LIM-homeodomain (LIM-HD) proteins regulate tissue specific expression of target genes and are required for the normal development and maintenance of numerous organs (reviewed in Hunter and Rhodes,2005). Lhx2 is a member of this family of transcription factors, which consists of 12 known genes structurally characterized by two amino terminal zinc-finger motifs (LIM domains) and a carboxy terminal homeodomain (Kadrmas and Beckerle,2004; Hunter and Rhodes,2005). Functional inactivation of Lhx2 in mice results in a cerebral cortex of reduced size as well as abnormal development of the cortical hem, telencephalon, olfactory system, and cortical patterning defects (Porter et al.,1997; Bulchand et al.,2001; Monuki et al.,2001; Hirota and Mombaerts,2004). Lhx2 is also required for normal limb and liver formation; Lhx2−/− mice develop liver fibrosis and erythropoiesis is incomplete resulting in death by severe anemia prior to birth (Nohno et al.,1997; Porter et al.,1997; Rodriguez-Esteban et al.,1998; Wandzioch et al.,2004). The known sites of Lhx2 expression correspond well with these phenotypes.
A conspicuous phenotype of the Lhx2−/− mouse is the lack of eyes. Although the optic vesicles evaginate normally from the brain, the optic cups, and consequently eyes, never form (Porter et al.,1997). Lhx2 is a member of a group of eye field transcription factors (EFTFs) required for eye formation. Functional inactivation of the EFTFs Pax6, Rx (Rax), Six3, Optx2 (Six6), or Lhx2 results in frogs, fish, rodents, and/or humans with no, or abnormal, eyes (Hill et al.,1991; Mathers et al.,1997; Porter et al.,1997; Chow et al.,1999; Zuber et al.,1999; Wawersik et al.,2000; Loosli et al.,2001,2003; Tucker et al.,2001; Carl et al.,2002; Andreazzoli et al.,2003; Lagutin et al.,2003; Voronina et al.,2004). Experiments in Xenopus predict that Lhx2 forms a regulatory feedback network with other EFTFs (Zuber et al.,2003). Despite the dramatic eye phenotype observed in the Lhx2−/− mouse and other evidence suggesting a critical role for Lhx2 in eye formation, its pattern of expression during eye development has not been reported.
In Drosophila, an analogous network of transcription factors is required for compound eye formation (Wawersik and Maas,2000; Kumar and Moses,2001). As in vertebrates, inactivation of fly EFTF homologs results in eyeless flies. Lhx2 has the highest sequence homology to the Drosophila selector gene apterous (ap) (Cohen et al.,1992; Xu et al.,1993). The expression pattern of ap is similar in many respects to that of mammalian Lhx2, as both are expressed in the nerve cords, olfactory organs, brain, and limbs of mice and flies (Rincon-Limas et al.,1999). Some functional conservation is also evident since human Lhx2 can rescue the ap phenotype (Rincon-Limas et al.,1999). However, in contrast to Lhx2−/− mice, ap flies have no eye phenotype. These results imply at least a partial divergence of ap/Lhx2 function during evolution.
Here, we report the cloning, structural characterization, and expression pattern of Xenopus laevis Lhx2. Using sequence comparisons and phylogenetic analysis, we demonstrate that Lhx2 has been highly conserved throughout evolution. Using in situ hybridization, we generate the first detailed analysis of Lhx2 expression during eye development and compare its expression pattern to that of other EFTFs required for eye formation. Finally, we identify differences between the expression patterns of mammalian and amphibian Lhx2 during early embryogenesis. Consistent with the relatively rapid evolution of the LIM domain containing gene family, our results suggest a partial functional divergence during vertebrate evolution.
RESULTS AND DISCUSSION
Isolation of a Xenopus laevis Lhx2 Homolog
Degenerate RT-PCR was used to amplify a fragment of the XLhx2 cDNA, which was then used to isolate a full-length cDNA coding for the predicted amino acid sequence presented in Figure 1. The XLhx2 cDNA encodes a predicted protein of 398 residues with a molecular weight of 44.2 kDa. The Xenopus protein shares greatest overall sequence identity with chicken Lhx2 (93%), followed by the axolotl (91%), mouse (86%), human (86%), fugu (84%), rat (84%), and fly (31%) homologues (Fig. 1). The relatively low overall sequence identity between XLhx2 and ap is due to the highly divergent amino and carboxy terminal sequences as well as the sequence between the highly conserved LIM domains and homeodomain. The canonical LIM domain contains 50–60 amino acids with the consensus sequence CX2CX16–23HX2CX2CX16–23 CX2 C/D/H (Kadrmas and Beckerle,2004). The LIM domains of XLhx2 follow this pattern. The first Xenopus Lhx2 LIM domain is identical with all the reported vertebrate homologues but Fugu, which differs at only two of the 53 residues. All, including ap, have the consensus sequence, CX2CX17HX2 CX2CX18CX2D. The second 54 amino acid LIM domain of XLhx2 differs by only one residue with the chicken, two residues with the axolotl, mouse, human, and rat, and three residues with the Fugu sequences yet all, including ap, have the same consensus sequence, CX2CX18HX2CX2CX18CX2H. The first and second LIM domains of ap are 57 and 56% identical to those of XLhx2, respectively. Finally, the Lhx2 homeodomain is remarkably well conserved among all species. The XLhx2 homeodomain is 98% identical to all the vertebrate homologues and 90% identical to that of ap (Fig. 1). Taken together, these data support the conclusion that the isolated cDNA codes for a Xenopus Lhx2 homolog.
XLhx2 Is Expressed in the Eyes During All Stages of Development
Lhx2−/− mice lack eyes (Porter et al.,1997). Previous studies have only reported mouse and chick Lhx2 eye expression at a single developmental time point (Xu et al.,1993; Rodriguez-Esteban et al.,1998; Retaux et al.,1999; Rincon-Limas et al.,1999). Therefore, we first used in situ hybridization to better define the spatial and temporal expression patterns of XLhx2 during eye formation.
We first detected XLhx2 at stage 12.5, consistent with previous RT-PCR results (Zuber et al.,2003). Initial expression was detected as a single stripe in the neuroectodermal layer of the anterior neural plate consistent with the location of the eye field (Fig. 2A,B). At stage 15 (Fig. 2C), expression is more intense, but reduced medially when compared to stage 12.5. At stage 18, when the optic vesicles are forming, the single band of expression is further resolved into the two eye primordia (Fig. 2D).
To more precisely determine the expression pattern of XLhx2 during eye formation, we performed in situ hybridization on sectioned embryos. We observed XLhx2 expression in the presumptive retinal pigment epithelium (RPE), neural retina, and optic stalk of stage 24 optic vesicles (Fig. 2E). At the onset of lens formation (stage 28), XLhx2 expression is distinctly absent from the lens placode, yet continues to be expressed in both the presumptive RPE and neural retina (Fig. 2F).
XLhx2 expression is detected throughout the retina at stage 32 with relatively strong expression in the peripheral retina (Fig. 2G). A few hours later, at stage 35/36, XLhx2 expression is restricted in the central retina to the presumptive inner nuclear layer (INL; Fig. 2H). By stage 41, retinal XLhx2 expression is restricted to the ciliary marginal zone (CMZ) and a subset of (possibly amacrine) cells in the INL (Fig. 2I).
In summary, at early developmental stages, XLhx2 is expressed in all cells of the developing retina. Once retinal progenitors begin to differentiate, Lhx2 expression becomes restricted to the retinal progenitor cells of the CMZ and a subset of INL cells. Potential roles for Lhx2 in the RPE and optic stalk are also suggested by the expression pattern.
These results are consistent with the functional analysis in mouse, demonstrating that Lhx2 is required in the early stages of eye development (Porter et al.,1997). Lhx2−/− mice develop optic vesicles but not optic cups, suggesting Lhx2 function is required for normal eye formation sometime between these developmental steps. Consistent with this interpretation, XLhx2 is detected prior to eye cup formation (stage 27 in Xenopus) (Nieuwkoop and Faber,1994). The absence of XLhx2 expression in the lens placode is of particular interest. The lens placode, which normally develops between optic vesicle and cup formation, does not form in the Lhx2−/− mouse (Porter et al.,1997). Together, these results suggest that lens placode formation requires Lhx2 in the optic vesicle. Functional Lhx2 in the optic vesicle may be necessary for the inductive interactions between the developing retina and lens, which have been proposed to control eye development (Grainger,1992; Saha et al.,1992). Confirmation of this interpretation will require a detailed analysis of Lhx2 expression during early lens formation in mouse.
Comparison of Eye Field Transcription Factor Expression Patterns During Eye Development
Lhx2 is part a genetic network of eye field transcription factors required for vertebrate eye formation (Zuber et al.,2003). In addition to regulating eye field specification, EFTFs can also regulate eye cell fate (Furukawa et al.,2000; Marquardt et al.,2001; Reza and Yasuda,2004; Wang and Harris,2005). The ability of EFTFs to alter retinal cell fate can be dramatically enhanced in combination with other transcription factors (Wang and Harris,2005). Therefore, we compared the expression pattern of Lhx2 (Fig. 2) with the reported expression patterns of Xenopus transcription factors Pax6, Rx1, Six3, and Optx2 during eye formation (Casarosa et al.,1997; Hirsch and Harris,1997; Perron et al.,1998; Zuber et al.,1999; Zhou et al.,2000; Ghanbari et al.,2001; Wang and Harris,2005). By stage 24, the optic vesicle has evaginated from the neural tube and makes contact with the overlying epidermis. The undifferentiated cells of the optic stalk (OS), presumptive RPE, and neural retina (NR) express all five genes. At stage 28, most cells of the neural retina are still multipotent, the vesicles invaginate to create the optic cup, and the lens placode is forming. All five genes continue to be expressed at this stage in both the presumptive RPE and NR. Pax6 and Six3 are also expressed in the lens placode. By stage 32, the lens vesicle is well formed, the cells within the retina are becoming post-mitotic and are migrating to their respective nuclear layers (Holt et al.,1988). At stage 32, all five genes are expressed in the neural retina, while only Pax6 is expressed in the lens vesicle. At stage 41, cells of the central retina are post-mitotic and form three nuclear layers: the outer nuclear, inner nuclear, and ganglion cell layers (ONL, INL, and GCL, respectively; Fig. 2J). Cells in the ciliary marginal zone are undifferentiated and pleuripotent (Wetts et al.,1989). At stage 41, only Rx1 is expressed in the ONL (Fig. 2L). Rx1 and Six3 are expressed in the outer part of the INL (Fig. 2K,L), while Lhx2, Pax6, and Optx2 are all expressed in the inner part of the INL (Fig. 2J–L). Pax6 and Optx2 are co-expressed throughout the GCL, while Six3 is expressed in only the outer part of the GCL (Fig. 2K,L). Only Pax6 is detected in the lens at stage 41, but just at its periphery (Fig. 2K). Interestingly, the CMZ is the only location in the mature retina where cells express all five EFTFs (Fig. 2J–L).
In summary, the expression patterns of Lhx2, Pax6, Six3, Rx1, and Optx2 are identical in retinal progenitors. By stage 41, however, these genes are expressed in overlapping, but non-identical patterns within the differentiated cells of the eye. These results are consistent with the hypothesis that specific combinations of the EFTFs may be required for eye cell differentiation and/or the maintenance of specific cell fates.
Divergent Lhx2 Expression Patterns in Vertebrates
Although the expression patterns of XLhx2 show similarity with those of higher vertebrates, significant differences were also observed. For example, XLhx2 is first detected at stage 12.5, prior to formation of the frog neural plate (Fig. 2A–C). In contrast, mouse Lhx2 is first detected at embryonic day 8.5, well after neural plate formation, when neural tube closure begins (Porter et al.,1997). XLhx2 is also clearly expressed in the Xenopus pineal gland from stage 25 to at least stage 37/38 (Fig. 3A and data not shown). Mouse E13.5 pineal gland, however, does not express detectable levels of Lhx2, and mouse pineal expression has not been reported at later developmental stages (Retaux et al.,1999; Bachy et al.,2001). The above differences were observed despite the fact that Lhx2 is detected in the developing olfactory system and other brain regions of both species (and chick) at similar developmental stages (Fig. 3A,B; Nohno et al.,1997; Rodriguez-Esteban et al.,1998; Rincon-Limas et al.,1999; Bachy et al.,2001,2002; Moreno et al.,2004,2005).
The livers of Lhx2−/− mice are reduced in size and develop a spontaneous and progressive hepatic fibrosis (Porter et al.,1997; Wandzioch et al.,2004). Mouse Lhx2 is expressed from embryonic day 9 in the septum transversum mesenchyme, which eventually forms the mesenchymal portion of the adult liver (Kolterud et al.,2004). Interestingly, we did not detect XLhx2 expression in or around the developing liver bud in whole mount or sectioned Xenopus embryos in any of the developmental stages tested (Fig. 3B,D,F and data not shown). In situ hybridizations for XLhx2 were performed in parallel with the liver marker XHex (Newman et al.,1997; Zorn and Mason,2001) to confirm the absence of detectable XLhx2 expression in the developing liver (compare Figs. 3B,D,F to C,E,G, respectively).
Xenopus laevis is a tetraploid with two closely related paralogs for most genes (Bisbee et al.,1977; Krotoski et al.,1985). Genome duplication took place no less than 15 million years ago, long enough to allow for evolutionary divergence (Bisbee et al.,1977). One possible explanation for the differing mouse and Xenopus expression patterns is that the isolated XLhx2 cDNA is an aberrantly expressed XLhx2 paralog. Indeed, two partial XLhx2 sequences, corresponding to residues 96 to 224 and 230 to 325 of a second highly homologous XLhx2 paralog, have been reported (Bachy et al.,2001). However, we observed no difference in the reported expression patterns of the two Xenopus Lhx2 paralogs. This excludes the possibility that differences in Xenopus paralog expression patterns explain the differences between frog and higher vertebrate Lhx2 expression patterns.
In summary, while the expression patterns of Xenopus and mouse Lhx2 shown some similarity (this study and Bachy et al.,2001,2002), divergent expression patterns were also observed in the early neural plate, pineal, and liver. These differences cannot be explained by differing paralogous expression patterns.
Phylogenetic Analysis Demonstrates XLhx2 Is the True Lhx2 Ortholog in Frog
LIM-HD proteins share significant amino acid identity. For example, the LIM-HD protein with the greatest identity to human Lhx2 is Lhx9; with 94, 91, and 100% identity in the LIM 1, LIM 2, and homeodomains, respectively (data not shown). Furthermore, not all Xenopus LIM-HD genes have been identified, and, in some cases, only partial sequences are available. For example, two distinct, yet incomplete Xenopus Lhx9 sequences have been reported (Bachy et al.,2001; Klein et al.,2002). Consequently, sequencing errors and/or splice variation of sequences in distantly related species could have resulted in the misidentification of the isolated cDNA as a Xenopus Lhx2, possibly explaining its divergent expression pattern when compared to true Lhx2 orthologs. Alignment of the XLhx2 protein sequence with all twelve human LIM-HD family members revealed that the frog sequence is more similar to human Lhx2 in the homeodomain, LIM domains, and in overall sequence identity than any other LIM-HD protein. Furthermore, the gene with the highest nucleotide and protein sequence homology to XLhx2 in the human genome is human Lhx2 (data not shown). Nevertheless, we used phylogenetic analysis to systematically determine the evolutionary relationship between the isolated sequence and members of the LIM-HD family of proteins (Fig. 4). Sequences were initially selected based on BLAST analysis. We then compared the protein sequences (where available) from human (Homo sapiens), mouse (Mus musculus), chicken (Gallus gallus), fish (Danio rerio, zebrafish; or Tetraodon nigroviridis, pufferfish), and frog (Xenopus laevis or Xenopus tropicalis) orthologs to that of the putative X. laevis Lhx2 protein. Genes for which full-length sequences were available (exceptions noted in Fig. 4) were used. Fifty-two LIM-HD family members were aligned using the CLUSTALW algorithm (Thompson et al.,1994). Phylogenetic trees were constructed using both distance (neighbor-joining) and maximum parsimony methods (see Experimental Procedures section for details). As expected, LIM-HD family members segregated into 12 groups (clades) corresponding to the eight LHX, two ISL, and two LMX genes. XLhx2 grouped exclusively with other vertebrate Lhx2 proteins with high bootstrap values regardless of the method used (100% for distance and 99% for maximum parsimony methods, respectively, calculated by analysis of 10,000 replicates). For clarity, not all available taxa were used in the analysis presented in Figure 4. However, XLhx2 always grouped in the Lhx2 clade with similarly high bootstrap values when broader alignments including LIM-HD proteins from other vertebrate and invertebrate species were added to the analysis (data not shown). These results indicate the isolated cDNA codes for the true Xenopus laevis Lhx2 ortholog.
These results demonstrate different expression patterns and suggest functional differences between mouse and Xenopus Lhx2. For example, mouse Lhx2 is expressed in the developing and adult liver and is required for normal liver formation (Porter et al.,1997; Kolterud et al.,2004). In contrast, Xenopus Lhx2 is not detected during frog liver development. It is interesting to note that Lhx2 is also undetectable in axolotl liver, suggesting that the differences observed here are not Xenopus laevis specific, but extend to other amphibia (Showalter et al.,2004). Whether all the differences in expression pattern are functionally relevant remains to be determined. This study, however, provides the groundwork for future experiments designed to define the precise role of Lhx2 during embryogenesis.
Xenopus laevis embryos were generated by in vitro fertilization, cultured in 0.1× MMR at 18°C to the appropriate developmental stage (Nieuwkoop and Faber,1994), anesthetized in 0.01% tricane, and fixed for 45 min in 4% paraformaldehyde. Fixed embryos were either incubated in 1× PBS containing 20% sucrose, mounted in O.C.T. compound (Tissue Tek, Sakura Finetek, Inc., Torrance, CA), and cryostat sectioned or placed in 50% methanol, overnight at −20°C, and then processed by in situ hybridization.
Cloning and In Situ Hybridization
A 700-bp fragment of the XLhx2 cDNA was PCR-amplified from stage-30 Xenopus eye cDNA using degenerate primers corresponding to the conserved rat and chicken Lhx2 protein sequences CAGCGGKI and NHNPDAKD (5′TGY GCI GGI TGY GGN GGN AAR AT3′ and 5′TCY TTI GCR TCI GGR TTR TGR TT3′, respectively). The fragment was subcloned into the pBSKS (+) vector (Stratagene, La Jolla, CA) at the EcoRV site and used to make 32P-labeled probe for library screening (Prime-It® II Random Primer labeling kit, Stratagene). The full-length cDNA, presented here was isolated from a lambda ZAP library (Stratagene) generated using stage 28–30 Xenopus head cDNA (courtesy of Richard Harland). The isolated full-length XLhx2 cDNA was used to generate digoxigenin-labeled (Roche Applied Science, Indianapolis, IN) RNA probe for in situ hybridizations carried out as previously described (Viczian et al.,2003; Zuber et al.,2003). The XHex cDNA template used for in situ hybridizations was RT-PCR amplified from RNA isolated from stage 33/34 Xenopus laevis gut tissue, using primers designed to XHex (accession no. U94837; Newman et al., 1997). The sequence of the 5′ and 3′ primers used were: 5′GCA CCC AAC TCC CTT CTA CA3′ and 5′ATC CTT TGT CGC CTT CAA TG3′, respectively. The PCR fragment was cloned into the pGEM-T easy vector (Promega, Inc., Madison, WI) and sequenced to confirm its identity. Sense and antisense DIG-labeled RNA in situ hybridization probes were generated by digestion with NcoI or SalI and transcription with Sp6 and T7 RNA polymerase, respectively.
Sequence and Phylogenetic Analysis
Analysis of the DNA and protein sequences were performed using the DNAStar Lasergene software package version 6.14 (DNAStar, Inc., Madison, WI) and PAUP version 4:Beta 10 release (Sinauer Associates, Inc., Sunderland, MA). Fly and vertebrate Lhx2 protein sequences were initially aligned with MegAlign using the ClustalW algorithm with a pairwise alignment gap penalty of 10, gap length penalty of 0.2, and the Gonnet Series Protein Weight Matrix (Thompson et al.,1994). Lhx2 orthologs were then optimally aligned by hand to generate Figure 1. For the phylogenetic tree, sequences were aligned with MegAlign using the ClustalW algorithm with a pairwise alignment gap penalty of 35, gap length penalty of 1.0, and the Gonnet Series Protein Weight Matrix (Thompson et al.,1994). PAUP version 4:b10 was used to generate phylogenetic trees using distance (neighbor-joining) and maximum parsimony algorithm methods (Saitou and Nei,1987; Swofford,2000). Bootstrap proportions were determined by analysis of 10,000 replicates for each method. The protein sequences of LIM-HD family members were retrieved from online databases and are available upon request from the authors.
We thank Nicole Patterson, Edward Pang, Dean Pask, Katie Baird, Aaron Zorn, and Barry Knox for assistance. We are also grateful to Drs. Gaia Gestri, André W. Brändli, and Hedyeh Ghanbari for sharing their unpublished XSix3 expression pattern data. This work was supported in part by E. Matilda Zeigler Foundation for the Blind, European Commission, and the Wellcome Trust, UK. M.Z. was sponsored by Burroughs Wellcome Fund through a Hitchings-Elion Fellowship. This work was also supported by Research to Prevent Blindness (Career Development Award to A.V. and M.E.Z. and an unrestricted grant to the Opthalmology Department of SUNY Upstate Medical University).