Projection Motifs and Wiring Logic of Medial Pulvinar Thalamocortical Axons in the Marmoset Monkey
- María Angélica Córdoba-Claros1,*,
- Pablo Rubio-Garrido1,*,
- Ruthnaldo R.M. de Lima2,
- Paulo Leonardo A.G. Morais2,
- Expedito S. do Nascimento Jr2,
- Jeferson S. Cavalcante2, and
- Francisco Clascá1
- 1Department of Anatomy & Graduate Program in Neuroscience, School of Medicine, Universidad Autónoma de Madrid, Madrid 28029, Spain
- 2Programa de Pós-graduação em Ciências Morfológicas, Universidade Federal do Rio Grande do Norte, Natal RN CEP 59078-900, Brazil
-
Author contributions: P.R.-G., E.S.d.N., J.S.C., and F.C. designed research; M.A.C.-C., P.R.-G., R.R.M.D.L., P.L.A.G.M., E.S.d.N., and F.C. performed research; M.A.C.-C., P.R.-G., and R.R.M.D.L. analyzed data; P.R.-G. and F.C. wrote the paper.
Abstract
The medial pulvinar thalamic nucleus (MPu) is an evolutionary novelty of the primate thalamus, prominently expanded in humans. Piecemeal data from studies in various monkey species indicate that MPu axons reach prefrontal, inferior parietal, cingulate, insular, or temporal areas; however, the precise wiring and functional logic of such brain-wide connections remain obscure. In marmoset monkeys (Callithrix jacchus) of both sexes, we visualized the axons originated from specific pulvinar domains by means of biotinylated dextran amine microinjections and compared them across multiple cases. In addition, by injecting retrograde tracers in the cortical areas targeted by the pulvinar axons, we investigated the organization of projection cells within MPu and the existence of long-range branched axons. Specific projection motifs reveal a caudal MPu subnucleus that innervates inferior and ventral temporal areas and a rostral MPu subnucleus that innervates temporal, ventral prefrontal, premotor, inferior posterior parietal, and cingulate areas. We demonstrate numerous MPu neurons that innervate through branched axons prefrontal and parietal or prefrontal and temporal areas; other cells with different projection patterns are closely intermingled with them. Our findings support the notion that MPu is a hub of the brain-wide networks that support complex visual and social cognition, sensory-guided reaching, working memory, and attention. Moreover, the finding of long-range branching MPu axons and dense terminal arborizations suggest that MPu cells may regulate functional connectivity among high-level cortical areas at different spatial scales. Besides, the anatomical “ground truth” provided by our study is relevant for functional imaging and distributed network modeling studies.
Significance Statement
The medial nucleus of the pulvinar complex is an evolutionary novelty of the primate thalamus and is uniquely expanded in humans. However, its functions are poorly understood, as data on its connections remain incomplete and fragmentary. Using high-resolution connection-labeling methods in marmoset monkeys, we mapped in full the thalamocortical axons arising from specific medial pulvinar domains and analyzed their branching patterns. Projection motifs reveal two main subnuclei, each targeting specific cortical networks. Moreover, we show that some pulvinar neurons simultaneously innervate distant prefrontal and parietal areas or prefrontal and temporal areas. Their connection patterns might allow medial pulvinar cells modulate, at different scales, functional connectivity in the cortical networks supporting object vision, social cognition, attention sensory-guided reaching, and working memory.
Introduction
The pulvinar is the largest nuclear group of the thalamus in anthropoid primates. While its cellular composition is basically homogeneous (Ma et al., 1998; Darian-Smith et al., 1999; Chan et al., 2021), four main pulvinar nuclei are traditionally delineated based on position, neurochemical markers, and connection patterns, namely, the inferior (IPu), lateral (LPu), medial (MPu), and anterior (APu) pulvinar nuclei (Olszewski, 1952; Jones, 2007a).
Studies have historically focused on IPu and LPu, collectively referred to as the “visual pulvinar.” These nuclei are interconnected with specific temporo-occipital visual areas, and some receive direct collicular or retinal inputs. Their cells can be driven by relatively simple visual stimuli, and their cortical projections can regulate local gain and/or functional connectivity between cortical areas involved in object vision and/or visuomotor behaviors (reviews in Baldwin et al., 2017; Gattass et al., 2018; Fan and Bourne, 2023).
MPu is massive in humans (Homman-Ludiye and Bourne, 2019). This nucleus virtually lacks subcortical excitatory inputs (Rovó et al., 2012), but its cortical connections are extremely widespread, reportedly reaching the inferior and ventral temporal, posterior parietal, prefrontal, insular, and cingulate areas (Burton and Jones, 1976; Stanton et al., 1977; Mufson and Mesulam, 1984; Asanuma et al., 1985; Baleydier and Mauguiere, 1985, 1987; Goldman-Rakic and Porrino, 1985; Insausti et al., 1987; Morán et al., 1987; Vogt et al., 1987; Yeterian and Pandya, 1989; Barbas et al., 1991; Baleydier and Morel, 1992; Romanski et al., 1997; Cavada et al., 2000; Buckwalter et al., 2008; Gamberini et al., 2020). Some MPu neurons may selectively respond to emotionally salient visual stimuli (Maior et al., 2010; Nguyen et al., 2013; Van Le et al., 2013), but most show no clear selectivity besides basic spatial tuning, and their activity seems instead related to sensory–motor transformations required for spatial attention and guidance of goal-directed eye (Benevento and Port, 1995; Dominguez-Vargas et al., 2017; Fiebelkorn et al., 2019; Schneider et al., 2023; Kaduk et al., 2024) and hand movements (Acuña et al., 1990; Grieve et al., 2000). Multisensory responses have been also reported (Mathers and Rapisardi, 1973; Vittek et al., 2023).
Because the MPu axons have not been directly visualized and mapped with modern methods (Benevento and Rezak, 1976, Burton and Jones, 1976; Trojanowski and Jacobson, 1976; Baleydier and Mauguiere, 1985; Romanski et al., 1997), current notions on thalamocortical MPu circuits are a composite from fragmentary observations, making it difficult to infer their global layout or functional logic. For example, single cortical tracer injections typically label several MPu cell clusters, which are also the main target of the reciprocal corticothalamic projections. The MPu thalamocortical cell clusters labeled by injections made in separate cortical injections usually remain separate; however, the clusters overlap in part when the two cortical injections are placed in areas that are themselves corticocortically interconnected, even if the areas in question are quite distant from each other (Baleydier and Morel, 1992; Shipp, 2003). Double-retrograde labeling studies concluded that MPu cell populations innervating different areas are separate (Asanuma et al., 1985; Baleydier and Mauguiere, 1985, 1987; Baleydier and Morel, 1992; Hardy and Lynch, 1992, Homman-Ludiye et al., 2020); hence, nonreciprocal interactions between areas via MPu are thought to depend on some corticothalamic axons spreading beyond the borders of their reciprocally projecting thalamocortical cell clusters (Selemon and Goldman-Rakic, 1988; Darian-Smith et al., 1999; Grieve et al., 2000; Shipp, 2003; Cappe et al., 2009).
In marmoset monkeys, we labeled, mapped, and compared the distribution and terminal morphology of the axons originating from different MPu zones. Furthermore, in view of the prevalence of branched thalamocortical axons revealed by modern single-cell labeling studies in rodents (review in Clascá, 2023), we reexamined the existence of MPu cells with branched axons simultaneously innervating two distant areas. Our results reveal the multiarea projection motifs of different MPu domains as well as the relative abundance and terminal arborization patterns of the thalamocortical axons in their target areas. Moreover, we show that MPu contains projection neurons whose branched axons simultaneously innervate prefrontal and parietal or prefrontal and temporal areas.
Materials and Methods
Animals
Marmoset monkeys (Callithrix jacchus) were chosen because of their relevance in neuroscience as small anthropoid primate models (review in Okano, 2021), for which a detailed stereotaxic atlas (Paxinos et al., 2012) and extensive data on corticocortical connections (Majka et al., 2020) are available. The present study reports observations made on a total of 17 adults (18–45 months of age) of both sexes (nine males, eight females) weighing 250–390 g (Table 1). Animals were obtained from the breeding colony at the Primatology Center of the Federal University of Rio Grande do Norte (UFRN Natal). Animals were housed under natural lighting, temperature, and humidity conditions, always in the company of other related conspecifics, with food and water available ad libitum. Experimental procedures involving live animals were all conducted at UFRN and the approved by the Animal Ethics Committee of the UFRN Biosciences Center (CEU–020/2012), in accordance with Brazilian Law and the guidelines published by the NIH Guide for Care and Use of Laboratory Animals (http:www.nap.edu/catalog/12910.html).
Animals and valid tracer injection experiments analyzed in the study
Anesthesia
Animals were first sedated with diazepam (1 mg/kg, i.m.). Anesthesia was then induced by inhalation of a mixture of 4% isoflurane in oxygen and subsequently adjusted (1–2%) to keep the animal areflexic while breathing spontaneously throughout the surgical procedure. The anesthetized animals were placed in a stereotaxic apparatus (Narishige). A water-circulated heating pad (TP-700, Gaymar) at 37°C was used to maintain body temperature. Saline was administered (1.5 cc/h, s.c.) to keep hydration, and the eyes were protected with Vaseline to prevent desiccation. Cardiac and breathing frequency were monitored throughout the surgery with a Vet-DL900 multiparametric touch screen (Delta Life). At the end of the surgical session, the isoflurane was discontinued, and animals recovered spontaneously. Buprenorphine hydrochloride (0.01 mg/kg/12 h, i.m.) was administered for the first 3 postoperative days for analgesia. Animals also received a single dose (0.1 mg/kg, s.c.) of the long-acting veterinary antibiotic Cefovecin sodium. After a survival period of 2 weeks to allow for the axonal transport to occur, animals were killed with a lethal dose of sodium pentobarbital (100 mg/kg, i.p.), before perfusion.
Iontophoretic BDA injections in the thalamus
To selectively label the axons arising from small clusters of thalamic neurons, we iontophoretically microinjected 10 kDa biotinylated dextran amine (BDA; Molecular Probes-Invitrogen; 10% in 0.01 M PB), pH 7.4, through glass micropipettes in the thalamus in 14 animals. To reduce the number of experimental animals, taking advantage of the fact that thalamocortical projections are exclusively unilateral, two injections were delivered, one in the right thalamus (R) and other in the left (L), and each cerebral hemisphere was subsequently analyzed as an independent experiment.
The scalp was cut open at the midline, a small craniotomy was drilled on the parietal bone, and the dura mater was opened. Borosilicate glass micropipettes (FHC; 10 µm internal tip diameter) loaded with BDA were positioned following the stereotaxic coordinates of Paxinos et al. (2012) marmoset brain atlas. Positive pulses (2–4 µA, 7″ on/off cycles, 20 min) were delivered using a Midgard continuous current source (Stoelting). Pipettes were left in place for 5 min after the injection, before removal. Finally, the exposed cortex surface was covered with hemostatic gelfoam, and the bone defect was occluded with acrylic cement. After allowing the animals to fully recover from the anesthesia, they were returned to their housing.
Tracer injections in the cortex
To reveal the position in the thalamus of the neurons innervating specific cortical regions, we injected retrograde tracers in the cortex. For this purpose, Fast Blue (FB; Polysciences, 1% in cacodylate buffer 0.1 M) or cholera toxin B subunit (CTB) conjugated with Alexa Fluor 594 (Life Technologies, Thermo Fisher Scientific, 1.5% in PBS buffer 0.1 M) was used. These tracers were injected, using different 10 µl syringes with ultrathin (36 gauge, 110 µm outer diameter) beveled metal needle tips (NanoFil, World Precision Instruments) mounted on a Narishige micromanipulator. In one experiment (Case 26), we made iontophoretic deposits in the cortex of BDA (as described above) or of low-salt CTB (List Biological Laboratories, 2% in 0,05 M Tris buffer) using glass micropipettes mounted on the stereotaxic micromanipulator and continuous current on/off pulses, 4 µA, for 40 min at each point of injection.
These cortical injections were made through small craniotomies. Since the marmoset cortex provides virtually no sulcal landmarks to locate the intended target areas, stereotaxic coordinates were used to locate the intended injection sites. At each site, the tracer was injected in 2–3 puffs, ∼100 nl each, over 3 min. After the injection was completed, the microsyringe tip was left in place for 5 min before removal, to limit tracer backflow. Several adjacent injections were usually made to impregnate a larger extent of the cortex. In one experiment (Case 26), two FB injections were interspersed with two iontophoretic injections of BDA to obtain simultaneous anterograde and retrograde labeling from the same region of the posterior parietal cortex.
Histological processing of the brain tissue
At the end of the survival period, animals were killed as described and immediately perfused through the left ventricle with saline (0.9%, 5 min), followed by 4% paraformaldehyde in PB (0.1 M, 30 min), pH 7.4, and 4% paraformaldehyde +15% sucrose in PB (0.1 M, 15 min), pH 7.4. The brain was exposed, and the dura mater was removed. The brain was then sectioned into three blocks adjusted to the stereotaxic coronal plane, removed from the skull, photographed, postfixed for 24 h in the last perfusion solution, and finally embedded in 30% sucrose in PB (0.1 M, 4°C).
Six parallel series of 40-µm-thick coronal sections were obtained on a freezing microtome (SM 2400, Leica) and collected in PB (0.1 M, 4°C). One series was used to visualize BDA. In these sections, peroxidase activity was blocked via incubation in 2% oxygen peroxide in PB (0.1 M) for 20 min. The free-floating sections were then incubated with an avidin–biotin–peroxidase (1:100; Vectastain Elite, Vector Laboratories, 12 h, 4°C) and, following several rinses in PB, the bound peroxidase activity was revealed using the glucose oxidase-DAB nickel sulfate-enhanced method (Shu, et al., 1988). The sections were serially mounted onto gelatinized glass slides and lightly counterstained with thionin for cytoarchitectonic reference.
In experiments with cortical injections of FB or CTB-Alexa Fluor 594, two series of sections were mounted immediately after cutting onto gelatinized glass slides and air dried in the dark.
In the experiment that received iontophoretic cortical injections of free CTB, we made it visible by incubating the tissue sections first in rabbit anti-CTB (Sigma-Aldrich #C3062; 1:500), overnight, and subsequently with goat anti-rabbit Alexa Fluor 594 (Thermo Fisher Scientific, #A-11011; 1:200), for 2 h at room temperature. In the same sections, BDA was visualized by incubating first in goat anti-biotin serum (Pierce, #31852, 1:100) for 2 h at room temperature and then with donkey anti-goat serum (Invitrogen, #A110055; 1:200). Multiple PBS rinses were intercalated between the above solutions.
In all cases, for the consistent delineation of nuclei and area borders across the brain, a second series of tissue sections were mounted and stained with cresyl violet (Nissl). Additional parallel series underwent free-floating acetylcholinesterase (AChE) histochemistry (Geneser-Jensen and Blackstad, 1971, modified by Cavada, et al., 1995) or with cytochrome-oxidase (CyO) histochemistry (Wong-Riley, 1979) and were subsequently mounted on gelatin-covered glass slides.
After each histological staining procedure was completed, sections were dehydrated in graded alcohols, defatted in 100% xylene, and coverslipped with DePeX (SERVA Electrophoresis). Additional section series were stored as backup at −20°C in a glycerine 30% and etilenglycol 30% in PB solution.
Analysis of the anterogradely labeled pulvinar axons
BDA labeling was analyzed throughout the whole cerebral hemisphere in 1:6 coronal sections (∼130 sections per hemisphere) using a Nikon Eclipse 600 microscope under 4–40× objectives and bright-field optics. Selected images were acquired using the XY motorized microscope stage (Proscan II, Prior Scientific Instruments) driven by the NIS-Elements image software (v3.22.15, Nikon). The position of each injection site was identified by the presence of a core of densely impregnated neuropil and cell somata. We assume that the anterograde transport of BDA occurred from the neurons within this core zone. To assess the full extent of the injection, we mapped the injection site across serial sections.
For each cerebral hemisphere, the labeled thalamocortical axon arborizations were plotted onto a standard computationally “unfolded” representation of the cortex surface (Fig. 1, based on Paxinos, et al., 2012 atlas). This flatmap is a low-distortion 2D renderization of the cortical hemisphere surface created with the CARET software (http://brainvis.wustl.edu/wiki/index.php/Caret:About) produced from the coronal section diagrams in this atlas. It includes, in addition to area boundaries, the outer contour all coronal section levels shown as winding but roughly parallel lines, one each 500 µm. We interpolated an additional line in between these contours (to match the sampling frequency in our series) and represented along the corresponding lines the location and relative density of the labeled axons observed in each of our sections. As dots representing a radial projection of the labeling onto the pial surface. Different in dot spacing and numbers were used to indicate variations in labeled arborization density. This process was iterated for every BDA-stained section (spaced ∼240 µm) in each experimental case. The resulting maps provide global but detailed snapshots of the labeled thalamocortical axons distribution across the whole cerebral hemisphere, which are extremely useful for comparisons across experiments.
Area localization and projection of coronal section contours on a two-dimensional map of the marmoset cerebral cortex surface. A, Lateral and (B) midsagittal schematic views of the brain. C, Surface of the whole cerebral hemisphere shown as a two-dimensional (2D) computationally “unfolded” reconstruction from the coronal sections of the Paxinos et al. marmoset brain atlas (Paxinos et al., 2012). For cross-reference, the lateral surface of the cerebral hemisphere is shaded in pale yellow, while the medial and ventral surfaces are shaded in gray. To minimize distortions and preserve relative area size, the medial and lateral surfaces of the frontal and occipital poles are split and spread apart (indicated by orange dashed arrows). The section profile of the corpus callosum (CC, brown) is shown in panels B and C. The position of the temporal pole (red star) and frontal pole (green star) are also indicated in the diagrams. The bottom of the lateral fissure and the rhinal sulcus are drawn in thick gray line. The inconstant and shallow temporal and intraparietal sulci are also represented, as thick gray dashed lines. Four coronal section levels are drawn (blue dashed lines) and marked 1–4. Note that, because of the nonlinear distortion applied in the 2D surface, the contour of these coronal sections are projected as curved or wavy lines on the flatmap. 1/2, areas 1 and 2 of the cortex; 3a, area 3a of the cortex (somatosensory); 3b, area 3b of the cortex (somatosensory); 4ab, area 4 of the cortex, parts a and b (primary motor); 4c, area 4c of the cortex (primary motor); 6DC, area 6 of cortex, dorsocaudal part; 6DR, area 6 of the cortex, dorsal and rostral part; 6M, area 6 of the cortex, medial part; 6Va, area 6 of the cortex, ventral part a; 6Vb, area 6 of the cortex, ventral part b; 8aD, area 8a of the cortex, dorsal part; 8aV, area 8a of the cortex, ventral part; 8b, area 8b of the cortex; 8C, area 8 of the cortex, caudal part; 9, area 9 of the cortex; 10, area 10 of the cortex; 11, area 11 of the cortex; 13a, area 13a of the cortex; 13b, area 13b of the cortex; 13L, area 13 of the cortex, lateral part; 13M, area 13 of the cortex, medial part; 19 DI, area 19 of the cortex, dorsointermediate part; 19M, area 19 of the cortex, medial part; 23a, area 23a of the cortex; 23b, area 23b of the cortex; 23V, area 23 of the cortex, ventral part; 24a, area 24a of the cortex; 24b, area 24b of the cortex; 24c, area 24c of the cortex; 24d, area 24d of the cortex; 25, area 25 of the cortex; 29a–c, area 29a–c of the cortex; 29d, area 29d of the cortex; 30, area 30 of the cortex; 31, area 31 of the cortex; 32, area 32 of the cortex; 32V, area 32 of the cortex, ventral part; 35, area 35 of the cortex; 36, area 36 of the cortex; 45, area 45 of the cortex; 46D, area 46 of the cortex, dorsal part; 46V, area 46 of the cortex, ventral part; 47L, area 47 of the cortex, lateral part; 47M, area 47 of the cortex, medial part; 47O, area 47 of the cortex, orbital part; ACo, anterior cortical amygdaloid nucleus; AI, agranular insular cortex; AIP, anterior intraparietal area of cortex; AO, anterior olfactory nucleus; APir, amygdalopiriform transition area; AuA1, auditory cortex, primary area; AuCBP, auditory cortex, caudal parabelt area; AuCM, auditory cortex, caudomedial area; AuML, auditory cortex, middle lateral area; AuR, auditory cortex, rostral area; AuRTL, auditory cortex, retrotemporal lateral area; CC, corpus callosum; DI, dysgranular insular cortex; Ent, entorhinal cortex; FST, fundus of superior temporal sulcus area of cortex; GI, granular insular cortex; Gu, gustatory cortex; IPro, insular proisocortex; LIP, lateral intraparietal area of the cortex; MIP, medial intraparietal area of the cortex; MST, medial superior temporal area of the cortex; OPAI, orbital periallocortex; OPro, orbital proisocortex; OPt, occipitoparietal transitional area of the cortex; PaIL, parainsular cortex, lateral part; PaIM, parainsular cortex, medial part; PE, parietal area PE; PEC, parietal area PE, caudal part; PF, parietal area PF; PFG, parietal area PFG; PG, parietal area PG; PGa, parietal area PGa; PGM, parietal area PG, medial part; Pir, piriform cortex; ProM, proisocortical motor region (precentral opercular cortex); ProSt, prostriate area; ReI, retroinsular area; S2E, secondary somatosensory cortex, external part; S2I, secondary somatosensory cortex, internal part; S2PR, secondary somatosensory cortex, parietal rostral area; S2PV, secondary somatosensory cortex, parietal ventral area; TE1, temporal area TE1 (inferior temporal cortex); TE2, temporal area TE2 (inferior temporal cortex); TE3, temporal area TE3 (inferior temporal cortex); TEO, temporal area TE, occipital part; TF, temporal area TF; TFO, temporal area TF, occipital part; TH, temporal area TH; TL, temporal area TL; TLO, temporal area TL, occipital part; TPO, temporo-parieto-occipital association area (superior temporal polysensory cortex); TPPro, temporopolar proisocortex; TPt, temporoparietal transitional area; STR, superior temporal rostral area; V1, primary visual cortex; V2, Visual Area 2; V3(VLP), Visual Area 3 (ventrolateral posterior area); V3A(DA), Visual Area 3A (dorsoanterior area); V4(VLA), Visual Area 4 (ventrolateral anterior area); V4T(MTC); Visual Area 4, transitional part; V5(MT), Visual Area 5 (middle temporal area); V6(DM), Visual Area 6 (dorsomedial area); V6A(PPM), Visual Area 6A (posterior parietal medial area); VIP, ventral intraparietal area of the cortex.
Besides, labeled axons from representative coronal sections were drawn camera lucida-like on microscope images acquired through 10× objectives using the Canvas X vectorial drawing software (ACDSee).
Hierarchical cluster analysis of the labeling patterns
As an unsupervised line of evidence to delineate nuclei boundaries within the pulvinar complex based on the similarity of thalamocortical projection motifs, we performed a hierarchical clustering analysis of the labeled axon distribution across cortical areas in each one of our 21 valid BDA micropopulation injection cases. For each experiment, we defined five integer levels of axon labeling abundance in each target area, from 0 (no labeling) to 4 (maximal labeling density). To normalize the intensity of labeling across experiments, we rated the labeling abundance in each area in comparison with that of all other areas labeled in the same experiment. Then, we compared among experiments the dissimilarity between these numerical values using the ClustVis web-based tool for analysis of high-dimensional biological data using Euclidean as the distance type and Average as the linkage type (Metsalu and Vilo, 2015).
Retrograde labeling analysis
Retrograde fluorescent labeling in the thalamus was examined in a complete series of sections. To prevent biases due to fluorescence fading, we conducted this analysis on TIFF panoramic (tiled composite) color images taken through 10× objectives and the appropriate filter cubes with a Nikon Eclipse microscope fitted with a DMX 1200 camera controlled by the NIS image software (Nikon). Nuclei boundaries and labeled cell somata were plotted on the images using Canvas X. We mostly followed Paxinos et al. (2012) brain atlas for nuclei delineation and nomenclature. Images from adjacent Nissl-, AChE-, or CyO-stained sections were routinely overlaid on the fluorescence images to aid in nuclei delineation.
In addition, selected regions of interest were examined using a Spectral Leica TCS SP5 confocal microscope by sequentially applying argon (488 nm) or diode-pumped solid–state (561 nm) laser lines to ensure complete channel separation. The tissue was imaged using 20 or 63× 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.
Measurement of thalamocortical axon varicosity laminar distribution and size
The Golgi-like quality of the metal-enhanced BDA labeling allowed a sharp visualization of axonal varicosities (putative synaptic boutons). To quantitatively estimate the laminar distribution of these varicosities, we applied the optical dissector stereological method (West et al., 1991) in three cortical areas. The labeling in each area in three different experimental cases was analyzed. Varicosities were examined using a 100× oil immersion plan-apochromatic objective (1.4 NA) mounted on an Olympus BX61 microscope equipped with a Proscan II stage and a Heidenham MT12 microcator controlled by the NewCAST stereology software (Visiopharm). First, we defined a 200-µm-wide columnar sampling area extended from the pial surface to the white matter. The sampling columns were placed on cortical regions containing the heaviest labeling. Columns were then segmented in 20 equal-sized domains (“bins”). The varicosity number was separately estimated for each bin, and with that data we also calculated the varicosity density per cortical layer, which were made visible by the thionin counterstain.
In addition, we estimated axonal varicosity sizes measuring their maximum projection area using Canvas X to measure camera lucida outlines of the boutons drawn at 1,000×. We measured a minimum of 300 boutons in each layer and cortical region, in three different animals. We conducted statistical analysis of the bouton size area using the statistical software SPSS 20 (IBM). The data obtained in each area and layer were considered as independent samples. First the normality of each sample was tested with the Shapiro–Wilk test. As not all samples were normally distributed, we compared the medians in pairs of samples using the Mann–Whitney U test. We also compared the distribution of the populations conducting a two-sample Kolmogorov–Smirnov test.
Results
Cyto- and chemoarchitectonic delineation of the pulvinar complex
The pulvinar complex occupies a large portion of the marmoset thalamus. It extends anteroposteriorly from interaural coronal plane (AP) +4.4 to +1.6 mm. As it is typical of anthropoid primates, the dorsal and caudal surfaces of the pulvinar complex are directly covered by the pia mater, as it bulges dorsomedially in the transverse fissure and caudally on the quadrigeminal cistern. Across these subarachnoid spaces, the complex is wrapped by the fimbria-fornix bundle.
The four major pulvinar nuclei (inferior IPu; lateral, LPu; medial, MPu; and anterior APu) are recognizable in marmosets (Fig. 2; Paxinos et al., 2012). Cytoarchitectonic and neurochemical differences allow a consistent delineation of some of these nuclei, while other nuclei borders are histologically ambiguous; for these, we rely mainly on differences in connectivity that will be discussed later.
Cyto- and chemoarchitectonic parcellation of the pulvinar complex and surrounding nuclei of the marmoset thalamus. Three representative coronal section levels are shown. Adjacent 40-µm-thick sections stained with either cresyl violet (A, D, G), CyO histochemistry (B, E, H), or AChE histochemistry (C, F, I) are shown. Thalamic nuclei are labeled on panels A–C. In addition, some fiber tracts important for nuclei delineation are labeled (in blue lettering) in all panels. Scale bar, 1 mm. APu, anterior pulvinar nucleus; ar, external medullary lamina, auditory radiation; bic, brachium of the inferior colliculus; bsc, brachium of the superior colliculus; CL, central lateral nucleus; ctt, corticotectal fibers; IPu, inferior pulvinar; IPuCL, inferior pulvinar, caudolateral part; IPuCM, inferior pulvinar, caudomedial part; IPuM, inferior pulvinar, medial part, IPuP, inferior pulvinar, posterior part; LD, laterodorsal nucleus; LGN; lateral geniculate nucleus; Lim, limitans nucleus, LP, lateral posterior nucleus; LPu, lateral pulvinar nucleus; MDc, mediodorsal nucleus, caudal pole; MGD, medial geniculate nucleus, dorsal part; MGV, medial geniculate nucleus, ventral part ml, medial lemniscus; MPuR, medial pulvinar nucleus, rostral division; MPuC, medial pulvinar nucleus, caudal division; Po/Sg, posterior/suprageniculate nucleus.
As previously described (Cusick et al., 1993; Stepiniewska and Kaas, 1997), IPu and its subnuclei can be delineated based on differences in CyO and AChE stains. A patch of heavy CyO staining identifies the medial caudal subnucleus (IPuCM) from the adjacent medial (IPuM) and caudolateral (IPuCL) subnuclei (Fig. 2E–H). Likewise, IPuCL can be delineated from IPuCM based on its heavier AChE staining (Fig. 2F–I). Between AP levels +4.30 to +3.20 mm, IPuCL borders the lateral geniculate nucleus; caudal to this level, IPuCL is laterally bordered by LPu. Studies in Old- and New-World monkeys have shown that the IPu subnuclei have each specific interconnections with subcortical and extrastriate cortical visual areas (Adams et al., 2000; Stepiniewska et al., 2000; Kaas and Lyon, 2007; Lyon et al., 2010; Arafune-Mishima et al., 2020) and that some subnuclei are split in two by the myelinated fiber bundles of the corticotectal tract (ctt; Fig. 2). The latter observation reveals that ctt myelin bundles are not a reliable indication of functional nuclei borders in the pulvinar.
The delineation of other pulvinar divisions in marmosets is less well established (Gutierrez et al., 2000; Paxinos et al., 2012; Baldwin et al., 2017; Homman-Ludiye et al., 2018; Chan et al., 2021). We identified LPu as a wide band of tissue with high CyO and AChE activity that extends along the lateral border of the pulvinar (Fig. 2). The central/ventral portion of this band is pierced by numerous myelinated ctt fiber bundles, which confer it a discontinuous, veined appearance. The zone pierced by the ctt bundles has been explicitly or tacitly equated to LPu in marmosets (Dick et al., 1991; Paxinos et al., 2012; Homman-Ludiye and Bourne, 2019; Chan et al., 2021). However, the high CyO and AChE region extends for over a millimeter dorsally to the ctt bundles, up to the dorsal surface of the thalamus (Fig. 2B,C,E,F; see also the AChE images in Paxinos et al., 2012 atlas). This zone has been labeled by some authors as LPu or a separate “dorsomedial” LPu subnucleus (Kaas and Lyon, 2007; Paxinos et al., 2012; Baldwin et al., 2017), while other authors included it in MPu (Dick et al., 1991, Homman-Ludiye and Bourne, 2019; Homman-Ludiye et al., 2020; Chan et al., 2021). At about +4.0 mm from the midline, the CyO and AChE staining decrease in the lateral-to-medial direction. This level roughly coincides with an abrupt shift in connectivity detected by the axonal tracing experiments which will be discussed in the following sections. Thus, we place the MPu/LPu border at this level, which is about a millimeter more medially than in the Paxinos et al. (2012) atlas. Along the anteroposterior dimension, MPu extends from the junction with the central lateral and lateral dorsal nuclei at its rostral end (∼AP level +4.3) to the caudal pole of the pulvinar. The ventral limit of MPu is constituted by APu at rostral levels and by the bsc and IPu at caudal levels. In Nissl-stained sections, MPu is basically homogeneous in cell size and packaging. The ventral part of the nucleus shows higher CyO and AChE staining (Fig. 2) and has been shown to have relatively high levels of calbindin expression (Paxinos et al., 2012; Homman-Ludiye and Bourne, 2019). However, since the tracing experiments discussed in the following sections revealed no significant differences in connectivity MPu, we do not subdivide the nucleus along its dorsoventral axis. In contrast, the tracing experiments show that connections change abruptly along the anteroposterior axis. Hence, we distinguish within MPu a rostral (MPuR) and a caudal (MPuC) subnuclei (Fig. 2A,D,G).
Our delineation of the border between APu and MPu in Figure 2 is tentative, given the absence of histological landmarks and limited sampling of this region by our connection-tracing experiments. Thus, we largely follow Paxinos et al. (2012) diagrams, although we include in APu the portion of the “posterior nucleus that is situated above the dorsal border of the ventroposterior complex, as recently proposed for cebus monkeys (Mayer et al., 2019). In addition, we apply the denomination “suprageniculate” (Sg) nucleus to the joint caudal pole of the ventral posterior, posterior and ventroposterior inferior nuclei, following in this regard the criterion of previous studies in macaque and cebus (Olszewski, 1952; Burton and Jones, 1976; Jones, 2007a; Mayer et al., 2019). This Sg nucleus is reportedly involved in thermalgesic sensory relay (Apkarian and Hodge, 1989; Craig et al., 1994; Davidson et al., 2008) and it selectively innervates somatosensory areas S2 and granular insular (Burton and Jones, 1976; Mufson and Mesulam, 1984). As for the small region labeled “suprageniculate” in Paxinos et al. (2012) atlas, we simply include it as part of the dorsal division of the medial geniculate nucleus.
From ∼AP +4.30 to +3.0 mm, the pulvinar complex is medially bordered by the pretectotalamic fiber lamina. This lamina contains scattered groups of elongated, relatively small thalamic cells interspersed among the fibers, which are collectively referred to as nucleus limitans (Lim). Caudally to AP +3.0 mm, the brachium of the superior colliculus (bsc) becomes the ventromedial limit of the pulvinar complex.
BDA labeling of the thalamocortical axons arising from discrete pulvinar domains
To directly visualize all axons arising from specific pulvinar domains, we made focal, stereotaxically guided iontophoretic microinjections of BDA. On the tissue sections, BDA deposits typically appeared as a dense black clump of neuronal bodies, neurites, and debris (Fig. 3A). We assume that the axonal transport occurred from the neurons situated within and immediately around this cluster. For each experiment, we reconstructed the extent of each deposit across serial sections. Thalamic nuclei boundaries were delineated using as reference the light thionin counterstain as well as the images from parallel alternate AChE-, Nissl-, or CyO-stained sections (Fig. 4A). To allow a direct comparison between experiments, we also plotted the position of each deposit on a standard stereotaxically adjusted series of coronal section diagrams (Fig. 4B).
A small BDA deposit in the pulvinar (Case 4L) labels thalamocortical arborizations in the prefrontal, posterior parietal, and temporal association cortices. A, Coronal section image showing the center of the BDA deposit. B, C, Low-magnification views of the anterograde labeling. In panel B labeled axon arborizations are visible in a coronal section of dorsolateral prefrontal areas. An asterisk (magenta) indicates a histological artifact. Panel C illustrates a coronal brain section showing simultaneous axon labeling in parietal (area PFG top) and temporal (area TE3 bottom) cortices. D–F, Representative higher-magnification columnar samples taken from dorsolateral prefrontal area 8aV (panel D), inferior parietal area PFG (panel E) or inferior temporal area TE3 (panel F). The location of the sampled columns is framed with magenta corners in panels B and C. G, Detail of the subpial plexus in area PFG Layer 1. H, High-magnification detail of labeled axons traversing the TRN and giving off a varicose branch (arrows). I, J, High-magnification images (taken at the sites indicated by blue arrows in panel B) show labeled axons branching in the subcortical white matter. Note that branches (blue arrows) emerge at roughly orthogonal angles toward the cortex (which is situated to the right) from axonal trunks that continue into the white matter. Scale bars: A–C, 1 mm; D–F, 100 µm; G, 250 µm; H–J, 25 µm. See caption to Figure 1 for cortical area abbreviations.
Iontophoretic BDA injection cases analyzed in this study. A, Bright-field images of coronal sections containing the center of the BDA injection site in all valid cases. Case numbers are indicated in the bottom-left corner of each image, and the coronal level in the bottom-right corner. For reference, pulvinar nuclei are identified and color-shaded as in panel B. Light thionin counterstain. Scale bar, 1 mm. CM, centromedian nucleus; Po/VP, posterior nucleus/ventral posterior nucleus; Spf, subparafascicular nucleus. For other abbreviations, see Figure 2. B, For direct comparison, all injection sites are plotted onto a series of standard coronal section diagrams of the marmoset pulvinar (modified from Paxinos et al., 2012). The various subnuclei are colored as in A. Diagrams are aligned to preserve the dorsoventral (DV) level between them. Coronal AP levels are indicated under each section.
The BDA injections labeled the thalamocortical axons with Golgi-like detail, up to their terminal arborizations. At the sampling frequency used in our analysis (40 µm sections separated by 200 µm gaps), the trajectory of the thick bundle of labeled axons arising from each injection site could be readily followed, from section to section, along their path in the white matter up to the terminal arborization regions. After leaving the injection site, the axons traversed the thalamic reticular nucleus (TRN) between planes +3 to +5 mm, at variable points, depending on the position of the BDA deposits. The axons gave off varicose and relatively simple branches in the TRN (Fig. 3H) and continued into the internal capsule and corona radiata. As they entered the cortical white matter, axons made a 90° turn to run in trajectories parallel to the cortex. Along these trajectories, axons emitted branches toward the cortical gray (Fig. 3I,J). The terminal intracortical arborizations were typically multifocal, waxing and waning in “column-like” fashion (Fig. 3B–F). In most areas, terminal plexuses were in L3b–L4 and superficial L1 (Fig. 3D,E,G).
Fifteen injection experiments, which together sampled most of MPu, were analyzed. The labeling patterns produced by injections situated rostrally or caudally to AP +3.0 mm were strikingly different. For this reason, we propose to consider these two portions of MPu as separate subnuclei, MPuR and MPuC. For brevity, we report in detail the labeling observed in a representative case for each of these two subnuclei. For the remaining experiments, we focus on the differences of their labeling pattern with that observed in the representative case.
As noted in the previous section, the delineation of some MPu borders remained unclear. Thus, we examined six further cases whose BDA injections were situated at or near the presumed location (Paxinos et al., 2012) of the borders of MPu with LPu (four cases), with APu (one case), or with Lim (one case).
BDA injections in the rostral subnucleus of the medial pulvinar (MPuR)
Six BDA deposits were located in MPuR. These experiments characteristically produced the simultaneous labeling of axonal arborizations at four very separate points of the cerebral hemisphere, namely, the inferotemporal, ventrolateral prefrontal, posterior parietal, and ventral temporal/rhinal cortices. Beyond these main foci, some labeled axon branches were often observed in the anterior cingulate, orbital, insular, and/or some of the anterior extrastriate visual areas.
Case 11R is described as representative for the MPuR experiments (Fig. 5). The BDA injection in this case involved a small central MPuR zone, centered at ∼AP +3.15 mm (Figs. 3A,B, 5E). Labeled axons exited the injection site in a rostral and lateral direction to cross the TRN at ∼AP +5.0 mm (Fig. 5D). The labeled axons continued along the superior and posterior bundles of the internal capsule and entered in the subcortical white matter under the posterior parietal cortex. There, numerous orthogonal branching points were observed in the white matter; above them, dense intracortical arborizations were present in some inferior parietal cortex areas (Fig. 5D,E). Axon trunks in the white matter continued in two separate tracts; one extended ventrally to reach the inferior and the ventral temporal cortices (Fig. 5C–F), while other continued in a rostral direction along the extreme capsule to reach the ventral prefrontal cortex (Fig. 5A–C). In between these distant cortical territories, no labeled axons entered the cortex.
A BDA deposit in MPuR labels widely separate thalamocortical axon arborization patches across the cerebral hemisphere. A–F, Camera lucida drawings of sections containing labeled axons at six coronal levels of the cerebral hemisphere from rostral (A) to caudal (F) in Case 11R. The corresponding coronal levels anterior (+) to the interaural plane are indicated (blue italic letters). Black arrowheads indicate area borders, according to Paxinos et al. (2012). The BDA injection site is indicated by a pair of green arrowheads in panel E. Labeled axons are shown as magenta lines. Both the labeled terminal arborization plexuses and white matter tracts are drawn. Note that, in addition to the cortex, terminal plexuses were present in the lateral amygdaloid nucleus (green arrow in C) and TRN (green arrowhead in D). Scale bar, 2 mm. G, Representation of the labeled thalamocortical arborizations onto the standard 2D cortex map (see Materials and Methods). For cross-reference, line contours corresponding to the coronal sections shown in panels A–F are indicated by blue dashed lines on the map and highlighted by arrowheads on both ends. The labeling in a complete series of sections, spaced from each other ∼200 µm, is represented. Areas containing the anterogradely labeled axonal arborizations are labeled using the abbreviations listed in Figure 1. H, Camera lucida drawings showing labeled axonal plexuses in opercular premotor area ProM, in posterior parietal area PG, in inferotemporal areas Te3, or in ventromedial temporal areas and TH and TF. In the latter, an arrowhead points to the shallow rhinal fissure. The pial surface is always to the top. The corresponding sampling points are indicated by asterisks in panels B–E. Scale bar, 500 µm.
Within the target areas, labeled axon arborizations usually concentrated heavily in small column-like domains largely sparing adjacent zones. Hence, in the cortical map visualizations, the labeling often shows a discontinuous appearance, concentrated in multiple patch-like or stripe-like domains several hundred microns-wide, separated by weakly labeled or unlabeled zones (Fig. 5G). The heaviest labeling was in the inferior temporal cortex area Te3 and adjacent zones of area Te2. Labeled arborizations were also abundant in ventral temporal cortex areas TL, TLO, and TH, with some axons reaching perirhinal areas 36 and 35. In the ventral prefrontal cortex, the heaviest labeling involved areas 45 and 47L and the proisocortical motor area (ProM). In the inferior parietal cortex, areas PF and PG and lateral intraparietal area (LIP) contained the heaviest labeling. The inferior temporal and inferior parietal arborizations extended into caudally adjacent zones of areas V4(VLA), V4T (MTC), or V3A (DA), respectively. The laminar distribution of the thalamocortical terminal arborizations was similar across areas; Layers L3b and L1 contained the densest arborizations (Fig. 5H).
In addition, sparse thalamostriatal axon branches were labeled in the caudal putamen and the tail of the caudate nucleus, as well in the dorsalmost portion of the lateral amygdaloid nucleus (Fig. 5C, green arrow).
A multifocal pattern of labeling essentially equivalent to that in case 11R was observed in the remaining five experiments that received BDA injections in MPuR, even though these injections involved each a different portion of the subnucleus (Fig. 6). The most noticeable variation among the experiments was a relatively heavier parietal and weaker temporal labeling in the most anteriorly located injection (Case 4L) compared with more caudal MPuR injections (Cases 13R, 12R, 11R, and 13L).
Comparison of the thalamocortical labeling pattern among BDA injections in different MPuR zones. A–F, Representation of the labeled thalamocortical axons plotted on standard cortex flatmaps in six different injection cases in MPuR. To ease comparisons, the flatmap in Case 11R is included. Each map represents the labeling observed in a complete series of sections (1:6). Areas containing the arborizations are labeled using the abbreviations listed in Figure 1. Note that despite differences in the position of the tracer deposits, the labeled axons distribution is remarkably similar across cases. G, Diagram comparing the position of the six BDA injection sites; conventions as in Figure 4B.
In thalamocortical axons, varicosities are the predominant location of synapses (White et al., 2004; Rodríguez-Moreno et al., 2020). These varicosities may be unevenly distributed along terminal axon branches; thus, their relative numbers might deviate from the apparent density of stained axon filaments (Ohno et al., 2012). For this reason, we directly counted varicosities as particles using nonbiased stereological sampling methods in the ventral frontal (45/47L), inferior parietal (PFG, PG), or inferior temporal (Te3) areas (Fig. 7a,b). The analysis showed that in these three cortical regions, ∼70% of MPuR axon varicosities are located in L4–L3, over 40% in L3b. Despite the relatively small volume of L1, varicosities in this layer were ∼15% of the total.
Quantitative analysis of MPuR axon varicosities laminar distribution and size in the ventral prefrontal, posterior parietal, or inferior temporal cortices. A, Stereological estimation of axon varicosities absolute number. Varicosities were counted along a transect of 20 bins from pia to white matter. Measurements were made in in high-density projection zones of the inferior temporal area Te3, inferior parietal areas PG and PFG, and ventrolateral prefrontal area 47L. For each cortical area, varicosities were counted in three sequential coronal sections from three different injection experiments, without normalization between experiments (see Materials and Methods). For reference, representative cortical column images of the labeled axons are illustrated. B, Comparison of the mean distributions (in percent) of varicosities per cortical layer. Varicosity number values obtained from the stereological analysis presented in panel A were compared here with the cytoarchitectonic laminar boundaries. Area color code as in panel A. C, Representative high-magnification images of labeled varicose axon branches in Layer 3b in each of the representative cortical areas. D–E, Comparison of varicosity size (measured as maximal 2D projection area, in µm2) distribution frequencies in MPuR axons labeled in L1 (D) or Layer 3b (panel E) in each the above three cortical areas. Kolmogorov–Smirnov test; *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001). F, G, Box plot charts comparing varicosity size distributions in L1 (F) or L3b (G). Median values are indicated by a horizontal black line. Atypic values are shown as circles. Statistically significant differences are indicated (Mann–Whitney test (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001).
The observed laminar distribution of MPuR axon varicosities is consistent with previous general descriptions of pulvinocortical axon arborizations in macaques and other monkeys (Benevento and Rezak, 1976; Burton and Jones, 1976; Trojanowski and Jacobson, 1976; Baleydier and Mauguiere, 1985; Romanski et al., 1997; Homman-Ludiye et al., 2020). By predominantly targeting L3–L4 and L1, MPu axon synapses might have preferential access to the pyramidal cells that originate corticocortical connections, as well as to the apical dendrites of some subcortically projecting pyramidal cell populations (Felleman and Van Essen, 1991; Chomsung et al., 2010; Markov et al., 2011).
In glutamatergic axons, differences in synaptic bouton size correlate largely with the variable functional impact of its synapses on the on postsynaptic cell responses (Sherman, 2016). For this reason, we measured and compared the size (maximal projection area) of axonal varicosities (putative boutons) in the labeled MPuR axons innervating the prefrontal, parietal, or temporal cortices. We found that varicosities have basically similar size range and distribution in the three areas; differences were minimal, although some were found to be statistically significant, probably due to the large sample size (Fig. 7D–G). For comparison, the varicosities in the cerebral cortex were twice as large as those labeled in TRN in the same experiments (mean, 0.56 µm ± 0.20 SD; see Fig. 3H). Moreover, the finding that MPuR axon varicosities are relatively large is consistent with observations after BDA labeling of pulvinar axons in a macaque (Rockland et al., 1999) or a galago (Marion et al., 2013). Taken together, the heavy convergence on specific layers/columns and relatively large size of axonal varicosities is consistent with functional reports that pulvinar signals powerfully impact the activity of their target cortical circuits (Wilke et al., 2010, 2013; Purushothaman et al., 2012; Zhou et al., 2016; Fiebelkorn and Kastner, 2019).
BDA injections in the caudal subnucleus of the medial pulvinar nucleus (MPuC)
Nine valid BDA deposits were in MPuC. All these experiments labeled thalamocortical axons predominantly or exclusively in the inferior and ventral temporal cortex areas. Some, but not all of them, produced sparse labeling of ventral prefrontal areas, yet none labeled axons in the inferior parietal or cingulate areas.
Case 22R (Fig. 8) is described as representative of MPuC injections. In this case, the BDA deposit involved a small dorsoventrally elongated zone, with its center about coronal plane AP +2.15 mm (Figs. 4A,B, 8F). Labeled axons leaving the injection site were visible as they traversed the TRN near the caudal end of this nucleus, about plane AP +3.0 mm, and gave off there some varicose branches (Fig. 8E). Subsequently, axons turned ventrally into the posterior thalamic radiation, extended along the white matter stem of the temporal cortex and finally arborized abundantly in ventral temporal areas TF, Te1, and A36. Less profuse axon branches extended into adjacent inferior and ventral temporal areas TE2, TEO, TFO, and A35. Varicose branches were observed mainly in L3b and L1. No labeling was found in the inferior parietal or frontal areas. Sparse varicose branches were present in the tail of the putamen, as well as in the lateral nucleus of the amygdala (data not shown).
A BDA deposit in MPuC labels thalamocortical axon arborizations in the ventral and inferior temporal association cortices. A–F, Camera lucida drawings of sections containing labeled axons at six coronal levels of the cerebral hemisphere from rostral (A) to caudal (F) after a BDA injection in MPuC (Case 22R). Graphic conventions as in Figure 4. The BDA deposit is indicated by a green arrow in panel F. Scale bar, 2 mm. G, Representation of anterogradely labeled terminal axon arborizations onto the standard 2D map of the cortical surface. The labeling in a complete series of sections is represented. Areas containing anterogradely labeled axonal arborizations are identified using the abbreviations listed in Figure 1. H, Camera lucida-drawn samples of the axonal labeling in perirhinal areas TF and 36. For cross-reference, the location of these columns is indicated (asterisks) in B and C. Other graphic conventions as in Figure 4. Scale bar, 500 µm.
Virtually identical patterns of labeling were observed in three other MPuC injection experiments which were situated posteriorly to plane AP +2.5 mm (Cases 21R, 22L, 21L; Fig. 9F–H⇓). The main difference with Case 22R (Fig. 9I) was a heavier labeling of A35 in these three cases. In contrast, two injections placed in MPuC anteriorly to plane AP +2.5 mm (Cases 17L and 28L, Fig. 9B,C) yielded a substantially different pattern, with dense labeled arborizations in inferior temporal area Te3 as well as in ventral temporal areas 35, 36, TL, and TLO (Fig. 10).
Comparison of the thalamocortical labeling pattern among BDA injections in different MPuC zones. A–I, Representation on standard cortex maps of the thalamocortical arborizations (magenta dots) labeled in nine different BDA injection cases. The labeling in a complete series of sections in each experiment is represented. Case 22R (Fig. 8) is included for direct comparison. The areas containing anterogradely labeled axonal arborizations are identified using the abbreviations in Figure 1. J, Diagram comparing the position of the BDA injection sites.
Direct comparison of the distribution across temporal cortex areas of the axons labeled by injections in different points of MPuR or MPuC. A, Detailed representation of the marmoset temporal cortex showing the relative size and position of its cortical areas. The line contour at the bottom indicates the lateral border of the hippocampal formation (which is not shown). The bottom of the lateral (ls), rhinal (rs), and the collateral (cs) sulci are indicated (blue lines). The cortex shaded in gray represents regions situated in the lips of the lateral sulcus or in the ventral aspect of the temporal lobe, while that shaded in white represent regions on the lateral surface of the hemisphere. Area name abbreviations as in Figure 1. B, Comparison of the labeling produced by a typical injection in MpuC versus one in LPu (Case 17L; see below). C, Comparison of the labeling distributions produced by injections in MPuR and MPuC. D–F, Comparison of labeling distribution among injections involving different zones of MPuC. Panel D compares an injection placed anteriorly versus one placed caudally. Panel E compares an injection placed laterally versus one placed medially. Panel F compares an injection placed dorsally versus one placed ventrally.
Two further injections in MPuC were also located at AP +2.5 mm. However, the two injections occupied the ventral tip of the nucleus, near its junction with Sg and IPuP (Cases 06R and 31L; Fig. 9D,E). The two experiments produced a pattern of labeling intermediate between that of MPuR and MPuC: heavy labeling of inferior temporal areas Te3 and Te2 perirhinal and ventral temporal areas, with some labeling of the ventrolateral prefrontal cortex (as MPuR injections), but no labeling of the inferior parietal cortex (as MPuC injections, Fig. 10). Overall, the experiments demonstrate that this ventral region shares the same basic projection motif than other MPu zones, which is markedly different from that of the adjacent inferior pulvinar nuclei (Baldwin et al., 2017; Arafune-Mishima et al., 2020).
Finally, the injection in Case 19R (Fig. 9A) involved the most medial tip of MPuC at about plane AP +2.7 mm. Labeled axons in this experiment targeted temporal pole areas Te1, TPo, and STR. Note that areas TPo and STR were not labeled by other medial pulvinar injections (Fig. 10). Likewise, Case 19 injection labeled ventromedial prefrontal areas 14, 32, and 25, which were never labeled in other medial pulvinar injection experiments.
Most MPuC, as well as some MpuR injections (Figs. 5C, 13A) labeled isolated varicose axon arborizations in the dorsal portion of the lateral amygdaloid nucleus, consistent observations in macaque studies (Jones and Burton, 1976; Romanski et al., 1997; Elorette et al., 2018).
BDA injections in other pulvinar nuclei surrounding MPu
As noted earlier, a border between MPu and dorsal LPu is not evident on cyto- and chemoarchitectural grounds. Interestingly, in three injection experiments which probed lateral portions of the region that Paxinos et al. (2012) identified as MPu, we obtained patterns of labeling that were typical of LPu, as described in the literature (Baldwin et al., 2017; Fan and Bourne, 2023). In these experiments, the labeled axon arborizations were concentrated in prestriate visual areas but spared the inferior and ventral temporal association cortex and the prefrontal cortex (Figs. 11, 12B–D). Hence, these experiments reveal a sharp connectivity-defined border between MPu and LPu at about parasagittal plane 4 mm.
A deposit in LPu labels multiple thalamocortical axon arborizations patches in posterior temporoparietal and occipital extrastriate visual areas. A–F, Representative sections showing the labeled axons at six coronal levels of the cerebral hemisphere from rostral (A) to caudal (F) after a BDA injection in LPu near the border with MPuR (Case 20 L). Graphic conventions as in Figure 4. Scale bar, 2 mm. G, Representation of anterogradely labeled axon arborizations onto the standard 2D map of the marmoset cortex. A complete series of sections is represented. Areas abbreviations listed in Figure 1. H, Camera lucida drawing of labeled axonal plexuses in areas V2 and V6. The two sampling sites are indicated by asterisks in panels E and F. Scale bar, 500 µm. Other graphic conventions as in Figure 4.
Comparison of thalamocortical labeling patterns among BDA deposits in different zones of LPu. A–D, Representation on standard 2D marmoset cortex maps of the thalamocortical arborizations labeled (magenta dots) in four different LPu injection cases. For direct comparison, the injection in case 20L (Fig. 11) is included. Labeled areas are marked using the abbreviations in the legend to Figure 1. E, Diagram comparing the position of BDA injection sites in LPu.
We describe Case 20L as exemplary of the injections situated laterally to this border. In this animal, the BDA deposit impregnated a vertically elongated zone at AP +3.10 mm and parasagittal ∼4.10 mm (Figs. 4A,B, 11C). The labeled axons exiting the injection site extended laterally and caudally to cross the TRN between AP +2.8 and +4.3. In the cortex, the densest arborizations were labeled in extrastriate visual areas V6(DM), 19, and V4(VLA). These arborizations concentrated in L3b and displayed a marked waxing and waning pattern, suggestive of a preference for specific columnar domains. Additional labeled axonal arborizations were observed in regions of area “V3 ventrolateral posterior” V3(VLP), area V2, and the “middle temporal crescent” area V4T(MTC). Sparser axons were present in the medial intraparietal area, LIP, the “fundus of superior temporal sulcus” area (FST), the occipitoparietal transition (OPt) area, the middle temporal area (V5/MT), as well as caudal parts of areas Te3 and PG (Fig. 10B). Some varicose branches were visible in the tail of the caudate nucleus.
A second experiment (Case 14L; Fig. 12B) was situated at the same parasagittal level, although at a slightly more rostral and ventral zone. The labeling pattern in this experiment was essentially the same as in Case 20L, albeit the labeled arborizations were more abundant in areas V4T(MTC), FST, OPt and V5/MT, and PG but less so in V4 (VLA), V6DM, 19, and V2.
A third injection (Case 14R) was revealing. The BDA deposit in this experiment (AP +2.85) was slightly more medial and dorsally extensive than the previous two and thus spread across the “invisible” border between MPu and LPu. As in the two previous LPu cases, labeling in this experiment was heavy in extrastriate areas V4(VLA), V3(VLP), and 19. However, Case 14R showed, in addition, abundant labeling of inferior temporal (Te3, TeO) and ventral temporal (TF, TFO) areas, as well sparse axons in some inferior parietal areas and in the ventral prefrontal cortex, which are areas typically target by MPuR axons but that were not labeled in Cases 20L and 14L. Our interpretation of the labeling pattern in Case 14R is that this injection extended across the MPuR–LPu border (Fig. 12E).
For comparison, we included a fourth experiment with a deposit situated at AP +4.0 near the ventrolateral border of LPu (Case 15L; Figs. 4, 12A). The labeling pattern in this experiment resembles that of Cases 20L and 14L, although here the labeled axons concentrated in areas V2 and V3. Projections to V4 were relatively sparse, and some labeled axons were present in a zone of V1 near the V2 border.
Together, the above experiments show that the projections labeled by the three BDA injections (Cases 20L, 14L, 14L), which were placed immediately lateral to the parasagittal level ∼4.0 mm, are dramatically different from those labeled by injections in MPu but resemble those labeled by an injection situated well within in LPu (Case 15L). Axons labeled by injections lateral to parasagittal +4.0 mm target mainly extrastriate areas V4, V3, V6, and V2 and strictly spare the anterior temporal, prefrontal, and anterior areas of the inferior parietal cortex. Their projection patterns are thus fully consistent with previous studies in marmoset and other monkey species that injected retrograde tracers in posterior extrastriate visual areas (Dick et al., 1991; Kaske et al., 1991; Baleydier and Morel, 1992; Baizer et al., 1993; Adams et al., 2000; Gutierrez et al., 2000; Soares et al., 2001; Shipp, 2003; Lyon et al., 2010; Gattass et al., 2014) as well as with those of the few studies which injected anterograde tracers in macaque LPu (Benevento and Rezak, 1976; Burton and Jones, 1976; Trojanowski and Jacobson, 1976). Thus, despite the absence of a clear cyto- or chemoarchitectonic border, we regard the shift in connectivity evidence enough to draw the border between MPu and LPu at parasagittal level ∼4.0 mm.
Our LPu injections were intended mainly as a probe to map the position of the MPu/LPu border and thus left most of LPu unexplored. In any case, the selective and heavy projections to posterior extrastriate visual areas revealed by these experiments align well with functional reports showing the involvement of LPu neurons in visual attention and visuomotor control networks as well as in the early stages of complex visual perception (reviews Baldwin et al., 2017; Gattass et al., 2018, Kagan et al., 2021).
The anterior pulvinar (APu) and the lateral posterior (LP) nuclei occupy the most anteroventral and anterolateral zones of the primate pulvinar complex (Jones, 2007a; Fig. 2). The injection in Case 04R impregnated a lateral and dorsal portion of APu, with some involvement of LP along its pipette track (Figs. 3, 13C). The site was thus located near, but beyond, the rostral limit of MPu (Paxinos et al., 2012). Labeled axons exiting the injection site traversed the TRN in a dorsal sector of this nucleus between ∼AP +5.5 and +4.2. After entering the subcortical white matter, the axons turned dorsally and posteriorly toward the dorsomedial parietal cortex. In contrast with MPuR injections, the labeled arborizations did not only target inferior parietal areas did not only target inferior parietal areas, but they arborized massively in the dorsomedial posterior parietal areas (PE, PGM, PEC), dorsomedial extrastriate visual areas (19 M, V6/DM), and posterior cingulate and retrosplenial areas (23a–b, 30). In most of these areas, the labeled axons targeted mostly L3b and L1; however, in areas prostriata (ProSt), 30, V2, and V1, the labeled arborizations were mostly limited to L1 (Fig. 13H). Occasional labeled axons were observed in somatosensory areas 2-1-3b and in some dorsal premotor areas, but no axons were labeled in the prefrontal or temporal areas. Overall, Case 04R reveals a projection motif that is strikingly nonoverlapping with that of MPuR. The only place where APu and MPuR projection may coincide is in the inferior parietal cortex (Fig. 13B,C,G).
A BDA deposit in APu labels thalamocortical axons mainly in posterior and medial parietal, cingulate, and retrosplenial cortices. A–F, Representative sections showing the labeled axons at six coronal levels of the cerebral hemisphere from rostral (A) to caudal (F) after a BDA injection in APu (Case 04L; see Fig. 3). Graphic conventions as in Figure 4. Scale bar, 2 mm. G, Representation of anterogradely labeled terminal arborizations drawn onto the standard 2D map. Areas containing labeled arborizations are marked using the abbreviations in Figure 1. H, Camera lucida drawing of labeled axonal plexuses in radial samples taken from posterior parietal area PE, from cingulate cortex area 23b, from retrosplenial area 30, or from a medial zone of primary visual cortex area V1. Samples correspond to the regions marked with an asterisk in panels D, E, and F. Other graphic conventions as in Figure 4.
APu is a large nucleus of the primate thalamus whose connectivity and functions remain underexplored. The fragmentary data from studies which injected retrograde tracers in different cortical areas in macaques and cebus monkeys (Schmahmann and Pandya, 1990, Shibata and Yukie, 2003; Buckwalter et al., 2008; Mayer et al., 2019) align well with the projections observed in Case 04R. In addition, the labeling observed in this case in the posterior cingulate and medial visual areas may also reflect the involvement of some LP neurons in the periphery of this BDA injection (Buckwalter et al., 2008).
The thalamocortical projection pattern revealed by the injection in Case 04R aligns well with the notion that APu is a subcortical hub of the cortical networks. That, based mainly on proprioceptive and visual information, create self-centered representations of the extrapersonal space and keep track of targets, hand position and eye movements for skilled sensory-guided reaching/grasping (Wilke et al., 2010; review in Borra and Luppino, 2017).
Along its medial aspect, MPuR is bordered by Lim, a small, fiber-rich nucleus often regarded as a posterior extension of the intralaminar nuclei (Jones, 2007b; Fig. 2A,D). One of our injection experiments (Case 12L; Fig. 14) was placed a portion of Lim at ∼AP +3.3, with minimal involvement of MPuR. We describe this case to note its sharp differences with the labeling observed after injections in MPuR. Sparse labeled axons in Case 12L were present in cortical territories not reached by MPu axons, such as the auditory, dorsolateral frontal, and retrosplenial areas (Fig. 14). Some sparse labeling was present in the ventral prefrontal and inferior temporal areas, probably because of the involvement of some MPu cells by this injection. Case 12L thus shows that, despite their close apposition, MPu and Lim neurons have largely different projection motifs.
A BDA deposit in the limitans nucleus labels loose axonal arborizations scattered across multiple areas. A–F, Representative sections showing the labeled axons at six coronal levels of the cerebral hemisphere from rostral (A) to caudal (F) after a BDA injection confined to the limitans nucleus (Lim, Case 12 L). Graphic conventions as in Figure 5. H, Camera lucida drawings of six labeled axons in columnar samples taken from areas 8aVv and TL. G, Representation of labeled axons onto the standard 2D map of the cortical surface. The labeling in a complete series of sections is represented. Abbreviations as in Figure 1.
Cluster analysis comparison of thalamocortical projection motifs
To identify in an unsupervised manner thalamocortical projection motifs as a basis for subnuclei identification and delineation, we performed, in addition to the visual comparison, a hierarchical cluster analysis of the cortical labeling (see Materials and Methods). To this end, for each injection experiment, we rated into five integer levels (4–0) the local labeling abundance observed in each target area. Subsequently, we compared the dissimilarity among cases using Euclidean as distance type and average as linkage type. The result is shown in Figure 15, which also provides a comprehensive and detailed tally of the labeling in all our BDA experiments.
Cluster analysis comparison of (dis)similarity between thalamocortical projection motifs across BDA injection cases. The distribution and relative abundance of labeling across cortical areas is compared among all the BDA injection experiments. A, On the left column, cortical areas containing labeled axons are listed. Areas are arranged according to the classification of main regions and subregions of the marmoset cerebral cortex in Paxinos et al. (2012). See the list in Figure 1 for area name abbreviations. On the right column, the distribution and relative abundance of thalamcortical axon labeling observed across cortical areas in every experiment is displayed using the graphic conventions shown in the legend at the bottom-right corner. These data are complementary to those displayed in Figures 5–14. Injection experiments are listed on the top and ordered according to the (dis)similarity of the labeling patterns that they produced across the cortical areas, as computed by a hierarchical cluster analysis (see Materials and Methods). Dissimilarity between clusters (Euclidean distance) is represented by the height of the vertical lines. For clarity, the five highest-level clusters in the dendrogram are highlighted with different colors. Axonal labeling in the striatum or amygdala is also shown at the bottom using the same graphic conventions, although their values were not taken into account in the hierarchical cluster analysis. B, The position of all the deposits is represented on standard serial coronal sections of the pulvinar complex. The case numbers and the zones of the pulvinar that contain experiments in the same high-level cluster in the analysis are shaded using the same colors as in the dendrogram in panel A. Hence, the diagram helps to visualize the partition of the pulvinar complex that emerges from the hierarchical cluster analysis of the thalamocortical projections.
The analysis consistently identified six high-level clusters. Remarkably, the experiments grouped in each cluster are situated in nonoverlapping pulvinar zones. These zones match, to a large extent, the subdivisions proposed by studies based on cyto- or chemoarchitectonic stainings (Paxinos et al., 2012; Homman-Ludiye et al., 2018; Fig. 15B). For example, the analysis robustly separates from each other the injections situated in APu, MPu, LPu, or Lim.
However, the analysis detected a high level of dissimilarity between the rostral (MPuR) and caudal (MPuC) parts of MPu, supporting their distinction as two different subnuclei, with their border at ∼AP + 3.0. Another robust finding was the position of the connection-defined border between MPu and LPu, confirming the impression gained from the visual comparison of labeling patterns produced by injections situated medially or laterally to the parasagittal plane +4.1 (Fig. 15A, green vs pale/dark blue cases). A third important finding is the confirmation that the projection motif of the cells situated in the most ventral portions of MPuC, near the corticotectal fiber bundle, clusters with that of other MPuC injections.
Interestingly, the analysis clearly separated Case 19R injection, situated at the most medial border of the nucleus, from all other MPu injections. This finding may have two nonmutually excluding explanations. First, a small population of cells at the medial edge of MPu may represent in marmosets the “medial MPu” subnucleus described in macaques by Romanski et al. (1997), which projects to medial frontal and temporopolar cortices. Alternatively, the injection in Case 19R may have involved some neurons at the most caudal end of the mediodorsal nucleus (Barbas et al., 1991; Ray and Price, 1993).
Double-retrograde labeling experiments reveal pulvinar neurons whose axons branch to innervate distant cortical regions
As noted above, BDA injection experiments in MPuR repeatedly labeled thalamocortical arborizations in the same widely separate points of the cerebral hemisphere. Moreover, in these experiments, we frequently observed branching axon segments in the subcortical white matter (Fig. 3I,J). These two observations raised the possibility that MPu might contain neurons whose axons branched to simultaneously innervate the widely separate target areas, in line with recent single-cell labeling studies in rodents (review in Clascá, 2023).
Thus, we devised a series of double-retrograde labeling experiments designed to test the presence of branched MPuR axons. In these experiments, we specifically targeted pairs of cortical areas found to receive dense MPuR projections after the BDA injections. Subsequently we mapped and compared, on the same sections, the location of the cell bodies labeled by each tracer in the pulvinar complex. To unambiguously identify double-labeled cells, we used fluorophores with nonoverlapping emission/absorption curves and examined the labeling using both standard epifluorescence and single Z-axis plane confocal microscopy.
To guide the cortical injections, we determined the stereotaxic coordinates of the cortical sites containing labeled thalamocortical arborizations in the MPuR injection experiments using Paxinos et al. (2012) marmoset brain atlas. Then, we used these coordinates to place the site of the craniotomies and the retrograde tracer injections in the cortex. To cover a relatively large cortical region while completely avoiding any white matter encroachment, at each of the selected target zones, we made 2–4 closely spaced tracer deposits, penetrating perpendicularly not deeper than 800 µm from the pial surface (Figs. 16A–C, 17A–C).
Double-retrograde labeling reveals pulvinar neurons whose branched axons simultaneously reach ventrolateral prefrontal and posterior parietal areas. A, B, Representative coronal section images taken at the center of tracer deposits in Case 18. CTB-Alexa Fluor 594 injections involved areas 6Va, 6Vb, and ProM (A), while the FB injections involved areas PG, AIP, and PFG (B). C, Lateral view diagram of the left marmoset cerebral hemisphere representing the extent of the CTB-Alexa 594 Fluor (red) or FB (blue) injections in Case 18. D–F, High-magnification confocal images showing two double-labeled MPuR thalamocortical cells (arrowheads) in Case 18 experiment. Scale bar, 25 µm. G–K, Coronal section diagrams of the thalamus showing the position of labeled cells in the same experiment. FB-labeled neurons are drawn in blue and CTB-Alexa Fluor–labeled neurons in red. Yellow stars indicate double-labeled neurons. Each dot represents a labeled cell body. See Figure 1 for abbreviations. Scale bar, 500 µm. L, Diagram representing the location of FB and CTB-Alexa Fluor 594 injections in Case 35R. In this experiment, FB injections involved areas 47L, 45, 6Vb, and 8Av, while CTB-Alexa Fluor injections involved areas PF and PFG. Note that the fluorescent tracers injected were reversed with the previous case (C). M–Q, Coronal section diagrams representing the neuronal cell bodies labeled in Case 35R. The tissue sections deviated noticeably from the standard coronal plane in the mediolateral axis; for this reason, we indicate, for each section, both the approximate AP level at the midline (bottom left) and at the lateral edge (bottom right) of the thalamus. Other conventions, as above.
Double-retrograde labeling reveals pulvinar neurons whose axons simultaneously reach the ventrolateral prefrontal and inferior temporal cortices. A, B, Representative coronal section images taken at the center of the tracer deposits in Case 19L. In this experiment, two CTB-Alexa Fluor 594 deposits were made in the lateral prefrontal cortex, involving 47 , 8aV, 6Vb, 6Va, and 45 (A). In the same hemisphere, three contiguous FB deposits involved areas TE3, TE2, and TEO (B). C, Diagram representing the extent of the CTB-Alexa Fluor 594 (red) and FB (blue) deposits in the cerebral hemisphere of Case 19L. D–K, Coronal section diagrams showing the position of labeled cells in the posterior thalamus. Each dot represents a labeled cell body. FB-labeled neurons are drawn in blue and CTB-Alexa Fluor–labeled neurons in red. Yellow stars indicate double-labeled neurons. See Figure 1 for abbreviations. Scale bar, 500 µm. L–N, High-magnification detail showing a double-labeled MPuR cell. Scale bars: A, B, 1 mm; D–K, 500 µm.
Case 18 received injections of CTB-Alexa Fluor 594 in lateral frontal areas (6V + ProM + 47L) and of FB in inferior parietal areas (PG + PFG; Fig. 16A–C). The cell population labeled by the frontal injection extended continuously from APu to MPu (Fig. 16G–K), with most of the cells clustered in a central zone of MPuR. In turn, the population labeled by the parietal injection was mainly located in APu, extending caudally into MPuR, and they grouped in two main clusters, one dorsal and one ventral, separated by a low-density zone. This zone was occupied by the large cluster of cells labeled by the frontal injection. Cell somata simultaneously containing the two tracers were observed in MPuR and APu; they were located at the intersection zones of the populations labeled by one or the other tracer (Fig. 16G–J).
Case 35 received pressure injections of FB in ventral prefrontal areas 45, 47L, and 6Vb and of CTB-Alexa Fluor 594 in inferior parietal areas PF and PFG. The overall distribution of the populations labeled by each of the two tracers was like that of Case 18, albeit the parietal-projecting population was shifted more anteriorly and ventrally in APu, and the number of cells labeled by the frontal injection was smaller. The double-labeled cells were again situated in the zones of overlap between the two populations and located both in MPuR and APu.
In marmoset Case 19, we combined a deposit of CTB-Alexa Fluor 954 in the ventral prefrontal cortex (areas 47L, 45, and 6V) with one of FB in temporal inferior areas (TE3, TEO; Fig. 17A–C). Cells labeled by the prefrontal injections were present in MPu and APu in a pattern equivalent to that of the previous two cases, although their number was much lower because of a general low efficacy to the retrograde transport of CTB-Alexa Fluor 954 in this experiment. In turn, cell labeled by the temporal FB injections were most numerous in MPuC along the entire rostrocaudal extent of this subnucleus. A smaller contingent of cells was labeled in MPuR, with occasional cells in APu (Fig. 17A–K). The spatial overlap between the frontal-projecting and the temporal-projecting populations was thus quite small. However, we also observed some few, but unambiguously double-labeled neurons in MPuR and APu (Fig. 17G,H,L–N).
Simultaneous double-labeling of thalamocortical cell somata and visualization of corticothalamic axon terminals
Previous anatomical studies in macaques had noted a correspondence between clusters of pulvinar cells retrogradely labeled from an area and the corticothalamic axons anterogradely labeled from the same area, suggesting the existence of closed, reciprocal thalamo-cortico-thalamic loops (Hardy and Lynch, 1992; Cavada et al., 2000, Grieve et al., 2000; Shipp, 2003). Thus, in one experiment, we investigated the spatial correlation between the pulvinar cell groups innervating a pair of areas, including the double-projection cells, with the corticothalamic axons originating from each of the two areas.
For the double-retrograde labeling assay, we pressure-injected FB in the inferior parietal cortex (areas PG and PFG) and iontophoretically delivered unconjugated CTB in the ventrolateral prefrontal cortex (areas 47L, 45, and 6V). The CTB simultaneously labeled both pulvinocortical cell somata and corticothalamic prefrontal axon terminals (Angelucci et al., 1996). In addition, to label the parietal corticothalamic axons, we interspersed two columnar iontophoretic deposits of BDA among the FB parietal injections. Unconjugated CTB and BDA were visualized using red- and green-fluorescent antibodies, respectively (Fig. 18A–F; see Materials and Methods).
Neurons with branched axons simultaneously targeting the prefrontal and posterior parietal areas are preferentially found along the borders of pulvinar domains innervated by either prefrontal or parietal corticothalamic terminals. A–C, Lateral view diagrams of the marmoset cerebral hemisphere (A), representing the tracer injections in Case 26. In this experiment, several adjacent CTB injections (in red, B) were made in the ventrolateral prefrontal cortex (areas 45, 47L, 6Va, and ProM). In addition, intermingled injections (C) of either FB (blue) or BDA (green) were made in the posterior parietal cortex (areas PFG and PG). D–F, Representative epifluorescence images of the tracer deposits. For reference, the inner L6 border and the L1/L2 border of the cerebral cortex are highlighted by white dashed lines on the epifluorescence images. G, H, Representative coronal section diagrams showing the retrogradely labeled cell somata (red dots, CTb; blue dots, FB; yellow stars, double-labeled cells). Thin green lines represent corticothalamic axon branches anterogradely labeled by the BDA injections. Pink-shaded areas indicate the plexus of corticothalamic axons labeled anterogradely from the prefrontal CTB injections (compare with panels Q and S). Abbreviations as in Figure 1. I–L, High-magnification detail of a MPuR region containing some double-retrogradely labeled (CTB + FB) cell somata (indicated by gray arrows) along with prefrontal corticothalamic axons (red stippling) and parietal corticothalamic axons (green fibers). M–P, High-magnification detail of an APu region containing some double-retrogradely labeled neurons. Q–S, High-magnification image of the region framed by a square in panel H. Note the numerous double-labeled cells indicated by gray arrows. Note also that the double-labeled cells seem to be preferentially located along the border between the territories innervated by the prefrontal or parietal corticothalamic axons. Scale bars: B–F, 1 mm; G–H, 500 µm; I–P, 25 µm; Q–S, 100 µm.
The prefrontal CTB and the inferior parietal FB deposits labeled numerous pulvinar cells whose distributions were like those in the cases described earlier and included numerous double-labeled cells (Fig. 18G–J,L–N,P–S). In addition, the unconjugated CTB and the BDA made visible, in the same tissue sections, through separate fluorescence channels, the corticothalamic axon terminals originating in one or the other area. As free CTB is transported retrogradely but also anterogradely, the terminal plexuses of CTB-labeled prefrontal corticothalamic axons were clearly visible in this experiment.
As expected from similar experiments in macaques (Selemon and Goldman-Rakic, 1988; Cavada et al., 2000; Shipp, 2003), the plexuses largely overlapped the clusters of CTB-labeled cell somata. Likewise, the axons labeled anterogradely from the parietal BDA deposits concentrated on the clusters of FB-labeled somata. Interestingly, most double-labeled neurons were located at or near the border between domains that received dense corticothalamic axons from the same two areas innervated by the bifurcated pulvinar axons (Fig. 18G,H,S). addition, this experiment revealed numerous other neurons labeled retrogradely from one area that were situated close to, or even within, the territory innervated by the corticothalamic axons from the other area.
Discussion
By means of BDA injections in the thalamus, we mapped with high resolution the distribution of thalamocortical projections from MPu and surrounding pulvinar domains. In addition, by injecting two different retrograde tracers at distant sites targeted by the pulvinocortical axons, we demonstrated the existence of numerous MPu neurons whose branched axons innervate widely separate cortical regions.
Medial pulvinar projections may simultaneously interact with several multidomain and single-domain cortical networks
As in macaques, the inferior and ventral anterior temporal cortices are the focus of MPu projections in marmosets (Fig. 19A; Benevento and Rezak, 1976; Burton and Jones, 1976; Baleydier and Mauguiere, 1985; Morán et al., 1987; Baleydier and Morel, 1992; Baizer et al., 1993; Romanski et al., 1997; Gutierrez et al., 2000). Functional studies of the inferotemporal cortex in macaques and marmosets have revealed several patch-like domains whose neurons show selectivity for object vision of a specific visual category. Neurons in caudally located patches encode simpler elements, while those near the temporal pole encode more global, individual-specific features (Gross, 2008; Freiwald and Tsao, 2010; Ku et al., 2011; Sato et al., 2013; Sliwa et al., 2016; Chang and Tsao, 2017; Schaeffer et al., 2020).
Pulvinar nuclei are wired as regional or inter-regional multidomain hubs in different cortical area networks. Cartoon summary of the projection motifs of the pulvinar nuclei that received BDA injections in this study. Each nucleus and its projections (arrows) are labeled in the same color code used in Figures 4 and 15. For a more comprehensive comparison across all pulvinar nuclei, the connections from IPu, which were not explored in the present study, are also represented (in gray shade) based on information from the literature (Baldwin et al., 2017). The cortical areas receiving the most robust projections from each nucleus are listed over the contour of a lateral view of the marmoset cerebral hemisphere. Arrowheads “disappearing” behind the brain contour indicate projections to areas situated in the medial or inferior surfaces of the cerebral hemisphere. The names of these medially or ventrally situated areas are shown in italics.
Hence, through their projection to inferotemporal areas, MPu cells might interact with cells encoding visual qualities and/or identities of objects and individuals (Pinsk et al., 2005; Orban et al., 2014; Schwiedrzik et al., 2015; Bao et al., 2020) and/or socially relevant interactions between them (Hung et al., 2015; Sliwa and Freiwald, 2017; Schaeffer et al., 2020). In addition, the projections to the ventromedial temporal cortex from MPuR and MPuC might interact with neurons supporting long-term memory visual entity consolidation, retrieval, and the sense of familiarity with previously seen entities (Martin-Elkins and Horel, 1992; Horel, 1996).
Marmoset MPuR axons also target mainly the ventrolateral prefrontal and ventral premotor areas and, less densely, some polar and lateral orbitofrontal areas as well (Fig. 19A), a pattern equivalent to that in macaques (Trojanowski and Jacobson, 1976; Asanuma et al., 1985; Goldman-Rakic and Porrino, 1985; Markowitsch et al., 1987; Selemon and Goldman-Rakic, 1988; Barbas et al., 1991; Morecraft et al., 1992; Romanski et al., 1997; Cavada et al., 2000; Cappe et al., 2009; Borra et al., 2023). Ventrolateral prefrontal neurons reportedly activate to the sight of specific objects, faces, and socially relevant facial expressions (Wilson et al., 1993; O’Scalaidhe et al., 1997; Tsao et al., 2008, Schwiedrzik et al., 2015; Shepherd and Freiwald, 2018; Schaeffer et al., 2020), while ventral premotor neurons activate both in ingestive/facial gestures made by the individual or seen in others, and area 8Av neurons perform similar functions regarding gaze direction (Ferrari et al., 2003; Petrides et al., 2005). In addition, some area 6Vb neurons directly innervate the brainstem motoneurons controlling the lower face mimic muscles (Morecraft et al., 2001). Thus, by targeting these frontal areas, the MPu axons may interface cortical circuits involved in the executive control of face and hand movements and social cognition.
Consistent with fragmentary data available for other monkey species (Asanuma et al., 1985; Baleydier and Mauguière, 1987; Schmahmann and Pandya, 1990; Baleydier and Morel, 1992; Hardy and Lynch, 1992; Baizer et al., 1993; Romanski et al., 1997; Gutierrez et al., 2000; Cappe et al., 2009, Mayer et al., 2019), we show that MPuR innervates the posterior inferior parietal cortex. These axons target mainly areas (PF, PFG, AIP, VIP) involved in visuomotor transformations for hand and mouth action guidance and for decoding the meaning of such actions observed in others. Less heavily, MPuR axons also target multisensory parietal areas PG and OPt, whose neurons encode egocentric representations of surrounding space for attention and the guidance of arm and/or eye movements (review in Borra and Luppino, 2017).
In sum, MPuR projections selectively and simultaneously target several key nodes of the frontoparietal and frontotemporal subnetworks that support visual recognition of individual object (including individual faces), gesture control and decoding, and visually guided hand–mouth actions. Meanwhile, MPuC projections are focused on the temporal node of these subnetworks (the inferior and ventral temporal areas), whose relevance for visual object recognition and recall is well documented. In addition, smaller projections target ventromedial prefrontal and temporopolar areas (from medial MPu) or some cingulate areas (from dorsal MPuR), where they may interface cortical subnetworks involved in self-referential/emotional functions (Saleem et al., 2008).
Individual pulvinar cells simultaneously innervate distant areas
Experiments in macaques using the double-retrograde labeling method reported that the clusters of pulvinar cell populations projecting to different cortical areas are usually separate. However, when injections are placed in areas that are heavily corticocortically interconnected, the clusters of retrogradely labeled cells partially intermingle (Asanuma et al., 1985; Baleydier and Mauguiere, 1985, 1987; Baleydier and Morel, 1992; Hardy and Lynch, 1992). This remarkable “replication” of the corticocortical and thalamocortical connection patterns was identified by Shipp (2003) as a fundamental principle of pulvinar connectivity. He and others proposed that such intermingling zones might be a possible site for nonreciprocal (open-loop) interactions between separate cortical areas, via the pulvinar (Selemon and Goldman-Rakic, 1988; Benevento and Port, 1995; Darian-Smith et al., 1999; Grieve et al., 2000; Cappe et al., 2009).
Present results align well with the replication principle. Our small BDA injections always labeled thalamocortical arborizations in several, often quite separate, cortical areas, consistent with the notion that pulvinar cells projecting to different areas may be intermingled. In addition, recent data on marmoset corticocortical connections (Majka et al., 2020) show that the areas containing axons labeled by our BDA injections are directly linked via corticocortical connections. Moreover, our double-retrograde tracer injections in some of the cortical spots labeled by BDA injections yielded partly overlapped pulvinar cell clusters.
Unlike in previous studies, however, our double-labeling experiments demonstrated numerous double-labeled neurons, probably because we maximized the chance of detecting them by specifically targeting pairs of cortical zones simultaneously labeled by the BDA injections in MPu. Consistent with this interpretation, in a single macaque experiment where two retrograde tracer injections involved prefrontal and inferior parietal areas roughly equivalent to those targeted in our experiments, Asanuma et al., (1985) observed some unambiguously double-labeled cells. While the prevalence of these long-range branching pulvinar neurons is difficult to determine from double-labeling experiments alone, some observations suggest that they may be quite numerous. First, the double-labeling method has an intrinsic low sensitivity and informs only about specific binary branching combinations (Schofield et al., 2007; Kim et al., 2018). Second, we observed numerous subcortical branching points in BDA-labeled axon segments (Fig. 3C; Rockland et al., 1999). Third, the complex arrangement of pulvinar neuron clusters innervating different areas makes it unlikely that BDA injections made in different points within MPuR would consistently produce the same projection motif (Fig. 6). Finally, single-cell labeling studies in rodents have shown that thalamocortical axons branching to innervate two or more areas are the norm in higher-order nuclei (reviewed in Clascá, 2023), including the LP nucleus (Nakamura et al., 2015), considered to be the rodent analog of the primate pulvinar (Baldwin et al., 2017).
At the same time, however, even in our experiments with “correctly” positioned double injections, many cells adjacent to the double-labeled ones remained single-labeled or unlabeled. Because all BDA MPuR injections labeled axons in the same cortical zones and the tracer presumably labeled all the axons arising from the injected zone, it follows that the cells not double-labeled by a given duo of cortical injections must either project to a different pair of areas among those targeted by MPuR axons or just to a single area. The implication is that cells with diverse scales of axon branching divergence (e.g., long-range or regional, between neighboring areas) are intermingled in MPuR. These pulvinar neurons might thus provide a cellular substrate for the coupling of functional activity between specific cortical areas at different scales across the cerebral hemisphere (Fig. 20).
Branched MPuR neuron axons may functionally couple cortical areas at different spatial scales. Idealized diagram representing the coexistence within MPuR of thalamocortical projection neurons (color circles) whose branched axons innervate two areas of the same cortical region (regional/single-domain coupling) or two widely separate regions (prefrontal/parietal/temporal; inter-regional/multidomain coupling). Inside MPuR, the projection neurons are placed within domains that receive corticothalamic axons arising from prefrontal, parietal, or temporal (represented as ovals in pale color shade). The inter-regionally branching axons (dashed color lines) arise mainly from MPuR neurons located mainly in overlap/border zones between different corticothalamic domains. Conceivably, these MPuR neurons might integrate corticothalamic inputs from two distant cortical areas, in combinatorial fashion. Gray double-arrow lines represent the existence of heavy corticocortical connections between prefrontal, parietal, and/or temporal areas.
Fast, simultaneous shifts in spike firing and/or membrane oscillation phase across specific, but spatially separate, cortical cell populations are thought to underlie brain-wide functions, such as conscious perception, attention, cognition, and working memory (Dehaene and Changeux, 2011; Saalmann et al., 2012; Bastos et al., 2015; Zhou et al., 2016; Fiebelkorn et al., 2019; Jaramillo et al., 2019; Roelfsema, 2023). The divergent wiring of long-branched MPuR axons suggests the possibility that some pulvinar neurons might facilitate fast activity coupling between distant frontal/parietal and frontal/temporal areas. Moreover, together with other intermingled pulvinar neuron populations with different targets and/or divergence scales, they might provide a rich combinatorial repertoire for the flexible dynamic regulation of topology and/or dimensionality in fronto-parieto-temporal networks (Kim et al., 2018; Shine et al., 2023).
On a different timescale, the consistent “replication” of shared pulvinar inputs and direct corticocortical connections among functionally related cortical areas hints at some kind of developmental relationship. Because thalamocortical axons reach their target areas well ahead of the establishment of corticocortical connections and are known to influence the specification and refinement of the cortical circuits (Sur and Rubenstein, 2005; Reillo et al., 2017; Mundinano et al., 2018), it is also conceivable that long-branched MPuR axons might play some instructive role in the correct development of the various fronto-parieto-temporal networks with which they are connected in the adult.
Footnotes
- Received September 25, 2024.
- Revision received January 6, 2025.
- Accepted January 18, 2025.
-
We acknowledge Prof. John F. Araujo (UFRN) and Profs. Cecilia Hedin-Pereira, Mônica S. Rocha, Ricardo Gattass, Juliana G.M. Soares, and Bruss R.C. Lima (Universidade Federal de Rio de Janeiro, UFRJ) for their support in the initial stages of this project and Begoña Rodriguez (UAM) for the expert neurohistological help. This work was supported by PVE 400730/2014-6 Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq, Brasil) to J.S.C. and PHB-H2012-0011 Ministerio de Educación (Spain) to F.C.
-
Dedication: This paper is dedicated to the memory of Prof. João Guedes da Franca (UFRJ, †2019) who actively participated in the early steps of this project.
-
↵*M.A.C.-C. and P.R-G. equally share first authorship.
-
The authors declare no competing financial interests.
- Correspondence should be addressed to Francisco Clascá at francisco.clasca{at}uam.es.
- Copyright © 2025 the authors
