Cell death is now recognized as a key factor in embryonic development. We can add it to the list of processes such as cell adhesion, cell migration, and cell proliferation whose complex coordination is required for proper patterning and differentiation of the embryo. In particular, a precise controlled balance between cell proliferation and apoptosis is required for normal development.
The vertebrate limb is among the best developmental examples illustrating how correct morphogenesis depends on programmed cell death, the paradigm being the separation of the digits. In amniotes, the digits primarily differentiate from the mesenchyme of the fan-shaped hand or footplate where prechondrogenic condensations are the first sign of the thickened digital rays separated by the flattened interdigital tissue. Individualization of the digits is then progressively achieved partly through the morphogenetic role of removal of the interdigital tissue by cell death (Saunders et al., 1962; Hinchliffe, 1982). However, sculpturing of the hand plate by removal of the interdigital tissue (as occurs in reptiles, birds, and mammals) is not the only way of making free digits, as exemplified by the amphibian forelimb, where the digits form by differential growth without participation of cell death (e.g., as demonstrated in Xenopus by Cameron and Fallon, 1977). In some cases such as the bird wing, cell death also plays a role in limb reconstruction and remodeling (e.g., removal of the ulnare cartilage from the wrist; Hinchliffe and Hecht, 1984) and specialization of the digits (e.g., digit 4; Feduccia et al., 2005). In amniotes, differential growth also participates in the separation of the digits (Salas-Vidal et al., 2001). Therefore, both cell death and cell proliferation are crucial processes during the morphogenesis of the digits.
During limb development programmed cell death is a predictable and dynamic process. It is probably controlled by many signals (e.g., Bmp, Shh, Fgf, Dkk; Zuzarte-Luis and Hurle, 2002, 2005; Bastida et al., 2004; Grotewold and Ruther, 2002) that result in distinct temporal and spatial areas in which cells die. The regions of cell death in the avian developing limb were extensively studied and mapped in classic studies, mostly in the chick wing (Saunders et al., 1962; Hinchliffe and Ede, 1973). Four main areas of cell death were identified: the anterior necrotic zone (ANZ) along the anterior border of the bud, the posterior necrotic zone (PNZ) along the posterior border, the opaque patch (OP; Fell and Canti, 1934), in the middle of the bud and the interdigital necrotic zones (INZ) in the interdigital spaces. Mammalian areas of cell death during limb development have also been studied, particularly in the mouse and in the rat (Milaire, 1971, 1977). In the mouse, the areas of cell death appear comparatively later than in the chick. Milaire described the “foyer préaxiale primaire” (fpp) as a group of dying cells located at the preaxial end of the apical ectodermal ridge (AER, the distal thickening of the limb bud ectoderm) at the beginning of the footplate stage (Milaire, 1971, 1977). Later, at early stages of digital outgrowth he described two other areas of cell death in the marginal subridge mesoderm, one preaxially and the other postaxially positioned at either end of the AER. Milaire called these areas the “foyer marginal I” and the “foyer marginal V,” respectively (Milaire, 1971, 1976, 1977). However, later he referred to these areas as the anterior and posterior marginal necrotic zones (AMNZ and PMNZ; Milaire, 1992).
It is important to note that the cells in all of these areas die by apoptosis as evidenced by the cytology of the dying cells (Hurle and Hinchliffe, 1978) and by the terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) assay (Zakeri et al., 1993). In addition, internucleosomal DNA fragmentation, representing molecular evidence for apoptosis, has been detected in DNA extracts from INZ of chick and mouse limb buds (Garcia-Martinez et al., 1993; Toné et al., 1994; Mori et al., 1995). Paradoxically, the areas of programmed cell death in the developing limb are named necrotic because they were described before the introduction of the term apoptosis (Kerr et al., 1972). Ideally, the zones should be renamed (ANZ becoming AAZ), but one obstacle to a change in their name to remove the term necrotic is the broad acceptance and use of the current terms within the field. However, if a consensus can be obtained by workers in this field, the present authors consider a change should be made.
Very interestingly, modifications in the pattern of cell death have been found to correlate with changes in the final definitive morphology of the limb. For example, across species, different patterns of cell death in the INZ correlate with different final morphologies of the interdigital region, and INZ inhibition correlates with survival of interdigit webbing (Saunders and Fallon, 1967; Hinchliffe and Ede, 1967, Hurle and Climent, 1987; Hurle et al., 1996). Within a single species, modifications in the amount of cell death by naturally occurring mutations have been shown to correlate with syndactyly or other modifications in digital pattern (Hinchliffe and Ede, 1967, 1973; Dvorak and Fallon, 1991; van der Hoeven et al., 1994; Zakeri et al., 1994; Aoto et al., 2002). In general, the areas of cell death are considered to play an important function in regulating the quantity of mesenchymal cells and, therefore, the number of digits (Hinchliffe, 1982).
Another crucial factor that controls correct size and shape in a developing structure is the spatiotemporal distribution of proliferative pools of cells (Garcia-Bellido et al., 1994). The pattern of cell proliferation is organ-specific and dynamic. During limb development, it has been claimed that the progress zone, a region of mesenchyme immediately subjacent to the apical ectodermal ridge (for review, see Tickle, 2003), maintains a high proliferation rate. However, studies of labeling index showed little difference along the proximodistal axis of the limb (Janners and Searls, 1970; Searls and Janners, 1971; Summerbell and Wolpert, 1972) until fairly late in development, after the formation of the proximal skeletal elements, and correlating with their differentiation.
Lately, much attention has been paid to the study of the molecular mechanisms that trigger and control the process of programmed cell death during limb development (Zuzarte-Luis and Hurle, 2002, 2005; Bastida et al., 2004; Grotewold and Ruther, 2002), however, accounts of the spatiotemporal distribution of the areas of cell death during limb development often appear confused between birds and mammals, and between fore- and hindlimbs. Equally, although much effort has been made to unravel the molecular basis of cell proliferation, few studies have been devoted to the analysis of the spatiotemporal patterns of proliferation during limb bud development.
The widespread use of genetic manipulation in mouse produces a great diversity of phenotypes that often include effects on the limb. Changes in the number of digits (polydactyly or oligodactyly) are frequent, as well as variable degrees of syndactyly. These malformations are frequently explained by modifications in the proliferation rates and/or cell death (Zakeri et al., 1994; Zou and Niswander, 1996; Sidow et al., 1997; Jiang et al., 1998; Sun et al., 2002; Grotewold and Rüther, 2002; Francis et al., 2005; Johnson et al., 2005; Lu et al., 2006). Therefore, this study was stimulated by the need for a comprehensive description of the spatial and temporal distribution of cell death and cell proliferation during limb development. We believe our work will be useful in establishing correlations between an observed phenotype and the type of developmental modification that may have caused it.
Cell Death and Proliferation in the Chick Wing Bud
The results of our study are shown in Figures 1 and 2. Each row corresponds to one Hamburger and Hamilton stage (HH; Hamburger and Hamilton, 1951) as indicated on the left. To provide a good visualization of the shape of the limb, for each stage we present a picture of the limb hybridized for Msx2. Although the correlation between Msx2 and cell death has been extensively suggested (Krabbenhoft and Fallon, 1992; Graham et al., 1994; Gañan et al., 1998), the choice of Msx2 was arbitrary, based solely on its peripheral pattern of expression that permits an excellent visualization of the shape of the limb.
Next to the picture of the limb hybridized for Msx2, two consecutive frontal and two consecutive transverse sections are shown. One of the two equivalent sections was subjected to the TUNEL assay and the other for immunohistochemistry with the anti-phosphorylated histone H3 (pH3) antibody. In most cases, the limb used for the in situ hybridization was the contralateral of the one used for the frontal sections. The whole limb was sectioned and analyzed. The sections shown in the figures are the most representative in showing the areas of cell death.
In the wing mesenchyme, cell death was imperceptible by the TUNEL assay up to the stage 21HH. Stages 17HH to 19HH were analyzed but are not shown because no cell death was detected. Cell death in the mesenchyme was also absent at stage 20HH (Fig. 1). At stage 21HH, a group of apoptotic cells was seen at the anterior border of the bud constituting the well-defined ANZ (Saunders et al., 1962; Fig. 1). At this stage, dying cells extended along the anterior border of the bud from the body wall to the distal tip, reaching the anterior end of the AER. This anterior area of cell death continued through stage 22 and 23, but remained proximal not extending beyond the anterior end of the AER (Fig. 1). The ANZ was at its maximum around stage 23HH but then declined at stage 24HH. The transverse sections showed that the apoptotic cells were located progressively deeper into the mesoderm during stages 21 to 23, and always restricted to the dorsoventral boundary. Later, from stage 25HH to stage 27HH, dying cells were again seen at the anterior border, at the junction between the zeugopod and the autopod (Fig. 2).
From stage 22HH, the TUNEL assay detected another group of dying cells located in the middle of the bud (Fig. 1, transversal section). This corresponded to the OP (Fell and Canti, 1934; Saunders et al., 1962, Hinchliffe, 1981). In the transverse sections, this group of dying cells localized deep in the central mesoderm, with a slightly ventral bias. The OP was detected up to stage 26HH (Figs. 1, 2). When it first started, its location was proximal at the junction between the limb bud and the body wall. Then it spread distally so that, by stage 24HH, it mainly occupied the prospective zeugopod, between prospective radius and ulna. By stage 27HH, only a few isolated dying cells remained in the OP (Fig. 2).
At stage 23HH, dying cells were detected at the posterior border of the bud: the first sign of the PNZ (Saunders et al., 1962; Hinchliffe, 1981). This was a rapidly developing area of cell death that peaked at stage 24HH when it occupied the space between the posterior marginal blood sinus and the posterior border (Fig. 1). Only some scattered cells were seen anterior to the marginal sinus. Cell death at the posterior border was observed up to early stage 26HH (Fig. 2 and Saunders et al., 1962).
By stage 28, cell death was again detected at the posterior border of the limb but, at this time, the shape of the developing limb has notably changed since a well-defined autopod had emerged. Cell death at this stage localized to the posterior margin of the autopod, probably in relation with sculpturing the posterior margin of the wing bud and digit 4 and persisted at least up to stage 32HH (Fig. 2). This area of cell death is not the PNZ.
Interdigital cell death, constituting the INZ (Saunders et al., 1962; Hinchliffe, 1981) began at stage 29–30HH in the proximal part of the second interdigital space, separating digit 3 from digit 4 (Fig. 2). A few hours later at stage 31–32HH, interdigital cell death was massive in the two interdigital spaces of the wing autopod as well as along the anterior margin of digit 2, the most anterior wing digit, and posterior margin of digit 4, the most posterior wing digit. Cell death at the anterior and posterior margins of the autopod is likely involved in shaping the anterior and posterior borders of the hand plate.
Concomitantly with the study of cell death, we have studied the distribution of mitosis within the limb bud. Figures 1 and 2 show pictures of sections consecutive or equivalent to those shown for the TUNEL assay processed for anti-pH3 immunohistochemistry. Up to stage 24HH, mitotic cells were found scattered in a uniform distribution across the limb bud, with the exception of the areas of cell death in which mitotic cells were scarce (Fig. 1).
From stage 24, mitotic cells were less abundant in the core (future chondrogenic) region of the bud, as observed both in frontal and transverse sections (Figs. 1, 2). Very interestingly, beginning at stage 27 the distal mesoderm shows a high number of mitotic cells (Fig. 2). Once the tips of the growing digits protruded from the hand plate, the distal high level of cell proliferation was restricted to the digital tips up to stage 31, while the interdigital zones lacked mitotic cells. Progressively, the chondrogenic skeletal elements showed a clear decrease in proliferation in each central region, corresponding to the region where chondrocyte differentiation occurs (Fig. 2). These observations were corroborated by bromodeoxyuridine (BrdU) labeling, using pulses of 30 min (not shown).
Of interest, the absence of mitotic cells coincided with the areas of programmed cell death when they were well defined such as the INZ. This finding was always observed when consecutive sections of the same specimen were analyzed. Therefore, we asked whether the decrease in proliferation could be detected before the establishment of the area of cell death. However, the distribution of mitotic cells was normal in the prospective areas of cell death one stage before their establishment, indicating that cessation of proliferation and programmed cell death occurred concomitantly or at least with less than one stage (approximately 12 hr of incubation) of difference.
Cell Death and Proliferation in the Developing Chick Leg
The results of our study are shown in Figures 3 and 4. Each row corresponds to one stage as indicated on the left, following the same arrangement as explained above for the wing.
In the leg mesenchyme, cell death is first detectable by the TUNEL assay at stage 22HH, when a cluster of apoptotic cells showed up at the anterior border of the bud (Fig. 3). This group corresponded to the well-defined ANZ of the chick leg that, as in the wing, spread out from the body wall to the anterior end of the AER, restricted to the dorsoventral boundary (Saunders et al., 1962; Hinchliffe, 1981). From stage 22HH, apoptotic cells were always seen at the anterior border of the developing leg bud, at least up to stage 28HH (Figs. 3, 4). As can be seen in Figures 3 and 4, the location of the ANZ in the leg bud became progressively displaced distally, so that, from stage 24HH, it occurred at the level of the prospective zeugopod and, from stage 26HH, at the level of the proximal autopod.
By stage 23, the OP was detected in the core of the leg bud (Fig. 3). It was visible up to stage 26HH, first extending along the prospective stylopod and zeugopod, and then, from stage 24HH, being restricted to the zeugopod, between prospective tibia and fibula.
Although there has been some debate about whether the PNZ occurs in the leg, we detected a group of dying cells at the posterior border of the leg bud at stage 23HH, very similar to the wing PNZ (Fig. 3). The leg PNZ occurs even more briefly than in the wing bud and became barely detectable by stage 24HH.
As in the wing, apoptotic cells were detected at the posterior margin of the autopod after the foot plate was differentiated, at stage 28 (Fig. 4). Properly interdigital cell death, the INZ, was first detected at stage 29–30 in the proximal part of the first, second, and third interdigital spaces, then in its distal part. Proximal and distal clusters of cell death in the interdigital spaces were first separated but later joined together, the whole interdigit exhibiting cell death (Pautou, 1974; Saunders and Fallon, 1967). The INZ reached its maximum at stages 31HH (day 7 and a half of incubation) and subsequently progressively declined (Fig. 4).
Analysis of proliferation was performed in consecutive or comparable sections to those used for the TUNEL assay. The immunohistochemistry with the anti-pH3 antibody initially showed scattered labeled cells uniformly distributed all across the whole of the limb buds. As occurred in the wing bud, fewer mitotic cells were seen in the core of the bud from stage 24HH (Fig. 3). Also, the absence of mitosis from the areas of cell death as well as from the chondrogenic condensations, as they mature, was significant. And also, as described for the wing, a high level of mitotic cells was observed in the distal mesoderm from stage 27HH and then localized in the growing tips of the digits as the intervening interdigital spaces became devoid of mitotic cells (Fig. 4).
Staging of Mouse Embryos
While the Hamburger and Hamilton stages of development have been adopted universally for the chick embryo, the method most widely used for staging the mouse embryo is gestational age. Several tables of development have been published for the mouse embryo (Theiler, 1989) and even specifically for the limb (Wanek et al., 1989; Martin, 1990) but have not been generally adopted by investigators. Staging the embryo by gestational age has the disadvantage that embryos within a single litter may vary considerably in their developmental stage. This variability is a considerable difficulty for our study. Therefore, to make our study thoroughly available to investigators outside the limb field but at the same time to avoid the inherent variation in the development in the same gestational age, we decided to use gestational age but restricted by some hallmarks that best define the forelimb and hindlimb at each developmental stage. Therefore, we only used the embryos that met these characteristics within each litter. These characteristics are shown in Table 1. They are based on previously published studies (Theiler, 1989; Wanek et al., 1989; Martin, 1990), the Edinburgh mouse atlas project (emap) and our own experience, and are exhibited by the majority of the embryos of the corresponding gestational age. The characteristics are the same for both forelimb and hindlimb but with the development of the hindlimb being delayed compared with that of the forelimb.
Table 1. Mouse Embryonic Stages With Corresponding Morphological Characteristic of Fore- and Hindlimba
HH, Hamburger and Hamilton stage.
The forelimb protrudes opposite somites 7–12
Hindlimb becomes distinct at the end of this stage
Forelimb bud enlarged and symmetrical Corresponding to chick wing bud stage 20HH
The hindlimb is opposite somites 23–28
Forelimb bud enlarges, still quite symmetrical Corresponding to chick wing bud stage 22HH
Hindlimb bud enlarged and symmetrical Corresponding to chick leg bud stage 20HH
Forelimb bud elongates Corresponding to chick wing bud stage 23-24HH
Hindlimb bud enlarges Corresponding to chick leg bud stage 21-22HH
Two regions become clearly distinct: the narrower proximal part and the circular distal part (handplate) with a sharp boundary between them
Hindlimb bud elongates Corresponding to chick leg bud stage 23-24HH
The circular contour of the handplate becomes polygonal due to the angles forming by the growing digital tips
Two regions become clearly distinct: the narrower proximal part and the circular distal part (footplate) with a sharp boundary between them
Digital condensations and interdigital spaces appear distinct in the handplate.The anterior digit not defined yet
The circular contour of the footplate becomes polygonal due to the angles forming by the growing digital tips
Distal contour of the interdigital spaces show a clear indentation.The five digits clearly defined.
Digital condensations and interdigital spaces appear distinct in the footplate. The anterior digit not defined yet
Fingers separate distally
Distal contour of the interdigital spaces show a clear indentation. The five digits clearly defined
Distal separation between the digits increased.Reduced and thinner interdigital spaces
Fingers separate distally
The digits are almost completely separated. Anterior and posterior marginal zones no longer visible
Distal separation between the digits enlarged. Reduced and thinner interdigital spaces
Digits completely separated and radially divergent. Digits slightly bent ventrally
The digits are almost completely separated. Digits slightly bent ventrally
Digits continue completely separated. The four posterior digits arrange in parallel. Digital tips sharpened
Digits completely separated and radially divergent. Anterior and posterior marginal zones no longer visible
Digits lie in parallel and begin to fuse
Digits continue completely separated. The four posterior digits arrange in parallel
Digits parallel and fusion proceeds
Digits lie in parallel and begin to fuse
Cell Death and Proliferation in the Developing Mouse Forelimb
In the mouse forelimb, cell death in the mesenchyme was not detectable by the TUNEL assay up to the embryonic stage 10.5 dpc (E10.5) when a group of dying cells was observed in the core of the bud, at the proximal level (Fig. 5). This area of cell death, although not as well defined, was considered equivalent to the OP of the chick limb and persisted up to E12 (Fig. 5). In some sections (see for example the frontal section of E11.5 in Fig. 5), the OP appeared to enter into the proximal autopod, something never observed in the developing chick limb. To clarify this point, we performed in situ hybridization for Sox9 in equivalent sections (Fig. 9A,B). Sox9, the earliest known marker of cells committed to the chondrogenic lineage (Bi et al., 1999; Akiyama et al., 2002), was used to visualize the cartilage condensations to localize the apoptosis in relation to the skeletal elements. The comparison of the pattern of TUNEL labeling and that of Sox9 expression showed that the OP in the mouse was restricted to the space between the primordia of radius and ulna, a pattern analogous to that in the chick embryo.
At E11, shortly after the OP was visible, another group of apoptotic cells was observed within the anterior mesenchyme. The dying cells in this area were somewhat scattered in distribution and formed an ill-defined group. The transverse sections revealed a central position in the middle of the bud, just under the anterior AER at the dorsal–ventral boundary (Fig. 5). With the differentiation of the digits in the E12.5 hand plate, the location of this area of cell death was seen to correspond to prospective digit 1 (Fig. 6). This area of cell death was described by Milaire who called it the “foyer préaxiale primaire” (fpp, Milaire, 1971, 1976). Because of its anterior location, the fpp can be compared with the ANZ of the chick limb bud; however, in the mouse it appeared comparatively later, and in a more distal position. Because the spatial and temporal differences between the fpp and the ANZ are marked, we believe that the mouse fpp should not be termed the mouse ANZ.
At stage E12.5, just when the hand plate lost its circular contour due to the protrusion of the growing digital tips, dying cells were observed in the marginal subridge mesoderm, preferentially at the anterior and posterior margins of the autopod (Fig. 6). These dying cells formed a thin band rimming the periphery of the hand plate, just under the ridge, extending approximately 100–150 microns deep into the mesoderm. The band was clearly seen anteriorly and posteriorly and, eventually, by E13, became a continuous from the anterior to the posterior extremes of the autopod (Figs. 6, 9C). This cell death was described by Milaire as two areas he called the AMNZ and the PMNZ (Milaire, 1992). Our observations indicated that, over time, these two marginal areas become a single continuous zone of cell death, we call the Sub AER Marginal Zone (SAMZ). It is interesting to note that the density of dying cells in this area was low (Fig. 9C,D) and that there is no comparable area of cell death in the chick limb bud.
Between E13 and E13.5, the SAMZ became restricted to the interdigital spaces and vanished from the digital tips (Fig. 6). The remnants of the SAMZ in the interdigits formed the beginning of the interdigital cell death areas (INZ) that grew rapidly between E13 and E13.5 as extensions of the SAMZs toward the proximal apex of the interdigit. By E13.5, between all the digits large-scale cell death had filled most of the interdigital space (Fig. 6). As the interdigital spaces regressed, cell death continued at the lateral margins of the digits as well as in the anterior margin of digit 1 and posterior margin of digit 5. The cell death at the anterior–proximal and posterior–proximal border of the autopod was conspicuous in the mouse hand plate. By 14.5, only a residual interdigital space remained at the most proximal level, and by E15.5, the digits appeared completely separated. Interdigital cell death, therefore, occurred between E13.5 and E15.5 (Fig. 6).
To study cell proliferation, we processed sections consecutive to those used for the TUNEL assay for the anti-pH3 immunohistochemistry. In the figures, each TUNEL picture is accompanied by its parallel anti-pH3 staining. Mitotic cells were found uniformly distributed across the whole limb bud up to E11 when the central core of the limb started to show a decrease in the number of mitotic cells (Fig. 5). As described for the chick, mitotic cells were seldom found in areas of programmed cell death, especially in the INZs (Fig. 6). The exclusion of mitotic cells from well-defined areas of cell death was clear, but it proved difficult to assess in areas in which the density of dying cells was low, as in the SAMZ. Finally, also comparable to the chick limb bud, the chondrogenic elements showed a diminished proliferation rate clearly detected from E12.5 (Fig. 6). Later, the developing digit tips continued to show a relatively high proliferation rate (Fig. 6).
Cell Death and Proliferation in the Developing Mouse Hindlimb
Cell death was first recognized by the TUNEL assay at E11 when just a few scattered dying cells were seen in the core of the bud, forming in the same position as in the chick OP but much less well defined (Figs. 7, 8). This area lasted up to E12 but was always less prominent than its corresponding OP in the forelimb. As for the forelimb, its locations corresponded to the zeugopod in the space between the two forming skeletal elements.
At E11.5, apoptotic cells were detected in the anterior mesoderm where they persisted through to E13 forming a well-defined cluster that overlapped the prospective digit 1 and corresponded with Milaire's fpp. This area was denser and much better defined than its corresponding forelimb fpp. The transverse sections showed how the dying cells were located in a central position in the dorso–ventral axis, immediately under the anterior AER (Figs. 7, 8).
At E13.5, apoptotic cells were detected at the periphery of the autopod, in a band similar to that we have named the SAMZ in the forelimb (Fig. 8). From E14, cell death ceased at the digital tips but remained at the distal interdigital spaces from where the cell death spread proximally from the SAMZ into the whole interdigit following a course very similar to that described for the forelimb. In the mouse hindlimb, the INZ runs from E14 through E15.5 (Fig. 8). The study of proliferation in the mouse hindlimb gave comparable results to those described for the forelimb, but with the normal delay corresponding to the delayed development of the hindlimb in relation to the forelimb.
Perinatal Digit Morphogenesis in the Mouse
As we have just described, during fetal development, the digits of the mouse fore- and hindlimbs become progressively separated due to interdigital apoptosis as well as differential growth of the digital tips (Milaire, 1977; Salas-Vidal et al., 2001). By E15.5, the digits appeared completely separated and showed a radial disposition as a consequence of the circular shape of the digital plate within which they differentiated. However, during the next day (E16.5), the digits became arranged in parallel and joined together due to the fusion of the maturing epidermis, a process that has been described in detail by Maconnachie (1979). The limb ectoderm consists of a basal cuboidal layer covered by a superficial squamous layer called the periderm. From E12.5, the basal layer starts to stratify to make three to four layers that form the stratum basale that remains covered by the monolayer periderm, now called the stratum superficiale (Sengel, 1976). Transverse sections through E17 fetal embryos (Fig. 10F) and newborn fore- and hindlimbs (not shown) demonstrated that the tissue joining the digits was exclusively the maturing epidermis that, during late fetal development, has adopted the stratified squamous morphology typical of mature skin.
Curiously, at birth (postnatal day 0, P0), all the mouse digits are joined by the epidermis up to the distal terminal phalanx as shown in Figure 10A, giving the appearance of a soft tissue syndactyly. This finding is important to note because it might mislead researchers, particularly if analyzing malformed or mutant limbs. The digits will progressively become free again during the first week of postnatal development through a process of peeling off of the interdigital epidermis and without participation of cell death (Fig. 10). Regression of the connecting epidermis commenced during the third day of postnatal life (Fig. 10B, P3) and progressed up to the end of the first week when only some remnants continued at the basis of the digits (Fig. 10C–E). We have observed the same process in the rat (not shown). However, the digits of chick and quail embryos remain separated once they become individualized in the fetal period. As is well known, humans also are born with completely separated digits. Therefore, the process of epidermal fusion of digits in the late fetal period appears to be a particularity of at least some murine species (Maconnachie, 1979).
Cell Death and Proliferation in the Limb Ectoderm
The study of cell death and proliferation in frontal and transversal sections permitted a precise evaluation of the presence of apoptotic or mitotic cells in the ectoderm. We consistently have observed cell death in the AER of the four types of limbs analyzed in the present study, as was reported by Jurand (1965) and Todt and Fallon (1984, 1986). Apoptotic cells were observed in transverse sections through the AER and in frontal sections that passed through a sufficient length of the AER (see for example 21HH in Fig. 3 and E10 in Fig. 5).
Cell death occurred in the AER throughout the period studied, in chick and mouse, from the earlier stages of AER formation before any detectable cell death appeared in the mesoderm up to the flattening of the AER and functional decay of its activity. The distribution of apoptotic cells in the AER was uniform along its anteroposterior extension, although an anterior bias can be observed in the early wing bud stages. Figure 9E–G shows a stage 25HH leg bud stained with the vital dye Neutral red in ovo, demonstrating cell death all along the extension of the AER as well as in the ANZ. In the mouse, we did not observe an anterior bias in the distribution of apoptotic cells within the AER at any stage analyzed.
Apoptotic cells were never found in the dorsal or ventral non-AER ectoderm as assessed in the transverse sections through all stages and types of limbs analyzed in this report. Although ectodermal cell death appeared restricted to the AER, mitotic cells were observed evenly scattered throughout the whole limb ectoderm, including the AER and the dorsal and ventral ectoderm. In the AER, mitotic and apoptotic cells occurred concomitantly.
Cell Death in Joint Formation
Formation of the joints involves several steps, including, first, the condensation and establishment of a three-layered mesenchyme and, second, vascular invasion and cavitation accompanied by cell death (described in the rat and chick by Mitrovic, 1971, 1977; Archer et al., 2003). Recently, the possibility of a role of cell death in joint formation has been questioned (Ito and Kida, 2000; Archer et al., 2003).
Cell death associated with the differentiation of the digital joints was consistently observed in the developing mouse limb bud autopod. In the forelimb, transverse bands of dying cells were clearly seen in the frontal sections at the level of the prospective interphalangeal joints at E13.5 and continued up to E14.5 (Fig. 6). In the hindlimb, the line of dying cells in the interphalangeal joints was first observed at E14.5 and continued up to E15.5 (Fig. 8). The metacarpophalangeal joints were also marked by cell death (arrow in the frontal section of the E15.5 forelimb in Fig. 6).
In this study, we report in atlas format the areas of cell death and patterns of cell proliferation during an extensive period of development of the mouse and chick fore- and hindlimbs. Our study confirms and expands previous studies. It has the advantage of using modern techniques, presenting actual photographs, and gathering all data in a single report.
In the mouse and chick, the areas of cell death are predictable (programmed) and dynamic; these are diagrammatically summarized in Figure 11. Such zones of cell death are frequently termed “physiological,” because they occur during normal unperturbed development (Lockshin and Zakeri-Milovanovic, 1984). Once cell death begins in a particular area, it usually progresses through an increase in the number of dying cells as well as the phagocytes engaged in the clearance of the dying cells. From its peak, cell death declines and completely disappears once the cell debris has been removed by resident neighboring cells (phagocytes) and/or by the invading immunocompetent macrophages (Hurle and Hinchliffe, 1978; Cuadros et al., 1993; Wood et al., 2000). In addition, at least in the interdigital spaces there is evidence of detachment of rods of macrophages loaded with cell debris and peridermal cells into the amniotic cavity (Hurle and Fernandez-Teran, 1983, 1984).
The analysis of cell proliferation shows that mitotic cells are distributed uniformly within the early limb bud without any obvious zone of increased proliferation under the distal tip as claimed for the “progress zone” (Summerbell and Wolpert, 1972). Of interest, however, at later stages, mitosis remains comparatively high first in the distal digital plate and then in the undifferentiated mesenchyme distal to the tip of the digits where the phalanges are forming, while it decreases proximally, in particular in the cartilaginous skeletal rudiments. Areas of cell death show proliferation at a very low level, but this finding is markedly reduced in comparison with adjacent tissues.
Patterns of Cell Death, Differences, and Similarities Between Chick and Mouse
In the early chick limb bud, two areas of programmed or physiological cell death are prominent: the ANZ and the OP (summary diagram in Fig. 11; see also diagram in Hinchliffe, 1982, Fig. 1.1 and micrographs in Zuzarte-Luis and Hurle, 2002; Fig. 1). Another area, the PNZ, occurs at the posterior border (stage 24 and 25HH). It has been reported that the chick leg bud, in contrast to the wing, possesses no PNZ (Saunders and Fallon, 1967; Brewton and MacCabe, 1988). However, our study demonstrates the presence of cell death in the posterior border of the leg bud in a pattern very similar to that in the wing bud, but it is small scale and transient (mostly occurring during stage 23HH). Generally, other avian species apart from the duck (Pautou, 1974) have been studied little, but it appears their cell death patterns are rather similar to the chick, with the exception of absence or reduction of the wing PNZ.
Studies of early mammalian limb buds are mainly restricted to mouse and rat with rather similar cell death patterns (Milaire, 1977). Our report accounts for the mouse where areas of cell death are also seen at the anterior border, the fpp, and in the middle of the bud, the OP. The mouse fpp show marked temporal and spatial differences with the chick ANZ. In the mouse, it is first detected just before the appearance of the digital plate and overlapping the early rudiment of digit 1 in both fore- and hindlimbs, as described by Milaire (1963) who termed this area the fpp, rather than the ANZ. In the chick, the ANZ is the first area of mesodermal cell death detected, from stage 21 in wing and stage 22 in leg and, despite a decline at stages 24 and 25, it persists at the junction between the zeugopod and autopod up to stage 27 (wing) and 28 (leg). Thus, there are sufficient differences to raise doubt as to whether in fact the fpp is equivalent to the ANZ. Therefore, we propose that the term fpp should be used to refer to the anterior distal area of cell death in the mouse limb buds and thus avoid suggesting correspondence with the chick ANZ that is much earlier and proximal. Of interest, cell death was not detectable in the posterior border mesoderm of the mouse fore- or hindlimb, indicating that the PNZ may be specific to the chick limb bud, particularly the wing.
Another notable difference between areas of cell death in mouse and chick is the peripheral band of dying cells in the autopod, named here the SAMZ (sub-AER marginal zone). This area is only detected in the mammalian limb bud and occurs in a period in which the AER is ceasing expression of Fgf8 (Sun et al., 2002). It has been suggested that the SAMZ depends on bone morphogenetic protein (BMP) signaling from the AER (Wang et al., 2004) because overexpression of Noggin (a potent BMP antagonist) in the AER cells results in the proximal displacement of the band of dying cells. The INZ are highly conserved in the two species, as in all amniotes, and will be discussed below.
In summary, two areas are highly comparable between birds and mammals: the OP and the INZ. Another area, the ANZ in chick and the fpp in mouse, markedly differ in their precise location and the time course varies with the species and the type of limb (see diagram in Fig. 11). Finally, the PNZ is specific to the chick, and the SAMZ is specific to mammals.
Morphogenetic Significance of Areas of Cell Death
The morphogenetic significance of these areas of cell death is difficult to infer. The ANZ and PNZ of the chick limb bud have been correlated with the evolutionary reduction in the number of digits typical of the normal pentadactyl limbs. Birds lose digit 1 and 5 in the wing (Feduccia et al., 2005) and severely reduce digit 5 in the leg. However, an alternative view proposes that the remaining digits in the bird wing represent digits 1, 2, and 3 by virtue of an anterior shift in the genetic developmental pathway that directs digit identity (Wagner and Gauthier, 1999; Vargas and Fallon, 2005). In either case, a correlation between the reduction in digit number and the ANZ and PNZ is supported by the absence of these areas in the polydactylous avian mutants talpid2 and talpid3 (Hinchliffe and Ede, 1967; Dvorak and Fallon, 1991). In accord with this interpretation, the mouse limb bud lacks the PNZ and has a small late fpp, which correlates with its broad digital plate from which five digits arise (Hinchliffe and Ede, 1973). Likewise, the fpp of the mouse limb is absent in several polydactylous mutants (Aoto et al., 2002). Therefore, it seems reasonable to assume that the ANZ and the fpp have a similar function in reducing the pool of cells and, therefore, the number of digits. Its anterior location could also contribute to the reduced proximodistal growth of the most anterior digit, which in mammals is the only one that forms two phalanges instead of the normal three typical of the rest of the digits.
Experimental modification of patterns of cell death in the fpp, using teratological agents, strongly suggest that it occurs normally under the influence of the local cell environment (Scott, 1981; Klein et al., 1981; Nakamura et al., 2000). It has been suggested that the ANZ could expand after a drop in the level of the morphogen diffusing from the zone of polarizing activity (Hinchliffe, 1981). Indeed the anterior mesoderm in chick requires the posterior mesoderm to survive and dies when it is deprived of Shh signaling (Bastida et al., 2004).
The OP is another area that is common to both species and types of limbs. It has been described in the region of the prospective elbow and knee joint and has been suggested as contributing to the morphogenesis of this joint (Fell and Canti, 1934). It has also been suggested that it may participate in the separation of the two cartilage elements of the zeugopod (Dawd and Hinchliffe, 1971), a notion that is supported by our study for the mouse limb bud.
Milaire (1976) suggested that the SAMZ may contribute to a reduction of the proximodistal length of the digital rays I and V. This contribution indeed could be the case because it starts earlier at the anterior and posterior borders and, therefore, has less effect in the middle of the bud. Although cell death is a critical factor in normal limb development and there are clear correlations of altered cell death patterns with changes to limb morphogenesis, it is still not clear how such cell death is controlled in molecular terms and whether it is a secondary effect of the genetic pathways that drive limb development.
Cell Death and Cell Proliferation in the Ectoderm
Apoptotic cell death is consistently seen in the AER, both in mouse and in chick, where it is observable from the earlier stages of limb budding, before any detectable cell death has appeared in the mesoderm (see for example, Figs. 3, 5, and Todt and Fallon, 1984, 1986). It is of interest to note that, although cell death is clearly noticeable in the AER, it is very seldom seen in the rest of the ectoderm.
The distribution of cell death along the anteroposterior axis of the AER has attracted much interest. Milaire and Rooze (1983) indicated that it was more prominent in the anterior part of the AER extending up to the middle, whereas posterior cell death was much more confined. However, others have reported no differences in physiological cell death distribution between the preaxial and postaxial sides of the AER at E14 and E15 rat limb buds (Antalikova et al., 1989), and in our study, we have not noticed a clear consistency in the anterior predominance of cell death in the AER, except for the early chick wing bud.
The morphogenetic relevance of cell death in the AER is presently unknown. Milaire and Rooze (1983) suggested that the areas of cell death in the AER correlated with the less active portions, because in several stages, it is more prominent at its anterior and posterior ends than in the central portion, supposed to be the most active part of the AER. They also described modification, including absence, retarded or excessive cell death, in the AER of different mutations in mouse or rat (Milaire and Rooze, 1983).
As mentioned earlier, polydactylous limb buds displayed a substantially lower physiological cell death than normal, not only in the mesoderm but also in the AER, which has been correlated with prolonged inductive effect possibly implicated in the polydactyly (Hinchliffe and Ede, 1967; Dvorak and Fallon, 1991; Bose et al., 2002; Francis et al., 2005). Intensive cell death in the AER has been reported to be caused by exogenous retinoic acid administration and, therefore, to be implicated in the teratogenic effects of retinoic acid treatment (Sulik and Dehart, 1988; Ferretti and Geraudie, 1995). Also, aberrant cell death of the AER is considered the cause of the ectrodactyly in the Dactylaplasia mutation in the mouse (Seto et al., 1997). Increase in cell death in the AER has also been correlated to the lack of limbs in the serpentine reptiles (Raynaud, 1990), although most recently, the lack of limbs in snakes has been correlated with specific changes in the expression domains of Hox genes (Cohn et al., 1997; Cohn and Tickle, 1999).
Interdigital Cell Death
The INZ is the best-studied area of cell death during limb development and the one that best illustrates the modeling function of programmed cell death (Saunders et al., 1962; Milaire, 1976; Hinchliffe, 1981). The INZ in the leg autopod of the chick embryo has been shown to develop in a precise pattern (Pautou, 1974, describes 14 distinct stages for both chick and duck). The INZ is also found in humans (Menkes et al., 1965) and has also been described in several other amniotes such as rats (Ballard and Holt; 1968), ducks (Pautou, 1974), and reptiles (lizards, Goel and Mathur, 1978; turtles, Fallon and Cameron, 1977).
However, it is important to note that cell death is not always an integral part of the mechanism of free digit formation, because in amphibians, it is absent (or nearly so) from the forelimbs of Xenopus laevis, Bufo americanus, Ambystoma mexicanum, Ambystoma maculatum, and Taricha torosa (Cameron and Fallon, 1977). An exception is provided by the presence of limited cell death positioned in the interdigits of limb buds of the urodele amphibian the Seepage salamander; but here it appears to have a relatively minor role: it flattens the interdigit web but does not free the digits (Franssen et al., 2005). Also, in the amniote species with conspicuous interdigital cell death it is detected well after the digital primordium protrudes from the hand plate, transforming the circular contour to a polygonal one, confirming that the initial separation between the digits is due to differential growth (Salas-Vidal et al., 2001). Thus, the process of freeing the digits in the autopodial plate results from the concomitant participation of cell death and differential growth. We observed this correlation but whether these two processes are linked remains to be seen.
However, the INZ has a clear role in removing interdigital tissue, and when this removal is (partially) inhibited, either naturally as in duck (Pautou, 1974) and water hen (Hurle and Climent, 1987) or in spontaneous or induced mutants (i.e., Hinchliffe and Ede, 1973; Hinchliffe and Thorogood, 1974; Zou and Niswander, 1996; Sidow et al., 1997; Jiang et al., 1998; Francis et al., 2005; Lu et al., 2006), then interdigital webs survive into the definitive limb. Particularly interesting regarding mechanisms is the study by Lu et al. (2006) showing that increased Fgf signaling from the mouse AER leads to absence of interdigital cell death without modification of the normal pattern of expression of BMPs.
Curiously, in the mouse and rat, the separation of the digits is followed by their posterior union by a fused layer of epidermal cells in a process similar to the fusion of the eyelids (Maconnachie, 1979). At birth, the digits appear parallel joined through their length, a feature that is not shared by most other mammals. It is unknown whether this secondary digital fusion has any functional meaning. We have shown here that the digits would definitively separate during the first week of development through a process of peeling off of the fused epidermis. Because the mouse embryo is a major model for analysis of limb development, this postnatal morphogenesis of the digits needs to be considered when evaluating the morphology of the mouse neonate autopod.
Cell Proliferation in Limb Development
Here, we have evaluated cell proliferation activity by determining the mitotic cells by anti-pH3 immunohistochemistry. The pattern of mitosis obtained correlates well with the distribution of BrdU-incorporating cells during normal development and has been proved as a good method to evaluate cell proliferation.
The limb bud mesenchyme is initially a highly proliferative tissue. Classic studies indicated that the cell generation time was approximately 10 hr during wing development and that it remained reasonably constant. Janners and Searls (1970) described a decrease in the proliferative index (the percentage of cells in the division cycle) throughout wing development. In proximal regions of the wing, the proliferative index was 100% until stage 21, decreased from stage 21 to stage 23, and then remained constant at 25% in the chondrogenic and 75% in the myogenic regions. Our observations fit well with these results. We have detected a uniform distribution of mitoses all across the limb bud in both chick and mouse. Then the number of mitoses decreases in the core of the limb where cells condense in a first step for cartilage differentiation. The decrease in proliferation is seen concomitantly or even before the initiation of cartilage differentiation (in the zeugopod) that occurs at 24HH in the chick wing (Dessau et al., 1980). Using either BrdU staining or proliferating cell nuclear antigen (PCNA) mRNA expression, low proliferative activity has been shown to occur in the center of the bud and higher activity in the limb margins (Kohler et al., 2005).
At later stages, the cells at the distal tip of the limb maintain their high initial mitotic rate. This finding is detected just before the differentiation of individual digits and is maintained during the period of rapid digital growth and specification of the phalangeal elements. This distal band of increased mitosis at the distal tip is reminiscent of the “progress zone.”
Here, we have observed consistently that the proliferation is absent from the areas of cell death, particularly well seen when the area is ample, such as the INZ. This exclusion is more difficult to evaluate in smaller areas or areas in which the density of dying cells is low, as the SAMZ. This finding may indicate the presence of local cues that control both processes. The relationship between cell cycle and apoptosis has been studied extensively in the past years. However, only one study has examined this relationship using a developmental system such as the limb bud (Toné and Tanaka, 1997). They found that approximately 20 to 24 hr before apoptosis, cells destined to die passed through the last S-phase. Their study of the DNA content allowed them to conclude that the cells withdraw from the cell cycle at G2-phase. It would appear that commitment to die is accompanied by withdrawal from the cell cycle.
Held and Saunders (1965), studying the PNZ of the chick wing bud, noticed that cells ingested by phagocytes had not incorporated tritiated thymidine administered after stage 21. This finding was taken as an indication that cells first stop growth and then die. Pollak and Fallon (1976) showed that the decrease in synthesis of DNA, as assessed by autoradiographic analysis, was significant by stage 22HH, two stages (approximately 20 hr) before overt cell death. It is a fact that the cells destined to die do not differ by any cytological criteria from their nonmoribund neighbors until obvious apoptosis has begun. However, these cells have stopped proliferating. Finally, it is worth noticing that the experimental perturbation of cell cycle kinetics in limb development can alter gene expression and pattern formation (Ohsugi et al., 1997).
In summary, our data demonstrate that there are spatiotemporal differences in the pattern of proliferation of mesenchymal cells in limb buds. In the early limb bud, the rate of cell proliferation was high and relatively uniform. Subsequently, more mitotic cells were observed in the periphery than in the center of the bud, both in chick and in mouse. At later stages, mitosis remained high in the distal tip, coincident with the time the phalanges of the digits are being laid down. Proliferating cells are practically absent from the areas of cell death.
In the chick, the stages we have analyzed ranged from 17HH (2.5 days of incubation) to 32HH (7.5 days of incubation). Therefore, our study spanned the period of limb morphogenesis from initial limb budding to the beginning of separation of the digits. In the mouse, our analysis started at E9.5, coincident with the initial development of the forelimb, to postnatal day 6 (P6). In the chick, we analyzed every Hamburger and Hamilton stage within this range, and in the mouse, the analysis was performed at half-day intervals.
Chick and Mouse Embryos
Normal hen eggs were obtained from local sources, mostly from Granja Santa Isabel (Córdoba, Spain). The eggs were incubated and opened as described (Ros et al., 2000). Chick embryos were staged in ovo according to the Hamburger and Hamilton tables of chick development (Hamburger and Hamilton, 1951).
The mouse embryos used in this study were of the strains CD1 and C57BL/6/Jico, originally obtained from CRIFFA (Barcelona, Spain) and bred locally at the Animal House in the University of Cantabria. Noon of the day the vaginal plug was observed was designated as embryonic day 0.5. The embryos were obtained by caesarean section from pregnant mice and separated from the placenta in cold phosphate buffered saline (PBS). Within a single litter, there was a considerable variation in the developmental stage of the embryos, and, therefore, the embryos were staged according to the external characteristics of the limbs defined for each developmental stage as explained in the text (Table 1).
Mouse and chick embryos were fixed overnight in 4% paraformaldehyde and thoroughly washed in PBS, and the limb buds were dissected out and photographed. For each embryo, the right limbs were processed for paraffin wax embedding and sectioning, while the left limbs were processed for whole-mount in situ hybridization or Neutral Red staining.
The whole limb was serially sectioned at 7 μm, and the sections sequentially were placed one by one on three different slides. Therefore, a set of three slides of adjacent sections was obtained for each limb sectioned. One of these slides was used for the TUNEL assay and another for the proliferation assay. The third slide was left as a reserve in case a procedure failed. The limbs were sectioned in the three orthogonal planes, which in relation to the axis of the limb are called frontal, transverse, and longitudinal. At least three specimens were analyzed for each stage.
Cell Death Analysis
To detect programmed cell death by apoptosis, we used the terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) assay, which has been validated for this purpose (Zakeri et al., 1993, 1994). The TUNEL assay was performed using the Apoptag Fluorescein Direct In Situ Apoptosis Detection Kit (Intergen) and following the manufacturer's instructions. Our study of cell death was completed in selected stages with the use of the vital dye Neutral Red, according to the procedure of Hinchliffe and Griffiths (1986), to demonstrate in whole limb buds the areas of cell death by this alternative method.
Several methods have been used for the study of proliferation patterns such as in vivo labeling with BrdU or [3H]thymidine to measure DNA synthesis, immunohistochemistry for the PCNA, or even direct counting of mitotic cells in appropriately stained sections. Recently, use has been made of the anti-pH3 antibody that recognizes histone H3 after it becomes phosphorylated on serine 10, when the chromosomes condense during prophase. The phosphorylation remains until telophase when the histone H3 is dephosporylated by specific phosphatases. Therefore, the anti-pH3 antibody recognizes chromosomes during mitosis and provides a valuable method for identifying mitotic cells and was the agent we selected for our study.
The proliferation was assayed by the anti-pH3 antibody. For immunohistochemistry Superfrost/Plus slides (Fisher Scientific, Rockville, MD) were used to prevent detaching of the sections from the slides in the process of antigen retrieval. The slides were deparaffinized and rehydrated routinely. Antigen retrieval was performed in R-buffer A (10× Pickcell laboratories) in the 2100 Retriever for at least 2 hr and 20 min, as recommended by the manufacturer. Sections were incubated overnight with primary antibody (rabbit polyclonal Phospho H3 from Upstate Biotechnology, Lake Placid, NY) diluted at 1/100. After washing and blocking, sections were incubated for 1 hr in the biotinylated secondary antibody. After washing, the slides were incubated in ABC solution (Vector Laboratories, Burlingame, CA) and developed in diaminobenzidine.
Our study was completed with analysis of BrdU incorporation in selected stages to corroborate our results following standard procedures (not shown). We did not select the incorporation of BrdU to the limb cells as a method to evaluate cell proliferation, because it has been shown to suppress interdigital cell death, and we wanted to study both processes, cell death and proliferation, in the same limb (Toné et al., 1983).
In Situ Hybridization
Whole-mount in situ hybridization and hybridization to tissue sections were performed according to standard protocols.
We acknowledge the excellent technical assistance of Marisa Junco.