The LIM-Homeodomain Gene Family in the DevelopingXenopus Brain: Conservation and Divergences with the Mouse Related to the Evolution of the Forebrain

Abstract

A comparative analysis of LIM-homeodomain (LIM-hd) expression patterns in the developing stage 32 Xenopus brain is presented. x-Lhx2, x-Lhx7, andx-Lhx9 were isolated and their expression, together with that of x-Lhx1 and x-Lhx5,was analyzed in terms of prosomeric brain development and LIM-hd combinatorial code and compared with mouse expression data. The results show an almost complete conservation of expression patterns in the diencephalon. The Lhx1/5 and Lhx2/9subgroups label the pretectum/ventral thalamus/zona limitans versus the dorsal thalamus, respectively, in alternating stripes of expression in both species. Conversely, strong divergences in expression patterns are observed between the telencephalon of the two species for Lhx1/5 andLhx2/9. Lhx7 exhibits particularly conservative patterns and is proposed as a medial ganglionic eminence marker. The conservation of diencephalic segments is proposed to mirror the conservative nature of diencephalic structures across vertebrates. In contrast, the telencephalic divergences are proposed to reflect the emergence of significant novelty in the telencephalon (connectivity changes) at the anamniote/amniote transition. Moreover, the data allow the new delineation of pallial and subpallial domains in the developing frog telencephalon, which are compared with mouse subdivisions. In the pallium, the mouse combinatorial expression of LIM-hd is notably richer than in the frog, again possibly reflecting evolutionary changes in cortical connectivity.

The comparative study of forebrain development was long impaired by difficulties in defining homologous territories between distant species (Striedter, 1997). However, the study of developmental genes has revealed two fundamental aspects of vertebrate brain development.

First, there is a large conservation in the way a topographical organization of structures and connections is set up under the control of genes expressed according to a grid of longitudinal and transverse compartments. As proposed in the prosomeric model, the expression of developmental factors divides the diencephalon into anteroposterior segments that prefigure adult functional units (Puelles and Rubenstein, 1993; Puelles, 1995). The principles of this model are verified in the vertebrate phylum: prosomeric subdivisions are reported in lamprey [vertebrate agnathe (Pombal and Puelles, 1999)], zebrafish [gnathostome (Wullimann and Puelles, 1999; Hauptmann and Gerster, 2000)], Xenopus [tetrapod (Milan and Puelles, 2000)], chick [amniote (Figdor and Stern, 1993)], and mouse (mammal), five species representing major transitions in vertebrate evolution. Such prosomeric (transverse/anteroposterior) subdivisions are more controversial in the telencephalon. Nevertheless, the expression of major regulators (Dlx-2, Pax-6, Emx-1) along the telencephalic dorsoventral axis also reveals strong similarities of molecular profiles between mouse and chicken pallium/subpallium (Puelles et al., 2000). Therefore, subdivisions and field homologies in the telencephalon are also postulated.

Second, there are significant differences in the relative size of cerebral areas, in the connections between areas, and in neuronal phenotypes among vertebrates, especially in the forebrain. A major evolutionary trend is the progressive involvement of the cortex in the processing of thalamic sensory information in tetrapods (Marin et al., 1998a; Reiner et al., 1998). Unknown changes in cell specification gene expression patterns must have allowed this functional diversification, although the general Bauplan to build a brain has been conserved across vertebrates. As emphasized by Striedter (1997), a major step in increasing the number and complexity of forebrain connections was achieved at the anamniote/amniote transition.

To approach the question of the evolutionary changes in regionalization and connectivity in vertebrate forebrain, we used LIM-homeodomain (LIM-hd) factors as functional determinants of cell identity. They govern not only regional specification, but also axonal projection patterns and neurotransmitter phenotypes, by using a LIM-hd combinatorial code well described in the spinal cord (Jessel, 2000). Moreover, LIM-hd gene function in neural development seems to be conserved across phylogeny (Hobert and Ruvkun, 1998). Therefore, discrete changes in regional and/or combinatorial LIM-hd expression would be susceptible to alter connectivity patterns between areas and to be selected in the evolutionary process. Here, we used the mouse/Xenopus comparison to analyze differences in LIM-hd expression domains between anamniotes and amniotes, and we tried to correlate these changes with known changes in morphogenesis and connectivity of forebrain structures. The Lhx1/5,Lhx2/9, and Lhx6/7/8 subgroups were selected for their predominant forebrain expression in mouse (Rétaux et al., 1999). We isolated new Xenopus orthologs (x-Lhx2, x-Lhx7, and x-Lhx9), we analyzed their expression together with that of x-Lhx1/5 in terms of both prosomeric organization and combinatorial LIM-hd expression, and we compared the resulting patterns with the mouse.

MATERIALS AND METHODS

RT-PCR cloning. Total RNA from brains of stages 32 or 39/40 embryos were reverse transcribed to cDNA with avian myeloblastosis virus reverse transcriptase (Boehringer Mannheim, Mannheim, Germany) and used as templates for PCR reactions (Qiagen, Hilden, Germany) using the following oligonucleotide primers: for Lhx7/8, Fdeg78: AARGTIAAYGAYYTITGYTGGCAYGT and Rdeg78: TGICKIGCICKRCARTTYTGRAACCA; for Lhx2/9, Fdeg9: TIGCIGTIGAYAARCARTGGCAY(ACT)T and 32, MAYTTIGCYCTIGCRTTYTGRAACCA (where I is an inosine residue); for Lhx1 and Lhx5, FX1: TGCCTTCTATTCTCCTAATCCGCCC; RX1: CAGCTTAGGCTACCACACTGCCG; FX5: GGATTTCACTGGACTTGGCTTCTGC and RX5: GTTGGAATCAGGCGTACAAGCACC. The various primer combinations led to the amplification of single bands. After these fragments were subcloned (700 bp to 1 kb) into pGEM-T (Promega, Madison, WI), sequencing of several independent clones revealed the presence of various fragments. Analysis and alignments performed using the ClustalX program identified the new clones as the Xenopus orthologs of mouseLhx2, Lhx9, and Lhx7 genes (GenBank accession numbers AJ311711, AJ311712, AJ311713, AJ311714, andAJ311715). Xenopus and mouse Lhx1 andLhx5 were reisolated using primers designed in the already published sequences (Taira et al., 1992; Fujii et al., 1994; Toyama et al., 1995; Sheng et al., 1997). The x-Dll3 plasmid was a gift of Nancy Papalopulu (Cambridge, UK).

In situ hybridization. The pGEM-T plasmids were linearized with restriction enzymes NdeI or NcoI (Promega) and used as template for RNA synthesis with T7 or SP6 polymerase (Stratagene, La Jolla, CA) in the presence of digoxigenin-11-UTP (Boehringer Mannheim) for antisense and sense control probes, respectively. Xenopus or mouse embryos were fixed overnight in MEMFA (0.1 m MOPS, pH 7.4, 2 mm EGTA, 1 mmMgSO_4, 3.7% formaldehyde) at 4°C and then progressively dehydrated in methanol and stored at −20°C until use. After rehydration, embryos were immersed in MEMFA at room temperature for 10 min and bleached in 6% H₂O₂ for 1 hr. Embryos were treated with proteinase K (10 μg/ml, 15 min) and fixed at room temperature for 20 min before prehybridization (1 hr at 65°C and 2–4 hr at 55°C). Hybridization was performed overnight at 55°C in a 50% formamide hybridization medium containing 1 μg of RNA probe. Hybridization was detected using an alkaline phosphatase-coupled anti-digoxigenin antibody (Boehringer Mannheim) diluted to 1:1500. Alkaline phosphatase staining was developed with NBT/BCIP (Boehringer Mannheim).

Two-color whole-mount in situ hybridization was performed with differently labeled RNA probes (fluorescein-UTP and digoxigenin-UTP), and subsequent visualization of transcripts in red and purple was obtained with INT/BCIP or NBT/BCIP (Boehringer). For histological observation, labeled embryos were embedded in gelatin/albumin and vibratome sectioned at 25 μm. Photographs were taken on a Leica microscope, scanned, and mounted for figures with Adobe Photoshop (images were corrected for brightness/contrast or cropping, but no other correction was made).

Immunohistochemistry. To compare gene expression patterns with the position of primary axon tracts, in situhybridization was combined with immunohistochemistry using a monoclonal antibody against α-acetylated tubulin (Sigma, St. Louis, MO).In situ hybridization was performed first as described above, brains were dissected out, and classical immunofluorescence staining was performed (primary antibody dilution 1:700).

RESULTS

The Xenopus orthologs of mouse Lhx1/2/5/7/9, named x-Lhx1/2/5/7/9, respectively, were isolated by degenerated or classical PCR, and their orthology relationships were assessed after alignments with ClustalX and distance method analysis (data not shown). All the sequences that we isolated fit well with the molecular phylogeny of the vertebrate LIM-hd factors (Failli et al., 2000). No divergent sequences have been found that might indicate the existence of additional members of the LIM-hd family, as a possible result of the partial genome tetraploidization that is known to have occurred in Xenopus laevis. We have also searched for additional genes with a PCR-based approach on cDNA and genomic DNA and have not been able to demonstrate the existence of specific paralogs in Xenopus, neither in the x-Lhx2/9 nor in the x-Lhx1/5 families.

For the sake of comparison with observations made in mouse mostly at embryonic day (E) 12.5 and E13.5, where neurogenesis is at its peak, regional specification is well advanced, and connectivity begins to be established in pioneering axonal pathways, we used Xenopusembryos at stages 32 and 39/40, in which similar events can be detected. The results were basically identical and therefore are presented on stage 32 embryos.

x-Lhx7 expression is well conserved

In mouse, the Lhx6/7/8 subgroup shows the most restricted brain expression among the LIM-hd family, being expressed in the medial ganglionic eminence (future pallidum) and the hypothalamus (Grigoriou et al., 1998) (Fig.1D). For comparison purposes we used the Xenopus ortholog of Lhx7. In the brain of stages 32 and 39/40 embryos, x-Lhx7 showed two main expression domains (Fig. 1A): a large, triangle-shaped domain at the ventral/anterior base of the telencephalon, just anterior to the optic recess, and a long thin band of expression in the diencephalon that followed the longitudinal axis of the brain. A small cluster of x-Lhx7-expressing cells was also consistently found in the ventroposterior part of the diencephalon (Fig. 1A,B, arrows). To establish whether these expression domains were homologous to the situation in mouse, and because distalless genes are established markers of the medial and lateral ganglionic eminences and of the telencephalic stalk region (Fig. 1D) in all studied vertebrate species, we performed double labeling withx-Dll3. The telencephalic domain of x-Lhx7expression was included into the x-Dll3 expression area and ran along the anterior border of the optic stalk (Fig.1B,C), suggesting that it could represent Xenopus medial eminence and aep/poa. MouseLhx7 is not expressed in the aep/poa (Fig.1D), but its paralog m-Lhx6 is (Lavdas et al., 1999). Thus, x-Lhx7 expression in the ventral telencephalon is identical to that of its mouse paralogs. The diencephalic expression territories of x-Lhx7 andx-Dll3 were also closely related (Fig.1D). As deduced from careful observations in toto and on sections, the most anterior part of their expression was colocalized, whereas in the dorsal aspect the thin band ofx-Lhx7 expression ran parallel to the larger band ofx-Dll3 expression (the ventral border of x-Dll3expression defines the alar/basal boundary). This x-Lhx7-positive, x-Dll3-negative domain therefore might correspond to the mammilary area. Moreover, x-Lhx7 was expressed in branchial arches and jaws (Fig.1C), just like its mouse ortholog. These observations showed that Lhx7 overall expression was well conserved between mouse and frogs and that x-Lhx7 might be considered as a marker of Xenopus medial ganglionic eminence, telencephalic stalk, and hypothalamus.

Fig. 1.

Lhx7 seems to be a general medial ganglionic eminence marker. A,B, Lateral whole-mount views of stage 32 brains labeled for x-Lhx7 expression (A) and double labeled for x-Lhx7 and x-Dll3(B). Colors for double labeling are indicated on each panel. The arrow points to the diencephalic group of cells discussed in Results. C, Transverse anteroposterior sections of an embryo double labeled forx-Lhx7 and x-Dll3. The orientation of sections is indicated by the white line inB. Arrowheads in B andC indicate the boundaries between ventral and dorsal telencephalon (vtel and dtel) and between ventral and dorsal thalamus (vt anddt), as revealed by x-Dll3 staining.e, Eye; ba, branchial arches;cg, cement gland; hyp, hypothalamus.D, Schematic color-coded (see box) comparison of Xenopus and mouse Lhx7expression patterns, with reference to x-Dll3 (to be compared with Dlx1/2/5/6 of the mouse). Note the strong conservation of patterns. For mouse, Lhx7 expression is drawn after Grigoriou et al. (1998).aep/poa, Anterior entopeduncular/preoptic area; cb, cerebellum; dt, dorsal thalamus; dtel, dorsal telencephalon;emt/spv, eminentia thalami/supraoptic paraventricular area; is, isthmus; lge, lateral ganglionic eminence; ma, mammilary area;mge, medial ganglionic eminence; mes, mesencephalon; met, metencephalon; or/os, optic stalk/recess; pt, pretectum;p1–p4, prosomeres 1–4;tec, tectum; tu, tuberal area;vt, ventral thalamus; vtel, ventral telencephalon.

x-Lhx2 and x-Lhx9 in pallial and subpallial territories of the telencephalon

Next we sought to integrate the two paralogs of the more complexLhx2/Lhx9 subgroup into this scheme. In mouse,Lhx2 labels the entire telencephalon (pallial and subpallial territories) and the hypothalamus, whereas Lhx9 expression is strictly pallial and is included in the Lhx2-expressing domain (Rétaux et al., 1999) (Fig.2H). Moreover,Lhx2 and Lhx9 are expressed in p4 and p2 prosomeres in mouse and show extensive expression throughout the pretectum and tectum.

Fig. 2.

Telencephalic expression ofLhx2 and Lhx9 shows major differences between Xenopus and mouse. A,B, Lateral whole-mount views of stage 32 brains labeled for x-Lhx2 (A) andx-Lhx9 (B) expression.C, Transverse anteroposterior sections of an embryo double labeled for x-Lhx2 (orange) andx-Lhx9 (purple). The orientation of sections is indicated by the white line inB. The arrow points to the ventral telencephalic domain of x-Lhx9 expression.D, E, Whole-mount (D) and anteroposterior transverse sections (E, orientation given by white line inD) of double labeled x-Lhx9(purple) and x-Dll3(orange) embryos. Arrows point to the ventral telencephalic domain of x-Lhx9 expression, andarrowheads indicate the boundaries between ventral and dorsal telencephalon (vtel anddtel) and between ventral and dorsal thalamus (vt and dt). F, A section through the telencephalon of an embryo double labeled forx-Lhx2 (orange) and x-Lhx7(purple). Arrows point to the nonoverlapping expression of the two genes in the basal forebrain.G, Anteroposterior transverse sections of double-labeledx-Lhx9 (orange) and x-Lhx7(purple) embryos. The arrow points to the sharp boundary between the two expression domains in the ventral telencephalon. H, Schematic color-coded (seebox) comparison of Xenopus and mouseLhx2/9 expression patterns, with reference tox-Dll3/mouse Dlx. Mouse Lhx2/9 expression is drawn after Rétaux et al. (1999). Note the major differences, especially in the telencephalon of the two species. In particular,x-Lhx9 is subpallial whereas m-Lhx9 is pallial, and x-Lhx2 does not cover the entire telencephalon whereas m-Lhx2 does. See Figure1D for abbreviations.

The two Xenopus paralogs x-Lhx2 andx-Lhx9 showed spectacular, striped-like patterns in the developing Xenopus brain (Fig.2A,B). In the telencephalon,x-Lhx9 expression domain was more restricted thanx-Lhx2, and double labeling was used to position their respective domains (Fig. 2C). The rather ventralx-Lhx9-positive domain (Fig. 2C,arrow) was included in the large x-Lhx2-positive area that covered almost the entire extent of the telencephalon. The situation was similar therefore to that found in rodents in terms of the extent of expression (broad x-Lhx2 domain, more restricted x-Lhx9 domain). However, it looked different from the mouse in terms of pallial/subpallial expression: inXenopus, x-Lhx9 was found anteroventrally, whereas it is strictly pallial in mouse. Another difference with the rodent is that x-Lhx2 did not cover the most basal part of the telencephalon (Fig. 2H, mge,aep, poa): x-Lhx7 andx-Lhx2 expression were mutually exclusive (Fig.2F, summary on Fig. 2H).

To further investigate this question, and because the functional telencephalic divisions of the developing Xenopus are poorly known and delimited, we next asked whether x-Lhx9(contrarily to its mouse ortholog) was indeed expressed in the subpallium by double labeling for x-Lhx7 (expressed in the medial ganglionic eminence) or x-Dll3 (a general ganglionic eminence, subpallial marker; see above). As observed in totoon Figure 2D and confirmed on sections in Figure2E, the x-Lhx9 telencephalic domain (Fig.2E, arrow) was included in thex-dll3 domain, suggesting that it labeled a part of the future basal ganglia. Moreover, the x-Lhx9 domain was strictly adjacent and did not overlap with the x-Lhx7 domain (Fig. 2G), suggesting that it did not label the medial eminence (future pallidum) but rather another subdivision of the subpallium. Finally, the dorsalmost telencephalic region that expressedx-Lhx2, but neither x-Lhx9 nor x-Dll3, had to be considered as pallial. Therefore, thex-Lhx2/x-Lhx9 situation was suggestive of a partial inversion of expression patterns between the two paralogs as compared with mouse. This was also supported by the fact that x-Lhx2(but not x-Lhx9) was expressed in the frog pineal gland (Fig. 2A, p), whereas the case is strictly the opposite in mouse (Rétaux et al., 1999). Nevertheless, the region of the frog basal telencephalon defined by x-Lhx7expression (mge, aep/poa) did not express a member of theLhx2/9 subgroup, which constitutes a major difference with the mouse. These results are summarized on Figure2H.

x-Lhx2 and x-Lhx9 in diencephalon and mesencephalon

x-Lhx9 and x-Lhx2 were expressed as bands with extremely sharp borders in the diencephalon and mesencephalon. Double labeling with x-Dll3, expressed in the diencephalic p3 prosomere (including the ventral thalamus) showed thatx-Lhx2/9 were expressed just anterior to p3, in p4, and just posterior to p3, in p2 (dorsal thalamus)(Fig.2D,E). The two paralogs therefore labeled p4 and p2 in frog and showed conservation of expression in most of the diencephalon as compared with mouse, including a ventral diencephalic/hypothalamic band of staining for x-Lhx2 (Fig.2A). By contrast, we again observed differences in regions posterior to p2. First, x-Lhx2 and x-Lhx9were not expressed in the pretectum (prosomere p1). This point was established by double labeling for α-acetylated tubulin to position the tract of the posterior commissure (TPC), one of the major early tracts that grows through prosomere p1, with regard tox-Lhx2/9 stripes of expression. The TPC ran just between the two x-Lhx2/9-expressing stripes of p2 and mesencephalon, respectively (see Fig. 5A). Second, in the mesencephalon itself, x-Lhx2 and x-Lhx9 were absent in the posterior tectum of the frog, as can also be defined by tubulin staining and morphological observation of the isthmus.

Fig. 5.

x-Lhx1 andx-Lhx2 expression relative to axonal tracts.A, B, Whole-mount lateral views of brains double labeled for α-acetylated tubulin and x-Lhx2(A) or x-Lhx1(B) expression. The left panelsshow dark-field micrographs of in situ hybridization. The middle panels show tubulin immunofluorescence micrographs. The right panels show combined pictures.tpc, Tract of the posterior commissure;tpoc, tract of the postoptic commissure;sot, supraoptic tract; mes, mesencephalon; met, metencephalon. Note the striking correspondence and close relationships between early axon tracts and LIM-hd expression. The tpc runs on the pretectal (p1 prosomere) band of x-Lhx1expression, just between the two bands of p2 and mesencephalic x-Lhx2 expression.

Finally, sections through the brains of embryos double labeled forx-Lhx2 and x-Lhx9 showed a perfect colocalization of the two paralogs in expressing bands of the diencephalon and mesencephalon (Fig. 2C), except in the dorsal part of p4. The results of the Xenopus/mouse comparison forLhx2/9 are recapitulated in Figure 2H.

Lhx1 and Lhx5 subgroup

Because mouse expression data for these two paralogs were not available in a precise manner, we first reexamined their expression patterns in E12.5 and E13.5 mouse brains (Fig.3A,B). Detailed data will be available elsewhere (S. Rétaux and I. Bachy, unpublished observations), and results compiling the expressions at E12.5 and E13.5 are summarized on Figure 3H(right panel). Both genes are expressed in the pallium, the hypothalamus, the ventral thalamus (p3), the zona limitans intrathalamica (zli), the pretectum, and the tectum. Interestingly, mLhx1/5 are absent in the developing basal ganglia and dorsal thalamus (p2) of the rodent.

Fig. 3.

Lhx1 andLhx5 label the zona limitans intrathalamica, and their diencephalic expression patterns are conserved between Xenopus and mouse.A–D, Whole-mount lateral views of E12.5 mouse embryos (A, B) and stage 32 Xenopusembryos (C, D) stained for m/x-Lhx1 andm/xLhx5, as indicated on panels. In A andB, arrows indicate the strong labeling in the pretectum (p1). In C, theasterisk points to the ventral telencephalic domain ofx-Lhx1 expression, and the arrows inC and D indicate the thin band of expression that surrounds the dorsal thalamus and is suggested to label the zona limitans intrathalamica (zli).E, Anteroposterior transverse sections (orientation given by white line in D) of double-labeled x-Lhx1 (orange) andx-Lhx5 (purple) embryos. Thearrow points to the thin band of expression of the two genes that envelops ventrally the dorsal thalamus (p2) and is suggested to be the zli.F, Two transverse sections through the telencephalon and diencephalon of x-Lhx5 (purple) and x-Dll3 (orange) double-labeled embryos. Arrowheads indicate the boundaries between ventral and dorsal telencephalon (vtel anddtel) and between ventral and dorsal thalamus (vt and dt). Arrow points to zli. Note that the thin band/zli is juxtaposed but not double labeled with x-Dll3. The dotted linemarks the limit between dorsal p3 (expressing x-Lhx1/5) and ventral p3/hypothalamus (expressing only x-Dll3).G, Two coronal hemisections (in the same plane of section) through the telencephalon of an embryo labeled forx-Lhx1 (left) or x-Lhx9(right) to show that the ventral telencephalicx-Lhx1 domain is included in the x-Lhx9domain. H, Schematic color-coded (seebox) comparison of Xenopus and mouseLhx1/5 expression patterns, with reference tox-Dll3/mouse Dlx. The diencephalic conservation can be opposed to the telencephalic divergences between the two species. For abbreviations, see legend to Figure1D.

x-Lhx1 and x-Lhx5 in the telencephalon

Xenopus orthologs showed rather complex expression patterns (Fig. 3C,D). Both x-Lhx1 andx-Lhx5 showed a single (but distinct) expression domain in the telencephalon. The x-Lhx1-positive area (Fig.3C,E,G,asterisks) was modest, located in the subpallium, and included in the x-Lhx9-positive region (Fig. 3G).x-Lhx5, on the other hand, labeled a large band in the telencephalon that was juxtaposed dorsally to thex-Dll3-defined subpallium (Fig. 3F). The ventral border of x-Lhx5 domain would thus follow the palliosubpallial border. The telencephalic x-Lhx5 domain did not cover the entire pallium (particularly, the dorsal aspect of the pallium was not labeled) and was included in thex-Lhx2-positive area (data not shown). x-Lhx5labeling did not cross the telencephalic/p4 border, as shown by examination of x-Lhx9/x-Lhx5 double-labeled brains and sections (Fig.4C,D). Therefore, we concluded that x-Lhx5 specifically labeled a subdivision of the pallium in Xenopus embryos (Fig. 3H, summary).

Fig. 4.

x-Lhx1/5 andx-Lhx2/9 expression domains are almost exclusive.A, B, Whole-mount brain (A) and transverse sections (B, orientation given by white line in A) of embryos double labeled for x-Lhx1(orange) and x-Lhx2(purple). The arrow inA points to the boundary between ventral diencephalicx-Lhx2 expression and the more dorsalx-Lhx1 domain. Note that except for the ventral telencephalic domain where both genes are expressed (asterisk in A) and the anterior tectum, the expression domains are exclusive. C,D, Whole-mount brain (C) and transverse sections (D, orientation given by white linein C) of embryos double labeled forx-Lhx5 (orange) and x-Lhx9(purple). Note that except for the anterior tectum, the expression domains are exclusive. The arrowsin B and D indicate the thin band ofx-Lhx1/5 expression that envelops ventrally theLhx2/9-expressing dorsal thalamus (p2). We suggest that this could represent thezona limitans intrathalamica.

x-Lhx1 and x-Lhx5 in the diencephalon and mesencephalon

In more posterior regions, x-Lhx1 and x-Lhx5colocalized in the majority of their expression domains. Both genes were expressed in prosomere p3 (ventral thalamus) (Fig. 3E), as shown with x-Dll3 double staining (Fig.3F). x-Lhx1 was expressed much higher and stopped at the dorsal boundary of the x-Dll3 domain, whereasx-Lhx5 ran more dorsally, resulting in the entire covering of dorsal p3. Both genes appeared to cross the alar/basal boundary (defined by the ventral border of the x-Dll3 diencephalic band) and to be expressed in a region of the hypothalamus (Fig.4B-D). Finally, in the ventral diencephalon the x-Lhx1/x-Lhx5 domain abutted thex-Lhx2 domain (Fig. 4B,arrow).

As shown by double labeling between the paralogs of thex-Lhx1/5 and x-Lhx2/9 families, neitherx-Lhx1 nor x-Lhx5 was expressed in the dorsal thalamus itself, but rather enveloped this structure with a strong expression in the pretectum (prosomere p1) and with a thin band of expression on the ventral and anterior side (Figs. 3, 4,arrows). Because this thin band was just dorsal tox-Dll3 but did not express x-Dll3(Fig. 3F, arrowhead), and because it was just adjacent to x-Lhx2/9 expressing dorsal thalamus (Fig.4B–D, p4, arrows), we suggest that this could represent the zona limitans intrathalamica. Moreover, the pretectal expression ofx-Lhx1/5 was established by double labeling for the TPC with α-acetylated tubulin (Fig.5B). Finally, the two paralogs were present in a thin band of the anterior mesencephalon, where they overlapped with x-Lhx2/9 expression (Fig.4B–D). In summary (Fig.3H), the mouse/frog comparison seems to indicate a good degree of conservation of expression in the pretectum, thalamus, and hypothalamus, but differences are found in the telencephalon, where the combinatorial expression of the two paralogs is notably different between the two species.

DISCUSSION

Xenopus/mouse LIM-hd expression: general conservation but telencephalic divergences

Our data suggest two general remarks. First, the current view for a common Bauplan of brain development among vertebrates fits well with our Xenopus/mouse comparison of expression patterns, which are similar (Fig. 6). The major trends are as follows: (1) Lhx7 is particularly conservative, (2)Lhx2/9 are predominant in the telencephalon and conserved in prosomeres p4 and p2, (3) x-Lhx1/5 show restricted telencephalic expression but are conserved in p3 and p1 and label thezona limitans intrathalamica, and (4) there is exclusive expression of one of the subgroups in p1–p4, with alternating expression of the paralogs of a given subgroup. Strong similarities are therefore present between mouse and Xenopus and suggest homology between most of these areas.

Fig. 6.

Schematic gene expression maps ofXenopus and mouse LIM-hd genes, with respect to subdivisions of the forebrain. A, A simplified phylogenetic tree of the LIM-hd family. The members studied in this paper are color coded, and their expression patterns inB and C are drawn in the same colors.B–D, Schematic color-coded (seeboxes) recapitulations of Xenopus and mouse LIM-hd expression patterns, with reference tox-Dll3/mouse Dlx. In B,Lhx2 (blue)/9(green) are compared with Lhx7(pink) and x-Dll3/Dlx(purple). In C,Lhx1 (red)/5(yellow) are compared with Lhx7(pink) and x-Dll3/Dlx(purple). In D, the additive expressions of Lhx1/5 in orange andLhx2/9 in turquoise are compared withLhx7 (pink) andx-Dll3/Dlx(purple). For abbreviations, see legend to Figure 1D.

Interestingly, the major differences between mouse and frog expression patterns happen to be located inside the telencephalon, the structure that is obviously the more divergent, particularly in terms of connectivity. An increasing number and complexity of forebrain connections were reached at the anamniote/amniote transition, where two important innovations emerged (Striedter, 1997): a cortical relay of thalamic information and a massive palliosubpallial projection, resulting in a higher involvement of the cortex in the processing of sensory information. The fact that four studied LIM-hd members (Lhx1/2/5/9) are expressed in various patterns in the mouse pallium, whereas only two members (one paralog of each subgroup:Lhx2 and Lhx5) are present in the amphibian pallium, might be functionally representative of the higher complexity, increased connectivity, and higher involvement of the mammalian cortex in perception, elaboration of movements, and other integrated functions.

LIM-hd genes define embryonic subdivisions inXenopus telencephalon

We suggest that LIM-hd expression allows to distinguish between telencephalic subdivisions in developingXenopus. Such subdivisions defined by gene expression become more and more precise in the developing telencephalon of mice or birds (Smith-Fernandez et al., 1998; Puelles et al., 2000). In contrast, they are poorly known in frogs and fishes, probably because of the fact that their telencephalon is small and less differentiated (Hauptmann and Gerster, 2000). We suggest that stage 32 Xenopustelencephalon is delimited by a line drawn from the optic stalk and running dorsally orthogonal to the brain axis. Inside the telencephalon, concurrent expression of LIM-hd factors and other genes such as x-Dll3 defines pallial and subpallial compartments (Fig. 6D) (Papalopulu and Kintner, 1993). Inside these, LIM-hd expression defines two pallial and three subpallial divisions.

We suggest that the x-Lhx7-expressing area corresponds to the medial ganglionic eminence. The frog pallidum is histologically poorly delineated. Only connectivity and immunohistochemical data suggest the existence of a pallidum in amphibians (Marin et al., 1998b), but GABAergic neurons have never been found (for review, see Reiner et al., 1998). Our finding of anLhx7-positive domain localized inside thedistalless-positive subpallium is an additional excellent argument in favor of the existence of this structure. The mammalian mge also expresses Lhx6 and Lhx8 (two paralogs of Lhx7; Lhx8 is probably caused by a rodent-specific duplication) (Fig. 6A) (Failli et al., 2000). We do not know whether the frog mge expresses any otherx-Lhx7 paralog. However, it is noteworthy that the mammalian mge also expresses Lhx2 and therefore presents a richer LIM-hd code. Functionally, members of the Lhx6/7/8 group might be involved in the tangential migration of GABA interneurons from the mge to the striatum and cortex in rodents (Marin et al., 2000;Anderson et al., 2001). The x-Lhx7 expression pattern might suggest that similar migrations occur in the amphibian telencephalon. In another respect, Lhx2 is strongly expressed in the proliferative zone of the rodent basal ganglia, which are hypoplasic inLhx2^−/−mice (Porter et al., 1997). The absence of x-Lhx2 in theXenopus cell-poor pallidum therefore would agree with a role for Lhx2 in cell proliferation control.

Two other subdivisions, expressing x-Lhx2 andx-Lhx1/2/9, can be delineated from LIM-hd expression inXenopus subpallium. Altogether, the three LIM-hd-deduced subpallial compartments might correspond to the three subdivisions proposed by Puelles et al. (2000) as the striatal, pallidal, and telencephalic stalk divisions of the basal forebrain. Among them, only the Lhx7-positive region can be attributed to the mge with some confidence. In the two other compartments the LIM-hd combinations are clearly different between Xenopus and mouse and might reflect the many differences in cell types and connectivity found in the basal ganglia of the two species (Reiner et al., 1998). However, we cannot exclude the possibility that cell migrations occur in frog telencephalon, as described in mouse, and could impair the interpretation of the results.

In the telencephalic pallium, four major subdivisions are found in birds and mammals: medial (hippocampus), dorsal (isocortex), lateral (olfactory cortex), and ventral (amygdala/claustrum) pallium (Puelles et al., 2000). These pallial divisions can be deduced from LIM-hd expression in mice, by comparing mediolateral extent and laminar patterns of Lhx1/2/5/9 expression (I. Bachy and S. Rétaux, unpublished observations). In Xenopus, only two pallial subdivisions were found: one expressing onlyLhx2, the other expressing Lhx2/5. Functionally, in mice Lhx2 regulates the formation of the cortical hem (Bulchand et al., 2001), and Lhx5 controls neural patterning in the hippocampus (Zhao et al., 1999), implying crucial roles for the LIM-hd family in patterning pallial subdivisions. The presence in the small Xenopus pallium of two LIM-hd-defined compartments suggests that only two distinct functional areas are found in anamniotes, when using these specific markers. Smith-Fernandez et al. (1998) observed an intermediate territory between Dlx- and Emx-positive domains in frog telencephalon, which is likely to correspond to the ventral pallium defined by Puelles et al. (2000) in amniotes. This intermediate territory does not correspond to one of our LIM-hd-defined compartments, because none of them is found in the Emx-negative, Dlx-negative portion of the frog telencephalon (our unpublished observations).

In conclusion, richer LIM-hd combinatorial expression in the mammalian pallium could reflect an enrichment in cortical connectivity. This will have to be functionally tested by overexpression experiments inXenopus, or, conversely, by analyzing in these terms the mouse lines in which LIM-hd genes have already been inactivated.

Finally, the topological relationships of the deduced telencephalic subdivisions are organized along the anteroposterior axis of the brain, with the “basal ganglia” complex in the anterior position (apparently ventral, because of the brain curvature) and the pallium in a more caudal position (apparently dorsal). This appears strongly similar to the situation shown by Smith-Fernandez et al. (1998) for the chicken telencephalic fate map.

LIM-hd expression defines conserved prosomeres inXenopus diencephalon

Prosomeres 1–4 show a remarkable and conserved alternation in expression of LIM-hd subgroups members between Xenopus and mouse. They can certainly be considered as prosomeric markers and should be taken into account for the definition of homologies in the diencephalon. Lhx1/5 label the pretectum (p1) and the ventral thalamus (p3), whereas Lhx2/9 label the dorsal thalamus and epithalamus (p2) in both species. A paralog inversion apparently occurred between Lhx2 and Lhx9 in the pineal gland. This inversion might also be true for the rest of the diencephalon, although we have no way to verify this because the expressions of the two paralogs are identical in this region. Such inversions are relatively common (Derobert et al., 2000; Zerucha and Ekker, 2000) and are representative of the conservation of function and redundancy of developmental genes. Finally, Lhx1/5conservatively label the zli, a morphological landmark that divides dorsal and ventral thalamus and probably corresponds to a forebrain organizing center (Braun et al., 2000; Garda and Martinez, 2000). Overall, the conserved alternating, stripe-like, and exclusive expression of the Lhx1/5 and Lhx2/9 subgroups in the diencephalon suggests two remarks. First, this could imply negative interactions in the regulatory sequences of their promoters, which have been well conserved in tetrapods and would be interesting to analyze. Because the two subgroups are coexpressed in some regions of the telencephalon and posterior brain, it also suggests that different regulatory modules are used to promote their expression in different areas. Second, the diencephalic conservation of the LIM-hd code fits well with the fact that the pretectum, epithalamus, and thalamus are conservative features of vertebrate brains (Butler and Hodos, 1996). In particular, a lemnothalamus (receiving direct sensory inputs) and a collothalamus (receiving sensory inputs through tectal relay) can be distinguished in both amniotes and anamniotes. However, a collothalamic projection to the pallium is a new feature in amniotes, but the emergence of this major new pathway is not correlated with any variation in LIM-hd gene expression between frogs and mice.

Concluding remarks

Among other functions in brain development, LIM-hd family members work through interactions with LIM-specific cofactors to govern pathfinding events: LIM-hd combinatorial expression determines the topography of motorneurons axonal projections, and the genetic manipulation of the LIM-hd code results in predictable changes in their projections (Sharma et al., 1998; Thor et al., 1999; Kania et al., 2000). It is highly possible that a similar role is played in the forebrain, through recruitment of similar genetic cascades. In this respect, the detailed comparison of expression patterns of several LIM-hd members between species known to present well characterized differences in brain connectivity and neuronal types is useful in elaborating functional hypotheses. In summary, we show a degree of conservation of LIM-hd expression between Xenopus and mouse that strengthens the idea of conservation of brain patterning through vertebrate evolution. Interestingly, the divergences of expression observed in the telencephalon can be correlated with the emergence of new neuronal circuits that occurred at the anamniote/amniote transition. Additional anatomical studies on intermediate species (birds and reptiles) and functional analysis are needed to further analyze the role of the LIM-hd family in vertebrate brain development and evolution.