx2-1 and Lhx6; ∼85% of prototypic neurons additionally coexpress PV and ER81 (Fig. 5A). Arkypallidal neurons are defined by their coexpression of FoxP2 and Meis2; they also express PPE (Fig. 5A). Together, prototypic and arkypallidal neurons account for >98% of all GABAergic GPe neurons. Virtually all GPe neurons innervating STN are of the prototypic cell type (Fig. 5B). Arkypallidal neurons innervate striatum but not STN (Fig. 5B).
Identified prototypic GPe neurons and arkypallidal neurons have different firing rates and patterns in dopamine-intact rats in vivo
Prototypic GPe neurons and arkypallidal neurons exhibit distinct firing rates and patterns in the dopamine-depleted parkinsonian brain in vivo (Mallet et al., 2012), but whether this holds true for these cell types in the normal brain is unknown. To address this, we first recorded individual GPe neurons in anesthetized, dopamine-intact rats during two extreme brain states (SWA and spontaneous activation), as evinced in frontal ECoGs (Mallet et al., 2008b). We then juxtacellularly labeled each neuron with neurobiotin for post hoc testing of molecular phenotype (Mallet et al., 2012). Given that prototypic neurons coexpress Nkx2-1 and Lhx6 (with most additionally expressing PV), and that arkypallidal neurons express FoxP2 (Fig. 5), we used a series of immunoreactions for these and other markers to assign a cell-type identity to each GPe neuron that was recorded in vivo and labeled with neurobiotin.
When the sampled population was considered as a whole, the spike firing of GPe neurons during SWA was extremely variable; some neurons exhibited vigorous, regular discharges (mean firing rate of up to ∼50 spk/s), whereas other neurons had sporadic activity and were almost quiescent (mean rate of ∼0.02 spk/s) (Fig. 6A–C). The firing rates and the fluctuations in firing regularity or “jitter” (as estimated from the SD of ISIs) of recorded GPe neurons varied almost continuously across the sampled population, and the relationship between average firing rate and SDISI was well described by a hyperbolic function (Fig. 6D), in line with a previous study in isoflurane-anesthetized rats (Deister et al., 2013). However, and most importantly, we found that GPe neurons with different molecular signatures tended to distribute within distinct and restricted aspects of this firing rate/pattern continuum (Fig. 6D). Indeed, all identified Nkx2-1+/PV+ GPe neurons (n = 31) fired at relatively high mean rates (>10 spk/s) and with comparatively regular patterns (CVISI < 0.9) during SWA (Fig. 6A,D,F). Identified Nkx2-1+/PV− neurons (n = 7) fired at significantly lower rates than PV+ GPe neurons, although their firing regularities were similar (p < 0.001 and p = 0.228, respectively, Mann–Whitney tests; Fig. 6B,D,F). As expected from our population-level cell counts (Figs. 1234–5), all recorded Nkx2-1+ neurons were also FoxP2−, thus further ascribing the prototypic cell-type identity to them. In general, the spike firing of prototypic GPe neurons (n = 38) was not strongly timed with the ongoing cortical slow oscillations (∼1 Hz) that prevailed during SWA (Fig. 6A,B), but there were a few exceptions (see below). In contrast to the firing of prototypic GPe neurons, identified FoxP2+ GPe neurons (n = 13) fired at low mean rates (<2 spk/s) and in a less regular manner (CVISI >1) during SWA (Fig. 6C,D,G). Five of these FoxP2+ neurons were tested for coexpression of PPE and, as could be predicted from our population cell counts, all five were also PPE+. All 13 FoxP2+ neurons tested negative for expression of PV. Eight of these FoxP2+/PV− neurons were additionally tested for Nkx2-1 expression; none of these neurons were Nkx2-1+, thus further ascribing an arkypallidal identity to the FoxP2+ GPe neurons recorded here. These arkypallidal neurons typically fired single spikes, or small groups of spikes, in time with the peaks of cortical slow oscillations (Fig. 6C).
Defining whether and how the spike firing of BG neurons is temporally related to the stereotyped cortical slow oscillation have utility for elucidating the potential contributions of different sets of inputs to BG activity (Magill et al., 2001; Walters et al., 2007). Although the power of slow oscillations in frontal ECoGs recorded with arkypallidal neurons or prototypic neurons was similar (Mann–Whitney test on ECoG power in 0.4–1.6 Hz band), the activity of arkypallidal neurons appeared more tightly coupled to the ongoing slow oscillations (Fig. 6A–C). To further examine the temporal relationship between activity in GPe and cortex during SWA, we used the Hilbert transform to analyze the instantaneous phase of the spiking of arkypallidal and prototypic neurons with respect to cortical slow oscillations (Nakamura et al., 2014). To qualify for these and related circular statistical analyses, a GPe neuron had to fire ≥40 spikes during the recording, a sampling criterion that helped to ensure accurate determination of circular means, and the significance of any phase-locked firing (Nakamura et al., 2014). This meant that 4 arkypallidal neurons with very low firing rates (means of <0.1 spk/s) were excluded from these analyses. The spikes of all qualifying arkypallidal neurons (n = 9) were significantly locked to the slow oscillations (Rayleigh's Uniformity Tests). A smaller proportion of PV+ prototypic neurons (n = 26 of 31 qualifying neurons) showed significant phase locking, but all PV− prototypic neurons (n = 7) were significantly locked. Other than this difference between PV+ and PV− prototypic neurons, their firing with respect to slow oscillations was similar, and so, the two groups of prototypic neurons were pooled for further analysis. When the spike train of each significantly phase-locked arkypallidal neuron (n = 9) and prototypic neuron (n = 34) was analyzed, both types of GPe neuron preferentially fired around the peak of cortical slow oscillations, defined as a phase angle of between >270° and <90° (Fig. 6H,I). However, compared with prototypic GPe neurons, the peak of activity of arkypallidal neurons was larger in the phase histogram (Fig. 6H,I), giving a first indication of tighter phase locking as a population. Circular plots of the preferred phases of these individual phase-locked arkypallidal and prototypic neurons confirmed the strong tendency of the former cell type to fire around the peaks of the slow oscillations (Fig. 6H,I). Indeed, although all arkypallidal neurons preferentially fired in time with peaks (Fig. 6I), not all prototypic neurons preferentially fired likewise; 20% of prototypic neurons fired more in time with the troughs of the slow oscillations (Fig. 6H). Although the mean angles of the preferred phases (i.e., the circular means) were similar for the two cell types (arkypallidal neurons, 343.4 ± 12.4°; prototypic neurons, 348.4 ± 10.2°; Watson–Williams F test), the vector lengths of arkypallidal neurons were significantly greater than those of prototypic neurons (p < 0.001, Mann–Whitney test), thus confirming that the preferred firing of individual arkypallidal neurons was comparatively more concentrated (i.e., more tightly locked).
Most prototypic neurons (24 of 38) were subsequently recorded during periods of spontaneous cortical activation, as evinced by a loss of ECoG slow oscillations (Fig. 6A,B). On average, the firing rate and regularity of prototypic neurons during cortical activation were significantly higher than those during SWA (p = 0.007 and p < 0.001, respectively, Mann–Whitney tests), with almost all of the tested neurons (23 of 24) responding with modest increases in their firing during transition from SWA to an activated brain state (Fig. 6F). Most arkypallidal neurons (11 of 13) were also recorded during spontaneous cortical activation (Fig. 6C). Both the average firing rate and regularity of arkypallidal neurons during activation were significantly higher than those during SWA (p < 0.001 and p < 0.001, respectively, Mann–Whitney tests), with almost all tested neurons (10 of 11) responding with pronounced increases in their firing rates during brain state transition (Fig. 6G). On average, arkypallidal neurons fired at significantly lower rates and regularities than prototypic GPe neurons during both SWA and activation (all p < 0.001, Mann–Whitney tests). However, during activation, there was a marked overlap in the firing properties of the two cell types (Fig. 6E), reiterating the necessity of controlling for brain state as well as the importance of unequivocally identifying the cell type of neurons recorded in GPe in vivo.
Together, these data show that, in the dopamine-intact brain in vivo, molecularly identified prototypic GPe neurons and arkypallidal neurons exhibit largely distinct firing rates and regularities (regardless of brain state), as well as dissimilar levels of temporal coupling with ongoing cortical slow oscillations during SWA.
The autonomous firing and intrinsic membrane properties of identified prototypic GPe neurons and arkypallidal neurons are distinct
The distinct firing rates/patterns of prototypic and arkypallidal neurons in the dopamine-intact brain in vivo could arise, at least in part, because these cell types have different intrinsic membrane properties. To address this possibility, we recorded the autonomous and driven firing of individual, molecularly identified GPe neurons in brain slices acutely prepared from juvenile (P20-P35) dopamine-intact rats. We first tested whether some definitive aspects of the molecular profiles of neurons in adult rat GPe held true for neurons in juvenile rat GPe. On average, Nkx2-1+ neurons and PV+ neurons constituted, respectively, just over two-thirds (69%) and one-half (52%) of all GPe neurons. Most (73%) Nkx2-1+ neurons coexpressed PV, whereas nearly all (99%) PV+ neurons coexpressed Nkx2-1. On average, FoxP2+ neurons comprised just over one-fourth (28%) of all GPe neurons. There was a high degree of overlap in the expression of FoxP2 and PPE; 94% of FoxP2+ neurons coexpressed PPE, and >99% of PPE+ neurons coexpressed FoxP2. Importantly, coexpression of Nkx2-1 and FoxP2, or coexpression of PV and FoxP2, by individual neurons in juvenile GPe was very rare (<1% of Nkx2-1+ or PV+ neurons coexpressed FoxP2 and vice versa). These cell counts from juvenile rat GPe are in good agreement with those from adult rat GPe (Fig. 5A) and suggest that the definitive molecular architecture of adult rat GPe neurons is largely established by the fourth postnatal week. Moreover, these findings confirm that PV expression and FoxP2 expression mark prototypic neurons and arkypallidal neurons, respectively, in the juvenile rat GPe. Thus, after electrophysiological recording in vitro, each GPe neuron was filled with biocytin, recovered, and tested for expression of PV and FoxP2 immunoreactivities.
The autonomous spike firing of synaptically isolated GPe neurons recorded in a perforated-patch configuration was highly variable; some neurons consistently engaged in robust, regular activity (mean firing rate of up to ∼55 spk/s), whereas other neurons fired only infrequently (∼1 spk/s) (Fig. 7A–D). The autonomous firing rates and jitter of recorded GPe neurons varied almost continuously across the sampled population, and the relationship between average firing rate and SDISI was well described by a hyperbolic function (Fig. 7D), similar to that reported in a recent in vitro study (Deister et al., 2013). Thus, in these respects, the diverse GPe neuron firing that we observed in vivo was recapitulated in vitro. In further keeping with our in vivo data, we observed that GPe neurons with different molecular signatures tended to distribute within distinct aspects of this autonomous firing rate/pattern continuum (Fig. 7D). All identified PV+ prototypic neurons tested (n = 14) were autonomously active, with most (79%) firing spikes at relatively high mean rates (>10 spk/s) and regularities (CVISI < 0.13) (Fig. 7A,D). Identified FoxP2+ arkypallidal neurons (n = 18) were also autonomously active, but at comparatively low mean rates (only one neuron fired at >10 spks/s) and regularities (only 44% of neurons had CVISI < 0.13) (Fig. 7C,D). A third, smaller group of PV−/FoxP2− GPe neurons (n = 5; presumably, PV− prototypic neurons) also displayed autonomous activity, but with firing rates and regularities that were often in between those of PV+ and FoxP2+ neurons (Fig. 7B,D). On average, arkypallidal neurons fired at significantly lower rates and regularities than PV+ prototypic neurons (p < 0.001 and 0.026, respectively, Kruskal–Wallis ANOVAs; Fig. 7E). The same held true when all prototypic neurons were pooled together (p < 0.001 and p < 0.001, Mann–Whitney tests). These observations again mirror our in vivo data. Input resistances, and action potential waveforms and thresholds, were similar across these groups of GPe neurons (Table 1). An analysis of autonomous firing showed that the average magnitude of the afterhyperpolarization potential (AHP) of the action potentials fired by arkypallidal neurons was almost twice that of prototypic neurons (Table 1). However, when the firing of arkypallidal and prototypic neurons was driven with current injections to an equivalent rate (∼40 spk/s), the AHP magnitudes of the cell types were similar (Table 1).
Somatic injections of depolarizing current pulses revealed that the “driven” firing of arkypallidal neurons was not as vigorous as that of prototypic neurons (Fig. 7A–C; Table 1). In line with this, the F-I response curves of these cell types was markedly different (Fig. 7F). More specifically, the driven firing rates of arkypallidal neurons were significantly slower across a range of current pulses compared with those of prototypic neurons (Fig. 7F), suggesting that arkypallidal neurons are comparatively less responsive to excitation. Upon somatic injections of hyperpolarizing current pulses, both prototypic and arkypallidal neurons exhibited a membrane potential “sag” (Fig. 7A–C). When prototypic and arkypallidal neurons were hyperpolarized to equivalent peak voltage deflections (to ∼−100 mV [Fig. 7A–C], or to ∼−90, −80, or −70 mV), their sag amplitudes were similar (Kruskal–Wallis ANOVAs). These data suggest that the subunit composition, density, and/or distribution of hyperpolarization-activated cyclic nucleotide-gated cation (HCN) channels along the somatodendritic membranes of the two cell types are also similar (Biel et al., 2009), although voltage-clamp recordings would be needed to resolve this in a more definitive manner. Compared with prototypic neurons, a smaller proportion of arkypallidal neurons had significant “rebound” firing at the break of hyperpolarizing currents pulses, and the mean rates of arkypallidal neuron rebound firing were also lower (Fig. 7A–C; Table 1). Together, these results support the conclusion that prototypic neurons and arkypallidal neurons differ in their autonomous firing and some other active membrane properties.
Prototypic and arkypallidal neurons exhibit different propensities to engage in autonomous “rate wandering”
Individual GPe neurons (of unknown cell type or types) can display a large degree of heterogeneity in the rate and regularity of their autonomous firing (Deister et al., 2013). Indeed, long-duration recordings in vitro have revealed that the firing of individual neurons can slowly (over a few minutes), spontaneously, and independently fluctuate or “wander,” such that one cell can eventually visit many of the rates and patterns seen in the population activity continuum (Deister et al., 2013). These data suggest that caution must be exercised when classifying GPe neurons on the basis of electrophysiological parameters alone, and further raise the possibility that the distinct autonomous firing of prototypic and arkypallidal neurons might not hold true across long sampling periods. To address this issue and test whether rate wandering is a feature common to both GPe cell types, we conducted long-duration (15–30 min) cell-attached recordings of molecularly identified neurons (Fig. 8). On average, the firing rates of prototypic and arkypallidal neurons in cell-attached recordings were, respectively, similar to those of these cell types recorded for shorter durations in a perforated-patch configuration (Mann–Whitney tests). However, the prolonged cell-attached recordings suggested that these cell types have different tendencies to engage in slow activity fluctuations. We observed that the autonomous firing of individual PV+ prototypic neurons (n = 8) did not fluctuate much over these long recording periods (Fig. 8A), and the ISI histograms of these neurons were typically symmetrical (Fig. 8B). In comparison, the firing of FoxP2+ arkypallidal neurons (n = 6) and PV−/FoxP2− prototypic neurons (n = 4) wandered to greater extents over time, and their ISI histograms were positively skewed and leptokurtic (Fig. 8C–F). Although most PV+ prototypic neurons (6 of 8) had ISI distributions with small values of kurtosis and skewness (i.e., values between −2 and 2), neither arkypallidal neurons nor PV− prototypic neurons had ISI distributions in this range (Fig. 8G). These data indicate that the overall consistency of autonomous firing rate/pattern is dependent on GPe cell type. To further examine the propensities of different cell types to engage in rate wandering, we partitioned the long-duration recording of each neuron into groups of 100 adjacent spikes and plotted the SDISI within each group against the corresponding mean firing rate (Fig. 8H,I). This analysis showed that the firing of individual arkypallidal neurons and PV− prototypic neurons wandered to a greater extent than that of individual PV+ prototypic neurons, but also that individual neurons visited only a limited aspect of the population activity continuum (Fig. 8H). When all neurons recorded in a cell-attached configuration were considered, it was evident that, although arkypallidal neurons engaged in modest activity fluctuations over time, their firing rates and regularities did not achieve those typical of PV+ prototypic neurons (Fig. 8I). In summary, these results show that the degree to which individual GPe neurons engage in rate wandering is cell-type dependent and further suggest that the different autonomous firing of prototypic and arkypallidal neurons is unlikely to arise from undersampling of slow activity fluctuations. Similarly, these in vitro results indicate that the different firing rates and patterns of these cell types in vivo cannot be readily explained by intrinsic rate wandering.
Prototypic GPe neurons have larger “persistent” sodium channel currents than arkypallidal neurons
The pace and precision of the autonomous firing of GPe neurons are critically dependent on a subthreshold (persistent) sodium current mediated by voltage-dependent sodium channels, as well as on other currents mediated by HCN channels, small-conductance calcium-activated potassium (SK) channels (with calcium provided by opening of high voltage-activated calcium (CavH) channels), and A-type potassium (Kv4) channels (Chan et al., 2004; Mercer et al., 2007; Deister et al., 2009). Moreover, computer simulations (Günay et al., 2008) suggest that the marked variability in the intrinsic membrane properties of GPe neurons, as defined here and reported previously (Nambu and Llinaś, 1994; Cooper and Stanford, 2000; Günay et al., 2008; Bugaysen et al., 2010; Chuhma et al., 2011; Miguelez et al., 2012; Deister et al., 2013), arises through neurons expressing different densities of the membrane conductances mediated by these and other types of ion channels. We next tested whether the distinct autonomous firing of prototypic and arkypallidal neurons was accompanied by cell-type-dependent differences in their persistent sodium channel currents (INaP). Using a whole-cell voltage-clamp recording configuration, we measured currents evoked in identified GPe neurons by somatic injection of slowly depolarizing ramps of voltage command (Fig. 9A,B) in the absence and in the presence of 1 μm TTX. Isolation of INaP was achieved by subtracting the currents recorded in the control and TTX-infused conditions (Fig. 9A,B). The peak amplitudes of INaP recorded in PV+ prototypic neurons (n = 6) were, on average, approximately double those in FoxP2+ neurons (n = 4), without any difference in channel activation kinetics between the two cell types (Fig. 9C,D) (p = 0.014 and p = 0.257, respectively, Mann–Whitney tests). This result suggests that the autonomous firing rates of prototypic neurons are higher than those of arkypallidal neurons because the former cell type expresses higher densities of the voltage-dependent sodium channels that mediate the subthreshold sodium conductance important for GPe neuron pacemaking. The smaller INaP of arkypallidal neurons might also partly explain why this cell type exhibited comparatively weak rebound firing (Günay et al., 2008).
Chronic dopamine loss differentially affects the firing of identified prototypic GPe neurons and arkypallidal neurons in vivo
A previous study in anesthetized parkinsonian rats predicted that chronic dopamine loss imbalances the activities of prototypic and arkypallidal neurons, such that their normal physiological phenotypes in vivo are teased further apart (Mallet et al., 2012). However, the previous use of spike-firing relationships as the sole classifier of cell type (Mallet et al., 2012) is not optimal because electrophysiological phenotypes in vivo are highly dynamic and can overlap, thus leading to ambiguities. Having defined the in vivo firing properties of prototypic and arkypallidal neurons in the dopamine-intact GPe here, we took the opportunity to test this important prediction in a final comparative analysis that used molecular expression profiles as the cell-type classifier. We thus quantified the extent to which chronic dopamine loss, induced by unilateral 6-OHDA lesions targeted to the midbrain (Mallet et al., 2008a; Mallet et al., 2012), alters the firing properties of molecularly defined arkypallidal and prototypic neurons in vivo. All arkypallidal neurons (n = 9) recorded in anesthetized 6-OHDA-lesioned rats were identified as such by their expression of FoxP2; all of these neurons also expressed PPE but also tested negative for PV expression. Prototypic neurons (n = 30) recorded in lesioned rats were initially identified by their lack of FoxP2 expression; 12 of 30 FoxP2− neurons were tested for Nkx2-1 expression and all were Nkx2-1+.
In 6-OHDA-lesioned rats, the spike firing of GPe neurons during SWA was again extremely variable; some neurons discharged at high rates (mean rate of up to ∼40 spk/s), whereas other neurons were almost quiescent (mean rate ∼0.01 spk/s) (Fig. 10A–C). Across the sampled cell population, there was again a firing rate/jitter continuum that was well described by a hyperbolic function (Fig. 10D). However, molecularly defined prototypic and arkypallidal neurons tended to distribute within distinct and restricted aspects of this continuum (Fig. 10D), in line with data from dopamine-intact animals. Thus, all PV+ prototypic neurons (n = 25) fired at relatively high mean rates (>8 spk/s) during SWA (Fig. 10A,D,F). Although PV− prototypic neurons (n = 5) fired at significantly lower rates than PV+ neurons, their firing regularities were similar (p = 0.026 and p = 0.148, respectively, Mann–Whitney tests) (Fig. 10B,D,F). Arkypallidal neurons fired at significantly lower mean rates (<3 spk/s; p < 0.001) than prototypic neurons during SWA (Fig. 10C,D,G).
During SWA, the average firing rate of prototypic neurons in lesioned rats was not different from that of neurons in dopamine-intact rats (p = 0.188, Mann–Whitney test). However, chronic dopamine loss profoundly altered the firing patterns of prototypic neurons. Indeed, on average, the firing regularity of prototypic neurons was significantly decreased (i.e., higher CVISI values than those of neurons in dopamine-intact rats; p < 0.001; see Fig. 10F), which was likely the result of their firing being strongly timed with ongoing cortical slow oscillations (Fig. 10A,B). In line with this, the firing of all prototypic neurons (n = 30) was significantly phase-locked to cortical slow oscillations in lesioned rats (Rayleigh's Uniformity Tests); and in further contrast to the situation in dopamine-intact animals, all prototypic neurons in lesioned rats preferentially fired around the oscillation troughs (i.e., at phase angles between >90° and <270°) (Fig. 10A,B,H). The mean angle of the preferred phases of these prototypic neurons (179.4 ± 3.4°) was significantly different from that of neurons in intact animals (p < 0.001, Watson–Williams F-Test; Figs. 6H and 10H). The vector lengths of prototypic neurons in lesioned animals were also significantly greater than those of neurons in dopamine-intact animals (p < 0.001, Mann–Whitney test), further indicating that the preferred firing of individual neurons was comparatively more tightly locked after dopamine loss.
Chronic dopamine loss had relatively subtle effects on the firing of arkypallidal neurons. Compared with arkypallidal neurons recorded during SWA in dopamine-intact rats, those recorded in lesioned rats had similar average firing rates and regularities (Fig. 10C,G). Moreover, the spikes of all qualifying arkypallidal neurons (n = 8) in lesioned rats were again significantly locked to cortical slow oscillations, and all neurons preferentially fired around the oscillation peaks (Fig. 10C,I). The mean angle of the preferred phases of these arkypallidal neurons (358.8 ± 8.1°; Fig. 10I) was similar to that of arkypallidal neurons in intact animals (Fig. 6I). Notably, however, the preferred firing of individual neurons was more tightly locked in lesioned rats, as demonstrated by significantly longer vector lengths (p = 0.036). The differential influence of 6-OHDA lesions on the firing of the two GPe cell types during SWA did not arise because of any systematic difference in the power of coincident cortical slow oscillations.
Most prototypic neurons (18 of 30) were subsequently recorded during periods of spontaneous cortical activation (Fig. 10A,B). In stark contrast to the scenario observed in dopamine-intact rats (Fig. 6), the average firing rate of prototypic neurons in lesioned animals during cortical activation was significantly lower than that during SWA (p < 0.001), with almost all of the tested neurons (16 of 18) responding with decreases in their firing during transition from SWA to an activated brain state (Fig. 10F). Furthermore, during cortical activation, prototypic neurons in lesioned rats were hypoactive compared with prototypic neurons in dopamine-intact rats (Figs. 6F and 10F; p < 0.001). Most arkypallidal neurons (6 of 9) were also recorded during spontaneous cortical activation (Fig. 10C). The nature and magnitude of the average responses of arkypallidal neurons to brain state transitions in lesioned rats (Fig. 10G) were indistinguishable from those exhibited by arkypallidal neurons in dopamine-intact rats (Fig. 6G). Because arkypallidal neurons in lesioned rats robustly increased their firing during brain-state transition (Fig. 10G), whereas prototypic neurons decreased their firing (Fig. 10F), a marked overlap in the firing properties of these two cell types emerged during cortical activation (Fig. 10E). Indeed, on average, the firing rates and regularities of arkypallidal neurons were not different from those of prototypic neurons in this brain state (p = 0.617 and p = 0.271, respectively, Mann–Whitney tests). Together, these data show: (1) chronic dopamine loss has a divergent impact on the firing of molecularly identified prototypic GPe neurons and arkypallidal neurons in vivo; and (2) these two cell types can be readily distinguished in lesioned rats by their firing properties during SWA but not by their firing rate and regularities during cortical activation.
Finally, our in vitro and in vivo electrophysiological recordings of identified neurons provide further context for our use of molecular and structural data to define GABAergic GPe cells as either prototypic or arkypallidal neurons (Fig. 5). The complex molecular profiles of GPe neurons in particular call for a careful evaluation of the merits of this dichotomous division; for example, Nkx2-1+/Lhx6+ prototypic neurons could be further subdivided into those that additionally express PV/ER81 and those that do not (Fig. 5). Importantly, though, our analyses at the population level show that not only do Nkx2-1+/PV+ neurons and Nkx2-1+/PV− neurons both innervate STN and striatum, but also that most of their electrophysiological properties are indistinguishable. Thus, the dichotomous grouping of GABAergic GPe neurons into prototypic and arkypallidal cell types is well supported by the current evidence and has clear conceptual utility.
Here, we provide evidence of arkypallidal and prototypic neurons in the dopamine-intact GPe. We demonstrate that these two cell types are distinguished and defined by a host of specialized molecular, structural, and physiological properties. Our data converge to support the concept that a division of labor in GPe, as actuated by arkypallidal and prototypic neurons, extends beyond the disease state and is likely critical for the implementation of normal function.
Molecular architecture of arkypallidal and prototypic neurons
Our analyses afford new insights into the molecular architecture of prototypic and arkypallidal neurons, plus impartial definitions of their population sizes. Prototypic neurons are identified by coexpression of Nkx2-1 and Lhx6 (with many also expressing PV and ER81) and comprise approximately two-thirds of all GPe neurons. In contrast, arkypallidal neurons coexpress FoxP2 and Meis2 and constitute just over one-fourth of all GPe neurons. The functional significance of this dichotomous cell classification is further emphasized by our findings that Nkx2-1+/Lhx6+ prototypic neurons are the major GPe cell population innervating STN and that FoxP2+ arkypallidal neurons do not innervate STN but rather innervate striatum. We observed that Lhx6 and PV are often coexpressed in prototypic neurons, but a recent study instead argues that mouse GPe contains two major cell populations that are distinguished by their expression of Lhx6 or PV (Mastro et al., 2014). This discrepancy might arise from species differences. However, this is unlikely because, in agreement with our results, other studies in mouse (Flandin et al., 2010; Nóbrega-Pereira et al., 2010) show that ER81, PV, and Lhx6 are coexpressed in a major subset of GPe neurons.
Selective expression of transcription factor combinations unambiguously discriminates prototypic from arkypallidal neurons. Our definition of Nkx2-1, Lhx6, FoxP2, and Npas1 expression has particular relevance for understanding the ontogeny of GPe neurons, and our results help place reports of developmental heterogeneity (Flandin et al., 2010; Nóbrega-Pereira et al., 2010) in the context of functionally defined cell types in adult GPe. Indeed, because adult prototypic neurons coexpress Nkx2-1/Lhx6, this cell type almost certainly derives from the medial ganglionic eminence (MGE) of the embryonic subpallium (Sussel et al., 1999; Flames et al., 2007; Flandin et al., 2010; Nóbrega-Pereira et al., 2010). Importantly, work in mice suggests that approximately one-fourth of GPe cells are generated independent of Nkx2-1 expression/function and, thus, do not arise from the MGE; many of these “non-MGE” neurons are Npas1+ (Flandin et al., 2010; Nóbrega-Pereira et al., 2010). Because almost all arkypallidal neurons are Npas1+, we speculate that the subpallial domain(s) of origin of this cell type lie outside of MGE. However, a minority of prototypic neurons are also Npas1+, in line with findings in juvenile mice that a subset of Nkx2-1+ GPe neurons coexpress Npas1 (Flandin et al., 2010). It follows that Npas1 is not a definitive lineage tracer for GPe neurons and does not strictly delineate arkypallidal or prototypic neurons in adult rat. Nevertheless, Npas1 expression is highly selective for pallidostriatal neurons.
Firing properties of GPe cell types in dopamine-intact brain
We provide the first working definitions of the in vivo firing properties of prototypic and arkypallidal neurons in the dopamine-intact brain. Across the whole population of recorded GPe neurons, firing rate and jitter were highly diverse but covaried almost continuously. Such a continuum of GPe neuron firing has been previously interpreted as evidence of a single cell type, albeit one with remarkably heterogeneous properties (Deister et al., 2013). Our recordings of identified neurons offer another perspective; each cell type discretely distributes at either end of this continuum and, on average, prototypic neurons fire at significantly higher rates and regularities than arkypallidal neurons. Thus, the distinct molecular and structural properties of prototypic and arkypallidal neurons are mirrored in their distinct firing properties. On the basis of firing rates/patterns during SWA, these two cell types can be distinguished with some confidence. However, there is more overlap in their firing properties during cortical activation. The activity profile of prototypic neurons is broadly similar to that of the majority of GPe units (of unknown cell type) recorded in awake dopamine-intact rats and monkeys; these units are typically classified as “high-frequency discharge” neurons that occasionally pause (DeLong, 1971; Anderson and Turner, 1991; Elias et al., 2007; Benhamou et al., 2012). Conversely, the firing rates/patterns of arkypallidal neurons are more akin to those of the 15%–25% of GPe units alternatively classified as “low-frequency discharge bursting” neurons (DeLong, 1971; Benhamou et al., 2012).
Our in vitro recordings of identified prototypic and arkypallidal neurons offer valuable insights into the intrinsic basis of the diverse firing observed in vivo. We determined that the autonomous firing of prototypic neurons was faster and had less jitter than that of arkypallidal neurons and that these differences cannot be readily explained by rate wandering (Deister et al., 2013). It is thus likely that the comparatively high firing rates and regularities of prototypic neurons in vivo partly arise from their ability to engage in more robust autonomous firing. The activity continuum that we and others observe in GPe in vitro (Deister et al., 2013) should be interpreted in light of prototypic and arkypallidal neurons exhibiting different autonomous firing properties. Taken with differences in other intrinsic properties (Table 1), our data lay foundations for defining the intrinsic physiological signatures of prototypic and arkypallidal neurons. Moreover, our finding of cell-type-dependent physiological diversity, together with the presumed sampling of both cell types in past in vitro studies, also helps explain why GPe neurons are reported to exhibit a wide array of intrinsic properties (Cooper and Stanford, 2000; Günay et al., 2008; Bugaysen et al., 2010; Chuhma et al., 2011; Miguelez et al., 2012).
Autonomous firing of GPe neurons depends on several membrane conductances, including INaP and those mediated by HCN, SK, Kv4, and CavH channels (Chan et al., 2004; Mercer et al., 2007; Deister et al., 2009). Computer simulations suggest that distinct intrinsic properties similar to those of prototypic and arkypallidal neurons can be generated by varying the densities of these and other conductances (Günay et al., 2008). Notably, GPe neurons capturing a key feature of the arkypallidal phenotype (i.e., relatively low rates of autonomous firing) are predicted to have relatively low INaP (Günay et al., 2008). Our finding that arkypallidal neurons have comparatively smaller INaP offers one candidate mechanism for the diverse autonomous firing of the two cell types and agrees well with simulation predictions. Because arkypallidal neurons exhibited larger AHP magnitudes (during autonomous firing at least) and lower firing rates, one might also expect that they express higher densities of CavH and Kv4 channels, respectively (Günay et al., 2008).
Firing of GPe neurons in vivo arises from a complex interplay between their intrinsic membrane properties and inputs, particularly those arising from striatum and STN. On average, the high-frequency firing of prototypic neurons was positively, but weakly, coupled to the peaks of cortical slow oscillations. Because GABAergic striatal projection neurons and glutamatergic STN neurons also fire most around these peaks (Magill et al., 2001; Sharott et al., 2012), the preferred firing of prototypic neurons better reflects periodic excitation from STN. However, the robust autonomous activity of prototypic neurons would ensure they continue to fire at high rates outside of rhythmic excitations. The low-frequency firing of arkypallidal neurons was strongly coupled to oscillation peaks, which again presumably results from rhythmic excitation by STN. However, their “weaker” autonomous activity, and inputs from prototypic neurons (Mallet et al., 2012), would ensure only modest firing outside of periodic excitations. Increased firing of prototypic and arkypallidal neurons upon transition to activated brain states probably arises because of increased STN output rather than altered striatal output (Magill et al., 2001; Mallet et al., 2005; Sharott et al., 2012).
Impact of dopamine loss and wider implications
Chronic dopamine loss markedly changed the activity of prototypic neurons, and their preferred firing phases became consistently aligned to slow oscillation troughs. Because both striatopallidal and STN neurons in 6-OHDA-lesioned rats are hyperactive, but continue to fire most during oscillation peaks (Mallet et al., 2006, 2008a; Zold et al., 2012), the altered firing phases of prototypic neurons likely arise because their input is now tipped in favor of inhibition from striatum. Abnormally efficacious inputs from striatum might also explain why prototypic neurons are hypoactive during cortical activation. After lesions, arkypallidal neurons continue to fire in time with oscillations peaks (albeit more tightly synchronized) and still increase their firing during activation. Therefore, firing of arkypallidal neurons is not as obviously constrained/shaped by striatum. This raises the possibility that prototypic and arkypallidal neurons are disparately innervated by striatum and/or STN. In support of this, computational modeling suggests that not only do prototypic neurons receive larger inputs from striatum, and smaller inputs from STN, compared with arkypallidal neurons, but also that these disparate connection strengths lead to the emergence of their “anti-phase” firing in parkinsonism (Nevado-Holgado et al., 2014). In conclusion, the divergent impact of dopamine loss on the activity of prototypic and arkypallidal neurons reiterates the importance of these dichotomous cell types for GPe function and dysfunction.
This work was supported by the Medical Research Council United Kingdom Awards UU138197109 and MC_UU_12020/5, Parkinson's UK Grant G-0806 and Monument Trust Discovery Award J-0901, Wellcome Trust Investigator Award 101821 to P.J.M., Association France Parkinson Ph.D. fellowship to A.A., Fondation de France, and Agence Nationale de la Recherche Grant 08-JCJC-0087 and Grant LABEX BRAIN ANR-10-LABX-43. A.S. was supported in part by the European Union Marie Curie European Re-integration Grant SNAP-PD. K.C.N. was supported in part by the Human Frontier Science Program (LT000396/2009-L). S.V.A. was supported in part by a Wellcome Trust Doctoral Studentship in Neuroscience 099711. F.N.G. was supported in part by a University of Oxford Clarendon Fund Scholarship. We thank S. Arber, A. Buchberg, C. Cozzari, and S. McKnight for gifts of antibodies; S. Butt, N. Doig, J. Kaufling, L. Magno, V. Pachnis, and N. Kessaris for insightful scientific discussions; the Bordeaux Imaging Center of the University of Bordeaux Segalen; and E. Norman, B. Micklem, L. Conyers, C. Johnson, C. Poujol, and S. Marais for technical assistance.
The authors declare no competing financial interests.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License Creative Commons Attribution 4.0 International, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.
- Correspondence should be addressed to either of the following: Dr. Peter J. Magill, Medical Research Council Brain Network Dynamics Unit, Department of Pharmacology, University of Oxford, Mansfield Road, Oxford OX1 3QT, United Kingdom, ; or Dr Jérôme Baufreton, Institut des Maladies Neurodégénératives, Unité Mixte de Recherche Centre National de la Recherche Scientifique 5293, 146 Rue Léo Saignat, 33076 Bordeaux Cedex, France.
This article is freely available online through the J Neurosci Author Open Choice option.