Our digits are some of the most sophisticated skeletal structures hominid evolution has brought about. In a delicate interplay with muscles and tendons they enable us to handle fine instruments, write, paint or perform music. Like most complicated structures, their development and function are governed by complex molecular mechanisms and thus are highly sensitive to pathologic changes. Malformations of hands and/or feet occur frequently in the general population and may present as isolated traits and/or as a part of complex malformation syndromes (Table 1). Numerous conditions show digit malformations amongst other features such as the large group of chondrodysplasias or pleiotropic syndromes such as Holt-Oram syndrome (MIM#142900) or ulnar–mammary syndrome (MIM#181450), that show the combination of limb malformations with heart defects or hypoplastic breast, respectively (see Table 1 for explanation of medical terms). In other conditions limb malformation occurs as an isolated trait and no other organs are affected. Among these the brachydactyly syndrome family was shown to be extremely valuable for our understanding of normal digit development.
Table 1. Glossary of Medical Terms
Affects the distal (digits) and intermediate (radius/ulna) elements
Shortening of hands/feet due to short phalanges and/or metacarpals/metatarsals
Shortening of middle phalanx
Abnormal development of cartilaginous tissue, usually accompanied by disproportionately short stature
Loss of one allele
Formation of additional phalanges
Congenital shortening or smaller in size
A disorder of tissue development
Bones that make up our palm
Degenerative joint disease
A digit consists of three phalanges, the thumb has two
A single gene influences multiple traits
Refers to the formation of additional digits
Fusion of phalanges/joints
Combination of too many (poly) with fusion (syn) of fingers (dactyly)
Fusion of bones
The brachydactylies are characterized by digit shortening caused by hypoplasia and/or aplasia of individual phalanges or metacarpals/metatarsals. There are five groups of brachydactylies, BDA-E, with individual subtypes (Bell, 1951; Mundlos, 2009; Fig. 1). BDA1 is characterized by hypoplasia/aplasia of all middle phalanges, whereas only phalanges of digits 2 and 5 are affected in BDA2. In BDB, the terminal phalanges are small or absent and there may be fusion of the distal interphalangeal joint. BDC presents with a more complex and variable picture but mainly features short metacarpal 1, short phalanges of digits 2, 3, and 5, as well as formation of additional phalanges in digit 1. BDE is characterized by variable shortening of the metacarpals. A genetic cause for most of these syndromes has been elucidated; however, the functional consequences of many causative mutations are still unclear (for detailed review, see Mundlos, 2009).
Limb development starts with the emergence of the limb buds, consisting at that time point of undifferentiated mesodermal cells covered by ectoderm. Outgrowth of the limb bud is under the control of fibroblast growth factor (Fgf) signaling from the apical ectodermal ridge (AER), a thickened ectodermal structure that runs along the dorsoventral border of the limb bud (Niswander et al., 1993; Sun et al., 2002). Besides the AER, other signaling centers such as the zone of polarizing activity (ZPA) in the posterior mesenchyme and the dorsal ectoderm specify the three spatial axes of the limb bud (reviewed in Tickle, 2003, 2006). These patterning mechanisms in the end generate a morphogenetic field that conveys the information for patterning the skeletal elements as well as other structures in the limb (see e.g., Zeller et al., 2009).
Skeletogenesis starts with the emergence of precartilaginous condensations in the core of the limb bud, where mesenchymal cells undergo changes in their adhesive capacities and form distinct aggregations that prefigure the future skeletal elements. These condensations (“anlagen”) develop from proximal to distal as a continuous rod that only later becomes subdivided into the individual skeletal elements by the formation of joints (Thorogood and Hinchliffe, 1975; Shubin and Alberch, 1986). The individual elements will subsequently grow and undergo differentiation to form the final skeletal elements of the limb. The formation of the limb skeleton can thus be subdivided into four major steps: (i) patterning of the skeletal elements, (ii) formation of the individual condensations, (iii) elongation and segmentation of the condensations, (iv) growth and differentiation (Fig. 2). As patterning mechanisms will be discussed elsewhere in this issue, we will focus on the subsequent steps. These steps are characterized by distinct molecular or morphological events. It is, however, of vital importance at this point to note that they are intertwined and not separable as the processes partially take place in parallel.
In the past years, we and others have begun to unravel the pathomechanism of digit malformations using in vitro systems or animal models (see Fig. 3). Especially the use of knock-in strategies to generate exact copies of a human mutation has proven useful, although it has to be pointed out that many mutations that act in a dominant manner in humans, as the great majority of brachydactyly associated mutations do, are recessive in the mouse indicating that conclusions drawn from these models have to be carefully evaluated when transferred to human development. These studies, sparked by the aim of understanding the pathogenesis of a human syndrome, have added to our understanding of limb development and have confirmed the relevance of developmental mechanisms unraveled in chicken or mouse for human limb development.
LIMB PATTERNING AND ITS IMPACT ON THE FORMATION OF THE DIGIT CONDENSATIONS
After the establishment of the three spatial axes the limb bud grows out and acquires its typical shape. This outgrowth and shaping requires proliferation being tightly coordinated with controlled cell death as well as tissue differentiation. The hand plate of the vertebrate limb (the autopod) is particularly suited for the analysis of such intricate events, because the main signaling centers controlling morphogenesis are well defined and the digit condensations that will differentiate to a unique digit identity provide a well-defined readout. As a prerequisite for the formation of the digit condensation the mesoderm of the autopod undergoes proliferative expansion governed by signaling from ZPA-derived Sonic hedgehog (Shh), which in this context has a dual role coupling the expansion of the limb mesenchyme with the specification of anterior-posterior digit identities (Towers et al., 2008; Zhu et al., 2008). Proliferative expansion of the mesenchyme is of vital importance, because deletion of Mycn from the limb mesenchyme in the mouse thus decreasing mesenchymal proliferation leads to severely reduced size of the skeletal elements (Ota et al., 2007). Of interest, only digit 5 was excluded from this effect, which is in accordance with the model proposed by Towers et al. (2008) where inhibition of proliferation in chicken wing buds led to loss of specifically anterior digits. In humans, heterozygous mutations in MYCN result in Feingold syndrome (MIM#164280) characterized by absent/hypoplastic middle phalanges of the second and fifth digits (van Bokhoven et al., 2005). Why, however, specifically these digits are affected remains obscure.
Expansion and distal outgrowth of the limb bud requires the propagation of a distal progenitor pool, which is mainly under the control of Fgf signaling from the AER as well as ectodermal Wnt signaling. Signaling by both these factors inhibit differentiation of e.g. cartilage (Rudnicki and Brown, 1997; Moftah et al., 2002) and account for the maintenance of naïve mesenchymal progenitors, however in regulating cellular differentiation both factors have distinct functions (reviewed in Yang, 2009). A failure in AER signaling leading to a collapse in the progenitor cell pool causes severe limb malformations such as the group of ectrodactyly syndromes also called split-hand/foot malformations (reviewed e.g., in Duijf et al., 2003). The distal progenitors will subsequently, under the influence of different signaling cues, differentiate to the lineages found in the growing limb, namely, the dermis in the most superficial layers and cartilage in the center, with the soft connective tissues arising in between (Stark and Searls, 1973, 1974; Pearse et al., 2007; Lu et al., 2008).
CONDENSATION OF THE DIGIT ANLAGEN
Condensation of the cartilage anlagen goes along with the appearance of visible structures exhibiting different morphology from mesenchymal cells. This process requires the initial specification of precartilaginous cells (chondroprogenitors), which are not yet fully committed toward the cartilage lineage, and the following determination of a subset of these cells to form the definite cartilaginous condensation (Akiyama et al., 2005). The limb mesenchyme appears to have an intrinsic chondrogenic potential as a default fate, as can be seen in mesenchymal micromass cultures (see e.g., DeLise et al., 2000). The initial specification of mesenchymal cells to chondroprogenitors is, however, restricted to the core of the limb bud. This can be mainly attributed to the inhibitory effect of signals emanating from the ectoderm, because ectoderm removal leads to ectopic chondrogenesis (Hurle and Ganan, 1986). The dorsal and ventral limb ectoderm is a source for several members of the Wnt family of diffusible signaling molecules (Witte et al., 2009) and there is substantial evidence suggesting that a gradient of Wnt/β-catenin signaling within the limb mesoderm has a key role in restricting chondrogenesis to the limb core (Hill et al., 2005, 2006; ten Berge et al., 2008).
Other negative signals that repress chondrogenesis are present between the digit anlagen (interdigital space) to ensure the ordered formation of the digits. Hoxd13 has been shown to suppress chondrogenesis directly and indirectly through the regulation of Raldh2, encoding the enzyme responsible for the formation of biologically active retinoic acid from the precursor retinol. In the mouse mutant synpolydactyly homolog (spdh) a mutation in Hoxd13 results in the down regulation of Raldh2, and, consequently, in low tissue concentration of retinoic acid. This in turn causes expression of Sox9 in the interdigital mesenchyme, the formation of interdigital condensations and, as a consequence, polydactyly (Kuss et al., 2009).
The precartilaginous condensation is characterized by the expression of Sox9, a transcription factor essential for chondrogenesis, as mice with inactivated Sox9 alleles do not form any cartilage (Bi et al., 1999). This is followed by the differentiation into chondrocytes characterized by the expression of Sox5 and Sox6 (Smits et al., 2001) as well as specific matrix proteins as Collagen type 2 alpha 1 (Col2a1) and aggrecan. Sox9 directly regulates the expression of Col2a1 (Lefebvre et al., 1997; Ng et al., 1997). However, the expression of Sox9 alone is not sufficient for definitive determination of the chondrogenic cell fate, because cells expressing Sox9 in vivo can still adopt other cell fates (Akiyama et al., 2005). Heterozygous mutations in SOX9 cause campomelic dysplasia, a generalized lethal chondrodysplasia with sex reversal (Wagner et al., 1994). However, the specific importance of fine-tuning of SOX9 regulation and hence the pace of chondrogenesis during digit development was recently highlighted by the discovery of potential regulatory mutations affecting SOX9 expression as a cause for brachydactyly–anonychia in four families with Cooks syndrome (MIM#106997; Kurth et al., 2009).
Bone morphogenetic protein (BMP) signaling is a major positive regulator of chondrogenesis (Duprez et al., 1996; Francis-West et al., 1996; Merino et al., 1999; Pizette and Niswander, 2000). BMPs and the closely related growth and differentiation factors (Gdfs) belong to the transforming growth factor-beta (Tgfβ) superfamily and in mammals comprise 21 members. BMPs and Gdfs bind to heteromeric receptor complexes consisting of type I and type II serine-threonine kinase receptors. Binding of ligand results in phosphorylation and activation of type I receptor by the type II receptors, which leads to a conserved pathway involving the phosphorylation of downstream signaling components, the Smad proteins 1, 5, and 8. Phosphorylated Smads associate with the common co-factor Smad4, translocate to the nucleus and activate the transcription of target genes. BMP/Gdf signaling is fine-tuned at several levels, for example by secreted antagonists that interfere with ligand-receptor binding, such as Noggin (NOG; reviewed e.g., in Hoffmann and Gross, 2001; Hartung et al., 2006). Sox9 is a major target of BMP signaling in chondrogenesis (Healy et al., 1999; Zehentner et al., 1999; Pan et al., 2008, 2009). Sox9 in turn positively regulates expression of NOG, providing negative feedback (Zehentner et al., 2002).
Although BMPs play a role in the initial global patterning process, their main role appears to be the local control of cartilage condensation (comprehensively reviewed in Robert, 2007). There are several genes encoding members of the BMP superfamily expressed in the autopod mesenchyme, with Bmps 2, 4, and 7 being most prominent, but also expression of Bmp5 and Bmp6 was reported (Zuzarte-Luis et al., 2004; Bandyopadhyay et al., 2006). Genetic experiments indicate that Bmp2, 4, and 7 appear to act in partial redundancy in the mouse limb; however a critical threshold of BMP dosage appears to be required to allow chondrogenesis (Bandyopadhyay et al., 2006). Bmps are expressed mainly in the interdigital mesenchyme and the perichondrium whereas their receptors, in particular BmpR1b, are expressed in the condensing chondrocytes. Chondrogenesis in the limb is dependent on BMP signaling, since its inhibition in the chicken limb by implantation of Noggin beads, or retroviral overexpression of dominant-negative BMP receptors, equally interfere with chondrogenesis (Zou et al., 1997; Merino et al., 1998). In the mouse ablation of both type I receptors (BmpR1a and BmpR1b) causes a severe generalized chondrodysplasia in which most skeletal elements that form through endochondroal ossification are absent or rudimentary (Yoon et al., 2005; Ovchinnikov et al., 2006). Homozygous inactivation of BmpR1b alone leads to severe brachydactyly in the mouse with hypoplasia/aplasia of the proximal and medial phalanges (Baur et al., 2000; Yi et al., 2000). Gdf5, which signals through BmpR1b, is also required for digit formation as demonstrated by the brachypodism (bp) mouse (see Fig. 3). Genetic experiments show that chondrogenesis in the digits is controlled by both Gdf5-dependent and -independent processes and that, reciprocally, Gdf5 acts through BmpR1b and other receptors (Baur et al., 2000).
In the mouse Gdf5 shows strong expression in the early autopod in mesenchyme flanking the condensation and later in the developing joints (see Fig. 2). Overexpression of Gdf5 in chicken or mouse limbs enlarges the initial condensations and increases cartilage nodules in limb explant cultures (Francis-West et al., 1996; Francis-West et al., 1999a; Tsumaki et al., 1999) pointing toward a positive role in chondrogenesis. Due to its restricted expression pattern in developing joint, Gdf5 has early on been ascribed a role in joint development (Storm and Kingsley, 1996), based on the observation that mice with mutations in Gdf5 (brachypodism, bp; Storm et al., 1994; see also Fig. 3) show joint fusions. However, analysis of the digit phenotype in brachypodism mice revealed a complex phenotype involving reduced chondrogenesis, reduced growth and impaired maintenance of the cartilaginous anlagen (Storm and Kingsley, 1999; Takahara et al., 2004). Altogether this points toward a pivotal role for Gdf5 in the regulation of the size of the initial cartilaginous condensation by means of the commitment of mesenchymal cells to the chondrogenic lineage.
The findings in mice are complemented by a series of mutations identified in humans. Heterozygous mutations in GDF5 (also termed cartilage derived morphogenetic protein 1, CDMP1) result in dominant brachydactyly type C (BDC, MIM#113100; Polinkovsky et al., 1997; Savarirayan et al., 2003). BDC is characterized by complex malformations of the digits including brachymesophalangy, hyperphalangy and shortening of metacarpals (Fig. 1). Mutations in the prodomain of GDF5 cause a recessive form of BDC, presumably by means of a gene dosage effect (Schwabe et al., 2004). GDF5 is pivotal for chondrogenesis in humans, because homozygous loss of function mutations in GDF5 cause recessive generalized acromesomelic chondrodysplasia of the Grebe (MIM#200700), Hunter-Thomson (MIM#201250), and Du Pan (MIM#228900) type (Thomas et al., 1996, 1997; Faiyaz-Ul-Haque et al., 2002). Furthermore, recessive homozygous mutations in the high affinity receptor BMPR1B also cause a severe form of acromesomelic chondrodysplasia with brachydactyly (Demirhan et al., 2005).
Brachydactyly A2 (BDA2, MIM#112600) is a less severe form of brachydactyly characterized mainly by short middle phalanges in the second and fifth digits (Fig. 1). BDA2 is caused by heterozygous mutations in BMPR1B (Lehmann et al., 2003) that are thought to confer dominant-negative properties to the resulting BMPR1B protein. Of interest, BDA2 has also been shown to be associated with dominant mutations in GDF5 (Seemann et al., 2005; Kjaer et al., 2006; Lehmann et al., 2006; Ploger et al., 2008), the ligand for BMPR1B (Nishitoh et al., 1996). These mutations result in a partial loss of function by specifically interfering with the binding of GDF5 to BMPR1B but not to its other receptor BMPR1A (Seemann et al., 2005; Ploger et al., 2008). This partial and specific loss of function is likely to explain the difference in phenotype between BDC and BDA2. Furthermore, it indicates that the fine-tuning of BMP signaling by means of its receptors is of specific importance for proper digit development in humans. This is corroborated by a recent finding that duplications of potential regulatory elements of BMP2 were identified in a subset of patients with BDA2 (Dathe et al., 2009). An increase in BMP2 signaling would be expected to result in a higher activation of the BMPR1A receptor and thus in a similar dysbalance as decreased signaling by means of the BMPR1B receptor.
ELONGATION OF THE INITIAL CONDENSATION BY DISTALLY DIRECTED CHONDROGENESIS
The positional information conveyed by Shh has to be translated to the establishment of a certain digit identity, i.e., the number of phalanges a digit has as well as the individual shape of the phalanges. Shh itself not only acts as a morphogen, but the spatial expansion of former ZPA cells also contributes to digit identities (Ahn and Joyner, 2004; Harfe et al., 2004). Moreover, it is thought that the interdigital mesenchyme memorizes positional information through local BMP signaling (Dahn and Fallon, 2000; Drossopoulou et al., 2000). Studies performed in the chicken have shown that the number of phalanges in a digit is determined by the duration of AER/Fgf signaling (Sanz-Ezquerro and Tickle, 2003). Prolonged AER signaling led to the development of additional phalanges while precocious ablation of the AER caused a cease of further growth resulting in loss of phalanges. AER ablation results in apoptosis in the underlying mesoderm (Rowe et al., 1982). Chemical inhibition of Fgf signaling during the phase of digit elongation caused the premature termination of outgrowth by induction of the formation of a terminal phalanx and hence loss of the penultimate phalanx (Sanz-Ezquerro and Tickle, 2003). It was thus proposed that one function of Fgf signaling is to inhibit the formation of a distal phalanx to sustain digit elongation in coordination with its role in maintaining the distal progenitor pool (Casanova and Sanz-Ezquerro, 2007). This phenotype, presence of a terminal phalanx but absence of more proximal structures is seen in human brachydactylies type A1, A2 and in some cases of BDB1. However, analyses of mouse models for BDA1 and BDB1 (Raz et al., 2008; Gao et al., 2009) have found no indication of AER/Fgf involvement in the pathogenesis of these syndromes. It thus appeared likely that another mechanism must be involved in the elongation of the digits, namely by driving chondrogenesis in a distally orientated manner.
However, the mechanism controlling the recruitment of distal undifferentiated progenitors to the growing cartilaginous condensation remained elusive. Two landmark studies have shed light on the molecular mechanisms responsible for this lineage commitment. The labs of Juan Hurle and John Fallon both demonstrated that a population of cells immediately in front of the growing digit condensation, the digit crescent (DC, Montero et al., 2008) or phalanx-forming region (PFR, Suzuki et al., 2008) controlled the incorporation of mesenchymal precursors into the condensation. This population was positive for the BMP receptor BmpR1b and the chondrogenic lineage marker Sox9 and showed high activity of the canonical BMP signaling pathway as it was positive for phosphorylated Smads 1, 5, and 8 (pSmad1/5/8). Both labs showed that this signaling was dependent on cues from the posterior interdigital mesenchyme, corroborating a role for this tissue in the specification of digit identities as proposed by Dahn and Fallon (2000). Suzuki et al. (2008) elegantly demonstrated that there is a specific signature of BMP/Smad1/5/8 signaling in the DC/PFR for the different digits of the chicken foot, implicating the DC/PFR in the establishment of digit identity. However, the digit having the highest number of phalanges (digit 4) did not show the highest level of BMP/Smad1/5/8 signaling. This corroborated findings from Bandyopadhyay et al. (2006), suggesting that an anterior–posterior gradient of interdigital BMPs is not involved in the determination of digit identities. Rather, BMP signaling is required for initial chondrogenesis as outlined above and BMP/pSmad signaling appears to be the driving force for the outgrowth of the digit condensations by means of distally directed chondrogenesis.
Montero et al. (2008) furthermore showed that the integrity of the DC/PFR required AER signaling, which links the DC/PFR to the elongation model proposed by Sanz-Ezquerro and Tickle (2003). These studies altogether suggest that the outgrowth of digit condensations depends on a timely and spatially defined input of strong BMP/Smad1/5/8 signaling. The final length of the digit and the number of phalanges appear to be defined by a combination of the intensity of the signaling input as well as its duration. Thereby an intricate balance between maintenance/proliferation of the distal mesenchyme and specification of chondroprogenitors has to be maintained to prevent either the depletion of the progenitor pool or a failure in their specification leading presumably to differentiation along other cell lineages. Figure 4 shows the DC/PFR in the mouse in digit 3 at day E13.5. It can be seen that in wild-type mice, pSmad1/5/8 activity coincides with and slightly precedes the prespecification of mesenchymal cells to chondroprogenitors characterized by the expression of Sox9 (Fig. 4).
Recent studies addressing the pathomechanism of brachydactylies A1 (BDA1, MIM#112500) and BDB1 (MIM#113000) have begun to unravel the genetic mechanisms governing digit elongation driven by chondrogenic commitment of distal mesenchymal precursors (Gao et al., 2009; Witte et al., 2010b). In studying a mouse model for BDA1 (see Fig. 3), caused in humans by mutations in Indian hedgehog (IHH; Gao et al., 2001), Gao et al. (2009) showed that this condition was caused by an impaired distal outgrowth of the digit condensation due to reduced commitment of distal progenitor cells. Ihh is produced by cells within the cartilage condensation and then diffuses across the growth plate, thereby forming a gradient (Vortkamp et al., 1996). This gradient is modulated by binding to the hedgehog receptor patched (Ptc1) itself, but also by binding to antagonists like hedgehog interacting protein 1 (Hip1). In the growth plate, Ihh exhibits short-range signaling to the proliferative chondrocytes expressing high levels of Ptc1, and long-range signaling to the periarticular perichondrium of the joints, where it induces parathyroid hormone-like hormone (Pthlh). Gao et al. (2009) showed that a BDA1 mutation in Ihh (Ihh p.E95K) led to a reduced binding affinity of Ihh to both Ptc1 and Hip1, thereby increasing the Ihh signaling range. This resulted in reduced short-range Ihh signaling in the growth plate chondrocytes but conversely caused elevated long-range Ihh signaling in the periarticular perichondrium, leading to increased expression of Pthlh. Because Pthlh is a negative feedback regulator of Ihh expression (Vortkamp et al., 1996), the increased dosage of Pthlh prevented the induction of Ihh expression in distal, newly formed condensations and finally led to a drastic decrease of Ihh signaling in distal mesenchyme. This combination of events leads to small or absent middle phalanges, the BDA1 phenotype. The study demonstrated a role for Ihh in directing chondrogenesis different from its well-known role in controlling hypertrophic differentiation of chondrocytes in the growth plate.
Witte et al. (2010b) analyzed a mouse model for human BDB1 (see Fig. 3). BDB1 is caused by mutations in ROR2 and results in hypoplastic/absent terminal phalanges (Oldridge et al., 2000; Schwabe et al., 2000). ROR2 encodes a receptor tyrosine kinase (Masiakowski and Carroll, 1992). BDB1 mutations lead to the expression of a truncated protein that lacks parts of the intracellular domains. Ror2 is known as an alternative Wnt co-receptor mainly for Wnt5a (reviewed in Minami et al., 2010). Wnt5a signaling by means of Ror2 can counteract the canonical Wnt/β-catenin pathway (Mikels and Nusse, 2006), while protein constructs representing human BDB1 mutations lose this inhibitory effect (Winkel et al., 2008; Witte et al., 2010a). Using the BDB1 mouse model (Ror2 p.W749X), Witte et al. (2010b) demonstrated the presence of a DC/PFR in the mouse (see also Fig. 4). This signaling center was absent in BDB1 (Ror2 p.W749X) mice and reduced in BDA1 (Ihh p.E95K) mice coinciding with the stronger phenotype of the BDB1 mouse model compared with the BDA1 mouse model. This implicated a DC/PFR-like structure in the elongation of the digital rays in mouse and in humans and suggested that this was based on a similar mechanism as proposed for the chick. It also suggested that both Ihh and Ror2 are upstream of BMP signaling in the PFR. Ror2 in this context seems to be required for controlling Wnt/β-catenin signaling in the distal mesenchyme, which is known to oppose BMP signaling.
SEGMENTATION OF THE DIGITS
Concurrent with the elongation of the digit condensations a segmentation into the metacarpals/-tarsals and the individual phalanges (in mammals 2 in the first digit and 3 in digits 2–5) takes place. This process occurs in a proximal to distal sequence and eventually results in the formation of the interphalangeal joints. As a first step cells change their shape, become flat and form the so called interzone. These cells start to express Gdf5, an early marker of joint formation in chicken and mouse (Chang et al., 1994; Francis-West et al., 1999b; Storm and Kingsley, 1999). Gdf5 expression, however, is not limited to these cells but is also present in progenitor cells surrounding the joint anlage. Genetic evidence suggests that Gdf5 expressing cells constitute joint progenitors in the limb that will later on develop into the cartilaginous joint surface, the joint capsule and ligaments. In concordance with its pro-chondrogenic role, GDF5 appears to inhibit the formation of joints in the vicinity of its expression (Storm and Kingsley, 1999). This might contribute to the correct spacing of joints, which is supported by the appearance of hypersegmentation (i.e., the emergence of supernumerary joints) in patients with BDC (GDF5 loss of function).
Abnormal joint formation and joint fusions are observed in individuals with symphalangism (SYM1; MIM#185800) and/or the more severe form of this condition, multiple synostosis syndrome (SYNS1; MIM#186500). These syndromes were shown to be caused by mutations in the BMP antagonist NOGGIN (NOG; Gong et al., 1999; Dixon et al., 2001; Takahashi et al., 2001; Brown et al., 2002) indicating that inhibition of BMP signaling is pivotal for the process of joint formation. Corroborating this, disruption of the mouse Noggin gene results in enlarged condensations and joint fusions (Brunet et al., 1998). Of interest, a certain subset of mutations in GDF5 were also shown to be associated with joint fusions (Seemann et al., 2005; Dawson et al., 2006; Wang et al., 2006; Yang et al., 2008; Seemann et al., 2009). These mutations generally result in an activation of BMP signalling, for example by an increased binding of GDF5 to BMPR1A. In another mutation GDF5 was shown to become insensitive for repression mediated by NOGGIN (Seemann et al., 2005, 2009). Thus, mutations in humans conferring a gain of function in GDF5 either by increased activity or decreased inhibition lead to joint fusion phenotypes, while dominant loss of function mutations result in BDC.
Moreover, a mutation in BMPR1B has been identified in patients showing a BDC/SYM1-like phenotype (Lehmann et al., 2006); however, the functional consequence of this mutation and its difference to mutations in BMPR1B causative for BDA2 remain unclear. Altogether it appears that the fine-tuning of the GDF/BMP pathway is required for both, the formation of phalangeal condensations, and the following elongation and segmentation. Disturbances in this signaling network can cause, dependent on the direction the signaling network shifts, distinct or even opposite phenotypes. Furthermore, specific mutations in NOG, that do not result in an apparent loss of protein function as shown by in vitro testing, do not cause SYM1 or SYNS1, but brachydactyly type B2 (BDB2; MIM#611377; Fig. 1; Lehmann et al., 2007). However, the functional consequence of these mutations remains to be elucidated.
These observations demonstrate that joint formation, segmentation and the formation and growth of individual phalanges are tightly coupled processes. But what is more important for the brachydactyly pathogenesis, joint formation or growth of the phalangeal condensation? Detailed studies using mouse models of BDA1 and BDB1 delivered new insights into the pathology and sequence of these events.
Human BDA1 is characterized by missing middle phalanges, while BDB1 shows a variable phenotype with short/absent medial and/or distal phalanges (Fig. 1). This is often combined with distal symphalangism in the milder cases. The BDB1 phenotypes correlate with the type and position of the mutation (Schwabe et al., 2000) and are caused by differential stability and intracellular localization of the resulting truncated protein product (Schwarzer et al., 2009). Witte et al. (2010b) analyzed mouse mutants harboring a targeted insertion of a human mutation in the endogenous Ror2 locus (Ror2 p.W749X), causing severe BDB1 phenotypes. Homozygous RorW749X/W749X mice lack the medial phalanges, this phenotype has previously been considered to be a defect in the formation of the last interphalangeal joint (Raz et al., 2008). Witte et al. (2010b) showed that preceding the failure of joint formation, elongation of the phalangeal condensation was severely impaired. This led to a primordium for the two distal phalanges that did not reach the critical size for insertion of an additional joint. This finding was corroborated on a genetic basis using the BDA1 mouse model (Ihh p.E95K), in which the phenotype is also caused by impaired elongation of the phalangeal condensations (Gao et al., 2009). Heterozygous IhhE95K/+ mice display shortened middle phalanges in digits 2 and 5; hence the critical size for segmentation was reached. However, in compound heterozygotes IhhE95K/+ / RorW749X/+, the phenotypic severity in digit 2 was increased and the middle phalanx was further shortened and now showed fusion to the terminal phalanx (Fig. 5) thus showing a distal symphalangism, similar to mild cases of BDB1.
These results provide further genetic support for a model in which the formation of a novel joint needs a critical distance from the previously formed joint (Hartmann and Tabin, 2001; Archer et al., 2003). In the chicken, the elongation of the digit ray is carried by elongation and segmentation of the penultimate phalanx. It is noteworthy that each digit appears to have its own critical threshold for segmentation, because the length of phalanges in the chicken foot is different from anterior to posterior digits (Sanz-Ezquerro and Tickle, 2003). Segmentation is under negative control of signals from the previous joint, as well as from signals from the AER. Those progenitors that have not undergone differentiation to cartilage at the time the AER degenerates, will automatically undergo differentiation into a tip structure (i.e., the terminal phalanx; Casanova and Sanz-Ezquerro, 2007). It appears that these mechanisms can also be transferred to the mouse, because impairment of outgrowth in both BDA1 and BDB1 mice results in a shortening of the penultimate phalanx (p2) up to the point where a threshold is reached that prevents formation of an additional joint. In BDB1 mice, apparently all remaining material that condenses distal to the proximal phalanx undergoes differentiation to a terminal phalanx-like structure in a mode as proposed for the chick.
Of interest, Ihh itself seems also to be involved in joint formation, because mice devoid of Ihh exhibit joint fusions in the digits (St-Jacques et al., 1999; Koyama et al., 2007). Ihh can act synergistically with Wnt/β-catenin signaling in this process (Mak et al., 2006), which itself is sufficient and necessary to induce the formation of synovial joints in a paracrine manner (Hartmann and Tabin, 2001; Guo et al., 2004).
GROWTH AND MAINTENANCE OF DIGITS AND JOINTS
An important feature of the initial condensation is the withdrawal of most cells undergoing differentiation to cartilage from the cell cycle (Janners and Searls, 1970; Summerbell and Wolpert, 1972). During chondrogenesis in explant culture systems, BMP signaling induces the cell cycle inhibitor p21 (Carlberg et al., 2001), indicating that BMPs couple cell cycle withdrawal with cartilage differentiation. However, the BMP pathway target Sox9 can also bind and activate the p21 promoter (Panda et al., 2001). Once the condensations of the phalanges are constituted and separated by joints, the cartilage growth plate is established, a structure that drives growth of most of our skeleton until the end of puberty. In the growth plate, chondrocytes undergo a stereotype set of differentiation steps from small, round reserve zone chondrocytes adjacent to the joints, to highly proliferative columnar chondrocytes, and finally to large hypertrophic chondrocytes that eventually undergo apoptosis and become replaced by trabecular bone. This process is accompanied by the coordinated formation of the bone collar (reviewed in Karsenty and Wagner, 2002). Two main mechanisms regulate the growth rate of the individual skeletal element: the pace of differentiation of chondrocytes and their proliferation rate. The proliferation of chondrocytes is regulated by a variety of signaling systems including Ihh and BMP/GDF signaling. Both, Ihh expressed in prehypertrophic growth plate chondrocytes, and BMPs/GDFs, expressed mainly in the perichondrium and the prospective joints, are positive regulators of proliferation (Francis-West et al., 1999a; St-Jacques et al., 1999; Buxton et al., 2001; Minina et al., 2002).
Ihh has a dual role in the growth plate as it promotes proliferation and also controls the pace of hypertrophic differentiation (Vortkamp et al., 1996; St-Jacques et al., 1999). This predicts that mutations in IHH can affect both, the condensation/elongation of cartilage elements as discussed above, as well as the growth of those elements by reducing proliferation and/or altering chondrocyte differentiation. In accordance with this presumption, BDA1 patients often exhibit a short stature (Farabee, 1903; Gao et al., 2001). Hence, in addition to the condensation defect, a growth plate defect might contribute to further shortening of the digits in BDA1. This is supported by findings in the BDA1 mouse model showing that reduced Ihh downstream signaling in the growth plate led to decreased proliferation and thus shortened skeletal elements (Gao et al., 2009).
Ihh is known to regulate the pace of hypertrophic differentiation in the growth plate by forming a negative feedback loop with parathyroid hormone-like hormone (Pthlh, also called parathyroid hormone-related peptide Pthrp; Lanske et al., 1996; Vortkamp et al., 1996). Ihh is expressed in prehypertrophic chondrocytes in the growth plate and by means of long-range signaling induces the expression of Pthlh in the periarticular perichondrium (Vortkamp et al., 1996; Koziel et al., 2004). Pthlh diffuses across the growth plate and signals by means of its receptor parathyroid hormone 1 receptor (Pth1r) to prevent further chondrocytes from undergoing differentiation to prehypertrophic cells, thus in turn reducing the expression of Ihh. Mice with inactivated alleles of Pthlh/Pthrp show shortened skeletal elements caused by a premature differentiation of growth plate chondrocytes in conjunction with a reduced proliferation rate (Karaplis et al., 1994). In humans loss-of-function mutations in PTHLH cause autosomal dominant brachydactyly type E with short stature (BDE, MIM#113300), a condition characterized by shortened metacarpals/metatarsals and in some cases also shortened phalanges (Klopocki et al., 2010; Maass et al., 2010; Fig. 1).
GDF5 has been involved in cartilage condensation and proliferation, but it appears to be also required for maintenance. Homozygous inactivation of Gdf5 in brachypodism mice results in aberrant apoptosis in digit cartilage, which might contribute to the brachydactyly phenotype in these mice (Takahara et al., 2004). This might also contribute to the phenotype of the recessive acromesomelic chondrodysplasias (see above), however in the case of dominant BDC, caused by haploinsufficiency, this appears unlikely. Inactivation of BmpR1a function in the joints by means of a GDF5 driven Cre recombinase resulted in a gradual erosion of joint cartilage, similar to human osteoarthritis. Thus, BMP receptor signaling is required not only for the formation of cartilage condensations and joints, but also for ongoing maintenance of articular cartilage after birth (Rountree et al., 2004).
DIGIT MALFORMATIONS: BMP SIGNALING AT CENTER STAGE
This review aims at integrating human digit malformation syndromes in the growing body of knowledge about limb development inferred from model systems such as the chick or mouse. Previous studies into human malformation syndromes have shown that the knowledge about mutational mechanisms is important to categorize and diagnose these conditions. Furthermore, careful analysis of the developmental pathology can provide novel insights into the mechanisms that govern digit development. The brachydactylies have been exceptionally instructive for our understanding of digit development as they represent a spectrum of defects that cover a broad range of developmental processes including digit condensation, digit growth, and joint formation. The majority of human brachydactylies (BDA2, BDC, SYM) are caused by mutations in genes encoding components of the BMP pathway. BDA1, caused by mutations in IHH, can neatly be integrated into this network, because BMP and Ihh signaling are interconnected at various points (see e.g., Pathi et al., 1999; Minina et al., 2001, 2002). BMP/Smad signaling can directly activate the Ihh promoter (Seki and Hata, 2004). In turn, Ihh can promote the expression of Bmps (Pathi et al., 1999; Minina et al., 2001), and Ihh-deficient mice show reduced expression of Bmp4 in distal mesenchyme during digit elongation (Witte et al., 2010b). BDB1, caused by mutations in ROR2, can also be linked to this signaling network, because ROR2 modulates Wnt/β-catenin signaling (Mikels and Nusse, 2006), which acts antagonistically to BMP signaling in chondrogenesis (Fischer et al., 2002; Akiyama et al., 2004). Furthermore, Ror2 has been shown to modulate Gdf5 signaling by inhibition of Smad1/5 signaling and by activating a Smad-independent pathway (Sammar et al., 2004). This led to the proposition that brachydactylies represent a molecular disease family, whose different yet partially overlapping phenotypes can be explained by shifts in a unifying molecular network (exemplified in Fig. 6), which altogether regulates and fine-tunes chondrogenesis and cartilage maturation in the digits (Mundlos, 2009).
No distinctive subdivision between the different steps of digit development can be made, because they are interdependent and functionally related to each other. Global patterning is translated to local paracrine signaling that guides chondrogenesis. Moreover, patterning information that controls digit identities is at least partially transduced by means of mesenchymal BMPs feeding into the DC/PFR driving distal elongation. Segmentation of digits is regulated by interplay of signals from the condensation itself, the pace of distal elongation by means of the DC/PFR and inhibition of joint formation from the AER. The forming joints are again signaling centers that influence the proliferation and differentiation of the adjacent cartilaginous growth plates. In the line of this the brachydactylies, which show features that might represent changes in digit identities (e.g., absent medial phalanges in BDA1), cannot be considered as defects in the establishment of digit identity patterning information. Rather, they represent a defect in the perception of these signals and thus a failure in the transfer of this program into actual local chondrogenesis, the neglected link between patterning of digit structures and the final shape and appearance of the digits.
SUMMARY: DIGIT FORMATION AND MALFORMATION
Integrating information from human syndromes and their analysis in animal models has shed new light on the contribution of individual key players to different phases of digit development (Fig. 7). During emergence of the initial cartilage elements in the limb bud (from E10.5 to E12.5 in the mouse), chondrogenesis appears to be restricted to the core of the limb bud due to suppressive ectodermal Wnt/β-catenin and Fgf signaling (Hill et al., 2005; ten Berge et al., 2008), with condensation apparently following a “passive” mechanism. Chondrogenesis is positively regulated by BMP signaling from the mesenchyme by means of BMPs 2, 4, and 7 and Gdf5. Disturbance of BMP signaling at this phase leads to reduced size of the initial condensation resulting in brachydactyly syndromes. After establishment of the initial condensation, the DC/PFR assumes an organizer-like function and “active” elongation of the condensation is driven by means of strong BMP/pSmad1/5/8 signaling in this structure (Montero et al., 2008; Suzuki et al., 2008). This is performed in interplay with AER/Fgf signaling to keep the balance between maintenance of progenitors and chondrogenic differentiation. The DC/PFR itself is under positive influence from Ihh and under negative control by Wnt/β-catenin signaling, which itself is controlled by Ror2 (Witte et al., 2010b). Thus, Wnt/β-catenin signaling controls the overall size of the initial condensations but also fine-tunes digit elongation by negatively regulating the cellular activity in the DC/PFR. A disruption of BMP signaling in the DC/PFR leads to BDA1 and BDB1. Joint formation is under control of Ihh, Wnt/β-catenin and BMP signaling. Elevation of BMP signaling by activating mutations in GDF5 or inactivating mutations in the antagonist NOGGIN lead to SYM1 or SYNS1. Finally, the individual skeletal elements establish the growth plates and expand. Disruption of the IHH-PTHLH feedback loop results in BDE (mutation in PTHLH); defective IHH signaling in BDA1 might contribute to the growth defects observed in this condition.
With the emergence of novel technologies allowing genome wide mutation analysis like exome sequencing, region capture followed by next generation sequencing, or array-CGH, novel mutations that complement the overall picture will be identified. So far, most interest has been on the function of genes/proteins within the signaling networks, while genomic regulatory events that ultimately control the entire process have been neglected. We are now beginning to realize how complex the genomic regulation of the key players and pathways that govern digit development is. Human mutations may help to unravel this next level of complexity by identifying genomic regions of importance. The first regulatory mutations affecting putative regulatory elements of such genes have just been identified and many more are likely to follow. It will be a major challenge to analyze the molecular consequences of such genomic variations in the mouse and to translate them into molecular consequences.
S.S. and S.M. were funded by grants from the Deutsche Forschungsgemeinschaft (DFG).