MMP-2 plays an essential role in producing epithelial-mesenchymal transformations in the avian embryo



To investigate the roles that matrix-degrading proteases may have in development of the chicken embryo, we documented the expression pattern of matrix metalloprotease-2 (MMP-2, 72-kDa type IV collagenase or gelatinase A) and perturbed its function in vitro and in vivo. MMP-2 is expressed as neural crest cells detach from the neural epithelium during an epithelial-mesenchymal transformation (EMT) but is rapidly extinguished as they disperse. It is also expressed in the sclerotome and in the dermis at the time that the EMT is initiated, and also as these cells migrate, and is down-regulated once motility has ceased. These patterns suggest that MMP-2 plays a role in cell motility during the EMT and during later morphogenesis. Inhibitors of MMPs, including BB-94 and TIMP-2 (tissue inhibitor of metalloprotease-2), prevent the EMT that generates neural crest cells, both in tissue culture and in vivo, but do not affect migration of the cells that have already detached from the neural tube. Similarly, knockdown of MMP-2 expression in the dorsal neural tube using antisense morpholino oligos perturbs the EMT, but also does not affect migration of neural crest cells after they have detached from the neural tube. On the other hand, when somites in culture are treated with TIMP-2, some mesenchymal cells are produced, suggesting that they undergo the EMT, but show greatly reduced migration through the collagen gel. MMP-2 is also expressed in mesenchyme where tissue remodeling is in progress, such as in the developing feather germs, in the head mesenchyme, in the lateral plate mesoderm, and in the limb dermis, especially in the regions where tendons are developing. Comparisons of these expression patterns in multiple embryonic tissues suggest a probable role for MMP-2 in the migration phase of the EMT, in addition to mesenchyme dispersion and tissue remodeling. Developmental Dynamics 229:42–53, 2004. © 2003 Wiley-Liss, Inc.


Proteases have long been recognized to play an important role in cell migration and tissue remodeling in various pathological conditions, especially tumor cell invasion and metastasis (Liotta, 1987; Liotta et al., 1988; Ossowski, 1988a, b; Werb, 1989, 1997; Alexander and Werb, 1991; Monsky and Chen, 1993; Stetler-Stevenson et al., 1993). Because morphogenetic movements and tissue remodeling are prominent developmental processes, it is likely that proteases have a significant role in embryogenesis as well (Werb, 1997). Several developmental events have been correlated with, or shown to depend on, proteolytic activity, including branching morphogenesis (Talhouk et al., 1991; Sympson et al., 1994; Lim et al., 1995; Lelongt et al., 1997), syntrophoblast invasion of the uterine wall (Harvey et al., 1995; Lefebvre et al., 1995; Alexander et al., 1996), neural crest cell migration (Valinsky and Le Douarin, 1985; Erickson and Isseroff, 1989; Agrawal and Brauer, 1996), endocardial cushion cell migration (McGuire and Orkin, 1992; McGuire and Alexander, 1993; Alexander et al., 1997; Cai et al., 2000), and craniofacial histogenesis (Chin and Werb, 1997).

The major classes of matrix-degrading proteases (Werb, 1989, 1997; Sternlicht and Werb, 2001) are the matrix metalloprotease family (matrixins; Matrisian, 1990, 1992); the adamalysin-related membrane proteases (the ADAM family; Wolfsberg et al., 1995); the tissue serine proteases, such as uPA and thrombin; and the BMP1 family of metalloproteases. To gain insight into the role that matrix-degrading proteases may have in development, we documented the expression patterns in the chick embryo of matrix metalloprotease-2 (MMP-2, 72-kD type IV collagenase or gelatinase A). MMP-2 degrades a variety of extracellular matrix (ECM) molecules, including collagen type I, IV, V, VII, X, XI, elastin, gelatins, laminin, aggrecan, and vitronectin (Alexander and Werb, 1991; Parsons et al., 1997; McCawley and Matrisian, 2001).

One morphogenetic event during development that is likely regulated in part by proteases is the epithelial-mesenchymal transformation (EMT). The EMT is comprised of several steps, which include the loss of cell-cell cohesion between individual cells in the epithelium (“activation”) and the stimulation of cell motility (”migration”) in order for the mesenchymal cells to disperse (Hay and Zuk, 1995; Hay, 1995; Erickson, 1997). This two-step model has been particularly well documented in the emigration of cardiac cushion cells (Brown et al., 1999; Boyer et al., 1999). The molecular basis for this transformation is currently far from clear, but our observations reported here suggest that proteases are involved in this event.

In the chick embryo, we find that the expression pattern of MMP-2 varies spatially and temporally, suggesting that it is developmentally regulated. MMP-2 is expressed in epithelial tissues just at the time that they begin to undergo the EMT, suggesting that MMP-2 is instrumental in the cell dispersal phase of the EMT. Chemical inhibitors of MMPs prevent neural crest dispersion from the neural tube, both in tissue culture and in vivo. Additionally, antisense morpholino oligos against MMP-2 phenocopy the drug results. However, none of these treatments affect motility of the neural crest cells once they have undergone the EMT. In contrast, experimental perturbation of matrix metalloproteases in somites in culture using the naturally occurring MMP inhibitor, TIMP-2 (Brew et al., 2000), suggests that MMP-2 may not play a significant role in the EMT itself, but inhibits cell motility once mesenchyme is formed. MMP-2 is also expressed in mesenchyme where tissue remodeling and morphogenesis are occurring. These data suggest, at a minimum, that MMP-2 plays an important role in embryonic EMTs, in addition to the previously documented function in tissue remodeling.


MMP-2 Expression in Cells of the Neural Crest

In whole-mount in situ preparations, beginning at stage 12, there is patchy expression of MMP-2 in the cranial neural tube (Fig. 1a) and weak expression in a population of cells that emigrates from the neural tube (Fig. 1b, arrows), which are likely to be neural crest cells based on their position. HNK-1 labeling strongly suggests that some of these are neural crest cells (Fig. 2a). However, MMP-2 expression is rapidly extinguished just a short distance from the neural tube (e.g., Fig. 2a-c). There are also cells at the vagal level that are expressing MMP-2, which are probably vagal neural crest cells based on their unique position (Fig. 1d,e,i) and their HNK-1 immunoreactivity (Fig. 2d-f). At later stages in development, the mesenchyme of the head and branchial arches, which is derived from both mesoderm and neural crest, is clearly expressing MMP-2 (Fig. 1d,e,j,n) and some of these cells are apparently neural crest cells, based on HNK-1 labeling or the known location of neural crest cells from chick/quail chimera studies (Fig. 2g-I; Noden, 1982; Lumsden et al., 1991; Cai et al., 2000).

Figure 1.

Matrix metalloprotease-2 gene (MMP-2) expression in whole-mount preparations. a-c: At stage (st) 12, MMP-2 expression is seen in the head mesenchyme (arrows in b), the center of the somites (s; arrowheads in b,c), and the lateral plate (lp) mesoderm. d-i: At stage 15, MMP-2 is expressed in cells that are undergoing the epithelial-mesenchymal transformation (EMT), including cells in the somitocoel (s), the sclerotome (scl), the myotome (my) and the neural crest (arrows). j-m: By stage 19, the dermatome is undergoing the EMT and MMP-2 is expressed in the mesenchymal dermis (arrows in k,m). All that remains of expression in the sclerotome is a small dorsal “hook” (j, arrows in l, arrowhead in m). n: The same pattern persists in the somites and head mesenchyme at stage 21 (arrows indicate message in branchial arches). nt, neural tube; a, anterior; p, posterior; d/m, dermamyotome; drg, dorsal root ganglion.

Figure 2.

Some neural crest cells express matrix metalloprotease-2 gene (MMP-2). Whole-mount in situ hybridiztions of MMP-2 expression (left column) were sectioned and labeled with HNK-1 (middle column) and the labeling patterns superimposed (right column) to reveal that early migratory neural crest cells express MMP-2. f,o: Higher magnifications of the boxed regions in d and m, respectively, to show better the MMP-2 expression in the neural crest. a-f: Stage 12 (st) embryos sectioned at progressively posterior axial levels. g-o: Stage 16 embryos. Although the mesenchyme around the eye (h-j) does not label with HNK-1, it is neural crest-derived based on chick/quail chimera studies (Johnston et al., 1979). mes, mesenchyme; nt, neural tube; scl, sclerotome; ret, retina; dm, dermamyotome.

At the trunk level, there is no detectable MMP-2 label in the dorsal neural tube, when observed as whole-mount preparations (e.g., Fig. 1f). In sections through these whole-mount preparations, however, weak expression is observed in the neural crest cells that have spread between the neural tube and somite (Figs. 1g, 2m-o). Once neural crest cells have reached the dorsal edge of the somites, we can no longer detect MMP-2 expression. Our results are in contrast to Cai et al. (2000), who reported no expression of MMP-2 in chick neural crest.

Analysis of conditioned medium from neural crest cultures by gelatin zymography reveal both the proform and activated form (72 kDa and 62 kDa, respectively) of MMP-2 (Fig. 3a). MMP-2 activity is detected in 1-day-old cultures but is lost by 72 hr. This loss is not owing to the simultaneous production of an MMP-2 inhibitor, because mixing of day-3 culture medium with day-1 medium does not result in a loss of activity (Fig. 3b). Loss of activity is also not a result of neural crest cell death or poor health, because examination of cultures at the time of harvesting the medium revealed healthy cells that by day 6 were differentiating into melanocytes.

Figure 3.

a: Gelatin zymogram showing metalloprotease activity in neural crest cells. Expression of matrix metalloprotease-2 (MMP-2) is high during the first two days of culture and then rapidly disappears. b: The loss of activity is not due to the production of an inhibitor produced by day 3, because mixing of supernatant from day 1 and 3 does not result in the loss of activity.

Inhibitors of MMP activity, BB-94 and KB8301, do not appear to reduce substantially the distance that neural crest cells migrate from the neural tube when neural crest cells have already undergone the EMT (migratory pieces; Fig. 4b,d,j,l). However, when segments of neural tube from which the neural crest have not yet begun to emigrate (premigratory pieces) are exposed to inhibitor, very few neural crest cells escape (Fig. 4a,c,i,k). These latter studies suggest that MMPs may be affecting the late events of the EMT, but not neural crest cell motility per se. A much less dramatic effect on neural crest migration is observed when neural tubes are cultured in TIMP-2 (data not shown).

Figure 4.

Inhibitors of matrix metalloproteases (MMP) prevent the detachment of neural crest cells from the neural tube but do not inhibit motility. a-p: Two MMP inhibitors, BB-94 and KB8301, and their respective controls, BB-1722 and DMSO, were incorporated into collagen gels with neural tubes taken from the premigratory level (a,e,i,m,c,g,k,o) or migratory level (b,f,j,n,d,h,l,p). There were no discernible differences between the presence or absence of serum.

Embryos in ovo incubated in BB-94 show a dramatic reduction in neural crest cell detachment from the neural tube, although neural crest cells appear to differentiate, based on the acquisition of the HNK-1 epitope in cells that are retained in the dorsal neural tube (Fig. 5a,c,e,g,i). Because BB-94 inhibits MMPs other than MMP-2 (MMP-1, -7, -9, MT1-MMP; product information from British Biotech Pharmaceuticals, Ltd.), albeit at lower activities, we also inhibited MMP-2 activity specifically by selectively preventing its translation using antisense morpholino oligos (hereafter referred to as morpholinos). We electroporated morpholinos into the dorsal neural tube at two stages: stage 12 at the level of the segmental plate, where neural crest cells would not begin to migrate for at least another 12 hr (Loring and Erickson, 1987); and in stage 16 embryos, where neural crest cell migration had already commenced. The former set of experiments was designed to produce knockdown of MMP-2 protein before the time that neural crest cells have begun to detach from the neural tube. The latter experiment would assess the affect of MMP-2 on cell motility by introducing morpholinos into the premigratory crest at a time when migration was well under way. We reasoned that by the time the morpholinos have effected a knockdown of MMP-2, many neural crest cells that had taken up the morpholino would have detached from the neural tube and dispersed. When an MMP-2-specific morpholino was electroporated into one side of the neural tube of stage 12 embryos, neural crest cells accumulated in the dorsal neural tube or just outside the neural tube, and there was a reduction in the number of neural crest cells migrating on the experimental side of the embryo (Fig. 6a-f). However, when morpholinos were introduced into older embryos, morpholino-positive cells disperse and are localized in the sensory (dorsal root) ganglia (Fig. 7). We could not determine whether there was a reduction in the total number of migratory cells under these conditions. We used two control morpholinos, the standard control manufactured by GeneTools and a so-called nonsense control, for which there is no match in GenBank. Neither control morpholino resulted in the disruption of neural crest cell migration (Kos et al., 2003). The morpholino reduces MMP-2 protein levels, because antibody generated against MMP-2 fails to immunoreact with morpholino-positive neural crest cells in culture (Fig. 6g,h).

Figure 5.

BB-94 applied to embryos inhibits neural crest cell detachment from the neural tube. a,c,e,g,i: BB-94 was applied to stage 10 embryos in EC culture, then fixed 24 hr later, sectioned, and labeled with HNK-1. b,d,f,h,j: Control embryos were incubated in BB-1722 (inactive control for BB-94), and patterns of neural crest migration were compared with stage-matched experimental embryos. Note that in the BB-94-treated embryos, neural crest cells either fail to migrate or are retarded in their migration compared with the controls, and many HNK-1-positive cells accumulate in the dorsal neural tube (arrows).

Figure 6.

Matrix metalloprotease-2 (MMP-2) morpholinos inhibit the epithelial-mesenchymal transformation. a-f: Morpholinos were electroporated into the right side of the neural tube of stage 12 (st) embryos and fixed 24 hr later, at stages 16 (a,b), 17 (c,d), and 18 (e,f). Morpholinos fluoresce red (left column) and HNK-1 is labeled with a green (Cy-2) secondary antibody. Merged figures showing the morpholinos and HNK-1 labeling are in the right column. Neural crest cell migration is relatively normal on the left side. On the right side, few neural crest cells are migrating and many morpholino-containing/HNK-1-positive cells are trapped in the neural tube (yellow in the superimposed figures indicated by white arrowheads) or in the lumen (arrows). g,h: Stage 12 embryos were electroporated with morpholino (red), the neural tubes (nt) explanted in culture, and after 24 hr immunolabeled with an antibody raised against MMP-2 (green). Few neural crest cells containing morpholino (arrows) detach from the neural tube, but those that do contain no, or greatly reduced levels of, MMP-2 immunoreactivity.

Figure 7.

a-h: When morpholinos (red; a,c,e,g) are electroporated into older embryos (stage [st] 16) and fixed 24 hr later, some neural crest cells (green; b,d,f,h) are still able to migrate. This finding may be because the morpholinos do not knockdown matrix metalloprotease-2 (MMP-2) expression until after the neural crest cells have already detached from the neural tube.

MMP-2 Expression in the Somites and Other Mesoderm

Whole-mount in situ hybridization beginning at stage 12 reveals low levels of MMP-2 expression in the center of the most posterior somites (i.e., the most recently formed somites; Fig. 1b,f). The expression gradually spreads throughout the somites as they age (or as one looks more anteriorly in the same embryo). Sections through the whole-mount preparations reveal the label is initially in the very central cluster of cells in the somitocoel (Fig. 1c,g), which is always mesenchymal. Expression then extends throughout the sclerotome as this tissue undergoes the EMT and disperses (Figs. 1h, 2d,k,p). Thus, expression is correlated with migration of the sclerotome but is not seen distinctly in the epithelium from which the sclerotome arises. Gradually, MMP-2 expression is extinguished, except for an intensely labeled band of sclerotome cells at the posterior margin of the somite (Fig. 1j,l, arrowhead in m). Somites cultured in the naturally occurring MMP inhibitor, TIMP-2, undergo an EMT, as evidenced by the appearance of some mesenchyme, but fail to disperse (Fig. 8), suggesting MMP-2 promotes cell migration in this cell type. Similarly, when embryos are treated with BB-94 (Fig. 5, left panel), the sclerotome undergoes an EMT but fails to disperse as far as the control embryos (Fig. 5, right panel).

Figure 8.

Tissue inhibitor of metalloprotease-2 (TIMP-2) inhibits the dispersion of the sclerotome. a,c,e: A somite in TIMP-2 at 12, 24, and 48 hr of culture in serum-free conditions. b,d,f: The corresponding controls. g,i,k: A somite in TIMP-2 at 12, 24, and 48 hr of culture in serum-containing conditions. h,j,l: The corresponding controls.

At about stage 18, as MMP-2 expression is lost in the sclerotome of the oldest somites, it is seen for the first time in the dermis and is correlated with that portion of the dermatome that is undergoing the EMT (Fig. 1j,k,m,n). Expression is initially observed at the ventrolateral edges of the somites (i.e., the edge of the somite that borders the intersomitic space). As development proceeds at a given axial level (or as one observes more anterior axial levels of the embryo), expression in each dermatome spreads from the intersomitic edges into the center of the dermatome and also dorsally. This pattern of expression reflects the progression of the EMT across the dermatome (Tosney et al., 1994). Because expression is high in the dermis but is barely detectable in the dermatome (e.g., Fig. 2k,m), MMP-2 expression is correlated with dermis migration rather than the initial stages of the EMT.

We as well as Cai and coworkers (2000) never see expression in the myotome, which was reported previously (Yang et al., 1996). The myotomes appear to form normally in BB-94-treated embryos.

As feather buds begin to form along the dorsal midline after stage 30, MMP-2 expression becomes concentrated in the buds and is diminished in the rest of the dermis, correlated with the migration of dermal cells into the forming feather buds (Fig. 9e). MMP-2 expression is also found in the lateral plate mesoderm, which is mesenchymal, and expresses MMP-2 as early as stage 12 (Fig. 1a,c).

Figure 9.

a-j: Matrix metalloprotease-2 gene (MMP-2) expression in older embryos. f,g,h: MMP-2 expression in the limb bud is largely in the dermis and is similar to patterns of tendon development (Kardon, 1998; Schweitzer et al., 2001). aer, apical ectodermal ridge. d,e: MMP-2 is also expressed in the dermis of the developing feather germs.

MMP-2 Expression in the Limb

The first expression in the limb occurs at stage 18 in the mesenchyme immediately beneath the ectoderm (Fig. 1j, 9f). It is conspicuously absent from the apical ectodermal ridge. The most pronounced expression of MMP-2 in the limb appears in stage 24 hind limbs in the dorsoproximal (i.e., at the level of the zygopod) midline of the limb (Fig. 9b), which sectioning reveals is in the dermis just subjacent to the ectoderm (Fig. 9g). As the hind limb continues to extend, expression in the dermis is seen more distally. By stage 27, the dorsal dermis still expresses MMP-2, but expression is concentrated in two regions on either side of the midline at the stylopod level, and then forms an arc across the limb at the border with the autopod (Fig. 9c,i). As the digits begin to emerge (after stage 30), MMP-2 expression in the dermis is concentrated around the forming digits (Fig. 9d,i,j). Sections reveal that this expression is always in the dermis immediately beneath the ectoderm (Fig. 9h) and is precisely correlated with the regions where tendons are forming (Kardon, 1998; Schweitzer et al., 2001).


In this study, we have determined the expression pattern of MMP-2 in the chicken embryo. The developmentally regulated expression is generally correlated with the EMT that generates the neural crest, the sclerotome, and dermatome, suggesting that MMP-2 is critically involved in the transformation of epithelia to mesenchyme, and also plays a role in the later dispersion of mesenchymal tissues. Experimental perturbation with drugs and knockdown with antisense morpholino oligos gives additional support for MMP-2 having a role in triggering the EMT that generates the neural crest, and for playing a later role in the dispersal of the sclerotome.

Proteases and the EMT

The first step in the EMT is the loss of cell-cell cohesion, primarily by removal of cadherins. This step has been coined the “activation” step (Boyer et al., 1999). Although several mechanisms have been proposed for the loss of cadherins during the EMT, one possibility is that proteases cleave cadherins and subsequently release them from the cell surface (Werb, 1997; Werb and Yan, 1998). Recently, several cell surface proteins (DiStefano et al., 1993; Pan and Rubin, 1997; Black et al., 1997; Kahn et al., 1998), including cadherins (Volk et al., 1990; Paradies and Grunwald, 1993; Lochter et al., 1997; Herren et al., 1998; Bergers and Coussens, 2000), have been shown to be removed by proteolytic “sheddase” activity (Hooper et al., 1997). Although MMP-2 cleaves several cell surface proteins (reviewed in McCawley and Matrisian, 2001), there is no report of its processing cell-cell adhesion molecules, such as cadherins. Moreover, it is expressed late in the process, as the cells begin to disperse, not during the initial “activation” stage.

Although MMP-2 is expressed primarily in mesenchyme as it is generated during the EMT [e.g. in the neural crest cells, in the lateral plate mesoderm, the dermis, the sclerotome, and endocardial cushion cells (Alexander et al., 1997; Cai et al., 2000)] and, therefore, predicted to have a role in the dispersal of the mesenchyme, it is possible that we could not detect low levels of expression in epithelia by in situ hybridization and, therefore, might have an earlier role in the activation stage. In fact, MMP-2 was seen occasionally in the somitic epithelium, although expression was generally patchy and inconsistent (e.g., Fig. 1g). In other published studies that examine the distribution of matrix metalloproteases throughout the early embryo (MMP-2 in the mouse, Reponen et al., 1992; Kinoh et al., 1996; MMP-2 in the chick, Cai et al., 2000; MMP-9 in the mouse, Canete-Soler et al., 1995; stromelysin-3 in the mouse, Lefebvre et al., 1995), no MMPs have been shown to be expressed in epithelia before the EMT, nor were proteases expressed in epithelia in the developing kidney (Lelongt et al., 1997), tooth (Sahlberg et al., 1992), or head (Chin and Werb, 1997). Similarly, in situ preparations of breast tumors show that MMP-2 is expressed in the fibroblasts surrounding the tumor but not in the epithelial carcinoma (Gilles et al., 1997).

Our results suggest that MMP-2 plays a role in the second or “migration” phase of the EMT that generates the neural crest, but once the EMT is complete, MMP-2 does not regulate neural crest cell motility. When premigratory-level neural tubes were treated with the MMP-inhibitor BB-94, no neural crest cells detached from the neural epithelium. Similarly, knockdown of MMP-2 expression in vivo using morpholinos results in the accumulation of HNK-1-positive cells in the dorsal neural tube and a reduction in the number of neural crest cells that disperse. However, when BB-94 is applied to migratory-level neural tubes, where neural crest cells have already undergone the EMT, it does not appear to affect neural crest motility (Fig. 4). These results suggest that MMP-2 and other metalloproteases are involved in the later dispersal phase of the EMT but not in the migration of mesenchyme once the EMT is completed. These curious results suggest that regulation of dispersal from the epithelium differs from later mechanisms that control cell motility. One possibility is that MMP-2 helps to degrade the basal lamina, allowing neural crest cells to escape the epithelium, but has no affect on motility per se.

Proteases and Cell Motility

Our in situ results suggest that MMP-2 is important in the initial dispersal of neural crest during the EMT and dispersal of the sclerotome and dermis after the EMT. Previous studies also showed that MMP-2 is expressed in migrating chick endocardial cushion cells (Alexander et al., 1997; Cai et al., 2000). MMP-2 is expressed in the mesenchyme of the mouse embryo (Reponen et al., 1992; Robbins et al., 1999), although the cranial neural crest was the only MMP-2-expressing cell population in those studies that overlapped with chicken MMP-2 expression. Direct inhibition of MMPs in mouse neural crest cultures (Robbins et al., 1999) or in chick neural crest cultures (Cai and Brauer, 2002) further show that this class of matrix-degrading enzymes may stimulate motility. Moreover, in the mouse mutant Patch, in which craniofacial mesenchyme does not disperse properly, MMP-2 activity is depressed (Robbins et al., 1999). Of interest, our cell culture studies and embryo studies suggest that once neural crest cells are fully mesenchymal, they are not affected by loss of MMP-2 activity. Thus, MMP-2 is likely only involved in the very earliest stages of dispersal. The rapid down-regulation of MMP-2 in the neural crest once they detach from the neural tube, as visualized by in situ hybridization, supports this notion. However, our perturbation of somites with BB-94 or TIMP show that the EMT is not affected, but dispersal is greatly reduced, suggesting that MMP-2 acts later in the somite. Similarly, in situ studies show that expression of MMP-2 is high in the dermis and the sclerotome through the process of dispersion.

Proteases may stimulate cell migration in the organism in several ways. Tumor cells are known to produce proteases, and there is good evidence to suggest that degradation of the ECM allows tumor cells to break through restraining basal laminae and disperse in loose connective tissue (McCawley and Matrisian, 2001). However, degradation of basal lamina components can also directly affect motility. For example, MMP-2 cleaves Ln-5, and this cleaved form is a potent stimulus of motility in the absence of any affect on adhesion (Giannelli et al., 1997). Similarly, MMP-2 degradation of type IV collagen is required for endocardial cushion cell migration (Song et al., 1998), and MMP-2 degradation of chondroitin sulfate proteoglycans stimulates Schwann cell motility (Krekoski et al., 2002). Finally, proteases have been localized to the focal contact (Hebert and Baker, 1988; Pollanen et al., 1988), which is the site of cell adhesion to the ECM, or to other regions of the lamellipodium (Chen, 1989), and they have been proposed to break the cell adhesions to allow cells “to take another step.” For example, β4 integrin is cleaved by MMP-7, resulting in decreased adhesion of prostate tumor cells (von Bredow et al., 1997).

MMP-2 and Matrix Turnover During Tissue Remodeling

Numerous studies have suggested that the matrix metalloproteases are involved in matrix processing and turnover during development and are required for tissue morphogenesis. In particular, MMPs are expressed in tissue mesenchyme that controls branching phenomena, such as in the lung (Lim et al., 1995) and in salivary gland (Reponen et al., 1992), kidney tubulogenesis (Lelongt et al., 1997), and mammary gland branching and involution (Talhouk et al., 1991; Sympson et al., 1994), or in other epithelia where basal laminae are undergoing remodeling (Lefebvre et al., 1995). MMPs are also associated with chondrogenesis (Chin and Werb, 1997; Vu et al., 1998), osteogenesis (Lefebvre et al., 1995), angiogenesis (Brooks et al., 1996; Vu et al., 1998), and in the complex process of craniofacial morphogenesis (Chin and Werb, 1997). It is not certain how the turnover of matrix is involved in tissue morphogenesis, but could include processing of ECM so that it is more adhesive, release of growth factors from the matrix, or changes in gene expression (Werb, 1997). We did not observe MMP-2 expression in most of the tissues where metalloproteases have been found previously in other organisms, especially in chondrocytes or osteocytes. However, we did see quite complex expression patterns in the craniofacial mesenchyme, and also in the dermis of the limb, where MMP-2 is coincident with known sites of tendon morphogenesis (Kardon, 1998; Schweitzer et al., 2001). One possibility is that the matrix molecules produced by tendon, such as the tenascins, are processed extracellularly, and MMP-2 may be required for proper tendon development.


Documenting the expression patterns of MMP-2 throughout the embryo has allowed us to draw some generalizations about the function of this protease during embryonic morphogenesis, especially its role in the late events of the EMT. Experimental studies on cell cultures and in the embryo using MMP inhibitors and using morpholino gene knockout strategies (Kos et al., 2001) to perturb specifically MMP-2 function similarly show that it is instrumental in triggering some steps in the EMT.


Cell Culture

Fertile quail eggs (Coturnix japonica) obtained from the Animal Sciences Department (University of California at Davis) were incubated for 56 hr at 37°C and then used to generate neural crest cultures, as described previously (Erickson and Goins, 1995; Reedy et al., 1998). Briefly, thoracic-level neural tubes were enzymatically digested and then cultured in 35-mm Petri dishes (tissue culture dishes; Corning) in Ham's modified F-12 medium (Gibco) supplemented with 10% fetal calf serum (Gibco), 4% chick embryo extract, and 1% penicillin/streptomycin solution (Gibco). Once the neural tubes stick to the dish, a relatively pure population of neural crest cells emigrates onto the plastic substratum (Loring et al., 1981). For some experiments, neural tubes were cultured in three-dimensional collagen gels in which inhibitors of metalloproteases had been incorporated. MMP inhibitors included KB8301 (10 μM; Pharmingen), BB-94 (Batimastat; 5 μM; British Biotech Pharmaceuticals, Ltd.), 1,10-phenanthroline (5 × 10-5; Sigma), and TIMP-2 (2 μg/ml; gift from Jim Quigley).

Zymography of Quail Neural Crest Cell Cultures

Quail neural crest cultures were generated on tissue culture plastic, as described above. Culture medium was then changed every 24 hr with F12 medium that contained 4% CEE but not fetal calf serum, and was assayed for proteolytic activity by using zymography. Conditioned culture medium was electrophoresed in 10% polyacrylamide minigels into which 3 mg/ml of gelatin (Sigma) was cross-linked to detect matrix metalloproteases. After electrophoresis, gels were washed 3 times for 5 min each with 2.5% Triton X-100 in TBS and then rinsed 6-8 times in ddH2O. Finally, gels were incubated for approximately 16 hr at 37°C in 100 mM Tris, pH 8.0. Areas of proteolysis were visualized as clear bands in a blue background after staining with 0.125% Coomassie blue. Gels were dried and photographed by using a Nikon 35-mm camera and light box.

Chicken MMP-2 cDNA Probes

The following forward and reverse primers, respectively, were used to obtain an MMP-2 cDNA fragment by polymerase chain reaction from stage 20 chick tissue (5′ ACT CCT GTA CAA GTG CAG 3′ and 5′ CAG AAT GTA GCA ACG AGA 3′). A 474-bp cDNA fragment was isolated (corresponding to bp 1204-1678) and subcloned into pBluescript-SK- (Stratagene). The 474-bp MMP-2 fragment was sequenced, and a GenBank search confirmed its identity to chicken MMP-2. Digoxigenin-labeled riboprobes were generated using T7 or T3 RNA polymerase and digoxigenin-labeling RNA mix (Boehringer Mannheim).

Whole-Mount In Situ Hybridization

Fertile White Leghorn chicken eggs (Gallus gallus) obtained from Western Scientific (Sacramento) were incubated at 37°C in a forced-draft egg incubator for varying periods of time and staged according to the criteria of Hamburger and Hamilton (1951). Embryos of the appropriate stages (stages 11-stage 34) were fixed in 4% paraformaldehyde overnight at 4°C with rocking. The embryos were then processed for in situ hybridization as described previously (Kos et al., 2001), except that the embryos were developed in NTMT buffer (100 mM NaCl, 100 mM Tris-HCl, pH 9.5, 25 mM Mg Cl2, 0.5% Tween-20, 2 mM levamisole) containing 4.5 μl/ml of a 75 mg/ml nitroblue tetrazolium stock (NBT) and 3.5 μl/ml of a 50 mg/ml 5-bromo-4-chloro-3-indoyl phosphate stock (BCIP). After sufficient signal had developed, embryos were washed for 12 hr in several changes of PBT and then fixed for 24 hr in 4% paraformaldehyde/0.2% glutaraldehyde at 4°C. Embryos were photographed by using a Leica stereomicroscope equipped with a video camera (Optronics). Some embryos were later embedded in paraffin and sectioned at 10 μm. To determine the distribution of neural crest cells in sections of in situs, we labeled with HNK-1, as described previously (Kos et al., 2001).

Inhibition of MMP Activity In Vivo

To inhibit MMP activity in the embryo, we explanted stage 10 embryos into whole-embryo culture by using the technique of Chapman and coworkers (Chapman et al., 2001). One milliliter of a 5 μM solution of BB-94 was applied to the top of each embryo, which was then allowed to develop for another 24 hr before fixing, sectioning, and labeling with the HNK-1 antibody. We analyzed 23 embryos treated with BB-94 and 17 embryos with BB-1722 (control for BB-94; British Biotech Pharmaceuticals, Ltd.).

We also used morpholinos to knock down expression of MMP-2 specifically. Morpholinos were designed by GeneTools, Inc. (ACTGTGAGTCTTCATTCTGCCTATA) and incorporated into the premigratory neural crest by electroporation, as described previously (Kos et al., 2001, 2003). Morpholinos were used at 1 μM. We did these electroporations either in stage 12 embryos and fixed them 24 hr later (stage 16-19; n = 26), or in stage 16 embryos and fixed 24 hr later (stage 21; n = 16). The former set of experiments were designed to produce knockdown of MMP-2 expression before the time that neural crest cells have begun to detach from the neural tube. The latter experiment introduced morpholinos into the premigratory crest at a time when migration was well under way. Thus, by the time that the morpholinos have effected a knockdown of MMP-2, many neural crest cells will have been able to detach from the neural tube and will have dispersed. Control morpholinos included the standard control marketed by GeneTools and a nonsense control (5′-CCTCTTACCTCAGTTACAATTTATA), which is a sequence that does not match any in GenBank. These controls are described in detail in the methods study by Kos et al. (2003).

To determine whether the morpholinos were reducing the level of MMP-2 protein, we incorporated morpholino into the entire dorsal neural tube of stage 12 embryos and then excised the neural tubes and cultured them, as described above. After 18 hr, the cultures were fixed with 4% paraformaldehyde, washed, and immunolabeled with a rabbit polyclonal antibody against MMP-2 (5 μg/ml), kindly provided by Jim Quigley, followed by an Alexa 488-conjugated goat anti-rabbit secondary (1:2,000; Molecular Probes). We analyzed the cultures for MMP-2 immunoreactivity in morpholino-positive and -negative neural crest cells.


We thank Drs. Peter Armstrong, Sherry Rogers, and Ronelle Hall for reading various drafts of this manuscript and Jim Quigley for sharing reagents, unpublished data, and his insights into chick proteases. We also thank Robert Kos for help in preparing the figures. C.A.E. was supported by a grant from the NIH.