Association of Neovascular Age-Related Macular Degeneration with Specific Gene Expression Patterns in Peripheral White Blood Cells

Abstract

Purpose.: Inflammation probably plays a major role in the pathogenesis of age-related macular degeneration (AMD). The authors evaluated whether AMD is associated with gene expression patterns in white blood cells (WBCs) and whether such a pattern may serve as a biomarker for the disease.

Methods.: Microarray analysis of gene expression in peripheral WBCs was performed on patients with neovascular AMD (NVAMD; n = 16) and controls (n = 16). Results were validated using quantitative real-time RT-PCR (QPCR) on another set of patients (n = 14) and controls (n = 16), respectively. QPCR findings were evaluated using receiver operator characteristic (ROC) curves and correlated with genotyping for the major risk single nucleotide polymorphisms (SNPs) for AMD in the genes for complement factor H and LOC387715.

Results.: NVAMD-associated expression was identified for eight sequences (false discovery rate [FDR] = 0%) and 167 sequences (FDR = 10%), respectively. There was an enrichment of genes involved in antigen presentation among the AMD-associated genes (P = 0.0029). QPCR confirmed increased expression (1.6- to 4.3-fold) of four genes (HSPA8, IGHG1, ANXA5, VKORC1) in association with NVAMD (P = 0.02–0.0002). Area under the curve for these genes according to ROC analysis ranged from 0.776 to 0.815. Gene expression was not associated with genotyping for risk SNPs or WBC counts.

Conclusions.: NVAMD is associated with altered gene expression in peripheral WBCs that is not underlined by the major risk SNPs for the disease. Such altered expression may potentially serve as a biomarker for the disease. These data support the involvement of systemic immune response in the pathogenesis of AMD.

Multiple types of evidence implicate both systemic and local inflammation in the pathogenesis of age-related macular degeneration (AMD). Among the evidence suggesting local inflammation in AMD are the presence of inflammatory mediators in drusen ^1,2 and the presence of macrophages in the choroid of eyes affected by AMD. ^3,4 Involvement of systemic inflammation in the disease is reflected by association of single nucleotide polymorphisms (SNPs) in complement components such as factor H, C3, C2, and B factor and SNPs in a chemokine receptor and, potentially, Toll-like recep-tor-3. ^5–13 Furthermore, antiretinal autoantibodies are present in the sera of patients with AMD, ^14–16 and altered plasma levels of inflammatory markers such as C-reactive peptide and complement components may also be associated with AMD. ^17–19 In addition, perturbed macrophage function is thought to lead to the development of features resembling AMD in mice strains deficient in chemokine receptors or their ligands. ^20,21 Macrophages can also modulate the development of neovascularization in mice models of laser-induced choroidal neovascularization. ^22–24  

Microarray analysis of gene expression in blood cells detected disease-specific expression patterns in several pathologic conditions—among them lupus, ²⁵ multiple sclerosis, ²⁶ stroke, ²⁷ schizophrenia, ²⁸ and Huntington's disease ²⁹ —in which inflammation may play a role. Characterizing such expression signatures may provide insight into the pathogenesis of the disease and may potentially serve as a surrogate biomarker for diagnosis and management of the disease. 

In view of the data suggesting the involvement of inflammation and white blood cells (WBCs) in the pathogenesis of AMD, it is conceivable that gene expression signature in WBCs may also exist in AMD and that it may reflect the involvement of these cells in the disease. To that end, we have characterized gene expression patterns in WBCs from patients with AMD. The feasibility of using expression patterns as a biomarker for neovascular AMD (NVAMD) was then evaluated and compared with that of using major risk SNPs for the disease for the same purpose. 

Materials and Methods

Patients

Blood samples were drawn from patients with NVAMD and controls older than 60 years who were evaluated in the Department of Ophthalmology of the Hadassah-Hebrew University Medical Center in Jerusalem, Israel, for routine eye examinations or for pathologic conditions other than AMD. The authors adhered to the tenets of the Declaration of Helsinki. Institutional Ethics Committee approval was obtained, and each patient signed an informed consent form. AMD was diagnosed and graded according to the AREDS trial classification, ³⁰ and choroidal neovascularization was diagnosed based on ophthalmoscopy and fluorescein angiography. Only patients with active choroidal neovascularization according to fluorescein angiography were included in the study, and none of the patients had subretinal or subretinal pigment epithelium hemorrhage larger than 50% of the lesion size. Mean lesion size according to fluorescein angiography was 4057 ± 1411 μm (range, 1600–7700 μm). Inclusion criteria for the control group included clear media that enabled ophthalmoscopy and absence of intermediate-size drusen, multiple small drusen, or retinal pigment epithelial abnormalities characteristic of AMD (AREDS category I). Persons with severe systemic diseases such as malignancies, active ischemic heart disease, uncontrolled diabetes or pulmonary disease, or autoimmune diseases were excluded from the study and the control groups. 

White Blood Cell Separation and RNA Extraction

Four milliliters of blood placed in tubes containing EDTA was used for WBC separation and complete blood count (CBC). CBC was performed at the central laboratory of the Hadassah Medical Center using an automated system. For RNA extraction, 8 mL hypotonic lysis buffer (155 mM NH₄Cl [Gadot, Or Akiva, Israel], 10 mM CH₂O₃ · NH₃ [Sigma-Aldrich, St. Louis, MO], 0.1 mM EDTA [J.T. Baker, Phillipsburg, NJ], pH 7.4) was added to the blood. The sample was stored on ice for 10 minutes and was then subjected to centrifugation at 2000g at 4°C for 10 minutes. Supernatant was discarded, and the previous stage was repeated. The pellet of white blood cells was resuspended in 1 mL reagent (TRI; Sigma-Aldrich). Total RNA was extracted according to the manufacturer's instructions. Possible remnants of DNA were degraded (DNA-free; Ambion, Austin, TX), and RNA samples were purified (RNeasy MinElute Cleanup Kit; Qiagen, Hilden, Germany). Samples were then stored at −80°C until further use. 

Microarray Analysis

Microarray experiments were performed as we have previously described. ^31,32 Briefly, an indirect fluorescence-labeling method was applied using 20 μg purified RNA as a template for cDNA synthesis with reverse transcriptase (SuperScript II; Invitrogen, Karlsruhe, Germany) and incorporation of aminoallyl-UTP (Sigma-Aldrich) during first-strand cDNA synthesis followed by coupling monoreactive Cy3 or Cy5 fluorescent dye (Amersham Biosciences Inc., Piscataway, NJ). The total amount of dye incorporation (measured in picomoles of dye per probe) and the ratio of unlabeled to fluorescence-labeled nucleotide in the probe were assessed by measuring probe absorbance at 260 nm, 550 nm, and 650 nm to assess DNA, Cy3, and Cy5 concentrations, respectively. 

A reference sample design was applied ³¹ by which sample (either NVAMD or control) RNA was labeled with Cy3 and reference RNA was labeled with Cy5. The analysis included 32 microarrays (16 NVAMD samples and 16 controls). Four patients underwent photodynamic therapy (PDT) >2 months before blood drawing for the study. A human array generated from an oligonucleotide set by Operon Biotechnologies Inc. (Cologne, Germany), version 3.0, which contains 35,035 oligonucleotide probes representing approximately 25,100 unique genes, was used for all experiments. The fluorescent probe was placed on the slide and was then incubated at 42°C for 16 to 22 hours. Posthybridization washes were performed, followed by scanning with the Axon scanner and the image analysis software (GenePix Pro 4.1 Microarray; Axon Instruments Inc., Union City, CA). 

Microarray analysis, including background correction, filtration, and normalization, was performed using a microarray software (TM4) package developed by the TIGR Institute (Rockville, MD), ³³ followed by analysis using the MIDAS (Microarray Data Analysis System) program. SAM (Significance Analysis of Microarrays) algorithm ³⁴ and LIMMA (Linear Models for Microarray data) ³⁵ algorithm through the R program (http://www.r-project.org/) were used to assess significance. Both algorithms provide FDR (false discovery rate) as an estimate of the significance of the results. 

Functional annotation of genes represented on the array was performed using Web-based software (http://bioinfo.vanderbilt.edu/gotm, http://fatigo.bioinfo.cipf.es). Potential enrichment of specific functional classes among genes with significantly high or low expression levels associated with AMD was then assessed using Fisher's Exact test. 

Quantitative Real-Time RT-PCR

A first-strand synthesis kit (Reverse-iT; ABgene, Epsom, UK) was used to prepare cDNA from 1 μg total mRNA through reverse transcriptase polymerase chain reaction using anchored oligo dT primers. The expression levels of four genes—ANXA5, HSPA8, IGHG1, VKORC1—were assessed in WBCs from 14 patients with NVAMD and 16 controls that had not been analyzed by microarray. Eleven patients underwent PDT >2 months before blood drawing. GAPDH served as the endogenous control to which each sample was normalized. Reactions were performed using either the dye chemistry (SYBR Green; Applied Biosystems, Foster City, CA) or the fluorogenic 5′ nuclease chemistry (TaqMan; Applied Biosystems) technique (Supplementary Table S1). Each tube contained 10 μL PCR mix and 2.8 μL primers for SYBR Green or 1 μL TaqMan assay. Optimal amounts of cDNA were calibrated for each primer. A total volume of 20 μL was completed by double distilled water. Samples were prepared in triplicate, and calculations were performed on the average value. Reactions were carried out and analyzed (ABI Prism 7000 and 7900HT systems; Applied Biosystems). 

Genotyping

DNA was extracted from 200 μL whole blood using a DNA kit (FlexiGene; Sigma) according to the manufacturer's protocol. Subjects who were studied using QPCR were genotyped for the LOC387715 rs10490924 and complement factor H (CFH) rs1061170 SNPs. A population of 163 patients with AMD and 104 controls was also genotyped for the VKORC1_2255 polymorphism. Genotyping was performed using restriction enzyme analysis of PCR (for rs10490924 and VKORC1_2255) or sequencing (for CFH) (Supplementary Table S2). Fragments were resolved on 2% agarose gel and visualized by ethidium bromide marking under an ultraviolet light lamp. 

ROC Analysis

Receiver operating characteristic (ROC) curves assist in evaluating the feasibility of using a certain test for diagnostic purposes by measuring the area under the ROC curve (AUC). The closer the AUC is to 1, the more specific and sensitive is the test. A ROC curve was calculated for each of the gene expression assays evaluated by QPCR and for the genotyping results of LOC387715 and the CFH SNPs. Calculations were performed with Web-based software (http://www.rad.jhmi.edu/jeng/javarad/roc/JROCFITi.html). For input data purposes, patients with AMD were defined as “1” and controls as “0.” When evaluating gene expression, the RQ values (relative expression derived from QPCR) of individuals in each group were entered and processed. For the evaluation of genotypes, homozygotes for the wild-type allele were categorized as “1,” heterozygotes as “2,” and homozygotes for the polymorphism as “3,” and data were processed using an algorithm appropriate for discrete parameters. 

Statistical Analysis

Statistical analysis for associations was performed using the χ² test. For all other comparisons the Student's t-test was used. All tests were performed using biostatistics software (InStat; GraphPad, San Diego, CA). 

Results

Microarray and QPCR Analysis

Microarray and statistical analysis using the SAM algorithm identified eight sequences showing altered expression in WBCs in patients with NVAMD (mean age, 79.4 ± 6.9 years; range, 65–91 years) compared with age-matched controls (mean age, 75.7 ± 6.8 years; range, 64–90 years) at an FDR of 0%. Each of these sequences demonstrated increased expression in patients. There were 159 additional sequences showing altered expression at a less stringent significance level of 10% (data not shown). When implementing an alternative analysis algorithm, LIMMA, 53 differentially expressed sequences were identified at an FDR of 20% after adjusting the P values for multiple comparisons using the Benjamini-Hochberg algorithm. Of these sequences, 33 were also identified by the SAM analysis at FDR 10%. Table 1 shows the 16 known genes from the 33 sequences identified by both algorithms. 

Table 1.

 

Genes Identified as Upregulated in Microarray Analysis of RNA from WBCs of Patients with NVAMD and Controls According to Both SAM and LIMMA Algorithms

Symbol	Name	GenBank Accession No.	Unigene No.	Fold Change (AMD/control)
CCNB1	Cyclin B1	NM_031966	Hs.23960	2.3
ANXA5 *	Annexin 5	NM_001154.2	Hs.480653	2
CSE1L	Chromosome segregation 1-like	NM_001316.2	Hs.90073	2
EIF5A	Eukaryotic translation initiation factor 5A	NM_001970.3	Hs.534314	1.6
TPD52	Tumor protein D52	NM_005079.2	Hs.368433	2
FANCG	X-ray repair, complementing defective in Chinese hamster 9; XRCC9	NM_004629.1	Hs.591084	1.6
HLA-DQA2	HLA class II histocompatibility antigen, DQ(6) alpha chain precursor	NM_020056.2	Hs.591798	1.6
IGHG1 *	IgG heavy chain locus	AF013620.1	Hs.510635	2
ISOC1	CGI-111	NM_016048.2	Hs.483296	2
TBC1D7	DKFZp686N2317	NM_016495.2	Hs.484678	1.8
NSUN2	FLJ20303	NM_017755.4	Hs.481526	1.6
ACN9	DC11	BC028409	Hs.592269	1.5
VKORC1 *	Vitamin K epoxide reductase complex, subunit 1	NM_024006	Hs.324844	1.8
TXNDC5	ERP46	NM_030810	Hs.150837	1.9
RBBP4	Retinoblastoma binding protein 4	BC053904	Hs.647652	2.1
ZDHHC4	Zinc Finger, DHHC domain containing 4; hypothetical protein FLJ10479	NM_018106	Hs.5268	1.5

 Fold change represents the ratio between the average expression levels of patients and those of controls. FDRs were 10% for SAM and 20% for LIMMA.

*  Genes identified by SAM algorithm at FDR of 0%.

QPCR validation experiments were performed on samples from 16 controls and 14 patients with NVAMD that had not been evaluated by the microarrays. Three genes (ANXA5, VKORC1, IGHG1) identified by both the SAM and LIMMA algorithms as having NVAMD-associated expression were included in this analysis, as was one gene identified only by SAM (HSPA8; 1.7-fold change AMD/control). These genes were selected for QPCR analysis because their known function suggests them as candidates for involvement in the pathogenesis of AMD. ^36,37 QPCR showed significantly higher mRNA levels in patients with NVAMD for each of the genes ranging from 1.6-fold to 4.3-fold compared with the controls (P = 0.02–0.0002), thereby confirming microarray results (Fig. 1). 

Figure 1.

 

Relative mRNA levels of four genes in patients with NVAMD (n = 14) and controls (n = 16) according to real-time QPCR. RQ, relative quantification of gene expression in arbitrary units, determined by ΔΔCt calculation; bars, average expression ± SEM. Light bars: control. Dark bars: AMD.

Functional analysis was performed to evaluate whether groups of genes belonging to the same biological pathway show altered expression in WBCs from patients with NVAMD. By applying the GOTM algorithm, we found enrichment for genes involved in antigen presentation among genes showing altered expression levels in NVAMD (P = 0.0029). Genes included in this group were HLA-DQA1, HLA-DQA2, and HLA-DQB1. These genes showed increased mRNA levels in patients with AMD compared with controls ranging from 1.6- to 1.8-fold. 

Correlation of Blood Counts and Genetic Variation with Gene Expression Levels

Blood counts and genotyping for major risk SNPs for AMD were performed to assess whether differences in numbers of WBCs between patients and controls or the risk SNPs for the disease underlie gene expression alterations. There was no significant difference in the average (±SD) number of WBCs (8.5 ± 2.7 vs. 7.2 ± 1.8), lymphocytes (2.7 ± 2 vs. 2 ± 0.8), monocytes (0.6 ± 0.2 vs. 0.6 ± 0.3), granulocytes (5.4 ± 2 vs. 4.3 ± 1.4), eosinophils (0.2 ± 0.2 vs. 0.2 ± 0.1), or basophils (0.014 ± 0.036 vs. 0.057 ± 0.064) between patients with NVAMD and controls, respectively. 

Although the SNPs in complement factor H (rs1061170) and LOC387715 (rs10490924) are associated with AMD in the Israeli population, ^38,39 there was no association between mRNA levels in WBCs of the four genes tested by QPCR and genotyping of the SNPs (Fig. 2). 

Figure 2.

 

mRNA levels of four genes evaluated by QPCR according to genotyping for risk SNPs for AMD. (A) mRNA levels according to genotyping for rs1061170 SNP in CFH. (B) mRNA levels according to genotyping for the rs10490924 SNP in LOC387715. None of the expression level differences between the genotypes was significant. RQ, relative quantification of gene expression. Bars, average expression ± SEM. Dark bars: risk allele. Light bars: wild-type allele.

Receiver Operating Characteristic Curve Analysis

To assess the feasibility of using measurements of mRNA levels of differentially expressed genes in WBCs as biomarkers for AMD, we have fitted QPCR results for the four genes tested on ROC curves (Figs. 3A–D). AUC was then calculated for each gene. AUC was 0.815 for HSPA8, 0.803 for IGHG1, 0.776 for VKORC1, and 0.781 for ANXA5. 

Figure 3.

 

ROC curves based on QPCR results for four NVAMD-associated genes (A–D) and on genotyping for risk SNPs for AMD (E, F). The graphs show the true-positive rate against the false-positive rate for distinction between patients with AMD and unaffected persons at different possible cut-points of a diagnostic test. AUCs ranged from 0.776 to 0.815 for the QPCR data and 0.632 to 0.818 for genotyping. Gray lines: 95% confidence interval of the fitted ROC curve.

Then we compared measurements of gene expression levels in WBCs with genotyping for risk SNPs for the disease in CFH and LOC387715 as biomarkers for AMD. This was performed by ROC analysis of the same data set used for QPCR analysis. AUCs for CFH (rs1061170) and LOC387715 (rs10490924) were 0.818 and 0.632, respectively (Figs. 3E, 3F). 

Assessment of SNP in VKORC1

A SNP in the VKORC1 gene (VKORC1_2255 SNP), one of the genes demonstrating increased expression associated with NVAMD according to our findings, was previously associated with coronary heart disease and stroke. ⁴⁰ To evaluate potential association of this SNP with NVAMD, we genotyped 163 patients with NVAMD and 104 unaffected controls. There was no association between this SNP and AMD because allele frequencies were similar in both populations (data not shown). 

Discussion

This study demonstrated specific alterations in gene expression patterns in peripheral WBCs from patients with NVAMD. These alterations were reproducible in two different data sets of patients and controls using two methodologies and were not associated with altered enumeration of WBCs or with major risk SNPs for the disease. Measurements of altered gene expression in WBCs facilitated a noninferior distinction between patients with NVAMD and controls compared with that obtained by using genotyping for major risk SNPs for the disease. 

Many of the genes that showed AMD-associated expression are involved in inflammation, a process that has a major role in the pathogenesis of AMD. ⁴¹ For example, mRNA for the immunoglobulin heavy chain gamma 1 gene was increased in AMD in both microarray and QPCR analysis. This gene encodes the Fc area of the IgG antibody, the most abundant antibody in the serum and a major participant in the adaptive immune response. Such increased expression levels of IgG are in accordance with reports of higher levels of antiretina autoantibodies in retinas and sera of patients with AMD compared with controls, ^14–16 and of the presence of immunoglobulins within drusen, the hallmark of AMD. ^42,43  

Functional analysis showed enrichment for genes involved in antigen presentation among genes demonstrating AMD-associated expression. These polymorphic genes, essential for regulation of the immune response, are involved in a variety of inflammatory and autoimmune diseases. Specific alleles of such HLA genes have been associated with the risk for AMD. ⁴⁴ Abnormal antigen presentation of retinal or drusen composites may contribute to the pathogenesis of AMD by fueling the self-inflicted attack of the immune system against the retina. 

Increased levels of ANXA5 mRNA transcripts were also found in the WBCs of patients with AMD. ANXA5, which plays a role in the regulation of blood clotting, has been found in atherosclerotic plaques ⁴⁵ and is proposed to have anti-inflammatory functions. ⁴⁶ Interestingly, other annexins were previously identified in drusen. ⁴⁷ ANXA5 levels may increase in AMD as part of a healing response. 

VKORC1 mRNA levels were also increased in the WBCs of patients with AMD. The product of this gene is an enzyme that activates vitamin K, an essential cofactor in many stages of the clotting cascade, and is involved in angiogenesis. ⁴⁸ A SNP in the encoding gene was associated with coronary heart disease and stroke, ⁴⁰ yet genotyping failed to identify an association between this SNP and NVAMD. Extensive cross-talk between the coagulation and complement systems is known to exist. ⁴⁹ Thus, the altered expression of genes involved in coagulation in the WBCs of patients with AMD, as we have described, may be of importance in the context of complement activation, angiogenesis, and hemorrhaging, all of which are characteristics of AMD. Altered expression of genes involved in angiogenesis, such as VKORC1, in the WBCs of patients with NVAMD supports the idea that subpopulations of WBCs, such as macrophages, may modulate the growth of choroidal neovascular vessels in patients with NVAMD. ^22–24  

SNPs in several genes have been associated with the risk for AMD. Yet we show that altered gene expression in WBCs is not associated with major risk SNPs for AMD in complement factor H and on chromosome 10q26 (LOC387715) and that such expression patterns distinguish between patients and controls at least as well as genotyping for the risk SNPs. These data suggest that though polymorphisms in genes involved in the process may in part account for the pathogenesis of the disease, other factors also play important roles in AMD. 

The existence of specific gene expression patterns in the WBCs of patients with AMD—in addition to the insight they provide to the pathogenesis of the disease—may serve as a biomarker for AMD. AMD is often undiagnosed until the late stages and is then often associated with substantial visual loss. ⁵⁰ Diagnosis of the disease at an earlier stage may facilitate the commencement of treatment and periodic follow-up that can improve visual outcome for patients with AMD. ³⁰ There is also a need for biomarkers that will correlate with disease progression and response to therapy. Thus, a blood test that facilitates detection of such biomarkers may serve as an important tool in the treatment algorithm for AMD. 

Although several biomarkers for the disease have been suggested—among them measurement of inflammatory mediators and markers for oxidative injury in the serum ^{14,17,18,51–53} —such markers were not demonstrated to be of value for the diagnosis of AMD in a clinical setting. 

Further research is required to assess whether measurement of gene expression will be useful for this purpose and to evaluate whether altered gene expression in WBCs reflects the involvement of these cells in NVAMD or is secondary to the disease process or the therapeutic interventions some patients underwent before they were enrolled in the study. 

Supplementary Materials

Supplementary Table S1 - (PDF) 

Supplementary Table S2 - (PDF) 

Footnotes

 Supported by a grant from the Hadassah-Hebrew University Medical Center.

Footnotes

 Disclosure: M. Lederman, P; A. Weiss, P; I. Chowers, P