Fibulin 2, a Tyrosine O-Sulfated Protein, Is Up-regulated Following Retinal Detachment*

Background: Fibulin 2 is an ECM protein of basement membranes and elastic tissues. Results: Fibulin 2 is up-regulated during retinal detachment. Conclusion: Up-regulation of fibulin 2 following detachment points to its role in the adhesion of the RPE to the Bruch membrane and prevention of RPE migration. Significance: This is the first study addressing the role of fibulin 2 in retinal detachment. Retinal detachment is the physical separation of the retina from the retinal pigment epithelium. It occurs during aging, trauma, or during a variety of retinal disorders such as age-related macular degeneration, diabetic retinopathy, retinopathy of prematurity, or as a complication following cataract surgery. This report investigates the role of fibulin 2, an extracellular component, in retinal detachment. A major mechanism for detachment resolution is enhancement of cellular adhesion between the retina and the retinal pigment epithelium and prevention of its cellular migration. This report shows that fibulin 2 is mainly present in the retinal pigment epithelium, Bruch membrane, choriocapillary, and to a lesser degree in the retina. In vitro studies revealed the presence of two isoforms for fibulin 2. The small isoform is located inside the cell, and the large isoform is present inside and outside the cells. Furthermore, fibulin 2 is post-translationally modified by tyrosine sulfation, and the sulfated isoform is present outside the cell, whereas the unsulfated pool is internally located. Interestingly, sulfated fibulin 2 significantly reduced the rate of cellular growth and migration. Finally, levels of fibulin 2 dramatically increased in the retinal pigment epithelium following retinal detachment, suggesting a direct role for fibulin 2 in the re-attachment of the retina to the retinal pigment epithelium. Understanding the role of fibulin 2 in enhancing retinal attachment is likely to help improve the current therapies or allow the development of new strategies for the treatment of this sight-threatening condition.

Fibulins are members of a seven-member family of extracellular matrix (ECM) 3 proteins that are important constituents of elastic fibers and basement membranes (1)(2)(3)(4)(5). They are divided into two classes based on their size (6). Both classes are characterized by calcium-binding EGF repeats and a C-terminal fibulin-specific module. Class I includes fibulins 1, 2, and 6, which are the larger members due to the presence of an additional N-terminal domain (absent in class II fibulins). Class II members include the shorter members fibulins 3, 4, 5, and 7 and are characterized by fewer calcium-binding EGF repeats (6).
Fibulins are known to play an essential role in cell adhesion and migration through their association with several ECM proteins and cell surface receptors such as integrins and syndecans (6,7). Mutations in several fibulins result in human diseases (8). Mutations in fibulin 4 and fibulin 5 cause a connective tissue disorder called cutis laxa, which is characterized by cutaneous abnormalities such as loose skin (9 -11). Fibulin 1 disruptions cause synpolydactyly in humans, which is characterized by malformation of distal limbs due to bone fusion (12). Fibulin 2 is down-regulated in breast cancer, colorectal cancer, lung cancer, Kaposi sarcoma, esophageal squamous cell carcinoma, nasopharyngeal carcinoma, and childhood acute lymphoblastic leukemia (13)(14)(15)(16)(17). The down-regulation of fibulin 2 in these cancers is either due to methylation of its promoter, gene deletion, or simply down-regulation of the levels of protein (13,14). It was further shown in nasopharyngeal carcinoma that re-introduction of fibulin 2 expression in these tumors suppressed tumor growth in nude mice by inhibiting cell proliferation, migration, and invasion of these tumor cells (17).
Fibulin mutations have also been associated with visual defects. Mutations in fibulin 3 have been shown to cause malattia leventinese (ML), a type of inherited macular degeneration that manifests in early childhood with drusen formation in honeycomb form that makes it distinct from age-related macular degeneration (AMD) (18). Mutations in fibulins 5 and 6 were found to associate with AMD, which is characterized by drusen deposits beneath the retinal pigment epithelium (RPE) and central vision loss late in life (19,20). Determination of whether mutations in fibulin 2 were associated with inherited retinal diseases led investigators to study fibulin 2 mutations in 429 controls and 402 AMD patients. Although numerous mutations were identified in the fibulin 2 gene, statistical analysis showed that these mutations were not significantly associated with AMD (19).
Fibulin 2 protein contains four major domains as follows: the N-terminal domain; three anaphylatoxin modules; 11 calciumbinding EGF repeats; and a C-terminal fibulin-type module. Its transcript has been previously identified in mouse ocular tissues (21,22). However, little is known about the function of this protein in the eye.
In this report, we show the presence of fibulin 2 in the RPE, Bruch membrane, and choriocapillaris of ocular tissue. We provide in vitro and in vivo evidence showing that fibulin 2 is posttranslationally modified by sulfation at tyrosines 192, 196, and 198, and in vitro elimination of these sulfated tyrosines resulted in increased cellular proliferation and migration but did not influence its secretion. Most importantly, we show that fibulin 2 is up-regulated following experimental retinal detachment and adhered to and inhibited the migration of the retinal pigment epithelial cell line ARPE19 in in vitro adhesion and migration assays. Therefore, we conclude that the up-regulation of fibulin 2 in vivo during retinal detachment suggests a role for it in the tight association between the retina and the RPE that involves a combination of its adhesive and anti-migratory characteristics, thereby allowing reattachment to the retina.

EXPERIMENTAL PROCEDURES
Recombinant Clone and Antibodies-A recombinant mouse fibulin Fbln2 clone was purchased from Genecopea. This clone was a full-length clone with a C-terminal Myc tag. The anti-Fbln2 antibody was either obtained from a commercial source (catalogue no. GTX105108, 1:1000 dilution, GeneTex) or was a kind gift from Dr. Mon-Li Chu (Thomas Jefferson University, Philadelphia, PA, dilution 1:2000) (23). The anti-Myc antibody was from Cell Signaling (catalogue no. 2276S, dilution 1:1000); the anti-fibronectin antibody was purchased from Santa Cruz Biotechnology (catalogue no. 9068, 1:200 dilution); and the anti-actin-HRP antibody was from Sigma (catalogue no. A3854, 1:25,000 dilution). The anti-sulfotyrosine antibody (PSG2, dilution 1:5000) was described previously (24) and has previously been used to enrich for tyrosine-sulfated proteins in epididymal homogenates of mice and was used as recommended (25).
Cell Lines, Transfection, and Establishment of Permanent Transfectants-The cell lines used were as follows: mouse photoreceptor cell line 661W (26); human RPE cell line ARPE19 (27); and human embryonic kidney epithelial cell lines HEK293 and HEK293T (28). HEK293T cells were transiently transfected using calcium phosphate transfection methods (29,30). Permanent transfectants were generated by transfection into HEK293 cells and selection with 1 mg/ml geneticin (Invitrogen).
Human Donor Eyes-Human donor eyes from a normal 72-year-old Caucasian male were obtained from Lions Eye Institute (Tampa, FL) and were dissected to obtain the retina, RPE, sclera containing choriocapillaries (sclera/CC), and optic nerve tissues. Lysates were prepared from these tissues as described previously (31).
Mouse Eyes-Mouse eyes were dissected at postnatal day 25 into retina, RPE, and choroid and sclera (PECS) fractions, and lysates were prepared from these tissues as described previously (31).
Immunoblotting and Immunoprecipitation-Protein extracts were prepared from mouse and human ocular tissues, 661W cells, ARPE19 cells, and from either transiently transfected HEK293T or permanently transfected 293 cells. Protein was estimated, fractionated, and transferred to membranes and immunoblotted as described previously (31).
For the matrix cytoplasmic lysate (MCL) fractions, cells were scraped from the plates, and lysates, which contained both the matrix and cytoplasmic fractions according to previously published protocols (32), were prepared.
For the trypsin-treated MCL fractions, media were removed; cells were washed with phosphate-buffered saline (PBS) and trypsinized, following which the cells were scraped from the plates, and lysates were prepared as described above. For immunoprecipitation, 500 g of protein extracts were incubated with the desired antibody for 12 h, precipitated by centrifugation, eluted in 1ϫ Laemmli buffer (33), and fractionated by SDS-PAGE as described previously (31).
Fractionation of 661W cells were done according to previously published protocols (32). Briefly, 661W cells were grown to confluence in regular media and then switched to serum-free media. Media were harvested, and cells were successively treated with detergent and then 6 M urea. Lysates in the detergent fraction were enriched for nuclear and cytoplasmic proteins, and lysates in the urea fraction and media were enriched for ECM proteins.
Site-directed Mutagenesis-Site-directed mutagenesis was performed using QuikChange II XL kit (Agilent Technologies) to Myc-tagged Fbln2 clone (WT) to create single (Y192F, Y196F, and Y198F) and triple mutants (TM) as per the manufacturer's recommendations.
Metabolic Labeling, Barium Hydroxide Hydrolysis, and Thin Layer Electrophoresis (TLE) Analysis-HEK293T cells were transiently transfected with WT, single, and TM Fbln2 clones. Eight hours after transfection, labeling media containing sulfate-free Joklik-modified Eagle's media (Sigma) containing 2% dialyzed fetal bovine serum and 0.15 mCi/ml of Na 2 SO 4 (PerkinElmer Life Sciences) were added. Forty eight hours later, media were collected. Fbln2 was then immunoprecipitated from the media and fractionated by SDS-PAGE. Following transfer onto a PVDF membrane and autoradiography, the radioactive bands corresponding to Fbln2 were cut out and subjected to barium hydroxide hydrolysis and TLE analysis according to previously published methods (25,37).
Retinal Detachment in Mice-Experimental retinal detachment was induced in P60 C57BL/6 mice via subretinal injections of 1 l of saline as described previously by Nour et al. (38).
Experimental design included subretinal injection of five animals, whereas two animals served as uninjected age-matched controls. Sham injections involved inserting a needle through the sclera but not penetrating the subretinal space. The experiment was done three independent times. Protein expression on immunoblots was quantitated using the Kodak molecular imaging software version 4 (Kodak Molecular Imaging Systems). Measurements included those of background intensity, gaussian fitting the identified bands, and measurement of net intensity of each band. Finally, the net intensity values obtained for Fbln2 protein on immunoblots were divided by the values for actin to get the "levels of Fbln2." Deglycosylation of Myc-tagged Fbln2-HEK293T cells were transiently transfected with WT and TM Fbln2 clones. Eight hours after transfection, labeling media containing sulfate-free Joklik-modified Eagle's media (Sigma) containing 2% dialyzed fetal bovine serum and 0.15 mCi/ml Na 2 SO 4 (PerkinElmer Life Sciences) were added. Forty eight hours later, media were collected, and WT and TM Fbln2 clones were immunoprecipitated from the media and subjected to deglycosylation overnight at 37°C, using the protein deglycosylation mix (New England Biolabs). The samples were then electrophoresed, followed by autoradiography and immunoblotting with anti-Myc antibody. Proteins were quantitated using Kodak Molecular Imaging software version 4, and net intensity was calculated for each sample. Bands corresponding to WT and TM Fbln2 were excised and counted for radioactivity. Finally, Fbln2 disintegration/min/Fbln2 pixels was calculated for each WT and TM Fbln2 clone.
Purification of Fbln2-Myc-tagged Fbln2 protein was affinity-purified from media of permanent transfectants using anti-Myc-agarose beads (Sigma) according to manufacturer's instructions. Media from empty vector (pcDNA3.1) permanent transfectants were passed through the affinity column and served as negative controls for the adhesion assays. The amount of purified Fbln2 eluted from the column was quantitated by SDS-PAGE alongside the known concentration of BSA standards and visualized using Coomassie Blue dye.
Adhesion Assay-For adhesion assays, 96-well plates were coated with 0.5 g/ml BSA, Fbnc, Myc-tagged Fbln2 protein, or equivalent volumes of pcDNA3.1 vector eluent from the affinity column and allowed to air dry. The plates were then blocked with blocking buffer (3% BSA in DMEM without FBS) for 1 h at 37°C. ARPE19 cells were dissociated and washed twice with DMEM without FBS and plated at 2 ϫ 10 4 cells/well and allowed to adhere for 3 h. Nonadherent cells were washed away with PBS, and attached cells were fixed in 10% paraformaldehyde for 10 min and stained with crystal violet dye. The dye was then solubilized in 33% acetic acid, and absorbance was read at 550 nm. Absorbance values from the pcDNA3.1 vector eluent were then divided from the absorbance values of each of the purified proteins, and this value (normalized to vector) was plotted for each of the purified proteins.
Migration Assay-Migration assays were performed in 24-well CytoSelect TM cell migration assay kit containing 8-m pore size polycarbonate membrane inserts (Cell Biolabs Inc.) according to product instructions. In migration assays comparing permanent transfectants of pcDNA 3.1 (vector), wild-type Fbln2 (WT), or the triple Tyr to Phe mutant (TM), two independent permanent transfectants of each cell type were used for the assay. The cells were pretreated with 2 g/ml mitomycin C for 45 min, and 2 ϫ 10 5 cells were plated in the upper chamber in DMEM without FBS and the lower chamber containing DMEM plus 10% FBS. Cells were incubated for 24 h at 37°C, following which the cells that had migrated were fixed and stained with crystal violet dye, and absorbance was determined at 550 nm. The same number of cells used in the migration assay for each of the permanent transfectants was also plated and grown for 24 h in DMEM and immunoblotted with anti-actin and anti-Myc antibody.
For migration assays of ARPE19 cells, the cells were pretreated with 2 g/ml mitomycin C for 45 min, and 1 ϫ 10 5 cells were plated in the upper chamber in DMEM F-12 without FBS media and conditioned media from permanent transfectants pcDNA 3.1 (vector), or wild-type (WT) Fbln2 containing 10% FBS was added to the lower chamber. In the immunodepletion experiments, media from wild-type Fbln2 were subjected to immunoprecipitation with anti-Myc or IgG antibodies to remove Myc-tagged Fbln2 from the media. These media were then supplemented with 10% FBS and added to the lower chamber. The rest of the experiment was as described above.
Statistical Analysis-Statistical analysis was done using GraphPad Prism software using one-way or two-way analysis of variance (ANOVA) or Student's t test.

Fibulin 2 Is Present in the Retina, RPE Bruch Membrane, and
Choroid-IHC analysis on sections of P30 pigmented mouse eyes showed that Fbln2 is observed in the CC, BrM, and sclera (green label, Fig. 1, A, C, and D). Labeling, although weaker, was observed in the retina, mainly around blood vessels, as observed by the Fbln2 and endothelial marker CD31 co-staining (arrows, Fig. 1A). Equivalent amount of rabbit IgG as control showed no labeling in the eye (Fig. 1B). Higher ϫ100 magnification images showed Fbln2 labeling in BrM (arrowheads, Fig. 1, C and D) and choriocapillaries ( Fig. 1, C and D). Also, a lumen in the choriocapillary region is visible by the staining of the endothelial marker anti-CD31 label (Lu, Fig. 1, C and D, red). Because the presence of pigment masks the expression of Fbln2 in the RPE of C57BL/6 eyes, immunostaining for Fbln2 was done on P30 BALB/c albino mouse eyes. Minor Fbln2 labeling is observed in the RPE (asterisk, Fig. 1E).
Further confirmation of the IHC findings came from immunoblot analysis of two independent lysates of mouse retinas and PECS. Fbln2 was mainly observed in the PECS lysates (lanes 1 and 2, Fig. 1F), whereas retinal lysates showed reduced amounts of Fbln2 (lanes 3 and 4, Fig. 1F). Interestingly, we also observed Fbln2 in conditioned media from 661W cells, a cone-specific cell line (lane 5, Fig. 1F) (26).
Because it is difficult to fractionate mouse PECS into its individual components, lysates were prepared from human donor eyes divided into optic nerve region, sclera/CC, RPE, and retina. Fig. 1G shows the presence of Fbln2 in the optic nerve region, sclera/CC, and RPE lysates (lanes 1-3), and immunoreactivity is only obvious in the retina upon prolonged exposure (data not FIGURE 1. Ocular localization of Fbln2. Immunohistochemical labeling was performed on frozen P30-pigmented C57BL/6 mouse eye sections using anti-Fbln2 antibody (green label), anti-CD31 (red label), and rabbit IgG (green label) as negative control and DAP1 (blue). Bright field images were taken at ϫ20 (A and B). Mouse sections were double-stained with Fbln2 (green) and CD31 (red) and imaged at ϫ20 magnification (A). Sections stained with rabbit IgG (green) were imaged at ϫ20 magnification (B). Higher magnification images (ϫ100) of sections double-stained with Fbln2 (green) and CD31 (red) superimposed on bright field images (C) or without the bright field are shown (D). Asterisks indicate RPE layer; arrows show the presence of two blood vessels in the retina; and arrowheads indicate Bruch membrane. All blood vessels, including choriocapillaries, are shown by the endothelial marker CD31 (red). A lumen of a vessel (lu) is shown at ϫ100 magnification (C and D). shown), in agreement with observed reduced levels in the mouse retina (Fig. 1F).
Although very little Fbln2 was observed in retinal lysates ( Fig.  1, F and G), high levels of Fbln2 were detected (lane 5, Fig. 1F) in conditioned media from 661W cells (26,39). Fbln2 produced by 661W cells exists both in the media and in the fraction that contains both matrix and cytoplasmic proteins, herein named MCL. When the conditioned media and MCL fractions were probed with anti-Fbln2 antibody, the MCL fraction was found to harbor two isoforms of Fbln2, a 195-kDa and a 160-kDa isoform (lane 3, arrowheads, Fig. 1H), although the 195-kDa isoform was mainly detected in the media (lanes 1 and 2, Fig.  1H). To differentiate between external or matrix fractions and the internal or cytoplasmic fractions, 661W cells were subjected to brief trypsin treatment to remove ECM or matrixattached Fbln2. As a control, another 661W sample was treated with saline. Lysates were then electrophoresed alongside conditioned media on SDS-PAGE. Interestingly, immunoblot analysis of these lysates showed that the 195-kDa Fbln2 (upper arrowhead) isoform was eliminated from the MCL fraction after trypsin treatment leaving the 160-kDa band (lower arrowhead) unaffected (compare lanes 3 and 4, Fig. 1H). However, media prior to trypsin treatment showed the presence of the 195-kDa Fbln2 band (lanes 1 and 2, Fig. 1H). Trypsin digestion of the 195-kDa isoform of Fbln2 in MCL proved that this isoform is located outside the cell, consistent with its ECM or matrix association and thereby its susceptibility to trypsin cleavage. The trypsin resistance of the shorter 160-kDa isoform suggests that this isoform is still present inside the cell and is not secreted into the ECM fraction of 661W cells.
To further separate the MCL fraction into cytoplasmic and ECM fractions and to verify the extracellular status of the 195-kDa isoform of Fbln2, 661W cells were fractionated according to a previously established method (32) wherein the 661W cells were grown to confluence in regular media and then switched to serum-free media. After removing the media, cells were treated successively with a detergent solution, followed by urea solution. It has been previously established that the detergent fraction is enriched in cytoplasmic and nuclear proteins, and the urea and media fractions are enriched in ECM proteins (32).
The lysates and media were then electrophoresed alongside conditioned media on SDS-PAGE. Immunoblot analysis with an ECM-specific marker, Fbnc (32), shows that the 195-kDa Fbln2 isoform is enriched in the media (only 2% of the total media was run), and the ECM-urea fraction (arrowheads, lanes  1 and 3, Fig. 1I) is similar to Fbnc. The 160-kDa Fbln2 isoform is enriched in the cytoplasmic and nuclear fractions, similar to the nuclear matrix protein p84 (lane 2, Fig. 1I).
Fibulin 2 Is a Tyrosine-sulfated Protein-In the course of isolating retinal tyrosine-sulfated proteins by a PSG2 affinity chromatography column (24) followed by tandem mass spectrometric analysis, Fbln2 was identified (data not shown). To verify that Fbln2 is indeed tyrosine-sulfated, it was immunoprecipitated from postnatal day 1 mouse eyecup lysates. Fig. 2A shows the 195-kDa isoform of Fbln2 was specifically pulled down by the anti-Fbln2 antibody but not by mouse IgG (compare lanes 2 and 3, Fig. 2A, IB:Fbln2). This 195-kDa isoform was also identified by PSG2 to be tyrosine-sulfated (lane 2, Fig. 2A, IB:PSG2).
Because Fbln2 was observed in the media and MCL fractions of 661W cells (Fig. 1H), conditioned media and MCL were subjected to immunoprecipitation with equivalent amounts of anti-Fbln2 and IgG antibodies. Although only the 195-kDa isoform was pulled down from the conditioned media with anti-Fbln2 antibody and not IgG (arrow, compare lanes 4 and 5,  To confirm that the Myc-tagged clone was in fact Fbln2, it was immunoprecipitated from the conditioned media of two independent transfectants with either anti-Fbln2 antibody or equivalent amounts of rabbit IgG. Immunoblotting the pulled down product with anti-Myc antibody showed the presence of the Myc tag, hence confirming that the identified 195-kDa band is indeed Myc-tagged Fbln2 (arrowhead, lanes 4 and 5, Fig. 3B, IB:Myc). This protein band was not observed in the IgG immunoprecipitants (lane 3, Fig. 3B, IB:Myc). Furthermore, probing with PSG2 confirmed that the 195-kDa Fbln band in the conditioned media is tyrosine-sulfated in vitro (arrowheads, lanes 4 and 5, Fig. 3B, IB:PSG2). No fibulin-2-related PSG2 immunoreactivity was observed in IgG controls (lane 3, Fig. 3B, IB:PSG2). Again, the same nonspecific band is detected in the immunoprecipitants from the media of Fbln2-transfected HEK293T cells (asterisk, lanes 3-5, Fig. 3B, IB:PSG2).
Although PSG2 recognizes sulfotyrosine residues in many sequence contexts (24,25), the presence of tyrosine sulfate is commonly confirmed after metabolically labeling with radioactive sulfate followed by sulfamino acid analysis. Therefore, Fbln2-transfected HEK293T cells were allowed to grow in the presence of Na 2 35 SO 4 , and conditioned media were subjected to immunoprecipitation with anti-Myc and then fractionated by SDS-PAGE and autoradiographed. Autoradiographic analysis showed that the immunoprecipitated Fbln2 protein incorporated the radioactive sulfate (lane 4, Fig. 3C, AR), although no label was immunoprecipitated from the vector media (lane 3, Fig. 3C, AR). After autoradiographic analysis, the blot was probed with anti-Myc antibody, which confirmed that the radioactive band is the 195-kDa isoform of Fbln (lane 4, Fig. 3C, IB:Myc). The radiolabeled Fbln2 band and an equivalent area of the membrane from the vector immunoprecipitated lane (lane 3, Fig. 3C) were cut out and subjected to barium hydroxide hydrolysis followed by TLE analysis using sulfated amino acids as standards (Fig. 3D). The co-localization of the radioactive tyrosine sulfate from Fbln2 hydrolysates (lane 2, Fig. 3D, AR) with the nonradioactive tyrosine sulfate standard detected by ninhydrin staining of the TLE plate (lane 2, Fig. 3D, NHD) confirmed the presence of tyrosine sulfate in Fbln2. There was no tyrosine sulfate signal detected in the vector hydrolysates (lane 1, Fig. 3D, AR). Fibulin 2 Is Sulfated at Tyrosines Residues 192,196, and 198-As it has been confirmed that Fbln2 is sulfated, it became necessary to identify the target tyrosine(s). The sulfotyrosine predicting program "Sulfinator" (40) identified residues 192, 196, and 198 as potential sites of tyrosine sulfation on Fbln2. Therefore, site-directed mutagenesis was performed on a Myc-tagged Fbln2 clone (WT) to generate four independent mutants. Single mutants, Y192F, Y196F, or Y198F had the individual tyrosine residues at 192, 196, and 198 mutated to phenylalanines (Fig.  4A), respectively. A TM was also generated that had all three potential tyrosine residues (192, 196, and 198) mutated to phenylalanines (Fig. 4A). All clones as well as pcDNA 3.1 only (vector) were independently transfected into HEK293T cells, and  Fbln2 (lane 4), single mutants Y192F, Y196F, and Y198F (lanes 5-7, respectively) and the triple mutant TM (lane 8) and were grown in the presence of radioactive sulfate. Conditioned media were used in an immunoprecipitation experiment with anti-Myc antibody followed by SDS-PAGE and immunoblotting. The blots were exposed to x-ray film overnight to produce an autoradiogram (AR), followed by immunoblotting with anti-Myc antibody. D, radiolabeled WT and TM Fbln2 were immunoprecipitated from conditioned media with anti-Myc antibody and subjected to deglycosylation (lanes 2 and 4) or left untreated (lanes 1 and 3). The samples were SDS-PAGE fractionated and immunoblotted with anti-Myc antibody. The blots were autoradiograhed and then probed with anti-Myc antibody. The radiolabeled bands were then excised from the blots and counted using liquid scintillation counter. E, radiolabeled WT, single mutant clones Y192F, Y196F, and Y198F, and the triple tyrosine sulfate mutant TM proteins were immunoprecipitated from conditioned media, electrophoresed, transferred to PVDF membranes, excised, subjected to alkaline hydrolysis, and analyzed by TLE. Plates were sprayed with ninhydrin (NHD), followed by exposure to an x-ray film to produce the autoradiogram. MAY 9, 2014 • VOLUME 289 • NUMBER 19 JOURNAL OF BIOLOGICAL CHEMISTRY 13425 products were immunoprecipitated from conditioned media with an anti-Myc antibody. Immunoprecipitants were electrophoresed and immunoblotted with anti-Myc and PSG2 antibodies. As shown in Fig. 4B (lanes 3-7), equivalent levels of all the different recombinant proteins (mutants and wild-type) were pulled down, whereas no Myc reactivity was observed in immunoprecipitants from vector transfectants (lane 2, Fig. 4B, IB:Myc). Immunoblotting with PSG2 demonstrated that the WT clone had the most prominent tyrosine sulfate signal (arrowhead, lane 3, Fig. 4B, IB:PSG2). Each of the single mutants (Y192F, Y196F, or Y198F) showed a reduced tyrosine sulfate signal (arrowhead, lanes 4 -6, Fig. 4B, IB:PSG2), although no signal was detected for the TM. A nonspecific protein band was detected in all lanes of the immunoblot (asterisk, lanes 2-7, Fig. 4B, IB:PSG2).

Fibulin 2 and Retinal Detachment
To confirm that the lack of tyrosine sulfate recognition by PSG2 of the TM clone was due to the fact that only three tyrosine sulfate residues (192, 196, and 198) existed in Fbln2, all independent mutant clones along with WT and vector were transiently transfected into HEK293T, and cells were grown in the presence of Na 2 35 SO 4 . Media from these clones were immunoprecipitated with an anti-Myc antibody. Autoradiographic analysis showed that equivalent radioactivity levels were incorporated by WT and single mutants, although slightly less radioactivity was incorporated by the TM clones (arrowhead, lanes 4 -8, Fig. 4C, AR), and no radioactivity was incorporated into vector-transfected samples (lane 3, Fig. 4C, AR). To further confirm that the radioactive immunoprecipitants are Fbln2, blots were probed with anti-Myc antibody, which showed that comparable levels of protein were immunoprecipitated from the media (lanes 4 -8, Fig. 4C, IB:Myc). Our data suggest that comparable levels of 35 S incorporated in the WT and single mutant clones are due to the fact that Fbln2 is a glycoprotein, and carbohydrates are also sulfated (41). In addition, Fbln2 has been shown to contain four N-linked glycosylation residues at the N-terminal end of the protein (23). To prove that carbohydrates are sulfated in WT and TM Fbln2 clones, we again radiolabeled WT and TM Fbln2. One set of WT and TM was subjected to deglycosylation, and another set was left untreated. An ϳ15-kDa drop in size after deglycosylation of WT and TM Fbln2 proved the presence of carbohydrate residues in the protein (lanes 2 and 4, Fig. 4D). The radiolabeled bands were cut out and counted by liquid scintillation. Counting showed that deglycosylation of WT and TM Fbln2 decreased the incorporation of the 35 S label (3.1 ϫ 10 Ϫ5 dpm/Fbln2 pixel for nontreated WT Fbln2 versus 1.55 ϫ 10 Ϫ5 dpm/Fbln2 pixel for deglycosylated WT Fbln2 and 1.04 ϫ 10 Ϫ5 dpm/Fbln2 pixel for nontreated TM Fbln2 versus 0.5 ϫ 10 Ϫ5 dpm/Fbln2 pixel for deglycosylated TM Fbln2).
In addition, to distinguish between carbohydrate sulfation versus tyrosine sulfation, the radiolabeled Fbln2 bands of WT, single mutants, and TM Fbln2 were subjected to barium hydroxide hydrolysis (Fig. 4E). Barium hydroxide causes the hydrolysis of peptide bonds and the alkali-labile sulfated carbohydrate residues in glycoproteins to be precipitated as barium sulfate. This leaves the tyrosine-sulfated residue intact, which is then identified by co-localization with a tyrosine sulfate standard after neutralization and thin layer electrophoresis analysis (42). As shown in Fig. 4E, radioactive spots were detected in the WT clone (lane 5, Fig. 4E, AR) and all of the single mutants (lanes 1-3, Fig. 4E, AR) but not in the TM or vector clone ( lanes  4 and 6, Fig. 4E, AR). This conclusively demonstrates that the three tyrosines (192, 196, and 198) are the only sites of tyrosine sulfation in Fbln2.
Tyrosine Sulfation of Fibulin 2 Does Not Affect Its Secretion-Because Fbln2 is a secreted protein, and tyrosine sulfation is a post-translational modification found mainly in secreted proteins (43), the role of sulfation in the secretion of Fbln2 was investigated. For this purpose, two independent permanent transfectants of either WT or TM Fbln2 were used. Fbln2 protein levels in the media and lysates of tyrosine-sulfated WT (lanes 1 and 2, Fig. 5A) and nontyrosine-sulfated TM clones (lanes 3 and 4, Fig. 5A) were compared by immunoblotting with an anti-Myc antibody. Quantitative analysis of the immunoblots presented in Fig. 5A showed that both WT and TM secreted ϳ80% of total expressed Fbln2 in the media fraction, although ϳ20% of it was sequestered in the cytoplasmic lysate fraction (Fig. 5B). The similar distribution of tyrosine-sulfated WT or unsulfated Fbln2 TM clones suggests that sulfation plays no if any role in Fbln2's secretion.
Sulfation Is Involved in the Growth Inhibitory Role of Fibulin 2-It has been previously shown that Fbln2 slows growth rate and cell migration in nasopharyngeal carcinoma cell lines (17). To investigate the growth regulatory role of the tyrosine sulfation of Fbln2, growth curves were generated for permanent transfectants independently expressing empty vector, WT, or TM Fbln2s. Cell numbers were counted 48 and 96 h postplating of 5 ϫ 10 4 cells on day 0 (Fig. 6A). Comparing the growth curves of two independent clones of each vector shows that expression of WT and TM Fbln2 inhibited growth by 64 and 54%, respectively, after 48 h in comparison with vector transfectants (Fig. 6A). Comparing the curves of WT and TM Fbln2 at 48 h showed insignificant differences in growth between them. However, at 96 h post-plating, nontyrosine-sulfated TM Fbln2 inhibited growth by 74%, whereas WT Fbln2 inhibited growth by 84% (Fig. 6A) compared with the vector transfectants. There was also a significant difference in the growth of WT and TM Fbln2 after 96 h, with the WT Fbln2 inhibiting growth by 39% more than the TM Fbln2 (p Ͻ 0.01,   Fig. 6A). These data suggest that although tyrosine sulfation plays a role in the growth inhibitory characteristics of Fbln2, it is not the sole determinant. Because Fbln2 has previously been shown to have an antimigratory function in many cell types (16,17,44), the role of Fbln2 in the migratory capabilities of the permanent transfectants was investigated using the Transwell system (45) and two independent permanent transfectants expressing vector, WT, or TM Fbln2s. Cells were incubated with 2 g/ml mitomycin C for 45 min, a dose established to be sufficient to inhibit cell division (data not shown). Fig. 6B shows that WT inhibited migration by 54% compared with vector (p Ͻ 0.0001), and TM Fbln2 inhibited cell migration by 42% compared with vector (Fig. 6B, p Ͻ 0.0001). In addition, the WT inhibited cell migration by 31% compared with TM Fbln2 (p Ͻ 0.001, Fig. 6B). Because the TM clones expressed higher levels (ϳ2.36) of Myc-tagged Fbln2 compared with the WT clones (Fig. 6C), it is reasonable to suggest that the unsulfated TM Fbln2 is not as efficient as WT Fbln2 in inhibiting the migratory capabilities of the cell.
Fibulin 2 Is Up-regulated Following Experimental Retinal Detachment-It has been previously shown that Fbln2 expression can be induced in response to injury in skin, aorta, and liver (5,46,47). Furthermore, Farjo et al. (21) showed that Fbln2 transcripts increased in the retina following experimental retinal detachment (ERD). To determine whether Fbln2 protein is up-regulated as a result of injury from the needle penetrating the eye, sham injections were performed on the eyes of mice by inserting a needle through the sclera but not penetrating the subretinal space. Levels of Fbln2 were examined at 8, 24, 72, and 168 h after injury. Levels of Fbln2 in the PECS were not significantly changed under these conditions compared with un-injected controls (lanes 1-5, Fig. 7, A and B).
To study the Fbln2 levels in response to ERD and to also determine RPE's response to injury, experimental retinal detachments were induced in mice, and levels of Fbln2 were determined in the retina and RPE.
Although levels of Fbln2 were not significantly changed in the PECS at 8 and 24 h after ERD (lanes 1-7, Fig. 7, C, D, and E), at 72 h post-ERD Fbln2 was significantly up-regulated (compare lanes 1-5 with lanes 6 and 7, Fig. 7, C and F). The levels of Fbln2 continue to be significantly up-regulated at 168 h post-ERD compared with controls (compare lanes 1-5 with lanes 6  and 7, Fig. 7, C and G). Fbln2 expression was not changed in retinal extracts at all time points tested (data not shown).
Because Fbln2 is an ECM protein expressed in RPE and Bruch membrane (Fig. 1) and has been shown to be a proadherent protein (48), the ability of ARPE19 cell, a human RPE cell line (27), to adhere to purified Fbln2 was tested. For this purpose, WT Fbln2 was purified from conditioned media of permanent transfectants using anti-Myc-agarose beads. Conditioned media from vector permanent transfectants similarly isolated served as a negative control. ARPE19 adherence in presence of Fbnc was used as a positive control. Adherence assays showed that a significantly higher number of ARPE19 cells were able to adhere to immobilized WT Fbln2 (Fig. 8A, p Ͻ 0.0025) and Fbnc (Fig. 8A, p Ͻ 0.0001), compared with a BSA control. . Growth curves are plotted on a logarithmic scale. Results were obtained from three independent experiments. B, in vitro migration assays were used to compare two permanent transfectants of pcDNA3.1 vector, WT Fbln2, and TM Fbln2 clones after preincubating cells with 2 g/ml mitomycin C for 45 min. Twenty four hours after plating, the degree of cell migration was assessed as described under "Experimental Procedures." Statistical significance was calculated using oneway ANOVA analysis using Bonferroni post-tests. Results present the mean values (Ϯ S.E.). The experiment was performed three times. C, to compare the levels of recombinant Fbln2 and actin levels in the permanent transfectants, 50 g of protein were electrophoresed and immunoblotted (IB) using anti-Myc antibody and anti-actin antibody. The experiment was performed three times.
It was observed that following ERD in rabbits, RPE cells are exposed to numerous cytokines that may modulate their behavior and cause RPE to dissociate from the Bruch membrane and migrate (49). Furthermore, collected subretinal fluid from detached human retinas caused the proliferation and migration of RPE cells (50) in culture. Therefore, the effect of Fbln2 on the migratory capabilities of ARPE19 was tested. Absorbance values after 24 h showed that migration of ARPE19 cells was significantly inhibited by media from WT Fbln2 (ϳ33%) (Fig. 8B, p Ͻ 0.0058) compared with ARPE19 cells growing in vector media. To verify that the anti-migratory function was due to WT Fbln2, media from the lower chamber in each Transwell were analyzed by SDS-PAGE followed by immunoblotting with anti-Myc antibody. As shown in Fig. 8B, WT Fbln2 was present in the lower Transwell chamber. Immunoblotting for Fbnc was used to demonstrate that equivalent amounts of conditioned media were used (Fig. 8B, IB:Fbnc). Results presented in Fig. 8B show that the suppression of migration of ARPE19 is likely due to the presence of Fbln2. To support the previous conclusion, an equi-volume of conditioned media from WT Fbln2 permanent transfectants was subjected to immunoprecipitation with anti-Myc or mouse IgG antibodies. Because WT Fbln2 is Myctagged, it is immunodepleted by immunoprecipitation with anti-Myc antibody. Fig. 8C shows the presence of Fbln2 in the conditioned media after mouse IgG immunoprecipitation as a control (Fig. 8C, Fbln2ϩ) and 3-fold reduction after anti-Myc immunoprecipitation (Fbln2Ϫ). Migration of ARPE19 cells was analyzed in the presence and absence of Fbln2-immunodepleted media. Migration was inhibited by 26% (Fig. 8C, top, p Ͻ 0.028) in the presence of Fbln2. This result confirms that the presence of Fbln2 was responsible for inhibition of migration of ARPE19 cells.

DISCUSSION
In the course of identifying tyrosine-sulfated proteins in ocular tissues using a PSG2 affinity column (24) followed by tandem mass spectrometry, Fbln2 was detected. Because Fbln2 has not been studied in any detail in the eye, we focused our studies on determining its cellular localization, its potential function in the eye, and the role of tyrosine sulfation in its function. In the mouse eye, Fbln2 was detected mainly in the RPE, BrM, and choriocapillaries. In the photoreceptor cell line 661W, Fbln2 existed in two isoforms, a 195-kDa isoform and a 160-kDa isoform. The presence of the 195-and 160-kDa isoforms of Fbln2 has been reported previously. Only the 195-kDa isoform has been shown in breast cancer cell lines (16), Schlemm canal cells (51), mouse ear (52), and aorta (53), although both the 195-and 160-kDa isoforms have been reported in nasopharyngeal cell lines (17) and mouse fibroblasts (32).
Sequence analysis of Fbln2 revealed four potential N-linked glycosylation sites (23), and so the 195-kDa isoform is likely a FIGURE 8. A, adhesion assays were performed using the ARPE19 cells grown on either purified WT, Fbnc, BSA or eluents obtained from passing media from pcDNA 3.1 transfected cells on the affinity column. About 0.5 g/ml protein (in the case of pcDNA 3.1 vector, equi-volumes were used) was immobilized on 96-well plates. ARPE19 cells were dissociated and allowed to adhere in the presence of these proteins for 3 h, followed by washing away nonadherent cells, staining the remaining cells with crystal violet, and measuring absorbance at 550 nm. The absorbance of each of the purified protein samples was divided by the absorbance value for the pcDNA 3.1. This resulting "normalized to vector" was plotted using GraphPad Prism. Statistical significance was calculated using one-way ANOVA analysis using Bonferroni post-tests comparing Fbnc and WT Fbln2 to BSA and to each other. Results present the mean values (Ϯ S.E.). Each experiment was performed three times in triplicate. B, in vitro migration assays were performed comparing the relative migration ability of ARPE19 cells grown in the presence of conditioned media from permanent clones of pcDNA 3.1 vector or WT Fbln2 transfectants. ARPE19 cells were grown in the upper wells of the transwell plates, and conditioned media were placed in the lower chambers. Twenty four hours later, cells that had migrated were stained with crystal violet, and absorbance was measured at 550 nm. Graphs were plotted using GraphPad Prism. Statistical significance was calculated using one-way ANOVAs using Bonferroni post-tests. Results present the mean values (Ϯ S.E.). The experiment was done three independent times. C, in vitro migration assays were performed comparing the relative migration ability of ARPE19 cells grown in the presence of conditioned media from WT Fbln2 permanent transfectants after immunoprecipitation with IgG (Fbln2ϩ) or anti-Myc antibody (Fbln2Ϫ). ARPE19 cells were grown in the upper wells of the transwell plates, and conditioned media were placed in the lower chambers. Twenty four hours later, cells that had migrated were stained with crystal violet, and absorbance was measured at 550 nm. Graphs were plotted using GraphPad prism. Statistical significance was calculated using one-way ANOVA analysis using Bonferroni post-tests. Results present the mean values (Ϯ S.E.). glycosylated isoform of Fbln2, whereas the 160-kDa isoform represents the nonglycosylated immature isoform. In addition, it is the large 195-kDa isoform of Fbln2 that is present outside the cell as determined by trypsin accessibility of this isoform in 661W, whereas the 160-kDa isoform is protected from trypsin suggesting that it is present inside the cell. The 195-kDa isoform itself is present in two fractions as follows: a tyrosinesulfated form in the media and an unsulfated form in the MCL fraction. One possibility for the presence of the two forms could be due to the ability of the 661W cells to produce the two forms independently. Alternatively, the unsulfated form may result from specific sulfatase action post-secretion. The absence of any ECM-based tyrosine sulfatase activity favors the former possibility. Regardless, the presence of the two forms emphasizes the importance of Fbln2 and the role tyrosine sulfation can play in the function of a protein.
Although tyrosine sulfation has been shown to influence the ability of the cell to secrete some proteins (54,55), the lack of tyrosine sulfation of Fbln2 did not affect its secretion. Both the tyrosine-sulfated and unsulfated forms of Fbln2 inhibited growth compared with vector-transfected permanent transfectants. However, the tyrosine-sulfated form of Fbln2 was additionally able to inhibit growth compared with the unsulfated isoform by 39% at 96 h, suggesting a role of Fbln2's sulfation in regulating cellular growth. Such a role in regulating growth has been previously studied in the glycosaminoglycans in the ECM, where it has been shown that sulfation of glycosaminoglycans is important for the sequestration and regulation of growth factors and associated signaling (56 -58). Experimental desulfation resulted in decreased sequestration and enhanced growth factor signaling (59). It is conceivable that such effects could also be assigned to tyrosine sulfation of a protein.
Tyrosine-sulfated Fbln2, reduced migration of HEK293T cells by 54%, and the unsulfated counterpart was able to reduce the migration rate by 42% compared with the vector control. When comparing the migration differences between two independent WT Fbln2-expressing permanent clones that only differ in expression levels, we observed a 68% decrease of migration with a 2-fold increase in WT-fibulin 2 (data not shown). Comparing the tyrosine-sulfated Fbln2 to the non-tyrosinesulfated TM showed a 31% inhibition of cell migration by the WT Fbln2. Moreover, the higher levels of Fbln2 expressed by the non-tyrosine-sulfated TM clone compared with the tyrosine-sulfated WT clones indicates that removing tyrosine sulfation from Fbln2 removes anti-migratory effects of WT Fbln2. The ability of Fbln2 to reduce growth rate and migration was described previously in breast and nasopharyngeal cancer cells (16,17). This may be due to the fact that Fbln2 is an ECM protein that forms molecular bridges by binding other ECM proteins such as fibronectin, laminins, perlecans, versicans, aggrecan, and nidogens (60 -64). In doing so, it may sequester growth factors essential for growth and migration of the cell. Future studies will address whether we observe a difference in growth factor sequestration between wild-type and unsulfated Fbln2s.
One observation in our studies is that the outcome of eliminating tyrosine sulfation resulted in a 39% increase in growth rate and a 31% increase in migration rate compared with its tyrosine-sulfated counterpart. Other studies where tyrosine sulfation was eliminated as in the protein P-selectin glycoprotein ligand-1 showed that elimination of tyrosine sulfation reduces but does not eliminate binding and rolling on endothelial cells (65,66). Similarly, studies on Factor VIII, where a naturally occurring mutation changes the tyrosine sulfate residue to phenylalanine, moderately reduces binding to the von Willebrand factor and causes mild hemophilia (67). Studies on the chemokine receptor CCR3 that contains two tyrosine sulfate residues have shown that elimination of tyrosine sulfation from both residues or one residue does not eliminate binding to its ligands eotaxin-1, -2, or -3, but it only modulates the selectivity of the receptor to the three chemokines (68). Therefore, our studies and other studies suggest that tyrosine sulfation is required for optimum function, and lack of it does not altogether eliminate the function.
Retinal detachment is the physical separation of the RPE from the outer segments of photoreceptors. It is a common occurrence as a consequence of old age, trauma, and cataract surgery and as a complication of diseases such as proliferative diabetic retinopathy, proliferative vitreoretinopathy, and uveitis (69). Large population-based studies put the annual incidence of retinal detachment around 1 in 10,000 people and a lifetime risk of 3% at the age of 85 years (70,71). Retinal detachment can lead to the exposure of the quiescent RPE cells (72) to multiple cytokines, which may modulate their behavior and cause them to dissociate from the Bruch membrane (73), undergo major morphological changes on the apical surface (74), or change their morphology from flat cuboid to columnar in appearance and enlarge in size (75). RPE proliferation and migration after retinal detachment has been observed in many species (49,50,76,77). An extreme condition that results due to retinal detachment occurs during proliferative vitreoretinopathy, where RPE cells trans-differentiate into myofibroblasts and proliferate into the vitreous humor to form sub-retinal membranes (78,79). It is feasible that one of the reasons most RPE cells adhere back to the photoreceptor layer and do not migrate following detachment is due to up-regulation of anti-proliferative and anti-migratory proteins such as Fbln2. This is supported by our findings that Fbln2 is up-regulated in the PECS fractions at 72 and 168 h post-ERD. Studies on Efemp1 (Fbln 3) knock-in mice showed that the R345W mutation leads to the photoreceptor outer segment detachment from the RPE (36). Therefore, considering the function of Fbln2 as an adherent ECM protein and its anti-migratory capabilities, it is likely that it may have a similar role in RPE-photoreceptor adherence. Our future studies will analyze the effects of endogenous fibulin 2 on RPE migration and adhesion by studying its effect in the fibulin 2 Ϫ/Ϫ mice.
Previous studies using surface plasma resonance assays have shown that mouse Fbln2, with an 'RGD' motif in the N-terminal cysteine-free domain, can adhere to cells by direct binding to multiple integrins (48). It can also modulate the adhesiveness of the ECM to the cells by binding other 'RGD' containing ECM proteins such as collagen IV, fibronectin and laminin (60,62). Our in vitro adhesion assays show that ARPE19 cells can adhere to purified Fbln2. Future studies will focus on identifying the RPE integrins involved in this process. Current therapies to treat retinal detachment involve surgical and nonsurgical techniques. The nonsurgical procedures involve cryotherapy using a freezing probe to seal the retina to the RPE; diathermy uses heat to seal the retina to the RPE; for laser therapy and pneumatic retinopexy, a gas bubble is introduced to allow the retina to float back into position, after which the hole is sealed with a laser. Surgical procedures involve scleral buckle surgery, where a silicon band is sewn to the sclera and pulled tight, then the sclera buckles and causes the retina to stick back to the RPE and vitrectomy, and then the scar tissue that is responsible for pulling the retina from the RPE is removed.
In summary, this report demonstrates a role for the ECM protein Fbln2 in ocular tissues and provides evidence for its up-regulation following retinal detachment. A more definitive role of Fbln2 in retinal detachments will be addressed in future studies involving the use of Fbln2 knock-out animals.