Relation of koniocellular layers of dorsal lateral geniculate to inferior pulvinar nuclei in common marmosets

Abstract

Traditionally, the dorsal lateral geniculate nucleus (LGN) and the inferior pulvinar (IPul) nucleus are considered as anatomically and functionally distinct thalamic nuclei. However, in several primate species it has also been established that the koniocellular (K) layers of LGN and parts of the IPul have a shared pattern of immunoreactivity for the calcium-binding protein calbindin. These calbindin-rich cells constitute a thalamic matrix system which is implicated in thalamocortical synchronization. Further, the K layers and IPul are both involved in visual processing and have similar connections with retina and superior colliculus. Here we confirmed the continuity between calbindin-rich cells in LGN K layers and the central lateral division of IPul (IPulCL) in marmoset monkeys. By employing a high-throughput neuronal tracing method, we found that both the K layers and IPulCL form comparable patterns of connections with striate and extrastriate cortices; these connections are largely different to those of the parvocellular and magnocellular laminae of LGN. Retrograde tracer-labeled cells and anterograde tracer-labeled axon terminals merged seamlessly from IPulCL into LGN K layers. These results support continuity between LGN K layers and IPulCL, providing an anatomical basis for functional congruity of this region of the dorsal thalamic matrix, and calling into question the traditional segregation between LGN and the inferior pulvinar nucleus.

Graphical abstract

Primate visual thalamus is traditionally segregated into a relay nucleus (dorsal lateral geniculate nucleus, LGN, and association nuclei including inferior pulvinar (IPul) and lateral pulvinar (LPul). We show in the marmoset brain that koniocellular layers of LGN and the central lateral (CL) division of IPul show common patterns of calbindin immunoreactivity and connectivity with the visual cortex, implying a functional congruity and anatomical continuity.

Introduction

The primate dorsal thalamus includes sensory and non-sensory nuclei that make reciprocal connections with the cortex. The nuclei of the dorsal thalamus are segregated based on multiple factors including cytoarchitecture, histochemistry, anatomical connections and function. Two dorsal thalamic nuclei are implicated in visual signal transmission to the cortex; these are the dorsal lateral geniculate nucleus (LGN) and the pulvinar complex (Jones & Hendry, 1989; Sherman & Guillery, 2006; Saalmann & Kastner, 2011). The LGN is typically treated as a primary relay of retinal signals to cortex (e.g. Preuss, 2007) whereas the visual sub-nuclei of the pulvinar (inferior pulvinar – IPul, and lateral pulvinar - LPul) are seen as association areas that act as a bridge between cortical areas, and a sensory link for visual signals from superior colliculus (SC) to cortex. However, parallel to the view that the LGN and visual pulvinar have distinct roles, both regions contain neurochemical subpopulations of cells that have similar cortical and subcortical connectivity (Stepniewska et al., 2000) and play a similar role in synchronizing oscillatory activity of thalamocortical loops and modulating visual cortical activation (Jones, 2001; Cheong et al., 2011; Saalmann & Kastner, 2011). Therefore, there is a possibility that cohesive functional groups of cells exist across these two thalamic nuclei.

As a means of identifying different functional subpopulations of cells in both LGN and pulvinar, differences in calcium binding protein expression have been used. In the LGN, calbindin is primarily expressed in cells located in the koniocellular (K) or interlaminar regions of all primate species studied (Casagrande, 1994; Hendry & Yoshioka, 1994; Goodchild & Martin, 1998). The K layers differ from the principal parvocellular and magnocellular layers of LGN in their cortical and subcortical connectivity. The K-cells receive input from widefield retinal ganglion cells, are the main target of projections from SC to LGN, and send projections to the superficial laminae of V1 as well as V2 and other higher order visual cortical areas including MT and DM (Leventhal et al., 1981; Yukie & Iwai, 1981; Hendry & Yoshioka, 1994; Martin et al., 1997; Beck & Kaas, 1998; Goodchild & Martin, 1998; Hendry & Reid, 2000; Solomon, 2002; Sincich et al., 2004; Szmajda et al., 2008). While K-cells do play a role in visual signal transmission, studies in marmoset have also implicated K-cells in synchronising cortical activity and thalamo-cortical oscillations (Cheong et al., 2011; Pietersen et al., 2017).

Of the two visual sub-nuclei in the pulvinar (lateral and inferior pulvinar), the inferior pulvinar can be further divided into four sub-regions based on calbindin immunoreactivity: central lateral (IPulCL), central medial (IPulCM), middle (IPulM), and posterior (IPulP) (Stepniewska & Kaas, 1997). The IPulCL, IPulCM and IPulP contain cells immunoreactive for calbindin whereas calbindin labeling is largely absent in IPulM (Cusick et al., 1993; Stepniewska & Kaas, 1997; Kaas & Lyon, 2007). Cortical and subcortical connectivity also differs across the four subregions of IPul. IPulCM, IPulM and IPulP are reciprocally connected with “dorsal stream” visual cortical areas such as the middle temporal area and are thought to contain crude maps of visual space. In contrast, IPulCL makes reciprocal connections with “ventral stream” cortical areas including V1, V2 and V4 and contains a retinotopic representation of the contralateral hemifield (Kaas & Lyon, 2007). Furthermore, the major pathway for signals from SC to reach cortex is mediated by IPul. The bulk of projections from SC to IPul are sent to IPulCM and IPulP, with a sparser projection to IPulCL (Stepniewska et al., 2000; Kwan et al., 2019). While there are few studies directly examining the functional differences between the sub-regions of IPul, the involvement of IPul in synchronising cortical activity is well-established (Bender & Butter, 1987; Wilke et al., 2010; Purushothaman et al., 2012; Saalmann et al., 2012; Zhou et al., 2016).

Based on the many similarities drawn between LGN K-cells and IPul, in this study we asked whether subpopulations of K-cells in LGN and the neighbouring IPulCL sub-region could constitute a single group of cells based on common patterns of calcium binding protein expression and cortical connectivity. Experiments were carried out on common marmosets, which are small diurnal New World monkeys increasingly being used as a primate model for visual system structure and function (Solomon & Rosa, 2014; Mitchell & Leopold, 2015). The K-cell layers in marmosets are relatively large and well-segregated from the main parvocellular and magnocellular layers facilitating study of their anatomical connections. Our evidence supports anatomical continuity between the ventral K-layers and the IPulCL, challenging the complete segregation between LGN and division IPulCL of the inferior pulvinar.

Materials and methods

Immunohistochemistry for calbindin-positive cells

Immunohistochemistry for visualisation of calbindin in the thalamus was conducted on tissue from two male adult common marmosets (Callithrix jacchus) acquired from the National Health and Medical Research Council (NHMRC) shared breeding facility in Australia. Procedures were approved by the Institutional Animal Experimentation and Ethics Committee at the University of Sydney and conformed to the Australian NHMRC policies on the use of animals in neuroscience research. Following up to 96 hours in which the marmosets were anaesthetized via intravenous infusion of sufentanil citrate (6–12 μg·kg⁻¹·h⁻¹, Sufenta Forte, Janssen-Cilag, NSW, Australia) for unrelated experiments, the animal was euthanized (intravenous infusion of 300–600 mg kg⁻¹ sodium pentobarbitone, Lethabarb, Virbac, NSW, Australia) and perfused transcardially with 0.9% saline followed by 4% paraformaldehyde in 0.1 M phosphate buffer (PB, pH 7.4), then with 10% glycerol in PB. The brain was removed and immersed in 20% glycerol for 24–72 hours, then coronally sectioned at 50 μm thickness on a freezing microtome. Sections were first processed with fluorescent Nissl stain (NeuroTrace blue-fluorescent, 1:100; Molecular Probes), pre-incubated in phosphate buffered saline containing 5% normal donkey serum and 0.5% Triton X-100 for 1 hour at room temperature, then incubated with primary antibodies (rabbit anti calbindin, Swant CB-38, 1:20000) for 3 to 5 days under slow agitation. Sections were then incubated with secondary antibodies (donkey anti-rabbit Alexa 488, 1:250) for 2 hours. Sections were mounted onto slides and coverslipped with Vectashield mounting medium (Vector Laboratories, Burlingame, CA, USA). Sections were imaged using a fluorescent microscope. High power image stacks through the sections were taken at regions of interest (ROIs) within LGN and IPul. Calbindin positive cell counts and soma diameter were measured from ROIs in Fiji (Schindelin et al., 2012).

Tracing study using a high-throughput neurohistology pipeline

Tracing studies to visualize thalamocortical connections were performed in eight female adult common marmosets (Callithrix jacchus), six animals were acquired from the Japanese Central Institute for Experimental Animals and two were acquired from Japan National Institute for Basic Biology. Case information is shown in Table 1. All experimental procedures were approved by the Institutional Animal Care and Use Committee at RIKEN and conducted in accordance with the Guidelines for Conducting Animal Experiments at RIKEN Center for Brain Science. For experiments involving participation of Australian researchers, all protocols were approved with a field work license from Monash University and conformed to the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.

Table 1.

Case information and injection sites for marmosets used in tracing study

Animal ID	Sex	Age	Weight (g)	Tracer	Injection area	Injection site center
						ML (mm)	AP (mm)	Depth (mm)
M820	F	6y 11mo	400	FB	V1	+1.80	−8.05	0.95
				AAV-GFP	V1	+7.25	−8.15	1.10
M822	F	7y	355	FB	V2	+6.60	−6.00	0.70
M919	F	8y 8mo	345	FB	V2	+0.70	−4.70	0.50
				AAV-GFP	DM	+1.30	−1.00	1.00
				AAV-tdTom	V2	+1.00	−3.50	0.80
M920	F	4y 7mo	347	AAV-GFP	V1	+0.80	−7.50	0.50
				AAV-tdTom	DM	+1.00	−3.20	1.00
				DY	DM	+4.00	−3.00	1.00
M1144	F	10y 3mo	353	FB	V2	+2.80	−4.50	1.10
				AAV-GFP	V1	+3.00	−8.50	1.20
				AAV-tdTom	V2	+2.55	−7.25	1.11
M1146	F	7y 1mo	394	FB	V1	+2.00	−10.00	0.80
				AAV-GFP	V2	+8.80	−7.00	2.00
M1147	F	4y 11mo	335	AAV-tdTom	V2	+3.50	−5.50	0.70
M1148	F	4y 3mo	343	FB	V2	+5.15	−4.00	1.00

Details of surgical procedures were described elsewhere (Lin et al. 2019). Briefly, during surgery, the animal was anesthetized initially with an intramuscular (i.m.) injection of ketamine (100 μl) and maintained with either intermittent injections of alfaxalone (0.01 μl/kg, i.m.) or 2–4% isoflurane inhalation. Throughout surgery, the animal’s heart rate, body temperature, and oxygen saturation were constantly monitored.

For tracer injections, a small cranial window was made using a dental drill (NSK UMXL-DT). Each tracer was delivered using Nanoject II injector (Drummond, USA) with equal volume at depths of 1200 μm, 800 μm, and 400 μm, controlled with Micro4 (WPI, USA), to fill the entire cortical column. Two types of anterograde tracer, AAV-TRE3-tdTomato (AAV-tdTom, 0.3 μl) and AAV-TRE3-Clover (AAV-GFP, 0.3 μl), and two types of retrograde tracer, Fast Blue (FB, 0.3 μl 5% solution in distilled water; Funakoshi; Tokyo, Japan) and Diaminido Yellow (DY, 0.3 μl 2% suspension in distilled water; American Custom Chemical Corporation, USA), were pressure injected at a rate of 20 μl/min in V1, V2 and DM in the right hemispheres. Details of tracer injection were described previously (Reser et al., 2009; Alegro et al., 2017). A summary of all injections is shown in Table 1.

After extraction of the injection pipette, the cranial window was cleaned and sealed with dental acrylic and the head skin was sutured back in place. The animal received injections of analgesics (Marcaine, Astra Zeneca, 100μl/100g), anti-inflammatory (carprofen 100μl/100g) and Hartmann’s solution. The animal received daily oral intake of anti-inflammatory (Oral Metacam; 0.05 mg/kg, Boeringer Ingelheim) for the first three days following the surgery and was monitored throughout an incubation period of 4 weeks.

A high-throughput neurohistology pipeline customized for marmosets was adopted to study the mesoscale connectivity (Lin et al., 2019). Briefly, after tracer injection and the 4-week incubation period, the animal was euthanized and perfused. After the brain was fixed with 4% paraformaldehyde, ex vivo MRI scanning was performed on 6 out of the 8 brains involved in the present study. All the brains were then embedded in freezing agent (Neg-50™, Thermo Scientific Richard-Allan Scientific), and cryosectioned coronally at 20 μm using a tape-transfer method (Pinskiy et al., 2013, 2015) customized for marmoset brain. Consecutive sections were treated differently in sequence: 1) either coverslipped directly after dehydration (to view anterograde and retrograde label), 2) processed for Nissl substance, 3) processed with silver staining for myelin, 4) processed for immunohistochemical treatment for cholera-toxin subunit B. Sections processed in the last two ways were not considered in the current study. Therefore 80 μm separated sections were treated the same way. An automatic staining system (Tissue-Tek Prisma, Sakura, Netherlands) was used to process the sections for Nissl substance with Cresyl Violet. All sections were then mounted with coverslips using an automated system (Tissue-Tek Glas, GLAS-g2-S0, Sakura, Netherlands).

Brain sections were next scanned in a Nanozoomer 2.0 HT (Hamamatsu, Japan) as 12-bit RGB images, with a resolution of 0.46 μm/pixel. Unstained sections were scanned for fluorescence imaging; and the Nissl-stained sections were scanned with bright-field imaging. The RAW images were processed in a high-performance computational infrastructure (Lin et al., 2019). Images of individual brain sections were isolated and compressed into JPEG2000 format for economic data storage and subsequent analyses.

LGN and IPul subregion segmentation

For experiments involving calbindin staining, the laminae of LGN and subdivisions of IPul were manually segmented based on cytoarchitecture from the calbindin-stained sections, taking Paxinos et al. (2012) as the reference model. For brains involved in tracing studies, principal laminae of LGN and subdivisions of IPul were manually segmented based on Nissl sections, whereas the borders between subdivisions of IPul were delineated using a combination of Nissl-stained sections and the established IPul segmentation from calbindin-stained sections (Figure 1). Subdivisions of IPul were not always distinguishable; in the most rostral sections the size of IPul decreases and borders between subdivisions vanish. The majority of our analysis was conducted on brain sections containing both IPul and LGN, largely in line with rostral parts of IPul. In those sections we did not attempt to further segment IPul and would only study the entire region. Where subregions of IPul are identifiable, they are referred to as central lateral (IPulCL), central medial (IPulCM) and middle (IPulM) using Paxinos et al. (2012) as a reference. The K layers were annotated by segmenting the interlaminar regions of LGN.

Figure 1.

A-B. Photomicrographs of coronal sections through LGN and pulvinar in the left hemisphere. Case MY154. Soma stained for Nissl substance using NeuroTrace blue (NTb) are shown in green. Soma stained for calbindin (CaBP) are shown in magenta. C-D. Drawings outlining the LGN laminae and IPul sub-regions (central lateral inferior pulvinar – IPulCL; central medial inferior pulvinar – IPulCM; middle inferior pulvinar – IPulM). The brachium of the superior colliculus (bsc) is also indicated. The blue dashed line indicates the traditional border between LGN and IPul. Neurons with CaBP labelling are shown as magenta dots. E. Photomicrograph of a Nissl stained coronal section through LGN and pulvinar in a different preparation (case MY156), with LGN and IPul sub-regions indicated. Rectangles outline regions shown in higher magnification in F and G. Scale bar = 200 μm. F. High-power magnification of the upper rectangle in E, showing labelling for CaBP. Neurons labelled for CaBP indicated with white arrows. The border between IPulCM and IPulM is shown in white. Note absence of CaBP labelling in IPulM. G. As in F, a high-power view of the lower rectangle in E indicating CaBP labelling in IPulCL. Scale bar = 100 μm.

Tracer-labeled neuron detection

An automatic cell detection routine was developed to locate retrograde tracer-labeled cell profiles in fluorescent sections. In short, a mask of the brain section was calculated, and a Mexican-hat filter was applied to quench background and foreground noise. Colour and intensity thresholds were applied to weakly filter the pixels. A series of morphological operations was performed to identify individual cells and clusters of cells. Individual cell profiles were separated from the clusters using a fast, unsupervised method (Pahariya et al., 2018). Coordinates of the cell centroids were recorded to represent individual cell location in the brain section. Manual proofreading was performed to remove false positive detections.

The number of cells in each section was estimated based on the distribution of profile area and the section thickness using a recursive reconstruction method (Rose & Rohrlich, 1987). From 298 profiles sampled from two 20-μm sections with FB labeled cells, the profile-to-cell ratio was estimated to be 1.75. For the gap of 80 μm between sections, we assumed linear relationship of labeled cell numbers across sections. Therefore, the number of cells within each brain region was estimated as

$$\frac{{\sum }_S\left(\#\mathrm{of}\mathrm{profiles}\right)}{1.75}\times \frac{80\mu m}{20\mu m}$$

where S is the total number of brain sections containing a particular brain region.

Anterograde tracer-labeled axon terminals were identified manually for each brain section within areas with fluorescent label intensities higher than an empirically defined threshold and were distinguished from the passing axons, which formed elongated strands of fluorescent signal.

Whole-brain reconstruction and injection analysis

For each case involved in tracing study, a computational routine was established to reconstruct the whole brain in 3D from all brain sections (Lee et al.., 2018a). Briefly, in 6 animals where ex vivo pre-sectioning MRI of the brains were available, the Nissl stack was reconstructed by a series of jointly optimized rigid motions using the MRI as a guide, and the resulting reconstructed stack was diffeomorphically registered to the marmoset brain atlas (Hashikawa et al., 2015; Woodward et al., 2018). In the 2 animals where MRI guidance was unavailable (M820, M822), reconstruction was performed using the atlas as a shape prior (Lee et al.., 2018b). The fluorescent sections were then aligned and registered to the reconstructed Nissl stack through a series of jointly optimized rigid motions. The calculated transformation was applied to the cell coordinates and terminal areas, so that the tracer-labeled neurons were mapped onto the segmented brain regions on Nissl sections. The lateral view of fluorescent sections was generated by maximal intensity projection through each cross-hemisphere line from the 3D reconstructed brain.

After 3D reconstruction of the whole-brain fluorescent sections, the center and spread of each tracer injection was computationally quantified. By mapping the marmoset brain atlas (Woodward et al. 2018) to individual brains, we verified the brain region where the injection was placed, and whether the eventual spread of injection extended beyond the region. All injections, with their extent, were mapped to a common atlas template to reveal their relative locations (Fig. 2).

Figure 2.

Surface reconstructions of all injection site locations, transposed onto a marmoset brain atlas (Woodward et al., 2018) in lateral (top row) and dorsal view (bottom row). A. Locations of 9 anterograde tracer injections in V1, V2 and DM. Top panel shows a magnified lateral view from right hemisphere (left inset); and bottom panel magnified dorsal view (left bottom inset). The center of each injection is shown as a dot. Animal ID and tracer information for each case are shown in the lateral view. Dot diameter is proportional to the extent of the injection, determined via 3D reconstruction of the fluorescent sections. Dotted lines indicate the bounds of tracer spread. B. Locations of 7 retrograde tracer injections in V1, V2 and DM. Injection site information is represented as in A. Scale bar = 5 mm.

Results

Calbindin-positive cells merge seamlessly from K layers of LGN to IPulCL

Immunohistochemical staining for calbindin was conducted in two animals, with sections containing both LGN and IPul examined. In LGN, cells positive for calbindin were restricted to the K layers, especially K1, and K3 (Figure 1A–D). In more anterior sections, IPul subdivisions could not be distinguished (e.g. Figure 3G), but three IPul subdivisions become apparent in more posterior sections as demonstrated in the line drawings in Figure 1C and D. Using both Nissl and calbindin staining it was possible to distinguish central lateral IPul (IPulCL), central medial IPul (IPulCM) and medial IPul (IPulM). The calbindin-positive cells were most abundant in anterior IPul and IPulCL, adjacent to LGN (Figure 1C, D & G). Calbindin labeling becomes sparser in IPulCM and is completely absent from IPulM (Figure 1G). A total of 18 regions of interest (ROIs) were examined at high power magnification in order to count and measure the size and density of CaBP positive cells. Seven ROIs were in the deepest koniocellular layer of LGN (K1), 5 in IPulCL and 6 in IPulCM. Within the ROIs a total of 320 CaBP positive cells were observed. As shown in Figure 4H, the mean diameter of measured CaBP cells in K1 (n = 166, mean = 13.6±2.2 μm), IPulCL (n = 101, mean = 14.2±1.9 μm) and IPulCM (n = 53, mean = 13.9±2.1 μm) was not significantly different across groups (χ² = 5.55, df = 2, p = 0.06, Kruskal-Wallis non-parametric analysis of variance). As demonstrated by micrographs in Figure 1A, B and E, neither Nissl or calbindin staining revealed clear cytoarchitectonic boundaries between LGN and IPul. Although the distribution of calbindin-immunoreactive cells we find here is essentially identical to that reported in previous studies (Paxinos et al. 2012); the neuropil labeling (especially in IPulCM) was relatively weak and some presumed non-specific label was evident in brachium of the superior colliculus (bsc, Fig. 1A,B). We do not have a clear explanation of this difference to previous results, it may be related to fixation or processing conditions. Despite these differences, in light of the similarity of calbindin reactivity between cells in anterior IPul and in IPulCL more posteriorly, we refer to them collectively as IPulCL below. In summary, the calbindin-positive cells in LGN K layers and IPulCL appear to form a homogeneous population.

Figure 3.

Maximum intensity projection of fluorescent signals onto right hemisphere shown in lateral view from a whole-brain reconstruction (left column), and series of coronal sections through the LGN and IPul (right 4 columns) of case M820 (A), M919 (F) and M1144 (K) showing tracer injections in V1, V2 and DM. A. Lateral view from right hemisphere of case M820. Green label shows AAV-GFP injection in V1. Blue label shows FB injection in V1. Scale bar: 5 mm. B-E Sequential coronal sections through LGN and IPul from posterior (B) to anterior (E). Contours of regions are shown in light grey. Blue arrow indicates examples of FB-labelled cells. Green arrow indicates a region containing AAV-GFP-labelled axon terminals. F. Lateral view of right hemisphere of case M919. Red label shows the AAV-tdTom injection in V2; green label shows the AAV-GFP injection location in DM; blue label shows FB injection location in V2. Note that the injection of AAV-GFP leaked into white matter (Table 2). G-J coronal sections from the animal in F, labelled as in B-E. K. Lateral view case M1144. Red label shows the AAV-tdTom injection in V2; green label shows the AAV-GFP injection in V1; blue label shows the FB injection in DM. L-O coronal sections from this animal, labelled as in B-E. Note that one section was lost between L and M. Scale bar: 1 mm. Abbreviations: K1, K3: LGN koniocellular layers 1 and 3.

Figure 4.

A-E. Drawings of fluorescent coronal sections through LGN and IPul from posterior (A) to anterior (E) for the case of M822 with FB injection in V2. Detected FB-labelled cells are represented as dots. LGN lamina and IPul subdivisions taken from the adjacent Nissl sections and established IPul segmentation (as in Figure 1). Abbreviations as Figure 1. AP and d_ip (distance from most anterior section containing IPul) values are shown beneath each section. F. Frequency histogram of retrogradely labelled cell soma radii as given by automatic cell detection. G. Frequency histogram of estimated cell soma radii after profile-to-cell correction. H. Diameter of CaBP positive cells as measured in regions of interest in either K1, IPulCL or IPulCM I-K. Bar graphs showing the corrected cell counts for retrograde tracer-labelled neurons in IPulCL, K layers, M layers and P layers of LGN after injections in V1 (I), V2 (J), and DM (K). NB: cell counts for the M and P layers in I and K layers in J are <10. Bar heights represent average number of cells across different cases; error bars show the standard deviation.

K layers and IPulCL form similar connection patterns with visual cortices

To identify the connectivity of LGN and IPul with visual cortices, retrograde and anterograde tracer injections were placed in either V1, V2 or DM of 8 animals. Table 1 summarizes the tracer injections made for each animal and Figure 2 provides a schematic representation of all injection site locations. Figure 3A, F and K are examples of whole brain reconstructions of the fluorescent sections from three animals revealing tracer-labeled neurons at the injection sites and destinations. In some cases (e.g. Figure 3F, AAV-GFP), tracer trajectory can be followed through the brain. Coronal sections containing both LGN and IPul were analysed to determine the location of retrogradely labeled cells and anterogradely labeled processes. Figure 3 shows example images of these coronal sections from 3 animals, overlaid with outlines of the LGN layers and IPul. From left to right the sections become more anterior, in this progression the LGN layers become more prominent and the IPul becomes smaller. Table 2 summarises the observed connections following each injection of anterograde or retrograde tracer.

Table 2.

Summary of all detected connections.

Tracing direction	Injection center	Leakage to	Animal ID	Connection
				IPulCL	K1	K2-K4	M layers	P layers
Anterograde	V1	-	M820	●	●	●	●	●
		-	M920	●	●	●	●	●
		V2	M1144	●	●	●	●	●
	V2	-	M919	●	●
		-	M1144
		V1	M1146
	DM	WM	M919	●	●	●	●	●
		V2	M920	●	●		●
		V2, WM	M1147	●	●
Retrograde	V1	-	M820	240	37	197	240	388
		-	M1146	107	178	766	147	1024
	V2	V1	M822	469	5	107	0	2
		-	M919	1408	7	32	0	0
		DM	M920	295	2	5	0	9
		-	M1148	114	5	0	0	2
	DM	-	M1144	329	0	2	0	0

WM: white matter. ●: label detected.

All retrograde tracer injections resulted in labeled cells in IPulCL. Similarly, all V1 and V2 retrograde tracer injections led to labeled cells in the LGN K layers. The single instance of DM injection only led to very sparse labeling in K layers other than K1 (Figure 4, 5A–C). An example of results from the labeled cell detection protocol used is shown in a series of section reconstructions through LGN and IPul in figure 4A–E, where the injection was placed in V2. We next quantified the number of retrogradely labeled cells in individual regions of LGN and IPulCL. The observed number of cell profiles was corrected for the profile area and inter-section distance to estimate the actual number of cells (Figure 4F–G). As expected, V1 retrograde tracer injections led to robust soma labeling in all the K (589 ± 502 cells), M (194 ± 66) and P (706 ± 450) LGN layers as well as in IPulCL (174 ± 94) (Figure 4H). V2 injections resulted in consistent labeling in K LGN layers (41 ± 50 cells) and IPulCL (572 ± 576 cells). In 3 cases of V2 retrograde tracer injection a small number of retrogradely labeled cells was detected in the P layers (a total of 6 cells across 3 cases). In one case (M822) this label could be attributed to leakage of the tracer into V1 (Figure 2B). It is also possible that these cells are part of the small proportion of K-cells present in the principal LGN laminae (Goodchild & Martin, 1998; Hendry & Reid, 2000). In the single case of DM injection, 329 cells were detected in IPulCL and 2 cells in K3 only. No labeled cells were detected in the M or P layers following DM injection (Figure 4K).

Figure 5.

A-C. Grids representing each experiment where retrograde tracer was injected into either V1 (A), V2 (B) or DM (C). The number of retrogradely labelled cells counted in each coronal section (shown in green) are organised based on their location in IPul or K layers (grid columns) and the d_ip value of the section (grid rows). Empty white grids indicate a lack of labelling. Grey grids indicate where data was unavailable, either due to missing sections or absence of IPul in the section. D-E. Grids representing each experiment where anterograde tracer was injected into either V1 (D), V2 (E) or DM (F). Blue squares indicate where axon terminals where observed in each coronal section. Grids are organised as in A-C.

Anterograde tracer labeled axon terminals in LGN and IPul were manually identified (Figure 3). Observed projection patterns following anterograde tracer injections in V1, V2 and DM are summarised in Table 2. All anterograde injections in V1 and DM yielded projections to IPulCL and K layers while only 1 case of V2 injection resulted in anterograde label of axon terminals in IPulCL and K1. Anterograde label was found in the P and M layers following all instances of V1 injection, no cases of V2 injection and 2 cases of DM injection.

Visual cortices-connected neurons transitioned continuously from K1 to IPulCL

In our observation of sections containing both LGN and IPul it became apparent that as sections moved from posterior to anterior, both retrograde and anterograde labels gradually shifted laterally from IPul to LGN. Three examples of this transition are shown in Figure 3. Since all retrograde tracer injections resulted in labeling in IPulCL and K layers, we graphically displayed the number of labeled cells in individual K layers and IPulCL in consecutive fluorescent sections in Figure 5A–C. We have used the most anterior section with IPul still visible as a means of aligning sections across different animals. The most anterior section containing IPul is assigned a “distance from anterior IPul” (d_ip) value of 0, with more anterior sections having positive values and posterior sections having negative values. In this way it is possible to visualise the transition of label from IPul to LGN while also having access to the number of retrograde labeled cells per section. For both cases of V1 retrograde tracer injection, labeled cells transitioned from IPulCL to K1, K2, K3, and eventually to K4, in a sequential manner (Figure 5A). In all four cases of retrograde tracer injections in V2, labeled cells were consistently detected in K layers and IPulCL, with label in IPulCL showing up more posteriorly, and label in K layers appearing more anteriorly. On at least one section (same anterior-posterior position) in each case, labeled cells were detected both in K1 and IPulCL, reflecting the continuity of V2-projecting cells between these two immediate adjacent structures (Figure 5B). Finally, in the single case of DM retrograde tracer injection, labeled cells were only found in IPulCL and K3 (Figure 5C). However, from section images for this animal in Figure 3L–O it is clear that these labeled cells were distributed along the border between LGN and IPulCL. It is possible that DM-projecting cells started anteriorly in LGN and extended posteriorly into IPulCL, but more experiments would be needed to confirm this single observation.

Figure 5D–F shows the pattern of anterograde labeling of axon terminals in K layers and IPulCL. In all three cases of anterograde tracer injection in V1, labeling showed a similar pattern as the V1 retrograde tracing, where the axon terminals were detected first in IPulCL in more posterior sections before continuing to K1, K2, K3 and K4 sequentially (Figure 5D). In one out of three cases of V2 injections, we observed axon labeling in IPulCL and K1. In the other two cases, neither IPulCL nor K layers were labeled (Figure 5E). Finally, in all cases of anterograde tracer injection in DM, axon terminal labeling was observed in IPulCL and K1 together in multiple sections, showing the continuity of the DM neuron projections in these two adjacent structures (Figure 5F). The continuity is also shown for one case (M919) in the AAV-GFP labeled axon terminals in Figure 3G. Note that in this case, labeled axon terminals were also seen in the other layers in more anterior sections (see Figure 3G–J).

In sum, all anterograde and retrograde tracer injections resulted in labeling in K1 and IPulCL, with the exception of one single case of retrograde tracer injection in DM. By displaying numbers of retrograde labeled cells and the presence of anterogradely labeled axon terminals for individual sections (Figure 5), we confirmed our observation from individual coronal sections (Figure 3) that the labeled cells and axon terminals are continuous between K1 and IPulCL.

Discussion

Thalamocortical neurons with different chemical identities of are implicated with distinct roles in thalamic synchronization (Jones, 1998). In visual thalamus of macaque and owl monkeys, a core of parvalbumin-positive neurons is largely constrained to the principal layers of LGN and pulvinar; while a matrix of calbindin-positive neurons is present in K layers of LGN and adjacent IPul (Jones, 1998; Xu et al., 2001). The idea that the K layers and IPul could form a continuous part of the thalamic matrix is expressed in the following statement made by Jones (1998) in relation to Figure 4 in that paper: “This is especially noticeable posteriorly where, as the [LGN] becomes enveloped in the enlarging [IPul] nucleus, calbindin/CAMKII-cells extend uninterruptedly across the intervening medullary lamina, to become continuous with the larger population of similar cells in [IPul].” (Jones, 1998, p. 336). Yet the cortical projection patterns of these two regions were rarely studied in parallel to each other (e.g. Cusick et al., 1993; Warner et al., 2012; Kwan et al., 2019), likely because they are customarily classified respectively as first-order and higher-order thalamic nuclei (Saalmann, 2014). In the present study we confirmed that in marmosets, calbindin-positive cells are abundant in the LGN K layers and adjacent regions of IPul. We also found that these cells form a cytoarchitechtonically continuous population: a result that is consistent with previous observations in other primate species, as stated above. The LGN layer K1 and IPulCL show shared patterns of connectivity with visual areas V1, V2 and DM. Neurons projecting to cortex and axon terminals projecting from cortex showed a rostro-caudal organisation that seamlessly crossed the border between layer K1 and IPul. These results suggested that the K1 layers of LGN and IPul are anatomically continuous and may underpin a single functional population of cells in this region of the thalamus.

In primates, calbindin labeling in the thalamus is associated with a role in cortical modulation and synchrony (Jones, 2007). Indeed, both the K LGN layers and vision-related pulvinar have been implicated in the modulation and synchronisation of cortical activity (Saalmann et al., 2012; Klein et al., 2016; Zhou et al., 2016; Pietersen et al., 2017). Similarities in function between the calbindin labeled K cells and IPul cells in primate thalamus, as well as the observed continuity in calbindin labeling between the two structures are evidence that they make up a single functional population of thalamic neurons.

The continuity between K LGN cells and IPul also extends to the cortical connections of these thalamic structures. The most compelling evidence for this came from the matched extrastriate inputs and outputs. In the current study, we showed that in all but one case of anterograde or retrograde tracing from V2 and DM, labeling was either seen in both K1 and adjacent IPul, or not at all (Table 2). Our results are consistent with previous reports which examined LGN and IPul independently (Bullier & Kennedy, 1983; Beck & Kaas, 1998; Hendry & Reid, 2000; Kaas & Lyon, 2007). However, our findings further provided evidence that these connections shift along a rostro-caudal gradient from LGN layer K1 to the adjacent IPulCL. One inconsistency we found following anterograde tracer injection into V2 was the absence of labeling in either LGN or IPul in two out of three cases. It is possible that tracer injection into different cytochrome oxidase stripe types of V2 led to the inconsistent results we have seen. Studies in macaque have shown that IPul projects to the thick cytochrome oxidase stripes of V2 (Levitt et al., 1995). Further studies with cytochrome oxidase staining of the cortex would be required to definitively determine if this is the case in marmosets.

We have also noted that the proportion of retrogradely labeled cells differed between LGN and IPul. Following V1 injections, more cells were labeled in LGN than in IPul (Fig. 4H) while the opposite was the case following tracer injections in V2 and DM (Fig. 4I–J). This result is not surprising considering the main role of both structures in visual signal transmission. The LGN is the primary relay of visual signal to V1. Destruction of V1 in primates yields extensive retrograde degeneration of LGN (Cowey & Stoerig, 1989; Hendrickson et al., 2015), with near complete loss of P and M cells (Mihailovic et al., 1971), leading to “cortical blindness” (Brindley et al., 1969; Stoerig et al., 2002). On the other hand, pulvinar, along with only the K layers of LGN, are implicated in residual visual capacity following damage to V1 because they project directly to extrastriate cortex (Cowey & Stoerig, 1989; Cowey et al., 1994).

In analysing the relationship between the K layers of LGN and IPul it is important to consider the subcortical inputs to these nuclei. There is evidence that the K layers of LGN as well as all subregions of IPul (with exception of IPulM) receive input from SC (Harting et al., 1991; Stepniewska et al., 2000; Kwan et al., 2019), although the projection into IPulCL was sparse (Stepniewska et al., 2000; Kwan et al., 2019). The input to LGN K layers comes from the upper sublayers of the stratum griseum superficiale of SC (Harting et al., 1991). Previous primate studies suggested a deeper layer origin of SC cells projecting to IPul (Graham and Casagrande, 1980; Huerta and Harting, 1983; Kwan et al., 2019). However, it is still unclear which layers of SC project to IPulCL, which is the pulvinar region of focus in the present study. Furthermore, the K layers of the LGN are known to receive other subcortical inputs including the thalamic reticular nucleus, parabigeminal nucleus, and nucleus of the optic tract (Wilson et al., 1995; Zeater et al., 2019). Comparatively little is known about similar subcortical inputs to inferior pulvinar. Such studies would be useful, particularly in comparing the adjacent K1 LGN layer and IPulCL and attempting to determine if they are operating in parallel or a single functional group of cells. Improved understanding of the developmental timing of IPulCL and the LGN K layers would also be important in this context. Despite these limitations, our findings support the suggestion that a “matrix” of calbindin positive cells across thalamic nuclei plays a role in synchronising the thalamocortical network (Jones, 2001).

Abbreviations

LGN

dorsal lateral geniculate nucleus

IPul

inferior pulvinar

IPulCL

inferior pulvinar, centro-lateral part

IPulCM

inferior pulvinar, centro-medial part

IPulM

inferior pulvinar, middle part

IPulP

inferior pulvinar, posterior part

P layers

parvocellular layers

M layers

magnocellular layers

K layers

koniocellular layers

CaBP

calbindin

PV

parvalbumin

DM

dorsomedial visual area

FB

Fast Blue

SC

superior colliculus