recently been confirmed by secondary structure predictions of the Antennapedia (Antpj homeodomain using NMR a n a l y ~ i s ( ~and - ~ ) the engrailed (en) homeodomain-DNA complex by x-ray crystallography@). Therefore, much of the interest in this family stems from the idea that proteins containing a homeodomain function as transcriptional regulatory factors in which the target specificity is mediated at least in part by the sequence-specific DNA binding properties of the homeodomain(5). Homeodomain roteins have been identified in many different species( ','I. They may be grouped into different families or classes based on sequence identities both within the homeodomain and the flanking regions. The largest group is the Antp-class and examples of other groupings are the engrailed, paired, even-skipped, and POU classes('). The Antp-class comprises homeodomains which share approximately 60% identity to the version of the homeodomain first identified in the Drosophilu Antennapedia protein. However a distinguishing feature of the Antp-class is that within this background of variation all members also have an identical core sequence of 12 amino acids between positions 4-55 in alpha-helix I11 of the home~domain('.~).The NMR, crystallographic and transcription data reveal that it is this region of the homeodomain that makes contact with DNA and is involved in determining many of the sequence-specific DNA binding characteristics of the h~meodomain(~). Therefore members of this class are believed to have a high general affinity for very similar target recognition sequences, and there is a distinct opportunity for competitive or cross-regulatory interactions between members of this class and their target sequences. In addition to the identities within the homeodomain, Antp-class proteins also usually have a short motif of 5-6 amino acids, containing the sequence Y-P-W-M, which is encoded in the exon just 5' of the homeodomain exon(8-10). The Antp-class genes are organised into chromosomal clusters. In Drosophilu all Antp-clads genes are located on chromosome 3 in the Antennapedia and Biihorux complexes, which to ether are known as the In vertebrates the Homeotic complex, H0M-C(F1312). most extensive anal sis of Antp-class genes has been done in the ~ n o u s e (and l ~ human(L4715), ~ where there are four un-linked complexes, HOX-l through Hox-4. Emerging data on the related genes in other vertebrates suggests that the organisation of the Antp-class Hox genes into four separate clusters is a common property of all vertebrates, not just of mammals. Parsimony analysis of homeodomains in Drosophila HOM-C(") and mouse H ~ x ( ~ ~ ) c o m p l esuggests xes that the tandem arrangement of these homeobox genes along the chromosome arose by the amplification and duplication of a single ancestral gene. In vertebrates the initial duplication phase was followed by a second series of steps whereby the chromosomal regions were further expanded and subsequently diverged to generate the
P,
Summary One of the most remarkable recent findings in developmental biology has been the colinear and homologous relationships shared between the Drusuphila HOM-C and vertebrate Hox homeobox gene complexes. These relationships pose the question of the functional significance of colincarity and its molecular basis. While there was much initial resistance to the validity of this comparison, it now appears the HoxlHOM homology reflects a broad degree of evolutionary conservation which has reawakened interest in comparative embryology and evolution. The evolutionary conservation of protein motifs in many gene families (including those for growth factors, secreted and membrane bound signalling factors, adhesion molecules, cytoplasmic receptor kinases, nuclear receptors and transcription factors) has lead to speculation on the extent to which these homology relationships represent common developmental processes and underlying molecular mechanisms. Structural identities in a protein may indicate the hiochemical/molecular function that a protein plays in cellular and developmental processes, without reflecting a conserved role in a cascade of developmental events. However, the analysis of genes encoding transcription factors has provided evidence suggesting that there are gene complexes in arthropods and vertebrates which are true homologues and which may share common roles in the specification of regional identity along embryonic AP axis. These genes comprise the Hux/HOirM-C homeotic complexes. This review will detail some of the evidence for this proposed relationship and will speculate on the functional implications. Introduction: The HOM-C and Hox Gene Complexes The homeobox gene family is a large and increasingly well characterized group of genes which contain sequences that display some degree of similarity to a short 180bp motif, termed the homeobox, as it was originally identified in several Drosophila homeotic selector and segmentation genes(','). The homeobox encodes a domain of 60 amino acids related to the helixturn-helix regions present in several yeast and prokaryotic DNA binding proteins. This relationship has
four independent Hox c l ~ s t e r s ( l ~ This - ~ ~ )process . was not restricted to the Hox clusters, but involved large chromosomal regions flanking the complexes, as evidenced by the conservation of chromosomal order of not just the Hox clusters but also linked genes including members of the keratin, collagen, and nerve growth factor Homologues and Paralogues in the Hox/HOM-C Complexes One consequence of the generation of Hox clusters by duplication of an ancestral complex is that specific genes in each of the different complexes are evolutionarily related to each other, forming a subfamily or paralogous group. Comparisons of the degree to which this identity has been maintained among paralogues, therefore provides insight into the evolutionary constraints placed on the Hox network. Figure 1 illustrates the similarities that define the paralogous relationships for a specific subfamily. The homeodomain of this subfamily has approximately 80% identity with that of the Antp homeodomain. However, it is clear that these three proteins (Hox-2.6/1.4/4.2) are further related to each other based on the position and identity of changes in their sequence when compared directly with the Antp sequence (Fig. 1A). Note the common substitutions of P/S/”r/A/a/v at positions 1/4/6/7/11/13 of the homeodomain in all members. The sequence identity extends outside of the homeodomain in two additional regions of the proteins. The first region can be referred to as an ‘extended homeodomain’. Little identity exists on the carboxy terminal portion of the protein flanking the homeodomain. However on the amino terminal side of the homeodomain there is a region of identity which extends upstream of the adjacent splice site into the exon encoding the hexapeptide or Y-P-W-M motif. The second region of identity outside of the homeodomain resides in the amino terminal region of the proteins where there is a block of 20-30 conserved amino acids (Fig. 1B). There are 13 paralogous groups in the human and mouse Hox clusters that have been identified b similar sequence comparisons of related members(1”1s7.Figure 2 shows a summary of these comparisons in an alignment of the four Hox complexes. The numbers at the bottom represent the different paralogous groups and the vertical rows above each number represent related genes in a subfamily. It is clear that all of the paralogous groups are not represented in each cluster. This may reflect the fact that some members of a cluster were not duplicated during the events that led to the formation of multiple complexes. However, analysis of the relevant chromosomal regions has not been performed at the sequence level, and it is equally possible that these gaps represent genes which were originally duplicated and have become highly diverged or lost during vertebrate evolution.
Sequence comparisons of the mouse Hox genes have been extended to those in lower vertebrates and it is apparent that there are closely related homologues. For example, the mouse Hox-2.1 gene has an equivalent in zebrafish that displays nearly 90% amino acid identity over the sequence of their entire predicted proteins(10*22). This high degree of conservation between species does not apply to all of the Hox genes, however, and it is often difficult to clearly assign homology relationships. For example, Xenopus Xhox7A is the homologue of the mouse Hox-2.6 gene but alignment of their two proteins reveals several deletions and large blocks of divergent amino acidd2”).In fact the Xenopzis X h o x l A protein is only homologous to HOX-2.6in the same regions in which the mouse paralogues are related to each other(20323-2’). It is only on the basis of its chromosomal position in a cluster of other Xenopus genes, which have more extended homology to mouse Hox-2 members, that a true homology relationship between Hox-2.6/XhoxlAis confirmed, However, an interesting point is that even in these cases of relatively high divergence the identity is conserved in the amino terminal region of the proteins and in the extended homeodomain region. This suggests that these identities may represent constraints to maintain important functional domains of the proteins. It is perhaps not surprising that hornologues exist in other vertebrates, but the observation that sequence identities are shared between mouse and Drosophila Antp-class genes suggests that the Hox genes might be true vertebrate homologues of the Drosoplzila IJOM-C homeotic g e n e ~ ( l”).~ . Figure ~ ~ 1shows that the specific groupings of amino acids in the homeodomain and amino terminal region of the Hox proteins in paralogous group 10 are also conserved in the Drosophila deformed (DfD) protein(20.2z~25). Overall the Dfd protein is 40% identical to HOX-2.6 and the multiple regions of identity between the sequences are colinear in the respectivc proteins (Fig. 1C). This strongly supports the idea that the paralogous group containing Hox-2.6/1.4/3.5/4.2 is a vertebrate Djd family. No one member of the four mouse genes is more related to Dfd than another, so it is not possible to designate a specific Dfd homologue. It is possible to confidently define homologous relationships between mouse Hox genes and the Drosophila labial (lab), proboscipedia (pb De ormed (Dfd) and abdominal-B ( Ahd-B) genes 19-21. 3,25-29), Furthermore when the mouse paralogues are aligned with their most closely related Drosophila counterparts, it is apparent that the linear order of related genes along the chromosome is the same in both the Hox and HOMC c o m p l e ~ e s ( ’ ~ ~ These ~~,~~ clear , ~ ~correlations ). are indicated by the grey shading and bold brackets in Figure 2. It is more difficult to define distinct homology relationships (dashed brackets, Fig. 2) between specific mouse groups and the Drosophilu abdominal-A (abdA), Ultrabithorax (Ubx), Antennapedia (Antp),and Sex combs reduced (Scr) genes, because they all share a / J
2f
Fig. 1. Homologues and paralogues of the vertebrate Deformed subfamily. A. Comparison of the homeodomains of three mouse genes, the Xenopus XhoxlA and Drosophilu Deformed homeodomains with that of Antp. Dashes represent identical amino acids. B. Amino acid indentities in the amino terminal regions of the proteins in (A). C. Alignment of the mouse, Xenopus and human homologues of the HOX-2.6gene. D. Alignment of the mouse Hox-2.6 protein and Dfd showing colinear regions of identity.
high degree of identity within the homeodomain. This does not necessarily imply that there are no direct homologues, but rather that the present criteria for
distinguishing ancestral relationships between these genes is not sufficient. For example, Hux-2.1has 12 of 14 residues at its amino terminus identical to those in
Drosophila ANT-C (3)
Hox-2 (1 1) Mouse HOX2 (17)
Human
Hox-1 (6)
Mouse
HOX1 (7)
Human
Hox-3 (15) Mouse HOX3 (12)
Human
Hox-4 (2)
Mouse
HOX4(2)
Human
Paralogous Subgroups Fig. 2. Homology relationships between the mouse and human Hox complexes and the Drosophila HOM-C complex. The mouse nomenclature for the genes are listed above the boxes and the human nomenclature below the boxes. The 13 paralogous groups are marked at the bottom. The vertical rows represent evolutionarily related genes. The grey shading and bold brackets identify groups where clear homology relationships are defined. Thc dashed brackets define groups that are related with lowere degrees of confidence. The chromosomal locations of the respective human and mouse clusters are indicated in brackets at the right. (Based on refs 15, 19, 29).
Scr, but Hox-2.2 and Antp also share most of these same sequences. These structural relationships strongly support the idea that the mouse Hox and Drosophila HOM-C homeobox complexes are derived from a common evolutionary ancestor of both arthropods and vertebrates. This ancestral organism appears to have contained a cluster of at lcast five Antp-class homeobox genes related to lab, pb, Dfd, Scr/Antp/Ubx/abd-A, and Abd-B genes (Figure 3 ) . The striking similarities in HoxIHOM-C presumably reflect some degree of selective presure for retention of the complexes. perhaps related to maintenance of their expression properties(12.'9'2' >30) An interesting observation from the extended structural comparisons is that homologues across species (Drosophila to human) and paralogues within a species all share the same identities in the amino terminal and extended homeodomain regions of their proteins. The implication is that these domains are essential for the function of these specific groups. However, the possibility exists that genes within a paralogous group may be partially redundant and show
functional compensation. Do all members of a paralogous group have the same function and target genes, and if not where does the specificity reside? Most of the experimental evidence suggests that the extended homeodomain region determines the specificity of DNA binding and target specificity, a5 changes as small as one amino acid to the homeodomain can alter the ~ p e c if ic ity ( ~ '-However ~~). no clear role for the identities at the amino terminal end of the protein have been put forward. Protein domain swapping experiments suggest that non homcodomain regions are involved in transcriptional activation. and it is possiblc that the amino terminal domain functions to modify this activity. However, another possibility comes from experiments on the POU class of homeodomains. It has recently been shown that despite nearly identical protein sequence in the homeodomain between Oct-1 and Ort-2, the alternation of a few amino acids can change the ability of this domain to interact with thc HSV transactivator protein VP16("736). This factor forms part of a multiprotein-DNA complex with the POU homeodomain proteins and are needed for transcriptional activation of a targct gene. These
arthropods A M 4
AM-A
Ubx
Antp
Scr
Dfd
PB
Lab
0 0 0 25
2 4
23
2 2
2 1
26
27
2.8
2 9
vertebrates
5’
posterior late
3
-b anterior
early
Fig. 3. Theoretical evolution of the H0xIHOM-C complexes from a common ancestor. At the bottom is a diagram illustrating that colinear pattcrns of expression have also been conserved in vertebrates, and this suggests a common role for these clusters in the regional specification of axial diversity.
experiments illustrate the importance of co-factors or accessory proteins in homeodomain activity, and show that even highly related sequences can have dramatically different target specificity. Therefore the amino terminal identities in the paralogous groups could serve to modify interactions with similar types of cofactors, and caution should be used in assuming that the minor sequence differences in the homeodomain of paralogues are not functionally significant. Differences between Hox and HOM-C Complexes While attention is often focussed on the homologies, the differences between the HoxlHOM-C complexes are equally informative. There is not a direct one-toone correspondence between the Hox and HOM-C members. For example, in Drosophilu the four genes abd-A, Ubx, An@ and Scr could correspond to the four paralogous groups (6-9) in mouse which are in a similar position in the respective clusters. but as mentioned above there is little data to support definitive relationships. It is equally possible that these separate families of genes represent amplification and diver-
gence events, from a single Antp related gene in the commoii ancestor, which have occurred independently in arthropods and vertebrates. Independent evolutionary differences between mouse and Drosophilu can be seen by examining mouse paralogous groups 1-5 and 11 (Fig. 2). In the Dromphila BX-C complex, there are no Antp-class segment identity genes upstream of Abd-B, but in the mouse it has recently been shown that there are five It has paralogous groups equally related to therefore becn argucd that either Drosophilu has lost these members or that this expanded Abd-B family represents a recent amplification s ecific to vert e b r a t e ~ ( ~Similarly ~). the Hox-2. subfamily is often proposed to be related to the Drosophilu Zen gene, which is not a true homeotic segment identity gene(37).This is largely based on relative positions in the complexes; however, there is n o good structural identity between this paralogous group and Zen. This suggests that this a vertebrate specific subfamily or that the Droso hilu counterpart has been lost during evolution(2 1. There are many other differences in the structural organisation of the Drosophilu and mouse complexes. The mouse Hox clusters have approximately 10 genes all with the same transcriptional orientation tightly grouped in a region spanning 120 kb. The Antp gene alone is this size and the region of HOM-C span greater than 600 kb. In Drosophilu, Dfd has been inverted and is oriented in the opposite transcri tional direction from the other Antp-class genes(11325 . In Drosophilu, HOM-C is divided into two separate complexes and there are many other specialized genes (some containing homeoboxes, bicoid, ftz) located in the complexes. This raises the question as to which type of organisation more accurately reflects the ancestral cluster. Genetic and molecular studies in another insect, the red flour beetle (Tribolium custuneum), have provided some insight into the problem. In this system there is a cluster of Antp-cluss genes structurally and functionally homologous to the Drosophila ANT-C and BX-C complexes, which are grouped in a single chromosomal ~ l u s t e r ( ” ~Several ~ ~ ) . of the specialised genes such as ftz appear to be missing and it will be interesting to determine if there are other members in the A h d - B and Zen positions. Overall the organisation in the red flour beetle appear9 to be more similar to the mouse than of Drosophilu. Therefore the Drosophilu HOM-C complex may not be the ideal choice for comparison with the vertebrate Hox family, as they have become more highly diverged incorporating many genes of alternate and specialised function. Further studies are needed to investigate this complex in other representatives of common evolutionary origin. In this regard it is interesting that a clustered group of at least four Antpclass genes, including members of the lab, Dfd, ScrlAntp, and AntplAbd-B like families, have been found to be conserved in the nematode C. eleguns (reviewed in ref. 40). While the homologies are not as
7,,1.5p.l
R
P
clearly documented, it appears that some form of this homeobox gene cluster existed in nonsegmented ancestors and must have played an important function in developing metazoans.
Relationship of Evolution to Function On the premise that the conservation of HoxIH0M-C structural organisation reflects a functional requirement, what would be the nature of this common functional role in development? Segmentation is widely believed to have evolved independently in different species and the ancestral An@-class cluster appears to be present in non-segmented organisms. Insight into this dilemma has come from the analysis of the spatially-restricted patterns of Hox gene expression during development. The patterns of expression are very diverse but a hallmark of Hox expression, as analyzed by in situ hybridisation in mouse embryos, is that different genes have sharp boundaries of expression along the anteroposterior (A-P) axis (reviewed in refs 13,41). These boundaries can vary from tissue to tissue, hence the patterns do not simply mark a common axial position in all tissues of the embryo unless these differences can be accounted for by tissue migrations after establishing the axial level. However, there is a correlation between these A-P boundaries of expression and the chromosomal organisation of the Hox clusters (see Fig. 3 ) . Genes in the Abd-B paralogous group have boundaries of A-P expression that are posterior relative to other members of the complexes. Moving along the chromosome from these Abd-B paralogues, each successive member in an individual cluster has a rogressively more anterior boundary of expression( 9,21.41,422). Hence the gene order is the same as the A-P order of expression along the embryonic axis. This property has been termed colinearity, b analogy to the relationship first described by Lewis(43x between genetic function and organisation of the Bithorax complex in Drosophila. Lewis proposed that this colinearity was part of a combinatorial code used to generate regional diversity. The colinear and highly ordered domains of expression have been observed in mouse embryos in an increasinp variety of contexts. These include: neural tuhe(1',2 .28,42.44), neural crest and branchial arches(28'45546), mesoderm(29,41,42,51-53) and gonads(s4). Furthermore this colinearity is not just limited to spatially-restricted domains of expression, but also is observed with respect to temporal pattern^(",^^.'^) and to the response of Hox genes to retinoic acid in cultured ~ e l l s ( ~ ~and -j~) Xenopus embryo^(^^'^^). Together the large number of tissues with colinear domains suggest that the Hox patterns could be part of a molecular mechanism for the specification of positional values along the embryonic axis. The different combinations of Hox genes would provide a molecular framework or code(28'33) to generate regional
P
diversity in a manner analogous to that proposed by ~). in mouse Hox Lewis for D r o ~ o p h i l u ( ~Mutations genes have directly demonstrated that these genes have normal roles in patterning axial structures which correlate with the spatially-restricted patterns of e x p r e ~ s i o n ( ~ ~ .~ "Th ~ e~ *common ) functional link between Hox and HOM-C therefore may lay in the mechanisms of axial patterning, not ~egmentation(~*). Recently it has been shown that mouse Hox genes do have patterns of expression which are restricted to segments in the hindbrain(28,44~4s,61) and that these patterns can be conserved in other vertebrates(62.6'). These findings suggest that the Hox genes might have a role in segment identity in the hindbrain, like their Drosophila counterparts. However, this is not a general property of the Hox genes as expression is observed in many nonsegmented structures even with in the neural tube. Therefore, association of the Hox/HOM-C with processes in segmentation may represent a co-opting of this system from its fundamental role in axial patterning.
Relationship of Evolution to Hox Regulation Despite the structural similarity between paralogous genes in the mouse, the recent phenotypes arising in Hox r n ~ t a n t s ( ~ ' ,demonstrate ~) that at least some functions of an individual member must be unique. If paralogous members are expressed in very different domains then there is little possibility for redundancy. Many studies have directly compared the patterns of paralogous members (for review see ref. 41), and the results vary considerably depending on the tissue. In the hindbrain and neural crest components of the branchial arches the Hox paralogues have identical anterior boundaries and segmentally-restricted domains of ex ression. with the exception of the lubiul subfamily(*' P ,45). There are, however, differences in the levels of expression within these domains between some members. This clearly implies that the regulatory regions imposing the segment-restricted domains are the same in each of the four clusters and have been conserved following the duplication of the ancestral complex in vertebrates which gave rise to the Hox clusters. However, it is important to point out that even though genes may have identical A-P expression boundaries they may vary with respect to other axes and that not all cells in the domain will be positive for each of the paralogous members. The Hox-2 genes have dynamic dorsoventral (D-V) restrictions in the neural tube that alter with time during development(@). Members of other Hox clusters also display D-V variations which are different from those of Hox2(41,s2,64). In the limbs and somites, a different pattern also emerges compared with the hindbrain. Paralogous members of a group may have ver different A-P expression boundaries in somites(41;4,53.65), and expression domains of members of the HOX-4and I3o.x-1
complexes map to different regions of the axis in the limb(47-50). These variations in distribution and timing argue that many of the regulatory controls for the separate Hox complexes have not been rigidly conserved. Regulation in a general sense could form an important basis for maintaining the linkage of Hox genes, because the spatially-restricted domains of expression are important for gene function. Is this concept supported by current data? There are examples of complex transcription patterns reflected in multiple transcripts from the same gene, multiple promoters, differential splicing and polyadenylation, and interspersed regulatory regions. One possible explanation for these properties is that each Hox gene may not simply function as an independent gene and might share regulatory regions with other members of the complex@6).Experiments in transgenic mice have shown that some isolated Hox/lczc 2 genes can function normally outside of their chromosomal environment(67-6y), but there is also evidence for enhancers that work over a distance and they may influence several Thcrefore there is a strong possibility that one of the constraints on the organisation of the Hox family is a need to preserve the relationships or positions of genes in the clusters in order to achieve the finely regulated patterns of expression.
Conclusions The dramatic similarities in structure, organisation and colinear expression between the vertebrate Hox and Drosophila HOM-C complexes clearly argue for a common evolutionary origin. A broad variety of genetic, molecular, and descriptive evidence suggests that the common functional basis for this conservation reflects the possibility that this set of genes is part of a molecular mechanism for generating regional diversity along the A-P axis. The offsets in some tissues and not others, and the different morphological properties of the cellular expression domains suggest that the H o x genes can not have a simple uniform patterning role in all embryonic contexts. This implies that different sets of target genes must be utilized in somites versus neural tube, and it is perhaps better to think of the functional role of these Hox genes as not involving a single defined output but in more general terms of providing regional differences. It is clear that the Hox family is going to provide an interesting paradigm for examining regulatory hierarchies in vertebrates and the extent to which gene families can help to define common processes in evolution. To properly build upon these findings more work is needed in comparative evolution to determine what the organisation of homeotic complexes are like in primitive organisms. Are the vertebrate clusters most closley associated with functional roles in the development of the nervous system, segmentation, or axial polarity? Despite the exciting observation that colinear-
ity exists there is still not a good explanation of its molecular basis. l t is hoped that investigations into the regulatory requirements using transgenic mice and biochemical approaches will address this problem and open the way to identifying other parts of the regulatory cascade in pattern formation.
References 1 Gehring, W. (1987). Homeoboxes in the study of development. Science 236,
124-1252, 2 Scott, M. P., Tamkun, J. W. and Hartzell, G. W., cd. (1989). The structure and function of the homeodomain. Eiochim. Eioyhys. Actu. 959, 25-48. 3 Otting, G., Qian, Y. Q.? Billeter, M.,Muller, M., Affulter, M., Gehring, W. J. and Wuthrich, K. (1990). Protein-DNA contacts in the structure of a homeodomain-DNA complex determincd by nuclear magnetic resonancc spectroscopy in solution. EhfBO J. 9 . 3085-3092. 4 @an, Y. Q., Billeter, M., Otting, G., .Wuller, M., Gehring, W. J. and Wuthrich, K. (1989). The structure of the Antennapcdia homeodomam determined by NMR spectroscopy in solution: comparison with prokaryotic repressors. Cell 59. 573-580. 5 Gehring, W., Muller, M., Affolter, M., Percival-Smith, A., Rilleter, M., Qian. Y., Otting, G. and Wuthrich. K. (1990). The structure of the homeodomain and its functional implications. TIG 6. 323-329. 6 Kissinger, C., Liu, B., Martin-Blanco, E., Kronherg, T. and Pabo, C. (1990). Crystal structure of an fngrailed homeodomain-DNA complex at 2.8A resolution: A fraoiework for understanding homeadomain-DNA interactions. Cell 63, 579-590. 7 Holland, P. and Hogan. B. (1986). Phylogenetic distribution of antennapedialike homeo boxes. Nature 321, 2.51-253. 8 Mavilio, F., Simeone, A., Giampaolo, A., Faiella, A., Zappavigna, V., Acampora, D., Poiana, G.. Russo, G . , Peschle, C. and Boncinelli, E. (1986). Differential and stage-related expression in embryonic tissues of a new human homco box gene. Noturc 324, 664-667. 9 Regnlski, M., Harding, K.. Kostriken, R., Karch, P., Levine, M. and McGinnins, W. (1Y85). Homeo box genes of the Anlennapedia and Bithorax complexea of Dvosophila. CeIl43, 71-80, 10 Krumlauf, R., Holland, P., McVey, J. and Hogan, B. (1987). Dcvclopmental and spatial patterns of cxprcssion of the mouse homeobox gene, Hox 2. Developmmt 99, 603-617. 11 Akam, kl. (1987). The molecular basis for nictameric pattern in the Drosophila emhryo. DeveIopment 101, 1-22. 12 Akam, kl. (1989). Hox and HOM: homologous gene clusters in insects and vertebrates. Cell 57. 347-349. 13 Kessel, M. and Gruss, P. (1990). Murine developmental control gcnea. Science 249. 374-379. 14 Roncinelli, E., Acampora, D., Pannese, M., D’Esposito, M., Somma, R., Gaudino, G., Sturnaiuolo, A., Cafiero, M., Fairlla, A. and Simeone, A. (1989). Organization of human class I homeobox genes. Genome 31, 745-756. 15 Boncinclli, E., Simeone, A.. Acampora, D. and Maviliu, F. (1991). HOX gene activation by retinoic acid. TIG 7. 329-334. I6 Akam. M., Dawson, 1. and Tear, G. (1988). Homeotic genes and the control of scgment diversity. Det’elopmenf 104. 1 17 Kappen. C., Schugart, K. and Ruddle, 9). Two stcps in the evolution of antennapedia-class vertebrate homeobox Genes. Proc. Nafl.Acad. Scf. U S A 56. 54.59-5463. 15 Schugart, K., Kappen, C. and Ruddle, F. (1989). Duplication of large gcnomic regions during the cvolution of vertebrate homeobox genes. Proc. Null. Acad. Sci. USA 86. 7067-7071, 19 Graham, A., Papalopulu, N. and Krumlauf, R. (1989). The murine and L)rosophila homeobox clusters have common featurcs of organisation and cxpression. Cell 57, 367-378. 20 Graham, A., Papalopulu, N., Lorimer, J., McVeg, J., Tiiddenham, E. and Krumlauf, R. (1988). Characterization of a muvine homeo box gene, Hox 2.6. related to the drorophila deformed gene. Genes Dev. 2, 14241438. 21 Duhoule, D. and Uolle, P. (1989). ‘The structurai and functinal organization of the murine HOX gene family rcscmbles that of Drosophila homeotic genes. EMBO J. 8. 1497-1505, 22 Njolstad, P. R., Molven, A. and Fjose, A. (1988). A zebrafish homologue of the murine Ifox-2.1 genc. FEES Lett. 230. 25-30. 23 Featherstone, M. S., Baron, A., Gaunt, S. J., Mattei, M.G. and Duboule, D. (1988). Hox-5.1 defines a humeobox-containing gene locus on moure chromosome 2. Proc. Nail. Acad. Sci. USA 85, 4760-4764. 24 Harvey, R. P., Tahin, C. J. and Melton, D. A. (1986). Embryonic cxprcssion and nuclear localization of Xenopus homeobox (Xhox) gene products. E M B O J. 5, 1237-1244. 25 Regulski, &McGinnis, I., N., Chadwick, R. and McCinnis, W. (1987).
Developmental and molecular analysis of Dcformed: A homeotic gene controlling Drosopltzla head development. EMBO J. 6, 767-777. 26 Ruhnck, M., Larin, Z., Cook, M., Papalopulu, N., Krumlauf, N. and Lehrach, H. (1990). A yeast artificial chromosornc containing the mouse homeobox cluster Hox-2. Proc. Natl. Acad. Sci. USA 87, 4751.4755, 27 Frohman, M., Boyle, M. and Martin, G. (1990). Iqolation oI the mouse Hox2.9 gene: analysis of embryonic expression suggests that positional information along the anterior-postcrior axis is specified by mesoderm. Development 110, 589-607. 28 Hunt, P., Culisano, M., Cook, M., Sham, M., Faiella, A., Wilkinson, D., Boncinelli, E. and Krumlauf, R. (1991). A distinct Hox code for the branchial rcgion of the head. 8Nature353, 861-864. 29 Izpisua-Belmonte, J., Falkenstein, H., Dolle, P., Renucci, A. and Dubnule, D. (1991). Murine genes related to the Drosopkila A b d B homeotic gene are sequentially expressed during development of the posterior part of the body. EiV.IRO. .I. 10, 2279-2289. 30 McGinnis, W. and Krumlauf, K. (19%). Homeobox genes and axial patterning. Cell 68, 283-302. 31 Treisman, J., Gonczy, P., Vashishtha, M.,Harris, E. and Desplan, C. (1989). A singlc amino acid can determine the DNA binding specificity of homeodomain proteins. Cell 59, 553-562. 32 Kuziurd, M. A. and McGinnis. W. (1989). A homeodomain substitution changes the regulatory specificity of the deformed protein in Drosophila embryos. Cell 59, 563-571. 33 Mann, R. S. and Hogness, D. S . (1990). Functional dissection of Ultrabithorax proteins in D. melanogaster. Cell 60, 597-610. 34 Gibson, G., Scbier, A., LeMotte, P. and Gehring, W. J. (1990). The specificities of Sex combs reduced and Antennapedia are defined by a distinct portion of each protein that includes the homeodomain. Cell 62. 1087-1103. 35 Kristie, T. and Sharp, P. (1990). Interactions of the Uct-1 POU subdomains with specific DNA sequences and with the IISV alpha-truns-activator protein. Genes Dev. 4, 2383-2396. 36 Stern, S., Tanaka, M. andHerr, W. (1989). The Uct-I homeodomain directs formation of a multiprotein-DNA complex with the HSV transactivator VP16. Nulure 341, 624-630. 37 Kushlow, C. and Levine, M. (1988). Combinatorial expression of aftz-zen fusion promoter suggests the occurrence of cis interactions between genes ofthe ANT-C. EMBO J. 7. 3479-3485. 38 Beeman, R. (1987). A homeotic gene cluster in the red flour beetle. Nature 327, 247-249. 39 Stuart, J . , Brown, S., Beeman, K. and Dcnell, R. (1991). A deficiency of the homeotic complex of the beetle Tribolium. Nature 350. 72-74. 40 Kenyon, C . and Wang, B. (1991). A cluster of Antennapeditirlarr homeobox genes in a nonsegmented animal. Science 253, 516-517. 41 Gaunt, S. J. (1991). Exprcssion pattcrns of mouse Hox genes: Clues to an understanding of developmental and evolutionary strategies. BioEssays. 13, 505-513. 42 Gaunt, S. J., Sharpe, P. T. and Duboule, D. (1988). Spatially restricted domains of homeo-gene transcripts in mouse embryos: relation to a segmented body plan. Development 104 (Supplement: Mechanisms of Segmentation), Eds: V. Frcnch, P. Ingham, J. Cooke and J. Smith), 169-179. 43 Lewis, E. (1978). A gcnc complex controlling segmentation in Drosophilu. ivutNre276, 565-5711. 44 Wilkinson, D., Bhatt, S., Cook, M., Boncinelli, E. and Krumlauf, R. (1989). Segmental expression of hox 2 homeobox-containing gcncs in thc developing mouse hindbrain. IVaiure 341. 405-409. 45 Hunt, P., Wilkinsun, D. and Krumlauf, H. (1991). Pattcrning thc vcrtebrate head: murine Hox 2 genes mark distinct subpopulations of premigratory and migrating ncural crest. Development 112. 43-51. 46 Hunt, P. and Krumlauf, R. (1991). Deciphering the Hox Code: Clues to patterning the branchial region of the head. Cell 66, 1075-1078. 47 Dolle. P., Izpisua-Belmontc, J. C., Falkenstein, H., Renucci, A. and Duhoule, D. (1989). Coordinate cxpresaion of thc murinc Hox-5 complex homoeobox-containing genes during limb pattern formation. Nature 342, 767-
772. 48 Izpisua-Belmonte, J.-C., Tickle, C., Dolle, P., Wolpert, L. and Duboule, D. (1991). Expression of homeobox Hux-4 genes and thc spccification of position in chick wing development. Narure 350: 585-589. 49 Nohno, T., Noji, S., Koyama, E., Ohyama, K., Myokai, F., Kuroiwa, A., Saito, T. and Tanapchi, S. (1991). Involvement of the Chox-4 chicken
homeobox genes in determination of anteroposterior axial polarity during limb development. Cell 64, 1197-1205. 50 Yokouchi, Y., Sasaki, H. and Kuroiwa, A. (1991). Homeobox gene expression correlated with the bifurcation process of limb cartilagc developmcnt. Nature 353, 443.445. 51 Gaunt, S. J. (1988). Mouse homeohox gene transcripts occupy different but ovcrlapping domains in emhryonic germ layers and organs: a comparison of Hox3.1 and HOX-1.5.Development 103, 135-144. 52 Gaunt, S. J . , Coletta, P. L., Pravtcheva, D. and Sharpc, P. T. (1990). Mowe Hox-3.4: homeobox sequence and embryonic expression patterns compared with other members of the Hox gene network. Devcltpnvnt 109, 329-339. 53 Kessel, M. and Gruss, P. (1991). Homeotic transformations of murine prevertebrae and concommitant alteration of Hox codes induced by retinoic acid. Cell 67: 89-104. 54 Dolle, P., Izpisua-Bclmonte, J., Brown, J., Tickle, C. and Duboule, D. (1991). Hox-4 genes and the morphogcnesis of mammalian genitalia. G e n w Del,. 5, 1767-1776. 55 Simeone, A., Acampora, D., Arcioni, L., Audrews, P. W., Roncinelli, K. and Mavilio, F. (1990). Sequential activation of HOX2 homeohox genes by retinoic acid in human embryonal carcinoma cells. Nature 346, 763-766. 56 Simeone, A., Acampora, D.; Nigro, V., Faiella, A., D'Esposito, M., Stornaiuulo, A., Mavilio, F. and Boncinelli, E. (1991). Differential regulation by retinoic acid uf the homeobox genes uf the four HOX loci in human embryonal carcinoma cells. Mechanisms of Development 33, 215-227. 57 Papalopulu, N., Lovell-Badge, R. and Krumlauf, R. (1991). The expression of murine Hox-2 genes is dependent on the differentiation pathway and displays collinear sensitivity to retinoic acid in F9 cells and Xenopus embryos. Nucl. Acids Res. 19, 5497-5506. 58 Krumlauf, R., Papalopulu, N., Clarke, J. and Holder, N. (1991). Rctinoic acid and the Xerropus hindbrain. Seminars In Developmental Biology 2, 181188. 59 Chisaka, 0. and Capeechi, M. (1991). Regionally restricted developmental defects resulting from targcttcd disruption of the mousc homcobox gene horl.5. Nafurv 350, 473-479. 60 Lufkin, T., Dierich, A., LeMeur, M., Mark, M. and Chamhon, P. (1991). Disruption of the Hox-1.6 homeobox gene results in defects in a region corresponding to its rostra1 domain of expression. Cell 66, 1105-1119. 61 Murphy, P., Davidson, D. and Hill, R. (1989). Segment-specific expression of a homcobox-containing gene in the mouse hindbrain. Nature 341, 156-159. 62 Maden, M., Hunt, P. N., Eriksson, U., Kuroiwa, it., Krumlauf, R. and Summerbell, D. (1991). Retinoic acid-binding protein and homeobox expression in the rhombomeres of the chick embryo. Development 111, 35-44. 63 Sundin, 0. and Eichele, G. (1990). A homeo donlain protein reveals the metameric nature of the developing chick hindbrain. Gcner Dev. 4, 1267.1276. 64 Graham, A., Maden, M. and Krumlauf, R. (1991). 'Ine murine Hox-2 genes display dynamic dorsoventral patterns of expression during central nervous system devcloprnent. Developnzent 112. 255-264. 65 Gaunt, S. J., Krumlauf, R. and Duboule, D. (1989). Mouse homeo-genes within a subfamily, Hox-1.4, -2.6 and -5.1, display similar anteroposterior domains of expression i n thc cmbryo. hut show stage- and tissue-dcpendent differences in their regulation. Development. 107. 131-141. 66 Sham, M A € . , Hunt, P., Nonchev, S., Papalopulu, N., Graham, A., Boncinelli, E. and Krumlauf, K. (1992). Analysir of the niurine Hox-2.7 gene: conserved alternative transcript> with differential distributions in the nervous system and the potential for shared regulatory regions. EMBO J. 11, In press. 67 Puschel, A., Balling, R. and Gruss, P. (1990). Postion-specific activity of thc Hox 1.1 promoter in transgenic mice. Development 108. 435-4442, 68 Puschel, A.. Balling, R. and Gruss, P. (1991). Seprrrate elements cause lineage restriction and specifv boundaries of Hux-1 .I expression. Devclopmenl 112, 279-288. '69 Whiting, J., Marshall, H., Cook, M., Krumlauf, R., Rigby, P., Stott, D., and Allemann, R. (1991). Multiple spatially-specific enhancers are required to reconstruct the pattern of HOX-2.6gene expression. Genes Dev. 5 . 2048-2059.
Robb Krumlauf is at the MRC Lab of Eukaryotic Moleculai- Genetics, National Institute for Medical Research, The Ridgeway, Mill Hill, London NW7 l A A , UK.
P,
Summary One of the most remarkable recent findings in developmental biology has been the colinear and homologous relationships shared between the Drusuphila HOM-C and vertebrate Hox homeobox gene complexes. These relationships pose the question of the functional significance of colincarity and its molecular basis. While there was much initial resistance to the validity of this comparison, it now appears the HoxlHOM homology reflects a broad degree of evolutionary conservation which has reawakened interest in comparative embryology and evolution. The evolutionary conservation of protein motifs in many gene families (including those for growth factors, secreted and membrane bound signalling factors, adhesion molecules, cytoplasmic receptor kinases, nuclear receptors and transcription factors) has lead to speculation on the extent to which these homology relationships represent common developmental processes and underlying molecular mechanisms. Structural identities in a protein may indicate the hiochemical/molecular function that a protein plays in cellular and developmental processes, without reflecting a conserved role in a cascade of developmental events. However, the analysis of genes encoding transcription factors has provided evidence suggesting that there are gene complexes in arthropods and vertebrates which are true homologues and which may share common roles in the specification of regional identity along embryonic AP axis. These genes comprise the Hux/HOirM-C homeotic complexes. This review will detail some of the evidence for this proposed relationship and will speculate on the functional implications. Introduction: The HOM-C and Hox Gene Complexes The homeobox gene family is a large and increasingly well characterized group of genes which contain sequences that display some degree of similarity to a short 180bp motif, termed the homeobox, as it was originally identified in several Drosophila homeotic selector and segmentation genes(','). The homeobox encodes a domain of 60 amino acids related to the helixturn-helix regions present in several yeast and prokaryotic DNA binding proteins. This relationship has
four independent Hox c l ~ s t e r s ( l ~ This - ~ ~ )process . was not restricted to the Hox clusters, but involved large chromosomal regions flanking the complexes, as evidenced by the conservation of chromosomal order of not just the Hox clusters but also linked genes including members of the keratin, collagen, and nerve growth factor Homologues and Paralogues in the Hox/HOM-C Complexes One consequence of the generation of Hox clusters by duplication of an ancestral complex is that specific genes in each of the different complexes are evolutionarily related to each other, forming a subfamily or paralogous group. Comparisons of the degree to which this identity has been maintained among paralogues, therefore provides insight into the evolutionary constraints placed on the Hox network. Figure 1 illustrates the similarities that define the paralogous relationships for a specific subfamily. The homeodomain of this subfamily has approximately 80% identity with that of the Antp homeodomain. However, it is clear that these three proteins (Hox-2.6/1.4/4.2) are further related to each other based on the position and identity of changes in their sequence when compared directly with the Antp sequence (Fig. 1A). Note the common substitutions of P/S/”r/A/a/v at positions 1/4/6/7/11/13 of the homeodomain in all members. The sequence identity extends outside of the homeodomain in two additional regions of the proteins. The first region can be referred to as an ‘extended homeodomain’. Little identity exists on the carboxy terminal portion of the protein flanking the homeodomain. However on the amino terminal side of the homeodomain there is a region of identity which extends upstream of the adjacent splice site into the exon encoding the hexapeptide or Y-P-W-M motif. The second region of identity outside of the homeodomain resides in the amino terminal region of the proteins where there is a block of 20-30 conserved amino acids (Fig. 1B). There are 13 paralogous groups in the human and mouse Hox clusters that have been identified b similar sequence comparisons of related members(1”1s7.Figure 2 shows a summary of these comparisons in an alignment of the four Hox complexes. The numbers at the bottom represent the different paralogous groups and the vertical rows above each number represent related genes in a subfamily. It is clear that all of the paralogous groups are not represented in each cluster. This may reflect the fact that some members of a cluster were not duplicated during the events that led to the formation of multiple complexes. However, analysis of the relevant chromosomal regions has not been performed at the sequence level, and it is equally possible that these gaps represent genes which were originally duplicated and have become highly diverged or lost during vertebrate evolution.
Sequence comparisons of the mouse Hox genes have been extended to those in lower vertebrates and it is apparent that there are closely related homologues. For example, the mouse Hox-2.1 gene has an equivalent in zebrafish that displays nearly 90% amino acid identity over the sequence of their entire predicted proteins(10*22). This high degree of conservation between species does not apply to all of the Hox genes, however, and it is often difficult to clearly assign homology relationships. For example, Xenopus Xhox7A is the homologue of the mouse Hox-2.6 gene but alignment of their two proteins reveals several deletions and large blocks of divergent amino acidd2”).In fact the Xenopzis X h o x l A protein is only homologous to HOX-2.6in the same regions in which the mouse paralogues are related to each other(20323-2’). It is only on the basis of its chromosomal position in a cluster of other Xenopus genes, which have more extended homology to mouse Hox-2 members, that a true homology relationship between Hox-2.6/XhoxlAis confirmed, However, an interesting point is that even in these cases of relatively high divergence the identity is conserved in the amino terminal region of the proteins and in the extended homeodomain region. This suggests that these identities may represent constraints to maintain important functional domains of the proteins. It is perhaps not surprising that hornologues exist in other vertebrates, but the observation that sequence identities are shared between mouse and Drosophila Antp-class genes suggests that the Hox genes might be true vertebrate homologues of the Drosoplzila IJOM-C homeotic g e n e ~ ( l”).~ . Figure ~ ~ 1shows that the specific groupings of amino acids in the homeodomain and amino terminal region of the Hox proteins in paralogous group 10 are also conserved in the Drosophila deformed (DfD) protein(20.2z~25). Overall the Dfd protein is 40% identical to HOX-2.6 and the multiple regions of identity between the sequences are colinear in the respectivc proteins (Fig. 1C). This strongly supports the idea that the paralogous group containing Hox-2.6/1.4/3.5/4.2 is a vertebrate Djd family. No one member of the four mouse genes is more related to Dfd than another, so it is not possible to designate a specific Dfd homologue. It is possible to confidently define homologous relationships between mouse Hox genes and the Drosophila labial (lab), proboscipedia (pb De ormed (Dfd) and abdominal-B ( Ahd-B) genes 19-21. 3,25-29), Furthermore when the mouse paralogues are aligned with their most closely related Drosophila counterparts, it is apparent that the linear order of related genes along the chromosome is the same in both the Hox and HOMC c o m p l e ~ e s ( ’ ~ ~ These ~~,~~ clear , ~ ~correlations ). are indicated by the grey shading and bold brackets in Figure 2. It is more difficult to define distinct homology relationships (dashed brackets, Fig. 2) between specific mouse groups and the Drosophilu abdominal-A (abdA), Ultrabithorax (Ubx), Antennapedia (Antp),and Sex combs reduced (Scr) genes, because they all share a / J
2f
Fig. 1. Homologues and paralogues of the vertebrate Deformed subfamily. A. Comparison of the homeodomains of three mouse genes, the Xenopus XhoxlA and Drosophilu Deformed homeodomains with that of Antp. Dashes represent identical amino acids. B. Amino acid indentities in the amino terminal regions of the proteins in (A). C. Alignment of the mouse, Xenopus and human homologues of the HOX-2.6gene. D. Alignment of the mouse Hox-2.6 protein and Dfd showing colinear regions of identity.
high degree of identity within the homeodomain. This does not necessarily imply that there are no direct homologues, but rather that the present criteria for
distinguishing ancestral relationships between these genes is not sufficient. For example, Hux-2.1has 12 of 14 residues at its amino terminus identical to those in
Drosophila ANT-C (3)
Hox-2 (1 1) Mouse HOX2 (17)
Human
Hox-1 (6)
Mouse
HOX1 (7)
Human
Hox-3 (15) Mouse HOX3 (12)
Human
Hox-4 (2)
Mouse
HOX4(2)
Human
Paralogous Subgroups Fig. 2. Homology relationships between the mouse and human Hox complexes and the Drosophila HOM-C complex. The mouse nomenclature for the genes are listed above the boxes and the human nomenclature below the boxes. The 13 paralogous groups are marked at the bottom. The vertical rows represent evolutionarily related genes. The grey shading and bold brackets identify groups where clear homology relationships are defined. Thc dashed brackets define groups that are related with lowere degrees of confidence. The chromosomal locations of the respective human and mouse clusters are indicated in brackets at the right. (Based on refs 15, 19, 29).
Scr, but Hox-2.2 and Antp also share most of these same sequences. These structural relationships strongly support the idea that the mouse Hox and Drosophila HOM-C homeobox complexes are derived from a common evolutionary ancestor of both arthropods and vertebrates. This ancestral organism appears to have contained a cluster of at lcast five Antp-class homeobox genes related to lab, pb, Dfd, Scr/Antp/Ubx/abd-A, and Abd-B genes (Figure 3 ) . The striking similarities in HoxIHOM-C presumably reflect some degree of selective presure for retention of the complexes. perhaps related to maintenance of their expression properties(12.'9'2' >30) An interesting observation from the extended structural comparisons is that homologues across species (Drosophila to human) and paralogues within a species all share the same identities in the amino terminal and extended homeodomain regions of their proteins. The implication is that these domains are essential for the function of these specific groups. However, the possibility exists that genes within a paralogous group may be partially redundant and show
functional compensation. Do all members of a paralogous group have the same function and target genes, and if not where does the specificity reside? Most of the experimental evidence suggests that the extended homeodomain region determines the specificity of DNA binding and target specificity, a5 changes as small as one amino acid to the homeodomain can alter the ~ p e c if ic ity ( ~ '-However ~~). no clear role for the identities at the amino terminal end of the protein have been put forward. Protein domain swapping experiments suggest that non homcodomain regions are involved in transcriptional activation. and it is possiblc that the amino terminal domain functions to modify this activity. However, another possibility comes from experiments on the POU class of homeodomains. It has recently been shown that despite nearly identical protein sequence in the homeodomain between Oct-1 and Ort-2, the alternation of a few amino acids can change the ability of this domain to interact with thc HSV transactivator protein VP16("736). This factor forms part of a multiprotein-DNA complex with the POU homeodomain proteins and are needed for transcriptional activation of a targct gene. These
arthropods A M 4
AM-A
Ubx
Antp
Scr
Dfd
PB
Lab
0 0 0 25
2 4
23
2 2
2 1
26
27
2.8
2 9
vertebrates
5’
posterior late
3
-b anterior
early
Fig. 3. Theoretical evolution of the H0xIHOM-C complexes from a common ancestor. At the bottom is a diagram illustrating that colinear pattcrns of expression have also been conserved in vertebrates, and this suggests a common role for these clusters in the regional specification of axial diversity.
experiments illustrate the importance of co-factors or accessory proteins in homeodomain activity, and show that even highly related sequences can have dramatically different target specificity. Therefore the amino terminal identities in the paralogous groups could serve to modify interactions with similar types of cofactors, and caution should be used in assuming that the minor sequence differences in the homeodomain of paralogues are not functionally significant. Differences between Hox and HOM-C Complexes While attention is often focussed on the homologies, the differences between the HoxlHOM-C complexes are equally informative. There is not a direct one-toone correspondence between the Hox and HOM-C members. For example, in Drosophilu the four genes abd-A, Ubx, An@ and Scr could correspond to the four paralogous groups (6-9) in mouse which are in a similar position in the respective clusters. but as mentioned above there is little data to support definitive relationships. It is equally possible that these separate families of genes represent amplification and diver-
gence events, from a single Antp related gene in the commoii ancestor, which have occurred independently in arthropods and vertebrates. Independent evolutionary differences between mouse and Drosophilu can be seen by examining mouse paralogous groups 1-5 and 11 (Fig. 2). In the Dromphila BX-C complex, there are no Antp-class segment identity genes upstream of Abd-B, but in the mouse it has recently been shown that there are five It has paralogous groups equally related to therefore becn argucd that either Drosophilu has lost these members or that this expanded Abd-B family represents a recent amplification s ecific to vert e b r a t e ~ ( ~Similarly ~). the Hox-2. subfamily is often proposed to be related to the Drosophilu Zen gene, which is not a true homeotic segment identity gene(37).This is largely based on relative positions in the complexes; however, there is n o good structural identity between this paralogous group and Zen. This suggests that this a vertebrate specific subfamily or that the Droso hilu counterpart has been lost during evolution(2 1. There are many other differences in the structural organisation of the Drosophilu and mouse complexes. The mouse Hox clusters have approximately 10 genes all with the same transcriptional orientation tightly grouped in a region spanning 120 kb. The Antp gene alone is this size and the region of HOM-C span greater than 600 kb. In Drosophilu, Dfd has been inverted and is oriented in the opposite transcri tional direction from the other Antp-class genes(11325 . In Drosophilu, HOM-C is divided into two separate complexes and there are many other specialized genes (some containing homeoboxes, bicoid, ftz) located in the complexes. This raises the question as to which type of organisation more accurately reflects the ancestral cluster. Genetic and molecular studies in another insect, the red flour beetle (Tribolium custuneum), have provided some insight into the problem. In this system there is a cluster of Antp-cluss genes structurally and functionally homologous to the Drosophila ANT-C and BX-C complexes, which are grouped in a single chromosomal ~ l u s t e r ( ” ~Several ~ ~ ) . of the specialised genes such as ftz appear to be missing and it will be interesting to determine if there are other members in the A h d - B and Zen positions. Overall the organisation in the red flour beetle appear9 to be more similar to the mouse than of Drosophilu. Therefore the Drosophilu HOM-C complex may not be the ideal choice for comparison with the vertebrate Hox family, as they have become more highly diverged incorporating many genes of alternate and specialised function. Further studies are needed to investigate this complex in other representatives of common evolutionary origin. In this regard it is interesting that a clustered group of at least four Antpclass genes, including members of the lab, Dfd, ScrlAntp, and AntplAbd-B like families, have been found to be conserved in the nematode C. eleguns (reviewed in ref. 40). While the homologies are not as
7,,1.5p.l
R
P
clearly documented, it appears that some form of this homeobox gene cluster existed in nonsegmented ancestors and must have played an important function in developing metazoans.
Relationship of Evolution to Function On the premise that the conservation of HoxIH0M-C structural organisation reflects a functional requirement, what would be the nature of this common functional role in development? Segmentation is widely believed to have evolved independently in different species and the ancestral An@-class cluster appears to be present in non-segmented organisms. Insight into this dilemma has come from the analysis of the spatially-restricted patterns of Hox gene expression during development. The patterns of expression are very diverse but a hallmark of Hox expression, as analyzed by in situ hybridisation in mouse embryos, is that different genes have sharp boundaries of expression along the anteroposterior (A-P) axis (reviewed in refs 13,41). These boundaries can vary from tissue to tissue, hence the patterns do not simply mark a common axial position in all tissues of the embryo unless these differences can be accounted for by tissue migrations after establishing the axial level. However, there is a correlation between these A-P boundaries of expression and the chromosomal organisation of the Hox clusters (see Fig. 3 ) . Genes in the Abd-B paralogous group have boundaries of A-P expression that are posterior relative to other members of the complexes. Moving along the chromosome from these Abd-B paralogues, each successive member in an individual cluster has a rogressively more anterior boundary of expression( 9,21.41,422). Hence the gene order is the same as the A-P order of expression along the embryonic axis. This property has been termed colinearity, b analogy to the relationship first described by Lewis(43x between genetic function and organisation of the Bithorax complex in Drosophila. Lewis proposed that this colinearity was part of a combinatorial code used to generate regional diversity. The colinear and highly ordered domains of expression have been observed in mouse embryos in an increasinp variety of contexts. These include: neural tuhe(1',2 .28,42.44), neural crest and branchial arches(28'45546), mesoderm(29,41,42,51-53) and gonads(s4). Furthermore this colinearity is not just limited to spatially-restricted domains of expression, but also is observed with respect to temporal pattern^(",^^.'^) and to the response of Hox genes to retinoic acid in cultured ~ e l l s ( ~ ~and -j~) Xenopus embryo^(^^'^^). Together the large number of tissues with colinear domains suggest that the Hox patterns could be part of a molecular mechanism for the specification of positional values along the embryonic axis. The different combinations of Hox genes would provide a molecular framework or code(28'33) to generate regional
P
diversity in a manner analogous to that proposed by ~). in mouse Hox Lewis for D r o ~ o p h i l u ( ~Mutations genes have directly demonstrated that these genes have normal roles in patterning axial structures which correlate with the spatially-restricted patterns of e x p r e ~ s i o n ( ~ ~ .~ "Th ~ e~ *common ) functional link between Hox and HOM-C therefore may lay in the mechanisms of axial patterning, not ~egmentation(~*). Recently it has been shown that mouse Hox genes do have patterns of expression which are restricted to segments in the hindbrain(28,44~4s,61) and that these patterns can be conserved in other vertebrates(62.6'). These findings suggest that the Hox genes might have a role in segment identity in the hindbrain, like their Drosophila counterparts. However, this is not a general property of the Hox genes as expression is observed in many nonsegmented structures even with in the neural tube. Therefore, association of the Hox/HOM-C with processes in segmentation may represent a co-opting of this system from its fundamental role in axial patterning.
Relationship of Evolution to Hox Regulation Despite the structural similarity between paralogous genes in the mouse, the recent phenotypes arising in Hox r n ~ t a n t s ( ~ ' ,demonstrate ~) that at least some functions of an individual member must be unique. If paralogous members are expressed in very different domains then there is little possibility for redundancy. Many studies have directly compared the patterns of paralogous members (for review see ref. 41), and the results vary considerably depending on the tissue. In the hindbrain and neural crest components of the branchial arches the Hox paralogues have identical anterior boundaries and segmentally-restricted domains of ex ression. with the exception of the lubiul subfamily(*' P ,45). There are, however, differences in the levels of expression within these domains between some members. This clearly implies that the regulatory regions imposing the segment-restricted domains are the same in each of the four clusters and have been conserved following the duplication of the ancestral complex in vertebrates which gave rise to the Hox clusters. However, it is important to point out that even though genes may have identical A-P expression boundaries they may vary with respect to other axes and that not all cells in the domain will be positive for each of the paralogous members. The Hox-2 genes have dynamic dorsoventral (D-V) restrictions in the neural tube that alter with time during development(@). Members of other Hox clusters also display D-V variations which are different from those of Hox2(41,s2,64). In the limbs and somites, a different pattern also emerges compared with the hindbrain. Paralogous members of a group may have ver different A-P expression boundaries in somites(41;4,53.65), and expression domains of members of the HOX-4and I3o.x-1
complexes map to different regions of the axis in the limb(47-50). These variations in distribution and timing argue that many of the regulatory controls for the separate Hox complexes have not been rigidly conserved. Regulation in a general sense could form an important basis for maintaining the linkage of Hox genes, because the spatially-restricted domains of expression are important for gene function. Is this concept supported by current data? There are examples of complex transcription patterns reflected in multiple transcripts from the same gene, multiple promoters, differential splicing and polyadenylation, and interspersed regulatory regions. One possible explanation for these properties is that each Hox gene may not simply function as an independent gene and might share regulatory regions with other members of the complex@6).Experiments in transgenic mice have shown that some isolated Hox/lczc 2 genes can function normally outside of their chromosomal environment(67-6y), but there is also evidence for enhancers that work over a distance and they may influence several Thcrefore there is a strong possibility that one of the constraints on the organisation of the Hox family is a need to preserve the relationships or positions of genes in the clusters in order to achieve the finely regulated patterns of expression.
Conclusions The dramatic similarities in structure, organisation and colinear expression between the vertebrate Hox and Drosophila HOM-C complexes clearly argue for a common evolutionary origin. A broad variety of genetic, molecular, and descriptive evidence suggests that the common functional basis for this conservation reflects the possibility that this set of genes is part of a molecular mechanism for generating regional diversity along the A-P axis. The offsets in some tissues and not others, and the different morphological properties of the cellular expression domains suggest that the H o x genes can not have a simple uniform patterning role in all embryonic contexts. This implies that different sets of target genes must be utilized in somites versus neural tube, and it is perhaps better to think of the functional role of these Hox genes as not involving a single defined output but in more general terms of providing regional differences. It is clear that the Hox family is going to provide an interesting paradigm for examining regulatory hierarchies in vertebrates and the extent to which gene families can help to define common processes in evolution. To properly build upon these findings more work is needed in comparative evolution to determine what the organisation of homeotic complexes are like in primitive organisms. Are the vertebrate clusters most closley associated with functional roles in the development of the nervous system, segmentation, or axial polarity? Despite the exciting observation that colinear-
ity exists there is still not a good explanation of its molecular basis. l t is hoped that investigations into the regulatory requirements using transgenic mice and biochemical approaches will address this problem and open the way to identifying other parts of the regulatory cascade in pattern formation.
References 1 Gehring, W. (1987). Homeoboxes in the study of development. Science 236,
124-1252, 2 Scott, M. P., Tamkun, J. W. and Hartzell, G. W., cd. (1989). The structure and function of the homeodomain. Eiochim. Eioyhys. Actu. 959, 25-48. 3 Otting, G., Qian, Y. Q.? Billeter, M.,Muller, M., Affulter, M., Gehring, W. J. and Wuthrich, K. (1990). Protein-DNA contacts in the structure of a homeodomain-DNA complex determincd by nuclear magnetic resonancc spectroscopy in solution. EhfBO J. 9 . 3085-3092. 4 @an, Y. Q., Billeter, M., Otting, G., .Wuller, M., Gehring, W. J. and Wuthrich, K. (1989). The structure of the Antennapcdia homeodomam determined by NMR spectroscopy in solution: comparison with prokaryotic repressors. Cell 59. 573-580. 5 Gehring, W., Muller, M., Affolter, M., Percival-Smith, A., Rilleter, M., Qian. Y., Otting, G. and Wuthrich. K. (1990). The structure of the homeodomain and its functional implications. TIG 6. 323-329. 6 Kissinger, C., Liu, B., Martin-Blanco, E., Kronherg, T. and Pabo, C. (1990). Crystal structure of an fngrailed homeodomain-DNA complex at 2.8A resolution: A fraoiework for understanding homeadomain-DNA interactions. Cell 63, 579-590. 7 Holland, P. and Hogan. B. (1986). Phylogenetic distribution of antennapedialike homeo boxes. Nature 321, 2.51-253. 8 Mavilio, F., Simeone, A., Giampaolo, A., Faiella, A., Zappavigna, V., Acampora, D., Poiana, G.. Russo, G . , Peschle, C. and Boncinelli, E. (1986). Differential and stage-related expression in embryonic tissues of a new human homco box gene. Noturc 324, 664-667. 9 Regnlski, M., Harding, K.. Kostriken, R., Karch, P., Levine, M. and McGinnins, W. (1Y85). Homeo box genes of the Anlennapedia and Bithorax complexea of Dvosophila. CeIl43, 71-80, 10 Krumlauf, R., Holland, P., McVey, J. and Hogan, B. (1987). Dcvclopmental and spatial patterns of cxprcssion of the mouse homeobox gene, Hox 2. Developmmt 99, 603-617. 11 Akam, kl. (1987). The molecular basis for nictameric pattern in the Drosophila emhryo. DeveIopment 101, 1-22. 12 Akam, kl. (1989). Hox and HOM: homologous gene clusters in insects and vertebrates. Cell 57. 347-349. 13 Kessel, M. and Gruss, P. (1990). Murine developmental control gcnea. Science 249. 374-379. 14 Roncinelli, E., Acampora, D., Pannese, M., D’Esposito, M., Somma, R., Gaudino, G., Sturnaiuolo, A., Cafiero, M., Fairlla, A. and Simeone, A. (1989). Organization of human class I homeobox genes. Genome 31, 745-756. 15 Boncinclli, E., Simeone, A.. Acampora, D. and Maviliu, F. (1991). HOX gene activation by retinoic acid. TIG 7. 329-334. I6 Akam. M., Dawson, 1. and Tear, G. (1988). Homeotic genes and the control of scgment diversity. Det’elopmenf 104. 1 17 Kappen. C., Schugart, K. and Ruddle, 9). Two stcps in the evolution of antennapedia-class vertebrate homeobox Genes. Proc. Nafl.Acad. Scf. U S A 56. 54.59-5463. 15 Schugart, K., Kappen, C. and Ruddle, F. (1989). Duplication of large gcnomic regions during the cvolution of vertebrate homeobox genes. Proc. Null. Acad. Sci. USA 86. 7067-7071, 19 Graham, A., Papalopulu, N. and Krumlauf, R. (1989). The murine and L)rosophila homeobox clusters have common featurcs of organisation and cxpression. Cell 57, 367-378. 20 Graham, A., Papalopulu, N., Lorimer, J., McVeg, J., Tiiddenham, E. and Krumlauf, R. (1988). Characterization of a muvine homeo box gene, Hox 2.6. related to the drorophila deformed gene. Genes Dev. 2, 14241438. 21 Duhoule, D. and Uolle, P. (1989). ‘The structurai and functinal organization of the murine HOX gene family rcscmbles that of Drosophila homeotic genes. EMBO J. 8. 1497-1505, 22 Njolstad, P. R., Molven, A. and Fjose, A. (1988). A zebrafish homologue of the murine Ifox-2.1 genc. FEES Lett. 230. 25-30. 23 Featherstone, M. S., Baron, A., Gaunt, S. J., Mattei, M.G. and Duboule, D. (1988). Hox-5.1 defines a humeobox-containing gene locus on moure chromosome 2. Proc. Nail. Acad. Sci. USA 85, 4760-4764. 24 Harvey, R. P., Tahin, C. J. and Melton, D. A. (1986). Embryonic cxprcssion and nuclear localization of Xenopus homeobox (Xhox) gene products. E M B O J. 5, 1237-1244. 25 Regulski, &McGinnis, I., N., Chadwick, R. and McCinnis, W. (1987).
Developmental and molecular analysis of Dcformed: A homeotic gene controlling Drosopltzla head development. EMBO J. 6, 767-777. 26 Ruhnck, M., Larin, Z., Cook, M., Papalopulu, N., Krumlauf, N. and Lehrach, H. (1990). A yeast artificial chromosornc containing the mouse homeobox cluster Hox-2. Proc. Natl. Acad. Sci. USA 87, 4751.4755, 27 Frohman, M., Boyle, M. and Martin, G. (1990). Iqolation oI the mouse Hox2.9 gene: analysis of embryonic expression suggests that positional information along the anterior-postcrior axis is specified by mesoderm. Development 110, 589-607. 28 Hunt, P., Culisano, M., Cook, M., Sham, M., Faiella, A., Wilkinson, D., Boncinelli, E. and Krumlauf, R. (1991). A distinct Hox code for the branchial rcgion of the head. 8Nature353, 861-864. 29 Izpisua-Belmonte, J., Falkenstein, H., Dolle, P., Renucci, A. and Dubnule, D. (1991). Murine genes related to the Drosopkila A b d B homeotic gene are sequentially expressed during development of the posterior part of the body. EiV.IRO. .I. 10, 2279-2289. 30 McGinnis, W. and Krumlauf, K. (19%). Homeobox genes and axial patterning. Cell 68, 283-302. 31 Treisman, J., Gonczy, P., Vashishtha, M.,Harris, E. and Desplan, C. (1989). A singlc amino acid can determine the DNA binding specificity of homeodomain proteins. Cell 59, 553-562. 32 Kuziurd, M. A. and McGinnis. W. (1989). A homeodomain substitution changes the regulatory specificity of the deformed protein in Drosophila embryos. Cell 59, 563-571. 33 Mann, R. S. and Hogness, D. S . (1990). Functional dissection of Ultrabithorax proteins in D. melanogaster. Cell 60, 597-610. 34 Gibson, G., Scbier, A., LeMotte, P. and Gehring, W. J. (1990). The specificities of Sex combs reduced and Antennapedia are defined by a distinct portion of each protein that includes the homeodomain. Cell 62. 1087-1103. 35 Kristie, T. and Sharp, P. (1990). Interactions of the Uct-1 POU subdomains with specific DNA sequences and with the IISV alpha-truns-activator protein. Genes Dev. 4, 2383-2396. 36 Stern, S., Tanaka, M. andHerr, W. (1989). The Uct-I homeodomain directs formation of a multiprotein-DNA complex with the HSV transactivator VP16. Nulure 341, 624-630. 37 Kushlow, C. and Levine, M. (1988). Combinatorial expression of aftz-zen fusion promoter suggests the occurrence of cis interactions between genes ofthe ANT-C. EMBO J. 7. 3479-3485. 38 Beeman, R. (1987). A homeotic gene cluster in the red flour beetle. Nature 327, 247-249. 39 Stuart, J . , Brown, S., Beeman, K. and Dcnell, R. (1991). A deficiency of the homeotic complex of the beetle Tribolium. Nature 350. 72-74. 40 Kenyon, C . and Wang, B. (1991). A cluster of Antennapeditirlarr homeobox genes in a nonsegmented animal. Science 253, 516-517. 41 Gaunt, S. J. (1991). Exprcssion pattcrns of mouse Hox genes: Clues to an understanding of developmental and evolutionary strategies. BioEssays. 13, 505-513. 42 Gaunt, S. J., Sharpe, P. T. and Duboule, D. (1988). Spatially restricted domains of homeo-gene transcripts in mouse embryos: relation to a segmented body plan. Development 104 (Supplement: Mechanisms of Segmentation), Eds: V. Frcnch, P. Ingham, J. Cooke and J. Smith), 169-179. 43 Lewis, E. (1978). A gcnc complex controlling segmentation in Drosophilu. ivutNre276, 565-5711. 44 Wilkinson, D., Bhatt, S., Cook, M., Boncinelli, E. and Krumlauf, R. (1989). Segmental expression of hox 2 homeobox-containing gcncs in thc developing mouse hindbrain. IVaiure 341. 405-409. 45 Hunt, P., Wilkinsun, D. and Krumlauf, H. (1991). Pattcrning thc vcrtebrate head: murine Hox 2 genes mark distinct subpopulations of premigratory and migrating ncural crest. Development 112. 43-51. 46 Hunt, P. and Krumlauf, R. (1991). Deciphering the Hox Code: Clues to patterning the branchial region of the head. Cell 66, 1075-1078. 47 Dolle. P., Izpisua-Belmontc, J. C., Falkenstein, H., Renucci, A. and Duhoule, D. (1989). Coordinate cxpresaion of thc murinc Hox-5 complex homoeobox-containing genes during limb pattern formation. Nature 342, 767-
772. 48 Izpisua-Belmonte, J.-C., Tickle, C., Dolle, P., Wolpert, L. and Duboule, D. (1991). Expression of homeobox Hux-4 genes and thc spccification of position in chick wing development. Narure 350: 585-589. 49 Nohno, T., Noji, S., Koyama, E., Ohyama, K., Myokai, F., Kuroiwa, A., Saito, T. and Tanapchi, S. (1991). Involvement of the Chox-4 chicken
homeobox genes in determination of anteroposterior axial polarity during limb development. Cell 64, 1197-1205. 50 Yokouchi, Y., Sasaki, H. and Kuroiwa, A. (1991). Homeobox gene expression correlated with the bifurcation process of limb cartilagc developmcnt. Nature 353, 443.445. 51 Gaunt, S. J. (1988). Mouse homeohox gene transcripts occupy different but ovcrlapping domains in emhryonic germ layers and organs: a comparison of Hox3.1 and HOX-1.5.Development 103, 135-144. 52 Gaunt, S. J . , Coletta, P. L., Pravtcheva, D. and Sharpc, P. T. (1990). Mowe Hox-3.4: homeobox sequence and embryonic expression patterns compared with other members of the Hox gene network. Devcltpnvnt 109, 329-339. 53 Kessel, M. and Gruss, P. (1991). Homeotic transformations of murine prevertebrae and concommitant alteration of Hox codes induced by retinoic acid. Cell 67: 89-104. 54 Dolle, P., Izpisua-Bclmonte, J., Brown, J., Tickle, C. and Duboule, D. (1991). Hox-4 genes and the morphogcnesis of mammalian genitalia. G e n w Del,. 5, 1767-1776. 55 Simeone, A., Acampora, D., Arcioni, L., Audrews, P. W., Roncinelli, K. and Mavilio, F. (1990). Sequential activation of HOX2 homeohox genes by retinoic acid in human embryonal carcinoma cells. Nature 346, 763-766. 56 Simeone, A., Acampora, D.; Nigro, V., Faiella, A., D'Esposito, M., Stornaiuulo, A., Mavilio, F. and Boncinelli, E. (1991). Differential regulation by retinoic acid uf the homeobox genes uf the four HOX loci in human embryonal carcinoma cells. Mechanisms of Development 33, 215-227. 57 Papalopulu, N., Lovell-Badge, R. and Krumlauf, R. (1991). The expression of murine Hox-2 genes is dependent on the differentiation pathway and displays collinear sensitivity to retinoic acid in F9 cells and Xenopus embryos. Nucl. Acids Res. 19, 5497-5506. 58 Krumlauf, R., Papalopulu, N., Clarke, J. and Holder, N. (1991). Rctinoic acid and the Xerropus hindbrain. Seminars In Developmental Biology 2, 181188. 59 Chisaka, 0. and Capeechi, M. (1991). Regionally restricted developmental defects resulting from targcttcd disruption of the mousc homcobox gene horl.5. Nafurv 350, 473-479. 60 Lufkin, T., Dierich, A., LeMeur, M., Mark, M. and Chamhon, P. (1991). Disruption of the Hox-1.6 homeobox gene results in defects in a region corresponding to its rostra1 domain of expression. Cell 66, 1105-1119. 61 Murphy, P., Davidson, D. and Hill, R. (1989). Segment-specific expression of a homcobox-containing gene in the mouse hindbrain. Nature 341, 156-159. 62 Maden, M., Hunt, P. N., Eriksson, U., Kuroiwa, it., Krumlauf, R. and Summerbell, D. (1991). Retinoic acid-binding protein and homeobox expression in the rhombomeres of the chick embryo. Development 111, 35-44. 63 Sundin, 0. and Eichele, G. (1990). A homeo donlain protein reveals the metameric nature of the developing chick hindbrain. Gcner Dev. 4, 1267.1276. 64 Graham, A., Maden, M. and Krumlauf, R. (1991). 'Ine murine Hox-2 genes display dynamic dorsoventral patterns of expression during central nervous system devcloprnent. Developnzent 112. 255-264. 65 Gaunt, S. J., Krumlauf, R. and Duboule, D. (1989). Mouse homeo-genes within a subfamily, Hox-1.4, -2.6 and -5.1, display similar anteroposterior domains of expression i n thc cmbryo. hut show stage- and tissue-dcpendent differences in their regulation. Development. 107. 131-141. 66 Sham, M A € . , Hunt, P., Nonchev, S., Papalopulu, N., Graham, A., Boncinelli, E. and Krumlauf, K. (1992). Analysir of the niurine Hox-2.7 gene: conserved alternative transcript> with differential distributions in the nervous system and the potential for shared regulatory regions. EMBO J. 11, In press. 67 Puschel, A., Balling, R. and Gruss, P. (1990). Postion-specific activity of thc Hox 1.1 promoter in transgenic mice. Development 108. 435-4442, 68 Puschel, A.. Balling, R. and Gruss, P. (1991). Seprrrate elements cause lineage restriction and specifv boundaries of Hux-1 .I expression. Devclopmenl 112, 279-288. '69 Whiting, J., Marshall, H., Cook, M., Krumlauf, R., Rigby, P., Stott, D., and Allemann, R. (1991). Multiple spatially-specific enhancers are required to reconstruct the pattern of HOX-2.6gene expression. Genes Dev. 5 . 2048-2059.
Robb Krumlauf is at the MRC Lab of Eukaryotic Moleculai- Genetics, National Institute for Medical Research, The Ridgeway, Mill Hill, London NW7 l A A , UK.
