ARMS2 Is a Constituent of the Extracellular Matrix Providing a Link between Familial and Sporadic Age-Related Macular Degenerations

Abstract

Purpose.: SNPs in chromosomal region 10q26 harboring PLEKHA1, ARMS2, and Htra1 showed the strongest association with age-related macular degeneration. Recent evidence suggests that in patients homozygous for the risk allele, the lack of synthesis of the poorly characterized ARMS2 is causative of this disorder. The present study was undertaken to gain an understanding of the genuine (patho)physiological role of this protein.

Methods.: ARMS2-interacting proteins were identified by using a yeast two-hybrid system and validated by coprecipitation. Immunofluorescence was applied to reveal the localization of ARMS2 in transfected cells and in human eyes. Western blot analyses were performed on extra- and intracellular fractions of ARMS2-expressing cells to demonstrate the secretion of ARMS2.

Results.: Contrary to previous reports, this study showed that ARMS2 is a secreted protein that binds several matrix proteins. Notably, ARMS2 directly interacts with fibulin-6 (hemicentin-1). Mutations in the fibulin-6 gene have been demonstrated to cause familial AMD. ARMS2 also interacts with further extracellular proteins, several of which have been implicated in macular dystrophies. Although ARMS2 apparently lacks any classic targeting sequence, it is translocated to the endoplasmic reticulum in cultured cells before secretion. ARMS2 is mostly confined to choroid pillars in human eyes, representing a part of extracellular matrix and corresponding to the principal sites of drusen formation.

Conclusions.: The pivotal role of the extracellular matrix in the progression of AMD is underlined by the abnormal deposition of extracellular debris in the macula, observed frequently in affected individuals. The results have shown that ARMS2 may be necessary for proper matrix function.

Age-related macular degeneration (AMD) is the leading cause of blindness in the elderly, affecting more than 50 million people worldwide. ¹ The disease primarily impairs the central part of the human retina, called the macula, thereby causing a stepwise loss of sharp vision. ²  

The disease is typically manifested by the accumulation of lipoproteinaceous deposits (diffuse basal deposits and focal drusen) between the monolayer of the retinal pigment epithelium (RPE) and the underlying Bruch's membrane. In later stages, the lesion also extends to the photoreceptors, and in the most severe wet form, it combines with choroidal neovascularization and causes retinal detachment. 

AMD is typically a complex, multifactorial disease. Age and genetic predisposition are the primary risk factors. In regard to genetic susceptibility, linkage analyses identified two major risk alleles implicated in AMD; a common variant of the complement factor H (CFH) ³ and a frequent polymorphism at chromosome 10, region q26. This latter region harbors three adjacent genes: PLEKHA1 (also called TAPP1), ARMS2 (Age-related maculopathy susceptibility 2, also called LOC387715) and Htra1 (also called PRSS11). In their recently published work, Fritsche et al. ⁴ revealed a deletion/insertion in the ARMS2 gene that occurs only in the established risk haplotype. Consequently, ARMS2 is not synthesized in individuals homozygous for the indel variant and may therefore represent the sought after risk gene within this locus. Furthermore, the identified SNP in ARMS2 which changes alanine to serine at codon 69 (A69S) can be considered as a surrogate marker for the downstream deletion/insertion. The ARMS2 gene consists of only two exons, which encode a poorly characterized 107 amino acid protein with no similarities to known protein motifs. 

Individuals being homozygous for the risk alleles of both CFH and ARMS2 have a ∼50-fold increased risk for AMD. Consequently, most AMD cases are attributable to the presence of the risk variant(s) of CFH and/or ARMS2 genes. Beside these two major susceptibility genes, several further gene variants have been implicated in the susceptibility to AMD. Although some are present only in certain Mendelian AMD cases, others confer a considerable population risk. These include additional members of the complement system ^5–10 and constituents/regulators of the extracellular matrix (ECM) such as elastin-associated protein fibulin-6 (hemicentin-1). ¹¹  

In this work, we sought to investigate the biological function of ARMS2. We selected this gene first because it exhibits the highest association with AMD, even higher than CFH, ¹² and second, since only speculations are available about its function. Yeast two-hybrid screening along with immunohistochemical experiments indicate for the first time a role for ARMS2 as a matrix protein and put it into a plausible biological context. The data obtained from these experiments enables us to propose a protein interaction map, which links ARMS2 to proteins already implicated in AMD and other macular dystrophies. 

Materials and Methods

Phylogenetic Analysis

C termini of human fibulin-1 to -7 containing the last two EGF-like domains were aligned (AlignX, Vector NTI suite 10; Invitrogen, Carlsbad, CA). Phylogenetic analysis was performed with the PHYLIP package version 3.67 ¹³ (http://evolution.genetics.washington.edu/phylip.html, provided in the public domain by the Department of Genome, University of Washington, Seattle, WA). The phylogenetic tree was generated by using the distance matrix method. The resulting tree was statistically analyzed by bootstrapping with 100 random data sets. The alcohol dehydrogenase sequence was used as the outgroup. 

Plasmid Constructs, Transfection

The full-length ARMS2 cDNA was amplified from human placental total RNA by RT-PCR and cloned into the pCMV-MCS vector (Stratagene, La Jolla, CA) enabling high-level eukaryotic expression of the untagged protein. The risk variant of ARMS2 was generated by in vitro mutagenesis (Stratagene), according to the manufacturer's recommendations. Hemagglutinin (HA)-epitope tagged fibulin-6 was generated by PCR amplification of the corresponding prey plasmid exploiting the HA-tag sequence present in the prey vector pACT2 (Clontech, Palo Alto, CA) and cloned into pcDNA3 vector (Invitrogen). ARMS2 was subcloned into pDEST/N-SF-TAP which contains an N-terminal tandem STREPII and a FLAG tag ¹⁴ (Gateway recombination system; Invitrogen). 

For transfection, the cells were cultured under standard conditions and transfected at ∼80% confluence (Effectene; Qiagen, Hilden, Germany). To confirm the secretion of ARMS2, HEK293 cells were transiently transfected with plasmid constructs coding for normal or risk variant ARMS2. Culture media were replaced with serum-free medium 24 hours after transfection and collected directly after the cells were incubated for a further 24 hours. Thereafter, the adherent cell layers were washed with 5 mM EDTA in PBS for 5 minutes, to promote the solubilization of extracellularly attached proteins. After the EDTA-containing supernatants were collected, the cells were lysed in lysis buffer (50 mM Tris-HCl [pH 7.4]; 250 mM NaCl; 25 mM EDTA; 1% NP-40; 10% glycerol; protease inhibitor cocktail; Roche, Mannheim, Germany). Proteins from extracellular fractions were precipitated in ice-cold acetone and the precipitates were dissolved in lysis buffer. Equal total protein amounts (2 μg) of all extracellular and cellular fractions were separated by SDS-PAGE. Actin, TIM23, and ARMS2 were detected by Western blot analysis, with mouse monoclonal anti-actin (Sigma, St. Louis, MO), anti-TIM23 (BD Biosciences, Heidelberg, Germany) and rat monoclonal anti-ARMS2 antibodies (6B5), respectively. Western blot signals were scanned with a transmission scanner (GS-710 Calibrated Imaging Densitometer; Bio-Rad Hercules, CA) and quantified (QuantityOne V.4.2 software; Bio-Rad). 

Copurification

For the copurification experiments, HEK293 cells were transfected with plasmids coding for N-terminal SF-TAP tagged ARMS2 ¹⁴ and the HA tagged fibulin-6 fragment. For the control experiments, the ARMS2-coding plasmid was replaced with the corresponding empty vector. The cells (along with their surrounding ECM) were lysed 48 hours after transfection in 500 μL lysis-buffer, containing 30 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1% NP-40, protease inhibitors (Roche), and phosphatase inhibitor cocktail II (Sigma). Equal amounts of each lysate (3.5 mg) were incubated with 10 μL settled resin (Strep-Tactin Superflow; IBA, Göttingen, Germany) for 3 hours at 4°C in spin columns (GE Healthcare, Piscataway, NJ). The resin was washed three times with 500 μL lysis buffer and finally with 500 μL TBS buffer (30 mM Tris-HCl [pH 7.4], 150 mM NaCl). After elution in 200 μL desthiobiotin elution buffer (IBA) the eluates were neutralized and concentrated to a volume of 30 μL by using centrifugal filter units (Microcon YM-3; Millipore, Billerica, MA), and separated by SDS-PAGE using 10% to 20% Tris glycine gradient gels (NuPage; Invitrogen). The gels were blotted onto PVDF membranes, and protein–protein interactions were revealed by Western blot analysis using anti-ARMS2 (6B5) and anti-HA (C12A5; Roche) antibodies. 

For the inverse copurification experiments, HEK293 cells were transfected with plasmids coding for untagged ARMS2 and HA-tagged fibulin-6 fragment, respectively. For control experiments, the fibulin-6 plasmid was replaced by the corresponding empty vector. Cell lysates were prepared just described. Equal amounts of each lysate (∼3.5 mg) were incubated with 20 μL settled anti-HA agarose resin (Roche) for 3 hours at 4°C in spin columns (GE Healthcare). The resin was washed with 500 μL lysis buffer, twice with 500 μL washing buffer (30 mM Tris-HCl [pH 7.4], 150 mM NaCl, 0.5% NP-40), and finally with 500 μL TBS buffer (30mM Tris-HCl [pH 7.4], and 150 mM NaCl). After elution in 100 μL 200 mM glycine [pH 2.0], the eluates were neutralized. The eluates were subsequently analyzed by Western blot analysis as described herein. 

Yeast Two-Hybrid Screen

A GAL4-based yeast two-hybrid system (Matchmaker; Clontech) was used to screen a human oligo-dT primed placental cDNA library in pretransformed PJ69–4α yeast cells for proteins interacting with ARMS2. Full-length ARMS2 was fused to the DNA binding domain and transformed into PJ69–4A yeast cells. Before screening, the bait plasmid was assayed for autoactivation. No growth was observed in yeast cotransformed with ARMS2 and empty prey plasmid, even when plated onto less stringent triple-dropout media and incubated for 2 weeks. Positive clones were obtained by cell-to-cell mating with an efficiency of >5%, resulting in a total number of ∼2 × 10⁷ diploid cells on selection plates lacking tryptophan, leucine, histidine, and adenine. Isolated prey plasmids, along with either the original ARMS2 bait construct or with that containing the risk variant of the gene, were back-transformed into PJ69–4A cells and tested for growth. Cotransformed yeasts capable of growth on selective media were considered as positives. 

Generation of anti-ARMS2

Rat monoclonal antibodies to ARMS2 were originally raised against two peptides corresponding to the predicted amino acid sequence (NP_001093137; http://www.ncbi.nlm.nih.gov/protein/). The antibodies used in this work are against the peptide LRESVLDPGVGGEGASDKQRSKLS (37-61) located in the first exon of ARMS2. This peptide was conjugated to ovalbumin and keyhole limpet hemocyanin through an N-terminal cysteine residue and used for immunization. Although this manuscript was under review, a new set of monoclonal antibodies were obtained from a second round of immunization in which the entire ARMS2 protein was used instead of a selected peptide. It is important to note that the same results were achieved when experiments were repeated with these new antibodies. The specificity of all antibodies was tested thoroughly with both positive and negative controls (for details see Fig. 1 and Supplementary Fig. S1). 

Immunocytochemistry

ARPE-19 cells were grown on glass coverslips. Selected cultures were treated with 20 μM nocodazole for 1 hour to propagate the dispersal of mitochondrial clusters. Mitochondria were visualized by using a red mitochondrial stain (MitoTracker Deep Red 633; Molecular Probes, Eugene, OR). After fixation with 4% paraformaldehyde for 5 minutes, the coverslips were washed twice with PBS for 5 minutes, and permeabilized with 1% Triton X-100 in PBS for 10 minutes. The cells were washed with 0.1% Tween 20 in PBS twice for 3 minutes, and unspecific binding was blocked by 5% normal goat serum (NGS) in the same solution for 1 hour. Primary antibodies were diluted in PBS containing 1% NGS and 0.1% Tween 20 and incubated for 1 hour. After intense washing, the cells were incubated with secondary antibody and DAPI for 1 hour. Unbound antibodies were removed by six washing steps of 0.1% Tween 20 in PBS, and the coverslips were mounted in antifade medium (FluorSave; Calbiochem, San Diego, CA). The primary antibodies used were rat anti-ARMS2 (6B5, 1:100 dilution) and mouse anti-calnexin (1:500 dilution; Abcam, Cambridge, MA). The secondary antibodies used were Alexa Fluor 488-conjugated goat anti-rat IgG (1:2000, Molecular Probes), and Alexa Fluor 568-conjugated and cross-adsorbed goat anti-mouse IgG (1:2000; Molecular Probes). Data visualization and acquisition were performed with either a fluorescence (Axioscope equipped with an AxioCam HRc camera; Carl Zeiss Meditec, Oberkochen, Germany) or a laser scanning confocal microscope (LSM 510; Carl Zeiss Meditec). 

Histology and Immunostaining of Donor Eyes

Sixteen human donor eyes (five female, seven male; mean age, 67 years) were used for evaluation of ARMS2, fibulin-1, and fibulin-6 expression patterns in normal eyes. Sampling was approved by the local ethics committee in compliance with the tenets of the Declaration of Helsinki. All donors gave their informed consent. Donor eyes were fixed 0 to 60 hours postmortem in 6% buffered formalin and embedded in paraffin (Microm, Walldorf, Germany). Porcine and equine eyes were processed as described. ^15,16 Antigen retrieval was performed at 99°C for 15 minutes in citrate buffer pH 6.0 (DAKO, Glostrup, Denmark). ARMS2 was detected by using anti-ARMS2 antibodies (5C6, 1:5 dilution) and biotinylated anti-rat secondary antibodies (Linaris, Wertheim-Bettingen, Germany). To localize fibulin proteins anti-fibulin-1 (B-5, 1:10 dilution, Santa Cruz Biotechnology, Santa Cruz, CA) and anti-HMCN-1 (1:10 dilution; Proteintech Group, Chicago, IL) were used. Negative controls comprised nonreactive hybridoma supernatant at a similar dilution; omission of the primary antibody; peptide blocking (for ARMS2); and using porcine and equine eyes, where ARMS2 is not expressed (for ARMS2). Sections were stained (Vectastain Elite ABC:HRP kit; Linaris, coupled to Vector VIP; Linaris). Nuclei were counterstained with Mayer's hematoxylin (Merck, Darmstadt, Germany). 

Results

Localization of ARMS2 in Human Eyes

To investigate ARMS2 expression in situ, we generated a set of monoclonal antibodies (mAbs) and thoroughly screened them for selectivity and specificity against ARMS2 (Fig. 1 and Supplementary Fig. S1). Formalin-fixed, paraffin-embedded sections from human donor eyes (n = 16) were stained for ARMS2 with these mAbs. Strong immunostaining was observed in the regions around the capillaries of the choroid (Figs. 1A, 1B), corresponding to the intercapillary pillars. A gradient of ARMS2 was frequently observed, the highest concentration exhibited in the matrix adjacent to Bruch's membrane. Very faint and diffuse ARMS2 staining was also detected in RPE and retina in some donor eyes. Labeling was absent in control sections with no primary antibody or when the antibodies were preblocked with the corresponding peptide (Figs. 1E, 1F). Similarly, no staining was seen in the eyes of the pigs or horses, species that lack ARMS2 (Figs. 1G, 1H). In addition, fibulin-1 and -6, two interacting partners of ARMS2 identified in this study, were also found to be localized to the pillars and exhibited a similar staining pattern (Figs. 1C, 1D). 

Figure 1.

 

Expression of ARMS2 protein in human eyes. Donor eyes were immunostained with an mAb (5C6) for ARMS2, followed by biotin-peroxidase VIP detection. (A) A low-magnification imaging showing the enrichment of ARMS2 in the intercapillary pillars (arrowheads). ARMS2 labeling was also occasionally observed outside the capillary layer, extending toward Sattler's layer. High-magnification image of ARMS2 labeling in choroid pillars (B). Localization of fibulin-1 (C) and -6 (D) in the same region. Negative control experiments were performed by omitting the primary antibody (E), blocking the primary antibody with incubation of 24 μg/μL of the corresponding peptide for 30 minutes (F), or using sections from species in which ARMS2 is absent (G, pig; H, horse). Sections were treated the same way as those used to detect ARMS2. The lumens of capillaries are marked () in high-magnification images. INL, inner nuclear layer; OPL outer plexiform layer; ONL, outer nuclear layer; OS, outer segments; RPE, retinal pigmented epithelium, BM, Bruch's membrane; CC: choriocapillaris. Scale bar, 20 μm.

Localization of ARMS2 to the Endoplasmic Reticulum in Cultured Cells

Since most ARMS2 was found to be localized to the intercapillary pillars in human eyes, representing a part of the ECM, we sought to examine the localization of ARMS2 at a cellular level using in vitro methods. As none of the cell lines tested so far exhibited sufficient endogenous expression for analysis, cultured ARPE-19 cells were transfected with plasmids coding for either the normal or the risk variant (A69S) full-length, untagged, ARMS2. Immunocytochemical analyses of the cells using anti-ARMS2 antibodies revealed very strong colocalization with the endoplasmic reticulum (ER; Fig. 2A). Colocalization with mitochondria was observed only in areas where staining for ER and mitochondria overlapped. Treating cells with nocodazole, an inhibitor of microtubule polymerization, resulted in the dispersion of mitochondrial clusters, whereas ER and ARMS2 staining remained unaffected (Fig. 2B). No differences in localization were found between the normal and risk variant form of ARMS2 (Supplementary Fig. S2). Nontransfected cells exhibited a much weaker staining but with the same distribution (data not shown), corresponding to low-level intrinsic expression. When a small tag was fused to the ARMS2 N-terminal, the localization remained unaffected, whereas adding the same tag to the C-terminal end of the protein abolished ER colocalization (data not shown). This indicates that the integrity of the C terminus is required for proper targeting. To exclude the possibility that this staining is specific to ARPE-19 cells, the same staining was replicated in several other cell lines (HeLa, HEK293, NIH-3T3, MDCK, and RFL-6, data not shown). 

Figure 2.

 

Confocal images of ARPE-19 cells transfected with plasmids coding for the untagged, nonrisk variant ARMS2. For each panel, a single cell was photographed, using different fluorochromes and was subsequently pseudocolored. (A) To assess the colocalization of ARMS2 with mitochondria and the ER, the same field was photographed using FITC, Cy3, and Cy5 filter sets for ARMS2, ER (calnexin), and mitochondria, respectively. Each row represents two from these three images pseudocolored in green (left), or in red (middle), and the merged image of these two (right). Cell nuclei were stained with Hoechst 33258 (blue, shown only in merged images). Note the large extent of overlap between staining for ARMS2 and ER (top), compared with that between staining for ARMS2 and mitochondria (middle). In addition, there is a slight overlap between the ER and mitochondria (bottom), which explains the observed small overlap of ARMS2 and mitochondria. (B) In nocodazole treated cells only a faint, blurred mitochondrial staining can be observed, while the colocalization between ARMS2 and ER remains unaffected.

Accumulation of ARMS2 in the Extracellular Space

To confirm the secretion of ARMS2, HEK293 cells were transiently transfected with plasmid constructs coding for either the normal or the risk variant (A69S) ARMS2. Cell culture medium, ECM, and cell lysates were screened for the presence of the ARMS2 protein. The abundance of ARMS2 was compared with that of actin in all fractions (Fig. 3). Most of both normal and risk variant ARMS2 was detected in the extracellular fraction (for risk variant, see Supplementary Fig. S3). Treating cells with EDTA led to further release of matrix-bound ARMS2. Although this treatment also resulted in a more intense actin signal, as a consequence of cell damage, the increase observed in the signal intensity of ARMS2 can be mainly attributed to the existence of an EDTA-sensitive ARMS2 pool. It has been reported recently that ARMS2 is localized in mitochondria. Therefore, we also probed the fractions from ARMS2-expressing cells with anti-TIM23 antibody, a marker for mitochondrial inner membranes. The presence of TIM23 exactly replicated the distribution of intracellular actin among the different pools (intracellular, extracellular, and extracellular with EDTA). Consequently, these results do not support a colocalization of mitochondria and ARMS2, but demonstrate the secretion of ARMS2 (Fig. 3). 

Figure 3.

 

ARMS2 subjected to secretion. HEK293 cells were transiently transfected with plasmid constructs coding for normal variant ARMS2. The culture medium was replaced with serum-free medium 24 hours after transfection. The supernatants were collected after 24 hours' incubation (S). Extracellularly attached proteins were released from cells by EDTA (S_EDTA) treatment. Finally, the cells were lysed (L). Similar total protein amounts of all extra- and intracellular fractions were separated by SDS-PAGE, and ARMS2 was detected by Western blot analysis. The relative abundance of ARMS2 was compared with the relative abundance of actin in all fractions (±SEM from three independent experiments). The major proportion of ARMS2 was detected in the extracellular fraction using two independent anti-ARMS2 antibodies. (As both antibodies yielded the same results, only data obtained from the anti-peptide antibody are presented here). Washing the cells with EDTA revealed the existence of an EDTA-sensitive ARMS2 pool. Probing with antibodies against TIM23, a mitochondrial inner membrane protein, showed a distribution similar to that of intracellular actin, whereas a known secreted protein (FBLN1) was detected exclusively in the supernatant fraction. These latter controls provide further evidence of the genuine secretion of ARMS2. It is important to note that ARMS2 can be readily detected also in the cell lysate fraction when exposed for a longer period. Shorter exposure was selected here only for better quantification.

Interaction of ARMS2 with ECM Proteins

We chose to focus the present study on the interacting partners of ARMS2 and reasoned that a strategy aimed at identifying these proteins would be beneficial in understanding its function. For this purpose, we performed a yeast two-hybrid screening, with full-length ARMS2 corresponding to the A allele used as a bait. As ARMS2 expression is readily detectable only in connective tissue and placenta (NCBI UniGene, Hs.120359; UNIGENE (http://www.ncbi.nlm.nih.gov/UniGene; provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD) we selected a placental library to seek for interacting partners. The identified binding partners of ARMS2 and their subcellular localizations are summarized in Table 1. To control for the specificity of the yeast two-hybrid interaction, positive clones were retested by back-transformation into yeast (Supplementary Fig. S4). Most of the interacting partners were ECM proteins, many of which were described as being enriched in basement membranes. Among the interactors we identified fibulin-1 and -6. Fibulins, especially fibulin-6, are clearly associated with AMD, suggesting a functional, and consequently a pathophysiological, relationship between the two proteins. Furthermore, some of the binding partners of ARMS2 bind to each other or are linked to further matrix elements. These data, combined with previously published work, support the presence of an ARMS2 interactome (Fig. 4). Within this protein network seven proteins have been implicated in the development of macular dystrophies (circled proteins). These include ARMS2, ¹⁷ fibulin-3, ¹⁸ fibulin-5, ¹⁹ fibulin-6, ²⁰ TIMP (tissue inhibitor of metalloproteinase)-3, ²¹ MMP (matrix metalloproteinase)-9, ²² and elastin. ^23,24  

Table 1.

 

Directly Interacting Proteins of ARMS2 Found by Yeast Two-Hybrid Screening

Interacting Partner	Symbol	Accession	Localization
Collagen type I α1	COL1A1	NP_000079	ECM
Collagen type IV α2	COL4A2	NP_001837	BM
Elastin microfibril interface-located protein 2	EMILIN-2	NP_114437	CBM
Fibronectin 1	FN1	NP_997639	ECM
Fibulin-1 isoform D	FBLN1	NP_006477	BM
Fibulin-6 (ARMD1, Hemicentin-1)	FBLN6	NP_114141	BM
Pregnancy-specific β-1-glycoprotein 9	PSG9	NP_002775	SEC
Pregnancy-specific β-1-glycoprotein 11	PSG11	NP_002776	SEC
Williams-Beuren syndrome chromosome region 27 protein	WBSCR27	NP_689772	Unknown

 Note the high number of extracellular proteins. For the majority of the preys, several independent clones were isolated. BM, basement membrane; ECM, extracellular matrix; CBM, cochlear basilar membrane; SEC, secreted.

Figure 4.

 

ARMS2 interacts with ECM proteins. Proteins encoded by genes known to be related to macular degenerations are circled. Fibulins are delineated by the dashed box. Note that three of five fibulin genes shown have been implicated in maculopathies, and two of five directly interact with ARMS2. Reference(s) for interactions described earlier are indicated in small boxes. ^21,51–60 ARMS2, age-related macular susceptibility 2; COL, collagen; ELN, elastin; EMILIN, elastin microfibril interface-located protein; FBLN, fibulin; FN1, fibronectin 1; MMP9, matrix metalloproteinase-9; PSG, pregnancy-specific β-1 glycoprotein; TIMP3, tissue inhibitor of metalloproteinase 3; WBSCR, Williams Beuren syndrome chromosome region 27.

Sequence analysis of these prey proteins showed that the minimal fragment of the corresponding fibulin and collagen proteins necessary and sufficient to bind ARMS2 are located close to their C-terminal moiety (Fig. 5A). We also compared the similarity among the ARMS2-binding domains of fibulin-1 and -6 and the corresponding regions of the other fibulin proteins (Fig. 5B). According to these data, fibulin-1 shows the highest similarity to fibulin-2, but it is also closely related to fibulin-6. The known macular dystrophies in which fibulins are implicated do not exhibit clustering within the similarity tree. It has been speculated that the alanine-to-serine substitution creates a new putative phosphorylation site and breaks a predicted α-helix, ²⁵ therefore we also extended our analysis to this variant. No differences in interaction quality between the normal versus risk variant (A69S) of ARMS2 were apparent at the level of the targeted yeast two-hybrid experiments when a series of pair-wise interactions were tested using either ARMS2 variant and one of the prey plasmids (Supplementary Fig. S4). Specificity of binding among ARMS2 and fibulin-6 was also confirmed by co-immunoprecipitation of these proteins (Fig. 6). Furthermore, we have shown that both fibulin-1 and -6 localize to the intercapillary pillars of the choroid, the area exhibiting the strongest ARMS2 staining in human eyes (Figs. 1C, 1D, respectively). 

Figure 5.

 

(A) Predicted domain structure (SMART; http://smart.embl-heidelberg.de) according to UniProt database (http://www.expasy.org) for ARMS2 interacting proteins Fibulin-6 (top, HMCN1_HUMAN) and fibulin-1 (bottom, FBLN1_HUMAN). ARMS2 interacts at the C-terminal ends of both fibulins. Horizontal bars: the alignment of the respective shortest prey clone to the full-length fibulin sequences with the position of the starting amino acids indicated (5498 and 578 for fibulin-6 and -1, respectively). Arrow: location of the previously identified mutation in fibulin-6 causing familial AMD. The corresponding AA substitution is also shown. VWFA, von Willebrand factor, type A; Ig-like C2-type, immunoglobulin subtype 2; TSP type-1, thrombospondin, type 1; nidogen G2, G2 nidogen and fibulin G2F; EGF-like, EGF-like calcium-binding; ANATO, anaphylatoxin). (B) Phylogenetic tree of human fibulins (FBLN1-FBLN7). The branch lengths and numbers correspond to the bootstrapping values. Alcohol dehydrogenase was used as the outgroup.

Figure 6.

 

ARMS2 interacts specifically with fibulin-6. (A) For copurification of ARMS2 and fibulin-6, HEK293T cells were transfected with HA-fibulin-6 and either SF-ARMS2 (lanes 1, 3) or the empty expression vector (lanes 2, 4). From total cell lysates including ECM (input, lanes 1, 2), ARMS2 was Strep/Flag purified (P, lanes 3, 4), and Western blots were probed with anti-ARMS2 (6B5) antibody (top) or with anti-HA antibody (bottom). (B) For the reverse copurification, HEK293T cells were transfected with untagged ARMS2 and either HA-fibulin-6 (lanes 1, 3) or the empty pcDNA3 expression vector (lanes 2, 4). From total cell lysates including ECM (input, lanes 1, 2) fibulin-6 was HA-purified (P, lanes 3, 4) and Western blot probed with anti-HA antibody (top) or anti-ARMS2 (6B5) antibody (bottom).

Discussion

AMD is a severe burden, on both the individuals with the disorder and the global health economy. The total annual loss in gross domestic product due to patients with AMD in the United States has been estimated to be approximately $25 billion. ²⁶ Hence, the development of novel therapeutic approaches is urgently needed. Understanding the genuine molecular mechanisms underlying the critical steps in the pathogenesis of macular dystrophies is an essential prerequisite to effective intervention. 

Comparing all known genetic factors, the risk variant of the chromosomal region 10q26 shows the highest association with AMD in various ethnic groups. This region encompasses three genes: PLEKHA1, ARMS2, and Htra1. However, the linkage disequilibrium (LD) among these genes hampers assessing their individual impact on AMD in SNP association studies. Nevertheless, together with an SNP in the promoter region of Htra1 (rs11200638), a coding SNP in the first exon of the ARMS2 gene (A69S, rs10490924) showed the strongest association with AMD. Coding for a secreted protease, Htra1 seemed a very attractive candidate gene a causative factor in AMD. This expectation was apparently affirmed, when DeWan et al. ²⁷ found a link between the Htra1 promoter polymorphism and the wet form of AMD. Recent reports however, claim that this polymorphism has no significant effect on Htra1 expression, ²⁵ whereas a variant of ARMS2 was found to be causative of AMD. ⁴  

Our data corroborate the importance of ARMS2 for developing AMD, because we can functionally link ARMS2 to a network of interacting extracellular proteins, in which many members are known to be involved in macular dystrophies. Most prominently, we found a direct binding between ARMS2 and fibulin-6. The data suggest that fibulin-6, implicated in rare familial AMDs, ^20,28,29 is functionally linked to ARMS2, which exhibits the highest association with the common form of AMD. Fibulin-6 mutations should therefore receive more attention in future studies, even if fibulin-6 is apparently not directly associated with the far more frequent sporadic cases. ³⁰ Of interest, the role of further fibulin proteins in the pathomechanism of macular dystrophies has been repeatedly demonstrated. ^18,19 Fibulins have overlapping binding sites for several basement membrane proteins, like tropoelastin, fibrillin, fibronectin, and proteoglycans, which are also important constituents of the elastic layers of Bruch's membrane. ^31,32 They are believed to participate in the assembly and stabilization of ECM structures and to regulate vasculo- and fibrogenesis ^33,34 and also inhibit angiogenesis. ³⁵  

It is tempting to speculate about a functional link between compromised elastic fibers and the development of AMD because (1) the ARMS2 interacting partners fibulins and EMILIN-2 (elastin microfibril interface located protein), are primarily known for their association with the elastic network, (2) the concentration of serum elastin-derived peptides are significantly higher in patients with AMD than in control subjects, ²⁴ (3) the loss of elasticity is the key hallmark of aging, and (4) age is the primary risk factor for AMD. In addition, smoking, which is the most studied modifiable environmental risk factor for AMD, was also found to provoke anti-elastin autoimmunity. ³⁶  

Furthermore, all the other interacting partners of ARMS2 found in the present study were also extracellular proteins. Hence, our work challenges recent results suggesting a mitochondrial function of ARMS2. ^4,25 Besides, the mitochondrial localization of ARMS2 has already been brought into question by Wang et al. ³⁷ They found that ARMS2 is mainly distributed in the cytosol, not in mitochondria. Unfortunately, this work was limited to intracellular organelles, therefore incapable of revealing ARMS2 secretion. 

Kanda et al. ²⁵ demonstrated the colocalization of tagged ARMS2 with mitochondria using immunocytochemical and biochemical methods after overexpression in cultured cells. However, the resolution of the immunocytochemical methods used does not allow precise discrimination between different organelles and it has been demonstrated that even a small tag can alter the localization of a protein. Our experiments showed that tagging ARMS2 N-terminally did not affect its localization, whereas adding the same tag C-terminally resulted in a diffuse cytoplasmic localization (data not shown). Using high-resolution confocal microscopy with three fluorochromes we were able to simultaneously detect ER, mitochondria, and ARMS2 in an in vitro system similar to that used by Kanda et al. These images unambiguously revealed that ARMS2 is colocalized with ER and not with mitochondria. This supports the theory emerging from our data, which presents ARMS2 as a secreted protein. ER localization of ARMS2 may thus serve as a prerequisite for subsequent secretion. Most proteins destined for secretion are translocated to the ER co- or posttranslationally. ³⁸ Nevertheless, some proteins are secreted in a nonclassic way independent from the ER-Golgi network. ^39,40 In spite of the clear accumulation of ARMS2 in ER, it apparently lacks any classic targeting signal sequence. ⁴¹ We hypothesize that ARMS2 translocates into the ER posttranslationally and uses the classic secretory route after passage. 

The in vivo colocalization of ARMS2 with mitochondria in photoreceptors was reported by Fritsche et al. ⁴ The results of our similar experiments also contradicted this finding. In contrast to Fritsche et al., we used a nonfluorescent immunolabeling method to circumvent the difficulties originating from the intense autofluorescence of retinal/choroidal structures (e.g., lipofuscin granules, drusen, and capillary walls). This approach enabled the extension of the analysis to the choroid, and revealed ARMS2 labeling in intercapillary pillars far more robust than that of the retina. These anatomic structures are thought to play a pivotal role in the pathogenesis of AMD. Of importance is the nonrandom distribution exhibited by autofluorescing drusen, ⁴² found to be located internal to the pillars, but not over the surface covering the vessel lumen. Hence, the strongest ARMS2 staining corresponds to the principal sites of drusen depositions. Because patients homozygous for the risk-haplotype do not express ARMS2, ⁴ this observation raises the possibility that the protein confers “local” protection against drusen formation. Accordingly, these patients (i.e., lacking ARMS2 expression) would exhibit higher susceptibility to the development of drusen and AMD. Our data also showed that fibulin-1 and -6 are enriched in pillars, akin to fibulin-5 described earlier. ³²  

ARMS2 also binds COL1A1, COL4A2, and fibronectin. COL1A1 is a constituent of type I collagen, a chief component of the collagenous layer in Bruch's membrane. ⁴³ Furthermore, type I collagen exhibits a potent angiogenic activity and is capable of inducing the expression of further angiogenic genes. ^44,45 Type IV collagen is a major component of the basement membrane of the choriocapillaris, but not that of the RP where immunolabeling has been found to be either weak or absent. ⁴⁶ This finding is in accordance with our results showing that ARMS2 is mainly localized in the intercapillary area of the choroid in human eyes, making the interaction between ARMS2 and COL4A2 conceivable. Mice carrying a mutant COL4A2 gene exhibit pleiotropic defects including those in eyes. ⁴⁷ Fibronectin fibrils cover the cell surface and are indispensable for the formation of blood vessels. Similar to type I collagen, its expression increases with age. ⁴⁴ These three genes are expressed in tissues throughout the body, and to our knowledge their mutations are not directly related to retinal dystrophies. 

Finally, ARMS2 interacts with pregnancy-specific glycoproteins. These proteins are the most abundant fetal proteins in the maternal bloodstream during late pregnancy and function as immunomodulators, essential for a successful pregnancy. ^48,49 Further experiments should be performed to clarify the physiological significance of this interaction. 

Taken together, our results suggest that ARMS2 is a constituent of the ECM. The central finding of our study is that ARMS2 directly interacts with ECM proteins, such as fibulin-6. This finding links a gene with high risk for common AMD to a rare familial AMD gene. Protein expression and interaction studies provided evidence that ARMS2 is a secreted protein that is processed through the ER. After secretion, ARMS2 is bound to diverse ECM proteins. Our work also provides further evidence that the nonsynonymous SNP (rs10490924) has no marked physiological influence on protein function. Accordingly, the indel variant occurring in the same haplotype as rs10490924 and the concomitant loss of ARMS2 synthesis may play a pivotal role in the pathomechanism of AMD. The localization of ARMS2 suggests that this protein is protective against drusen formation. This work sheds new light on the physiological function of ARMS2 and may contribute to our understanding of the pathomechanism of this sight-threatening disorder. 

Supplementary Materials

Supplementary Figure S1 - (PDF) 

Supplementary Figure S2 - (PDF) 

Supplementary Figure S3 - (PDF) 

Supplementary Figure S4 - (PDF) 

Supplementary Figure S5 - (PDF) 

Footnotes

 Supported by RETNET Grant MRTN-CT-2003-504003 (MU); EVI-GENORET (Functional Genomics of the Retina in Health and Disease) Grant LSHG-CT-2005-512036 (MU); Interaction Proteome Grant LSHG-CT-2003-505520 (MU); and Deutsche Forschungsgemeinschaft Grant SFB 571 A5 (CD).

Footnotes

 Disclosure: E. Kortvely, None; S.M. Hauck, None; G. Duetsch, None; C.J. Gloeckner, None; E. Kremmer, None; C.S. Alge-Priglinger, None; C.A. Deeg, None; M. Ueffing, None

The authors thank Stephanie Schoeffmann for excellent technical assistance, Nora Heisterkamp for providing the human placental yeast library, Ronald Roepman's group for help with the yeast two-hybrid technique, Peter Hutzler for help with confocal microscopy, Naomi Chadderton for critical reading of the manuscript, and Siegfried Priglinger for providing human donor eyes.