Cellular invasion of the chicken corneal stroma during development: Regulation by multiple matrix metalloproteases and the lens



Avian corneal development requires cellular invasion into the acellular matrix of the primary stroma. Previous results show that this invasion is preceded by the removal of the fibril-associated type IX collagen, which possibly stabilizes matrices through interfibrillar cross-bridges secured by covalent crosslinks. In the present study, we provide evidence for the expression of three matrix metalloproteinases (MMPs) in early corneas, two of which act cooperatively to selectively remove type IX collagen in situ. In organ cultures, MMP inhibitors (either TIMP-2 or a synthetic inhibitor) resulted in arrested development, in which collagen IX persisted, and the stroma remained compact and acellular. We also show that blocking covalent crosslinking of collagen allows for cellular invasion to occur, even when the removal of type IX collagen is prevented. Thus, one factor regulating corneal invasion is the physical structure of the matrix, which can be modified by either selective proteolysis or reducing interfibrillar cross-bridges. We also detected another level of regulation of cellular invasion involving inhibition by the underlying lens. This block, which seems to influence invasive behavior independently of matrix modification, is a transient event that is released in ovo just before invasion proceeds. Developmental Dynamics 232:106–118, 2005. © 2004 Wiley-Liss, Inc.


The avian cornea develops in a series of discrete steps that have been reviewed previously (Hay and Revel, 1969; Hay, 1980; Linsenmayer et al., 1998). In brief, beginning at 3 days of incubation, an initial acellular extracellular matrix (termed the primary stroma) is deposited by the corneal epithelium on its basal surface. At 4 days, mesenchymal cells from the periphery migrate centrally on the undersurface of the primary stroma and form an epithelium that becomes confluent by 4.5 days. Approximately a day later, the compact primary stroma rapidly swells, and is invaded by another wave of periocular mesenchymal cells. These cells differentiate into corneal fibroblasts and secrete elements of the secondary, or mature, corneal stroma. At 12 days, the secondary stroma begins to condense and ultimately becomes transparent.

Knowledge of the mechanisms that regulate the invasion of the primary stroma is still rudimentary; however, it is likely that developmental changes in this matrix are critical. Morphologically, the primary stroma is detectible as a compact acellular matrix (∼10 μm thick) composed of orthogonal layers of thin, uniform-diameter, collagen fibrils (Trelstad and Coulombre, 1971; Bard and Bansal, 1987) and a variety of noncollagenous glycoproteins and proteoglycans (Conrad and Dorfman, 1974; Tucker, 1991; Kaplony et al., 1991; Doane et al., 1996). Each layer, which is only one or two fibrils thick, is composed of interstitial fibrils of collagen types I and II coassembled as heterotypic molecules (Hendrix et al., 1982; Linsenmayer et al., 1990). These fibrils interact with several fibril-associated collagens (Svoboda et al., 1988; Gordon et al., 1996), termed FACITs (Gordon et al., 1989; Shaw and Olsen, 1991), which are thought to play a role in the structural organization of connective tissues. FACITs have one or more triple-helical domains that lie along the fibril surface and interact with the underlying fibrillar collagens. Other domains extend outward from the fibril; these are thought to interact with other matrix components. In the primary stroma, a major FACIT is collagen type IX (Van der Rest and Mayne, 1987; Svoboda et al., 1988; Fitch et al., 1988b). This is the prototype FACIT, which in cartilage has been extensively characterized biochemically. In cartilage, type IX molecules are aligned with fibrils such that intermolecular covalent crosslinks can form between collagen IX and the type II collagen molecules within the fibril (Wu and Eyre, 1984; Van der Rest and Mayne, 1988; Brewton and Mayne, 1994). Moreover, crosslinks involving type IX collagen can potentially link together fibrils directly through crosslinks to collagen II in adjacent fibrils, or indirectly by means of crosslinks between molecules of collagen IX on adjacent fibrils (Wu et al., 1992; Miles et al., 1998; Eyre et al., 2002, 2004). In addition, the amino-terminal globular domains of collagen IX molecules can associate in vivo, possibly also forming a bridge between adjacent fibrils (Douglas et al., 1998).

In the cornea, studies on the acquisition of type IX collagen by the primary corneal stroma suggest a structural arrangement similar to that of cartilage, in which collagen IX, along with collagen I and II, are deposited early in development (Fitch et al., 1988b; Svoboda et al., 1988), with type IX collagen anchored to the stromal matrix by covalent crosslinks (Fitch et al., 1994). Thus, the compact nature of the primary stroma might be maintained by interfibrillar cross-bridges composed of type IX collagen.

This acellular primary stroma provides a substratum for two sequential waves of migration by the periocular mesenchymal cells. These cells, which are of neural crest origin (Johnston et al., 1979), are initially located around the periphery of the cornea. At 4 days of development, the first wave of these cells begins to migrate centripetally between the lens and the undersurface of the primary stroma, using both the anterior surface of the lens and the posterior surface of the stroma as surfaces for migration (Bard et al., 1975). By 4.5 days, they have formed a confluent epithelial monolayer, termed the corneal endothelium (Hay, 1980).

The second wave of migration begins at approximately 5.5 days, when the mesenchymal cells invade the primary stromal matrix itself. Immediately before this event, the matrix rapidly swells, and the collagenous layers separate (Coulombre, 1965; Hay, 1980; Linsenmayer et al., 1990). Within a day, the cells have reached the center of the cornea, and by 7 days, through further migration and proliferation, they completely populate the stroma. These cells differentiate into the corneal fibroblasts, which synthesize the components of the mature stroma.

Previous studies suggest that changes in the collagenous structure of the primary stroma are involved in regulating the invasion of this matrix. Just before invasion, there is a rapid loss of collagen type IX, but not of fibrillar collagen type II, from the primary stroma (Fitch et al., 1988b). This finding suggested a mechanism that selectively removes the FACIT collagen, leaving the fibrils intact. Although transcriptional regulation of type IX collagen occurs at approximately this time (Fitch et al., 1995), these changes in expression cannot account for the abrupt removal of the type IX collagen, including molecules already incorporated into the matrix. Instead, this likely involves proteolysis by one or more extracellular matrix metalloproteinases (MMPs). Analyses by immunofluorescence and immunoblotting show that type IX collagen is degraded into fragments at the time of corneal swelling. Zymograms of corneal extracts taken before swelling detect only a latent (proenzyme) form of MMP-2, whereas those taken at the time of swelling also show the active form of the enzyme (Fitch et al., 1998).

These observations support a model in which (1) interfibrillar crosslinks involving molecules of collagen types IX and II stabilizes the matrix, keeping it tightly compact and impenetrable to invasive cells; (2) developmentally regulated proteolysis of type IX molecules destabilizes the primary stroma; and (3) the matrix swells, driven by the hydration and expansion of hyaluronic acid (Toole and Trelstad, 1971), a glycosaminoglycan that is present in the primary stroma and is known to exert considerable hydrostatic pressure (Toole and Trelstad, 1971; Trelstad et al., 1974; Toole, 1991). This swollen matrix is then permissive for cellular invasion into its expanded interfibrillar spaces.

This model was tested by inhibiting MMP activity with a tissue inhibitor of metalloproteinase (TIMP-1) in organ cultures of primary stroma-stage anterior eyes (Cai et al., 1994). Such cultures normally undergo the subsequent events of corneal development, including the selective disappearance of type IX collagen, swelling of the primary stroma and its invasion by mesenchymal cells. We found that exogenous TIMP-1 partially inhibited the loss of collagen IX, but matrix swelling and mesenchymal cell invasion nevertheless occurred. Analysis by confocal microscopy revealed that discrete regions that retained collagen IX remained compact and acellular, whereas extensive cellular invasion had occurred into the surrounding matrix that lacked type IX collagen. Although TIMP-1 is a potent inhibitor of MMP-2, more recent findings demonstrated that this TIMP inhibits other MMPs weakly (Brew et al., 2000). The failure of TIMP-1 to completely inhibit the degradation of type IX collagen and the subsequent swelling and invasion of the corneal matrix suggested that other proteinases, perhaps unaffected by TIMP-1, might also be expressed and active at this time in corneal development.

In the present study, we have examined this possibility and demonstrate by reverse transcription-polymerase chain reaction (RT-PCR) that two other MMPs, in addition to MMP-2, are expressed in early corneas. At least one of these (MT3-MMP) is relatively unaffected by TIMP-1 but is strongly inhibited by TIMP-2 (Will et al., 1996; Matsumoto et al., 1997; Hotary et al., 2000). Incubation of sections of primary stroma-stage anterior eyes with recombinant MMP-2, MT3-MMP, or both together, showed that each enzyme alone had little or no effect on immunoreactivity for collagen IX but that, together, they selectively removed the molecule, as it is arranged in situ, from the corneal stroma. Treatment of organ cultures with exogenous TIMP-2, or with other synthetic inhibitors of MMPs, more completely blocked the loss of immunoreactivity for type IX collagen, and also inhibited both corneal swelling and invasion. The role of interfibrillar covalent crosslinks in controlling cellular migration was also investigated in organ cultures in which collagen crosslinking was prevented by an exogenous inhibitor of crosslink formation. In such cultures, stromal invasion occurred even when removal of collagen IX was also inhibited.

Finally, other observations suggest that cellular invasion of the developing cornea is subject to another level of regulation in addition to changes in the primary stroma. In previous studies using these organ cultures, we found that early corneal development could be experimentally manipulated by culturing the corneas in the presence or absence of the lens (Zak and Linsenmayer, 1985). Whereas primary stroma-stage corneas explanted without the lens developed as described above, corneas explanted with the lens failed to develop beyond the primary stroma stage. Collagen IX persisted, and the stroma remained compact and acellular. This finding is somewhat enigmatic, because in ovo the lens is present throughout corneal development and suggests that either modified culture conditions and/or additional tissues are required in vitro for necessary developmental interactions to occur. Here, a culture system is described in which further development proceeds in the presence of the lens, and also reveals an inhibitory effect of the lens on invasive behavior of mesenchymal cells that is independent of modifications to the primary stroma.


Expression of MMPs in Developing Corneas

Our previous studies using TIMP-1 suggested that, in addition to the MMP-2 identified by zymography (Fitch et al., 1998), other enzymes also participate in early avian corneal morphogenesis. To assess which MMPs might be involved, we performed RT-PCR on mRNA from 6- to 7-day embryonic corneas using gene-specific primers for the MMPs that have been identified in chickens (i.e., MMP-2 [Aimes et al., 1994], MMP-9 [Hahn-Dantona et al., 2000], MMP-13 [Lei et al., 1999], CMMP [Yang and Kurkinen, 1998], and the membrane type MT3-MMP [Yang et al., 1996]). The 6- to 7-day corneas were examined because they represent a stage at which stromal swelling and cellular migration has just been initiated. The results were compared with those using cDNA from 14-day sternal cartilage, a tissue containing many of the same matrix components found in the primary stroma-stage corneas. As shown in Figure 1, 6- to 7-day corneas expressed mRNAs for MMP-2, MT3-MMP, and CMMP but gave little or no product for MMP-9 and MMP-13. In contrast, 14-day cartilage contained mRNAs for MMPs -2, -9, and -13, and MT3-MMP but not for CMMP. (As a positive control, cornea and cartilage each gave a signal for mRNA encoding type IX collagen, known to be present in both.) Therefore, the early developing cornea potentially has activities for at least three MMPs: MMP-2, MT3-MMP, and CMMP.1

Figure 1.

Expression of known avian matrix metalloproteinases (MMPs) in embryonic chicken tissues. Reverse transcription-polymerase chain reaction was performed using mRNA from 6- to 7-day corneas (lanes 1–7) and 14-day cartilage (lanes 8–14) and gene-specific primers for MMP-2 (lanes 1 and 8), MMP-9 (lanes 2 and 9), MMP-13 (lanes 3 and 10), CMMP (lanes 4 and 11), MT3-MMP (lanes 5 and 12), and the α1 chain of type IX collagen (lanes 6 and 13). Lanes 7 and 14 show a negative control from which primers were excluded. The left-hand lane shows DNA size markers. The cornea shows expression of MMP-2, CMMP, and MT3-MMP but not of MMP-9 and MMP-13. The cartilage tissue gave products for MMP-9 and -13, along with MMP-2 and MT3-MMP but not for CMMP.

Digestion of Ocular ECM Components In Situ by Exogenous MMPs

Of the matrix proteases shown to be expressed in early corneas, two (MMP-2 and MT3-MMP) are cloned, well-characterized, and available as recombinant enzymes (from mammalian genes; see Experimental Procedures section). We asked whether these proteases are able, when acting individually or together, to remove collagen IX, as it occurs in situ in the primary stromal matrix, from preswelling-stage corneas. For this, we incubated unfixed frozen sections of anterior eyes from 5-day chicken embryos with various dilutions of active MMP-2, MT3-MMP, or both in combination (1:1) and analyzed the treated sections for immunoreactivity for collagen types I, II, IV, and IX. As a control, fibronectin (FN), a matrix glycoprotein that, like collagen IX, also becomes diminished when the primary stroma is invaded (Doane et al., 1996; Kurkinen et al., 1979; and unpublished observations) was also analyzed. FN is reported to be a substrate for a wide variety of MMPs, including MMP-2 and MT3-MMP (Corcoran et al., 1996; Seiki, 1999). The results for FN and collagen types II and IX after digestion by 0.1 μg/ml of enzyme are shown in Figure 2.

Figure 2.

Digestion of unfixed frozen sections of 4.5-day anterior eyes with active exogenous enzymes. A–L: Sections of embryonic chicken eyes were incubated with buffer alone (A–C), active matrix metalloproteinase (MMP) -2 (D–F), active MT3-MMP (G–I), or a 1:1 mixture of MMP-2 and MT3-MMP (J–L). The sections were then reacted with monoclonal antibodies for collagen type II (left column, A,D,G,J), collagen type IX (middle column, B,E,H,K), or fibronectin (right column, C,F,I,L). In these and subsequent images, bound antibodies were identified by rhodamine-labeled second antibody (red/orange color); cells were revealed by Hoechst nuclear dye (green color). AC: Buffer control. Collagen type II (A) and type IX (B) and fibronectin (C) were all localized in the compact acellular primary stroma, collagen II and fibronectin throughout the stroma, and collagen IX in its characteristic pattern marking the outer third of the stroma. DF: MMP-2. Collagen type II (D) and type IX (E) were resistant to an overnight digestion with MMP-2; in contrast, this enzyme removed immunoreactivity for fibronectin (F). GI: MT3-MMP. Collagen II (G) was unaffected, and collagen IX (H) was only slightly diminished by MT3-MMP digestion, whereas fibronectin was again removed by the treatment (I). JL: MMP-2 + MT3-MMP combined. Collagen type II (J) was largely unaffected by digestion by both enzymes together, but immunoreactivity for collagen IX (K) and fibronectin was greatly diminished or eliminated (L). Scale bar = 500 μm in A (applies to A–L).

After an overnight incubation at 37°C, exogenous MMP-2 alone (Fig. 2D–F) had no effect on the pattern or apparent intensity of collagen types II (Fig. 2D) or IX (Fig. 2E) as compared with the buffer controls (Fig. 2A,B). Nor did this enzyme noticeably affect the immunoreactivity for collagen types I or IV (not shown). However, MMP-2 essentially eliminated corneal immunoreactivity for FN (Fig. 2F) to background levels. Likewise, digestion of sections with MT3-MMP alone (Fig. 2G–I) also eliminated FN immunoreactivity (Fig. 2I) but had little or no effect on collagen type II (Fig. 2G), I, or IV (not shown), and it caused just a barely discernible reduction of the signal for collagen type IX (Fig. 2H). Digestion with both enzymes together (Fig. 2J–L), however, resulted in essentially complete removal of immunoreactivity for collagen IX (Fig. 2K), as well as that for FN (Fig. 2L) but had relatively little effect on collagen types II (Fig. 2J), I, or IV (not shown). Therefore, although neither exogenous MMP-2 nor MT3-MMP alone removed collagen type IX as it is arranged structurally in the corneal primary stroma, the action of both enzymes together is sufficient to achieve this result. Among the collagens analyzed, the sensitivity to degradation in situ appears selective for type IX collagen. The sensitivity of FN to both enzymes made it useful as a positive control for enzyme activity.

Effect of MMP Inhibitors on Primary Stroma-Stage Corneal Organ Cultures

Previous studies on the effect of exogenous TIMP-1 on organ cultures of primary stroma-stage corneas showed that this inhibitor only partially blocked the removal of type IX collagen and failed to inhibit swelling and invasion of the primary stroma (Cai et al., 1994). Because TIMP-1 is known to be a potent inhibitor of MMP-2 but not of certain other MMPs, this finding raised the possibility that other proteases might be involved in regulating these developmental events.

At least one such MMP, MT3-MMP, was demonstrated to be expressed in the developing cornea (and is likely active; see Discussion section); previous studies have shown that MT3-MMP (Matsumoto et al., 1997; Hotary et al., 2000) is strongly inhibited by TIMP-2, as well as by various synthetic inhibitors of MMP activity. We therefore cultured primary stroma-stage anterior eyes (defined as corneas without lens, with a surrounding margin of periocular mesenchyme and its underlying pigmented and neural epithelium), explanted at 4.5–5 days incubation, in the presence of exogenous TIMP-2 (Fig. 3C,D). Other cultures were incubated with the synthetic MMP inhibitors BB-2516 (Marimastat; not shown) or GM-6001 (Ilomastat; Fig. 4E,F). Cultures containing any one of these three inhibitors behaved similarly in that the corneal stromas retained strong immunoreactivity for collagen type IX as well as collagen II, failed to swell and remained largely, if not completely, acellular (Fig. 3C,D). Control cultures showed a marked or complete reduction of immunoreactivity for collagen type IX (Fig. 3B) but no change for collagen II (Fig. 3A); they also became swollen and invaded by cells (Fig. 3A,B). Thus, an MMP activity that is sensitive to TIMP-2 (and to synthetic MMP inhibitors), most likely MT3-MMP, is essential for these developmental events to occur.

Figure 3.

Effect of tissue inhibitor of metalloproteinase (TIMP) -2 on organ cultures of 4.5- to 5-day anterior eyes incubated for 2 days. A–D: Corneas with a margin of adjacent peripheral tissue containing periocular mesenchymal cells were explanted into culture and incubated with standard medium (A,B) or medium supplemented with TIMP-2 (C,D). Sections were reacted with antibodies for collagen types II (A,C) or IX (B,D). A,B: Control. After 2 days of culture in standard medium, the corneal stroma has retained immunoreactivity for collagen type II (A) but lost immunoreactivity for type IX (B). The swollen matrix has become invaded by mesenchymal cells. C,D: TIMP-2. In the presence of TIMP-2, cultures remained compact and acellular and retained immunoreactivity for both collagen types II (C) and IX (D). Scale bar = 50 μm in A (applies to A–D).

Figure 4.

Effect of inhibition of stromal matrix crosslinking on corneal development in culture: mesenchymal cell invasion in the absence of collagen IX proteolysis. A–H: Corneal organ cultures from 4.5- to 5-day chicken embryos were established in normal medium (A,B) or in medium supplemented with β-aminopropionitrile (βAPN, C,D), matrix metalloproteinase (MMP) inhibitor GM-6001 (E,F), or both βAPN and GM-6001 together (G,H). Explants cultured in medium supplemented with βAPN were from embryos treated in ovo with βAPN; all other explants were from embryos treated with Hanks' balanced salt solution. After 2 days of incubation, cultures were terminated and sections were reacted for collagen type II (left column, A,C,E,G) or type IX (right column, B,D,F,H). A,B: Control. Cultures incubated in standard medium exhibited a swollen, invaded stroma, with retention of collagen II (A) and a loss of collagen IX (B). C,D: βAPN. The addition of βAPN to the medium did not interfere with the developmental events. The stroma swelled, and cells invaded a matrix in which collagen II (C) was retained, and collagen IX (D) was lost. E,F: GM-6001. Explants cultured with the MMP inhibitor GM-6001 had a compact stroma that retained both collagen types II (E) and IX (F) and into which cells failed to invade. G,H: βAPN + GM-6001. Cultures incubated in both βAPN and GM-6001 exhibited some minimal degree of stromal swelling that varied in extent within and between individual cultures, and invasion of mesenchymal cells into a stroma that retained collagen type IX (H) as well as collagen II (G). Scale bar = 100 μm in A (applies to A–H).

Corneal Invasion in the Absence of Collagen IX Proteolysis

At least two possible mechanisms, which are not mutually exclusive, could explain the correlation between reduced collagen IX and invasion of the matrix by the periocular mesenchymal cells. In one, the stimulus for migration results from the elimination of physical constraints imposed by interfibrillar cross-bridging by type IX collagen. Thus, the removal of type IX collagen would eliminate interfibrillar cross-bridging, converting a restrictive ECM to one that is permissive for cellular invasion. Alternatively, the stimulus might be chemotactic, with proteolysis generating a cleavage product or cryptic site that serves to stimulate cell migration, as has been described in other systems following proteolysis of, for example, collagen I (Seandel et al., 2001) and laminin (Koshikawa et al., 2000).

We examined these alternatives by inducing a physically altered stromal matrix in the absence of matrix proteolysis. For these studies corneas were explanted from embryos treated with β-aminopropionitrile (βAPN), an agent that inhibits the formation of crosslinks of the type that covalently bind type IX collagen to itself and to collagen type II (Tanzer, 1976; Wu and Eyre, 1984). Inhibition of crosslink formation in these cultures produced a stroma in which collagen IX was present but not covalently bound to other components of the matrix, as demonstrated by its extraction from tissue sections that had not been fixed before processing for immunohistochemistry (not shown). This would ease the physical constraints imposed on the stromal matrix by the cross-bridging molecule (type IX collagen). To test the necessity of proteolytic cleavage for invasion to occur, some of the cultures also received an MMP inhibitor (either TIMP-2 or GM-6001).

To block crosslink formation, the embryos were first treated in ovo with βAPN or carrier (HBSS) at approximately 4 days of incubation, a time when the primary stroma is undergoing rapid assembly (and ∼1.5 days before swelling). Then, at 4.5–5 days of incubation the anterior eyes were removed and cultured (without attached lens) in medium containing either βAPN, or an MMP inhibitor, or both together. The results are presented in Figure 4.2

Invasion of the corneal stroma occurred in all cultures in which the primary stroma was disrupted—either by selective proteolytic removal of the stromal type IX collagen seen in the controls cultured in the absence (Fig. 4A,B) or presence (Fig. 4C,D) of the crosslink inhibitor βAPN or, most importantly, by a reduction of covalent crosslinks in cultures in which proteolysis was also reduced by an inhibitor of MMP activity (Fig. 4G,H). (In such cultures, most of the uncrosslinked type IX collagen was extracted from the tissue during processing for immunohistochemistry unless the tissue was first fixed [either en bloc or on the section] before the application of antibodies.) In this latter group, cells invaded a matrix that retained uncrosslinked collagen type IX. Corneal invasion was blocked only in cultures where normally crosslinked collagen IX was retained due to the inhibition of its proteolysis (Fig. 4E,F). These results indicate that the major function of collagen IX in regulating cellular invasion is to maintain the primary stromal matrix in a compact assembly, through its formation of interfibrillar cross-bridges. Disruption of this conformation can induce stromal matrix modifications that allow cellular invasion.

However, in explants cultured with inhibitors of both crosslink formation and MMP proteolysis, the extent of cellular invasion was usually not as great as that observed in explants cultured without the MMP inhibitors (e.g., compare Fig. 4G,H with Fig. 4A,B). This finding might be due to incomplete inhibition of crosslinking in, for example, matrix that was deposited before the addition of βAPN, to a weak inhibition of migration by the retained collagen IX (see Discussion section), or to a reduction in the extent of matrix swelling (also suggested in a comparison of Fig. 4G,H with Fig. 4C,D). Nevertheless, these results show that cell invasion of the primary stromal matrix can occur in the absence of MMP-mediated proteolysis, indicating that the generation of a stimulatory proteolytic fragment is not essential for invasion.

Involvement of Additional Regulation of Cellular Invasion: Effect of the Lens and Peripheral Tissue

The results described above show that disruption of the primary stromal matrix is necessary for cellular invasion. However, they do not determine whether this event is normally sufficient for invasion to occur. Here, we show that another regulatory event is also necessary.

As described above, explants of 4.5- to 5-day anterior eyes cultured in vitro without the lens undergo all of the subsequent steps of normal development. However, when the lens is included in similar cultures (Zak and Linsenmayer, 1985; see also Bard et al., 1975), development is arrested (i.e., the corneal stroma retains its type IX collagen and remains compact and acellular). This finding implicates either a deficiency in the culture conditions and/or the limited amount of tissue explanted as being responsible for the failure of further development in the presence of the lens. This theory was tested by altering the culture conditions and varying the amount of tissue included in the explants and the age at explantation.

Because previous explants of anterior eyes were cultured on pieces of Nuclepore filter at the air/medium interface, a trivial explanation for the inhibitory effect of the lens could be that the underlying lens, by raising the surface of the cornea, increases the surface tension on the cornea, which might physically prevent stromal swelling and mesenchymal cell invasion. To test this possibility, anterior eyes containing the lens were cultured submerged in medium, thus minimizing surface tension. Such submerged cultures behaved essentially the same as those grown at the air/medium interface. Although submerged explants from which the lens had been removed developed as previously described (e.g., Fig. 4A,B), submerged cultures with the lens (Fig. 5A) still underwent none of these changes.

Figure 5.

Effect of lens and additional peripheral tissue on development of 4.5- to 5-day explants cultured for 2 days. Asterisks indicate the lens. A: Culture of anterior eye + lens without additional peripheral tissue, showing a compact, acellular stroma with retained collagen type IX. B: Culture of “whole globe” (including lens and additional peripheral tissue) reacted for collagen type IX, showing a loss of this collagen from a swollen, but acellular, stroma. C: Culture of “whole globe” reacted for type II collagen, showing the extent of expansion of the stroma that remained acellular. Scale bar = 500 μm in A (applies to A–C).

The need for additional tissue was investigated by establishing cultures that included the lens and substantially more peripheral ocular tissue (termed “whole globes”). In such cultures explanted at 4.5–5 days, further, albeit incomplete, development proceeded. In approximately half of such cultures (7 of 13, in three experiments), the stroma lost type IX collagen (Fig. 5B), retained its collagen II (Fig. 5C) and underwent swelling, but the periocular mesenchymal cells remained at the periphery, leaving the expanded stroma completely acellular (Fig. 5B,C).3

Thus, in such cultures, stromal swelling and mesenchymal cell invasion appear to be independent, the latter not being the inevitable result of the former. Whereas the inhibitory effect of the lens on matrix proteolysis can be overcome by the inclusion of more peripheral tissue, such cultures revealed a persistent inhibition of invasive behavior. This inhibition is not released in vitro in cultures explanted at 4.5–5 days but clearly must be released in ovo at some point before stromal invasion is initiated. To determine when this event occurs and whether complete corneal development can proceed in vitro in the presence of the lens, we examined submerged cultures of explants identical to those just described but initiated from embryos one half day older (5.5 days), a time just preceding stromal swelling and cell invasion. The results are shown in Figure 6. In such cultures, all of the corneas analyzed after 2 days incubation (six of six, three experiments) had undergone a complete sequence of development, having stromas that had little, if any, type IX collagen (Fig. 6B) but ample collagen II (Fig. 6C) and were swollen and cellular. This finding indicates that the invasive behavior of the periocular mesenchymal cells becomes enabled immediately before cellular invasion begins.

Figure 6.

Development of 5.5-day explants of “whole globes” incubated for 2 days in organ culture. Asterisks indicate the lens. A: Unincubated anterior eye embedded at 5.5 days, reacted for type IX collagen. The primary stroma has not yet swollen, is acellular, and is immunoreactive for collagen IX. B: Explant of 5.5 day “whole globe” cultured for 2 days and then reacted for type IX collagen. Immunoreactivity for collagen IX is much diminished in the corneal stroma, which has swollen and become invaded by cells. C: A similar section of a cultured 5.5 day “whole globe,” reacted for collagen type II, showing the retention of this molecule, and the extent of the swollen, invaded stroma. Scale bar = 500 μm in A (applies to A–C).

As a control, to ascertain whether the cultured 5.5 day eyes were indeed explanted before the onset of stromal invasion, we examined corneal sections from contralateral eyes and from other embryos embedded at the same time the cultures were initiated. Of the seven corneas analyzed (each from a different embryo), six had a compact, acellular primary stroma that was immunoreactive for collagen type IX (Fig. 6A). The seventh lacked type IX collagen and showed cells just beginning to invade the stromal periphery (not shown). Thus, it is likely that most, if not all, of the 5.5–day cultures were explanted before overt stromal invasion had begun.


These results suggest that MMP activity plays a major role in driving three sequential events in early corneal development: (1) the loss of collagen type IX, which facilitates (2) stromal swelling, which in turn, is necessary for (3) mesenchymal cell invasion. Other observations suggest that the removal of the type IX collagen requires the action of multiple MMPs, and still others indicate that at least one independent event is required for cellular invasion to occur—the activation of motility itself.

Functionally, treatment of primary stroma-stage corneal explants with TIMP-2, or with a synthetic inhibitor of MMP activity, can completely block all three events. Moreover, the selective effect of TIMP-2, vs. the less-effective TIMP-1, suggests that the critical activity may be that of MT3-MMP, found by RT-PCR to be expressed in early corneas and reported to be substantially more sensitive to inhibition by TIMP-2 than by TIMP-1 (Matsumoto et al., 1997). Further evidence for proteolytic activity of an MT-MMP comes from our previously published zymograms (Fitch et al., 1998). Studies on a variety of mammalian systems have shown that one role in vivo of most of the MT-MMPs (Will et al., 1996; Seiki et al., 2003), including MT3-MMP (Kang et al., 2000), is to proteolytically activate MMP-2. Indeed, a common assay for demonstrating MT-MMP activity uses this activating function as revealed by zymography (Seiki, 1999). In our previous study, zymograms of corneas taken before invasion begins showed only the latent form of MMP-2; the active form of the enzyme was detected only in corneas undergoing invasion. The timing of this activation of MMP-2 indicates that the acquisition of the MT3-MMP correlates with the onset of stromal swelling and invasion.

We also found that MT3-MMP, acting in conjunction with MMP-2, might be directly involved in the selective removal of collagen IX from the primary stroma. For this determination, we used an assay in which unfixed tissue sections of primary stroma-stage corneas were digested with purified enzymes—in this case constitutively active recombinant MT3-MMP and MMP-2—and then examined by immunofluorescence histochemistry for the removal of specific components. This assay allows an assessment of the sensitivity to proteolytic degradation of ECM molecules as they occur in their native conformation and supramolecular organization within tissues (Fitch et al., 1988a). In the present study, neither MT3-MMP nor MMP-2 alone caused an appreciable loss of immunoreactivity for collagen type IX, but both enzymes used together essentially abolished the signal for type IX collagen in the cornea. In contrast, the signals for the fibrillar collagen types I and II, which in the cornea occur together (Hendrix et al., 1982; Linsenmayer et al., 1990), were not appreciably diminished. Thus, the combined action of MMP-2 and MT3-MMP appears to be sufficient for the selective removal of collagen type IX. These results suggest that CMMP, the other MMP expressed in early corneas (but not tested here), is not essential for this task.

We do not yet know whether both enzymes act directly on the type IX collagen molecule, perhaps cleaving it at opposite ends to extricate it from the matrix, or whether one of the enzymes might digest a closely apposed molecule, thus exposing cleavage sites on collagen IX to the action of the other enzyme. The degradation products that are subsequently extruded into the anterior chamber of the eye, however, are large enough to retain immunoreactivity with at least one of the monoclonal antibodies against collagen IX (Fitch et al., 1998).

Both MMP-2 and MT3-MMP have been reported also to cleave certain fibrillar collagens (e.g., type I collagen) in solution or in three-dimensional matrices (Aimes and Quigley, 1995; Matsumoto et al., 1997; Seiki, 1999). In the early cornea, however, collagen IX is removed selectively, while fibrillar collagen types I and II in the stroma remain present. To explain this finding, a variety of mechanisms can be envisioned. For example, the coassembly of collagens I and II in the corneal stroma into heterotypic fibrils (Hendrix et al., 1982; Linsenmayer et al., 1990) may lower the sensitivity of each to proteolysis; alternatively (or in addition), collagen IX might be the preferred substrate due either to its cleavage sites being inherently more sensitive than those of the fibrillar collagens, or to its more exposed location at the surfaces of fibrils.

In this enzyme assay, as a positive control for the activities of MMP-2 and MT3-MMP, we also monitored FN, an ECM component reported to be a substrate for a wide variety of MMPs (Corcoran et al., 1996; Seiki, 1999). We found that both enzymes independently degraded FN. Indeed, in ovo, immunoreactivity for FN does fall somewhat when the primary corneal stroma swells and becomes invaded (Kurkinen et al., 1979; Doane et al., 1996; and unpublished observations), but this decline of immunoreactivity for fibronectin is notably slower than that for collagen IX. The reason for this difference is likely due to replacement of the degraded FN by stromal cells, which appear to synthesize FN during their invasion (Fitch et al., 1991).

Various studies (Will et al., 1996; Kang et al., 2000; Seiki et al., 2003) have demonstrated that MMP-2 activation occurs at the cell surface in a complex that includes, along with MMP-2, an MT-MMP (in this case, presumably MT3-MMP) and TIMP-2. This finding could explain our paradoxical evidence (unpublished) by RT-PCR for expression of TIMP-2 (shown here to inhibit invasion) in corneas that are undergoing swelling and invasion.

Whether proteolysis of collagen IX is mediated at the cell surface or by soluble enzymes released into the acellular stroma (or both) is as yet unclear. Previous observations (Fitch et al., 1988b) have shown that the loss of immunoreactivity for collagen IX occurs in ovo at the same time as, or shortly before, stromal swelling and mesenchymal cell invasion, and our current observations on the cultured “whole globes” that undergo swelling but remain acellular suggest that collagen IX proteolysis is not a cell surface event. This finding is difficult to reconcile with a mechanism driven solely by membrane-bound enzymes, as seems to be the case in a variety of mammalian systems (Hotary et al., 2000; Koike et al., 2002). It is possible, however, that type IX collagen proteolysis is mediated by soluble MMP-2, released from the cell surface into a TIMP-2–poor matrix after its activation by membrane-bound MT3-MMP (Kang et al., 2000), and joined by a secreted soluble isoform of MT3-MMP, generated by alternative splicing as described previously for the mammalian enzyme (Matsumoto et al., 1997; Shofuda et al., 1997).

Inhibition of covalent crosslink formation was sufficient to allow invasion into corneal cultures, which occurs even when type IX collagen removal is also blocked by MMP inhibitors. Therefore, the critical feature of a stromal matrix receptive to cellular invasion appears not to require the proteolysis of type IX collagen per se; rather, the important modification is the disruption of interfibrillar stabilization, allowing the cells to invade a more penetrable stromal matrix. Mechanistically, there are two implications of this result. First, mesenchymal cell invasion of the cornea may not require a stimulatory peptide generated by MMP activity, possibly acting as a chemoattractant, as reported for a variety of mammalian systems (e.g., Koshikawa et al., 2000; Seandel et al., 2001). Second, corneal invasion can occur into a matrix in which collagen type IX is retained. This observation appears to differ from the findings of others who report that tissues containing type IX collagen are inhibitory to invasion by neural crest cells and to neurite extension (Ring et al., 1995, 1996), functions that were attributed to its covalently bound chondroitin sulfate glycosaminoglycan (GAG). However, recent observations (unpublished) of primary stromas of 5-day anterior eyes exhibit little, if any, immunoreactivity for this GAG chain.4 It is presently unclear whether this apparent difference in type IX collagen in the cornea is analogous to the differences in the extent or type of glycosaminoglycan modification of collagen IX reported previously in cartilages (Ayad et al., 1991; Yada et al., 1992).

Finally, modification of the compact primary stroma does not seem to be the only requirement for corneal invasion; there also appears to exist an event that activates cell motility. For this activation, the lens likely plays a regulatory role, as it does in the development of the avian anterior eye (Zinn, 1970; Beebe and Coats, 2000). In this case, the lens is likely inhibitory. Whereas anterior eyes with attached lens explanted at 4.5–5 days invariably failed to develop further, the inclusion of more peripheral tissue (“whole globes”) allowed some of the cultures to lose collagen IX and undergo swelling, but nevertheless remain acellular. This finding suggests that matrix modification alone does not invariably result in invasion, which likely was inhibited by the lens.

This inhibition of migration is, of course, transient in ovo, and this is the case also in vitro. In cultures of “whole globes” explanted from slightly older embryos (just before the onset of migration), stromal invasion does take place even in the presence of the lens. One explanation for this finding is that the lens in vitro maintains an inhibitory effect on the activation of the migratory phenotype that in ovo is lost just before the onset of invasion. The nature of such an activation event and the mechanism of its inhibition by the lens is presently unknown but could involve integrin switching or activation (Hynes, 2002), or other cell surface events, and/or perhaps an intracellular reorganization of the cytoskeleton essential for cell motility (Raftopoulou and Hall, 2004). It is evident, at any rate, that there exists a multiplicity of controls of mesenchymal invasive behavior in the developing cornea, occurring at the levels of the cells themselves as well as the extracellular matrix they invade.



White Leghorn chicken eggs were obtained from Spafas (North Franklin, CT) and incubated at 38°C. Embryos were staged according to Hamburger and Hamilton (Hamburger and Hamilton, 1951). Various ocular tissues were removed for organ culture, embedded for immunohistochemistry, and processed for RNA extraction and analysis by RT-PCR.

Analysis of mRNA Expression by RT-PCR

RT-PCR was performed and analyzed as described previously (Fitch et al., 1995), using primers chosen from the published sequences for avian MMP-2 (Aimes et al., 1994), MT3-MMP (Yang et al., 1996), putative MMP-9 (Hahn-Dantona et al., 2000), CMMP (Yang and Kurkinen, 1998), MMP-13 (Lei et al., 1999), and collagen type IX (Ninomiya and Olsen, 1984; Nishimura et al., 1989). For MMP-2, the forward primer covered nucleotides #1262-1281 (CTATGATGATGACCGCAAGT);the reverse primer covered #1663-1681 (GGCCAGAATGTAGCAACGA). MT3-MMP was amplified by primers that annealed to #646-663 (GGCAAGAGAGACGTGGAT) and #1263-1282 (GCCCTCTCCAGAAGTAGGTA). Primers for avian MMP-9 were for #367-376 (TACCGGGTGATGACCTACTC) in the forward direction and covered #645-662 (CGGGTCTTCACCTCTAAGC) in the reverse direction. For CMMP, the forward primer covered #561-580 (GGACATTGCCCTCGCTATTT) and the reverse primer for #1113-1131 was TTATATCCGCTGACGACCCA. Primers for MMP-13 were for #782-801 (TTTGGGCTATGAATGGCTAT) and for #1098-1117 (TAGTATGCAGGATGCGGACA). The primers for collagen type IX were as described previously (Fitch et al., 1995). Messenger RNA (poly A +) from 7-day embryonic corneal tissue, prepared as previously described (Fitch et al., 1995), was used as a template for amplification, and the results were compared with those obtained with mRNA isolated from sternal cartilage from 14 day embryos, used as a control.

Digestion of ECM Components in Sections of Anterior Eyes by Recombinant MMPs

Eight-micrometer cryostat sections of unfixed anterior eyes were applied to 12-spot glass slides (see below) and then digested in a closed humid chamber for 18–22 hr at 37°C with various concentrations of recombinant MMP-2 (PF-023, Calbiochem), MT3-MMP (catalytic domain, #475939, Calbiochem) or both enzymes together. Recombinant human MMP-2 and MT3-MMP were purchased (Calbiochem) as active enzymes (MMP-2 as an activated zymogen; MT3-MMP as the constitutively active catalytic domain). They were solubilized in digestion buffer (50 mM Tris, 150mM NaCl, 5mM CaCl2) at 100 μg/ml. This stock solution was then diluted to 1 μg/ml, 0.1 μg/ml, and 0.01 μg/ml with digestion buffer, and 20 μl of each was applied to sections mounted on spot slides. Sections treated with both enzymes in concert received a 1:1 combination (by volume) of each enzyme at each dilution. Slides were incubated overnight (18–22 hr) at 37°C, and the reactions were terminated by immersion in phosphate-buffered saline (PBS) washes (3). Because observations from early experiments indicated that the 0.1 μg/ml concentration was the lowest concentration that digested the “control” substrate (fibronectin) within the overnight time period, later experiments used the 0.1 μg/ml concentration only. This corresponded to a molarity of approximately 4.3 μM for MT3-MMP and 1.6 μM for MMP-2. After digestion, the slides were rinsed in PBS and analyzed by immunohistochemistry for the presence of collagen types II and IX (and X as a negative control), and for fibronectin, using the monoclonal antibodies and protocol described below.

Organ Culture

Embryos were removed after 4.5–5 days or 5.5 days of incubation, rinsed in Hanks' balanced saline solution (HBSS), and staged according to Hamburger and Hamilton (Hamburger and Hamilton, 1951). Anterior eyes, defined as the cornea with a surrounding ∼4 mm margin of periocular mesenchymal tissue and its underlying pigmented epithelium, were explanted, with or without the attached lens, into organ culture, and oriented endothelium/lens-side down on a polycarbonate membrane filter support (Nuclepore or Millipore). They were fed with Dulbecco's high-glucose modified Eagle's medium (DMEM) or with a mixture (1:1) of DMEM and medium F12 (both from Gibco/BRL). The media were supplemented with 1–5% fetal calf serum (HyClone) plus penicillin and streptomycin (100 U/ml and 100 μg/ml) and ascorbic acid (50 μg/ml, all from Gibco/BRL). Initially, such cultures were maintained in organ culture dishes (Falcon) as described previously (Cai et al., 1994); in later experiments, for convenience, most of the cultures were maintained in either four-chamber culture slides (Falcon) or on transwell supports in 12-well tissue culture dishes (Costar). (The explants developed equally well in all of these systems.) Medium was added to the cultures to a level that either just covered the upper surface of the explant or to a level that submerged the explant, as described in the Results section, and was changed every 1–2 days as needed.

For some experiments, more extensive portions of 4.5–5 day ocular tissue were explanted into submerged organ cultures. The largest of these explants, termed “whole globes” and described in the Results section, included all of the tissues of the anterior eyes (described above) plus most of the remainder of the ocular globe, except for a ∼5- to 10-mm-diameter hole in the posterior end (around the optic nerve) extirpated to facilitate fluid exchange. These explants were placed in chamber slides completely immersed in medium and were fed daily.

Some explants were cultured in the presence of recombinant or synthetic MMP inhibitors, including TIMP-1 (Calbiochem, 50 nM and 400 nM, in HBSS), TIMP-2 (Calbiochem, 50 nM, in HBSS), GM-6001 (Galardin, or Ilomastat, Calbiochem, 20μM in HBSS), or BB-2516 (Marimastat, Calbiochem, 5μM, in DMSO/HBSS).

Analysis of collagen degradation in ocular organ cultures was performed by immunohistochemistry, using collagen-specific monoclonal antibodies, as described previously (Fitch et al., 1988a). For this investigation, unfixed anterior eyes (4.5–5 days or 5.5 days) or 2–3 day organ cultures (i.e., 7–8 days postincubation) were soaked in 7–8% sucrose in PBS for 5–20 min, embedded in OCT (Miles Laboratories, Elkhart, IN), and frozen in liquid nitrogen. The blocks were stored at −20°C until sectioned. The 8- to 10-μm frozen sections were mounted on 12-spot slides (Thermo Electron Corp., Pittsburgh, PA) previously coated with Biobond adhesive (Electron Microscopy Sciences, Ft. Washington, PA), dried for 4–24 hr, and stored at −20°C.

In some experiments, to disrupt the dense interfibrillar packing of the primary stroma, collagen crosslink formation was inhibited. For this investigation, embryos were first treated in ovo at 4 days with the lathyrogen β-aminopropionitrile (βAPN; Aldrich Chemical Company, Inc., Milwaukee, WI), 10 mg/ml, in HBSS, 0.1 ml/egg. Controls received HBSS alone. Anterior eyes were then explanted, without the lens, into organ culture at 4.5–5 days as described above. Cultures in the lathyritic group received medium supplemented with βAPN, 50 μg/ml, with or without an MMP inhibitor (either TIMP-2 or GM-6001, as described above). Controls received standard medium without βAPN but with or without TIMP-2 or GM-6001. Explants were removed after 2–3 days of culture, and rapidly embedded in OCT for analysis by immunohistochemistry. Because of the possibility of extraction of uncrosslinked matrix components from sections of lathyritic eyes during processing (Gross, 1974), sections were routinely fixed on the slide with Histochoice fixative as described below.

Immunofluorescence Histochemistry

The pattern of anti-collagen immunoreactivity in 8-μm frozen sections was revealed using a modification of an indirect immunofluorescence procedure described previously (Fitch et al., 1982). Sections were blocked with 1% bovine serum albumin (BSA) or 5% powdered milk + goat IgG (Sigma; 100 μg/ml in PBS) for 5–20 min and then covered with a drop of the primary antibody for 1–4 hr at room temperature or overnight at 4°C. Some sections from unfixed tissue were first fixed on the slide for 10 min in Histochoice (Amresco, Solon, OH), a noncrosslinking fixative that preserves epitopes on immobilized tissue components, before being incubated in primary antibody. Sections were then washed thoroughly with PBS and incubated with a rhodamine-conjugated goat anti-mouse IgG second antibody (1 hr, room temperature), washed in PBS, and mounted in glycerol/PBS (95:5) containing Hoechst 33258 dye (Sigma, 1 μg/ml) to stain nuclei.

In some experiments, unfixed sections were extracted on the slide by immersing them in 1 M NaCl in PBS for 30 min, at room temperature. They were then washed briefly in PBS, blocked, and reacted with antibody as described.


Monoclonal antibodies against FN (VA13 and B3/D6; see Gardner and Fambrough, 1983) and collagen type I (DD4; see Swasdison et al., 1992), type II (II6B3; see Linsenmayer and Hendrix, 1980), and type IX (2C2 and 4D6; see Irwin et al., 1985) were used to identify FN or collagen type-specific epitopes in tissue sections. The antibodies against FN and type IX collagen were purchased from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA). Antibody IX-2C2 (termed “anti–IX-L/S”) reacts with both the long and short isoforms of collagen IX. Its epitope is located at the carboxy terminus of the HMW fraction of pepsin-digested collagen IX (Irwin et al., 1985), in the NC2 domain probably near its junction with the Col2 domain. Antibody IX-4D6 (termed “anti–IX-L”) is specific for the long form of the molecule (Irwin et al., 1985; Brewton et al., 1991). It reacts with the amino terminus of HMW in the long form-specific NC4 domain, near the junction between Col 3 and NC4. I-DD4 and II-II6B3 were characterized previously as having epitopes in the triple-helical domain of their respective molecules. Finally, sections used for negative controls were reacted with a monoclonal antibody against type X collagen (Schmid and Linsenmayer, 1985), which is not found in early ocular tissue. The antibodies were used in the form of either undiluted supernatant from spent hybridoma cultures or ascites fluid diluted 1/200–1/500 (10–20 μg/ml) with PBS containing 1% BSA.


The monoclonal antibodies against type IX collagen (2C2 and 4D6) and fibronectin (VA13 and B3/D6) obtained from the Developmental Studies Hybridoma Bank were developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences (Iowa City, IA). We thank Eileen Gibney for her expert technical assistance.

  • 1

    CMMP is relatively uncharacterized but has been reported to cleave type I collagen and gelatin [Yang and Kurkinen, 1998]. Because its sensitivity to the TIMPs is unknown, and it is not available as a recombinant protein, it was not evaluated further.

  • 2

    In many but not all organ cultures, the corneal epithelium became hypertrophic and noticeably thickened. This hypertrophic epithelium, especially evident in Figure 4C–H, can secrete a secondary matrix within the epithelial layer that contains type II collagen (see Fig. 4C,E,G) but little if any type IX (Fig. 4D,F,H). This finding can also be seen in Figure 6C.

  • 3

    The remaining explants in this group [not shown] behaved similarly to the anterior eyes + lens with the “normal” amount of peripheral tissue [described above], in that their stromas retained collagen type IX immunoreactivity and remained compact and acellular. Cultures of “whole globes” from which the lens was removed before explantation showed that invasion into a swollen matrix had proceeded (not shown), similar to that observed in explants of anterior eyes without the lens [e.g., Fig. 4B].

  • 4

    As a control, the ocular vitreous humor, which contains collagen IX, known to have an attached GAG chain [Yada et al., 1990], and the ciliary epithelial cells both showed strong immunoreactivity for this specific GAGmoiety.