Main

Homeobox genes are master developmental control genes that act at the top of genetic hierarchies regulating aspects of morphogenesis and cell differentiation in animals. These genes contain a common sequence element of 180 bp, the homeobox, which was first discovered in the fruit fly Drosophila. Subsequently, the homeobox was shown to occur in all metazoa ranging from sponges to vertebrates and also in plants and fungi, and has thus been evolutionary conserved throughout the three kingdoms of multicellular organisms(1). The homeobox encodes a 60-amino acid homeodomain that is responsible for sequence-specific DNA binding of much larger homeodomain proteins. These proteins are transcriptional transregulators, i.e. they activate or repress the expression of target genes.

In vertebrates, the family of homeobox genes is large: about 170 different homeobox genes have already been cloned, which may represent more than half of the homeobox genes actually present in a given vertebrate genome. Thus, more than 0.2% of the estimated 100,000 genes per genome may possess a homeobox(2). The family of vertebrate homeobox genes can be conveniently divided into two subfamilies: 1) the clustered homeobox genes known as Hox genes or class I homeobox genes and 2) the nonclustered or divergent homeobox genes; the latter are scattered throughout the genome and fall into a number of groups based on sequence similarities as for example the vertebrate Pax, Msx, Emx, and Otx genes named after their homologs in the fly, i.e. the paired, muscle segment homeobox, empty spiracles, and orthodenticle genes, respectively (reviewed inRefs. 1 and 3).

This review will focus on the role of homeobox genes in normal mammalian development, as well as on their involvement in congenital malformations and tumorigenesis. As, due to space constraints, it is impossible to provide an extensive overview of the existing literature on homeobox genes, we decided to emphasize recent advances in the understanding of the involvement of Hox genes in 1) patterning embryonic structures such as the axial skeleton, the limbs, the genital and digestive tracts; 2) craniofacial morphogenesis and the development of the nervous system; and 3) the pathogenesis of human congenital malformations. The second part of the review will provide a brief survey of nonclustered homeobox genes and their roles in morphogenesis, focusing on head development, and describe the available data on human dysmorphologies caused by mutations in these genes. The last section will summarize the current knowledge on the oncogenic potential of deregulated homeobox genes.

HOX GENES: THE UNIVERSAL MOLECULAR ARCHITECTS OF BODY DESIGN

The Greek word “homeo” means alike, and the Drosophila homeotic (or HOM) genes are so named because of their ability, when mutated, to transform one of the insect's body segment into the likeness of another. For example, loss-of-function mutants of the Ultrabithorax (Ubx) gene lead to the transformation of the third thoracic segment carrying halteres (small balancers, see Fig. 1) toward a second thoracic segment with wings, thus generating four-winged flies (reviewed inRefs. 1 and 4). Pioneering work by E. B. Lewis(5) demonstrated that the normal function of homeotic genes is to assign distinct spatial (or positional) identities to cells in different regions along the fly's anteroposterior axis, and that it is the combination of homeotic genes (or homeotic code) expressed in a given cell which “tells” it that it belongs to the fly's head, thorax, or abdomen.

Figure 1
figure 1

Genomic organization and colinear expression patterns of Drosophila HOM genes and mammalian Hox genes. Schematic representation of the Drosophila homeotic complex (HOM-C), the four human Hox complexes and a hypothetical ancestral homeotic complex are displayed showing their possible phylogenic relationships[modified from Favier and Dollé(13)]. Each gene is represented by a colored box. The expression domains of HOM/Hox genes are schematized in a fly and in the CNS and prevertebrae of a human fetus (extrapolated from data in the mouse). For the sake of clarity, the partial overlap between HOM gene transcripts in thoracic and abdominal segments of the fly and overlapping expression domains of mammalian Hox genes along the body axis are not represented; hence, each color is meant to show the anteriormost expression domain of a given subfamily. HOM gene abbreviations are: lab, labial; pb, proboscipedia; Dfd, Deformed; Scr, Sex combs reduced; Antp, Antennapedia; Ubx, Ultrabithorax; abd-A, abdominal-A; Abd-B, Abdominal-B.

The mammalian Hox genes are defined by virtue of their homology with the genes of the homeotic complex (HOM-C) of Drosophila. Analysis of mouse and human Hox genes indicates that there are at least 39 of them organized in four clusters, HoxA, HoxB, HoxC, and HoxD, each localized to a different chromosome (e.g. human chromosomes 7, 17, 12, and 2, respectively), and comprising 9-11 genes (Fig. 1). On the basis of sequence similarities and relative position in the complex, the individual Hox genes within the different clusters can be aligned with each other and with genes of the Drosophila HOM-C cluster. These similarities suggest that the four mammalian clusters probably arose from a single ancestral complex, predating the divergence between arthropods and chordates 600 million years ago, first by the expansion of the cluster through lateral gene duplication (caused for example by unequal crossing over during meiosis) and second, during the transition from cephalochordates to vertebrates, through duplication of the clusters by chromosomal duplication or polyploidization [see references in Manak and Scott(3)].

In mammalian embryos, the earliest expression of Hox genes can be detected at gastrulation. Hox genes are expressed in all three germ layers with overlapping domains that extend from the caudal end of the embryo to a sharp anterior limit which is specific for each Hox gene. The Hox genes are arranged in the same order along the chromosomes as they are expressed along the anteroposterior axis of the embryo,i.e. the genes that are located 5′ in the clusters are expressed most posteriorly, whereas the more 3′ located genes are progressively expressed in more anterior regions. Significantly, this correlation, termed “spatial colinearity,” also holds for Drosophila homeotic genes (reviewed inRef. 6)(Fig. 1).

Hox genes in axial skeleton patterning. In the absence of any obvious naturally occurring mutation in vertebrate Hox genes, it was the development of transgenic mice and gene targeting technologies which permitted the generation of both gain-of-function and loss-of-function germ line mutations in the mouse(7, 8).

In Drosophila, ectopic expression of a homeotic gene often results in posterior homeotic transformations. For example, gain-of-function mutants of Ubx (Fig. 1) lead to the transformation of the second into the third thoracic segment with the transformation of the wings into a second pair of halteres. Similarly, ectopic expression of the Antennapedia (Antp) gene resulting from a spontaneous chromosomal inversion, placing the protein-coding region of this gene under the control of a heterologous promoter, leads to the conversion of the antenna into mesothoracic (second thoracic) legs (reviewed inRef. 1). The first evidence of a functional evolutionary conservation of the mammalian Hox genes was provided by two gain-of-function experiments performed by M. Kessel et al.(9) at the Max Planck Institute in Göttingen and by T. Lufkin et al.(10) in our group. We constructed a transgene in which the coding sequence of Hoxd-4 had been fused to the promoter region of the Hoxa-1 gene. This construct was then integrated into the mouse genome. In wild-type mouse embryos,Hoxd-4 is expressed up to the level of the fifth somite, which participates in the formation of the atlas (C1 in Fig. 2D). Placing Hoxd-4 under the control of Hoxa-1 cis-regulatory elements drives its expression into an ectopic anterior domain that spans the four occipital somites, which normally merge to form the basi- and exoccipital bones (B and E in Fig. 2D). In“Hoxa-1 promoter-Hoxd-4”(Hoxd-4+) transgenic newborns, the exoccipital bones are replaced by one to four ectopic ossified structures (A1-A4) resembling the neural arches of vertebrae, and the basioccipital, normally a flat bone, acquires a cylindrical shape reminiscent of a vertebral body (B* in Fig. 2E)(10). Thus,Hoxd-4 ectopic expression leads to posterior homeotic transformations, similarly to homeotic gene gain-of-function mutations in Drosophila. This experiment also provided evidence that at least part of the HOM/Hox gene network has been coopted to impart morphologic identities to segmented structures in animal groups employing radically different developmental strategies, and that some aspect of gene function has been conserved since the divergence of arthropods and chordates. This phylogenetic conservation of function is strengthened by the ability to rescue loss-of-function mutations in the Drosophila homeotic gene by means of Hox genes [see Lutz et al.(11) and references therein]. It is also noteworthy that the malformations observed in mice ectopically expressing Hoxd-4 correspond to normal features of agnathans, the most primitive vertebrates, which had occipital vertebrae instead of occipital bones. That change in the expression of a single Hox gene is sufficient to induce an atavistic change may help in understanding the genetic events that have been instrumental in the evolution of the shape of the present-day vertebrates from more primitive ones (reviewed inRef. 12).

Figure 2
figure 2

Examples of homeotic transformations resulting from loss-of-function and gain-of-function mutations of Hox genes.(A-C) Comparison between a wild-type (A) and Hoxa-2 null fetuses (B and C) of the skeletal structures present in the middle ear region at birth. (A) and(B) correspond to lateral views and (C) to A medial view. Note that the pterygoquadrate element (Q) represents an atavistic reptilian feature. G and G* wild-type and mutant gonial bones, respectively (this bone later becomes a part of the adult malleus); I and I2, orthotopic and ectopic incudes, respectively; M and M2, orthotopic and ectopic mallei, respectively; MC and MC2, orthotopic and ectopic Meckel's cartilages, respectively; Q, pterygoquadrate element; S, stapes; T and T2, orthotopic and ectopic tympanic bones, respectively (modified from Rijli et al.(40).(D-F) Lateral views of the occipitocervical junction in newborn mice. (D) Wild-type mouse: the vertebra and occipital bone anterior expression domains of Hoxd-3 and of Hoxd-4 are indicated in yellow and green, respectively. (E) “Hoxa-1 promoter-Hoxd-4 (Hoxd-4+)transgenic mouse expressing Hoxd-4 in the four occipital somites showing the posterior homeotic transformation of the exoccipital (E) and basioccipital(B) bones into ectopic neural arches (A1-A4) and into an altered cylindrical skeletal element (B*), respectively.(F) Hoxd-3 knock out mouse showing the occipitalization of the atlas (C1). A1-A4, ectopic neural arches;AC1, anterior arch of the atlas; B and B*, normal and homeotically transformed basioccipital bones, respectively; C, calvaria; C1-C4, cervical vertebrae 1 to 4; E, exoccipital bone; O, otic capsule.

To date, numerous loss-of-function mutations in mammalian Hox genes have also been reported (reviewed inRef. 13). The strategy consists of targeted gene disruption by homologous recombination. In brief, a wild-type allele is substituted by a nonfunctional allele in the genome of an ES cell, which is then injected into a mouse blastocyst to generate a chimera, from which mutant animals are derived through germline transmission of the mutation. During the last few years, more than half of the Hox genes have been functionally inactivated. In the majority of these studies, the null mutant phenotype comprised at least one vertebral transformation as for example: “occipitalization” of the atlas(e.g. Hoxd-3 inactivation; Fig. 2F)(14); transformation of the axis to an atlas identity evidenced by the formation of a ventral arch at the level of the second cervical vertebra (e.g. Hoxb-4 inactivation)(15); conversion of the first lumbar vertebra to a thoracic vertebra, thus producing a supernumerary 14th pair of ribs(e.g. Hoxc-8 inactivation)(16); transformation of the first sacral into a lumbar vertebra (e.g., Hoxd-11 inactivation)(17, 18). It is noteworthy that the axial transformations almost always correspond to anteriorizations such that the transformed vertebra(e) acquires the morphology of its anterior wild-type neighbor. Similarly, loss-of-function mutations of homeotic genes in the fruit fly lead to homeotic transformations of body segments to an anterior fate (see above). Moreover, null mutant mice for a given Hox gene usually show homeotic transformations in the anteriormost region where that Hox gene is normally expressed, and not within regions where a more 5′ gene is expressed.

Altogether, the results of gain- and loss-of-function mutations support the notion of a hierarchy of homeotic gene function in both mice and flies: in general more posterior-acting Hox and homeotic proteins are dominant with respect to function over more anterior-acting proteins. This phenomenon, which has been termed “posterior dominance” (or “posterior prevalence”) in mice and “phenotypic suppression” in flies, suggests that the morphogenetic program at a given level of the main body axis is dictated by the most posterior (i.e. most 5′-located in the cluster) Hox or homeotic gene expressed at that axial level(reviewed inRef. 6).

The recent analysis of double and triple Hox mutants has led to a partial revision of the “posterior prevalence” model of Hox gene function. A quantitative aspect has been integrated in the model(6), which takes into account the possibility that tandem duplication and tetraploidization of Hox genes in mammals might have resulted in partially redundant functions(1822): that inactivation of a single Hox gene results in vertebral transformations only in its anteriormost expression domain may reflect the fact that it is quantitatively preponderant in this region; however, the simultaneous disruption of two (or more) other Hox proteins expressed at the same axial level, but present individually in lesser amounts, may lead to the same phenotype. Therefore, vertebral specification is more likely achieved by the functional cooperation of a “combination” of Hox genes expressed at a given axial level (the combinatorial Hox(23) rather than by the function of a single gene.

Hox genes in limb morphogenesis and patterning of reproductive and digestive tracts. The mammalian Hox gene family contains 15 genes related to the Drosophila gene, Abd-B(Fig. 1). In addition to their expression domains along the prevertebral column, most of these Abd-B-related genes are expressed with overlapping domains in the developing fore and hind limbs consistent with a role in specification of the digit pattern(24). Indeed, knock-out experiments of murine Abdominal B- related genes resulted in size reduction, changes in shape, and/or delayed ossification of skeletal elements of both forelimb(e.g. Hoxd-9, Hoxd-9, Hoxd-11)(17, 21, 25, 26) and hind limb(e.g. Hoxa-10)(18). These results indicate that these genes control initially the allocation and growth of prechondrogenic condensations and, subsequently, the ossification sequence of the cartilage models (reviewed inRef. 13). However, homeotic transformations were never observed. For instance, disruption of the Hoxd-13 gene, which is specifically expressed in the presumptive digit territory, results in truncation of most metacarpal and metatarsal bones, shortening or lack of certain phalanges, as well as formation of a supernumerary digit rudiment in the forelimb(27)(Fig. 3, A and B). Targeted disruption of Hoxa-13 resulted in an autopodal phenotype, distinct from that of its paralogue Hoxd-13, including the lack of the first digit in both fore and hind limbs and alteration of some carpal/tarsal elements, suggesting that these two genes are not functionally equivalent for autopodal development(28).

Figure 3
figure 3

Examples of limb defects resulting from Hox gene disruptions. Dorsal view of the forelimb autopod (carpals, metacarpals, and phalanges) in adult wild-type (WT) and Hox mutants(genotypes as indicated). The digits are numbered in roman numerals, digit I(thumb) being the most anterior and digit V the most posterior. (A) and (B) Hoxd-13 mutant abnormalities include a supernumerary carpal bone (the postminimus, pm) in the heterozygotes(d-13+/-) and a supernumerary postaxial digit, distal to the postminimus in homozygotes (d-13-/-, arrow); homozygote mutants also typically display an absence of the 2nd phalanx in digits II and V. (C) Hox-13+/-/Hoxd-13-/- compound mutant showing polydactyly and severe deformations, truncations and fusions of digits in the form of a “terminal phalangeal” arch(`TA'). Note that Hoxa-13 heterozygote mutants have a near-normal autopod skeleton, except for minor abnormalities of the phalanges of digit I. I-V, digits one to five; M, metacarpal bones;P1, P2, and P3, first, second, and third phalanges respectively; pm, postminimus, R, radius; `TA', terminal phalangeal arch; U, ulna. (Modified from Dolléet al.(27) and Fromental-Ramain et al.(28).

On the other hand, compound Hox mutant phenotypes have also revealed partial redundant functions in limb patterning, as for vertebral specification (see above), between paralogous(28, 29) and nonparalogous(18) genes. For instance,Hoxa-13/Hoxd-13 compound mutants displayed alterations of growth and patterning of the autopods much more severe than those observed in each single mutant (i.e. almost complete lack of chondrogenic patterning;Fig. 3, A-C) showing that these gene products can also partly compensate each other(28).

Two human inherited limb abnormalities have been recently described, caused by mutations in the HOXD13 and HOXA13 genes(30, 31). Human synpolydactly is a semidominant syndrome and results from in-frame insertion of short poly(alanine) stretches in the N-terminal region of HOXD13(30). Intriguingly, the human limb phenotype is more severe (both at the homozygous and heterozygous states) than the Hoxd-13 gene disruption phenotype in mice(27) (Fig. 3, A and B). On the other hand, synpolydactly limb abnormalities resemble those observed in Hoxa-13/Hoxd-13 compound mutants(28)(Fig. 3C). A plausible molecular explanation is that the mutated human protein is transcriptionally inactive, while keeping all of its DNA binding properties, thus exhibiting dominant negative properties. Through binding site occupancy, it may interfere with the functions of other homeoproteins normally involved in digit morphogenesis, as for example HOXA13, leading to alterations more severe than those due to Hoxd-13 inactivation alone(28, 32). Human hand-foot-genital syndrome is an autosomal dominant syndrome due to a nonsense mutation in the HOXA13 homeodomain(31), which affects primarily first digit (both thumb and great toe) as well as Müllerian duct, ureteral and urethral development(causing didelphic or bicornuate uterus, ectopic ureteral openings and hypospadias). That heterozygotes for the Hoxa-13 null mutation in the mouse show defects that are apparently less severe than those seen in the hand-foot-genital syndrome supports the previous suggestion that the HOXA-13 mutation results in a protein exerting a dominant negative effect(31) (X. Warot, C. Fromental-Ramain, P. Chambon, and P. Dollé, unpublished observations).

Several studies have revealed roles for other Abd-B-related Hox genes in the developing genitourinary tract and terminal part of the digestive tract. Both male and female Hoxa-10 null mutants are hypofertile. Mutant male subjects exhibit cryptorchidism, caused by abnormal formation of inguinal canal and by a failure of shortening of the gubernaculum(33, 34). In addition, they display a malformation of the vas deferens that resembles a partial homeotic transformation to an epididymis(35). Partial homeosis of the vas deferens to an epididymis-like morphology is also observed in Hoxa-11 null mutants(36). Interestingly, Hoxa-10 null mutant females display an anterior transformation at a corresponding level of the reproductive tract, because the proximal part of the uterus is transformed into an oviduct-like structure(35). These data clearly indicate an important function of some Hox genes in defining regional fates along the anteroposterior axis of the Wolffian and Müllerian ducts.

The most “posterior” murine genes, Hoxd-12 and Hoxd-13, were also found to have a specific function in the morphogenesis of the terminal part of the digestive tract. Indeed,Hoxd-12 and Hoxd-13 null mutants display a disorganization of the smooth muscles layers forming the internal anal sphincter, resulting in rectal prolapsus in some mutants(37).

Hox genes in hindbrain patterning and craniofacial development. The hindbrain or rhombencephalon, is transiently divided along its anteroposterior axis into a series of segments (seven in mice and man) called rhombomeres(38). This segmental organization of the hindbrain determines the segmental migration of the NCC from the neurectoderm to populate and pattern the pharyngeal arches. In situ hybridization analyses have revealed that, in general, part of the combination of Hox gene expressed in a given rhombomere is also expressed in the NCC migrating from that rhombomere, suggesting that Hox genes may be instrumental in the patterning of the branchial region of the head (reviewed inRef. 39).

To validate this hypothesis, we and others functionally inactivated the Hoxa-2 gene(40, 41). NCC emigrating from the first two rhombomeres and caudal mesencephalon normally populate the first(or maxillomandibular) arch where they give rise to the dentary, maxilla, squamosal, tympanic, malleus, and incus bones and to the Meckel's cartilage. NCC emanating from the fourth rhombomere normally populate the second (or hyoid) arch and form the stapes, styloid bone, and lesser horn of the hyoid bone. Hoxa-2 is the “anteriormost” Hox gene, as it is the only member of this family to be expressed in the second rhombomere. However, at this level of the anteroposteior axis its expression is restricted to the neurectoderm; thus, the NCC of the first pharyngeal arch do not express any Hox gene. In contrast, Hoxa-2 is expressed in the NCC of the second pharyngeal arch. In the Hoxa-2 null fetuses the NCC-derived skeleton of the second pharyngeal arch is selectively lacking (e.g. the stapes, Fig. 2, A-C). In the place of second arch skeletal elements, an ectopic caudal set of first arch skeletal elements is present, mostly as a mirror image of its orthotopic counterpart. This ectopic set comprises: 1) within the middle ear region, a supernumerary incus, malleus, truncated Meckel's cartilage, and tympanic bone (Fig. 2, A-C); 2) outside the middle ear region, a small supernumerary squamosal bone(40). These data from skeletal analysis, combined with gene expression data, indicate that disruption of the Hoxa-2 gene results in a homeotic transformation of second to first pharyngeal arch identity. Such a transformation reveals that the morphogenetic program of the NCC derived from the first two rhombomeres corresponds to a ground (or default) skeletogenic patterning program (GPP) which is common to mesenchymal NCC of at least the first and second pharyngeal arches, and does not require Hox gene expression. In wild-type mice, the GPP is respecified by Hoxa-2, which, like Drosophila homeotic genes, acts as a selector gene to yield the second arch-specific morphogenetic program. Interestingly, an atavistic skeletal structure corresponding to the reptilian upper jaw (or pterygoquadrate) cartilage is developed from the second arch of Hoxa-2 mutants (Fig. 2C); thus in the absence of this gene, the second arch NCC appears to execute a GPP which corresponds to that of the therapsid phase of mammalian evolution(12, 40).

Additional inactivation studies also supported important roles of Hox genes in patterning hindbrain and pharyngeal arches. For instance, targeted inactivation of Hoxa-3 leads to hypoparathyroidism and thymic and thyroid hypoplasia(42). Interestingly, these defects are also observed in the Di George syndrome (which, however, is not due to a Hox gene mutation) [See Daw et al.(43) and references therein]. Defects of rhombomeres 4 and 5, cranial nerve and inner ear abnormalities were found in Hoxa-1 null mutants(44, 45), and selective facial nerve motor nucleus deficiencies were observed in Hoxb-1 and Hoxb-2 knockout mice(4648). Features of these phenotypes, such as paralysis of facial muscles, resemble those of Bell's palsy and Moebius syndromes in humans.

DIVERGENT HOMEOBOX GENES IN NORMAL AND ABNORMAL HEAD DEVELOPMENT

As already pointed out, Hox genes are not expressed in the neurectoderm anterior to rhombomere 2 and thus cannot participate in the development of the anterior hindbrain, midbrain, or forebrain (i.e. diencephalon, striatum, cerebral hemispheres). Likewise, Hox gene expression is lacking in the non(overtly)-segmented paraxial mesoderm which gives rise to craniofacial striated muscles, and is also absent in the NCC of the first pharyngeal arch and frontonasal mass which form most of the skull bones as well as the mesenchymal component of the teeth.

In flies, the control of head specification is dependent on homeobox gene not present within a cluster including the orthodenticle (odt) and empty spiracles (ems) genes. Drosophila sequences were used to identify vertebrate homologues, and this approach led to the cloning of the Emx-1 and Emx-2 genes related to ems, and of the Otx-1 and Otx-2 genes related to odt(reviewed inRef. 49). In the mouse, the two Emx and the two Otx genes are expressed in discrete, overlapping regions of the developing forebrain and midbrain, which often coincide with anatomical landmarks, and loss-of-function studies have shown that they play an important role in the patterning of these structures(5052). Therefore, both the sequence conservation and the related expression patterns of the two gene families suggests that cephalization was established in a primitive ancestor of both flies and humans. De novo mutations in the human EMX2 gene have been reported in patients with schizencephaly, an extremely rare congenital disorder characterized by a full-thickness cleft within the cerebral cortex; eventually, large portions of the cerebral hemispheres may be lacking, resulting in an holohemispheric cleft filled with cerebrospinal fluid. The phenotype of these patients suggests a requirement of the EMX2 protein for the correct formation of the cerebral cortex(53). Interestingly, Otx-1 null mice show spontaneous epileptic behavior associated with subtle brain abnormalities, suggesting that mutations in the human OTX1 gene might be responsible for some cases of epilepsy associated with cortical dysgenesis(54).

The Engrailed homeobox genes En-1 and En-2 are both expressed in a domain spanning the first rhombomere and the midbrain. Targeted gene disruption experiments have shown that En-1 has a critical role in the specification of its entire region of expression, whereas En-2 function is restricted to cerebellar foliation. However, the En-1 null phenotype, agenesis of the tectum and cerebellum, is completely rescued by insertion of the En-2 cDNA into the En-1 locus, suggesting that the distinct phenotypes of the En-1 and En-2 mutations reflect differences in the temporal expressions of the corresponding proteins, rather than differences in their biochemical activity(55).

The Pax family consists of nine unlinked genes; for example, each human PAX gene is located on a different chromosome (reviewed inRef. 56). Pax3, 4, 6, and 7 encode, in addition to the characteristic paired domain (a 128-amino acid DNA binding domain), a full-length homeodomain. In mouse embryos, Pax genes are widely expressed in the CNS and, as with the Otx, Emx, and En genes, their domains of expression in the brain suggest a role in its regionalization. Pax6 is also expressed in the optic vesicle (a prosencephalic derivative) and in the presumptive lens. Heterozygotic mutations in PAX6 have been reported in families with eye defects such as: 1) aniridia, a panocular disorder in which the development of the iris, cornea lens, and retina are disturbed; 2) Peters' anomaly, a defect of the anterior chamber of the eye with corneal malformations and attachment of the lens to the central aspect of the cornea; and 3) isolated foveal hypoplasia [e.g. see Glaser et al.(57) and Azuma et al.(58) and references therein]. It is noteworthy that mutations in PAX6 involve the inactivation or complete deletion of the PAX6 gene; hence, their autosomal dominant nature is not due to the presence of a dominantnegative mutant protein that could interfere with the function of the normal protein or related proteins, as appears to be the case for HOXA13, HOXD13, and PIT1 mutations (see below). Instead, it must reflect haploinsufficiency, a condition in which the amount of protein produced from a single functional allele is not sufficient to control the expression of downstream genes. In this respect, it is noteworthy that a putative case of PAX6 homozygous mutation resulted in anophthalmia and severe brain defects(57).

The WS is a dominantly inherited syndrome associated with sensorineural hearing loss and pigmentary disturbances which is responsible for ≈2% of all cases of congenital deafness. Loss-of-function mutations of the PAX3 gene have been found in WS type I and WS type III (reviewed inRef. 59).

The patterning of the first pharyngeal arch is partially reflected by the number, size, and shape of the teeth. Expression of Msx-1 and 2 in the mouse fetus and the phenotype displayed by Msx-1 null mice (anodontia, i.e. complete tooth agenesis) supports a critical role for these genes in early stages of odontogenesis (reviewed inRef. 60). These and other homeobox genes such as Distal-less 1 and 2 (Dlxl, Dlx2) and Goosecoid (Gsc), are expressed in restricted overlapping fields in the developing mandible. Based on these findings an“odontogenic homeobox code” that specifies tooth position and type(incisor, canine, premolar, molar) has been proposed(61). Interestingly, it was recently demonstrated that selective tooth agenesis (i.e. lack of all permanent second premolars and third molars) in a family with autosomal dominant tooth agenesis was caused by a single point mutation in the human MSX1 gene. MSX1 is not, however, linked to the more common human hypodontia, agenesis of the lateral incisors and second premolars [for a review, see Thesleff and Nieminen(60) and references therein]. Mutation analysis in families with Rieger's syndrome (an autosomal-dominant disorder characterized by hypodontia, abnormalities of the anterior chamber of the eye, and a protuberant ombilicus) has led to the identification of a novel homeobox gene, RIEG, whose mutations are responsible for the abnormalities observed in the Rieger syndrome(62).

Mutations in the homeobox of the human MSX2 gene have been associated with a rare form of craniosynostosis, called the Boston type. It is, however, noteworthy that more frequent craniosynostotic syndromes(i.e. Crouzon, Jackson-Weiss, Pfeiffer, and Apert syndromes) are caused by point mutations in fibroblast growth factor receptor genes (reviewed inRef. 63).

The transcription factor Pit-1, a member of the POU family of homeoproteins, is synthesized only in the anterior pituitary gland and regulates expression of growth hormone, prolactin, and β-glycoprotein subunit of the TSH-β. Mutations in the gene encoding Pit-1 have been identified in patients with combined pituitary hormone deficiency, in which there is no production of growth hormone, prolactin, and TSH, resulting in mental retardation and growth deficiency [for a review, see Rhodes et al.(64) and references therein]. The mutant Pit-1 can still bind to its DNA-binding site in target genes, but unlike the normal transcription factor, it does not activate transcription and, moreover, it prevents the normal protein from binding to DNA. These findings account for the dominant nature of the disease. Mutations in the POU3F4 gene cause deafness with fixation of the stapes, which represents the most frequent X-linked form of hearing impairment(65). The homeobox gene mutations causing human birth defects are summarized in Table 1.

Table 1 Homeobox gene mutations causing human birth defects (see the text for references)

THE ONCOGENIC POTENTIAL OF DEREGULATED HOMEOBOX GENES

The demonstration of cell lineage-specific patterns of Hox gene activation in human and murine leukemic cell lines supports the hypothesis that Hox gene expression can regulate normal hematopoietic differentiation (reviewed inRef. 66). Notably, expression of all HoxA cluster members was reported predominantly within cells of myelomonocytic origin. Very recently, two groups have reported the involvement of HOXA9 in the t(7;11) (p15;p15) chromosomal translocation, a rare but recurrent chromosomal rearrangement associated with AML. The t(7;11) produces a chimeric NUP98/HOXA9 protein containing the amino-terminal half of the nuclear pore complex protein NUP98 fused to the HOXA9 protein, which may promote leukemogenesis through inhibition of HOXA9- mediated differentiation [see references in Lawrence et al.(66)]. Hoxa-9 and Hoxa-7 have also recently been shown to be activated by proviral integration in a mouse model of myeloid leukemia(67). The induction of leukemogenesis was strongly correlated with the simultaneous proviral insertion into the Pbx-1-related Meis gene, a finding that is particularly interesting given that Hox proteins appear to cooperatively bind DNA with Pbx proteins (see below).

Pbx-1 is a divergent homeobox gene that was identified as the chromosome 1 partner of the t(1;19) translocation in human preB-cell ALL. The t(1;19) results in the fusion of a portion of PBX1 including the homeodomain, with a truncated EA2 protein. Pbx-1, as well as the E2A-Pbx-1 fusion, cooperatively bind DNA in vitro with several other Hox proteins, thus further suggesting that the oncogenic effects of Pbx proteins involve the formation of Pbx-Hox heterodimers [see references in Nakamura et al.(67).

HOX11 is another example of divergent homeobox gene originally isolated from a human leukemia. Overexpression of this gene occurs in rare cases of human T cell ALL that have the translocations t(10;14) (q24;q11) or t(7;10) (q35;q24), in which the HOX11 gene is activated due to juxtaposition with promoter elements from the T cell receptor α andβ genes, respectively [see references in Nakamura et al.(67)].

The mixed-lineage leukemia gene (MLL/HRX/ALL1) is disrupted by translocations involving the chromosomal region 11q23 in a group of acute leukemias characterized by a mixed lymphoid-myeloid phenotype and a high prevalence in children under the age of 1 y. MLL, does not possess a homeodomain; it is mentioned here because of its homology to the Drosophila trithorax gene, a regulator of the expression of HOM-C genes during embryogenesis. Loss-of-function data support a functional conservation in mammals, thus suggesting that MLL may act in hematopoietic malignancies by altering Hox genes expression [see Yu et al.(68) and references therein].

As to solid tumors, only PAX3 and PAX7 have been so far involved in some forms of rhabdomyosarcoma characterized by t(2;13) (q35;q14) and t(1;13) (q36;q14) chromosomal translocations, respectively (reviewed inRef. 69). The homeobox gene mutations involved in tumorigenesis are summarized in Table 2.

Table 2 Homeobox gene mutations causing human cancer (see the text for references)

CONCLUSION

The study of mouse homeobox genes mutations have clearly established the fundamental role of their products in patterning and organogenesis. It is likely that, in humans, mutations in such important genes are responsible for some of the early cases of spontaneous abortion and that, in the future, an increasing number of congenital malformation syndromes will be correlated with Hox gene mutations. It is also noteworthy that environmental factors such as retinoids [e.g. see Kessel an Gruss(23)] might exert some of their teratologic effects through alterations of Hox gene expression.