Chronic oxidative stress upregulates Drusen-related protein expression in adult human RPE stem cell-derived RPE cells: a novel culture model for dry AMD.

PURPOSE
The goal of this study was to examine changes in the expression of transcripts and proteins associated with drusen in Age-related Macular Degeneration (AMD) after exposing human retinal pigment epithelium (hRPE) cells to chronic oxidative stress.


METHODS
Primary adult human RPE cells were isolated from cadaveric donor eyes. The subpopulation of RPE stem cells (RPESCs) was activated, expanded, and then differentiated into RPE progeny. Confluent cultures of RPESC-derived hRPE and ARPE-19 cells were exposed to a regimen of tert-butylhydroperoxide (TBHP) for 1-5 days. After treatment, gene expression was measured by quantitative PCR (qPCR), protein expression was assessed by immunocytochemistry and transepithelial resistance and cell toxicity were measured.


RESULTS
hRPE cells exposed to a regimen of TBHP for 5 days upregulate expression of several molecules identified in drusen, including molecular chaperones and pro-angiogenic factors. 5-day TBHP treatment was significantly more effective than 1-day treatment at eliciting these effects. The extent of hRPE response to 5-day treatment varied significantly between individual donors, nevertheless, 6 transcripts were reliably significantly upregulated. ARPE-19 cells treated with the same 5-day stress regime did not show the same pattern of response and did not upregulate this group of transcripts.


CONCLUSIONS
RPESC-derived hRPE cells change significantly when exposed to repeated oxidative stress conditions, upregulating expression of several drusen-related proteins and transcripts. This is consistent with the hypothesis that hRPE cells are competent to be a source of proteins found in drusen deposits. Our results suggest that donor-specific genetic and environmental factors influence the RPE stress response. ARPE-19 cells appear to be less representative of AMD-like changes than RPESC-derived hRPE. This adult stem cell-based system using chronic TBHP treatment of hRPE represents a novel in vitro model useful for the study of drusen formation and dry AMD pathophysiology.

The cell type responsible for the formation of drusen components has been debated and remains unclear. Electron micrographs have shown that the RPE can exude cell components from their basal membranes on top of and into drusen, suggesting that drusen are, at least in part, the products of RPE cells [15,16]. In a prior study, we found that αB-crystallin was highly expressed in RPE cells immediately adjacent to drusen, further demonstrating that in AMD, the RPE cells change to express proteins associated with the disease [17]. Evidence for RPE production of drusen proteins during AMD progression is accumulating [12,[18][19][20][21]. Cultured fetal human RPE exposed to human serum were recently found to upregulate apolipoprotein E (APOE), APOJ, Vitronectin, and serum amyloid P (SAP), with evidence of complement activation [19].
Chronic oxidative stress is understood to be an important contributing environmental factor to the development of AMD [22]. Single exposure of cultured primary hRPE cells to hydrogen peroxide for 24 hours caused an upregulation of inflammatory factors IL-6, NF-κB, and phosphorylation of p38 MAPK, ERK, and JNK [23]. Other studies have used ARPE-19 cells, a transformed cell line originating from the RPE of a 19year-old adult male donor [24], to study the impact of oxidative stress. Several similarities and differences between hRPE and ARPE-19 have been documented previously [25][26][27]. ARPE-19 cells exposed to single or repeated (up to 5 days) oxidative stress treatment upregulate several anti-apoptotic, inflammatory and DNA-repair proteins, consistent with some of the pathological changes associated with AMD [23,[28][29][30][31][32][33]. However, upregulation of the most common drusenrelated proteins in AMD has not been reported in primary adult hRPE or in ARPE-19 cells in response to oxidative stress.
In this study, we used a recently described adult stem cell present within the human RPE layer to produce homogenous hRPE cultures [34]. These RPESCs, which constitute roughly 10% of primary human RPE isolates, can be activated to self-renew, expanding into a large cell population that can be differentiated into progeny exhibiting the characteristics of mature RPE (hRPE). The ability to generate highly pure RPESC-derived hRPE cultures improves the interpretation of results.
Confluent cultures of RPESC-derived hRPE cells were subjected to a multi-day exposure to TBHP, shown previously to activate senescence in ARPE-19 [28]. TBHP exposure has been demonstrated to disrupt junctional integrity of the RPE and cause lipid peroxidation of the membrane bilayer as well as the oxidation of glutathione, endoplasmic reticulum Ca 2+ release, increased intracellular calcium ([Ca 2+ ] i ), and increased mitochondrial inner membrane permeability [35][36][37][38][39][40][41][42]. TBHP has also been demonstrated to oxidize lipids in the membrane bilayer, thereby inducing membrane leakage leading to apoptosis and cell lysis [43]. We hypothesized that repeated exposure to TBHP would model the chronic oxidative stress previously associated with AMD [22]. Here we compared, for the first time, single and multi-day exposure to TBHP in differentiated RPESC-derived hRPE from 6 donors and the ARPE-19 cell line to ascertain the effect on expression of some of the major proteins associated with drusen and AMD (Appendix A). This large group of proteins and transcripts identified from the literature is hereafter termed 'drusen-related' in this report.
Repeated stress treatment of hRPE in vitro was found to be more effective than a single treatment at inducing upregulation of the major AMD-associated drusenrelated protein transcripts. In contrast, ARPE-19 cells under chronic stress conditions showed a different pattern of response, with many of the major drusen-related protein transcripts being down-regulated. The observation that highly pure populations of hRPE respond in this manner, even in the absence of choroidal endothelial cells or neural retina, supports the evidence that the RPE is an important source of drusen components.

RESULTS
The experimental approach is presented as a diagram (Fig. 1A). hRPE cells were isolated and then plated into 24-well plates (Fig. 1B). The sub-population of RPESCs was activated to self-renew, dividing approximately once per day [34]. The resulting cultures were differentiated by lowering the serum concentration to produce a confluent culture of cells with the cobblestone epithelial morphology of mature hRPE (Fig. 1C) that exhibited a TER measurement of 200-250Ω cm 2 . The total time from initial donor harvest to the initiation of a stress experiment was 60 days (Fig. www.impactaging.com 1A). ARPE-19 cells were cultured in the same conditions as donor-derived hRPESCs and produced monolayers that closely resembled the cobblestone morphology of hRPE cells (Fig. 1E, F).

Changes in cell monolayer integrity in response to oxidative stress
Light microscopic inspection revealed that hRPE and ARPE-19 cell morphology appeared minimally affected by one or two day exposure to TBHP, but after additio-nal treatments, a moderate increase in cell size, loss of epithelial morphology, and disruption in ZO-1 immunoreactivity were observed, as illustrated at day 5 ( Fig. 1D,G,I).
LDH release, which indicates plasma membrane compromise, was greatest following the first TBHP exposure for both hRPE and ARPE-19 cells ( Fig. 2A). hRPE cells released significantly more LDH than ARPE-19 cells during the first two days (Student's ttest, **, p<0.01), but both cell types responded similarly over the next three days of TBHP treatment. www.impactaging.com As a measure of integrity of the hRPE and ARPE-19 cell monolayers, we measured the TER daily, immediately prior to the next TBHP exposure. Substantial resistances (200-250Ω cm 2 ) such as those generated by the hRPE in this system are consistent with the generation of tight junctions [45], as is the observation of ZO-1 immunoreactivity (Fig.1H). Similar significant decreases in TER (Student's t-test, *, p<0.05, **, p<0.01) were observed in hRPE and ARPE-19 cells over the 5-day treatment period (Fig. 2B).

Chronic TBHP treatment upregulates drusenrelated protein and mRNA expression in hRPE cells
The expression levels of 21 previously identified drusen-related protein transcripts (Appendix A) were measured after 1 and 5 days of 500μM TBHP exposure (Fig. 3). A single TBHP exposure was not sufficient to upregulate any drusen-related transcripts significantly when compared to vehicle-treated controls in the hRPE. However, following 5 days of TBHP treatments, 11 drusen-related transcripts were upregulated relative to hRPE treated with TBHP for 1 day (Welsh's t-test, *, p<0.05; **, p<0.01), including some that were unchanged or down-regulated after 1 day.
To analyze total expression response profiles following 1 and 5 days of TBHP treatment across all six hRPE donors, we treated days of treatment as one variable and total change in protein expression (including transcripts that did not change significantly when compared individually) as a second variable, and performed a twoway ANOVA.
This analysis revealed a highly significant difference in the expression level of this group of transcripts between 1 and 5 days of treatment (p<0.01). hRPE (n=4) release a greater percentage of total intracellular LDH in response to TBHP treatment than ARPE-19 cells (n=4) in the first two days, but both cell types responded similarly over the following days of TBHP treatment (Student's t-test, **, p<0.01). (B) The maximum achievable TER of the hRPE was significantly greater than that of the ARPE-19 under these culture conditions (Student's t-test, **, p<0.01). The TER decreased after multiple TBHP treatments (Student's t-test, *, p<0.05, **, p<0.01; Error bars represent SEM).

www.impactaging.com
Interestingly, we found that there was considerable variability between the different donor hRPE expression responses ( Table 2) to 5-day TBHP treatment (Fig. 4). Nevertheless, following 5 days of TBHP, the different donor hRPE cells consistently demonstrated a significant upregulation of 6 transcripts relative to vehicle-treated controls (Fig. 4A). While some of the hRPE upregulated more than these 6 ( Table 2), only APP, APOJ, BACE1, CC9, VEGF A, and Annexin 2 were significantly upregulated across all 6 hRPE donors (Wilcoxon Matched Pairs Test, *, p<0.05). The remaining 15 transcripts showed a variable response depending on the individual donor (Fig. 4B).

hRPE and ARPE-19 cells respond differently to oxidative stress
When ARPE-19 cells were exposed to TBHP for 1 day, we detected only marginal differences in the drusenrelated transcript expression profile (Fig. 5). αAcrystallin and CC9 were significantly upregulated compared to vehicle-treated controls 22 hours following this initial TBHP exposure (Student's t-test, *, p<0.05). However, both of these markers decreased to below control levels after 5 days of treatment (Fig. 5). Only βA4-crystallin and Vitronectin were significantly upregulated by the ARPE-19 compared to vehicletreated controls after 5 days of TBHP treatment (Student's t-test, *, p<0.05). In contrast to the hRPE, the total expression response profile (Appendix A) of the ARPE-19 was not found to differ significantly between 1 and 5 days by two-way ANOVA (p=0.9161).
We then compared the mean expression level of each transcript in hRPE and ARPE-19 cells after 5 days of TBHP. Significant differences were observed in βSand βA4-crystallins, APP, BACE1, VEGF A, PEDF, Annexin 2 (Welsh's t-test, p<0.05), αB-crystallin, APOJ, CC9 (p<0.01), and Vitronectin (p<0.001). We then compared the total drusen-related transcript expression profile (Appendix A) of each individual hRPE cell line to that of ARPE-19 cells after 5 days of TBHP and found that the hRPE expression response differed from that of the ARPE-19 in all cases (2-way ANOVA, p<0.01). This evidence demonstrated that none of the individual hRPE lines tested behaved similarly to ARPE-19 in this expression response assay. . hRPE cells differ in response to acute and chronic oxidative stress. hRPE cells were exposed to 500μM TBHP for 1 and 5 days. qPCR for drusen-related transcripts was controlled with S18 ribosomal RNA and GAPDH as housekeeping genes. None of the transcripts assessed were upregulated by hRPE after 1 day (n=4) of TBHP compared to controls. However, hRPE significantly upregulated 11 drusen-related transcripts following 5 days (n=6) of TBHP compared to hRPE treated for 1 day (Welsh's t-test, *, p<0.05; **, p<0.001). Error bars represent standard error of the mean (SEM). . Individual donor variation in response to 5 days TBHP and a set of 6 consistently upregulated transcripts. hRPE (n=6) were exposed to 500μM TBHP for 2 hours each day for 5 days and mRNA was quantified using qPCR. (A) Relative quantitation identified 6 transcripts with significantly higher mean expression levels after TBHP treatment compared to controls (Wilcoxon Matched Pairs Test *, p<0.05). hRPE mRNA was quantified relative to housekeeping genes (S18 ribosomal RNA and GAPDH) as well as to the mean of vehicle-treated control hRPE for each transcript across all donor cell lines. (B) 15 of the drusen-related transcripts did not show consistent upregulation after TBHP treatment. hRPE donors are indicated by color. The bar represents the geometric mean.
www.impactaging.com Representative photomicrographs of vehicle-treated hRPE from donor 5(A-E) and ARPE-19 cells (K-O). hRPE (F-J) from donor 5 and ARPE-19 (P-T) cells were exposed to 500μM TBHP or vehicle in medium for 2 hours/day for 5 days. Changes in cellular morphology can be seen from observing the changes in distribution of cytoplasmic αB-crystallin (G,Q) and the tight-junction associated marker ZO-1 (H,R). Increased expression of Aβ (F), APOJ (I) and APOE (J) was observed in hRPE exposed to TBHP, but not in ARPE-19. Expression is heterogeneous, with some cells expressing notably more than others in the field. Nuclei are labeled with DAPI (blue). White arrowheads indicate pigment deposits. Scale bar, 100μm. ARPE-19 cells were exposed to 500μM TBHP for 1 and 5 days. qPCR for drusen-related transcripts was controlled with S18 ribosomal RNA and GAPDH as housekeeping genes. ARPE-19 cells significantly upregulated αA-crystallin and CC9 after 1 day of TBHP treatment (Student's t-test, *, p<0.05). After 5 days of TBHP, ARPE-19 cells down-regulated αA-crystallin and CC9 to below control levels and significantly upregulated βA4-crystallin and Vitronectin (*, p<0.05). Error bars represent standard error of the mean (SEM). www.impactaging.com

TBHP treatment results in altered drusen-related protein expression
To confirm that this treatment elicited changes in drusen-related protein expression, hRPE and ARPE-19 cultures were fixed following a 5-day TBHP exposure or vehicle and probed side by side for several of the drusen-related proteins (Fig. 6). hRPE and ARPE-19 cells exposed to 5 days of vehicle control typically had a consistent epithelial morphology with uniform ZO-1 immunoreactivity (Fig. 6C,M; ZO-1, red).
αBcrystallin (Fig. 6B,L; green) was expressed in the cytoplasm in many untreated cells. Aβ (Fig. 6A,K), APOJ (Fig. 6D,N), and APOE (Fig. 6E,O) were expressed at relatively low levels in control conditions. In hRPE (Fig. 6F-J) and ARPE-19 ( Fig. 6P-T) cultures exposed to TBHP for 5 days, the ZO-1 immunoreactivity (red) was irregular (Fig. 6H) or absent (Fig.  6R), suggestive of disrupted tight junctions. Altered cellular morphology was further illustrated by the irregular distribution of αB-crystallin (Fig. 6G,Q), which appeared to be upregulated in some cells and not others. Expression of Aβ ( Fig. 6F,P) and APOJ (Fig.  6I,S) was increased heterogeneously in the hRPE and decreased in the ARPE-19 cells. While the observed increase in Aβ expression in hRPE was primarily localized to the nucleus, extracellular and cytoplasmic deposits were also found (Fig. 6F). Expression of APOE (Fig. 6J,T) was not generally upregulated in this model, with the exception of hRPE from donor 5 (Fig. 6J). Visualization of APOE expression was minimal in the ARPE-19 cells (Fig 6O,T) under stress and control conditions. Hence, changes in the protein expression mirrored changes in transcript expression observed after TBHP exposure.

DISCUSSION
This study demonstrates for the first time that hRPE cells respond to repeated TBHP exposure by changing to upregulate several proteins commonly found in AMD-associated drusen. These results establish the novel, RPESC-based hRPE in vitro system as a model of drusen formation and dry AMD with potential value as a tool for drug discovery.
Significant risk factors for the development of AMD include age, cigarette smoking, high body mass index (BMI) and consumption of high fat diets, all of which increase chronic stress to the body [52][53][54][55][56][57]. The RPE layer remains quiescent during life; the same cells present at birth are generally exposed to ≥50 years of oxidative damage from the environment before the onset of AMD. While it has long been suspected that the RPE cell layer expresses many of the proteins detected in drusen [12,[15][16][17][18][19], the specific cell type or types responsible for the deposition of most of these proteins within drusen remains undetermined. Here we show that chronic exposure of hRPE cells to 500μM TBHP for 2 hours/day for 1-5 days induced cell dysfunction and damage that manifested in the upregulation of several drusen-related proteins, disrupted cell morphology, decreased cell viability and decreased epithelial integrity, all of which resemble the pathology associated with RPE dysfunction in AMD.
We assessed the TER and the level of LDH in the media comparing TBHP stress to vehicle-treated controls for hRPE cells. Interestingly, the large change in LDH release on day 1 (49% of the total available), which indicates severely compromised membrane integrity, appears at variance with the relatively small observed decrease in TER (12%). This apparent inconsistency could be explained if the mechanism of LDH release was temporary, for example via membrane blebs or transient pores present during TBHP exposure, followed by membrane re-sealing prior to the TER measurement. The precise mechanism of hRPE LDH release under TBHP conditions is not completely understood and is worthy of further investigation.
Upregulated expression of many drusen-related protein transcripts in stressed hRPE occurred in a timedependent manner. Moreover, the pattern of expression was heterogeneous, with some cells expressing more protein than others. This observation is interesting in light of the heterogeneous distribution of drusen deposits observed in vivo. Similarly, in a prior study [17], we found αB-crystallin to be expressed heterogeneously in RPE cells surrounding drusen in AMD patients. The mechanisms behind these heterogeneous focal points of expression within the epithelial layer are unclear and remain an important point of future study.
The substantial variability in expression patterns observed between hRPE cell lines, presented in Table 2 and Figure 4, was surprising. The unique genetic makeup and history of exposure to environmental stress are likely contributors to the variable response of individual hRPE preparations in vitro [58], suggesting that studies of the RPE stress response should utilize several donors to account for donor variability. In addition, as our approach utilizes the sub-population of RPESCs as precursors for the hRPE cultures, some of the variation observed could be due to the particular sub-population of cells activated.
To further examine the variability in expression responses, we investigated whether the mean magnitude www.impactaging.com of drusen-related transcript expression change by the RPE was related to the age of the donors (Table 1). Age was not found to correlate with expression, but our study was limited by a sample size of six. Given a larger cohort of human donors, age, genetic, and environmental factors that have a significant impact on the expression of the transcripts might be identifiable using this model. Despite the variability in individual hRPE response, we were able to identify a set of six drusen-related transcripts that were most likely to be upregulated by hRPE following a 5-day TBHP exposure across all individuals: APP, BACE1, APOJ, CC9, VEGF A, and Annexin 2. This set could be used to develop a reliable screen for AMD using donor-derived hRPE cells.
Of the six-transcript set, two, APP and BACE1, are involved in production of Aβ. Indeed, in this RPESCderived hRPE disease model, we found consistent increases in Aβ peptide (1-40 and 1-42 isoforms) via immunocytochemistry, and in the mRNA transcripts of its precursor (APP) and processing enzyme (BACE1). Previous groups have demonstrated that Aβ is expressed in drusen [13,14,59,60], preferentially from patients with an AMD diagnosis [13], and that Aβ is cytotoxic to RPE cells and causes aberrant complement system activation [61][62][63][64]. While Aβ is traditionally thought of as a secreted or cytoplasmic product, we found a strong nuclear localization. This is consistent with prior studies showing that oxidative stress-induced DNA damage can lead to nuclear localization of intracellular Aβ42, binding to the p53 promoter, and apoptosis in guinea-pig primary neurons [65]. Hence, it is possible that there is a nuclear mechanism for Aβ-mediated RPE cell death and dysfunction.
The set also includes CC9, and while it is known that Aβ positively regulates the complement cascade [66,67] and CC9 expression [68] in other systems, this is the first time that the upregulation of CC9 by chronic TBHP exposure has been demonstrated, supporting the role of oxidative stress in modulating the complement system in hRPE. Annexin 2, a calcium-regulated membrane-binding protein, and APOJ (Clusterin) a molecular chaperone, were also among the set of 6 transcripts reliably upregulated by chronic TBHP in this study, although the mechanism responsible is unknown.
The sixth transcript of the set is VEGFA, known for its important pro-angiogenic role and contribution to the progression of AMD to the wet or neovascular form. While multiple pathways can lead to VEGF A upregulation in AMD, its upregulation in this model is possibly via TBHP-mediated upregulation of AP-1 [69][70][71][72][73][74] which is known to induce VEGF A [75]. Interestingly, both complement activation [76,77] and Annexin 2 activity [78] have been observed to increase VEGF A expression, suggesting that the set of 6 transcripts could positively interact to help produce a consistent upregulation of the set.
It was unexpected that αB-crystallin and TIMP3 were not consistently upregulated in hRPE cells ( Table 2) as both of these proteins have been strongly associated with drusen [11,17]. Perhaps regulation of TIMP3 and αB-crystallin expression depends on signals other than oxidative stress, such as those from the choroidal vasculature or the retina.
Notably, our data show that the response of hRPE and ARPE-19 cells to the same oxidative stress regimen is significantly different, further reinforcing prior studies Fold change in drusen-related mRNA expression in hRPE cells organized by source following 5 days of TBHP treatment for 2 hours/day. Drusen-related transcript expression differs by individual. Gene expression is relative to the housekeeping genes S18 and GAPDH as well as to vehicle-treated control hRPE from the same donor.   26,79]. The ARPE-19 line was derived from the RPE of a 19-year-old male donor, raising the possibility that reduced lifetime environmental stress has resulted in a blunted stress response [24]. Other possibilities are that the mutations involved in the spontaneous immortalization of the ARPE-19 contribute to their resistance to oxidative stress, or that the ARPE-19 cell line may have come from a donor that was genetically less susceptible to oxidative damage. Our study suggests that ARPE-19 should be used cautiously for AMD disease modeling.

αA-crystallin αB-crystallin βB1-crystallin βB2-crystallin βS-crystallin βA4-crystallin
These findings are consistent with the hypothesis that cumulative oxidative stress is an important pathophysiologic mechanism of drusen formation and AMD in the elderly. Further exploration and modulation of the stress response in this hRPE model may provide new targets for the development of novel therapeutics to slow the development of drusen as well as the advancement of dry AMD.  [45] containing 10% fetal bovine serum (FBS), and plated at a density of approximately 150,000 cells/well on human placental extracellular matrix-(ECM, 10μg/ml, BD Biosciences) coated 24-well plates (Corning) (Fig. 1A).

Appendix A. Quantitative PCR Primer Sequences
Once the primary (passage zero) cells reached confluence (20 days), the FBS concentration was reduced to 5%.
The expansion protocol involved dissociating primary cultures in 0.25% trypsin and plating onto ECM-coated 24-well plates (Corning) or 12-well polyester transwell membrane inserts (Costar, 0.4μm); thus the minor population of RPESCs was activated to proliferate and self-renew [34]. Quantitative PCR (qPCR). Purified total RNA was converted to cDNA using the high-capacity RNA to cDNA conversion kit (Applied Biosystems) followed by qPCR with gene-specific primers using a SYBR green reporter kit (Applied Biosystems). qPCR primers (Appendix A) were designed using the PrimerQuest online software from Integrated DNA Technologies and were designed to span exon-exon junctions within the coding region. Additional primers (Annexin 2, βB1crystallin, PEDF, and BACE1) were obtained from the PrimerBank database, Harvard Medical School [46][47][48].
Data Analysis. Graphpad Prism version 5.0b was used to generate graphs and analyze cytotoxicity, TER, and mRNA expression data. Error bars in all graphical figures represent the standard error of the mean (SEM).
To visually display the different RPE conditions and sources on the same scale in Figures 3, the expression levels of each hRPE sample were first quantified relative to the mean delCT (the CT normalized to housekeeping genes) of the vehicle-treated control RPE from the same donor prior to comparing to other donors ( Table 2). ARPE-19 cell expression data is graphically represented in Figure 5 in the same way.
To represent the variance in expression responses between the donors in Figure 4, the mean delCT values of TBHP or vehicle-treated control hRPE from each donor were quantified relative to the mean delCT of all untreated control hRPE from all donors for each mRNA transcript. Significance was determined using the nonparametric Wilcoxon Matched Pairs Test for comparing paired values of different numerical ranges, the Student's t-test, Welsh's t-test, and the two-way analysis of variance (ANOVA).