Molecular cell identities in the mediodorsal thalamus of infant mice and marmoset

Abstract

The mediodorsal thalamus (MD) is a higher-order nucleus located within the central thalamus in many mammalian species. Emerging evidence from MD lesions and tracer injections suggests that the MD is reciprocally connected to the prefrontal cortex (PFC) and plays an essential role in specific cognitive processes and tasks. MD subdivisions (medial, central, and lateral) are poorly segregated at the molecular level in rodents, leading to a lack of MD subdivision-specific Cre driver mice. Moreover, this lack of molecular identifiers hinder MD subdivision- and cell-type-specific circuit formation and function analysis. Therefore, using publicly available databases, we explored molecules separately expressed in MD subdivisions. In addition to MD subdivision markers, we identified several genes expressed in a subdivision-specific combination and classified them. Furthermore, after developing medial MD (MDm) or central MD (MDc) region-specific Cre mouse lines, we identified diverse region- and layer-specific PFC projection patterns. Comparison between classified MD marker genes in mice and common marmosets, a non-human primate model, revealed diverging gene expression patterns. These results highlight the species-specific organization of cell types and their projections in the MD thalamus.

Introduction

The mediodorsal thalamus (MD) plays a critical role in cognitive function (Mitchell, 2015). For instance, impairments in working memory and reversal learning have been reported in MD-lesioned rats and primates (Bailey et al., 2005; Block et al., 2007; Chudasama et al., 2001; Floresco et al., 1999; Hunt et al., 1998a, 1998b). MD comprises various cellular subgroups that divergently project to distinct brain areas, including the prefrontal cortex (PFC), basolateral amygdala, and striatum (Alcaraz et al., 2016; Deschenes et al., 1995; Elena Erro et al., 2002; Krettek et al., 1977). Although several studies investigated the function of MD subdivisions by targeted physical and irreversible ablation, these lesion studies produced unclear results between individual animal or experiment (Mitchell, 2015). In contrast, similar cognitive dysfunctions as those reported for MD damage have been reported following PFC damage in rodents and non-human primates (Kahn et al., 2012; Ongur et al., 2000). Furthermore, several brain imaging studies show decreased functional connectivity between the MD and PFC in the schizophrenia patients exhibiting cognitive deficits, suggesting that a subpolulation of MD projects to PFC and plays distinctive roles on cognitive function. Recent optogenetic and chemogenetic studies on MD-PFC neural circuits aimed to reveal the neuronal activity in the MD in the context of working memory and cognition (Bolkan et al., 2017; Parnaudeau et al., 2013). However, these approaches still involve necessitate stereotaxic injections into the MD without targeting discrete subdivisions.

Identifying MD subdivision-specific molecular markers is crucial to enable the dissection of MD cell-specific circuit formation and function. Based on simple histological staining such as Nissl substrate, the MD can be divided into cytoarchitectonically distinct subnuclei. In rodents, the MD is typically subdivided into three distinct regions: the medial MD (MDm) (closest to the midline), central MD (MDc), and lateral MD (MDl) (bordering the rostral intralaminar nuclei and internal medullary lamina) (Krettek et al., 1977; Mitchell et al., 2013; Van Eden, 1986). In primates, MD subdivisions can be morphologically parsed into a magnocellular subdivision (MDmc) (corresponding to the MDm in rodents), parvocellular subdivision (MDpc) (corresponding to the MDc in rodents), pars multiformis, and densocellular subdivisions (MDmf [rostral MDl] and MDdc [caudal MDl], corresponding to the MDl in rodents) (Barbas et al., 1991; Giguere et al., 1988; Goldman-Rakic et al., 1985). Although the corresponding subdivisions are speculated based on anatomical location, the connectivity of different subdivisions to the PFC differs between primates and rodents (Parnaudeau et al., 2018). For example, the MDc in rodents and the MDmc in primates project to the orbitofrontal cortex, while the MDm in rodents and the MDpc in primates project to the medial PFC. In addition to the different organization of MD projections, the mPFC itself is also differently organized differently in rodents and marmosets, which makes it even more difficult to compare the MD-mPFC connectivity among the two species (Schaeffer et al., 2020).

To elucidate the homology between these subdivisions across species, unveiling the molecular identities of MD subdivisions is critical. Therefore, in the present study, we explored molecules that were expressed explicitly in MD subdivisions and compared patterns of gene expression and connectivity of MD subdivisions between rodents (mice) and primates (common marmosets, Callithrix jacchus).

Materials and Methods

Animals

All procedures in mice were performed in accordance with protocols approved by the Institutional Animal Care, and Use Committee of RIKEN Wako and Kobe branches (W2020-2-021). ICR (CD1) mice were purchased from Japan SLC. Tacr3-IRES-Cre mice (accession No. CDB0056E: http://www2.clst.riken.jp/arg/mutant%20mice%20list.html) were generated by the Laboratory for Animal Resources and Genetic Engineering (RIKEN Center for Biosystems Dynamics Research). The SERT^Cre knock-in mice used in the present study were provided by Dr. Zhuang (Zhuang et al., 2005). Ai9 (The Jackson Laboratory, JAX#007909, RRID: IMSR_JAX#007909) and R26R mice (Jackson Laboratory, JAX#003309, RRID: IMSR_JAX:003309) were used as reporter mice. The day of birth was designated as postnatal day 0. Both sexes were analyzed and ≥3 individual mice were analyzed in each experiment. Genotyping of tail clips was confirmed by genomic PCR. The following primers were used for the genotyping of Tacr3-IRES-Cre mice: FW: 5’-GAC TCC ACC TAC TTA GAG-3’, REV1: 5’-TCG GCC AGT AAC GTT AGG-3’, and REV2: 5’-CTC CCT ACT CTT TGA GGC TG-3’ (502 bp and 878 bp, respectively).

All procedures in marmosets were performed in accordance with protocols approved by the Institutional Animal Care, and Use Committee of RIKEN Wako branches (W2020-2-022). Marmosets were derived from a breeding colony at RIKEN Center for Brain Science.

In situ hybridization (ISH)

Mice were deeply anesthetized with a lethal dose of pentobarbitone (150 mg/kg). After three failed attempts to elicit a foot withdrawal reflex, the animals were transcardially perfused with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS). Brains were removed and post fixed in the same fixative for 1h at 4°C and 2 h at room temperature and equilibrated with 30% sucrose in 4% PFA at 4°C overnight. Coronal sections of 28 μm thickness were generated with a sliding microtome (SM2020R, Leica). ISH was performed as described previously (Grove et al., 1998). Antisense probes were generated from the template plasmids (FANTOM clone, B230383K19 for Tacr3, K530020M21 for Gnb4, C730036B16 for Necab1, G530130E02 for Ptk2b, E230015I18 for Ucp2, B230325C03 for Shox2, 0610025O15 for Cck and GenBank clone, and Z48800 for Gbx2). Tnnt1 (nucleotides 525–875, NM_001277903.1) was subcloned into a pGEM-T-Easy vector (Promega) and used as a template. The expression patterns of each gene were confirmed in three independent brains.

Neonate marmosets were deeply anesthetized with a lethal dose of pentobarbitone (100 mg/kg). After three failed attempts to elicit a foot withdrawal reflex, the animals were transcardially perfused with 4% PFA in PBS. Brains were removed and fixed with 30% sucrose in 4% PFA at 4°C for several days until sinkage. Coronal sections of 28 μm thickness were generated with a sliding microtome (SM2020R, Leica). ISH was performed as previously described (Mashiko et al., 2012). Probes and images were obtained from the Marmoset Gene Atlas (https://gene-atlas.brainminds.riken.jp) (Shimogori et al., 2018).

Stereotaxic virus injection

The AAV-DIO-EYFP was produced as previously described (Kinoshita et al., 2019). Tacr3^Cre heterozygous mice at P2 or P3 were used for stereotaxic virus injection. The mice were anesthetized on ice for several minutes. After confirmation of the lack of reflex by pinching, pups were placed on a stereotaxic apparatus maintaining hypothermia by ice during surgery. A purified AAV solution (200 nl/site, injection rate 50 nl/min) was injected into the right side of the MD (A/P −2.3 mm, M/L +0.33 mm from bregma, D/V −2.2 mm from the skull surface). Experiments were carried out with KDS Legato 130 (KD Scientific) and stereotaxic frame (Muromachi Kikai).

Immunohistochemistry

Mice were deeply anesthetized (150 mg/kg) and transcardially perfused with 4% PFA in PBS. Brains were removed and post fixed in the same fixative for 2h at 4°C and equilibrated with 30% sucrose in PBS at 4°C overnight. Coronal sections (50 μm thickness) were generated using a sliding microtome (SM2020R, Leica). Sections were blocked for 1h at room temperature with blocking buffer (0.1% Triton X-100 and 5% normal goat serum in PBS) and incubated with rabbit polyclonal anti-RFP antibody (Abcam, Cat#ab62341, RRID: AB_945213), rabbit polyclonal anti-GFP antibody (MBL International, Cat#598, RRID: AB_591819) and/or rat-monoclonal anti-Ctip2 antibody (Abcam, Cat#ab18465, RRID: AB_2064130) at 1:200 in blocking buffer at 4°C overnight. After washing three times in PBS for 10 min, sections were incubated with Alexa-594-conjugated anti-rabbit IgG secondary antibody (Thermo Fisher Scientific, Cat# A11037, RRID: AB_2534095), Alexa-488-conjugated anti-rabbit IgG secondary antibody (Thermo Fisher Scientific, Cat#A11034, RRID: AB_2576217) and/or Alexa-488-conjugated anti-rat IgG secondary antibody (Thermo Fisher Scientific, Cat#A11006, RRID: AB_2534074) at 1:400 in blocking buffer for 2 h at room temperature. After washing three times in PBS for 10 min, sections were counterstained with 1 μg/mL of 4’, 6-diamidino-2-phenylindole (DAPI) (11034-56, Nacalai Tesque).

X-gal staining

Brains were removed and post fixed in the same fixative for 2h at 4°C and equilibrated with 30% sucrose in PBS at 4°C overnight. Coronal sections (28 μm thickness) were generated with a sliding microtome (SM2020R, Leica). After washing in PBS for 10 min, sections were incubated with staining buffer (1 mg/mL X-gal, 5 mM K₃Fe(CN)₆, K₄Fe(CN)₆, 5 mM EDTA, 2 mM MgCl₂, 0.01% deoxycholate, and 0.02% NP-40 in PBS) at 37°C overnight.

Images

Mouse ISH images and X-gal-stained images were acquired using a digital slide scanner (NanoZoomer S210, Hamamatsu Photonics). Marmoset ISH images were obtained from the Marmoset Gene Atlas (https://gene-atlas.brainminds.riken.jp). Fluorescent images were acquired by fluorescence microscopy (BZ-9000, KEYENCE) or confocal microscopy (FV3000, Olympus). The images were imported into Adobe Photoshop (Adobe) or ImageJ (NIH) to adjust brightness and contrast.

Results

Gene expression patterns in mouse MD subdivisions

To investigate the molecular markers labeling each MD subdivisions in mice, we investigated gene expression patterns using ISH in mice on postnatal day (P) 15. This is the earliest stage of development detected by Nissl staining in each subdivision (Fig. 1a). In mice, MD subdivisions were distributed along the medial-lateral and anterior-posterior axes. The anterior part of the MD consisted of the MDm (medial) and MDl (lateral), whereas the posterior part of the MD consisted of the MDm, MDc (central), and MDl (Fig. 1a). Based on multiple tracer analysis in rodents (already well established in rat), each MD region was identified to have a specific target area in the PFC (Alcaraz et al., 2016). A summary of the connections from the MD to the PFC identified by retrograde tracer injections into the PFC is illustrated in Fig. 1a (Alcaraz et al., 2016). Projections from the MDm to the prelimbic cortex, infralimbic cortex, agranular insular area (AID), and basolateral amygdala; from the MDc to the orbital frontal cortex and AID; and from the MDl to the prelimbic cortex, anterior cingulate area, and orbitofrontal cortex (OFC) were observed (Alcaraz et al., 2016; Murphy et al., 2018).

Figure 1.

(a) Nissl staining in P15 mouse brain and schematic of the projection patterns from each MD subdivision (MDm, MDc, and MDl) to the PFC as based on a previous study (Alcaraz et al., 2016). AC, anterior cingulate cortex; PrL, prelimbic cortex; IL, infralimbic cortex; OFC orbitofrontal cortex; AID, agranular insular cortex. (b-f) Representative images of in situ hybridization revealed expression patterns in mice MD subdivisions at P15. Five groups were categorized based on the expression patterns in MD subdivisions. (b) Group 1: Tacr3 and Gnb4 were expressed in the anterior and posterior MDm. (c) Group 2: Tnnt1 in the MDc. X-gal staining revealed Cre-mediated recombination in the MDc of SERT^Cre;R26R mice at P6. (d) Group 3: Necab1 and Gbx2 were expressed in the anterior and posterior MDm and MDl. (e) Group 4: Ptk2 and Ucp2 were expressed in the MDc and in a subregion of the posterior MDl. (f) Group 5: Shox2 and Cck were broadly expressed throughout the entire MD.

Single-cell anterograde tracer labeling provides a detailed projection mapping of MD subdivisions to specific regions and layers of the PFC (Kuramoto et al., 2017; Winnubst et al., 2019). Although tracer injections into the MD revealed projections to parvalbumin-positive (PV+) cells in the dorsal anterior cingulate cortex (dACC) layer 2/3 (Delevich et al., 2015) and vasoactive intestinal peptide (VIP) cells in layer 1b in the prelimbic cortex (PrL) (Anastasiades et al., 2021), the MD subdivisions that project to these specific PFC cell populations have not been systematically investigated. To search for genes expressed in the MD, we used a publicly available ISH database (Allen Brain Atlas: https://portal.brain-map.org) (Lein et al., 2007) and our ISH data to further explore MD-expressing marker genes (Shimogori et al., 2010). Expression patterns were classified into five groups based on the number of molecules differentially expressed in MD subdivisions (Table 1). Group 1 comprised genes showing MDm-subregional expression in both anterior and posterior MD portions (Fig. 1b), Group 2 comprised expression in anterior and posterior MDm and MDl portions (Fig. 1c), Group 3 MDc expression (Fig. 1d), Group 4 expression in the MDc and posterior/dorsal MDl (Fig. 1e), and Group 5 the entire MD (Fig. 1f and Table 1). MDm markers, such as Tacr3 (tacykinin receptor 3) and Gnb4 (guanine nucleotide binding protein beta 4), were expressed in both the anterior MDm (Fig. 1b, upper panel) and posterior MDm (Fig. 1b, lower panel). Group 2 marker genes, such as Necab1 (N-terminal EF-hand calcium binding protein 1) and Gbx2 (gastrulation brain homeobox 2), were expressed in the anterior MDm and MDl (Fig. 1c, upper panel) and in the posterior MDm and MDl but not in the MDc (Fig. 1c, lower panel). Group 3 comprised MDc markers, including Tnnt1 (slow skeletal muscle troponin T) and Slc6A4 (solute carrier family member 4) (Fig. 1d). Slc6A4 (SERT: serotonin transporter) was only transiently expressed in the MDc (until P4) (Fig. 1d) (Narboux-Nême et al., 2008). In order to perform Slc6a4 lineage analysis, we employed Slc6A4 reporter mice (SERT^Cre line) and crossed them with R26R mice (Narboux-Neme et al., 2008). X-gal staining in SERT^Cre × R26R mice revealed labeling in the MDc at P6 (Fig. 1d). We further identified Group 4 genes that were expressed in the MDc and posterior/dorsal part of the MDl, such as Ptk2b (protein tyrosine kinase 2 beta) and Ucp2 (uncoupling protein 2), implying the presence of further anterior/posterior MDl subdivisions based on these molecular expression patterns (Fig. 1e). Genes expressed in the entire MD, such as Shox2 (short stature homeobox 2) and Cck (cholecystokinin), were classified as Group 5 (Fig. 1f).

Table 1.

Molecular profile in mouse MD subdivisions. Mouse genes were categorized into five groups from their expression patterns.

Group1 (MDm)	Group2 (MDm + MDl)	Group3 (MDc)	Group4 (MDc + MDl)	Group5 (MDm + MDc + MDl)
Gnb4	Col25a1	Gabra4	Fam20c	Cck
Kitl	Gbx2	Lynx1	Plekhg1	Ppm1a
Rab37	Htr2c	Nefh	Ptk2b	Shox2
Tacr3	LOC432748	Ptprj	Ucp2	Tshz1
	Necab1	Slc6a4
	Stmn1	Tnnt1
	Stxbp6

Visualization of MDm circuits using Tacr3^Cre mice

In order to examine the specificity of the molecular markers of MD regions for afferent fibers and projection targets, we generated a Tacr3-IRES-Cre mouse line (Tacr3^Cre line) to label the MDm population. Using the CRISPR/Cas9 system, the IRES-Cre cassette was inserted into the 3’-UTR of the Tacr3 gene locus (Fig. 2a). To confirm Cre-mediated recombination in Tacr3-expressing cells, Tacr3^Cre mice were crossed with Ai9 reporter mice (Madisen et al., 2010). tdTomato-positive cells were explicitly detected in the anterior and posterior MDm, and ventral hippocampal (vHip) neurons, with scattered expression in the cortex and ventral tegmental area (VTA), consistent with the observation of Tacr3 signals from our ISH analysis at P10 (Fig. 2b and c). Additional expression was observed in the paraventricular thalamus (PVT, Fig. 2b), which exhibited a weak ISH signal (Fig. 1b). To further investigate detailed projections from the MD to the PFC, we performed stereotaxic injections of AAV-DIO-EYFP virus to specifically label MDm neurons in Tacr3^Cre mice at P2 (Figs. 2d–i). Detailed analysis of the PFC region of virus-injected mice at P14 revealed layer-specific axon terminals in prelimbic cortex (PrL) layers 1b and 3, layers 1 and 3 and deep layers of the infralimbic cortex (IL), and the medial orbital cortex (MO) (Figs. 2g–i). Taken together, these results indicate that Tacr3-positive MDm neurons project to specific layers in mPFC subdivisions.

Figure 2.

(a) Scheme of the targeting strategy for Tacr3-IRES-Cre mice. The guide RNA was designed based on the 3’UTR of the Tacr3 gene locus. The IRES-Cre-bpA cassette was inserted into the 3’UTR of the Tacr3 gene locus. (b) Representative immunohistochemistry images with anti-RFP antibody in Tacr3^Cre;Ai9 at P16. Cre-mediated recombination confirmed by crossing Tacr3^Cre with Ai9 mice. tdTomato-positive neurons observed in the anterior and posterior MDm and ventral hippocampus of Tacr3^Cre;Ai9 mice. Sections counterstained with DAPI. Scale bars, 200 μm and 500 μm. (c) Tacr3 expression in cortex, hippocampus and ventral tegmental area (VTA) at P10, as revealed by ISH. (d–i) Stereotaxic injection of AAV-DIO-EYFP into MD of Tacr3^Cre at P2 revealing the projection from Tacr3-positivie MD neurons to the PFC. Representative immunohistochemistry images with anti-GFP and anti-Ctip2 antibodies in virus-injected-Tacr3^Cre at P14. Projections from the MDm to the PFC. Scale bars, 500 μm. Strong projections observed in the mPFC (ventral part of the PrL and IL) (e) and agranular insular cortex (AID) (f) depicted in high magnification images. Scale bar, 200 μm. (g–i) High magnification images in the PrL (g), IL (h), and MO (i) revealed differentially distributed GFP-positive axon terminals in the cortical layers of each region. Cortical layers were identified based on immunohistochemistry with anti-Ctip2 antibody and DAPI staining patterns. MO, medial orbital cortex. Scale bar, 100 μm.

Visualization of MDc circuits using SERT^Cre mice

In order to distinguish between MDc and MDm projections, MDc projections were visualized using SERT^Cre mice crossed with Ai9 reporter mice (Fig. 3). Consistent with the ISH results and analysis of lacZ staining of SERT^Cre × Rosa26 mice (Fig. 1d), we observed SERT tdTomato-positive MDc glutamatergic neurons and some scattered tdTomato-positive neurons in the anterior MDm (Fig. 3a). As previously reported, strong tdTomato fluorescence was observed in deep-layer neurons in the PFC (Fig. 3b) (Soiza-Reilly et al., 2019). Despite considerable innervation of the OFC by tdTomato-positive fibers, no obvious axon terminals were observed in the PrL and IL (Fig. 3b). Despite the need for technical improvements to more specifically label MDc projections in SERT^Cre mice, our data suggested that the SERT^Cre mouse line is a useful tool for labeling MDc neurons and that MD subdivisions project to distinct target regions and layers, underpinning their divergent connectivity and function.

Figure 3.

(a, b) Representative immunohistochemistry images with anti-RFP antibody in Sert^Cre;Ai9 mice at P9. Scale bars, 200 μm. (a) Cre-mediated recombination confirmed by crossing Sert^Cre mice with Ai9 mice. tdTomato-positive neurons observed in the MDc of Sert-Cre;Ai9 mice. (b) Projections from the MDc to PFC in Sert^Cre;Ai9 mice. Strong projections were observed in the OFC but not in the mPFC, and depicted in high magnification images.

Gene expression patterns in common marmoset MD

In non-human primates, interconnectivity is observed between the MDmc region and OFC, the MDpc region and medial to dorsal PFC, and the lateral multiform part and premotor cortical area (Barbas et al., 1991; Giguere et al., 1988) (Fig. 4a). These afferents innervate layer 3 and layer 5 in primates, while reciprocal connections originate from layer 5/6 (Giguere et al., 1988; Krettek et al., 1977; Van Eden, 1986). Based on the distinct connectivity of MD subdivisions to the PFC in primates, we compared the molecular expression patterns of MD subdivisions in the mouse and common marmoset (Callithrix jacchus). The common marmoset is a small New World monkey that has recently emerged as a novel animal model in neuroscience research (Okano, 2021). A brain mapping project in Japan, Brain/MIND (Brain Mapping by Integrated Neurotechnologies for Disease Studies), focused on structural and functional mapping of the common marmoset brain, provides a comprehensive gene expression mapping in the marmoset brain based on ISH over the Marmoset Gene Atlas (https://gene-atlas.brainminds.riken.jp) (Shimogori et al., 2018).

Figure 4.

(a) Nissl staining in the neonate marmoset brain and schematic of the projection patterns from each MD subdivision (MDmc, MDpc, MDmf, and MDdc) to the PFC. Projections from the MDmc to the OFC, the MDpc to the mediodorsal PFC, and the MDmf and MDdc to premotor cortical areas. (b–d) Representative in situ hybridization images against genes expressed in marmoset MD subdivisions at P0. (b) Group 1: TRH, PYGL, and HAPLN3 expressed in the MDm. (c) Group 2: NECAB1, GBX2, and TNNT1 expressed in the MDm and MDl. (d) Group 3: MEIS2, PCP4L1, and TSHZ1 expressed in the MDc.

According to the marmoset atlas (Paxinos et al., 2012), and similar to mice, MD subdivisions are generally categorized into the MDm, MDc, and MDl. We first analyzed mouse MD gene expression in the neonatal marmoset brain, which is equivalent to mouse P15 (Mashiko et al., 2012). We first searched for Group1 (MDm) marker genes using a gene expression profiling system (GePS), which systematically and automatically identified gene expression patterns in specific regions of the neonatal marmoset brain (Kita et al., 2021). We identified 18 marmoset genes in the MDm (Table 2), the majority of which showed disperse expression. For example, thyrotropin-releasing hormone (TRH), glycogen phosphorylase L (PYGL), and hyaluronan and proteoglycan link protein 3 (HAPLN3) were expressed in the MDm in a scattered pattern with regional expression within the MDm (Fig. 4b). Other mouse MDm markers were not defined in the marmoset MD. For example, TACR3 was not detected in marmoset MDm, despite showing similar expression patterns in the hippocampus and hypothalamus of mice and marmosets (data not shown). We next searched for Group 2 marker genes (expressed in the MDm and MDl, 28 genes, Table 2) in marmosets and observed that both NECAB1 and GBX2 were expressed in the anterior MDm and MDl, similar to the pattern observed in the anterior and posterior MD in mice (Fig. 4c). Further, TNNT1 exhibited Group 2 expression patterns in the marmoset MD (Fig. 4c), although in mice its expression was detected in the MDc (Fig. 1d).

Table 2.

Molecular profile in marmoset MD subdivisions. Marmoset genes were categorized into seven groups from their expression patterns.

Group1 (MDm)	Group2 (MDm + MDl)	Group3 (MDc)	Group4 (MDc + MDl)	Group5 (MDm + MDc + MDl)						Group6 (MDl)	Group7 (MDm + MDc)
CACNA2D3	ALDH2	ADD2	UCP2	AAK1	CCDC184	GABRA5	LY6H	PLEKHG1	SORL1	APRT	CDKL5
CHRM3	ANK2	ARHGAP32	ADAMTS2	AARS1	CCK	GABRB3	LYNX1	PNISR	SPOCK1	ARL10	CNKSR3
CYB5R1	ATP2C1	EGR1	CACNA2D1	ABAT	CD47	GAD1	MAL2	PODXL2	SRGAP2	BHLHE40	CRH
FBXW7	CCK	GABRA3	CADM1	ABLIM3	CDC34	GAP43	MAP2K1	PPM1G	SRPK2	CACNG5	GPR161
GABRA5	CDH6	GRIN1	CBLN1	ACE	CDH8	GIT1	MARK1	PRDX2	SRSF5	CCN2	KCTD1
GNAL	CERK	IGSF21	CHRNA4	ACHE	CDHR1	GLRB	MBOAT7	PRDX6	ST6GALNAC5	CCNDBP1	MAPK11
GRIN3A	DGCR2	ITPR1	CUX1	ACLY	CHD3	GLYR1	MEF2A	PRICKLE1	STMN4	CLVS2	NECTIN1
GRM8	GABRG2	KCNQ3	GAD2	ADARB1	CHRNB2	GNAS	METTL9	PRKCD	STXBP6	CSDC2	NPNT
HAPLN3	GBX2	KIT	GJD2	ADCYAP1R1	CIAPIN1	GPM6B	MICU1	PRKCG	SUSD4	DNAJC7	NTN4
HTR7	GFAP	MAGEL2	GNB4	ADGRL1	CIT	GPR153	MRAP2	PRPH	SYN1	EIF4H	PDZD2
MAP2K2	MAPK1	MEIS2	LMO1	ADM	CLASP2	GRID2IP	MTCH2	PRR12	SYNGAP1	ENTPD2	PMP22
NRSN2	MAPK3	N4BP2	NDNF	AFF2	CLIP2	GRM4	MTMR9	PTPN3	SYNJ1	EPHA5	PTK2B
PYGL	MRPL38	PCP4L1	NETO1	AGPAT1	CLMN	GRM5	NBEA	PTPRO	SYNPR	FARSA	SCUBE1
RTN4R	MSMO1	RELN	NR4A1	AKAP8L	CLMP	GSG1L	NCDN	RAB1A	TBC1D24	GPRIN3	SLC6A3
SCN1B	NECAB1	TSHZ1	NRXN3	ALDH5A1	CLSTN2	GSK3A	NDUFB7	RASD2	TGFA	GPX3
SCN8A	NFXL1	ZADH2	OTX2	AMPH	CNTNAP4	GSTA4	NECAB2	RASGRP2	TIMP2	GRM7
SDHB	OPRM1	ZNF521	PCDH19	ANK3	COL25A1	GTF2I	NECAP1	RASSF4	TM2D3	IRS4
TRH	PAK6		ROBO1	ANOS1	CORO2B	GUCY1A1	NEFM	RBFA	TMEM179	KLHL1
	PKIA			ANP32A	CPLX3	HRH3	NLGN4X	RBFOX1	TMX4	KMT5A
	PLCH2			ANXA2	CPNE6	HS6ST3	NPTX2	RCAN2	TOMM20	LYPD1
	SCN1B			AP3B2	CPNE7	HSD17B12	NPTXR	REEP2	TOMM70	OPCML
	SLC17A6			APBA1	CRTAC1	HSPD1	NR1D1	RORA	TOX4	PCSK5
	SLC6A7			ARCN1	CSPG5	HTR7	NREP	RTN4RL2	TPPP3	PDE1B
	SPARC			ARL6IP5	CTNNA2	IDH2	NRIP3	SAMD14	TRPC3	PITPNC1
	TNNT1			ARNTL	CTNNB1	IGFBP3	NRN1	SCG2	TSPAN2	PPM1A
	TSPYL2			ATP1B2	CTTN	IGFBP6	NRXN2	SCN1B	USP11	PRDX3
	TSPYL4			ATP2A2	DEXI	INA	NSDHL	SCN3B	USP9X	RADIL
	VAV3			ATP6V0A1	DGKZ	INPP5A	NSG2	SDK2	VAT1	RANGAP1
				ATP6V1C1	DLAT	INSYN1	NTNG1	SEMA6D	VDAC2	SHOX2
				ATP8A2	DLG4	INSYN2A	NTRK2	SEMA7A	VGF	SNCA
				ATRX	DLGAP4	JMJD1C	NTS	SERPINE2	VSTM2B	SV2B
				AUTS2	DMTN	KCNAB2	NUTF2	SERPINF1	VWC2	SYN2
				B3GALT2	DPP6	KCNH2	NYAP2	SETBP1	WDR47	TMEM215
				BACE1	DPYSL5	KCNH8	OGFRL1	SETD2	WLS
				BBS2	DSCAM	KCNMA1	OPRK1	SEZ6	WNK2
				BDNF	EDEM3	KCTD17	PACS1	SFXN1	WNT4
				BRAF	EFNA5	KCTD2	PAK3	SHANK2	WNT7B
				BRINP1	EFNB3	KITLG	PAK5	SHOC2	ZDHHC14
				BRSK2	EPC1	L1CAM	PAM	SIPA1L2	ZIC1
				BTBD3	EPHA4	LEF1	PAMR1	SLC12A5	ZIC4
				CA10	FAM3C	LGALS1	PCDH10	SLC17A6	ZNHIT1
				CACNA1A	FAM43A	LGI4	PCDH17	SLC24A3
				CACNA1C	FAM49A	LHFPL3	PCDH9	SLC29A4
				CACNA1E	FAM49B	LHFPL6	PCP4	SLC2A13
				CACNA1G	FHOD3	LHX9	PDE5A	SLC39A10
				CACNA1H	FIGN	LIMK1	PEX5	SLC39A6
				CACNG8	FNBP1L	LIN7A	PHACTR1	SLC8A1
				CALB2	FOS	LRP8	PI4KA	SLC9A6
				CAMKV	FOXP2	LRRC3B	PITPNA	SMG7
				CBLN4	FXYD6	LRRTM1	PJA2	SNAP91

We searched for Group 3 marker genes using the GePS and identified 17 genes expressed in the MDc (Table 2). The majority of these marker genes were dispersely expressed with some regional specificity. Meis homeobox 2 (MEIS2) and teashirt zinc finger homeobox 1 (TSHZ1) were expressed in the MDc in a scattered manner, whereas Purkinje cell protein 4 like 1 (PCP4L1) was more strongly expressed only in the lateral portion of the posterior MDc (Fig. 4d). For Group 4 marker genes (genes expressed in the MDc and MDl, 18 genes, Table 2), we identified GNB4, UCP2, and NDNF (Fig. 5a). GNB4, categorized as a Group 1 (MDm) marker gene in mice, was observed in the MDc and anterior MDl in marmosets (Fig. 5A, green arrow). NDNF was expressed in the MDc and MDl with regional specificity in marmosets (Fig. 5a, magenta arrows), but its expression was not evident in the mouse MD (data not shown), highlighting the differences in marmoset and mouse MD marker expression. We also searched for genes expressed in all MD subdivisions (291 genes expressed in the entire thalamus were excluded, Group 5, Table 2). Our search for genes expressed in the MDl but not in the MDm and MDc in marmosets using the GePS revealed 33 genes (Group 6) including short stature homeobox 2 (SHOX2), synaptic vesicle glycoprotein 2B (SV2B), and protein phosphatase Mg²⁺/Mn²⁺ dependent 1A (PPM1A) (Fig. 5b and Table 2).

Figure 5.

(a–c) Representative in situ hybridization images against genes expressed in marmoset MD subdivisions at P0. Scale bar, 1 mm. (a) Group 4: GNB4, UCP2, and NDNF expressed in the MDc and MDl. Green and magenta arrows indicate subregional expression, respectively. (b) SHOX2, SV2B, and PPM1A expressed in the MDl. (c) PTK2B, NTN4, and SLC6A3 expressed in the MDm and MDc. (d–f) Pseudo-colored ISH images of ADCY2 (CL), SHOX2 (MDl), and NTN4 (MDm and MDc) revealing the MDl borders. Overlaid pseudo-colored ISH images revealing the border between MDl and adjacent regions.

Finally, we searched for genes expressed in the MDm and MDc (not identified in the mouse, Group 7, 14 genes, Table 2). We observed that PTK2B, netrin 4 (NTN4), and solute carrier family 6 member 3 (SLC6A3, also known as DAT) were expressed in MDl. In contrast, Ntn4 and slc6a3 were not detected in the mouse MD, whereas Ptk2b showed different expression in mouse MD than Group 4 (Table 1). The MD is separated from the ventral thalamic nuclei by the central lateral thalamic nuclei (CL), which are easily distinguished with simple Nissl staining (Jones, 2007). We used a CL marker, adenylate cyclase 2 (ADCY2), as a landmark to confirm MDl marker genes. Pseudo-colored ADCY2 ISH was overlaid on a pseudo-colored SHOX2 ISH, which revealed a non-overlapping expression pattern (Fig. 5d). Overlay of MDm and MDc marker genes, such as NTN4, delineated precise MDl patterns in the non-overlapping region (Fig. 5e). The NTN4 and SHOX2 ISH combination labeled the entire MD, suggesting that we identified marker genes that delineated all the marmoset MD subdivisions (Fig. 5f). Collectively, we identified dynamic gene expression profiles in mouse and marmoset MD subdivisions and cell types, highlighting partially overlapping patterns between these species. Nevertheless, many genes were differentially expressed in mice and marmosets MD subdivisions, supporting the concept that heterogeneous MD cell types and diverse connectivity patterns to distinct target regions may underpin functional differences between MD subdivisions.

Discussion

The mediodorsal thalamus (MD) is a higher-order nucleus located in the central thalamus. MD subdivisions are connected to specific PFC regions and layers, which play important roles in cognitive control, working memory, and decision-making. We identified several molecular markers of mouse MD subdivisions, which could be used to label MD-PFC projections using a combination of genetically modified animals and in utero electroporation. We also compared MD marker expression in common marmosets to define the molecular similarities and differences between the mouse and marmoset MD. Notably, the genes expressed in MD subdivisions partially overlapped in mice and marmosets, but the majority exhibited different expression patterns. These results suggest that MD subdivisions in mice and marmosets differ at the molecular level.

MD subdivisions in evolution

The MD is widely recognized as the largest nucleus in the medial thalamus in rodents. Its size is relatively larger in humans, and this volumetric increase is thought to parallel the volumetric increase of the PFC in humans. MD subdivisions are well segregated in primates based on cytoarchitecture and chemoarchitecture (Jones, 2007). Our ISH analysis revealed many genes showing differential expression patterns in MD subdivisions of marmosets and mice, suggesting that MD subdivisions can also be described based on gene expression patterns in addition to anatomical segregation.

In primates, MDmc (MDm in rodents) neurons project to the OFC, and MDpc (MDc in rodents) neurons project to the medial PFC. In mice, MDm neurons project to the ventral PrL and IL, and MDc neurons project to the OFC. These results suggest that the cell types in these two MD subdivisions may differ between rodents and primates (Kuroda et al., 1998; Morecraft et al., 1992; Ray et al., 1993). Indeed, few genes exhibited the same expression patterns in the MDm and MDc in both mice and marmosets. For example, Gnb4, which was expressed in the mouse MDm, was found in the marmoset MDc, while Tnnt1, in the mouse MDc, was expressed in the marmoset MDm and MDl. A recent transcriptomic analysis delineated comprehensive gene profiling of MD subdivision neurons in mice (Phillips et al., 2019). Although the Marmoset Gene Atlas provides abundant gene expression profiles, comprehensive analyses such as single-cell-based RNA sequences will facilitate further comparisons of gene expression patterns between mouse, marmoset, and other non-human primate MD neurons.

The MDl can be further subdivided into pars multiformis subdivisions (MDmf or rostral MDl) and densocellular subdivisions (MDdc or caudal MDl) in primates (Barbas et al., 1991; Giguere et al., 1988; Goldman-Rakic et al., 1985). However, this cytoarchitectonic definition may not translate to rodents. A previous report demonstrated that anterior MDl neurons project predominantly to the dorsal part of the mPFC (Alcaraz et al., 2016), whereas the posterior MDl contains dorsal mPFC-projecting neurons and OFC-projecting neurons, suggesting that MDl neurons can be further subdivided based on projection patterns. In this study, we observed that Ptk2b and Ucp2 were expressed in the posterior MDl but not in the anterior MDl (Fig. 1e), indicating the presence of further anterior/posterior subdivisions of MDl as shown by ISH analysis.

The majority of genes expressed in the neonatal marmoset MD were not restricted only to the MD, and their expression domain often showed expression in neighboring dorsal thalamus nuclei, including the Pulvinar (Pul) (data not shown). The majority of genes expressed in the Pul were also expressed in other dorsal thalamic nuclei, which suggests a possible shared identity and potentially common origin; this needs further investigation.

Specific MD subdivision-PFC circuits

In this study, we established a Tacr3^Cre mouse line that genetically labels MDm neurons. Single-cell morphological analysis revealed that the MDm contained mPFC-projecting neurons and AID-projecting neurons, indicating that MDm neurons have heterogeneous projection patterns (Kuramoto et al., 2017). Stereotaxic virus injection in Tacr3^Cre mice visualized projections from the MDm to the mPFC and AID (Fig. 2). SERT^Cre mice labeled MDc neurons, which enabled visualization of the MDc-OFC circuit (Fig. 3). Taken together, in combination with optogenetic tools, Tacr3^Cre and SERT^Cre mice will enable high-resolution dissection of MDm-mPFC and MDc-OFC-specific circuit function. Derivation of an MDl-specific Cre mouse line will provide novel insights into MD subdivision-specific neural circuits and their specific functions.

The roles of MD subdivisions in cognitive function

The critical roles of MD-PFC neural circuits in cognitive function was demonstrated using MD lesions in rodents and non-human primates. These studies, especially in non-human primates, provided MD subdivision-specific functions based on lesions of each MD subdivision. For example, ablation of the MDmc induced PFC hypoactivity and disrupted cognitive function (Parker et al., 1998). Furthermore, selective damage in the MDmc disrupted learning and memory (Gaffan et al., 2000; Mitchell et al., 2007). Similar deficits in cognitive function were observed in animals with bilateral ablation of the OFC and ventral PFC (Baxter, Browning, et al., 2008; Baxter et al., 2007). However, these deficits were not observed in animals with ablation of the dorsolateral PFC (Baxter, Gaffan, et al., 2008).

In rodents, ablation of the MD disrupts working memory, which depends on mPFC function (M’Harzi et al., 1991; Stokes et al., 1990a, 1990b), suggesting that the MD-mPFC circuit underscores working memory in rodents. Recently, the function of this circuit in working memory was confirmed using optogenetics. Inhibition of the MD-mPFC circuit disrupted performance in a delayed non-matching-to-sample T-maze test. In contrast, inhibition of the MD-OFC circuit did not affect performance in this task (Bolkan et al., 2017). These results suggest that each MD subdivision contributes to higher cognitive functions, but their precise roles remain unclear. Therefore, the elucidation of MD subdivisions and cell types in both rodents and primates is necessary to further investigate the specific roles of MD subpopulations in cognitive function.

Cell type specific MD dysfunction in schizophrenia

The MD has been implicated in the pathogenesis of schizophrenia. Strong evidence for the relationship between the MD and schizophrenia is based on the smaller MD volume in patients with schizophrenia (Young et al., 2000). Despite controversy regarding volumetric MD reduction in schizophrenia due to technical limitations (Cullen et al., 2003; Young et al., 2000), a recent study using high-resolution MRI analysis of a large sample provided support for smaller MD volume in patients with schizophrenia (Pergola et al., 2017). Several neuroimaging studies have reported disruption of the MD-PFC neural circuit in patients with schizophrenia (Anticevic et al., 2014; Giraldo-Chica et al., 2017). In addition, a previous study demonstrated that decreased neuronal activity in the MD disrupted the MD-PFC circuit and cognition in mice (Parnaudeau et al., 2013). These results strongly suggest that the MD-PFC circuit contributes to cognitive symptoms in schizophrenia.

Several risk genes for schizophrenia have been identified in schizophrenia models as pathogenic mechanisms underpinning schizophrenia, such as 22q11.2 deletion syndrome (22q11DS). This syndrome results from the hemizygous deletion of a part of human chromosome 22, a genomic locus is conserved in the mouse genome. Therefore, a mouse model (Df(16)1/+) carrying this genomic deletion has been established (Lindsay et al., 1999), which strongly supports the presence of genetic risk factors for schizophrenia. For example, in this schizophrenia model, Dgcr8 disruption, which controls miRNA biogenesis, induced behavioral abnormalities similar to those in schizophrenia (Stark et al., 2008). Although Dgcr8 is ubiquitously expressed, the thalamus-specific Dgcr8 deletion identified in 22q11DS disrupted the thalamus-amygdala neural circuit and fear memory (Eom et al., 2017). In addition, several risk genes from schizophrenia Genome Wide Association Study (GWAS) data show the expression in marmoset MD (PAK6 in Group 2 and ATP2A2, CACNA1C, CNTNAP4, DGKZ, NLGN4X, SNAP91, SRPK2 and SYNGAP1 in Group 5 of Table 2).

In the present study, we established an MDm-Cre mouse line, Tacr3^Cre (Fig. 2). A Tacr3^Cre mouse line in combination with a floxed mouse line with schizophrenia risk genes (MD-specific conditional knockout mouse of risk genes) may reveal the sufficiency of MD-specific gene ablation to disrupt the MD-mPFC neural circuits responsible for the behavioral impairments observed in schizophrenia. As described above (Parnaudeau et al., 2013), it will also be useful to investigate the involvement of MD neuronal activity in cognitive function using MD-Cre mouse lines combined with manipulation of neuronal activity. These specific analyses of MD-PFC circuits will provide novel insights into the pathogenesis of schizophrenia.

This is the first report revealing the detailed molecular organization of MD subdivisions in mice and marmosets. These molecular markers allowed us to analyze subdivision-specific circuits and functions by means of transgenic mouse model. Moreover, a subdivision-specific promotor AAV line will help us understand primate MD subdivision-specific circuit formation and function in non-human primates. In the future, these technologies can provide further insights into species-specific MD function and dysfunction.