Molecular anatomy of the thalamic complex
The anatomic distinction between the thalamus and prethalamus was clearly indicated by our correlation analysis. However, we observed some deviant patterns within the thalamic structures derived from prosomere 2 (i.e., the thalamus and epithalamus or habenula), questioning some classic groupings of its nuclei. The latter were originally defined according to their topography relative to the internal medullary lamina. This landmark simply represents the way in which efferent thalamic fibers and various afferent fibers navigate across the thalamic mass and may not necessarily show a direct relationship with regard to patterning field effects.
The most striking result was obtained for the MHb, which displayed the most distinctive gene expression pattern within the thalamic complex. We observed that many transcripts are expressed exclusively in the MHb. The singular molecular profiles of the MHb and LHb are supported by a previous gene expression study (Quina et al.
2009) and reflect different functions of these two nuclei. Indeed, the MHb and LHb differentially participate in neuronal circuits (Aizawa et al.
2011). These two nuclei are currently classified as the epithalamus, a derivative of the dorsal-most progenitor domain in the alar territory of prosomere 2 that lies within the range of roof plate FGF8 signaling (Puelles et al.
2012). However, the MHb locus is the only site in prosomere 2 that contacts the insertion of the choroidal tela at the diencephalic roof, a position that is accompanied by a retarded and particularly protracted mode of neurogenesis (McAllister and Das
1977; Angevine
1970). This may explain its particular molecular characteristic. In contrast, LHb neurons have rather early birthdates and a short neurogenetic period compared with MHb and other neurons of the thalamic complex (McAllister and Das
1977; Angevine
1970). Recently, some embryologic data in the rat were reported, suggesting that the LHb (or a portion of it) may in fact originate outside the habenular territory in a caudoventral region of the thalamus, and its neurons secondarily migrate upward into the habenular domain (Puelles et al.
2015; Beretta et al.
2013). In that case, a significant part of the LHb might be regarded as not primarily epithalamic, and this observation may offer another explanation why the LHb was correlated more with prethalamic and thalamic nuclei than with MHb in our analysis. Either way, the results of our analysis support the view of intrinsic heterogeneity of the epithalamus (Geisler et al.
2003).
In our analysis, the RT, ZI, and PG were molecularly differentiated from the thalamus and from the MHb and LHb, consistent with the classic opinion that they all derive from prosomere 3 (Puelles et al.
2012). The RT, which develops at the rostral part of the prethalamus, differs from the PG and ZI with regard to the expression of a set of specific genes. The distinct molecular profile of the RT is consistent with connectivity differences between the RT and other prethalamic nuclei (Guillery et al.
1998; Mitrofanis and Mikuletic
1999; Bourassa and Deschênes
1995), possibly underpinned by their differential relative dorsoventral and anteroposterior origins within alar prosomere 3.
Partly new organizational properties of the thalamus proper emerged from our analysis. We found strong evidence of the existence of three nuclear groups, which we called medial, lateral, and posteroventral, and some small individual regions with unique gene expression patterns (Fig.
6). This especially refers to the posterior and anterior PV, PF, and AD/AM/AV, which are conventionally classified as a midline nucleus, caudal intralaminar nucleus, or anterior nuclei, respectively (Puelles et al.
2012; Van der Werf et al.
2002). Our divisions agree with groupings based on thalamic cell birthdates (i.e., neurogenetic patterns), which reportedly follow lateromedial, caudorostral, and ventrodorsal gradients (Angevine
1970; McAllister and Das
1977). The lateral, medial, and posteroventral groups appear to be homologous to rostrodorsal, caudoventral, and ventral compartments, respectively, which was previously proposed for the embryonic thalamus by González et al. (
2002).
The proposed lateral group (green in Fig.
6) includes (1) nuclei that are classified as first-order (sensory) nuclei (VPL/VPM, MGN, DLG, and VA/VL), which are populated by C-type neurons and receive inputs from cortical layer 6, sensory lemniscus pathways, or deep cerebellar nuclei and project to layers 3–4 of single or adjacent cortical areas (Clascá et al.
2012); (2) associative nuclei (LP, Po, MD, MGN, VPPC, Sub, and VM) that have reciprocal connections with several cortical areas, receive afferents from the ZI, and are populated by M-type neurons that never arborize in cortical layer 4 (Deschênes et al.
1998; Power et al.
1999); (3) and the LD although this nucleus is sometimes classified with anterior nuclei because of its reciprocal connections with part of the anterior cingulate cortex (Shibata and Naito
2005). The lateral group also shares the expression of some genes with medial nuclei, LHb and PF or exclusively with the PF.
The proposed medial group (yellow in Fig.
6) includes several median and paramedian midline nuclei and rostral intralaminar nuclei, which all have reciprocal connections with the frontal cortex, send collaterals to the striatum, and receive afferents from the midbrain, brainstem, and pallidal areas (Krout et al.
2001,
2002; Groenewegen et al.
1999; Van der Werf et al.
2002).
Finally, the proposed caudally located posteroventral group (light blue in Fig.
6) includes small nuclei that are adjacent to the alar–basal boundary (SPF, PP, PIL, and SG). These nuclei receive auditory inputs and project to the amygdala and temporal part of the dorsal pallidum (Winer et al.
2002; Ledoux et al.
1987). We also found that many genes that are expressed in the posteroventral group, particularly in the PP, are shared with the IGL. This might support the thalamic origin of IGL nuclei.
As we noted above, some thalamic nuclei exhibit unique gene expression patterns, and this corresponds well to specific functions that are conferred by these nuclei. The PV serves as a nodal point between brain regions that are involved in emotional and motivational circuitry and receives strong aminergic inputs from the brainstem and peptidergic inputs from hypothalamic neurons, a property not shared by other thalamic nuclei (Li and Kirouac
2008,
2012). The PF is characterized by receiving afferents from midbrain and brainstem areas that are involved in processing sensory and motor information and has a large efferent connection with the striatum, with relatively few cortical collaterals (Van der Werf et al.
2002; Krout et al.
2002; Puelles et al.
2012). The AD is a key thalamic relay of the head-direction system, possesses a unique set of connections (via the mamillothalamic and fornix tracts), and displays characteristic electrophysiological properties (Van Groen and Wyss
1995; Taube
1995). The AM and AV also have specific functions and connectivity related to the circuit of Papez (Aggleton et al.
2010; Shibata and Naito
2005). We hypothesize that these unique thalamic nuclei either derive from distinct progenitor domains, or the mechanisms of their terminal specifications are different, possibly involving signals that spread from either the prethalamus (PV and AD/AV/AM) or pretectum (PF). These unique nuclei (and their presumptive primordia) are localized at extreme ends of the thalamus where they may be selectively accessible to morphogenetic factors. Another possibility is that these nuclei acquire their specific properties during late embryonic or early postnatal development because of singular functional conditions.
Differential specification of thalamic complex nuclei by transcription factors
Our analysis of transcription factor expression in the thalamic complex supports the idea that its molecular identity and internal regionalization can be defined by variously overlapping combinations of active transcription factors, rather than being characterized by a single factor for each thalamic nucleus. Similar conclusions were drawn using high-throughput analysis of transcription factors in the developing subpallium, hypothalamus, and spinal cord (Del Barrio et al.
2013; Flames et al.
2007; Shimogori et al.
2010).
In the prethalamus, many transcription factors that are characteristic of early embryonic stages (e.g.,
Dlx1/2/5/6,
Gli3,
Foxd1, and
Lhx1 on embryonic day 11.5–13.5) are no longer expressed in the adult, suggesting that they play a role only in the early specification of prethalamic neurons (Jones and Rubenstein
2004; Nakagawa and O’Leary
2001; Puelles and Martinez
2013). However, several transcription factors, such as
Isl1,
Six3,
Essrg, and
Meis2, continue their expression in the adult mouse and human prethalamus, whereas
Pax6 is present only in mouse (Ehrman et al.
2013; Lavado et al.
2008; Toresson et al.
2000; Pratt et al.
2000). In the epithalamus,
Pou4f1 is specifically and highly expressed, and its involvement in the activation of habenula-specific genes has been previously reported (Quina et al.
2009). Expression of this gene is also conserved in the human epithalamus. Thalamus-specific transcription factors show diverse and overlapping expression patterns that align with the proposed nuclear groups. For example,
Foxp1,
Foxp2,
Gbx2, and
Tox are all expressed in the medial group. Whereas
Foxp2 and
Gbx2 are also expressed in some of the lateral nuclei,
Tox and
Foxp1 are not expressed in any of them. All anterior nuclei are characterized by the expression of
Prox1, but each of these nuclei also expresses some specific transcription factors that are not expressed by others. For example, the AD selectively expresses
Id4 and
Arnt2, whereas AM expresses
Rreb1 and
Foxp1. Neurons in the PF strongly express
Etv1 but do not express
Zic1 or
Id4, which are expressed in many thalamic nuclei. The PV is characterized by the high expression of
Hopx,
Tox, and
Gbx2. Several transcription factor genes are also broadly expressed in the adult mouse and human thalamus, particularly
Tcf7l2 and
Lef1, the expression of which in the mouse thalamic area is maintained throughout development and in adulthood (Nagalski et al.
2013).
Tcf7l2 is expressed in all thalamic and epithalamic nuclei, and
Lef1 is expressed in most nuclei, with the exception of the PV and posteroventral group. We suppose that the transcription factors that are specific for different groups of nuclei or unique nuclei can serve as molecular markers that stably and selectively demarcate these areas and might participate in the regionalization of the thalamic complex and its parts.
TCF7L2 as a terminal selector in prosomere 2
The aforementioned TCF7L2 and LEF1 are effectors of Wnt/β-catenin signaling. Our results showed that the promoters of some thalamus- and epithalamus-specific transcription factors, namely
Etv1,
Foxp2,
Gbx2,
Mef2c,
Nr4a2,
Pou4f1,
Rora,
Zic1, and
Znf804a, are responsive to the thalamic TCF7L2-S isoform and/or LEF1, at least using a cell line. Several of these transcription factors are known to be involved in brain development. POU4F1 (alias BRN3A) together with NR4A2 (alias NURR1) regulates the coordinated expression of habenula-enriched genes, and habenular connections are lost in
Pou4f1
−/− embryos (Quina et al.
2009). GBX2 is a marker of postmitotic thalamic neurons, and the thalamus is disrupted in
Gbx2
−/− embryos (Li et al.
2012; Chen et al.
2009). PROX1 has been shown to regulate adult neurogenesis in the hippocampus as a target of the Wnt/β-catenin pathway (Karalay et al.
2011). We recently identified other genes (i.e.,
Cacna1g,
Kcna6,
Calb2,
Gabra3,
Cacna2d2, and
Kcnh8) with conserved LEF/TCF motifs in their promoter regions that were highly expressed in the thalamus and regulated by β-catenin (Wisniewska et al.
2010,
2012). They are likely directly regulated by TCF7L2 as a β-catenin effector. Data in the Allen Mouse Brain Atlas indicate that the expression of these genes is restricted to different sets of thalamic nuclei. Given the structural and functional complexity of the thalamus, it is rather implausible that TCF7L2 can regulate the same genetic programs by itself in all of the nuclei. Gene regulation by TCF7L2 is known to be context-dependent (Frietze et al.
2012; Boj et al.
2012). We hypothesize that the thalamus-specific isoform TCF7L2-S and/or LEF1 can eventually regulate various thalamic and epithalamic transcription factors, with expression patterns restricted to some areas. TCF7L2 and LEF1 may represent terminal selector genes of thalamic neurons, analogous to AST-1 in dopaminergic neurons (Flames and Hobert
2009) and UNC-3 in cholinergic motor neurons (Kratsios et al.
2012) of
Cenorhabditis elegans or Pet-1 in mouse serotonergic neurons (Liu et al.
2010; Alonso et al.
2013). However, our hypothesis has yet to be confirmed with further studies in knockout mouse models.
Possible relevance to schizophrenia and other neuropsychiatric disorders
Several lines of evidence indicate thalamic dysfunction and thalamocortical disconnectivity in psychiatric conditions such as schizophrenia (Behrendt
2003; Woodward et al.
2012; Alelú-Paz and Giménez-Amaya
2008; Ellison-Wright and Bullmore
2010; Pratt and Morris
2015; Popken et al.
2000; Byne et al.
2009; Kumari et al.
2010), bipolar disorder (Radenbach et al.
2010), major depression (Young et al.
2004; Greicius et al.
2007; Li et al.
2013), and autism spectrum disorder (Nair et al.
2013). The aforementioned disorders are considered to be of neurodevelopmental origin; therefore, understanding the development of the thalamic complex may help to reveal their etiology. Two of the transcription factors that were identified here to be specific for thalamic subregions have been associated with schizophrenia, bipolar disorder and autism in genome-wide association studies: TCF7L2 (Alkelai et al.
2012; Iossifov et al.
2014; Winham et al.
2014) and ZNF804A (Consortium SWGotPG
2014; Williams et al.
2011a,
b; O’Donovan et al.
2008). In particular, the rs1344706 single nucleotide polymorphism (SNP) in intron two of
ZNF804A was the first variant to reach an unequivocal genome-wide significance for schizophrenia (O’Donovan et al.
2008) with later meta-analyses confirming the association and extending it to a broader psychosis phenotype (Zhu et al.
2014; Sun et al.
2015; Williams et al.
2011a,
b). Interestingly, the subregions in the mouse thalamic complex where
Znf804a is expressed overlap with the regions that are affected in schizophrenic patients, i.e.: the prethalamic reticular nucleus (Pratt and Morris
2015), MD (Alelú-Paz and Giménez-Amaya
2008; Byne et al.
2009), pulvinar nucleus (Byne et al.
2009; Kumari et al.
2010) considered to be an equivalent of the LP in rodents (Baldwin et al.
2011), and MGN (Kumari et al.
2010). The function of this transcription factor and the potential mechanism by which it might increase risk for schizophrenia is still not known. Morphometric analysis of schizophrenic brains revealed a possible effect of
ZNF804A genetic variation on the frontal cortical regions and thalamus (Nenadic et al.
2015), suggesting an involvement of ZNF804A in the development of these areas of the brain. Even less is known about the relationship between TCF7L2 and psychopathologies, except that
Tcf7l2 haploinsufficient mice show anxiety-like behavior and altered fear learning (Savic et al.
2011). It is important to note here that TCF7L2 (Transcription Factor 7-Like 2, HMG-box) has been initially abbreviated as TCF4 (T cell factor 4), and it should not be confused with a basic helix-loop-helix transcription factor TCF4 (Transcription Factor 4, alias E2-2, ITF2), which is also associated with schizophrenia. Further research is needed to elucidate the role of ZNF804A and TCF7L2 in thalamus development and in the etiology of psychiatric disorders.