ABSTRACT
We have isolated a new murine homeobox gene, Rpx (for Rathke’s pouch homeobox), that is dynamically expressed in the prospective cephalic region of the embryo during gastrulation. Early expression is seen in the anterior midline endoderm and prechordal plate precursor. Expression is subsequently activated in the overlying ectoderm of the cephalic neural plate, suggesting that inductive contact with Rpx-expressing mesendoderm is required for this expression. Subsequently, Rpx expression is extinguished in the mesendoderm while remaining in the prospective prosencephalic region of the neural plate ectoderm. Ultimately, transcripts become restricted to Rathke’s pouch, the primordium of the pituitary, which is known to be derived from the most anterior ectoderm of the early embryo. Down regulation of Rpx in the pouch coincides with the differentiation of pituitary-specific cell types. Rpx is the earliest known marker for the pituitary primordium, suggestive of a role in the early determination or differentiation of the pituitary. Since Rpx is expressed so dynamically and so early in the anterior region of the embryo, and since its early expression domain is much more extensive than the region fated to form the pituitary, it is likely that Rpx is involved in the initial determination of the anterior (prechordal) region of the embryo.
INTRODUCTION
In recent years, the elucidation of the genetic program involved in vertebrate embryonic patterning has been initiated, in part, through the study of homeobox genes. These genes, originally identified as members of the homeotic gene clusters in Drosophila (HOM-C), occupy key positions in the regulatory gene hierarchy responsible for defining the embryonic body plan in the fly. It has become clear that members of the vertebrate Hox homeobox gene family are structurally and functionally equivalent to their HOM-C counterparts, and play analogous roles in defining positional identity along the rostrocaudal axis (reviewed by McGinnis and Krumlauf, 1992). In mice, both Hox loss-of-function and gain-of-function result in defects in morphogenesis and, in some cases, apparent homeotic transformation (for example see: Kessel et al., 1990, Ramirez-Solis et al., 1993). These genes do not provide spatial cues along the entire axis however, since they are not detectably expressed anterior to the hindbrain. Accordingly, the identification of regulatory genes that control the determination and specification of anterior structures of the vertebrate embryo, including the forebrain, is of fundamental importance, particularly since the forebrain is determined in the absence of the inductive influences of the notochord.
Recently, murine counterparts of several homeobox genes, expressed in the head structures in Drosophila, have been characterized that are promising candidates for cephalic regulatory genes. These include members of the Distal-less (Dlx) (Price et al., 1991; Robinson et al., 1991, 1994; Porteus et al., 1991; Simeone et al., 1994) and Nkx (Price et al., 1992) gene families, which are expressed prominently in the head and forebrain and show homology to the Distal-less (Dll) and NK genes of Drosophila. Murine relatives of the Drosophila orthodenticle (otd) and empty spiracles (ems) genes (Otx1, Otx2 and Emx1, Emx2, respectively) display nested patterns of expression in the developing rostral central nervous system (CNS; Simeone et al., 1992, 1993). Since it has been established that otd and ems participate in a regulatory network required for head formation in Drosophila that is distinct from the HOM-C regulatory network at work in the embryonic trunk, it is highly likely that the conserved murine genes perform similar functions in the morphogenesis of the head and rostral CNS (reviewed by Finkelstein and Boncinelli, 1994).
While intriguing parallels exist between the Drosophila head segmentation genes and their vertebrate counterparts, with the exception of Otx2, which is expressed in the epiblast and becomes restricted to the anterior region of the embryo during gastrulation, the expression of the vertebrate genes is detected relatively late in embryogenesis (i.e. after 8.5 days p.c.). Consequently, while these genes, like the Hox genes, may play a role in the specification or elaboration of regional identity, they are probably not involved in the very early determinative events which initiate patterning during gastrulation.
Several candidate vertebrate genes have been isolated that may play a role in these early events. These include putative transcription factors such as Brachyury (T) (Herrmann, 1992), the fork head genes XFKH1 (Dirksen and Jamrich, 1992) and HNF-3β (Sasaki and Hogan, 1994; Ang et al., 1993), and the homeobox genes goosecoid (gsc) (Cho et al., 1991; Blum et al., 1992) and lim-1 (Taira et al., 1992; Barnes et al., 1994; Fujii et al., 1994). Signalling molecules that appear to be directly involved in early inductive events, such as noggin (Smith and Harland, 1992), nodal (Zhou et al., 1993) and sonic hedgehog (Echelard et al., 1993), have also been identified. These genes are expressed in the Spemann’s organizer region in the blastopore lip of Xenopus embryos, and in the comparable region in the node of murine embryos. Subsequent expression in the chordamesoderm of the notochord and/or prechordal plate, structures known to be required for both anteroposterior and dorsoventral patterning of the embryonic CNS, further implicates these genes as important developmental regulators.
During a screen for novel homeobox genes that are expressed during gastrulation, we identified a gene that is restricted in its expression to the anterior region of the mouse embryo. Early in gastrulation its expression is detected in the anterior endoderm and prechordal plate precursor, followed by activation in the overlying cephalic neural plate. Throughout early postimplantation development, this gene displays a very limited expression domain in the anterior neuroectoderm of the embryo, initially encompassing the prechordal area and prospective prosencephalon, and ultimately becoming restricted to Rathke’s pouch, the primordium of the pituitary. This gene encodes a homeodomain that is similar to that of the prd gene of Drosophila and has been called Rpx for Rathke’s pouch homeobox gene. We propose that Rpx is involved in the determination of the pituitary gland and that it may participate in establishing an early embryonic field in the anterior prechordal neural plate of the embryo that is later subdivided into smaller developmental units during organogenesis.
MATERIALS AND METHODS
Animals and embryos
Mice used were FVB/N or C57B/6J. Embryos were obtained from superovulated females. Noon of the day when the vaginal plug appeared was considered 0.5 days post coitum (p.c.).
Embryoid body formation
Embryoid bodies were generated by aggregation of D3 (Doetschman et al., 1985) embryonic stem (ES) cells according to Robertson (1987). Samples of aggregates were removed from continuous culture at various time points and total RNA was isolated.
PCR amplification of homeobox sequences
Template for PCR reactions was phage DNA prepared from a 7.5 day p.c. mouse embryo cDNA library (kindly provided by J. Rossant). PCR reactions were carried out according to Mackem and Mahon (1991). Degenerate primers that were derived from the predicted N and C terminus of the Xenopus Mix-1 homeodomain (Rosa, 1989) were synthesized with linkers containing restriction sites (XhoI, SalI, ClaI). The sequences were as follows (IUPAC code): TFFTQAQLD (5′-GACTCGAGTCGACATCGATACNARGCNCARCTNGA-3′) and IQVWFQN (5′-GACTCGAGTCGACATCGATTYTGRAACCANACYTGDAT-3′). Purified reaction products of 140 bp were cloned into the SalI site of pGEM3Z (Promega), and the inserts were sequenced (Sanger et al., 1977). The insert from pMix7d1 contained the Rpx homeobox. A longer cDNA sequence was isolated by PCR from the library using a non-degenerate oligo homologous to a region from the homeobox in combination with a primer specific for the λgt10 phage. This clone, pPrd3, contained 400 bp of sequence from the 3′ end of the gene extending from the middle of the homeobox to the putative poly(A) tail.
RNA isolation and analysis
Total RNA was prepared with RNAzol (Tel-Test, Inc.) according to the manufacturer’s instructions. Poly(A) RNA was prepared using POLY(A)QUIK push columns (Stratagene). Northern blots of total RNA (50 μg/lane) were prepared according to standard methods and were hybridized to random-primed 32P-labeled probes (Feinberg and Vogelstein, 1983) in Rapid-hyb buffer (Amersham Life Science) at 65°C. The probe used was a 538 bp cDNA fragment (nucleotides 604-1142). The sizes of the Rpx transcripts were determined by comparison to RNA markers of known size (BRL). Blots were stripped and reprobed with rat glyceraldehyde-3-phosphate dehydrogenase (GADPH) or actin to control for loading. Northern blots of embryoid body RNA were also reprobed with Oct-3 (700 bp cDNA sequence generated by PCR) and Endo-B (pUC9B7, Singer et al., 1986) labeled restriction fragments.
Isolation of Rpx cDNA clones
Three oligo(dT)-primed λgt10 mouse embryo cDNA libraries representing different stages of development, 7.5 days p.c., 8.5 days p.c. (Fahrner et al. 1987) and 12.5 days p.c. (Joyner and Martin, 1987), were screened using standard methods (Sambrook et al., 1989) with either the 140 bp insert from pMix7d1 or the 400 bp insert from pPrd3 as probes. One Rpx cDNA was obtained from each library. A mixed-primed cDNA library was constructed using poly(A) RNA prepared from 8.5- to 9-day old embryos. First strand synthesis was primed with either random oligos or with oligo(dT). After second-strand synthesis, both sets of reaction products were ligated to EcoRI adapters, combined at a ratio of 2:1 (random:oligo(dT) primed), and cloned into λZAPII (Stratagene). Two positive clones were identified after screening with a 2.5 kb genomic DNA fragment overlapping with, and extending 5′ to, the known cDNA sequence. cDNA clone inserts were subcloned into plasmid Bluescript KS II and sequenced.
RACE PCR
The 5′ ends of the Rpx mRNAs were obtained by the modified RACE method termed SLIC (Edwards et al. 1991) using the 5′-AmpliFINDER RACE and the PCR Optimizer kits (Clontech). First strand cDNA was made by random priming 2 μg of poly(A)-enriched RNA isolated from 8.5- to 9-day old embryos. The single-stranded anchor oligonucleotide was ligated to the 3′ end of the cDNA by T4 RNA ligase and amplified using a gene-specific primer as partner. This PCR product was further amplified with a nested Rpx-specific primer. Southern analysis revealed two bands hybridizing to the 5′ coding end of Rpx which were purified and cloned into pCR vector (Invitrogen). Seven independent clones were sequenced, all of them overlapped with the Rpx cDNA sequence.
In situ hybridization
In situ hybridization with 35S-labeled riboprobes was performed as previously described (Mackem and Mahon, 1991). Serial sections from several embryos for each stage were examined. Labeled sense and antisense riboprobes were synthesized from pPrd3 (Rpx), pSK75 (Brachyury or T) (Hermann, 1992), p3 (gsc) (Blum et al., 1992), and pmPit1 (Pit-1) (Camper et al., 1990). Sense control probes in all cases gave background labeling. Exposure to NTB-2 (Kodak) emulsion varied from 2-4 days for Pit-1 and T, and 2.5 to 3 weeks for Rpx hybrids. After development, the slides were photographed in dark-field optics, stained in Geimsa and rephotographed in bright-field optics. In some cases, slides were stained in Hoechst and photographed under simultaneous dark-field and UV illumination according to the method of Sundin et al. (1990).
RESULTS
Isolation of cDNA clones containing Rpx
A PCR-based screen was utilized to isolate novel homeobox genes expressed during gastrulation. Using a 7.5 day p.c. cDNA library as template with primers specific for the Xenopus homeobox gene Mix-1, one clone (pMix7d1) containing the novel Rpx homeobox sequence was isolated. A longer clone (pPrd3), containing sequences extending 3′ to the poly(A) tail was obtained by PCR from the cDNA library and was used to screen cDNA libraries for full-length clones.
A total of five overlapping cDNA clones were isolated from three different mouse embryo oligo(dT)-primed cDNA libraries and from one mixed random-primed/oligo(dT)-primed cDNA library. The sequence of the longest Rpx clone contained a single open reading frame that would encode a predicted polypeptide of 185 amino acids. The nucleotide sequence and conceptual translation deduced from this and other overlapping cDNA clones, as well as 5′ RACE analysis, is shown in Fig. 1. This cDNA represents the full length or near full length transcript since its size is close to that of the predominant message (910 nt) seen on northern blots (see below and Fig. 3), taking into account the addition of poly(A) tails. In addition, all three reading frames are terminated upstream of the putative initiation codon.
A RACE PCR amplification procedure was used to verify the position of the 5′ end of the message. Two distinct products were obtained. The major product terminates several bp from the 5′ end of the cDNA, supporting the contention that this cDNA represents the entire smaller mRNA transcript. The other RACE-PCR product ends 380 bp upstream of the putative AUG codon (1123 bp from the poly(A) signal), consistent with the predicted size of the less abundant, longer transcript visualized on northern blots. Both start sites are indicated in Fig. 1.
The putative amino acid sequence of Rpx contains a homeodomain near the C-terminal end which is similar to the paired (prd) class of homeodomain sequences of Drosophila (Bopp et al. 1986). Comparison with other known homeodomain proteins reveals that the Rpx homeodomain belongs to a subset of prd-like homeodomains (Fig. 2A) that differ from canonical prd homeoproteins in two respects. First, the amino acid at position 9 of the recognition helix is a glutamine instead of a serine. This residue is known to be critical for the sequence specificity of DNA binding (Treisman et al., 1989). Second, Rpx lacks a paired box, another conserved DNA binding domain typically found N-terminal to the homeodomain in prd family members (Walther et al., 1991). In these respects, Rpx is more related to the Xenopus genes, Mix-1 (Rosa, 1989) and Xanf1 (Zaraisky et al., 1992); the mouse genes, Chx10 (Liu et al., 1994), S8 (Opstelten et al., 1991), MHox (Cserjesi et al., 1992), Phox2 (Valarché et al., 1993), and Cart-1 (Zhao et al., 1993); the Drosophila gene, aristaless (al) (Schneitz et al., 1993), and the C. elegans gene, ceh-10 (Hawkins and McGhee, 1990). With the exception of Xanf1, which may be the Xenopus ortholog of Rpx, the homeodomains encoded by these genes are not extremely conserved and share only 50-55% homology.
The predicted Rpx protein also contains the octapeptide sequence present in some, but not all, prd-like homeodomain protein sequences. Interestingly, The Rpx octapeptide (FSIESILG) is more similar to the octapeptide domains of four non-prd-like homeobox genes (Noll, 1993) and the recently described Chx10 (Liu et al., 1994) than to those of canonical prd-domain genes (consensus HSIDGILG) (Noll, 1993).
The homeodomain sequences of Rpx and Xanf1 are 80% identical, and sequence comparison across the entire coding regions reveals additional regions of homology concentrated near the N terminus around the initiator methionine and the octapeptide sequence (Fig. 2B). These regions potentially encode domains involved in protein:protein interaction, such as transactivation or dimerization. The extended homology overlapping the octapeptide sequence in particular is rich in serine:threonine residues (9/38 residues). In addition, the relative position of these motifs within the coding regions and the size of the predicted proteins are practically identical between the two genes. Overall, the Rpx and Xanf1 proteins show 48.1% sequence identity and 70.8% similarity, taking into account conservative changes. Although the homology is not extremely high, it is possible that Rpx is the ortholog of Xanf1 (see Discussion).
Developmental profile of Rpx expression
Northern blot analysis of total RNA isolated from microdis-sected embryos was performed to ascertain when Rpx was expressed during embryogenesis. It was found that at 8.5 days p.c., the earliest time point analysed, there were at least two hybridizing bands: a predominant broad band of approximately 910 bases, and a minor species of 1.2 kb (Fig. 3). At later stages of development, only the smaller species is detected. At this level of analysis, transcripts were only detected up to 9.5 days p.c. and were specifically localized to the head but not the body.
These data indicated that Rpx was expressed early in postimplantation development. Since it was not technically feasible to isolate significant quantities of RNA earlier than day 8, northern analysis was performed on differentiating embryoid bodies to gain further insight into the developmental profile of Rpx expression (Fig. 4). Embryoid bodies are aggregates of ES cells that differentiate when cultured in the absence of feeder cells and the differentiation inhibitor, LIF. Differentiating embryoid bodies are believed to represent a good in vitro model system for the study of preimplantation and early postimplantation development (Doetschman et al., 1985). After one day postaggregation, Rpx expression was virtually undetectable. Expression was progressively activated after 2-3 days, but was dramatically upregulated at 8 days to levels roughly comparable to those seen in 8.5 day p.c. embryos.
Since the kinetics of in vitro differentiation attained by embryoid body aggregates can vary slightly depending on the passage number of the ES cells and other culture variables, the degree of differentiation was independently assessed by reprobing northern blots for expression of the POU domain gene Oct-3 (Rosner et al., 1990; Scholer et al., 1990) and the Endo B cytokeratin gene (Singer et al., 1986), marker genes for undifferentiated and differentiating cells, respectively. Rpx is strongly upregulated as Oct-3 begins to be downregulated, and after Endo B is activated. In vivo, Oct-3 is expressed in preimplantation stage embryos and in the epiblast of postimplantation stage embryos, and is turned off during gastrulation as cells enter the primitive streak (Rosner et al., 1990; Scholer et al., 1990). By comparison, the hybridization data suggests that Rpx is activated at a stage comparable to gastrulation in these embryoid body cultures.
In situ hybridization analysis of Rpx expression during gastrulation and neurulation
The spatial pattern of expression of Rpx was examined by in situ hybridization to tissue sections of mouse embryos. Whole-mount in situ hybridization was also performed but, due to the low abundance of Rpx transcripts, was less reliable than hybridization of radiolabeled probes to tissue sections. Hybridization of Rpx probes to pre-gastrula stage embryos was negative, although transcripts could be detected by northern blot in undifferentiated ES cells (data not shown). Since ES cells are thought to resemble undifferentiated epiblast cells, it is conceivable that levels of transcripts below the limits of detection are present in these early embryos. Rpx transcripts were first detected by in situ hybridization at gastrulation (Figs 5, 6). At 7 days p.c., expression was detected in the endoderm/prechordal plate underlying the prospective cephalic neural plate (Figs 5A-D, 6). A few hours later (7.5 days p.c.), expression was detected in the ectoderm juxtaposed to the endoderm/prechordal plate, which will become the cephalic neural plate (Figs 5E,F, 6G,I, 7). This pattern of expression suggests that Rpx may be induced in the anterior neural plate by the adjacent Rpx-expressing mesendoderm of the prechordal plate. By the early head fold stage, Rpx expression has been extinguished in the underlying mesendodermal cells, and is limited to the anterior neuroectoderm (Fig. 7C,D), underscoring the dynamic nature of Rpx expression during this important period.
These data indicated that Rpx was expressed very early in an anterior-restricted domain along the developing embryonic axis. In order to define more precisely the temporal and spatial domain of Rpx expression, a comparative analysis of Rpx expression with that of Brachyury (T) (Fig. 6) and the homeobox gene goosecoid (gsc) was performed. The T gene is required for the proper development of the notochord, and has been shown to be strongly activated along the entire primitive streak at gastrulation and, at later stages, in the developing notochord (Herrmann, 1992). The murine gsc gene has been shown to be transiently expressed during gastrulation in the anterior-most region of the node (Blum et al., 1992) and, in other vertebrate species, in the prechordal plate (Izpisua-Belmonte et al., 1993; Steinbeisser and De Robertis, 1993).
During early gastrulation, 6.5-6.75 days p.c., Rpx expression was seen in a very restricted region of the endoderm, which in cross section was directly across from the T expression domain in the primitive streak. At the earliest stage examined (6.5 days; Fig. 6A,B), just as cells begin to delaminate from the primitive ectoderm to form mesoderm, T expression was visualized in the primitive ectoderm of the prospective primitive streak. The Rpx-expressing region (Fig. 6A) in the endoderm layer was clearly anterior – proximal to both T (Fig. 6B) and gsc (data not shown) expression domains in the ectoderm and, by inference from later expression patterns, may represent precursor cells of the prechordal plate or head process. Attempts to confirm this by colocalization with gsc transcripts, which are reported to be present in prechordal plate in both frog (Steinbeisser and De Robertis, 1993) and chick (Izpisua-Belmonte et al., 1993) were not definitive, although hybridization slightly over background was seen in this region with gsc probes (data not shown). It may be that in the mouse, gsc is not expressed at very high levels in this region. It was clear, however, that the Rpx hybridization domain was anterior/proximal to the prominent gsc expressing region of the node. Alternatively, it is possible that Rpx expression at this early stage was in the primitive visceral embryonic endoderm, a cell layer that is essentially replaced by definitive endoderm during the process of gastrulation, and as such presages the prechordal plate primordium.
By 7-7.5 days p.c., Rpx was clearly expressed in the mesendoderm of the prechordal plate primordium (Fig. 6C,D), which has a raised, columnar appearance at the midline (Tam et al., 1982). Although at 7 days p.c. the Rpx-expressing region was distinctly separated from the domain of T expression in the primitive streak and notochordal plate, by 7.5 days, the Rpx expression domain in the prechordal plate directly abutted the T expression domain of the notochordal plate (Fig. 6E-J). By this stage, the Rpx gene was activated in the ectoderm overlying the prechordal plate (Figs 6G,I, 7A,B).
By 8-8.5 days p.c., Rpx expression was limited to the neuroectoderm of the prospective prosencephalon (Fig. 8). In 8.5 day embryos, expression was also detected in the oral ectoderm and in a very small region of the foregut endoderm that briefly lies in contact with the oral ectoderm at this stage of development (Fig. 8C,D). Expression in the foregut endoderm was extremely transient and, because it was seen only in an area of direct contact with ectoderm, may result from inductive signals emanating from the juxtaposed Rpx-expressing ectoderm.
Expression in the developing pituitary
Dramatic restriction of Rpx expression was seen at 9 days p.c. Expression was no longer detectable in the neuroectoderm, but was exclusively restricted to the layer of ectodermal cells that will give rise to Rathke’s pouch, the primordium of the adenohypophysis, which ultimately becomes the anterior and intermediate lobes of the pituitary (Fig. 9A,B). Rathke’s pouch originates from a cohort of cells within the oral ectoderm that come to lie in contact with the floor of the diencephalon (Schwind, 1928). Classical embryological experiments in several vertebrate species suggest that reciprocal inductive interactions between the embryonic brain and the pituitary primordium in this region of contact are crucial for the development and proper differentiation of the adenohypophysis (derived from Rathke’s pouch), and of the median eminence of the hypothalamus (Pehlemann, 1962; Etkin, 1967; Wantanabe, 1982a,b; Daikoku et al., 1982, 1983). Subsequently, the neurohypophysis, derived from an outpocketing of the diencephalic neuroectoderm adjacent to the pouch, gives rise to the posterior or neural lobe of the pituitary, which contains the axon terminals of neurosecretory neurons which reside in the hypothalamus. Thus, the developmental programs of these essential components of the hypothalamic-pituitary axis are interdependent.
The cells of Rathke’s pouch primordium invaginate and eventually detach from the oral ectoderm to form a definitive pouch structure by 11.5 days p.c. Rpx is strongly expressed in virtually all of the cells of the pouch at this stage (Fig. 9C,D). Around 12.5-13 days, definitive pituitary cell types begin to differentiate in the developing pouch in a stereotypical spatial and temporal order (reviewed in Simmons et al., 1990; Japón et al., 1994), beginning with corticotrophs that appear in the ventral pouch at 12.5 days and thyrotrophs in the rostral tip at 13.5 days, followed by melanotrophs at 14.5 days, and somatotroph, lactotrophs, and gonadotrophs at 15.5-16 days p.c. Rpx was found to be downregulated spatially and temporally co-incident with this differentiation, such that by 15.5 days, expression was no longer detectable (Fig. 9E,F).
The progressive downregulation of Rpx occurs in a distinct ventral to dorsal direction (Fig. 10), mirroring the general direction of differentiation of the pouch, beginning at 12.5-12.75 days within the cells of the so-called rostral tip (Fig. 10A), which produce a transient population of thyrotroph cells early in development (Lin et al., 1994). Rpx transcripts disappear from the entire ventral portion that will give rise to the anterior lobe by 13.5 days (Fig. 10C,D). At 14.5 days p.c., only sparse labeling was seen in the dorsal part of the pouch (the incipient intermediate lobe), particularly in areas in contact with the neuroectoderm of the diencephalon or posterior lobe as well as in cells lining the lumen of the pouch (Fig. 10E).
The extinction of Rpx expression in the anterior lobe coincided temporally and spatially with the activation of the pituitary-specific POU domain transcription factor Pit-1 (Fig. 10D), which has been shown to occur at 13.5 and 15 days of embryogenesis in the mouse and rat, respectively (Dollé et al., 1990; Simmons et al., 1990). Pit-1 is required for the expression of growth hormone (GH), prolactin (PRL) (Ingraham et al., 1990; Fox et al., 1990; Castrillo et al., 1991), and β-thyroid stimulating hormone (β-TSH) genes (Lin et al., 1994). In addition, Pit-1 plays a poorly understood role in the proliferation and/or maintenance of pituitary cell types in development, for the Snell dwarf (dw) mutant mouse, which lacks Pit-1 gene function (Li et al., 1990; Camper et al., 1990), has a severely hypoplastic pituitary that lacks three pituitary cell lineages – thyrotroph, somatotroph, and lactotroph (Slaughbaugh et al., 1981).
Comparative in situ hybridization of neighboring sections revealed that Pit-1 expression was first detectable in the ventral region of the developing anterior pituitary where Rpx expression had turned off at 13.5 days (Fig. 10D). By 14.5 days p.c., Pit-1 was almost uniformly expressed in the ventral anterior lobe, where Rpx expression was no longer detectable (Fig. 10F). These reciprocal expression domains may indicate potential regulatory relationships between Rpx and Pit-1.
DISCUSSION
We report here the isolation and characterization of a new murine homeobox gene that is expressed dynamically during gastrulation in the future cephalic region of the embryo. Because this gene is ultimately restricted in its expression to Rathke’s pouch, the precursor of the mammalian pituitary gland, the gene has been called Rpx for Rathke’s pouch homeobox. Rpx is a member of a subfamily of homeobox genes related to the prd homeobox gene of Drosophila that differ most notably in that they lack the paired box found in Drosophila prd, gooseberry and vertebrate Pax genes.
Expression of Rpx is initially detected during gastrulation in mouse embryos in the midline endoderm/mesendoderm that forms the prechordal plate precursor. The prechordal plate presumably is involved in the induction of anterior head structures, including the forebrain, and prechordal plate precursor cells have been shown to possess neural inducing activity in chick embryos (Izpisua-Belmonte et al., 1993). Other potential developmental regulatory genes have been reported to be expressed in the prechordal plate, including lim-1 (Barnes et al., 1994) and Otx2 (Simeone et al., 1993; Ang et al., 1994), HNF-3β (Sasaki and Hogan, 1993; Ang and Rossant, 1993; Weinstein et al., 1994) and gsc (Izpisua-Belmonte et al., 1993). However, Rpx is, to our knowledge, the only gene that is exclusively expressed in this region at 6.5-7 days. Most of these other genes are also simultaneously expressed, often at higher levels, in the notochord and primitive streak (lim-1, HNF-3β) or in the ectoderm (Otx2). Gsc is expressed in the prechordal plate in both chick and frog embryos, but our attempts to colocalize gsc with Rpx transcripts indicate that the expression of gsc in this region in the mouse is extremely low. The zebrafish gene hlx-1 is expressed in the early prechordal plate (Fjose et al., 1994), but expression of the apparent mouse ortholog (Dbx) has not been reported there (Lu et al., 1992).
The amino acid sequence of Rpx is most similar to that of Xenopus Xanf1, a homeobox gene reported to be expressed in the anterior neural folds of the frog embryo (Zaraisky et al., 1992). These two genes may be orthologs, for they show 48.1% total homology and 80% homology in the homeodomain. Since the homology is not extremely high relative to other orthologous homeobox genes, it is also possible that there may be another gene that is more similar. However, it has been found that Xanf1 and a very similar Xenopus gene, Xanf2 are both expressed in anterior neural plate and in the pituitary primordium in the frog (Mathers et al., 1995).
Several features of the Rpx expression pattern suggest that it is responsive to inductive interactions. At approximately 7.5 days p.c., corresponding to the neural plate stage, transcripts of Rpx can first be detected in the rostral ectoderm overlying the Rpx-expressing prechordal plate in a manner suggesting that Rpx expression in the ectoderm is induced by signals emanating from the Rpx-expressing cells in contact with it. It has, for example, been shown that grafted gsc-expressing cells are capable of inducing gsc expression in host tissues (Izpisua-Belmonte et al., 1993). This can be experimentally addressed in the mouse through the use of recombinant tissue explants of gastrulation stage embryos (Ang and Rossant, 1993).
By extrapolation from the limited fate map available for the gastrulating mouse embryo, the Rpx expression domain in the cephalic neural plate of the 7.5 day embryo is limited to the prospective prosencephalic region and perhaps part of the mesencephalon (Tam, 1989; Tam and Beddington, 1992). Certainly by 8.5 days, expression is restricted to the prosencephalon, and serves as the most anterior restricted marker for this region. Rpx expression is also detected in the oral ectoderm at this stage and in a region of the rostral foregut endoderm that lies in contact with it, again suggestive of inductive interactions.
Restriction of expression to Rathke’s pouch occurs at 9-9.5 days p.c. Fate mapping in chick (Couly and Le Douarin, 1988), frog (Eagleson and Harris, 1989; Kawamura and Kikuyama, 1992), and mouse (Osumi-Yamashita et al., 1994) has shown that the primordium of the adenohypophysis is derived from the anterior ridge of the neural plate. This assignment is consistent with the Rpx expression domain, which becomes progressively anteriorly restricted within the cephalic neural plate during development. Alternatively, expression of Rpx may be biphasic, with the early prechordal plate/neural plate expression being distinct mechanistically from the later expression in the pouch.
The adenohypophysis does not form in chick embryos after surgical removal of the anterior neural ridge, suggestive of commitment to pituitary development prior to neural tube closure (elAmraoui and Dubois, 1993). Rpx is the earliest known gene expressed in the ridge and pituitary primordium, suggesting involvement in the early determination or differentiation of this organ. Its expression precedes that of the α-glycoprotein subunit gene (αGSU), the first hormone gene reported to be expressed in the developing pituitary. αGSU is expressed in the placode that gives rise to Rathke’s pouch by day 11 in the rat (10.5 days p.c. in mouse), and is thought to be a marker for most or all of the progenitor cells of the pituitary (Simmons et al., 1990). Most obviously, the Rpx gene could be required for the proper differentiation of the adenohypophysis. Alternatively or in addition, Rpx expression could be crucial for the development of other tissues known to interact with the pouch via cell-cell contact, such as the median eminence of the hypothalamus and associated posterior lobe of the pituitary, by regulating the expression of signaling molecules in the pouch.
Rpx expression in the developing pituitary is transient and is extinguished as differentiated pituitary cell types appear. In the anterior lobe, this extinction coincides temporally and spatially with the transcriptional activation of the Pit-1 gene, raising the possibility that Pit-1, or one of its regulators, may negatively control Rpx expression in this part of the developing gland. This possibility is being investigated by the analysis of the transcriptional regulation of the Rpx gene.
The role Rpx plays in development is currently under analysis through the generation of both gain-of-function and loss-of-function mouse models. However, since the expression of Rpx is detected several days before the formation of a definitive Rathke’s pouch, and since the initial Rpx expression domain encompasses a much larger region than that fated to form Rathke’s pouch, it seems likely that Rpx plays a role in the early establishment of the anterior region of the embryo. Rpx may be instrumental in defining a larger embryonic field that is eventually divided into smaller developmental units through the interaction with other developmental control genes. In this respect, it is interesting that the Otx2 expression domain overlaps with that of the Rpx gene in the anterior neural plate at early stages, and it has been shown that this expression depends on signals from the anterior mesendoderm (Ang et al., 1994). Rpx potentially may regulate, or be regulated by, these signals. As development proceeds and Rpx expression becomes more anterior-restricted to Rathke’s pouch, Otx2 expression regresses from its most anterior extent, and is localized to discrete regions of the forebrain (Simeone et al., 1992), raising the possibility that Rpx and Otx2 may be part of an interactive regulatory gene network involved in defining anterior cell fates.
ACKNOWLEDGEMENTS
We thankfully acknowledge our colleague Milan Jamrich for help with initial PCR reactions and for fruitful scientific discussions and review of this manuscript. We thank Janet Rossant, Brigid Hogan and Gail Martin for the generous gifts of cDNA libraries and Sally Camper, Bernard Herrmann and Eddy de Robertis for sharing plasmids for Pit-1, T and gsc, respectively.
Note added in proof
While this manuscript was in press, it came to our attention that the sequence of Rpx is the same as that of the Hesx1 gene recently reported by Thomas et al., J. Biol. Chem. 270, 3869-3875 (1995).