By use of in situ hybridization experiments, the mouse homeobox genes Hox-3.1 and Hox-1.5 are compared in the temporal and spatial patterns of their embryonic transcripts. Transcripts of both genes are first detected at about days, although the appearance of Hox-3.1 transcripts apparently follows Hox-1.5 after a small delay. Hox-3.1 and Hox-1.5 transcripts occupy domains which are different, although overlapping, along the anteroposterior axis of the embryo. The domains are first established within the ectoderm and mesoderm germ layers at days, but subsequently they persist within the nervous system, the prevertebral column and within at least some of the organs (the thyroid, lung, stomach, mesonephric and metanephric kidneys) at days. In discussion, two different mechanisms are considered by which positional information might first be generated within the germ layers.
We found earlier (Gaunt et al. 1986) that transcripts of the mouse homeobox gene Hox-1.5 (Mo-10) were restricted to posterior parts of the ectoderm and mesoderm germ layers in -day embryos. Restriction to posterior ectodermal structures (posterior nervous tissue) and posterior mesodermal structures (such as mesonephric tubules and somites) persisted during organogenesis at 9| days. In view of these results, and also of the part known to be played by homeobox genes in pattern formation during Drosophila embryogenesis (e.g. Gehring, 1987), we suggested that expression of the Hox-1.5 gene might provide a positional cue during the development of ectoderm- and mesoderm-derived structures in the mouse. This suggestion was supported by the fact that the transcript domain for Hox-1.5 differed spatially, at least within the nervous system, from that already described for Hox-3.1 (Mo-EA, Awgulewitsch et al. 1986).
From the above and more recent studies, it is becoming apparent that mouse homeobox gene transcripts occupy different, although overlapping, spatial domains in both the nervous system and prevertebrae of midgestation embryos. In these tissues, the homeobox genes characterized have transcript domains which overlap posteriorly, but which extend to different anterior limits. Within the central nervous system, these anterior limits are in the anterior (preotic) myelencephalon (Hox-1.5:Gaunt et al. 1986; Gaunt, 1987; Fainsod et al. 1987), the posterior myelencephalon (for Hox-2.1:Krumlauf et al. 1987; Utset et al. 1987; Holland & Hogan, 1988; for Hox-1.4 and Hox-1.2:Toth et al. 1987), the anterior spinal cord (Hox-6.1:Sharpe et al. 1988) and the cervical spinal cord at the level of the third to fifth prevertebrae (Hox-3.1:Awgulewitsch et al. 1986; Utset et al. 1987; Holland & Hogan, 1988). In the prevertebral column, the anterior limits are inclusive of the first prevertebrae (Hox-1.5:Gaunt, 1987), at the level of the seventh to eighth prevertebrae (for Hox-6.1:Sharpe et al. 1988; for Hox-1.3:Dony & Gruss, 1987), and the twelfth prevertebra (Hox-3.1:Holland & Hogan, 1988).
In addition to nervous and prevertebral tissue (and the somites of early embryos that give rise to the prevertebrae), other organs display homeobox gene transcripts in a way that is spatially restricted. This has been demonstrated most clearly for Hox-1.3 (Dony & Gruss, 1987) and Hox-2.1 (Holland & Hogan, 1988). It has not yet been established, however, that homeobox gene transcripts within the organs occupy different but overlapping domains. Thus, it remains to be examined whether or not different organs exhibit different arrays of homeobox gene transcripts depending upon their position along the anteroposterior axis of the body.
Although Hox-1.5 transcripts are first detected in posterior regions of the germ layers at 7| days (Gaunt, 1987), it is not yet known whether different homeobox genes already occupy different transcript domains at this early time or, indeed, whether different genes commence expression synchronously or sequentially (Gaunt, 1987). In the present paper, comparison is made between Hox-3.1 and Hox-1.5 in the temporal and spatial patterns of their embryonic transcripts. These genes were chosen for comparison specifically because their transcript domains provide the most widely separated anterior boundaries. It is now shown that the transcripts of these two genes occupy different, but overlapping, spatial domains both within the germ layers at 7I-8| days, and in various organs at 12| days.
Materials and methods
Preparation of embryo sections
Embryos were obtained from natural matings between F1 (CBAXC57BL/6) mice. For aging of embryos, midday on the day of the vaginal plug was designated day of pregnancy. Fixation, embedding and sectioning of embryos was as previously described (Gaunt et al. 1986; Gaunt, 1987), but sections were cut at a thickness of 7 μm.
Preparation and use of35S-labelled RNA probes
For Hox-1.5, production and alkaline hydrolysis of 35S-labelled RNA probes was as previously described (Gaunt, 1987). The mouse homeobox gene Hox-3.1 has been cloned by Awgulewitsch et al. (1986) and by Breier et al. (1986). In addition, Hox-3.1 genomic DNA has been independently cloned, and the homeobox sequenced, by D. Duboule et al. (unpublished data). A clone of Hox-3.1 genomic DNA in pBR322 was received as a gift from Denis Duboule. From this, the 330 bp Pstl −Aval homeobox-containing fragment (Breier et al. 1986) was subcloned into pGEM-3 (Promega). This fragment has previously been found to hybridize specifically to Hox-3.1 mRNA (2-7 kb) in Northern blots (Breier et al. 1986). For production of 35S-labelled RNA probes to Hox-3.1 mRNA (antisense probes), the recombinant plasmid was cut with /fzndlll and then used in a T7 RNA polymerase reaction, as already described for Hox-1.5 (Gaunt et al. 1986; Gaunt, 1987). For production of 35S-labelled Hox-3.1 sense (control) RNA probe, plasmid was cut with EcoRI and then used in an SP6 RNA polymerase reaction. Prior to use in hybridizations, Hox-3.1 RNA probes were hydrolysed at pH 10-2 for 60 min (Cox et al. 1984).
Methods used for in situ hybridization, RNase A digestion, washing, autoradiography and staining of sections were all as previously described (Gaunt et al. 1986).
Sagittal and near-sagittal sections were prepared and examined by in situ hybridization for Hox-3.1 and Hox-1.5 transcripts. During sectioning of an embryo, the midline was indicated by the presence of the allantois at the posterior end. For each embryo at 7| days only about eight adjacent sections included the allantois, and it was usually found that such a small number of sections readily permitted comparative hybridization of only two different probes. For most embryos, the probes compared were Hox-3.1 antisense versus Hox-1.5 antisense. For remaining embryos, and with the inclusion of at least one such embryo per experiment, the probes compared were antisense (Hox-3.1 or Hox-1.5) versus sense (Hox-3.1 or Hox-1.5). Sense (control) probes consistently produced only background levels of hybridization, whilst antisense probes on adjacent sections gave specific hybridization, as described below. A comparison of Hox-1.5 sense versus antisense probes hybridized to parallel sections of an 8-day embryo has already been published (Gaunt, 1987).
Figs 1, 2 and 3 illustrate variation in the developmental stage reached by days. In ‘early -day’ embryos (Fig. 1), anterior migration of mesoderm had not yet separated completely the ectoderm and endoderm layers, and posterior migration of mesoderm had produced only a small allantois. In ‘late -day’ embryos (Fig. 2), the first signs of head-fold formation were apparent and the allantois was well developed.
When tested by in situ hybridization, early -day embryos (six such embryos were studied) showed no detectable Hox-3.1 (Fig. 1A) or Hox-1.5 (Fig. 1B) transcripts. Late -day embryos (five such embryos were studied) showed Hox-3.1 transcripts localized mainly or entirely within the allantois (Fig. 2A), while Hox-1.5 transcripts were located both in the allantois and in the ectoderm and mesoderm layers of the posterior half of the embryo (Fig. 2B). Parasagittal sections at this stage (not shown), not including the allantois, showed no Hox-3.1 transcripts but continued to show Hox-1.5 transcripts over at least six sections to either side of the allantois. Two of the -day embryos studied, which were at a stage of development intermediate between the ‘early’ and ‘late’ embryos, showed Hox-1.5 transcripts (Fig. 3B) but little or no evidence of Hox-3.1 transcripts (Fig. 3A). It is therefore probable that Hox-1.5 transcripts begin to accumulate earlier than Hox-3.1 transcripts. However, the precise time interval between first expression of these two genes is at present uncertain. It is important to note that in the early stages of Hox-1.5 expression (Fig. 3B) transcripts were detected over the complete posterior half of the ectoderm and mesoderm layers: no embryos were found in which Hox-1.5 transcripts, like the earliest-detected Hox-3.1 transcripts (Fig. 2A), were restricted entirely to the allantois.
8- and -day embryos
Fig. 4 compares the transcript domains of Hox-3.1 and Hox-1.5 at 8 days. In the embryo shown, the head fold was well developed and three somites had already been formed. Transcripts of both genes were seen in posterior regions of ectoderm and mesoderm, including the allantois, but the transcript domain for Hox-1.5 (Fig. 4B) extended to a more anterior position than that for Hox-3.1 (Fig. 4A). For Hox-3.1, the anterior limits of transcripts within both ectoderm and mesoderm were situated posterior to the position of the earliest-developing somites. For Hox-1.5, both within ectoderm and mesoderm tissues, the anterior limits of transcripts extended anterior to the position of first somite formation. It therefore appears likely that the Hox-1.5 transcript domain within mesoderm tissue includes somite one.
Fig. 5 compares the transcript domains of Hox-3.1 and Hox-1.5 in an 8-somite-stage (8i-day) embryo. At this stage, it was apparent that the Hox-1.5 transcript domain (Fig. 5B) extended anteriorly into that part of the neural ectoderm which was becoming segmented as neuromeres, and which would normally have formed the hindbrain (Rugh, 1968). Although not shown very precisely in Fig. 5B, the anterior limit of transcripts was usually found at this stage to coincide with a neuromere constriction on the ventral surface of the neural ectoderm (Gaunt et al. 1986; Gaunt, 1987). The Hox-1.5 transcript domain within mesoderm included somites two to eight (Fig. 5B) although, as previously described (Gaunt, 1987), the intensity of labelling within mesoderm was now becoming less than that in the neural ectoderm. In a parallel section of the same embryo, Hox-3.1 transcripts were clearly restricted to ectoderm and mesoderm posterior to the region of early somite formation (Fig. 5A).
The -day embryo
Fig. 6 shows a comparison of Hox-3.1 and Hox-1.5 transcript domains as detected by in situ hybridization on adjacent or nearby sections taken from the same -day embryo. The anterior boundaries of these domains within the nervous system at about this stage have already been described elsewhere (for Hox-3.1:Awgulewitsch et al. 1986; Utset et al. 1987; Holland & Hogan, 1988; for Hox-1.5:Gaunt, 1987; Fainsod et al. 1987), and are indicated here by large arrows on Fig. 6A. For Hox-3.1, the boundary is at the level of the fifth prevertebra in the ventral third of the spinal cord, and at about the level of the third prevertebra in the centre of the spinal cord. For Hox-1.5, the boundary is within the preotic hindbrain. For both genes, there is a gradual anterior-to-posterior reduction in the intensity of labelling along the length of the spinal cord (not shown). .
Fig. 6C and D compare Hox-3.1 and Hox-1.5 transcripts in tissues that surround the pharynx. No specific labelling of this region was given by the Hox-3.1 probe (Fig. 6C), but Hox-1.5 transcripts were clearly present in the most anterior prevertebrae, including prevertebra 1 (Fig. 6D). The thyroid duct (Fig. 6A) leads from the pharynx to the thyroid gland which, at 12| days, lies at the base of the duct, deep in the pharyngeal floor (Fig. 6A; Rugh, 1968). Strong labelling with the Hox-1.5 probe was detected over the thyroid gland (Fig. 6D). Less-intense labelling with this probe was also seen generally over tissue which formed the floor of the pharynx (Fig. 6D). Labelling of the pharyngeal floor extended anterior to the thyroid duct, but did not extend over the tongue. The thyroid gland consists of both endoderm- and mesoderm-derived components (Hilfer, 1968) but sections of thyroid tissue at days did not pèrmit clear distinction between these cell types. In some sections, although not apparent in Fig. 6D, the duct leading into the thyroid from the pharynx showed only low levels of labelling when compared with the surrounding tissue. Posterior to the thyroid region, labelling with the Hox-1.5 probe extended to include a small area of tissue in the most anterior (atrial) region of the heart (not shown). Most of the heart tissue, however, showed only background labelling after hybridization with the Hox-1.5 probe. No specific labelling of any part of the heart was given by the Hox-3.1 probe (not shown).
Fig. 6E and F compare Hox-3.1 and Hox-1.5 transcripts in prevertebrae 10-14 and adjacent tissues. Hox-3.1 transcripts were seen in prevertebra 12 but not in any of the more anterior prevertebrae (Fig. 6E). Prevertebrae 13-16 were labelled strongly by the Hox-3.1 probe, and labelling persisted at a lower intensity over all more posterior prevertebrae. This finding is similar to that described for Hox-3.1 by Holland & Hogan (1988). All prevertebrae were labelled by the Hox-1.5 probe, and Fig. 6F shows Hox-1.5 transcripts in prevertebrae 10-14. The lung was positive for Hox-1.5 (Fig. 6F) but not Hox-3.1 transcripts (Fig. 6E). Hox-1.5 transcripts in the lung were restricted to mesodermal components and were not detected in the endodermally-derived lining epithelium. This distribution is therefore similar to that already described for transcripts of Hox-2.1 (Krumlauf et al. 1987), Hox-1.3 (Dony & Gruss, 1987) and Hox-6.1 (Sharpe et al. 1988).
Fig. 6G,H compare Hox-3.1 and Hox-1.5 transcripts in the liver, stomach, mesonephric and metanephric kidneys. Neither gene was seen to be expressed in the liver, but both Hox-3.1 and Hox-1.5 transcripts were detected in the mesonephric and metanephric kidneys. These findings for the kidney are as earlier reported by Awgulewitsch et al. (1986) for Hox-3.1, and by Gaunt et al. (1986) for Hox-1.5.
The distributions of Hox-3.1 and Hox-1.5 transcripts in the kidney were similar, with no obvious spatial restriction within the tissue. The stomach was positive for Hox-1.5 (Fig. 6H) but not Hox-3.1 transcripts (Fig. 6G). As noted above for the lung, Hox-1.5 transcripts in the stomach were restricted to mesodermal components and were not detected in the endo-dermally-derived lining epithelium. A similar pattern of distribution within the stomach has been reported for transcripts of Hox-2.1 (Holland & Hogan, 1988) and Hox-1.3 (Dony & Gruss, 1987).
In this paper, the mouse homeobox genes Hox-3.1 and Hox-1.5 are compared in the temporal and spatial patterns of their embryonic transcripts. Transcripts of both genes were first detected by in situ hybridization at about 7| days, although the appearance of Hox-3.1 transcripts apparently followed Hox-1.5 after a small delay. Hox-3.1 and Hox-1.5 transcripts occupied different, but overlapping, domains along the anteroposterior axis of the embryo. This was found both within the ectoderm and mesoderm germ layers at 74-8) days, and in at least some of the organs at 12) days. These results extend earlier findings that Hox-3.1 and Hox-1.5 transcripts occupy different but overlapping domains within the nervous system and prevertebrae of midgestation embryos (for Hox-3.1’. Utset et al. 1987; Holland & Hogan, 1988; for Hox-1.5’. Gaunt et al. 1986; Gaunt, 1987; Fainsod et al. 1987). At the present time, the term ‘overlapping’ used here to describe transcript domains is only strictly applicable to gross regions of tissue such as, for example, an area of the spinal cord, prevertebral column or kidney. It seems likely that transcript domains of mouse homeobox genes also overlap within individual cells, but it is important to note that this is not yet proven.
Table 1 compares the transcript patterns now demonstrated in organs of midgestation mouse embryos with that shown by another mouse homeobox gene, Hox-6.1 (Sharpe et al. 1988). It is seen that the transcript domain for Hox-6.1 differs from, but overlaps with, the transcript domains for Hox-1.5 and Hox-3.1. Presented in the form shown in Table 1, the data suggest that it is the position of an organ along the anteroposterior axis that determines the range of homeobox gene transcripts present. This seems reasonable for Hox-3.1 and Hox-1.5 since it is position along the anteroposterior axis of -to -day embryos that determines presence or absence of Hox-3.1 and Hox-1.5 transcripts. It is important to note, however, that the organs listed in Table 1 may also differ in their positions of origin along the lateral axis of the embryo. Position along lateral, or even dorso-ventral, axes might possibly be additional factors that influence the expression of some homeobox genes. Thus, it is not clear at the present time that all homeobox genes and organs can be expected to conform to the simple hierarchical pattern shown in Table 1. Our recent additional observations on the distributions of Hox-1.2, 1.3 and 1.4 transcripts in -day embryos do conform to the pattern of Table 1 (S. J. Gaunt, P. T. Sharpe & D. Duboule, unpublished data). However, Hox-2.1 does not conform closely, since expression is found in the lung, stomach and metanephric kidney but is weak or absent in prevertebrae (Holland & Hogan, 1988). It is of interest to note that the mesenchymal component of the thyroid gland, unlike that of the other organs described, is derived from neural crest cells (Le Douarin, 1982). Holland & Hogan (1988) previously observed Hox-2.1 expression in autonomic ganglia derived from the neural crest.
The transcription patterns of Drosophila homeotic genes show two important points of similarity with those now described for homeobox genes of the mouse. First, the transcription domains are established early within the germ layers (ectoderm and mesoderm) of the developing Drosophila embryo (Levine et al. 1983; Akam & Martinez-Arias, 1985) with vestiges of this original pattern remaining in descendant structures, such as the imaginai discs of third instar larvae (Akam, 1983; Wirz et al. 1986). Second, the transcripts of most Drosophila homeotic genes occupy different, but overlapping, spatial domains. This is seen most clearly for Antennapedia and for genes of the Bithorax complex (BX-C) (e.g. Harding et al. 1985). It is an interesting possibility, therefore, that homeobox genes might regulate cellular determination along the anteroposterior axis of the mouse by similar mechanisms to those proposed by Lewis (1978) for genes in the BX-C of Drosophila (compare Table 1 of this paper with fig. 4 of Lawrence & Morata, 1983). In Lewis’s model, and its interpretion by Lawrence & Morata (1983), each gene of the BX-C plays its main role in that segment where it is expressed most anteriorly, but it may also play a lesser role in more posterior segments. Thus, it is the total array of homeotic genes active in a particular group of cells which may determine the final phenotype. In Lewis’s model, the BX-C gene products within each segment of Drosophila promote the development of structures appropriate to that segment and suppress structures characteristic of more anterior parts. Holland & Hogan (1988) have previously suggested that mouse homeobox genes, like homeotic genes of Drosophila, might act combinatorially to effect positional specification.
If mouse homeobox genes provide positional cues for the specification of tissue determination during embryogenesis (Gaunt, 1987), then the patterns of homeobox gene expression now observed at and days may be directly linked by cell lineage. The scheme of homeobox gene activity presented as Table 1 might thus be used to construct a fate map for the germ-layer-stage embryo. Such a conclusion is implicit in the earlier work of Dony & Gruss (1987) and Holland & Hogan (1988). Two points of caution should be considered in any comparison of the patterns of homeobox gene expression at early and later stages of embryogenesis. First, expression at days may be subject to later modification associated with the focusing of transcript boundaries (Gaunt, 1987). Second, many organs begin to cease expression of homeobox genes in the second half of gestation (Dony & Gruss, 1987).
Some aspects of vertebrate development, especially of development within the mesoderm, clearly occur in an anterior-to-posterior sequence. Anterior mesoderm, for example, precedes posterior mesoderm both in times of movement away from the primitive streak and of segmentation to somites. Thus, somite 1 (labelled by the Hox-1.5 probe) forms at about 8 days, while somites 16 and 17 (suggested by Holland & Hogan, 1988, to be the precursors of prevertebra 12, and thus probably the first to be labelled by the Hox-3.1 probe) form at days (Rugh, 1968). It is now demonstrated that both Hox-1.5 and Hox-3.1 transcripts are first detected by in situ hybridization at about 71 days, but with Hox-3.1 transcripts apparently following Hox-1.5 after a small delay. These data show clearly that the time course over which homeobox genes are first expressed does not correspond to the time course of somite formation. Unfortunately, the data now presented, and our knowledge of cell behaviour in the -day embryo, are too imprecise to indicate whether or not activation of homeobox genes corresponds in its time course with that of mesoderm invagination through the primitive streak. In one popular hypothesis (discussed by Gaunt, 1987) it is during invagination through the primitive streak that mesoderm cells first acquire their positional address; ectoderm cells then acquire positional information secondarily by induction from the underlying mesoderm. The data obtained so far are probably most consistent with an alternative model in which all homeobox genes are activated as a single event, but with a slight anterior-to-posterior delay between different genes. This event would commence at about the time that mesoderm completes its migration to the anterior tip of the embryo. Such a model is in-keeping with the finding from Northern blotting experiments that several different homeobox genes of Xenopus are activated in late gastrulae (Carrasco et al. 1984; Harvey et al. 1986; Condie & Harland, 1987).
Although the body plan of the mouse might be specified by the spatial patterns of homeobox gene expression first established at about 7| days, and then subsequently focused in their transcript boundaries (Gaunt, 1987), the important question remains as to what is the underlying mechanism by which this early pattern is generated. In Drosophila, the spatial patterns of homeotic gene expression probably arise, in a complex series of steps (reviewed by Akam, 1987), from at least two gradients of morphogenetic material which are already present in the egg (Frohnhofer & Nûsslein-Volhard, 1986; Lehmann & Niisslein-Volhard, 1986). In the mouse embryo, however, there are probably no such anteroposterior determinants until first appearance of the primitive streak (discussed by Gaunt, 1987). Thus, the mouse embryo may differ from the fly embryo in that it must generate de novo during gastrulation a gradient of positional information. Two alternative mechanisms might account for this. In a ‘cascade model’, the product of one homeobox gene might, after an interval required for its accumulation, activate the next homeobox gene along the anterior-to-posterior sequence. If activation occurred in migrating mesoderm cells, but only during the period of time that they remained within the vicinity of the primitive streak, then the overall pattern of gene activity generated along the anteroposterior axis of the embryo would resemble that shown in Table 1. Breier et al. (1986) have already suggested some essential features of a cascade model. In an alternative ‘morphogen gradient model’, a gradient of morphogen within the epiblast or newly formed germ layers might underlie the generation of positional information. In this model, homeobox genes would differ in the thresholds of morphogen concentration above which they became activated. Retinoic acid, a likely morphogen in the chick limb bud (Thaller & Eichele, 1987), is one candidate for possible involvement. This substance induces homeobox gene activity in mouse teratocarcinoma cells (e.g. Breier et al. 1986), but by a mechanism which may not simply be a consequence of its effect upon differentiation (Deschamps et al. 1987).
I thank Denis Duboule for the generous gift of Hox-3.1 DNA, and Don Powell for many stimulating discussions about the establishment of homeobox gene transcription patterns.