Development of Congenital Stromal Corneal Dystrophy Is Dependent on Export and Extracellular Deposition of Truncated Decorin

Abstract

Purpose.: Congenital stromal corneal dystrophy (CSCD) is an autosomal dominant condition with clouding of the cornea due to stromal opacities. It is caused by mutations in the decorin gene (DCN) leading to the expression of a truncated form of decorin. In an attempt to replicate this condition in mice, a knock-in mouse strain, 952delT Dcn, was created.

Methods.: Mice were constructed by targeted mutation. Sequencing of genomic DNA confirmed correct genotype. Mouse and human corneas, including corneas from patients with CSCD, and primary keratocyte cultures were subjected to Western blot analysis, transmission electron microscopy, and immunofluorescence microscopy.

Results.: Histologically, the mice did not show any organ pathology. Corneas were clear, and the electron-lucent deposits observed in CSCD were not present. Furthermore, while nearly equivalent amounts of normal and truncated decorin are present in CSCD corneas, truncated decorin was hardly detectable in the mouse corneas. By immunofluorescence analysis of corneas from 952delT Dcn homozygous mice, decorin was found only in keratocytes. In primary cultures of mouse corneal explants, truncated decorin was retained intracellularly in contrast with human corneal explants where truncated decorin was exported into the culture medium. Immunofluorescence analysis revealed that native mouse decorin localized to the Golgi complex, whereas the truncated decorin accumulated in the endoplasmic reticulum (ER).

Conclusions.: The ER retention of truncated decorin may explain why the mouse corneas remained clear. The consequences of the decorin mutation are different in mice and humans, and 952delT Dcn knock-in mice are therefore not a suitable model for CSCD.

Decorin belongs to a family of small leucine-rich proteins. It participates structurally in the spacing of collagen fibrils by maintaining appropriate interfibrillar distance and regulates lateral growth of the fibril. Decorin also is known to be involved in cell proliferation and in extracellular signaling processes through epidermal growth factor receptor in tumor cells, and through insulin-like growth factor receptor in renal cells.^1,2 

Mutations in human decorin are associated with autosomal dominant congenital stromal corneal dystrophy (CSCD).^3–6 All mutations described so far are located in the last exon of the gene and result in the production of a truncated form of decorin lacking the 33 C-terminal amino acids. The CSCD is characterized by the presence of small opacities in the corneal stroma, seen by electron microscopy as amorphous areas with small filaments organized in an irregular manner. Experimental evidence suggests that these opacities are the result of deposition of decorin in the amorphous areas.⁷ 

The decorin gene is highly conserved in mammals. Mouse Dcn is approximately 80% identical to human DCN. Decorin null mice⁸ have clear corneas, and are reported to have normal fertility and no gross phenotype alterations. The skin is somewhat fragile, and there is abnormal collagen morphology in skin and tendons. Recently, a transgenic 952delT Dcn mouse was reported.⁹ The mutation corresponds to the human 967delT DCN first described to be associated with CSCD. The mutated mouse gene was inserted into a wild-type background Dcn^+/+, under the control of a chicken β-actin (CAG) promotor. Expression from the mutated gene was restricted to corneal tissue by mating with Kera-cre expressing mice. Expression from the mutated gene also was studied in Dcn^+/− and Dcn^−/− background. Abnormal keratocyte morphology¹⁰ and disrupted stromal structure⁹ were observed in these mice, but amorphous deposits of proteoglycans as seen in CSCD were missing. In HEK293 cells transfected with the mutated murine decorin gene, export of truncated decorin was severely impaired.¹⁰ 

We report here a novel mouse where a mutation corresponding to the human 967delT mutation (952delT) was inserted into the native decorin gene by homologous recombination. This targeted mutation reproduces the human genotype. In contrast with the mice described by Chen et al,⁹ our mice had clear corneas. When comparing corneal tissue from our 952delT Dcn mice with human CSCD, we observed nearly equivalent amounts of normal and truncated decorin in human CSCD corneas, while truncated decorin was hardly detectable in mouse corneas. Truncated decorin was retained intracellularly in cells from mouse corneal explants while in cells from human CSCD corneas truncated decorin was exported into the culture medium. Export of truncated decorin, thus, seems to be a prerequisite to produce the amorphous deposits of decorin typical of CSCD. 

Methods

Construction of Mice

The Dcn 952delT mice were constructed by the use of homologous recombination, which was performed by genOway, Lyon, France. In silico bioinformatic analysis predicted that the introduction of the corresponding human CSCD mutation (967delT) into the mouse decorin gene (952delT) would result in activation of an equivalent premature stop codon and deletion of the C-terminal 33 amino acids of the decorin protein. A construct was made consisting of intron 7 and exon 8 of the mouse decorin gene with the 952delT mutation flanked at the 3′ end by a validated FRT-neomycin-FRT-LoxP cassette, and a single LoxP site at the 5′end, positioned in intron 7 (Fig. 1). The linearized targeting construct was transfected into 129Sv/Pas ES cells by electroporation. A total of 760 G418 resistant clones was isolated, and of these, four ES cell clones were shown by PCR screening, direct sequencing, and Southern blot analysis to have the targeting vector with the correct 952delT mutation inserted by homologous recombination. Recipient blastocysts were obtained from pregnant C57BL/6J females. The four ES cell clones were injected into C57BL/6J blastocysts, and a total of 83 blastocysts were reimplanted into OF1 pseudo-pregnant females. In the offspring, contribution of the recombinant ES cells to each individual could be assessed using coat color markers. Seven male chimeras displayed a chimerism rate of 50% to 80%. Four highly chimeric males were selected for further breeding, and were bred with C57BL/6J Flp recombinase expressing deleter mice to excise the neomycin selection cassette. One animal revealed the presence of the excised allele and was, therefore, heterozygous for the floxed Decorin allele. This animal was bred further with C57BL/6J mice to obtain a pure line of neo-excised point mutation mice devoid of the Flp-transgene. Mice heterozygous for the 952delT mutation (B6.129S-Dcn^tm1Geno/Ub), hereafter called heterozygous knock-in (HetKI), and mice homozygous for the 952delT mutation, named homozygous KI (HomoKI) were established. DNA was purified from ear punch tissue by Wizard Genomic DNA purification kit (Promega, Madison, WI, USA). Mice carrying the mutation were screened by PCR (primers DcnKI forward: CCTCAATTCATTTGTGTCCAGTTAGGCC and DcnKI reverse: ACAAGGAGAGGAAAAAGGGAGGACCC, giving fragments of 254 base pairs (bp) for wild-type (wt) and 370 bp for KI alleles, respectively (Supplementary Fig. S1). In addition, a decorin knock-out mouse was generated by mating with Cre expressing mice. Exon 8 was thereby removed, which resulted in degradation of the decorin mRNA (data not shown). Mice were backcrossed into C57BL/6JBomTac (Taconic, Lille Skensved, Denmark), for more than 10 successive generations. The animal project was approved by the Norwegian Food Safety Authority, and adheres to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Housing and genotyping was according to Federation of Laboratory Science Associations (FELASA) guidelines. 

Figure 1

 

Construction of mice. A targeting vector consisting of exons 7 and 8 with the 952delT point mutation in exon 8 was created. By homologous recombination, the recombined locus was obtained. The PGK/Neo (phosphoglycerate kinase I/neomycin resistance) cassette was excised by breeding with ubiquitous Flp-expressing mice. In the resulting mice, production of the truncated decorin protein is under control of the endogenous promotor.

Autopsy and Light Microscopic Studies of Mice

Animals were euthanized by CO₂ inhalation, immediately placed at 4°C, and autopsy then was performed within 30 minutes. Lung, heart, liver, kidney, spleen, skin, and gastrocnemius tendon were sampled for histopathology. The samples were fixed in 4% formaldehyde in PBS, embedded in paraffin, sectioned at 3 μm, and stained with hematoxylin-eosin. 

Transmission Electron Microscopy

Corneal tissue was fixed in 2% glutaraldehyde/0.2 M cacodylate buffer, embedded in Epon and postfixed in osmium tetroxide (OsO₄). Ultrathin sections were stained with 2% uranyl acetate and Reynold's lead citrate before examination by transmission electron microscopy (JEM 1230; Jeol, Tokyo, Japan). 

Establishment of Primary Keratocyte Cultures From Mouse and Human Corneal Tissue

Human CSCD corneas were obtained after consent from patients undergoing corneal transplantation. Healthy corneas were obtained from eyes that were enucleated because of malignant melanoma, after consent from the donor. This part of the project was approved by the Regional Committee for Medical and Health Research Ethics, Western Norway (IRB# 00001872). 

Mouse and human corneal explants were cut into small pieces, seeded on tissue culture dishes and grown in Dulbecco's modified Eagle's medium (DMEM) nutrient mixture F12 HAM (Sigma-Aldrich Corp., St. Louis, MO, USA) or Amniochrome (Lonza, Verviers, Belgium) and maintained as described.¹¹ 

Homogenization of Corneal Tissue

Corneas were collected in Allprotect Tissue Reagent (Qiagen, Hilden, Germany) and kept at −20°C until use. The tissue was homogenized in lysis buffer by use of TissueLyser II (Qiagen) and subsequently O- and N-deglycosylated as described previously.⁷ 

Harvest of Culture Media and Homogenization of Cultured Cells

Cells were grown to confluence in 6-cm dishes in DMEM nutrient mixture F12 HAM (Sigma-Aldrich Corp.) supplemented with 5% fetal calf serum medium, 2 mM L-Glutamine, 100 U/mL penicillin, and 0.1 mg/mL streptomycin. At days 1, 3, 5, 8, and 12 after confluence, cells were washed with PBS, and incubated for 24 hours with serum free medium. Cell medium was collected and concentrated approximately 50-fold by acetone precipitation, and subsequently O- and N-deglycosylated.⁷ Cells were harvested in ice-cold PBS, pelleted and lysed in lysis buffer before O- and N-deglycosylated as described.⁷ 

Western Blot Analysis

Protein samples were subjected to gel electrophoresis (12% NuPAGE Novex Bis-Tris gel; Life Technologies, Carlsbad, CA, USA), and transferred to polyvinylidine fluoride (PVDF) membranes according to the manufacturers protocols. After blocking, membranes were incubated overnight at 4°C with rabbit anti-mouse decorin antibody (LF-113), kindly supplied by Larry W. Fisher¹² at a 1:200 dilution in 1% nonfat dry milk, 1% glycine, 1% BSA in PBS/0.01% Tween 20. After washing in PBS/0.01% Tween 20, membranes were incubated with horseradish peroxidase (HRP)-conjugated goat anti-rabbit secondary antibody (DAKO, Glostrup, Denmark) and development of signal was performed by Immun-Star WesternC Kit (Bio-Rad, Hercules, CA, USA) as a substrate. Images were obtained using the XRS BioRad chemiluminescent detection system. Molecular weight marker was MagicMark XP Western Protein Standard (Life Technologies). 

Immunofluorescence Analysis

Mouse and human primary keratocytes were grown on 18-mm diameter glass coverslips (Marienfeld Laboratory Glassware, Lauda-Königshofen, Germany) coated with 0.1 mg/mL poly-D-lysine (Sigma-Aldrich Corp.) until 5 days after reaching confluence. Staining was performed as described previously.¹³ The polyclonal antibody against mouse decorin (LF-113) was visualized using secondary Alexa Fluor 488 conjugated goat anti-rabbit F(ab')₂ fragments (Jackson ImmunoResearch, West Grove, PA, USA). 

The mouse GM130 antibodies (BD Transduction Laboratories, Lexington, KY, USA) detecting the Golgi apparatus in mouse and human cells and the mouse anti-Ribophorin 1 antibodies (a kind gift from Gerd Kreibich, New York University, New York, NY, USA) detecting the endoplasmic reticulum (ER) in mouse cells were both visualized using secondary TRITC-conjugated goat anti-mouse F(ab') fragments (Jackson ImmunoResearch). The goat antibody against human decorin (R&D Systems, Minneapolis, MN, USA) was visualized by Alexa Fluor 594–conjugated donkey anti-goat F(ab')₂ fragment (Jackson ImmunoResearch). Endoplasmic reticulum staining in human cells was obtained using antibodies against the signal peptidase complex subunit SPC-25, kindly provided by Stephen High (University of Manchester, UK). Images were obtained using a scanning confocal microscope (Leica SP2 AOBS; Leica, Wetzlar, Germany). 

Mouse eyes were fixed in ice-cold 96% ethanol, or in 3% paraformaldehyde in phosphate buffer for 20 hours and embedded in paraffin. Sections (6 μm) were deparaffinized in xylene, rehydrated in graded ethanol, and stained with antibodies as described above. 

Sequencing of mRNA

RNA was isolated from mouse corneas using RNeasy Mini Kit and TissueLyserII (Qiagen). Complementary DNA was synthetized using High Capacity RNA-to-cDNA kit (Applied Biosystems, Carlsbad, CA, USA) following the manufacturers protocol. Sequencing of the decorin gene and confirmation of the correct point mutation was by PCR, using primers spanning three overlapping regions of DNA. Primers used were dcn81F:ACAGAAGCGGTAACGAGCAG, dcn629R:TTTTCAGGCAGTTCCTTTAGTTG, dcn502F: AAGGACTTGCATACCTTGATCC, dcn970R: CCATAACGGTGATGCTGTTG, dcn831F: GATCCCTCAAGGTCTGCCTAC, dcn1408R: GTGGAAAAGGCATGCACAG. 

Results

Clinical Examination and Autopsy

Mice were constructed by homologous recombination of the modified decorin gene into ES cells, which then were introduced into the C57BL/6J mouse strain. The knock-in mice carry the mutation corresponding to the human point mutation in their genome, and the mutant gene is present in all cells. Production of normal and truncated decorin protein is regulated by the endogenous promotor. Heterozygous mice will have a genotype that is a replication of the human CSCD genotype. All strains of mutant mice, including homozygous and heterozygous carriers for the 952delT mutation and the decorin knock-out allele, bred normally and had healthy pups. No embryonal lethality was observed. Neither gross examination nor histopathological examination revealed pathological findings. By slit-lamp biomicroscopy corneas were clear, and red reflex was completely normal in all strains. Animals up to the age of 1 year were examined. In the human disease, the corneas are opaque, with multiple small opacities and red reflex is severely reduced. In some animals, unrelated to genotype, changes in the eyes were discovered that included microphthalmic eyes, macroscopically absent eyes, and eyes where the cornea was gray, and seemed adherent to the lens. These are changes common to inbred BL6 mice (JAX NOTES Issue 463, Fall 1995), and are not caused by the mutation in the decorin gene. 

Transmission Electron Microscopy

Electron microscopic studies of corneas from mice heterozygous or homozygous for the 952delT decorin mutation did not reveal the amorphous deposits that are present in the human corneas. Careful comparisons of TEM pictures were done at magnifications ranging from 2.5K to 50K. Figure 2 shows a representative overview of cross sections at 10K magnification of the different mouse corneas and a cross section of a representative human cornea (20K) with CSCD. The amorphous, electron lucent areas characteristic for CSCD as shown in the human specimen could not be found in any of the mouse specimens. In some of the cross sections, areas with irregularly spaced fibrils were seen, but such areas also were observed in the wt specimens. 

Figure 2

 

Transmission electron microscopy. Pictures show cross sections of mouse corneas (wt, heterozygous, and homozygous for the 952delT decorin mutation), and human corneas from a patient with CSCD (far right). *The area with electron lucent deposits characteristic for CSCD is marked in the human cornea. Areas like these were not identified in the corneas from mice.

Western Blot Analysis of Mouse and Human Corneal Lysates

Corneas from mice and humans with and without mutations in the decorin gene were homogenized, and O- and N-deglycosylated. In the CSCD lysate, a double band of wt and truncated decorin was visible. The amount of the protein core was approximately equal between the wt and truncated decorin forms. By contrast, only one decorin band corresponding to the wt form was seen in the corneal lysates from the wt and heterozygous KI mouse corneas. In the corneal lysate from homozygous KI mice, only a faint band was visible, probably corresponding to a very low level of the truncated form of decorin (Fig. 3). 

Figure 3

 

Decorin expression in corneal lysates. Corneal lysates were O- and N-deglycosylated, and the samples were separated on a 12% Bis-Tris gel before blotting onto a PVDF membrane. Membranes were incubated with antibodies against human or mouse decorin. In human samples from healthy individuals, only wt decorin is detected, whereas in samples from individuals with CSCD there is an additional band of truncated decorin.

Immunofluorescence Analysis of Corneal Tissue

When murine corneas were immunostained with decorin antibody (LF113), a strong decorin signal was observed mainly in the extracellular part of the stroma in corneas from wt and heterozygous animals, whereas in corneas from animals homozygous for the 952delT Dcn mutation, decorin staining was seen only in the cytoplasm of the keratocytes (Fig. 4). 

Figure 4

 

Murine corneas immunostained with decorin antibody (LF113). The specimens were stained in parallel to minimize intersample variation of intensity due to different staining. All exposure times were kept constant. A strong decorin signal is observed extracellularly in the stroma in corneas from wt and heterozygous animals, whereas in corneas from animals homozygous for the 952delT decorin mutation decorin staining is seen mainly in the cytoplasm of the keratocytes. Scale bars: 10 μm.

RNA Analysis of Mouse Corneal Tissue

Sequencing of cDNA from heterozygous 952delT Dcn mice revealed the presence of the mutation in the decorin mRNA, and that approximately equal amounts of wt and mutated mRNA were present in corneal tissue (data not shown). 

To explain the low levels of truncated decorin in the corneas from mice carrying the 952delT mutation, we looked for the possibility of mRNA degradation, alternative polyadenylation or association with inactive polysomes due to the inserted mutation. By DNA fragment analysis, increased degradation of mRNA was ruled out (data not shown). Northern blot analysis showed mRNA of equal length from wt, heterozygous, and homozygous KI animals, thus indicating that the same polyadenylation site was used (data not shown). By gradient centrifugation, we compared the amount of decorin mRNA associated with active polysomes and inactive mRNP complexes. Both mRNA of wt and mutated decorin associated to a similar degree with active polysomes, indicating that the translation of both forms of decorin mRNA is similar (data not shown). 

Decorin Expression in Lysates of Primary Keratocyte Cultures and Immunofluorescence of Intact Primary Keratocyte Cultures

As decorin is a secreted protein, we looked for the amount of normal and truncated decorin in the cell medium from primary corneal explants. Cells secrete more decorin when confluent. Cell medium was concentrated by acetone precipitation before deglycosylation. Human corneal cells secreted wt and truncated decorin, whereas murine cells only secreted wt decorin. As shown in Figure 5E, no truncated decorin was detected in the cell medium from mice homozygous for the 952delT Dcn mutation. All the truncated decorin appeared to be retained intracellularly. We also incubated the cells in presence of proteasome inhibitor MG132 and lysosome inhibitor leupeptin, without seeing an increase in the extracellular amount of truncated decorin (data not shown). To test if the lack of secretion is due to the murine cell system, we transfected murine NIH3T3 fibroblasts with vectors containing the human decorin gene with and without the 967delT DCN mutation. This experiment showed that wt and truncated human decorin were secreted by the murine cells, indicating that the lack of secretion is not a species-specific feature (data not shown). Figure 6 shows that murine wt decorin localizes to the Golgi complex, whereas truncated murine decorin is associated with the ER. In contrast, decorin staining of human keratocytes from normal and CSCD cells shows the presence of decorin in ER and Golgi compartments. In the CSCD samples, extracellular deposits that were strongly reactive with decorin antibody were detected indicating that truncated decorin participates in the formation of extracellular aggregates of decorin. Western blot analysis showed that truncated murine decorin was N-glycosylated (data not shown). 

Figure 5

 

Decorin secretion from human and murine keratocyte cultures. Keratocyte cultures were established from corneal explants from humans without (A) and with (B) CSCD, and from wt mice (C), and mice heterozygous (D) and homozygous (E) for the 952delT decorin mutation. Keratocytes were grown in DMEM F12 culture medium until confluence. At indicated days after confluence, cell cultures were incubated with serum-free medium for 24 hours, and thereafter cell medium and cells were harvested and deglycosylated. The figure shows that whereas the human truncated decorin protein is secreted to the cell medium, the mouse truncated protein is retained intracellularly. <A background band that should not be mistaken for truncated decorin. The standard is MagicMark XP Western Protein Standard. Loading reference for cell medium is a Coomassie stained gel showing bands at approximately 90 kDa. For loading reference of cell pellets, the membranes were stripped and reprobed with antibody against GAPDH (37 kDa).

Figure 6

 

(A, B) Immunostaining of keratocyte cultures. Keratocyte cultures from wt mice and mice homozygous for the 952delT decorin mutation (A) or from normal human corneas or corneas with CSCD (B) were grown to five days after reaching confluency. The cells were fixed in paraformaldehyde and permeabilized by 0.05% Triton. Double-staining in (A) was done with antibodies against decorin (LF113) and marker proteins of cis-Golgi (GM130) or ER (anti-Ribophorin 1). The wt decorin localizes to the Golgi complex assumingly on its way to secretion, whereas the truncated decorin seems retained in the ER. (B) Double-staining was done against decorin (hDCN antibody) and cis-Golgi (GM130) or the ER (SPC-25). In contrast with the mouse cells, in human keratocytes decorin is present in the ER and Golgi complex. In addition, large decorin deposits are visible extracellularly, most likely representing decorin aggregates. Scale bars: 5 μm.

Discussion

In CSCD, the cornea becomes opaque due to extracellular deposits of decorin. In the present work we have developed a mouse by targeted mutation with a 952delT Dcn mutation that corresponds with the human 967delT DCN mutation that is found in CSCD. Surprisingly, heterozygous and homozygous mice had clear corneas and very little truncated decorin was detected in corneal tissue. We were unable to find evidence for a major increase in degradation of either mRNA or protein to explain the low levels of truncated decorin. Also, mutated decorin mRNA seemed to be appropriately associated with ribosomes. However, there was a distinct difference in that export of truncated mouse decorin protein was impaired, and the protein was detected only in the ER and not in the Golgi apparatus. In contrast, truncated decorin in human CSCD corneas was present in nearly equivalent amounts as wt decorin, and was exported extracellularly in corneal cell cultures. Immunostaining of the corneal cell cultures even demonstrated extracellular aggregates of decorin. The band corresponding to truncated decorin is weaker in corneal lysate (Fig. 3) than in lysates from cells in culture (Fig. 5). The volume of keratocytes constitute 10% to 17% of the cornea,¹⁴ and with truncated decorin protein present only intracellularly, this can explain the faint bands seen on Western blots of corneal lysate. 

Chen et al.⁹ have previously developed a 952delT transgenic mouse where expression of the mutated gene was controlled by a CAG-promoter. In these mice, truncated decorin also was localized intracellularly in keratocytes. By transfection of HEK293 cells, impaired export of the truncated decorin was observed, and decorin was retained in the ER causing ER stress. Chen et al.¹⁰ suggested that ER stress is a pathophysiological mechanism involved in the development of CSCD in humans. However, we were unable to detect any upregulation of proteins associated with ER stress in cell extracts from mouse and human primary keratocyte cultures (Supplementary Fig. S2). 

We find that the human condition is not replicated in 952delT mice, since the truncated form of murine decorin is retained in the ER, and the amorphous stromal deposits formed by aggregated decorin⁷ are not generated. However, the truncated form of human decorin is exported from transfected mouse cells, and a possible way to create a mouse model of CSCD could be to introduce the human decorin gene with the 967delT mutation into the mouse genome. 

The impaired export of truncated decorin appears to be caused by retention of the protein in the ER. Exit from the ER secretory pathway requires that proteins pass several quality control checkpoints.^15,16 Proteins can be retained in the ER and shuttled into the proteasomal degradation pathway if they are not associated appropriately with and folded by chaperones (heat shock proteins and calnexin/calreticulin), ER retention signals like arginine framed tripeptides are exposed due to misfolding of the protein, or exit signals are lost. Reuptake into the ER may occur by proteins having the C-terminal H/KDEL sequence. Such proteins may bind to KDEL receptors and thereby be retrieved from the ER-Golgi intermediate compartment and cis-Golgi.¹⁷ An interesting difference between truncated mouse and human decorin are the C-terminal amino acids that are introduced as a consequence of the frame shift mutation, PTRL in mouse, and LIRV in human decorin. TRL is a conserved C-terminal element in several mouse and human proteins,¹⁸ that may bind to PDZ (PSD-95, discs large, zonula occludens-1) domain–containing proteins.^19–22 In these cases, the function of the PDZ domain–containing protein is to mediate transport to and positioning at the plasma membrane. The cystic fibrosis transmembrane conductor regulator (CFTR) has been extensively studied. The PDZ domain–containing proteins also may be involved in ER retention. Coexpression of the ion channel protein Kv1.4 with SAP97 leads to ER retention of Kv1.4, while Kv1.4 is transported through the Golgi apparatus when expressed together with PSD-95.²³ Thus, the TRL element of truncated mouse decorin should be examined further to determine if it has a role in ER retention. 

Acknowledgments

Unni Larsen and Jorunn Skeie Bringslid provided technical support, and Bjørn Christensen assisted with dissecting corneas. The authors thank Larry W. Fisher for providing the LF-113 antibody, and Stein Ove Døskeland for fruitful discussions. Transmission electron microscopy and scanning confocal microscopy was performed at the Molecular Imaging Center, and mice were kept at the Vivarium, University of Bergen, Bergen, Norway. 

Supported by Grants 911296 (AEC) and 911746 (ER) from Western Norway Regional Health Authority, grants from the Meltzer Research Fund, Kirsten Orning Olsen and Inger Holms memorial funds, and Odd Fellow Research Fund. 

Disclosure: A.E.C. Mellgren, None; O. Bruland, None; A. Vedeler, None; J. Saraste, None; J. Schönheit, None; C. Bredrup, None; P.M. Knappskog, None; E. Rødahl, None