A gene with sequence similarity to Drosophila engrailed is expressed during the development of the neural tube and vertebrae in the mouse

Attempts to identify genes that regulate mammalian development have been limited by the inaccessibility of the embryo, the difficulties of classical genetic analysis and the scarcity of consistent fine details in the phenotype. In organisms such as Drosophila, genetic analysis and recombinant DNA technology have led to the identification and isolation of numerous genes involved in the establishment, and differentiation, of the metameric pattern (Akam, 1987). Several of these genes have been found to share a conserved region, the ‘homeobox’, 183 bp long which codes for 61 amino acids and has sequence similarities to the DNA-binding domain of some regulatory proteins in prokaryotes and yeast (Laughon & Scott, 1984; Shepherd et al. 1984). Genes containing homeobox sequences have been identified, and in several cases shown to be developmentally expressed, in a wide range of metazoans, including vertebrates (Levine et al. 1984; McGinnis et al. 1984). Thus, the homeobox-containing genes may have regulatory functions in vertebrate embryogenesis.

The Drosophila genes engrailed and invected contain related homeobox sequences that have diverged significantly from the majority and thus constitute a separate class. Genetic analysis has shown that engrailed plays a crucial part in segmentation (Morata & Lawrence, 1975; Kornberg, 1981). In situ hybridization has shown that, commensurate with this function, the pattern of engrailed expression at the cellular blastoderm stage is a series of stripes marking the anterior margin of each larval parasegment (Kornberg et al. 1985; Ingham et al. 1985). engrailed is also expressed in the posterior compartments of other structures, for example in imaginal disks, and in the ventral nervous system (Kornberg et al. 1985). invected is expressed in the same pattern (Coleman et al. 1987).

Two genes in the mouse, En-1 and En-2, contain the engrailed-like homeobox. cDNA clones corresponding to both genes have been isolated: En-1 has been mapped to chromosome 1 and En-2 to chromosome 5 (Hill et al. 1987; Joyner & Martin, 1987). The similarity between the mouse genes and their Drosophila counterparts extends outside the homeobox region for 100 bp in the 5′ direction and 96 bp in the 3′ direction (Joyner & Martin, 1987). The conservation of sequence in and around the homeobox suggests conservation of function in this region in the mouse and fly. In view of the correlation, in Drosophila, between the pattern of engrailed expression and its role in maintaining spatial organization, we have investigated the pattern of expression of the En-1 and En-2 genes in the mouse embryo by in situ hybridization.

En-1 and En-2 cDNAs were isolated from a cDNA library (kindly provided by Drs Brigid Hogan and Karen Fahmer) synthesized from 8·5-day post coital (p.c.) mouse mRNA cloned into λgt10(Fahmer et al. 1987). These cDNAs were selected using Drosophila engrailed cDNA kindly provided by Dr W. Gehring. The En-1 and En-2 cDNAs and the gene-specific subclones were cloned into the vector pTZ18U containing the T7 promoter (Mead et al. 1986) (Pharmacia PL Biochemicals).

Plasmids (see Results, Fig. 1) were linearized with EcoRI or BamHI as appropriate and labelled RNA was prepared by transcription with T7 RNA polymerase (Bio-rad) in the presence of ³⁵S UTP (1400 Ci mmol^-1, NEN). The preparation was treated with RNase-free DNase (Pharmacia) and purified, by phenol-chloroform extraction and cold-ethanol precipitation. The RNA was degraded to approximately 100 bases average length by partial alkali degradation, precipitated in cold ethanol for 3h at —20°C, washed, dried in vacuo and redissolved in hybridization mix (Wilkinson, 1987a) at a concentration of approximately 1ngμl^-1 per kb length of cloned fragment (2 ×10⁵ disints min^-1μl^-1).

Mouse embryos were derived from matings of Swiss mice. Development is assumed to have begun at midnight on the night of mating which is designated 0 days p.c. Pregnant females were killed by cervical dislocation. The embryos were removed into ice-cold phosphate-buffered saline (PBS) transferred immediately to ice-cold 4 % paraformaldehyde in PBS and fixed in this solution at 4°C overnight.

The fixed embryos were processed routinely for histology and imbedded in wax (to facilitate handling, 9·5-day embryos were preimbedded in 1·5% agar after fixation, then processed as for older embryos). Sections, 6μm thick, were cut and floated onto poly-lysine-coated slides then incubated at 56°C for 1·25 h to fix them to the slides. We found it convenient to place three groups, of two or three sections each, on each slide so that combinations of probes and controls could be used on the same slide.

In situ hybridization was performed by the method described by Wilkinson et al. (1987a), with the inclusion of high-stringency washing steps described by Wilkinson et al. (1987b). Briefly, the prehybridization treatment included steps to further fix the material, to partially digest the sections with proteinase K in order to improve access of the probe to the mRNA and to block basic groups by acetylation in order to reduce nonspecific binding of the nucleic acid probes. Each group of sections was treated with 7 μl of the appropriate ³⁵S-labelled RNA probe in hybridization mix (for concentrations, see above) at 50°C overnight. The slides were washed under stringent conditions and treated with RNase to remove unhybridized, nonspecifically bound probe, then dehydrated. Autoradiography was performed with Ilford K5 liquid emulsion; exposures were between 2 and 3 weeks. Sections were stained lightly with methyl green or toluidine blue and mounted under a coverslip using DPX mountant. Sections adjacent to those used for in situ hybridization were stained routinely using haematoxylin and eosin, and examined together with the hybridized sections to analyse the distribution of label.

Sections were examined under a Leitz Orthomat microscope using dark- or bright-field illumination and photographed using Kodak Technical Pan film rated 25–50 ASA.

We have examined, by in situ hybridization, the distribution of RNA transcripts from the En-1 and En-2 genes in mouse embryos. For this, we have used three probes derived from cDNAs of the two genes (Fig. 1). Two probes were derived from specific regions of these cDNAs. In order to obtain an independent indication of the distribution of En expression in the embryo, a nonspecific probe, which recognizes both genes, was constructed using a region of the En-1 cDNA with partial homology to the En-2 gene. We have confirmed that this latter probe recognizes both genes by Southern analysis: under the conditions used (2×SSC, 68°C) the probe hybridized to both genes to approximately the same degree (data not shown). As negative controls, we have used two sense probes, complementary to the En-1 and the En-(1 +2) probe sequences. The patterns of hybridization were examined in transverse, frontal and sagittal sections through a total of seven 12·5-day embryos. This series of experiments was supplemented with limited studies on sections through three 9·5day, and two 10·5-day, embryos.

The tissue distribution of label observed with each probe is summarized in Table 1 and detailed below. While both specific probes labelled the hindbrain and midbrain region, several additional domains of the embryo labelled with the En-1 probe. This is despite the fact that the En-2 probe labelled the brain region more intensely than the En-1 probe. With the exception of the thoracic body wall (see below), the combined hybridization patterns of the En-1 and En-2 probes matched the pattern of hybridization of the En-(1 +2) probe (Table 1). With the exception of label in red blood cells (see below), all the labelling patterns described were observed only with En antisense probes and not with either of the control probes. General views of sections through 12·5-day embryos treated with the En-1 and En-(1 +2) antisense, and the En-(1 +2) sense, probes are shown in Fig. 2; Fig. 2A also serves to help the reader locate the views illustrated in other figures.

All three anti-sense En probes labelled the hindbrain and midbrain in discrete domains in the germinal and intermediate zones in 12·5-day embryos (Fig. 3). The roof of the brain around, and including, the isthmus of rhombencephali, the developing pons and much of the midbrain floor were heavily labelled. However, only the En-(l+2) and En-1 probes labelled the germinal and intermediate zones in the anterior floor of the hindbrain (Fig. 3E).

The following tissues were labelled with the En-1 and En-(1 +2) probes, but not with the En-2 probe.

Sagittal sections through 9·5-day and 10·5-day embryos showed a band of expression in the ventral part of the neural tube in the cervical and upper thoracic regions. The precise posterior extent of this domain was not established, but transverse sections of the caudal part of the tail showed no labelling over the neural tube. Sections through 12·5-day-old mice showed a similar band of label that extended throughout the trunk, from the junction with the hindbrain to at least the level of the 30th segment posterior to the atlas (Fig. 4); again, the neural tube in the posterior part of the tail was unlabelled (Fig. 6).

The neural tube comprises two cellular regions, a germinal, proliferative zone which lines the lumen of the tube and from which cells migrate into the more lateral intermediate zone where they differentiate into neurones and glial cells. Transverse sections showed that expression was confined to bilaterally symmetrical regions in the ventral part of the cord (Fig. 5). From 9·5 to 12·5 days of development, labelling was most intense in a restricted, midlateral, region of the germinal zone which occupied the same position at different developmental stages and at different axial levels in the 12·5-day embryo. At 9·5 and 10·5 days, only the germinal zone was present, but at 12·5 days expression extended ventrolaterally into the intermediate zone in streaks which appeared to mark the positions of individual cells. No histological boundary was evident between the labelled and unlabelled regions of the germinal and intermediate zones.

The pattern of expression in somites and their derivatives varied with embryonic stage. At 9·5 days, the dermatome showed a moderate degree of labelling in the trunk, but was barely labelled above background in the caudal part of the tail: the remainder of the somite was unlabelled. At 10·5 days, the dermatome was still labelled. At this stage, the cervical and thoracic vertebrae are beginning to form and the axial sclerotome comprises alternate bands of loose and dense tissue which will contribute to the centrum and intervertebral disc, respectively (these alternate bands are not to be confused with the loose and dense sclerotome which characterize the rostral and caudal halves of the somites at earlier stages of development). A few labelled mesenchymal cells were observed in the lateral parts of each loose band, most prominently at its junction with the more dense material. The more posterior somites, which represent earlier stages of axial development, were unlabelled.

At 12·5 days, the dermatome of the tail somites was clearly labelled (Fig. 6) and the segments in the trunk displayed a complex pattern of expression. Labelling was confined to three regions. First, the axial sclerotome of the cervical and thoracic regions displayed alternate bands of labelled and unlabelled sclero-tomal tissue. The loosely organized tissues of the developing centra were labelled and the dense bands of sclerotome that contribute to the formation of the intervertebral discs were unlabelled (Fig. 7). Second, throughout the trunk, label was observed in discrete blocks of mesenchymal tissue (Fig. 7) immediately lateral to each ventral primary ramus, identified by their position and shape as the developing costal processes. These labelled blocks of tissue could be most clearly seen in frontal and sagittal sections (Figs 7, 8A,B and 9) where they were continuous, at approximately the notochordal level, with labelled, nonsegmented mesenchymal tissue immediately lateral to the segmental dense sclerotome and mesial to the unlabelled myotome. At a more dorsal level, this nonsegmented mesenchymal tissue appeared to be continuous with labelled subectodermal derivatives of the dermatome. Third, the dorsal parts of the developing neural arches in the cervical and thoracic regions were lightly labelled. In several instances using the En-(1 +2) probe, we observed that the label was most intense in the anterior part of the arch (Fig. 8C,D). This labelled region clearly included the body of the arch. In adjacent sections stained with haematoxylin and eosin loose mesenchymal cells were observed in close association with the neural arches: at the resolution afforded by ³⁵S, we cannot exclude the possibility that some of these cells were also labelled. The body of the arch showed no evidence of anteroposterior differentiation at the histological level.

The mesenchyme of the face was labelled in both 9·5- and 12·5-day embryos in a restricted region anterior to the forebrain, above the developing snout. The lateral parts of the face, for example primordia of the vibrissae, were not labelled above background.

The loose mesenchyme of the thoracic body wall and the ectoderm of the forelimb bud were labelled in several, but not all, 9·5-day and 12·5-day embryos examined. Since label was not consistently observed in these regions, it is possible that this was an artefact. The same regions did not, however, label above background in sections treated with control probes.

Finally, nucleated, embryonic, red blood cells were frequently labelled with the three anti-sense probes. However, the same intensity of labelling was observed with control probes, suggesting that the antisense hybridization was an artefact.

No En expression was detected in any other tissues in the 12·5-day embryo, though all the major organs and tissues present at this stage were clearly visible.

These results reveal two broad features of En expression in the mouse embryo between 9·5 and 12·5 days of development. First, En-2 is expressed in a small subset of cells expressing En-1. Second, both En genes are active only in restricted locations within otherwise non-expressing tissues. In this discussion, we focus attention on the implications of these aspects of En expression for our understanding of the control and function of the En genes in mouse embryogenesis.

The expression of En-1 and En-2 is not exclusive. Northern analysis has shown that both genes are transcribed in the brain (Joyner & Martin, 1987). A recent in situ hybridization study has described in detail the expression of En-2 (Davis et al. 1988). Our observations agree with the results of these studies and further show that transcripts of both mouse En genes are found in the developing brain, at least in some cases, apparently in the same cells. However, the present results, again extending the information available from Northern analysis (Joyner & Martin, 1987), demonstrate a complex and diverse pattern of En-1 expression. By comparison with the limited domain of En-2 activity the diverse expression of En-1 indicates that the two genes are subject to different controls.

Our results suggest that the activity of these genes is controlled according to cell position as well as cell type. In several instances, expression is restricted to particular locations within otherwise non-expressing tissue. Examples are, first, the difference in expression between rostral and caudal parts of the neural arches and, second, the restricted domains of expression in the midbrain, hindbrain and neural tube. In these examples, the boundaries between expressing and non-expressing cells lie within a single tissue, but do not correspond to any obvious boundaries at the histological level.

In very general terms, therefore, the control of En expression in the mouse shows some similarities to the control of the engrailed family in Drosophila. The expression of engrailed and invected is not exclusive (Coleman et al. 1987). The boundaries between cells expressing engrailed and non-expressing cells do not correspond to boundaries of tissue type (Kornberg et al. 1985; Ingham et al. 1985). This is not to say that the analogy will likely hold at any more detailed level. There is no reason to suppose that the mouse En genes are regulated by control sequences homologous to those associated with their Drosophila counterparts. It is of interest, in this context, to compare the pattern of En expression with the distribution of transcripts from the gene, int-1 (Wilkinson et al. 1987b). int-1 is the mouse homologue of the fly gene wingless (Rijsewijk et al. 1987) which has been implicated in the control of engrailed expression (DiNardo et al. 1988). The restriction of int-1 expression to the developing nervous system implies that it is not generally responsible for controlling En expression in the mouse.

What is the function of the En genes in the mouse? In Drosophila, there is a strong relationship between expression and function in homeobox-containing genes (Akam, 1987). We will therefore assume that RNA expression, as detected by in situ hybridization, marks En function. In order to test this point in the future, it will be important to investigate experimentally the effects of changes in En expression during mouse embryogenesis.

In Drosophila, engrailed is one of a set of genes that act to maintain differences between cells of different polyclonal lineages. These differences, in turn, prevent the cells of adjacent lineages from mixing, and thus effectively structure the embryo into compartments. engrailed is expressed in the posterior compartment of each segment and functions to maintain anteroposterior compartment boundaries, for example, during segmentation (Morata & Lawrence, 1975; Lawrence & Morata, 1976; Kornberg, 1981).

Do the mouse En genes play an analogous role? Several examples of extensive mixing between marked cell lineages in pregastrula vertebrate embryos have effectively excluded the possibility that spatial restriction of polyclones is widespread in vertebrates (see, for example, Kimmel & Warga, 1987; Dale & Slack, 1987). However, the possibility remains that the mechanisms that control the integrity of groups of cells with like developmental commitments in insects have their counterpart in vertebrates where such commitments appear to originate later in development and in larger groups of cells that are not necessarily of the same lineage.

The restriction of En expression to particular domains within morphologically undifferentiated tissues is consistent with a role in preventing adjacent groups of cells from mixing. The alternate bands of En-1 expression in the developing axial skeleton provide a striking example. It is unlikely, however,that this expression plays a role in the primary segmentation process. At the time of segmentation, chick somites are already differentiated into rostral and caudal halves; 6–8 h later, the cells of the two halves are differentiated in ways that prevent them from mixing (Stem & Keynes, 1986) in a manner that is at least superficially similar to the spatial restriction of cell lineages during segmentation in the fly. At the level of sensitivity of the present experiments, En does not appear to be expressed in any half-segment pattern prior to the beginning of vertebral development. Moreover, even at this late phase, the pattern of expression of En-1 does not follow the contributions made by adjacent somites to the vertebrae as determined in chick-quail chimaeras (Bagnall et al. 1988). For example, we have observed no expression boundary within the developing centrum, although the expression of En-1 in the anterior parts of the developing neural arches may correspond to the contribution of caudal parts of the somite to this structure (Bagnall et al. 1988). We cannot exclude the possibility that an earlier phase of En-1 expression may play a role in the process of segment determination in the presomitic mesoderm of the mouse. However, it appears that if the function of En-1 in the developing axial skeleton is in any way analogous to that of engrailed in Drosophila it may be to define or maintain patterns of tissue organization during the complex cell rearrangement and differentiation processes that are involved in vertebra formation.

It is not surprising that the expression of En-1 during segmentation in the mouse differs from that of engrailed in Drosophila. Segmentation almost certainly arose independently in the vertebrates and insects (Cooke, 1981). Moreover, the establishment of the segmental body plan in arthropods appears to occur in the epidermis, whereas in the vertebrates it occurs in the mesoderm. It may be for this reason that, in Drosophila, engrailed is expressed mainly in the epidermis and only transiently in mesodermal cells (Kornberg et al. 1985; Ingham et al. 1985).

A second example of the restriction of En expression to domains without morphological boundaries is in the developing nervous system. Davis et al. (1988) have noted this feature in En-2 expression in the brain and have shown that, within this limited domain, expression begins in the neural folds, continues in the germinal zone, and is found later in cells migrating to the intermediate zone. The present results indicate that En-1 has a similar, though more extensive, pattern of expression in the central nervous system. In particular, in the neural tube, En-1 transcription is active in a small region of the germinal layer and its derivatives in the intermediate layer. Like En-2 (Davis et al. 1988), En-1 remains restricted in its expression to the same domain in the germinal layer over several days of development. These observations are consistent with the view that both genes are involved in maintaining, perhaps establishing, spatial differentiation in the central nervous system. The same may be true of engrailed which is expressed in a subset of cells in the ventral neural system of the fly, but here expression is in a reiterated pattern quite different from the expression of the mouse gene (Ingham et al. 1985; DiNardo et al. 1985; Coleman, et al. 1987).

Several studies have demonstrated the expression of homeobox-containing genes in the developing vertebral column and neural tube during mouse embryogenesis (Awgulewitsch et al. 1986; Dony & Gruss, 1987; Gaunt, 1987,1988; Krumlauf et al. 1987; Breier et al. 1988; Holland & Hogan, 1988; Sharpe et al. 1988). However, only Hox 3.1 and Hox 1.3 have been reported to be expressed in some parts of the sclerotome and not in others (Dony & Gruss, 1987; Breier et al. 1988). Hox 1.3, in particular, is expressed in the developing ribs and centrum in the thoracic region in a pattern that shows some similarities to the pattern of En-1’. however, unlike En-1, Hox 1.3 is also expressed in the lung and gut (Dony & Gruss, 1987). The relations between patterns of expression of the mammahan homeobox-containing genes is not clear, though they are all developmentally regulated and expressed in restricted, often complex, patterns.

Taking a broad view of our results, we see that the En genes fit this general picture. More specifically, they are expressed in a few situations where tissues, whose primary fate has been decided, undergo limited cell rearrangements and further differentiation. This can be seen, for example, in the development of vertebrae which involves complex rearrangements of sclerotomal cells and in the development of the intermediate zone of the spinal cord following cell migration from the germinal zone. This coincidence may reflect a role for the En genes in maintaining the spatial integrity of tissues in the mouse analogous, in principle, to the role of engrailed in Drosophila. Whether the genes play such a role will only become clear when En expression can be genetically or experimentally modulated during mouse embryogenesis.

We are indebted to Professor W. Gehring for giving us the Drosophila engrailed cDNA clone, Drs B. Hogan and K. Fahmer for giving us the mouse cDNA library and Drs A. McMahon and D. Wilkinson for teaching us the subtleties of the in situ hybridization technique. It is also a pleasure to thank Professor M. Kaufman, and Drs M. Snow and B.Gregg for their expert advice on the development of the axial skeleton.