Regulation of Eye Size by the Retinal Basement Membrane and Vitreous Body

Abstract

purpose. Congenital high myopia is an early-onset enlargement of the eye globes that carries a high risk for retinal detachment. The genetic basis for congenital high myopia has frequently been connected to mutations in genes encoding extracellular matrix proteins of the vitreous body (VB) and the inner limiting membrane (ILM). Why defective or missing VB and ILM proteins lead to an increase in eye size is unknown. The present study introduces the chick embryo as a model to study the role of ILM and VB in regulating eye size.

methods. The ILM and VB were disrupted by injecting collagenase into the eyes of E5 chick embryos. The digestion of VB and ILM proteins was monitored by Western blot and immunocytochemistry. Eye size was assessed up to 9 days after the enzyme injections.

results. Intraocular injection of collagenase led to the disruption of the ILM and the VB by digesting their collagen constituents. Once disrupted, the ILM and the collagen II fibrillar network failed to regenerate despite continued synthesis of VB and ILM proteins. ILM and VB disruption resulted in eye enlargement of 50% within 4 days. The increase in eye size was greatly reduced by reconstituting the ILM.

conclusions. The present data show that the ILM and the VB play major roles in the early regulation of eye size. The authors speculate that the integrity of the vitreoretinal border is an important factor in preventing congenital high myopia.

The vertebrate eye includes several structural compartments that are particularly rich in extracellular matrix (ECM) proteins, such as the cornea, sclera, and vitreous body and the basement membranes of the lens, retina, and pigment epithelium. There is a wealth of literature showing that the ECM plays a dominant role in regulating eye size. By using blurring or defocusing goggles in monkeys, chicks, and tree shrews, it was found that negative defocus is a driving force for changes in eye size. ^{1 2} The size changes occurred rapidly and were correlated with degradation or de novo synthesis of ECM proteins in the sclera. ^{3 4 5 6} The current hypothesis states that negative defocus and the expanding or shrinking scleral ECM are responsible for refraction-dependent myopia. ^{2 7}  

Several syndromic human diseases that feature congenital high myopia result from mutations in genes that encode ECM proteins that are prominent in the vitreous and inner limiting membrane (ILM), the basement membrane at the vitreoretinal border, but at best are only minor components in the sclera. In Stickler and Wagner syndromes, for example, the vitreous is in a liquid rather than a gelatinous state, because of dominant mutations of collagen II and V/XI, ⁸ and in Knobloch syndrome, caused by mutations of the basement membrane protein collagen XVIII, an impaired ILM results in early posterior vitreous detachment. ⁹ Common to all these mutations is early-onset high myopia combined with a high risk of retinal detachment. The early onset and the defects in the vitreous rather than the scleral ECM in many of the congenital high myopia cases are not easily explained by the current sclera-myopia model, but point to a function of the ECM of the vitreous body (VB) and ILM in regulating early eye size. 

In the present study, we show that the experimental disruption of the ILM and VB in early chick embryos leads to eye enlargement at a stage when the sclera is still forming, and the retina does not process visual information. Previous studies in this laboratory have already implicated the ILM as an important factor in retinal histogenesis and ganglion cell survival. ^{10 11 12} The present data show that the ILM and VB extracellular matrices are not only important for proper retinal development, but also have an important function in the early regulation of eye size. 

Experimental Procedures

Antibodies

The antibodies to chick laminin-1, nidogen-1, perlecan, collagen XVIII, collagen II, fibrillin, agrin, and collagen IX have been described previously. ¹³ The antibodies are mouse monoclonals (mAbs) and are available from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City). mAb 3F1 to collagen II, mAb 2B9 to collagen IX, and mAb 3H11 to laminin are IgA, IgM, and IgGs respectively, and the proteins were colocalized to determine overlapping or nonoverlapping distributions of the proteins. The mAb to tenascin (mAb M1) was obtained from the Development Studies Hybridoma Bank. Polyclonal antisera to collagen IV, collagen II, fibronectin, and laminin-1 were obtained from Rockland Immunochemicals (Gilbertsville, PA), Chemicon (Temecula, CA), and Invitrogen (Carlsbad, CA). The rabbit antiserum to perlecan was described previously. ¹⁴ A nuclear counterstain was used (SYTOXgreen; Invitrogen, Eugene, OR). 

Disruption and Reconstitution of the ILM

The ILMs and VBs were disrupted by injecting 0.5 μL of 50 U/mL (∼30 μg/mL) collagenase (type VII; cat. no. C0773; Sigma-Aldrich. St. Louis, MO) into embryonic day (E)5 chick eyes ^{10 11} (n = 60). Usually, the right eye of embryos was injected, and the left eye served as the control. However, identical data were obtained by injecting the left eye and using the right eye as the control. For ILM regeneration experiments, collagenase was injected as described, followed by a second injection of 2 μL of mouse laminin-1 (Invitrogen-Gibco, Grand Island, NY) and α2-macroglobulin at 1 mg/mL each ¹¹ (Sigma-Aldrich; n = 30) 10 hours later. The embryos were killed 7 hours to 10 days after the injections, and the heads were fixed in 4% paraformaldehyde and cryoprotected in 25% sucrose. Twenty-micrometer cryostat sections through the experimental and the contralateral control eyes of each embryo were labeled with mAbs to collagen II, IX, fibrillin, fibronectin, and tenascin for detection of the fibrillar structure of the VB. Antibody staining for laminin-1, nidogen-1, collagen XVIII, and agrin was used to determine the presence or absence of the ILM. For control purposes, eyes were also injected with 1 μL of each PBS, Proteus vulgaris chondroitinase ABC (cat. no. C3667; 50 U/mL; Sigma-Aldrich) or testicular hyaluronidase (cat. no. 106500; 1 mg/mL; Roche Diagnostics, Indianapolis, IN). 

The possibility of leakage of peptides injected into the vitreous cavity was tested by localizing a 45-kDa agrin peptide that was injected into the VB at E5. The peptide included an myc tag for immunohistochemical detection. ¹⁵ Embryos injected with the peptide were killed 1 to 2 days after intraocular injection of 2 μL of 100 μg/mL peptide, and cryostat sections through the eyes were labeled with an anti-c-Myc mAb (1:500; Santa Cruz Biotechnology, Santa Cruz, CA) followed by a Cy3-labeled goat anti-mouse secondary antibody (Jackson ImmunoResearch, West Grove, PA). A second test used IgG to laminin-1, secreted by the 3H11 hybridoma cell line, ¹⁶ which was injected into E5 eyes. The distribution of the antibody was detected in cryostat sections 1 to 3 days after injection by labeling with a Cy-3-labeled goat anti-mouse IgG (Jackson ImmunoResearch). A third test for leakage involved two biotin-based tracers (N-hydroxysuccinimidobiotin; biotin-NHS; and 6-(biotinamidocaproylamido)capronic acid N-hydroxysuccinimide ester; biotin-X-NHS; both from Sigma-Aldrich) that indiscriminately and covalently label proteins. The tracers were injected at 1 mg/mL in dimethyl sulfoxide (DMSO) into eyes of E5 embryos. The labeled proteins were detected 1 to 2 days after injection by incubating cross-sections of paraformaldehyde-fixed eyes with Cy3-labeled streptavidin (Jackson ImmunoResearch) for 1 hour. 

Histology

Heads and retinas from chick embryos were fixed in 4% paraformaldehyde and processed for cryostat-sectioning and immunostaining as described. ¹¹ For double-labeling of collagen II (IgA) and collagen IX (IgM) as well as for collagen II and laminin-1 (IgG) α- (Sigma-Aldrich), μ- and γ-specific secondary antibodies (Jackson ImmunoResearch) were used. The specimens were mounted in 90% glycerol and examined with an epifluorescence or a confocal microscope (Flowview; Olympus, Lake Success, NY). 

In Vitro Collagenase Digestion and Western Blot

VBs were collected, homogenized and cleared by centrifugation, and the supernatant run out on 3.5% to 15% gradient SDS gels. ^{10 11} Western blot analysis were probed for collagen II, laminin-1, fibronectin, and perlecan using the listed polyclonal antibodies. Collagen IX, fibrillin, and collagen XVIII were detected with the listed mAbs. For in vitro digestion experiments, five E10 chick VBs were incubated with 5 μL of 50 U/mL (∼30 μg/mL) collagenase for various times at 37°C. The samples were brought to 8 M urea and used as samples for SDS-PAGE. For controls, samples were incubated for the same period with PBS instead of enzyme. For semiquantitative measurements, 40 μL of each VB sample was loaded onto the gels. Because the concentrations of individual ECM proteins and the total protein concentration in VB change during development, identical volumes were loaded when comparing experimental and control eyes. The relative abundance of individual proteins was estimated from digitized Western blots by densitometry and measuring the size of the bands, as described previously. ¹³  

Measurement of Eye Size

Eye sizes were compared by weighing of the VBs and by measuring the area of cross-sectional profiles of the eyes. For weighing of the VB from eyes injected with collagenase, the liquefied VB was aspirated with a 1-mL syringe tipped with a 21-gauge hypodermic needle (BD Biosciences, Rutherford, NJ) and transferred into sample tubes. For the control eyes, the eyes were opened and the VB gel transferred with forceps into the sample tubes. At embryonic stages, the chick VB is a compact gel that is readily transferred as one piece. In contrast to the adult VB, the embryonic chick vitreous has no liquid compartment. For cross-sectional area measurements and for determining axial length and horizontal or vertical diameters of the eyes, the heads of the embryos were fixed in 2.5% glutaraldehyde and 0.01% picric acid overnight. To avoid tissue shrinkage, the heads were cryostat sectioned without cryoprotection along the nasotemporal or the dorsoventral axis. The heads were sectioned up to the center of lenses of both eyes and the frozen tissue blocks photographed with a dissecting microscope, similar to the method described by Pickett-Seltner et al. ¹⁷ Cross-sectional areas and eye length or diameters were measured from the digital images by using a drawing tablet (Sigma-Aldrich Scan). Eyes with intravitreal bleeding and eyes that became microphthalmic after injury during the injection were excluded from the evaluation. Results were expressed as mean ±SEM. The data were analyzed by Student’s t-test, and P < 0.05 was considered significant. 

Results

Protein Composition of the Embryonic Chick VB

The proteins of the embryonic chick VB included several collagen family members, such as collagen II and IX (Fig. 1b) ; a series of large extracellular matrix glycoproteins, such as fibrillin (Figs. 1a 1b) , fibronectin, and tenascin (Fig. 1b) ; and several proteins typically found in serum, such as albumin and transferrin. ¹³ Collagen II, collagen IX, fibrillin, tenascin, and fibronectin were organized in fibrillar networks that filled the VB chamber (Fig. 1a) . The VB ECM proteins were particularly abundant at the vitreoretinal border (shown in Figs. 1a 2b 2d ; 4d, 4e; 5a). Western blot analysis (Fig. 1b)confirmed that the proteins had the expected molecular masses of 220 and 440 kDa for monomeric and dimeric fibronectin, 350 kDa for fibrillin, and multiple bands of tenascin with the 200-kDa monomer. Collagen II appeared in a complex banding pattern around 150 kDa. The bands represent unprocessed, half-processed, and fully processed protein. ^{18 19} Chick VB also included a splice variant of collagen II that lacks exon II and that also exists in unprocessed, semiprocessed, and fully processed versions. ¹⁸ Chick VB collagen IX is a proteoglycan that appeared in blots as a smear of more than 700 kDa because of the heterogeneity of the glycosaminoglycan chain attached to the core protein. ²⁰ In addition, the embryonic VB contained all ILM proteins known at present, such as laminin, agrin, nidogen, perlecan, and collagens IV and XVIII. ¹³  

Disruption of the ILM and VB in the Embryonic Chick Eye

The ILM and VB in chick embryos were dissolved by injecting ∼15 ng of clostridial collagenase into the vitreous of E5 eyes (n = 60 ^{10 11 12} ). Liquefaction of the VB and the disruption of the ILM occurred within 5 hours of enzyme injection. As shown in Figures 2and 3 , the collagenous proteins of VB and ILM of the injected eyes were digested, and no longer detectable (compare Figs. 2a and 2b ), whereas all noncollagenous proteins, such as fibrillin, tenascin, and fibronectin were still present (Fig. 2 , compare 2c and 2d). In vitro digestion experiments followed by Western blot analysis confirmed that the collagenase digested the collagen constituents of the VB, whereas the noncollagenous components were left intact (Fig. 2e ; compare control samples, con, with collagenase-treated samples, coll). The time course of collagen digestion showed that soluble collagen II and XVIII underwent extensive proteolytic degradation within 30 minutes of enzyme incubation. The highly glycosylated collagen IX proteoglycan was much more resistant to the enzyme, and within a 3-hour digestion period, neither a decline in staining intensity nor a shift of the collagen IX smear to lower molecular weights was detectable (Fig. 3) . A complete digestion of collagen IX required approximately 7 hours of incubation (Fig. 2e) . As shown above, the noncollagenous proteins of the ILM and VB were unaffected by the enzyme (Fig. 2e) . 

Once the ILM and VB had been enzymatically disrupted, the eyes were unable to reassemble the missing ILM and to establish a new VB gel ^{10 11 12} (n = 60). Accordingly, the ILM and a fibrillar network of collagen II and IX were undetectable in the vitreous chamber even 4 days after enzyme injection (Figs. 4a 4b 4c ; compare with the control retinas shown in Figs. 4d 4e 4f ). The distribution and staining intensity of the scleral extracellular matrix proteins were not affected by intravitreal collagenase injection and appeared the same in experimental and control eyes. Furthermore, as in control eyes, there was a continuous increase in scleral collagen II and IX labeling between 1 and 4 days after collagenase injection (Figs. 4a 4b ; 5). The avian sclera is mainly a cartilaginous structure with collagen II and IX as its main ECM components, and it has only a thin outer fibrous layer, whereas the sclera in mammals is a dense connective tissue with collagen I as its main component and little if any collagen II and IX. Although collagen II and IX were detectable from day 5 onward, a defined scleral cartilage was present in both control and experimental chick eyes only from day 9 onward (compare SC in Figs. 4a 4bwith 4d, 4e, and 5). It is of note that discontinuities in the scleral cartilage reflect a normal process of scleral cartilage formation, and these discontinuities were not noticeably different in experimental and control eyes. 

The failure to reconstitute the ILM and the fibrillar network of the VB after collagenase treatment was not due to the inability of the eyes to synthesize ILM and VB proteins. Western blot analysis showed that the ILM and VB collagens reappeared within 2 days after collagenase injection. As shown in Figure 4g 4days after enzyme treatment, the concentrations of soluble collagen II (Coll2; lanes 1, 2), collagen IX (Coll9; lanes 1, 2) and laminin (LN-1; lanes 1, 2) in VB were similar in experimental and control eyes (lanes C in Fig. 4g ). The fact that ILM and VB proteins were readily detectable in Western blots but undetectable by immunocytochemistry showed that soluble ECM proteins were washed out during the labeling procedure and that these proteins were only detectable in tissue sections when they were networked into an insoluble, fibrillar VB or an insoluble basement membrane. 

ILM-Dependent Deposition of Collagen II

As shown previously, injecting mouse laminin-1 into the collagenase-treated eyes induced the reassembly of a new ILM. ^{11 12} We found by double-labeling with antibodies to laminin and collagen II that the reassembly of the ILM also facilitated the deposition of collagen II. In control E6 eyes (Fig. 5a) , a high density of collagen II fibrils was detected in the VB along the laminin-positive ILM. In contrast, in E6 eyes that were injected with collagenase at E5, collagen II and laminin were no longer detectable (Fig. 5b) . The ILM reappeared within 6 hours when the collagenase was chased with an intraocular injection of mouse laminin-1 ^{11 12} (n = 30). The new ILM assembled 50 μm deep to the vitreous surface (Fig. 5c)and remained at this ectopic position throughout further development (Fig. 5e) . The ectopic location of the reconstituted ILM was due to the retraction of the neuroepithelial end feet, which carry the cellular receptors for ILM assembly. ^{11 12} The neuroepithelial cells are the precursor cells that give rise to neurons and glia cells in the retina. No collagen II fibrillogenesis was detectable within 24 hours after ILM reconstitution (Fig. 5c) . However, significant deposition was detected 2 days after collagenase–laminin injections (Fig. 5d) , and there was a high degree of overlap of laminin and collagen II labeling along the reconstituted ILM 4 days after injection (Fig. 5e) . There was very little collagen II deposition without ILM reconstitution (Fig. 5f) . Despite extensive collagen II fibril formation along the reconstituted ILM, the collagen II network that normally fills the VB chamber was not reestablished, and the vitreous remained liquid. As shown in Figure 4g , the reassembly of the ILM by laminin-1 did not alter the concentration of ECM proteins in the VB, and the concentrations of vitreous collagen II (Coll2; lane 3), collagen IX (Coll9; lane 3), fibrillin, fibronectin, and tenascin (not shown) 4 days after the collagenase–laminin-1 injections were indistinguishable from those of control eyes and from eyes injected with collagenase only (Fig. 4g , lanes C). As expected, the concentration of vitreous laminin-1 was greatly increased (Fig. 4g ; LN-1; lane 3). 

Increase in Eye Size after ILM and VB Disruption

Four days after collagenase injection, the injected eyes were visibly larger than the contralateral control eyes (Fig. 6a) . The size increase occurred whether the collagenase was injected into the right or left eye. The size disparity was quantified by comparing the weight of the VB of experimental and control eyes. Four days after collagenase injection, the weight of the VBs of experimental eyes was approximately 50% greater than the weight of VB from the control eyes (Fig. 6b ; n = 36; P < 0.0001). In contrast, 4 days after control injection of PBS (n = 8), chondroitinase (n = 4), and hyaluronidase (n = 8) into the E5 chick eyes, we found no difference in the size of the experimental and control eyes (Fig. 6b) . To determine the time course of the excessive eye growth, VB weights were determined at different time intervals after collagenase injection (Fig. 6c) . The VB weights of experimental eyes, expressed as the percentage of weights of the control eyes, were approximately 120% at day 2 after collagenase injection (n = 9), 130% at day 3 (n = 10), 150% at day 4 (n = 26) and 160% at day 9 (n = 10) compared with the average weight of VBs of the control eyes. The greatest size increase occurred between days 3 and 4. 

We also compared cross-sectional areas, axial length, and equatorial (i.e., vertical or horizontal) diameters of experimental and control eyes (Fig. 7a) . As measured from horizontal sections 7 days after collagenase injection, the cross-sectional area of experimental eyes was on average 26% greater than that of the contralateral control eyes (Figs. 7a 7c ; n = 10; P < 0.0001). The axial length was on average 19% longer (P < 0.001), whereas the horizontal diameter was only 8% longer (Figs. 7a 7d ; P < 0.004). In every experimental case (n = 10), the increase in axial length was greater than the increase in the horizontal diameter (P < 0.0001; two-tailed, nonpaired t-test). Similar data were obtained from eyes sectioned along the vertical axis (Fig. 7b) : the cross-sectional area in experimental eyes was 29% greater (n = 12; P < 0.0001), the axial length was 16% longer, and the vertical diameter 8% longer. The difference in axial length versus vertical diameter was statistically significant (P = 0.0002). The calculated increase in volume due to the 25% increase in horizontal and vertical cross-sectional areas is close to the 50% volume increase of the VB as measured by comparing the weight of the VB from experimental and control eyes. 

Our data suggest that removal of the ILM and VB reduced the resistance of the eye to intraocular pressure and led thereby to excessive ocular enlargement. Previous experiments have already demonstrated that intraocular pressure plays a role in ocular enlargement during embryonic life. ²¹ According to our hypothesis, the reconstitution of the ILM should prevent the excessive eye growth after collagenase injection. To this end, we compared the sizes of eyes treated by injection of collagenase with those of eyes treated with collagenase followed by laminin-1 12 hours later. As shown in Figure 6d , the weight of VBs from eyes 4 days after the collagenase/laminin-1 injections was on average only 13% greater than that of the controls (n = 15), much less than the 50% increase found in embryos injected with collagenase alone (Fig. 6d ; n = 36). We also injected collagen I into the vitreous of collagenase-treated eyes, trying to reconstitute the VB gel. Injections of collagen I attenuated the extensive eye growth (Fig. 6d ; n = 13), but the injected collagen I did not form a solid VB gel. Co-injection of laminin-1 and collagen I led to eye sizes similar to those injected with laminin alone (Fig. 6d ; n = 15). 

We also established the timing of photoreceptor differentiation for the stages in which the experiments were performed, to rule out visual processing as a cause of excessive eye enlargement (not shown). A histologically separate photoreceptor layer appeared in chick retina from E9 onward, and in situ hybridization has shown that the opsins in chick are expressed from day 14 onward. ²²  

Leakage of Injected Proteins from the Vitreous Cavity to the Periocular Tissues

Regulation of eye size and development of myopia in hatched chicks has previously been associated with changes in the scleral extracellular matrix. ^{3 4 5 6 7} It is conceivable that in our experiments, the 15 ng of intraocularly injected collagenase not only disrupted the ILM and VB but also affected the collagens of the nearby sclera, Bruch’s membrane and the choroid. Therefore, we tested whether proteins smaller or similar sized as the 100-kDa collagenase can leak from the vitreous cavity into the periocular connective tissue. To this end, we injected a recombinant 45-kDa, myc-tagged peptide from the N-terminal part of agrin into the vitreous and traced its distribution during the following 2 days. The peptide included a laminin-binding domain that resulted in binding to all basement membranes. ¹⁵ When injected into an intact eye, the peptide bound to the ILM of the injected eye (Fig. 8a) . When the peptide was injected into eyes in which the ILM and VB had been destroyed by collagenase and the ILM reconstituted by laminin-1, it localized to the reconstituted ILM (Fig. 8c)but did not label the adjacent BM of Bruch’s membrane or blood vessels of the choroid. Staining of adjacent sections with antibodies to nidogen confirmed the presence of a regenerated ILM and the nearby Bruch’s membrane (Fig. 8d) . Controls showed that there was no myc-labeling in the chick tissue (Fig. 8b) . Similar data were seen when hybridoma cells that secreted a mAb to laminin were injected into the vitreous cavity. The hybridoma cells survived, proliferated, and, during the 2-day survival, continuously secreted 150 kDa IgG into the vitreous. As expected, the anti-laminin IgG bound to the laminin in the ILM, the lens capsule and the corneal BM. There was some diffusion of IgG in between the lens fibers and little labeling along the BM of Bruch’s membrane in the anterior most segment of the eye, but no leakage of the IgG from the eye to the posterior Bruch’s membrane and to the sclera was detectable (Fig. 8e) . Similar data were obtained by repeating previous experiments injecting two low molecular weight derivatives of biotin into the embryonic chick eyes ²³ : the two tracer very strongly labeled proteins of the VB, the ILM and the retina, but never leaked into and labeled the periocular tissues (Figs. 8f 8g) . 

Discussion

Protein Composition of the Chick VB

The VB is the largest structural compartment of the eye, and it is a very highly hydrated ECM (∼98% water wt/wt). Consistent with earlier reports on human, bovine, and mouse VB, ^{24 25} the chick VB contains several collagens, including type II, IX, V/XI, IV, and XVIII. Collagen II forms the principal fibrillar network that confers the gel-like properties of the VB. ²⁶ The type II collagen fibrils are heterogeneous and include type IX and type V/XI collagen. Type IX collagen is a chondroitin sulfate proteoglycan that localizes to the surface of the collagen II fibrils and binds large quantities of water, ²⁰ whereas type V/XI may form a central core that provides a template for fibril assembly. ²⁷ Collagens IV and XVIII participate in the assembly of the ILM. In addition to the collagenous components, the VB also includes several glycoproteins that are associated with fibrillar structures, either by forming their own fibrillar system, such as fibrillin, ²⁸ or by adhering to the collagen II fibrils, such as fibronectin and tenascin. Finally, the VB includes several proteins that are concentrated in serum, such as albumin and transferrin. ²⁹ The nonpigmented ciliary epithelium and the lens epithelial cells are the major source of all VB proteins. ^{13 30 31} As shown previously, the VB proteins are most abundant in the embryonic and least abundant in the adult eye. The decline in VB protein concentration occurs shortly after hatching in chick and shortly after birth in humans. ¹³ The chick vitreous is nearly all gel at hatching, but approximately 70% liquid at 1 year of age. ¹³ Similarly, the gradual degradation of VB proteins combined with an insufficient de novo synthesis leads to the age-dependent liquefaction of the VB later in human life. ^{13 32 33 34 35 36}  

Disruption and Partial Reconstitution of the VB

As shown in this study, injection of collagenase leads to the rapid liquefaction of the chick VB and the lysis of the ILM. A time-course study showed that the enzyme digested collagen II and XVIII within 30 minutes, followed by collagen IX hours later. Noncollagenous VB proteins were spared by the enzyme. Consistent with earlier reports, the degradation of the collagen II fibrils was the main reason for the liquefaction of the VB gel. ²⁶ Western blot analysis showed that between 2 and 4 days after collagenase injection, the concentrations of collagen II and IX and all noncollagenous VB proteins were at normal levels, yet the vitreous never reassembled into a gel. The inability to re-establish a fibrillar structure was obviously not due to the absence or the insufficient supply of the protein. Deposition of insoluble and thus detectable collagen II fibrils in the VB chamber only occurred after the ILM was reconstituted. Double labeling revealed that the ILM reassembly preceded the collagen II fibril deposition by 24 hours, and the collagen II fibrils that appeared 2 days after ILM reconstitution colocalized exactly with the reassembled ILM. The finding suggests that the presence of an ILM is a requirement for collagen II fibril deposition in the vitreous cavity. The colocalization of collagen II fibrils and the ILM also indicates binding between ILM proteins and the collagen II fibrils. The binding may be directly between collagen II fibrils and one of the ILM proteins, or, mediated by proteins that bind both to the ILM and to the collagen fibrils. A candidate is opticin, ³⁷ a glycoprotein with binding sites for collagen II and collagen XVIII, a component of the ILM. ^{38 39} Consistent with the involvement of collagen XVIII in VB fibril assembly is the fact that mice with a targeted mutation of collagen XVIII have fewer collagen II fibrils inserting into the ILM, ³⁹ and patients with Knobloch syndrome, a deletion mutation of collagen XVIII, have early-onset posterior vitreous detachment. ^{9 40} Although collagen II was deposited along the reconstituted ILM, it did not form a normal fibrillar network within the vitreous chamber. We assume that the displacement of the reconstituted ILM deep into the retina precluded a direct connection of the ECM proteins of the vitreous chamber with the ILM and did not sufficiently anchor the intravitreal fibrils to the retina. 

Eye Enlargement after ILM and VB Disruption

The enzymatic disruption of the ILM and VB led to an eye enlargement of 50% within 4 days. The eye enlargement was predominantly axial, thus resembling most human myopias. Unlike the postnatal eye, in which vision and the growth of the sclera play key roles in regulating eye size, ^{2 7} the collagenase-induced eye enlargements in our experiments occurred before the completion of scleral development and before maturation of visual function in the retina. Furthermore, the intraocular injection of collagenase did not visibly interfere with the deposition of ECM proteins in the sclera. Finally, tracing experiments showed that intraocularly injected laminin-binding peptides and protein do not leak from the vitreous cavity into the adjacent periocular tissues, even after the ILM and VB had been disrupted (Fig. 8) . The data are consistent with the results of biotin tracer experiments showing that the label did not leak from the injected eyes. ²³ From the present data, it is very unlikely that the injected collagenase may have led to eye enlargement by leaking from the vitreous cavity and digesting the ECM of the sclera. The strongest support in favor of the ILM’s and the VB’s playing a role in regulating early eye size comes from the fact that the regeneration of the ILM and partial reconstitution of the vitreous ECM with laminin-1 greatly attenuates the size increase after addition of collagenase. The data also show that the myopia in the experimental eyes is not due to the retraction of the neuroepithelial cells, since ILM regeneration greatly attenuated the size increase in collagenase-treated eyes, but did not reverse neuroepithelial cell retraction. ^{11 12} Although the data implicate the ILM and VB in the explanation of our results, other mechanisms, such as a blockage of the aqueous outflow and an increased intraocular pressure cannot be entirely excluded. Efforts to measure the intraocular pressure failed because of the softness of the embryonic eyes. 

By implicating a defective ILM and VB in excessive eye growth, our results may provide an explanation of why human mutations of collagen II and V/XI, proteins that are associated with VB fibril assembly, lead to congenital high myopia. ^{41 42 43} We propose that the pathologic high myopia seen with these mutations is due to the inability to assemble a normal VB gel or because of a defective connection between the VB with the ILM. We propose that the experimental paradigm described herein may serve as a model system for the study of the mechanisms underlying congenital high myopia. The importance of introducing chick embryos as a model for high myopia is emphasized by the lack of myopia models in mice. Mice with collagen II mutations do not show myopia. ⁴⁴ Further, it is well known from form-deprivation experiments that mice do not regulate their eye sizes to nearly the same extent as do chicks, monkeys, and tree shrews. ⁴⁵ However, the alterations at the vitreoretinal surface that are observed in collagen II mutant mice are already obvious during embryonic development, ⁴³ and we propose that congenital high myopia in humans, as in our chick model, develops prenatally. 

When the ILM was reconstituted but the VB remained in a liquefied state, the eye enlargement was greatly reduced, indicating that ILM disruption is an important factor in causing excessive eye growth under the conditions in these experiments. Patients with Knobloch syndrome (collagen XVIII deficiency) have early-onset high myopia with a great risk of retinal detachment. ^{9 40} Because collagen XVIII is a constituent of the ILM ^{38 39} but not of the sclera, this is consistent with a critical role for the vitreoretinal border in regulating eye size. 

How Does the Disruption of the ILM Lead to Excessive Eye Growth?

On the basis of our experiments, we propose that the ILM and the cortical VB provide mechanical strength to withstand the pressure within the vitreous chamber while the sclera is still developing. In the absence of the cortical ECM, the eye globe expands to a larger size than normal. Consistent with this concept, reconstitution of the ILM after its disruption greatly attenuated the size increase of the eyes (Fig. 6d) . Support for the role of the basement membrane in stabilizing tissue borders comes from the fact that mutations of basement membrane proteins cause herniation and rupturing of blood vessels ^{12 46} and penetration of cortical and retinal cells through breaks in the pial and retinal basement membranes. ^{12 47}  

 Supported by National Science Foundation Grant IBN0240774.

 Submitted for publication November 18, 2005; revised January 31, February 27, March 21, and March 30, 2006; accepted May 26, 2006.

 Disclosure: W. Halfter, None; U. Winzen, None; P.N. Bishop, None; A. Eller, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Distribution and abundance of ECM proteins of the chick VB. A montage fluorescence micrograph of a cross section through an E8 chick eye stained for fibrillin shows the fibrillar network that fills the vitreous cavity (a). R, retina; L, lens. Western blot analyses (b) show the abundance of collagen II (II), collagen IX (IX), fibrillin (Fib), fibronectin (FN), and tenascin (TN) in E10 VB. The specific protein bands are marked by arrows. The multiple bands for collagen II represent the unprocessed, the partially processed, the fully processed and a spliced variant of collagen II. The smear for collagen IX is due to high glycosylation of the protein. The SDS PAGE for fibrillin, fibronectin, and tenascin were run in nonreducing conditions. Blots show the 220-kDa monomeric and the 440-kDa dimeric versions of fibronectin. For tenascin, the monomeric 200-kDa form and several oligomeric versions were detectable. Scale bar, 150 μm.

Figure 2.

 

Disruption of the collagen II (coll2) network in E5.5 chick eyes 7 hours after the intraocular injection of collagenase (a). The contralateral control eye is shown in (b) for comparison. The fibrillin (fibri) network was not affected in the collagenase-injected eyes (c), appearing identical with control eyes (d). Western blot analysis (e) confirmed that incubation of VB with collagenase for 7 hours at 37°C in vitro led to the digestion of collagen II (Coll2) and collagen IX (Coll9) but left fibrillin (Fibri) and fibronectin (Fibro) intact. The blots show control VB (con) and collagenase-treated VB (coll) side by side. Scale bar, 100 μm.

Figure 3.

 

Time course of degradation for collagen II (Coll2), collagen XVIII (Coll18), and collagen IX (Coll9) after incubating E10 chick VB with collagenase in vitro for 30, 60, 120, and 180 minutes. C, control. Collagen II and XVIII were rapidly degraded, as is apparent from the shift of the collagen II protein bands to lower molecular weights and the decreased staining by a polyclonal antiserum, or by the disappearance of the stained collagen 18 band recognized by a monoclonal antibody. The highly glycosylated collagen IX proteoglycan was more resistant to collagenase degradation, and no changes in staining intensity and molecular weight were detectable in the 3-hour digestion.

Figure 4.

 

Inability of the VB and ILM to regenerate despite continued ILM and VB protein supply. Fluorescence micrographs show cross sections of retinas (R) 4 days after collagenase injection (a–c). Sections through the corresponding control retinas are also shown (d–f). The sections were stained for collagen II (coll2; a, d), collagen IX (coll9; b, e), and laminin (LN-1; c, f). Collagen II (a), collagen IX (b) and laminin (c) were undetectable in the VB and ILM 4 days after the enzyme injection. Note the abundance of collagen II (a) and IX (b) in the sclera (SC) and note the presence of laminin staining along the basement membrane of Bruch’s membrane in both the experimental and the control eye (*; c, f). Western blot analyses (g) compared the abundance of soluble collagen II (Coll2), collagen IX (Coll9), and laminin-1 (LN-1) in VB from control eyes (C), in two different VB samples from eyes 4 days after collagenase injection (lanes 1 and 2), and in a VB sample from eyes that had been injected with collagenase followed by laminin-1 to reconstitute the ILM (lane 3). The concentrations of the proteins in experimental and control VBs were similar. Scale bar, 50 μm.

Figure 5.

 

ILM-dependent deposition of collagen II fibrils at the vitreoretinal border. Fluorescence micrographs show cross sections of retinas (R) 1 (c), 2 (d) and 4 days (e) after collagenase and laminin-1 injections at E5. The sections were double-stained for collagen II (Coll2; red) and laminin (LN-1; green). The E6 control retina shown in (a) illustrates the high density of collagen II fibrils along the ILM. The E6 retina from an eye that had been injected at E5 with collagenase alone (b) shows the total absence of laminin and collagen II labeling at the vitreoretinal border (asterisks). Note that the labeling of collagen II in the sclera (SC) and of laminin in Bruch’s membrane (BM) was not affected by intraocular collagenase injection (compare a and b). A chase injection with laminin-1 led to the reconstitution of a new ILM within 1 day after its disruption (c). The new ILM was located 50 μm deep from the vitreous surface rather than at the vitreoretinal border as in a control retina (a; *). There was no collagen II deposition in the vitreous and retina 1 day after ILM reconstitution (c). Two days after ILM reconstitution, however, collagen II was deposited along the regenerated ILM (d). By day 4, laminin and collagen II labeling completely overlapped along the new ILM (e). A retina 4 days after injection with collagenase alone showed few and very minute laminin and collagen II spots (f). The sclera developed normally, whereas the ILM and VB were entirely dissolved (e, f). Scale bar, 50 μm.

Figure 6.

 

Increase in eye size after the enzymatic disruption of the ILM and VB. The size difference between the enzyme-injected (right eye) and the noninjected contralateral control eyes (left eye) 4 days after collagenase injection at E5 is shown in (a). The size difference was quantified (b) by comparing the weights of the VB from the experimental eyes (exp.; green bar; Coll) and their contralateral control eyes (cont.; red bar). VB weights from experimental and control eyes were also determined after injections of PBS (b, PBS), chondroitinase (b, Chon), and hyaluronidase (b, Hya). All measurements were performed 4 days after eye injections at E5. A time course study showed differences in VB weights between experimental and control eyes from day 2 onward (c). The first statistically significant size increase between control and experimental eyes was detected at day 4. Reconstitution of the ILM or VB matrix by chase injections with laminin-1 or collagen I attenuated the size increase after collagenase injection, as determined by VB weight measurements 4 days after ILM and VB disruption at E5 (d). While the disruption of the VB and ILM by collagenase resulted in a 50% increase in eyes size (Coll), eyes with a reconstituted ILM (LN) were only 13% larger than normal. Similar data were obtained with chase injections of collagen I (Coll1) and laminin-1 plus collagen 1 (LN+Coll1). Data are expressed as the mean; error bars, SEM. Except for (c), data sets that were not significantly different do not have a probability. Scale bar, 1 cm.

Figure 7.

 

Increase in eye size 7 days after collagenase injection at E5. The heads of embryos were sectioned along the horizontal (a, N, T: nasotemporal) or vertical (b; D, V: dorsoventral) axes. The experimental eyes (left in the micrograph) were larger than the contralateral control eyes (right in the micrographs). Area measurements showed a 26% increase in cross-sectional area of the experimental (green bar) versus the control (red bar) eyes (c). Measurements of the axial length (Ax) and the horizontal (H) diameter showed a 19% increase in the axial length, and an 8% increase in the horizontal diameter. Similar data were obtained by measuring eyes sectioned along the dorsoventral axis (D, V) as shown in (b). Scale bar, 1 mm.

Figure 8.

 

Proteins injected or synthesized in the vitreous cavity of the chick embryo do not leak into the adjacent tissues. A myc-tagged, 45-kDa agrin peptide that includes a laminin binding site was injected into E5 chick embryo eyes. The distribution of the peptide was detected 24 hours after injection with an anti-myc antibody. When the agrin peptide was injected into a normal eye (a), it bound to the ILM of the injected eye. There was no myc labeling in the noninjected control eye (b). When injected into an eye that had been treated with collagenase and laminin-1, the peptide bound to the reconstituted ILM. There was no leakage into the adjacent tissue including the nearby Bruch’s membrane (*). Staining of the adjacent section for nidogen (d) confirms the presence of the regenerated ILM and the adjacent Bruch’s membrane (*). Arrows: the vitreous surface of each retina (R). Hybridoma cells that secreted a mAb to laminin (mAb 3H11) were injected into E5 chick eyes (e). During the 2 days of survival, the cells proliferated and secreted the anti-laminin IgG into the vitreous cavity. The bound antibody was detected with Cy3-labeled anti-mouse IgG. The antibody labeled the ILM (arrow), the lens capsule (L), and the corneal basement membrane (BM) and a few micrometers along the BM of Bruch’s membrane at the ciliary margin. There was no leakage of antibody into the Bruch’s membrane at the posterior retina (*) or the sclera. When two low molecular weight biotin-based tracers (biotin-NHS, f, and biotin-X-NHS, g) were injected into the vitreous, the vitreous and the retinal cells were strongly labeled, but there was no labeling of the adjacent periocular tissues. The bound tracers were visualized in cross sections of the eyes with fluorescent streptavidin 24 hours after tracer injection. Scale bar: (a–d, f, g) 100 μm; (e) 150 μm.