Main

More than 200 y ago, scientists such as Buffon, Lamarck, Darwin, and Goethe reflected on how the developmental history of each organism related to the history of life on Earth(1–3). Their ideas defined one of the fundamental questions of biology in the 19th and 20th centuries, namely, investigation of the relationships between evolution, development, and morphology. This is a daunting task in organisms in which development can be studied directly, and until recently it was a nearly insurmountable assignment in humans. However, the rapid development of molecular-based tools and their application to the analysis of the human genome now make it possible to effectively explore the impact of evolution and development on human morphology. In this review, we demonstrate that the molecular analysis of human birth defects can be powerful approach for understanding human development and potentially for exploring human evolution.

THE LIMB AS A MODEL DEVELOPMENTAL SYSTEM

In 1975, King and Wilson(4) predicted that the major biologic differences between primates, specifically humans and chimpanzees, were probably owing to "regulatory mutations" in the genome. Thus, a key to understanding the morphologic differences among primates, including humans, was to understand the differences in the spatial and temporal patterns of gene expression during development. This was predicted on the observation that approximately 99% of the genetic sequence between chimpanzees and humans is conserved, and this echoed the conclusions of many earlier scholars who had lacked access to the emerging tools of molecular biology(5,6).

Our approach to reconstructing the genetic and developmental basis of hominid (i.e. erect, bipedal primates comprising recent humans together with extinct ancestral forms) morphology begins by characterizing defects in human developmental programs, that is birth defects. Although the etiology of most birth defects remains unknown, a substantial proportion of defects is caused by mutations in genes that are components of developmental programs(7). These represent so called errors of morphogenesis(8). Consequently, characterization of the molecular basis of birth defects can reveal the genetic constituents of these developmental programs. Characterization of these genes and their regulatory elements is a prerequisite to exploring how the altered genetic control of developmental programs may be responsible for morphologic differences among species.

Ideally, the genes comprising the developmental program to be investigated should be conserved, whereas the morphologic phenotypes produced by the same program are widely varied. Such tremendous morphologic variation is observed in the vertebrate fin-limb, from the expansive wings of the condor or the pectoral fins of the humpback whale to the extraordinary fingers of aye-aye, the stubby fins of a lungfish, and the wings of a bat. The functional diversity of these appendages is enormous, with fins and limbs being used for locomotion, feeding, communication, and defense. Despite the remarkable variation in size, shape, and function, the anatomic pattern of the tetrapod limb has been highly conserved, even used repeatedly. This suggests that the developmental programs of the limb have also been conserved while being repeatedly modified to generate dramatically different morphologies(9). Thus, the tetrapod limb is a paradigmatic structure for studying variation in the regulatory elements of genes in developmental programs and the effects on morphologic changes among species.

It has been suggested that extant primates exemplify a graded series representative of the evolution of the hominid hand(10). The most primitive primate hand can bring the tips of the digits together only by flexion of the palm at the metacarpal-phalangeal joints. This movement of the hand is called convergence, and it is characteristic of the hand of the tree shrew. The only way that a tree shrew can grasp an object is to place two convergent paws together to mimic a prehensile hand (similar to a squirrel grasping a nut with its forepaws). Tarsiers exhibit a more advanced degree of prehensility and can grasp a small object with one hand. Prehensile use of the hand is apparent in all monkeys. True opposability appears for the first time among extant primates in Old World monkeys and apes (including humans). The variation in hand morphology among primates has been related to differences in adaptive behaviors such as feeding and the mode of locomotion. Some authors have gone so far as to argue that the changing morphology of the hominid hand has driven the evolution of human sensory systems, cognition, and culture(11). Thus, understanding the genetic control of the evolution of hand morphology could provide compelling new information about the evolution of primates.

Modifications of the limbs are also one of the most important events in human evolutionary history. The use of tools has been correlated with the expansion in size of the hominid brain, and tool making is often regarded as the milestone that marked the emergence of the genus Homo. The fossil record of hominids suggests that the changes observed in the primate hand consisted largely of an adaptive shift from a power-grasping hand to a precision-grasping hand. This conclusion is based on morphometric analyses of primate hands, particularly the first metacarpal(12). The first metacarpal of extant primates limited to using a power grasp has a narrow head relative to its length. In contrast, the head of the first metacarpal of humans and other members of the genus Homo is broad relative to its length. This difference is the result of the anatomic and functional consequences of additional muscle attachments found in humans that are needed to perform a precision grasp. These morphologic changes identified in the hominid fossil record could be the consequence of changes in the regulation of genes controlling development of tetrapod limbs.

Another important rationale for analyzing birth defects of the limb is that they are second only to congenital heart malformations as the most common birth defects observed in infants(13). Indeed, the first human trait to be interpreted in terms of a Mendelian segregation pattern was a form of multiple digital reductions (i.e. brachydactyly)(14). Nevertheless, of the several hundred human malformation syndromes characterized by defects in limb patterning(15), only a handful have been mapped to specific chromosomal segments, and considerably fewer have been characterized at the molecular level(16) (Table 1). However, the broad availability of resources for characterizing genes causing human birth defects has accelerated the pace of discovery. These findings are rapidly being translated into a more lucid understanding of the genetic control of limb development. This success is, in large part, contigent on researchers taking advantage of the meticulous descriptions of morphologic abnormalities compiled by clinicians over the last 100 y. This underscores the importance, if not necessity, of appropriating adequate resources to train clinical geneticists and to comprehensively document the phenotypes of individuals with birth defects. Additionally, under the weight of the ever-increasing discoveries of genes that cause birth defects, the paradigm that birth defects are caused only by teratogens is beginning to shift.

Table 1 Genes causing primary patterning defects of the limb
Full size table

MOLECULAR BASIS OF LIMB DEVELOPMENT

The tetrapod limb is one of the best understood classical models of morphogenesis(17). More than 50 y of experiments in which limbs were surgically manipulated has identified three major signaling centers that interact with each other to control growth and patterning of the limb (Fig. 1). These centers are: 1) the ectodermal cells overlying the limb bud that control formation of the dorsal/ventral (D/V) axis; 2) a specialized region of ectoderm, the apical ectodermal ridge (AER), extending from anterior to posterior along the D/V boundary of the limb bud that in conjunction with immediately proximal mesodermal cells, the progress zone (PZ), controls proximal/distal (P/D) growth; and 3) a group of mesodermal cells in the posterior margin of the limb bud, the zone of polarizing activity (ZPA), that controls anterior/posterior (A/P) patterning of the limb bud. Over the last 10 y, gene misexpression and loss-of-function studies have led to the isolation and characterization of many of the molecular mediators of these centers(18). These data indicate that the signaling pathways and transcriptional control elements that coordinate limb development in model organisms such as Drosophila and chick seem to be conserved in mammals(19,20), including humans.

Figure 1
figure 1

Schematic illustration of a limb bud. The apical ectodermal ridge (AER; red) extends from anterior to posterior along the dorsal/ventral boundary of the growing limb bud. Immediately proximal to the AER is a region of rapidly proliferating mesodermal cells (dark blue) called the progress zone (PZ). Located in the posterior mesoderm is an important signaling center (yellow) called the zone of polarizing activity (ZPA). The AER, PZ, and ZPA are interconnected so that limb patterning and growth are partly dependent on their coordinated function. (Modified from Jorde LB, Carey JC, Bamshad M, White RW 1999 Medical Genetics, Mosby, London)

Full size image

Limb development begins with decisions that are made very early during embryogenesis. These decisions determine the number of limbs, the position of limbs with respect to the body axis, and the identity (i.e. upper versus lower) of limbs. The signal that initiates induction of the limb bud seems to arise in the intermediate mesoderm. A candidate for this signal, fibroblast growth factor 10 (FGF10), is expressed in the cells fated to become the chick wing or leg(21). Once initiated, proliferating cells from lateral plate mesoderm become bone, cartilage, and tendon; cells from the somitic mesoderm contribute to muscle, nerve, and blood/lymphatic vessels of the developing limb.

Before differentiation of the AER, two genes, Radical fringe (r-Fng) and Wnt-7a, are expressed in the dorsal ectoderm. In the ventral ectoderm and mesoderm, expression of r-Fng and Wnt-7a is blocked by the product of the gene Engrailed-1 (En-1), a homeodomain-containing transcription factor(22). The AER forms at the interface of r-Fng-expressing and non-r-Fng-expressing cells(23,24). Expression of Wnt-7a in the dorsal ectoderm induces the underlying mesoderm to express another transcription factor, Lmx-1, that promotes the mesoderm to adopt dorsal characteristics(25). Mesoderm in which Wnt-7a expression is blocked by En-1 becomes ventralized. Thus, the process of AER formation and D/V patterning are interconnected and coordinated by En-1. In the mouse, functional inactivation of Wnt-7a results in ventralization of the dorsal surface(26) (e.g. ventral pads on both side of the foot).

Mediation of P/D growth by the AER is controlled, in part, by fibroblast growth factors (e.g. FGF2, FGF4, FGF8) that stimulate proliferation of an underlying population of mesodermal cells in the PZ(27). Cells that exit the PZ first are fated to become more proximal elements, whereas cells that leave later become distal elements of the limb. Maintenance of the AER is dependent on a signal from the ZPA. The signaling molecule of the ZPA is Sonic hedgehog (Shh)(28). Transplantation of the ZPA to the anterior portion of the limb bud or ectopic expression of Shh in the anterior limb bud results in duplication of limb elements along the A/P axis. Thus, the ZPA also specifies positional information along the A/P axis of the limb bud. However, it is not clear whether Shh regulates A/P patterning directly or indirectly.

The formation of the limb is controlled by a complex set of molecules that interactively promote axis formation, stimulate growth, and pattern the individual skeletal elements. Genes encoding these molecules are activated by proteins that act like master switches to turn on developmental programs. The proteins encoded by Hox genes and T-box genes exhibit the characteristics of a master switch, and if perturbed, disrupt the normal architecture of the human limb(29). Furthermore, molecular characterization of birth defects caused by mutations in these genes is revealing insights about their function in tetrapod limb development.

BIRTH DEFECTS OF THE LIMB

Classification of limb defects. Defects of the limb can be categorized using a variety of classification schemes (e.g. anatomic location of defect, specific limb axis perturbed, type of molecule disrupted, isolated versus syndromic). Because our knowledge of the pathogenetic basis of limb defects is limited, all of these classifications are somewhat arbitrary. For example, anatomically similar defects can be caused by mutations in many different genes. This is illustrated by the mapping of genes on at least four different chromosomes that cause split hand/foot malformations. Classification of defects by anatomic location can be useful for developing strategies for surgical palliation or fitting prosthetic devices. Categorization of limb defects by the axis perturbed is useful for generating hypotheses about causative pathogenetic mechanisms, although many limb defects affect more than one axis, making it difficult to identify the primary disturbance. Moreover, no specific classification is appropriate for all applications. We have chosen to organize limb defects by the type of molecule disrupted (Table 1). This facilitates the integration of human limb defects into models of the control elements of developmental systems. In this review, we will largely limit our discussion to the role of selected transcriptional activators in limb development.

HOX genes. Hox genes encode transcription factors that contain a 60-amino acid DNA-binding domain called a homeodomain(30). These genes compose the homeotic gene complex (HOM-C) in Drosophila, the organism in which they were first identified. Four copies of HOM-C (i.e. Hox clusters A through D) are found in mammals. Each 100-kb gene cluster is located on a different chromosome, and within each cluster Hox genes are numbered from 1 to 13. However, the human and the mouse have only 39 Hox genes, and thus not each Hox cluster contains 13 genes. Hox genes occupying the same position in different clusters are called paralogues (e.g. Hoxa-13 and Hoxd-13 are paralogues).

Hox genes partly control patterning of many different embryonic structures including the axial skeleton, limbs, genital ridge, digestive tract, and CNS. In developing forelimbs and hindlimbs, Hox genes from clusters A and D are expressed in nested patterns along the P/D and A/P axes, respectively. This suggests that Hox genes may subdivide the limb into domains and activate specific downstream target genes contingent upon their position along the axes. This is supported by misexpression studies demonstrating that ectopic anterior expression of several of the Hox genes can produce ectopic posterior skeletal elements(31). Furthermore, loss-of-function mutations indicate that mice lacking a specific Hox gene are missing some of the skeletal elements in which that Hox gene would be normally expressed.

Although the loss of one Hox gene typically produces only subtle limb defects, elimination of two or three paralogues (e.g. Hoxa-13 and Hoxd-13) produces more pronounced defects(32). For example, the size and number of digits is directly dependent on the level of Hox proteins encoded by the Hoxa and Hoxd genes expressed most posteriorly in the limb. This suggest that Hox proteins pattern the limb in a quantitative, dose-dependent fashion rather than by unique combinatorial codes in which each Hox protein makes a qualitative contribution(33).

At one time, it was thought that mutations in HOX genes would be embryonically lethal or that because of functional redundancy there would be no detectable phenotype. However, mutations in HOX genes have been described in two syndromes characterized by limb defects, synpolydactyly (SPD) and the hand-foot-genital syndrome (HFGS). SPD is distinguished by ¾ syndactyly of the hands and 3 4 syndactyly of the feet, with duplication of a digit within the syndactylous web(15). In most individuals with synpolydactyly, an imperfect trinucleotide repeat sequence in HOXD13 that normally encodes a 15-residue polyalanine tract is expanded, producing a tract with 22 to 29 residues(34). The penetrance of SPD and the severity of the limb defects are positively correlated with the size of the expansion. This observation and the different phenotype of mice lacking Hoxd-13(35) led to the hypothesis that expansion of the alanine tract produces a gain of function in the mutant Hoxd-13 protein. This suggests that HOXD13 mutations may cause functional inactivation of the neighboring 3′ HOX genes. This is supported by the finding that simultaneous disruption of Hoxd-13, Hoxd-12, and Hoxd-11 results in mice with digital fusions and duplications(36), findings similar to those observed in SPD.

More recently, two kindreds have been reported in which individuals have deletion mutations in HOXD13 that produce frameshifts(37). In contrast to the presumed gain-of-function caused by expansion of the polyalanine tract in HOXD13, these frameshifts are more likely to result in loss-of-function of HOXD13. Moreover, the limb defects caused by frameshifts in HOXD13 are distinctive because many affected individuals exhibit duplications of the base of the 1st and 4th metatarsals. Similar defects of the lower limb are not found in individuals with classical SPD. This extends the range of abnormal limb defects caused by HOXD13 mutations to include the most anterior elements of the autopod (i.e. hands and feet). Furthermore, deletion of the entire HOXD cluster has been reported in two individuals, each of whom has a single bone in the zeugopod and monodactyly in the autopod of all four limbs(38).

At least three different mutations in HOXA13 have been reported in families with HFGS(39,40). The limb anomalies in HFGS are characterized by reduction of the 1st metacarpal and metatarsal and carpal/tarsal fusions. Similar digital reductions are observed in the mouse mutant, hypodactyly, which is caused by mutations in Hoxa-13(41).

T-box genes. T-box genes are members of a rapidly growing and highly conserved family of transcription factors that share a region of homology to the DNA-binding domain (T-box) of the mouse Brachyury (or T) gene product(42–44). Most T-box genes have been found through experiments designed to identify genes that have specific activities in embryonic development or that cause developmental defects(42). Phylogenetic analyses suggest that the origin of the T-box gene family predates the divergence of arthropods and chordates from a common ancestor more than 600 million years ago(45). Thus, T-box genes are likely to play critical roles in the development of all animal species.

During embryogenesis, the spatial and temporal expression pattern of each T-box gene is unique but overlaps with the expression domains of other T-box genes(46). This indicates that T-box genes are differentially regulated during development(46). At least five T-box genes are expressed in the developing vertebrate limb. These include Tbx2, Tbx3, Tbx4, Tbx5, and Tbx15(47,48). In the chick and mouse, Tbx5 and Tbx4 are exclusively expressed in the vertebrate forelimb and hindlimb, respectively. This suggests that T-box genes may specify limb identity in tetrapods(48–51).

Direct evidence for a role of TBX3 and TBX5 in controlling the A/P axis of the upper limb comes from the analysis of human limb defects. Defects of the anterior elements of the upper limb are the most common abnormalities observed in individuals with mutations in TBX5 that cause Holt-Oram syndrome (HOS)(52–54) (Fig. 2A). Mutations in TBX3 cause deficiencies or duplications of the posterior elements of the upper limb in ulnar-mammary syndrome (UMS)(55) (Fig. 2B). In the mouse forelimb, Tbx5 is expressed throughout the mesoderm, whereas Tbx3 is expressed along the anterior and posterior margins of the limb bud mesoderm(56). However, along the anterior margin of forelimb mesoderm, Tbx3 expression does not extend as far distally(48). TBX3 and TBX5 arose from an ancestral gene, Tbx2/3/4/5, common to bony fishes and tetrapods. Thus, the temporal and spatial differences in expression between Tbx3 and Tbx5 in the forelimb could reflect functional divergence to control formation of the A/P axis of the forelimb.

Figure 2
figure 2

Mutations in TBX3 cause defects of the posterior skeletal elements of the upper limb (A, B) in ulnar-mammary syndrome. Abnormalities can range from hypoplasia of the ulna and radius and absence of digits 3, 4, and 5 (A) to hypoplasia of the phalanges of digit 5 (B). Defects of the anterior skeletal elements of the upper limb in Holt-Oram syndrome are caused by mutations in TBX5 (C, D). These defects can range from a shortened humerus with absence of the radius, ulna, and digit 1 (C) hypoplasia of digit 1 (D).

Full size image

TBX3 is composed of at least seven exons with an open reading frame of 2172 bp that is predicted to encode a protein containing 723 amino acid residues(53) (Bamshad M, Le T, Watkins WS, Dixon ME, Kramer BE, Roeder AD, Carey JC, Root S, Schinzel A, Van Maldergem L, Gardner M, Lin RC, Seidman CE, Seidman JG, Wallerstein R, Moran E, Sutphen R, Campbell CE, Jorde LB, unpublished observations) (Fig. 3). Mutations in TBX3 produce upper limb defects ranging from duplication of the 5th digit to complete absence of the hand and forearm. Additional features of UMS include defects of the teeth, genitals, and apocrine glands including breasts(58). Novel mutations have been characterized in 10 families with UMS(53) (Bamshad M, Le T, Watkins WS, Dixon ME, Kramer BE, Roeder AD, Carey JC, Root S, Schinzel A, Van Maldergem L, Gardner M, Lin RC, Seidman CE, Seidman JG, Wallerstein R, Moran E, Sutphen R, Campbell CE, Jorde LB, unpublished observations) (Fig. 3). At least two mutations in TBX3 disrupt the T-box and produce a protein fragment that presumably has reduced, if any, DNA-binding activity. Two missense mutations in exon 2 produce nonconservative amino acid substitutions in the T-box. Both of these substitutions are near sites that seem to be critical for Xenopus T-box proteins to interact with each other(59). Five mutations in TBX3 are located 3′ of the region encoding the T-box. Thus, if these alleles are transcribed and translated, the mutant TXB3 proteins may bind DNA, but their function must be substantially impaired. Taken together, these data suggest that UMS results from functional haploinsufficiency of TBX3 protein. Additionally, the observation that the majority of mutations in TBX3 do not disrupt the T-box underscores the existence of an additional functional domain(s) in TBX3.

Figure 3
figure 3

Genomic structure of exons encoding the open reading frames of TBX3 (above) and TBX5 (below). Each gene is composed of exons that contain untranslated sequence (open boxes), highly conserved T-box sequence (solid boxes), and protein-encoding sequence (striped boxes). The locations of TBX3 mutations found in 10 families with ulnar-mammary syndrome and TBX5 mutations in five families with Holt-Oram syndrome are indicated by arrows.

Full size image

Analysis of the clinical features of families with UMS shows no obvious phenotypic differences between those who have missense mutations versus those who have deletions or frameshifts (Bamshad M, Le T, Watkins WS, Dixon ME, Kramer BE, Roeder AD, Carey JC, Root S, Schinzel A, Van Maldergem L, Gardner M, Lin RC, Seidman CE, Seidman JG, Wallerstein R, Moran E, Sutphen R, Campbell CE, Jorde LB, unpublished observations). In other words, mutations producing TBX3 alleles, which could make TBX3 proteins that presumably would maintain DNA-binding activity, do not produce more severe developmental defects. This is consistent with the wide range of types and severity of birth defects among individuals with UMS from the same kindred. These data indicate that the clinical variability observed within and among families with UMS may be caused by epistatic modifiers of TBX3.

TBX5 is composed of at least eight exons with an open reading frame of 1581 bp that is predicted to encode a protein containing 512 amino acid residues(53,54). Mutations in TBX5 produce upper limb abnormalities ranging from triphalangeal thumb to complete absence of the limb. Nonsense, frameshift, and splice-site mutations that cause HOS are scattered throughout TBX5 (Fig. 3; Bamshad M, unpublished observations). The majority of these mutations result in premature termination and presumably make a truncated TBX5 protein. Some of these mutations are predicted to eliminate portions of the T-box, whereas others leave it intact. The phenotypes of individuals with varying types of mutations seem to be indistinguishable. These data indicate that functional domains outside of the T-box are critical for the normal function of the TBX5 protein and suggest that HOS results from a haploinsufficiency of normal TBX5 protein.

The multiple roles of GLI3. Mutations in the gene encoding GLI3, a zinc-finger transcription factor, cause limb defects in four different disorders(60). GLI3 encodes at least seven conserved domains including DNA-binding, zinc-finger, and microtubular anchor domains. GLI3 may be regulated so that it can have either activator or repressor functions. Greig cephalopolysyndactyly is characterized by craniosynostosis, anterior and posterior polydactyly, and syndactyly. It is caused by deletions of GLI3(61) and mutations in the carboxyl-terminal portion of GLI3(62) that abolish DNA-binding activity. Each of these mutations eliminates the activator and repressor functions of GLI3 and results in functional haploinsufficiency.

Mutations in the region between the zinc-finger and the microtubular anchor domains produce a protein in which the amino-terminal is cleaved so that it can migrate to the nucleus and repress transcription. Such mutations in GLI3 cause hypothalamic hamartomas, visceral anomalies, and posterior polydactyly in Pallister-Hall syndrome(63). Mutations 3′ of the microtubular anchor domain produce a protein retaining both repressor and activator functions and have been found in individuals with isolated posterior polydactyly(64). Thus, mutations in GLI3 alter the balance between its activator and repressor function and can cause either anterior or posterior polydactyly in three disorders with features that overlap but are clearly distinguishable from each other. Preliminary data suggest that mutations in GLI3 also cause anterior polydactyly type 4, characterized by duplications of the 1st digit of the hands and feet (Sutphen R and Bamshad M, unpublished observations).

Split-hand/split-foot malformations. Split-hand/split-foot malformations (SHFM) are a genetically heterogeneous group of limb-defect disorders characterized by hypoplasia or absence of the central elements of the autopod(15) (Fig. 4). However, the spectrum of patterning defects observed in SHFM ranges from cutaneous syndactyly of two digits to absence of the entire autopod. Genes causing SHFM have been localized to small critical intervals on chromosomes Xq26(65), 7q21(66), and 10q24(67). Despite these efforts, no gene causing SHFM has been cloned.

Figure 4
figure 4

SHFM syndromes are characterized by defects of the medial skeletal elements of the feet (A) and hands (B). These defects may be isolated or accompanied by abnormalities of the skin, hair, teeth, nails, palate, and mammary glands in syndromic SHFM disorders.

Full size image

SHFM can be observed as an isolated finding or in combination with other anomalies, such as tibial aplasia, craniofacial defects, and genitourinary anomalies. The prototypic example of an SHFM accompanied by multiple organ defects is ectrodactyly with ectodermal dysplasia and cleft-lip/cleft-palate syndrome (EEC). Ectrodactyly is an older term synonymous with SHFM and when literally translated means "aborted digit." EEC is characterized by anomalies of skin, hair, teeth, and nails, urogenital defects, mammary defects, nasolacrimal abnormalities, and cleft palate with or without cleft lip(68). A gene causing EEC has been mapped to chromosome 3q(69,70). The overlapping features of UMS and EEC suggest that TBX3 and the gene causing EEC may participate in the same developmental program(s).

SHFM has also been associated with cleft palate in the absence of cleft lip and ectodermal findings (ECP)(71). Thus, it has been suggested that ECP and EEC may be allelic, and that the presence of an orofacial cleft is dependent on the action of modifying genes. The cosegregation of these defects is important as some of the developmental programs controlling growth and patterning of the vertebrate limbs and face may be similar. Identification of the genes producing defects of the limbs and face in EEC and ECP will illuminate these similarities.

EVOLUTION AND THE GENETIC CONTROL OF LIMB DEVELOPMENT

One of the goals of biology over that last 100 y has been to understand the developmental and molecular mechanisms controlling, in part, evolutionary patterns. To this end, it has been posited that the dramatic differences in body plans that distinguish phyla or subphyla may have been caused by changes in gene number that permitted the establishment of completely novel developmental strategies(72). The most profound innovations in animal design occurred during the Cambrian period, and no features that distinguish higher taxa have since arisen. This suggests that dramatic innovations in developmental programs controlling animal body plans are no longer tolerated. One explanation of this phenomenon is that as additional genes were recruited into existing developmental programs, changes in key regulatory genes had increasingly disruptive effects on an organism. Thus, some developmental programs, such as those controlling body plans, may have become less flexible to upstream changes. However, the developmental program of the vertebrate limb has remained relatively flexible, as illustrated by the widely varying strategies used to build winged vertebrates (e.g. birds, bats, manta rays).

Less dramatic alterations in morphology may have been produced by modification of trans-acting regulatory factors, cis-acting control elements, or downstream targets of existing developmental programs(4). Regulatory elements have also been found in introns and the 3;-untranslated regions of genes. Naturally arising sequence differences in control elements of genes have been correlated with variation in gene expression patterns(73). For example, subtle variations in the expression patterns of Hox genes correspond to the variation of morphologic patterns between insect groups(74). Promoter variation and its association with regulatory function of the Ldh-B locus in killifish has been correlated to altered physiologic characteristics such as swimming performance(75). However, demonstrating the direct role of gene regulation in the adaptive evolution of animals has been challenging. In fact, given the differing anatomies and evolutionary histories of arthropod and vertebrate appendages, the origin of the similarities of their developmental programs is somewhat enigmatic.

Molecular population genetic analysis of the regulatory regions of genes in the developmental programs of closely related species eliminates some of this ambiguity. Comparison of homologous genes is advantageous because the modifiers of developmental programs conserved among closely related species can be identified, whereas regulators unique to a particular species may also become apparent. Additionally, the identification of modifiers between multiple taxa is more likely to represent alteration of a shared developmental program rather than evolutionary convergence.

Because vertebrate appendages are valuable models for understanding general mechanisms of pattern formation, the evolutionary control of vertebrate limb morphology has been widely investigated(76). The transition between osteolepiform fishes (having paired pectoral and pelvic fins with an internal skeleton) and land-dwelling vertebrates (e.g. Acanthostega, the most primitive known limbed tetrapod) resulted in tetrapods with 6-, 7-, and 8-digit limbs. Thereafter, fixation of the pentadactyl tetrapod limb occurred independently at least twice and probably preceded the evolution of wrist and ankle joints(77). However, the developmental programs of limb construction may not have been conserved in all tetrapods, as the spatial expression patterns of the genes controlling D/V patterning in Xenopus seem to be different compared with the chick and mouse(78).

Over the last 10 y, the most widely used strategies for studying variation in developmental programs include comparison of the spatial expression patterns of genes among species during embryogenesis, ubiquitous functional inactivation of a gene, or misexpression of a gene at an inappropriate place or time. However, instead of concentrating on the differences among developmental programs, these studies have focused on the elements of developmental programs that seem to be highly conserved among animals. Thus, despite the successful application of these molecular-developmental approaches, there have been only limited direct comparisons of the functional consequences of changes in spatial and temporal expression patterns of genes involved in limb development. Micro-array technologies may make it possible to further explore the dynamic relationships among different genes by simultaneously comparing the expression patterns of many different genes in the developing limb. Nevertheless, it is difficult to envision how this approach could be used to study the evolution of developmental patterns in many animals, especially primates. This underscores the need to further study the functional consequences of genetic variation of trans-acting regulatory proteins, cis-acting control elements, and downstream targets of genes causing morphologic changes during animal development.

EVOLUTION OF DEVELOPMENTAL SYSTEMS IN PRIMATES

The typical hand of a nonhuman primate has long and narrow digits that are slightly curved, slender terminal phalanges with a small apical tuft, metacarpal bones that are short compared with the digits, and a diminutive thumb. In contrast, humans have relatively straight and short digits, wide terminal phalanges, and a long thumb with a broad tip. In various primates, the hand has been adapted for specialized functions and thus deviates from the typical morphology. For example, digital reductions are found in primates (e.g. Colobus) with locomotory patterns in which the hands are used more as suspensory hooks than grasping pincers. The hand of the spider monkey (Ateles) has lost its thumb entirely. This morphologic phenotype mimics the changes observed in some individuals with mutations in TBX5. Similarly, the fourth digit of the aye-aye has become markedly elongated to facilitate feeding. Given that all of these species are members of the same order (and thus very closely related), it seems reasonable that some of these morphologic differences among primate hands are caused by altered regulation of shared developmental programs controlling the construction of the vertebrate upper limb.

Functional consequences of the differential regulation of developmental programs among primates have not been previously demonstrated. Nevertheless, conserved sequence motifs (so called phylogenetic footprints) among closely related species can reveal the identity of cis-acting regulatory elements(79). Furthermore, the functional significance of the variation in these elements can be directly tested using a variety of standard biochemical methods. For example, phylogenetic footprints have been identified in the noncoding regions within the globin gene cluster in primates. These sequence motifs were subsequently demonstrated to specifically bind trans-acting proteins. Comparisons of the level of polymorphism and fixed substitutions among species also foster inferences about the adaptive evolution of sequences controlling gene expression. However, functional studies of the effects of sequence variants in control elements on DNA transcription must take into account the effects of complex relationships between elements in developmental programs(57). In other words, the effects of specific sequence variants may be dependent on the state of the transcriptional control system. Despite this limitation, characterizing the regulatory elements of genes (e.g. HOXD13, TBX3, TBX5) known to play a role in patterning the skeletal elements of vertebrate upper limb among primate species with varying hand morphologies may reveal important insights about the evolution of Homo.

CONCLUSIONS

It is firmly established that mutations in genes encoding signaling proteins, receptor molecules, and transcription factors can alter the normal architectural pattern of the limb to cause human birth defects. The genetic analysis of birth defects allows us to dissect the roles of these molecules in developmental programs. Molecular evolution of the regulation of the genes coordinating developmental programs underlies, in part, morphologic differences within and between species. Systematic comparisons of the regulatory elements of genes in developmental programs and correlation of identified differences with functional changes and morphologic variation will begin to address questions about the molecular evolutionary history of these developmental programs. This may provide us with the insights necessary to further understand how to reconstruct the evolution of the primate limb. If this is successful, we will have a better understanding of the forces that shaped our past and will likely shape our future.