Proteomic Analysis of Exfoliation Deposits

Abstract

purpose. To increase knowledge of the biochemical composition of lenticular exfoliation material (XFM) by using proteomic approaches.

methods. Anterior lens capsules from patients with and without exfoliation syndrome (XFS) were homogenized in formic acid and subjected to cyanogen bromide (CNBr) cleavage, and the pattern of chemically generated fragments was compared by SDS-PAGE after silver staining. Unique XFS bands not present in control cases were excised, digested with TPCK-trypsin, and the resultant peptides sequenced with quadrupole time-of-flight mass spectrometry (MS). In parallel experiments, CNBr-fragmented XFM was separately digested in solution with trypsin and elastase, and the resultant peptide mixture was analyzed by liquid chromatography coupled to tandem MS followed by identification through homology searches at nonredundant protein databases. Immunolocalization of the MS-identified components were performed in XFS versus control samples by using conventional immunohistochemical methods and light microscopy.

results. In addition to fibrillin-1, fibronectin, vitronectin, laminin, and amyloid P-component, which are well-known extracellular matrix and basement membrane components of XFM, the proteomic approaches identified the multifunctional protein clusterin and tissue inhibitor of metalloprotease (TIMP)-3 as well as novel molecules, among them fibulin-2, desmocollin-2, the glycosaminoglycans syndecan-3, and versican, membrane metalloproteases of the ADAM family (a disintegrin and metalloprotease), and the initiation component of the classic complement activation pathway C1q. In all cases, classic immunohistochemistry confirmed their location in XFM.

conclusions. A novel solubilization strategy combined with sensitive proteomic analysis emphasizes the complexity of the XFS deposits and opens new avenues to study the molecular mechanisms involved in the pathogenesis and progression of XFS.

Exfoliation syndrome (XFS) is an age-related disorder that constitutes the most common identifiable cause of glaucoma, accounting for most of the cases in some countries and causing both open-angle and angle-closure glaucoma. ¹ The hallmark of the disease is the pathologic production and accumulation of an abnormal fibrillar extracellular material (XFM) in many intraocular tissues. ² The most prominent XFM accumulations are usually found in the anterior segment of the eye, leading to numerous clinical complications. ^{2 3} Ultrastructural evidence suggests that, among other intraocular cell types, XFM is mainly produced by epithelial cells of the iris, lens, and ciliary body. ^{3 4} However, the exact composition of the pathologic material, as well as the molecular mechanisms responsible for its excessive production and progressive accumulation remains undetermined. Iridolenticular friction leads to loss of XFM from the anterior lens surface and disruption of the iris pigment epithelium, resulting in pigment deposition in the trabecular meshwork, which also produces XFM locally. The primary cause of chronic pressure elevation appears to be the active involvement of trabecular cells and Schlemm’s canal cells in particular, in the generalized pathologic process with subsequent degenerative changes of Schlemm’s canal and adjacent tissues. ²  

Current knowledge of the molecular components of exfoliation material (XFM) is mostly based on data obtained via immunolabeling techniques. By these means, a number of extracellular matrix proteins have been demonstrated as an integral part of the deposits including fibrillin-1, ^{5 6} laminin, entactin/nidogen, fibronectin, vitronectin, ⁷ elastin, ^{7 8} and serum amyloid P component (SAP). ^{7 9} Also present as intrinsic components of the XFM are the glycosaminoglycans (GAGs) heparan sulfate and chondroitin sulfate components of proteoglycans (PGs), the broadly distributed nonprotein constituents of basement membranes as well as the modulators of extracellular matrix (ECM) formation such as the growth factor TGF-β1 and its latent-form binding proteins LTBP-1 and LTBP-2. ¹⁰ In only one published study was an attempt made to identify the XFM components using SDS-PAGE analysis of anterior lens capsules. ¹¹ The study described the presence of ∼14- and 16-kDa bands that were absent in control cases, but did not provide further molecular identification of these proteins. 

A major challenge to the biochemical analysis of the components of XFM has been the difficulty of solubilizing the deposited fibrillar material. In the present studies taking advantage of newly developed proteomic approaches, we performed cyanogen bromide/formic acid (CNBr/FA) peptide solubilization followed by quadrupole time-of-flight mass spectrometry (Q-ToF-MS) together with liquid chromatography–tandem mass spectrometry (LC-MS/MS). These studies unveiled novel components of the XFM, including cell adhesion molecules and ECM proteins, nondescribed PGs, complement proteins, matrix metalloproteases and specific inhibitors, as well as the presence of the multifunctional protein clusterin as a major component of the deposits. Immunochemical studies corroborated the MS findings and further demonstrated the presence of additional classical pathway activation-derived complement proteins. The association of the newly described components of XFM with binding, stabilization, cross-linking, and ECM degradation support the notion of XFS as a complex ECM disorder and open new avenues to explore pathogenetic and mechanistic aspects of the disease and the search for potential disease biomarkers. 

Materials and Methods

Anterior lens capsules were obtained from patients undergoing cataract surgery. The exfoliation group consisted of those patients who had XFM on the lens capsule, as determined by the slit lamp examination. Eyes with the presence of capsular fibrosis, posterior synechiae, or other debris on the anterior lens surface and a history of uveitis were excluded from the study. The control group consisted of those patients having cataract surgery who on slit lamp examination did not have XFM on the lens, iris, or the trabecular meshwork nor iris transillumination defects or loss of papillary ruff. All participants signed informed consents and the pertinent study protocol was approved by the NYU and NYEEI Institutional Review Boards and adhered to the tenets of the Declaration of Helsinki for experiments involving human tissue. 

All chemicals, unless otherwise indicated, were purchased from Sigma-Aldrich (St. Louis, MO). Rabbit polyclonal antibodies to complement component C1q and the activation-derived fragments C3c and C4c were purchased from Dako (Carpinteria, CA). Rabbit polyclonal antibody to the activation-derived neoepitope of complement terminal complex C5b-9 was from Calbiochem (San Diego, CA). Monoclonal antibody immunoreactive with the activation-derived Bb factor of the alternative pathway was obtained from Quidel (San Diego, CA). Polyclonal antibodies recognizing the multifunctional proteins vitronectin and clusterin, which are also known for their role as complement membrane attack (MAC) inhibitors, were purchased from Chemicon International (Temecula, CA) and Santa Cruz Biotechnology (Santa Cruz, CA), respectively. Mouse monoclonal antibodies to the cell adhesion protein desmocollin-2 (clone 7G6) were from Zymed Laboratories (Invitrogen Corp., San Francisco, CA). Antibodies recognizing metalloprotease members of the ADAM (a desintegrin and metalloprotease) family were purchased from different providers: rabbit polyclonal anitbodies anti-ADAM19 and anti-ADAM21 from Chemicon International, and mouse monoclonal anti-ADAMTS-8 from Abcam Inc., (Cambridge, MA). Rabbit polyclonal antibody to the tissue inhibitor of metalloprotease (TIMP)-3, was purchased from Abcam, Inc.. Antibodies to the various ECM components were obtained from different commercial sources: monoclonal anti-fibronectin (clone FBN11) from Oncogene Research (San Diego, CA); rabbit polyclonal anti-fibulin-2 from Santa Cruz Biotechnology; the monoclonal antibodies anti fibrillin-1 (clone 69) and anti-entactin/nidogen (clone ELM1) from Chemicon; monoclonal anti-chondroitin sulfate (clone CS-56) from Sigma-Aldrich; polyclonal antibodies versus syndecan-3, and versican from Santa Cruz Biotechnology; and polyclonal anti-serum amyloid P component (SAP) from Dako. 

Lens Capsule Preparation and Electrophoretic Analysis

Solubilization Strategies.

Lens capsules of XFS cases in triplicate were subjected to the action of different solubilizing agents and the remaining XFM evaluated by microscopic visualization after hematoxylin staining and/or SDS-PAGE. For this purpose, capsules were separately incubated with 99% formic acid (FA; 1–24 hours), hexafluoroisopropanol (HFIP, 1 hour), 6 M guanidine HCl (24 hours; Pierce Biotechnology, Inc., Rockford, IL), Laemmli sample buffer containing 8 M urea (24 hours), or a mixture of 50% acetonitrile (ACN)/0.1% trifluoroacetic acid (TFA; 1 hour). 

Cyanogen Bromide/FA Fragmentation.

In duplicate independent experiments, separate pools of XFS lens (n = 3 for each experiment) and control capsules (n = 3 for each experiment) were washed with PBS containing a protease inhibitor cocktail (Complete; Roche Applied Science, Indianapolis, IN) to remove epithelial cells, and homogenized in 100 μL cyanogen bromide (CNBr, Sigma-Aldrich) in 70% FA at an approximate CNBr-to-protein ratio of 10:1 (wt/wt). After overnight incubation at room temperature, the material was lyophilized, resuspended in 250 μL of 20 mM Tris-HCl (pH 8.2) containing 8 M urea, heated at 95°C for 10 minutes, and further incubated overnight at room temperature with continuous stirring. Sample aliquots (20 μL) were combined with equal volumes of 2× Tris-Tricine SDS sample buffer (Bio-Rad, Hercules, CA) containing 8 M urea and 0.1 M dithiothreitol (DTT), and separated in a 4% to 16% linear gradient SDS-Tris-Tricine gel electrophoresis. The resultant electrophoretic pattern was visualized by silver stain (Silver Quest Silver Staining Kit; Invitrogen Corp., Carlsbad, CA) according to the vendor’s protocol. Molecular weight standards (GE Healthcare, Piscataway, NJ) were myosin (220 kDa), phosphorylase b (97 kDa), bovine serum albumin (66 kDa), ovalbumin (45 kDa), carbonic anhydrase (30 kDa), soybean trypsin inhibitor (20.1 kDa), lysozyme (14.3 kDa), aprotinin (6.5 kDa), and insulin β- (3.5 kDa) and α- (2.5 kDa) chain. 

LC Coupled with Q-ToF-MS

The silver-stained protein bands of interest were excised from the gel under a laminar flow tissue-culture hood to minimize contamination and were destained for 15 minutes with a mixture of 30 mM potassium ferricyanide and 100 mM sodium thiosulfate, to remove silver ions and enhance sensitivity for the subsequent MS analysis. ¹² The gel slices were cut into ∼1-mm³ pieces, digested with TPCK-trypsin (1:50 wt/wt; Roche), and the resultant peptides were dried under vacuum. After solubilization in 20 μL of 2% ACN/0.1% FA, samples were subjected to liquid chromatography tandem mass spectrometry (LC-MS/MS) ¹³ (CapLC; Waters, Milford, MA, with a Micromass Q-ToF mass spectrometer; Waters). Samples were separated by using a 75-μm internal diameter × 15-cm C18 analytical column (PepMap; LC Packings-Dionex Corp. Sunnyvale, CA) equilibrated in 2% ACN/0.1% FA, at 400-nL/min flow rate obtained with a home-made 1:10 pre-column flow splitter, and a 70-minute 5% to 90% ACN/0.1% FA gradient though a 10-μm internal diameter emitter (Picotip; New Objective, Inc., Woburn, MA). Automatic switching between MS and MS/MS modes were controlled by commercial software (MassLynx ver. 4.0; Micromass), dependent on both signal intensity and charge states from MS to MS/MS and on either signal intensity or time from MS/MS to MS. The MS/MS data were processed (MassLynx; Micromass) and the deconvoluted spectra directly used to search in batches the human NCBI nonredundant protein database using a commercial search program ¹⁴ (Mascot; Matrix Science, London, UK). 

LC Coupled with Ion Trap MS/MS

In separate duplicate experiments, independent pools of XFS (n = 6) and control (n = 6) lens capsules were solubilized with CNBr/FA, as described earlier. After the chemical cleavage, each group was divided into two aliquots for further proteolytic digestion and lyophilized. Samples were dissolved in either 100 mM Tris-HCl buffer (pH 8.2) or ammonium bicarbonate (pH 8.5) containing 2 mM calcium chloride for trypsin or elastase digestion, respectively. Proteolytic degradation was performed overnight at 37°C in a thermomixer at 1200 rpm, with either TPCK-trypsin as for LC-Q-ToF MS/MS, or with porcine elastase (1:50 wt/wt; Roche). Samples were subsequently analyzed with an ion trap mass spectrometer (LCQ; Thermo Fisher Scientific, Waltham, MA). ¹⁵ One-dimensional LC was then performed (Ultra Plus II Proteomic System; MicroTech, Vista, CA equipped with RP-capillary columns packed with 5 μm C18, 300Å particles). Samples were loaded onto the column equilibrated in 99% mobile phase solvent A (5% ACN/1% FA) and 1% solvent B (95% ACN/0.8% FA), and peptides were eluted with a 100-minute linear gradient (1%–95% B) directly into the mass spectrometer equipped with an LCQ nanospray ion source (Thermo Finnigan) and 10 μm (ID) noncoated nanospray emitters (SilicaTip PicoTip; New Objective, Inc.), with the electrical contact made through a liquid junction at the PEEK union (PerkinElmer Life and Analytical Sciences, Inc., Wellesley, MA). The spray voltage was set at 1.5 kV and the heated capillary at 160°C. 

MS/MS spectra were acquired with commercial software (Xcalibur 1.3; Thermo Fisher Scientific) by the following method: A full MS scan followed by three consecutive MS2 scans of the top three ion peaks from the preceding full scan was obtained by using dynamic exclusion (four repetitions in 1.5 minutes were excluded for 3 minutes). Data were analyzed on a commercial system (Bioworks 3.1, a new Beta test site version; Beckman Coulter, Fullerton, CA; with the SEQUEST algorithm to determine cross correlation scores between the acquired spectra and a protein database; Thermo Finnigan). The following parameters were used for the search analyses (TurboSEQUEST): No enzyme was chosen for the protease, as not all proteins are digested to completion; molecular weight range, 400 to 4500; threshold, 1000; monoisotopic, precursor mass, 1.4; group scan, 10; minimum ion countZ 20; charge state, auto; peptideZ 1.5; fragment ions, 0; differential amino acid modifications, Cys 57.0520. Results were filtered using SEQUEST cross-correlation scores greater than 1.5 for +1 ion, 2.0 for +2 ions, and 2.5 for +3 ions. 

Immunohistochemical Studies

Fifteen anterior lens capsules from eyes with XFS (mean age, 72.38 years; range, 68–80) and eight control lens capsules (mean age, 74.50 years; range, 61–83) were used for immunohistochemical analysis on 6-μm cryostat sections. After equilibration to room temperature, the sections were fixed for 5 minutes in 10% paraformaldehyde, rinsed in PBS, and incubated for 1 hour with the pertinent primary antibodies (Table 1)followed by the corresponding biotinylated anti-rabbit or anti-mouse secondary antibodies (BioSource International, Camarillo, CA). In negative control experiments, the primary poly- or monoclonal antibodies were replaced by equivalent dilutions of normal, nonimmune rabbit or mouse serum, respectively (Sigma-Aldrich) or equal concentrations of an unrelated primary antibody. In all cases incubation with the secondary antibodies was followed by reaction with the ABC complex (Dako), color-developed with diaminobenzidine/H₂O₂ and the sections counterstained with hematoxylin. 

Results

XFM Extraction, CNBr/FA Solubilization, SDS-PAGE and Q-ToF-MS

A major obstacle in the biochemical identification of the XFM components is the difficulty in dissolving the deposited fibrillar material. As illustrated in Figure 1 , XFM remained unaltered after 24 hours in PBS (Fig. 1A)or after incubation of XFS lens capsules with various solubilizing agents, including elevated concentrations of the chaotropic agents urea and guanidine (Fig. 1B 1C , respectively), high polarity solvents like HFIP (Fig. 1D) , strong acids as FA (Fig. 1E 1F)or hydrophobic solvents such as 50% ACN/0.1% TFA (data not shown) successfully used to solubilize fibrillar amyloid deposits. ¹⁶ In all these cases a significant proportion of the XFM remained basically unaltered rendering very low yield of extracted proteins. To enhance the release of material from the deposits and obtain biochemical information about their composition, we used traditional protein chemistry techniques newly adapted for proteomic approaches and performed CNBr/FA solubilization, followed by analysis of the resultant fragments by SDS-PAGE and MS. Chemical cleavage of proteins by CNBr and FA, widely used in protein chemistry studies for four decades, results in cleavage of the peptide bonds Met-X, ^{17 18} and Asp-Pro, ¹⁹ respectively. The limited frequency of these peptide bonds results in the generation of large peptide fragments adequate for subsequent analysis, even in the presence of incomplete digestion products. 

The CNBr/FA chemical cleavage of the samples successfully released enough material to allow visualization in SDS-PAGE after a sensitive silver staining methodology was used that was compatible with the subsequent MS analysis. As illustrated in Figure 2 , a ∼2-kDa band was clearly visible on the XFS sample (Fig. 2A , arrow), whereas it was either nonexistent or present in trace quantities in the normal control samples. Q-ToF MS analysis of this band yielded the sequence ASSIIDELFQRFFTR, which was identified as a peptide fragment of clusterin comprising residues 183 to 194 of the protein primary sequence (P < 0.05). Immunostaining with specific antibodies confirmed the presence of clusterin in the XFM (Fig. 2C) . 

Identification of Unique Components of XFM by LC Coupled with Ion Trap MS/MS

To confirm the presence of clusterin and obtain further information about the composition of XFM, we subjected lens capsules of XFS and control samples, solubilized with CNBr/FA, to in-solution elastase and trypsin digestion, as described in the Methods section, and directly analyzed them with LC-MS/MS. Results from searches of human databases identified several proteins in the XFS samples that were either not present or at a significantly lower concentration in normal control subjects. Among the molecules identified as constituents of the lens capsules with XFM were ECM and basement membrane proteins such as fibronectin, fibrillin-1, fibulin-2, vitronectin, laminin, and SAP, cell adhesion molecules such as desmosomal cadherins (desmocollin-2), PG core proteins like syndecan-3 and versican, metalloproteases of the ADAM family (ADAM-19, ADAM-21, and ADAMTS-8), tissue inhibitors of metalloproteases such as TIMP-3, as well as the C1q complement component and the multifunctional glycoprotein clusterin. The aforementioned XFM components were not identified in normal control samples, suggesting that they were either absent or at much lower concentration, thereby falling below the limit of detection of the MS analysis. Reproducible results were obtained in duplicate experiments; moreover, LC-MS/MS after trypsin and elastase digestion yielded comparable peptide profiles, further validating the proteomic approach. 

Immunohistochemical Studies

A series of immunohistochemical studies were performed to corroborate the presence of the novel components identified by MS and to validate further the LC-MS/MS approach. As indicated in Figures 3A and 3B , in association with the previously described constituents fibronectin, vitronectin, SAP, fibrillin-1, SAP, entactin/nidogen, elastin, and TIMP-3, ^{7 9 20} immunohistochemistry also confirmed the presence of fibulin-2, versican, syndecan-3, desmocollin-2, ADAM-19, ADAM-21, and ADAMTS-8 in XFM. 

Based on the identification of clusterin, a fluid phase inhibitor of the complement activation pathway, as a main component of the XFM (Fig. 2) , and the additional demonstration of C1q in the LC/MS-MS analysis, additional immunohistochemical analyses were conducted to investigate the presence of other activation products of the complement system. As illustrated in Figure 4 , the deposits not only stained with antibodies to the recognition component C1q corroborating the MS data, but also with antibodies recognizing the activation derived products of the classical complement pathways C3c and C4c. On the contrary, no staining was observed for the neoepitope of the C5b-9 complex or for Bb, the activation-derived and central component of the alternative pathway. 

Discussion

The biochemical analysis of the fibrillogranular material that progressively accumulates on the lens, iris, cornea, zonules, and other areas of the eye in XFS ²¹ has been difficult to pursue due to technical limitations primarily related to its poor solubility and the limited amount of material that can be retrieved from the lesions. The current knowledge of XFM deposits therefore basically relies on immunolabeling approaches that have demonstrated a complex glycoprotein–proteoglycan composition with a predominance of elements of the basement membrane and elastic fiber systems. Cross-linking processes mediated by tissue transglutaminase seem also to participate in XFM fiber formation and stabilization, likely contributing to the matrix mechanical strength and resistance to enzymatic degradation. In this sense, patients with XFS have been shown to have upregulation of the gene codifying for transglutaminase-2, the most widespread member of the transglutaminase gene family and the major cross-linker in ocular connective tissues. ²² Our data further support the existence of highly cross-linked molecules in the XFS deposits; a variety of treatments (e.g., the use of chaotropes, detergents, strong acids, organic solvents, and mixtures of them) consistently yielded minimal amounts of extracted material, a clear indication that disrupting ionic and/or hydrophobic interactions—strategies that were successful in the solubilization of fibrils from cerebral and systemic amyloid deposits ^{16 23 24} —do not produce the same effect in XFM. The combination of low pH and CNBr chemical cleavage targeting Met-X and Asp-Pro peptide bonds, however, succeeded in releasing soluble material for further analysis via SDS-PAGE/Q-ToF and LC-MS/MS, both powerful techniques that allow the identification of complex protein samples while providing detection with high sensitivity and mass accuracy. As a result, this proteomic approach successfully rendered the first biochemical data on XFM, corroborating the presence of previously immunoidentified components while revealing novel matrix-associated elements, reinforcing the notion that XFS correlates with a progressive accumulation of ECM components. ²⁵  

Extracellular Matrix, Cell Adhesion, and Basement Membrane Components of XFM

Besides retrieving the already known components of XFM fibronectin, laminin, vitronectin, SAP, and fibrillin-1 (reviewed in Ref. ²⁶ ), our proteomic approach identified additional ECM, cell adhesion, and basement membrane components, among them the glycoproteins fibulin-2 and desmocollin-2 and the PG core proteins syndecan-3 and versican. Fibulin-2 is a 1157-amino-acid glycoprotein that belongs to a six-member family of extremely versatile ECM molecules highly conserved in evolution, even in distant species. ²⁷ All members of the fibulin family share an elongated structure containing several copies of epidermal growth factor-(EGF)-like modules distributed in tandem. Fibulin-2, like fibulin-1, has three additional domains with homology to the complement anaphylotoxins C3a, C4a, and C5a. These anaphylotoxic molecules, generated enzymatically during activation of their parent precursors C3, C4, and C5, respectively, exert potent vasoactive and inflammatory effects in vivo. Of all fibulins, fibulin-2 is the one with the most complex structure, forming disulfide-linked dimers mediated by the presence of an extra cysteine residue. The fibulins share binding sites, most of them calcium dependent, for several basement-membrane proteins, among them elastin, fibrillin-1, fibronectin, and PGs. As a consequence of these widespread interactions, fibulins are believed to function as intramolecular bridges that stabilize the organization of supramolecular ECM structures, such as elastic fibers and basement membranes. Taking into account its high-affinity binding to elastin and fibrillin-1 ²⁷ and that it is the only member of the fibulin family with reported binding activity for fibrillin-1 and perlecan, ²⁸ the presence of fibulin-2 in XFM is not a surprising finding. 

Among the novel components of XFM revealed by the studies presented herein, one of them belongs to a category of proteins never previously associated with these deposits. Desmocollin-2 is a member of the cadherin family of cell adhesion molecules and constitute one of the major glycoproteins in the desmosomal epithelial cell–cell junctions essential in the maintenance of tissue integrity and architecture. ²⁹ All desmocollins contain five cadherin domains in the extracellular region of the molecule which, through the calcium-mediated cell–cell interactions, play an active role in cell signaling, proliferation, differentiation, and migration. Three distinct desmocollin isoforms have been described, each protein the product of a separate gene. Desmocollin-2, which we identified in XFM, is ubiquitous in desmosome-containing tissues and is the only isoform expressed in corneal epithelia. ³⁰ Although no other cadherin has so far been associated with the deposits, it is interesting to note that some of the ECM components known for many years to constitute the deposits (e.g., fibronectin and vitronectin) also exhibit cell adhesion properties. Whether other desmosomal components or additional members of the cadherin family are also present in XFM at concentrations below the detection level of the methods used in our studies remains to be elucidated. 

It is known that XFM also contain PG moieties, as indicated by histologic staining with PAS, Alcian blue, and ruthenium red, ^{31 32} and by its immunoreactivity with antibodies for heparin-, chondroitin-, keratin-, and dermatan-sulfate GAG side chains. ^{7 32} The biochemical diversity of PG molecules, which is only beginning to be uncovered, arises from the presence of different core proteins associated with GAG chains of various lengths and types. Of the different core protein structures of the central nervous system (CNS) PGs, our studies identified versican and syndecan-3 in the XFM deposits. Versican, named for its highly interactive and versatile functions ³³ and also known as chondroitin sulfate proteoglycan core protein 2, is a structural molecule with major involvement in cell growth and differentiation. Based on sequence alignments, versican is considered a member of the lectican family of PGs, which includes also aggrecan, neurocan, and brevican. These PGs share common structural features such as a hyaluronan-binding domain at the N terminus as well as two EGF-like domains, one lectin-type-like and one complement regulatory protein–like domain. In contrast to the limited distribution of aggrecan to the cartilage and that of neurocan and brevican to the brain, versican is more ubiquitous. It is known to interact with other ECM molecules, including fibulin-2 and -1. ³³ It has been proposed that in the assembly of elastic fibers, fibulin serves as a bridge between versican and fibrillin, forming the highly ordered multimolecular structures essential in fiber formation. ^{34 35} In the eye, versican is thought to play a role in regulating aqueous humor outflow and intraocular pressure via the human trabecular meshwork. ³⁶ Syndecans, on the other hand, are type I integral membrane PGs containing both chondroitin- and heparan-sulfate groups and are involved in cell-ECM adhesion, growth factor binding, and activity regulation, as well as signal transduction pathways. Many of them function through binding interactions with diverse extracellular ligands including, but not limited to, growth factors and their receptors, cytokines, ECM proteins, lipoproteins, proteases, and their inhibitors. Although syndecans are transmembrane proteins, they undergo ectodomain shedding, a physiological process associated with their biological functions ³⁷ as well as with their presence in ECM and codistribution with fibronectin, laminin, and collagen. ³⁸ Although we identified only syndecan-3 in XFM, expression of all four members of the syndecan family was shown in cells of the anterior segment of the normal human eye. ³⁸  

ECM Degrading Enzymes and Their Inhibitors in XFM

ECM matrix metabolism and turnover is greatly orchestrated by closely related metalloprotease families: MMPs and metalloprotease-disintegrins (ADAMs), the latter also comprising an extended subfamily containing thrombospondin-like domains, the ADAMTSs. The activity of all these Zinc-dependent proteases with documented roles in connective tissue organization, coagulation, inflammation, angiogenesis, and cell migration is selectively regulated by endogenous TIMPs. MMPs comprise a family of more than 20 enzymes, many of them showing disregulated activity and abnormal levels in aqueous humor and differential gene expression in anterior segment tissues from patients with XFS. ^{25 39} Unlike the mammalian ADAMs, which with few exceptions are all transmembrane proteins, the ADAMTSs are secreted molecules, some of which bind to ECM components (reviewed in Ref. ⁴⁰ ). In fact, many of the ADAMTSs functions relate to their participation in the processing of ECM components, including procollagens ⁴¹ and PGs. Activity against PGs such as aggrecan, versican, and brevican, form the basis for the designation of some of the ADAMTS family members as aggrecanases. ^{42 43} In contrast to other metalloproteases, ADAMs are particularly important for the cleavage-dependent activation of diverse cell surface molecules including Notch, amyloid precursor protein (APP) and transforming growth factor-α; based on their homology to snake venom disintegrins, they also interact with integrin receptors. ⁴⁰ As a result, and not surprisingly, ADAMs play important roles in the development of the nervous system, regulating proliferation, migration, differentiation, and survival of various cell types, as well as axonal growth and myelination. Among the more than 30 known mammalian ADAMs, at least 17 of them have been described in the nervous system, including ADAM19 and -21, both exhibiting high mRNA expression in the CNS ⁴⁴ and both identified in our proteomic analysis as components of the XFM. ADAMTS-8, also identified in our studies, is one of the family members with highest expression in brain tissues. Although its function remains still poorly understood, ADAMTS-8 together with ADAMTS-1 are the only members of the ADAMTS subfamily with antiangiogenic properties. ⁴⁵  

Under physiological conditions, the activity of all matrix metalloproteases is precisely regulated by specific endogenous TIMPs in a 1:1 stoichiometry. Four TIMP family members have been described (TIMP-1, -2, -3, and -4) which, with very few exceptions, inhibit all matrix metalloproteases, although TIMP-3 appears to be more efficient for the ADAMs and ADAMTS than for MMPs (reviewed in Ref. ⁴⁶ ). TIMP-3 is composed of 187 amino acids and, as do all TIMP family members, contains the VIRAK consensus sequence necessary for inhibitory activity as well as six disulfide bonds arranged in three knotlike structures. TIMP-3 shows <40% amino acid homology with the other TIMP members. It is poorly soluble and is found primarily associated with the ECM in a variety of tissues widely distributed throughout the body; TIMP-3 binds tightly to sulfated GAGs, and these interactions may explain its location in the XFM. TIMP-1, -2, and -4, on the contrary, are soluble proteins present in many biological fluids. TIMP-1 and -2 have been shown to be elevated in aqueous humor from patients with XFS. ^{39 47} The identification of TIMP-3 by proteomic analysis confirms previous immunohistochemical studies in XFS (Schlötzer-Schrehardt U et al. IOVS 2002;43:ARVO E-Abstract 3369). Of note, defects in TIMP-3 are the cause of Sorsby fundus dystrophy, a rare autosomal dominant macular disorder with an age of onset in the fourth decade and characterized by loss of central vision from subretinal neovascularization and atrophy of the ocular tissues. ⁴⁸ It has been suggested that an imbalance in the MMP/TIMP ratio in aqueous humor may promote the abnormal matrix accumulation characteristic of XFS. ^{25 39 47} Our findings identifying novel protease partners and protease inhibitors in matrix metabolism as components of the XFM, suggest that this delicate balance is even more complex was than originally believed. 

Complement Components in XFS

Sequential activation of the complement system occurs mainly by the classical or the alternative pathways. ^{49 50 51} The former, triggered through binding to C1q, results in the subsequent cascade activation of C4, C2, C3, and C5, leading to the final assembly of the cytolytic C5b-9 MAC through the generation of hydrophobic binding sites which allow its insertion in cellular membranes. The alternative pathway evolves around C3 activation through the specific components factor B, factor D, and properdin; once C3 is activated, the cascade proceeds as in the classical pathway toward the formation of the final MAC. Because the complement system is a powerful mechanism, activation of which has the potential to damage normal host cells severely, it is under the tight molecular control of regulatory molecules at various points in the cascade. MAC formation is modulated by both the cell membrane-associated CD59 and the circulating and multifunctional proteins clusterin and vitronectin which exert their inhibitory function by binding to the C5b-7 complex preventing its insertion into lipid membranes and precluding the MAC assembly. ^{52 53} Our findings suggest the possibility that XFS may be associated with some degree of complement activation in vivo, as previously proposed (Ovodenko B et al. IOVS 2005:46:ARVO E-Abstract 3763). Whether this activation is directly mediated by the deposits or it correlates in any way with viral infections postulated as nongenetic factors contributing to XFS pathogenesis ⁵⁴ remains to be elucidated. The colocalization in XFM of the recognition component C1q and the activation-derived fragments C3c and C4c together with the absence of Bb indicate activation of the classical pathway. Whether the lack of progression to the final stages of MAC assembly suggested by the absence of C5b-9 immunoreactivity relates to the inhibitory effect exerted by vitronectin and clusterin remains to be determined. The presence of complement inhibitors together with the activated components are common findings at sites of activation in many disorders of unrelated origin including immune complex deposits in autoimmune diseases, ⁵⁵ fibrillar amyloid lesions associated with Alzheimer’s disease ⁵⁶ and other amyloidoses, ^{57 58 59} as well as abnormal accumulations of extracellular material present in drusen formation in age-related macular degeneration. ⁶⁰ It is noteworthy that a study of a limited number of samples of aqueous humor from individuals with XFS reported an elevated concentration of the proinflammatory C3a anaphylotoxin, ⁶¹ also suggesting complement activation in vivo. Complement anaphylotoxins, generated at early stages of the system activation, exert potent local effects that ultimately result in cytokine-mediated cellular injury leading to a self-perpetuating cycle of inflammatory events. ⁵⁰ This local inflammation could in some way correlate with the low-grade inflammatory conditions observed in XFS and indicated by the gene expression analysis demonstrating upregulation of proinflammatory cytokines. ²⁶ Whether these effects are relevant elements in the pathogenesis of XFS and they are further enhanced by molecules containing anaphylotoxin-like domains (e.g., fibulin-2, as indicated earlier) is not known. Certainly, codeposition of fibulin-2 with activation products of the complement classical pathway in XFM suggests a potential mechanistic connection. 

Clusterin in XFM

The multifunctional glycoprotein clusterin is a novel component of XFM, as reported (Ovodenko B et al. IOVS 2005:46:ARVO E-Abstract 3763) and recently confirmed. ⁶² Under our experimental conditions, clusterin was identified by both, Q-ToF and LC-MS/MS experiments and its prevalence in the deposits was suggested by the number of hits obtained in LC-MS/MS, by its recovery in XFM proteolytic fragments visible in standard gel electrophoresis and by the consistently strong immunoreactivity of the deposits. Clusterin, a ubiquitous molecule also known as apolipoprotein J, has been implicated in a wide variety of physiological and pathologic processes, including lipid transport, apoptosis, stabilization of cell–cell and cell–matrix interactions, prevention of complement activation and stabilization of protein folding following stress-induced denaturation. ^{63 64} It is mainly a secreted glycoprotein although nonglycosylated intracellular forms, both cytoplasmic and nuclear, originated by alternative splicing mechanisms have also been described. ^{65 66} Clusterin mRNA is present in almost all mammalian tissues ^{67 68} and protein expression occurs in nearly all body fluids. ^{68 69 70} In the eye, clusterin is expressed in most ocular cells and tissues ⁶² including corneal and conjunctival epithelium, ^{71 72} corneal endothelium, ⁷³ ciliary body, and retina. ^{74 75 76} It is also localized to extracellular structures such as ocular basement membranes and stromal fibers ⁶² and present in both aqueous and vitreous humors. ^{62 75} In plasma, clusterin is preferentially located in high-density lipoprotein particles, actively participating in the mechanism of reverse-cholesterol transport. ⁷⁷ As discussed earlier, secreted clusterin also acts as a fluid phase inhibitor of the MAC formation, ^{49 50 51} a mechanism that may have relevance in XFS (Ovodenko B et al. IOVS 2005;46:ARVO Abstract 3763). However, the most important and puzzling biological function of clusterin appears to pertain to its enhanced expression during cellular stress. It participates in apoptosis signaling and oxidative stress mechanisms ⁶³ and is an extracellular chaperone with the ability to bind a wide variety of partly unfolded stressed proteins and several hydrophobic ligands through its unstructured, molten, globule-like regions of the molecule. ^{78 79} As some of the multiple examples of this activity, clusterin interaction with soluble Aβ Alzheimer’s amyloid in vitro has been shown to preclude its typical fibrillization and neurotoxicity, a protective effect seen also with other amyloid molecules and prion fragments (reviewed in Ref. ⁶³ ). Paradoxical to its protective chaperone activity, clusterin has been found codeposited in all amyloid lesions tested so far, irrespective of their location ⁶³ including forms restricted to ocular tissues. ⁸⁰ Because of its widespread presence in all amyloid deposits, clusterin is considered one of the so-called amyloid-associated proteins, a heterogeneous group of unrelated components that includes, among others, SAP, vitronectin, ECM proteins, GAGs, and complement proteins. These molecules colocalize with the amyloid lesions but are not a structural part of the final fibril, and it is still debatable whether they are innocent bystanders or their presence plays a key role in the mechanism of protein aggregation and fibrillogenesis. Interestingly enough, all these molecules are also components of XFM, suggestive of related mechanisms in the formation of XFS and amyloid deposits. 

The synthesis of clusterin at both the mRNA and protein levels in different cell types is stimulated by TGF-β1. ^{65 81 82 83 84 85} This modulatory effect is mediated by the consensus AP-1 site, the cognate transcription factor AP-1 and protein kinase C, ⁶⁵ with c-Fos playing a negative regulatory role in clusterin gene expression. ⁸⁶ TGF-β1 is significantly elevated in the aqueous humor of patients with XFS, and it is considered to be a key mediator in the formation of the fibrillar deposits. ¹⁰ However, this elevation does not translate in clusterin gene expression; on the contrary, gene expression analysis performed on combined tissues from XFS eyes—namely, the entire lens epithelium, iris tissue, and ciliary processes, indicated an overall downregulation of clusterin mRNA compared with age-matched control eyes. ²⁵ Moreover, RT-PCR and in situ hybridization analysis displayed downregulation of clusterin in tissues of the anterior segment, irrespective of the presence of glaucoma, whereas no differences were observed in tissues of the posterior segment. ⁶² Correlating with these data, the same investigators found low levels of clusterin in the aqueous humor of XFS as well as significant downregulation of clusterin mRNA by TGF-β1 in human nonpigmented ciliary cells in vitro. ⁶² Low levels of clusterin in the presence of strong clusterin immunoreactivity in XFM may be explained by the late onset of the disease and the slowly but chronic accumulation of clusterin in XFM over the years of disease progression, even in the presence of downregulated local synthesis. However, the existence of elevated TGF-β1 with decreased clusterin mRNA and protein in XFS is difficult to reconcile and warrants further studies. It may indicate the involvement of still undefined regulatory factors and/or the existence of particular binding interactions still uncovered in XFS. Alternatively, it may reflect a secondary event with negative feedback in clusterin mRNA production. 

Conclusions

XFS deposits represent a complex mixture of glycoproteins and PGs assembled in a supramolecular fibrillar structure. New molecules are being uncovered by biochemical and molecular biological approaches while new information about the cellular pathways involved is being provided by the analysis of gene expression patterns in XFS tissues. The nature and diversity of the molecules forming part of XFM undoubtedly indicate that XFS is a disorder that correlates with the excessive production, impaired catabolism, and subsequent accumulation of abnormal cross-linked ECM material. ²⁶ Matrix proteolytic imbalances, ^{25 39} low-grade inflammatory processes, ²⁶ and the presence of ischemia/hypoxia, ⁸⁷ together with dysregulated cellular and oxidative stress mechanisms, ^{26 88 89 90} also appear to be significant contributors to the pathobiology of XFS. All the newly available information highlights even more the complexity of the mechanisms underlying the pathogenesis of XFS. The complex biochemical composition, poor solubility, and fibrillar appearance of the deposits, the presence of common codeposited chaperone proteins, and the ability of the aggregated–fibrillar material to elicit some degree of complement activation, inflammatory response, and oxidative stress are XFS features also shared by an expanding group of clinically heterogeneous entities, collectively known as “protein folding disorders.” The group include, among others, Alzheimer’s, Parkinson, Huntington, and prion diseases; cerebellar ataxias; cataracts; type-II diabetes; and a variety of systemic amyloidosis. ⁹¹ In these disorders, through mechanistic pathways poorly understood, soluble proteins normally found in biological fluids change their native conformation and form insoluble structures that accumulate in the form of either intra- and extracellular aggregates or fibrillar lesions. Whether XFS is a new member of the protein folding disorders superfamily is a challenging hypothesis that awaits further confirmation. Certainly, the study of XFS will benefit from the expanding field of cutting-edge proteomic and genomic approaches while awaiting development of cell culture and animal models of the disease. 

²  Contributed equally to the work and therefore should be considered equivalent authors.

 Supported in part by the Joseph Cohen Research Fund of the New York Glaucoma Research Institute, New York, New York; the Fight For Sight Foundation, New York, New York; and National Institute of Neurological Disorders and Stroke Grants NS38777 and P30 NS050276, National Institute on Aging Grant AG10491, and National Center for Research Resources Grant S10 RR017990.

 Submitted for publication April 11, 2006; revised September 13, 2006; accepted February 2, 2007.

 Disclosure: B. Ovodenko, None; A. Rostagno, None; T.A. Neubert, None; V. Shetty, None; S. Thomas, None; A. Yang, None; J. Liebmann, None; J. Ghiso, None; R. Ritch, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Primary Antibodies Employed in the Immunohistochemical Analysis of XFS Cases

Antibody	Dilution	Source
Rabbit polyclonal anti-ADAM19	1:100	Chemicon International Temecula, CA
Rabbit polyclonal anti-ADAM21	1:100	Chemicon International
Mouse monoclonal anti-ADAMTS-8	1:100	Abcam, Inc., Cambridge, MA
Rabbit polyclonal anti-C1q	1:100	Dako, Carpinteria, CA
Rabbit polyclonal anti-C3c	1:100	Dako
Rabbit polyclonal anti-C4c	1:100	Dako
Rabbit polyclonal anti-C5b-9	1:100	Calbiochem, San Diego, CA
Mouse monoclonal anti-Bb	1:50	Quidel, San Diego, CA
Rabbit polyclonal anti-clusterin (H-330)	1:400	Santa Cruz Biotechnology, Santa Cruz, CA
Rabbit polyclonal anti-vitronectin	1:100	Chemicon International
Mouse monoclonal anti-chondroitin sulfate (clone CS-56)	1:100	Sigma-Aldrich, St. Louis, MO
Rabbit polyclonal anti-syndecan-3 (M-300)	1:100	Santa Cruz Biotechnology
Rabbit polyclonal anti-versican (H-56)	1:100	Santa Cruz Biotechnology
Rat monoclonal anti-entactin/nidogen (clone ELM1)	1:100	Chemicon International
Mouse monoclonal anti-fibrillin-1 (clone 69)	1:100	Chemicon International
Rabbit polyclonal anti-fibulin-2 (H-250)	1:100	Santa Cruz Biotechnology
Mouse monoclonal anti-fibronectin (clone FBN11)	1:100	Oncogene Research
Mouse monoclonal anti-desmocollin 2/3 (clone 7G6)	1:50	Zymed-Invitrogen Corp. S. San Francisco, CA
Rabbit polyclonal anti-TIMP-3	1:50	Abcam, Inc.
Rabbit polyclonal serum amyloid P component	1:100	Dako

Figure 1.

 

Exfoliation material extraction and electrophoretic analysis. XFS lens capsules were subjected in triplicate sets to the action of different solubilizing agents, and the remaining XFM was evaluated by microscopic visualization after hematoxylin staining and SDS-PAGE analysis. The remaining XFM after (A) PBS, 24 hours; (B) 2% SDS/8 M urea, 24 hours; (C) 6 M guanidine-HCl, 24 hours; (D) HFIP, 1 hour; (E) 99% FA, 1 hour; and (F) 99% FA, 24 hours.

Figure 2.

 

Identification of clusterin by Q-ToF LC-MS/MS and immunohistochemistry. Deconvoluted MS/MS spectrum of a triply charged ion of clusterin-derived peptide with the sequence ASSIIDELFQDRFFTR. (A) Anterior lens capsule specimens of XFS and control samples were subjected to CNBr/FA chemical cleavage and analyzed by using a linear 4% to 16% gradient SDS-PAGE, and the separated proteins were visualized by silver staining. Lane C: pooled specimens of control subjects (n = 3; ages: 62, 68, and 70 years); Lane EX: pooled specimens from patients with lenticular XFM (n = 3; ages: 69, 73, and 76 years). Note the presence of an extra band below the marker of 3.5 kDa (arrow) in the EX lane that is not visible in the control lane. (B) Partial mass spectrum of the triply charged precursor ion with observed monoisotopic mass 1943.98. The peptide ASSIIDELFQDRFFTR has a calculated monoisotopic mass of 1943.974. (C) Immunohistochemical analysis of anterior lens capsules of XFS indicates the positive immunoreactivity of XFM with anti-clusterin antibodies.

Figure 3.

 

Immunohistochemical analysis of extracellular matrix components and novel matrix metalloproteases on XFM. Frozen sections of PAS positive anterior lens capsules of XFS were immunostained with a panel of antibodies recognizing extracellular matrix and basement membrane components. Incubation with the primary antibodies specified in Table 1 , was followed by the corresponding biotinylated secondary antibodies and by ABC complex. Color was developed with diaminobenzidine/H₂O₂. (A) XFM labeled by antibodies to fibrillin-1, fibronectin, fibulin-2, nidogen/entactin, serum amyloid P component, and vitronectin. Also shown is the immunoreactivity for the desmosomal component desmocollin-2. No staining was seen when the primary antibody was omitted (indicated in the figure as control sample). (B) XFM is labeled by antibodies to versican, syndecan-3, chondroitin sulfate, ADAM-19, ADAM-21, ADAMTS-8, and TIMP-3. Also shown is the histologic staining PAS showing the GAG components of XFM.

Figure 4.

 

Immunohistochemical analysis of activation-derived complement components on XFM. Frozen sections of anterior lens capsules of XFS and control eyes were immunostained with a panel of antibodies recognizing activation-derived components of the classical and alternative complement pathways. Incubation with the primary antibodies was followed by the pertinent biotinylated secondary antibodies and the ABC complex. Color was developed with diaminobenzidine/H₂O₂. The XFM was immunoreactive with C1q as well as the classical path activation-derived fragments C3c and C4c, whereas Bb, the first activation-generated component of the alternative pathway was not present. Also negative was the labeling for the neoepitope of the activation-generated C5b-9 complex, suggesting the lack of successful completion of MAC assembly.

The authors thank the ophthalmologists of the Exfoliation Syndrome Study Group who provided additional capsulorrhexis specimens: Mark Jofe, MD, Regina Smolyak, MD, David Pinhas, MD, Tal Raviv, MD, and Daniel Laroche, MD (New York Eye and Ear Infirmary), and Leonard Bley, MD (Laser and Microsurgery Institute, Brooklyn, NY).