Digit identity has been studied using the chick embryo as a model system for more than 40 years. Using this model system, several milestone findings have been reported, such as the apical ectodermal ridge (AER), the zone of polarizing activity (ZPA), the Shh gene, and the theory of morphogen and positional information. These experimental results and models provided context for understanding pattern formation in developmental biology. The focus of this review is on the determination of digit identity during limb development. First, the history of studies on digit identity determination is described, followed by descriptions of the molecular mechanisms and current models for determination of digit identity. Finally, future questions and remarkable points will be discussed.
Humans have five distinct digits along the anterior-posterior (thumb to little finger) limb axis (Fig. 1). When we look at our hands and feet, we can see that the thumb/big toe is the widest and shortest among the five digits, while the little finger is the narrowest and smaller than the middle three digits, and the middle three digits – digits 2, 3, and 4 – appear similar to each other, although their morphology and length are slightly different. These morphological and anatomical differences between digits are called “digit identity”. Hands and feet develop from a small population of somatic lateral plate mesoderm cells, called the limb bud, which grows out from the body wall. The limb is characterized by three axes: the anterior-posterior (AP) axis described above, the proximal-distal (PD) axis (shoulder/thigh to fingers), and the dorsal-ventral (DV) axis (knuckles to palm). The study of digit identity is designed to uncover the mechanisms of AP axis determination in the limb bud. The limb skeleton is organized into three anatomical regions: (i) the stylopod (humerus/femur), (ii) the zeugopod (ulna and radius/tibia and fibula), and (iii) the autopod, which consists of carpal/tarsal bones, metacarpal/metatarsal bones, and phalanges (Fig. 1). Phalanges are classified into three types according to their position along the PD axis: proximal phalanx, middle phalanx, and terminal phalanx, which has nails or claws/hoofs. Humans have two phalanges in the thumb/big toe, a characteristic morphological criterion of digit 1 in all amniotes (Vargas & Fallon 2005a,b; Wagner & Vargas 2008; Woltering & Duboule 2010). The posterior four digits have three phalanges, so other criteria are necessary to distinguish each digit. The mouse has five digits, as well, and the morphological differences in each are similar to those in human digits, with the middle three digits (digits 2, 3, and 4) appearing similar (Fig. 2). To distinguish among digits 2, 3, and 4 in the mouse limb morphologically, it is necessary to assess metacarpal/metatarsal articulation in the carpals/tarsals at least after E16.5, when the ossified tip of the terminal phalanx begins to be visible (Verheyden et al. 2005). In the forelimb, the digit 2 metacarpal articulates with the trapezoid and central carpal bones. Similarly, the digit 3 metacarpal articulates with the capitate bone, and the digits 4 and 5 metacarpals with the hamate bone. The digit 1 metacarpal articulates with trapezium bone. In contrast, in the hindlimb, the digit 2 metatarsal articulates with the intermediate and navicular bones and is located proximal to the intermediate bone, while the digit 3 metatarsal articulates with the lateral cuneiform bone and the digit 4 metatarsal with the cuboid bone.
The avian forelimb is a highly derived structure that is modified for flight and is not representative of amniote limb development in general (Fig. 2). The avian wing has three digits that were thought to be digits 2, 3, and 4 for most of the last century. The reasons for this have been reviewed extensively recently (see for example, Young & Wagner 2011; Young et al. 2011 and will not be discussed here). However, Tamura et al. reported, using cell lineage analysis, that the chick wing has digits 1, 2, and 3; this finding fits nicely with paleontological findings (Tamura et al. 2011; see also Towers et al. 2011). Transcriptome data for each digit in the chick forelimb and hindlimb also suggest that gene expressions in digit 1 in the wing are similar to those of digit 1 in the leg (Wang et al. 2011). In contrast, the chick foot has four digits, which are easily distinguishable as digits 1, 2, 3, and 4 (Fig. 2). Compared with the chick wing, the chick leg is more closely representative of amniote limb structures and is a highly useful model for studying digit identity determination because of its anatomy.
Importantly, it is easier to identify each digit morphologically in the chick embryo than in the human or mouse. Digit identity is characterized by three major morphological criteria: number, size, and shape of the phalanges (Suzuki et al. 2008). All three criteria are necessary to judge each digit identity. To identify each digit by these criteria, first, it is easy to count phalanx numbers in the hindlimb. To establish this, we have to examine terminal phalanx formation, which occurs after St. 36 (Hamburger & Hamilton 1951). Similar morphological criteria should also be applied to the mouse digits. Gli3−/− mice have unidentifiable digits because of their abnormal phalanx morphology (Litingtung et al. 2002). Therefore, the identities of digits in Gli3−/− mice cannot be determined by morphological criteria, although attempts to determine digit identity in these mice have been reported (e.g., Lopez-Rios et al. 2012). With regard to caution about counting phalanx numbers, the chick hindlimb is an exceptionally useful model system for studying digit identity determination. In the hindlimb, digits 1, 2, 3, and 4 have two, three, four, and five phalanges, respectively, so this criterion is unambiguous and therefore easy to use as a first step (Fig. 2).
Discovery of the ZPA and the Shh gene
A milestone in the study of digit identity is the discovery of the zone of polarizing activity (ZPA), located in the mesoderm along the posterior edge of the limb bud. This small group of cells establishes posterior limb identity. In 1968, Saunders and Gasseling discovered the polarizing property of these cells (Saunders & Gasseling 1968). They were grafting the region that undergoes cell death – called the posterior necrotic zone (PNZ) – to the anterior border of a host limb bud. In this paper, they did not name it ZPA. In 1969, Wolpert published a theoretical basis for the ZPA hypothesis (Wolpert 1969). He proposed that positional information that establishes the polarity of the limb depends on the distance from the posterior PNZ zone. This model provides a fundamental basis for understanding of pattern formation in developmental biology. Subsequently, Saunders used the name “zone of polarizing activity” (ZPA) for the first time in an abstract in 1972 (Saunders 1972); and subsequently co-authored a paper in 1973, mapping the ZPA's polarizing activity by induced polarized anterior supernumerary digits, with the most posterior supernumerary digit closest to the ZPA graft (MacCabe et al. 1973). In 1975, Tickle et al. published a paper on the specification of digits by positional information, in which they started to interpret their data based on the model put forward in Wolpert's theoretical paper (Tickle et al. 1975). They proposed that digit identity is specified by a morphogen gradient that directly reflects positional information provided by the ZPA.
In 1993, Riddle et al. established a milestone in the study of digit identity (Riddle et al. 1993) with the discovery that the Shh gene, expressed at the posterior border of the limb, is a ZPA factor (Fig. 4A). SHH protein released from beads was found to induce a mirror-image duplication of digits in the chick limb bud (López-Martínez et al. 1995). In support of the morphogen hypothesis, SHH protein induced more posterior digits with higher dosage (Yang et al. 1997). This result suggested that SHH functions as a morphogen to determine digit identity in a dose-dependent manner along the AP axis in the limb. Thus, the ZPA hypothesis was elegantly supported by the discovery of Shh gene expression in the chick limb. In Shh−/− mice, only digit 1 was formed in the hindlimb, and no digits were formed in the forelimb (Chiang et al. 2001). Similarly, in the chick mutant oligozegodactyly (ozd), in which Shh is not specifically expressed in the limb bud, only digit 1 formed as well in the hindlimb; these had two phalanges with the precise morphology of digit 1 for each phalanx (Ros et al. 2003). These results are not predicted by the morphogen model, indicating that digit 1 in the hindlimb is formed in a Shh-independent manner. However, similar to Shh−/− mice, no digits formed in the ozd wing. Interestingly, Amano and Tamura have suggested that digit 1 in the chick forelimb is formed independently of ZPA function because an implanted ZPA never induced an independent ectopic digit 1. When an additional digit 1 did form it was associated with the original digit 1 (Amano & Tamura 2005).
Approaches to the mechanisms of SHH function
Harfe et al. reported that concentration of and time of exposure to SHH determine digit identity in both mouse and chick embryos (Harfe et al. 2004). Surprisingly, using a LacZ cell lineage tracing approach, it was shown that digits 5, 4 and the posterior length of digit 3 in the mouse were derived from the descendants of Shh-expressing cells. It was proposed that the most posterior digit, digit 5, received the highest concentration of and experienced the longest exposure to SHH in the mouse, that digit 4 received a shorter exposure to SHH than did digit 5, that digit 3 had a shorter exposure and received a lower concentration of SHH, and that digit 2 received the lowest concentration of SHH among the posterior four digits. This genetic conclusion was supported by inhibition of SHH signaling with cyclopamine treatment in the chick embryo (Scherz et al. 2007). When cyclopamine was applied at St. 20 (E3.5) in the hindlimb, no digit 4 formed, and digit 3 was partially transformed to digit 2. In contrast, when cyclopamine was used at St.22/23(E4.0), only digit 4 was malformed. Thus, anteriorization of digit or posterior digit loss was induced in a SHH exposure, time-dependent manner. These observations suggest that more posterior digit formation and posterior digit identity determination require longer exposure to SHH.
Towers et al. integrated growth with SHH function and proposed a “growth and morphogen model” (Towers et al. 2008). In that paper using the chick wing, the authors concluded that Shh promotes both anterior-posterior expansion of the limb bud and more posterior digit specification during limb development. Specifically, in this model, more SHH protein over longer times progressively specifies more posterior digit identity from prospective digit 1 identity to digit 3 identity. At the same time, SHH also induces expansion of the digit-forming field to control digit number.
Two separate approaches in the mouse do not support the “growth and morphogen model” in the chick wing
Gli1 is a direct target of SHH signaling, and Gli1 expression has been used as a read-out of SHH signaling activity in target cells. A Gli1 reporter line with a knock-in CreERT2 in the mouse showed an interesting and instructive result (Ahn & Joyner 2004). When these investigators crossed with the R26R allele to detect SHH signaling activity at the limb bud stages, they found that Gli1 expression was highest at the position of prospective digit 5 at E10.5. However, its activity was the highest at the position of prospective digit 4 at E11.5. Interestingly, only weak Gli1 reporter expression was detected at the position of prospective digit 5 in Gli2−/− mice in which a normal digit 5 was formed. In that study, the authors proposed that the most posterior digit, digit 5 in the mouse, is determined by a high SHH concentration for a short time at the early limb bud stages.
Zhu et al. reported a similar conclusion (Zhu et al. 2008). They knocked out Shh by Hoxb6/CreERT at different stages of mouse limb development. They found the order of condensation of the posterior four digits was 4→2→5→3. When the duration of expression of Shh was shorter, digits did not form in reverse order. For example, when Shh expression was inhibited from E10.5, only digit 3 was lost. In addition, when Shh expression was inhibited from E10, digits 3 and 5 were lost. When Shh expression was inhibited from E9.5, digits 3, 5, and 2 were lost. A significant point is that digits were progressively lost without homeotic transformation of digit identity by inhibition of time of exposure of SHH. In light of the results, the authors suggest that digit identity is determined by SHH at limb bud stages. They concluded that Shh function has two phases. The first is transient: the early patterning phase regulates digit identity, as Ahn et al. indicated. The second is an extended growth-promoting phase, during which the prospective digit precursor cell region expands from digits 4→2→5→3 in this order. It is an interesting question how the sequence of each digit precursor region develops in the mouse limb and whether the “growth and morphogen model” by Towers et al. fits with this experimental result.
SHH is known to be a mitogen (Kenney & Rowitch 2000). A missing part of the puzzle that was not mentioned is whether ZPA cell descendants respond to the gradient of SHH as a mitogen for proliferation. How is proliferation induced in the cell descendants of the ZPA? In the mouse, if these descendant cells respond to the gradient of SHH, the posterior digit precursor cells may expand differently along the gradient. However, in the mouse limb, the prospective digit precursor regions expand in the order, digit 4→2→5→3. This means that cell proliferation in the mouse limb is not controlled directly along the proposed mitogen gradient of SHH. In this context, when and how proliferation and digit identity “specification” occur along the SHH gradient in the mouse cannot be explained by the “growth and morphogen model”.
SHH signaling components and digit identity
SHH is secreted from the ZPA and determines the AP axis in the limb bud. The SHH protein is cholesterol-modified at the C-terminus and palmitate-modified at the N-terminus to become the mature and active form (Porter et al. 1996; Pepinsky et al. 1998). Palmitate addition is catalyzed by a putative acyltransferase called Skinny hedgehog (Ski) (Chamoun et al. 2001). Mice expressing Shh that lack palmitate modification show a reduction in digit 2 in the hindlimb (Chen et al. 2004a). Dispatched (mDisp1) is necessary for the release of SHH from the ZPA and restricts the spread of the SHH gradient (Li et al. 2006). Loss of mDisp1 activity results in loss of digit 2, as well (Harfe et al. 2004). These results indicate that digit 2, which is not derived from SHH descendants, is also the most sensitive to decreased SHH diffusion.
Patched is a receptor for SHH and usually inhibits Smoothened (Smo) (Stone et al. 1996) (Fig. 3). In the absence of SHH, SMO is held in intracellular trafficking vesicles. GLI1, 2, 3 are transcriptional factors downstream of SHH signaling (Büscher & Rüther 1998). During limb bud development, Gli3 is necessary for determination of digit number and identity (Litingtung et al. 2002). Gli3−/− mice show polydactyly, and the resulting digits look like posterior digits, but as noted above they are unidentifiable because of the size and shape of the phalanges (Litingtung et al. 2002). In the absence of SHH, GLI3 is located in the primary cilium in a complex with KIF7 and SUFU (Fig. 3A). GLI3 is phosphorylated by protein kinase A (PKA) (Wang et al. 2000; Hsu et al. 2011). Phosphorylated GLI3 is recognized by β-TRCP, and GLI3 ubiquitination is induced, resulting in partial degradation in the proteasome (Bhatia et al. 2006). This short form of GLI3, GLI3R, inhibits transcription of target genes (Wang et al. 2000).
In the presence of SHH, SMO is released from inhibition by Patched (PTC), and it is phosphorylated by GRK2 (Chen et al. 2011). Subsequently, SMO makes a complex with KIF3A and β-ARRESTIN and inhibits SUFU (Kovacs et al. 2008). Under these conditions, GLI3 is maintained in a long-activator form called GLI3A (Litingtung et al. 2002). GLI3A forms in direct proportion to SHH signal, and it induces expression of target genes (Fig. 3B). Conditional knockout of Ptc1 or Kif3a causes polydactyly with loss of AP polarity (Liu et al. 2005; Butterfield et al. 2009) due to upregulation of SHH signaling in the entire limb bud. In the normal limb bud, the GLI3A:GLI3R ratio is high at the posterior side in the presence of SHH, and it is proposed that the GLI3A:GLI3R ratio determines digit identity along the AP axis by Litingtung et al. Interestingly, as predicted by this model, the Gli3Δ699 mouse, which mimics the processed GLI3 isoform GLI3R mouse, has biphalangeal digits unique to digit 1 (Hill et al. 2009). This suggests that digits have anterior identity in the presence of only GLI3R. In contrast, constitutive activation of SHH signaling in the limb results in polydactyly, with posterior digit-like structures that have four phalanges in the chick hindlimb (Litingtung et al. 2002). Taken together, there is support for the hypothesis that the GLI3A:GLI3R ratio is crucial for determination of digit identity downstream of SHH signaling.
Necessity for SHH and its downstream target
In 2002, Shh−/−;Gli3−/− compound knockout mice were reported (Litingtung et al. 2002; te Welscher et al. 2002). Surprisingly, Shh−/−;Gli3−/− mice had digits in the autopod, and they were polydactylous, with unidentifiable digits. The compound Shh−/−;Gli3−/− null was indistinguishable from the Gli3−/− null mouse limbs, even though the Shh gene was present. These observations indicate that the autopod has an intrinsic potential to form digits, even if Shh is not expressed, or the downstream genes are not present. In the Wolpert ZPA model, the gradient of SHH along the AP axis induces particular digits with distinguishable identities. However, the foregoing genetic data indicate that digits without identity can form in the absence of Shh and that Shh is necessary for precise digit number to form, as well as determination of digit identity. Interestingly, Shh−/−;Gli3+/− mice, which express only GLI3R, had two biphalangeal digits, identified as digit 1 in the hindlimb. These data again support the idea that the GLI3A:GLI3R ratio is critical for determining digit identity. Presently we do not understand how digit primordia start to form by intrinsic mechanisms.
Digits can form without Shh; however, Shh expression is necessary for digit identity. Shh expression is undetectable after St. 26 (Dunn et al. 2011), which is at the interface of ankle (tarsals) determination and autopod (metatarsals) initiation by AER removal (Rowe & Fallon 1981). Furthermore, GLI3R is the major form of GLI3 in the autopod (Chen et al. 2004b). Therefore, it is likely that SHH is not directly functioning in the autopod. Harfe et al., proposed that limb bud cells acquire a “memory” downstream of SHH and that this memory would control digit identity at later stages (Harfe et al. 2004). Such a memory would either be a part of the SHH signaling pathway or a separate secondary signal pathway that would directly determine digit identity in the emerging autopod stages.
Secondary signaling involving digit identity determination downstream of Shh
HoxD genes show nested expression along the AP axis in the autopod, and each expression pattern is correlated with condensed cartilage of the digit primordium (Yokouchi et al. 1991). In Shh−/− mice and chick ozd mutants, Hoxd11, 12, and 13 expression is downregulated (Chiang et al. 2001; Ros et al. 2003). Therefore, SHH signaling is necessary for 5′HoxD expression in the autopod. Morgan et al. reported that misexpression of Hoxd11 in the limb bud induced posterior homeotic transformation of digit 1 to digit 2 in the chick hindlimb (Morgan et al. 1992). This transformed digit has three phalanges with morphology very similar to that of normal digit 2. Hoxd11 is not expressed normally around digit 1. Therefore, this posteriorization of digit 1 to digit 2, is thought to be a gain-of-function of Hoxd11. This was the first report that identified gene expression directly controlling a particular digit identity. Hoxa13−/−;Hoxd13−/− mice show severe hypoplasia of digit formation (Fromental-Ramain et al. 1996). HoxA and D genes would also be involved in the formation and condensation of the digit primordium (Kmita et al. 2005). Furthermore, unique methods of exogenous gene manipulation were used in this gain-of-function experiment using the avian retroviral Replication-Competent Avian sarcoma-leukosis Virus (ASLV) long terminal repeat (LTR) with Splice acceptor (RCAS) system to introduce the Hoxd11 gene into the limb bud (Morgan et al. 1992). By using this system, an exogenous transgene can be expressed at the restricted area at a particular developmental time in the chick embryo. Similarly, we have reported that misexpression of Tbx3 and Tbx2 with an RCAS retrovirus induces posteriorization of digits 2 and 3 to digits 3 and 4 (Suzuki et al. 2004). Tbx3 and Tbx2 are transcription factors and are expressed in interdigit 3 and interdigit 4, but not in the digit primordium itself. Transcripts of Hoxd11 and Tbx3/2 genes are expressed in the interdigits rather than in the digit primordia. It is possible to infer from these and other data (Dahn & Fallon 2000) that the “memory” of SHH is retained in the surrounding mesenchyme of the digit primordium, interdigit, and determines the size and shape of the phalanges of each digit (Fig. 4A).
Interdigits determine each digit identity
Dahn and Fallon reported that interdigits (IDs) are necessary for and determine each digit identity in the chick limb (Dahn & Fallon 2000). They found that digit 2 and digit 3 were transformed to digit 1 and digit 2, respectively, after removing ID2 or ID3 in the chick hindlimb. These observations support a model that the interdigit is necessary for digit identity determination of the digit immediately anterior to it. In addition, when they inserted tantalum foil at the center of the digit 3 digital ray after bisection of the digit 3 digital ray along the PD axis, a bisected anterior digit 3 digital ray was transformed to digit 2, which had three phalanges with the complete size and shape of the normal digit 2, whereas a bisected posterior digit 3 developed as digit 3, which has four phalanges. Here, the digital rays are composed of the proximal cartilaginous digit primordium, the noncondensed vascular mesenchyme, and the distal-most avascular mesenchyme underlying the proximal-distal axis signaling center, the apical ectodermal ridge (AER). Very importantly, digit identity is not a fixed property, but a labile one, (i.e. digit identity is not determined), before the digit primordium starts to condense. When the digit 2 digital ray was implanted in ID3, digit 2 was transformed to digit 3. These results demonstrate that the interdigit has positional information for determining digit identity and for supporting the corollary that the digital ray does not have positional information when each digit primordium starts to develop.
Bone morphogenetic protein 2 (BMP2), 4, and 7 are expressed in the interdigits (Suzuki et al. 2004), and they are associated with apoptosis in interdigits (Gañan et al. 1996; Pajni-Underwood et al. 2007), resulting in separated digits in human, mouse, and chick. Importantly, when beads soaked with the BMP antagonist, Noggin, were implanted in ID3 at St. 27, before interdigital apoptosis occurs, digit 3 identity was transformed to digit 2 (Dahn & Fallon 2000), as is the case with ID3 removal. Thus, interdigital BMP signaling is involved in digit identity determination before inducing apoptosis at later stages. Misexpression of Tbx3 or Tbx2 also induces upregulation of BMP2, 4, and 7 in the interdigit (Suzuki et al. 2004). This observation is consistent with a role for interdigital BMP signaling in digit identity determination. Taken together, digit identity has not yet been determined when each digit primordium starts to develop, and interdigits possess positional information, including BMPs, to determine anterior digit identity. However, compound knockout mice in combination with Bmp2/4/7 do not show transformation of digits in the mouse (Bandyopadhyay et al. 2006). This is interesting, but these double compound null mutants do not take into account all of the BMP family members tested, for example, Bmp5 (Zuzarte-Luís et al. 2004) Gdf5/6 (Settle et al. 2003).
The next question is how interdigital BMP signaling functions to determine digit identity. What is the target tissue of interdigital BMP signaling? We reported that the cells located distal to the condensing cartilage of the digit primordium, called the phalanx forming region (PFR), received interdigital BMP signaling (Suzuki et al. 2008) (Fig. 4B). Further, we showed that the cells located under the AER are continually incorporated into the PFR (Fig. 4C). Thus, the hypothesis is that the PFR is the location in a digital ray where digit identity is being determined by the posterior interdigit. The cells at the PFR are constantly changing. As distal elongation occurs, older more proximal cells (that were PFR cells) form cartilage, while new cells emerge from the influence of the AER to become the PFR. Cells that were the PFR received positional information from the interdigit and form particular phalanx on condensation into cartilage.
Sox9 expression and phosphorylated SMAD1/5/8 are detected at the PFR (Fig. 4C). This means that a phalanx is being determined and starts to take on the characteristics of the cartilage lineage at the PFR. Interestingly, we found using an in vivo reporter assay that each PFR showed unique SMAD1/5/8 activity levels. When we implanted ID3 at the posterior side of the digit 1 digital ray, digit 1 was transformed to digit 3. This was accompanied by an activity level of SMAD1/5/8 in digital ray 1 characteristic of digital ray 3. The transformed digit 3 had four phalanges with the complete size and shape of the phalanges of a normal digit 3. These results are reminiscent of the digital homeotic transformation caused by misexpression of HoxD11 in the chick foot (Morgan et al. 1992). Thus, unique SMAD1/5/8 activity at each PFR is a characteristic of digit identity determination.
Future questions about digit identity determination
An important question is whether the morphogen theory is supported when the most recent data on limb patterning are taken into account. For example, digit 3 needs the longest Shh expression in the mouse (Zhu et al. 2008). Digit 5 identity may be determined under the condition of high SHH activity for a short time (Ahn & Joyner 2004). In addition, it is crucial that mouse digits were not transformed to other digits when the timing of Shh expression and activity are changed (Scherz et al. 2007; Zhu et al. 2008). Moreover, digit identity has not been determined when the digital rays of the autopod appear as discrete entities separated by the interdigits (Dahn & Fallon 2000; Suzuki et al. 2008). Rather, interdigits have positional information for determining each digit identity through the PFR region that is located in the digital ray distal to the condensed cartilage of the digital ray (Suzuki et al. 2008).
Taken together, the recent observations discussed here imply that the function of SHH as a morphogen in mouse or chick limb development cannot be explained as currently proposed. The challenge is to integrate the genetic and molecular events in the limb bud with the genetic and molecular events in the autopod to produce a coherent understanding of the mechanisms governing digit determination and morphogenesis.
I would like to thank to Dr Atsushi Kuroiwa, Yo-ichi Shiraishi, and Dr John F. Fallon for helpful discussions. This work was supported by JSPS KAKENHI Grant Numbers 23127504 and 23770244.