Abstract
The most caudal part of the striatum in rodents, the tail of the striatum (TS), has many features that distinguish it from the rostral striatum, such as its biased distributions of dopamine receptor subtypes, lack of striosomes and matrix compartmentalization, and involvement in sound-driven behaviors. However, information regarding the TS is still limited. We demonstrate in this article that the TS of the male mouse contains GABAergic neurons of a novel type that were detected immunohistochemically with the neurofilament marker SMI-32. Their somata were larger than cholinergic giant aspiny neurons, were located in a narrow space adjacent to the globus pallidus (GP), and extended long dendrites laterally toward the intermediate division (ID) of the trilaminar part of the TS, the region targeted by axons from the primary auditory cortex (A1). Although vesicular glutamate transporter 1-positive cortical axon terminals rarely contacted these TS large (TSL) neurons, glutamic acid decarboxylase-immunoreactive and enkephalin-immunoreactive boutons densely covered somata and dendrites of TSL neurons, forming symmetrical synapses. Analyses of GAD67-CrePR knock-in mice revealed that these axonal boutons originated from nearby medium spiny neurons (MSNs) in the ID. All MSNs examined in the ID in turn received inputs from the A1. Retrograde tracers injected into the rostral zona incerta and ventral medial nucleus of the thalamus labeled somata of TSL neurons. TSL neurons share many morphological features with GP neurons, but their strategically located dendrites receive inputs from closely located MSNs in the ID, suggesting faster responses than distant GP neurons to facilitate auditory-evoked, prompt disinhibition in their targets.
SIGNIFICANCE STATEMENT This study describes a newly found population of neurons in the mouse striatum, the brain region responsible for appropriate behaviors. They are large GABAergic neurons located in the most caudal part of the striatum [tail of the striatum (TS)]. These TS large (TSL) neurons extended dendrites toward a particular region of the TS where axons from the primary auditory cortex (A1) terminated. These dendrites received direct synaptic inputs heavily from nearby GABAergic neurons of the striatum that in turn received inputs from the A1. TSL neurons sent axons to two subcortical regions outside basal ganglia, one of which is related to arousal. Specialized connectivity of TSL neurons suggests prompt disinhibitory actions on their targets to facilitate sound-evoked characteristic behaviors.
Introduction
The most caudal part of the striatum in rodents, the tail of the striatum (TS), has recently gained much attention because of its structural uniqueness (Gangarossa et al., 2013b, 2019; Miyamoto et al., 2018, 2019), input/output connectivity (Jiang and Kim, 2018; Miyamoto et al., 2018), functional significance in sound-evoked behaviors (Li et al., 2021), and response patterns to psychostimulants (Gangarossa et al., 2013a, 2019; for review, see Valjent and Gangarossa, 2021).
The striatum has a dichotomous structural organization, striosomes and matrix compartmentalization. In histological sections, striosomes can be detected as numerous island-like domains spreading inside the striatum, whereas the space surrounding striosomes is the matrix. Our recent studies demonstrated that there are two exceptional regions where this pronounced compartmentalization is absent (Miyamoto et al., 2018). The first region is located at the most lateral part of the rostral striatum close to the white matter. This “striosome-free space” receives inputs from the primary motor and sensory cortices, whereas the compartmentalized regions of the striatum receive inputs from association cortices. Another region lacking compartmentalization is found in the ventral half of the TS. This part can be recognized as a trilaminar profile showing heterogeneous immunoreactivity for substance P (SP), enkephalin (Enk), and tyrosine hydroxylase (TH; Miyamoto et al., 2018, 2019). The trilaminar part matches the region showing both the distinctive expression pattern of dopamine receptor subtypes (Gangarossa et al., 2013b; Miyamoto et al., 2018) and the unique neurochemical responses to various stimuli (Gangarossa et al., 2013a, 2019).
Functionally, the TS has been related to processing in sensory-evoked behaviors, such as decision-making in sound-driven discrimination tasks (Znamenskiy and Zador, 2013). A recent study further demonstrated that innate defensive behaviors toward looming sounds, mimicking approaching threats, are mediated through corticostriatal circuits targeting the D2 receptor-rich region of the TS (Li et al., 2021). Importantly, this region corresponds to the intermediate division of the trilaminar part of the striatum (Miyamoto et al., 2018), which is characterized by a high proportion (>80%) of D2 receptor-expressing neurons, selective innervation from the primary auditory cortex (A1), and poorer innervation by dopaminergic axons (Miyamoto et al., 2018, 2019).
Apart from these recent studies of the TS, classical anatomical studies noted a peculiar profile of extremely long dendrites extending from a large soma in the striatum (Bolam et al., 1985). Both the soma and dendrites of these neurons are densely covered by boutons that are immunohistochemically labeled for glutamic acid decarboxylase (GAD), the GABA-synthetic enzyme. This leads to a dendritic appearance similar to railroads consisting of parallel arrays of GAD-positive boutons that abut densely on long immunonegative dendrites. Because no immunohistochemical markers have been available thus far to visualize their large somata or dendritic ramification, they remain an unidentified but intriguing constituent of the striatum.
In this study, we demonstrate that the antibody SMI-32 visualizes this neuronal population in a specific manner. SMI-32 recognizes a nonphosphorylated epitope at 168,000 and 200,000 molecular weight subunits of neurofilament proteins (Sternberger and Sternberger, 1983) and has been used preferentially to obtain clear immunohistochemical images of pyramidal cells in the infragranular layers of the neocortex (Campbell and Morrison, 1989; Kaneko et al., 1994; van der Gucht et al., 2001). We characterized the morphological features of SMI-32-labeled neurons in the TS at both the light and electron microscopic (EM) levels and further examined their connectivity with other brain regions using anterograde and retrograde tracers. The present results support recent findings illuminating the functional significance of the TS in auditory-evoked specific behaviors.
Materials and Methods
Tissue preparation.
All surgical procedures were conducted according to the National Institutes of Health (NIH) Guidelines for the Care and Use of Laboratory Animals (NIH publication 86-23), and all protocols were approved by the Institutional Animal Care and Use Committee at Kumamoto University. All efforts were made to minimize the number of animals used and their suffering.
Fourteen male C57BL/6J mice (weight, 21–26 g; age, 7–8 weeks) were used for histological analysis, and 15 male C57BL/6J mice of the same age were subjected to tracer injection experiments. For fixation of brains, animals were deeply anesthetized by inhalation of isoflurane and perfused via the ascending aorta with 0.01 m PBS, pH 7.4, followed by 50 ml of 4% paraformaldehyde in 0.1 m phosphate buffer (PB), pH 7.4, or a mixture of 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 m PB for EM at room temperature. Brains were removed from the skull and stored overnight in the same fixative at 4°C; then, the fixative was replaced by PBS containing 0.1% sodium azide.
Tracer injections.
The retrograde tracer used was Fluoro-Gold (FG; 4% dissolved in saline; Fluorochrome), and the anterograde tracers used were 10 kDa biotinylated dextran amine (BDA; 2.5% in saline; Thermo Fisher Scientific) or phaseolus vulgaris-leucoagglutinin (phal; 2.5% in 10 mm PB, pH 8.0; Vector Laboratories). Each animal was deeply anesthetized by inhalation of isoflurane and positioned in a stereotaxic frame (model SR-5 M-HT, Narishige Scientific Instrument Lab). A burr hole was drilled in the appropriate position of the skull, and a glass microelectrode (outside tip diameter, 40–50 µm) containing the tracer solution was stereotaxically inserted into the brain. FG was injected into the substantia nigra (bregma, −3.2 mm; lateral, 1.25 mm; depth, 4.5 mm), entopeduncular nucleus (bregma, −1.3 mm; lateral, 1.8 mm; depth, 3.9 mm), ventral medial nucleus (VM) of the thalamus (bregma, −1.6 mm; lateral, 0.7 mm; depth, 3.8 mm), and rostral part of the zona incerta (ZI; bregma, −1.5 mm; lateral, 1.3 mm; depth, 3.9 mm). Following injections into the ZI, the distribution of parvalbumin (PV) neurons around the injection site was examined immunohistochemically: the rostral sector of the ZI was devoid of PV neurons, whereas the more caudal part, consisting of dorsal and ventral sectors running horizontally in two rows, was discriminated by the presence of many PV neurons residing in the ventral sector (Kolmac and Mitrofanis, 1999). BDA and phal were injected into the medial division of the medial geniculate nucleus (MGM; bregma, −2.9 mm; lateral, 1.9 mm; depth, 3.3 mm). In another set of experiments, phal was injected into the A1 (bregma, −2.69 mm; lateral, 3.9 mm; depth, 1.0 mm) of GAD67-CrePR mice. Injection coordinates were derived from the mouse brain atlas of Paxinos and Franklin (2008). Tracer was injected iontophoretically into the targeted site by passing a positive-pulsed 5–7 µA duty cycle (2 s on/2 s off) for 20 and 5 min, respectively. After surgery, the wound was closed and topical analgesics (2% lidocaine gel; Fujisawa) were applied to the wound. Mice were housed singly in small compartments that were temperature (23°C) and light (12 h light/dark cycle) controlled. After a survival period of 7 d, the mice were perfusion fixed as described above.
Immunohistochemistry.
Serial 40-µm-thick coronal sections were cut using a vibrating microtome (model TTK-3000, Dosaka) from the brain block that contained the entire striatum. After cryoprotection in 25% sucrose in PBS, sections placed on aluminum foil were rapidly frozen in liquid N2 vapor, rapidly thawed in 25% sucrose in PBS, and then processed for triple-fluorescence immunohistochemistry, as previously described (Fukuda and Kosaka, 2000; Miyamoto and Fukuda, 2015) using slight modifications. The primary and secondary antibodies used are listed in Tables 1 and 2, respectively. Briefly, sections were incubated with a blocking solution containing 5% normal donkey serum (Jackson ImmunoResearch) or 1% bovine serum albumin (Sigma-Aldrich), 0.3% Triton X-100 and 0.1% sodium azide in PBS at 20°C overnight, followed by a mixture of three different primary antibodies diluted in the blocking solution at 20°C for 7 d, and an appropriate set of secondary antibodies labeled with fluorescent dyes of different wavelengths overnight. The long incubation period with the primary antibodies was essential to improve the permeation of the antibodies into the deep part of the 40-μm-thick sections and thus to obtain confocal images of constant and sufficient quality throughout the depth of the section (Fukuda et al., 1998; Fukuda and Kosaka, 2000). When the sections were processed for GAD immunostaining, Triton X-100 was omitted from the incubation medium of all steps because it reduced the GAD immunoreactivity in the soma (Fukuda et al., 1996, 1997).
List of the primary antibodies used
List of the secondary antibodies used
Confocal laser-scanning microscopy.
Immunostained sections were mounted in Vectashield (Vector Laboratories) and examined using a confocal laser-scanning light microscope (CLSM; model C2, Nikon) at 4× [Plan Apo; numerical aperture (NA) = 0.2], 10× (Plan Fluor; NA = 0.3), 20× (Plan Fluor; NA = 0.5), 40× (Plan Fluor; NA = 0.75), and 60× (Plan Apo VC; NA = 1.4) objectives. Single laser beams (488, 543, and 633 nm) and a filter set of BA 515/30, BA590/50, and 650 LP were alternately used to collect images for different fluorescence signals. The size of each frame was 1024 × 1024 pixels, and images of the optical slices were acquired from the section surface to the bottom at the preset optimal step size and stored as a stacked file for each frame using the three single laser beams alternately at each z-position of the stage to collect images for different fluorescence signals.
Immunoelectron microscopy.
Dual-label immunoelectron microscopy was performed as previously described with slight modifications (Fukuda, 2017; Shigematsu et al., 2019). Sixty-micrometer-thick coronal sections were cut using a vibrating microtome from the brains fixed for EM. After rapid freezing/thawing as described above, sections were treated with 1% NaBH4 in PBS for 30 min to quench the aldehyde group remaining in the tissue (Kosaka et al., 1986). Sections were incubated with blocking solution that did not contain detergent for 1 h, followed by incubation with a mixture of primary antibodies containing rabbit anti-PV (1:5000) and mouse anti-Enk (1:400) at 4°C for 4 d and biotinylated anti-mouse IgG (1:200; Jackson ImmunoResearch) at 4°C overnight. Immunoreactivity was first visualized for Enk, for which sections were treated using an ABC Standard Kit (Vector Laboratories), followed by color development using a mixture of 0.02% diaminobenzidine (DAB) and 0.04% nickel ammonium sulfate. Then, sections were rinsed in PBS containing 0.1% sodium azide to inactivate peroxidase, rinsed twice with azide-free PBS, and incubated with rabbit peroxidase-anti-peroxidase complex (1:500; Jackson ImmunoResearch) in PBS at 4°C overnight. After rinsing in PBS, sections were treated with 0.05% DAB for visualization of PV immunoreactivity. Sections were postfixed with 1% OsO4 in PB for 1.5 h on ice, stained en bloc with 1.5% uranyl acetate for 1 h, dehydrated with ethanol, cleared with propylene oxide, and flat embedded between slides and coverslips using epoxy resin. After taking photographs and removing coverslips, ultrathin sections were cut and stained with uranyl acetate and lead citrate, and then examined with an electron microscope (model H-7700, Hitachi). A correlative CLSM-EM examination was performed as previously described (Fukuda, 2017; Shigematsu et al., 2019), using anti-PV, anti-SP, and anti-Enk antibodies. The density of both bouton contacts and synaptic contacts per unit length of dendrite was measured with a CLSM and electron microscope, respectively.
GAD67-CrePR knock-in mice.
The generation of GAD67-CrePR knock-in mice was described previously (Esumi et al., 2021). GAD67-CrePR mice express a mifepristone-inducible Cre recombinase under the control of the GAD67 gene. To label GAD67-positive cells, 0.006 mg/g body weight mifepristone (Cayman Chemical) was administered to GAD67-CrePR mice. Mifepristone was dissolved in 100% ethanol to 50 mg/ml and then diluted in corn oil (Wako) to 5 mg/ml. To detect Cre-mediated recombination in GAD67-CrePR knock-in mice, mice that express tdTomato reporter on Cre-mediated recombination were used (stock #007909 with an ICR background, The Jackson Laboratory; Madisen et al., 2010). GAD67-CrePR knock-in male mice were mated with R26-tdTomato (Ai9) reporter female mice. As was shown in our previous study, GAD67-positive cells were sufficiently labeled with tdTomato 24 h after the injection (Esumi et al., 2021). Five GAD67-CrePR knock-in mice were fixed for CLSM 3 d after mifepristone injection.
Experimental design and statistical analysis.
Two types of neurons labeled with SMI-32 were counted in three mice using a computer-assisted neuron tracing system (Neurolucida, MBF Bioscience). In each animal, two sections at 1.08 and 0.56 mm from bregma were selected for counting cells at rostral positions in the striatum, and eight sections between −1.12 and −1.68 mm from bregma, which were separated from each other at a distance of 0.08 mm, were selected for counting cells at caudal positions in the striatum. CLMS images were taken using a 20× objective. Labeling intensity in somata was measured as gray levels of fluorescence signals (no signal, 0; maximum level, 255) using a 20× objective and the open-source software ImageJ (version 1.50; NIH). CLSM settings such as laser power, gain of photomultiplier, and size of the confocal iris were kept constant during the acquisition of images for quantification of gray levels in somata. To examine the cell body size of each neuron, the largest circumference observable in serial optical slices in CLSM images taken with a 20× objective was measured using Neurolucida. Distribution of dendrites in each division of the trilaminar part was analyzed quantitatively using Neurolucida. After reconstructions of dendrites of 13 cells from three mice, lengths of dendritic segments located in each division were measured using the Neuroexplorer program of Neurolucida. In each neuron, the sum of dendritic lengths in each division was divided by the total dendritic length of that cell to yield the proportion of dendritic length in each division. To quantify the density of bouton contacts on TS large (TSL) neurons, PV immunolabeling was used to visualize neuronal somata and dendrites, because SMI-32 labeled mainly filamentous structures inside cells, but it did not fill the cytoplasm (see Fig. 2A), leading to underestimation of labeled profiles. Somata and dendrites of PV-positive neurons were identified as those of TS large neurons when they were located in the appropriate positions in the trilaminar part of the TS and were surrounded by numerous Enk-positive elements. In high-resolution CLSM analysis, confocal images of VGluT1-positive and VGluT2-positive boutons were acquired using 60× objective, and the numbers of boutons that contacted cell bodies (n = 11 from three mice), proximal dendrites (n = 13 from three mice), and distal dendrites (n = 11 from three mice) were counted and divided by the perimeter of soma and length of dendrite, respectively.
All quantitative data are expressed as the mean ± SD. All data in graphs are shown as dots in addition to lines showing mean and SD values. The statistical analysis was performed using the public program R, and differences were considered significant for p < 0.05.
Results
SMI-32-immunoreactive neurons in the striatum
The antibody SMI-32 visualized particular neuronal types in many brain regions, including pyramidal cells in the infragranular layers of the neocortex, principal neurons of the globus pallidus (GP), and Purkinje cells in the cerebellum, all of which were consistent with previous observations (Sternberger and Sternberger, 1983; Campbell and Morrison, 1989; Kaneko et al., 1994; van der Gucht et al., 2001). Prominent immunoreactivity in somata and dendrites resulted in the Golgi-like appearance of labeled neurons (Figs. 1B, 2). At higher magnification, SMI-32 immunoreactivity was observed as filamentous structures inside the cytoplasm, indicating the specificity of the antibody that recognizes neurofilament proteins (Fig. 2A).
SMI-32-positive large neurons in the TS. A, Low-power image of a coronal section cut through the TS and stained with SMI-32. The trilaminar part of the TS is divided into medial (M), intermediate (I), and lateral (L) divisions according to the labeling pattern in triple-immunofluorescence shown in C–F. The frame area is enlarged in B. SMI-32 immunoreactivity in other brain areas such as labeling in pyramidal cells in the neocortex and principal neurons in the GP (also see B) is consistent with that seen in previous studies. Str, Striatum. Scale bar, 500 µm. B, Somata of SMI-32-positive large neurons (arrows) are located in the M division of the trilaminar part and extend dendrites laterally, reaching the I division. The blue dashed line demarcates a region where background immunoreactivity for SMI-32 is slightly higher than that in surrounding regions because of labeling in thin processes of dendrites ramifying therein. Scale bar, 100 µm. C–F, Triple-immunofluorescence image in C consists of TH (red), SMI-32 (green), and PV (blue) immunoreactivities, which are shown separately in D, E, and F, respectively. Three divisions of the trilaminar part are defined according to the lower TH immunoreactivity in the intermediate division (D). The intensity of diffuse labeling of not only SMI-32 but also PV is slightly higher in the intermediate division. Scale bar, 200 µm. G, All SMI-32-positive large neurons contained in one hemisphere were reconstructed and superimposed on the image of TH immunoreactivity. Somata are located in the medial division, whereas skewed dendrites are directed toward the intermediate division where immunoreactivity for TH is lower than that in surrounding regions. The images of TH immunostaining in D and G are taken from neighboring sections. Scale bar, 200 µm.
Morphological features of SMI-32-positive large neurons in the TS. A, Double-immunofluorescence image consists of SMI-32 (green) and PV (blue) immunoreactivities. SMI-32-immunoreactive filamentous structures are clearly seen inside both soma and dendrites. Scale bar, 20 µm. B, C, Comparison of type 1 (B, arrow) and type 2 (C, arrowheads) neurons. The latter are much smaller in soma size and weaker in staining intensity. Image in C was taken from the rostral striatum where numerous bundles of fibers passing through are also labeled. A similar staining pattern in fiber bundles is also visible in the dorsal half of the striatum in Figure 1A. Scale bar, 100 µm. D, Reconstruction of three cells contained in Figure 1G. Most of the dendrites are directed to the intermediate (I) division of the trilaminar part. Multiple contours in the intermediate division are tracings of outlines showing areas of low TH signal, acquired from serial 40-µm-thick sections. Scale bar, 200 µm. E, Distribution of the two types of neurons along the rostrocaudal axis. Type 1 neurons are located only in caudal locations, whereas type 2 neurons are found at both rostral and caudal levels. F, Quantitative analysis of the cell body perimeter. Type 1 neurons (n = 32 cells) were significantly larger than type 2 neurons (n = 246 cells; ***p < 0.001, Welch's t test). G, Quantitative analysis of immunoreactivity in somata. Gray levels of type 1 neurons (n = 32 cells) were significantly higher than those of type 2 neurons (n = 246 cells; ***p < 0.001, Welch's t test). Error bars show SDs.
In the striatum, neurons of a novel type were detected with SMI-32. They had large somata located in the ventral part of the TS close to the globus pallidus, and their long dendrites extended laterally (Figs. 1B,G, 2B,D). The location of somata corresponded to the medial division of the trilaminar part of the TS (Fig. 1B; Miyamoto et al., 2018, 2019), whereas laterally oriented dendrites reached the intermediate division, but did not invade the lateral division, of the trilaminar part (Figs. 1B,G, 2D). The definitions of these three divisions (medial, intermediate, lateral) in the TS were originally based on a combination of SP and Enk immunoreactivities (Miyamoto et al., 2018), both of which show differential immunoreactivity in the three divisions (see Fig. 8A, Enk immunoreactivity). Our recent study (Miyamoto et al., 2019) further demonstrated that the labeling pattern of TH immunoreactivity also distinguishes the three divisions such that the intermediate division can be identified by its low immunoreactivity for TH (Fig. 1C–F). In Figure 1G, all SMI-32-positive large neurons that were contained in one hemisphere were reconstructed and superimposed on the TH image. Their skewed dendrites came together to the intermediate division and occupied it collectively.
In addition to this large type of neuron (Fig. 2B), there was another population of SMI-32-positive neurons that were smaller in soma size and showed much weaker immunoreactivity (Fig. 2C). They had short, multipolar dendrites and were distributed throughout the striatum. These apparent morphological differences between the two types were further confirmed quantitatively (Fig. 2E–G). In this analysis, SMI-32-positive neurons were classified by dendritic morphology, not by the location or size of the soma: those with skewed dendrites extending laterally were classified as type 1, whereas those with short dendrites radiating in all directions were classified as type 2. The perimeters of the soma were measured in all SMI-32-positive neurons contained in single sections at both rostral and caudal levels of the striatum. All SMI-32-labeled neurons located at two rostral positions (1.08 and 0.56 mm rostral to bregma) were type 2 and had small somata (Fig. 2E). In contrast, the caudal striatum contained both type 1 and type 2 neurons. The average perimeter of type 1 neurons (70.8 ± 13.9 µm, n = 32 cells) was significantly larger than that of type 2 neurons (40.3 ± 4.4 µm, n = 246 cells; p = 1.1 × 10−13, Welch's t test; Fig. 2F). The immunoreactivity in the soma also differed between the two types; the mean gray level of type 1 neurons (115.3 ± 45.3, n = 26 cells) was significantly higher than that of type 2 neurons (44.9 ± 17.8, n = 59 cells; p = 2.1 × 10−8, Welch's t test; Fig. 2G). Because of their characteristic features, the present study focuses on type 1 neurons, which we term TSL neurons, in the following description. Observations of serial coronal sections covering the entire striatum revealed that TSL neurons were distributed in the caudal positions between −1.16 and −1.68 mm from bregma, which corresponded exactly to the location of the trilaminar part of the striatum (Miyamoto et al., 2019).
The dendritic morphology of TSL neurons was further analyzed quantitatively (Fig. 3). Some TSL neurons had one or two short dendrites that were coursed medially and entered the GP. On average, the proportions of dendritic length located in the GP, medial division, intermediate division, and lateral division were 7.5 ± 14.5%, 36.2 ± 13.1%, 53.6 ± 16.3%, and 2.6 ± 6.6%, respectively (Fig. 3; n = 13 cells in three mice). In one case (cell 13), a long dendrite entering into the GP was observed, with a proportion of 56% in length.
Quantification of the length of dendritic segments of TSL neurons located in each division of the TS. Data are expressed as the proportion of dendritic length in each division to the total dendritic length of each cell. Medially oriented dendrites entering the GP are also included in this quantification.
The total number of TSL neurons in a hemisphere was 21.0 ± 4.1 (n = 4 animals). The proportion of this number to all neuron numbers in the medial division (759.5 ± 86.7, n = 3 mice; the data were calculated from the same sources used in the study by Miyamoto et al., 2018) was 2.7%. Thus, in addition to the specialized morphological features, TSL neurons were also characterized by the small number. As recognizable in Figure 1G, dendrites originating from this small number of somata as a whole covered the intermediate division that received inputs from the A1 (see Fig. 13C).
Colocalization of other neurochemicals
Previous studies have established a population of large neurons in the striatum, termed giant aspiny neurons, that consists of cholinergic interneurons. Thus, we examined the colocalization relationship between SMI-32 immunoreactivity and choline acetyltransferase (ChAT) immunoreactivity in TSL neurons (Fig. 4A). There were no double-labeled cells in either TSL neurons (n = 42) or ChAT-positive neurons (n = 147). Moreover, the perimeter of the soma of TSL neurons (76.6 ± 13.9 µm, n = 42 cells) was significantly larger than that of ChAT-positive neurons (59.5 ± 7.4 µm, n = 147 cells; p = 7.5 × 10−10, Welch's t test; Fig. 4B,C). Therefore, TSL neurons were the largest neurons observed in the mouse striatum.
Colocalization of other neurochemicals. A–C, Comparison between TSL neurons and ChAT-positive neurons. Dual-immunofluorescence image in A consists of ChAT (magenta) and SMI-32 (green) immunoreactivities. There is no overlap between SMI-32-positive neurons (arrows) and ChAT-positive neurons. Another distinctive feature is that SMI-32 neurons are larger than ChAT-positive neurons, as shown in B. Scale bars: A, 50 µm; B, 20 µm. C, Quantitative analysis of the soma size of the two populations. The mean size of SMI-32 neurons (n = 42) was significantly larger than that of ChAT-positive neurons (n = 147 cells: ***p < 0.001, Welch's t test). D, E, PV immunoreactivity in TSL neurons. Dual-immunofluorescence image in D consists of SMI-32 (green) and PV (magenta) immunoreactivities; the former is shown separately in E as a monochrome image. SMI-32-positive TSL neurons (arrows) show immunoreactivity for PV, but not all PV neurons are immunoreactive for SMI-32, as indicated by arrowheads. Scale bar, 50 µm. F–I, Dual-immunofluorescence images in F and H consist of SMI-32 (green) and GAD (magenta) immunoreactivities, with the latter shown in G and I as monochrome images. GAD immunoreactivity inside the soma indicates that this cell is a GABAergic neuron. Note the dense accumulation of GAD-immunoreactive boutons on surfaces of both soma and dendrites at proximal (F, G) and distal (H, I) sites. Scale bars, 10 µm.
Immunohistochemical colocalization was further analyzed using several antibodies that are known to visualize representative neuronal populations in the striatum except for medium-sized spiny neurons. All TSL neurons showed immunoreactivity for PV (Fig. 4D,E). In contrast, not all PV-positive neurons were labeled with SMI-32. Among 193 PV-positive neurons, 22.5% expressed PV alone, and 77.4% coexpressed SMI-32 of either type 1 or type 2. Other immunohistochemical markers of striatal neurons including nitric oxide synthase, somatostatin, calretinin, and calbindin did not label TSL neurons (data not shown). To determine whether TSL neurons are GABAergic neurons, we tested an antibody against GAD, a GABA-synthetic enzyme. We found that TSL neurons showed intense immunoreactivity for GAD in somata (Fig. 4F,G).
It was noteworthy that both somata and dendrites of TSL neurons were surrounded by numerous GAD-immunoreactive boutons (Fig. 4F–I), which was not a general feature of PV-positive neurons in the remaining part of the striatum, suggesting involvement of TSL neurons in specialized circuitry in the striatum (see below).
Principal neurons of the GP were also labeled with SMI-32, and their dendrites formed a dense plexus that was confined within the GP (Fig. 1B; see also Fig. 14). On very rare occasions, however, pallidal neurons that were located close to the border of the striatum extended one or two dendrites laterally toward the striatum, where these dendrites intermingled with dendrites of TSL neurons located in the striatum. The cell body size was compared between TSL neurons and GP neurons. The average perimeter of TSL neurons was 69.0 ± 8.6 µm (n = 36 cells from three mice), which was significantly larger than that of GP neurons (62.4 ± 9.1 µm, n = 32 cells from three mice; p = 0.003 in t test).
Input structures
Neurons in the striatum generally receive massive glutamatergic inputs from the neocortex and thalamus. Axon terminals originating from the former are visualized as VGluT1-immunoreactive boutons, whereas those from the subcortical regions, most likely from the thalamus, are visualized as VGluT2-immunoreactive boutons (Fujiyama et al., 2001). Because the intermediate division (Miyamoto et al., 2018; Li et al., 2021) and medial division (Gangarossa et al., 2013b) of the trilaminar part of the TS selectively receives inputs from the A1, most VGluT1-positive boutons around skewed dendrites of TSL neurons are thought to originate from the A1. However, VGluT1-positive boutons rarely made direct contact with TSL neurons on both the soma and dendrites (Fig. 5A). In this analysis, an important technical point was that SMI-32 labeled mainly filamentous structures inside cells, not filling the cytoplasm. Thus, we used PV immunolabeling to facilitate precise detection of bouton contacts in high-resolution CLSM analysis. Using this procedure, we confirmed the presence of narrow free space excluding VGluT1-positive boutons along the surface membrane of TSL neurons (Fig. 5A, middle column). In contrast, VGluT2-positive boutons made direct contact with both somata and dendrites of TSL neurons (Fig. 5A, right column). The density of bouton contacts per unit length (Fig. 5B–D) was significantly higher for VGluT2 than VGluT1 in cell body (n = 11 cells each; p = 1.46 × 10−9, t test), proximal dendrite (n = 13 each; p = 9.93 × 10−6, t test), and distal dendrite (n = 11 each; p = 8.49 × 10−6, t test).
Glutamatergic input structures of TSL neurons. A, Triple-immunofluorescence images (left column) consist of PV (green), VGluT1 (blue), and VGluT2 (red) immunoreactivities, which are shown in the middle and right columns with a different color set using green for PV and magenta for VGluT1 and VGLuT2 to see the mode of contact more clearly. VGluT1-positive boutons rarely contact TSL neurons, whereas VGluT2-positive boutons make contact with both soma and dendrites of a TSL neuron. Arrows indicate PV-positive boutons abutting on TSL neurons. Scale bars: left column, 10 µm; middle and right columns, 5 µm. B–D, Comparisons of the numerical density of bouton contacts on soma (B), proximal dendrite (C), and distal dendrite (D) between VGluT1-positive and VGluT2-positive boutons. ***p < 0.0001, t test.
To identify the origin of these VGluT2-positive boutons, we injected anterograde tracers into the medial geniculate nucleus (MG), the auditory relay nucleus of the thalamus, based on previous findings that the TS is innervated by neurons in the MG (LeDoux et al., 1985). As shown in Figure 6, anterograde tracers injected into the medial division of the MG were detected in axons and varicose swelling along axons in the trilaminar part of the TS (Fig. 6B2). At higher magnification, labeled boutons of relatively large size contacted dendrites of TSL neurons (Fig. 6C), and these boutons were positive for VGluT2 (Fig. 6C, inset). In contrast, injections into either the dorsal or ventral division of the MG did not lead to labeling in the TS.
Anterograde tracer labeling to determine the source of VGluT2-positive boutons on TSL neurons. A, Injection of phal (red) is into the medial division of the MG, which is located just medial to the dorsal (MGD) and ventral (MGV) divisions of the MG. IP, interpeduncular nucleus; PAG, periaqueductal gray; RN, red nucleus; SC, superior colliculus; SN, substantia nigra. Scale bar, 1 mm. B1, SMI-32 immunoreactivity in the trilaminar part of the TS. The arrow indicates a TSL neuron located in the medial (M) division, which extends dendrites to the intermediate (I) division. The frame area is superimposed on the phal image in B2 and enlarged in C. Scale bar, 100 µm. B2, Axons and axon terminals originating from the MGM are distributed broadly in the GP as well as in the trilaminar part of the TS. C, High-power triple fluorescence image consisting of SMI-32 (green), Phal (red), and VGluT2 (blue) immunoreactivities. Two large boutons make contact with a dendrite of a TSL neuron. Scale bar, 10 µm. Insets, Enlargement of the circled area. The Phal-positive bouton makes direct contact with the dendrite and shows VGluT2 immunoreactivity. Scale bar, 1 µm.
Next, we examined possible sources of GABAergic terminals that densely surrounded both somata and dendrites of TSL neurons (Fig. 4F–I). Somata of TSL neurons were located in the medial division of the trilaminar part where SP-positive GABAergic terminals originating from the striatonigral direct pathway neurons densely accumulate (Miyamoto et al., 2019). However, only a small number of SP-positive boutons contacted TSL neurons (Fig. 7B). This is in sharp contrast with the finding in Enk immunohistochemistry (Figs. 7C, 8). The medial division is generally characterized by low immunoreactivity for Enk when viewed at low magnification (Fig. 8A). However, observations at higher magnification revealed that Enk-immunoreactive boutons formed dense clusters around the somata of TSL neurons (Fig. 8B). Dense accumulation of Enk-positive boutons was further observed along dendrites in both the medial and intermediate divisions of the trilaminar part (Fig. 8B–E), which could be observed even in low-magnification images (Fig. 8A, framed area). It is intriguing that Enk-positive boutons not only contacted TSL neurons but were also distributed in a wider space surrounding TSL neurons (Fig. 8B,E). This staining pattern indicates that axon terminals originating from Enk-expressing striatopallidal neurons were concentrated in a particular space around TSL neurons. Remarkably, accumulation of Enk-positive boutons was not observed around SMI-32-negative PV neurons even if they were located adjacent to TSL neurons (Fig. 8F–I, arrowhead). This suggests that TSL neurons are incorporated in a particular circuitry in which Enk-positive GABAergic axon terminals originating from indirect pathway-type medium spiny neurons (MSNs) have critical roles as major input sources.
SP- and Enk-immunoreactive boutons on TSL neurons. A, Triple-immunofluorescence image for PV (green), SP (blue), and Enk (red), showing bouton contacts on soma and dendrite. Scale bar, 10 µm. B, C, A different color set consisting of magenta (SP and Enk) and green (PV) is used. Note the sparse contacts made by SP-positive boutons (B), which are contrasted with dense contacts by Enk-positive boutons (C). D–G, Framed area is further examined in a correlative CLSM-EM quantitative analysis. D, Enk-positive boutons (magenta) indicated by arrows (1–7) make contact with a PV-positive dendrite (green). Two boutons (arrowheads, a, b) are detached from the dendrite and are excluded from the counting. Crossed arrow in D indicates PV-positive axonal bouton on the dendrite. Asterisks in D and E indicate the same dendrite. Scale bar, 5 µm. E, An EM image of the same structure that is shown in D. Enk and PV immunoreactivities are shown using DAB-nickel and DAB, respectively. Because of the omission of detergent from the incubating medium, DAB signals in EM sections cut at the deeper part of the specimen are generally weakened, and thus PV immunoreactivity in the dendrite can be seen only at the bottom left part (PV+). All seven axon terminals (arrows 1–7) corresponding to boutons in D form symmetrical synapses with the TSL dendrite (inset), whereas two boutons (arrowheads, a, b) do not make contact with the dendrite. Scale bar: D, 2 µm; (in D) D inset, 0.2 µm. F, G, Quantitative analyses of bouton (D) and synaptic (E) contacts indicate that the densities of contacts by Enk-positive elements are significantly higher than those by SP-positive elements in both the light (F) and electron (G) microscopic analyses.
Encasement of TSL neurons by Enk-positive boutons. A, Low-power image of Enk immunoreactivity. The GP is a target region of Enk-positive indirect pathway neurons; thus, it shows intense immunoreactivity for Enk. Three divisions of the trilaminar part [medial (M), intermediate (I), lateral (L)] are characterized by differential immunoreactivity for Enk with graded intensity from the medial to lateral. Thus, the M division generally shows very low immunoreactivity for Enk, but intense labeling along string-like structures goes through the medial division toward the intermediate division, as seen in a framed area. Scale bar, 100 µm. B, Double-immunofluorescence image of the framed area in A, consisting of SMI-32 (green) and Enk (red) immunoreactivities. Numerous Enk-positive boutons accumulate around TSL neurons. Note that these boutons not only densely cover SMI-32-positive somata and dendrites, but also are distributed in the surrounding area. Scale bar, 20 µm. C–E, Double-immunofluorescence image in the intermediate division of the trilaminar part. Again, numerous Enk-positive boutons not only make direct contact with the distal segments of dendrites but also distribute diffusely in surrounding tissue close to the dendrites. Scale bar, 20 µm. F–I, Accumulation of Enk-positive boutons is very selective, avoiding a SMI-32-negative PV neuron (arrowhead) located close to TSL neurons (arrows). Scale bar, 20 µm.
EM analysis
The morphological features of TSL neurons were further examined at the EM level (Figs. 9, 10). In dual-labeled immunoelectron microscopy, TSL neurons were labeled with anti-PV antibody. DAB was used as a chromogen to visualize PV immunoreactivity, whereas DAB-nickel was used as a chromogen to visualize Enk immunoreactivity. DAB-nickel provided a much more electron-dense signal than DAB as has been shown in previous studies (Acsády et al., 2000; Fukuda, 2017; Shigematsu et al., 2019).
Electron micrographs of a dendrite of a TSL neuron and surrounding structures double immunolabeled for PV and Enk, the immunoreactivities of which are visualized using DAB and DAB-nickel. A, A PV-positive dendrite receives many symmetrical synaptic contacts from both Enk-positive (asterisks) and Enk-negative (open circles) axon terminals. It also receives asymmetrical synapses (squares). Framed areas (b–d and f) are enlarged in B–D and F. B–D, Enk-positive axon terminals (asterisks) form symmetrical synapses (arrowheads) with the dendrite of the TSL neuron shown in A. Immunonegative axon terminals (open circles) also form symmetrical synapses (arrowheads). In D, a rare occasion of a spine extending from the dendrite of the TSL neuron is shown by the arrow. E1, Enlargement of the framed area in D. The spine (sp 1) receives a symmetrical synapse (arrowhead) from an axon terminal on the right side (open circle). The left-sided terminal (square) forms asymmetrical synapses (white arrowheads) with the dendritic shaft of the TSL neuron and another spine of unidentified origin (sp 2). E2, Enlargement of E1. Double arrows indicate adherens junction. F, The TSL dendrite receives asymmetrical synapses (white arrowheads). G1–G11, Serial ultrathin sections of the framed area (g) in A. Enk-positive elements (asterisks) contain numerous synaptic vesicles but do not form synapses with surrounding structures including a neighboring spine (sp), which is targeted by Enk-positive (dot) and Enk-negative (open circle) axon terminals forming symmetrical synapses (arrowheads).
Immunoelectron micrographs of a soma of a TSL neuron and surrounding structures double immunolabeled for PV and Enk in the medial division of the TS. A, A large soma is lightly stained for PV using DAB as a chromogen, whereas Enk immunoreactivity is detected in axon terminals surrounding the soma, using DAB-nickel as the second chromogen showing denser signals. The perimeter of the soma is 115 μm, indicating that this cell is a TSL neuron (Figure 4C). The large perikaryon is filled with organelles such as numerous mitochondria (m) and rough endoplasmic reticulum (rER). Axon terminals surrounding the soma include both Enk-positive (asterisks) and negative ones, part of which in framed areas are enlarged in B and C. Arrows indicate Enk-positive elements around the soma with some distance away from it. B, Enk-positive axon terminals make direct contacts with the soma, forming symmetrical synapses (arrowheads). C, A PV-positive large axon terminal (white dot) forms a symmetrical synapse (arrowhead) with the soma of TSL neuron that is also labeled for PV.
Dendrites of TSL neurons were surrounded by numerous axon terminals (Fig. 9A). Both Enk-positive and Enk-negative axon terminals densely contacted dendrites and formed symmetrical synapses (Figs. 9A–D). In addition, axon terminals forming asymmetrical synapses were also observed (Fig. 9F). On rare occasions, a spine was observed along a dendrite of a TSL neuron, receiving symmetrical synaptic input (Fig. 9E1,E2). Regarding Enk-positive boutons scattering around but some distance away from TSL neurons, some formed symmetrical synapses with a spine of unidentified origin (Fig. 9G9–G11), but others did not make synaptic contacts with any structures, even after thorough examination of serial ultrathin sections (Fig. 9G9–G11).
The large somata of TSL neurons contained many organelles, such as mitochondria and rough endoplasmic reticulum (Fig. 10A). Somata of TSL neurons were surrounded by numerous axon terminals forming both symmetrical and asymmetrical synapses. Symmetrical synaptic terminals included both Enk-positive (Fig. 10B) and Enk-negative ones. PV-positive axon terminals were occasionally detectable based on moderate labeling with DAB (Fig. 10C). Overall, ultrastructural features in both somata and dendrites of TSL neurons were similar to those of principal neurons in the globus pallidus (Kita, 1994).
The above observations at both the light and electron microscopic levels were further investigated quantitatively using a correlated CLSM-EM (Fig. 7D–G). In CLSM analysis (Fig. 7D,F), the averaged density of Enk-positive boutons on TSL dendrites was 0.60 ± 0.16/µm (n = 5), which was much larger than that of SP-positive boutons (0.01 ± 0.13/µm, n = 4; p < 0.001, Mann–Whitney U test). These data were in good correspondence with the density of synapses analyzed in serial ultrathin sections in EM (Fig. 7E,G): the density of synapses formed between Enk-positive axon terminals and a TSL dendrite per unit length was 0.60 ± 0.19/µm (n = 3), which was larger than the density of synapses between SP-positive axon terminals and a TSL dendrite (0.00 ± 0.00/µm, n = 3; p = 0.012, Mann–Whitney U test).
Visualization of medium spiny neuron axons using GAD67-CrePR knock-in mouse
To identify the origins of Enk-positive boutons on dendrites of TSL neurons, we used GAD67-CrePR knock-in mice to visualize the whole structure of labeled MSNs (Figs. 11, 12). To facilitate analysis, the number of labeled neurons was kept small by administering a minimal dose of mifepristone to the animals. Clear, intense labeling for tdTomato immunoreactivity was obtained in axons originating from MSNs that were characterized as such by the presence of massive spines along dendrites (Fig. 11B). This labeling method was combined with multiple immunofluorescence using anti-RFP, anti-PV, and anti-Enk antibodies; anti-RFP antibody provided bright, specific labeling for tdTomato. We found that MSNs located in the intermediate division of the trilaminar part gave rise to axons that approached nearby dendrites of TSL neurons, formed large boutons, and made multiple contacts with TSL dendrites (Figs. 11C1–C3,F, 12). The number of contacts between axons from a single MSN and dendrites of a single TSL neuron reached as many as 17 (Fig. 12). Importantly, Enk immunoreactivity was confirmed inside these boutons (Fig. 11D,E). Therefore, the origins of ENK-positive boutons on TSL dendrites were MSNs that were located in the vicinity of TSL dendrites in the intermediate division of the trilaminar part. Figure 12 further indicates that an MSN in the intermediate division sent axons not only to the intermediate division but also to the medial division, where labeled axon made multiple contacts with a proximal portion of a TSL dendrite. Thus, dense accumulation of ENK-positive boutons on both the proximal and distal parts of TSL dendrites can be ascribed to axons originating from MSNs located in the intermediate division of the TS. We did not follow tdTomato-labeled axons further into the GP, because we focused on the connectivity of MSN axons and TSL neurons.
Visualization of an MSN located in the intermediate division of the trilaminar part of the TS using GAD67-CrePR knock-in mouse. A, Low-magnification image of the trilaminar part, labeled for tdTaomato (red), Enk (green), and PV (blue). A cell showing intense immunoreactivity for tdTomato is located in the intermediate division (i). G, Globus pallidus; m, medial division; l, lateral division. Scale bar, 200 µm. B, Enlargement of the tdTomato-positive cell. Note the presence of massive spines along dendrites, indicating that this cell is an MSN. An axon (ax) emitting from the cell body bifurcates many times and arborizes around the cell to form local axon collaterals. Arrow indicates a point where the axon was cut at the section surface. The corresponding position in a neighboring section is shown by arrow in C1. Scale bar, 50 µm. C1–C3, Continuation of the axonal arborization originating from the cell in B. C1, C2, The axon forms many large boutons (C1) that surround and make contact with a dendrite of a TSL neuron (C2). C3, In triple-immunofluorescence image, numerous Enk-positive elements accumulate around the dendrite, a characteristic profile of TSL dendrite. Scale bar, 10 µm. D, E, Enlargement of the boutons indicated by arrows in C2, viewed from three different angles. The images clearly indicate that Enk-positive elements are contained inside the tdTomato-positive boutons. Scale bar, 1 µm. F, A reconstruction of the tdTomato-labeled MSN from serial sections (soma and dendrites, green; axon, dark blue). Dendrites of four TSL neurons (TSL 1–4) are also reconstructed, and spheres of different colors show the sites of contacts between the MSN axon and individual dendrites. The dendrite shown in C belongs to TSL1, with which the axon makes as many as 15 contacts (red spheres). Scale bar, 100 µm.
A reconstruction of a tdTomato-labeled MSN that is located in the intermediate division and is different from the cell demonstrated in Figure 11. Soma and dendrites are shown in orange, and axon is shown in dark blue. Axon makes as many as 17 contacts (green spheres) with the dendrite of TSL2, which is the same dendrite shown in Figure 11F. Note that an axon of this MSN also targets the proximal dendrite of TSL 5 in the medial division of the trilaminar part, where the axon makes 13 contacts with the dendrite. Scale bar, 100 µm.
Then, we examined the possibility that these MSNs in the intermediate division in turn received inputs from the A1 and, if so, how dense the connection was (Fig. 13). We injected an anterograde tracer phal into the A1. Labeled axons were distributed mainly in the intermediate division of the trilaminar part (Fig. 13A,C), which was consistent with previous observations (Miyamoto et al., 2018; Li et al., 2021). Of 19 tdTomato-labeled MSNs that were located in the intermediate division, all 19 cells were contacted by phal-positive boutons (Fig. 13E–G). Moreover, the number of contacts a single MSN made with labeled boutons was 5.6 ± 3.2 (mean ± SD; range, 3–15; n = 19 cells). Dendrites of labeled MSNs in the intermediate division were mostly confined within the intermediate division (Figs. 11, 12). Together, we obtained direct evidence for the serial connectivity originating from the A1 through MSNs toward TSL neurons (Figs. 11–13), in which multiple bouton contacts mediated the two connections.
Anterograde tracer study applied to GAD67-CrePR knock-in mouse. A–H, Triple-immunofluorescence images consist of tdTomato (red), phal (green), and PV (blue) immunoreactivities (A, E), which are shown separately in B and F, C and G, and D and H, respectively. A–D, Insets, Injection site centered into the A1. m, Medial division; i, intermediate division; l, lateral division. Scale bars: (in A) A–D, 200 µm; (in A, inset) A–D, Insets, 400 µm. E–H, Enlargement of the framed area in A. Axons labeled for phal make direct contacts (arrowheads) with a dendrite of an MSN that has many spines. Scale bar, 10 µm.
Retrograde labeling
Experiments using a retrograde tracer, FG, were performed to explore projections from TSL neurons. Injections of retrograde tracers into the substantia nigra, entopeduncular nucleus, or ventral lateral nucleus of the thalamus did not result in labeling of TSL neurons, although somata in several other brain regions, including the striatum and subthalamic nucleus, were labeled. In contrast, when the rostral ZI and the VM were injected, multiple somata of TSL neurons were labeled (Fig. 14). In both ZI and VM injections, somata in the globus pallidus near the medial division of the trilaminar part of the TS were also labeled in addition to TSL neurons.
Retrograde tracer studies. A, D, E, F, After the injection of FG into the rostral part of the ZI (D), sections were double immunostained for SMI-32 (green) and FG (magenta; A), which are separately shown in E and F. Retrograde labeling was observed in a soma of TSL neuron (arrow) as well as several neurons in the GP and surrounding area (arrowheads). B, C, Labeling in the TSL neuron is enlarged and shown separately for SMI-32 (B) and FG (C). G, J, K, L, After the injection of FG into the VM of the thalamus (VM; J), sections were double immunostained for SMI-32 (green) and Fluoro-Gold (magenta; G), which are separately shown in K and L. Retrograde labeling was observed in a soma of TSL neuron (arrow). H, I, Labeling in TSL neuron is enlarged and shown separately for SMI-32 (H) and FG (I). Scale bars: A, E–G, K, L, 100 µm; C, I, 20 µm; in D, J, 500 µm.
Discussion
The present study demonstrated a new type of neuron in the striatum of the mouse. The morphological features of this neuron are summarized in a schematic drawing (Fig. 15). These neurons were termed TSL neurons based on their location and large size in the striatum. TSL neurons have many unique features, including long, laterally oriented dendrites reaching the region targeted by the A1. They are GABAergic neurons and receive Enk-positive GABAergic axon terminals heavily on both somata and dendrites. These terminals originated from nearby MSNs in the intermediate division of the TS, where virtually all MSNs received inputs from the A1. TSL neurons also receive afferents from the MGM. The projection targets of the TSL neurons were the rostral ZI and VM of the thalamus. Although TSL neurons share many morphological features with GP neurons, the strategic location of their skewed dendrites extending toward the intermediate division of the TS will facilitate immediate responses to the A1 inputs through the activation of nearby MSNs, leading to faster disinhibitory actions on targets than distant GP neurons that receive striatal inputs after axonal conduction delay.
Schematic drawing of the disinhibitory pathway in which TSL neurons are incorporated. TSL neurons, shown in green in the medial division (M) of the trilaminar part of the TS, extends dendrites toward the intermediate division (I), where the TSL neuron receives dense GABAergic synaptic inputs from enkephalin-expressing MSNs. Axons from the A1 make direct contact with MSNs, but they rarely contact dendrites of TSL neurons. Consequently, when auditory signals activate neurons in the A1, GABAergic TSL neurons will be inactivated by GABAergic inputs from medium spiny neurons, leading to disinhibitory responses in the target areas of TSL neurons including the VM nucleus of the thalamus and rostral part of ZI. Neurons in the VM send axons diffusely in layer 1 of the neocortex (Herkenham, 1979; Kuramoto et al., 2015). TSL neurons also receive glutamatergic inputs from the medial division of the medial geniculate nucleus (MGM). L, Lateral division.
Technical considerations
The antibody SMI-32 labels neurofilament proteins and has been used as a clear marker to detect particular types of neurons in various brain regions. An advantage of using SMI-32 in many studies is apparently based on the fact that it visualizes dendrites explicitly in a Golgi-like fashion, because neurofilaments are cytoskeletal protein enriched in dendrites of some but not all neuronal populations. However, it should be taken into consideration that SMI-32 does not label all dendrites, such as thin processes and dendritic spines, if present, because neurofilaments are not localized diffusely in the cytoplasm. All TSL neurons colocalized PV, which is a cytosolic protein. Thus, morphological details of the TSL neurons could be observed in PV immunofluorescence images, confirming a lack of spines on most of the dendrites and the presence of numerous bouton contacts on TSL neurons. Furthermore, we performed EM analysis to confirm both the sparseness of dendritic spines and actual synaptic contacts between diverse axon terminals and TSL neurons. Thus, these two approaches yielded consistent results.
Unexpectedly, the examination of synapses by EM analysis showed that some Enk-positive boutons surrounding TSL neurons (Figs. 8, 9) did not have postsynaptic targets, although these boutons contained numerous synaptic vesicles (Fig. 9G1–G11). This may suggest extrasynaptic actions of GABA and/or Enk or so-called volume transmission (Zoli et al., 1999) of auditory signals that might need to continue after the cessation of stimuli.
The large type neurons in the striatum
According to previous studies, there are at least three types of large neurons in the striatum (Bolam et al., 1981; Chang and Kitai, 1982; Penny et al., 1988). The first are cholinergic neurons (Satoh et al., 1983; Bolam et al., 1984a,b), which are generally called giant aspiny neurons. The second are characterized by the presence of many spines on their somata (Chang and Kitai, 1982), and the third are large, striatonigral type 2 neurons (Bolam et al., 1981, 1985). TSL neurons defined in the present study differ from these types. TSL neurons did not colocalize ChAT or have somatic spines, and the substantia nigra was not a target. Moreover, the localization of somata in the medial division of the TS was also a distinctive feature of TSL neurons; striatonigral type 2 neurons were found in the ventral part of the rostral striatum at the level of the anterior commissure (Bolam et al., 1985) as well as in the nucleus accumbens (Penny et al., 1988). However, striatonigral type 2 neurons and TSL neurons have many features in common: both are PV positive (Bennett and Bolam, 1994) and are ensheathed by numerous Enk-immunoreactive (Penny, 1988) and GAD-immunoreactive boutons (Bolam et al., 1985), suggesting similarities in functional aspects; future studies will be needed to explore unknown striatal circuitry.
Connectivity and functional implications
The most distinctive feature of TSL neurons was the encasement of soma and dendrites with GAD-positive boutons. EM analysis confirmed the formation of symmetrical synapses at these contact sites. TSL neurons extend dendrites toward the intermediate division of the trilaminar part and receive dense synaptic inputs from GABAergic axon terminals. The intermediate division receives inputs from the A1 (Fig. 13; Miyamoto et al., 2018), but VGluT1-positive cortical axons rarely contact TSL neurons. This synaptic organization strongly suggests that glutamatergic excitatory inputs from the A1 have an effect on TSL neurons through activation of MSNs that in turn form GABAergic synapses densely with TSL neurons. In fact, we provided direct evidence for these connections (Figs. 11–13). Remarkably, axons of Enk-expressing MSNs made numerous contacts with TSL dendrites in the vicinity. A recent behavioral study using an optogenetic approach showed that the activities of both TS-targeting A1 neurons and neurons in the intermediate division are critical for sound-evoked defense behaviors (Li et al., 2021). In this situation, TSL neurons will be inactivated through the activation of MSNs, leading to disinhibition in target areas of TSL neurons.
This disinihibitory circuit in which TSL neurons are embedded is similar to the canonical indirect pathway circuit in striatopallidal connections: cortical inputs activate MSNs that inactivate pallidal neurons. Moreover, the present results revealed many morphological similarities between GP neurons and TSL neurons except for the soma size. Both are positive for PV and SMI-32, have common ultrastructural features, and receive numerous GAD-positive and Enk-positive boutons on soma and dendrites. Thus, a question may arise of whether TSL neurons might be just ordinary GP neurons located around the GP/striatum border. However, according to the general principle in morphological science, a characteristic structure usually reflects a specific function executed by that structure. The most obvious difference between the two populations lies in dendritic morphology: TSL neurons elongate dendrites selectively to the place where they can receive inputs directly from immediately located MSNs, whereas GP neurons confine their dendrites within the nucleus, receiving striatal inputs from axons traversing a distance of ∼500 µm (Fig. 1A). Based on known conduction velocities of 0.4–0.8 m/s for the rat striatopallidal axons (Park et al., 1982), it will take ∼0.6–1.3 ms for striatal signals to reach the GP, which is in a range of synaptic delay (Nicholls et al., 2000). Thus, sound-driven cortical activities can evoke the fastest disinhibitory actions through TSL neurons, apart from the issue of whether TSL neurons belong to the GP or striatum. Moreover, both the coverage of the A1-targeted area by the convergence of TSL dendrites (Fig. 1G) and a large number of bouton contacts between MSNs and TSL dendrites support the functional significance of the characteristic structures TSL neurons form inside the TS.
The present study described two regions that were targeted by TSL neurons: the VM nucleus of the thalamus and rostral ZI. The VM nucleus belongs to the nonspecific nuclei of the thalamus, sending axons diffusely into layer 1 of multiple neocortical areas (Herkenham, 1979; Kuramoto et al., 2015). Thus, if excitatory inputs from the A1 to the TS have prompt disinhibitory effects on the VM nucleus, it follows that cortical neurons are broadly activated to keep animals awake (Honjoh et al., 2018), which is the brain state required for defensive and survival behaviors.
The present study also revealed that TSL neurons received direct inputs from the MGM. This is consistent with the results of a classic study showing that neurons in the MGM target the ventral part of the caudal striatum as well as the lateral amygdala, with only slight innervation to the cerebral cortex (LeDoux et al., 1985). In contrast to the ventral division of the medial geniculate nucleus that relays major sound information to the A1, the MGM does not have tonotopy and is critical in fear conditioning. However, lesion experiments suggest differences between the amygdala and caudal striatum. Lesions in the amygdala abolish fear conditioning, whereas these effects were not observed after lesions in the caudal striatum (LeDoux et al., 1990). In the present study, labeling after injection into the MGM was detected in VGluT2-positive boutons on dendrites of TSL neurons, indicating that inputs from the MGM can elicit excitatory responses in TSL neurons. Thus, firing of MGM neurons will result in suppression of the target areas of TSL neurons, a direction opposite to A1-derived disinhibitory actions. This raises the possibility of integration through converging inputs to TSL neurons, which might be reflected in behavioral changes.
Footnotes
This work was supported by Japan Society for the Promotion of Science KAKENHI Grants 16J02300 and 19K16262 to Y.M., and Grants 16KT0174 and 18H02529 to T.F.
The authors declare no competing financial interests.
- Corresponding should be addressed to Takaichi Fukuda at tfukuda{at}kumamoto-u.ac.jp
