Late Outgrowth Endothelial Progenitor Cells in Patients with Age-Related Macular Degeneration

Abstract

purpose. To evaluate the feasibility of isolating and expanding endothelial progenitor cells (EPCs), in the form of late outgrowth endothelial progenitor cells (OECs), from the peripheral blood of an aged population, particularly patients affected by different forms of AMD.

methods. Peripheral blood mononuclear cells were collected from young control subjects (n = 18) and from elderly subjects with non-AMD/low-risk dry AMD (n = 15), high-risk dry AMD (n = 6), or neovascular AMD (nvAMD; n = 32); cultured in established conditions; and observed for appearance of OEC clusters and growth characteristics on expansion. Expression of VEGF receptor-2 (KDR) in OECs after expansion was determined by Western blot. Plasma samples of study subjects were analyzed for CRP and VEGF levels.

results. OEC cultures were successfully generated from a similar number of subjects in each group. After adjustment for all other variables, subjects with high-risk dry AMD had a 5.6-fold higher number of OEC clusters per 20 mL blood, and subjects with nvAMD had a 5.1-fold high number than did subjects with non-AMD/low-risk dry AMD (P < 0.05). High-risk dry AMD generated 63 times more (NS) and nvAMD 32-times more (P < 0.05) OECs on expansion of clusters than did non-AMD/low-risk dry AMD. Population doubling occurred significantly faster in cultures from nvAMD eyes compared to non-AMD/low-risk dry AMD eyes. In addition, a significant correlation between the number of OEC clusters, expanded OECs and levels of KDR was demonstrated.

conclusions. An OEC population was isolated and expanded from the blood of elderly control and AMD-affected patients and demonstrated significantly higher number of initial OEC clusters and expansion potential of OECs in patients at risk for or already affected by nvAMD. OECs may be used for further phenotypic, genetic, and functional analyses in patients with nvAMD.

Age-related macular degeneration (AMD) is the leading cause of irreversible visual impairment and blindness in the Western World in those older than 50 years. It takes one of two forms: dry or wet. Most vision impairment in AMD is caused by conversion of the dry into the wet or neovascular form of the disease (nvAMD), characterized by invasion of pathologic new vessels under the macula (choroidal neovascularization; CNV). ^{1 2} Although various fundus features apparent on clinical examination signify increased risk, the mechanisms that cause conversion from dry into wet AMD remain unknown. ^{3 4 5}  

Until recently, it was assumed that all neovascularization develops from activation, migration, and proliferation of resident endothelial cells. In CNV formation, these cells were thought to arise from the choriocapillaris (for review, see Ref. ⁶ ). This paradigm changed when Asahara, et al. ⁷ described that peripheral blood contains a population of bone marrow–derived circulating endothelial progenitor cells (EPCs) that differentiates into endothelial cells at sites of postnatal vasculogenesis and pathologic neovascularization. Several studies in animal models of nvAMD now provide evidence that these EPCs may also be major contributors to the formation and growth of CNV. ^{8 9 10 11 12} In particular, EPCs were recruited at high frequency into the developing CNV in a rodent model caused by overexpression of vascular endothelial growth factor (VEGF) within the outer retina. ⁹ In addition, cells expressing the EPC marker CD133 were identified in specimens of surgically excised CNV. ¹³  

Much controversy remains about the identity and regulation of recruitment of EPCs into developing neovascularization. Many studies have suggested that the absolute number of circulating EPCs in peripheral blood is a major determinant of their participation in neovascularization. Some investigators therefore use flow cytometry to analyze the frequency of rare cells expressing a combination of specific surface antigens such as CD34, CD133, and VEGFR-2. ^{2 14 15 16 17 18 19 20 21 22 23} However, the first study to further isolate CD34⁺CD133⁺VEGFR-2⁺ cells and subject them to endothelial cell assays recently showed that they are in fact hematopoietic progenitor cells, but are not EPCs and do not have vessel formation ability. ²⁴ Of interest, the frequency of CD34⁺CD133⁺ circulating cells dramatically decreases with age, and the cells are often undetectable in the blood of elderly patients in the age group susceptible to nvAMD. ¹⁹  

Another method defines EPCs by cell culture of isolated peripheral blood mononuclear cells (PBMCs) that form outgrowth clusters after short-term (<7 days) culture in an angiogenic growth factor–enriched medium. ²⁵ The cells identified in this assay are called early endothelial progenitor cells (eEPCs) because of their early appearance in culture. However, eEPCs express hematopoietic markers and lack long-term proliferative potential, suggesting that they are in fact monocyte-derived cells that manifest some angiogenic properties. ^{26 27} A recent report further supports this hypothesis by demonstrating that eEPCs do not form vessels in vivo and even differentiate into macrophages. ²⁸ The frequency of eEPCs also appears to correlate inversely with age, and subjects older than 70 years possess few if any of these cells. ²⁹  

In addition to these eEPCs, Lin, et al. ³⁰ have identified a second functional subset of EPCs that they term late outgrowth endothelial progenitor cells (OECs), cells that appear during the second week of PBMC culture. OECs appear to share bone-marrow origination with eEPCs; however, they differ from eEPCs in three key properties: the long-term ability to proliferate in culture, ^{28 30 31} the expression of markers of endothelial but not hematopoietic phenotype ^{30 31 32} and the ability to form perfused vessels in vivo. ²⁸ The latter finding was highlighted in another recent report that confirms formation of robust vascular networks in vivo by OECs expanded from human cord blood. ³³  

The present study set out to test the possibility of isolating an EPC population circulating in the peripheral blood of elderly people and patients with different stages of AMD, especially nvAMD, and to expand systematically and characterize those cells. If successful, EPCs could be isolated from the blood of patients affected by AMD and be expanded for further phenotypical, genetic, and functional analyses, to provide a better understanding of their possible involvement in the development and severity of CNV. 

Methods

Study Subjects

Participants included a group with nvAMD (n = 32), age-matched non-AMD/dry AMD group (n = 21), and a young control group (n = 18). The clinical diagnosis of nvAMD was defined as choroidal neovascularization and/or subretinal fibrosis involving the macula in at least one eye. Diagnosis of dry AMD was defined as the presence of drusen greater than 63 μm in at least one eye. Subjects with non-nvAMD were further divided into a non-AMD/low-risk dry AMD group (n = 15) and a high-risk dry AMD group (n = 6) based on the simplified severity scale for the development of nvAMD/AREDS (Age-Related Eye Disease Study) report no. 18. ⁴ All participants were recruited from the National Eye Institute Clinic in Bethesda, Maryland. Young control and aged non-AMD participants were normal volunteers. Exclusion criteria included age <55 for the aged subjects, <18 or >40 for the young subjects and the presence of other retinal disease and presence of hematologic disease for all subjects. All participants underwent a complete medical history and all aged subjects received an ophthalmic clinical evaluation. The protocol for collection and use of human blood samples was approved by the NEI Institutional Review Board, and all participants gave informed consent to participate in the study. The study adhered to the guidelines set forth in the Declaration of Helsinki. 

OEC Isolation and Cell Culture

Peripheral blood (32 mL) was collected in heparin sodium tubes (Vacutainer CPT; BD Labware, Bedford, MA). After density gradient centrifugation, PBMCs were collected, washed, and resuspended in endothelial cell growth medium (EGM)-2 (Cambrex, Walkersville, MD). Cells were plated at a density of 2 × 106 cells/cm² in 24-well plates precoated with fibronectin (BD Labware). Medium was changed daily for 7 days and on alternate days thereafter, according to the protocol of Lin et al. ³⁰ OEC clusters that appeared between 7 and 30 days of culture were enumerated by visual inspection. Subconfluent cells were trypsinized and replated in vessels coated with human fibronectin at a concentration of 10 μg/cm² (Chemicon, Temecula, CA). OECs were further subpassaged and expanded until cell senescence, as determined by morphology changes, decrease in proliferation and positive staining for senescence-associated β-galactosidase (BioVision Research Products, Mountain View, CA) was reached. At the onset of senescence, the total number of expanded OECs was counted for each successfully generated OEC culture, using a counting chamber (Neubauer; Paul Marienfeld GmbH & Co. KG, Lauda-Königshofen, Germany) and trypan blue stain (Invitrogen) for exclusion of dead cells, according to the manufacturer’s instructions. 

For representation of a growth kinetic curve and calculation of population doubling times (PDTs), 21 different OEC cultures (young control, n = 5; non-AMD/low-risk dry AMD, n = 4; high-risk dry AMD, n = 3; and nvAMD, n = 9) were selected at random and counted at the beginning and end of each passage, starting after collection of subconfluent clusters until the onset of senescence of the final OEC culture. PDT was calculated as the time interval between cell seeding and harvest at each passage, divided by the number of population doublings (PDs) for that passage (PD was derived from the equation PD = log₂ [C _h/C _s], where C _h is the number of viable cells at harvest and C _s is the number of cells seeded at each passage). 

Human umbilical vein endothelial cells (HUVECs; Cambrex Corp., East Rutherford, NJ) and human microvascular endothelial cells (HMVEC-dermal cells; Cambrex) were cultured in EGM-2 on fibronectin-coated vessels. Characterization for endothelial cell attributes was performed on early-passage OECs (passages 2–4), HUVECs, and HMVEC-dermal cells (passages 2–8). 

Immunophenotyping for Endothelial Surface Marker Expression

For analysis by fluorescence-activated cell sorting, mAbs against CD31-FITC, CD34-FITC, CD105-FITC, CD14-FITC, CD146-phycoerythrin (PE), c-kit-PE (all BD-PharMingen, San Jose, CA), VEGFR-2-PE (R&D Systems Inc., Minneapolis, MN), VE-Cadherin-FITC (Serotec, Raleigh, NC), and CD133-PE (Miltenyi Biotech, Auburn, CA) were used. Isotype-matched IgG antibodies were used as the control. OECs, HUVECs, and HMVEC-dermal cells were serum-starved for 18 hours in endothelial cell basal medium (EBM)-2+0.1% BSA to prevent receptor downregulation by cytokines or factors contained in EGM-2. The cells were then trypsinized and incubated at 4°C for 30 minutes with primary or isotype control antibody, washed, and identified by flow cytometry (FlowJo software; Tree Star Inc., Ashland, OR). 

Uptake of DiI-Acetylated Low-Density Lipoprotein, Staining for Ulex europaeus Lectin, and In Vitro Tube Formation Assay

Expanded OECs grown on fibronectin-precoated culture slides were incubated with 10 μg/mL DiI-Ac-LDL (Invitrogen-Molecular Probes, Eugene, OR) at 37°C for 1 hour. The cells were washed, fixed in 4% paraformaldehyde, and further incubated with 10 μg/mL lectin from Ulex europaeus (Sigma-Aldrich, St. Louis, MO) for 1 hour. The slides were washed and mounted, and fluorescent images were collected. 

The cells were seeded on 96-well plates coated with 60 μL solidified synthetic matrix (Matrigel; BD Labware) at a density of 2 × 10⁴ cells per well. Tube formation was assessed after incubation at 37°C for 16 to 18 hours. 

Western Blot Analysis for VEGFR-2 (KDR) Expression

OECs and HUVECs were serum starved in EBM-2+0.1% BSA 18 hours before collection of the cells, to prevent downregulation of KDR by VEGF or other factors contained in EGM-2 medium and/or serum. Western blot analysis was performed as previously described. ³⁴ Membranes were incubated with rabbit mAb against VEGFR-2 (Cell Signaling Technology, Beverly, MA) or goat polyclonal Ab to β-actin (Abcam, Cambridge, MA), and the intensity of the obtained protein bands was measured (Kodak ID Image Analysis, Rochester, NY). 

VEGF and CRP Plasma Levels

Heparin plasma was separated by density gradient centrifugation, and samples were immediately frozen and stored at −80°C. The levels of heparin plasma VEGF were measured by enzyme-linked immunosorbent assay (Quantikine ELISA kit; R&D Systems). CRP plasma levels were then analyzed at the Pierce Biotechnology facility (Searchlight Multiplex ELISA; Rockford, IL). 

Statistical Analysis

All statistical tests have a type-I error probability of 0.05 (SAS, ver. 8.02; SAS Cary, NC, or R, ver. 2.3.1). Several model-building steps were used, including higher-order terms (quadratic or cubic) or transformations (logarithmic or exponential), to achieve the best fit for a given model. 

Baseline covariate imbalances were compared by using a generalized linear model of a multinomial distribution (PROC GENMOD; SAS). Time to first OEC cluster formation was modeled with a standard ANOVA (PROC GLM; SAS). The probability of no OEC clusters was modeled by standard logistic regression (PROC GENMOD; SAS). 

The regression model for the number of OEC clusters was fit for all 71 subjects, including those with no OEC clusters. A similar method was used for the regression model of total number of OECs after expansion. 

A paired Student’s t-test was used for analysis of differences in PDTs and VEGF and CRP blood levels among separate groups. 

Results

Clinical Characteristics of Study Subjects

The clinical characteristics of all study subjects are summarized in Table 1 . Factors thought to decrease (age, ^{19 29} smoking, ^{16 35} and cardiovascular disease ^{25 35 36 37} ) or increase (statin therapy ^{38 39 40} ) the number and/or function of circulating EPCs (as determined by flow cytometry or eEPC assay) were evaluated for each group. Positive family history of AMD was also assessed. Although there is a slight imbalance from the non-AMD/low-risk to the high-risk/dry AMD and nvAMD categories, smoking history and sex were the only predictors significantly overrepresented in any one group (P = 0.01 for sex and P = 0.03 for smoking). Smoking is a variable known to be associated with the incidence of nvAMD. ⁴¹  

Endothelial Characterization of OECs Derived from Study Subjects, Karyotyping, and Cell Senescence

OEC clusters were defined as the following: well-circumscribed monolayers of cobblestone-appearing cells, emerging only after 7 and up to 30 days of culture and proliferating for up to 15 passages. Expanded early-passage OEC cultures were verified for endothelial cell phenotype and characteristics. Immunophenotyping by flow cytometry revealed expression of the endothelial cell-surface antigens CD31, CD146, KDR, VE-cadherin, and CD105, but not the hematopoietic cell-specific surface antigen CD14 (Fig. 1A) . Both OECs and control endothelial cells expressed CD34, a cell-surface marker previously identified on endothelial and hematopoietic progenitor cells. Furthermore, OECs incorporated DiI-acetylated low-density lipoprotein (DiI-Ac-LDL), bound lectin from U. europaeus (Fig. 1B) , and formed capillary-like structures in synthetic matrix (Matrigel; BD Biosciences; Fig. 1C ). These phenotypic and functional characteristics were similar for OEC cultures derived from all analyzed groups (data in Fig. 1is from OEC cultures derived from patients with nvAMD). 

The combination of these results also demonstrates that the cells expanded from control and patient peripheral blood were in fact endothelial cells and not hematopoietic or monocytic cells. Furthermore, methods used to define OEC phenotype and function, as well as the results obtained, were in accordance with those used and obtained in other studies. ^{30 31 32 42 43 44}  

It has been suggested recently that human embryonic stem cells may be prone to the development of chromosomal anomalies while in continuous culture in vitro. ⁴⁵ Whether long-term cultures of stem or progenitor cells derived from the adult organism are chromosomally stable is presently unknown. Therefore, the karyotype of eight OEC long-term cultures was analyzed (Cytogenetics Laboratory, Obstetrics and Gynecology, Georgetown University Hospital, Washington, DC) between passages 2 and 15 and was found to be consistent with that in normal males and females, respectively, with no numerical or structural chromosome abnormalities (data not shown). 

All serially passaged OEC cultures eventually demonstrated morphologic changes consistent with senescence (enlarged and flattened morphology of cells, numerous binucleated cells), a marked decrease in proliferation rate and an increase in the proportion of cells positive for senescence-associated β-galactosidase (Fig. 1D) , a histochemical marker of replicative senescence in many different cell types including endothelial cells. ⁴⁶ Cultures were terminated at this stage. 

Difference in the Number of OEC Clusters and Growth Kinetics of OECs between Study Groups

We observed distinct differences in the time of initial OEC cluster appearance, the number of OEC clusters, cluster size, growth rate, and capacity for expansion between cultures derived from different study subjects. Representative photomicrographs of different OEC appearances are given in Figure 2 . Figures 2A and 2Bgive an example of formation of one typical OEC cluster in a single-culture well (Fig. 2B)but not in any of the other inspected wells (Fig. 2A) . Figures 2C and 2Deach show extensive and early OEC cluster formation from another study subject in almost all inspected wells. These clusters tended to proliferate rapidly and merged into confluent cultures. Figure 2Egives an example of a cluster appearing late in culture and presenting a less tightly packed cluster morphology. Figure 2Fgives a representation of an OEC culture appearance after the expansion of the clusters. 

Cultures from all studied subjects were observed for time of initial cluster appearance. No statistically significant difference in days to initial cluster appearance was observed across the disease categories (P = 0.55). Average time of initial cluster appearance was 17.1 ± 1.7 (range, 10–30) days in young control subjects and 14.9 ± 1.2 (range, 9–23) days in non-AMD/low-risk dry AMD, 16.2 ± 2.5 (range, 10–26) days for high-risk dry AMD, and 14.5 ± 1.4 (range, 7–30) in nvAMD subjects. 

To assess the frequency of OEC clusters among different groups, we counted the number of clusters emerging from each culture between day 7 and 30. Figure 3Adisplays the number of OEC clusters observed in all participants. Several subjects had no OEC clusters appear, but this finding was not differentiated by disease group (P = 0.99), as clusters did not develop in 27.8%, 26.7%, 0%, and 28.1% of the young control, non-AMD/low-risk dry AMD, high-risk dry AMD, and nvAMD subjects, respectively. 

Figure 3Bdisplays the final total number of OECs after expansion of the initial OEC clusters (all clusters merged). Overlaid on the plots in Figures 3A and 3Bis a box-and-whisker diagram displaying the median (solid gray horizontal line), quartiles (top and bottom of the box) and 95th quantile (end of whisker) with the mean for each group indicated by the horizontal dashed line. Of interest, a highly elevated number of clusters (9.4–31.2 clusters/20 mL blood) was present in one (16.7%) patient in the high-risk dry AMD group and seven (21.9%) patients with nvAMD, but were not found in any of the other groups. 

Standard logistic analysis of subjects in whom the number of clusters was above or below the upper quintile (and quartile) did not provide significant statistical analysis (data not shown). An analysis of the raw OEC cluster count was therefore undertaken. Most important, this analysis also took into account several variables, particularly factors commonly present in the analyzed groups (e.g., cardiovascular disease or statin use) or disease category (e.g., smoking) that have been shown to influence EPC frequency and function. Results from fitting a regression model of the number of OEC clusters and the number of postexpansion OECs, taking into account all possible variables, including disease categories, are shown in Table 2 . The effects of age, sex, family history of AMD, smoking, presence of cardiovascular disease, and current statin therapy for hyperlipidemia, as well as disease categories (non-AMD/low-risk dry AMD, high-risk dry AMD, and nvAMD) on the observed OEC characteristics are listed, including associated probability and confidence intervals. Results from all subjects, including those with no OEC cluster formation, are included in the analysis. 

There was no detectable difference between the young control subjects and the non-AMD/low-risk dry AMD subjects (P = 0.25). 

None of the clinical variables, including sex and smoking, which were unequally balanced among analyzed groups (Table 1) , significantly affected the number of OEC clusters or the total number of OECs after expansion. However, disease category was a uniquely significant predictor of the number of OEC clusters (P = 0.038), with the nvAMD and high-risk dry AMD categories being significantly different from the non-AMD/low-risk dry AMD category (P = 0.0015 for nvAMD and P = 0.036 for high-risk dry AMD). With adjustment for all other predictors, on average, subjects with nvAMD subjects had 5.11 times as many OEC clusters as subjects with non-AMD /low-risk dry AMD (95% CI 1.12– 28.26). Similarly, on average, subjects with high-risk dry-AMD had 5.62 times as many OEC clusters as those with non-AMD/low-risk dry AMD (95% CI 1.87–13.94). 

A similar analysis of the total number of OECs after cluster expansion was performed. Again, the only significant predictor was disease category, with the average number of cells after expansion in nvAMD being 31.79 times higher than in non-AMD/low-risk dry AMD subject (P = 0.01, 95% CI 2.32–435.0). Although the difference was not statistically significant, there were substantially higher cells after expansion in high-risk dry AMD than in non-AMD/low-risk dry AMD. 

In a further comparison of the proliferative kinetics among groups, we counted the expanded OECs from 21 different cultures at the beginning and end of each passage. Growth kinetic curves for cultures >10⁷ total number of OECs after expansion were chosen for representation of nvAMD (37.5% of cultures) and high-risk dry AMD (83.3% of cultures), whereas growth curves of <10⁷ total number of OECs after expansion were chosen to represent young control (88.9% of OEC cultures) and non-AMD/low-risk dry AMD (86.7% of cultures; Fig. 3C ). Next, PDTs were calculated for each group (Fig. 3D) . Whereas there was a nonsignificant trend toward decreased PDT in nvAMD compared with the young control (P = 0.054) and in high-risk dry AMD compared with non-AMD/low-risk dry AMD (P = 0.079), PDTs were significantly decreased in nvAMD (P = 0.027) compared with that in non-AMD/low-risk dry AMD. 

Finally, growth characteristics were plotted against each other, to determine whether there were positive relationships between time to initial cluster appearance, number of OEC clusters, and number of expandable OECs. Figure 4Adisplays the relationship between time to first appearance and the number of OEC clusters. Figure 4Bdemonstrates a similar, but stronger relationship between the time of first OEC cluster appearance and the total number of OECs after expansion. Figure 4Cdisplays the relationship between the number of OEC clusters and the total number of OECs after expansion. There was a strong relationship between the number of OEC clusters and the total number of OECs arising from those clusters (P < 0.001). However, the curve appeared to be flattening, indicating that there is some threshold number of clusters (∼5) beyond which formation of a large number of OECs cannot be increased by a higher initial number of OEC clusters. Figure 4Ddemonstrates a significant relationship between the number of expandable OECs per cluster and the time to initial cluster appearance. 

Correlation of KDR Expression and OEC Growth Characteristics

We determined whether the different OEC cultures could be distinguished by distinct expression levels of KDR, one of the VEGF receptors known to be expressed in endothelial progenitor cells and to be important in the long-term survival of endothelial ⁴⁷ and hematopoietic progenitor cells. ^{48 49}  

Actively proliferating, nonsenescent OECs were collected from 25 different study subjects, randomly through all control and disease categories, and analyzed by Western blot analysis. A 210-kDa protein corresponding to KDR was detected in HUVECs and to a various extent in all analyzed OECs (Fig. 5A) . A densitometric analysis was performed to quantify KDR expression of OEC cultures compared with HUVECs. 

Figure 5Bdisplays a significant relationship between KDR expression and the number of OEC clusters per 20 mL blood. Using a quadratic curve for a reasonable fit to the data (R ² = 0.46 is satisfactory for only 25 data points), the disease category was a strong influence on the number of OEC clusters, as in Table 2 , and on KDR expression (P = 0.022). Figure 5Cdisplays a significant relationship (P = 0.042) between KDR expression and the total number of OECs after expansion. 

Correlation of VEGF and CRP Levels with OEC Growth Characteristics

Plasma VEGF levels were analyzed in relation to OEC growth characteristics (Figs. 6A 6B) and for each group (Fig. 6C) . No correlation was found between VEGF plasma levels and the initial number of OEC clusters or between VEGF and the final number of expanded cells (Figs. 6A 6B) . VEGF levels did not differ significantly among the groups (Fig. 6C) . 

Similarly, no correlation was found between CRP plasma levels and OEC growth characteristics (Figs. 6C 6D) . Although there was a trend toward higher CRP levels in the old and diseased groups (Fig. 6F) , the levels were not significantly different in the no-AMD/low-risk dry AMD (P = 0.14), high-risk dry AMD (P = 0.08), and nvAMD (P = 0.07) groups when compared to that in the young control group. 

Discussion

We analyzed the feasibility of isolating late OECs from the peripheral blood of old subjects, with or without AMD. We report that OECs were present and expanded in these subjects and that 17% of patients with high-risk dry AMD and 22% of patients with established nvAMD had a high frequency of OEC clusters (9.4–31.2 clusters per 20 mL blood), whereas patients with no AMD or low-risk dry AMD had no or only low OEC cluster formation. In total, 83% of patients with high-risk dry AMD and 37.5% of patients with nvAMD achieved a final culture of more than 10⁷ OECs, whereas only 13% of those with no-AMD/low-risk dry AMD and 11% of young control subjects had cultures of more than 10⁷ expanded OECs. On average, when adjustment was made for all other clinical variables, high-risk dry AMD had a 5.6-fold and nvAMD had a 5.1-fold higher number of OEC clusters per 20 mL blood than did non-AMD/low-risk dry AMD, and high-risk dry AMD generated 63-fold higher and nvAMD 32-fold higher OECs on expansion of clusters than did non-AMD/low-risk dry AMD. These findings should help expand our knowledge on the pathogenic mechanisms of AMD and may be relevant to the diagnosis and treatment of this disease. 

Until recently, CNV in AMD was thought to arise exclusively from cells inside the choriocapillaris. However, we and others have demonstrated that in animal models of nvAMD, bone-marrow–derived cells are recruited from the circulation to participate in CNV and provide up to 50% of the vascular cells. ^{8 9 10 11 12} Formation of nvAMD may therefore be dependent on a combination of two processes: (1) angiogenesis—that is endothelial migration or sprouting from resident mature choroidal blood vessels, and (2) vasculogenesis, defined as recruitment of EPCs from the peripheral circulation to the site of CNV. 

The nature and role of human EPCs in nvAMD has not been studied to date. It even appears that EPCs may not be detectable in the age group susceptible to AMD. ^{19 29} Also, as detailed earlier, there is now much controversy regarding methods traditionally used to identify or quantify EPCs from human origin, such as the eEPC assay or flow cytometric analysis of CD34⁺CD133⁺KDR⁺ cells. This population seems to be composed of hematopoietic cells or progenitors rather than true EPCs. Another functional method consists of the isolation and long-term culture of the highly proliferative late OEC population. However, OECs are relatively well characterized only in cord blood and, to a much lesser extent, in young adult blood, because cord blood OECs are much more frequent and proliferative than those from adult peripheral blood and therefore are easier to investigate. ^{31 33}  

Our study showed that an age-related decline in bone-marrow–derived endothelial cells does not seem to apply to OECs, since the number of OECs in the non-AMD/low-risk dry AMD population (mean age over 70) was very similar to that in our young control group. Culture and growth characteristics such as initial time of OEC cluster appearance and number of OEC clusters in the young control population studied herein were highly consistent with a previous study of OEC cultures from 18 healthy volunteers aged 22 to 50 years. ³¹  

A previous study on the role of OECs in disease demonstrated a transient increase in OEC outgrowth (as well as KDR/CD133-positive EPCs, measured by flow cytometry) in patients who suffer acute vascular trauma. ⁵⁰ Of interest, we also found a significantly increased frequency of OEC clusters in subjects with high-risk dry AMD or nvAMD (with a mean age of ∼75 years). 

The proliferative capacity (total number of OECs after expansion of OEC clusters) of our OEC cultures was also consistent with prior studies. ^{31 43 51} Of interest, large numbers of OEC clusters were more likely to appear quickly and give rise to an elevated number of progeny (as evidenced by a significant relationship between number of clusters and time of initial appearance and total number of expanded OECs and number of clusters). 

Similar, albeit more pronounced, differences in growth characteristics were demonstrated for cord blood, known to be enriched in progenitor cells, ⁵² when compared with blood from healthy adults (aged, 22–50 years). Cord blood gave rise to a higher number of OEC clusters than adult blood, and cord-blood–derived OEC clusters appeared earlier in culture, were larger, and had higher proliferative potential than did adult blood–derived OECs. ³¹ When further from our study are compared to published data, PDTs of OEC cultures derived from young control and non-AMD/low-risk dry AMD were very similar to values given for adult blood, whereas the PDTs for nvAMD were significantly faster and closer to values for cord blood OEC cultures. 

The question remains whether the number of in vitro expanded OEC progeny is simply a function of the initial number of clusters (the more clusters, the more expanded OECs), or whether there are OEC clusters that have more aggressive proliferation potential than others, thus generating the essential part of the progeny, and that are more frequent in one population than another. Furthermore, will every cluster generate OECs that participate in neovascularization, or are there clusters of a particular size, time of appearance, growth rate, and potential, and so forth, that will home to and sustain CNV growth while others will not? Could we eventually differentiate those clusters not only by their morphology and growth characteristics, but also by their phenotype or increased possession of properties vital to the angiogenic process (such as adhesion, migration, or invasion)? What are the mechanisms responsible for the increased presence of this type of OEC in the peripheral blood of patients at risk of or already affected by nvAMD? Most interesting, will it be possible to identify the progenitors to those OECs by using a specific marker or a combination of markers on samples of peripheral blood? To try to address all these question, a single OEC cluster or single-cell culture and clonal analysis (based on the technique described by Ingram, et al. ³¹ ), as well as a detailed characterization (endothelial marker profile, more profound analysis of presence or loss of early stem cell markers, analysis of markers suitable for identifying aggressive OECs, as well as in vivo and in vitro angiogenic assays) of the early unexpanded OEC cluster should be included in future studies. 

Of interest, when we examined the nvAMD group in the present study, we found only a fraction of patients with a high frequency of OEC clusters or a final culture count of more than 10⁷ OECs. This observation begs the question of why all patients do not demonstrate an increased frequency of OECs and a high number of expanded progeny. A possible explanation is that the OEC outgrowth and culture assay may not be sensitive enough to detect differences in all the samples. Alternatively, CNV formation may rely predominantly on EPCs only in a fraction of patients with nvAMD, whereas in others, CNV formation would essentially be derived from the local angiogenic process. 

Finally, production and mobilization of progenitor cells from the bone-marrow into the systemic circulation may be a transient or intermittent process, a rather plausible explanation, because it would be consistent with the process of sudden conversion from dry to nvAMD, and the intermittent activity and self-limited time course of established nvAMD. Unfortunately, the factors that regulate the release and recruitment of OECs have not been studied to date. 

We propose different factors that may be at the basis of OEC involvement in AMD. First, EPC or OEC mobilization from the bone-marrow may be transiently triggered by factors released from the diseased eye into the peripheral blood. In EPCs (as defined by surface marker expression or eEPC assay) several cytokines, for example VEGF, ⁵³ GM-CSF, ⁵⁴ or erythropoietin, ⁵⁵ were shown to affect acute mobilization in animal models of neovascularization. Increased blood levels of these mobilizing cytokines were also detected in acute cardiovascular events ^{2 56 57 58} and increased blood levels of VEGF were shown to parallel an increase in circulating CD34⁺ EPC counts in patients with acute myocardial infarction. ²⁰ In the adult blood, as reported for EPCs, systemic VEGF levels may decline with age. ¹⁹ Circulating VEGF resides in different compartments of the peripheral blood, such as in plasma, platelets, and neutrophils. ⁵⁹ Because VEGF release changes with clotting time of blood samples, reliable measures of VEGF blood levels, especially from serum specimens, are difficult. ⁶⁰ In our study, plasma VEGF levels were similar to sample values indicated for the detection assay and to published results, ^{19 61} but did not significantly differ between groups of different age or disease status. Only one study provides data on VEGF blood levels in patients with AMD, demonstrating mildly increased VEGF in AMD compared with the age-matched control, but not between dry and nvAMD. ⁶² However, investigators seem to have detected VEGF values in serum rather than the indicated plasma samples (values are in the range of sample serum rather than plasma values), making interpretation and comparison to our results difficult. 

Still, findings from both studies do not rule out a central role of systemic VEGF in OEC mobilization, since blood levels are difficult to measure and could peak only transiently preceding conversion from dry to nvAMD or progression of nvAMD, and return to baseline during the remainder of the disease. Alternatively, OEC mobilization could rely on the additive or predominant effect of other factors previously shown to affect mobilization of EPCs (e.g., erythropoietin or GM-CSF). 

Of interest, although plasma levels of VEGF did not correlate with OEC growth characteristics, the number of OEC clusters and the number of OECs after expansion significantly correlated with KDR (the membrane-bound receptor to VEGF) expression levels of OEC cultures (as well as with advanced AMD categories). The apparent discrepancy between VEGF plasma levels and KDR protein expression on OECs can be explained as follows: Signaling through the KDR receptor increases cell survival and maintenance of hematopoietic progenitor cells, ⁴⁸ malignant cells ⁶³ and endothelial cells. ⁴⁷ Increased levels of KDR were found in OEC cultures with increased proliferation (and survival) potential. We therefore propose that OECs with elevated KDR expression either participate in or orchestrate the neovascular progression in nvAMD, because they survive longer within ocular tissues (that locally release VEGF during disease) than OECs with low KDR expression. This fact is especially interesting considering the recent impact of anti-VEGF modalities on the treatment of nvAMD. 

In addition or as an alternative to the cytokine-directed explanation of mobilization, other processes or a combination of such processes as age, cardiovascular disease, genetic alterations or differences, smoking or medication, and chronic infection or systemic inflammation may be responsible for bone-marrow alteration and an increased mobilization of progenitor cells and OECs and thus may trigger conversion from dry to nvAMD. 

Macrophage chemoattractants, elevated levels of inflammatory mediators, and activated complement components are found in drusen samples from patients with AMD, ^{64 65} and VEGF-expressing macrophages aggregate in areas of drusen and areas of established CNV. ^{66 67} In addition, genetic factors, such as polymorphisms of complement factor H ⁶⁸ and CX3CR1, ^{69 70} the chemokine CX3CL1 receptor, are related to inflammation and are risk factors for AMD. In patients with unstable angina, and increased number of eEPCs correlates with increased serum levels of the inflammatory marker CRP. ³⁶ In another study, the eEPC clonogenic potential was related to CRP (and VEGF) levels, even in a healthy population. ⁷¹ Some studies suggest that elevated levels of CRP are also associated with AMD, especially the neovascular form. ^{72 73 74 75} However, although there was a trend toward higher CRP levels in the aged and diseased groups, analysis of participants in this study did not show increased plasma CRP levels paralleling OEC frequency or expansion potential. It is still possible that inflammatory stimuli either precedes OEC mobilization or presents independent of CRP increase. 

Alternatively, infectious agents such Chlamydia pneumoniae (C. pneumoniae), Cytomegalovirus (CMV), and Helicobacter pylori are known to cause chronic infection and persistent inflammation and are possible contributors to the pathogenesis of vascular diseases such as atherosclerosis, ^{76 77} coronary heart disease, ^{76 78} and acute myocardial infarction. ⁷⁸ In line with this hypothesis, a significant serologic association was demonstrated for C. pneumoniae and AMD ⁷⁹ and for CMV and neovascular, but not dry, AMD or control. ⁸⁰ In addition, C. pneumoniae and CMV have been shown to infect and persist in bone marrow and hematopoietic progenitor cells such as CD34-positive cells. ^{81 82} Similarly, chronic smoking can cause stimulation of the bone-marrow with phenotypic changes in leukocytes. ⁸³ In the animal model of CNV, chronic nicotine exposure increased the size and vascularity of CNV. ⁸⁴  

In support of increasing age as a trigger for OEC mobilization, a study showed aged mice to have much greater ability to mobilize hematopoietic stem cells than their young counterparts have. ⁸⁵ In favor of genetic differences, strains of mice with aggressive angiogenic potential had high levels of circulating EPCs. ⁸⁶ In this case, it remains to be determined whether it is the same genetic alteration that increases peripheral angiogenesis and also causes increased bone marrow production and release of EPCs, or whether the higher levels of EPCs actually directly cause the higher level of neovascularization. 

In our study, neither sex, nor age, positive family history for AMD, cardiovascular disease, use of statin medication, nor history of smoking correlated significantly with OEC growth characteristics. Positive serology for infectious agents or genetic differences among study participants were not investigated, but should be the topic of future analysis. 

Finally, based on data derived from our six patients with high-risk dry AMD, it seems that OEC growth characteristics in subjects with dry AMD at elevated, not low risk, of developing nvAMD are similar to those in nvAMD. Therefore, the factor or factors triggering OEC mobilization may already be present in patients at high risk of the development of neovascularization. A larger number of patients with high-risk dry AMD should be examined to confirm our findings and explore possible ocular or systemic factors common to high-risk dry and nvAMD. EPCs may even be involved in the conversion from dry to wet AMD, but this notion is purely speculative. 

In addition to their in vitro and in vivo vasculogenic potential, EPCs have provide evidence of mediating paracrine effects in recent studies. Thus, EPCs of different origin were shown to release growth factors and cytokines relevant to angiogenesis. ^{26 27 32} The secretory potential of OECs should be analyzed to address the question of whether they indirectly support CNV formation, in addition to or instead of contributing directly. 

In conclusion, this is the first study to demonstrate that EPCs, in the form of OECs, can be isolated and expanded successfully from the peripheral blood of aged subjects, particularly patients with AMD. It also provides evidence of differences in OEC growth characteristics depending on the time of OEC cluster appearance in culture, the initial number of clusters, and the KDR (or VEGFR-2) expression of OECs. Of interest, the number of OEC clusters and their expandable progeny seems to increase with advanced stages of AMD. In addition to the observation that bone-marrow–derived cells participate in CNV formation in animals, our results further suggest that EPCs may be involved in the formation and growth of CNV in patients affected by nvAMD. We propose different pathophysiologic mechanisms as a basis for future studies on the involvement of EPCs in AMD. A more detailed analysis of the undifferentiated or unexpanded OEC progenitor or OEC cluster as well as subsequent culture and expansion of OECs and their progeny will provide material for further analysis of the phenotype, genetics, and in vitro and in vivo function of these cells and for identification of the mechanisms involved. 

³  Contributed equally to the work and should be considered equivalent first authors.

 Presented at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2006.

 Supported by Grant BFR 04/109 from the Ministry of Culture, Higher Education and Research of Luxembourg (MT).

 Submitted for publication July 27, 2007; revised November 21, 2007, and January 15, 2008; accepted April 25, 2008.

 Disclosure: M. Thill, None; N.V. Strunnikova, None; M.J. Berna, None; N. Gordiyenko, None; K. Schmid, None; S.W. Cousins, None; D.J.S. Thompson, None; K.G. Csaky, 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.

 

Clinical Characteristics of Study Subjects

Characteristic	Young Control	Non-AMD/Low-Risk Dry AMD*	High-Risk Dry AMD*	nvAMD
Subjects (n)	18	15	6	32
Male	7 (38.9)	3 (20)	4 (66.7)	21 (65.6)
Age (y)
Mean ± SEM	28.6 ± 1.6	71.3 ± 2.6	74.5 ± 2.9	77.1 ± 1.3
(Range)	(19–39)	(55–85)	(63–83)	(63–92)
Race
White	15 (83)	15 (100)	6 (100)	32 (100)
Other, †	3 (7)	0 (0)	0 (0)	0 (0)
Family history of AMD	2 (11)	1 (6.7)	3 (50)	12 (37.5)
Smoking
Current	0 (0)	1 (6.7)	0 (0)	2 (6.3)
Former	0 (0)	3 (20)	2 (33.3)	16 (50)
Current and former combined	0 (0)	4 (26.7)	2 (33.3)	18 (56.3)
Presence of cardiovascular disease	0 (0)	7 (46.7)	3 (50)	19 (59.4)
Current statin therapy for hyperlipidemia	0 (0)	2 (13.3)	3 (50)	13 (40.6)

 Data are the number and the percentage of patients in each category.

*  Low-risk, 0–3 risk factors; high-risk, 4 risk factors; according to the AREDS simplified severity scale for AMD.⁴

†  Includes two people of Indian and one of Chinese origin.

Figure 1.

 

Phenotype and functional analysis of early-passage OEC cultures derived from nvAMD subjects, representative of OEC cultures derived from patients of all analyzed groups. (A) Representative flow cytometry histogram plots for surface antigens in OECs, HUVECs, and HMVEC-dermal cells. (B) Uptake of DiI-Ac-LDL (yellow) and positive staining for Ulex europaeus lectin (green). (C) Representative photomicrograph of tube formation in OECs. (D) Senescent OEC culture as determined by positive SA β-galactosidase staining (blue).

Figure 2.

 

Representative photomicrographs of OEC clusters. Arrowhead: cluster outlines. (A) One culture well exemplar presented a few spindle-shaped single cells derived from a PBMC culture, but no OEC cluster. (B) The same study subject presented a single typical OEC cluster in another observed well at day 16 after initiation of culture. (C, D) Extensive and early OEC cluster formation in another study subject at day 12 with details of two distinct (C) and merging (D) OEC clusters. (E) OEC cluster appearing later in culture (day 17) and presenting a less tightly packed cluster morphology. (F) Final appearance of a representative expanded OEC culture. Magnification, ×50.

Figure 3.

 

Growth characteristics of OEC cultures. (A) Number of OEC clusters isolated per 20 mL of peripheral blood from young control and older subjects without AMD or those with low-risk dry AMD, high-risk dry AMD, or nvAMD. (B) Final number of OECs after expansion. (A, B) Results represented by dot plots. Overlaid on the plots is a box-and-whisker diagram displaying the median (solid gray horizontal line), quartiles (top and bottom of the box), and 95th quantile (end of the whisker), with the mean for each group (horizontal dashed line). (C) Growth kinetic curves for OEC cultures. Growth kinetic curves for cultures containing >10⁷ OECs after expansion were chosen for representation of nvAMD (37.5% of nvAMD cultures) and high-risk dry AMD (83.3% of cultures). Growth curves <10⁷ total number of OECs after expansion were chosen to represent young control (88.9% of cultures) and non-AMD/low-risk dry AMD (86.7% of cultures). (D) PDTs from 21 different subjects randomly selected from all groups. Results represent the mean ± SEM for five young control, four non-AMD/low-risk dry AMD, three high-risk dry AMD, and nine nvAMD. *P = 0.027 (Student’s t-test).

Table 2.

 

Predictors and Their Effect on the Number of OEC Clusters and the Total Number of OECs after Expansion, with Accompanying Significance Level and 95% CI

Characteristic	OEC Clusters/20 mL Blood			Total OECs after Expansion
		Relative Risk	P	Relative Risk	P
Age increase by 1 year	0.99 (0.94, 1.05)	0.85	0.91 (0.74, 1.13)	0.38
Male	1.66 (0.82, 3.36)	0.16	0.34 (0.042, 2.69)	0.31
Family history of AMD	1.67 (0.66, 4.20)	0.27	7.07 (0.32, 157.2)	0.22
Smoking (current or ever)	1.69 (0.55, 5.25)	0.36	19.73 (0.45, 860.4)	0.12
Cardiovascular disease	1.18 (0.44, 3.15)	0.74	0.078 (0.0031, 1.95)	0.12
Current statin therapy for hyperlipidemia	1.00 (0.35, 2.92)	0.99	0.66 (0.0091, 48.33)	0.85
AMD severity
Non-AMD/low-risk dry AMD vs. young control	1.22 (0.09, 15.78)	0.88	9.62 (0.0003, 341415.9)	0.67
Non-AMD/low-risk dry AMD vs. high-risk dry AMD	5.62 (1.12, 28.26)	0.036*	62.87 (0.55, 7235.4)	0.087
Non-AMD/low-risk dry AMD vs. nvAMD	5.11 (1.87, 13.94)	0.0015*	31.79 (2.32, 435.0)	0.01*

*  Significant at the 5% level.

Figure 4.

 

(A) Significant relationship between time to first appearance and the number of OEC clusters. (B) Similar relationship between the time of first OEC cluster appearance and the total number of after expansion OECs. (C) Significant relationship between the initial number of OEC clusters and the final number of expanded OECs.

Figure 5.

 

Western blot analysis for KDR/VEGFR-2 in OEC cultures. (A) Representative extract of Western blot in actively proliferating, nonsenescent OEC cell cultures, as well as in HUVECs collected between passages 3 and 9; top lanes: KDR protein; bottom lanes: corresponding β-actin loading control. (B, C) KDR expression in HUVECs and OECs calculated by band density measurement. (B) Significant relationship between KDR expression and the number of OEC clusters per 20 mL blood, with the disease category being a strong influence on the number of OEC clusters and on KDR expression. (C) Significant relationship between KDR expression and the total number of OECs after expansion.

Figure 6.

 

VEGF and CRP blood levels in study subjects. Lack of significant correlation between the VEGF plasma levels and the number of OEC clusters per 20 mL of blood (A) or total number of expanded OECs (B). (C) VEGF plasma levels by group with no significant difference among groups. Lack of significant correlation between CRP plasma levels and number of OEC clusters per 20 mL of blood (D) or total number of expanded OECs (E). (F) CRP plasma levels by groups with no significant difference in aged or diseased categories when compared with young control (P = 0.14, 0.08, and 0.07 for no AMD/low-risk dry AMD, high-risk dry AMD, and nvAMD, respectively).

The authors thank the nurses and staff of the NIH Eye Clinic and the volunteers and patients who participated in this study.