| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
,2
Department of Oral and Developmental Biology, Harvard School of Dental Medicine, Boston, Massachusetts, USA;
* Departments of Pathology and
Medicine, University of Hamburg, Hamburg, Germany;
Schepens Eye Research Institute and Department of Ophthalmology, Harvard Medical School, Boston, Massachusetts, USA; and
Cardiovascular Research Division of Massachusetts General Hospital, Boston, Massachusetts, USA
3Correspondence: Department of Oral and Developmental Biology, Harvard School of Dental Medicine, 188 Longwood Ave., Boston, MA 02115, USA. E-mail: bjorn_olsen{at}hms.harvard.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: type VIII collagen • corneal endothelial cells • proliferation
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
2
3
4)
. As a member of the short-chain collagen family, collagen VIII molecules are characterized by a short triple-helical domain flanked by a shorter N-terminal nontriple helical NC2 domain and a longer C-terminal nontriple helical NC1 domain (5
, 6)
. Two distinct polypeptides, 
1(VIII) and 2(VIII), have been identified. The two polypeptides show homology to each other and to the 1 chain of type X collagen (7
8
9)
; recent evidence suggests they form two types of homotrimeric molecules in tissues (10
, 11)
. Type VIII collagen is a component of basement membrane-associated matrices in most organs. However, its most prominent location is in Descemet’s membrane (DM), a thick basement membrane of the corneal endothelium, in which collagen VIII molecules (1(VIII) and 2(VIII) homotrimers) are major constituents and assemble into a characteristic hexagonal lattice structure (12)
. This lattice structure forms a highly ordered array in a region called the anterior banded zone (ABZ) (13
, 14)
. The ABZ is first secreted in utero by the endothelial cells and DM consists entirely of this embryonic banded portion at birth (15)
.
Type VIII collagen has been proposed to play a structural role in DM by creating a matrix underneath the endothelial cells that can resist compression while maintaining an open porous structure (6)
. Although this collagen has been well characterized structurally, its functional importance is still unclear. In humans, mutations in the 2(VIII) gene result in corneal endothelial dystrophies, namely, Fuchs’ endothelial dystrophy (FECD) and posterior polymorphous corneal dystrophy (PPCD) (16)
. In Fuchs’ dystrophy, DM is markedly thickened by additional abnormal layers posterior to the normal banded layer and the presence of wart-like excrescences. PPCD can be associated with vesicular-like lesions, localized or diffuse thickenings of DM and focal replacement of normally amitotic endothelial cells by abnormally proliferating cells (16
, 17)
.
To obtain better insight into the precise role of collagen VIII in organ development, we have generated mice deficient in 1(VIII) and 2(VIII) collagen chains. We report here that mice lacking type VIII collagen appear to develop normally and have a normal life span. They exhibit no histological abnormalities in most organs. However, Col8a1–/–/Col8a2–/– null mice develop a distinct phenotype of dysgenesis of the anterior segment of the eye with a globoid, keratoglobus-like protrusion of the anterior chamber. The corneal stroma is diffusely thin, similar to what is seen in human keratoglobus. Descemet’s membrane is markedly thinned and lacks the anterior banded zone. The corneal endothelial cells are enlarged and reduced in number. Finally, mutant corneal endothelial cells show a decreased ability to proliferate in response to different growth factors in vitro, suggesting that collagen VIII may function as an enhancer of growth factor-induced proliferation of cells.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Construction of the targeting vectors
Col8a1 targeting vector
A murine 129/Sv lambda genomic library (Stratagene, La Jolla, CA, USA) was screened using a Col8a1 cDNA as a probe. A clone of 17 kb was isolated, restriction mapped, and partially sequenced. An 11 kb EcoRI/SalI fragment was subcloned into pBluescript. This fragment spans the whole coding region of the Col8a1 gene. A positive selectable cassette was introduced at the BamHI site to create the replacement vector pUB8a1KO. This cassette contains stop codons in all three frames, an internal ribosomal entry site (IRES), followed by the EGFP gene with an SV40 polyadenylation signal and a neomycin phosphotransferase gene (NEOr) with polyadenylation signal, under control of the mouse PGK promoter. The cassette was generated as follows: the EGFP/poly A gene was amplified by PCR using plasmid pEGFP-N1 (Clontech, Palo Alto, CA, USA) as a template (sense primer: 5'-ACTATGGTGAGCAAGGGCGA-3'; antisense primer: 5'-AATTTACGCCTTAAGATACAT-3'), blunted, ligated downstream of an internal ribosomal entry site. A PGK/NEO/polyA gene was blunted and ligated downstream of the EGFP gene. Finally, BamHI sites were introduced at both ends of the cassette using synthetic BamHI linker (Biolabs, Beverly, MA, USA). The structure of the construct was confirmed by sequencing.
Col8a2 targeting vector
Col8a2 genomic clones were isolated by screening the 129/Sv lambda library with a Col8a2 cDNA clone. One 10 kb clone, which spans both coding exons, was subcloned into pBlueskript (pNFm129-28) and characterized by restriction enzyme mapping and sequencing. A 1.3 kb SacI/KpnI fragment was subcloned to generate plasmid pNFm129-24. A positive selectable cassette containing the neomycin phosphotransferase gene (NEOr) with polyadenylation signal, under the control of a mouse PGK promoter, was introduced into a unique restriction site PflMI in exon 2. pNFm129-24 was then religated into vector pNFm129-28. Finally, the herpes simplex virus thymidine kinase gene, with polyadenylation signal and controlled by a PGK promoter, was cloned in the antisense direction of the NEOr gene for negative selection.
Transfection of ES cells, screening for homologous recombination, and genotyping
Col8a1
To obtain Col8a1 chimeric mice, 50 µg of the Col8a1KO construct was first linearized with EcoRI, electroporated into J1 embryonic stem cells, and selected for resistance with G418 as described (18)
. 198 G418-resistant ES cell clones were analyzed for homologous recombination by long template PCR choosing the sense primer inside the NEOr gene (NEO-FOR: 5'-CAGCGCATCGCCTTCTATCGC-3') and the antisense primer outside of the construct (LTa1-REV: 5'-ACAGCTACTAAGACAGGTGCAGAC-3') to amplify a 2 kb fragment. For further analysis, DNA of positive clones were digested with Bgl II, separated on a 0.8% agarose gel, and transferred to a nylon membrane (Hybond N+; Amersham Pharmacia Biotech, Piscataway, NJ, USA). Two probes, a 1.4 kb Sac I/EcoRV fragment chosen from a region downstream of the construct (external probe) and a 600 bp XhoI/Not I fragment inside the EGFP gene (internal probe), were labeled with 32P dCTP by random primer extension using the High Prime Labeling KIT (Roche, Mannheim, Germany). Of 198 G418-resistant embryonic stem cell clones analyzed, two contained the correct targeting event. Both independent clones were injected into C57BL/6J blastocysts to obtain chimeric mice. Male chimeras were mated to C57BL/6J mice and offspring screened for transmission of the disrupted allele. Genomic DNA was obtained from tail biopsies by digestion with proteinase K (19)
and screening performed by Southern blots. Further genotyping was performed by PCR to detect the mutated (sense primer 8A1-KO-FOR: 5'-GTGGGGGTGGGGTGGGATTAGATA-3'; antisense primer 8A1-KO-REV: 5'-CTCGGCCCAAGAACCCCAGGAAC-3') and the wild-type alleles (sense primer 8A1-FOR: 5'-CGGGAGTAGGAAAACCAGGAGTGA-3'; antisense primer 8A1-REV: 5'-GGCCCAAGAACCCCAGGAACA-3'). The mice were backcrossed for 8–10 generations into C57BL/6J mice.
Col8a2
To obtain Col8a2 chimeric mice, 50 µg of the construct was linearized, electroporated into J1 embryonic stem cells, and selected for resistance to G418. ES clones were screened using Kpn I digested DNA for Southern blot with an XbaI/PstI probe or by Long Template PCR using sense primer (sense primer f-8KO: 5'TGCATACTCATATTAGCACTGCCTGATACTAGC-3') upstream of the vector and an anti-sense primer (Neo1 : 5'-GCTGCTAAAGCGCATGCTCCAGACTGCCTT-3') within the NEO cassette. Nine of 218 ES clones showed evidence of homologous recombination; of these, four showed homologous recombination without random integration. One clone was injected into C57BL/6 blastocysts and four male chimeras were obtained and mated to C57BL/6J females. Further genotyping was performed by PCR to detect the mutated (sense primer NF-7: 5'-CCGGTAAAGTATGTGCAGC-3'; antisense primer f-NEO: 5'-CAGCGCATCGCCTTCTATCGC-3') and the wild-type alleles (sense primer NF-7: 5'-CCGGTAAAGTATGTGCAGC-3'; antisense primer NF-10: 5'-ATCCTGGGAACATTGCAGG-3'). The mice were backcrossed for at least 12 generations into C57BL/6J mice. Col8a1–/–/Col8a2–/– mice were generated by mating Col8a1 with Col8a2 knockout mice.
Northern blot analyses
Total RNA was extracted from heads of newborn mice with TRI reagent (Sigma, St. Louis, MO, USA). 20 µg of RNA was separated on a 1% agarose gel using a glyoxal-based system (Ambion, Austin, TX, USA) and blotted by downward capillary transfer on a nylon filter. Hybridization was performed at 42°C using a DNA probe containing a 1.2 kb BamHI/EcoRV fragment of 3'UTR of the Col8a1 gene and a 0.5 kb EcoRV/SpeI fragment of 3'UTR of the Col8a2 gene. The ß-actin probe was amplified as a 540 bp PCR-product with intron flanking primers (sense: 5'-GTGGGCCGCTCTAGGCACCAA-3'; antisense: 5'-CTCTTTGATGTCACGCACGATTTC-3'; Clontech).
Histology and immunohistochemistry
All major internal organs (brain, lung, liver, heart, kidney, three different parts of the stomach, duodenum, jejunum, ileum, colon, spleen, testes, ovary), skeletal muscle, aorta, bladder, trachea, skin, thymus, pancreas, salivary glands, and all major joints were dissected from 6-wk, 4-month, 6-month, 1- and 2-year-old mice and fixed overnight in 4% paraformaldehyde in phosphate-buffered saline (PBS) or Bouins fixative. Tissues were dehydrated, embedded in paraffin, and sections 5 µm in thickness were stained with hematoxylin and eosin (H&E) and observed in a light microscope. Kidney sections were stained with periodic acid-Schiff reagent (PAS). Whole eyes were fixed in 10% neutral buffered formalin, dehydrated, then infiltrated and embedded in Historesin (Leica, Heidelberg, Germany) or paraffin. Sections (3 µm) were cut and stained with H&E or PAS. Pictures were taken with an Axioscope microscope (Zeiss, Oberkochen, Germany) and the depth of the anterior chamber was measured by the computer program Axiovision 3.1.
For immunohistochemistry, corneas of adult mice and heads of embryos were fixed in 4% buffered PFA, infiltrated with 20% sucrose and frozen in OCT. Sections of wild-type and Col8a1–/–/Col8a2–/– eyes 5 µm thick were rehydrated, treated with proteinase XXIV (Sigma), blocked in 5% goat serum in PBS, and incubated for 1 h at 37°C with rabbit anti-EGFP antibodies at a dilution of 1:500 (Molecular Probes, Eugene, OR, USA). A nonimmune rabbit serum was used as a control. Fluorescein-anti-rabbit IgG was applied as secondary antibody (DAKO, Glostrup, Denmark). Pictures were taken at various magnifications with an Axioscope fluorescence microscope (Zeiss).
Electron microscopy
Mice at the age of 2 to 6 months were killed and the cornea of four wild-type, four Col8a1–/–/Col8a2–/–, three Col8a1–/–, three Col8a2–/–, and two Col8a1+/–/Col8a2+/– mice were dissected and fixed at room temperature for 2 h in 1.25% formaldehyde, 2.5% glutaraldehyde, and 0.03% picric acid in 0.1M cacodylate buffer pH 7.4. After postfixation with 1% osmiumtetroxide and 1.5% potassiumferrocyanide for 1 h in the dark, specimens were stained en bloc in 1% uranyl acetate for 30 min, dehydrated, and embedded in Epon/Araldite (Embed 812: Electron Microscopy Sciences, Fort Washington, PA, USA). Ultrathin sections were contrasted with uranyl acetate and lead citrate. Pictures were taken in an JEOL 1200EX electron microscope.
Cell culture and characterization of corneal endothelial cells
Mouse corneal endothelial cells (mCECs) were released from isolated Descemet’s membrane using a modification of a published method (2)
. In brief, corneas of four wild-type and four knockout mice (backcrossed for eight generations to C57Bl/6 with their second grade control littermates) were dissected and washed in 1x Antibiotic/Antimycotic (Sigma) in HBSS. Descemet’s membrane with endothelium was peeled off and digested in 0.2% dispase II (Roche) and 0.2% bacterial collagenase (Sigma) in HBSS for 30 min at 37°C. Cells were maintained on gelatin-coated plates in DMEM, 10% serum and 1% glutamine, 100 U/mL penicillin, 100 µg/mL streptomycin (Irvine Scientific, Santa Ana, CA, USA) at 37°C and 5% CO2. For cell characterization, cells were grown on glass coverslips and fixed with methanol for 10 min on ice for immunostaining or with 10% formalin for hematoxylin staining. Immunostaining was performed at room temperature using the mouse monoclonal anti-Collagen VIII antibody 8C (Seikagaku, Tokyo, Japan) with an isotype-matched control mouse IgG1 (DAKO) and rabbit anti-EGFP antibodies (Molecular Probes) with nonimmune rabbit IgG as a control. Secondary antibodies were fluorescein-anti-mouse IgG and fluorescein-anti-rabbit IgG (DAKO). Pictures were taken at 20x magnification with an Axioscope fluorescence microscope (Zeiss). RT-PCRs of Col8a1, keratin 14, decorin, and GAPDH on RNA of corneal endothelial cells and renal tubular epithelial cells as non-ocular epithelial control cells were performed as described before (20)
.
Proliferation assays
mCECs of passages 2–6 (3x103 cells/well) were plated on gelatin-coated 96-well plates and maintained in DMEM medium supplemented with 10% FCS for 24 h to allow cell attachment. Medium containing 0.1% serum was added for 24 h to synchronize the cells in G0 phase. DNA synthesis was measured by [3H]thymidine incorporation: 2 µCi/mL [3H]thymidine was added for 16 h in the presence of either 10% FCS, 25 ng/mL EGF, 25 ng/mL NGF, 25 ng/mL bFGF, or 0.1% FCS (unstimulated control). Concentrations of growth factors were as published (21
22
23)
. At the end of the incubation period, cells were washed twice with PBS and trypsinized. The cell suspensions were subsequently harvested onto a filter paper using an automated cell harvester (Dynatech Labs., Chantilly, VA, USA) before [3H]thymidine incorporation was measured in a ß-scintillation counter. Results were plotted as means of 12 different values. Experiments were repeated three times with two different cell preparations. Statistical significance between individual groups was tested using the nonparametric unpaired Mann-Whitney U test. A P value of <0.05 was considered significant.
Attachment assays
Attachment assays were performed by coating 96-well tissue culture plates with gelatin for 1 h. Corneal endothelial cells were suspended in serum-free DMEM containing 1 mg/mL BSA and plated at a density of 6000 cells/well. Nonadherent cells were rinsed off with PBS after 24 h, and the attached cells were fixed with 10% formalin, washed, and stained with hematoxylin. After washing the cells three times with PBS, the absorbance at 630 nm was measured in a microplate reader (Dynatech Labs.). Means represent the OD at 630 nm of 24 different values. The experiment was repeated three times with two independent cell preparations.
Whole mount staining of mouse corneas
Col8a1–/–/Col8a2–/– mice at the age of 4 to 8 months of both mouse lines (derived from the two different ES cell lines) were compared with wild-type cousins of the same line. Ten corneas of wild-type and 14 corneas of Col8a1–/–/Col8a2–/– mice were removed, fixed in ice cold methanol for 10 min, and washed and permeabilized in 1% Triton-X100 for 10 min at room temperature. After blocking with 4% BSA in PBS, corneas were incubated with a 1:150 dilution of rabbit anti-ZO-1 antibody (Zymed Laboratories, San Francisco, CA, USA) for 2 h at room temperature. Specimens were washed, incubated with fluorescein-conjugated anti-rabbit IgG (DAKO) for 1 h, dilution 1:2000, and cut four times to allow them to flatten. Images were collected using 10x magnification and morphometrically analyzed using the computer program KS300.1.
Quantitative RT-PCR
For PCR amplifications of Col8a1, Col8a2, and ß-actin, first-strand cDNA was used with components of the SYBER Green Core PCR Reagent Kit (PE ABI, Foster City, CA, USA) on an iCycler iQ real-time PCR system (Bio-Rad). The cycling parameters were 50°C, for 3 min, 94°C for 10 min, followed by 50 cycles of 94°C for 30 s, 60°C for Col8a1, 65°C for Col8a2 for 30 s, and 72°C for 30 s. Three separate experiments were performed to quantitate the expression of each gene. In each experiment, three or more identical 20 µL PCR reactions of the gene of interest and the control gene ß-actin were run using the primers for ß-actin:sense: 5'-GTGGGCCGCTCTAGGCACCAA-3'; antisense: 5'-CTCTTTGATGTCACGCACGATTTC-3'; (Clontech), for Col8a1: sense: 5'-TCTGCCACCTCAAATCCCTCCTCA-3', antisense: 5'-TCTCCGCGCAAACTGGCTAACG-3', for Col8a2: 5'-GAGCGACGCGGAGTTCTG-3', antisense: 5'-CCCCCTGCATGGCGTCTGTG-3'. Relative expression levels of Col8a1 mRNA and Col8a2 mRNA in Col8a1–/–/Col8a2–/– organs were calculated separately for each of the three experiments. The relative expression values represent average values from three repeats of the experiments ± standard deviation. Threshold cycles were determined using the default threshold levels (10-fold standard deviation from cycle 2 to 10) and the average threshold cycles of the gene of interest were normalized for amplification of ß-actin, assuming an amplification efficiency of 2.0 in all reactions. The reaction products were analyzed on 3% Nusieve gels to verify the absence of nonspecific PCR products.
Statistical analysis
All values are presented as means ± standard deviation (SD). All experiments were repeated a minimum of three times. Statistical analyses were performed using SSPS version 10.1.3 for Windows and Microsoft Excel 97. Statistical significance between the wild-type and Col8a1–/–/Col8a2–/– values of Fig. 4B
and Fig. 7
were tested using the nonparametric unpaired Mann-Whitney U test. A P value of <0.05 was considered significant. Data of Fig. 4A
were analyzed by 1-way ANOVA with Tamhane post-tests. Level of significance was set at P < 0.05.
|
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
A). Of 198 G418-resistant ES cell clones analyzed, two were found to have undergone homologous recombination at the Col8a1 locus; use of an internal probe demonstrated that a single copy of the targeting vector had been integrated into the genome (Fig. 1B-D
). Both stem cell lines were used to produce chimeras.
|
The Col8a2 gene was targeted with a positive/negative selection scheme using neomycin resistance for positive selection and thymidine kinase for negative selection (Fig. 2
A). Long template PCR and Southern blot analyses of DNA from 218 ES cell clones revealed four positive clones; an internal probe revealed a single integration (Fig. 2B-D
). One ES cell line was used to produce chimeras.
|
Col8a1/Col8a2-deficient mice are viable, fertile, and show no major abnormalities
Homozygous mutant Col8a1 and Col8a2 mice were crossed to generate double heterozygous mice, which in turn were intercrossed to generate double null mutants. To confirm the disruption of both alleles of both genes, the absence of Col8a1 and Col8a2 transcripts was checked by Northern blot analysis. Total RNA of heads of newborn mice showed no detectable collagen VIII-specific transcripts (Fig. 1E
, Fig. 2E
). Collagen type VIII-deficient mice appeared healthy, and no significant difference in growth rate was observed in control and collagen VIII-deficient mice. Collagen VIII-deficient mice were fertile and lactated normally, and no reduction in litter size was noted, indicating that neither fetal nor maternal collagen VIII is essential for embryonic development. Mice heterozygous for Col8a1, Col8a2, or for both null mutations were mated to check for Mendelian transmission with the following results. Of 325 Col8a1 mice, 80 (25%) offspring were wild-type, 158 (49%) heterozygotes, and 87 (27%) were homozygotes; of 395 Col8a2 mice, 107 (27%) were wild-type, 186 (47%) heterozygotes, and 102 (26%) were homozygotes. Of 422 Col8a1+/–/Col8a2+/– mice 27 were homozygotes, and 22 offspring were wild-type. Thus, the null mutations do not cause embryonic lethality, and mice completely lacking type VIII collagen are viable. Collagen VIII-deficient and control mice were aged up to 23 months and showed no significant difference in their life span.
All major internal organs, skeletal muscle, skin, salivary glands, and all major joints were dissected from 6-wk to 2-year-old mice and checked for macro- or microscopic changes, but no major anatomical abnormalities were found. Collectively, these results indicate that disruption of both collagen VIII genes does not affect fertility, development, or the life-span of mice in cages.
Lack of collagen VIII leads to an abnormal anterior chamber of the eye
Sections of eyes were made at the age of postnatal 2, 3, and 4 days, at 6, 7, 8, and 12 wk, 6 months, and 1 and 2 years. Macroscopic investigation revealed a keratoglobus-like depth of the anterior chamber as seen in Fig. 3
A, B. To confirm the increased depth of the anterior chamber, sections through the centers of 14 wild-type and 21 Col8a1–/–/Col8a2–/–eyes were made and the distances between the lens and the corneal endothelial cell layer were measured along the central axis of the eye (Fig. 3E, F
, Fig. 4
B). Col8a1–/–/Col8a2–/–eyes revealed a mean depth of 392 ± 60 µm in contrast to wild-type eyes with 223 ± 49 µm (P<0.001). Differences were already seen at the age of 6 wk. This difference remained in aging mice and did not change with age. Sections of 10 Col8a1–/– eyes revealed a mean depth of 324 ± 48 µm, and in 8 sections of Col8a2–/– eyes a mean depth of 243 ± 65 µm was found (Fig. 4B
).
|
DM was markedly, and the corneal stroma slightly, thinner in collagen VIII null mice than in wild-type mice (Fig. 3G, H
, Fig. 4A, C
). To rule out the possibility of errors due to oblique sections through the cornea, we measured the thickness of the corneal epithelial layer, since the measured thickness of all these layers (DM, corneal stroma, and epithelium) would be increased proportionally in nonperpendicular sections. Compared in this way, the thickness of the corneal stroma was reduced by 21.5 (±5.3)% in Col8a1–/–/Col8a2–/– eyes, by 12.3 (±4.3)% in Col8a1–/– eyes, and not reduced in Col8a2–/– eyes. The epithelium was the same thickness in wild-type and mutant mice (Fig. 4C
). No anterior synechias, guttata (wart-like expansions), or regions of corneal cloudiness were observed (Fig. 3C, D
). Endothelial cells did not continue onto the iris surface and Schlemm’s canal as seen in PPCD in humans, and the trabecular meshwork was visible and open.
Ultrastructural abnormalities in Col8a1–/–/Col8a2–/– corneas
Transmission electron microscopy revealed that DM in wild-type mice was 1.3 µm thick and showed two layers: a small anterior dense and a posterior more homogenous layer comparable to, but fainter, than the anterior banded and posterior nonbanded layers in human corneas (24)
. In contrast to human corneas, there were abundant polymers of type VIII collagen within the posterior layer of wild-type mice with periodic banding at 100 nm intervals. In 2-year-old mice, DM showed an uneven appearance with focal thickening. This is different from human membranes, which increase uniformly in thickness with age (25)
. Even wild-type mice showed attenuated endothelial cells at this age. This is probably due to age-related loss of endothelial cells and an impaired ability of corneal endothelial cells to proliferate. Col8a1–/–/Col8a2–/– mice showed a markedly thinner DM of 0.7(±0.1) µm in contrast to the wild-type membrane of 1.3 ± (0.2) µm already at the age of two months (Fig. 4A
, Fig. 5
, P<0.001). The electron-dense anterior zone and any banding were lacking. To confirm that the difference in thickness was not due to a tissue processing artifact, we performed transmission electron microscopy of frozen sections and verified the results. Descemet’s membranes of Col8a1–/– or Col8a2–/– mice revealed a thickness of 1.0 (±0.1) µm in both cases (each P<0.001 compared with wild-type). A dense posterior layer was visible, but banded structures were only detectable in Col8a1–/– mice.
|
Corneal endothelial cells were cuboidal and contained large and flattened nuclei in Col8a1–/–Col8a2–/– eyes. Endothelial cell borders interdigitated and overlapped normally. At the age of 2 years, there were no additional pathologic layers of material between endothelial cells and DM and no subepithelial fibrocellular tissue or subepithelial fluid as seen in corneal dystrophies in humans. There were no iridocorneal adhesions or lens abnormalities.
Decreased proliferation of corneal endothelial cells in Col8a1/Col8a2 null mice
To test the hypothesis that the thinning of the cornea may be due to a decrease in cell proliferation during corneal development, we isolated Col8a1–/–/Col8a2–/– and wild-type corneal endothelial cells (mCECs) and compared their ability to proliferate in response to various growth factors and 10% serum. The isolated corneal endothelial cells showed the expected paving-stone mosaic of polygonal cells (26)
. mCECs of Col8a1–/–/Col8a2–/– mice were immunostained against EGFP (Fig. 6
A, B). Staining mCECs with antibodies against markers specific for vascular endothelial cells such as CD31/PECAM-1 or vWF was negative (data not shown). RT-PCR assays were performed on RNA of corneal endothelial cells of wild-type and Col8a1–/–/Col8a2–/– mice and renal tubular epithelial cells as described (20)
. Wild-type mCECs were positive for Col8a1 RNA (Fig. 6G
), a corneal endothelial specific marker, and decorin RNA (Fig. 6H
), a marker for corneal endothelium and epithelium but not for stromal fibroblasts. Col8a1–/–/Col8a2–/– mCECs were positive for decorin RNA. All mCECs were negative for keratin14 RNA (Fig. 6I
), a marker for corneal epithelium. GAPDH was used as a control (Fig. 6J
). Immunostaining with antibodies against collagen VIII (Fig. 6C, D
) showed a punctate staining pattern in wild-type cells, consistent with previous descriptions of stained collagen VIII-synthesizing macrophages (27)
, at least up to passage 6. Only faint background staining was visible with anti-collagen VIII in mCECs of Col8a1–/–/Col8a2–/– mice and in wild-type mCECs with isotype matched control IgG1.
|
Proliferation of cells lacking type VIII collagen and control cells was measured by [3H]thymidine incorporation (Fig. 7
A). In basal medium (0.1% FCS) lacking added growth factors, mCECs did not proliferate, and depletion of type VIII collagen had no effect on incorporation of thymidine (wild-type: 8927±863 cpm; Col8a1–/–/Col8a2–/–:10,768±1050 cpm). Incubation of mCECs with 10% serum for 16 h strongly induced proliferation of control and Col8a1–/–/Col8a2–/– mCECs, but [3H]thymidine incorporation was significantly less in Col8a1–/–/Col8a2–/– mCECs (42,237±5493 cpm vs. wild-type 57,666±11,823 cpm, P<0.05). To rule out the possibility that the decrease in counts for Col8a1–/–/Col8a2–/– mCECs was due to an impaired ability of these cells to attach, we performed attachment assays with cells that had been attached for 24 h. There was no significant difference in attachment between wild-type (OD 0.17±0.02) and Col8a1–/–/Col8a2–/– mCECs (OD 0.16±0.03) after 24 h (Fig. 7B
). After incubation with 25 ng/mL EGF, proliferation was reduced by 32% (17,963±2552 cpm vs. 12,181±1705 cpm; P<0.00001), with NGF by 37% (17,963±3140 cpm vs. 11,029±1617 cpm, P<0.00001), and with bFGF by 25% (21,526±2721 cpm vs. 16,168±1956 cpm; P<0.0001).
Corneal endothelial cells in Col8a1–/–/Col8a2–/– mice display polymegathism and polymorphism
One typical manifestation of corneal dystrophies is the appearance of large and polymorphic endothelial cells. These findings are most apparent by in vivo microscopy such as specular microscopy or by scanning electron microscopy (28)
. To test whether polygonal or polymegal cells are present in Col8a1–/–/Col8a2–/– mice, we stained whole mounts of corneas with an antibody against ZO-1, a peripheral membrane protein associated with tight junctions between endothelial cells (29)
. To describe accurately normal or diseased endothelial cells, three variables have been suggested (24)
: 1
) cell size, 2
) the spread of cell size in terms of the standard deviation, and 3
) heterogeneity of the endothelial population. The mean cell size was 936 ± 73 pixels for wild-type cells and 1095 ± 76 pixels for Col8a1–/–/Col8a2–/– cells (P<0.005, comparing Col8a1–/–/Col8a2–/– and 10 wild-type corneas). The histogram of cell size is skewed to the right for Col8a1–/–/Col8a2–/– corneas because of the presence of more large cells (Fig. 8
).
|
Expression of 1(VIII) collagen in the developing and postnatal eye
To gain further insight into the expression pattern of 1 (VIII) collagen, we had inserted the EGFP gene into the Col8a1 replacement vector so that EGFP would be expressed under the control of the Col8a1 promoter in the targeted Col8a1 null allele (see Materials and Methods). Since the expression of EGFP was insufficient for detection by direct fluorescence microscopy, we enhanced the fluorescence using immunofluorescence detection with an anti-EGFP antibody (Fig. 9
). Wild-type mice and nonimmune rabbit serum were used as negative controls (Fig. 9D, G, H
). We observed strong staining in the condensation of mesenchymal cells forming the future corneal stroma and endothelium at E14.5 (Fig. 9A, B
). During E15.5 (Fig. 9C, D
) to E17.5 (Fig. 9E, F
), staining remained in most posteriorly condensed mesenchyme, seen as a layer of mesothelium (corneal endothelium) and in more regularly arranged future stromal cells. Stromal and endothelial staining was seen until the first postnatal day (Fig. 9G, H
). Additional staining was detected in the pigmented epithelium underneath the developing retina and in the pigment epithelium of the ciliary processes at the growing edge of the optic cup (Fig. 9B, I
) (30)
. In the adult cornea, strong staining was evident in corneal endothelial cells (Fig. 9K
) and in the trabecular meshwork (Fig. 9L
); no staining was apparent in the stroma or the corneal epithelial cell layer.
|
Up-regulation of Col8a1 mRNA in the aorta of Col8a2–/– mice
Functional redundancy within protein families has been proposed for several mice that show only minor phenotypes following targeted inactivation of genes encoding different extracellular matrix proteins (15
, 31
32
33)
. This raises the question of whether the targeted deletion of Col8a1 gene function might have an effect on the transcriptional activity of the Col8a2 gene and vice versa. Col8a1 and/or Col8a2 are expressed in the eye, aorta, kidney, brain, heart, muscle, and lung (2
, 3
, 34
35
36)
. Real-time RT-PCR analyses of RNA prepared from adult Col8a1–/– mice showed a decrease of Col8a2 mRNA in almost all tissues (eye: 31±0.5%, spleen: 54±0.9%, aorta: 47±1.5%, kidney: 42±1.1%, brain: 69±2.2%, lung: 51±1.8%, liver: 50±0.6%, muscle: 55±3.6%, intestine: 63±4.8%, and testis: 90±6.9%), except for the heart (103±2.6%) (Fig. 10
A). In contrast, Col8a1 mRNA in Col8a2–/– mice decreased in the eye (41±1.2%), kidney (57±1.5%), lung (39±1.4%) muscle (72±2.2%), liver (49±1.1%), spleen (19±0.9%) and testis (75±3.3%), but there was a strong increase of Col8a1 mRNA in aorta (228±3.6%) and a slight increase in the heart (127±3.9%) (Fig. 10B
). Transcript levels did not seem to be altered in the brain (110±1.1%) and intestine (111±2.4%).
|
These findings suggest that Col8a1 mRNA is up-regulated in the aorta of Col8a2–/–mice in response to the loss of Col8a2, but Col8a1 does not compensate for Col8a2 in other tissues. In fact, Col8a1 expression is reduced in most tissues of Col8a2–/– mice; the same is true for Col8a2 expression in Col8a1–/– mice.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
38
39)
. This developmental process is regulated by a complex and incompletely understood series of interactions between the mesenchymal cells, cytokines and the extracellular matrix (ECM). Overexpression of TGF and EGF or TGFß1 in the ocular lens or lack of TGFß2 leads to multiple anterior segment defects in mice (39
, 40)
. Lack of 1(I) collagen leads to abnormally thin corneas with disorganized collagen fibrils (41)
, and mice deficient in lumican exhibit deregulated assembly of collagen fibrils in the cornea (42)
. The data presented here demonstrate that mice with inactivated Col8a1 and Col8a2 genes manifest a dysgenesis of the anterior segment of the eye with a globoid, keratoglobus-like protrusion of the anterior chamber: the depth of the anterior chamber is increased and the corneal stroma is thin. A thin Descemet’s membrane separates the corneal stroma from the corneal endothelial cells, which are enlarged and decreased in number.
The globoid, keratoglobus-like protusion of the anterior chamber with a thin corneal stroma suggests that expression of type VIII collagen during corneal development is important for establishing the appropriate number of stromal fibroblastic layers. We suggest that the reduced number of these layers (see Fig. 5
) in collagen VIII null mice may be due to a reduced ability of periocular mesenchymal cells to proliferate and/or migrate in the absence of collagen VIII. Collagen VIII is expressed by a number of rapidly proliferating cells such as different tumor cells and endothelial cells during angio- and vasculogenesis and throughout the development of hemangiomas (34
, 43
44
45
46
47
48
49
50
51
52)
. Pericellular collagen VIII may therefore function during development to help maintain cells in a proliferative state. Consistent with this hypothesis is the finding that collagen VIII-deficient corneal endothelial cells derived from collagen VIII-expressing periocular mesenchyme show reduced proliferation in vitro (compared with wild-type cells) and are in vivo reduced in number and therefore enlarged (Figs. 7
, 8)
. Collagen VIII has also been suggested to act as a chemotactic factor for mesenchymal cells such as smooth muscle cells and to provide a substrate for enhanced migration (50
, 53)
. It is possible, therefore, that the presence of collagen VIII during migration of periocular mesenchymal cells into the developing cornea (see Fig. 9
) provides a migratory substrate for the cells. A reduced migration of cells into the developing cornea could conceivably result in a reduced number of layers of corneal fibroblasts and collagen layers in the collagen VIII null eyes. Obviously, a reduction in proliferation and migration could contribute to the thinning of the cornea seen in knockout eyes.
Mutations in VSX1, a member of the Vsx group of vertebrate paired-like homeodomain transcription factors (54)
, have been reported in patients with keratoconus or PPCD (55)
. One of the mutations resulting in keratoconus (R166W) alters the homeodomain and impairs binding of VSX1 to DNA. Other sequence changes are missense mutations close to the homeodomain; these could also affect DNA binding. Since mice lacking collagen VIII display keratoglobus, a condition that in humans resembles keratoconus with a generalized globoid protrusion of a clear, diffusely thin cornea and a deep anterior chamber, it is tempting to speculate that collagen VIII is downstream of VSX1 during the development of the cornea.
Thinning of the cornea is also seen in mutations affecting the synthesis of other matrix components of the cornea. For example, mice deficient for the small proteoglycans lumican or keratocan have thin corneas, but show a decreased iridocorneal angle, unlike the collagen VIII-deficient mice in which this angle is increased (5
, 42)
. Furthermore, mutations in keratocan in humans cause the more severe, recessively inherited form of cornea plana (56)
. Thus, it is likely that keratoglobus in the collagen VIII-deficient mice is caused by a combination of corneal thinning and the absence of collagen VIII from the Descemet’s membrane, and not by corneal thinning alone.
The posterior corneal phenotype of Descemet’s membrane thinning and reduced number of endothelial cells in collagen VIII knockout mice is different from what is seen in patients with PPCD and FECD, both conditions for which mutations in COL8A2 have been reported (16)
. In FECD, a primary degenerative disease that affects the corneal endothelium, Descemet’s membrane is markedly thickened by additional abnormal layers of material and wart-like outgrowths (57)
. The basic endothelial cell abnormality in FECD is not known, but altered synthesis and/or assembly of extracellular matrix components are suspected (58)
. In PPCD, endothelial cells often retain their ability to divide, become multilayered and extend onto the trabecular meshwork (59)
. Localized or diffuse thickenings of Descemet’s membrane with deposits of abnormal collagenous material and flattened, enlarged and widely spaced endothelial cells are also a common finding (17)
. In normal eyes, the anterior zone of banded fibrils in the Descemet’s membrane reaches its maximal thickness by the time of birth and appears to remain unchanged thereafter. However, in corneal endotheliopathies secretion of type VIII collagen appears to be resumed, resulting in an abnormal accumulation of wide-spaced collagen in a posterior collagenous layer subjacent to the endothelial cells (15)
. Such alterations have not been observed in collagen VIII KO mice, even in old mice. The reduced number and abnormal size distribution of corneal endothelial cells in the mice is reminiscent of the reduced endothelial cell number reported for familial cases of PPCD (30)
. However, the cellular mechanisms are likely quite different. In the mice, the reduced number of corneal endothelial cells is most likely the result of a decrease in proliferation of the endothelial cells during embryonic development, whereas excessive apoptosis has been proposed as a mechanism in the case of FECD (60
, 61)
.
Thus, the pathogenetic mechanisms are probably quite different between mice with abnormalities caused by lack of collagen VIII and corneal endotheliopathies, even when some aspects of the phenotypic outcome look similar. The most likely explanation for these differences is that the reported mutations in COL8A2 in the human corneal dystrophies cause abnormal collagen VIII to be deposited in the Descemet’s membrane, in close proximity to the endothelial cells. The presence of this abnormal collagen VIII may result in increased endothelial cell proliferation and/or increased apoptosis. In contrast, complete lack of collagen VIII results in decreased endothelial proliferation and a thin Descemet’s membrane without the anterior banded collagen VIII-containing zone. The thinning of the Descemet’s membrane in the Col8a1–/–/Col8a2–/– double mutant was about twice the thinning seen in the single mutants (Fig. 4A
). In addition, banded structures within the Descemet’s membrane were completely absent in the Col8a2–/– mutant eyes. This suggests that 1(VIII) and 2(VIII) homotrimeric molecules may be localized within different regions of Descemet’s membrane and that the banded structures depend primarily on the presence of 2(VIII) homotrimers. Finally, our data clearly demonstrate that while collagen VIII is a major component of the anterior portion of Descemet’s membrane, collagen VIII molecules are not essential for the formation of this membrane during eye development.
The differential effects on transcript levels of the expressed collagen VIII polypeptide in mice homozygous for targeted alleles of Col8a1 or Col8a2 are consistent with the conclusion that collagen VIII molecules are homotrimers of 1(VIII) or 2(VIII) chains and that different tissues/cells regulate the proportion between the two types of proteins in different ways. In most tissues, the absence of one type of collagen VIII molecule results in a reduced expression level of the other, but in the aorta, the loss of 2(VIII) collagen results in a 2-fold increase in 1(VIII) mRNA levels. In contrast, the lack of 1(VIII) collagen leads to a reduced level of 2(VIII) mRNA, even in the aorta. The basis for this striking difference between the two collagen VIII proteins in the aorta is not clear, but one possibility that needs to be considered in future studies is that lack of 2(VIII) collagen results in up-regulation of 1(VIII) collagen in smooth muscle cells in the aortic wall. Collagen VIII is dramatically up-regulated in smooth muscle cells after damage to arterial endothelium and intima (53
, 62
, 63)
. It is therefore possible that absence of 2(VIII) collagen in aorta generates "microdamage" and a "collagen VIII response" by smooth muscle cells. Detailed future studies comparing the structures and functional properties of the aortas of the two single collagen VIII null mice described here should provide insights into this and other possibilities.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
2 Present address: Novartis Institutes for Biomedical Research, Cambridge, MA 02139, USA.
Received for publication October 7, 2004. Accepted for publication March 30, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
0.7 µm in electron micrographs. Descemet’s membrane of Col8a1–/– and Col8a2–/– mice had an average thickness of 1 µm. Col8a1+/–/Col8a2+/– mice had no differences in thickness of Descemet’s membrane compared with wild-type mice. B) Depth of the anterior chamber. The depth was distinctly increased in Col8a1–/–/Col8a2–/–, moderately increased in Col8a1–/– and normal in Col8a2–/– mice. C) Thickness of corneal stroma compared with wild-type mice. The corneal stroma was clearly thinned in Col8a1–/–/Col8a2–/–, moderately thinned in Col8a1–/– and normal in Col8a2–/– mice. No differences in the thickness of the epithelial layer were seen between the mutant and wild-type eyes.
