Human fetal tissue from terminated pregnancies was obtained from the joint MRC/Wellcome Trust-funded Human Developmental Biology Resource (HDBR)¹ Gerrelli et al., 2015). All tissue was collected with appropriate maternal consent and approval from the Newcastle and North Tyneside NHS Health Authority Joint Ethics Committee. Fetal samples ranging in age from 8 to 16 PCW were used. Ages were estimated from foot and heel to knee length measurements according to Hern (1984). Six samples in total were studied, 2 at 8PCW, 3 at 10 PCW, 1 at 14 PCW, 1 at 15 PCW and 1 at 16 PCW. Samples were dissected, fixed in buffered 4% paraformaldehyde solution and embedded in paraffin and then sectioned according to standard protocols applied by the HDBR staff.²

This was carried out on 8 μm thick paraffin sections. Antigen retrieval involved boiling in 10 mM citrate buffer pH6 for 10 min. After washing in Tris buffered saline pH7.6 (TBS), sections were incubated with primary antibody (diluted in 10% normal blocking serum in TBS overnight at 4°C. Details of primary antibodies are found in Table 1. Following 3 washes in TBS, sections were incubated with HRP-conjugated secondary antibody for 30 min (ImmPRESS HRP IgG [Peroxidase] Polymer Detection Kit, Vector Labs) washed and then developed with diaminobenzidine solution (Vector Labs supplied kit) washed, dehydrated, and mounted using DPX (Sigma-Aldrich). For double immunofluorescence, the Tyramide Signal Amplification (TSA) method was used permitting double staining using same species antibodies. At the secondary antibody stage, sections were incubated with HRP-conjugated secondary antibody for 30 min (ImmPRESS HRP IgG [Peroxidase] Polymer Detection Kit, Vector Labs) and then incubated in the dark for 10 minutes with fluorescein tyramide diluted at 1/500 (TSA) fluorescein plus system reagent (Perkin Elmer, Buckingham, United Kingdom) or OPAL 520 diluted 1;700 (Akoya Biosciences, Marlborough MA, United States) leaving fluorescent tags covalently bound to the section. Sections were then boiled in 10 mM citrate buffer pH6 to remove all antibodies and unbound fluorescein then incubated first in 10% normal serum then with the second primary antibody for 2 h at room temperature. Sections were again incubated with HRP-conjugated secondary antibody followed by CY3 tyramide for 10 min ([TSA] CY3 plus system reagent, Perkin Elmer) or OPAL 570 (Akoya Biosciences). Sections were dyed with 4’,6-diamidino-2-phenylindole dihydrochloride (DAPI; Thermo Fisher Scientific) and mounted using Vectashield Hardset Mounting Medium (Vector Labs). Extensive washing of sections was carried out between all incubations (for more details see Alzu’bi et al., 2022).

RNA in situ hybridization experiments were performed using the RNAscope^® technology, which has been previously described (Wang et al., 2012). Paired double-Z oligonucleotide probes against target RNA for SOX14, ZIC4, and CNTNAP2 were designed and supplied by ACD Bio Techne (Abingdon, United Kingdom) and described in Table 1. The RNAscope Reagent Kit (ACD Bio Techne) was used according to the manufacturer’s instructions but with slight modifications. In brief, 8μm-thick paraffin sections were baked on a heating pad for ten minutes at 60°C, dewaxed in Xylene, and then boiled with target retrieval buffer (ACD) for 20 min at 95°C. Protease digestion was carried out at 40°C for 30 min, followed by probe hybridization for 2 hours at 40°C with target probes. The hybridized signals were amplified by a cascade of signal amplification molecules and detected with the RNAscope 2.5 HD detection kit (Fast Red). Slides were counterstained with 50% hematoxylin and positive signals showed as red chromogenic dots in the cytoplasm or nucleus. Each sample was quality controlled for RNA integrity with a probe specific to the housekeeping gene GAPDH. Negative control background staining was evaluated using a probe specific to the bacterial dapB gene (Alzu’bi et al., 2022).

In situ hybridization was carried out first as described above except that the hybridized signals were detected with either Cy3 tyramide diluted 1:500 or Opal 570 diluted 1:700. Then Immunofluorescent staining was carried out according to previously described protocols (Alzu’bi et al., 2022) with the antibodies used described in Table 1. Briefly, sections were boiled in 10 mM citrate buffer pH6, followed by incubation with primary antibody (diluted in 10% normal blocking serum in Tris buffered saline [TBS] pH 7.6) overnight at 4°C. Sections were then incubated with HRP-conjugated secondary antibody for 30 min (ImmPRESS HRP IgG [Peroxidase] Polymer Detection Kit, Vector Labs). Signals were detected by incubation with fluorescein tyramide or OPAL 520 as described above. Sections were counterstained with 4′, 6-diamidino-2-phenylindole dihydrochloride (DAPI; Thermo Fisher Scientific, Cramlington, United Kingdom) and mounted using Vectashield Hardset Mounting Medium (Vector Labs, Peterborough, United Kingdom).

Three dimensional reconstructions of the human embryonic and fetal brain, derived from micro computerized tomography scans of whole fetuses, were used to deduce the plane of sectioning and included with the figures presented. The reconstructions were provided by the Human Developmental Biology Resource at https://hdbratlas.org/fetal-stages/8pcw.html, https://hdbratlas.org/fetal-stages/10pcw/, and https://hdbratlas.org/fetal-stages/13pcw.html.

Brightfield images were captured using a using Leica SCN400 Slide Scanner. Fluorescent images were obtained with a Zeiss Axioimager Z2 apotome. Processing of images, which involved adjustment of brightness, color balance and sharpness with minimal removal of artifacts, was achieved using Adobe Photoshop software.

The thalamus was clearly recognizable in haemotoxylin and eosin stained sagittal sections as an ovoid structure dorso-anterior to the mesencephalic flexure by 8 post-conceptional weeks (Figure 1A). The precise delineation of the thalamus could be ascertained by examining expression of SHH alongside that of 3 transcription factors, ZIC4, GBX2 and FOXP2. SHH expression marks the Zona limitans (ZLI; the embryonic boundary between p2 (containing the thalamus) and p3 (containing the prethalamus; Kiecker and Lumsden, 2004). It was strongly expressed in cells within the ventricular zone (VZ) at the boundary and in a smaller number of cells along the ZLI (Figure 1B). The alar portion of p2 was characterized by strong expression of ZIC4 throughout including the dorsal most epithalamus and right up to and including the rostral thalamus close to the ZLI (Figure 1C) although expression was relatively limited in the prethalamus, unlike in the mouse where strong expression in the prethalamus is reported at the equivalent stage of development (E11.5-E12.5) (Li et al., 2018). GBX2 and FOXP2 immunoreactivity was confined to the thalamus and excluded from the epithalamus and showed largely uniform expression throughout (Figures 1D,E) although FOXP2+ cells were more prevalent in the VZ of the thalamus than GBX2+ cells, confirming our previous study (Alzu’bi et al., 2019). OTX2 was also expressed in the thalamus but showed a different pattern of expression, being strongly expressed in both the VZ and post-mitotic zones. This could be predicted from animal models where it has been shown to promote a glutamatergic identity for progeny of thalamic progenitor cells (Puelles et al., 2006). OTX2 was expressed uniformly throughout the thalamus in medial sections, but in lateral sections became confined to regions close to the ventricles (Figures 1F,G). This suggests that OTX2 is expressed by thalamic progenitor cells and more immature, possibly migratory thalamic neurons with expression downregulated as the neurons mature. SP8, a transcription factor associated with GABAergic neurons derived from the ventral telencephalon (Waclaw et al., 2006; Ma et al., 2013; Alzu’bi et al., 2017a) was confined in its expression to the GABAergic prethalamus and not observed in the thalamus in this plane of section (Figure 1H).

Coronal sections (locations shown in Figures 2A–A”) confirmed these observations, SHH expression was clearly observed in a small group of cells in the thalamic VZ just dorsal to the hypothalamic sulcus (Figure 2B). Gene expression in this small region differed from the VZ both dorsal and ventral to it, showing absence of immunoreactivity for NR2F2 and PAX6 (Figures 2C,D). NR2F1 and ZIC4 were expressed throughout prosomere 2 including the VZ and the epithalamus (Figures 2C,E) whereas FOXP2 immunoreactivity was confined to the thalamus except for the VZ and excluded from the epithalamus. Expression of FOXP2 was also relatively weak in more lateral regions of the thalamus, compared to ZIC4 and NR2F1. PAX6 was expressed throughout the VZ of p2 (Figure 2D). However, there was a ventral low to dorsal high gradient of expression. Some PAX6+ cells were present in the subventricular zone of the thalamus (SVZ). Ventral to the ZLI, PAX6 immunoreactivity was high in the VZ of the prethalamus, and PAX6 expressing post-mitotic cells were present throughout the prethalamus, medially to laterally (Figure 2D). NR2F2 was also expressed strongly in the prethalamus, both ventricular and post-mitotic layers (Figure 3B) as well as in the thalamic VZ in a ventral high to dorsal low gradient, that is, the opposite of PAX6. It was also expressed in post-mitotic cells in more medial locations in ventral parts of the thalamus.

Figures 2D–F’ compare more anterior and posterior sections of thalamus for expression of three transcription factors that are key to thalamic development. FOXP2 expression was seen throughout the thalamus, but was absent from the epithalamus and weakly expressed in more lateral regions of the anterior thalamus (Figures 2D,D’). ZIC4, on the other hand, was strongly and uniformly expressed throughout the anterior thalamus and epithalamus, but showed weaker expression posteriorly, confined to the epithalamus and medial and dorsal thalamic areas close by Figures 2E,E’. Localization of OTX2 immunoreactivity in prosomere 2 was more similar to the FOXP2 expression pattern, except that is was expressed in the epithalamus, strongly expressed in the ventricular zone, and was more weakly expressed laterally, especially anteriorly (Figures 2F,F’).

As predicted from animal experiments (Puelles et al., 2013) expression patterns of transcription factors and other molecules clearly delineate p2 from p1 and p3, with p2 and p3 being separated by the SHH positive ZLI. The thalamus can also be distinguished from the epithalamus at this stage, with only ZIC4, OTX2 and PAX6 being expressed in the epithalamus. The thalamic VZ is divided into zones, with stronger PAX6 expression dorsally and stronger NR2F2 expression close to the ZLI. In post-mitotic zones, ZIC4 is expressed in a high anterior to low posterior gradient, whereas the gradient of expression of FOXP2 and OTX2 opposes this.

In sagittal sections at 10 PCW (Figure 3A) the thalamus had become more elongated dorsally to ventrally, compared to 8 PCW and appeared bean shaped (Figure 3B). Its full extent was clearly delineated by FOXP2 immunoreactivity, but by this stage GBX2 expression was confined either to dorsal posterior regions of the thalamus, or close to the 3rd Ventricle (Figure 3B). FOXP2 was also expressed in numerous other brain regions nearby, while GBX2 was also expressed by cells of the basal forebrain and basal ganglia, presumably precursors of cholinergic interneurons of the basal ganglia, and non-cholinergic neurons of the basal forebrain as observed in mouse (Chen et al., 2010). Other gene expression patterns revealed the boundaries of the thalamus. PAX6 and NR2F2 were strongly expressed in the prethalamus (Figures 3C,D) although weak NR2F2 immunoreactivity was also observed in the ventral posterior thalamus. GAD67 immunoreactive neurons and axons were excluded from the thalamus but were abundant in the prethalamus, subthalamus, hypothalamus, basal ganglia and pretectum (Figure 3E). Similarly, OTX2 expression marked the pretectum and thalamic eminence, but could not be seen in the thalamus in these sections (Figure 3F).

Coronal sections (Figure 4A) confirmed these findings. Figure 4B clearly shows FOXP2 immunoreactivity throughout the extent of the thalamus but excluded from the habenula (derived from the epithalamus) with NR2F2 strongly expressed in the prethalamus, but also in the thalamic VZ. As at 8 PCW, at a relatively anterior level, ZIC4 was expressed throughout both thalamus and habenula, in mitotic and post-mitotic cells. Co-staining with FOXP2 revealed that ZIC4 showed a relatively homogenous expression, whereas FOXP2 showed relatively stronger expression in mediodorsal regions (Figure 4C).

Horizontal sections (Figures 5A,B) further confirm and expand upon these findings. FOXP2 expression delineated the extent of the thalamus whereas GBX2 immunoreactivity was confined to medial regions just lateral of the VZ but also to posterior-lateral parts of the thalamus (Figures 5C,D). Interestingly, in regions where GBX2 expression was reduced, calbindin (CALB1) immunoreactivity was observed (Figure 5E). In the ventricular zone, OLIG3 expression confirmed the extent of p2 as observed in mouse (Figure 5F; Takebayashi et al., 2002; Vue et al., 2007) and suggested by human ScRNAseq studies (Kim et al., 2023). NR2F2 was also expressed in the thalamic VZ in a gradient with strong expression posteriorly and weak anteriorly and was also expressed by cells in the posterior-lateral thalamus, but not in exactly the same locations as where GBX2 immunoreactivity was observed (Figures 5D,G). PAX6 was also expressed in a gradient in the VZ with strong expression posteriorly and weak anteriorly (Figure 5H). PAX6 expression was reduced in the VZ just at the boundary with the pretectum, but was expressed in the VZ and some post-mitotic cells posterior of the boundary.

In summary, the boundaries of the thalamus can be clearly delineated by the expression patterns of various mRNA and proteins. The VZ of the thalamus continues to show some evidence of a protomap, with gradients of expression of both PAX6 and NR2F2 observed. Although defined thalamic nuclei were not identifiable at this stage, post-mitotic cells also show evidence of regionalized gene expression by 10 PCW, with GBX2 expression, in particular, becoming confined to medial and posterior locations and CALB1 immunoreactivity occupying complementary lateral locations. FOXP2 expression, while ubiquitous, was stronger in posterior and medial locations.

We considered two series of sections through the thalamus, one extensive set from a 14 PCW specimen stained for multiple mRNAs and proteins and cut in a plane intermediate between horizontal and coronal, the exact plane deduced by comparing the histological sections with virtual sections from a 3-D MRI image of the fetal brain supplied by the Human Developmental Biology resource (Figures 6A,E,H, 7A,D,G, 8A)³ and a smaller set of coronal sections from a 15 PCW immunostained for a limited number of proteins (Figure 9). By 14 PCW it was clear that the thalamus was becoming divided into discrete regions identifiable by patterns of gene expression and containing relatively cell dense and cell poor regions (Figures 6, 7). FOXP2 localization provided the most striking visualization of putative thalamic nuclei (Figures 6B,F,I, 10B). In more dorsal and anterior sections (Figure 6A) the extent of the pulvinar complex was delineated by moderate to strong FOXP2 expression, with expression stronger in the lateral and inferior pulvinar regions compared to medial pulvinar and the lateral geniculate nucleus (LGN), which we were able to observe as a discrete nuclei by this developmental stage (Figure 6B). The LGN occupied a more dorsal location lateral to the main body of thalamus than where it is found in the adult thalamus. As has been previously described, the late developing pulvinar gradually displaces the LGN in a latero-ventral direction (Hitchcock and Hickey, 1980). The LGN had not adopted a laminar structure by 14 PCW also confirming what was reported previously (Hitchcock and Hickey, 1980). The LGN which also showed strong expression of SP8 (Figure 6C). ZIC4 was no longer expressed in all of the anterior part of p2 but was still strongly expressed in the putative habenula and the antero-dorsal edge of the thalamus at this level, as well as some medial parts of the pretectum. It was also strongly expressed in the LGN (Figure 6D).

In sections located more posteriorly and ventrally (Figures 6E–G) distinct boundaries between thalamic nuclei, as well as between thalamus and pretectum, were observed. FOXP2 expression was very strong in the putative centromedian nucleus (CM) and parafascicular complex (pF). It was also moderately expressed close to the midline and in the MGN, but immunoreactivity was less in lateral parts of the thalamus (Figure 6F). Strong ZIC4 expression was confined to the anterior thalamus and LGN, where FOXP2 expression was low, but was absent from the CM-PF where FOXP2 expression was high (Figures 6F,G). The most posterior and ventral set of sections (Figures 6H–J) offered a different perspective. Here, the paraventricular complex stained strongly for FOXP2, otherwise there seemed to be a gradient of increasing FOXP2 from medial to lateral, distinguishing a putative ventromedial nucleus from more lateral structures (Figure 6I). ZIC4 showed more widespread expression at this level but was strongest in the anterodorsomedial parts of the section (Figure 6J).

NR2F1 and NR2F2 are involved in the patterning of the human telencephalon and show partly complementary, partly overlapping patterns of expression during early fetal development (Alzu’bi et al., 2017a). In anterior and dorsal sections (Figure 7A) NR2F1 showed high expression in the diencephalon which was largely exclusive to the thalamus but with relatively higher expression in the LGN (Figure 7B) confirming a previous scRNAseq/spatial transcriptomic study in mid-gestation human brain (Kim et al., 2023). NR2F2, showed a partly complementary pattern, with strong immunoreactivity in the pretectum, but it was also expressed in the medio-posterior parts of the thalamus (Figure 6D), as was observed at 10 PCW (Figure 5G). In sections located more posteriorly and ventrally (Figures 7D–F) again, NR2F1 expression was strong in the thalamus and weak in the pretectum (Figure 7E). Expression was particularly marked in the LGN and in an anterior region that lacked FOXP2 immunoreactivity but showed strong expression of ZIC4 (Figures 6F,G). NR2F2 expression was strong in the pretectum and extended into the most ventroposterior half of the thalamus, covering the CM-pF complex and also areas lateral to it, but excluding the LGN (Figure 7F). In the most posterior and ventral set of sections NR2F1 showed a similar pattern of expression as seen more dorsally but was also detected in the thalamic reticular formation, which predominantly originates from p3 (Puelles et al., 2013; Puelles and Rubenstein, 2003) and forms a lateral boundary to the thalamus (Figure 7H). The reticular formation was likewise strongly immunoreactive for NR2F2, which was also expressed in the pretectum and midbrain but extended expression into the dorsomedial and ventrolateral regions of the thalamus as well (Figure 7I).

The observation of SP8 expression in the LGN at 14PCW was a surprising finding so we then further examined SP8 immunoreactivity at 8 PCW in comparison to thalamic markers (Figures 10A–D) and found evidence of low levels of expression within the thalamus, close to the border with the prethalamus anteriorly, where SP8 was expressed strongly (Figure 10A) and pretectum posteriorly, where SP8 was absent (Figure 10C). A defined group of SP8+ cells could be seen ventrolaterally in posterior sections (Figure 10C). We also examined several levels at 14 PCW and found it to be a reliable marker for the LGN. At more ventral levels, it was also expressed along the posterior ventral boundary and in medial anterior regions. Both areas expressed GBX2 and in a double labeled preparation we found, in anterior regions, SP8 was co-expressed with GBX2 in a majority of cells (Figures 10E,F). However, in posterior ventral regions, SP8 and GBX2 immunoreactivity were expressed in the same region but mostly in different cells (Figures 10E,F). The medial geniculate nucleus (MGN) only expressed GBX2, but there was a small grouping of cells just medial and posterior to the MGN that strongly expressed SP8 exclusively (Figure 10F). This was in addition to strong expression in the thalamic reticular nucleus (Figure 10E). This is the first time, to our knowledge, that SP8 expression has been described in the developing thalamus.

GBX2 expression was weak at this stage of development and not always easy to detect, but in Figures 8C, 10E–G, it can be seen to be expressed anteriorly and medially, and in restricted locations posteriorally, as described at 10 PCW (Figure 5D). It also forms a complementary pattern of expression with CALB1 (Figures 8C,D) as described at 10 PCW, except that neither marker is expressed in the CM and the medial geniculate nucleus (MGN) where there is strong immunoreactivity for FOXP2. However, FOXP2 and GBX2 were both expressed in the putative medial dorsal nucleus (Figures 8B,C).

Cells lining the third ventricle remained positive for transcription factors expressed earlier in development. NR2F1 expression was homogenous throughout the thalamus (Figure 11A), however, NR2F2 expression was high close to the midbrain boundary but reduced in cells in more anterior locations (Figures 11B,C) whereas PAX6 showed higher expression dorsally (Figure 11D) and low expression ventrally and close to the boundary with p3 (arrow; Figure 11E) similar to the pattern observed at 10 PCW (Figure 5). Employing tritiated thymidine labeling, neurogenesis in the human thalamus is reported to be complete by 15 PCW (Rakic and Sidman, 1968). We found immunoreactivity for the cell division marker KI67 in the ventricular zone of the thalamus at both 12 and 15 PCW (Figures 11F,G) although this may be an indicator of gliogenesis as well as neurogenesis. Therefore, thalamic neurogenesis may be fairly protracted, and cell identity may be determined by a protomap until well into the second trimester.

Coronal sections of 15 PCW thalamus offer a different perspective, giving more access to anterior and ventral regions. In the most anterior sections, GBX2 expression was confined to the more medial regions of the thalamus, except for in the most dorsal regions where it extends laterally into the putative anterior nucleus (Figure 9A). FOXP2, on the other hand, was confirmed to be co-expressed with GBX2 medially, was not expressed in the anterior nucleus, but did show strong expression in the putative ventral anterior nucleus, more ventrally and laterally (Figure 9B). OTX2 was strongly expressed in the ventricular zone of both the epithalamus, choroid plexus and thalamic eminence, but exhibited no expression in the thalamus itself (Figure 9C). NR2F2 was also strongly expressed in the VZ of the thalamic eminence and epithalamus, and also in the reticular formation and regions of the hypothalamus, but similarly was not expressed in the thalamus at this anterior location (Figure 9D).

For sections from more towards the middle of the thalamus, GBX2 was most strongly expressed medially in the pF and MD (Figure 9E). Expression was completely absent from the putative habenula, which was also devoid of FOXP2 immunoreactivity (Figure 9F), but showed OTX2 immunoreactivity throughout its ventricular surface, unlike the thalamus (Figures 9G,G’). FOXP2 was expressed throughout the thalamus, but expression was strongest in the CM, and weakest in lateral regions. It was also expressed in the subthalamic nucleus (Figure 9F). Strong NR2F2 expression was observed the reticular formation and zona incerta between the thalamus and subthalamus (Figure 9H). However, it was also expressed throughout the thalamus except for the most dorso-lateral regions confirming observations made at 14 PCW. Interestingly, OTX2 immunoreactivity within the thalamus, although weaker than NR2F2, was also most highly expressed in post-mitotic cells of the MD (Figure 9G).

At 10 PCW no GAD67 immunoreactive neuronal cell bodies were observed in the thalamus (Figure 3D) in agreement with a recent ScRNAseq study (Kim et al., 2023). However, by 14 PCW three potential markers for GABAergic neurons or their precursors, NR2F2, SOX14, and GAD67, were all generally expressed in the more posterior parts of the thalamus. Expression of all four was greatest in the pretectum and midbrain, and appeared in a gradient across the thalamus, high at the thalamo-pretectal border and declining towards dorsal and anterior thalamus (Figures 7C,F,I, 10D, 12). These patterns of expression generally overlapped, suggesting a migration of cells from pretectum and midbrain into posterior regions of the thalamus.

GAD67 exhibited particularly striking expression in the CM-pF, as has been previously been shown, at 16 PCW (Alhesain et al., 2023; Figures 12B,D,F,G) where it showed clear co-expression with SOX14 in individual cells (Figure 12B’). However, GAD67 was not expressed just lateral to these nuclei where NR2F2 and SOX14 expression was observed (Figure 12B”). NR2F2 expression was seen in the FOXP2 negative reticular formation, along with lower levels of GAD67 and SP8 immunoreactivity. However, no SP8+ positive neurons were present in the thalamus, except in the dorsal LGN as described above.

We conclude that, between 10 and 14 PCW, the thalamus is partially invaded by SOX14+ and possibly NR2F2+ cells of pretectum/midbrain origin, however, certain nuclei, for instance the medial dorsal nucleus and medial pulvinar, remain devoid of these cells even by 16 PCW, suggesting migration has stopped. However, in some locations, maturation of these cells is quicker than in others. For instance the SOX14+ cells of the CM express GAD67 at this stage, whereas those of the LGN do not.

Figure 13A demonstrates that CNTNAP2 was expressed at 8 PCW, being present in post-mitotic cells in the thalamic mantle but generally not expressed in the thalamic ventricular zone. This confirms supplementary data provided by Kim et al. (2023) in their human thalamus scRNAseq study suggesting localization to post-mitotic excitatory neurons at 4-8 PCW. It is also similar to what was observed in the ganglionic eminences and the cerebral cortex, with the exception of the cortical hem and thalamic eminence where expression was high (Figure 13A). In the thalamus, expression was highest medially, close to the ventricle, but this may simply reflect higher cell density.

A series of sections at 14 PCW, running from anterior and dorsal to more posterior and ventral, Figures 13B–F show that CNTNAP2 expression, by the developmental stage when thalamic nuclei are emerging, no longer showed uniform expression across the thalamus in agreement with Ding et al. (2022). In the dorsalmost section (Figure 13B) habenular structures strongly expressed CNTNAP2. The pulvinar complex shows differential levels of staining in different areas. The medial pulvinar exhibited strong expression, as did more lateral regions of the lateral pulvinar. However, the inferior pulvinar exhibited relatively low expression. Moving more ventrally and anteriorly (Figure 13C) the habenula, medial pulvinar, LGN and the border with the pretectum all showed high expression, however, the MGN and an undefined region ventromedial to the MGN exhibited low expression. Figure 13C’ shows a high magnification image of a region at the edge of where multiple markers for GABAergic neurons are expressed at this level of the thalamus (Figures 12C,D). CNTNAP2 mRNA expression is visualized as red dots predominantly around faint blue nuclei (arrowheads) and is not seen in conjunction with green immunofluorescence for GAD67+ GABAergic neurons (asterisks) suggesting CNTNAP2 is predominantly expressed by glutamatergic neurons in the thalamus.

Moving further ventrally and posteriorly (Figures 13D,E) CNTNAP2 exhibited widespread strong expression throughout the thalamus with the exception of the parafascicular nucleus and close to the 3rd ventricle (Figure 13D) in agreement with Ding et al. (2022). The CM-pF complex strongly expressed FOXP2 (Figures 8B) which is proposed to downregulate expression of CNTNAP2 (Spiteri et al., 2007) so we double labeled an adjacent section for FOXP2 and CNTNAP2 (Figures 13D’,D”). In the MGN, we observed double labeled cells in the more medial parts of the nucleus, but FOXP2+ only cells in lateral parts (Figure 13D’). The LGN exhibited CNTNAP2 expression only, with no expression of CNTNAP2 and FOXP2 in its most ventral part. We confirmed co-expression of CNTNAP2 and FOXP2 in the CM, including within individual cells, but only FOXP2 was expressed in the pF, and only CNTNAP2 was expressed lateral and dorsal to CM-pF. In the most posterior and ventral section (Figure 13E) expression of CNTNAP2 was highest in ventral and posterior regions, and also close to the third ventricle.

Although FOXP2 has been proposed to downregulate CNTNAP2 expression, we found that only in some FOXP2 expressing thalamic nuclei was CNTNAP2 expression reduced, notably the pF and parts of the MGN. In the medial pulvinar, CM and parts of the MGN, co-expression was predominant.

There is an anterior ventral to posterior dorsal gradient of SHH concentration across the developing thalamus, as this morphogen is secreted from the anteriorly located ZLI and ventrally located basal plate (Kiecker and Lumsden, 2004; Nakagawa, 2019). SHH is known to sometimes suppress PAX6 expression in the forebrain (Chi et al., 2017) and so it was not surprising to observe higher PAX6 expression in the VZ posteriorly and dorsally, and absent from the ZLI. However, PAX6 was expressed strongly in the prethalamus of p3 which would be expected to receive a relatively high dose of SHH, indicating that co-expression and regulation of specific receptors and regulators of SHH signaling pathways in target cells is also important (Carballo et al., 2018). PAX6 expression was stronger and more persistent than that observed in mouse where weak expression dorsally was only observed up to E10.5 (Parish et al., 2016) the age of onset of neurogenesis in the mouse thalamus (Angevine, 1970). It is not clear why PAX6 expression extends throughout human thalamic neurogenesis which ranges from as early as 6PCW to at least 15 PCW (Rakic and Sidman, 1968; Alzu’bi et al., 2019; Workman et al., 2013).

We observed a different gradient of expression of NR2F2 in the thalamic VZ. Here expression was highest posteriorly and ventrally. Thus we have identified four zones to the human thalamic progenitor zone; anterior/ventral which is NR2F2+/PAX6-; anterior/dorsal which is NR2F2-/PAX6+; posterior/ventral which is NR2F2++/PAX6- and posterior/dorsal which is NR2F2+/PAX6++ (Figure 14A). Other transcription factors that might be predicted to display gradients of expression from mouse studies include OLIG2 expressed ventrally and DBX2 expressed dorsally (Vue et al., 2007). A more extensive future study of their expression patterns in human might be informative. We have observed very weak expression of OLIG2 immunoreactivity in or near the VZ, but unfortunately we only looked at 14 PCW. Cell lineage tracing studies in mouse have demonstrated that neuroprogenitors located in specific regions of the thalamic ventricular zone give birth to neurons destined to populate discrete thalamic nuclei (Shi et al., 2017). Thus anterior and dorsal thalamic VZ give rise to glutamatergic neurons that will populate anterior and medial higher order cognitive nuclei, whereas middle and ventral thalamic VZ populate higher order dorsolateral sensorimotor nuclei, and first order sensory nuclei located ventrally, posteriorly and laterally, respectively. It seems likely that a combinatorial protomap of transcription factor expression guides specification of thalamic neurons.

Previous studies have shown that a variety of transcription factors can display universal expression in thalamic post-mitotic neurons, only to show restricted expression to specific nuclei as development proceeds. For instance, in mouse Gbx2 is expressed by all thalamic neurons after leaving the cell cycle (Chen et al., 2009) but becomes restricted in expression to mostly medial and anterior nuclei, but also certain lateral and ventral nuclei posteriorly, in both mice and monkeys (Jones and Rubenstein, 2004). We have demonstrated that this also the case in human. GBX2 immunoreactivity became restricted to medial and posterior regions of the thalamus as early as 10 PCW, and by 14 PCW was confined to medial locations, the medial geniculate nucleus and other nearby posterior nuclei (Figures 4, 5, 10, 14B,C).

We observed some evidence of a gradient of increasing FOXP2 expression from anterior to posterior as early as 8 PCW, to also being widespread but stronger medially and posteriorly at 10 PCW, to being strongly expressed only in specific nuclei (e.g., CM, MGN, paraventricular thalamus) and absent from the anterior nuclei by 14 PCW (Figures 14B,C). An increasing expression from anterior to posterior was previously reported in mouse (Ebisu et al., 2017) but a medial to lateral gradient was not. Using transgenic mice expressing mutant Foxp2, Ebisu et al. (2017) demonstrated that FOXP2 directs development of posterior nuclei and interacts with GBX2 in counter gradients of expression. Posterior nuclei were smaller, while intermediate nuclei expanded along with thalamic territories expressing Gbx2 and Cadh6. Anterior nuclei were unaffected. Our evidence suggests that something similar is occurring in human, although it is interesting that this interaction to form a protomap would be occurring in post-mitotic cells as these transcription factors are not highly expressed in ventricular zone progenitors, even at 8–10 PCW (Figures 2–5, 14B). From their spatial transcriptomic study at mid-gestation, Kim et al. (2023) reported two classes of thalamic excitatory neurons EN1 and EN2, with FOXP2 being a marker for EN2 which were enriched in higher order nuclei and posterior first order nuclei, such as the pulvinar, MD, ventral lateral nucleus, LGN, and MGN. This has been partially confirmed by our observations, although we observed only moderate FOXP2 expression in the LGN, and weak expression in more lateral parts of the MD.

Conversely, ZIC4 expression exhibited a partially opposing gradient of expression at 8 PCW, being stronger anteriorly than posteriorly, and maintaining expression laterally where FOXP2 expression was weaker at 10 PCW (Figure 14B). By 14 PCW expression was confined to anterior and medial structures, but also the LGN (Figure 14C). According to supplementary data provide by Kim et al. (2023) ZIC4 may be a marker for EN1 cells (see above) which were exclusively found in anterior nuclei but intermingled with EN2 cells in the LGN mid-gestation. In mouse, it is known that Zic4 is preferentially expressed in LGN and Foxp2 in MGN during post-natal development, and that they play roles in guiding development of visual and auditory pathways, respectively (Horng et al., 2009). Interestingly, it has been shown in mouse that expression of Zic4 in progenitor cells with reduced Pax6 expression gives rise to LGN neurons that maintain Zic4 expression (Li et al., 2018). As we have shown that the human thalamus also has a ventral and posterior ZIC4+ PAX6- domain early in development, we can also surmise that this might be the location for production of LGN neurons, as has been demonstrated in mouse by cell lineage tracing studies (Shi et al., 2017). However, human seems to differ significantly from mouse in not expressing ZIC4 as strongly in the prethalamus (Li et al., 2018).

NR2F1, NR2F2, and SP8 also exhibited varying levels of expression across the thalamus. Although it was possible to attribute different combinations of transcription factor expression to different emerging nuclei and detect counter gradients of expression; these patterns revealed more about the location of the nuclei rather than predicting their future functional roles in cognition, sensory and motor integration or as first order relay nuclei (Figure 14C).

Detection of SP8 immunoreactivity in the thalamus was a surprising finding. This may be non-specific staining. The antibody has been characterized by protein array assay (only one interaction peak, the antigen, but not by Western blot).⁴ However, the staining patterns observed mostly concurred with specific and expected patterns of staining in human fetal telencephalon achieved with a different antibody (now unavailable; Alzu’bi et al., 2017a). We employed the new antibody as a marker of GABAergic neurons originating either from the prethalamus or caudal ganglionic eminence (CGE). We did observe strong SP8-like immunoreactivity in the prethalamus at 8PCW, but careful observation found SP8+ cells within the posterior and ventral margins of the thalamus in particular. These may have migrated into the thalamus from the prethalamus, however, we found the largest concentration posteriorly close to the ventral border with the SP8 negative pretectum, suggesting they could have been generated in the thalamus.

At 14 PCW SP8 immunoreactivity was a strong marker for the reticular formation which develops from the prethalamus. However, it was also a reliable marker for LGN but not the MGN. Potentially, the group of SP8+ cells observed posteriorly at 8PCW could have been the nascent LGN, although its location appears too ventral for this developmental stage. SP8 immunoreactivity showed strong co-localization with GBX2 at the cellular level in the anterior medial thalamus, but in posterior regions it was co-expressed in the same nuclei but not so much in the same cells. We have provided evidence for a potential novel role for SP8 in thalamic development, specifying the differentiation of distinct thalamic nuclei in concert with other transcription factors. This may be human specific as it has not been previously reported in other studies in other species.

It is known that in the primate thalamus there is a far higher proportion of GABAergic interneurons found within the nuclei, as opposed to thalamic reticular neurons that provide inhibition to thalamic neurons from a location outside of the thalamus (Braak and Weinel, 1985; Hunt et al., 1991; Arcelli et al., 1997). In both rodents and primates, the source of these interneurons has been demonstrated as being either the rostral midbrain for those expressing SOX14 and OTX2 (80% in rodents, 90% in marmoset) or from forebrain, using cell tracing studies (Jager et al., 2021) but not from the dorsal thalamic ventricular zone in either rodents or primates.

In rodents, the forebrain origin of GABAergic interneurons is the prethalamic (p3) progenitor zone (Jager et al., 2021) although the rostralmost thalamic ventricular zone may give rise to GABAergic projection neurons of the intergeniculate leaflet and ventral LGN (Vue et al., 2007; Delogu et al., 2012). This is also the case in marmosets but in human, it has long be argued that late born thalamic neurons can derive from the medial CGE (Letinic and Rakic, 2001) a source of SP8 and NR2F2 expressing GABAergic neurons in the telencephalon (Alzu’bi et al., 2017a; Alzu’bi et al., 2017b) and this has been recently confirmed by transcriptomic studies of cell lineage (Bakken et al., 2021; Kim et al., 2023). However, in situ hybridization studies have demonstrated a complete absence of expression of FOXG1, a marker of telencephalic neurons, in the thalamus at 15 PCW (Ding et al., 2022) suggesting these neurons arrive later in development.

We have demonstrated that SOX14, OTX2 and GAD67 expressing cells appear in the posterior two thirds of the thalamus by 14 PCW. It has previously been observed that there is a posterior to anterior appearance of interneurons across the thalamus in non-human primates and carnivores (Jones, 2002) and that interneurons arrive late in the thalamus; mid-gestation in monkeys and at birth in ferrets (Jones, 2002). In human a ScRNAseq/spatial transcriptomic study has identified a class of SOX14+/OTX2+ GABAergic neurons derived from the midbrain that appear in the thalamus after 10 PCW and inhabit the entire thalamus by mid gestation, but only selected nuclei at 14 PCW (Kim et al., 2023). Our observations confirm that appearance of interneurons in the thalamus proceeds posterior to anterior, is co-incident with the emergence of thalamic nuclei, but again stresses the point that human thalamic development is taking place relatively earlier than in other species.

We found no evidence for interneurons progressing much further anteriorly between 14 and 16 PCW, which leaves open the question as to whether anterior nuclei are populated by midbrain origin GABAergic neurons much later in development, or whether these nuclei are populated by cells of prethalamic or telencephalic origin. The anterior nuclei have the highest proportion of interneurons in humans and macaques (Hunt et al., 1991; Dixon and Harper, 2001; Popken et al., 2002) and this may be achieved by supplementation from the CGE. On the other hand, whereas in rodents midbrain origin interneurons preferentially cluster in the posterior and lateral nuclei and forebrain origin neurons in the anterior and medial (Jager et al., 2021) analysis of non-human primates and human data suggests a more homogenous distribution of the two interneuron classes across the thalamic nuclei (Jager et al., 2021; Bakken et al., 2021; Kim et al., 2023). Making some pause in the migration a more likely explanation.

Although SOX14+ presumed GABAergic interneuron precursors were spread throughout the posterior two thirds of the thalamus at 14 PCW, a large proportion did not co-express GAD67, suggesting that they were not functionally mature as GAD67 expression is activity dependent (Lau and Murthy, 2012). However, the CM-pF showed precocious expression of GAD67. This suggests these nuclei may mature more quickly than others. In the adult brain, the CM receives both multimodal sensory input and afferents from the ascending reticular activating system, while projecting primarily to sensorimotor cortical areas, however, it connects indirectly via other thalamic nuclei with a large proportion of the forebrain and is thus proposed to play a role in arousal and attention (Jang et al., 2014; llyas et al., 2019; Van der Werf et al., 2002; Kinomura et al., 1996). In development, thalamocortical innervation occurs first in the subplate of the sensorimotor cortex (Alzu’bi et al., 2019; Krsnik et al., 2017) and projections from the CM-pF may initiate early co-ordinated activity in thalamus and cortical subplate that drives development (Molnár et al., 2020; Molnár et al., 2023).

CNTNAP2 expression was uniform across the thalamus at 8 PCW, as was the case for a number of genes and proteins we examined at this stage. This suggests that CNTNAP2 plays a significant role in thalamic development, but that different regions of the thalamus are not yet following distinct developmental pathways. By 14 PCW, CNTNAP2 was found to be expressed by glutamatergic, not GABAergic, neurons and its expression varied between the different thalamic nuclei. Broadly speaking, this partially confirmed our hypothesis that this neurodevelopmental susceptibility gene would be expressed in higher order thalamic nuclei involved in cognitive processes, showing strong expression in the medial pulvinar, MD and laterodorsal nucleus, for instance. However, expression was also detected in the LGN, as well as ventrolateral regions, demonstrating it can be expressed by first order thalamic centers.

FOXP2 negatively regulates expression of CNTNAP2 (Spiteri et al., 2007; Vernes et al., 2007) providing a possible explanation as to why mutations in both genes disrupts language development (Lai et al., 2001; Whitehouse et al., 2011). It might be predicted that thalamic nuclei that strongly express FOXP2 would fail to express CNTNAP2, but we did not find this always to be the case. The centromedian nucleus (CM) in particular, showed strong expression of both, suggesting that FOXP2 regulation of CNTNAP2 expression is context specific. Intriguingly, three thalamic regions primarily shown to suffer reductions in volume in schizophrenia patients, the MD, pulvinar and CM (Kemether et al., 2003) all highly expressed both CNTNAP2 and FOXP2. It has been suggested that evolution of language acquisition, including expansion of the pulvinar, could be driven by the human specific isoform of FOXP2 (Ebisu et al., 2017). Perhaps the ability to negatively or positively regulate expression of CNTNAP2 is crucial to this process.

CNTNAP2 is robustly expressed in the ganglionic eminences and the neurons they produce, i.e. GABAergic basal ganglion neurons and cortical interneurons (Peñagarikano et al., 2011; Gordon et al., 2016; Ding et al., 2022) where it is proposed to play roles in neurogenesis, neurite growth and cell migration (Peñagarikano et al., 2011; Gao et al., 2018; Hali et al., 2020). However, in the thalamus expression was not found in GAD67+ cells but in presumptive GAD67- glutamatergic neurons. It was not found to be expressed in the ventricular zone, suggesting it does not play a role in neurogenesis in the thalamus, although it was highly expressed in the VZ of the epithalamus/habenula. As the habenula is considered to play a crucial role in regulating negatively motivated behavior and is implicated in several psychiatric illnesses including major depression (Hu et al., 2020) a better understanding of its development and expression of susceptibility genes would be worth pursuing. In the thalamus, CNTNAP2 may still play crucial roles in directing migration of thalamic neuroblasts to the correct thalamic location, and in directing axon outgrowth.