Collagen XVIII Short Isoform Is Critical for Retinal Vascularization, and Overexpression of the Tsp-1 Domain Affects Eye Growth and Cataract Formation

Abstract

Purpose.: Collagen XVIII deficiency leads to anterior and posterior eye defects in Col18a1^−/− mice, and overexpression of its C-terminal endostatin domain under a K14 promoter leads to cataract. We studied the consequences of K14-driven overexpression of the thrombospondin-1 (Tsp-1)–like domain, and also the roles of the three collagen XVIII isoforms in mice specifically lacking either the promoter 1–derived short or the promoter 2–derived medium/long isoforms.

Methods.: Two transgenic lines were generated and compared to Col18a1^−/− and promoter 1 and 2 knockouts. Enucleated eyes were analyzed histopathologically, immunohistochemically, biochemically, and ultrastructurally. IOP was measured by noninvasive tonometry, and the anterior chamber was studied in vivo using a slit-lamp and optical coherence tomography.

Results.: Overexpression of the Tsp-1 transgene in an FVB/N background resulted in increased axial length, and substantial incidences of cataract, lens subluxation, phthisis, retinal ablation, corneal vascularization, and intraocular hemorrhages. The FVB/N Col18a1^−/− mice were affected similarly. The findings in the knockout and transgenic lines were milder in a C57BL/6JOlaHsd (B6) background. Studies with the promoter-specific knockouts revealed the short isoform as the sole variant in the lens capsule and inner limiting membrane, while the ciliary body, iris, and Bruch's membrane contained short and medium/long isoforms. Lack of the short isoform, but not of the medium/long isoforms, caused aberrant retinal vascularization.

Conclusions.: An excess of the collagen XVIII Tsp-1 domain is deleterious in the eye, possibly by impairing certain functions of the full-length molecule. Moreover, the short isoform is the critical variant in the development of the posterior eye structures.

Introduction

Collagen XVIII is a heparan sulfate proteoglycan and an ubiquitous basement membrane (BM) component. It has an essential role in the eye, as mutations in its gene result in Knobloch syndrome, characterized by severe myopia, vitreoretinal degeneration, and anterior eye defects. ^1–7 Collagen XVIII deficiency in mice results in delayed regression of the hyaloid vasculature, and abnormal retinal vascularization through the optic nerve head and from persistent hyaloid vessels. ^8,9 Besides defects associated with blood vessels, Col18a1^−/− mice also exhibit abnormalities in the anterior part of the eye, where atrophy of the ciliary body (CB) and fragility of the iris can be detected. ^10,11 The vision of the Col18a1^−/− mice is attenuated increasingly with age. ¹² While eye defects are the hallmark of mutations in the collagen XVIII gene, these mutations also may lead to occipital encephalocele in some patients and abnormalities in other organs in individual cases. ^2,3,13  

The properties of collagen XVIII have been studied, especially with respect to its C-terminal noncollagenous (NC) domain, endostatin, shown to possess antiangiogenic and antitumorigenic properties. ¹⁴ In contrast, very little is known about the functions of the other domains of collagen XVIII, which has an intriguing multidomain structure, and occurs in three isoforms resulting from the use of two gene promoters. The upstream promoter 1 encodes the “short form,” which is present in most vascular epithelial BMs and lacks sequences corresponding to exon 3 of the 43-exon gene. The medium and long forms are encoded by promoter 2, and contain some (medium) or all (long) of the exon 3 sequences. All three isoforms contain a thrombospondin-1 (Tsp-1) homology domain in their N-terminal NC portion, while the N-termini of the medium and long isoforms have a further domain of unknown function, and the long isoform alone a domain sharing structural identity with the extracellular domain of frizzled (Fz) receptors (Fig. 1). ^15–18  

Figure 1

 

Polypeptide structures of collagen XVIII isoforms and the transgenic K14-Tsp-1 construct. All three isoforms (long, medium, and short) are shown. Collagenous sequences are shown in white, noncollagenous domains common to all variants in black, noncollagenous sequences common to the long and medium variants in gray, and a noncollagenous sequence specific to the long variant with a brick pattern. Predicted heparan sulfate (HS) attachment sites and Fz, Tsp-1, and endostatin domains are marked. Isoforms not produced in Col18a1^P1/P1 and Col18a1^P2/P2 mice, and sequences corresponding to the anti–medium/long-18 and anti–all-18 antibodies are indicated. The transgene product is depicted at the bottom. A cDNA encoding the 325-residue N-terminal domain of the collagen XVIII short isoform followed by an HA-tag was inserted between the K14 promoter and a polyA tail.

In the eye, where it is crucial, collagen XVIII is expressed in the inner limiting membrane and the outer plexiform layer of the retina, in the posterior pigment epithelium BM of the iris, and the BM zone between the iris epithelia and stroma, in the nonpigmented ciliary body BM, in the BM zones of the iris, and in the ciliary body capillaries. It also is expressed in the lens capsule, Bruch's membrane, and some of the choroidal capillary BMs. In the developing eye it is present in the vasa hyaloidea propria and tunica vasculosa lentis. ^8,11,19,20  

In view of our previous findings that K14 promoter-driven overexpression of the endostatin domain leads to broadened BMs in the anterior eye compartment and lens abnormalities, ²¹ we here decided to study the effects of similar overexpression of N-terminally located sequences of collagen XVIII. Hence, we generated a new mouse model by overexpressing the Tsp-1 domain common to all three isoforms using the K14 promoter (K14-Tsp-1). The effects of mutant collagen XVIII was investigated in two genetic backgrounds. FVB/N mice were chosen originally for the generation of transgenic mice. However, because early photoreceptor cell death is characteristic of this parental strain, and since it is known from previous studies with Col18a1^−/− mice that the background influences the penetrance of the phenotype, transgenic mice were bred also in a B6 background. ^22,23 The Col18a1^−/− mice of both backgrounds were characterized further. Moreover, to understand how the different collagen XVIII isoforms contribute to the eye structures, we studied the eyes of two recently generated promoter-specific knockout mouse lines ²⁴ (Aikio M, Elamaa H, Vicente D, et al., unpublished results, 2013). 

Materials and Methods

Mouse Lines

Transgenic mice overexpressing the Tsp-1 domain of collagen XVIII (K14-Tsp-1) were produced by generating cDNA constructs encoding the N-terminal sequence of the short isoform in a PCR using mouse cDNA clone SXT-5 ¹⁵ as a template and the partially complementary primers 5′-GAACGCGGATCCCTGGGGAGATGGCGCCCAGGTGGC-3′, covering the transcription start site, and 5′-TTCCGCGGATCCCTAAGCGTAGTCTGGGACGTCGTATGGGTACTTTATCAAGCCCCTTCCGGGGTT-3′, adding an C-terminal HA-tag and creating a stop codon at the end of the last N-terminal noncollagenous domain (NC11). ¹⁷ The ensuing PCR fragment was digested with BamHI, ligated into the K14 promoter expression cassette, ²⁵ and the insert was injected into fertilized mouse oocytes of the FVB/N strain. The transgenic founder mice were mated with FVB/N mice to generate separate lines and bred to homogeneity. Two independent transgenic founders, named K14-Tsp-1(1) and K14-Tsp-1(2), were obtained. The K14-Tsp-1(1) transgenic mice also were mated with C57BL/6JOlaHsd (B6) mice (Harlan Laboratories, Boxmeer, The Netherlands) and bred to homogeneity. The genotypes of the transgenic mice were confirmed by Southern blotting and PCR with the specific primers 5′-TGGAAAGTGCCAGACCCGCC‐3′ and 5′-CTACCACCTGGGCAGCATCT‐3′ or 5′-GCATTTGTTGGGCAGTGGAC‐3′ and 5′-TCCAGCTGTGAAGTGCTTGG‐3′. Promoter-specific knockouts of collagen XVIII were generated by inactivation of either exon 1 (Col18a1^P1/P1 ) or exon 3 (Col18a1^P2/P2 ), as described ²⁴ (Aikio M, Elamaa H, Vicente D, et al., unpublished results, 2013). The Col18a1^−/− mouse line has been described previously. ⁸ All procedures were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, and were approved by the Laboratory Animal Centre of the University of Oulu or the Animal Board of the State Provincial Office of Southern Finland. 

Analysis of Transgene Expression With Western Blotting

Eyes were snap-frozen in liquid nitrogen and total proteins of whole eye samples were isolated using the TriPure isolation reagent (Roche Diagnostics, Espoo, Finland). Total protein (2 μg) was subjected to 10% SDS–polyacrylamide gel and transferred to a nitrocellulose membrane. After blocking in 5% nonfat milk, the membrane was incubated overnight at 4°C with a polyclonal antibody against the anti-HA-Tag (1:250, sc-805; Santa Cruz Biotechnology, Inc., Heidelberg, Germany), followed by a horseradish peroxidase (HRP)–linked secondary antibody (Jackson ImmunoResearch Europe, Ltd., Newmarket, Suffolk, UK). Enhanced chemiluminescence reagents (GE Healthcare Europe, Helsinki, Finland) were used for detection. 

Histologic Examination of Eyes

Eyes from mice of different ages were obtained and fixed in Davidson's fixative (one part of glacial acetic acid, 3 parts of ethyl alcohol, 2 parts of 10% neutral buffered formalin, 3 parts of distilled water) for 24 hours. After fixation, the samples were kept in 70% ethanol, embedded in paraffin, sectioned at 5 μm, and stained with hematoxylin and eosin (H&E) by routine methods. The paraffin sections were immunostained using the HistoMouse SP kit (Invitrogen Life Technologies, Ltd., Paisley, UK) according to the manufacturer's instructions. 

IOP Measurements

We measured IOP bilaterally with the TonoLab rodent tonometer (Icare Finland Oy, Espoo, Finland) in gently restrained, unanesthetized mice. The mice were trained for these measurements, which were performed from 8 AM to 10 AM for all genetic backgrounds. The mean of 6 readings with optimal variability grade was recorded. 

Slit-Lamp Analysis

Eyes of transgenic mice and strain-matched controls were analyzed at multiple ages in vivo with a hand-held slit-lamp biomicroscope (Clinite; Middleton, Manchester, UK) at 22°C. For this, the mice were anesthetized (maximum 5 mice at the time) with a subcutaneous injection of a solution (0.1 mL/10 g) prepared by mixing one part of Hypnorm (0.315 mg/mL fentanyl citrate and 10 mg/mL fluanisone) with two parts of distilled water and one part of Dormicum (5 mg/mL midazolam). To minimize the formation of cold cataracts, anesthetized mice were kept on heated mats and the examination was done within 15 minutes of anesthesia. After the examination, the eyes were enucleated and analyzed under a dissection microscope. 

Transmission Electron Microscopy (TEM)

Eyes were fixed in 1% glutaraldehyde and 4% formaldehyde in 0.1 M phosphate buffer, pH 7.4, for 12 hours, after which the specimens were postfixed in 1% osmium tetroxide, dehydrated in acetone, and embedded in Epon LX 112 (Ladd Research Industries, Williston, VT). Thin sections were analyzed using a Philips CM100 transmission electron microscope, and images were captured using a Morada CCD camera (Olympus Soft Imaging Solutions, Münster, Germany). 

Optical Coherence Tomography (OCT)

Mice for the OCT examinations were anesthetized by subcutaneous injection of the above Hypnorm-Dormicum solution (0.1 mL/10g). The OCT was performed with Cirrus HD-OCT (Carl Zeiss Meditec, Jena, Germany). 

Examination of Collagen XVIII Variant Expression in the Mouse Eye

Expression of collagen XVIII was studied in control, Col18a1^P1/P1 , Col18a1^P2/P2 , and Col18a1^−/− mice at postnatal day 7. Eyes were snap-frozen and embedded in Tissue-Tek OCT Compound (Sakura Fine Technical Co., Ltd., Tokyo, Japan). For the immunofluorescence staining, 10 μm frozen sections were fixed with ethanol, blocked with 3% BSA-PBS, pH 7.3, for 30 minutes at room temperature, and stained overnight at 4°C with anti–all-18 and anti-medium/long-18²⁴ (Saarela J, Rehn M, Vainio S, Pihlajaniemi T, unpublished results, 1998) antibodies (10 ng/μL), and then a Cy3-conjugated secondary antibody (Jackson ImmunoResearch Europe, Ltd.). Cell nuclei were visualized with 4′,6′-diamino-2-phenylindole hydrochloride (Sigma-Aldrich, St. Louis, MO). Images were captured with an Olympus IX81 laser confocal microscope (Olympus Corporation, Tokyo, Japan). 

Staining of Whole Mount Preparations of Mouse Retinas

Eyes were harvested on postnatal day 21, and whole mount retinas were prepared and stained as described. ⁹ Rat anti-mouse CD31 (PECAM-1) antibody (BD Biosciences, Vantaa, Finland), rat anti-mouse CD-34 antibody (BD Biosciences) and Cy3 or Alexa 488-conjugated secondary antibodies (Jackson ImmunoResearch Europe, Ltd.) were used for the immunofluorescence staining. Images were captured with an Olympus IX81 laser confocal microscope. 

Statistical Analysis

We used SPSS 16.0 software (IBM Corporation, Armonk, NY) for the statistical calculations. The results are presented as medians (25th percentile, 75th percentile) in box plots, where the whiskers represent data points lying within 1.5 interquartile ranges of the median. Differences between 2 groups (each mutant line versus controls) were tested for statistical significance with a 2-tailed Student's t-test. A P value less than 0.05 was considered significant (*P < 0.05, **P < 0.01, ***P < 0.001). 

Results

Transgenic Mice Overexpressing the Tsp-1 Domain of Collagen XVIII

A cDNA encoding the first 325 N-terminal amino acid residues of the mouse collagen XVIII short isoform and an HA-tag for specific detection of the transgenic protein was generated by PCR (Fig. 1, lower panel). While representing most of the N-terminal noncollagenous portion of the short isoform, the Tsp-1 domain also is included in the more extensive noncollagenous N-termini of the medium and long isoforms (Fig. 1). The cDNA was inserted into an expression vector with a K14 promoter and the plasmid was injected into fertilized oocytes to create transgenic mice lines. The transgenic founder mice were mated with FVB/N mice to generate separate lines and bred to homogeneity. The transgenic mice were identified by Southern blotting and PCR (data not shown), and two independent transgenic founders, named K14-Tsp-1(1) and K14-Tsp-1(2), were obtained. The K14-Tsp-1(1) transgenic mice also were mated with B6 mice and bred to homogeneity. 

The transgene expression was assessed by Western blot analysis of eye protein extracts, which revealed a 50-kDa fragment with an anti HA-tag antibody in the FVB-K14-Tsp-1(1) transgenic mice, corresponding to the expected size of the transgene product (Fig. 2A). The transgene was found in immunohistochemical staining to be expressed in the basal cells of the corneal epithelium, the corneal endothelium, and the lens capsule in the K14-Tsp-1(1) and K14-Tsp-1(2) mouse lines (data not shown, and Figs. 2B, 2C, respectively), showing a similar expression pattern to that reported previously for the K14 promoter. ^21,26 Positive labeling for the HA-tag also was detected occasionally in the vitreous of the transgenic mice (Supplementary Fig. S1B). In view of the transgene expression pattern, the two new, independent mouse lines were used to assess the consequences of overexpression of the collagen XVIII Tsp-1 domain in the eye, and the observed phenotypic alterations assessed relative to null mice. The phenotypic changes observed in the lines K14-Tsp-1(1) and K14-Tsp-1(2) were highly similar. 

Figure 2

 

Transgene expression patterns, weights, and volumes of the K14-Tsp-1 transgenic and Col18a1^−/− eyes. (A) Transgene expression in the eye was analyzed by Western blotting. The anti-HA antibody recognized a major 50-kDa fragment in the K14-Tsp-1(1) transgenic mouse (Lane 2), while the control mouse eye was negative (Lane 1). Immunohistochemical stainings of eye samples from FVB/N controls (B) and K14-Tsp-1(2) (C) mice with the HA-tag antibody. The basal layer of the corneal epithelium (arrow), corneal endothelium (red arrowhead), and lens epithelium (black arrowhead) is stained in the transgenic mouse, but no staining was detected in the control mouse. The transgenic K14-Tsp-1 mice and Col18a1^−/− mice in an FVB/N background developed buphthalmos. (D) The eye of a 10-week-old control FVB/N mouse. (E) Enlargement of the eye of a 10-week-old K14-Tsp-1(2) mouse. Comparison of eye volumes of K14-Tsp-1(1), K14-Tsp-1(2), and FVB/N control mice at age of 12 months (F), and weights at ages of 6 to 12 months (G). (H) The axial length is greater in the K14-Tsp-1 transgenic eyes than in the FVB/N control eyes. Comparison of eye volumes (I) and weights (J) in Col18a1^−/− and FVB/N control mice at age 3 months.

Overexpression of the K14-Tsp-1 Transgene Leads to Buphthalmos and Phthisis Bulbi

Eyes of K14-Tsp-1(1) and K14-Tsp-1(2) mice in the FVB/N background had buphthalmos, that is, enlargement of the eye, by the age of 2 weeks (not shown). The eye enlargement continued gradually until it reached a plateau at around 12 months (Figs. 2D–E, at 10 weeks of age). To confirm our initial observations, we measured the volumes, weights, and axial lengths (AL) of K14-Tsp-1(1) and K14-Tsp-1(2) mouse eyes, and compared them to age-matched FVB/N control eyes. The eye volumes of the adult transgenic mice (12 months) were significantly larger than those of the control mice (mean ± SD, 39.8 ± 3.9 mm³ for K14-Tsp-1(1), 43.2 ± 5.9 mm³ for K14-Tsp-1(2), and 22.5 ± 2.4 mm³ for control FVB/N mice; P < 0.001, Fig. 2F). The eyes of the adult transgenic mice (6–12 months) also were significantly heavier (average weights, 44.2, 50, and 30 mg for FVB-K14-Tsp-1(1), FVB/N-K14-Tsp-1(2), and control mice, respectively, Fig. 2G). Also, the average ALs of the K14-Tsp-1(1) and K14-Tsp-1(2) eyeballs were significantly greater than for the FVB/N controls (4.19 ± 0.13 and 3.97 ± 0.28 mm vs. 3.66 ± 0.16 mm, Fig. 2H). 

In addition, in vivo and postmortem analyses pointed to a high incidence of phthisis bulbi, that is, shrinkage of the eye, among the K14-Tsp-1(1) and K14-Tsp-1(2) mice relative to the age-matched controls by the age of 9 months or more (5/18, 5/16 vs. 0/16, respectively). Phthisis was seen as an end stage of the eye enlargement. These eyes were excluded from the size analyses. In contrast to the FVB/N background, the average eyeball AL of the K14-Tsp-1(1) mice with a B6 background was the same as in the controls (3.73 ± 0.08 vs. 3.66 ± 0.11 mm, P = 0.22, not shown). Two of the 18 B6-K14-Tsp-1(1) mouse eyes analyzed had gone into phthisis, whereas none of the 16 B6 control mice or any of the FVB/N control mice had ocular phthisis. 

Lack of Collagen XVIII Leads to Buphthalmos and Phthisis Bulbi in a FVB/N Background

Since our original analysis of Col18a1^−/− mice was performed in a C57BL/6J background, ⁸ we decided, given that FVB/N gave a more severe phenotype for the Tsp-1 transgenics, to reanalyze the Col18a1^−/− mice in a similar genetic context. This revealed a significantly larger eye volume (24.7 ± 4.6 vs. 18.7 ± 2.3 mm³, P < 0.001, Fig. 2I) and a significantly greater eye weight (29 ± 4.7 and 23 ± 1.5 mg, respectively, P < 0.001, Fig. 2J) relative to controls even at the age of three months. In addition, as many as 8 of the 14 Col18a1^−/− eyes analyzed were exhibiting phthisis bulbi by 9 months. These eyes were excluded from the size analyses. Thus, the null phenotype is significantly more severe in the FVB/N background than in the C57BL/6J background investigated previously. ^8,10,11  

Changes in IOP

We hypothesized that one reason for the increased eyeball size would be increased IOP, which would stretch the developing eye. For that reason we measured IOPs of the mice at ages between 2 and 6 months. At the age of two months, there were no statistically significant differences in the average IOPs between K14-Tsp-1(1) and K14-Tsp-1(2) mice when compared to control FVB/N mice (24.4 ± 4.9, 20.0 ± 3.3, and 20.4 ± 5.6 mm Hg, respectively, Supplementary Fig. S2A). Surprisingly, at the age of six months, the average IOP of the K14-Tsp-1(2) mice was lower than in controls (14.6 ± 4.5 vs. 19.0 ± 4.3 mm Hg, respectively, P < 0.01), but the IOP of K14-Tsp-1(1) mice did not differ from that seen in controls (22.6 ± 9.0 vs. 19.0 ± 4.3 mm Hg in controls, Supplementary Fig. S2B). In the case of the Col18a1 ^−/− mice, at the age of 3 months their average IOP was lower relative to control mice (14.8 ± 6.2 vs. 18.7 ± 3.8 mm Hg, P = 0.054, Supplementary Fig. S2C), and at the age of 6 months the difference was statistically significant (13.7 ± 6.8 vs. 19.0 ± 4.3 mm Hg, P < 0.05, Supplementary Fig. S2D). 

Anterior Abnormalities

The transgenic mice and their age-matched controls were subjected to in vivo slit-lamp examinations and postmortem analysis under a dissection microscope at the age of 9 months or more to assess possible anterior abnormalities. These were easy to observe in the FVB/N background due to the transparency of the tissues. Cataract was present in the majority of the transgenic mice, where 89% of the K14-Tsp-1(1) and 94% of the K14-Tsp-1(2) eyes showed nuclear, cortical, or capsular changes in the lenses. Cataracts also were seen in 56% of the control eyes, but the changes in the control lenses in vivo were subtle relative to the mature cataracts seen in most of the K14-Tsp-1(1) (83%) and K14-Tsp-1(2) (88%) eyes, and none had had a mature cataract. Cold cataracts can develop in mouse eyes under general anesthesia, the speed of development being temperature-dependent. ²⁷ To minimize the formation of cold cataracts, mice were anesthetized as small groups, kept on heated mats, and examined within 15 minutes of anesthesia. Moreover, striking opacity of the eyes can be seen in K14-Tsp-1 mice without anesthesia (i.e., during IOP measurements), which further supports our finding of more frequent and more severe cataract formation in mutant mice. In addition, subluxation of the lens (Figs. 3D, 4C), corneal vascularization (Fig. 3E), folds in Descemet's membrane, intraocular hemorrhage, and hyperemia of the iris (not shown) were seen in some of the transgenic mouse eyes, whereas none of the above could be observed in the FVB/N controls (Figs. 3A–C). Filamentous material was observed posterior to the iris in some K14-Tsp-1(2) eyes, connecting the (subluxated) lens to the area around the ora serrata (Fig. 3F), while corneal opacities were seen in 17% of the K14-Tsp-1(1) eyes. 

Figure 3

 

Anterior abnormalities in the eyes of K14-Tsp-1 transgenic and Col18a1^−/− mice. Subluxation of the lens (arrow), [D]) and corneal vascularization (inset, [E]) were seen in the eyes of the transgenic mice, whereas neither of these findings were observed in the FVB/N controls (A–C). Filamentous material was observed posterior to the iris in some K14-Tsp-1(2) eyes, connecting the (subluxated) lens to the area around the ora serrata (black arrow and white arrow in inset, [F]). Some additional abnormalities were found in the Col18a1^−/− mouse eyes. Excessive material was bulging from the anterior lens capsule of a phthisic eye (inset, [G]). Phthisic eyes occasionally showed folds in the Descemet's membrane (inset, [H]) and hemorrhagia (arrow, [I]).

Figure 4

 

Subluxation of the lens in the K14-Tsp-1 mice. The OCT demonstrated a deeper anterior chamber in the K14-Tsp-1 mice (B) than in the FVB/N controls (A). Lens subluxation was detected in over 30% of the transgenic eyes either by OCT (C), in a slit-lamp examination or upon postmortem analysis.

Although the eyes of the transgenic B6-K14-Tsp-1(1) mice appeared grossly normal, slit-lamp examination showed that the transgenic mice had excessive mucous discharge and increased incidence of lashes growing/bending towards the cornea (not shown). Whereas none of the control B6 eyes had corneal opacities, over 30% of the 18 B6-K14-Tsp-1(1) eyes had corneal opacities (mostly central), usually accompanied by corneal vascularization. Clinically, the anterior chamber (AC) of the K14-Tsp-1(1) mice appeared somewhat deeper than average. Altogether an AC of normal depth was detected in less than 30% of the transgenic eyes, whereas only 19% of the control eyes analyzed appeared to have a slightly deeper AC when examined with a hand-held slit-lamp. When the eyes were dissected and prepared for histologic analyses it became evident that the B6-K14-Tsp-1(1) mice also had more severe cataracts in their lenses than the controls, although the incidence of cataracts was similar (17% and 13%, respectively). The slit-lamp examination results are summarized in the Table. 

Table

 

Results of Slit-Lamp Examination

	FVB/N			C57BL/6OlaHsd
	Control	K14-Tsp-1(1)	K14-Tsp-1(2)	Control	K14-Tsp-1(1)
Cataractic changes	9 (16)	16 (18)	15 (16)	2 (16)	3 (18)
Mature cataract	0 (16)	15 (18)	14 (16)	0 (16)	0 (18)
Subluxation of lens	0 (16)	2 (18)	3 (16)	0 (16)	0 (18)
Corneal vascularization	0 (16)	6 (18)	3 (16)	0 (16)	4(18)
Folds in Descemet's membrane	0 (16)	0 (18)	2 (16)	0 (16)	0 (18)
Corneal opacity	0 (16)	3 (18)	0 (16)	0 (16)	6 (18)
Intraocular hemorrhagia	0 (16)	1 (18)	2 (16)	0 (16)	0 (18)
Hyperemia of the iris	0 (16)	1 (18)	4 (16)	0 (16)	0 (18)
Phthisis bulbi	0 (16)	5 (18)	5 (16)	0 (16)	2 (18)

 Figures in parentheses represent the numbers of eyes analyzed.

The eyes of the FVB/N-Col18a1^−/− mice were analyzed only post mortem, since the changes seen in postmortem analysis of the K14-Tsp-1(1) and K14-Tsp-1(2) mice had been similar to those in the in vivo slit-lamp examination. As in the transgenic mice, there were several abnormalities in the collagen XVIII null mouse eyes. Some of the eyes not affected by phthisis showed excessive material bulging from the anterior lens capsule (Fig. 3G), while most of the phthisic eyes did not have a visible AC and those that did showed folds in Descemet's membrane (Fig. 3H). The phthisic eyes showed irregular vascularization on either the iris or the lens (in the FVB/N mice it was hard to distinguish which because of tissue transparency, not shown) and occasional hemorrhage (Fig. 3I). 

The TEM revealed an excessive accumulation of fibrillar material in the aqueous humor near the ciliary body in the FVB/N-K14-Tsp-1 transgenic mice (Figs. 5A, 5B), and some transgenic mice had inflammatory cells at the site of this fibrillar material (Fig. 5C), on the top of the lens capsule (Fig. 5D) and inside the lens capsule (not shown). Occasionally, the lens capsule was seen to be loose and folded (Fig. 5E), and electron-dense spots resembling intraocular calcium deposits were seen in the lens capsule, and in the space between lens and ablated retina (Fig. 5F). 

Figure 5

 

The TEM revealed accumulations of fibrillar material and inflammatory cells in the eyes of the transgenic mice and abnormalities in the lens capsule. Excess fibrillar material (arrow) was found in the aqueous humor near the ciliary body in the transgenic mice (B) relative to the FVB/N control mice (A). There was some accumulation of inflammatory cells at the sites of this fibrillar material (arrows, [C]) and on the top of the lens capsule (arrows, [D]), and folding of the lens capsule (arrow, [E]) in the transgenic mice. (H) The electron–dense spots found in the lens capsule, and in the space between lens capsule and ablated retina of the transgenic eye (arrowheads) resemble calcium deposits. Scale bars: 2 μm in (A, B, F), 10 μm in (D), and 20 μm in (C, E). L, lens capsule; R, retina.

The impression of buphthalmos led us to quantify AC depth in the transgenic mice by OCT. This demonstrated subluxation of the lens in the FVB/N-K14-Tsp-1 transgenic mice, as over 30% of the K14-Tsp-1(1) and the K14-Tsp-1(2) eyes showed exceptionally deep Acs, with the iris flopping backwards and the lens displaced from its normal position behind the iris (Fig. 4C). Although the ACs of the B6-K14-Tsp-1(1) eyes appeared deeper than those of the controls in the slit-lamp examination, OCT did not reveal any statistically significant AC differences (not shown). In the FVB/N background, however, the ACs of the transgenic mice were on average deeper than in the littermate controls, even with the cases of subluxed lenses excluded (Figs. 4A, 4B). 

Posterior Findings

Although the photoreceptors are lacking in FVB/N mice, the remaining cell populations formed a well-organized structure. Interestingly, most of the K14-Tsp-1(1) and K14-Tsp-1(2) eyes in the FVB/N background showed histologic and TEM abnormalities in the retina and vitreous. Most of the nonphthisic eyes showed either peripheral retinal degeneration or atrophy (Fig. 6B), or else ablation that could vary from a small local lesion to a panretinal effect (Fig. 6C). These were not observed in any of the control mice (Fig. 6A). The ablations were judged to have occurred in vivo rather than being an artifact due to eye enucleation, since some organized structures (collagens, blood vessels) could be detected in the space between the lens capsule and the ablated retina (Fig. 6D). Only one K14-Tsp-1(1) mouse eye in the B6 background had peripheral retinal atrophy and one had a subtle local misplacement of the photoreceptor nuclei. Two phthisic B6-K14-Tsp-1(1) eyes also had total retinal ablation (not shown), and histologic analysis showed total retinal ablation in all of the FVB/N-Col18a1^−/− retinas analyzed (Figs. 7A–C). In addition, the cataractous lenses had a disorganized appearance (Fig. 7C). 

Figure 6

 

Posterior findings. The transgenic eyes showed peripheral retinal degeneration (arrow, [B]) and local or total ablation of the retina (arrow, [C]). These were not observed in any of the FVB/N control mice studied (A). Organized structures at the ultrastructural level could be detected in the space between the lens capsule and ablated retina in the transgenic eyes (D). Scale bars: 500 μm in (A, B), 2 μm in (D). C, collagen.

Figure 7

 

Histologic overview of Col18a1^−/− eyes as compared to controls. (A) Representative photograph of FVB/N control eye section. (B–C) Representative photographs of eye sections of Col18a1^−/− mice illustrating retinal detachment (arrows), and the disorganized appearance of a cataractous lens (arrowhead, [C]). ON, optic nerve.

Distinct Expression Patterns of Collagen XVIII Short and Medium/Long Isoforms

To understand how the different collagen XVIII isoforms contribute to structures in the eye, we determined the localization of Col18a1-encoded proteins in mice that were deficient for the short form (which is transcribed using promoter 1 and called Col18a1^P1/P1 ), or mice that were deficient for the medium and long forms (which are transcribed using promoter 2 and called Col18a1^P2/P2 ). Eyes were stained with an antibody that labels all isoforms (anti–all-18) and an antibody that labels only the medium and long isoforms (anti–medium/long-18, antibody epitopes marked in Fig. 1). In control FVB/N mice, the anti–all-18 antibody labeled the BM zones of the ciliary body, iris epithelia, lens capsule, inner limiting membrane, and Bruch's membrane (Figs. 8A, 8E). In Col18a1^P2/P2 mice, which lack the medium and long isoforms, labeling with anti–all-18 antibody was similar to FVB/N control mice, indicating that the short isoform is present in all of these locations (Figs. 8C, 8G). In Col18a1^P1/P1 mice that lack the short isoform, the anti–all-18 antibody did not label the lens capsule and inner limiting membrane, indicating that the medium and long isoforms are not highly expressed at these sites, and that the signal in control mice was detecting the short isoform (Figs. 8B, 8F). Consistent with these data, the anti–medium/long-18 antibody labeled the ciliary and iris epithelia, and Bruch's membrane, but did not label the lens capsule or inner limiting membrane in FVB/N control and Col18a1^P1/P1 mice (Supplementary Figs. S3A, S3B, S3E, S3F). This antibody showed no labeling in Col18a1^P2/P2 mice (Supplementary Figs. S3C, S3G), and both antibodies showed no specific labeling in Col18a1^−/− eyes that were used as negative controls (Figs. 8D, 8H, Supplementary Figs. S3D, S3H). It should be noted that the anti–medium/long-antibody gave unspecific staining in some of the posterior eye structures, but specific staining was observed in the Bruch's membrane (Supplementary Figs. S3E–H). In addition, the vessels around the lens capsule (tunica vasculosa lentis [TVL]) were positive for collagen XVIII. The positive signal in TVL was seen in Col18a1^P2/P2 mice with the anti–all-antibody (Fig. 8C) and in Col18a1^P1/P1 with the anti–medium/long-antibody (Supplementary Fig. S3B), indicating that this structure contains both isoforms. 

Figure 8

 

Immunofluorescence stainings of FVB/N control eye and Col18a1^P1/P1 , Col18a1^P2/P2 , and Col18a1^−/− mouse eyes at postnatal day 7 with the anti–all-18-antibody (recognizing all isoforms of collagen XVIII) showed distinct expression patterns for the short and medium/long isoforms. (A) In the anterior part of the FVB/N control mouse eye collagen XVIII is localized to the BM zones of the ciliary body and iris epithelia (red arrow and red arrowhead, respectively) and to the lens capsule (green arrow). (B) In the Col18a1^P1/P1 mice the collagen XVIII immunosignal can be seen in the epithelia of the ciliary body and the iris (red arrow and red arrowhead, respectively). (C) The anti–all-18 staining pattern in the Col18a1^P2/P2 mice is similar to that in the FVB/N control mice, that is, collagen XVIII positivity could be seen in the epithelia of the ciliary body and the iris (red arrow and red arrowhead, respectively), and in the lens capsule (green arrow). In addition, the TVL is positive for collagen XVIII (green arrowhead). (D) The eyes of the Col18a1^−/− mice are negative for collagen XVIII. In the retinas of the FVB/N control (E) and Col18a1^P2/P2 (G) mice, collagen XVIII is localized to the inner limiting membrane (red arrow) and Bruch's membrane (green arrow). (F) In the Col18a1^P1/P1 mice the staining is restricted to Bruch's membrane (green arrow). (H) No collagen XVIII immunosignal could be seen in the Col18a1^−/− retina. I, iris; ILM, inner limiting membrane; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; CH, choroid. Scale bar: 100 μm.

Abnormal Outgrowth of Retinal Vessels is Due to Lack of the Short Isoform

The availability of isoform-specific null mice allowed us to investigate the significance of the short and medium/long collagen XVIII isoforms for generation of the markedly delayed and aberrant outgrowth of retinal vessels observed previously in the Col18a1^−/− mice completely lacking collagen XVIII. ⁸ Staining with the blood vessel markers CD31 and CD34 showed irregular growth of capillaries in the retina of 21-day-old Col18a1^P1/P1 individuals (Fig. 9B), similar to that observed in the Col18a1^−/− (Fig. 9D), whereas the retinal vessels of the Col18a1^P2/P2 mice (Fig. 9C) were comparable with the B6 control eyes (Fig. 9A). Thus, the broader expression pattern observed for the short isoform in the retina is reflected in the important role that this isoform has in retinal vessel formation, whereas the medium/long isoforms are not required for this. 

Figure 9

 

Abnormal outgrowth of retinal vessels is due to the absence of the short isoform. Staining of flat-mounted retinas from 21-day-old mice with the blood vessel markers CD31 and CD34 showed irregular growth of capillaries in the Col18a1^P1/P1 mice (B), similar to that observed in the Col18a1^−/− mice (D), whereas the retinal vessels of the Col18a1^P2/P2 mice (C) were comparable to those in the B6 control eyes (A). Scale bar: 100 μm.

Conclusions

A total of 21 mutations in the COL18A1 gene has been reported in Knobloch patients, most of which are predicted to create premature stop codons, thus, often affecting the C-terminal end of the protein, which is common to all collagen XVIII isoforms, while one missense mutation has been shown to reduce the binding affinity of the C-terminal endostatin domain for some extracellular matrix components. ^2,28 The present analysis of mice specifically lacking either the short or the medium/long collagen XVIII isoforms indicated that the epithelia of the ciliary body, and the iris and Bruch's membrane express all the isoforms, whereas the short form is the sole isoform in the lens capsule and the inner limiting membrane. Importantly, we found that absence of the short isoform was sufficient to cause the aberrant vascularization of the retina previously reported for mice lacking all isoforms of collagen XVIII. ⁸ In contrast, the normal retinal vascular development in mice lacking the medium/long isoforms suggests that these are not crucial for normal development of the posterior eye structures. Nevertheless, the fact that the medium/long isoforms are present in the anterior eye structures and Bruch's membrane supports the notion that mutations affecting all isoforms, as is the case in most Knobloch patients, lead to a more severe form of the disorder. ²⁹ However, to our knowledge no mutations affecting the third exon of COL18A1, as is specific to the medium/long isoforms, have been reported in Knobloch patients, and the present findings suggested that if such mutations were to exist, the patients might not be characterized by sufficiently severe eye abnormalities to be categorized clinically as suffering from the Knobloch disorder. 

Overexpression of the Tsp-1 domain, as found with all isoforms and identified as representing most of the noncollagenous N-terminus in the case of the short variant, also led to serious eye abnormalities, with cataract in most lenses and phthisis in many cases. In addition to its sites of expression in the corneal and lens epithelia, the transgene product also was detected in the vitreous, possibly reflecting its secretion and accounting for the pronounced phenotypic consequences. We also extended our characterization of the eyes of Col18a1^−/− mice to include analyses in the FVB/N background, which is known to harbor a retinal deficiency. ²² A significantly more severe phenotype, including buphthalmos and phthisis, was observed in the FVB/N-Col18a1^−/− eyes than had been reported previously for the same genetic modification in the C57BL/6J background. ⁸  

Transgene expression had an effect on eye size in the FVB/N background, in that both transgenic mouse lines exhibited larger eyes than control mice, with no statistical difference between the two transgene groups. The occurrence of larger eyes may be related to myopia and myopic changes. It is well known that myopic eyes have a greater risk of ruptures of the retina developing with accompanying ablation. ³⁰ Indeed, almost all the transgenic FVB/N mice showed peripheral thinning of the retina and some had either peripheral ablation or total retinal ablation. The incidence of total retinal ablation or phthisis was increased in the FVB/N-Col18a1^−/− mice, and this often was accompanied by intraocular inflammation and degradation of the lens. It already has been shown in the C57BL/6 background that absence of collagen XVIII results in ciliary body atrophy and abnormal vascularization of the retina through persistent hyaloid vessels. ^8,10,11 The vascular pattern of the FVB/N transgenic mice retina was not studied here, and, hence, tractional ablation due to persistent vitreal vessels cannot be excluded. 

Lack of collagen XVIII in the FVB/N background resulted in larger eyes and a tendency to lower IOP at the age of 6 months. Interestingly, it has been shown previously that lack of collagen XVIII in the C57BL/6J background also results in lower IOP at the age of 6 months, which coincides well with the findings in this study. ¹¹ The IOPs of the transgenic mice showed some variability compared to control mice, but overall were not significantly increased. Since individual mice were not studied longitudinally, we cannot exclude the possibility that some of the transgenic mice may have suffered from increased IOP, which leads to enlargement of the eye globe, and eventually to lens subluxation, retinal detachment, and phthisis. However, especially in the line K14-Tsp-1(2), we found no indication of increased IOP and at the age of 6 months the level was lower than in controls. Also to be considered is whether the observed excessive material near the ciliary body, the changes in the lens capsule, and the subluxation of the lens would be related to pseudoexfoliation syndrome. In the Nordic countries this syndrome is associated strongly with two single-nucleotide polymorphisms in the lysyl oxidase gene LOXL1, an important regulator of connective tissue maturity and turnover, but this is known to be population-dependent and other connective tissue–related genes may be involved. ³¹ Thus, it is possible to conceive that the lack of collagen XVIII in mice could result in strain-dependent phenotypic changes. However, pseudoexfoliation has not been reported to be a hallmark of Knobloch syndrome in humans (caused by mutations in the collagen XVIII gene), whereas vitreoretinal degeneration, retinal detachment, zonular weakness (ectopic lenses), cataract, phthisis, and high myopia are observed and suggested to be pathognomonic. ² Moreover, retinal degeneration and enlargement of the eye are not associated typically with pseudoexfoliation syndrome. All in all, the findings in the transgenic mice, including cataract, lens luxation/subluxation, retinal degeneration and detachment, and phthisis, coincide well with those observed in Knobloch syndrome in humans. ^1–7  

Although not identical, the transgene phenotypes were similar in many respects to those observed for Col18a1^−/− in the FVB/N background. These effects extended beyond the sites of expression of the transgenic product, namely the cornea and lens capsule. The C-terminal endostatin domain of collagen XVIII molecules is known to bind to a number of BM components, ^32,33 but very little is known about the properties of the other domains. The Tsp-1 domain of collagen XVIII is homologous with the extracellular matrix protein thrombospondin-1 known to be able to modulate cell-matrix interactions and have anti-angiogenic properties. ³⁴ We consider it possible that the excess amount of the collagen XVIII Tsp-1 domain produced in the cornea and lens of transgenic mice could interfere with the functions of the full-length collagen XVIII produced in these mice, for example, by competing for the binding activities of the endogenous protein, with deleterious effects on its functions. Hence, overexpression of the N-terminal Tsp-1 domain or the C-terminal endostatin domain of collagen XVIII would lead to deleterious effects with respect to the lens, but the buphthalmos and strong phthisic tendencies were not observed in the latter case. This difference is likely to reflect different roles for the N- and C-terminal domains in BMs in which collagen XVIII resides. 

Altogether, our findings demonstrated the critical role of the short collagen XVIII isoform in the eye, the known mutations seen in Knobloch patients are predicted to lead to absence of the collagen XVIII protein, but this hypothesis has been confirmed only in two patients. ³⁵ Our work with transgenic mice overexpressing either the Tsp-1 domain or the C-terminal endostatin domain ²¹ suggested that, even though truncated protein is produced in Knobloch patients as a consequence of missense mutations, the consequences may be deleterious. 

Supplementary Materials

pdf S1 

Acknowledgments

The authors thank the Biocenter Oulu transgenic and electron microscopy core facilities for their services, Jukka Veijola for performing the OCT measurements, and Jaana Peters, Jaana Träskelin, and Aila White for technical assistance. 

Supported by the Health Science Council of the Academy of Finland (Grant 138866 and Centre of Excellence 2012-2017 Grant 251314), and by the Sigrid Jusélius Foundation. 

Disclosure: M. Aikio, None; M. Hurskainen, None; G. Brideau, None; P. Hägg, None; R. Sormunen, None; R. Heljasvaara, None; D.B. Gould, None; T. Pihlajaniemi, None