Abstract
All pathways targeting the thalamus terminate directly onto the thalamic projection cells. As these cells lack local excitatory interconnections, their computations are fundamentally defined by the type and local convergence patterns of the extrinsic inputs. These two key variables, however, remain poorly defined for the “higher-order relay” (HO) nuclei that constitute most of the thalamus in large-brained mammals, including humans. Here, we systematically analyzed the input landscape of a representative HO nucleus of the mouse thalamus, the posterior nucleus (Po). We examined in adult male and female mice the neuropil distribution of terminals immunopositive for markers of excitatory or inhibitory neurotransmission, mapped input sources across the brain and spinal cord and compared the intranuclear distribution and varicosity size of axons originated from each input source. Our findings reveal a complex landscape of partly overlapping input-specific microdomains. Cortical layer (L)5 afferents from somatosensory and motor areas predominate in central and ventral Po but are relatively less abundant in dorsal and lateral portions of the nucleus. Excitatory inputs from the trigeminal complex, dorsal column nuclei (DCN), spinal cord and superior colliculus as well as inhibitory terminals from anterior pretectal nucleus and zona incerta (ZI) are each abundant in specific Po regions and absent from others. Cortical L6 and reticular thalamic nucleus terminals are evenly distributed across Po. Integration of specific input motifs by particular cell subpopulations may be commonplace within HO nuclei and favor the emergence of multiple, functionally diverse input-output subnetworks.
SIGNIFICANCE STATEMENT Because thalamic projection neurons lack local interconnections, their output is essentially determined by the kind and convergence of the long-range inputs that they receive. Fragmentary evidence suggests that these parameters may vary within the “higher-order relay” (HO) nuclei that constitute much of the thalamus, but such variation has not been systematically analyzed. Here, we mapped the origin and local convergence of all the extrinsic inputs reaching the posterior nucleus (Po), a typical HO nucleus of the mouse thalamus by combining multiple neuropil labeling and axon tracing methods. We report a complex mosaic of partly overlapping input-specific domains within Po. Integration of different input motifs by specific cell subpopulations in HO nuclei may favor the emergence of multiple, computationally specialized thalamocortical subnetworks.
Introduction
The thalamus receives direct connections from an astonishing variety of systems, from the cerebral cortex and retina to the caudal spinal cord. Remarkably, such input pathways terminate directly onto the same projection cells whose axons carry the thalamus output signals. The projection cells lack local recurrent excitatory connectivity; thus, while inhibitory large-scale modulation between nuclei may occur (Crabtree, 2018), thalamic computations are essentially determined by the inputs that each cell receives. Delineating input “motifs” is thus essential to understanding thalamus coding and its overall role in cortical function (Acsády, 2017, 2022; Halassa and Sherman, 2019).
Input motifs in the so-called “first-order relay” (FO; Sherman and Guillery, 1996) nuclei are relatively simple and consistent across each nucleus. Neurons in FO nuclei invariably receive focal, topographic and synaptically powerful glutamatergic synapses from a single source (a sensory or cerebellar pathway); transmission of this “driver” signal is modulated by abundant but weaker synapses from cortical layer (L)6 (glutamatergic) and thalamic reticular nucleus (TRN; GABAergic; Pinault, 2004; Reichova and Sherman, 2004).
In contrast, there is evidence that input motifs may be highly diverse and region-specific in the remainder of the thalamic nuclei, referred to as “higher-order relay” (HO) nuclei (Sherman and Guillery, 1996). As in the FO nuclei, neurons in HO nuclei receive modulatory L6 and TRN inputs but, in addition, they receive many or all of their “driver” inputs from cortical L5b cells (Reichova and Sherman, 2004; Mease et al., 2016). These L5b axon terminals may converge onto the same thalamic cells from separate cortical areas (Sampathkumar et al., 2021). The L5b inputs can also converge with one or more synaptically powerful inputs from subcortical sensory and/or motor structures (Groenewegen et al., 1990; Parent and Hazrati, 1995; Stepniewska et al., 2000; Bokor et al., 2005; Rovó et al., 2012; Groh et al., 2014; Bickford et al., 2015; Timbie et al., 2020). Differences in the strength/structure of the various input synapses as well as in the neurotransmitters they release (excitatory/inhibitory) may further expand the range of variation among motifs (Halassa and Acsády, 2016; Acsády, 2017, 2022). As a result of their diverse input motifs, thalamic cells perform input-output computations in a region-specific manner (Trageser and Keller, 2004; Lavallée et al., 2005; Groh et al., 2014; Ahissar and Oram, 2015).
Delineating the “input landscapes” may thus be key to understanding the functional organization of the HO nuclei. However, as thalamic input systems have been investigated largely in isolation, a cohesive picture is lacking, even for the best studied nuclei. Here, we set out to map the sources, intranuclear distribution and axon varicosity sizes (as a proxy for synaptic strength) of all inputs reaching the posterior nucleus (Po), a representative and extensively studied HO nucleus of the rodent thalamus.
Rodent Po has been reported to receive L5b input from the primary somatosensory area (S1; Bourassa et al., 1995; Veinante et al., 2000a; Grant et al., 2012; Sumser et al., 2017; Hayashi et al., 2021), secondary somatosensory area (S2; Liao et al., 2010), and the motor cortex (Rouiller et al., 1991; Prasad et al., 2020). These L5b inputs drive Po cell firing with high efficacy (Reichova and Sherman, 2004). In addition, Po receives spinal trigeminal nucleus (Jacquin et al., 1989; Pierret et al., 2000; Veinante et al., 2000b), dorsal column nuclei (DCN; Lund and Webster, 1967), spinal cord (Iwata et al., 1992; Gauriau and Bernard, 2004), and superior colliculus inputs (Gharaei et al., 2020) which are also able, under certain conditions, to drive Po cell spiking. Besides, many Po cells receive inputs from the zona incerta (ZI; Power et al., 1999; Barthó et al., 2002) and the anterior pretectal nucleus (APT; Bokor et al., 2005), which provide strong, focal, and temporally precise inhibition (Halassa and Acsády, 2016).
We show that the Po neuropil is a heterogeneous tridimensional mosaic of partly overlapping input domains. Cells in different parts of Po may thus perform widely different computations.
Materials and Methods
Animals
Experiments were performed on adult (60–120 d old, 25–35 g body weight) wild-type C57BL/6 mice of both sexes. Animals were bred in the Animal Facilities of the School of Medicine of the Autónoma de Madrid University. All procedures involving animals were conducted under protocols approved by the university Ethics Committee and the competent Regional Government agency (PROEX175/16), in accordance with the European Community Council Directive 2010/63/UE. Animals were housed under standard colony conditions with food and water ad libitum under a 12/12 h light/dark cycle. Efforts were made to minimize the number of animals required. In total, twenty-six mice were used for retrograde tracing experiments, 52 were used for anterograde BDA axon labeling and seven further mice were used for immunohistochemistry. The numbers of valid cases in each of the various experiments were too small to analyze potential sex differences, but each set of experiments contains data from animals of both sexes.
In addition, the brains from two mice from the L5 Cre-expressing Rbp4-Cre:Ai14 strain (Gong et al., 2007) were used to examine the distribution of cortical L5 pyramidal cell axons within Po. Procedures to obtain these two brains were performed in the animal facilities of the University of Oxford (Unite Kingdom) under Animals (Scientific Procedures) Act 1986 project license with local ethical approval by the Central Committee on Animal Care and Ethical Review (ACER) and the Animal Welfare and Ethical Review Body (AWERB) at the University of Oxford. Tg(Rbp4cre)KL100Gsat/Mmucd (Rbp4-Cre; Jackson Laboratories) mice were crossed with B6;129S6-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J (Ai14) to constitutively drive fluorescent protein expression in cortical L5 pyramidal neurons (Grant et al., 2016; Hoerder-Suabedissen et al., 2018, 2019; Hayashi et al., 2021).
Anesthetic procedures
In the tracing experiments with postinjection survival, anesthesia was induced with an intraperitoneal injection of ketamine (0.075 mg/g body weight) + xylazine (0.02 mg/g body weight), and subsequently maintained throughout the surgical procedure with isoflurane in oxygen (0.5 –1%). At the end of the surgery, isoflurane was interrupted, and animals recovered promptly. Ibuprofen (120 mg/l) was added to the drinking water to ensure analgesia during the postoperative period.
At the time of killing, animals were overdosed with sodium pentobarbital (0.09 mg/g body weight, i.p.).
Retrograde axonal tracer experiments
Retrograde tracers were microinjected in the thalamus to map across the brain and spinal cord the neuronal populations targeting specific thalamic nuclei. Animals were positioned in a stereotaxic apparatus (David Kopf Instruments) and placed on a water-heated pad at 37°C. The scalp was sectioned at the midline and retracted, and small craniotomies were opened in the parietal bones. Two different retrograde fluorescent tracers, one in each hemisphere, were injected through borosilicate glass micropipettes (WPI; outer tip diameter: 10–15 µm), positioned under stereotaxic guidance either in Po (bregma −1.8 mm posterior, 1.2 mm lateral, and 3.2 mm ventral; Paxinos and Franklin, 2012) or in ventral posteromedial nucleus (VPM; −1.7, 1.7, and 3.3 mm, respectively). Fast Blue (FB; diamidino compound 253/50, Polysciences Europe GmbH; 0.1% w/v in 0.1 m cacodylate buffer; pH 7.4) or Fluoro-Gold (FG; hydroxystilbamidine methanesulfonate H-22845, Invitrogen Fluorochrome LLC; 0.75% w/v in 0.1 m phosphate buffer; PB) were used as tracers. FB was pressure-injected (0.01–0.1 µl) using a precision electro-valve system (Picospritzer II, Parker Hannifin) while FG was microiontophoresized with a Midgard Precision Current Source catalog #51590 (Stoelting) by applying 4- to 5-µA positive continuous current cycles (7 s on/off) for 3–5 min. Pipettes were left in place for 10 min after the injection, before being removed. Finally, muscle and skin were disinfected with povidone iodine and sutured, and the animals were allowed to recover and returned to their cage.
Anterograde axonal transport experiments
To label axon terminals from the neuronal populations in various brain and spinal cord structures identified by the retrograde tracing experiments, lysine-fixable 10 kDa biotinylated dextran amine (BDA; Invitrogen; 3% w/v solution in 0.01 m PB, pH 7.4) was selectively iontophoretically injected in different regions of the forebrain, brainstem or spinal cord. A positive current was applied using a Dual Current 260 source (World Precision Instruments, WPI) or Midgard Current Source (Stoelting). Injection parameters are summarized in Table 1.
Injection parameters applied to selectively inject BDA in brain regions that innervate Po
Forebrain regions were targeted following the stereotaxic coordinates of Paxinos and Franklin (2012) atlas. To target the Gracilis (Gr) and Cuneatus (Cu) nuclei, the head was secured in ventroflexed position. After a skin incision, the muscles of the back of the neck were dissected laterally from the midline, and the atlantooccipital membrane was opened to expose the dorsum of the lower medulla. The micropipette was inserted in Gr or Cu (between 0.2 and 0.5 mm caudal to the obex). To inject BDA in the spinal cord, a midline skin incision was made over the upper thoracic vertebrae. The paravertebral muscles were separated and retracted, the ligamenta flava incised and the micropipette tip was lowered into the spinal cord. After injections, micropipettes were left in place for 10 min before removal and wound closure.
Tissue fixation and histologic procedures
Animals were allowed to survive for 7 d and were then perfused transcardially with 30 ml of saline, followed by 100 ml of 4% paraformaldehyde (PFA; diluted in 0.1 m PB, pH 7.4). Brains were then removed from the skull and postfixed overnight at 4°C in the same solution, and cryoprotected by embedding in 30% sucrose in 0.1 m PB, at 4°C, for 48 h. In retrograde transport experiments, the whole spinal cord was extracted and processed likewise.
In the retrograde fluorescent labeling experiments, three parallel series of 50-µm-thick coronal sections of the forebrain, brainstem and spinal cord were obtained on a freezing microtome (SM 2400; Leica). One series was mounted onto gelatin-coated glass slides, air dried, dehydrated in graded ethanol, defatted in xylene and coverslipped with DePex (Serva). These sections were analyzed under an epifluorescence light microscope (Eclipse 600; Nikon) for retrograde labeling analysis. The second series was histochemically stained for cytochrome oxidase activity (CyO; Wong-Riley et al., 1978) to help cytoarchitectonic localization of the labeling. The third series was stored at −20°C as a backup.
In the BDA axonal labeling experiments, brains were freeze-sectioned in the coronal plane at 60 µm, and sections were collected in two parallel series. In the first, after peroxidase activity blocking by incubation in H2O2 0.66% (w/v) in 0.1 m PB for 15 min, sections were incubated for 2 h in avidin-biotin-peroxidase (1:100; Vectastain Elite, Vector Laboratories) diluted in 0.1 m PB. After washing, peroxidase was visualized using the glucose oxidase-3-3'diaminobenzidine (DAB; Sigma-Aldrich) nickel sulfate-enhanced method (Shu et al., 1988). These sections were then counterstained with CyO histochemistry (Wong-Riley et al., 1978) for cytoarchitectonic localization of the labeling. Sections were finally mounted and coverslipped as above. A second series was used either for immunolabeling as explained below or were kept in antifreeze solution at −20°C as a backup.
Immunolabeling
All antibodies used in this study are commercially available and their host species, manufacturer, and dilutions are indicated in Table 2. Immunohistochemistry against vesicular glutamate transporter type 1 (vGLUT1), type 2 (vGLUT2), vesicular GABA transporter (vGAT) or glutamate decarboxylase 65 and 67 (GAD65/67) was performed in some brains. Following peroxidase activity blockage, sections were serially incubated in: (1) 2% Triton X-100 (Sigma-Aldrich) + 5% normal goat serum (NGS) + 1% bovine serum albumin (BSA) in 0.1 m PBS at room temperature (RT) for 2 h; (2) guinea pig anti-vGLUT1, guinea pig anti-vGLUT2, rabbit anti-vGAT or rabbit anti-GAD65/67 polyclonal antibodies (Table 2), 2% Triton X-100, 5% NGS and 1% BSA in 0.1 m PBS with at 4°C for 48 h; (3) biotinylated goat anti-guinea pig or goat anti-rabbit IgG, 2% Triton X-100, 5% NGS in 0.1 m PBS at RT for 2 h; (4) in ABC Elite (Vector Laboratories) 1:100 in 0.1 m PBS containing 2% Triton X-100 at RT for 2 h. Multiple PBS rinses were intercalated between the above solutions. Finally, peroxidase activity was made visible using the glucose oxidase-diaminobenzidine method (Shu et al., 1988). In experiments involving immunolabeling against GABA, sections were pretreated for antigen retrieval in sodium borohydride 1% (w/v) for 30 min, RT.
List of antibodies used in this study
For some experiments, simple or double immunofluorescence labeling was performed, in different combinations. The following antibodies were applied (Table 2): guinea pig anti-vGLUT2, rabbit anti-biotin and two different rabbit anti-GAD65/67. Besides, in tissue sections from the Rbp4-Cre:Ai14 brains, a rabbit anti-red fluorescent protein (RFP) antiserum that also reacts against tdTomato was used.
In experiments involving immunolabeling against GAD65/67, sections were pretreated for antigen retrieval in sodium borohydride 1% (w/v) for 30 min, RT. In all experiments, sections were incubated, free-floating, in 2% Triton X-100 and 5% NGS in 0.1 m PB, 2 h, RT, rinsed in PB, and subsequently incubated in the preincubation solution plus the primary antibodies listed above (72 h, 4°C). Following multiple rinses, sections were then incubated in the corresponding secondary AlexaFluor-conjugated antibody (Table 2) in 0.1 m PB containing 2% Triton X-100 and 5% NGS, 2 h, RT.
After rinsing in 0.1 m PB, sections were finally mounted onto gelatin-coated glass slides, air dried, dehydrated graded ethanol, defatted in xylene, and coverslipped with DePex (Serva).
Retrograde labeling analysis
The distribution of the cells revealed by retrograde tracer labeling was examined under an epifluorescence microscope Eclipse E600 (Nikon) and mapped on JPEG images of the serial sections acquired with a Nikon DMA1200 digital camera and a motorized microscope stage (Proscan, Prior Scientific Instruments) controlled by the Nis-Elements BR 3.2 software (Nikon). Images were assembled into a single panoramic view using the Grab Large Image package of Nis-Elements BR 3.2 software.
First, we analyzed the location of the tracer deposits in the thalamus; 40× images of all sections containing the tracer deposit (spaced 150 µm) were examined to delineate the deposit consistently. For each deposit, we outlined a core zone of solid, heavy fluorescence where no individual cell profiles were evident. Results suggest tracer uptake occurred mostly within this zone. Nevertheless, for a conservative interpretation of the labeling results, we also delineated and included in our analysis the peripheral halo of fainter fluorescence where individual labeled cell profiles were visible. Retrograde tracer deposits were redrawn using the vectorial design software Canvas X (ACD Systems).
Retrogradely labeled cell somata were subsequently mapped across the brain and spinal cord. One out of six coronal tissue sections (∼200 sections; 300-µm gap) was examined in each valid case. Autofluorescence images of the tissue sections were acquired through a 4× objective and the Nikon B2A filter cube (ex. 450–490) for anatomic reference. Regions containing labeled cells were imaged using a 10× objective and the Nikon UV2A filter cube (ex. 330–380).
Subsequently, retrogradely labeled somata were counted using Canvas X software tools. For each deposit, the fraction represented by the cells labeled in a particular structure over the total of labeled neurons was calculated. It should be emphasized that, given the many variables involved in the effectivity of retrograde transport labeling (Schofield et al., 2007), labeled cell counts were intended here only as an indirect approximation of the relative anatomic weight of the various sources of input to the thalamic nuclei under investigation.
As a visual aid for comparison of the spatial distribution patterns of corticothalamic neurons in the tangential dimension of the cerebral cortex, labeled cell somata were plotted on standard flat cortex maps registered to stereotaxic coronal levels. We built this map from Paxinos and Franklin (2012) coronal diagrams. For each section, arc segments between cytoarchitectonic area borders measured on the section diagrams were plotted along a line. The lines were then stacked in order and aligned along the dorsal midline of the hemisphere (Fig. 7f–h). Additional details were derived from CyO-stained brain sections. In the present study, the position in each section of the labeled corticothalamic cells was radially projected on the pial surface and plotted along its matching level line.
Anterograde BDA labeling analysis
For BDA-labeled experiments, sections were systematically examined under brightfield optics. Sections containing the BDA deposit were imaged and nuclear boundaries were delineated following cytoarchitectonic references provided by CyO counterstain.
BDA-labeled axons terminals in the thalamus were analyzed and imaged using 10–40× objectives. Axon arborizations were labeled in Golgi-like detail and showed frequent varicose swellings. Since axonal varicosities have been shown to contain presynaptic organelles and their overall size correlates with the number and/or strength of synapses they establish (Rovó et al., 2012; Groh et al., 2014; Rodriguez-Moreno et al., 2020), we decided to measure and compare labeled axon varicosity sizes. For this purpose, varicosities were identified as such when their diameter was at least twice that of the adjacent axonal segments. As a proxy of bouton size measured on two-dimensional light microscope images, the major perimeter of each varicosity was focused in the z-axis at 100× and delineated, and its cross-sectional (maximal projection) area measured using the Polyline and Polygon tools of the Nis-Elements BR 3.2 software (Nikon). Varicosities with cross-section area below the microscope resolution limit (<0.2 μm2) were not included in the analysis. For each glutamatergic-labeled projection system, 100 randomly selected varicosities were measured, and 200 in the case of inhibitory system (total of 2200 varicosities for all cases). Varicosity size data for corticothalamic axons (Hoerder-Suabedissen et al., 2018) are included here for direct comparison with other sources of input to the thalamus.
The location and distribution of the BDA-labeled varicosities in Po were plotted using a motorized Nikon Eclipse 80i microscope with 40–60× brightfield objectives, Neurolucida controller, camera and software (v. 2020; MBF Bioscience). The CyO counterstain was used for thalamic nuclei delineation. In each section, all BDA-labeled varicosities were plotted using markers of Neurolucida software.
Confocal microscope analysis
Immunofluorescence analyses of axon terminals were carried using a Spectral Leica TCS SP5 confocal microscope by sequentially applying argon (488 nm), diode-pumped solid-state (561 nm) and/or helium–neon (633 nm) laser lines to ensure complete channel separation. Regions of interest were imaged using 10× objective. In addition, images of BDA-labeled axons were taken also using 20× (plus 3× digital zoom) and 40× objectives. For each region of interest, 10 image stacks were obtained moving the sample in the z-axis. Both image stacks and maximal projections were analyzed in separate and merged channels.
Experimental design and statistical analysis
Very small deposits of retrograde tracers often fail to label afferent pathways in consistent fashion. For this reason, out of a total of 52 retrograde tracer injections, we selected only eight injection experiments for detailed analysis (six injections in Po plus two in VPM) whose deposits were confined to the nuclei of interest and had produced robust retrograde labeling in distant brain regions. The retrograde labeling was analyzed separately for each case, as a n = 1 experiment class.
From the BDA-labeling experiments, we analyzed only those in which the tracer deposit was confined to the intended target structure (46 injections). Comparison of varicosity mean cross-sectional areas was performed using one-way ANOVA plus T3 Dunnet's multiple comparisons as a post hoc test to identify significant interactions. Two-sample Kolmogorov–Smirnov test (K-S) was used to compare size distributions of varicosities between different structures. Statistical analysis was computed using GraphPad Prism 5 software or SPSS Statistics software (v. 24; IBM). Data are presented as the mean ± SEM. The threshold level of significance was set at *p < 0.05, **p < 0.01, and ***p < 0.001.
Results
Distribution within Po of vesicular glutamate transporter-specific immunolabeled axon terminals
To compare the general distribution within Po of glutamatergic axon terminals from the cerebral cortex with those of subcortical origin, we immunostained adjacent tissue sections against either vGLUT1 or vGLUT2 (Fig. 1). The vGLUT1 protein is a well-established marker of the presynaptic vesicle pools in corticothalamic terminals. In contrast, vGLUT2 protein is present in glutamatergic brainstem and spinal presynaptic terminals, but absent from corticothalamic terminals (Fujiyama et al., 2001; Oliveira et al., 2003; Bopp et al., 2017). Although most thalamic projection neurons express vGLUT2 mRNA, their protein product is selectively directed to their axon terminals, outside the dorsal thalamus (Fremeau et al., 2001; Fujiyama et al., 2001).
Distribution within Po of vesicular glutamate transporter-specific immunolabeled axon terminals. a, b, Microphotographs showing in different coronal sections of Po the distribution of terminals marked by immunohistochemistry against the glutamate vesicular transporter type 1 (vGLUT1). Dorsal is to the top and medial to the left. High-magnification details are shown in a1. c–f, Microphotographs showing in different coronal sections of Po the distribution of terminals marked by immunohistochemistry against the glutamate vesicular transporter type 2 (vGLUT2). High-magnification details showing larger vGLUT2 terminals in dorsal domains (d1) and in the border with the ventral posteromedial nucleus (VPM; d2), smaller vGLUT2 terminals in the lateral portion of Po (d3) and domains with almost absence of these terminals in the ventral portion (e1). Arrowheads indicate focal clusters of larger immunopositive puncta. PC: Paracentral nucleus; PF: Parafascicular nucleus. Bregma level in mm is indicated in the upper right corner. Scale bars: 250 µm (a–f) and 25 µm (a1, d1–d3, e1).
Immunostaining for vGLUT1 produced heavy punctate neuropil labeling across Po (Fig. 1a,b). These puncta delineated, as empty spaces, cell body contours (Fig. 1a1). On the other hand, immunolabeling for vGLUT2 was markedly less dense overall than in the adjacent nuclei and was distributed in a variegated pattern. An extensive ventral and central portion of Po contained only very small, spherical, and weakly vGLUT2 labeled puncta (Fig. 1e1). In contrast, an elongated patch in the dorsal portion of the nucleus, and smaller clusters along the border with the VPM contained heavily stained and larger puncta (arrowheads in Fig. 1c–f), similar to those present in massive numbers in the adjacent VPM (Fig. 1d2). The distribution and shape of these large vGLUT2 puncta-rich Po domains were consistent across animals.
Distribution of L5b corticothalamic axon terminals within Po
The vGLUT2 immunolabeling pattern revealed that subcortical excitatory afferent systems to Po concentrate predominantly in some nucleus subdomains yet are scant in others. In contrast, as both L5 and L6 corticothalamic terminals contain vGLUT1, the heavy immunolabeling precluded a reliable identification of L5 corticothalamic terminals. To obtain a global and unobstructed view of the L5 axon terminals, we examined the thalamus of transgenic mice Rbp4-Cre;Ai14. In these animals, abundant L5b neurons in the dorsolateral cerebral hemisphere constitutively express high levels of the fluorescent protein tdTomato (Grant et al., 2016; Hoerder-Suabedissen et al., 2019; Hayashi et al., 2021; Fig. 2). The thalamus of Rbp4-Cre;Ai14 animals showed a profuse plexus of tdTomato-labeled varicose terminal branches and boutons in Po. However, these branches were noticeably less dense or virtually absent in some small dorsal and caudal Po domains (Fig. 2b,c,c2). In addition to Po, the lateral Po received abundant labeled arborizations; in sharp contrast, fluorescent axonal arborizations were virtually absent from the adjacent VPM and intralaminar nuclei. Besides, few isolated fluorescent neuron bodies are present in the caudomedial portion of Po in these animals. These highly specialized structure of Rbp4-Cre+ large boutons in Po were recently reconstructed using serial electron microscopy (EM; Hayashi et al., 2021) since a method to match fluorescence signal and EM reconstructions has been developed (Maclachlan et al., 2018).
Distribution within Po of fluorescent L5 (tdTomato) corticothalamic axon terminals in the Rbp4-Cre;Ai14 mice. a, Epifluorescence image of a representative coronal section of the cerebral cortex. Note the massive and ubiquitous labeling of L5 cells across motor (M1), primary somatosensory (S1HL, S1FL, S1DZ, S1BF, S1ULp), and second somatosensory (S2) cortical fields, in sharp contrast with the virtual absence of fluorescently labeled cells in L6. Dorsal is to the top and medial to the left. b, c, Distribution of tdTomato-labeled axon bundles and axonal varicosities (putative synaptic boutons) in two coronal sections of Po. PC: Paracentral nucleus. b1–c2, High-magnification detail taken from the regions indicated by a square in the images shown in panels b, c. Note the abundance of large boutons ventral Po (b1, c3) and their scarcity/smaller size in dorsal Po (c2). Note also the virtual absence of boutons in VPM, where only axon bundles are present (c3). AP: bregma level in mm. Scale bars: 500 µm (a), 250 µm (b, c), and 50 µm (b1, c2, c3).
Distribution within Po of GABAergic axon terminals
Local interneurons are essentially absent from mouse Po (Evangelio et al., 2018; Jager et al., 2021), but the nucleus receives robust extrinsic GABAergic input from TRN, ZI (Barthó et al., 2002), and the APT (Bokor et al., 2005). To determine whether these inhibitory pathways innervate Po in homogeneous fashion, we immunolabeled Po sections with antibodies against two different widely used markers of GABA-releasing terminals: the enzymes for GABA synthesis GAD65/67 (Erlander and Tobin, 1991; Fig. 3a,b) and the vGAT (Fig. 3c,d). Both molecular species are preferentially or exclusively localized at the synaptic vesicle pools of GABAergic neuron axons (Esclapez et al., 1994).
Distribution within Po of GABAergic axon terminals. a, b, Coronal sections of wild-type mice Po immunostained for GAD65/67. Dorsal is to the top and medial to the left. Note the uneven and relatively sparse neuropil labeling of Po, compared with that of the adjacent parafascicular (Pf) and VPM nuclei. c, d, Coronal sections of wild-type mice Po immunostained for vesicular GABA transporter (vGAT). High-magnification details (c1–c4) of the regions outlined by rectangles on the image in panel c are shown. Arrowheads indicate focal clusters of larger immunopositive puncta. PC: Paracentral nucleus; PF: Parafascicular nucleus. Bregma level in mm is indicated in the upper right corner. Scale bars: 250 µm (a–d) and 25 µm (c1–c4).
The labeling patterns produced by GAD65/67 and vGAT immunohistochemistry were identical (Fig. 3). Labeling was overall much fainter than in adjacent VPM and intralaminar nuclei, and was limited to neuropil immunopositive puncta. Small (<1 µm2) round and faintly labeled puncta were present throughout the nucleus (Fig. 3l). In addition, several separate focal clusters of larger immunopositive puncta (1–4 µm2), were also present (see Fig. 3, arrowheads). These clusters were more frequent and spread in the rostral portion of Po, and became limited to the dorsal third of the nucleus at more caudal levels. Additional smaller groups of large puncta were also present along the border with VPM (Fig. 3c1). VPM labeling consisted of large puncta (Fig. 3c2).
Convergence/divergence of GABAergic terminals with glutamatergic terminals of subcortical origin or cortical L5b terminals
Double-immunolabeling against vGLUT2 and GAD65/67 performed on Rbp4-Cre;Ai14 fluorescent tissue provided a direct visualization of the intricate convergence/segregation patterns that are established in the Po neuropil between L5b, glutamatergic subcortical, and GABAergic axon terminals (Fig. 4). Overall, these patterns revealed that large L5b terminals are present throughout Po, and predominate in its ventral and medial regions, where vGLUT2 and GABA puncta are scarce and smaller (Fig. 5). In contrast, low-density L5b labeling patches in the dorsal portions of Po are matched by similar patches of high vGLUT2 and high GAD65/67 puncta density/size. Higher magnification analysis shows that the three kinds of terminals remain always distinct despite close intermingling (Fig. 5). Overall, this multilabeling analysis of the Po neuropil reveals a variegated mosaic of small and partly overlapping microdomains. The thalamic cells located in different Po subregions may thus receive combinations of axon terminals that differ in the kind of signal that they carry (e.g., cortical or subcortical excitatory, and/or inhibitory) as well as in their presynaptic structure (e.g., bouton or neurotransmitter vesicle pool size). Such differences may entail different signal transmission dynamics in different axon terminals and allow diverse computations by the postsynaptic Po cells (Groh et al., 2014; Rodriguez-Moreno et al., 2020; Acsády, 2022).
Multilabeling comparison of glutamatergic subcortical (vGLUT2) and GABAergic (GAD65/67) puncta with the fluorescent corticothalamic L5 axons and boutons in Rbp4-Cre;Ai14 mouse thalamus. Multichannel confocal imaging of thalamus sections fluorescently immunolabeled for vGLUT2 (a–c; green), tdTomato (L5; d–f; magenta), GAD65/67 (g–i; blue), and their merge (j–l) in Rbp4-Cre;Ai14 coronal thalamus sections. Note that regions enriched in vGLUT2 and GAD65/67 are roughly complementary to those receiving L5 cortical axons. Note also the heavy GABA and vGLUT2 labeling in VPM. The dash lines square in panels a-c indicate the location of the high-magnification details in Figure 5. PC: Paracentral nucleus; PF: Parafascicular nucleus. AP: bregma level in mm. Scale bar: 250 µm.
Brain-wide retrograde tracer mapping of inputs to Po
While the immunofluorescence analysis revealed that excitatory and inhibitory inputs distribute unevenly within Po, it left unresolved the precise origin and local prevalence of the various afferent pathways. To this end, we injected retrograde tracers unilaterally into Po (Fig. 6) and mapped labeled cell somata across the whole brain and spinal cord. The tracer deposits consisted of a core of continuously bright fluorescence with no discernible cellular elements in it. This core ranged in diameter between 250–450 µm. We assume that this is the region from which virtually all tracer uptake and transport actually occurred, because all deposits with a core under 250 µm of diameter failed to produce consistent retrograde labeling or did not produce labeling at all. Beyond this core, there was a fading halo of faint fluorescence where discrete cellular elements were recognizable. This halo probably made minimal, if any, contribution to the observed labeling (Schofield et al., 2007).
High-magnification details of the distribution of glutamatergic subcortical (vGLUT2), GABAergic (GAD65/67) puncta and corticothalamic L5 axons and boutons. a–l, Comparison of vGLUT2 (a–d; green), tdTomato (L5-labeled cellls; e–h, magenta), and merge (i–l) dorsolateral Po, dorsomedial Po, ventral Po, and VPM domains in Rbp4-Cre;Ai14 mouse thalamus. m–x, Comparison of vGLUT2 (m–p; green), with GAD65/67 (q–t; blue), and its merge (u–x) in dorsolateral Po, dorsomedial Po, ventral Po, and VPM domains of the wild-type C57BL/6 mice. The sampled regions are indicated in Figure 4a–c. Scale bars: 10 µm (a–x).
Location of retrograde tracer deposits. Location of the six retrograde tracer deposits in Po (a; Po1, FB; Po2- Po6, FG) and in VPM (b; VPM1, FB; VPM2, FG). Epifluorescence images showing the center of the deposits are shown on the left. Deposits were drawn overlaying microphotographs in all Po coronal sections of Paxinos and Franklin (2012) atlas, and they were represented with halo and central zone. Anteroposterior, dorsoventral and mediolateral coordinates are indicated above, on the right or below, respectively. Adjacent serial sections were stained with cytochrome oxidase histochemistry to delimit thalamic nuclei. Eth: Ethmoid thalamic nucleus; LD: Lateral Dorsal thalamic nucleus; Ang: Angular nucleus; LP: Lateral Posterior nucleus; LPM, Lateral Posterior Medial nucleus; LPL: Lateral Posterior Lateral nucleus; OPC: Ovoid Paracentral nucleus. Scale bars: 250 µm.
Six cases with FG or FB deposits limited to Po were considered valid for analysis (Fig. 6a). Together, they covered about two-thirds of the nucleus. For comparison, we also analyzed the labeling produced by two further deposits in the adjacent VPM, a FO relay nucleus (Fig. 6b). We reconstructed the extent of each deposit on serial sections. Some deposits were completely separated (e.g., cases Po1 vs Po5 or Po6), while others overlapped among them to variable extent.
Retrograde tracer mapping of corticothalamic inputs
Retrograde tracer injections in Po or VPM labeled massive numbers of corticothalamic cells (Fig. 7a–d). In register with previous reports (Deschênes et al., 1998; Killackey and Sherman, 2003), neurons labeled in S1 after Po or VPM injections distributed in complementary laminar patterns. The Po injections labeled cells in L6b and L5b, while VPM injections labeled cells in L6a, with few additional cells in L6b (Fig. 7e). In addition, there were notable disparities in the tangential spread across cortical areas (Figs. 7j, 8). VPM injections labeled corticothalamic neurons almost exclusively in ipsilateral area S1 (Fig. 7i), while Po injections invariably labeled cells in several cortical areas such as motor area M1 and somatosensory areas S1 and S2. Additional cells were labeled by some Po injections in the secondary motor area (M2), and/or in the ectorhinal, insular and perirhinal areas (Figs. 7j, 8).
Area origins of L5a, L6a and L6b corticothalamic projections to VPM and Po. Epifluorescence microphotographs showing retrogradely stained neuron somatas in a coronal section of the cerebral cortex from a deposit located in VPM (a) and in Po (b). c, High-magnification detail of L5b corticothalamic-labeled cells. d, High-magnification detail of L6 corticothalamic-labeled cells. e, Cortical column details showing the corticothalamic input patterns in S1BF, S1mv, M1, and S2. f, g, Generation of an isometric surface map of mouse somatic motor and sensory areas. This map was created by aligning all the pial contours of Paxinos and Franklin (2012) coronal section diagrams as parallel lines along the dorsal midline of the cerebral hemisphere (black arrowheads and dashed black line on the map). Yellow dots indicate bregma. h, Flat map detail of the areas used in the subsequent figures to represent the corticothalamic labeling. i, j, Location of ipsilateral labeled cortical neurons: Separate maps are used to represent L5b (magenta), L6a (blue), and L6b (green) cells. Dots represent relative-labeled cell densities. Panels f, g represent the labeling in VPM1 and Po3 cases, respectively. A1: primary auditory area; Cg: cingulate cortex; Ect: ectorhinal area; FrA: frontal association; INS: insular (granular + dysgranular) cortex; M1: primary motor area; M2: secondary motor area; Prh: perirhinal area; S1: primary somatosensory area; S2: secondary somatosensory area; V1: primary visual area. Scale bars: 250 µm (a, b), 25 µm (c, d), and 150 µm (e).
Area origins of L5a, L6a, and L6b Corticothalamic projections to different Po domains. a, Direct comparison of the relative position of the tracer deposits. b, Location of ipsilateral retrogradely labeled cortical neurons in cases Po1, Po2, Po4, Po5, and Po6. Conventions as in Figure 7i,j.
Injections in different Po regions produced specific patterns of labeling across areas. L6b cells were labeled in large numbers across the motor and somatosensory areas. Additional L6b cells were labeled in nearby areas, as well as in the contralateral hemisphere. L6a cells were labeled in large numbers always in M1, and often in area S2. In contrast, labeling of L6a neurons in area S1 was relatively scarce, except in those deposits located in rostral portions of the nucleus (Po1, Po2, and Po3; Figs. 7j, 8).
Cells were labeled in L5b only in the ipsilateral hemisphere, mainly in S1, and often in S2 and M1. There was rough somatotopic order in this projection; cells were labeled in the macrovibrissae domain (cases Po1, Po2), snout domain (Po3), limbs domain (Po4, Po6), or trunk domain of S1 (Po5, Po6) in register with the position of each deposit and the known somatotopic organization of cell responses within Po (Diamond et al., 1992; Fig. 8b). Interestingly, the deposit in case Po5, which was centered approximately in the location of the region containing few L5 terminals in the Rbp4-Cre;Ai14 transgenic mouse (Fig. 2) labeled an unusually low number of L5b cells.
Retrograde tracer mapping of subcortical inputs
In addition to corticothalamic neurons, tracer deposits in Po labeled multiple cell populations across the diencephalon, brainstem, and spinal cord (Figs. 9, 10). Sizable numbers of retrogradely labeled neurons were always found in some structures known to use GABA predominantly or exclusively in their projections to the thalamus such as the TRN (Fig. 9a), ZI (Fig. 9b), and the APT (Fig. 9c). In addition, labeled neurons were consistently found, in variable numbers, in subcortical regions known to use mainly or exclusively glutamate in their excitatory projections to thalamus, such as the SC intermediate layers (Fig. 9d), the trigeminal complex (Fig. 9e–g) and the DCN (Fig. 9h). In some experiments, cells were labeled in the parabrachial complex (PB), the contralateral medial vestibular nucleus and some nuclei of the brainstem reticular formation, as well as in layers 1–2, 5, and the lateral spinal nucleus of the spinal cord (Fig. 9i).
Subcortical inputs to Po. Epifluorescence images showing retrogradely labeled cell bodies (arrows) in the ipsilateral reticular thalamic nucleus (TRN; a), zona incerta (ZI; b), anterior pretectal nucleus (APT; c), superior colliculus (SC; d), and contralateral principal (PrV; e, f) and interpolaris spinal (SpVI; g) trigeminal nuclei, gracilis nucleus (Gr; h), and dorsal horn of spinal cord (i). High-magnification details are shown as insets. A coronal section sketch is included, for reference. CuR: cuneatus nucleus, rotundus part; DMSpV: dorsomedial spinal trigeminal nucleus; DpG: deep gray layer; Gr: gracilis nucleus; InG: intermediate gray layer; InWh: intermediate white layer; Op: optic tract; PrV-DM: principal trigeminal nucleus, dorsomedial part; PrV-VL: principal trigeminal nucleus, ventrolateral part; ZId: dorsal zona incerta; Ziv: ventral zona incerta. Scale bars: 250 µm (a–i) and 25 µm (insets).
Summary of subcortical projections to Po and VPM. Graphic summary of the relative abundance of retrogradely labeled cells in the forebrain, midbrain, hindbrain, and spinal cord in all cases analyzed in this study. Ipsilateral (I) and contralateral (C) projections are represented separately for each case. Relative abundance of labeled cells is represented by shades of gray. Asterisks indicate the glutamatergic structures (red) and GABAergic structures (blue) that were quantified (Fig. 11).
In contrast, the subcortical neurons labeled by the VPM injections (which were both placed in the vibrissal domain of this nucleus) were present only in the trigeminal complex (principal and spinal interpolaris nuclei) and TRN (Fig. 10). Overall, the contrasting patterns of labeling after either Po or VPM deposits are in register with the notion that while VPM neurons are devoted specifically to the relay of excitatory tactile signals from the face and mouth, the Po cells may integrate, along with cortical inputs from L5b, diverse combinations of full-body tactile, motor, nociceptive, and attention-related signals conveyed via excitatory or inhibitory synapses (Yu et al., 2006; Groh et al., 2014; Ahissar and Oram, 2015).
Comparisons between deposits that involved different Po regions reveal local fluctuations in the prevalence of the various afferent systems (Fig. 10). For example, the deposits in the dorsal and caudal portions of the nucleus (Po4, Po5, Po6) labeled sizable numbers of cells in SC, while deposits involving ventral and anterior injections did not. Likewise, spinal cord and DCN cells were labeled in abundance only by dorsally and medially situated deposits, whereas cells in the principal trigeminal nucleus (PrV) were labeled selectively by deposits near the border with VPM. Labeling in APT was robust in all cases except one limited to the ventral and central portions of the nucleus (case Po3). ZI input was substantial in all cases, but more prominent in dorsocaudal deposits (Po5 and Po6). In contrast, labeling in the TRN was robust in all cases. Some few cells were labeled in the substantia nigra reticulata in the two cases (Po3 and Po6) whose injection halo encroached into the adjacent intralaminar nuclei, but not in other experiments.
Quantitative estimation of local input convergence
As a way to gauge the local convergence of different inputs, we compared the numbers of cells labeled in the various structures in each retrograde tracing experiment. We counted cells in the whole brain and spinal cord. To normalize for fluctuations in labeling efficacy or injection volume between experiments, cells in each structure were compared as percentages over labeled neuron totals. This parameter, often referred to as fraction of labeled neurons (FLN; Markov et al., 2011) provides an objective comparison of the relative anatomic weight of the pathways. Note, however, that FLN cannot be directly equated to functional impact without factoring in other parameters such as the local convergence and strength of the synapses, as well as the postsynaptic receptor mechanisms involved (Sherman and Guillery, 2002; Rodriguez-Moreno et al., 2020).
First, we compared the relative contributions of the various glutamatergic input pathways to the Po regions impregnated in each retrograde tracer experiment. We considered this analysis relevant because the ratio of cortical versus subcortical origins of excitatory inputs is widely used for functional classification of thalamic nuclei. According to this classification, FO are the thalamic nuclei that relay ascending excitatory signals from the subcortical systems to the cortex, while HO nuclei send back to cortex signals received from the cortex via collaterals of L5b axons. A third “integrator” category has been proposed for nuclei like Po, where at least some neurons receive both L5b and excitatory subcortical inputs and can integrate them in complex nonlinear fashion (Groh et al., 2014). Among the excitatory pathways we compared the numbers of cells labeled in pathways previously shown to drive Po neuron firing and specify their information content (hence referred to as “driver” pathways; Sherman and Guillery, 1996). We excluded corticothalamic L6 cells because we assume that their prevalence is homogeneous across Po, and their effect on thalamus cells essentially modulatory, fine-tuning variables such as temporal profile of the “driver” signals carried by other pathways (Sherman, 2016; Briggs, 2020).
As a first step, we compared labeled cell numbers in cortical L5b versus all glutamatergic subcortical structures (Fig. 11b). In the VPM control cases, subcortical sensory afferents represented always over ∼90% of the sum cortical L5b and subcortical labeled cells. Such massive prevalence may even be an underestimate, as the labeled L5b cells probably represent passing axons directed to Po that were impregnated at the injection site. In contrast, in the Po cases L5b cells prevailed (∼65–75%) in all cases except in two (24.7% in case Po5 and 42.7% in Po6) whose deposits overlapped the vGLUT2-rich domain in central-dorsal Po. Overall, these findings show that while most of Po is robustly innervated by L5b, all parts of the nucleus receive a sizable number of inputs from subcortical excitatory structures. The central and dorsal vGLUT2-rich domain of Po receives a proportion of subcortical excitatory inputs in a proportion nearing that of the “FO” VPM.
Quantitative estimation of convergence between input sources based on retrograde labeling patterns after injections in different Po subregions. a, Location of the retrograde tracer deposits. b, Percentage comparison of labeled cells sending putatively “driver” excitatory inputs to Po (L6 cells are not included). Cortical L5b (colored in white) versus subcortical structures (superior colliculus + sensory brainstem + spinal cord) shaded in gray. The total number of labeled cells counted in these structures in each experiment is indicated above. c, Percentage comparison of cells labeled among the four structures sending putatively inhibitory inputs to Po: anterior pretectal nucleus (APT, white), zona incerta (ZI, dash), reticular thalamic nucleus (TRN, gray shade) and substantia nigra pars reticulata (SNR, black). The asterisk used to indicate SNR labeling in some experiments probably reflects the involvement of the intralaminar nuclei and/or internal medullary lamina fibers. d, Percent comparison of L5b cells labeled in each cortical area over the total of putatively excitatory inputs (cortical + subcortical). e, Percent comparison of labeled cells in each subcortical structure over the total of putatively excitatory inputs (cortical + subcortical). Structures showing <2% of the total labeled cells were grouped as “others.” f, Correlation analysis between the absolute numbers of cells retrogradely labeled in subcortical structures providing Po long-range inhibitory inputs (APT + ZI + SNR) versus those labeled in subcortical structures providing putatively excitatory inputs. g, Correlation analysis between the absolute numbers of cells retrogradely labeled in subcortical structures providing long-range inhibitory inputs versus cells labeled in cortical L5b. h, Correlation analysis between the absolute numbers of cells retrogradely labeled in cortical L5b versus cells labeled in the trigeminal complex. i, Correlation analysis between the absolute numbers of cells retrogradely labeled cells in cortical L5b versus cells labeled in the spinal cord and DCN. Correlations between cases in Po are indicated by linear regression as well as by the coefficient of correlation (R2), data on VPM experiments are also shown for comparison. DCN: dorsal column nuclei; PrV: principal trigeminal complex; S1: primary somatosensory cortex; S2: secondary somatosensory cortex; SC: superior colliculus; SpV: spinal trigeminal complex; VC: vestibular complex.
Cortical L5b input to Po originated mainly from S1, finding at least 63% of all L5b-labeled cells in this area and up to 98.3% in case Po2 (74.1% of all putative “driver” inputs; Fig. 11d). Other Po domains also receive a significant percentage of L5b inputs from other cortical areas, such as the motor cortex (Po1, 16.5%; Po5, 19.5%) or S2 (Po3, 20.8%).
The trigeminal complex was labeled in all cases, but most robustly in experiments placed in rostral and ventral levels of Po. This input arose always from the SpV (oralis and interpolaris nuclei present 12.3–29.3% of subcortical cells), and, in the deposits injected in the anterior half of Po, also from PrV (Fig. 11e). Cells in the intermediate layers of SC were most numerous in those experiments with deposits in dorsal and caudal Po, being the majority (56.4%) among the glutamatergic subcortical cells labeled in case Po6 (32.3% of all putative “driver” inputs; Fig. 11e). DCN cells were labeled in all cases, yet they were markedly more numerous when the deposit overlapped the position of the dorsal vGLUT2-rich domain of Po (30.6% in Po4 and 21.6% in Po5 of all subcortical cells). Likewise, spinal cord neurons were a small but consistent fraction of the subcortical glutamatergic afferent mix. The most abundant labeling was observed in cases located in dorsal Po (up to 23.5% of all subcortical cells in Po5 and Po6; Fig. 11e).
Likewise, we compared the proportion of retrogradely labeled cells within the group of structures known to innervate the thalamus via GABAergic axons (Fig. 11c). TRN was robustly labeled in all Po injection cases, although its prevalence among GABAergic inputs fluctuated widely (32.8–82.7%) depending on the presence or absence of other inputs. Projections from APT were labeled after every retrograde tracer injection in Po. Their prevalence among other GABAergic inputs ranged from 8.0% (case Po3, involving ventral Po) to 49.6% (case Po4, involving dorsal-central Po). Neurons projecting to Po from the ZI were likewise labeled by every Po injection. They fluctuated between 5.6–29.7%, and were most abundant after injections involving dorsal and caudal Po portions. Remarkably, we found a positive correlation (R2 = 0.75) between the number of labeled cells in the extra-reticular inhibitory structures with the subcortical structures (Fig. 11f), but not with the number of labeled L5b cells (R2 = 0.17; Fig. 11g). In addition, a strong positive correlation (R2 = 0.88) was also found between L5b neurons and the number of cells in the trigeminal complex (Fig. 11h), but not with the number of cells labeled in the spinal cord and DCN (R2 = 0.43; Fig. 11i). In striking contrast with Po, VPM received GABAergic inputs exclusively from TRN (Fig. 11c).
Distribution within Po of afferent pathways
Retrograde tracing experiments informed about the sources and relative weight of the various input pathways to Po but did not delineate with precision the domains targeted or spared by each pathway. To this end, first we compared the fluorescent protein axonal labeling produced in Po by relatively large associated adenovirus vector (AAV) injections placed in some structures identified in the retrograde study (Fig. 12; Allen Institute for Brain Science, 2011; Oh et al., 2014; Harris et al., 2019).
Anterograde mass labeling of some subcortical inputs to Po with viral vector-mediated transfection. Two-photon tomography image samples taken from experiments in which viral vectors able to drive the expression of high levels of fluorescent protein were injected in some of the structures that originate inputs to Po. Injections sites in the anterior pretectal nucleus (APT; a), the zona incerta (ZI; c), the interpolaris spinal trigeminal nucleus (SpVI; e), or the superior colliculus (SC; g). Coronal section images showing fluorescently labeled axons in Po are shown in b, d, f, h. Images from the Mouse Connectivity Projection datasets can be found in http://connectivity.brain-map.org/projection/experiment/278401778 (a, b), /301539438 (c, d), /272738620 (e, f), and /128001349 (g, h). AP: bregma level in mm. CL: central lateral nucleus; PC: paracentral nucleus; PF: parafascicular nucleus; VL: ventrolateral nucleus; VPM: ventral posteromedial nucleus; ZId: dorsal zona incerta; ZIv: ventral zona incerta; DpG: deep gray layer; InG: intermediate gray layer; InWh: intermediate white layer; Op: optic tract. Scale bars: 1000 µm (a, c, e, g) and 500 µm (b, d, f, h).
This analysis provided several important insights. First, it showed that APT axons arborize heavily in the most rostral Po levels (at about ∼AP −1.50 mm), but far less so at more caudal levels, where they concentrate mainly in the dorsolateral corner of the nucleus (Fig. 12a,b). Remarkably, the dense APT terminal axon arborizations do not spread to other nuclei and display an uneven, clustered pattern. Second, viral vector injections in ZI consistently labeled axons targeting only the dorsal half of Po. Unlike APT, the ZI axon arborizations extend dorsally to the nearby laterodorsal and lateral Pos (Fig. 12c,d). Vector injections in the spinal trigeminal nucleus labeled axons in ventral and lateral portions of Po and provide substantially more profuse axon arborizations in VPM (Fig. 12e,f). Finally, axons from the intermediate/deep layers of the SC innervate only a dorsal fringe of Po but arborize densely in the adjacent lateral Po (Fig. 12g,h).
To analyze the fine morphology of axon arborizations and complete the delineation of their target territories, we made BDA microinjections in the various subcortical structures that innervate Po (Fig. 13; Extended Data Fig. 13-1). Some BDA injections were placed in the same structures labeled in the Allen Brain Connectome Dataset (Allen Institute for Brain Science, 2011). The distribution within Po of these BDA-labeled axon terminals was consistent with the viral transfection data and added topographic detail (Fig. 14; Extended Data Fig. 14-1).
Axons in Po labeled by BDA injections in glutamatergic subcortical structures. Microphotographs showing location of BDA deposits in the cuneatus (Cu; a) and gracilis (Gr; c) nuclei, ventrolateral part of principal trigeminal nucleus (PrV-VL; e), or pars caudalis of the spinal trigeminal nucleus (SpVC; g). Cytochrome oxidase counterstain. Confocal images of double immunofluorescence labeling within the contralateral Po of vGLUT2 terminals (green) and BDA axons (magenta) from Cu (b), Gr (d), PrV-VL (f), and SpVC (h). Insets are represented as white squares. High-magnification detail is shown for each case. Arrowhead indicates BDA axonal varicosities immunopositive for vGLUT2. AP: bregma level in mm. Scale bars: 250 µm (a, c, e, g), 100 µm (b, d, f, h), and 25 µm (b1, d1, f1, h1). Similar images from BDA injection cases involving the dorsomedial region of principal trigeminal nucleus, the oralis or interpolaris parts of the spinal trigeminal nucleus, the parabrachial nuclei, the superior colliculus, or the cervical spinal cord are provided as Extended Figure 13-1.
Extended Data Figure 13-1
Axonal arborizations in Po arising from subcortical glutamatergic structures. BDA deposits in the dorsomedial region of principal trigeminal nucleus (PrV-DM; a), oralis (SpVO; c) and interpolaris (SpVI; e) part of spinal trigeminal nucleus, parabrachial nuclei (PB; g), superior colliculus (SC; i) and a low cervical level of spinal cord (k). Details of the axonal arborizations in Po arising from PrV-DM (b), SpVO (d), SpVI (f), PB (h), SC (j) and spinal cord (l) are shown. Cytochrome oxidase counterstain. AP: bregma level in mm. Scale bars: a, c, e, g, i, k = 250 μm; b, d, f, h, j, l = 100 μm; insets = 25 μm. Download Figure 13-1, TIF file.
Axonal varicosities in Po labeled by BDA injections in glutamatergic subcortical input structures. Schematic representation of the location of axonal varicosities in Po and adjacent thalamic nuclei, plotted from the anterograde axonal labeling of BDA microiontophoretic injection sites in gracilis nucleus (a), cuneatus nucleus (b), ventrolateral region of principal trigeminal nucleus (PrV-VL; c), the caudalis part of the spinal trigeminal nucleus (SpVC; d), superior colliculus (SC; e), and cervical spinal cord (f). Individual axonal varicosities in Po are represented as red circles, and axonal varicosities outside Po are represented in gray color. Ang: angular thalamic nucleus; LD: laterodorsal nucleus; LP: lateral posterior nucleus; LPM: lateral posterior nucleus, medial part; LPL: lateral posterior nucleus, lateral part; PC: paracentral nucleus; PF: parafascicular nucleus; VL: ventrolateral nucleus; VPM: ventral posteromedial nucleus. AP: bregma level in mm. Scale bar: 250 µm. Additional diagrams in Extended Figure 14-1 illustrate the location of varicosities labeled in Po by BDA deposits in the dorsomedial region of principal trigeminal nucleus, the oralis or interpolaris parts of the spinal trigeminal nucleus or the parabrachial nuclei.
Extended Data Figure 14-1
Axonal varicosities labeled from subcortical glutamatergic structures. Location of axon varicosities labeled by BDA injections in the principal trigeminal nucleus (PrV-DM, a); pars oralis of the spinal trigeminal nucleus (SpVO, b); pars interpolaris of the spinal trigeminal nucleus (SpVI; c); and parabrachial complex (PB; d). Each red dot represents a varicosity in Po; varicosities outside Po are shown in gray ink. Abbreviations: Ang: angular thalamic nucleus; LD: laterodorsal nucleus; LP: lateral posterior nucleus; LPM: lateral posterior nucleus, medial part; LPL: lateral posterior nucleus, lateral part; PC: paracentral nucleus; PF: parafascicular nucleus; VL: ventrolateral nucleus; VPM: ventral posteromedial nucleus. AP: bregma level in mm. Scale bar: 250 μm. Download Figure 14-1, TIF file.
Moreover, double fluorescence labeling for vGLUT2 and BDA confirmed that the projections from both the Gr and Cu nuclei (n = 5) target the isolated vGLUT2-rich domain in the central and dorsal portion of Po in a focused manner. There they have large boutons (>1 µm2) immunopositive for this transporter (Fig. 13a–d). DCN axons target the dorsal and central region of the nucleus, similar to what is described in the rat Po (Villanueva et al., 1998). The distribution of labeled axon varicosities after BDA injections limited to the Cu or Gr indicate that axons from the Cu nucleus former terminate in more dorsal and lateral sectors of Po than those from the latter (Fig. 14a,b).
Likewise, boutons labeled from BDA injections in the principal (n = 4) and spinal trigeminal nuclei (n = 8) were immunopositive for vGLUT2 and formed small clusters in the lateral border of the nucleus, near VPM (Fig. 13e–h). Anterograde labeling experiments show that the trigeminal axons remain relatively scattered within ventral Po, being very scant in the dorsomedial third of the nucleus. Some trigeminal terminals mass into clusters of large boutons along the lateral border of Po (Fig. 14c,d; Extended Data Fig. 14-1).
In addition, the labeling produced in Po by large BDA injections in the intermediate layers of the SC (n = 2) innervate the dorsal third of Po, but not the ventral part of the nucleus (Extended Data Fig. 13-1; Fig. 14e). On the other hand, deposits in the parabrachial complex (PB; n = 2) innervate mainly the ventral and caudal portion of Po (Extended Data Figs. 13-1, 14-1d). Finally, BDA deposits in the vestibular complex (n = 3) show scarce axonal arborization in Po, with just a few varicosities (data not shown).
Vector transfection datasets did not include experiments involving the spinal cord. We examined the labeling produced in Po by large BDA injections (n = 2) in the central spinal gray matter laminae at about the third cervical level and found labeled axons in the dorsomedial portion of Po (Extended Data Fig. 13-1; Fig. 14f) a region matching that reported to respond to tactile stimulation of the limbs and trunk in rats (Diamond et al., 1992). Spinal cord terminals are more abundant in the caudal half of the nucleus.
Small deposits in different portions of the TRN (n = 4), labeled richly branched and relatively focal arborizations in Po. The position of these arborizations varied in register with that of the TRN injections. For example, a BDA deposit in a central domain of TRN labeled axonal arborizations in dorsolateral Po, along the border with VPM (Fig. 15a–c), while a deposit in a ventral domain of TRN labeled axons only in the ventral region of Po (Fig. 15d–f).
Axons labeled in Po by BDA injections in GABAergic projection structures. Microiontophoretic deposits in a dorsal (a) and ventral (c) portion of the thalamic reticular nucleus (TRN), the anterior pretectal nuclei (g) and in the dorsal (ZId; j) and ventral (ZIv; m) zona incerta. Axonal arborizations in the ipsilateral Po are represented in panels b, e, h, k, n, respectively. Cytochrome oxidase histochemistry counterstain. c, f, i, l, o, Location of axonal varicosities in Po and adjacent thalamic nuclei, plotted from the respective anterograde axonal labeling. Individual axonal varicosities in Po are represented in red ink; those outside Po are drawn in gray. Eth: Ethmoid thalamic nucleus; LD: laterodorsal thalamic nucleus; VPL: Ventral posterolateral nucleus; ZId, ZIv: dorsal and ventral divisions of the Zona Incerta. Scale bars: 250 µm (a, c, d, f, g, i, j, l, m, o), 100 µm (b, e, h, k, n), and 25 µm (insets).
The distribution of APT (n = 2) and ZI (n = 8) BDA-labeled axon terminals within Po was consistent with the viral transfection data (Fig. 15g–o) and with some previous studies in rat (Barthó et al., 2002; Bokor et al., 2005). APT axons are present across most of Po and leave dense clustered arborizations in the rostral region of the nucleus (Fig. 15g–i). ZI axons distributed mainly in rostral Po and more sparsely at caudal and ventral levels (Fig. 15j–o). The position of these arborizations generally aligns with the heavy concentrations of large GAD65/67 and vGAT puncta observed in this region (Fig. 3).
Overall, the labeled axon arborization patterns in the viral vector tracing datasets or BDA microinjections are consistent with the immunolabeling and retrograde tracing mapping results. Together, they delineate several subdomains within Po, each containing a specific combination of two or more overlapping excitatory (L5b, DCN, trigeminal, SC) or predominantly inhibitory (APT, ZI) long-range input systems. The input combinations might thus impinge on particular Po projection neurons (Groh et al., 2014) while not on others. In contrast, the glutamatergic corticothalamic L6 terminals and GABAergic TRN terminals appear to cover Po homogeneously and in rough topographic fashion; hence, they probably reach all Po neurons with similar prevalence.
Axonal varicosity sizes of the input pathways
In glutamatergic and GABAergic axon terminals, varicosities reflect local accumulations of mitochondria and synaptic vesicle pools at presynaptic sites (which may be single or multiple). Hence, varicosity size provides an indirect indication of important functional parameters such as synapse release probability and strength (Sherman and Guillery, 2002; Halassa and Acsády, 2016; Rodriguez-Moreno et al., 2020). Interestingly, some of the afferent inputs to Po investigated in the present study such as the glutamatergic L5b corticothalamic (Hoerder-Suabedissen et al., 2018) or the mainly GABAergic ZI and APT axons (Halassa and Acsády, 2016) had been previously shown to have unusually large varicosities in their terminal axon branches.
To compare in systematic fashion varicosity sizes across the various input systems targeting Po and as an indirect indicator of their synaptic strength, we measured and compared their maximal projection areas in axons labeled by BDA injections in different subcortical structures (Fig. 13; Extended Data Fig. 13-1). For comparison, we included in this analysis a previously published dataset of bouton sizes from BDA deposits located in L5b, L6a, and L6b (Hoerder-Suabedissen et al., 2018).
First, we compared the axonal varicosities from cortical and subcortical glutamatergic structures (Fig. 16). For each structure, we analyzed both the frequency of size distribution (Fig. 16e) and the mean size (Fig. 16f). Statistical data comparisons are summarized in Table 3. We found that the mean sizes of varicosities in axons originating from L6a and L6b in the barrel field of the S1 (S1BF; 0.60 ± 0.02 and 0.58 ± 0.03 µm2, respectively) were always significantly smaller than those originated in L5b (2.42 ± 0.23 µm2) or subcortical structures (T3 Dunnet's, p < 0.05 for all comparisons) and had a significantly different frequency in size distribution (K-S, p < 0.001 for all comparisons). The varicosities from nucleus Gr, Cu, ventrolateral region of the principal trigeminal nucleus (PrV-VL), and the interpolaris portion of the spinal trigeminal nuclei (SpVI), as well as the deep intermediate layers of the SC did not differ significantly in size from cortical L5b varicosities. The varicosities in the caudal portion of the spinal trigeminal nuclei (SpVC; 4.16 ± 0.34 µm2) were significantly larger than those from L5b (T3 Dunnet's, p = 0.003). Previous studies in rats have shown that ascending inputs from SpVI form large varicosities in Po (Veinante and Deschênes, 1999; Pierret et al., 2000; Lavallée et al., 2005). Other glutamatergic axons had significantly smaller axonal varicosities than cortical L5b, including some trigeminal subnuclei (oralis portion of the spinal trigeminal nuclei, SpVO, and dorsomedial part of the principal trigeminal nuclei, PrV-DM), parabrachial complex and axons from the central laminae of the spinal cord, being the mean varicosities size of the last similar to those reported in rat Po (Iwata et al., 1992).
Quantitative analysis of varicosity size from glutamatergic input structures
Variable varicosity sizes in glutamatergic input axons. Representative axonal varicosities labeled by selective BDA injections in L6a (a), spinal trigeminal nucleus pars oralis (SpVO; b), cortical L5b (c), and ventrolateral part of the principal trigeminal nucleus (PrV-VL; d) in Po. Scale bars: 10 µm. e, Size distribution of axonal varicosities in Po labeled different cortical and subcortical glutamatergic structures. Axonal varicosities were grouped into discrete intervals according to their maximal cross-sectional area. f, Box-plot diagram showing axonal varicosity sizes of all glutamatergic structures innervating Po. SpVC: caudalis spinal trigeminal nuclei; Cu: cuneatus; Gr: gracilis; SC: superior colliculus; SpVI: interpolaris spinal trigeminal nucleus; PrV-DM: dorsomedial part of the principal trigeminal nucleus.
Varicosity sizes were likewise measured and compared among GABAergic axon populations (Fig. 17), analyzing the frequency of size distribution (Fig. 17e) and the mean size (Fig. 17f). We included in this analysis the varicosities from TRN axons in VPM. Statistical data are summarized in Table 4. The analysis revealed that the largest varicosities are those arising from APT (1.96 ± 0.09 µm2) and ZI (1.80 ± 0.08 µm2). No significant differences were found between APT and ZI varicosities in the overall distribution of sizes. Both populations included, in addition to a large number of relatively small varicosities, a consistent fraction of unusually large varicosities (Fig. 17f). Accordingly, APT and ZI axons have been reported to terminate as large boutons on proximal dendrites in rat Po (Trageser and Keller, 2004; Lavallée et al., 2005; Wanaverbecq et al., 2008; Halassa and Acsády, 2016). Axonal varicosities in TRN axons across most of Po were significantly smaller (0.84 ± 0.03 µm2; T3 Dunnet's, p < 0.001 for all comparisons) and had a significantly different frequency in size distribution (K-S, p < 0.001). In contrast, the terminals from this nucleus were much larger along the lateral border of Po, near VPM (1.83 ± 0.06 µm2) as well as inside VPM (1.58 ± 0.06 µm2; T3 Dunnet's, p < 0.001).
Quantitative analysis of varicosity size from GABAergic input structures
Variable varicosity sizes in GABAergic input axons. Representative axon varicosities labeled by BDA injections in structures known to innervate the thalamus with GABAergic axons: anterior pretectal nucleus (APT; a), zona incerta (ZI; b), and thalamic reticular nucleus (TRN; c, d). Scale bar: 10 µm. e, Size distribution of axonal varicosities grouped into discrete intervals according to their maximal cross-sectional area. f, Box-plot diagram showing axonal varicosity sizes.
Discussion
We mapped the origin, intranuclear distribution and axon varicosity sizes of the various excitatory and inhibitory input pathways terminating in Po, a representative HO nucleus of the rodent thalamus. Results reveal that while some of these pathways (L6 and TRN) innervate all regions of Po with similar prevalence, other pathways (L5b, DCN, trigeminal, SC, spinal cord, APT, ZI) concentrate in some Po regions and are absent from others, thus creating a complex mosaic of partly overlapping domains (Fig. 18).
Input landscapes of the Po neuropil. Cartoon diagram of Po regions targeted by the various input systems. The approximate delineation of preferential input regions is based on the combined evidence from immunolabeling and tracing data. Higher color saturation is used to indicate regions containing more densely packed terminals. Empty areas indicate regions containing few or no terminals. a, Thalamic reticular nucleus (TRN) and L6 corticothalamic inputs cover Po in roughly homogeneous fashion. b, Cortical L5b inputs. c, Territories targeted by axons from dorsal column nuclei, the spinal cord, trigeminal complex or the intermediate/deep layers of the superior colliculus (SCi). d, Territories targeted by zona incerta and anterior pretectal nucleus.
Cortical Po inputs
The massive presence of vGLUT1-positive terminals in the Po neuropil and the retrograde cortical labeling produced by tracer injections show that the cortex is, by far, the largest source of afferents to Po (Graziano et al., 2008; Zhang et al., 2018). Most of these corticothalamic cells are located in L6b in areas S1, S2, M1, and ectorhinal cortices, the same regions innervated by Po axons (Ohno et al., 2012; Casas-Torremocha et al., 2019). The origin of L6b projection shows a coarse topographic order, and it includes a small number of neurons in the same areas of the contralateral hemisphere.
L5b inputs can readily drive Po cell firing, hence determining the information content of their output (Reichova and Sherman, 2004; Theyel et al., 2010; Mease et al., 2016). Our data show that L5b inputs to Po originate mainly from ipsilateral S1, and to a lesser extent, also from S2 and M1. We show that the neurons in some Po regions can receive convergent L5b inputs from two different cortical areas such as S1 and S2 (cases Po3, Po6; Figs. 6, 7) or S1 and M1 (case Po), while other regions receive massive L5b inputs from area S1 alone (cases Po2, Po4). Neurons in these regions are thus likely to be predominantly driven by L5b inputs and involved in multiarea integration or transcortical communication (Sampathkumar et al., 2021). In contrast, the scant labeling of L5b cells in case Po5, whose position overlaps a vGLUT2-rich, GAD65/67-rich and Rbp4-Cre:Ai14 neuropil-poor domain located in the dorsal third of Po (Figs. 1–3) suggests that neurons in this Po region may be relatively less driven by cortical input.
Excitatory subcortical Po inputs
Our connection-tracing data show that Po receives sizable direct inputs from structures in the diencephalon, brainstem and spinal cord (DCN, trigeminal sensory complex, and SC) known to use excitatory neurotransmitters in their projections to the thalamus. In addition, vGLUT2 immunolabeling reveals that these connections are unevenly distributed within Po, both in terms of local abundance and puncta size.
A small but robust projection from the contralateral DCN targets the dorsal and central region of the nucleus, in a pattern equivalent to that described in the rat Po (Villanueva et al., 1998). These DCN axons form large boutons. Direct spinal inputs likewise reach the dorsal and central portion in the caudal half of the nucleus.
Trigeminal complex axons reach the ventral and central portions of Po but they are scant in the dorsomedial portions of the nucleus. In addition, some trigeminal terminals form clusters of large vGLUT2-positive boutons along the lateral border of Po. Some cells in this region may extend their dendrites across the border and have functional properties that are intermediate between those of typical Po and VPM cells (Slézia et al., 2011). Studies in rats have shown that Po receives most of its ascending inputs from SpVI and that these axons form large varicosities (Veinante and Deschênes, 1999; Pierret et al., 2000; Lavallée et al., 2005). In register with their bouton morphology, trigeminal axons drive Po cells with short latency, as typical driver inputs (Trageser and Keller, 2004; Lavallée et al., 2005).
The intermediate layers of the SC innervate the dorsal third of Po in relatively sparse fashion, but not the ventral part of the nucleus. These observations are in consonance with the recent demonstration that colliculo-thalamic terminals are able to drive Po neuron firing (Gharaei et al., 2020).
Inhibitory subcortical Po inputs
Immunolabeling for specific markers of GABAergic axon terminals and connection tracing data consistently indicate that extrinsic GABAergic inputs are present across the Po neuropil. On the one hand, every region of Po is innervated by TRN axons. Small terminals from TRN richly arborized axons (Pinault et al., 1995) are probably the main contributor to the faint, fine-grained and ubiquitous neuropil of GAD65/67-positive terminals observed all over Po.
In register with previous studies in rats (Barthó et al., 2002; Bokor et al., 2005), we show that APT and ZI axons reach most of Po and leave dense clustered arborizations in the rostral region of the nucleus, that align with clumps of large GAD65/67 and vGAT puncta in this region. Boutons in APT and ZI axons are on average larger than other putatively GABAergic Po inputs, and a fraction of them are unusually large. These axons have been reported to terminate as large boutons on proximal dendrites, able to exert powerful sustained inhibition of Po cells (Trageser and Keller, 2004; Lavallée et al., 2005; Wanaverbecq et al., 2008; Halassa and Acsády, 2016). Our data show that the ventral part of Po receives few ZI projections, consistent with reports the ZI inhibitory effects are found only in a fraction of Po neurons, particularly those located in rostral and dorsolateral regions of the nucleus (Lavallée et al., 2005).
In addition, TRN boutons in VPM and the lateral border of Po display significantly larger varicosities than in the rest of Po. Again, this observation suggests that the neuropil composition along the Po-VPM border displays transitional features (Slézia et al., 2011). These large varicosities may represent branches from VPM-targeting axons or a specific TRN subpopulation (Lam and Sherman, 2007; Halassa et al., 2014; Crabtree, 2018; Li et al., 2020).
In contrast with Po, VPM receives GABAergic inputs exclusively from TRN; however, the GAD65/67 and vGAT neuropil labeling in VPM is far heavier than in Po. This may reflect the fact that the TRN axons innervating VPM show three times more varicosities than those targeting Po (Pinault and Deschênes, 1998).
Input landscapes of Po
Multilabeling for cortical L5b, GABAergic and vGLUT2 boutons (Fig. 4) reveals a striking contrast between the uniform, “monoculture-like” input pattern of VPM and the variegated mosaic of Po inputs. The approximate extent of the regions preferentially targeted or avoided by the various afferent systems can be further delineated from the connection tracing data (Fig. 18). For example, L6 and TRN seem to innervate the whole extent of Po in topographic fashion and with similar overall prevalence across the nucleus (Fig. 18a). These two pathways form relatively small axonal varicosities, target distal dendritic domains (Pinault and Deschênes, 1998) and have mainly modulatory effects (excitatory or inhibitory, respectively) on signal transmission by thalamic cells (Liu et al., 1995; Wang et al., 2001; Halassa and Acsády, 2016; Crabtree, 2018; Briggs, 2020).
Other input systems innervate limited Po territories (Fig. 18b–d). Most or all have synapses powerful enough to determine cell spiking, either in excitatory or in inhibitory manner (Sherman, 2012; Halassa and Acsády, 2016). As the territories innervated by these input sources are partly segregated, qualitatively and quantitatively different input combinations may thus be present in different Po subregions (Fig. 19).
The input mix in the Po neuropil varies across the nucleus. Idealized diagram of a coronal section of Po and VPM. The L6 and TRN inputs are both massive and homogeneously distributed across the two nuclei (thick gray arrows pointing from the outside to each nucleus). Vertical arrows indicating different Po portions represent other pathways whose prevalence in the local neuropil is markedly variable. Roughly estimated prevalence is indicated by different case size. An uneven pale red shading is used to note that the prevalence of L5b terminals (in red ink) varies widely across Po. Other glutamatergic (in black ink) and GABAergic (in blue ink) inputs are listed. In contrast with Po, the VPM neuropil input mix is dominated by trigeminal inputs. The camera lucida drawing of a typical mouse Po neuron (Porrero et al., 2016) is added for reference, at the same scale. Note that because of the size and radial arrangement of their dendrites, many Po cells might sample signals from adjacent input-specific domains. Scale bar: 250 µm. SCi: intermediate layers of the superior colliculus; Sp: spinal cord; DCN: dorsal column nuclei; APT: anterior pretectal nucleus; ZI: zona incerta; PrV: principal trigeminal nucleus; SpV: spinal trigeminal nucleus; PB: parabrachial nuclei.
The ventral Po subregion is massively innervated by large-bouton L5b axons from face and mouth-related regions of cortical areas S1, S2 and M1. Hence, this region nears the notion of “HO nuclei” receiving its “driver” afferents only from L5b cells (Guillery, 1995), as reported to occur in parts of the primate pulvinar nucleus (Rovó et al., 2012). Note, however, that this ventral Po region contains, in addition, some small-bouton axon terminals from the trigeminal complex, parabrachial nucleus and APT. In contrast, the L5b terminals sparser in dorsal and lateral Po, where subcortical boutons (DCN, spinal, SC, ZI, and APT) are more abundant. This observation aligns with reports that neurons under inhibitory APT and ZI influence as well as those able to dynamically integrate cortical and ascending sensory signals are more frequently found in dorsolateral Po than in ventral Po (Bokor et al., 2005; Lavallée et al., 2005; Groh et al., 2014).
In addition to the complex arrangement of the various neuropil domains, it is intriguing to consider that mouse Po neurons have bushy or stellate dendritic trees of ∼250 µm mean diameter (Fig. 19; see also Ohno et al., 2012). As a result, most Po cells, and particularly those near domain borders, may receive inputs from multiple sources, in graded combinations (Slézia et al., 2011). In fact, there is experimental evidence that such convergence occurs. For example, cortical L5b and trigeminal synaptic boutons have been shown to target the same Po neurons; hence, their signals can be nonlinearly integrated in the thalamic cells to code for temporal relationships between tactile and cortical signals (Groh et al., 2014; Ahissar and Oram, 2015). Likewise, L5b inputs from two separate cortical areas have been shown to converge on the same Po cells (Sampathkumar et al., 2021). Thus, in Po and in other HO thalamic nuclei, the complex patterns of input segregation/overlap may expand in combinatorial fashion the range of possible computations performed by the thalamocortical neurons.
Footnotes
The work in the laboratory of FC was supported by the European Union's Horizon 2020 Framework Program for Research and Innovation under the Specific Grant Agreement No. 945539 (Human Brain Project SGA3) and Spain's Ministerio de Ciencia e Innovación (PID2020-115780GB-I00). The work in the laboratory of Z.M. was supported by the Medical Research Council Grant G00900901 and Einstein Stiftung. S.H. was supported by the Daiichi Sankyo Foundation of Life Science, The Uehara Memorial Foundation and KAKENHI (Grant-in-Aid for Research Activity Start-up 19K23786; to S.H.). We thank Ms. Marta Callejo and Ms. Begoña Rodriguez for excellent technical help as well as suggestions for improvement by two anonymous reviewers.
The authors declare no competing financial interests.
- Correspondence should be addressed to Francisco Clascá at francisco.clasca{at}uam.es
