TALE transcription factors during early development of the vertebrate brain and eye

Authors

  • Dorothea Schulte,

    Corresponding author
    1. Institute of Neurology (Edinger Institute), University Hospital Frankfurt, J.W. Goethe University, Frankfurt, Germany
    • Correspondence to: Dorothea Schulte, Institute of Neurology (Edinger Institute) University Hospital Frankfurt, J.W. Goethe University, 60528 Frankfurt, Germany. E-mail: dorothea.schulte@kgu.de or Dale Frank, Department of Biochemistry, The Rappaport Family Institute for Research in the Medical Sciences, Faculty of Medicine, Technion-Israel Institute of Technology, Haifa 31096, Israel. E-mail: dale@tx.technion.ac.il

    Search for more papers by this author
  • Dale Frank

    Corresponding author
    1. Department of Biochemistry, The Rappaport Family Institute for Research in the Medical Sciences, Faculty of Medicine, Technion-Israel Institute of Technology, Haifa, Israel
    • Correspondence to: Dorothea Schulte, Institute of Neurology (Edinger Institute) University Hospital Frankfurt, J.W. Goethe University, 60528 Frankfurt, Germany. E-mail: dorothea.schulte@kgu.de or Dale Frank, Department of Biochemistry, The Rappaport Family Institute for Research in the Medical Sciences, Faculty of Medicine, Technion-Israel Institute of Technology, Haifa 31096, Israel. E-mail: dale@tx.technion.ac.il

    Search for more papers by this author

Abstract

Our brain's cognitive performance arises from the coordinated activities of billions of nerve cells. Despite a high degree of morphological and functional differences, all neurons of the vertebrate central nervous system (CNS) arise from a common field of multipotent progenitors. Cell fate specification and differentiation are directed by multistep processes that include inductive/external cues, such as the extracellular matrix or growth factors, and cell-intrinsic determinants, such as transcription factors and epigenetic modulators of proteins and DNA. Here we review recent findings implicating TALE-homeodomain proteins in these processes. Although originally identified as HOX-cofactors, TALE proteins also contribute to many physiological processes that do not require HOX-activity. Particular focus is, therefore, given to HOX-dependent and -independent functions of TALE proteins during early vertebrate brain development. Additionally, we provide an overview about known upstream and downstream factors of TALE proteins in the developing vertebrate brain and discuss general concepts of how TALE proteins function to modulate neuronal cell fate specification. Developmental Dynamics 243:99–116, 2014. © 2013 Wiley Periodicals, Inc.

INTRODUCTION

CNS development begins at the end of gastrulation, when signals from the notochord induce the overlaying ectoderm to form the neural plate. Through a series of morphological changes in cell shapes, the lateral edges of the neural plate rise and fuse at the midline to generate the neural tube. The first indication of regionalization and specification becomes apparent when the anterior neural tube subdivides into a series of brain vesicles. These vesicles constitute the morphological basis of distinct functional units of the brain. The telencephalic vesicle at the neural tube's most anterior end gives rise to the cerebral cortex, basal ganglia, and hippocampus. The adjacent diencephalic vesicle is the embryonic anlage of the thalamus, hypothalamus, and pineal gland. Lateral protrusions of the diencephalon, the optic vesicles, give rise to much of the adult eye structure. More posteriorly, the mesencephalic vesicle generates important centers of sensory and motor control, whereas the hindbrain gives rise to the cerebellum, pons, and medulla. TALE homeodomain proteins are expressed in each of these brain vesicles in the embryo and adult, suggesting that they participate in a multitude of different developmental processes, including growth of the primary neuroepithelium, neuronal cell fate decisions, cell migration, and formation of neuronal circuits.

TALE (three amino acid loop extension) proteins are atypical homeodomain containing transcription factors, characterized by the presence of three additional amino acids between the first and second helix of the homeodomain (Bürglin, 1997). These three amino acids are almost always a proline-tyrosine-proline motif inserted after position 23 in the homeodomain. In the animal kingdom, TALE transcription factors comprise three subclasses: (i) the MEINOX (i.e., Meis/Prep) gene family, which includes vertebrate Meis1–3, Prep1–2 (also known as Pknox1–2), Tgif1–2, as well as Drosophila homothorax (hth) and C. elegans unc-62; (ii) the PBC family, including vertebrate Pbx1–4, fly extradenticle (exd), and worm ceh-20; and (iii) the more distantly related Iroquois (Iro) family. The purpose of this review is to summarize current knowledge about the various roles of MEIS/PREP and PBC-type TALE proteins in early neural development and neuronal cell fate decisions, with particular attention given to hindbrain, midbrain, forebrain, and eye development. For a review on IRO protein function in the embryonic vertebrate and invertebrate nervous system, the interested reader is referred to Cavodeassi et al. (2001) or Gómez-Skarmeta and Modolell (2002).

HOX-DEPENDENT FUNCTIONS OF TALE PROTEINS: HINDBRAIN DEVELOPMENT

The hindbrain controls crucial physiological processes such as motor activity, respiration, sleep, and blood circulation. The embryonic hindbrain subdivides into seven metameric segmented units known as rhombomeres (r1–r7). The most anterior r1 borders the midbrain, while the most posterior r7 borders the anterior spinal cord. Each rhombomere has unique gene expression that promotes the region-specific differentiation of neurons. These compartmentalized units are highly conserved during vertebrate evolution and act as the template for the adult hindbrain structure and function. TALE and HOX family transcription factors play a key role in defining rhombomeric cell fate identities along the anterior–posterior (AP) axis. Signaling pathways, such as canonical Wnt, fibroblast growth factor (FGF) and retinoic acid interact with Meis/Pbx/Hox proteins in multiple ways to induce hindbrain development and subsequently control rhombomeric identity. Below, we examine the early dynamic interactions between Meis/Hox proteins and signaling pathways in these processes.

Meis, Pbx, and Paralogous Group 1–4 Hox Proteins Regulate Hindbrain Development

Meis1 and Meis3 TALE family homeobox proteins are required for proper hindbrain formation in Xenopus and zebrafish embryos (Dibner et al., 2001; Waskiewicz et al., 2001; Choe et al., 2002; Maeda et al., 2002). The meis3 gene is expressed as early as late gastrula stages in the presumptive hindbrain region, later localizing to the r2–r4 region (Salzberg et al., 1999; Dibner et al., 2001; Vlachakis et al., 2001; Waskiewicz et al., 2001). meis1 and meis2 gene expression is less restricted, expanding into more anterior neural plate regions, whereas embryonic prep gene expression is ubiquitous (Cecconi et al., 1997; Oulad-Abdelghani et al., 1997; Ferretti et al., 1999; Maeda et al., 2001, 2002; Sánchez-Guardado et al., 2011; Waskiewicz et al., 2001; Zerucha and Prince, 2001). Loss of Meis3 protein function by either dominant negative or anti-morph proteins, morpholino (MO) -based knockdown of expression, or genetic mutation, results in loss of the entire hindbrain region, accompanied by a loss in expression of a plethora of hindbrain markers, including the Hox paralogous group (PG) 1–4 genes. This is coincident with a posterior expansion of anterior forebrain markers and enlarged forebrain structures. Meis3 also suffices for hindbrain induction, as its over-expression induces ectopic hindbrain formation while repressing forebrain formation in both embryos and explants (Salzberg et al., 1999; Dibner et al., 2001, 2004; Vlachakis et al., 2001; Waskiewicz et al., 2001; Elkouby et al., 2010; Gutkovich et al., 2010). In zebrafish embryos, Meis3 synergizes with HoxB1b (HOXA1 in other vertebrates) protein to induce expression of hindbrain markers such as hoxb1a and krox20 (Vlachakis et al., 2001). Similarly, HoxD1 and Meis3 co-expression enhances hoxb3 and krox20 marker expression (Dibner et al., 2004). Moreover, hoxd1 is a direct target of Meis3 in Xenopus, acting downstream of Meis3 to induce hindbrain cell fates (Dibner et al., 2004; Gutkovich et al., 2010).

The PBC family protein Pbx4 (lazarus) directly interacts with Meis1 and Meis3 proteins in zebrafish, and perturbation of Pbx2 or Pbx4 activities eliminated r2–r6 cell fates (Pöpperl et al., 1995; Vlachakis et al., 2001; Waskiewicz et al., 2001, 2002). Likewise, Meis1 and Pbx1 proteins interact to regulate neural cell fate specification in Xenopus, and Pbx1 knockdown also disrupted hindbrain formation (Maeda et al., 2002). In Xenopus, Pbx1 and HoxD1 proteins strongly activated a heterologous mouse hoxb1 enhancer, but Meis3 had no additive effect with either protein when tested separately or together (Chan et al., 1996; Dibner et al., 2004). A Pbx protein partner that enhances Meis3 activity in Xenopus has not been identified. Of interest, in zebrafish, ectopic Meis1 expression rescued hindbrain formation in the absence of zygotic Pbx4 protein. This suggests a potential Pbx-independent mechanism of activity, although maternal Pbx protein involvement was not fully ruled out (Waskiewicz et al., 2001).

PG1 Hox proteins are key factors controlling early hindbrain specification. The PG1 Hox proteins are homologues of the Drosophila labial gene, and include HOXA1, HOXB1 and HOXD1 proteins. In all vertebrates, expression of the PG1 Hox genes precedes all other Hox genes in the presumptive hindbrain region at early gastrula stages and persists through neurula stages (Wilkinson et al., 1989; Frohman et al., 1990; Sundin et al., 1990; Murphy and Hill, 1991; Frohman and Martin, 1992; Godsave et al., 1994; Kolm et al., 1997; Rossel and Capecchi, 1999; McClintock et al., 2002; McNulty et al., 2005). PG1 genes are essential for correct hindbrain induction and segmentation. Combined knockdown of HoxB1a and HoxB1b proteins in zebrafish, or deletion of the Hoxa1 and Hoxb1 genes in mouse embryos led to significant reduction or complete loss of rhombomeres r4 and 5 (Lufkin et al., 1991; Chisaka et al., 1992; Studer et al., 1996, 1998; Gavalas et al., 1998; Rossel and Capecchi, 1999; McClintock et al., 2001, 2002). In Xenopus embryos, triple knockdown of all PG1 genes, hoxa1, hoxb1, and hoxd1 caused a complete loss of r2–6, and the entire hindbrain resembled the Hox nonexpressing r1 region (McNulty et al., 2005). This phenotype is reminiscent of Pbx2 loss-of-function in zebrafish embryos, arguing for cooperative function of Pbx2 and PG1 Hox proteins in the specification of r2–6 regions (Waskiewicz et al., 2002). Supporting this assumption, HOXA1 hexapeptide mutant proteins that fail to interact with PBX proteins cause severe hindbrain phenotypes in mice (Remacle et al., 2004). Moreover, the combined PG1 loss-of-function phenotype is synergistically stronger than that of each of the individual inhibitions, and in Xenopus, PG1 knockdown could be rescued by over-expression of HoxD1 protein alone. This suggests at least partial functional redundancy between PG1 members (McNulty et al., 2005).

Early PG1 protein expression is a prerequisite for the proper sequential expression of later, more posteriorly expressed PG2–4 Hox genes (Dibner et al., 2004; McNulty et al., 2005). In PG2 mutant mice, development of the r3/r4 region is disrupted, with poor border formation between r2/r3. Cells in r2 express only Hoxa2, the most anterior of all Hox genes, and Hoxa2 loss-of-function mutations cause a partial r2 to r1 fate change (Prince and Lumsden, 1994; Gavalas et al., 1997; Davenne et al., 1999; Barrow et al., 2000; Ren et al., 2002; Oury et al., 2006). R3 expresses Hoxa2 and Hoxb2. Single Hoxb2 mutant mice had no hindbrain segmentation defects, but in Hoxa2/Hoxb2 double mutant mice, the r2/3 and r3/r4 borders were lost, indicative of synergistic function of both PG2 genes (Davenne et al., 1999). Hox PG1 and PG2 genes are both expressed in r4, but with temporal differences. In multiple PG3 mutants, r5/6 identity was disrupted and r4-specific Hoxb1 expression was ectopically activated in r5/6 (Gaufo et al., 2003). In PG4 mutants, in contrast, hindbrain development was normal (Horan et al., 1995).

Signaling Pathways Induce and Shape the Hindbrain Primordium: Wnt/β-catenin, FGF, and Retinoic Acid

The canonical Wnt-signaling pathway regulates physiological processes in the embryo and adult. Abnormal Wnt-signaling is involved in many human diseases, such as cancer, aging, and degenerative disorders (reviewed in Moon et al., 2004; Clevers, 2006). The Wnt-ligand triggers a cascade of events leading to activated transcription of repressed target-genes. In the absence of Wnt-ligand, β-catenin is targeted for ubiquitin mediated degradation by the activities of Axin, APC, and glycogen synthase kinase-3β (GSK3β). Secreted Wnt proteins bind the Frizzled (Fz) and the low density lipoprotein receptor-related protein 6 (LRP6) receptor complex. Wnt-ligand binding triggers LRP6 phosphorylation, which ultimately prevents β-catenin degradation. Stabilized β-catenin protein then undergoes nuclear translocation, where it binds TCF/LEF proteins and directly activates gene transcription (MacDonald et al., 2009).

Wnt/β-catenin signaling is obligatory for the expression of most hindbrain and spinal cord Hox genes during neural plate specification in vertebrates, including the PG1 Hox proteins. Many Wnt ligands are expressed in the developing nervous system of vertebrate embryos (reviewed in Elkouby and Frank, 2010). Their role in posterior neural development was first shown in Wnt1 and Wnt3a deficient mouse embryos, where midbrain, hindbrain, and spinal cord structures failed to develop properly (McMahon et al., 1992; Augustine et al., 1995, 1993). Wnt3 null mouse embryos showed a similar phenotype of posterior truncation, with a dramatic expansion of forebrain markers and a loss Hoxb1 expression in r4 (Liu et al., 1999). A similar observation was seen in early stage chick embryos implanted with beads soaked with a soluble-inhibitory form of the Frizzled receptor (mFrz8-CRD), in which forebrain markers were expanded and hindbrain marker expression was inhibited (Nordström et al., 2002). In addition, in mFrz8-CRD treated posterior-neural explants, expression of hindbrain and spinal cord markers was also eliminated (Nordström et al., 2002, 2006). A requirement for Wnt-signaling for posterior neural development was confirmed in Xenopus embryos, as expression of a dominant-negative Wnt protein (inhibiting Wnt1, Wnt3a, and Wnt8 activity) induced an anteriorized phenotype. Moreover, like Wnt3/ mouse embryos, these embryos showed neural tube closure defects (McGrew et al., 1997). More recently, the requirement for mesodermally derived Wnt3a in hindbrain induction was revealed by knockdown experiments in Xenopus (Elkouby et al., 2010). Wnt3a morphant embryos exhibited neural convergent and extension defects, with caudal expansion of forebrain markers and depletion of hindbrain markers at neurula stages. Finally, targeting zebrafish embryos with specific Wnt3a- and Wnt8-MOs further supports a role in hindbrain and spinal cord development (Erter et al., 2001; Lekven et al., 2001).

Wnt3a signaling and Meis/Hox gene activity are tightly linked during hindbrain induction. In Wnt3a morphant gastrula stage Xenopus embryos, onset of expression of meis3 and hoxd1 is lost in the presumptive neural plate (Elkouby et al., 2010). In fact, meis3 and hoxd1 are direct β-catenin target genes (Elkouby et al., 2010; Janssens et al., 2010; Rieden et al., 2010). The Xenopus meis3 promoter/enhancer harbors two conserved β-catenin binding sites that are required to drive reporter expression in the hindbrain. These two sites are bound by β-catenin in vivo (Elkouby et al., 2010). Moreover, ectopic Meis3 expression can rescue the loss-of-hindbrain phenotype in embryos deficient in zygotic Wnt signaling (experimentally induced by either injection of Wnt3a-MOs or ectopic Dkk1 expression). Therefore, the hindbrain inducing activity of Wnt3a is largely mediated by Meis3. In addition, because correct expression of PG2–4 Hox genes depends on the proper induction of PG1 Hox genes, later hindbrain Hox gene expression is indirectly regulated by the Wnt pathway (Dibner et al., 2004; McNulty et al., 2005; Rieden et al., 2010). Thus in Xenopus, the direct induction of meis3 and PG1 Hox gene expression by the Wnt/β-catenin pathway is one of the earliest steps of the hindbrain developmental program, positioning Wnt signaling at the top of this hierarchy.

Of interest, Meis3 induces later neural plate specific wnt3a gene expression (Elkouby et al., 2012). Mutual positive regulation between a cell-autonomous transcriptional regulator, like Meis3, and a secreted protein, like Wnt3a, could theoretically lead to exuberant expression of both proteins and thus to progressive caudalization of the entire neural plate. This is prevented by a double negative feed-back loop: Wnt3a activates expression of the zinc-finger protein Teashirt 1 (Tsh1), which in turn negatively regulates Meis3 activity and thereby restricts its activity to the hindbrain region (Elkouby et al., 2012).

Fibroblast growth factors (FGFs) are signaling ligands, which regulate mesoderm and neuroectoderm formation and patterning during early vertebrate development (reviewed in Dorey and Amaya, 2010). FGFs signal through specific receptor tyrosine kinases located in the cell membrane. The FGF signaling pathway is crucial for the correct differentiation of many cells and tissues in the embryo and disrupted FGF signaling is associated with a wide spectrum of pathologies, such as skeletal defects, neurodegenerative diseases, metabolic disorders, and cancer.

FGF3/8 expression in the hindbrain is conserved among all vertebrates (Mahmood et al., 1995; Lombardo et al., 1998; Maves et al., 2002; Walshe et al., 2002). In zebrafish embryos, the fgf3 and fgf8 genes are expressed in the presumptive hindbrain primoridia region at 80–90% epiboly, before the onset of hindbrain segmentation marker expression (Maves et al., 2002; Walshe et al., 2002). At segmentation initiation, fgf3/8 transcripts are expressed broadly in the central r4 region, with r3/r5 overlap. When morphological segments have formed, fgf3 continues to be expressed in r4, but fgf8 expression is extinguished, with new shifted expression to the more anterior isthmus region. Early FGF3/8 signaling in r4 forms a primary hindbrain inducing center (Maves et al., 2002; Walshe et al., 2002). This was first observed in chick and zebrafish embryos, where transplanted r4 cells induced ectopic expression of r5/6 markers surrounding the graft, a phenotype strongly resembling that of ectopic FGF3/8 expression (Graham and Lumsden, 1996; Maves et al., 2002). FGF3 and FGF8 activities are functionally redundant, because a strong loss-of-hindbrain phenotype is only seen upon knockdown of both proteins (Maves et al., 2002; Walshe et al., 2002). Hence, early fgf3 and fgf8 gene expression is crucial for initial hindbrain induction.

In Xenopus, zebrafish, and chick embryos, FGF signaling acts downstream of Wnt signaling (Domingos et al., 2001; Rhinn et al., 2005; Nordström et al., 2006; Elkouby et al., 2010). This is, at least in part, mediated by Meis3, as Meis3 protein directly activates fgf3 and fgf8 gene expression, and fgf3 expression in r4 is lost in Meis3 morphant Xenopus embryos (Gutkovich et al., 2010). Consistent with the frequent heterodimerization of Meis and Pbx proteins, Pbx2 and Pbx4 protein functions are also required for the onset of fgf3 and fgf8 expression in the zebrafish hindbrain (Waskiewicz et al., 2002). Notably, Meis3 protein cannot induce hindbrain cell fates in the absence of downstream FGF/MAPK signaling, arguing that Meis3-mediated activation of the FGF/MAPK pathway is a central aspect of its caudalizing activity (Ribisi et al., 2000; Aamar and Frank, 2004). Protein phosphatase 1 regulatory subunit (ppp1r14al) protein also participates in the specification of the r4 hindbrain inducing center. The ppp1r14al gene is directly activated by HoxB1b/Meis3 proteins in r4 of zebrafish embryos and Meis3/Pbx4/HoxB1b proteins are bound to its enhancer in vivo (Choe et al., 2011). Ppp1r14al, in turn, regulates activation of fgf3 expression in r4, as hindbrain formation is highly compromised by ppp1r14al protein knockdown.

Expression of PG2–3 Hox genes is also FGF-dependent. FGF3 indirectly activates Hoxa2, Hoxb2, and Hoxb3 expression by means of induction of the Pea/Krox20 transcription factors (Manzanares et al., 1999, 2002; Maconochie et al., 2001; Weisinger et al., 2010). In mice and zebrafish, the Krox20 promoter has a functional r3-specific Hox/Meis/Pbx binding site, although it does not appear to be a direct Meis3-target in Xenopus (Wassef et al., 2008; Stedman et al., 2009; Elkouby et al., 2010). In zebrafish embryos, krox20 transcriptional activation also requires the variant hepatocyte nuclear factor-1 (vHnf1) protein, which synergizes with FGF activity to induce mafb/kreisler/valentino and krox20 gene expression. vHnf1 protein in turn represses hoxb1 gene expression, limiting its expression to r4, thus enabling formation of more posterior rhombomeric fates (Wiellette and Sive, 2003; Hernandez et al., 2004). Subsequently, vHnf1 together with Mafb activate hoxa3 and hoxb3 expression in r5/6 to specify r5/6 identity (Manzanares et al., 1999; Hernandez et al., 2004). Thus, in the hindbrain, the r4 FGF-inducing center stands at a pivotal position, acting downstream to Meis/Pbx/Hox PG1 proteins, but upstream to PG2–3 Hox gene activity.

Retinoic acid (RA) is a vitamin A derivative that acts as potent signaling ligand during early vertebrate development. RA acts as a diffusible morphogen, and excess RA during embryogenesis perturbs development of the brain, limbs, heart, gut, and other organs. RA synthesis and levels are tightly regulated by the interplay and differential expression of RA synthesizing and RA degrading enzymes (reviewed in White and Schilling, 2008). RA binds RA receptors (RARs), which contact DNA as heterodimers with retinoid X receptors in transcriptional regulatory regions called retinoic acid response elements (RAREs). RA binding alters RAR conformation, which modulates binding of additional transcriptional regulatory proteins that can activate or repress gene expression.

In the vertebrate hindbrain, RA is crucial for formation of posterior rhombomeres. RA signaling inhibition causes a loss of the r4 to r7 regions, whereas excess RA activity induces an expansion of posterior rhombomeres, with a parallel loss of anterior rhombomeres (reviewed in White and Schilling, 2008). RA plays a crucial role by regulating Hox PG1–4 gene expression (Bel-Vialar et al., 2002). The Hoxa1, Hoxb1, Hoxa4, Hoxb4, and Hoxd4 genes all contain highly conserved RAREs that act both positively and negatively to modulate the correct spatial and temporal expression of these genes in the hindbrain (Whiting et al., 1991; Langston and Gudas, 1992; Pöpperl and Featherstone, 1993; Moroni et al., 1993; Marshall et al., 1994; Studer et al., 1994, 1998; Frasch et al., 1995; Dupé et al., 1997; Gould et al., 1998; Nolte et al., 2003). In Xenopus, the hoxd1 and hoxa1 genes are direct-targets of RA (Kolm et al., 1997). Similarly, highly conserved 3′ RAREs in the murine Hoxa1 and Hoxb1 genes are required for their initial activation in the r3/4 region (Langston and Gudas, 1992; Marshall et al., 1994; Frasch et al., 1995). Hindbrain RA levels are regulated by the temporal and spatial interplay of the RA synthesizing and degrading enzymes, aldehyde dehydrogenase 1 family member A2 (ALDH1a2/Raldh2), and Cyp26 (White and Schilling, 2008). In mice and Xenopus, mesodermal Raldh2 gene expression is regulated by HOXA1/PBX1/MEIS2 proteins, and all three proteins bind a conserved enhancer element in the Raldh2 gene in vivo (Vitobello et al., 2011). Consequently, Hoxa1/Pbx1/Meis2 mutants or knockdown embryos have RA-depletion hindbrain defects, resulting from poor RA synthesis due to the downregulation of mesodermal Raldh2 gene expression (Vitobello et al., 2011).

RA activates the more anterior Hox PG1 genes, followed by later activation of PG2–4 Hox genes. Initially, Hoxa1 acts with Meis/Pbx proteins and RA to activate hoxb1 expression. Hoxb1 in turn, triggers a cascade of events by means of Krox20, which leads to the activation of hoxa2, hoxb2, and hoxb3 expression. In addition, Meis3 activity is required for RA-mediated induction of hoxd1 gene expression, and Meis3 and RA activities mesh to induce hoxd1 and retinoic acid receptor RARa2.2 gene expression in the Xenopus hindbrain, thus fine tuning RA patterning activity (Dibner et al., 2004). In zebrafish embryos, early cyp26 and hoxb1b gene expression patterns overlap in the hindbrain, suggesting that RA levels must be tightly controlled in the r4 region (Kudoh et al., 2002). Further supporting this fine tune mechanism, RA acts to repress Hoxb1 expression outside of r4 by means of a 5′RARE, as shown in chick and mice (Studer et al., 1994). Later, RA directly activates more posterior Hox PG4 gene expression in r6/7 (Whiting et al., 1991; Moroni et al., 1993; Pöpperl and Featherstone, 1993; Morrison et al., 1997; Gould et al., 1998; Nolte et al., 2003;). RA signaling acts downstream to FGF signaling in both Xenopus and zebrafish. In zebrafish, ectopic FGF3 activity inhibits cyp26 expression while anteriorly expanding hoxb1b expression. Conversely, the inhibition of FGF signaling increased cyp26, while reducing hoxb1b expression (Kudoh et al., 2002). Remarkably, although the basic blue-print of the hindbrain is identical in all vertebrates, some aspects of hindbrain induction appear to differ among vertebrates. For instance in zebrafish, meis3 gene expression levels are highly sensitive to RA levels, but this was not observed in Xenopus, where meis3 expression levels along the AP axis were slightly shifted, but not significantly altered by modulating RA signaling levels (Kudoh et al., 2002; Dibner et al., 2004).

In summary, experimental observations in different vertebrate species support a model for initial hindbrain induction, in which mesodermal Wnt3a signaling initially acts upstream to Meis3/Pbx proteins (Fig. 1A). Meis/Pbx proteins subsequently induce PG1 Hox and FGF3/8 gene expression in r4, which further activates Hox gene expression. Mesodermal Raldh2 and neural Cyp26 levels control RA levels in the hindbrain. The combined activity of Meis/Pbx/Hox proteins is fine-tuned by RA levels in the hindbrain to specify differential Hox gene expression and segmentation of the hindbrain into distinct rhombomeres along the AP axis (Fig. 1B). This is the simplest unifying model for initial hindbrain induction, incorporating Wnt, FGF, and RA signaling and Meis/Pbx/Hox protein activities. Obviously this model oversimplifies the complex nuances of potential regional and temporal interactions between Meis/Pbx, the different Hox proteins and signaling pathways at different AP levels in the hindbrain. Further elucidation of the downstream Meis/Pbx/Hox target-genes needs to be carried out to fully understand the intricate nature of all these interactions and their effects on hindbrain gene expression and eventual pattern formation.

Figure 1.

Regulation of hindbrain induction and specification. A: Wnt3a from the paraxial-fated mesoderm (yellow) induces Meis3 expression and subsequent hindbrain cell fates in the neural plate (blue). B: Meis/Pbx proteins control hindbrain development. Meis3 acts by means of FGF3/8-induced MAPK signaling, in co-operation with RA signaling to induce PG1 Hox gene expression. PG1 Hox proteins subsequently activate expression of downstream PG2–4 Hox genes, which further specify the more posterior hindbrain regions. Meis3 induces later neural-specific Wnt3a gene expression. In a feed-back loop: Wnt3a/Meis3 proteins activate expression of the zinc-finger protein Teashirt 1 (Tsh1). Tsh1 in turn negatively restricts Meis3 activity and expression in the hindbrain. Controlling RA signaling, mesodermal (yellow) Hoxa1/Pbx1/Meis2 proteins regulate Raldh2 expression, while in the neural plate (blue) FGF3/8 controls Cyp26 expression. The balance of Raldh2 and Cyp26 enzymatic activities optimizes RA signaling levels in the hindbrain.

Upstream and Downstream to TALE and Hox Proteins in the Hindbrain

During hindbrain development, Meis/Pbx and Hox proteins interact at two distinct levels. Early in development, Meis proteins are required to initially activate PG1–4 Hox gene expression in the hindbrain. Later, Meis/Pbx/Hox protein combinations bind target genes to activate their transcription. Some of these genes are also Hox genes themselves, but non-Hox gene targets also lie downstream to Meis/Pbx/Hox (Rohrschneider et al., 2007; Choe et al., 2011). Meis/Pbx proteins act as transcriptional activators in hindbrain induction (Dibner et al., 2001). In zebrafish embryos, ChIP studies showed that Meis/Pbx proteins specifically bind the hoxb1a and hoxb2a promoters in their respective tissues of expression (Choe et al., 2009). These promoters were also enriched for histone H4 acetylation. In embryos ectopically expressing dominant negative Pbx proteins, Meis/Pbx activity was inhibited and histone acetylation was highly reduced. Furthermore, Meis/Pbx protein complex formation removes histone deacetylase (HDAC) from Hox regulated promoters. Additionally, Meis also appears to be crucial for recruiting CBP/p300 histone acetylase to Hox promoters. Thus, Meis functions as a direct transcriptional activator of target genes by controlling the accessibility of HDAC/CBP proteins to the hoxb1a promoter (Choe et al., 2009).

In rhombomeric segments, Hox gene expression is controlled by auto- and cross-regulatory binding of Meis/Pbx/Hox proteins to enhancer elements. For example, the murine Hoxb1 gene enhancer that drives expression in r4 harbors distinct Meis/Pbx and Hox/Pbx binding sites. HOXA1 together with Meis/Pbx proteins initially activate this enhancer, but later, HOXB1 itself, together with HOXB2, MEIS3, and PBX2/4 proteins, are required for expression maintenance in r4 (Pöpperl et al., 1995; Ferretti et al., 2005). Many additional examples of such interdependent, cross-regulatory loops are known. For instance, hoxd1, hoxb2, and hoxa2 all require Meis/Pbx for their expression (Ferretti et al., 2000; Elkouby et al., 2010). Hoxb2 expression is directly activated by HOXB1 in r4, and HOXB2 protein then drives Hoxb1 expression. Thus, Hoxb2 indirectly controls its own expression by means of Hoxb1 (Maconochie et al., 1997; Davenne et al., 1999; Jacobs et al., 1999; Ferretti et al., 2000; Gavalas et al., 2003). Hoxa2 expression in r4 is also regulated by conserved vertebrate enhancer elements that bind Meis/Pbx and the HOXB1/HOXA2 proteins (Tümpel et al., 2006, 2007; Lampe et al., 2008). In more anterior r2-r3, Meis3/Pbx proteins are required for the early neural expression of hoxa2 in Xenopus (Elkouby et al., 2010), but it is presently unresolved if this occurs directly or indirectly (Tümpel et al., 2008). More posteriorly, Hoxa3 expression in r5/6 is controlled by an element binding Meis/Pbx and HOXA3/HOXB3 proteins, as shown in mice and chicks (Manzanares et al., 2001). In r6/7, the Hoxb4 and Hoxd4 genes are initially activated through their RARE elements by RA (Whiting et al., 1991; Moroni et al., 1993; Pöpperl and Featherstone, 1993; Gould et al., 1998; Nolte et al., 2003), but later Hoxb4 expression is maintained by MEIS1/PBX and a separate HOXB4/HOXD4 responsive element (Gould et al., 1998, 1997; Serpente et al., 2005). Notably, the murine RARβ gene itself contains a HOXB4/HOXD4 responsive element, which can be bound by Meis/Pbx proteins together with Hox (Serpente et al., 2005). This provides a positive feed-forward loop to maintain RA levels and PG4 Hox expression in r6/7. Finally, adding an additional level of complexity, all rhombomeres fail to develop properly upon Meis3 knockdown in Xenopus, despite the fact that meis3 expression is restricted to r2–4 (Dibner et al., 2001; Gutkovich et al., 2010). This is likely the result of elimination of early fgf3 expression in r4 of Meis3-depleted embryos, which triggers nonautonomous effects. Thus any direct activation of meis/pbx/hox enhancer sequences in r5–7 could be mediated by additional Meis or Prep proteins.

In Xenopus, Meis3 also appears to directly auto-regulate its own expression. As determined by bioinformatics and ChIP assays, there are two functional Meis protein binding sites in the meis3 gene promoter in vivo. At late gastrula stages, Meis3 protein activates its own expression (Elkouby et al., 2012); later, at early neurula stages, Meis3 induces tsh1 gene expression. Tsh1 protein together with Meis3 binds the meis3 promoter and negatively regulates its expression. Meis3/Tsh1 interactions also control the ratio of terminally differentiated primary neurons vs. proliferating hindbrain neuron precursors through cell cycle modulation (Elkouby et al., 2012). Notably, a similar role has been proposed for Meis1/2 in the developing retina (Bessa et al., 2008; Heine et al., 2008).

Two studies in zebrafish identified novel hindbrain-specific target genes that lie downstream to Meis/Pbx/Hox PG1 proteins (Rohrschneider et al., 2007; Choe et al., 2011). One study compared gene expression in hoxb1a mutant vs. wild-type embryos, identifying eight novel genes expressed specifically in r4 (Rohrschneider et al., 2007). One of these genes, prickle1b was shown to be required for facial motor neuron migration. An additional study in zebrafish aimed at the isolation of Hoxb1b target genes led to the identification of 25 novel hindbrain specific genes (Choe et al., 2011). Other studies also identified individual Meis/Pbx/Hox target genes. The LIM domain only-1 (Lmo1) gene strictly depends on MEIS/PBX/HOX PG2 proteins for its expression in the mouse hindbrain (Matis et al., 2007). Furthermore, the paired-like Homeobox 2b (Phox2b) gene, an obligatory determinant of visceral motor neuron specification, is a direct target of MEIS/PBX/HOXB1–2 proteins in r4 (Samad et al., 2004). Finally in r4 in mice, the receptor tyrosine kinase EphA2 gene enhancer is directly regulated by HOXA1/B1 and PBX proteins, and its expression is lost in Xenopus Meis3 morphants (Chen and Ruley, 1998; Dibner et al., 2001). Ephrin ligands and Eph receptor proteins are expressed in alternating rhombomeres in complementary expression patterns, driving attraction/repulsion between different cell populations thus dividing the hindbrain into metameric rhombomeric segments (reviewed in Sela-Donenfeld and Wilkinson, 2005). Based on bioinformatic predictions, a recent study in zebrafish detected functional hindbrain enhancer sequences containing Meis binding motifs (Burzynski et al., 2012).

HOX-INDEPENDENT FUNCTIONS OF TALE PROTEINS: EYE, MIDBRAIN, AND FOREBRAIN DEVELOPMENT

The anterior border of Hox gene expression corresponds to the boundary between rhombomeres r1 and r2, mes-, so the di-, and telencephalon develop without any direct contribution by members of the Hox cluster. Nevertheless, PBC and Meis/Prep family proteins are widely expressed anterior of the r1/2 boundary, suggesting that they control cell differentiation programs independent of Hox proteins.

A (Largely) Conserved Role for Homothorax/Meis During Early Eye Development

At first glance, the arthropod compound eye and the vertebrate lens eye have little in common, except for both containing cells that detect light and convert it into an electrochemical signal. The compound eye consists of multiple ommatidia units, each composed of eight light sensing photoreceptors and accessory cells. The vertebrate retina is a multicellular, layered structure that lines the back of the eye. Vertebrate retina and invertebrate compound eyes not only differ considerably in their anatomical structure, but also develop from very different precursor cell domains in the embryo. The invertebrate eye primordium (eye imaginal disc) is an epithelial infolding, whereas the vertebrate neural retina develops from the optic vesicle, an out-pocketing of the neural tube.

Thus, it was quite surprising when it was discovered that the transcription factor Pax6 and its fly ortholog eyeless (ey) are not only necessary for eye development in both phyla, but also sufficient to trigger ectopic eye formation when misexpressed in certain regions of the embryo (Hill et al., 1991; Quiring et al., 1994; Halder et al., 1995; Chow et al., 1999; Gehring and Ikeo, 1999). Partial functional conservation was subsequently reported for other transcription factors, including the basic helix-loop-helix type proneural gene athonal (ato)/Ath5, which endows progenitor cells with the competence to exit the cell cycle and initiate neuronal differentiation (Brown et al., 2001; Kay et al., 2001; Wang et al., 2001). Recent work in Drosophila, zebrafish, chick, and mouse implicate Meis/Prep class proteins in this evolutionary conserved network that controls eye development.

The Drosophila eye imaginal disc is specified in the late embryo by the expression of the Pax6 orthologs, ey and twin of eyeless (toy) (Czerny et al., 1999). At early larval development, progenitor cells of the eye imginal disc proliferate rapidly and express the Meis/Prep ortholog homothorax (hth) together with ey and toy (Pai et al., 1998; Pichaud and Casares, 2000; Bessa et al., 2002). At the end of larval development, during third larval instar (L3), a wave of differentiation sweeps across the eye imaginal disc, which can be morphologically recognized as the morphogenetic furrow (MF; Fig. 2A). As the MF moves from posterior to anterior, it coordinates the cell cycle and initiates photoreceptor development (Wolff and Ready, 1991). hth in cooperation with ey promotes proliferation of cells ahead of the MF (Bessa et al., 2002; Lopes and Casares, 2010). More posteriorly, cells undergo a few rounds of rapid cell division before they transiently arrest their cell cycle in G1. hth modulates the kinetics of these cell divisions by regulation of cdc25/string (stg), a protein phosphatase that facilitates progression through the cell cycle (Lopes and Casares, 2010). As cells enter the transient G1-arrest, they downregulate hth and induce expression of the proneural gene ato (Bessa et al., 2002; Lopes and Casares, 2010). ato expression, in turn, is a prerequisite for further differentiation and the formation of mature ommatidia (Jarman et al., 1994). Hence, downregulation of hth and upregulation of ato are associated with the transition of immature progenitor cells to committed precursor cells in the fly eye anlage. Consequently, both factors affect eye development in opposing ways: hth keeps the cells in an undifferentiated, proliferating progenitor state and ato triggers neuronal differentiation (Bessa et al., 2002).

Figure 2.

Partial conservation of hth/Meis regulation in vertebrate and invertebrate eye development. A: Expression of Meis (blue) and Ath5 (brown) in the vertebrate retina (top panel, zebrafish as example) and hth (red) and ato (white) in the Drosophila eye imaginal disc. B: Regulatory relationship between Meis, Pax6, Athonal, and Hedgehog relative to the propagating wave of neurogenesis. See text for details. The images in (A) are courtesy of Joana Santos, Carla Lopes and Fernando Casares (CABD, Sevilla Spain).

These events are partially conserved in the developing vertebrate retina (Fig. 2B). Six types of neurons and one type of glia differentiate in the retina over an extended period and in subsequent waves, which are reminiscent of the developing Drosophila eye (Livesey and Cepko, 2001). Each wave initiates in the central retina in close proximity to the optic stalk, and progresses toward the retinal margin. In striking similarity to the Drosophila L3 eye imaginal disc, expression of the atonal-homologue 5 (Ath5/Atoh7) is closely associated with the initiation of retinal neurogenesis in all vertebrates (Brown et al., 2001; Kay et al., 2001; Wang et al., 2001). The spatial-temporal expression patterns of Meis1 and Meis2 parallel that of hth during early retinal development in some vertebrates. meis1 in fish and MEIS2 in chick are uniformly transcribed in the early optic vesicle when cells proliferate. Similar to the L3 eye imaginal disc, both proteins are downregulated in a central-to-peripheral wave that parallels the appearance of ath5-expressing cells (Bessa et al., 2008; Heine et al., 2008). Moreover, both Meis proteins promote rapid progenitor cell proliferation through regulation of c-myc and/or cyclinD1 (ccnd1), two major regulators of progression through the G1-phase of the cell cycle. Thus in flies, fish, and chicks, retinal progenitor cells ahead of the first neurogenic wave express a MEIS/Hth protein that maintains their proliferative state. Consequently, fish or chicks with compromised Meis1 or MEIS2 function are microphthalmic and mice, which carry a targeted deletion of the Meis1 locus that results in the expression of a truncated, dominant-negative protein, develop small eyes (Hisa et al., 2004; Bessa et al., 2008; Heine et al., 2008). Meis family proteins thus control retinal progenitor cell proliferation and ocular growth in vertebrates and invertebrates.

Some aspects of the regulation of Meis/hth are also conserved. In Drosophila, hth expression is downregulated in response to the BMP (bone morphogenetic protein) 2/4 ortholog decapentaplegic (dpp) and hedgehog (hh) signals (Bessa et al., 2002; Firth and Baker, 2009; Lopes and Casares, 2010). hh is produced by differentiated cells behind the furrow and signals short-range to induce dpp at the MF (Heberlein et al., 1993). Together, hh and dpp drive the forward movement of the MF, in part by cooperatively downregulating hth expression in the progenitor domain ahead of the furrow (Lopes and Casares, 2010). Similarly, the vertebrate hedgehog ortholog Sonic hedgehog (Shh) is expressed by cells of the optic stalk, as well as retinal ganglion cells when their differentiation spreads across the retina. Shh drives retinal neurogenesis, although the underlying mechanisms are still debated (Neumann and Nuesslein-Volhard, 2000; Zhang and Yang, 2001; Wang et al., 2005; Kay et al., 2005; Locker et al., 2006). In the chick retina, the central-to-peripheral wave of MEIS2 repression, which normally accompanies the production of ATH5-expressing cells, is accelerated when retinas were forced to overexpress SHH (Heine et al., 2009). In contrast to Drosophila, however, BMP2/4 signaling is not involved in MEIS2 downregulation (Heine et al., 2009). This is not surprising, as MF progression and ommatidia development in Drosophila can occur in the absence of dpp, arguing that hedgehog signaling may be the primary mechanism to ensure progression of the neurogenic wave across the eye anlage (Burke and Basler, 1996). The molecular details underlying the negative regulation of MEIS2 by hedgehog remain to be elucidated and it is still unclear if the loss of MEIS2 transcripts that occurs during normal retina development results from SHH secretion by the optic stalk, by newly born retinal ganglion cells or both. Nevertheless, these results argue for an evolutionary conserved regulatory relationship between hedgehog signaling and MEIS/hth expression.

Despite many similarities in MEIS/Hth protein function in the vertebrate and invertebrate eye, species specific differences also exist. For instance, the temporal sequence of Meis1 and Meis2 expression differs across species. In the chick retina, MEIS2 is present at high levels in progenitor cells ahead of the first neurogenic wave, while MEIS1 expression continues in proliferating progenitor cells throughout retinal development. In contrast, Meis2 does not seem to play a major role in retina development at comparable embryonic stages in mice and fish (Bumsted-O'Brien et al., 2007; Bessa et al., 2008; Heine et al., 2008) (D. Engelkamp and D. Schulte, unpublished results). Retinal progenitor cells in the chick thus express two MEIS family proteins at a developmental time in which mice and fish express only one. Considering that the eyes in these species vary greatly in their overall size and cellular composition, such differences may reflect species-specific strategies to adapt eye growth and retinal differentiation to divergent animal habitats.

Later Roles for TALE Proteins in Visual System Development

TALE homeodomain proteins also control later steps in eye development, such as cell fate specification and neuronal wiring. For example, 5′-TG-3′-interacting factor (TGIF), a more distantly related member of the MEIS/PREP subclass of TALE homeodomain proteins, is involved in the generation of retinal amacrine cells, interneurons that regulate parallel processing of visual signals in the retina (Wässle, 2004; Satoh and Watanabe, 2008). MEIS2 may also play a role in the specification of particular amacrine cell subtypes, although molecular details of the relationship between TGIF and MEIS2 remain unresolved (Bumsted-O'Brien et al., 2007; Satoh and Watanabe, 2008).

TALE proteins also contribute to the orderly projection of retinal ganglion cell axons into the brain. A principal feature of axonal projections in the visual system is their organization into topographic maps, whereby neighboring nerve cells in one structure form synapses onto closely located neurons in the target structure. Retinotectal map formation relies on the temporally and spatially controlled expression of cell surface proteins, such as Eph receptor tyrosine kinases and ephrins, and the transcriptional regulators that control their graded expression across the retina and tectum. Knockdown of pbx2 in a pbx4-mutant background in zebrafish altered patterning of the retina and tectum and caused retinal ganglion cell pathfinding errors (French et al., 2007). This can be explained by loss of expression of several factors, which were previously shown to control retinal ganglion cell patterning and retinotectal map formation, such as raldh2/aldh1a2 (Hyatt et al., 1996; Sen et al., 2005; Halilagic et al., 2007), tbx5 (Koshiba-Takeuchi et al., 2000), and hmx4/soho-1 (Schulte and Cepko, 2000). Similarly, knockdown of zebrafish meis1 led to partial ventralization and subtle changes in nasal-temporal identity of the retina (Erickson et al., 2010). This is likely due to altered gene expression profiles of the retinal patterning genes tbx5, vax2 (Schulte et al., 1999; Mui et al., 2002), foxg1a and foxd1 (Yuasa et al., 1996), several Eph/ephrin molecules, which are known transcriptional targets of these patterning genes, and altered BMP- and FGF signaling.

The Meis/Prep–cyclinD1 Relationship

The previously described observation that MEIS/Hth proteins modulate eye growth through the transcriptional regulation of cdc25/stg and the G1-S cyclin ccnd1 implicate Meis family proteins in maintaining the balance between proliferation and differentiation of eye progenitors. The regulation of two important mitogenic triggers by Meis family proteins is remarkable, considering that Meis-dysregulation has been associated with several forms of cancer (Geerts et al., 2003; Crijns et al., 2007). Indeed, cyclinD3 (Ccnd3) is a direct target of MEIS1 in a murine model for acute myeloid leukemia (AML) (Argiropoulos et al., 2010).

The regulatory relationship between MEIS-proteins and Ccnd1 took an interesting turn with the discovery that Meis2 itself is a transcriptional target of the DNA-associated form of CCND1 in the murine retina (Bienvenu et al., 2010). Besides its role as cell cycle regulator, CCND1 is also present in transcription factor complexes. CCND1 has been reported to associate with various transcriptional cofactors, including the histone actetyl transferases P/CAF and NcoA/SRC1a, the histone deactelylase HDAC3, or TFAII250, a component of the core transcription factor complex (Coqueret, 2002). Chromatin immunoprecipitation followed by DNA microarray analysis (ChIP-chip) from tissues of mice expressing epitope-tagged forms of CCND1 in vivo, revealed CCND1 occupancy in promoter regions of several abundantly expressed genes (Bienvenu et al., 2010). Strikingly, Meis2 was among the genes bound by CCND1 in the P0 mouse retina. Subsequent expression profiling revealed that Meis2 transcripts were strongly reduced in the retinas of mice mutant for Ccnd1, indicating that Meis2 is normally under positive control of Ccnd1. Due to the experimental design of the study (ChIP from whole P0 retinas), it is not yet resolved if this regulation occurs in all cells that co-express both proteins or is restricted to particular cell types. Nevertheless, these results support an intriguing model whereby CCND1 and MEIS2 engage in positive reciprocal regulation and mutually enforce each other's expression. This model has yet to be tested in tissues outside the murine retina and still has yet to be extended to other TALE proteins, but it may help to understand the full oncogenic potential of MEIS/PREP family proteins.

The Meis/Prep–Pax6 Relationship

Work on MEIS family proteins in the developing eye has also discovered a regulatory interaction between MEIS/PREP and PAX6 proteins. Pax6 and its fly ortholog ey are key regulators of eye development in Metazoans, but Pax6 also fulfills crucial functions in the development of other tissues and organs, including different brain regions, spinal cord, olfactory system, and pancreas (Hogan et al., 1988, 1986; Hill et al., 1991; Stoykova and Gruss, 1994; Ericson et al., 1997; St-Onge et al., 1997). In the vertebrate brain anlage, Pax6 is first expressed at the end of gastrulation in the entire anterior neural plate. Later, expression becomes progressively restricted to the optic vesicles and the overlying, nonneuronal surface ectoderm, where it is co-expressed with Meis1 and Meis2. Although the genomic organization of the Pax6 gene locus is highly complex, a short, 107 bp cis-regulatory element was identified that directs expression in the prospective mouse lens (Williams et al., 1998; Kammandel et al., 1999). This element is bound by MEIS1 and MEIS2 in the developing lens, and MEIS binding is required for its activity (Zhang et al., 2002). TALE homeodomain family proteins hence function upstream of Pax6 during lens formation. MEIS2 also directly or indirectly participates in the regulation of PAX6 in the chick retina (Heine et al., 2008). pax6 was also identified as Meis2 target in the optic cup and lens of medaka fisch (Conte et al., 2010). In Prep1-hypomorphic mice, Pax6 transcript levels were significantly reduced, arguing that also in these tissues Pax6 is downstream of Meis/Prep protein activity (Ferretti et al., 2006). Considering that Pax6 is a key regulator of ocular fate, these findings suggest that Meis/Prep may link retinal fate specification and growth control in the developing eye. The reverse regulation, namely control of meis1 by Pax6, also occurs, at least in zebrafish (Royo et al., 2012). An enhancer element of the zebrafisch meis1 gene, which directs expression in the embryonic neural retina and distinct brain regions, contains two conserved Pax6 consensus binding sites, which are required for normal retinal meis1 expression. Although the biological relevance of these sites is unclear, reciprocal regulation of Meis and Pax6 is possible.

The microRNAs miR-9 and miR-204, target Meis2 in the forebrain and lens respectively (Conte et al., 2010; Shibata et al., 2011; Hoffmann et al., 2012). In mice lacking mir9-2 and mir9-3, Meis2 expression in the telencephalon increased and its expression domain extended dorsally. Notably, the elevated Meis2 expression in these mice was accompanied by a concomitant decrease in Pax6 expression, which suggests that in this system Meis2 negatively regulates Pax6 (Shibata et al., 2011). In medaka fish, mir-204 directly regulates meis2 expression in the lens and retina, and mir-204 overexpression resembled the phenotypic alterations observed in meis2 loss-of-function models (Conte et al., 2010). In sharp contrast to the mouse telencephalon, however, mir-204 dependent loss of meis2 expression was paralleled by loss of pax6 expression in the medaka eye, indicative of a positive regulation of pax6 by meis2. Although both studies once more provide compelling evidence for a Meis-Pax6 transcriptional pathway, they also suggest that this pathway likely has been considerably modified in different physiological contexts.

Adding even more complexity to the regulatory relationship between Meis/Prep and Pax6, was the finding that sub-saturation binding of PREP1 to two evolutionary conserved, low affinity binding sites in an extended Pax6 lens regulatory region is critical for the correct timing and level of Pax6 expression in the lens (Rowan et al., 2010). Pax6 function in eye development is highly dependent on gene dosage. Mice carrying multiple copies of the Pax6 gene exhibit severe eye malformations, which share some features (such as microphthalmia and cataracts) with mice heterozygous for the Pax6/sey loss-of-function mutation (Schedl et al., 1996; van Raamsdonk and Tilghman, 2000). The study by Rowan and colleagues identified two novel binding sites for PREP1 in a Pax6 lens enhancer, which both exhibit phylogenetically conserved nucleotide substitutions compared with the optimal consensus motif for Meis/Prep proteins that had been identified by in vitro selection (Chang et al., 1997). Both motifs were bound by PREP1 at levels below saturation in the embryonic lens and PREP1 binding activated the corresponding enhancer synergistically in a dose-dependent manner. Dose-dependent activation of gene expression by PREP1 had previously also been observed for fatty acid binding protein 7 (FABP7)/brain lipid binding protein (BLBP) in Down syndrome (Sánchez-Font et al., 2003). The presence of multiple low-affinity binding sites that can be bound noncooperatively by PREP1 in the Pax6 lens enhancer may allow for tighter control over Pax6 expression levels than the presence of a single high-affinity site (Rowan et al., 2010). It is tempting to speculate that similar mechanisms may account for some of the divergent functions that have been described for TALE-proteins in different developmental contexts or disease. Whole genome approaches like ChIP-Seq or ChIP-chip will undoubtedly help to answer this question.

TALE Proteins in Midbrain Development

Meis and Pbx proteins also have fundamental roles in midbrain development. The midbrain is an important center of sensory-motor integration developing from the mesencephalic vesicle. The dorsal region, derived from the mesencephalic alar plate, comprises the superior and inferior colliculi in mammals or optic tecta in non-mammalian vertebrates. It is involved in visual and auditory information processing. The ventral midbrain, derived from the mesencephalic basal plate, comprises the substantia nigra and ventral tegmental area, two important sites of dopamine production. Midbrain development is tightly linked to the activity of the mid–hindbrain boundary organizer (MHB, or isthmic organizer), a group of cells located at the junction between the mesencephalic and metencephalic vesicles (Wurst and Bally-Cuif, 2001). Cells of the MHB organizer secrete long-range and short-range signaling molecules, which are necessary and sufficient for the development of the adjacent mid- and hindbrain structures (Itasaki and Nakamura, 1992; Crossley et al., 1996).

MHB maintenance in zebrafish requires the concerted activities of Pbx and Engrailed (Erickson et al., 2007). Global depletion of Pbx proteins in the neural tube did not perturb the initial establishment and positioning of the MHB or the onset of expression of MHB associated genes, but profoundly affected MHB maintenance. Pbx-depleted embryos lack visible MHB structures at late somite stages and develop stunted tecta. Pbx proteins bind to Engrailed2a (Eng2a), a homeodomain containing transcription factor associated with MHB activity, by means of a motif resembling the primary binding domain in Hox-proteins. Pbx and Engrailed cooperate in MHB maintenance, because mutation of the Pbx-binding motif in Eng2a renders the protein biologically inactive (Erickson et al., 2007). Again, this genetic interaction is evolutionary conserved: Drosophila engrailed (en) associates with exd in vitro and cooperates with exd and hth in compartment boundary formation in the embryo (Peltenburg and Murre, 1996; Kobayashi et al., 2003).

A role for Meis family proteins in midbrain development has also been reported (Erickson et al., 2007; Agoston and Schulte, 2009; Agoston et al., 2012). Meis2 is abundantly expressed in the mesencephalic alar plate of mice and chicks, the anlage of the tectum and superior colliculus respectively (Cecconi et al., 1997; Agoston and Schulte, 2009). MEIS2 is both necessary and sufficient for tectal development in chicks. In sharp contrast to PBX, however, MHB formation and maintenance were independent of MEIS2 (Erickson et al., 2007; Agoston and Schulte, 2009). Likewise, zebrafish embryos in which meis1 expression was knocked down had smaller tecta and exhibited retinotectal mapping defects, but did not show noticeable disturbance in the expression of MHB-associated genes (Erickson et al., 2007). PBX proteins thus likely act earlier than MEIS or have additional functions in the genetic cascade that regulates tectum development. In chicks, MEIS2 triggers ectopic tectal development by interacting with OTX2, a non-Hox homeodomain protein, which is essential for the development of all anterior brain structures, including the midbrain. Binding of MEIS2 to OTX2 competes for association of the transcriptional co-repressor TLE4/GRG4 with OTX2 in the tectal anlage, releases OTX2 from TLE-mediated repression and thereby allows tectal development to commence (Agoston and Schulte, 2009). Interaction between MEIS2 and OTX2 does not require either protein to be bound to DNA. This suggests that MEIS family proteins can modulate gene expression by at least two different mechanisms. They can directly act on promoter/enhancer regions of downstream genes and they can initiate the controlled assembly and disassembly of transcription regulator complexes independent of DNA. Interestingly, a very similar mode of action had been described previously for TGIF (Wotton et al., 1999). TGIF was originally identified for its ability to compete with retinoic X receptors (RXR) for binding to RXR-response elements in the promoter of retinoic acid responsive genes, thereby inhibiting retinoic acid induced transcriptional activation of these genes (Bertolino et al., 1995). However, TGIF also represses transcriptional activation downstream of the TGF-β signaling pathway (Wotton et al., 1999). TGF-β signals are transmitted into the cell by means of smad-proteins, which translocate into the cell nucleus and activate gene expression by interacting with transcriptional co-activators such as the histone acetyl transferases p300 and CBP (Shi and Massagué, 2003). TGIF competes with p300 for binding to SMAD2 and recruits transcriptional co-repressors like histone deacetylases into the complex, which leads to repression of TGF-β induced transcription (Heinzel et al., 1997; Wotton et al., 1999; Sharma and Sun, 2001). Although ultimately leading to very different outcomes—overriding transcriptional repression in the case of MEIS2 interaction with OTX2 or inducing repression in the case of TGIF interaction with SMAD2—in both contexts TALE homeodomain proteins regulate gene expression by changing the composition of preexisting transcription factor complexes.

MEIS2 downstream targets in the tectum include ephrin B1 (EFNB1, in chick) and EphA8 (in mice), two cell surface molecules involved in the proper development of retinotectal / retinocollicular projections (Shim et al., 2007; Agoston and Schulte, 2009). In the forebrain, expression of another Eph receptor, EphA7, is directly regulated by PBX1 (Pietri et al., 2012). Besides OTX2, MEIS2 also directly interacts with PAX3 and PAX7 in the chick, two paired box transcription factors, which function in the network that shapes the MHB (Matsunaga et al., 2001; Agoston and Schulte, 2009).

In contrast to the dorsal midbrain, the role of TALE proteins in the ventral midbrain is not well understood. A recent study implicates PBX proteins in dopaminergic cell differentiation in the ventral midbrain (Sgadò et al., 2012). The dopamine system is involved in a broad range of functions including motivation, reward processing and voluntary movements, and the ventral midbrain harbors several key components of this system. Pbx1 and Prep1 are expressed in mesencephalic dopaminergic neurons during embryogenesis and in the adult. In Pbx1-deficient mice, development of midbrain dopaminergic neurons is initially normal, but axon projections are misrouted at mid-gestation, which is likely due to loss of high-affinity netrin receptor DCC expression in these cells (Sgadò et al., 2012). These results point toward a role for PBX proteins in dopaminergic neuron development, although the precise role of TALE proteins in the dopamine system awaits further investigation.

TALE Proteins in Forebrain Development

TALE homeodomain proteins also regulate several aspects of forebrain development. Clues for a role for TALE proteins in forebrain development stem from expression analyses, some of which were already associated with the initial cloning of the genes (Roberts et al., 1995; Bertolino et al., 1996; Cecconi et al., 1997; Oulad-Abdelghani et al., 1997; Toresson et al., 1999, 2000; Schnabel et al., 2001; Coy and Borycki, 2010). In contrast to the hindbrain, midbrain, and retina, only few mechanistic details about TALE protein function in the forebrain are known. Rather, the involvement of TALE proteins in forebrain development has largely been deduced from the association of particular mutations with specific cortical defects in humans or mice.

Meis1 and Meis2 are expressed in the lateral and medial ganglionic eminences (LGE and MGE) of the ventral forebrain with partially overlapping expression domains (Toresson et al., 2000). In fact, Meis2 was identified in a subtraction cloning approach for genes enriched in the LGE (Toresson et al., 1999). MGE and LGE give rise to the basal ganglia, including the striatum, and are also a major source for GABA-ergic interneurons of the cerebral cortex, hippocampus, and olfactory bulb (Parnavelas, 2000). Several studies place MEIS proteins in the regulatory network that controls the development and function of striatal neurons and cortical interneurons. For instance, MEIS2 participates in the regulation of Dlx (distal-less homeobox) genes, which are required for neuronal differentiation and migration in the ventral telencephalon. Specifically, MEIS2 activates transcription of an enhancer in the Dlx5/6 cluster, which drives Dlx expression in postmitotic striatal projection neurons that derive from LGE progenitors (Ghanem et al., 2008). Conversely, Meis genes were identified by ChIP-chip as direct targets of the ARX (aristaless-related homeobox) protein (Friocourt and Parnavelas, 2011; Quillé et al., 2011). ARX is prominently expressed in GABA-ergic interneurons of the cortex and mutations in the ARX gene lead to a wide spectrum of brain phenotypes in humans, including a particular form of X-linked lissencephaly (Kitamura et al., 2002). Finally, expression of the D1A dopamine receptor, one of the predominant dopamine receptors on striatal neurons, is under direct regulation by MEIS2 and TGIF (Yang et al., 2000). Of interest, both TALE proteins compete for binding to overlapping sites upstream of the D1A transcriptional start site, exerting opposing functions. MEIS2 activates D1A transcription, while TGIF represses it. Hence, the balance between MEIS2 and TGIF expression levels may determine the abundance of D1A receptors in a given cell and consequently the cell's ability to respond to dopamine signals. The link between TALE proteins and dopaminergic neurotransmission in the striatum is likely of clinical relevance, as the dopamine system has been implicated in Parkinson's disease as well as in neuropsychiatric disorders such as schizophrenia, addiction, and obsessive-compulsive disorder. In fact, a very recent genome-wide linkage analysis identified MEIS2 and PBX1 as potential candidate genes involved in the pathogenesis of obsessive-compulsive disorder (Nestadt et al., 2012).

A second well studied example for a role of TALE homeodomain proteins in forebrain development concerns the involvement of TGIF in holoprosencephaly (Gripp et al., 2000). Holoprosencephaly is a congenital malformation of the forebrain, in which the primary telencephalic vesicle fails to divide into two cerebral hemispheres as a result of defective midline patterning. The human TGIF gene is located on chromsomome 18p11.3, a genomic region associated with the familial HE4 form of holoprosencephaly. Several heterozygous loss-of-function mutations in TGIF were identified in individuals affected with holoprosencephaly (Gripp et al., 2000). In contrast to humans, however, a single mutation of Tgif was not sufficient to cause holoprosencephaly in mice, but required simultaneous deletion of the related Tgif2 gene (Shen and Walsh, 2005; Taniguchi et al., 2012). As shown by a series of genetic experiments in mice, TGIF and TGIF2 function mechanistically to modulate the Shh-pathway by limiting Nodal-signaling by means of inhibition of SMAD2 (Taniguchi et al., 2012).

Third, the human PREP1 gene is located on chromosome 21q22.3 and thus in a genomic region, which is amplified in Down syndrome. PREP1 expression levels are imbalanced in Down syndrome fetal brains, leading to a moderate but significant dysregulation of genes located elsewhere in the genome. One such gene is FABP7/BLBP, which maps to chromosome 6 (Sánchez-Font et al., 2003). FABP7/BLBP is expressed in radial glia, which are both, general precursors for neurons and glia in the brain and key elements in guiding newborn neurons (Feng et al., 1994; Malatesta et al., 2000; Campbell and Götz, 2002). PREP1 may thus contribute to the neurological manifestation of Down syndrome. Supporting this assumption, elevated levels of PREP1 in fibroblasts from Down syndrome patients increase the sensitivity of these cells to genotoxic stress, such as UV irradiation (Micali et al., 2010).

Finally, hypothalamus production of gonadotropin-releasing hormone (GnRH) is regulated by a multimeric complex comprising the POU-transcription factor Oct1 together with Prep1 and Pbx1b (Rave-Harel et al., 2004). GnRH is a peptide hormone responsible for the release of follicle-stimulating hormone (FSH) and luteinizing hormone (LH) from the pituitary gland to control reproductive function in male and female. Interestingly, PREP1 and PBX1 proteins also participate directly in the regulation of FSH expression in the pituitary (Bailey et al., 2004). TALE homeodomain transcription factors thus emerge as modulators of the reproductive axis.

Perspectives

Although substantial progress has been made in understanding the many and often divergent roles of TALE homeodomain proteins in patterning, cell fate specification and differentiation in the vertebrate CNS, obtaining a comprehensive picture of the gene regulatory networks driven by this protein family continues to be a long-term goal. The Meis/Pbx genes are expressed in gastrula stage embryos at the earliest onset of neural development, and identification and characterization of their earliest direct-targets will be crucial for understanding the initial events in neuronal specification and patterning. Using combined strategies of RNA-seq, microarray, and ChIP-seq assays in different vertebrate embryo model systems such as zebrafish, Xenopus, chick, and mouse will hopefully identify new gene targets and further elucidate the Hox-dependent and -independent Meis/Pbx interactions that occur during brain development. Moreover, all TALE proteins are expressed as multiple splice variants, which often differ quite remarkably in their exon composition and hence may regulate gene expression in very different ways (Asahara et al., 1999). For instance, splice variants of hth that lack the homeodomain have different developmental functions than the full-length protein (Noro et al., 2006). It is, therefore, highly likely that some of the alternatively spliced forms of other TALE proteins also possess unique, still undiscovered activities during brain development. Finally, little is known to date about whether TALE proteins are posttranslationally modified in response to extracellular signals. Availability of PBX1 in the cell nucleus, for example, depends on phosphorylation by protein kinase A (Kilstrup-Nielsen et al., 2003). If and how the activity of other TALE proteins is subject to similar regulatory mechanisms and whether such modifications impact on CNS development, are still open questions.

ACKNOWLEDGMENTS

We thank members of the Frank and Schulte labs for helpful discussions, Fernando Casares for providing the images shown in Figure 2A and Susanne Mükusch for creating the figures. D.F. was funded by the Israel Science Foundation (#658/09) and D.S. was funded by the Deutsche Forschungsgemeinschaft. (Schu1218/3-1). The present review was made possible through funding by the EU-COST action BM-0805.

Ancillary