Collagen V is a dominant regulator of collagen fibrillogenesis: dysfunctional regulation of structure and function in a corneal-stroma-specific Col5a1-null mouse model

Collagen V is a member of a subclass of fibril-forming collagens that retain portions of the N-terminal peptide domain. It is involved in the regulation of fibril assembly and can be classified as a regulatory fibril-forming collagen (Birk and Bruckner, 2011). The major isoform of collagen V, [α1(V)]₂α2(V), co-assembles with collagen I to form heterotypic fibrils (Birk et al., 1988). Collagen V is a quantitatively minor component relative to collagen I, comprising 2–5% of the total collagen in most tissues, e.g. dermis, tendon and bone, but in the cornea, it constitutes 10–20% (Birk, 2001; Segev et al., 2006). Much of the elucidation of collagen I and V heterotypic fibril structure and analysis of its regulatory interactions have been done in the corneal stroma (Birk et al., 1988; Birk, 2001; Linsenmayer et al., 1993; Marchant et al., 1996). The stroma contains a homogeneous population of small-diameter fibrils with regular packing, organized as orthogonal lamellae. This ordered structure results in minimal light scattering and optical transparency (Hassell and Birk, 2010). The tightly controlled assembly of stromal structure and function makes it an ideal model for elucidating the mechanisms regulating tissue-specific collagen fibrillogenesis and matrix assembly.

A collagen-V-knockout mouse model homozygous for a targeted deletion in the Col5a1 gene was embryonic lethal, and did not develop past the onset of organogenesis (Wenstrup et al., 2004a). In the embryonic mesenchyme, few fibrils were assembled in the absence of collagen V and presence of normal amounts of collagen I. These data indicate a critical role for collagen V in the nucleation of fibril assembly in the low collagen concentration environment of the embryonic mesenchyme. This critical role is consistent with the embryonic lethality of Co5a1 ablation and provides an explanation for why patients homozygous for mutations that do not produce collagen V alpha 1 chains have not been described, whereas haplo-insufficiency in COL5A1 is common (Schwarze et al., 2000; Wenstrup et al., 2000). However, it is well known that collagen I can self-assemble into fibrils in vitro in the absence of collagen V. Fibrils with normal D-periodic cross-striations are assembled indicating that nucleators, such as collagen V, are not required when collagen concentrations are above relatively high critical levels. Collagen V has been shown to nucleate collagen fibril formation in self-assembly assays in vitro, in cell culture studies and in mouse models (Birk et al., 1990; Birk, 2001; Marchant et al., 1996; Wenstrup et al., 2004a; Wenstrup et al., 2004b).

The nucleation of fibril formation by collagen V provides a mechanism whereby the fibroblast can regulate fibril diameter in a tissue-specific manner by regulating the ratio of nucleators to a given collagen concentration. For instance, collagen V is a larger percentage of the total fibril-forming collagen in the corneal stroma than in dermis and tendon. The large number of nucleation sites in the cornea would contribute to the formation of small-diameter fibrils necessary for transparency, whereas the lower number in the tendon and the dermis leads to larger-diameter fibrils required for mechanical strength. In addition, the fibroblast can define the site of nucleation and control fibril organization.

The classic form of Ehlers–Danlos syndrome (EDS, types I/II) is a generalized connective tissue disorder with broad tissue involvement characterized by fragile, hyperextensible skin, widened atrophic scars, joint laxity, a high prevalence of aortic root dilation and other manifestations of connective tissue weakness (Beighton, 1992; Beighton et al., 1998; Malfait et al., 2010; Steinmann et al., 2002). The most common reported molecular mechanism in classic EDS is the functional loss of one COL5A1 allele (Schwarze et al., 2000; Wenstrup et al., 2000), with a haplo-insufficiency for collagen V. Abnormal fibril formation is seen in the classic EDS and the dermis of EDS patients contains large, irregular collagen fibrils (Hausser and Anton-Lamprecht, 1994; Vogel et al., 1979). Studies using fibroblasts from EDS patients with characterized mutations in COL5A1 and a mouse Col5a1^+/− model of classic EDS have demonstrated that heterotypic collagen I and V interactions are involved in the regulation of fibril diameter and fibril number (Segev et al., 2006; Wenstrup et al., 2004a; Wenstrup et al., 2004b; Wenstrup et al., 2006). The Col5a1 heterozygous mice are haplo-insufficient and have a phenotype comparable to that of classic EDS (Wenstrup et al., 2006). In the dermis, there were two populations of fibrils, a structurally aberrant one and one with normal fibril structure, but larger fibril diameters. In addition, there were fewer fibrils assembled than in wild-type controls. This suggested a concentration-limited nucleation of fibril assembly, with the abnormal fibrils resulting from unregulated assembly. In the corneas of these mice, all fibrils had normal structure, but with larger diameters than the wild-type controls. This indicated that in the high collagen V concentration cornea, collagen V did not become limiting. However, there were fewer nucleation events due to the reduced level of collagen V, with collagen I concentration comparable to that in the control, resulting in a single population of larger fibrils. These data suggest a role for collagen V in nucleation of fibril assembly.

The purpose of this investigation is to elucidate the role of collagen V in the regulation of corneal stromal fibrillogenesis. A conditional mouse model with a null mutation in Col5a1 targeted to the corneal stroma was generated using a Cre-loxP approach. This approach permits the study of the regulation of fibrillogenesis in the high collagen concentration environment of growing and mature tissues by avoiding the embryonic lethal phenotype encountered in the conventional gene deletion mouse model. The results demonstrate that the absence of collagen V results in dysfunctional nucleation and fibril assembly. This leads to a severe disruption in corneal fibrillogenesis, matrix assembly and function, defining a critical role for collagen V in the regulation of fibrillogenesis and development of a functional cornea.

To elucidate the functional roles of collagen V in the regulation of collagen fibrillogenesis, a conditional mouse line was created. Homozygous null mice carrying a standard targeted deletion of Col5a1 exhibit an embryonic lethal phenotype (Wenstrup et al., 2004a). To overcome this limitation, a cornea-stroma-specific Col5a1-null mouse line was generated using a Cre-loxP-mediated conditional gene-knockout approach (Fig. 1). The alpha 1 chain of collagen V was targeted because all collagen V isoforms contain this chain. The strategy was to flank exons 3 and 4 in the Col5a1 gene with loxP elements. Cre excision would result in a nonsense mutation with a premature stop codon. The excision of Col5a1 exon 3 and 4 generated a premature stop codon, thus truncated the synthesis of collagen V. Only a small, non-functional peptide can be potentially translated. This putative small peptide contains the signal peptide and a portion of the non-collagenous domain 3 (NC3). The truncated collagen V peptide would be unable to form a stable triple helix. In addition, the chain selection domain is absent. A similar deletion strategy was used in the standard targeted deletion (Wenstrup et al., 2004a). Cre was targeted to the corneal stromal keratocytes by crossbreeding Kera-Cre mice with Col5a1^flox/flox mice thus creating bitransgenic mice where Cre expression is targeted to the corneal keratocytes using the keratocan promoter sequence. This resulted in a targeted deletion of Col5a1 in the corneal stroma in Col5a1^Δst/Δst mice. Unlike the mid-gestation lethality of conventional collagen-V-knockout mice, Col5a1^Δst/Δst mice were viable and fertile. There was no significant difference in the body weight between wild-type mice and Col5a1^Δst/Δst mice at post-natal day (P)10 and P30 (data not shown).

As expected, floxed Col5a1 alleles were identified in genomic DNA isolated from tails of Col5a1^flox/flox and Col5a1^Δst/Δst, but not wild-type mice, using PCR analysis. A 150 bp PCR product was amplified from tail DNA in the wild-type mice, and a 272 bp PCR product from the floxed allele in the Col5a1^flox/flox and Col5a1^Δst/Δst mice. In addition, the expression of the Kera-Cre transgene was characterized. A Cre PCR product was amplified only in the bitransgenic Col5a1^Δst/Δst mice and not in the Col5a1^flox/flox or wild-type mice. These data indicate that the Col5a1^Δst/Δst mice contain the floxed Col5a1 alleles and express Cre (Fig. 2A). The tissue specificity of the Kera-Cre recombinase activity was determined by breeding a male mT/mG double reporter mouse with female Kera-Cre mice (Fig. 2B). Analysis of the offspring demonstrated corneal-stroma-specific Cre excision. To analyze the targeting of Cre double reporter mice were used. Stromal keratocytes expressed mEGFP resulting from Cre recombinase activity. However, the corneal epithelium and endothelium expressed the red fluorescence, indicating a lack of excision. There was no Cre recombinase activity in muscle and other tissues in the eye, including sclera and iris. These data demonstrate a specific targeting of Cre recombinase to the corneal stroma keratocytes using the Kera-Cre 4.2 mice.

The excision of exons 3 and 4 in the Col5a1^flox/flox by Cre-recombinase in the corneal stroma of Col5a1^Δst/Δst mice was demonstrated using qualitative PCR of corneal stroma DNA as a template (Fig. 2C). A pair of PCR primers located upstream of exon 3 and downstream of exon 4 was designed and a 391 bp PCR product was amplified from Col5a1^Δst/Δst corneal stroma DNA. By contrast, a large PCR product that included exons 3 and 4 (over 2500 bp) was amplified in the Col5a1^flox/flox and wild-type mice. These data demonstrate the targeted deletions of exons 3 and 4 in the corneal stroma of Col5a1^Δst/Δst mice.

Null expression of Col5a1 after Kera-Cre-mediated excision of exons 3 and 4 was confirmed using Real Time PCR (Fig. 2D). Col5a1 mRNA expression was decreased to near baseline in the Col5a1^Δst/Δst corneal stroma compared with control samples. This indicated a null genotype in the Col5a1^Δst/Δst corneal stroma. Previous reports have indicated anterior and posterior differences in corneal structure, and severity of mutant phenotypes in mouse models (Chen et al., 2010; Zhang et al., 2009). Therefore, collagen V expression in the anterior and posterior stroma was analyzed. Samples from the anterior and posterior stroma were isolated using laser capture micro-dissection and a qPCR analysis was performed. The expression of Col5a1 mRNA was completely abolished in the Col5a1^Δst/Δst anterior corneal stroma and was at or near baseline in the posterior stroma (Fig. 2D). These data indicate that the Col5a1^Δst/Δst corneal stroma is null for collagen V expression.

Analysis of collagen V in the corneal stroma was performed using immunoblots and immunofluorescence localization (Fig. 3). The α1(V) chain of collagen V was not detectable in the cornea from Col5a1^Δst/Δst adult mice (Fig. 3A,B), consistent with the mRNA data. By contrast, the α2(V) chain expression was decreased and the α1(I) chain of collagen I demonstrated comparable expression in Col5a1^Δst/Δst and control corneas (Fig. 3A). Reactivity against collagen V was homogeneously localized throughout the wild-type stroma by immunofluorescence microscopy. However, there was a virtual absence of collagen V reactivity in the Col5a1^Δst/Δst corneal stroma (Fig. 3C). The data indicate that no collagen V is expressed in the corneal stroma of Col5a1^Δst/Δst mice. By contrast, the Col5a1^flox/flox mice expressed the α1(V) chain of collagen V at a level comparable to that in wild-type mice, indicating that the genetic manipulation and insertion of loxP elements did not alter the function of the Col5a1 gene.

Grossly, adult Col5a1^Δst/Δst mice displayed conspicuous corneal stromal opacity (Fig. 4A). Structurally, the corneal opacity is associated with abnormal fibril structure and stromal disorganization (Fig. 4B,C). In mature (P30) wild-type control mice, collagen fibrils had small, homogeneous diameters when analyzed using transmission electron microscopy. The fibrils were regularly packed and formed well-organized, interwoven lamellae. However, the mature Col5a1^Δst/Δst stroma was composed of collagen fibrils with a heterogeneous fibril population containing fibrils with very large diameters. Structurally aberrant fibrils with irregular profiles were common. In addition, the fibrils were irregularly packed and lamellar organization was disrupted. This phenotype was comparable in P30 and P90 Col5a1^Δst/Δst mice.

A striking disruption in fibril structure in the Col5a1-null mouse cornea was demonstrated in ultrastructural analyses of the stroma. Compared with wild-type control mice, Col5a1^Δst/Δst mice displayed larger and less uniformed collagen fibrils throughout the cornea stroma, with disrupted fibril packing. However, there was an anterior–posterior difference in the fibril phenotype, with the anterior stroma containing larger and more heterogeneous fibrils than the posterior stroma (Fig. 5A). An analysis of fibril diameter distributions demonstrated significant differences in fibril structure between Col5a1^Δst/Δst and control corneas (Fig. 5B). In the anterior stroma, the mean stromal fibril diameter increased from 25.3±3.4 nm in the control corneas to 42.5±12.8 nm in the Col5a1^Δst/Δst corneas. The mean diameter of the fibrils in the posterior stroma increased to 34.7±6.7 nm in the Col5a1^Δst/Δst mice, compared with 27.5±3.3 nm in the control mice. Differences in both the anterior and posterior regions were statistically significant (P<0.0001). In addition to the shift to larger diameter fibrils in the Col5a1^Δst/Δst corneas, there was a substantial broadening of the diameter distribution compared with controls, indicating a more heterogeneous fibril population. Both the anterior and posterior Col5a1-null stroma had a population of fibril diameters with a normal distribution. However the range was 8–95 nm in the anterior stroma compared with 12–61 nm in the posterior stroma. In comparison, the range of fibril diameters in the anterior and posterior stroma from control corneas was 14–42 nm and 14–39 nm, respectively. Associated with the alterations in fibril structure, the Col5a1^Δst/Δst corneas contained fewer fibrils than in the control corneas. In both anterior and posterior regions of the stroma, there was a decrease in fibril density in the Col5a1-null stromas compared with the controls (Fig. 5C). A significant (P<0.0001) decrease in the number of fibrils per unit area (μm²) of 58% in the anterior stroma and 36% in the posterior stroma of Col5a1^Δst/Δst compared with control corneas was observed. This is representative of a decrease in the number of fibrils assembled in the Col5a1-null stroma. The decreased fibril density and larger-diameter fibrils in the Col5a1^Δst/Δst stromas were associated with a decrease in the inter-fibrillar space compared with that in control stromas (Fig. 5D). The decrease in inter-fibril space was significant in the anterior stroma (P<0.01) as well as the posterior stroma (P<0.05).

Corneal stromal architecture was assessed using non-linear optical imaging of second harmonic generated signals (SHG) (Fig. 6). The SHG signals can be used to access lamellar organization in the cornea (Morishige et al., 2006). The SHG signals are dependent on intrinsic order in tissues such as that derived from organized collagen fibrils in the corneal stroma (Han et al., 2005). Forward-scattered signals result from the collagen fibrils. The forward-scattered SHG signals showed similar small bands of interwoven lamellae with a uniform distribution in the anterior (Fig. 6A) and posterior stroma (data not shown) in the control mouse cornea. In comparison, in the Col5a1-null stromas of Col5a1^Δst/Δst mice, much larger bands of uneven fibers were observed, especially in anterior stroma. This is consistent with larger fibrils with a higher density packing of fibrils within lamellae. Backward-scattered SHG signals are sensitive to the order of fibrils, with order associated with low signal, and disorder associated with intense signal. Intense backscattered SHG signals were obtained in the Col5a1-null stroma, with signals generated by disorganized collagen-containing lamellae (Fig. 6B). By contrast, the control stroma had a low-intensity backscattered SHG signal, which is consistent with well-ordered stromal lamellae. These data demonstrate that disruption of fibril assembly and organization significantly impacted higher order lamellar structure with lamellar architecture being less ordered in the absence of collagen V.

A non-invasive analysis of corneal light scattering and stromal thickness was done using in vivo confocal microscopy. Grossly, the Col5a1^Δst/Δst mice exhibited cloudy corneas (Fig. 4A). Corneal opacity was analyzed in Col5a1^Δst/Δst and control mice at P60 (Fig. 7A). In Col5a1^Δst/Δst mice, backscattered light was observed throughout the corneal stroma. However, there was an anterior–posterior difference, with more haze in the anterior stromas. By contrast, the control stroma demonstrated little scattered light. Both control and Col5a1^Δst/Δst mice demonstrated scattered light at the epithelial surface as a result of reflection at this interface. A three-dimensional image was rendered from the corneal epithelium to the endothelium (Fig. 7B). The epithelium and endothelium from Col5a1^Δst/Δst and control mice were comparable in cell number and thickness. However, a significant decrease (P<0.05) in stromal thickness was observed in Col5a1^Δst/Δst compared with wild-type mice (Fig. 7C). This decrease of 14% is consistent with the differences observed in the immunostaining analyses (Fig. 3C). Scattered light in the stroma was quantitatively assessed (Fig. 7D). There was increased backscattered light in the anterior versus the posterior stroma, consistent with the structural phenotype; however, in both regions, backscatter was considerably greater than in the control stroma. Total pixel intensity of the three-dimensional volume was divided by stromal thickness and the average light scattering per micrometer thickness of the corneal stroma was calculated. In the Col5a1^Δst/Δst mice, the light scattering was 3.5-fold greater than in control mice. The observed difference was statistically significant (P<0.01).

In summary, a conditional knockout of Col5a1 was created and targeted to the corneal stroma. This resulted in a Col5a1-null stroma with no alteration in collagen I. The Col5a1^Δst/Δst mice had grossly cloudy corneas associated with scattered light in the stroma. This functional defect was associated with a dysfunctional regulation of fibril assembly. In the absence of collagen V, a broad heterogeneous population of fibrils was assembled with abnormal structures. The fibrils were poorly organized and lamellar architecture was disrupted. Collagen fibrillogenesis is tightly regulated and disruption at any of these levels is inconsistent with corneal transparency. In addition, fewer fibrils were assembled, and there was a decrease in stromal thickness. These data demonstrate key regulatory roles for collagen V in: (1) nucleation, where it determines the number of fibrils initiated; (2) initial fibril assembly, where it regulates fibril structure, i.e. diameter and circular cross-sectional profiles; and (3) fibril packing and regular orthogonal organization of lamellae. Regulation of these key steps is central to the normal development of corneal structure and function.

In the current study, we demonstrated collagen V as a central regulator of collagen fibril formation, matrix assembly and tissue function in the corneal stroma. In the absence of collagen V, fewer fibrils were assembled, fibrils had altered structures and fibril organization, as well as tissue architecture, was disrupted. Abnormal regulation of these key regulatory steps resulted in corneal transparency being compromised and a non-functional cornea. This is the first direct demonstration of a fundamental mechanism involving interactions of collagen I and collagen V in the regulation of collagen fibrillogenesis in tissues. Heterotypic collagen fibrils containing co-assembled collagen I and collagen V were initially described in the corneal stroma (Birk et al., 1988). It is now recognized that all fibrils are heterotypic co-assemblies of two or more fibril-forming collagens; a quantitatively major and minor collagen (Birk and Bruckner, 2011). The importance of collagen V has been suggested by studies that modulate its content in model systems, as well as EDS patients (Burrows et al., 1996; 1997; De Paepe et al., 1997; Nicholls et al., 1996; Toriello et al., 1996; Wenstrup et al., 1996; Wenstrup et al., 2006; Wenstrup et al., 2011; Bouma et al., 2001).

The cornea has served as a model for studies of collagen V function because of its high collagen V concentration (10–20%) compared with other connective tissues containing collagen I (2–5%), as well as its rigidly controlled fibrillar structure and tissue architecture, which is tightly coupled to function. To overcome the limitation of the embryonic lethality of the conventional collagen-V-knockout mice (Wenstrup et al., 2004a) and elucidate the regulatory roles of collagen V in fibrillogenesis, a conditional knockout of Col5a1 was targeted to the corneal stroma. A mouse line where loxP elements flank exons 3 and 4 of the Col5a1 allele was developed. Compared with wild-type mice, Col5a1^flox/floxmice had no differences in collagen V expression or in gross and microscope phenotype, indicating that the genetic modifications had no effect on the function of the Col5a1 allele. The conditional mouse model was used to generate a mouse line that is null for collagen V in the corneal stroma using a Kera-Cre mouse line that targets Cre expression to corneal keratocytes (Weng et al., 2008) and our data demonstrated stroma-specific Cre expression with no Cre activity observed in the corneal epithelium, endothelium, sclera or muscle. Breeding these Col5a1^flox/flox mice with the Cre-expressing mouse line results in bitransgenic mice where exons 3–4 of the Col5a1 gene are excised in stromal keratocytes. These Col5a1^Δst/Δst mice were collagen V null and developed cloudy corneas associated with altered fibrillogenesis. This is the only known mouse model where collagen-V-knockout tissues can be analyzed in mature mice, thus allowing an analysis of its roles in the regulation of fibrillogenesis in tissues and developing organ systems. In addition, the Col5a1^flox/flox conditional knockout mouse model allows the analysis of tissue-specific functions of collagen V by targeting Cre expression to different tissues. It also can be used to study the function of collagen V at different times in the lifecycle of the mouse or post injury, using drug-inducible Cre-expressing mice (Chen et al., 2004).

Collagen V nucleates fibril assembly to control fibril number and initial diameter. Our data demonstrated the assembly of fewer large-diameter fibrils with a heterogeneous diameter distribution in the Col5a1-null corneal stroma. This is in contrast to our previous work that demonstrated a virtual lack of fibril assembly in the mesenchyme of a traditional Col5a1^−/− mouse model (Wenstrup et al., 2004a). Our interpretation is that in mesenchyme with a low collagen I concentration, collagen V is required for fibril nucleation. Collagen I can self-assemble under physiological conditions in vitro after long lag phases and at relatively high concentrations. This indicates that nucleation of fibril assembly with collagen I alone is inefficient. Nucleators such as collagen V are not required when collagen concentrations exceed relatively high critical concentrations (Birk et al., 1990; Birk, 2001). In the current work, fibril assembly occurred in the mature corneal stroma, but lacked corneal-specific regulation, with a narrow distribution of small-diameter fibrils with near-circular fibril profiles. Our interpretation is that in the relatively high collagen concentration environment of the stroma collagen I inefficiently self-assembles. The absence of collagen V results in dysfunctional regulation of nucleation and initial fibril assembly, resulting in the aberrant fibrils observed. In the Col5a1^Δst/Δst stroma, there was also a decrease in the number of fibrils assembled. Our hypothesis is that the collagen V content defines the number of fibrils assembled, with a large number of nucleation sites leading to more small-diameter fibrils. Collagen V interacts with collagen I during assembly, subsequently becoming incorporated into heterotypic fibrils (Birk et al., 1990; Birk, 2001; Linsenmayer et al., 1993), thereby only functioning in one round of assembly. Therefore, normal regulation of fibril number and fibril diameter is a direct result of the ratio of collagen V to collagen I content. These data strongly support a regulatory role for collagen V in the determination of fibril number.

In normal corneal development, collagen fibrils are assembled with small diameters and do not undergo the lateral fibril growth observed in most connective tissues. The regulation of lateral growth involves small leucine-rich proteoglycans (SLRPs), such as decorin, biglycan, lumican, fibromodulin and keratocan (Chakravarti et al., 1998; Ezura et al., 2000; Jepsen et al., 2002; Liu et al., 2003; Zhang et al., 2009; Chen et al., 2010). The absence of SLRPs results in unregulated lateral fusion of fibrils giving rise to ‘cauliflower’ cross-sectional fibril profiles. These were not observed in the current study, and there is little observed interaction of stromal fibrils. This supports an interpretation where abnormal fibril structure is the result of dysfunctional regulation of initial assembly rather than abnormal lateral growth.

Corneal transparency also requires regular fibril packing and organization into orthogonal lamellae (Hassell and Birk, 2010) and a disruption in these parameters was observed in the Col5a1-null stroma. The factors regulating these steps have not been not fully elucidated. However, the interactions involving matrix molecules and the cell surface, as well as the interactions between matrix components, such as proteoglycans, have been implicated (Bredrup et al., 2005; Hassell and Birk, 2010; Zhang et al., 2009). During development, initial collagen assembly occurs in close association with the keratocyte surface (Birk and Trelstad, 1984; Birk and Bruckner, 2011). We speculate that collagen V is associated with the keratocyte surface. The NH₂-non-collagenous domain of collagen V was shown to interact with numerous matrix proteins that could generate such interactions (Symoens et al., 2010). An absence of collagen V would compromise control of assembly and deposition by dissociating these processes from their normal regulatory domain. For instance, mutations in tenascin X lead to an EDS phenotype in the presence of normal collagen V (Bristow et al., 2005; Mao et al., 2002). The mechanism might be related to the uncoupling of fibril assembly from the fibroblast surface, resulting from the absence of an indirect link to the fibroblast with the overall effect being dysregulation of collagen assembly. An absence of collagen V, resulting in dissociation of initial collagen assembly from the keratocyte surface, could impact regulation in different ways. In addition to the unregulated nucleation of assembly, less-efficient assembly of collagen I would be expected. Compartmentalizing the initial assembly steps within a micro-domain provides a mechanism whereby processes important in fibril assembly such as pro-collagen processing can be integrated under cellular control. Pro-collagen processing in the absence of collagen V requires further study. Support for this suggestion is provided by studies indicating that retention of the collagen I N-propeptide is associated with irregular fibril cross-sectional profiles in EDS and other genetic diseases, and in vitro systems with defects in processing of the NH₂-propeptide (Hulmes et al., 1989; Colige et al., 2004; Lenaers et al., 1971; Steinmann et al., 2002). This could contribute to the altered fibril structure in the absence of collagen V. Collagen V interactions with the keratocytes also permit control over positioning of newly assembled fibrils, and the disruption of stromal architecture observed in the absence of collagen V supports this suggestion.

In summary, a Col5a1^flox/flox conditional-knockout mouse line was established in this study. This enables us to explore the function of collagen V in different tissues and different developmental stages. Breeding Col5a1^flox/flox mice with cornea-stroma-specific Cre mice, we established a cornea-stroma-specific Col5a1 conditional-knockout mouse model. The Col5a1-null stroma exhibited a severe dysfunctional regulation of fibrillogenesis. The fibril diameter was increased and the fibril lamellae structure was disorganized. This structural abnormality resulted in corneal opacification and a loss of corneal function. These data indicate a key regulatory role for collagen V in the regulation of corneal collagen fibrillogenesis and development of function.

The Col5a1 sequence obtained from the Celera mouse genomic database was used to generate a Col5a1-targeting construct where exons 3 and 4 of the Col5a1 gene were flanked by loxP elements. The excision of exons 3 and 4 by Cre recombinase leads to a nonsense mutation with premature termination and only a non-functional truncated small peptide of Col5a1 could be translated. The detailed strategy is presented in Fig. 1. Basically, the targeting vector, which includes exons 3 and 4 flanked with loxP elements, a Neo cassette flanked with FRT sequences and a thymidine kinase (tk) negative selection sequence, was linearized and electroporated into 129 Sv/J mouse ES cells. After double selection with G418 and ganciclovir, ES cells containing the targeted Col5a1 allele were identified by Southern blot using SphI and SpeI digestion of ES cell genomic DNA. The positive ES cells heterozygous for the allele containing the integrated homologous recombinant were injected into wild-type mouse blastocysts. The chimeric males were mated with C57 BL/6 females and created targeted (Col5a1^ta/wt) mice. Then, Col5a1^ta/wt mice were crossed with a germline-specific Flp transgenic mouse to remove the Neo cassette and generate the heterozygous floxed Col5a1 (Col5a1^flox/wt) mice for conditional knockout. These mice were bred to homozygosity to generate homozygous Col5a1^flox/flox mice.

To generate cornea-stroma-specific conditional-knockout mice, female keratocyte-specific keratocan-Cre (Kera-Cre) transgenic mice (Weng et al., 2008) were bred with male Col5a1^flox/flox mice. The resulting female Col5a1^Δst/wt mice appeared normal and were bred with male Col5a1^flox/flox mice to create Col5a1^Δst/Δst mice; this avoids promiscuous excision of floxed alleles during spermatogenesis in male bitransgenic Kera-Cre/Col51a1^flox/wt mice (Weng et al., 2008). Genotyping was performed using a REDExtract-N-Amp™ Tissue PCR Kit (Sigma). The primers for Col5a1 genotyping were: Col5a1-F, TGGGATAGAGACAGGGCTTTG; Col-5a1-R, AGTCATTTGGTTCCCTCCCAG; and NeoR, ATCGCCTTCTTGACGAGTTC. The sizes of the Col5a1 floxed allele and wild-type allele are 272 bp and 150 bp, respectively. The primers for Kera-Cre are ra47, GCAGAACCTGAAGATGTTCGC and ra48, ACACCAGAGACGGAAATCCATC. To determine whether Cre-recombinase could use the inserted loxP sites to delete Col5a1 exons 3 and 4, primers from intron 2 (TAGCCTTTGATGCAGCTGGAGACT) and intron 3 (GCCCCTTCTCTGTTTTCTGCTC) were used to verify the excision of loxP sites by Cre recombinase in the mouse cornea stroma genomic DNA. The size of the excised allele is 391 bp.

Cornea was dissected from the mice at 6 weeks and was treated with Dispase II (Roche) to separate the cornea epithelia from the cornea stroma. The stroma of cornea was cut into small pieces and total RNA from cornea stroma and epithelia was extracted using the micro RNeasy Kit (QIAGEN). 4 ng total RNA per well was subjected to reverse transcription by using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and the real-time PCR was performed with SYBR Green PCR master mix (Applied Biosystems) on a StepOnePlus Real Time PCR system (Applied Biosystems). The primer sequences were as follows: Col5a1 forward primer, AAGCGTGGGAAACTGCTCTCCTAT and Col5a1 reverse primer, AGCAGTTGTAGGTGACGTTCTGGT; actin forward primer, AGATGACCCAGATCATGTTTGAGA and actin reverse primer, CACAGCCTGGATGGCTACGT. Each sample was run in triplicate and data were analyzed using StepOne software (Applied Biosystems). Actin was used as an internal control to standardize the amount of sample total RNA.

Laser-capture microdissection was used to dissect anterior cornea stroma and posterior cornea stroma. The detailed procedure followed the Arcturus protocol. Briefly, P30 mouse eyeballs from control Col5a1^flox/flox mice and Col5a1^Δst/Δst mice were dissected, snap frozen, serially sectioned onto PEN membrane glass slides (Arcturus). After 70% ethanol fixation, staining and dehydration, laser-capture microdissection was performed to separately collect anterior and posterior cornea stroma samples with the Arcturus XT™ system and software. The total RNA isolation was performed using the PicoPure RNA Isolation Kit (Arcturus) and the reverse transcription and real time PCR was performed as described above. The CT value was adjusted by the amplification efficiencies for each sample. Amplification efficiencies were measured by the default fit option of LinRegPCR while maintaining the cycle threshold as a data point within the measured regression line (Mienaltowski et al., 2008; Schefe et al., 2006).

Protein extracts were prepared in extracting buffer containing 50 mM Tris-HCl, pH 6.8, 1% SDS and proteinase inhibitor cocktail (Roche). 5 μg of cornea protein lysate was separated on a 4–12% Bis-Tris gel (Invitrogen) and transferred onto a Hybond-C membrane (GE Healthcare). The membrane was hybridized with anti-α1(V) antisera (Wenstrup et al., 2004a), which targeted a peptide sequence in exon 6, downstream from the exon 3–4 region targeted by homologous recombination DNA. Actin was used as a protein loading control. Actin antibody was purchased from Millipore (Billerica, MA).

Immunofluorescence staining of collagen V in frozen sections was done in P30 mice corneas. The whole eye was fixed in fixative containing 4% paraformaldehyde, embedded in OCT medium and frozen at −80°C. Frozen sections were cut at 5 μm using a HM 505E cryostat. Before immunostaining for collagen V, the sections were pretreated with testicular hyaluronidase. Anti-mouse collagen V antibody was used at 1:400. The secondary antibody was goat anti-rabbit IgG conjugated to Alexa Fluor 568 (Invitrogen) at 1:400. Vectashield mounting solution with DAPI (Vector Laboratories, Burlingame, CA) was used as a nuclear marker. Images were captured with a Leica CTR 5500 fluorescent microscope (Wetzlar, Germany) and Leica DFC 340 FX digital camera. Antibody incubations and image acquisition were done concurrently for control Col5a1^flox/flox mice and Col5a1^Δst/Δst mice sections, using identical procedures and settings to facilitate comparison.

To examine the specificity and efficiency of Cre excision, female Kera-Cre mice were bred with mT/mG reporter male mice, which were purchased from Jackson Labs. mT/mG mouse is a dual-fluorescent Cre-reporter mouse that normally expresses membrane-bound Tomato Red (mT) fluorescence protein prior to Cre-mediated excision that leads to consequent expression of membrane-bound EGFP (Muzumdar et al., 2007). Tissues were dissected at P30, embedded in OCT and 6 μm frozen sections were cut. Fluorescence images were taken using with a Leica CTR 5500 fluorescence microscope.

The corneas from Col5a1^Δst/Δst and control Col5a1^flox/flox mice at age P30 were used for ultrastructural analysis. The samples were prepared for transmission electron microscopy as previously described (Ansorge et al., 2009). Sections were examined and photographed at 80 kV using a JEOL 1400 transmission electron microscope with a Gatan Orius widefield side-mount digital camera.

Fibril diameter analysis was done as previously described (Chen et al., 2010). For the measurement of collagen fibril diameters in control mice and Col5a1 conditional-knockout mice, corneas from three different P30 mice from each phenotype were analyzed. Four non-overlapping cross-sectioned digital images were taken at 100,000× from the anterior stroma and posterior stroma of the central cornea of each specimen. Diameters were measured along the minor axis of cross-sections using an R&M Biometrics-Bioquant Image Analysis System (Nashville, TN). Data analysis and histogram were done with Microsoft Excel software. The same images were used to measure the inter-fibril space. Four non-overlapping regions of interest (ROI) from each digital image were converted to binary images, the noise outliers were removed, and then the inter-fibril area fraction was measured with ImageJ software (National Institutes of Health).

In vivo analysis of cornea stromal thickness and haze was performed in P60 mice under anesthesia using an in vivo confocal microscope (Heidelberg Retinal Tomograph – HRT II Rostock Cornea Module, Heidelberg Engineering, Germany) as described (Liu et al., 2010). Briefly, a drop of GenTeal Gel (Novartis Pharmaceuticals) was applied to the tip of the HRT II objective as lubricants. Subsequently, a series of images was collected from the central corneal region. A continuous z-axis scan was obtained through the entire cornea at 1–3 μm increments starting in front of the epithelium and ending below the endothelium. To generate depth intensity profiles, the pixel intensity in the central region (ROI) of each consecutive image was measured. The total haze of the corneal stroma was obtained by summing the pixel intensity of the ROI in the continuous planes between the subbasal epithelium and the anterior of the endothelium cells. The axial distance between these two planes represented the cornea stroma thickness. The images along the x–z planes were also used to determine the corneal stroma thickness. 3D images were reconstructed using AxioVision Imaging software.

Confocal imaging of second harmonic-generated (SHG) signals was used to analyze the orientation of collagen fibrils within the stroma. The detailed procedure was described previously (Liu et al., 2010; Morishige et al., 2006).

We would like to acknowledge many helpful discussions with Yayoi Izu-Takahashi. The expert technical assistance of Qingmei (Chris) Yao also is gratefully acknowledged.