Collagen fibril assembly during postnatal development and dysfunctional regulation in the lumican-deficient murine cornea



The transparent cornea is the outer barrier of the eye and is its major refractive surface. Development of a functional cornea requires a postnatal maturation phase involving development, growth and organization of the stromal extracellular matrix. Lumican, a leucine-rich proteoglycan, is implicated in regulating assembly of collagen fibrils and the highly organized extracellular matrix essential for corneal transparency. We investigated the regulatory role(s) of lumican in fibril assembly during postnatal corneal development using wild type (Lum+/+) and lumican-null (Lum−/−) mice. In Lum+/+ mice, a regular architecture of small-diameter fibrils is achieved in the anterior stroma by postnatal day 10 (P10), while the posterior stroma takes longer to reach this developmental maturity. Thus, the anterior and the posterior stroma follow distinct developmental timelines and may be under different regulatory mechanisms. In Lum−/− mice, it is the posterior stroma where abnormal lateral associations of fibrils and thicker fibrils with irregular contours are evident as early as P10. In contrast, the anterior stroma is minimally perturbed by the absence of lumican. In Lum+/+ mice, lumican is expressed throughout the developing stroma at P10, with strong expression limited to the posterior stroma in the adult. Therefore, the posterior stroma, which is most vulnerable to lumican-deficiency, demonstrates an early developmental defect in fibril structure and architecture in the Lum−/− mouse. These defects underlie the reported increased light scattering and opacity detectable in the adult. Our findings emphasize the early regulation of collagen structure by lumican during postnatal development of the cornea. Developmental Dynamics 235:2493–2506, 2006. © 2006 Wiley-Liss, Inc.


The highly regulated assembly of a stromal matrix containing precisely packed collagen fibrils with uniform, small diameters is a requisite for the development of a transparent, refractive cornea and mechanically stable anterior eye (Maurice,1957). The structural and functional properties of the adult cornea are acquired in multiple steps during fetal and postnatal development. These steps and their regulatory mechanisms have not been fully elucidated. It is important to define these in the mouse model system to understand corneal gene function and to fully utilize the power of knockout models. Toward this end, a recent study explored the growth and acquisition of transparency as the cornea matures between postnatal days 1 and 30 (Song et al.,2003). The decrease in light scattering is almost linear in this time frame as the cornea becomes transparent. At the same time the stroma thickens gradually, except around day 10, just prior to eyelid opening, when there is a sharp increase in thickness. A decline in thickness is seen after eyelid opening that recovers by postnatal day 20. Lumican, a major proteoglycan of the cornea, is critical to the establishment of normal corneal structure and function. In the lumican knockout mouse model, the corneal stroma develops 60% of its normal thickness and fails to acquire wild type transparent properties (Chakravarti et al.,1998,2000). These studies demonstrate that the sharp increase in stromal thickness observed before eyelid opening in the normal cornea does not occur in lumican-null mice (Song et al.,2003). These earlier studies suggest a prominent regulatory role for lumican in the development and maturation of corneal stromal architecture. Elucidation of the regulatory role(s) of lumican during stromal development is the focus of the current study.

Collagen fibril structure and organization are central to corneal structure and transparency. Newly assembled immature collagen fibrils assemble, grow, and mature during postnatal development of the cornea (Birk et al.,1996; Birk,2001). The corneal keratocyte is the major cell type in the corneal stroma that synthesizes the stromal extracellular matrix (Hamilton,1972; Beales et al.,1999; Fini,1999; Chakravarti et al.,2004). It secretes two fibril-forming collagens. Type I collagen is the most prevalent form and type V collagen comprises 10–20% of the total corneal fibrillar collagen. These collagens co-assemble into heterotypic fibrils (Birk et al.,1988,1990). The keratocytes also produce at least 4 different small leucine-rich proteoglycans: decorin, lumican, keratocan, and osteoglycin (Hassell et al.,1983; Blochberger et al.,1992; Funderburgh et al.,1997; Beales et al.,1999; Dunlevy et al.,2000). Interactions of these proteoglycans with the collagen fibril can modulate fibril structures (Vogel et al.,1984; Rada et al.,1993; Svensson et al.,2000). Therefore, the small leucine-rich proteoglycans are important regulators of matrix assembly. This family of proteoglycans also has been implicated in the regulation of cell growth and behavior; as reservoirs for growth factors; and in cytokine interactions with receptor tyrosine kinases (Ruoslahti and Yamaguchi,1991; Santra et al.,1997; Moscatello et al.,1998; Iozzo,1999; Vij et al.,2004,2005). In addition, in the corneal stroma these proteoglycans, particularly the keratan sulfate (KS) containing members (lumican, keratocan, and osteoglycin), are important in binding water and, therefore, in regulating stromal hydration, an important parameter in stromal swelling, interfibrillar spacing, and corneal transparency (Bettelheim and Plessy,1975; Doughty,2001).

Each of these four leucine-rich proteoglycans have been “knocked out” in the mouse (Danielson et al.,1997; Chakravarti et al.,1998; Saika et al.,2000; Tasheva et al.,2002; Liu et al.,2003) and all demonstrate abnormal connective tissue phenotypes in various organs and tissues. However, only mice deficient in lumican have a significant disruption in corneal collagen fibril architecture. The fibrils in adult lumican-null corneas show a wide range in diameter and contain a population of abnormally large fibrils with irregular contours (Chakravarti et al.,1998,2000). In addition, alterations in fibril packing are detected by X-ray diffraction analysis (Quantock et al.,2001). The fibril defects in the lumican-null cornea reside primarily in the posterior stroma where lumican content is normally high in the mature wild type mouse (Chakravarti et al.,2000). Decorin-deficient mice have no obvious corneal phenotype (Danielson et al.,1997). Keratocan- and osteoglycin-deficient mice have a modest increase in stromal fibril diameter, but maintain normal circular fibril profiles (Tasheva et al.,2002; Liu et al.,2003). In addition, the keratocan-deficient mice demonstrate fibril-packing irregularities and increased interfibrillar spacing (Meek et al.,2003).

The current study investigates the dynamics of collagen fibril assembly and structure through early postnatal development to the mature adult cornea. Collagen architectural changes that coincide with development and stromal growth in the wild type mouse are contrasted with the defective stromal development in the lumican-null mouse. This study provides essential data addressing collagen assembly and ultrastructure during postnatal stromal development, growth, and maturation of the cornea as well as the regulatory role(s) of lumican in these processes.


Postnatal Growth and Thickness of the Corneal Stroma

Previous in situ measurements of stromal thickness in living mice demonstrated a significantly thinner stroma in Lum−/− mice after eye opening (Chakravarti et al.,2000; Song et al.,2003). Our light microscopic analyses of thick sections of the cornea at postnatal day 10, 30, and 90 (P10, P30, P90) confirmed these in vivo results. These data indicate that after P10, the growing (P30) or adult (P90) Lum−/− stromas were approximately half as thick as in the Lum+/+ cornea (Fig. 1, page 7). These data suggested a structural basis of corneal thinning in the absence of lumican that is preserved after fixation and dehydration rather than one involving changes in hydration.

Figure 1.

Stromal thickness is reduced in growing and mature Lum−/− corneas. Light microscopy of thick sections (1 μm) of the cornea stained with methylene blue-Azur B. Compared to Lum+/+ mice, Lum−/− mice had a consistently thinner stroma (S) at 1 month (P30), a phenotype that was maintained in the mature 3-month-old (P90) Lum−/− corneas. Arrowheads indicate the boundaries of the stroma.

Corneal Collagen Fibril Structure During Postnatal Development and Maturation

Earlier structural studies were done using adult lumican knockout mice and demonstrated defects in collagen fibril structure and organization in the adult posterior stroma (Chakravarti et al.,1998,2000; Saika et al.,2000). Here, we focused on analyzing the developmental acquisition of this phenotype. Collagen fibrillogenesis is a multistep process, therefore, elucidation of the period in development when the defect first manifests itself will provide mechanistic insight into the collagen assembly steps that are altered in Lum−/− mice.

The murine cornea begins to acquire optical transparency after P10; this coincides with an increased matrix organization in the developing cornea seen as a gradual decrease in fibril spacing and an associated decreased backscattering of light (Quantock et al.,2001; Song et al.,2003). These processes involve the assembly of small-diameter collagen fibrils with a narrow diameter distribution. Fibril structure was analyzed in the stroma during postnatal development (P10, P30) and at maturity (P90) (Fig. 2). In the Lum+/+ mice, no obvious anterior-posterior difference in collagen fibril structure was observed. In contrast, in the Lum−/− mice, anterior-posterior differences in fibril structure were acquired with postnatal development. In P10 Lum−/− mice, the anterior stroma contained fibrils with relatively normal structures. At P30, the anterior stroma contained occasional thick fibrils and small irregularities in the fibril contour (data not shown), which became more pronounced by P90 (Fig. 2A). Collagen fibril structure was analyzed in the posterior stroma of postnatal corneas. In Lum+/+ corneas, small-diameter fibrils with circular profiles were assembled throughout postnatal development, growth, and aging. Morphologically the fibril structures at different stages were indistinguishable. However, fibril packing increased in regularity from P10 to P90. In the Lum−/− mice, fibril architecture was distinct from that seen in the Lum+/+ posterior stromas at these time points (Fig. 2B). Our previous study of the Lum−/− mouse demonstrated an increased fibril diameter and irregular contoured fibrils in the mature 7.5-month cornea (Chakravarti et al.,2000). The current analyses of the development of aberrant collagen fibrils demonstrated the early appearance of collagen anomalies, namely fused fibrils and fibrils with abnormally large diameters as well as altered diameter distributions by P10. By 1 month of development (P30), a population of large-diameter fibrils was consistently observed in the posterior stroma, with virtually all regions of the posterior stroma containing abnormal fibrils. At P90, the abnormal fibrils were larger and more irregular in profile than those seen earlier in development, with increased incidence of abnormal associations of collagen fibrils. After 3 months, the posterior stromal phenotype was relatively stable with no detectable changes between P90 and 1 year (data not shown).

Figure 2.

Spatial and temporal acquisition of abnormal collagen fibril structure in the Lum−/− corneal stroma. Ultrastructural analyses of fibril structure were done in the anterior and posterior stroma during early postnatal development (P10), growth (P30), and in mature adult (P90) corneas. A: Fibril structure and organization in the anterior stroma were comparable in Lum+/+ and Lum−/− mice. However, at P90, occasional large-diameter fibrils (arrow) were present in the Lum−/− corneas. B: Fibril architecture was markedly altered in the posterior stroma at all stages of corneal development, in the Lum−/− mouse, with the frequency of structurally aberrant fibrils increased from P10 to P90. In the P10 Lum−/− mouse cornea, abnormal fibrils, consisting of large fibrils and small clusters of laterally associated fibrils (arrows), were evident; at P30 and P90, large structurally aberrant fibrils with irregular cross-sectional profiles accumulated, with an increased incidence of laterally associated fibrils. At all 3 stages, the packing of the posterior stroma also was disrupted in the Lum−/− corneas compared to the Lum+/+ controls. P10, postnatal day 10; P30, 1 month; and P90, 3 months.

Spatial (Anterior-Posterior) Differences in Stromal Fibrillogenesis

Distinct structural, biochemical, and functional differences have been demonstrated along the anterior-posterior axis in human, rabbit, and bovine corneas (Bettelheim and Goetz,1976; Castoro et al.,1988; Freund et al.,1995). Therefore, development of the stromal matrix in the anterior and posterior stroma of the mouse was defined in Lum+/+ and Lum−/− mice at P10, P30, and P90.

No obvious anterior-to-posterior differences in collagen fibril structure were observed in Lum+/+ mice, while Lum−/− mice acquired a posterior enrichment of structural defects. The developmental and spatial acquisition of these defects were further characterized using analyses of fibril diameters. In both Lum+/+ and Lum−/− mice, the mean fibril diameters were larger in the posterior versus anterior stroma at all developmental stages studied (Fig. 3). Furthermore, the diameter distributions in the posterior stroma of the Lum−/− mice contained a distinct right-hand shoulder, consistent with the presence of two fibril subpopulations: one population of relatively normal fibrils with circular profiles and a second of large diameter fibrils with irregular profiles. A statistical analysis of median diameters in the 3 combined developmental stages for Lum+/+ versus Lum−/− mice was done. The median fibril diameters in the posterior stroma were an average of 4.1 nm larger than in the anterior stroma in Lum+/+ mice (P < 0.001, 95% CI: 3.3, 4.9) and 3.0 nm larger in the posterior stroma of Lum−/− mice (P < 0.001, 95% CI: 2.2, 3.7).

Figure 3.

Altered collagen fibril diameter distributions in the Lum−/− corneal stroma. Lum+/+ mice had small diameter fibrils within a narrow range in both the (A) anterior and (B) posterior stroma with the mean fibril diameters consistently larger in the posterior stroma. A: In the anterior stroma, the distribution of fibril diameters in the Lum−/− corneas was comparable to that observed in the Lum+/+ corneas, but the spread was somewhat greater at P30 and P90, with no major shoulder corresponding to large, irregular diameter fibrils observed. B: In the posterior stroma, Lum−/− mice demonstrated a distinct shoulder, indicative of larger-diameter fibrils. This shoulder was present at all 3 stages examined. The number of large, abnormal fibrils increased with postnatal development. Values are means ± sd (n) and the diameter range (min-max). P10, postnatal day 10; P30, 1 month; and P90, 3 months.

Fibril Assembly Is Altered in the Lum−/− Mouse

Two distinct fibril subpopulations, one relatively normal-appearing and a second composed of abnormal, laterally fused fibrils, were assembled in the Lum−/− cornea. A dysfunctional regulation of fibril growth associated with the absence of lumican presumably resulted in the aberrant, laterally associated subpopulation of fibrils. To determine if the normal-appearing fibril subpopulation in the Lum−/− mice was altered compared to the fibrils in the wild type corneas, the two subpopulations were analyzed separately in Lum+/+ and Lum−/− mice.

The dominant, normal-appearing fibril subpopulation was analyzed using data sets where the large structurally aberrant fibril subpopulation was removed. As described in the Experimental Procedures section, the structurally aberrant fibrils were identified using an established outlier detection rule (Hoaglin et al.,1986). The dominant, cylindrical fibril subpopulations had a symmetric distribution of fibril diameters. No significant differences in median fibril diameter were observed between Lum−/− and Lum+/+ mice and the difference among ages (P10, P30, P90) was only marginally significant (P < 0.058). The spread and heterogeneity of the fibril diameters in the cylindrical fibril subpopulation were characterized by analyses of sample inter-quartile ranges (Q3–Q1), as described in the Experimental Procedures section. The inter-quartile range encompasses the central 50% of the dominant fibril subpopulation and is used to measure the spread of the data that does not necessarily follow a normal distribution. For a normal distribution the inter-quartile range is the same as the interval (mean ± 0.67 sd) and the length of such inter-quartile range is 1.35 × sd. A significantly broader distribution of fibril diameters was observed in the Lum−/− versus Lum+/+ mice as determined from analyses of the inter-quartile ranges (Fig. 4). This broadening of the distribution indicated significant differences in diameter heterogeneity between Lum−/− and Lum+/+ distributions (P < 0.001), between different age groups (P < 0.001), and between the anterior and posterior stroma (P = 0.013). Moreover, the age-by-genotype interaction was significant (P < 0.001), indicating that changes in diameter variability associated with age had different patterns in Lum+/+ and Lum−/− animals (Fig. 4). At P10, the diameter ranges were similar for all genotypes and locations. At P30 and P90, diameter heterogeneity was significantly greater (P < 0.001) in Lum−/− compared to Lum+/+ mice in both the anterior and posterior stroma. In addition, at P90, diameter heterogeneity in the posterior stroma was greater than in the anterior stroma of Lum−/− animals (P = 0.004), but not in the Lum+/+ mice (Fig. 4).

Figure 4.

Diameter heterogeneity in the structurally normal fibril subpopulation in the Lum−/− corneal stroma. Analysis of the dominant, structurally normal fibril subpopulation in Lum−/− and Lum+/+ stromas during postnatal development. Lum−/− mice (solid circles) assembled fibrils with a significantly broader distribution compared to Lum+/+ mice (open circles). The difference between Lum−/− and Lum+/+ cylindrical subpopulations was significant at P30 and P90 in both the (A) anterior and (B) posterior stroma. Means of inter-quartile ranges (minimum and maximum values for central 50% of the fibril diameters) and their 95% confidence intervals are shown. For these analyses, the data sets were stripped of the larger, structurally abnormal fibrils (outliers) as described in the Experimental Procedures section.

The large, aberrant fibril subpopulation also was analyzed separately (Table 1). These analyses indicated a significant correlation between the presence of aberrant fibrils and the absence of lumican. With postnatal development and maturation of the corneal stroma, there was a significant increase in the number of large, aberrant fibrils (outliers). Aberrant fibrils were identified as outliers using standard statistical criteria as described in the Experimental Procedures section. In the posterior stroma where the large, abnormal fibrils predominate, a total of 204 outliers, including 44 extreme outliers, were identified. All extreme outliers, except one, were found in Lum−/− groups. These include 4 extreme outliers at P10 (0.2% incidence rate), 13 extreme outliers at P30 (0.6% incidence rate), and 26 extreme outliers at P90 (1.2% incidence rate). These results were analyzed by looking at all the data using a linear mixed effects model and asking what the chances are of finding abnormal fibrils. For the number of all posterior outliers, the generalized linear mixed effects model provided a good fit and predicted a total of 204.8 outliers. Significant differences in numbers of abnormal fibrils between Lum−/− and Lum+/+ posterior stromas were observed in each of the age groups analyzed (Table 1). The Lum−/− mice were more likely than Lum+/+ animals to have abnormally large-diameter fibrils, about 2.5 times at P10, 4.8 times at P30, and 4.0 times at P90. A significant increase in large, aberrant fibrils (outliers) was seen between P10 and both P30 and P90 in the Lum−/− stromas (Table 1). In the Lum−/− animals, the incidence rate of outliers was 1.8 times higher at P30 and 2.1 times higher at P90 as compared to P10. In the Lum+/+ posterior stroma, no significant differences in incidence rates of outliers were observed among different age groups. In the anterior stroma, only 37 outliers were identified in the entire data set and only one extreme outlier was observed. For the counts of all anterior outliers, the generalized linear mixed effects model provided a good fit and predicted a total of 37.4 outliers. Consistent with our experimental analyses, this model yielded no significant differences between Lum+/+ versus Lum−/− or among developmental stages in the anterior stroma.

Table 1. Analyses of the Number of Large, Abnormal Fibrils in the Posterior Stroma of Lum−/− and Lum+/+ Corneas
Incidence of large outliers in the posterior stroma
GenotypeAgeaTotal fibrilsObserved outliersObserved incidence rate (%)Standard error of incidence rate (%)
  • a

    P10, postnatal day 10; P30, 1 month; P90, 3 months.

  • b

    The incidence rate ratios were estimated from the fitted generalized linear mixed effects model, which accounts for correlation of data from the same animals and negatives. Since adjustments were made for these correlations, the estimates are close, but not exactly equal to the ratios of corresponding incidence rates reported.

Model-based results
GenotypeAgeaIncidence rate ratiob95% Confidence interval limitsP value
Lum−/− vs. Lum+/+P102.541.195.460.017
Lum−/− vs. Lum+/+P304.792.628.77<0.001
Lum−/− vs. Lum+/+P903.992.177.32<0.001
Lum−/−P30 vs. P101.841.123.010.015
Lum−/−P90 vs. P102.051.273.320.003
Lum−/−P90 vs. P301.120.731.710.615

Lumican Localization During Postnatal Development of the Cornea

Our previous study demonstrated increased localization of lumican in the posterior stroma of the adult mouse cornea (Chakravarti et al.,2000). Here, we addressed the developmental acquisition of this restricted spatial localization pattern and the relationship to structural defects in the developing Lum−/− stroma (Fig. 5). At P10, lumican was present homogeneously throughout the corneal stroma. By P45, lumican was localized almost exclusively in the posterior stroma. The posterior expression pattern was maintained in the mature (P90) cornea with some focal, heterogeneous expression in the anterior stroma of mature mice. Thus, the pronounced posterior fibril phenotype present in the Lum−/− cornea coincides with the developmental pattern of lumican deposition in the Lum+/+ cornea.

Figure 5.

Temporal and spatial distribution of lumican in the Lum+/+ cornea. Immunofluorescence micrographs showing the distribution of Lum+/+ corneas at P10, P45, and P90. At P10, there was a homogeneous distribution of lumican reactivity throughout the corneal stroma. At P45, the reactivity was primarily in the posterior stroma. However, the anterior stroma had anti-lumican reactivity above that observed in the Lum−/− control corneas. In the mature P90 cornea, the posterior stroma retained the higher level of lumican expression seen at P45. However, the anterior stroma showed heterogeneous regions with increased lumican reactivity. As expected, reactivity for lumican was not observed in the Lum−/− corneas.

Lumican is one of five small leucine-rich proteoglycans in the corneal stroma. To evaluate the role of lumican in the corneal stroma in context, the localization of the other small leucine-rich proteoglycans was analyzed over the same postnatal period (Fig. 6). Keratocan demonstrated a homogeneous localization throughout the mature (P90) stroma. Decorin and biglycan both demonstrated a comparable homogeneous stromal localization at P90. In contrast, osteoglycin was localized primarily to the epithelium and epithelial basement membrane zone with lower immuno-reactivity associated with the keratocytes and endothelium. Unlike lumican, the spatial localization of keratocan, osteoglycin, decorin, and biglycan did not change during postnatal development studied at P10, P30, and P90 (data not shown).

Figure 6.

Small leucine-rich proteoglycan localization during postnatal corneal development. Localization of keratocan, osteoglycin, decorin, and biglycan in P90 corneas using immunofluorescence microscopy. Keratocan demonstrated a homogeneous localization throughout the mature corneal stroma. No reactivity was observed in the epithelium or endothelium. Both decorin and biglycan demonstrated a comparable homogenous localization throughout the stroma, with biglycan reactivity substantially weaker than decorin. In contrast, osteoglycin localized to the epithelium and basement membrane zone (arrowhead). In addition, reactivity was associated with keratocytes and the endothelium. Arrowheads show osteoglycin reactivity in epithelium and basement membrane. Nuclei were counterstained with DAPI. Primary antibody was omitted in the negative controls. Arrows, endothelium; Ep, epithelium.

Semi-quantitative RT-PCR of total RNA with lumican specific primers indicated high lumican levels at P4 and P10 with a 30–45% decrease by 3–4 weeks. This expression pattern remained relatively stable at this level thereafter (Fig. 7A). However, there was a consistent small increase at 3 months coinciding with the period where the spatial localization of lumican within the stroma changes. Immuno-blotting of total protein extract from the cornea with an anti-lumican polyclonal antibody demonstrated high levels of the lumican core protein at P4 and P8 with decreased expression by P10 and P12 (Fig. 7B). The high lumican content during the early stages of postnatal development actually preceded the stage when an abnormal fibril phenotype appeared in the Lum−/− mice. This suggested that high initial levels of lumican were important in the regulation of stromal fibril assembly. At P10 and P12, there was a decrease in faster migrating bands. In general, the cornea is known to gain KS chains as it matures. We interpreted the faster bands at the younger stages as immature core proteins with few KS-side chains. The shift from fast to slower migrating bands is less likely to be due to a shift in the length of the KS-side chains, which in the mouse are short to begin with. The faster migrating bands were not observed, at the later stage, because with development these immature core proteins became modified by attachment of KS-chains and appeared as the slower bands. One would then expect to see increased intensity of the slower bands at P10 and P12, but we instead saw a decrease. We interpret this to mean that less core protein was made at the later stages and so lower amounts of the KS-modified proteoglycans were observed. This also followed the pattern of reduced lumican mRNA levels at these later stages. We do not believe that the reduced reactivity at P10 and P12 was due to reduced antibody penetration and staining of the core protein in the fully glycosaminoglycanated form, because this antibody, characterized in the past, has not shown reduced staining of mature proteoglycan versus deglycosylated core protein (Chakravarti et al.,1998).

Figure 7.

Lumican content during postnatal corneal development. A: Lumican mRNA was analyzed using semi-quantitative PCR. Lumican mRNA was highest at postnatal day 4 (P4) and P10. There was a decrease to 3 weeks (3W) and relatively stable expression thereafter. Lumican was amplified from total RNA extracts of the cornea. The amount of lumican amplicon was assessed relative to 18S RNA, with the relative amount at P4 set at 1.0. Error bars represent the standard deviation. B: Lumican core protein content was studied using semi-quantitative immuno-blot analyses. The amount of lumican core protein was assessed by immunoblotting total protein extracts from the cornea at P4, P8, P10, and P12.

We also investigated mRNA levels of the other corneal proteoglycans; keratocan, osteoglycin, decorin, and biglycan at these postnatal developmental stages (Fig. 8). Keratocan and decorin mRNA levels were constant throughout the developing stages analyzed. Biglycan and osteoglycin mRNA was high at P4 and P10, reduced by 3 weeks, after which the expression levels remained constant.

Figure 8.

Small leucine-rich proteoglycan mRNA during postnatal development of the mouse cornea. A: Keratocan and osteoglycin mRNAs were analyzed using semi-quantitative PCR. Keratocan mRNA was constant during the period studied while osteoglycin mRNA was highest at P4 and 10 followed by a decrease and stable levels from 3W to 6M+. B: Decorin mRNA levels were constant during the period analyzed. In contrast, biglycan levels were highest at P4, decreased to 3W, and were stable thereafter. Small leucine-rich proteoglycan mRNAs were amplified from total RNA extracts of the cornea. The amount of product was assessed relative to 18S RNA at postnatal day (P) 4 and, 3, 4, 6, and 8 weeks (W) as wells as 3 months (3M) and greater than 6 months (6M+). C: The amount of amplicon was assessed relative to 18S RNA, with the relative amount at P4 set at 1.0 for keratocan, osteoglycin, decorin, and biglycan. Error bars represent the standard deviation.


The cornea consists of a stratified, multilayered epithelium, a stroma, and a single layered endothelium that by virtue of its ion pump regulates stromal hydration. Although fully formed at birth, the cornea undergoes considerable, yet poorly defined pre- and postnatal development, growth, and maturation. During chicken development, the corneal stroma becomes compacted before hatching by dehydration and changes in fibril packing (Hay,1980; Connon et al.,2004). In the mouse, there is an initial decrease in thickness when endothelial pump-function ensues after eyelid-opening. Thereafter, the stroma thickens gradually until P30 (Song et al.,2003). In larger animals, such as the cat, the cornea grows for up to 2 years after birth (Moodie et al.,2001) and in the dog for up to the first 8 months (Montiani-Ferreira et al.,2003). In the human, eyelid opening–associated changes are prenatal, but the cornea grows and increases in thickness after birth up to 3 years of age. Therefore, the manifestation of many genetic defects during these early formative stages of the cornea is not unexpected. At maturity the fully functioning cornea, of which 90% is made up of collagenous stroma, provides three quarters of the refractive power of the eye (Pepose and Ubels,1992). Collagen fibrils and the anionic collagen-binding proteoglycans are organized into sheets or lamellae and form the bulk of the stroma. Adaptive changes in the ultrastructure of the stroma accompany the structural changes that occur before and after birth and are critical to the acquisition of corneal transparency.

Here we investigated the assembly and maturation of collagen fibrils during postnatal growth and development of the cornea in mice homozygous for the wild type allele or gene-targeted null allele of lumican. The data indicate a difference in fibril morphology and organization along the anterior-posterior axis, as noted in the chicken, rabbit, and human cornea (Bettelheim and Goetz,1976; Castoro et al.,1988; Freund et al.,1995). In the Lum−/− stroma, fibril structural defects arose early, were somewhat progressive, and primarily affected the posterior stroma where lumican expression became localized during normal corneal maturation (∼6 weeks).

First, our comparison of the anterior and posterior stroma at postnatal P10, P30, and P90 with respect to collagen fibril diameter revealed a small, but significant increase in fibril diameter from anterior to posterior stroma of the Lum+/+ mouse across all ages studied. In the human and rabbit, on average fibril diameters were smaller in the posterior stroma. In the human and rabbit studies, the current study examined very young to young adult mouse corneas. The observed difference in fibril diameter between the posterior and the anterior stroma may be simply age related. Alternatively, it may be species-specific with the extremely thin mouse cornea regulating collagen fibril assembly in a manner that is different from other larger species.

Across species, the overall organization and packing of collagen fibrils was consistently more regular in the posterior stroma. Our data, from the P10, P30, and P90 mouse cornea, suggest fibril organization in the posterior stroma increased at the later stage. A recent study of the developing chicken cornea showed that at the early stages, the anterior stroma was far more organized with a higher density of collagen fibrils than the posterior stroma. Fibril density increased in an age-dependent manner in the posterior stroma, ultimately catching up with the anterior stroma in collagen organization (Connon et al.,2004). In the human cornea, the anterior stroma was shown to be largely responsible for establishing corneal curvature. It was resistant to swelling and structurally more stable with the lamellae of collagen fibrils more interwoven than those in the posterior stroma (Muller et al.,2001). Presumably, these differences were modulated by various compositional differences along the anterior-posterior axis of the cornea. For example, the water content increased in the bovine cornea from the epithelium towards the corneal interior, alongside differences in the glycosaminoglycan content. The anterior stroma was rich in chondroitin sulfate, while the posterior stroma was rich in keratan sulfate (KS) which has greater water retentive power (Bettelheim and Goetz,1976; Castoro et al.,1988). Ample studies link keratan sulfate with the transparent refractive properties of the corneal stroma. Deficiencies in keratan sulfate metabolism were linked to macular corneal dystrophies (Hassell et al.,1980; Niel et al.,2003). During acquisition of corneal transparency, the developing cornea gains in KS, conversely wounding and loss of transparency, was associated with a decrease in KS (Funderburgh,2000). Thus, lumican, the most abundant keratan sulfate–containing proteoglycan of the corneal stroma is likely to be a key regulator of corneal transparency.

In this study, we demonstrated temporal as well as spatial differences in lumican expression in the developing mouse cornea. In the early neonatal stage (P10), there was a uniform distribution of lumican across the stroma. By 1 month, lumican began to show a preferential localization to the posterior stroma. Quantitative assessment of lumican in the mouse cornea indicated high levels of the lumican core protein during the early postnatal stages (P4, P8) with a substantial decrease observed by P10. At the mRNA level, we also noted high expression at P4 and P10. This was down-regulated by 3 weeks postnatal. Our findings on lumican distribution are in agreement with the temporal and spatial distribution of KS in the cornea and the general consensus that this proteoglycan plays a key role in the development and maintenance of corneal transparency.

We know that lumican regulates collagen fibril diameter and organization in the adult cornea (Chakravarti et al.,1998;2000; Saika et al.,2000). The spatial distribution of collagen fibrillar ultrastructural anomalies coincides with the posterior-rich distribution of lumican in the normal cornea, with transparency defects localized to the posterior stroma of the Lum−/− cornea. Given the temporal and spatial distribution of lumican, how is the structure and function of the cornea affected during its postnatal development and maturation in the Lum−/− mice? In our earliest study, corneal opacity was visible by slit lamp in the adult six-week-old animals (Chakravarti et al.,1998). In vivo confocal microscopy of the developing Lum−/− cornea indicated light scattering defects by three weeks (Chakravarti et al.,2000; Song et al.,2003). These functional anomalies were corroborated by the current identification of detectable structural anomalies at the early stages. The ultrastructural analyses of the postnatal cornea presented here demonstrated occasional abnormally associated fibrils and an increased range in the collagen fibril diameter distribution by P10. The number of abnormal fibrils and the large-diameter fibrils increased to maturity when the phenotype stabilized, arguing against a major regulatory role for lumican in the earliest, molecular assembly stages of fibrillogenesis.

Lumican-deficiency was associated with lateral fibril growth in the posterior stroma. Lateral growth is an important step in fibrillogenesis in virtually all soft connective tissues. However, it is inconsistent with corneal transparency. In other tissues, i.e., sclera, tendon, the down regulation of lumican expression was associated with the onset of fibril growth (Ezura et al.,2000; Chakravarti et al.,2003). This is congruent with a key role for lumican-fibril interactions in the regulation of initiation of lateral fibril growth. We propose a model (Fig. 9, page 7) where interactions with lumican stabilize the immature fibril intermediate and thereby restrict lateral growth. We hypothesize that these interactions are permissive for linear fibril growth since linear growth is required and occurs during stromal development (Birk et al.,1996). In addition, the initiation of fibril assembly does not appear to involve lumican since our data demonstrated no structural changes until after eye opening. This early step has been shown to be regulated by type I/V collagen interactions (Wenstrup et al.,2004; Segev et al.,2006). The absence of interactions between lumican and the fibrils lead to a dysfunctional regulation of fibrillogenesis, particularly in the posterior stroma. This allows abnormal lateral association and partial fibril fusion generating the structurally aberrant fibrils seen in the Lum−/− cornea. The spatial and temporal patterns in the lumican-deficient phenotype suggest that other regulatory interactions are involved in the coordinate regulation of stromal fibrillogenesis. Other small leucine-rich proteoglycans are good candidates. Interestingly, gene-targeted mice deficient in the other leucine-rich proteoglycans—osteoglycin (Tasheva et al.,2002), keratocan (Liu et al.,2003), and decorin (Danielson et al.,1997)—have milder corneal phenotypes with respect to collagen architecture and no disruption in transparency. However, these leucine-rich proteoglycans may interact with lumican in the regulation of collagen fibrillogenesis. The interactions of the closely related class I versus class II leucine-rich proteoglycans in the regulation of fibrillogenesis have been described (Ezura et al.,2000; Ameye and Young,2002; Chakravarti et al.,2003).

Figure 9.

Lumican in the regulation of corneal stromal collagen fibril formation. Schematic diagram presenting our model for the role of lumican in the regulation of collagen fibril assembly and growth. Collagen molecules assemble into fibril intermediates in close association with the keratocyte surface. In most connective tissues, these preformed fibril intermediates then undergo end-to-end linear growth and lateral associations leading to larger-diameter fibrils. In the cornea, only the linear growth takes place during normal development. However, during postnatal corneal development in Lum−/− mice, lateral growth occurs in a limited fibril subpopulation with restricted spatial and temporal distribution. A: In Lum+/+ mice, lumican (blue circles) associates with the shaft of newly assembled fibril intermediates. This interaction stabilizes the fibril intermediates and prevents lateral fibril growth as the intermediates grow linearly. B: In Lum−/− mice, in the absence of lumican, there is no stabilizing interaction and the immature fibril intermediates laterally and linearly associate. This unregulated growth yields abnormally large-diameter fibrils and laterally associated fibrils with irregular cross-sectional fibril profiles. However, the presence of normal corneal fibrils in the anterior stroma as well as intermixed with abnormal fibrils in the posterior stroma indicates that there are other regulatory interactions acting during postnatal development of the corneal stroma. The other closely related small leucine-rich proteoglycans, specifically keratocan, decorin, and biglycan, are potential candidate regulatory molecules as well as fibril-associated collagens type XII and XIV found with restricted temporal and spatial expression patterns during this period.

Lumican and keratocan are class II leucine-rich proteoglycans and lumican has been shown to regulate corneal keratocan synthesis (Carlson et al.,2005). Our demonstration of a predominantly cellular localization of osteoglycin during postnatal corneal development suggests that this proteoglycan is not directly involved. The anterior-posterior differences in the regulation of fibrillogenesis may be associated with differences in the composition and stoichiometry of other fibril-associated matrix macromolecules such as collagens type XII and XIV, which have demonstrated to have changing patterns during corneal development (Anderson et al.,2000; Young et al.,2002; Marchant et al.,2002).

In summary, our study demonstrates that in the Lum−/− mouse, the posterior stroma acquires collagen fibril structural disruptions by P10, while the anterior stroma is relatively unperturbed at this early stage. The posterior collagen fibril anomaly stabilizes in the mature adult, at which point the anterior stroma shows subtle structural defects. Our findings reinforce the concept that in the corneal stroma, the anterior and posterior regions have different mechanisms of development and maturation. Expression and accumulation of lumican early in corneal development may be essential for the maturation of collagen fibrils and associated with a properly hydrated and transparent cornea. After the early framework is established, the anterior stroma, primarily responsible for defining corneal curvature, may be under the control of other factors/interactions that are unaffected by lumican-deficiency.



Gene targeted mice deficient in lumican (lumtm1sc/lumtm1sc), here referred to as Lum−/−, in an outbred CD-1 background (Chakravarti et al.,1998) and wild-type mice (Lum+/+) in the same genetic background were used in these experiments between postnatal day 4 and 12 months. All animal studies were performed in compliance with IACUC approved animal protocols.

Immunofluorescence Microscopy

Corneas were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS, pH 7.3) for 30 min on ice as previously described (Ezura et al.,2000; Chakravarti et al.,2000).The tissues were cryoprotected with 2 M sucrose-PBS, and frozen in optimal cutting temperature compound (OCT; Tissue Tek, Miles Laboratories, Naperville, IL). Sections (6 μm) were cut and mounted on poly-L-lysine–coated slides, reduced with sodium borohydride, and nonspecific binding sites were blocked by incubation in 5% bovine serum albumin (BSA) in PBS overnight at 4° C. Sections were then incubated with rabbit anti-mouse polyclonal antibodies specific for the protein cores of lumican, keratocan (Chakravarti et al.,2000), decorin (LF113), biglycan (LF159) (Fisher et al.,1995), or osteoglycin (Ge et al.,2004). The secondary antibodies were goat anti-rabbit IgG dichlorotriazynyl amino fluorescein–conjugated at 1:150 (Jackson ImmunoResearch, West Grove, PA) or goat anti-rabbit IgG Alexa Fluor 568 or 488 at 1:400 (Molecular Probes, Eugene, OR). Each primary antibody was serially diluted from 1:50 to determine appropriate dilutions in the linear range for experimental analyses. The following primary antibody dilutions were used: anti-lumican 1:200; anti-osteoglycin 1:100; anti-keratocan 1:200; anti-decorin 1:200; and anti-biglycan 1:100. Negative control samples were incubated identically, except the primary antibody was excluded. To visualize nuclei, the slides were mounted in glycerol solution with 1 μg/ml Hoechst stain. Images were captured using a digital camera (Optronics, Goletta, CA), with set integration times and identical conditions to facilitate comparisons between samples.

Western Blots

Total protein was extracted from homogenized corneas using M-PER mammalian protein extraction reagent (Pierce Technology) at room temperature for 5 min, followed by adding the Halt protease inhibitor mixture (Pierce Technology). The proteins were resolved by SDS-PAGE in an 8% polyacrylamide gel and transferred to 0.2-μm pore size nitrocellulose membrane (Invitrogen) for immunoblotting. Lumican was detected using a polyclonal antibody against mouse lumican described above (Chakravarti et al.,2000). Actin was immunostained with a polyclonal antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) as a control for equivalent loading. Anti-rabbit (1:1,000) and anti-goat (1:2,500) HRP conjugated secondary antibodies were used with the appropriate primary antibodies. Membranes were developed using Super Signal West Pico, stable peroxide solution, and Luminol/Enhancer (1:1 ratio).

Semi-Quantitative PCR

Semi-quantitative RT-PCR analyses were done as previously described (Ezura et al.,2000; Young et al.,2002; Zhang et al.,2003). Primer sequences for lumican, keratocan, osteoglycin, decorin, and biglycan are given in Table 2. Classic II 18S internal standard (Ambion) was used as reference gene yielding a 324-bp product. The amplicon size; primer pair to competimer ratio yielding comparable product intensities for 18S and each proteoglycan; and the optimal PCR cycle number are also presented in Table 2. PCR programming was: 94°C 2 min followed by cycles of 94°C 20 sec, 60°C 20 sec, 72°C 40 sec, and a final extension at 72°C for 10 min. For each age, PCR was done with at least 6 different cDNA preparations from at least 2 independent mRNA samples. The mean density of proteoglycan bands was normalized to the 18S reference gene and presented as mean density ± standard deviation.

Table 2. Semi-Quantitative PCR
GenePrimer sequencesAmplicon size (bp)Primer/Competimer ratioOptimal cycle no.
Lum-F5′-GTC ACA GAC CTG CAG TGG CTC AT-3′6562:1327
Ogn-F5′-TGC TTT GTG GTC ACA TGG AT-3′2182:1228
Kera-F5′-CTT TCC CCG AAT CAA TGC TA-3′1752:1226
Dcn-F5′-AGG CAT TCA AAC CTC TCG TG-3′1812:1224
Bgn-F5′-GAC AAC CGT ATC CGC AAA GT-3′2132:1226

Transmission Electron Microscopy

Corneas from Lum−/− and age-matched Lum+/+ controls were used in these experiments at the postnatal stages indicated. Processing for transmission electron microscopy was as previously described (Birk and Trelstad,1984; Chakravarti et al.,2000). Briefly, whole eyes were placed in fixative and the corneas dissected. The corneas were fixed in 4% paraformaldehyde/2.5% glutaraldehyde/ 0.1 M sodium cacodylate, pH 7.4, with 8.0 mM CaCl2 for a total of 2 h on ice, followed by post-fixation with 1% osmium tetroxide for 1 hr, dehydrated to 50% ethanol, en bloc stained with ethanolic uranyl acetate and complete dehydration in a graded ethanol series, followed by propylene oxide. The corneas were infiltrated and embedded in a mixture of Embed 812, nadic methyl anhydride, dodecenylsuccinic anhydride, and DMP-30 (EM Sciences, Fort Washington, PA). Thick sections (1 μm) were cut and stained with methylene blue-azur B for light microscopy and selection of specific regions for further analysis. Thin sections were prepared using a Reichert UCT ultramicrotome and a diamond knife. Staining was with 2% aqueous uranyl acetate followed by 1% phosphotungstic acid, pH 3.2. Sections were examined and photographed at 80 kV using a Hitachi 7000 transmission electron microscope. The microscope was calibrated using a line grating.

Fibril Measurements

The central corneal stroma was divided into 4 equal regions. Analyses were done for both the anterior and posterior stroma. The anterior stroma was defined as the region subjacent to the epithelium and the posterior stroma was adjacent to the endothelium. Both regions were photographed in the central portion. Micrographs were taken at 48,700×. Calibrated micrographs from each region were randomly chosen in a masked manner from the different regions. The micrographs were digitized and all diameters were measured within a 1.6-μm2 mask. The mask was placed based on fibril orientation, i.e., cross-section and absence of cells. Diameters were measured along the minor axis of fibril cross-sections using a RM Biometrics-Bioquant Image Analysis System (Memphis, TN).

Statistical Methods

The following statistical analyses of fibril diameters were performed for the anterior and posterior stroma. In the anterior stroma, there were a total of 59 negatives from 29 animals (5 animals per age/genotype combination, except for the 10-day Lum+/+ group with 4 animals, usually 2 negatives per animal, 1 field per negative). In the posterior stroma, a total of 360 fields from 90 negatives from 28 animals (4 or 5 animals per age/genotype combination, usually 3 negatives per animal and always 4 fields per negative) were analyzed. The anterior and posterior fibril diameters were measured in the same 28 animals.

The cornea fibril diameters in each measured field in a negative were considered as a mixed distribution consisting of a dominant symmetric component contaminated with a relatively small percentage of outliers. To identify potential outliers in each measured field, a well-established resistant outlier detection rule was utilized (Hoaglin et al.,1986). This rule labels as an outlier any observation that falls below Q1–1.5(Q3–Q1) or above Q3 + 1.5(Q3–Q1), where Q1 and Q3 are the first and third sample quartiles, respectively. The first and the third sample quartiles, Q1 and Q3, are the values where 25 and 75% of the fibril diameters are less than or equal to Q1 and Q3, respectively (Rosner,1995). Thus, the interval [Q1, Q3] contains the central 50% of the observations, and the sample inter-quartile range (Q3–Q1) is representative of the data spread. This rule is implemented in the usual box-plot data presentation. Also, extreme outliers were defined as all observations below Q1–3(Q3–Q1) or above Q3 + 3(Q3–Q1) (Hoaglin et al.,1986).

The numbers of large outliers (above Q3 + 1.5(Q3–Q1)) per field were modeled in a generalized linear mixed effects model (Vonesh and Chinchilli,1997), with the assumption of the negative binomial distribution, which provided a better fit than the Poisson distribution (also often used to model count data). These models incorporated animal-to-animal and negative-to-negative variability, but without separating corresponding variance components. Two separate models were fit to the anterior and posterior data.

For the outlier-trimmed posterior cornea fibril diameter data, the assumption of a Gaussian distribution was reasonable. However, for the outlier-trimmed anterior cornea fibril diameters, it was adequate to assume symmetric, but not Gaussian dominant component. Therefore, non-parametric measures (not dependent on any distribution assumptions and robust with respect to the presence of outliers) were used to summarize all of the outlier-trimmed data from each negative. We used the sample median as a measure of location and the sample inter-quartile range (Q3–Q1) as a measure of spread. The negative-specific medians and square roots of inter-quartile ranges were separately modeled in a linear mixed effects model incorporating animal-to-animal and negative-to-negative variability. In these models, the anterior and posterior outlier-trimmed data were combined and accounted for the correlation of measures from the same animals. Based on examination of residuals from these models, the model assumptions were adequate for these data.


We acknowledge the expert technical assistance of Biao Zuo and Diana W. Menezes. We thank Dr. Magnus Höök, Center for Extracellular Matrix Biology, Texas A&M University Health Science Center, Houston, for kindly providing the antibodies against osteoglycin. This work was supported by grants from the National Institutes of Health, EY05129 (D.E.B.) and EY11654 (S.C.).