OTD/OTX2 functional equivalence depends on 5′ and 3′ UTR-mediated control of Otx2 mRNA for nucleo-cytoplasmic export and epiblast-restricted translation

Early brain development in higher vertebrates is a complex process characterised by the induction and maintenance of the anterior neuroectoderm, which subsequently becomes regionalised into forebrain, midbrain and hindbrain (Beddington and Robertson, 1999). Further differentiation of these territories leads to the generation of neuronal populations with a restricted identity and specific functional properties (Lumsden, 1990; Lumsden and Krumlauf, 1996; Rubenstein et al., 1998).

In the last decade, a considerable amount of data has been collected on the role of developmental control of genes involved in brain morphogenesis in mammals. Most of them are the vertebrate homologues of Drosophila genes that encode transcription factors or signalling molecules (Lemaire and Kodjabachian, 1996; Holland, 1999; Rubenstein et al., 1998). Among these is the orthodenticle gene family, which includes the Drosophila orthodenticle (otd) and the murine Otx1 and Otx2 genes (Simeone et al., 1992; Simeone et al., 1993; Finkelstein and Boncinelli, 1994).

Previous studies have indicated that the Otx genes play an important role in brain development [see reviews by Simeone and Acampora (Simeone, 1998; Acampora and Simeone, 1999; Simeone 2000; Acampora et al., 2001)]. These studies have established that Otx2 is required in the VE for early induction of anterior neural patterning and proper gastrulation, and subsequently, in the epiblast-derived tissues such as axial mesendoderm (ame) and anterior neuroectoderm for maintenance of forebrain and midbrain regional identities (Acampora et al., 1995; Acampora et al., 1998b; Rhinn et al., 1998; Acampora and Simeone, 1999; Beddington and Robertson, 1999). In particular, the analysis of embryos replacing Otx2 with the human Otx1 (hOtx1) cDNA has also revealed that while the hOtx1 mRNA is transcribed in both the VE and epiblast, the hOTX1 protein is synthesised only in the VE (Acampora et al., 1998b). This unexpected phenomenon has suggested the possibility that a differential post-transcriptional control may exist between VE and epiblast and that the Otx2 replaced region, including 213 bp of the 5′ untranslated region (UTR), exons, introns and the 3′ UTR, may contain regulatory elements required for epiblast-restricted translation of Otx2 mRNA. In terms of brain development this potential control of Otx2 mRNA translation might represent a crucial aspect since lack or reduction of OTX2 protein is invariably associated with lack or reduction of the forebrain and midbrain.

In this context, we have very recently reported that embryos carrying a 300 bp insertion of exogenous DNA from the λ phage (Otx2^λ) into the Otx2 3′ UTR, exhibit severe brain abnormalities and drastic reduction of OTX2 protein for an impairment of Otx2^λ mRNA to form efficient polyribosome complexes (Pilo Boyl et al., 2001).

However, although OTX1 and OTX2 appear to be equivalent in the VE and OTD/OTX1 properties largely overlap, it could not be predicted whether OTD is also equivalent to OTX2 in instructing proper brain morphogenesis.

Understanding how the genetic mechanisms that control brain morphogenesis have evolved represents a central aspect of comparative developmental biology. Despite the enormous morphological diversity between invertebrates and vertebrates, striking examples of functional conservation of genetic programmes for morphogenesis are found in these animal groups. Among these are the invertebrate HOM-C and vertebrate Hox genes which control the specification of axial patterning (Lewis, 1978; Krumlauf, 1994; Lumsden and Krumlauf, 1996), the short gastrulation/Chordin and decapentaplegic/Bmp4 genes, which control dorsoventral (DV) patterning (De Robertis and Sasai, 1996) and the Drosophila eyeless (ey) and vertebrate Pax6 genes, which control eye morphogenesis (Callaerts et al., 1997).

Recently, functional conservation among members of the otd/Otx group has been demonstrated in cross-phylum genetic rescue experiments (Acampora et al., 1998a; Leuzinger et al., 1998; Nagao et al., 1998; Sharman and Brand, 1998). Nevertheless, homeodomain-restricted homology between OTD and OTX2 proteins as well as the huge diversity existing between mouse and Drosophila early head morphogenesis suggest that otd should be unable to share functional equivalence with Otx2 in VE (specification of the early anterior patterning and proper gastrulation) and neuroectoderm (maintenance of forebrain and midbrain identities).

In order to assess whether OTD is equivalent to OTX2 and, importantly, whether translation of otd mRNA requires Otx2 post-transcriptional control in epiblast and neuroectoderm, we generated a first mouse model (otd²) where an otd cDNA replaces an Otx2 genomic region including 213 bp of the 5′ UTR, coding sequences, introns and the entire 3′ UTR. In a second mutant (otd^2FL), only the Otx2 coding sequence and introns were replaced by the otd coding sequence. While the otd² and otd^2FL expression was correctly driven by Otx2 transcriptional control, mRNA translation in epiblast and neuroectoderm occurred only in otd^2FL mutants, being otd² mRNA translation restricted only to VE. A third mutant (Otx2^2c) carrying an Otx2 cDNA in place of the Otx2 coding sequence and introns, was generated to compare OTD and OTX2 functions. Phenotypic analysis of otd², otd^2FL and Otx2^2c homozygous mutants indicated that OTD fulfilled OTX2 functions required in VE and neuroectoderm for murine brain development when the two genes undergo the same genetic modifications. A detailed molecular analysis has indicated that the Otx2 5′and 3′ UTRs deleted in otd² mutant were responsible for a severe impairment of both nucleo-cytoplasmic export and epiblast-restricted translation of the otd² mRNA.

Importantly, these molecular abnormalities were completely rescued in otd^2FL and Otx2^2c mutants and confirmed in hOtx1² embryos. In this study, we have shown that the otd mRNA is translated in epiblast and neuroectoderm only if the Otx2 post-transcriptional control is provided, and then in turn, the OTD protein can perform OTX2 functions required for murine head morphogenesis. Together with previous studies, this work provides more conclusive evidence indicating that this control may play a central role in development and evolution of the vertebrate head.

The gene replacement vectors were generated from the same plasmid (pGN31) used to produce Otx2^–/– mice (Acampora et al., 1995). For the otd² construct, a Bsu36I/XmnI fragment of the otd cDNA containing 56 bp of its 5′ UTR, the entire coding sequence and 450 bp of its 3′UTR was cloned in place of the lacZ gene in the former targeting vector. In this molecule, the otd cDNA was fused to the Otx2 5′UTR lacking 213 bp in front of the methionine.

For the otd^2FL allele, a chimeric fragment containing the 213 bp of the Otx2 5′UTR (missing in the otd² locus), the otd coding sequence and the Otx2 3′UTR from the stop codon up to a sequence 57 bp upstream of the Otx2 polyadenylation signal, was generated by PCR, sequenced and cloned in the same targeting vector. The control Otx2^2c locus was generated using the same criteria as for otd^2FL, the Otx2 cDNA coding sequence being the only difference. To facilitate both genotyping and molecular analysis, the Otx2 cDNA was a chimeric fragment made up of human sequences from the methionine to the NsiI site in exon 3 (Fig. 6A). Downstream of their 3′ UTR, all of the three molecules carried a SV40 polyadenylation signal to ensure transcription termination. The three targeting vectors were all electroporated into HM-1 embryonic stem cells. Homologous recombinant clones were identified by PCR using the same primers and conditions previously described (Acampora et al., 1995) (black arrows in Fig. 1A and Fig. 6A) and confirmed by hybridising HindIII digested genomic DNA with probes c and f (Fig. 1A,B, Fig. 6A,B and data not shown). Once established, all the mutant lines were kept in the B6D2 F1 genetic background.

Genotyping was performed by PCR using two primers specific for the wild-type allele that were common for all the mouse lines (black arrowheads in Fig. 1A and 6A) (sense primer, GTGACTGAGAAACTGCTCCC; antisense primer, GTGTCTACATCTGCCCTACC) and three different pairs of primers for each mutated allele, otd² (open arrowheads in Fig. 1A) (sense primer, ATCAAGACGCACCACAGTTCCT; antisense primer, TCCTTTAGCTGATCATAGGGCG), otd^2FL (open arrows in Fig. 1A) (sense primer, CTTCTGGCACAATCAGTACCAGC; antisense primer, TGCTGGTTGATGGACCCTTC) and Otx2^2c (open arrowheads in Fig. 6A) (sense primer, TCACTCGGGCGCAGCTAGATGTG; antisense primer, AGAGGAGGTGGACAAGGGATCT). These primers amplify a 223 bp wild-type fragment and 444, 429 and 371 bp long fragments corresponding to the otd², otd^2FL and Otx2^2c mutant alleles, respectively (Fig. 1C and Fig. 6C).

RNAse protection was performed as previously described (Simeone et al., 1993) using as probes, three fragments generated by PCR in corresponding regions (hatched boxes b, d in Fig. 1A and g in Fig. 6A). Phosphorimager scanning was performed to quantify RNA levels.

Embryos were processed for immunohistochemistry with αOTD (1:1000) and αOTX2 (1:1500) polyclonal antisera as previously described (Acampora et al., 1998b). The αOTD polyclonal antiserum was generated following the same procedure previously used for the αOTX2 antiserum (Briata et al., 1996). Total extracts were prepared from three independent pools of 10.5 and 7.5 d.p.c. embryos for each mutant, processed for standard western blot assay and probed with αOTD (1:5000) and αOTX2 antiserum (1:5000). Films were processed for densitometric scanning.

In situ hybridisation experiments on sections and whole embryos were performed as previously described (Hogan et al., 1994; Simeone, 1999).

Wnt1, Tbr1, Bf1, Fgf8, Six3, Gbx2, Nog, Chd, cer-l, Hesx1 and Lim1 probes are the same previously described (Acampora et al., 1997; Acampora et al., 1998b; Pilo Boyl et al., 2001). The otd probe (probe e in Fig. 1A and Fig. 6A) is a PCR fragment spanning the last 800 bp of the otd coding region. The Otx2 probe for in situ hybridisation on sections corresponds to probe a in Fig. 1A. The Otx2^2c-specific probe for in situ hybridisation experiments on sections corresponds to probe h in Fig. 6A.

Sucrose gradients were performed essentially as described (Loreni and Amaldi, 1992; Pilo Boyl et al., 2001). Nuclear and cytoplasmic mRNAs were purified as previously described (Pilo Boyl et al., 2001). Stability of Otx2, otd², otd^2FL and Otx2^2c mRNAs was determined by administering 15 μg/ml of actinomycin D to the heterozygous ES cell clones followed by RNA extraction at the time indicated in Fig. 8F.

In order to assess whether OTD is equivalent to OTX2 and whether this equivalence is subordinated to the Otx2 post-transcriptional control, we generated two different mouse models replacing Otx2 with otd.

In the first mutant locus (otd²) an otd cDNA containing 56 bp of the 5′ UTR, the full coding sequence, and 450 bp of the 3′ UTR region was introduced into an Otx2 disrupted locus and fused at its 5′ end to the Otx2 5′ UTR and at its 3′ end to the SV40 poly(A) cassette of the pGN targeting vector (Acampora et al., 1995) (Fig. 1A). Otx2 deleted or misplaced sequences included 213 bp upstream of the start methionine in the 5′ UTR, exons, introns and 3′ UTR (Fig. 1A).

In the second mutant locus (otd^2FL) the entire Otx2-5′ UTR and the Otx2-3′ UTR, excluding only the last 57 bp upstream of the polyadenylation signal, were fused to the methionine and stop codon of the otd coding sequence, respectively (Fig. 1A). The resulting heterozygotes (otd²/+ and otd^2FL/+) were healthy and fertile. Expression of the two mutated alleles was monitored in heterozygous embryos at 6.5 and 10.5 days post coitum (d.p.c.) by using Otx2 and otd allele-specific probes (probes a and e in Fig. 1A). At 10.5 d.p.c. the Otx2 (Fig. 1D,E), otd² (Fig. 1F) and otd^2FL (Fig. 1G) transcripts were properly co-expressed along the forebrain and midbrain of otd²/+ and otd^2FL/+ embryos, even though the amount of otd² transcripts appeared heavily reduced in otd²/+ embryos.

Similar experiments performed at 6.5 d.p.c., revealed that during gastrulation the amount of otd² mRNA was higher (Fig. 3 and Fig. 8).

All otd²/otd² embryos at 10.5 d.p.c. (23 out of 23) had a striking headless phenotype (Table 1; Fig. 2B); approximately one third of them (7 out of 23) also had a reduced body size. Moreover in otd²/otd² mutants, the maxillary process, the mandibular arch and their derivatives were strongly impaired or absent (Fig. 2B,G).

In contrast, otd^2FL/otd^2FL mutants exhibited an evident morphological rescue of the anterior defects observed in otd²/otd² embryos, even though, head and brain development still appeared compromised (Table 1; Fig. 2C-E,H-J). Nevertheless, phenotype analysis (see also below) indicated that most of the otd^2FL /otd^2FL embryos (≥80%) exhibited forebrain and midbrain identities as well as ocular and olfactory sense organs (moderate and mild phenotypes in Table 1). Moderate phenotype was characterised by exencephaly and forebrain reduction (Fig. 2D,H,I); mild phenotype by a variable anterior shift of the presumptive midbrain-hindbrain boundary (MHB), and a well developed forebrain and anterior sense organs (Fig. 2E,J). Only a residual fraction (approx. 15%) of otd^2FL/otd^2FL embryos exhibited a severe phenotype (Table 1; Fig. 2C). However, also these embryos showed midbrain-specific and, occasionally, forebrain-specific gene expression (data not shown). Only 1 out of the 28 otd^2FL/otd^2FL embryos scored, displayed a headless phenotype (Table 1).

These data suggest that, despite the huge diversity between the Drosophila OTD and the murine OTX2, the fly protein was able to compensate OTX2 functions that are required for head development.

The phenotype of otd²/otd² embryos revealed striking similarity with that of mice in which exactly the same Otx2 region had been replaced with a human Otx1 cDNA (hOtx1²) (Acampora et al., 1998b). In this mutant, the hOtx1 mRNA was transcribed in the VE and epiblast but translated only in the VE. Therefore transcription and translation was studied in otd² and otd^2FL mutants. A detailed analysis of otd² and otd^2FL mRNAs and proteins was performed at early streak (6.5 d.p.c.) and late streak (7.75 d.p.c.) stages in wild-type, otd²/+, otd²/otd², otd^2FL/+ and otd^2FL/otd^2FL embryos (Fig. 3). Genotypes were determined by hybridising adjacent sections with Otx2 and otd allele-specific probes (probe a and e in Fig. 1A, respectively). Two additional series of adjacent sections were assayed for the presence of OTX2 and OTD proteins with αOTX2 and αOTD antibodies. No staining was detected with either of the two antibodies on Otx2^–/– embryos (data not shown) (Acampora et al., 1998b).

Otx2 transcripts and protein distribution fully colocalised in the epiblast and VE of wild-type (Fig. 3A,A’), otd²/+ (Fig. 3C,C’) and otd^2FL/+ (Fig. 3G,G’) embryos at 6.5 d.p.c. and similar results were obtained at 7.75 d.p.c. (data not shown).

In otd²/+ and otd²/otd² embryos, otd² transcripts were detected in VE and epiblast at 6.5 d.p.c. (Fig. 3D,F) and in the anterior neuroectoderm and ame at 7.75 d.p.c. (Fig. 3L and data not shown). Also, as previously mentioned, their amount did not differ greatly from that of the Otx2 mRNA. In contrast with the otd² mRNA distribution in both VE and epiblast, at 6.5 d.p.c. the OTD protein appeared restricted to the VE of otd²/+ (Fig. 3D’) and otd²/otd² (Fig. 3F’) embryos and was not evident in the epiblast. Also at 7.75 d.p.c., OTD was not detected (Fig. 3L’).

Importantly, in otd^2FL/+ and otd^2FL/otd^2FL embryos, both otd^2FL mRNA and protein colocalised in the VE and epiblast at 6.5 d.p.c. (Fig. 3H,H’,J,J’) as well as in the neuroectoderm and ame at 7.75 d.p.c. (Fig. 3N,N’ and data not shown).

These data provide in vivo evidence that Otx2 5′ and 3′ UTRs are required for translation of the Otx2 mRNA in epiblast and derived tissues. Moreover, the finding that the otd² mRNA is not translated in the epiblast of otd²/+ embryos suggests that this control does not require OTX2. Nevertheless, on the basis of these data it cannot be excluded whether this control requires both Otx2 coding sequence and 5′ and 3′ UTRs in a cis configuration.

Previous data indicated that the earliest defects of Otx2^–/– embryos are related to an OTX2 requirement in the VE that is abnormally located at the distal tip of the embryo as revealed by the expression of Lim1, cerberus-like (cer-l), Hesx1 (Acampora et al., 1995; Acampora et al., 1998b; Ang et al., 1996). To assess whether the OTD protein could rescue these impairments, otd²/otd² embryos were studied at 6.5 and 7.5 d.p.c. In otd²/otd² embryos, otd² mRNA was detected in the VE and epiblast (Fig. 4B) and Lim1, cer-l and Hesx1 were correctly expressed (Fig. 4D,F,H).

Therefore, VE abnormalities and molecular impairments detected in Otx2^–/– embryos were rescued by the VE-restricted translation of the otd² mRNA. This suggests that OTD and OTX2 proteins exhibit functional equivalence in the VE.

At late streak-headfold stage, Otx2^–/– embryos were severely impaired in morphology and expression of early forebrain, midbrain and hindbrain markers (Acampora et al., 1995; Acampora et al., 1998b; Matsuo et al., 1995; Ang et al., 1996; Rhinn et al., 1998). In otd²/otd² embryos, otd² transcripts were restricted to the presumptive forebrain and midbrain (Fig. 4J), Six3 was expressed in the rostralmost neuroectoderm (Fig. 4L) and Gbx2 throughout the presumptive hindbrain (Fig. 4N).

Otx2 null embryos also showed severe impairments of the gastrulation process. Accordingly, Noggin (Nog) and Chordin (Chd) transcripts were either absent or only barely detected in disorganised cells (Acampora et al., 1995; Acampora et al., 1998b; Ang et al., 1996). As revealed by Nog (Fig. 4P) and Chd (Fig. 4R) expression, these defects were recovered in otd²/otd² embryos. This analysis was also performed in 6.5 and 7.5 d.p.c. otd^2FL/otd^2FL embryos which, as expected, did not show any abnormalities (data not shown).

Thus, the morphological defects and the molecular impairments that affect VE, rostral neuroectoderm, primitive streak, as well as the node and its derivatives of Otx2^–/– embryos are rescued by VE-restricted translation of otd² mRNA. It is noteworthy that the additional presence of OTD protein in the epiblast of otd^2FL/otd^2FL embryos does not improve the rescue observed when the OTD protein is restricted to the VE. This strengthens the idea that in Otx2^–/– embryos, Otx2-mediated impairments of the early anterior patterning and primitive streak are determined in the VE.

We previously showed that hOtx1²/hOtx1² embryos failed to maintain anterior patterning and exhibited headless phenotype (Acampora et al., 1998b).

Embryo morphology and lack of OTD protein in epiblast and neuroectoderm are very similar in otd²/otd² and hOtx1²/hOtx1² embryos. On this basis, maintenance of forebrain and midbrain identity should be lost in otd²/otd² mutants. To assess this issue, the expression of a number of regionally restricted markers such as Fgf8, Gbx2, Wnt1, Bf1 and Tbrain1 (Tbr1) was analysed together with otd at 8.5 and 10.5 d.p.c.

Compared to wild-type embryos (Fig. 5A-J), in otd²/otd² embryos at 8.5 and 10.5 d.p.c., otd² transcripts were undetectable along the neuroectoderm (Fig. 5A’); Fgf8 (Fig. 5B’,G’), Gbx2 (Fig. 5C’) and Wnt1 (Fig. 5D’,H’) expression was markedly displaced anteriorly; Bf1 (Fig. 5E’,I’) and Tbr1 (Fig. 5J’) were no longer detected in presumptive forebrain. In otd^2FL/otd^2FL embryos recovery of forebrain and midbrain regional identity was observed.

Indeed, compared to otd²/otd² embryos, otd^2FL expression was detected and maintained along the anterior neural plate (Fig. 5A’’,F’’); Fgf8 (Fig. 5B’’) and Gbx2 (Fig. 5C’’) expression domains were shifted posteriorly almost at their normal position; Wnt1 (Fig. 5D’’) transcripts were detected in a broader domain corresponding to the presumptive midbrain; Bf1 (Fig. 5E’’,I’’) was abundant in the rostral neuroectoderm, and Tbr1 (arrow in Fig. 5J’’), which labelled post-mitotic neuroblasts in the dorsal telencephalon, was properly expressed.

However, even though the forebrain and midbrain rescue determined by the fly OTD protein was evident in most of otd^2FL/otd^2FL mutants, it was never complete. Indeed, at 10.5 d.p.c. the midbrain territory defined by otd^2FL and Wnt1 expression domains (Fig. 5F’’,H’’) was smaller and this reduction was balanced by an enlargement of the rostalmost hindbrain. Moreover, Fgf8 (Fig. 5G’’) expression was abnormally expanded within the otd^2FL and Wnt1 domains.

Finally, it is noteworthy that all the embryos with a moderate phenotype (Table 1 and Fig. 2D) recovered forebrain and midbrain expression of Bf1, Tbr1, Wnt1 and otd^2FL (data not shown) and that 3 out of 6 embryos with a severe phenotype (Table 1 and Fig. 2C) exhibited spotted expression of Bf1 (data not shown).

Therefore, this analysis indicates that (i) the Drosophila OTD and the murine OTX2 proteins share functional equivalence also in the neural plate where OTD is able to maintain the anterior patterning and direct forebrain and midbrain regionalisation; and (ii) OTD/OTX2 equivalence in maintenance of forebrain and midbrain depends on the translation of the otd^2FL mRNA in the epiblast and neural plate and this process requires Otx2 5′ and 3′ UTRs.

The phenotypic features of otd^2FL/otd^2FL embryos are reminiscent of those previously described in mutant embryos with reduced level of OTX2 or both OTX2 and OTX1 proteins (Acampora et al., 1997; Suda et al., 1997; Pilo Boyl et al., 2001). This suggests that the level of OTD protein is not sufficient to fulfil all of the OTX2 requirements. Alternatively, it may suggest that OTD and OTX2 proteins share partial functional equivalence.

To address this issue, we have generated a third mouse model replacing Otx2 with an Otx2 cDNA (Otx2^2c) with the same strategy used to generate the otd^2FL recombined locus (compare Fig. 6A with Fig. 1A). The Otx2^2c recombined locus was therefore identical to the otd^2FL locus, except for the Otx2 coding sequence and, thereby, it should represent the most appropriate control of the otd^2FL mutated locus.

Otx2^2c/Otx2^2c and otd^2FL/otd^2FL embryos exhibited striking phenotypic similarities. Indeed, also Otx2^2c/Otx2^2c embryos were classified in three groups on the basis of head abnormalities (Table 2) (Fig. 6E-G,I-K). At 7.5 d.p.c. Otx2^2c mRNA and protein colocalised in the anterior neuroectoderm and ame of Otx2^2c/Otx2^2c embryos (data not shown) and Six3, Chd and Gbx2 were correctly expressed (data not shown). Maintenance of the anterior patterning was assessed at 8.5 and 10.5 d.p.c. by analysing the expression of Otx2^2c, Fgf8, Gbx2 and Bf1 as well as the distribution of the OTX2 protein. At 8.5 d.p.c., Otx2^2c/Otx2^2c embryos showed that Otx2^2c was transcribed along the presumptive forebrain and midbrain (Fig. 7A’), and Fgf8 (Fig. 7B’) and Gbx2 (Fig. 7C’) expression domains appeared slightly expanded anteriorly. At 10.5 d.p.c. in Otx2^2c/Otx2^2c embryos, Otx2^2c transcripts (data not shown) and protein (Fig. 7E’) colocalised and were less abundant in the posterior midbrain; Fgf8 (Fig. 7F’) was broadly expressed up to the presumptive position of the telen-diencephalic boundary; Gbx2 transcripts (Fig. 7G’) were slightly displaced anteriorly and a robust expression of Bf1 (Fig. 7D’,H’) was detected in the forebrain at 8.5 and 10.5 d.p.c.

Therefore, these findings and those reported for otd^2FL/otd^2FL embryos suggest that OTD and OTX2 functional properties may be fully equivalent. However, from our data it cannot be excluded that OTD is unable to fully rescue the Otx2 mutant phenotype when only the Otx2 coding sequence is replaced.

In order to determine the molecular events that were impaired in otd²/otd², otd^2FL/otd^2FL and Otx2^2c/Otx2^2c embryos as well as those that were rescued by the presence of the Otx2 5′ and 3′ UTRs, a detailed analysis was performed to assess protein levels and to determine whether otd², otd^2FL and Otx2^2c mRNAs were efficiently transcribed, processed and translated. Protein levels were assessed in western blot assays on three independent pools of wt, Otx2^+/–, otd²/+, otd^2FL/+ and Otx2^2c/Otx2^2c embryos at 7.5 and 10.5 d.p.c. At 7.5 d.p.c. OTD protein was detected only in otd^2FL/+ embryos (Fig. 8A) and its quantification indicated that, it was only 27% that of OTX2. A very similar result was obtained at 10.5 d.p.c. (data not shown). However, we were aware that the accuracy of this quantitative analysis could be affected by different affinity and specificity of the two antibodies (αOTD and αOTX2), mainly in denaturing conditions. Indeed, based on αOTD and αOTX2 calibration experiments performed on HeLa cell extracts transfected with Otx2- and otd-expressing vectors and normalised for Otx2 and otd mRNAs, we strongly suspected that the OTD level detected (27%) was underestimated (data not shown). This was also supported by the quantitative analysis performed in Otx2^2c/Otx2^2c embryos at 7.5 d.p.c. (Fig. 8B), which showed that the OTX2 level was 37% and 85% of that detected in wild-type and Otx2^+/– embryos, respectively. Hence, this analysis confirms that head abnormalities observed in otd^2FL and Otx2^2c homozygous embryos are probably the consequence of a reduction in the amount of OTD and OTX2.

Next, we analysed whether otd², otd^2FL and Otx2^2c mRNAs were efficiently transcribed and processed. In order to compare wild-type and mutant mRNAs in the same cells, this analysis was performed in heterozygous embryos.

In these experiments, allele-specific probes (Fig. 1A, Fig. 6A) designed in the same region and of similar length were employed in order to minimise experimental heterogeneity. Compared to the Otx2 mRNA, the cytoplasmic otd² mRNA detected in 10.5 d.p.c. embryos was considerably diminished (14%), while at 7.5 d.p.c. it was remarkably higher (32%) (Fig. 8E). In contrast, the amount of the nuclear otd² mRNA was abnormally high, being up to 50% that of the Otx2 mRNA at 10.5 d.p.c and 90% at 7.5 d.p.c. (Fig. 8E).

The same analysis was performed in otd^2FL/+ and Otx2^2c/+ embryos at 10.5 d.p.c. In these mutants cytoplasmic and nuclear mRNA levels decreased by approximately 60% and 50%, respectively (Fig. 8E).

Hence, as nuclear and cytoplasmic mRNAs underwent a similar reduction, nucleo-cytoplasmic export and processing should be unaffected in these embryos. In contrast, the remarkable accumulation of the nuclear otd² mRNA suggests that nucleo-cytoplasmic export is heavily affected. This implies that sequences within the Otx2 5′ and 3′ UTRs play a relevant role in this process.

In this context, the quantitative reduction observed in otd^2FL/+ and Otx2^2c/+ embryos should be ascribed to perturbation in transcription and/or stabilisation of the mutant mRNAs. Actinomycin D experiments performed on otd²/+, otd^2FL/+ and Otx2^2c/+ ES cell lines indicated that the half-life of the mutant mRNAs was very similar to that of the Otx2 mRNA (t_1/2=90 minutes) (Fig. 8F).

This suggests that lack of introns and insertion of the targeting vector may severely interfere with the transcription of the Otx2 locus and, thereby, be responsible for the partial rescue observed in otd^2FL/otd^2FL and Otx2^2c/Otx2^2c embryos.

Finally, the ability of otd², otd^2FL and Otx2^2c mRNAs to form efficient polyribosome complexes was studied. Cytoplasmic extracts from 10.5 d.p.c. otd²/+ heads were fractionated on a sucrose gradient and mRNA purified from each fraction. Distribution of β-actin and Otx2 mRNAs along the gradient was analysed with the mutant mRNA. According to the length of β-actin (350 amino acids) and OTX2 (289 amino acids) coding sequences, ≥50% of their mRNAs was concentrated in the first two fractions that contained polyribosome complexes with >10 and 8-10 ribosomes (Fig. 9A,B). Similarly, the otd² mRNA that encodes a 548 amino acid long protein should be concentrated in the fractions with the highest number of ribosomes, if efficiently translated. In contrast, approx. 60% of the otd² mRNA was distributed among the central fractions of the gradient containing ≤5-7 ribosomes. This finding indicates that the otd² mRNA is severely affected in its ability to form efficient polyribosome complexes.

The same experiment was performed on otd^2FL/+ and Otx2^2c/+ embryos (Fig. 9A,C,D). Importantly, in these embryos the otd^2FL (Fig. 9A,C) and the Otx2^2c (Fig. 9A,D) mRNAs were correctly distributed and ≥60% of the otd^2FL mRNA was concentrated in the first two fractions (Fig. 9C) and ≥60% of the Otx2^2c mRNA colocalised with the Otx2 mRNA. This result provides direct in vivo evidence that the Otx2 5′ and 3′ UTR sequences deleted in otd² mutants are essential for both otd^2FL and Otx2^2c mRNAs to be correctly translated.

Distribution and quantitative mRNA analysis as well as translational efficiency were also studied in the hOtx1² mutant. Indeed, also in this case lack of protein in the epiblast and its derivatives correlated with the deletion of the same Otx2 5′ and 3′ UTRs deleted in the otd² mutant. In this mutant, the cytoplasmic hOtx1 mRNA was approx. 40% of the Otx2 mRNA, while the hOtx1 nuclear mRNA accumulated up to approx. 140% of the Otx2 nuclear mRNA (Fig. 9E). This indicates that impairment of nucleo-cytoplasmic export, identified in otd² mutants, was also observed in hOtx1² mutant embryos. The ability of the hOtx1 mRNA to form polyribosome complexes was also greatly impaired and only ≤25% of its mRNA was detected in the first two fractions where, based on the length of the protein (355 amino acids), most of the hOtx1 mRNA should be concentrated.

Therefore, also in the hOtx1² mutant, heavy accumulation of nuclear mRNA and impairment in translation correlate as in the otd² mutant, with lack of Otx2 5′ and 3′ UTRs. This molecular analysis has demonstrated by both gain- and loss-of-function experiments in four different mouse models that the Otx2 5′ and 3′ UTRs are crucial in controlling nucleo-cytoplasmic export and translation of the Otx2 mRNA.

However, although these data cannot demonstrate whether the nucleo-cytoplasmic export is also affected only in the epiblast and neuroectoderm, they provide further support to the concept that murine brain development requires accurate epiblast-restricted translational control of the Otx2 mRNA.

Until recently, it has been generally assumed that the insect and vertebrate CNS are non-homologous structures that evolved independently (Garstang, 1928; Lacalli, 1994).

For example, the so-called auricularia hypothesis postulates that the chordate CNS is homologous to the entire outer ectoderm of the insect embryo, based on both anatomical studies made in echinoderms (particularly auricularia larvae) and urochordates as well as on comparative expression of the HOX genes (Garstang, 1928; Lacalli, 1994). In contrast to such hypotheses, and beginning with the highly debated proposal by Geoffroy Saint-Hilaire (Lacalli, 1994; Peterson, 1995; Arendt and Nübler-Jung, 1999), an increasing body of evidence has now accumulated that suggests a common evolutionary origin of the insect and vertebrate CNS (Arendt and Nübler-Jung, 1996; Arendt and Nübler-Jung, 1999; De Robertis and Sasai, 1996).

A growing number of molecular genetic studies support the idea of a common evolutionary origin of the CNS in different phyla. In Drosophila and vertebrates, the conserved families of HOM-C/HOX, otd/Otx, empty spiracles (ems)/Emx as well as en/En and wingless (wg)/Wnt genes play a fundamental role in the regional specification of the neuroectoderm destined to form nerve cord/posterior brain and anterior brain as well as in the establishment of boundary regions between them (Joyner, 1996; Lumsden and Krumlauf, 1996; Hanks et al., 1998; Araki and Nakamura, 1999; Reichert and Simeone, 1999).

Despite the overall morphological differences, expression pattern and mutant phenotypes for Drosophila otd and mouse Otx genes can have obvious parallels.

Nevertheless, OTD and OTX proteins are highly conserved only in the homeodomain, which represents roughly one-tenth of the whole OTD protein and about one-sixth and one-fifth of OTX1 and OTX2 proteins, respectively (Simeone et al., 1993). This suggests that Otx genes have either acquired new functions during evolution and these reside outside the homeodomain or, alternatively, that the functional properties of the non-homeodomain regions are common to both OTD and OTX but they are difficult to identify in the amino acid sequence. This study and previous reports (Acampora et al., 1998a; Leuzinger et al., 1998; Nagao et al., 1998) indicate that, despite the restricted homology in their amino acid sequence, OTD/OTX proteins share a remarkable and unexpected functional equivalence in mice and flies.

Indeed, although an OTD-mediated rescue of OTX1 functions could be expected in neuronal cells (Acampora et al., 1998a), it is totally unexpected, in our opinion, that the fly gene is able to rescue OTX2-dependent properties that are encoded by the VE and neuroectoderm for early induction of the anterior patterning and maintenance of forebrain and midbrain regionalisation. It may suggest that even though OTD/OTX proteins have gained new role(s) during evolution, their basic properties have been retained. In this sense it is noteworthy that the absence of the lateral semicircular duct of the inner ear of Otx1^–/– mice is never recovered either by otd or Otx2, thus indicating that the ability to specify this sensorial structure may, therefore, represent a new Otx1-specific property (Acampora et al., 1996; Acampora et al., 1998a; Acampora et al., 1999; Morsli et al., 1999). However, the molecular nature of functional equivalence between sequences outside the OTD/OTX homeodomain is still unclear and remains to be investigated in detail. Assuming that otd and Otx2 gene functions are conserved, it is unclear why the brain vesicles of protochordates have been so deeply and suddenly modified in a much more complex brain that has been maintained in its basic topography until mammals. A rather obvious answer to this question is that new genetic functions have been recruited and stabilised, thus creating new versions of pre-existing developmental pathways (Acampora and Simeone, 1999; Holland 1999; Acampora et al., 2001). Indeed, it might be that conserved functions such as those encoded by OTD/OTX proteins become able to perform new roles even while retaining an evolutionary functional equivalence because they acquire the ability to be expressed in new cell types. Based on this hypothesis, it is expected that drastic evolutionary events should act on the regulatory control (transcription and translation) of Otx-related genes rather than on their coding sequences.

Here, we report that OTD is functionally equivalent to OTX2 only when the OTD encoding mRNA is provided with the 5′ and 3′ UTRs of Otx2. This implies that besides the Otx2 transcriptional control, synthesis of OTD protein also requires the Otx2 regulatory control for nucleo-cytoplasmic export and, importantly, epiblast-specific translation of its mRNA. These functions were established when 213 bp of the 5′ UTR and the 3′ UTR were fused to the otd coding sequence. We have recently reported that mice carrying a 300 bp insertion in the Otx2-3′ UTR exhibited epiblast-restricted impairment in the translation of the mutated Otx2 mRNA (Pilo Boyl et al., 2001). Here we have also shown that hOtx1² mutants, which lack the same Otx2 sequences that were absent in otd² mutants, display identical molecular impairments. Together these data provided in vivo evidence of both gain and loss of protein synthesis in epiblast and neuroectoderm, depending on the presence of 5′ and 3′ UTRs of Otx2. While we have reported that the 3′ UTR may be relevant in the translation of the Otx2 mRNA (Pilo Boyl et al., 2001), we believe that also the 5′ UTR plays an important role in this control. This is supported by preliminary data indicating that mice lacking only 150 bp within the 213 bp long region deleted in otd mutants, develop a sharp headless phenotype (D. A. and A. S., unpublished results).

In sum, these data support the notion that otd/Otx functions have been established in a common ancestor of fly and mouse and retained throughout evolution, while regulatory control of their expression has been modified and re-adapted by evolutionary events that have led to the specification of the increasingly complex vertebrate brain (Sharman and Brand, 1998; Simeone, 1998; Acampora and Simeone, 1999; Reichert and Simeone, 1999).

A likely consequence of modification of regulatory control of gene expression may result in a greater number of molecular interactions. This may contribute to modifying relevant morphogenetic processes that, in turn, can confer a change in shape and size of the body plan as well as in the generation of cell types with new developmental potentials (Holland, 1999; Acampora and Simeone, 1999). On this basis, Otx gene duplication and subsequent or contemporary modification of regulatory control might have contributed to the evolution of the mammalian brain, for example by controlling the position of the midbrain-hindbrain boundary (MHB).

Previous work has indicated that Otx genes control the positioning of MHB (Simeone, 2000; Wurst and Bally-Cuif, 2001). Importantly, this control is strictly dependent on the level of OTX proteins in the neuroectoderm. Data presented in this study and those previously reported, clearly indicate that reduction in OTX levels is invariably reflected in anterior displacement of the MHB and progressive loss of forebrain and midbrain identities. Our data demonstrate that OTD protein may restore a quite normal positioning of the MHB, thus indicating that a crucial process in murine brain development can be performed by the invertebrate function.

Nevertheless, this process may occur only if the Otx2 regulatory control is provided. On this basis, we hypothesise that, once established, this control has been recruited and maintained throughout vertebrate head evolution.

We wish to thank Andrew Lumsden for critical reading of the manuscript and helpful discussions. We also thank A. Secondulfo and S. Piga for manuscript preparation, Amanda McGuigan and the staff for the excellent animal care. This work was supported by the MRC (Grant Number: G9900955), the Wellcome Trust (Grant Number: 062642/Z/00), the Italian Association for Cancer Research (AIRC), the EU BIOTECH programme (Number: BIO4-CT98-0309), the CNR Target Project on Biotechnology, the MURST-CNR Programme Legge 488/92 (cluster 02) and the Fondation Bettencourt Schueller.