Age-related macular degeneration (AMD) is the primary cause of central vision loss in developed countries. Genetic studies have repeatedly shown a strong association of a locus at chromosome 10q26 with the risk of developing both dry and wet AMD. This locus represents one of the two most significant genetic factors being identified in AMD. ^1–6 Three genes are located within the bounds of this locus, pleckstrin homology domain containing family A member 1 (PLEKHA1), age-related maculopathy susceptibility 2 (ARMS2, previously known as LOC387715), and HtrA serine peptidase 1 (HTRA1), all of which are associated with AMD. The most significantly associated haplotype includes single nucleotide polymorphisms (SNP) rs10490924 (nonsynonymous change A69S) in ARMS2 and rs11200638 in the promoter region of HTRA1. Due to a strong linkage disequilibrium (LD) in the region, it is difficult to identify with certainty the susceptibility variant(s) and related gene(s) by genetic statistical analysis alone. Analyzing and verifying biological consequences of the associated variants will help determine susceptibility gene(s) for AMD in this locus. 

A combinative insertion/deletion polymorphism (indel; consisting of a 443-bp deletion and an adjacent 54-bp insertion) has been identified in ARMS2 3′UTR and flanking region (Fig. 1). ⁷ The association of indel with AMD is equal to that of ARMS2 A69S and rs11200638 since these variants are in strong LD. ^7–10 The indel contains two AUUUA motifs that may mediate rapid mRNA turnover. ¹¹ Initially, the indel was correlated with a lower level of exogenous ARMS2 transcripts in cultured cells and endogenous ARMS2 protein in placentas, where ARMS2 is highly expressed. ⁷ A deficient expression of ARMS2 caused by the indel was thus suggested to functionally contribute to the onset of AMD. Results of subsequent attempts to replicate this effect of the indel on the stability of ARMS2 transcripts have been mixed. Some studies verified the effect of the indel ^12,13 while others found that genotypes at the indel do not change the level of ARMS2 transcripts in human retinas. ^10,14 It is still under debate whether the indel could really accelerate the degradation of ARMS2 transcripts. 

A nonsense change R38X in ARMS2 was reported to be inversely associated with AMD (Fig. 1). ¹² This relatively weak and inverse association has not been validated in large datasets. ^15–17 R38X introduces a premature stop codon and could potentially lead to lower ARMS2 expression due to nonsense-mediated mRNA decay. ¹⁸ Published data show that R38X is associated with a lower level of ARMS2 expression. ¹² A possible role of the decreased level of ARMS2 transcripts in AMD etiology is thus put in a paradoxical position. If both the indel and R38X, as reported, decrease the level of ARMS2 transcripts, it would be impossible to explain how the indel confers the risk of AMD, yet R38X is protective against the development of the disease. In an attempt to verify the reported effects of these variants and more importantly to exclude nonfunctional variants in this locus for future studies to focus on, we examined ARMS2 transcripts in human eye tissues and by in vitro assays. 

A total of 83 human retina–RPE–choroid samples from 83 unrelated Caucasian subjects (age 74.8 ± 14.5 years) without any known eye diseases were provided by the Florida Lions Eye Bank. Procedures for recruitment, requests for medical records, and consent forms were approved by the University of Miami Miller School of Medicine Institutional Review Board. This research adhered to the tenets of the Declaration of Helsinki. Eye tissues were retrieved and frozen in −80°C within 24 hours postmortem. Tissue columns (including neuroretina, RPE, and choroid) were punched from the macular region of frozen eyes as previously described. ^17,19 RNA was extracted from retina–RPE–choroid samples by using the RNeasy lipid tissue kit according to the manufacturer's instructions (Qiagen, Hilden, Germany). A nanodrop 8000 photospectrometer was used to measure the yield of RNA extraction. An Agilent 2100 Bioanalyzer was applied to monitor the quality of RNA. Samples with RIN (RNA integrity number) ≥ 7 were chosen for this study. 

The SuperScript III First-Strand Synthesis System (Invitrogen, Carlsbad, CA) was used for reverse transcription (RT) according to the manufacturer's instructions. Briefly, each RT reaction consisted of 500 ng purified RNA, random hexamer or Oligo(dT)20 Primers, deoxyribonucleotide triphosphates (dNTP), 1× RT buffer, MgCl₂ (5 mM), dithiothreitol (50 mM), SuperScript reverse transcriptase, and RNase OUT. RT reactions were incubated at 25°C for 10 minutes, at 50°C for 50 minutes, and at 85°C for 5 minutes, followed by the addition of 1 μL RNase H and incubation at 37°C for 20 minutes. The PCR was applied by the following three pairs of primers: 

ARMS2-forward: 5′-TTTTTCAAATCCCTGGGTCTCT-3′ 

ARMS2-reverse: 5′-AGAGAAAGGAGGGCAAGAAAAC-3′ 

Luciferase-forward: 5′-AATGAACGTGCTGGACTCCT-3′ 

Luciferase-reverse: 5′-GACACTCTCAGCATGGACGA-3′ 

RPE65-forward: 5′-AGACAATTAAGCAGGTTGATCTTTG-3′ 

RPE65-reverse: 5′-AAAGACTCCATGAAGAAAGGAACTT-3′ 

GAPDH-forward: 5′-TTAGCACCCCTGGCCAAGG-3′ 

GAPDH-reverse: 5′-CTTACTCCTTGGAGGCCATG-3′ 

The PCR products were separated in 2% agarose gel. The gel image was captured using the AlphaImager gel documentation system (ProteinSimple, Santa Clara, CA). The density of PCR bands was captured for semiquantification analysis. 

Quantitative RT-PCR was performed using Bio-Rad CFX 384 Real-Time System. We conducted two Bio-Rad (Hercules, CA) assays including ARMS2 and GAPDH. Within the 384-well plate, each assay was duplicated and each sample was repeated four times at different locations on the plate. The 10.0-μL PCR reaction mix contained PrimePCR Assay (0.5 μL), SsoAdvanced SYBR Green Supermix (5.0 μL), cDNA (1 μL corresponding to the cDNA reverse transcribed from approximately 10 ng RNA), and nuclease-free water (4.5 μL). The 384-well plate was then run on the CFX 384 at 95°C for 30 seconds, then 95°C for 5 seconds and 60°C for 15 seconds (for 45 cycles). After the PCR run was complete, quantitative gene expression data were acquired and analyzed using the Bio-Rad CFX manager software. The experiment was repeated three times. We also applied a custom-designed assay to specifically quantify a newly identified ARMS2 transcript alternative splice isoform as described in detail previously. ²⁰  

ARMS2 gene including flanking regions was amplified by PCR using forward primer 5′-TTCAAATCCCTGGGTCTCTGC-3′ and reverse primer 5′-AATGCTCAGGGGCTCCTATT-3′. A human genomic DNA heterozygous at the indel variant was used as a template. The PCR products (two bands: 3.5 kbp for wild-type [WT] allele and 3.1 kbp for the indel allele) were separated in 1% agarose gel and extracted, purified, and cloned, respectively, into a pcDNA3.1 plasmid vector (Invitrogen). Finally, the pcDNA3.1-ARMS2-R38X construct was generated using the Quickchange XL site-directed mutagenesis kit (Stratagene, La Jolla, CA) by applying forward mutagenic primer (5′-CCTTCATTTCCACTCTGTGAGAGTCTGTGCTAGACCC-3′) and a complementary reverse mutagenic primer (5′-GGGTCTAGCACAGACTCTCACAGAGTGGAAATGAAGG-3′) with pcDNA3.1-ARMS2-WT as a template. All the resulting constructs, including pcDNA3.1-ARMS2-WT, pcDNA3.1-ARMS2-indel, and pcDNA3.1-ARMS2-R38X, were verified by sequencing. 

Mouse embryonic fibroblasts (MEFs) were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (FBS) at 37°C in an incubator supplemented with 5% CO₂. Cells were seeded at 3 to 5 × 10⁵ cells per well in six-well plates. Sixteen hours after the plating, cells were cotransfected with one of the ARMS2 minigene vectors (2.5 μg/well) and psiCHECK2 vector (0.25 μg/well) using Lipofectamin LTX (Invitrogen) according to the manufacturer's instructions. Each minigene vector group contained three wells, and the experiment was repeated at least three times. After transfection for 48 hours, cells were harvested for RNA extraction using RNeasy kits (Qiagen). To evaluate the turnover of ARMS2 transcripts, cells were treated with actinomycin D (10 μM) for different periods after being transfected overnight (∼16 hours). RT-PCR and qRT-PCR were applied to measure the exogenous ARMS2 transcripts. RT-PCR products were also sequenced by capillary sequencing to verify their specificity. 

The 3′UTR and flanking region ARMS2 containing WT or the indel allele were amplified with forward primer 5′-CGGCTCGAGTGTCACTGCATTCCCTCCTGTCAT-3′ and reverse primer 5′-GCGGCCGCAAGCTTCTTACCCTGACTTCCA-3′ from a heterozygous human genomic DNA sample. The PCR products (two bands: 615 bp for WT allele and 226 bp for the indel allele) were separated in 1% agarose gel and extracted, purified, and cloned into psiCHECK2 (Promega, Madison, WI) at restriction sites Xho1 and Not1. The inserts containing either the WT allele or the indel allele at the 3′UTR of renilla luciferase gene in the construct were confirmed by sequencing. Human embryonic kidney–293 (HEK-293) and MEFs were maintained in DMEM with 10% FBS at 37°C in an incubator supplemented with 5% CO₂. Cells were seeded at 1 to 2 × 10⁵ cells per well in 24-well plates. Sixteen hours after the plating, psiCHECK2 vectors (50 ng/well) were transfected into cells by Lipofectamin LTX according to manufacturer's suggestions. Each vector group contained six wells, and the experiment was repeated at least three times. After transfection for 24 hours, cells were harvested by the addition of 100 μL passive lysis buffer. Renilla luciferase activities in cell lysate were measured using Dual Luciferase assays (Promega) by a TD-20/20 luminometer (Promega) and were normalized with the firefly luciferase activities. 

All data were normalized by inner controls, such as Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and firefly luciferase activities. Data are presented as mean ± SD. Statistically significant changes among treatments or genotypes were assessed by Student's t-tests at α = 0.05. 

ARMS2 gene is not annotated in the mouse genome, and ARMS2 transcripts are not detectable in mouse eye or other tissues. We therefore chose MEFs to examine the effects of variants on exogenous ARMS2 transcripts while avoiding potential interference of endogenous ARMS2 in human cells. MEFs were transfected with three minigene plasmid vectors including pcDNA3.1-ARMS2-WT, pcDNA3.1-ARMS2-indel, and pcDNA3.1-ARMS2-R38X. The differences among these vectors were only at the variant R38X or the indel. Because a haplotype with minor alleles at both R38X and the indel was not identified in a large dataset (1169 AMD cases and 707 controls) in our previous studies, a minigene carrying combinative R38X and the indel was not included in this experiment. ¹⁹ Recently, the ARMS2 gene transcription start site was found 220 bp upstream from the annotation. A novel alternative splice isoform of ARMS2 (isoform B) was identified through characterization of ARMS2 transcription in human eyes by our group. ²⁰ We designed primers for amplifying ARMS2 transcripts of both isoforms. Polymerase chain reaction results displayed only one specific band in MEFs after transfection for 48 hours (Fig. 2A). The sequence of this band perfectly matched the annotated ARMS2 transcripts (isoform A) (data not shown), suggesting that there was no alternative splicing of exogenous ARMS2 in MEFs. As shown in Figure 2A, we first applied RT-PCR to evaluate ARMS2 transcripts in MEFs. Semiquantification of the PCR bands indicated that there was no significant difference between minigenes of WT and the indel by normalization with either housekeeping gene GAPDH or renilla luciferase transcribed from cotransfected psiCHECK2 vectors (Fig. 2B). In contrast, the level of ARMS2 transcripts was lower in the R38X group than in the WT group (P < 0.01). Quantitative RT-PCR further verified that R38X, not the indel, correlated with a much lower level of transcripts in MEFs (P < 0.01; Fig. 2C). 

To test whether R38X and the indel really could mediate transcript decay or decrease transcript stability, as previously suggested, ^7,12 we treated MEF cells with actinomycin D, which inhibits transcription. After transfection overnight (∼16 hours), MEFs were treated with actinomycin D (10 μM) for 0, 6, 12, and 24 hours, respectively. Since obvious cell death was observed after treatment with actinomycin D for 24 hours, this group was not included in further analysis. RT-PCR (Figs. 3A, 3B) did not show an obvious overall decrease in ARMS2 transcripts in any but the R38X group after actinomycin D treatment for 12 hours. Semiquantification of RT-PCR demonstrated that the level of ARMS2 transcripts was lower in the R38X group than in the WT group only after actinomycin D treatment for 12 hours (14.7% vs. 21.8% of nontreated transcript level, respectively; P < 0.05). In contrast, there was no significant difference in the level of ARMS2 transcripts between WT and the indel groups after actinomycin D treatment for either 6 or 12 hours. Results from the more sensitive qRT-PCR displayed obvious decreases in ARMS2 transcripts in the three groups (WT, R38X, and the indel) after treatment with actinomycin D. However, only the R38X group displayed lower ARMS2 transcripts compared to the WT group (P < 0.05) after treatment with actinomycin D for 12 hours. These results suggest that R38X, not the indel, may enhance the degradation of ARMS2 transcripts. 

Dual luciferase assay was applied to further examine whether the indel interferes with gene expression. In the psiCHECK2 vectors, the 3′UTR of renilla luciferase was replaced by the 3′UTR of ARMS2 carrying either the WT or the indel allele. Firefly luciferase, coexpressing with renilla luciferase from the same vectors, was used as an internal control. HEK-293 cells and MEFs were transfected with psiCHECK2 vectors. Results indicated no significant difference in renilla luciferase activity between WT and the indel alleles in either HEK-293 cells or MEFs (Fig. 4). The results further suggest that the indel at ARMS2 3′UTR may not interfere with gene expression. 

We then applied qRT-PCR to evaluate the two ARMS2 transcript splice variants as described previously. ²⁰ Genotypes at R38X were assessed by Taqman assays (Life Technologies, Carlsbad, CA). Genotypes at the ARMS2 3′UTR indel were evaluated by PCR and gel assay as described previously. ^7,10 No sample was identified as homozygous XX at R38X or homozygous indel at the indel variant. By RT-PCR, we detected a higher level of RPE65 transcripts (Fig. 5A), which verified the inclusion of RPE layer in the sample preparations. Results of qRT-PCR indicated that there was no significant difference in age among the six groups carrying different genotypes at variants R38X and the indel (data not shown). We did not find statistically significant differences (P > 0.05) in the level of both ARMS2 transcripts among groups carrying different genotypes at the two variants using the Student's t-test (Fig. 5B). 

Verifying biological consequences of the genetically associated variants is one way to determine susceptibility gene(s) for AMD at the chromosome 10q26 locus. Theoretically, both the indel and R38X could influence the level of ARMS2 transcripts. The indel at 3′UTR and flanking region presumably deletes the original polyA signal of ARMS2 and inserts sequences containing two ATTTA fragments. When these sequences are transcribed, the AUUUA motif located in the 3′UTR could mediate mRNA degradation. ¹¹ However, it also should be noted that the AUUUA motif alone might not be sufficient to enhance mRNA degradation. ²¹ By examining protein level in placentas and applying an in vitro transcription assay, Fritsche et al. first identified the indel and reported that it causes a lower expression of ARMS2. ⁷ Yang et al. observed a similar effect of the indel by examining ARMS2 mRNA in placentas. ¹² The presumed correlation between the indel and gene expression was further verified by cDNA capillary sequencing and imaging of exogenous ARMS2 protein in cultured cells. ¹³ However, there is a possibility that the regulation of ARMS2 expression is different in retinas compared to placentas. Because the sequence of ARMS2 protein has not yet been fully established, conclusions from antibody-based experiments need further supporting evidence. If the indel influences the stability of ARMS2 transcripts, the direct evidence should be derived from examination of transcripts. 

Our group previously found that the indel is more complicated and actually contains two side-by-side indels. The effect of the indel on the level of ARMS2 transcripts appears insignificant in retinas (n = 52). ¹⁰ Kanda et al. further reported that none of the AMD-associated variants in ARMS2 change the level of ARMS2 transcripts using qRT-PCR. ¹⁴ The inconsistency regarding the effect of the indel suggests that there is a need to reevaluate the biological consequences of these variants. In this study, we first observed no significant difference in exogenous ARMS2 transcript levels between WT and the indel. After transcription is blocked, the indel allele does not change the rate of ARMS2 transcript degradation compared to the WT allele. Results of luciferase assay further suggest that the indel has no obvious effects on gene expression at the protein level. Furthermore, genotypes at the indel do not obviously change the level of ARMS2 transcripts in retina–RPE–choroid samples. All these results suggest that the indel may not influence the level of ARMS2 transcripts in vivo and in vitro. Since the indel and A69S are in strong LD, results from these experiments may also suggest that A69S does not change the level of ARMS2 transcripts. 

Yang et al. first reported the inverse association of R38X with AMD. ¹² This inverse association was somehow verified in a small Polish sample (90 wet AMD cases and 40 controls, P = 0.053). ¹⁵ Bergeron-Sawitzke et al. found that genotypes at R38X were not associated with AMD after examining 421 cases and 215 controls. ¹⁶ We analyzed three common coding variants of ARMS2, and found that the reported inverse association of R38X with AMD is insignificant after adjustment for sex and age in our dataset (1169 AMD cases and 707 controls). ¹⁷ Although genetic association of R38X in AMD might be either weak or insignificant, it is still necessary to examine its biological consequences in order to understand whether an assumed “loss of function” of R38X in ARMS2 might not be responsible for the etiology of AMD. ¹³  

For R38X to execute “nonsense-mediated mRNA decay,” the precondition is that the annotated translational start site must be correct or at least in-frame with the annotated site. If that is not the case, SNP rs2736911 will not result in a premature stop codon (R38X), and it will no longer be a nonsense variant. Yang et al. found that R38X correlates with a lower level of ARMS2 transcripts in placentas. ¹² In this study, by utilizing an in vitro transcription assay, we observed that R38X decreased the level of ARMS2 transcripts. This effect could be due to accelerated degradation of transcripts. However, the regulation of ARMS2 transcripts in retina–RPE–choroid samples might be more complicated. Our data suggest that genotypes at R38X do not obviously affect the level of ARMS2 transcripts in human retina–RPE–choroid samples. 

The functional role of ARMS2 protein in AMD remains an enigma. The results of these experiments show that R38X could mediate transcript decay, at least in vitro. R38X is thus a potential nonsense change, suggesting that the annotated translational site could be correct. Recently, a unique peptide fragment sequence matching the N-terminus of the predicted ARMS2 protein was identified by proteome-wide screening, ²² suggesting that ARMS2 protein could be translated from the annotated translational start site. It is known that ARMS2 is expressed at a low level in retinas. ² Previous reports also indicated that in some normal retinas, ARMS2 transcripts are not detectable by qRT-PCR, suggesting that a “loss of function” of ARMS2 may not be involved in AMD pathogenesis. ^10,13  

Another interesting phenomenon we observed is that only isoform A, not both isoforms of ARMS2 transcripts, was detected from transcription of minigenes in MEFs. This suggests that transgenic modeling of ARMS2 in mice might not be able to imitate the expression of ARMS2 in humans. 

In conclusion, our data showed that variant R38X, not the indel, decreases the stability of ARMS2 transcripts in vitro. Genotypes at R38X and the indel do not obviously affect the level of ARMS2 transcripts in human retina–RPE–choroid samples. These results suggest that variants R38X and the indel are less likely to play a pathogenic role in AMD by changing the level of ARMS2 transcripts. 

We thank all the individuals who donated eyes for the study and their families. We also thank the Florida Lions Eye Bank, which provides eye tissues for research purposes. 

Supported by BrightFocus Foundation Grant M2012048 (GW). 

Disclosure: E.A. Minor, None; B.L. Court, None; S. Dubovy, None; G. Wang, None