Mouse homeo-genes within a subfamily, Hox-1.4, -2.6 and -5.1, display similar anteroposterior domains of expression in the embryo, but show stage- and tissue-dependent differences in their regulation

In situ hybridization studies have provided much circumstantial evidence to support the hypothesis that the Antennapedia- (Antp-) like homeobox-containing genes of the mouse, like those of Drosophila (e.g. Gehring, 1987), may play a role in pattern formation during embryogenesis. Thus, mouse homeo-genes are expressed in spatially restricted domains within the developing embryo. Different homeo-genes are usually found to be expressed in different domains, and these lie within a series of partially overlapping transcript domains along the body axis (e.g. Gaunt, 1988; Gaunt et al. 1988; Holland and Hogan, 1988a). The domains are first established before organogenesis within the ectoderm and mesoderm germ layers at days gestation (Gaunt et al. 1986; Gaunt, 1987, 1988; Mahon et al. 1988), and subsequently they persist within the developing nervous system, the prevertebral column, and at least some of the organs at days.

Four separate clusters of Antp-like homeo-genes have been identified in the genome of the mouse. These are the Hox-1, Hox-2, Hox-3 and Hox-5 loci (e.g. Duboule and Dollé, 1989; Graham et al. 1989). The four clusters display similarities in their intergene spacing and nucleotide sequence, suggesting that they arose by duplication of a common ancestral cluster (Hart et al. 1987; Odenwald et al. 1987; Graham et al. 1988; Duboule et al. 1989). This ancestral cluster has ancient origins in animal evolution, and it is a related cluster of homeo-genes in Drosophila (Duboule and Dollé, 1989; Graham et al. 1989) that regulates pattern formation in this species. In both Drosophila and mouse, there is apparently a strict correspondence between the relative position of homeo-genes within their clusters and that of their transcript domains within the developing embryo (for Drosophila:Harding et al. 1985; for mouse Hox-1: Gaunt et al. 1988; Duboule and Dollé, 1989; for Hox-2: Graham et al. 1989; for Hox-5: Duboule and Dollé, 1989).

It is not yet known why the mouse genome contains four related clusters of Antp-like homeo-genes. Each gene within a cluster has homologues in other clusters, and the term ‘subfamily’ has been used as the collective name for such a grouping of homologous genes (Hart et al. 1987; Duboule et al. 1989). Initial studies by Gaunt et al. (1988) indicated that homeo-genes within the same subfamily (Hox-1.4 and Hox-5.1; Hox-1.2 and Hox-6.1) display similar or identical positions of their transcript domains in the developing central nervous system and prevertebral column. However, recent studies have identified two genes within another subfamily that do not show similar domains of transcription: Hox-2.5 (Graham et al. 1989) and Hox-5.2 (Duboule and Dollé, 1989).

In this paper, we compare the transcript patterns of three genes from the Hox-1.4 subfamily (Hox-1.4, Hox-2.6 and Hox-5.1). Some aspects of our findings have been reported in earlier in situ hybridization studies (for Hox-1.4: Toth et al. 1987; Gaunt et al. 1988; for Hox-2.6: Graham et al. 1988, 1989; for Hox-5.1: Featherstone et al. 1988) but we have attempted in this work to provide an accurate comparison of Hox-1.4, -2.6 and -5.1 transcript patterns as detected on adjacent or nearby sections of the same embryo. We show that all three genes are closely similar in the limits of their transcript domains along the anteroposterior axis. Between genes, however, we note some striking stage- and tissue-dependent differences in the relative abundance of their transcripts.

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. Embryos of and days (but not and days) were fixed within their deciduae. Methods used for fixation, embedding and sectioning (7μm thick sections) of embryos were all as previously described (Gaunt et al. 1986).

Labelled RNA probes (riboprobes) were prepared to 3’ regions of coding sequence after subcloning genomic DNA (Hox-1.4 and Hox-5.1) and cDNA (Hox-2.6) fragments into transcription vectors (pGEM-1, Promega; or Bluescript pKSM13+, Stratagene). For Hox-1.4 and Hox-5.1, riboprobes were prepared from a sequence (about 700 bp) which extends 3’ of the homeobox BglII site (BglII-HindIII fragments). For restriction maps of these genes, see Duboule et al. (1986, for Hox-1.4) and Featherstone et al. (1988, for Hox-5.1). Hox-2.6 riboprobes were prepared from a sequence (about 700bp) which extends 3’ of the BamHI site (at nucleotide 1147 in Fig. 2 of Graham et al. 1988). Each of the three riboprobes used in in situ hybridization gave its own characteristic pattern of labelling (as described in detail in this paper), showing that there was no significant cross-reactivity under the experimental conditions employed. Identical patterns of labelling to those now reported were also obtained by using a shorter Hox-1.4 probe that did not include any of the homeobox sequence (a probe of about 200 bp, representing the 3’ end of the 700bp probe: AvaI–HindIII fragment), and by using a longer (1·4 kb) Hox-2.6 probe that included the 700 bp sequence described above but which also extended 5’ up to the homeobox BglII site. Although not shown in this paper, specificity of the probes for their respective genes has also been established by us in Northern blotting experiments.

For use in in situ hybridization, ³⁵S-labelled antisense RNA probes were synthesised in a direction opposite to that of normal transcription, as described by Gaunt et al. (1986) and Gaunt (1987). Control (sense) probes, synthesised in the direction of normal transcription, gave no specific labelling in in situ hybridization. The three gene probes were used in in situ hybridization experiments at similar specific activities and concentrations. Methods used for alkaline hydrolysis of labelled probes, in situ hybridization and autoradiography were all as previously described (Gaunt et al. 1986).

Hox-1.4, -2.6 and -5.1 transcript domains in the central nervous system were examined and compared in near-sagittal sections (Figs 1,2).

For each of the three genes, the anterior boundary of transcripts was found in the mid-myelencephalon. Between genes, small differences were found in the exact position and shape of the boundaries. Thus, the Hox-1.4 boundary (Fig. 2A) was located slightly posterior to the Hox-2.6 (Fig. 2B) and Hox-5.1 (Fig. 2C) boundaries, and a characteristic feature of the Hox-2.6 boundary was a dorsal tongue of strong labelling which usually extended anterior to the most anterior position of the Hox-5.1 boundary. From comparison with adjacent sections hybridized to a Hox-1.5 probe (not shown), the tongue of Hox-2.6 labelling did not extend as far forwards as the position of the Hox-1.5 boundary (e.g. Gaunt et al. 1988).

Posterior to the hindbrain boundaries, Hox-2.6 and Hox-5.1 probes usually gave an approximately constant intensity of labelling along the entire length of the central nervous system, but transcripts of both genes were often seen to be most concentrated in dorsal parts of the spinal cord (Figs 1C,D). In contrast, Hox-1.4 transcripts showed a marked anterior-to-posterior fall in abundance over the anterior region of the spinal cord (Fig. IB). This fall in transcript abundance was anterior to prevertebra 1 (pv1) in dorsal parts of the spinal cord, but posterior to pv5 in ventral parts. Posterior to the level of pv6, the abundance of Hox-1.4 transcripts was low in comparison to those of Hox-2.6 and Hox-5.1.

Since the abundance of homeo-gene transcripts may vary across the lateral and dorsoventral axes of the spinal cord (e.g. Utset et al. 1987; Toth et al. 1987; Holland & Hogan, 1988b), a complete analysis of Hox-1.4, -2.6 and -5.1 transcripts in the central nervous system would require examination of nervous tissue in all planes of section. We have not yet undertaken such a detailed comparison.

For each of the three genes, transcripts were most abundant within the cervical and anterior thoracic prevertebrae (Figs 1, 3). As judged from the density of silver grains, the intensities of prevertebral labelling given by the Hox-1.4 and Hox-2.6 probes were clearly greater than that given by the Hox-5.1 probe (Table 1).

Hox-1.4 transcripts (Fig. 3A) were absent in pv1, weak in pv2, and then strong in pv3. Hox-2.6 (Fig. 3B) and Hox-5.1 (Fig. 3C) transcripts were weakly detected in pv1, especially in ventral parts, and were then strong in pv2. Thus, the anterior boundary of the Hox-1.4 transcript domain was apparently located slightly posterior to that of Hox-2.6 and Hox-5.1. Pv1 showed, when cut in most planes of section, a lower cell density than other cervical prevertebrae (Figs 3D, 4C). Lower cell density did not account for the low labelling intensities detected for Hox-1.4, -2.6 and -5.1 since pv1 displayed strong labelling in parallel control experiments using the Hox-1.5 probe (Fig. 4B; see also Gaunt, 1988).

For Hox-1.4, a gradual and progressive decline in the abundance of transcripts was noted posterior to pv7, but the level of labelling remained above background even within the lumbar region (Figs 1B, 3A). For Hox 2.6, the anterior-to-posterior decline was more gradual (Figs 1C, 3B). For Hox-5.1, labelling intensity had fallen to background level by pv14 but a second, more posterior domain of prevertebral labelling was detected in the lumbar region (Figs 1D, 3C). In this posterior Hox-5.1 domain, labelling intensity increased from pv20 to pv25, declined from pv27 to pv30, and was then weak in more posterior parts. For Hox-2.6, but not Hox-1.4, we also noted a very slight increase in the abundance of transcripts in the lumbar region.

In the pharyngeal region, in planes just lateral to sagittal, each of the three genes displayed a characteristic and reproducible pattern of transcripts. Hox-1.4, as described earlier (Gaunt et al. 1988), was detected in the thyroid gland, the mesodermal derivatives (but not the lining endothelium) of the trachea, and in a structure that we tentatively identify as the thymus (Fig. 5B). In a nearby section, Hox-2.6 transcripts (Fig. 5C) were detected in all of these Hox-1.4-labelled structures and, in addition, were found in the pharyngeal floor posterior and anterior to the thyroid duct. Hox-2.6 transcripts did not, however, extend over the tongue. This pattern for Hox-2.6 is similar or identical to the pattern detected earlier for Hox-1.5 (Gaunt, 1988). Hox-5.1 transcripts (Fig. 5D) were detected in the thyroid and ‘thymus’ glands and in the floor of the pharynx posterior and anterior to the thyroid duct. In contrast to the results for Hox-1.4 and Hox-2.6, Hox-5.1 transcripts were not detected in the trachea. Hox-5.1 transcripts were also absent in posterior parts of the pharyngeal floor. We consider, however, that the Hox-2.6 labelling found in this region (and also found for Hox-1.4, but not in the plane of section shown in Fig. 5B; see Gaunt et al. 1988) could simply be due to transcripts present in anterior parts of the tracheal wall.

Hox-1.4 and Hox-2.6 transcripts were detected in the mesodermal derivatives, but not the lining endothelium, of the lung (Fig. 3A,B). These findings are consistent with observations published earlier for Hox-1.4 (Gaunt et al. 1988) and for Hox-2.6 (Graham et al. 1988), and they are typical of the transcript distribution found within the lung for several homeo-genes (e.g. Gaunt et al. 1988). Transcripts within the lung were more abundant for Hox-2.6 than for Hox-1.4 (Table 1). The Hox-5.1 probe gave only background levels of labelling over the day lung (Fig. 3C; Table 1).

Hox-1.4 and Hox-2.6 transcripts were abundant in the mesodermal derivatives, but not the lining endothelium, of the stomach (Fig. 6B,C; Table 1). This finding is in agreement with earlier reports (for Hox-1.4; Gaunt et al. 1988; for Hox-2.6: Graham et al. 1988) and is typical of the transcript distribution found in the stomach wall for a variety of other mouse homeo-genes (e.g. Gaunt et al. 1988). For Hox-2.6, but not Hox-1.4, weaker labelling was also detected in the mesentery underlying the stomach (Fig. 6C). In contrast to the results for Hox-1.4 and Hox-2.6, Hox-5.1 transcripts in the stomach were restricted to patches located in a peripheral zone of the wall (Fig. 6D). An identical pattern of punctate labelling has been noted before for Hox-2.1 transcripts in the intestine (Holland and Hogan, 1988b). This pattern probably indicates that Hox-5.1 transcripts in the day stomach are restricted to neural-crest-derived parasympathetic ganglia of the enteric plexus (Holland and Hogan, 19886), and are absent from mesodermal components (Table 1). The descriptions now given for the distributions of Hox-1.4, -2.6 and -5.1 transcripts within the stomach are also applicable to the patterns of transcripts observed in more posterior regions of the intestine (not shown). The liver was not labelled above background by any of the three gene probes (Fig. 6).

Both mesonephric and metanephric kidneys were labelled by all three gene probes (Fig. 6B,C,D). The metanephric kidney is not included within the section shown for Hox-2.6 (Fig. 6C) but was seen to be labelled in neighbouring sections. The most intense labelling within kidney tissue was given by the Hox-5.1 probe; the least intense labelling was given by the Hox-1.4 probe (Table 1).

At days, the gonad lies on the inner surface of the mesonephros (Fig. 6A). In the embryo under illustration, the presence of developing testis cords (which contain primordial germ cells and Sertoli cells; Fig. 6A) indicate that the gonad is a testis. No specific labelling of the day testis was given by the Hox-1.4 or Hox-2.6 probes (Figs 6B,C). However, strong labelling of the testis was given by the Hox-5.1 probe. Hox-5.1 transcripts were located mainly within the peripheral mesenchymal tissue of the testis, and were less intense or absent within the testis cords (Fig. 6D). This distribution of transcripts within the developing testis is similar to that found earlier for Hox-5.2 (Dollé and Duboule, 1989) and Hox-6.1 (Sharpe et al. 1988).

The mesodermal component of the day trachea, lung and gut is derived from the lateral plate mesoderm of earlier stages (e.g. Hogan et al. 1985). It has been assumed (e.g. Gaunt, 1988) that cells that express a homeo-gene at days are derived from cells that express the gene at much earlier stages of development. Following this assumption, we anticipated that Hox-1.4 and Hox-2.6 transcripts, abundant in the day trachea, lung and gut mesoderm, would be readily detectable in the lateral plate mesoderm at to days. We further reasoned that the absence of Hox-5.1 transcripts in lateral plate mesoderm derivatives at days (e.g. mesodermal components of the trachea, lung and gut) might follow an earlier absence of Hox-5.1 expression in the lateral plate mesoderm at to days.

To test the above hypotheses, we examined Hox-1.4, -2.6 and -5.1 transcript distributions in serial transverse sections of and day embryos. Unexpectedly, we found that all three genes were expressed in both somitic and lateral plate mesoderm (Figs 7, 8). In all cases, transcripts were confined to more posterior parts. Hox-5.1 transcripts (absent from mesodermal derivatives of the gut wall at days) were clearly detected in lateral plate mesoderm as it contacted and surrounded gut endoderm at days (Fig. 8D). Examination of additional serial transverse sections taken at all positions through and day embryos (not shown) did not reveal any clear differences between the three genes in the spatial limits of their transcripts. Hox-5.1 transcripts at these stages were present in lateral plate mesoderm, somitic mesoderm and neurectoderm at greater abundance than Hox-2.6 or Hox-1.4 transcripts (Table 1).

From the above findings, we considered it likely that the founder cells of lung and stomach mesoderm, present within the lateral plate mesoderm at to days, may be strongly positive for Hox-5.1 transcripts, but that with continued development to days there is selective loss of Hox-5.1 transcripts within these cell lineages. In contrast, we considered it likely that Hox-1.4 and Hox-2.6 transcripts may increase in abundance within these cell lineages. To test these hypotheses, we examined the day lung and stomach for Hox-5.1 and Hox-2.6 transcripts.

In contrast to results obtained for the day lung (Fig. 3C), the mesodermal component of the lung at days was found to be labelled by the Hox-5.1 probe (Fig. 9E; Table 1). The labelling was weak, but was clearly greater than the background labelling seen over the adjacent lining endothelium. Hox-2.6 transcripts were abundant within mesodermal derivatives of the day lung (Fig. 9C; Table 1).

At days, labelling given by the Hox-5.1 probe was weakly but uniformly distributed over the mesodermal components of the stomach wall (Fig. 9F; Table 1). The pattern of labelling observed was clearly different to the punctate pattern confined to peripheral parts of the stomach wall at days (Fig. 6D). The lining endothelium of the day stomach was not labelled by the Hox-5.1 probe. Hox-2.6 transcripts were abundant within mesodermal derivatives of the day stomach (Fig. 9D; Table 1).

Within the prevertebral column of the day mouse embryo, the anterior boundary of the transcript domain for each of the three genes examined was most obvious at the junction of pv1 and pv2. The exact position of the boundary was not, however, identical for all three genes, and was slightly more posterior for Hox-1.4 (which showed weak labelling of pv2 relative to pv3) than for Hox-2.6 and Hox-5.1 (which also showed some weak labelling in pv1). All three genes showed greatest abundance of transcripts over cervical prevertebrae 3–7, and an anterior-to-posterior decline of transcripts over the thoracic prevertebrae.

Within the central nervous system, the anterior boundaries of the Hox-1.4, -2.6 and -5.1 transcript domains were located in the mid-myelencephalon. As in the prevertebral column, the boundary for Hox-1.4 was apparently located at a position slightly posterior to that of Hox-2.6 and Hox-5.1.

These observations are consistent with results published earlier for Hox-1.4 and Hox-5.1 transcript domains within the prevertebral column and central nervous system (for Hox-1.4: Gaunt et al. 1988; for Hox-5.1: Featherstone et al. 1988), and for Hox-2.6 within the central nervous system (Graham et al. 1988, 1989) but they are an advance on earlier work since they permit accurate comparison of the domains on nearby sections of the same embryo.

The findings are in accordance with an earlier suggestion (Gaunt et al. 1988) that mouse homeo-genes within a subfamily display similar transcript domains in the embryo. Some additional support for this is also given by the apparently similar transcript boundaries displayed in the myelencephalon by Hox-1.5 (e.g. Gaunt et al. 1988) and Hox-2.7 (Graham et al. 1989). However, two members of another subfamily, Hox-2.5 (Graham et al. 1989) and Hox-5.2 (Duboule and Dollé, 1989) have clearly shown marked differences in the anterior boundaries of their transcripts within the central nervous system. Taken together, the findings suggest that subfamily similarities in transcript patterns (as have been found for members of the Hox-1.2, -1.4 and -1.5 subfamilies) may be a feature only of those genes that are expressed in anterior parts of the embryo, and are not a feature of genes expressed in more posterior parts (such as Hox-2.5 and Hox-5.2).

In addition to the similarities noted above, there were also some striking differences between the transcript patterns of Hox-1.4, -2.6 and -5.1. At days, these were observed mainly within some pharyngeal, thoracic and abdominal organs. These differences in labelling patterns have a practical value in demonstrating that there is no significant cross-reactivity between the three gene probes used in in situ hybridization.

Within pharyngeal tissue, Hox-2.6 and Hox-5.1 transcripts, unlike Hox-1.4 transcripts, extended anterior to the thyroid duct. This may be a consequence of the slightly more anterior position observed for Hox-2.6 and Hox-5.1 transcripts in the myelencephalon, since it is neural crest cells from this region that give rise to the mesenchymal tissues of the pharyngeal floor (Le Douarin, 1982).

Within mesodermal derivatives of the day trachea, lung and gut, Hox-1.4 and Hox-2.6 transcripts, but not Hox-5.1 transcripts, were readily detected. Conversely, Hox-5.1 transcripts, but not Hox-1.4 or Hox-2.6 transcripts, were abundant in the mesenchymal component of the day gonad. In addition to these qualitative differences in the patterns of Hox-1.4, -2.6 and -5.1 expression, quantitative differences in transcript abundance were also noted. As shown, for example, in Table 1, the day prevertebral column, mesonephric and metanephric kidneys each displayed clear differences in the relative abundance of Hox-1.4, -2.6 and -5.1 transcripts.

In addition to these tissue-dependent differences noted at days in the relative abundance of Hox-1.4, -2.6 and -5.1 transcripts, stage-dependent differences were also found (Table 1). Thus, although Hox-5.1 transcripts were apparently absent from the mesodermal component of the lung and stomach at days, Hox-5.1 transcripts were clearly detectable in these tissues at days. Lung and stomach mesoderm are derivatives of the lateral plate mesoderm present at earlier stages (e.g. Hogan et al. 1985). Lateral plate mesoderm of and day embryos was labelled strongly by the Hox-5.1 probe. We conclude, therefore, that Hox-5.1 transcripts displayed a fall in abundance from to days in tissues destined to form the mesodermal components of the lung and stomach. In contrast, Hox-1.4 and Hox-2.6 transcripts within these tissues displayed a rise in abundance from to days. Similarly, Hox-5.1 transcripts displayed fall in abundance within tissues that give rise to the prevertebral column (since the Hox-5.1 probe gave strong labelling of somitic mesoderm at days, but weak labelling of prevertebrae at days), while Hox-1.4 and Hox-2.6 transcripts displayed a rise in abundance within these tissues. These observations suggest, but do not prove, that genes within the Hox-1.4 subfamily might be coordinately regulated in their patterns of expression.

Two quite separate mechanisms could account for these tissue- and stage-dependent differences in expression now observed between genes within the Hox-1.4 subfamily. As one possibility, homeo-gene expression might be modulated within a cell lineage in a tissue- and stage-dependent way. This assumes, for example, that transcripts from each of the three genes are present at to days within all cells of the posterior lateral plate mesoderm. Then, with further development and divisions of one cell to form, say, mesodermal components of the stomach, transcript abundance falls for Hox-5.1 but rises for Hox-1.4. As a second possibility, Hox-1.4, -2.6 and -5.1 might be expressed in separate cell populations within the lateral plate mesoderm. Mesodermal components of the stomach would then develop from Hox-1.4 and Hox-2.6, but not Hox-5.1, -positive subpopulations. The in situ hybridization technique does not permit distinction between these two possibilities since it does not permit us to localize homeo-gene expression within individual cells.

We do not favour the second of the mechanisms discussed above. Most of the cells within the lateral plate mesoderm would, in this model, presumably be of the Hox-5.1-positive subpopulation. Yet Hox-5.1-positive cells apparently contribute little or nothing to derivatives of lateral plate mesoderm at days. Results obtained in Drosophila show that expression of homeo-genes may be modulated in a tissue- and stagedependent way. This is consistent with the first of our above proposals for the mouse. As one example of modulation of homeo-gene expression in Drosophila, some epidermal cells of parasegment 5 (ps5) initially express both Antennapedia (Antp) and Ultrabithorax (Ubx) transcripts, but at later stages groups of the cells express either Antp alone or Ubx alone (Carrol et al. 1988). As a second example of modulation, the epidermal cells of ps3 apparently switch from expressing Antp at early stages to Sex combs reduced (Scr) at later stages (Martinez-Arias et al. 1987; Carrol et al. 1988). In the following section, we assume that the changing patterns of homeo-gene transcripts now observed in tissues of the developing mouse are due to modulations in gene expression.

It is of interest to consider two models that have been proposed to describe homeotic gene activity in Drosophila. Both models may be applicable to observations made on the expression of mouse homeo-genes. The two models are related, and the second is able to accommodate observations made on modulation of homeo-gene expression.

The Lewis model (Lewis, 1978; Lawrence and Morata, 1983), originally suggested for genes of the Bithorax complex, proposes that each homeotic gene is switched on according to position along the anteroposterior axis of the embryo. The genes are postulated to be expressed in a series of partially overlapping domains. This gives each body segment a different combination of active homeotic genes, and it is this combination that defines segment identity. As earlier discussed (Gaunt, 1988; Gaunt et al. 1988), the patterns of embryonic expression observed for several different mouse homeo-genes conform closely to these predictions of Lewis’s model. Indeed, the patterns for mouse genes fit better than do those of some Drosophila genes, especially those expressed in more anterior parts of the fly (e.g. Labial, Proboscipedia and Scr: Mahaffey et al. 1989).

The ‘open for business’ model (Akam et al. 1988; Peifer et al. 1988) conforms to the formal structure of Lewis’s model, but differs from it in proposing that each homeotic gene is made accessible for transcription, or ‘open for business’, according to position along the anteroposterior axis of the embryo (rather than being simply switched on as in Lewis’s model). The molecular mechanism by which this is achieved is unknown (but see Akam et al. 1988). Modulations of homeo-gene expression in Drosophila are not easily reconciled with Lewis’s model, but they can be accommodated by the new model (see Akam et al. 1988). This model may similarly be of value in describing some of the new features of mouse homeo-gene expression now reported in this paper. Thus, Hox-1.4, -2.6 and -5.1 might best be considered as ‘open for business’ within the same spatial domain along the anteroposterior axis of the mouse embryo. The stage- and tissue-dependent modulations in expression that we have seen in structures within the domain, such as mesodermal components of the stomach and lung, might then be a consequence of other factors (so far unknown) which determine whether or not, or to what extent, a gene ‘open for business’ is actually switched on.

We, and others, have suggested that mouse homeo-gene expression might be required for specification of tissue identities according to position along the anteroposterior axis (e.g. Gaunt et al. 1988). We have now provided evidence that Hox-1.4, -2.6 and -5.1, derivatives of a common ancestral gene, have evolved to display both similarities and differences in their patterns of expression. Although we do not discount the possibility that the similarities found might indicate some overlap and redundancy of function between these genes, we suggest below two possible ways in which the differences observed might usefully increase the variety of instructions available in the developing embryo.

First, although all three subfamily members show basically similar domains of transcription, some small differences in their anterior boundaries were noted. These might specify fine differences in position along the anteroposterior axis. Second, at any given position along the anteroposterior axis of the embryo, a multiplicity of different structures pursue different pathways of development. We consider, therefore, that some mechanism must exist to provide different developmental instructions to different tissues located at the same anteroposterior position. Theoretically, a single homeo-gene could provide a variety of different instructions by changes in the nature or abundance of its transcripts. But we suggest that stage- and tissue‐dependent differences in expression within a group of homeo-genes possessing basically similar domains of transcription (as we have now found for genes of the Hox-1.4 subfamily) could potentially provide a much wider variety of developmental instructions effective at the same anteroposterior position.

We propose that one or both of the above mechanisms could provide a selective advantage to species evolving a more complex body structure, and could explain, at least in part, the advantage for the vertebrates to have duplicated the ancestral homeo-gene cluster (Duboule and Dollé, 1989; Graham et al. 1989; Akam, 1989). In terms of this hypothesis, less complex species, including Drosophila, would not require more than a single cluster.

Following from this, we can suggest a possible explanation as to why subfamily similarities in expression domains are not also observed for homeo-genes expressed in more posterior parts of the mouse embryo (Hox-2.5 and Hox-5.2; Graham et al. 1989; Duboule and Dollé, 1989). Posterior parts, being less complex than anterior parts, may not require such a wealth of developmental instructions effective at similar positions along the anteroposterior axis. Following duplication of the ancestral homeo-gene cluster, there would therefore have been less selective pressure upon ‘posterior’ genes to conserve the position of their transcript domains. Boundaries for posterior genes could then shift to acquire new positions along the anteroposterior axis. We consider, therefore, that posterior genes of the Hox clusters might be viewed as forming collectively a unique, functional homeo-gene complex.