Meis2 competes with the Groucho co-repressor Tle4 for binding to Otx2 and specifies tectal fate without induction of a secondary midbrain-hindbrain boundary organizer

Otx2 is a member of a highly conserved family of homeodomain-containing transcription factors that control early development of anterior brain structures in different animal phyla. Mice mutant for Otx2 lack fore-and midbrain and D. melanogaster embryos mutant for the related gene orthodenticle (ocelliless - FlyBase) do not form the anterior-most part of the head (Acampora et al., 1995; Finkelstein and Perrimon, 1990). Otx2 controls multiple steps during vertebrate brain development, beginning with the specification of the anterior visceral endoderm, which acts as a head organizer by inducing Otx2expression in the anterior part of the adjacent neuroectoderm(Simeone, 2002). Subsequently,this Otx2 domain becomes subdivided into molecularly distinct units,which eventually give rise to the different brain structures of the fore- and midbrain. Thus, whereas Otx2 is required for the specification of general characteristics of the anterior vertebrate brain, additional factors need to cooperate with Otx2 in the specification of individual brain structures.

Meis1-3 belong to the TALE (three amino acid loop extension) family of homeodomain-containing transcription factors and function as regulators of cell proliferation and differentiation of several organs and tissues during development. Meis1 and Meis2, for instance, are involved in proximal-distal limb patterning, skeletal muscle differentiation, hindbrain,lens and retina development (Berkes et al.,2004; Bessa et al.,2008; Capdevila et al.,1999; Dibner et al.,2001; Heine et al.,2008; Mercader et al.,1999; Vlachakis et al.,2001; Zhang et al.,2002). To date, only mice mutant for Meis1 have been generated, and these display defects in angiogenesis and eye development(Azcoitia et al., 2005; Hisa et al., 2004). TALE homeodomain proteins form dimeric or trimeric complexes with other transcription factors. Complex formation not only influences the DNA-binding affinity and specificity of the interacting proteins, but has also been shown to control nuclear import of the partner protein(Mercader et al., 1999; Moens and Selleri, 2006; Vlachakis et al., 2001). Meis-interacting proteins have been isolated from non-neuronal tissue and from the hindbrain and include the Meis-related Pbx family, members of the Hox clusters, several homeodomain-containing proteins, and the myogenic bHLH proteins MyoD and Myf5 (Chang et al.,1997; Knoepfler et al.,1999; Shen et al.,1997; Swift et al.,1998; Vlachakis et al.,2001). Meis-interacting proteins in the anterior neural tube, by contrast, have thus far remained elusive. Here we identify Otx2 as a direct interaction partner of Meis2 in the tectal anlage and show that Meis2 is both necessary and sufficient for tectal fate specification.

pMIWIII-Meis2HA, pMIWIII-Meis2EnR and the RNAi targeting vector pSTRIKE-Meis2_siRNA were described previously(Heine et al., 2008). Each construct (1-2 μg/μl) was electroporated together with 0.5 μg/μl of a plasmid expressing enhanced green fluorescent protein(pMIWIII-GFP) into the right-hand wall of the neural tube of Hamburger-Hamilton stage (HH) (Hamburger and Hamilton, 1992) 9-11 chick embryos (White Leghorn) as described (Heine et al.,2008). To control for possible non-specific effects,pMIWIII-GFP or a fusion protein of the unrelated homeodomain(HD)-containing protein SOHo1 to the EnR domain were electroporated(Schulte and Cepko, 2000). Retroviral transfection employing RCASBP(B)-Meis2HA was used for widespread, long-term expression (Heine et al., 2008).

The cDNAs used to generate in situ probes were gifts from C. Cepko and C. Tabin (Harvard Medical School, Boston, MA, USA), D. O'Leary (Salk Institute,La Jolla, CA, USA), C. Logan (University of Calgary, Alberta, Canada), A. Pierani (Institut J. Monod, Paris, France) or were cloned from chick HH10-12 whole-head total RNA by RT-PCR with gene-specific primers (sequences available upon request). In situ hybridization was performed as described(Heine et al., 2008). Immunohistochemical detection of the misexpressed transgene after in situ hybridization was carried out with a rabbit antibody against GFP (1:5000;Molecular Probes, Eugene, OR, USA), followed by detection with a horseradish-peroxidase-coupled secondary antibody (1:1000; Roche Diagnostics,Mannheim, Germany) and visualization with the DAB Peroxidase Substrate Kit(Vector Laboratories, Burlingame, CA, USA).

Primary antibodies used were rabbit polyclonal anti-GFP (1:5000; Molecular Probes), rat polyclonal anti-HA (1:1000; Roche), rabbit polyclonal anti-Meis2(1:5000; gift of A. Buchberg, Kimmel Cancer Center, University of Philadelphia Medical School), goat polyclonal anti-Otx2 (1:40; R&D Systems,Minneapolis, MN, USA), mouse monoclonal anti-Pax6 purified IgGs [1:5000;Developmental Studies Hybridoma Bank (DSHB), IA], mouse monoclonal anti-Pax7(1:5; DSHB, IA), and mouse monoclonal anti-neurofilament RM270.7 (1:8000; gift of V. Lee, University of Pennsylvania Medical School, Philadelphia).

Approximately 30-40 HH15-18 chick tecta per experiment were resuspended in 10 mM Hepes pH 8, 10 mM KCl, 0.1 mM EDTA, 2 mM DTT and Complete protease inhibitor tablets (Roche) and lysed by adding Igepal (Sigma Aldrich,Steinheim, Germany) to a final concentration of 1%. Cell nuclei were collected by brief centrifugation. The supernatant contained the cytosolic fraction(`cyto'). The nuclei were reconstituted in 10 mM Hepes pH 8, 10 mM KCl, 0.1 mM EDTA, 2 mM DTT, 400 mM NaCl, 1% Igepal and Complete protease inhibitors and incubated for 15 minutes at 4°C under constant rotation. Cellular debris was removed by brief centrifugation (`nucl'). Cytosolic and nuclear fractions were used separately or combined (designated input, `in'). Lysates were pre-cleared by incubation with empty glutathione sepharose 4B beads (GE Healthcare-Amersham, Piscataway, NJ, USA) or empty Protein G-agarose beads(Roche) for 30-60 minutes under constant rotation at 4°C.

³⁵S-radiolabeled proteins were in vitro translated using the TNT T7/T3 Reticulocyte Lysate System (Promega, Mannheim, Germany) according to the manufacturer's instructions. For each pull-down experiment, 15 μl of radiolabeled protein were diluted in buffer BP (10 mM Hepes pH 8, 10 mM KCl,0.1 mM EDTA, 2 mM DTT, 150 mM NaCl, 0.4% Igepal and Complete protease inhibitors).

GST fusion proteins (in pGEX4T1) were purified following standard procedures. Immobilized GST fusion proteins were incubated with pre-cleared tectal lysates or with ³⁵S-radiolabeled proteins for 2 hours under constant rotation at 4°C. Following extensive washes in buffer BP, the protein complexes were analyzed by SDS-PAGE followed by autoradiography or western blot following standard procedures.

Pre-cleared tectal lysates were incubated with anti-Meis2 antibody overnight at 4°C under constant rotation. Protein G-agarose beads (Sigma Aldrich) were added for 4 hours at 4°C with rotation. After extensive washes, the immunoprecipitates were separated by SDS-PAGE and analyzed by western blot using a goat polyclonal Otx2 antibody (1:2000; R&D Systems). The input (`in') loading control corresponded to less than 10% of the total protein used in the precipitation.

The Gal4-responsive construct plGC-luc and the expression vectors pMC-Gal4Otx2, pMC-Gal4Otx2Δ1-100 and pKW-Tle4 were described previously(Eberhard et al., 2000; Heimbucher et al., 2007). Mouse NIH 3T3 fibroblasts were transfected in 24-well plates using Lipofectamine (Invitrogen, Karlsruhe, Germany) with 300 ng plGC-luc and 50 ng of the Renilla luciferase expression plasmid phRG-TK (Promega)together with empty vector and the expression vectors indicated. Luciferase activities were measured 48 hours after transfection with the Dual Glo Luciferase Assay System (Promega). For normalization, firefly luciferase values were standardized to Renilla luciferase activity. Experiments were performed in at least triplicate.

Meis2 transcripts were first detected in the mesencephalic vesicle at HH11-12 (13- to 16-somite stage; data not shown). From its onset, Meis2 expression was strong in the dorsal mesencephalon, but was absent from the diencephalon, midbrain-hindbrain boundary (MHB) region and rhombomere 1 (Fig. 1A,B). Additional domains of expression at these developmental stages included the neural retina, lens and rhombomeres 2/3(Fig. 1A,B)(Heine et al., 2008). Within the mesencephalic vesicle, Meis2 was strongly expressed in the alar plates (Fig. 1C,D, arrows),which give rise to the optic tecta, and more weakly in the mesencephalic basal plates, which give rise to the tegmenta. Expression was excluded from the roof plate and floor plate (Fig. 1C,D). By HH19, Meis2 transcripts were absent from the mesencephalic basal plates, with the exception of two lateral groups of cells close to the ventral midline, which were presumably newly generated neurons of the nucleus oculomotoris (Fig. 1D, arrowheads). In the late embryonic tectum, Meis2transcripts marked distinct tectal layers(Fig. 1E). Comparable expression was observed in mouse embryos(Fig. 1F). Meis1transcripts, by contrast, were present at very low levels in the mesencephalic vesicle (data not shown) (Shim et al.,2007).

The Meis2 splicing isoform 2a is prominently expressed in the anterior neural tube of HH11-15 chick embryos(Heine et al., 2008). Similar to the related Meis1a, we identified a transcriptional activation domain in the C-terminus of Meis2a (see Fig. S1 in the supplementary material)(Huang et al., 2005). To investigate a possible role for Meis2 in tectal fate specification,we introduced full-length Meis2a fused to an HA tag(Meis2HA) into the diencephalic alar plate at HH10-11 and monitored changes in diencephalic brain morphology 6 and 12 days later. Prominent bulges indicative of a diencephalic-to-mesencephalic fate change were readily observed in all electroporated embryos(Fig. 2A,A′,G,G′)(n=5/5 for E7; n=5/5 for E13). These ectopic bulges were populated by cells that were immunoreactive for the tectal markers Pax7 and Meis2 (Fig. 2B-B″,H). Pax7- and Meis2-expressing cells in the diencephalon were always restricted to the electroporated, right-hand side of the brain and were never found on the contralateral, non-electroporated side of the same embryo(Fig. 2E). Notably, Pax7- and Meis2-expressing cells were arranged in distinct laminae, which although not perfectly recapitulating the lamination of a normal optic tectum, nevertheless strongly resembled the cellular organization characteristic of the mesencephalic alar plate at the respective developmental stages (compare Fig. 2B″,C with 2H,I). Ectopic midbrain-like structures in Meis2HA-transfected brains also failed to express the diencephalic marker Pax6, a further indication of a true cell fate shift of diencephalic to mesencephalic tissue (Fig. 2D,D′). In addition, the structures induced by Meis2 transfection in the diencephalon possessed a superficial,neurofilament-immunoreactive axonal layer similar to the stratum opticum of the endogenous optic tecta (Fig. 2J,K). For a uniform misexpression of Meis2 throughout the mesencephalic vesicle (including the roof and basal plates), we turned to retroviral transfection with the replication-competent retrovirus RCASBP(B)-Meis2HA (Heine et al.,2008). This strategy resulted in a marked increase in tectal volume at the expense of tegmental structures and in the complete absence of the mesencephalic roof plate and dorsal midline-derived structures at mid- to late embryonic stages (Fig. 2L-N). Transfection of Meis2HA into the metencephalic vesicle, by contrast, never initiated the formation of tectum-like structures,nor induced corresponding molecular changes (see Fig. S2 in the supplementary material; data not shown).

Because of the abundance of Meis2-immunoreactive cells in the ectopic tectal structures, we examined whether they corresponded to cells that had retained expression of the Meis2HA transgene, or whether Meis2 expression in these cells reflected upregulation of the endogenous Meis2 protein. Labeling with an HA-specific antibody revealed only a few, randomly distributed Meis2HA-transfected cells at E7 and none at E13(Fig. 2F; data not shown). When Meis2- and HA-immunoreactive cells were monitored at different times following Meis2HA transfection, we found that expression of the electroporated Meis2HA protein decreased after 24 hours post-transfection, whereas the endogenous Meis2 protein continued to be expressed (see Fig. S3 in the supplementary material). Thus, Meis2 stabilizes the diencephalic-to-mesencephalic cell fate change through positive regulation of its own expression.

To test whether Meis2 is not only capable of inducing a diencephalic-to-mesencephalic fate shift, but is also required for normal tectal development, we expressed a fusion of Meis2a to the repressor domain of D. melanogaster Engrailed (Meis2EnR), which we had previously shown to function as a Meis function-blocking construct, or vectors expressing Meis2-specific small interfering RNAs (siRNAs) in the mesencephalic vesicle (Heine et al.,2008). Meis2EnR transfection into the midbrain vesicle resulted in a marked distortion of the optic tectum as compared with non-transfected or mock-transfected brains (n=5/7)(Fig. 3A,C). Coronal sections through Meis2EnR-transfected tecta stained with DAPI revealed that the lumen of the optic tectum, which is normally still prominent at this stage, was often filled with ectopic cell masses and the characteristic tectal lamination was markedly disturbed (Fig. 3A′,A″,C′,C″). Areas with aberrant tectal morphology also failed to express the tectal markers Pax7 and Meis2(Fig. 3B,D). When mesencephalic Meis2 expression was repressed by RNAi-mediated knockdown, the optic tecta displayed abnormalities and frequently failed to separate(Fig. 3F-H). At later embryonic stages, tectal lamination was disrupted in such specimens (compare Fig. 3A″ with 3H). In addition to these structural defects, the lateral longitudinal fascicle (LLF), a prominent axonal trajectory in the midbrain, was disrupted following transfection with Meis2 function-blocking constructs(Fig. 3J-N′). To control for possible non-specific effects, we targeted Meis2EnR or the Meis2-specific knockdown constructs to regions of the neural tube that do not express endogenous Meis2(Fig. 3E-E″,I). Meis2EnR transfection did not notably compromise diencephalic development, as evident in the persistent expression of the diencephalic marker Pax6 in transfected cells, nor did it affect axonal trajectories in the hindbrain (Fig. 3E-E″;data not shown). Similarly, diencephalic morphology was unaltered following Meis2 knockdown (Fig. 3I). Taken together, these results demonstrate that Meis2is necessary for normal tectal development and can induce the formation of tectal structures at the expense of diencephalic, tegmental or roof plate-derived structures when experimentally expressed in the anterior neural tube.

To assess early events following Meis2 transfection, the expression of several genes that are specific for the mesencephalic,metencephalic and diencephalic anlagen was monitored. The cell surface protein ephrin B1 (Efnb1) is expressed in the mesencephalic alar plate at early somite stages, making it a suitable marker for the tectal anlage(Braisted et al., 1997). When Meis2HA together with GFP was introduced into the neural tube at the level of the future diencephalon, Efnb1 transcripts were induced in patches of transfected cells (n=10/12, 24 hours post-transfection; n=8/9, 48 hours)(Fig. 4A-D; data not shown). Transfection of GFP alone did not induce Efnb1(n=0/14; data not shown). Pax6 expression, which we had found to be repressed by Meis2HA at 6 days post-transfection, was already lost 20 hours after Meis2HA misexpression(Fig. 4E-F″)(n=5/5).

Consistent with the loss of tectal features(Fig. 3), transfection of Meis2 function-blocking constructs, but not transfection of a control-EnR fusion protein, in the midbrain vesicle effectively repressed Efnb1expression (Fig. 4G-I; data not shown) (n=22/28; n=0/13 for SOHoEnR as control)(Schulte and Cepko, 2000). Similarly, Dbx1 transcripts, which are normally present in the dorsal mesencephalon, were undetectable in Meis2EnR-transfected regions of the neural tube (n=8/9) (Fig. 4J-K″). By contrast, expression of the homeodomain transcription factor Otx2, which specifies the anterior neural plate and its derivatives, and expression of the paired-box transcription factors Pax3 and Pax7, which mark the alar plate along much of the length of the neural tube, were unaffected by Meis2EnR 24 hours following transfection (Fig. 4L-M′; see Fig. S4 in the supplementary material). In addition, although transfection of Meis2-interfering constructs had induced aberrant foliation of the transfected tissue at mid-embryonic stages,this was not accompanied by ectopic induction of hindbrain-associated genes,such as Ath1 (Tnfsf4 - Mouse Genome Informatics), Irx2 or Zic1 (Fig. 4N-Q′; data not shown)(Ben Arie et al., 1997; Lin and Cepko, 1998; Matsumoto et al., 2004). We also analyzed whether loss of Meis2 function induces expression of the diencephalic marker Pax6 in the midbrain. The rationale behind this experiment was the observation that diencephalon and mesencephalon share expression of Otx2 but differ in the expression of Meis2. This raises the possibility that loss of Meis2 in the midbrain might entail diencephalic differentiation as a consequence of the persistent Otx2 expression. However, we did not observe ectopic Pax6expression in the midbrain under these conditions (data not shown).

Otx2 is essential for the development of all anterior brain structures, including the midbrain, and can induce midbrain development when ectopically expressed in rhombomere 1(Katahira et al., 2000). Transfection of Otx2 into rhombomere 1 rapidly induced Meis2expression, placing Meis2 downstream of Otx2 in tectal development (see Fig. S5 in the supplementary material). In addition,co-transfection of Meis2 function-blocking constructs together with Otx2 prevented the metencephalic-to-mesencephalic fate change, which is otherwise triggered by Otx2 misexpression (see Fig. S6 in the supplementary material). We conclude from these results (1) that Meis2 acts upstream of the tectal markers Dbx1 and Efnb1 but downstream of Otx2, and (2) that loss of Meis2 function disrupts normal tectal development, but is not sufficient to induce a mesencephalic-to-metencephalic or a mesencephalic-to-diencephalic transdifferentiation.

Midbrain development is under the control of the MHB organizer, a signaling center located just caudal to the Otx2/Gbx2 junction, which forms during late gastrulation. An interdependent loop of nuclear factors and secreted proteins, including Fgf8/17/18, Wnt1, Pax2/5 and En1/2, shape and maintain the MHB organizer(Wurst and Bally-Cuif, 2001). Ectopic expression of any of these genes in the diencephalic vesicle of chick or zebrafish embryos induces expression of other MHB-associated factors and initiates development of ectopic tectal structures(Araki and Nakamura, 1999; Crossley et al., 1996; Funahashi et al., 1999; Okafuji et al., 1999; Ristoratore et al., 1999). To investigate whether Meis2 is part of this molecular network at the MHB, we targeted Meis2 expression to different regions of the neural tube and assessed the expression of genes that are essential for MHB organizer activity at different times following transfection(Fig. 5; Table 1). Forced Meis2expression in the diencephalic vesicle failed to induce expression of the MHB marker genes En1 or Pax2(Fig. 5A,A′,C,D). We also never detected upregulation of Fgf8, a key molecule of MHB organizer function, at times up to 4 days following Meis2HA transfection(Fig. 5B,E,E′,H,J). Notably, from 48 hours post-transfection onwards, the diencephalic vesicles exhibited ectopic enlargements indicative of the beginnings of a diencephalic-to-mesencephalic transformation, yet these morphological alterations were not accompanied by ectopic Fgf8 expression(Fig. 5E-K). Furthermore, Meis2 transfection into the diencephalic vesicle did not induce Pax3 or Pax7, which also contribute to MHB maintenance and function (see Fig. S4 in the supplementary material)(Matsunaga et al., 2001). Conversely, expression of Fgf8, Otx2, En1, Pax2, Pax3, Pax5, Pax7,Wnt1 and Wnt3a was normal following targeted misexpression of the dominant-negative Meis2EnR in the MHB territory(Table 2; see Fig. S4 in the supplementary material). In contrast to other known factors, which can induce ectopic midbrain development in the diencephalon, Meis2 thus triggers the diencephalic-to-mesencephalic fate change without inducing a secondary MHB organizer.

Our observation that ectopic tectal development upon Meis2transfection was restricted to the anterior neural tube and thus to regions of the neural tube that express Otx2 (see Fig. S2 in the supplementary material) raises the possibility that both proteins cooperate in tectal development. As a first test of this idea, we performed pull-down experiments with a GST-tagged form of Meis2a using protein extracts of the HH15-18 (24- to 36-somite stages) chick mesencephalic alar plate. Meis2-GST, but not GST alone, readily precipitated Otx2 (Fig. 6A). Notably, complex formation also occurred when the experiment was performed in the presence of DNaseI(Fig. 6B). To test which protein domain(s) of Meis2 was involved in complex formation, we performed pull-down experiments with truncated forms of Meis2 fused to GST. Fusion proteins that lack the homeodomain (Meis2ΔHD_[1-190]) failed to precipitate Otx2, whereas precipitation of Otx2 was reduced when GST fusion proteins that lack the N-terminus including the MEINOX domain(Meis2ΔN_[199-400]) were used (compare Fig. 6A with 6C). Meis2-Otx2 interaction therefore involves the three-dimensional structure of the entire Meis2 polypeptide chain. The homeodomain of Meis2 appears to be essential for complex formation, whereas the N-terminal protein, including the MEINOX domain, might serve to stabilize the complex. In support of this, we found that EnR fusion constructs lacking the N-terminus including the MEINOX domain or the homeodomain failed to inhibit Efnb1 expression or tectal development upon transfection into the midbrain vesicle (data not shown).

Next, we investigated whether Meis2-Otx2 interaction was direct or required the presence of additional mesencephalic proteins. To this end, GST pull-down experiments were performed with Meis2-GST and ³⁵S-radiolabeled Otx2 produced in vitro by coupled transcription-translation. Meis2-GST, but not GST alone, precipitated Otx2 (Fig. 6D, left lanes). Again, complex formation was not disrupted by prior treatment with DNaseI (Fig. 6H). Next, truncated forms of Otx2 were tested for their ability to associate with Meis2-GST (Fig. 6D-G). Of these polypeptides, only those that contained 18 amino acids N-terminal of the homeodomain (corresponding to the sequence RKQRRERTTFTRAQLDVL) precipitated with Meis2. Notably, the eh1 (engrailed homology region 1) motif, which mediates binding of Otx2 to the Groucho repressor protein Tle4 (Grg4) (see below), was dispensable for Otx2-Meis2 complex formation (Heimbucher et al.,2007). In summary, these results demonstrate that Meis2 and Otx2 can directly interact in the absence of DNA via an 18 amino acid motif in Otx2.

Neuroepithelial cells of the tectal anlage co-express Meis2 and Otx2(Fig. 7A,A′). To investigate whether Otx2-Meis2-containing protein complexes exist in the tectal anlage in vivo, we performed co-immunoprecipitation experiments from protein extracts of native HH14-18 (20- to 36-somite stage) chick tecta with an antibody directed against the N-terminus of Meis2. As expected for transcriptional regulators, both proteins were predominantly found in the nuclear fractions. Upon immunoprecipitation, Otx2 was enriched in the precipitates when the Meis2-specific antibody was used, but not with an unrelated antibody (Fig. 7B). Both transcription factors are therefore not only able to interact in vitro,but also form higher order protein complexes in the mesencephalic alar plate in vivo.

Because we had previously observed that Meis2 functions as a transcriptional activator in an in vitro reporter assay (see Fig. S1 in the supplementary material), we speculated that binding of Meis2 to Otx2 might augment Otx2 transactivation activity. To test this, we assessed whether Meis2 was able to modulate the activity of an Otx2-dependent reporter in vitro. In this assay, the basal activity of a Gal4-responsive promoter is enhanced after co-transfection of an Otx2-Gal4 fusion protein(Fig. 7C)(Heimbucher et al., 2007). Meis2 alone (not fused to Gal4) was not able to activate the reporter. Yet,contrary to our initial expectation, Meis2 co-transfection with Otx2-Gal4 also did not elevate reporter activity above the level seen after transfection of Otx2-Gal4 alone (Fig. 7C).

Otx2 can function as transcriptional activator or repressor depending on the cellular context. Binding to the Groucho co-repressor protein Tle4 is sufficient to attenuate Otx2 transactivation activity and leads to effective repression of an Otx2-dependent promoter(Heimbucher et al., 2007; Puelles et al., 2004). Binding of Tle4 to Otx2 was mapped to the eh1 domain, an evolutionary conserved motif located C-terminal to the Otx2 homeodomain(Heimbucher et al., 2007). Although the Meis2-binding motif identified in the present study is separated from the eh1 region by the entire homeodomain, the Otx2 homeodomain, like the related Bicoid homeodomain, adopts a three-helical global fold in the non-DNA-bound state, which brings the amino acids N-terminal of helix 1 into close proximity of those C-terminal of helix 3(Baird-Titus et al., 2006)(NCBI Structure MMDB ID #42063). We therefore speculated that Meis2 binding to Otx2 might interfere with the ability of Otx2 to interact with Tle4. To test this, we monitored the activity of the Gal4-dependent reporter in the presence of Otx2-Gal4 alone, following transfection of Otx2-Gal4 together with Tle4 and upon co-transfection of increasing amounts of Meis2(Fig. 7D). As reported previously, Tle4 alone effectively repressed transactivation by Otx2-Gal4(Fig. 7D, bar 2)(Heimbucher et al., 2007). Interestingly, co-transfection of the Meis2-expressing plasmid restored reporter activity in a concentration-dependent manner(Fig. 7D, bars 3 and 4). To verify that this effect required Meis2 binding to Otx2, a Gal4 fusion to a truncated form of Otx2, Otx2Δ1-100-Gal4, was used. This protein,although lacking the entire N-terminus including the 18 amino acid Meis2-binding motif, can still activate the Gal4-dependent reporter and interact with Tle4 (Fig. 7E,bars 1 and 2) (Heimbucher et al.,2007). In contrast to full-length Otx2-Gal4, co-transfection of Meis2 with Otx2Δ1-100-Gal4 did not restore reporter activity,underscoring the importance of the 18 amino acid motif(Fig. 7E, bars 3 and 4). Direct interaction with Otx2 is therefore indispensable for the ability of Meis2 to interfere with Otx2-Tle4-mediated repression. In support of this, we found that Otx2 lacking the N-terminus including the 18 amino acid motif was ineffective in inducing ectopic midbrain structures in the metencephalic vesicle (data not shown).

To test whether Meis2 is also capable of binding to Tle4, we performed pull-down experiments with Meis2-GST together with ³⁵S-radiolabeled Tle4 and compared the efficiency of the precipitation with that of Meis2-GST together with ³⁵S-radiolabeled Otx2 under identical experimental conditions (Fig. 7F). Relative to the amount of radiolabeled protein added into the reactions (input, `in'),only a minor fraction of Tle4 could be detected in the precipitates, whereas Otx2 was highly enriched by the Meis2-GST fusion protein (compare `in' and PD in Fig. 7F, top and bottom panels). Meis2 thus binds robustly to Otx2, but only weakly, if at all, to Tle4. It is tempting to speculate that the weak binding of Meis2 to Tle4 might reflect a transient contact between the two proteins, which might take place when Meis2 releases Otx2 from Tle4-mediated repression.

Here we show that the TALE-homeodomain protein Meis2 is a key regulator of tectal development. In contrast to other known genes involved in tectal development, Meis2 initiates tectal fate specification without inducing a secondary MHB organizer. Instead, Meis2 binds to Otx2 in the absence of DNA, competes with the co-repressor Tle4 for binding to Otx2 and thereby restores Otx2 transcriptional activator function. As discussed below,these results suggest a model in which the balance between a co-repressor and a co-activator, which compete for binding to Otx2 in the mesencephalic vesicle, provides spatial and temporal control over the onset of tectal development. Our data thus argue for a novel, potentially DNA-independent function of TALE-homeodomain proteins: the controlled assembly and disassembly of transcription regulator complexes.

Tectum development is induced when an ectopic Fgf8 source is generated in the prosencephalon through transplantation of an ectopic MHB organizer or implantation of Fgf8-releasing beads into the lateral wall of the diencephalon (Crossley et al.,1996; Martinez et al.,1991). In addition to Fgf8, several transcription factors can trigger tectal development upon misexpression, including Otx2, Pax2/5, En1/2 and Pax3/7 (Araki and Nakamura,1999; Funahashi et al.,1999; Katahira et al.,2000; Matsunaga et al.,2001; Okafuji et al.,1999; Ristoratore et al.,1999). Unlike Meis2, expression of these proteins is not specific for the tectal anlage. Moreover, each of these proteins participates in the interdependent, positive maintenance loop at the MHB organizer and,consequently, induces ectopic expression of MHB marker genes, including Fgf8, when misexpressed. These molecules therefore evoke tectal development indirectly through formation of an ectopic MHB organizer. Meis2, by contrast, is unique as it can initiate tectal development without participating in MHB organizer function or maintenance. As we have recently shown, endogenous Meis2 expression is repressed when metencephalic development is experimentally induced through activation of the Ras-MAP kinase pathway in the mesencephalon and is upregulated concomitantly to the metencephalic-to-mesencephalic fate change that occurs when Ras-MAP kinase signaling is blocked in rhombomere 1(Vennemann et al., 2008). Meis2 expression must therefore be directly or indirectly under control of the MHB organizer. Notably, a single, transient transfection of Meis2 in the diencephalic alar plate at the 10- to 11-somite stage was sufficient to initiate long-term expression of endogenous Meis2in transfected cells (Fig. 5;see Fig. S3 in the supplementary material). Meis2, once induced, can therefore stabilize its own expression. Together, these results suggest a model in which regulation of tectal development by signals from the MHB is mediated via induction and subsequent auto-maintenance of Meis2expression.

Meis family proteins act as co-factors of other transcriptional regulators(Moens and Selleri, 2006). To date, Meis-interacting proteins have been isolated from non-neuronal tissue and the posterior hindbrain, yet Meis co-factors in the developing anterior brain have remained elusive. We performed GST pull-down experiments from tectal extracts or with in vitro translated proteins as well as co-immunoprecipitation experiments with native tectal proteins to demonstrate direct interaction of Meis2 and Otx2 during early midbrain development. Using deletion constructs of Otx2, we find that complex formation requires a short motif N-terminal of the Otx2 homeodomain. The region of the Otx2 polypeptide chain that contacts Meis2 thus differs from the tryptophan-containing hexapeptide that mediates cooperative DNA binding of Hox or myogenic bHLH proteins with TALE-homeodomain proteins(Knoepfler et al., 1999; Lu and Kamps, 1996). Meis family proteins can therefore interact with different protein motifs present in a variety of transcription factors.

Employing an Otx2-dependent reporter assay, we provide evidence that Meis2 competes with the co-repressor Tle4 for binding to Otx2. Tle4expression begins in the anterior primitive streak (thus preceding that of Meis2), is later strong in the anterior neural tube and decreases after the 20- to 25-somite stage (Sugiyama et al., 2000; Van Hateren et al., 2005). Tle4 binding to Otx2 was previously shown be required for the ability of Otx2 to repress Fgf8 anterior to the MHB, an important step in the formation and stabilization of the MHB organizer(Heimbucher et al., 2007). Overexpression of Tle4 in the mesencephalic vesicle, in turn,disrupts normal development and lamination of the optic tecta(Sugiyama et al., 2000; Sugiyama and Nakamura, 2003). Together with the results presented here, these data might allow us to reconstruct the probable temporal sequence of tectal fate specification in the embryo. In the anterior neural plate and anterior neural tube at early somite stages, Tle4 is co-expressed with Otx2 but Meis2 is missing. In the absence of Meis2, Otx2 and Tle4 can interact, prevent precocious tectal differentiation and inhibit Fgf8 expression anterior to the organizer, which stabilizes the MHB signaling center. Meis2 expression in the mesencephalic alar plate begins at HH11-12(13-16 somites) and is strong from the 20- to 22-somite stage onwards, at which time the MHB organizer is established(Fig. 1). Meis2 competes with Tle4 for binding to Otx2 in the tectal anlage, releases Otx2 from Groucho-mediated repression and thereby allows tectal development to commence.

If correct, two predictions can be drawn from this model. First, loss of Tle4 from the diencephalic vesicle (where Tle4 is co-expressed with Otx2 at early somite stages) should lead to derepression of tectal genes. Second, precocious and ectopic expression of Meis2 in the MHB territory may destabilize the Fgf8expression domain through premature restoration of Otx2 transcriptional activator function. Indeed, as previously demonstrated, transfection of a putative dominant-negative form of Tle4 - a truncation comprising only the first 203 amino acids of the protein - into the diencephalic vesicle causes widespread ectopic induction of En2 transcripts(Sugiyama et al., 2000). In addition, when we ectopically introduced Meis2 into the MHB region at the 4- to 6-somite stage, small ectopic patches of Fgf8 transcripts anterior to the normal Fgf8 expression domain at the MHB were visible(see Fig. S8 in the supplementary material). These ectopic patches of Fgf8 expression might correspond to cells that have escaped Fgf8 downregulation by Otx2-Tle4 during the period of MHB organizer formation owing to the precocious inactivation of the Otx2-Tle4 complex by Meis2HA.

Meis2 had to be transfected in excess to Tle4 in order to restore Otx2 transactivation in the Otx2-dependent reporter assays. This observation is consistent with the fact that between the 24- and 44-somite stages, Meis2 transcripts are abundant in the dorsal midbrain, whereas Tle4 expression is barely detectable (see Fig. S7 in the supplementary material). Tle4 thus appears to bind to Otx2 with higher affinity than does Meis2, which might allow for tight control over the tectum-inducing activity of Otx2. Recently, the spatial-temporal windows of Otx2 control over head, brain and body development were defined by Tamoxifen-induced deletion of Otx2(Fossat et al., 2006). Interestingly, Otx2 deletion at E10.5-12.5 resulted in a mesencephalic-to-metencephalic fate change without shifting the molecular MHB. Hence, the interaction of Otx2 with Meis2 in the tectal anlage reported here occurs at a similar developmental stage to that at which Otx2 is required for mesencephalic fate determination but not MHB organizer positioning.

Possible targets of Otx2/Meis2 include Efnb1 and Dbx1, both of which carry several potential consensus Bicoid- and Meis-binding sites upstream of their transcriptional start sites. Direct regulation of a midbrain-specific regulatory element of the EphA8 gene by Meis2 has also been demonstrated in mice (Shim et al.,2007). However, because Meis2 binding to Otx2 does not require either protein to be bound to DNA, Meis2-Otx2 interaction and restoration of Otx2 transcriptional activator function might in fact take place before both proteins have contacted the regulatory elements of downstream genes. Regulation of gene expression by putative transcription factors independent of DNA binding is not unprecedented. For instance, several Hox proteins can modulate gene expression by inhibiting the activity of CBP histone acetyltransferases (HATs) without forming DNA-binding complexes with CBP HAT(Shen et al., 2001). Meis family members might therefore affect gene expression by multiple,DNA-dependent and -independent mechanisms. This view is supported by the fact that despite the identification of Meis proteins as transcriptional co-factors, few direct target genes of these proteins have been reported to date.

In summary, the results reported here strongly suggest that Tle4 and Meis2 compete for binding to Otx2 in the mesencephalic vesicle and that the balance between these proteins provides spatial and temporal control over the onset of tectal differentiation. Formation of spatially and temporally distinct higher order protein complexes involving Meis proteins and known regulators of neural patterning or fate determination might serve as a simple, yet versatile,mechanism to subdivide broad territories into smaller functional units during brain development.

We are grateful to A. Buchberg, M. Busslinger, C. Cepko, T. Czerny, D. O'Leary, C. Logan, H. Rohrer, V. Lee and C. Tabin for reagents; to S. Nickel and A. Poplawski for experimental help; and to C. Ziegler for excellent technical assistance. Z.A. is a recipient of a pre-doctoral fellowship from the Studienstiftung des Deutschen Volkes.