Chemokine Expression in Retinal Pigment Epithelial ARPE-19 Cells in Response to Coculture with Activated T Cells

Abstract

Purpose.: To investigate the effects of T-cell–derived cytokines on gene and protein expression of chemokines in a human RPE cell line (ARPE-19).

Methods.: We used an in vitro coculture system in which the RPE and CD3/CD28–activated T-cells were separated by a membrane. RPE cell expression of chemokine genes was quantified using three different types of microarrays. Protein expression was determined by single and multiplex ELISA and immunoblotting.

Results.: Coculture with activated T-cells increased RPE mRNA and protein expression of chemokines CCL2 (MCP-1); CCL5 (RANTES); CCL7 (MCP-3); CCL8 (MCP-2); CXCL1 (GRO-α); IL8 (CXCL8); CXCL9 (MIG); CXCL10 (IP10); CXCL11 (ITAC); and CX3CL1 (fractalkine). CCL7, CXCL9, CXCL10, and CXCL11 were secreted significantly more in the apical direction. Using recombinant human cytokines and neutralizing antibodies, we identified IFNγ and TNFα as the two major T-cell–derived cytokines responsible for the RPE response. For CCL5, CXCL9, CXCL10, CXCL11, CXCL16, and CX3CL1, we observed a synergistic effect of IFNγ and TNFα in combination. CCL20, CXCL1, CXCL6, and IL8 were negatively regulated by IFNγ.

Conclusions.: RPE cells responded to exposure to T-cell–derived cytokines by upregulating expression of multiple chemokines related to microglial, T-cell, and monocyte chemotaxis and activation. This inflammatory stress response may have implications for immune homeostasis in the retina, and for the further understanding of inflammatory ocular diseases such as uveitis and AMD.

Introduction

Interactions between the eye and the immune system lie at the heart of many ocular diseases such as uveitis, diabetic retinopathy, and AMD. ¹ The healthy retina balances between providing immunosuppressive signals to maintain the ocular immune privilege that is important for normal vision, and protecting the ocular tissues from infection by expressing inflammatory factors. The RPE forms the outer blood-retinal barrier (BRB) and plays a major role in ocular immune regulation and the homeostasis of the retina. ^2–4  

Few immune cells are present in the healthy retina. ⁵ Retinal microglia are the resident inflammatory cells, and are normally kept in a deactivated state by TGFβ1 secreted from RPE cells. ⁶ They migrate into the retinal tissue during late embryonic development and are renewed by homeostatic proliferation. ⁶ During aging, microglia migrate from the inner retina to the subretinal space next to the RPE cells, where they achieve a state of activation. ^7,8 Peripheral monocytes do not pass the BRB of healthy retinas; but if the barrier is damaged, monocytes can migrate into the retina and differentiate into microglia. ^9–11 Leukocyte extravasation from the choriocapillaris into the retina is observed in aging, possibly reflecting age-related breakdown of the BRB. ¹²  

Chemokines are small cytokines involved in regulation of leukocyte trafficking and activation. They may be divided into four families based on the motif of their conserved cysteines: CXC, CC, XC, and CX3C. ¹³ Chemokines are highly redundant and their effects depend upon which chemokine receptors are present on local cells. Recent studies have shown that functionally distinct subtypes of monocytes or microglia differ in their expression of chemokine receptors. ¹⁴  

The aim of our study was to investigate the chemokine response in RPE cells, using the human RPE cell line ARPE-19 for in vitro coculture with activated T-cells, and to identify which T-cell–derived cytokines were responsible. The RPE is important for the immune status of the outer retina, and is in close proximity with bloodborne cytokines and peripheral blood cells. ³ Knowledge of the chemokine profile in RPE cells could help us understand how a systemic or local inflammation could affect leukocyte attraction to and activation in the retina. Our results show that the RPE responds to inflammatory stimuli by expressing a panel of chemokines, suggesting an important role in the regulation of resident retinal microglia and peripheral monocytes, T-cells, and/or NK cells during retinal inflammation. 

Materials and Methods

Cell Culture

The adult human RPE cell line ARPE-19 (American Type Culture Collection) was cultured in 6-well cell culture plates or on 0.2-μm membrane inserts (Anopore; Nunc A/S, Roskilde, Denmark) >6 weeks, until pigmented monolayers containing approximately 1.6 million cells had formed. Cells were initially cultured in Dulbecco's modified Eagle's medium with 10% fetal calf serum (BioWhittaker; Lonza Group Ltd., Basel, Switzerland), 300 μg/mL L-glutamine, 50 μg/mL gentamicin, and 2.5 μg/mL amphotericin B (all from Life Technologies, Carlsbad, CA). Medium was gradually changed to X-vivo 15 serum-free medium (Lonza Group Ltd.) containing 300 μg/mL L-glutamine, 100 μg/mL streptomycin, and 2.5 μg/mL amphotericin B, 2 to 4 weeks before experiments. 

T-Cell Purification and RPE–T-Cell Coculture

T-cells were purified from fresh whole blood from healthy, young volunteers, as previously described. ¹⁵ Verbal consent to blood sampling was considered adequate by the local ethics committee, and was obtained. T-cells were cultured in inorganic membrane, 0.2-μm membrane inserts (Nunc) over RPE cells in the ratio 2.5 T-cells to 1 RPE cell (apical coculture) or RPE cells were grown on the inserts and T-cells added to the bottom compartment (basolateral coculture) as previously described. ¹⁵ This ratio was chosen to achieve cytokine concentrations in the supernatant comparable with concentrations in vivo in areas with high local T-cell density. At the end of coculture, media was gathered and RPE cells were removed using a cell scraper. RNA or cell lysates for protein studies was prepared as previously described. ¹⁵ A different donor was used for each independent replicate. 

Recombinant Cytokines and Neutralizing Antibodies

RPE cells were grown on membrane inserts and cytokines added to the basolateral compartment in the concentration 200 ng/mL IFNγ, 40 ng/mL TNFα and/or 50 ng/mL lymphotoxin (LT)-α (all from R&D Systems, Minneapolis, MN). These concentrations were found by ELISA to yield approximately the same concentrations after 48 hours as found in 48-hour coculture media from RPE cells with activated T-cells (data not shown). For studies with neutralizing antibodies, RPE cells were cocultured with CD3/CD28-activated T-cells with 40 μg/mL mouse IgG2a anti-human IFNγ (R&D Systems) and/or 118 μg/mL anti-TNFα (Infliximab, Remicade; Janssen Biologics, Leiden, Netherlands). In a preliminary experiment, we found no effect of isotype controls (R&D Systems) on RPE cell gene expression using a DNA microarray (GEArray; SABiosciences, Valencia, CA; data not shown). Media was collected and RNA was purified as above. 

Microarrays

Three different oligonucleotide microarray types were used: two genome-wide microarray types from Affymetrix (Human Genome U133 Plus 2.0 and Human Gene 1.0 ST; Santa Clara, CA); and the pathway-specific DNA microarray (Oligo GEArray OHS810; SABiosciences), including 440 immunology-related genes. 

Whole genome DNA microarrays (Affymetrix GeneChips; Affymetrix) were processed at the Copenhagen University Hospital Microarray Center. Labeling of RNA samples was done according to Affymetrix protocol. For the U133 Plus 2.0 GeneChips (Affymetrix), data analysis was performed as previously described. ¹⁵ Among the identified differentially expressed genes were several chemokine genes, and the chemokine network was chosen for further investigation. Probe sets in genes of interest were matched against UCSC genome build 18 to verify that they were unique and targeted the genes in question. This was carried out using the genome software package ¹⁶ (GeneRegionScan; Fred Hutchinson Cancer Research Center, Seattle, WA) from the genomic data analysis software repository (Bioconductor; Fred Hutchinson Cancer Research Center). For the 1.0 ST GeneChips (Affymetrix), cell files were processed using the robust multichip average (RMA) algorithm in the expression array software (Expression Console version 1.1; Affymetrix). All microarray data are MIAME compliant, and raw data have been deposited in the Gene Expression Omnibus (GEO, accession numbers GSE17938 and GSE36331). Pathway-specific DNA microarrays (SABiosciences) were processed as previously described. ¹⁵  

ELISA

Secreted cytokines were quantified by sandwich ELISA. The standard curve was generated from recombinant proteins (BD Pharmingen, BD Biosciences, Rockville, MD). The following capture and biotinylated detection antibodies were used (all from BD Biosciences): IFNγ (2 μg/mL mouse mAb and 0.5 μg/mL biotinylated mouse mAb); TNFα (2 μg/mL mouse mAb and 0.5 μg/mL biotinylated mouse mAb). ELISA plates (Maxisorp; Nalge Nunc International) were coated with capture antibody in 0.1 M Na₂PO₄, pH 9.0 by incubating overnight at 4°C, and blocked with PBS with 10% FCS, pH 7.4 (30 minutes at RT). Samples and standards were added and incubated overnight at 4°C, then with biotinylated detection antibody in PBS with 10% FCS and 0.05% Tween-20, pH 7.4 (1 hour at RT), followed by HRP-conjugated streptavidin (BD Biosciences) 1:1000 in PBS with 10% FCS and 0.05% Tween-20, pH 7.4 (30 minutes at RT), and OPD substrate (DakoCytomation) according to manufacturer's recommendations. Absorbance was measured on a microplate reader (Discovery HT-R; BioTek Instruments, Inc., Winooski, VT) at 490 nm. For LTα (also known as TNFβ) and CXCL11 (IP-9/ITAC), the ELISA kits (Quantikine; R&D Systems) for TNFβ and CXCL11 were used. 

Multiplex ELISA

The remaining secreted chemokines were quantified with the multiplex assays (Luminex Chemokine Human 10-Plex Panel; Life Technologies). This detects CCL2 (MCP-1); CCL3 (MIP-1α); CCL4 (MIP-1β); CCL5 (RANTES); CCL7 (MCP-3) CCL8 (MCP2); CCL11 (eotaxin); CXCL1 (GRO-α); CXCL9 (MIG); and CXCL10 (IP-10). Manufacturer's instructions were followed. Briefly, we first performed a preliminary experiment to determine optimal dilutions for each sample type that yielded concentrations within the range of the standard curve. Supernatants from untreated RPE cells were subsequently assayed at 2× dilution, activated T-cells at 2× and 20× dilution, and coculture supernatants at 2×, 20×, and 100× dilutions. For each chemokine, the dilution that yielded a concentration within the standard curve was used. The assay was performed with duplicate measurements on 2 to 3 independent replicates. 

Immunoblotting

Membrane-bound CX3CL1 in RPE cell lysates and secreted IL8 in supernatants were detected by immunoblots as previously described. ¹⁵ The antibodies used were goat-anti-IL8 (0.2 μg/mL; R&D Systems) goat-anti-CX3CL1 (0.2 μg/mL; R&D Systems), horseradish-peroxidase (HRP)-conjugated rabbit anti-goat IgG (1:1000; R&D Systems) mouse-anti-beta-actin (1:10,000; Sigma, St. Louis, MO), and HRP-conjugated donkey anti-mouse IgG (1:1000; R&D Systems). 

Statistics

Statistical analyses were performed using graphing and statistical software (GraphPad Prism; GraphPad Software, La Jolla, CA), and P values less than 0.05 were considered significant. The 19 and 16 chemokine genes expressed on the U133 Plus 2.0 GeneChips (Affymetrix) and pathway-specific DNA microarray (SABiosciences), were analyzed with one-way ANOVA to test for overall differences between groups, and Student's t-tests for comparison of the groups RPE/- versus RPE/T. The 16 expressed chemokine genes on the 1.0 ST GeneChips (Affymetrix) and the multiplex ELISA were analyzed with one-way ANOVA and Tukey's multiple comparison tests to compare all eight and five groups, respectively.The singleplex ELISAs and the immunoblots were analyzed with t-tests. 

Results

Increased Chemokine Gene Expression in RPE Cells after T-Cell Coculture

To determine whether coculture with activated T-cells had an impact on chemokine regulation in RPE cells, we cultured RPE cells in membrane inserts alone or with activated T-cells added basolaterally for 48 hours, using a ratio of 2.5 T-cells per RPE cell. Chemokine gene expression in RPE cells was characterized with U133 Plus 2.0 GeneChips (Affymetrix; n = 2). Forty-two genes encoding chemokines were identified on the microarray, and the analysis was limited to these 42 genes. An expression level above the median level of expression for all genes on the array was considered to be of functional importance. A total of 19 chemokine genes were expressed in RPE cells and/or activated T-cells cultured either alone (RPE/-, T/-) or together (RPE/T, T/RPE; Fig. 1): CC-chemokines 2, 3, 4, 5, 7, 8, and 20; CXC-chemokines 1, 2, 3, 6, 8, 9, 10, 11, and 16; XC-chemokines 1 and 2; and CX3CL1. Overall chemokine gene expression was significantly changed (P < 0.001 in one-way ANOVA), and 11 genes were significantly upregulated in RPE cells after coculture. To test whether the polarized RPE cells responded differently to apical (rather than basolateral), exposure to T-cell–derived cytokines, additional experimental groups were included, where RPE cells were grown on plate bottoms, with or without T-cells added to the membrane insert. Apical T-cell exposure had a similar effect on chemokine gene expression in RPE cells as basolateral exposure (see Supplementary Material and Supplementary Fig. S1A). 

Figure 1. 

 

RPE cell gene expression of chemokines in response to basolateral co-culture with activated T-cells.

 

RPE cells were cultured in membrane inserts alone (RPE/-), or with activated T-cells added to the basolateral compartment (RPE/T), in the ratio 2.5 T-cells per RPE cell. T-cells were cultured alone (T/-) or in inserts above RPE cells (T/RPE). After 48 hours, RNA was analyzed by U133 Plus 2.0 GeneChips (Affymetrix). *P < 0.05. **P < 0.01. ***P < 0.001.).

To validate human genome expression array (Affymetrix) results, pathway-specific DNA microarrays (SABiosciences) that included probes for 36 chemokine genes, were used to analyze RPE cells exposed apically to activated T-cells in the ratio 1:2.5 (n = 3–8; see Supplementary Material and Supplementary Fig. S1B). The results confirmed our findings. Further, we observed a clear dose-response tendency in RPE cells cultured with sequentially lower numbers of T-cells (n = 1–2; see Supplementary Material and Supplementary Fig. S1B), and found that activation of the cocultured T-cells was important for the RPE cell gene expression, as resting T-cells did not induce chemokine gene expression in RPE cells (n = 3; see Supplementary Material and Supplementary Fig. S1B). 

Chemokine Protein Expression Correlates with Gene Expression

In order to confirm the chemokine expression that was significantly increased in either of the three microarray analyses, immunoblotting, multiplex, and singleplex ELISA were performed. Secretion of chemokines to the culture media was measured quantitatively using multiplex ELISA (CCL2, CCL3, CCL4, CCL5, CCL7, CCL8, CXCL1, CXCL9, and CXCL10; n = 2–3 independent replicates measured in duplicate); singleplex ELISA (CXCL11, n = 5–6 independent replicates measured in duplicate); and semiquantitatively with immunoblotting (IL8, n = 4; Figs. 2A, 2B). RPE cells were cultured on membrane inserts, and T-cells added to the basolateral side in the ratio 2.5 activated T-cells: one RPE cell. Media was collected separately from the apical and basolateral side of the RPE cells. The membrane-bound chemokine CX3CL1 was measured in whole RPE cell lysates (n = 2, Fig. 2B), as well as in culture supernatant (data not shown). The amounts produced in cocultures of RPE cells with activated T-cells was significantly higher than that produced in RPE cells or activated T-cells alone, for all 12 chemokines measured. CCL3 and CCL4 were mainly expressed by the activated T-cells and were found in significantly higher amounts on the basolateral side of the RPE cells. Both RPE and T-cells expressed the CCL5 gene, and both cell types likely contributed to the amount of protein found in the medium. The T-cells did not express significant amounts of the other chemokine genes (Fig. 1). CCL7, CXCL9, CXCL10, and CXCL11 were secreted in significantly higher amounts from the apical side of the RPE cell, while the remaining chemokines were secreted in equal quantities to both sides of the cell (Fig. 2C). CCL11 was also detected in coculture supernatants in very low amounts (data not shown). The commercially available antibodies to CXCL2 and CXCL3 were either not sensitive enough to measure the small amounts produced, or cross-reactive, and these two chemokines were not measured. 

Figure 2. 

 

Polarized expression of chemokine proteins from RPE cells. RPE cells were cultured alone (RPE/-), or cocultured with activated T-cells in the ratio 2.5 T-cells to 1 RPE cell in membrane inserts (RPE/T). After 48 hours, media was collected. (A) Secretion of CCL2, CCL3, CCL4, CCL5, CCL7; and CCL8, CXCL1, CXCL9, and CXCL10 was quantified by multiplex ELISA; and CXCL11 with singleplex ELISA. (B) Secretion of IL8 and membrane expression of CX3CL1 was determined semiquantitatively by immunoblotting of conditioned media and whole RPE lysates. Blots are representative of 2 to 4 independent experiments, with all replicates analyzed densitometrically. (C) Sketch summarizing results of chemokine secretion from RPE cells. The size of the arrow indicates amount of chemokine secretion. The arrow size for IL8 is estimated based on the detection limit (5 ng/lane) of the antibody used for the immunoblot.

 

Data are shown as mean with error bars indicating standard deviation (SD). *P < 0.05. **P < 0.01. ***P < 0.001.

IFNγ, LTα, and TNFα from T Cells Identified as Potential Cause of RPE Chemokine Expression

To determine the identity of the T-cell–derived cytokines influencing RPE cell chemokine expression, gene expression of 174 cytokines and their receptors was analyzed using U133 Plus 2.0 GeneChips (Affymetrix) data from RPE and T-cells (see Supplementary Material and Supplementary Table S1).Seven cytokine genes expressed in activated T-cells, whose corresponding receptors were expressed in RPE cells, were identified (Table 1). The most highly expressed of these were IFNG, TNFA, and LTA (also known as TNFB). Additionally, LTB, which after translation forms heterotrimers with LTα and binds the LTβ-receptor, ^17,18 was also expressed in T-cells. Likewise, FASL (CD95L) was expressed at very low levels in activated T-cells after coculture, and its receptor, FAS (CD95), was expressed in RPE; and TRAIL and its receptor gene, TRAILR2, were expressed in both T and RPE cells. However, these forms of lymphotoxin, FASL, and TRAIL are membrane-anchored and should not play a role in our transwell system. Finally, TGFB1 and its receptors TGFBR1, -2, and -3, were expressed in low levels in both T and RPE cells. RPE cells cultured alone also expressed both these ligands and receptor genes; however, it is unlikely that this cytokine could trigger the many observed changes in RPE chemokine expression. Secretion of IFNγ, TNFα, and LTα from RPE cells, activated T-cells, and cocultured cells was analyzed by ELISA on conditioned media (Fig. 3). RPE cells cultured alone did not secrete any of the three cytokines. Activated T-cells cultured alone secreted all three cytokines, and TNFα showed no change in secretion levels, while LTα showed a tendency toward increased secretion. IFNγ secretion was significantly increased after coculture with RPE cells. 

Figure 3. 

 

Expression of IFNγ, TNFα, and LTα from T-cells.

 

RPE cells and activated T-cells were cultured separately (RPE/- and T/-, respectively) or in coculture in the ratio 1 RPE cell to 2.5 T-cells separated by membrane inserts (RPE/T) for 48 hours, and media collected. The secretion of cytokines IFNγ, TNFα, and LTα by both cell types was quantified by ELISA. Each point is the mean of duplicate measurements. *P < 0.01. **P < 0.001.

Table 1. 

 

Cytokine/Receptor Pairs Expressed in T and RPE Cells

Gene	RPE/-	RPE/T	T/-	T/RPE
IFNG	—	—	20,500	36,200
IFNGR1	3,500	7,700	2,300	5,100
IFNGR2	2,800	9,400	(+)	(+)
TNFSF1: LTA, TNFβ	—	—	7,500	14,900
TNFSF2: TNFα	—	(+)	7,800	9,700
TNFRSF1A: TNFR1	1,600	3,700	(+)	(+)
TNFRSF1B: TNFR2	—	1,100	1,700	1,700
TNFRSF14: HVEM	1,600	1,800	1,000	800
TNFSF3: LTB, TNFc	—	—	5,200	7,100
TNFRSF3: LTBR	1,200	1,000	—	—
TNFSF6: FASL	—	—	1,100	1,300
TNFRSF6: FAS	3,800	9,900	2,400	2,700
TNFRSF6B: DCR3	—	—	—	(+)
TNFSF10: TRAIL	—	10,700	3,500	2,100
TNFRSF10A: TRAILR1	(+)	(+)	(+)	(+)
TNFRSF10B: TRAILR2	6,700	5,500	1,100	2,200
TNFRSF10C: TRAILR3	—	—	—	—
TNFRSF10D: TRAILR4	—	—	—	(+)
TNFRSF11B: OPG	—	(+)	—	—
TGFB1	1,100	1,600	2,300	1,800
TGFBR1	2,700	2,400	2,300	2,300
TGFBR2	2,600	(+)	2,800	2,300
TGFBR3	3,000	2,000	1,600	(+)

 Gene expression values of 174 cytokines were examined in U133 Plus 2.0 GeneChips (Affymetrix) data of RPE cells cultured alone (RPE/-) or with activated T-cells (RPE/T); activated T-cells cultured alone (T/-), or with RPE cells (T/RPE). Genes with average expression above the median value of all genes were considered functionally expressed. Expression values well below the median value are denoted as (-), values 25% to 100% of the median value are denoted as (+). Ligand rows are highlighted. For cytokine genes functionally expressed in T or RPE cells after coculture, the gene expression of the corresponding receptor(s) was also examined. Only ligand-receptor pairs that were expressed in T-cells and RPE cells, respectively, are shown.

IFNγ and TNFα Responsible for the Majority of T-Cell–Induced RPE Cell Chemokine Expression

Recombinant human cytokines were added to the basolateral side of RPE cells grown on membrane inserts, in end concentrations comparable with those measured in T-cell–conditioned media. After 48 hours, RNA was purified from RPE cells and gene expression quantified using pathway-specific DNA microarrays (SABiosciences) and validated with 1.0 ST GeneChips (Affymetrix; n = 2–3). A quantity of 50 ng/mL LTα had no effect on chemokine expression from RPE cells either alone or in combination with IFNγ and TNFα (data not shown). On the 1.0 ST GeneChips (Affymetrix), 16 chemokine genes were expressed above the median value in RPE cells in at least one treatment group (Fig. 4 and see Supplementary Material and Supplementary Fig. S2).A quantity of 200 ng/mL IFNγ and/or 40 ng/mL TNFα significantly increased gene expression of CCL2, CCL5, IL8, CXCL9, CXCL10, CXCL11, CXCL16, and CX3CL1 (compared with RPE cells cultured alone, Fig. 4 and Table 2). RPE cells grown on membrane inserts were cocultured with activated T-cells in the ratio 2.5 T-cells to 1 RPE cell, and neutralizing antibodies against IFNγ, TNFα, or both were added to the medium. Neutralization of IFNγ and/or TNFα significantly reduced expression of CCL3, CCL8, CXCL9, CXCL10, CXCL11, and CX3CL1 (compared with RPE cells cultured with activated T-cells, Fig. 4 and Table 2). 

Figure 4. 

 

Chemokine gene expression in RPE cells in response to IFNγ and/or TNFα, or activated T-cells with neutralizing antibodies against IFNγ and/or TNFα.

 

RPE cells were cultured in membrane inserts, and added either activated T-cells in the ratio 2.5 T-cells per RPE cell; cytokine(s); or activated T-cells + neutralizing antibodies to the bottom compartment. After 48 hours, RNA was analyzed by 1.0 ST GeneChips (Affymetrix). Note the different scales on the x-axes. Data are shown as mean with error bars indicating SD. ns, not significant. *P < 0.05. **P < 0.01. ***P < 0.001.

Table 2. 

 

Overview of the Effect of IFNγ and/or TNFα on RPE Chemokine Gene Expression

Gene	IFNγ versus RPE/-	TNFα versus RPE/-	IFNγ+TNFα versus RPE/-	Neutralizing Antibodies versus RPE/T
CCL2	—	***	***	—
CCL3	—	—	—	IFNγ and/or TNFα
CCL5	—	—	*	—
CCL7	—	—	—	—
CCL8	—	—	—	IFNγ+TNFα
CCL20	—	—	—	(IFNγ)
CXCL1	—	—	—	(IFNγ)
CXCL2	—	—	—	—
CXCL3	—	—	—	—
CXCL6	—	—	—	(IFNγ)
IL8	—	**	—	TNFα , (IFNγ)
CXCL9	—	—	***	IFNγ+TNFα
CXCL10	—	—	***	IFNγ+TNFα
CXCL11	—	—	***	IFNγ+TNFα
CXCL16	**	—	***	—
CX3CL1	—	—	***	IFNγ+TNFα

 The 16 chemokine genes considered functionally expressed in RPE cells, showing which genes were significantly upregulated by IFNγ, TNFα or a combination of the two, compared with RPE cells cultured alone (RPE/-); and which genes were significantly downregulated by adding neutralizing antibodies, compared with RPE cells cultured with activated T-cells (RPE/T). (IFNγ) denotes upregulation when anti-IFNγ was added, signifying that IFNγ downregulates this gene. * P < 0.05. ** P < 0.01. *** P < 0.001.

Taken together, the increased gene expression of CX3CL1 and CXCL9, 10, and 11 in RPE cells after T-cell coculture was completely explained by the synergistic effect of IFNγ in combination with TNFα. CCL2 and IL8 expression was partially explained by TNFα, CXCL16 partially by IFNγ, and CCL3, CCL5, and CCL8 partially by IFNγ+TNFα. The expression of CCL7, CXCL2, and CXCL3 was not regulated by IFNγ+/-TNFα. Interestingly, we found that IFNγ negatively regulated gene expression of CCL20, CXCL1, CXCL6, and IL8 addition of IFNγ-neutralizing antibody to the coculture upregulated expression of these chemokines. For IL8, we also found that though TNFα significantly increased IL8 transcription, IFNγ+TNFα had no effect, showing that addition of IFNγ negated the effect of TNFα. 

Discussion

In this study, we used a previously described in vitro model where ARPE-19 cells were exposed to activated T-cells through a membrane insert. ¹⁵ In this model, RPE cells will be directly affected only by secreted cytokines. In vivo, adhesion of leukocytes to the choroidal endothelium could potentially create high local cytokine concentrations in the choroid. We exclusively used the cell line ARPE-19, and our results may not be applicable to primary RPE cells. However, we used ARPE-19 cells cultured on plastic or semipermeable membranes in serum-free medium for at least 6 weeks, until pigmentation was macroscopically visible; conditions that have been shown to yield the expression profile closest to native RPE. ^19,20  

Chemokine Secretion by the RPE

Previous studies have examined the expression of chemokines from RPE cells under different conditions in vitro. Elner et al. exposed primary human RPE cells to conditioned medium from activated T-cells; found upregulation of CCL2, IL8, and CXCL10; and identified IFNγ and TNFα as the major inducers of these chemokines. ^21,22 Our results partly support these data. By analyzing gene expression of 174 cytokines and their corresponding receptors, we identified IFNγ, TNFα, and LTα as candidate T-cell–derived cytokines affecting the RPE cells (Table 1), and verified IFNγ and TNFα as the most important cytokines (Fig. 4). Interestingly, we observed that IFNγ downregulated four chemokine genes (CCL20, CXCL1, CXCL6, and IL8). While this has been reported previously in other cell types, ^23–26 it appears to be highly cell type-specific. To our knowledge, this has only been studied for IL8 in RPE cells, but with divergent results. Elner et al. found in one study that IFNγ reduced TNFα-induced IL8 secretion, ²¹ and in a later study that IFNγ potentiated the effect of TNFα on IL8 secretion from RPE cells. ²² These findings demonstrate the complexity of cytokine crosstalk and the difficulty in interpreting studies using one or a few defined cytokines. This is especially important in light of the recent interest in using neutralizing antibodies to a single cytokine for the treatment of inflammatory eye diseases. 

In studies of individual chemokines, CCL2, CCL5, and IL8 have been investigated especially, and numerous inflammatory and oxidative stressors have been found to induce expression in RPE cells. ^3,27–37 In a thorough study by Shi et al., human fetal (hf)RPE cells were exposed to a mixture of TNFα, IFNγ, and/or IL1β, and secretion of CCL2, CCL3, CCL4, CCL5, CCL7, CCL8, CXCL1, IL8, CXCL9, CXCL10, and CXCL11 were all increased. ³⁸ The increase was greatest after apical stimulation with all three cytokines, resulting in higher secretion to the apical side. Basolateral stimulation with TNFα+IFNγ alone resulted in a much lower and less polarized secretion. The hfRPE cells expressed receptors for TNFα, IFNγ, and IL1, and IL1β-stimulation accounted for a major part of the chemokine response. ³⁸ Our results partly support these data. Though the ARPE-19 cells expressed mRNA for the TNFα and IFNγ receptors in our system, they did not express IL1R mRNA. This could be due to differences in hfRPE and the ARPE-19 cell line. Furthermore, ARPE-19 cells expressed high amounts of IL1B mRNA after coculture (see Supplementary Material and Supplementary Table S1). Therefore, we do not believe that IL1β released from activated T-cells should play a large role in our system. Since the expression of all the T-cell–induced chemokines could not be explained by IFNγ and/or TNFα, other T-cell–derived cytokines must be important for the RPE cell response. These could include surface-cleaved forms of lymphotoxin, FASL, and TRAIL, or cytokines of receptors expressed weakly in RPE cells like IL1B and IL17A. 

We observed a significantly higher apical (compared with basolateral) secretion of CCL7, CXCL9, CXCL10, and CXCL11 from RPE cells after basolateral exposure to activated T-cells, while the chemokines CCL2, CCL5, CCL8, CXCL1, and IL8 were secreted to both sides of the epithelial layer, with no significant differences in amounts (Fig. 2). We did not monitor the transepithelial resistance of the ARPE-19 cell layer, and it is possible that microbreaks were present in the monolayer. However, the large differences in cytokine amounts on the apical and basolateral sides, even after 48 hours of coculture, indicates that the RPE cells created a relatively tight epithelial layer allowing only small amounts of cytokines to pass. 

CX3CL1 is one of only two chemokines known to exist in a membrane-bound form, but it can be cleaved from the cell surface by the action of ADAM10 and ADAM17. ³⁹ Both ADAM10 and ADAM17 were transcribed in RPE cells in our system, and CX3CL1 was secreted in predominantly the apical direction (data not shown).The soluble form has chemotactic properties, while the membrane-bound form serves to anchor CX3CR1-expressing cells in the tissue. ¹⁴ An increase in expression or a shift in expression from the membrane-bound to the secreted form could have important effects on the microglial response. To our knowledge, induction of CX3CL1 expression has not been identified in RPE cells before. Silverman et al. found constitutive and TNFα-induced expression of CX3CL1 in the retina, but this was mainly from endothelial cells, and the RPE was not identified as a source. ⁴⁰ Likewise, a study of CX3CL1 expression in primary RPE cells found no change after coculture with LPS-activated retinal microglia. ³⁶ In contrast, we observed increased CX3CL1 expression in RPE cells treated with activated T-cells or a combination of IFNγ and TNFα (Figs. 1, 4). 

Effects of RPE Chemokine Secretion

The implications of increased chemokine secretion from the RPE have not been thoroughly examined in vivo. Potentially, it could lead to chemotaxis, adhesion, activation, and degranulation of immune cells (monocytes, immature dendritic cells, T-cells, NK cells) from the peripheral blood as well as activation and proliferation of resident retinal microglia (see Supplementary Material and Supplementary Table S2 showing chemokine ligand-receptor pairs, Supplementary Table S3 showing cellular expression of chemokine receptors, and Fig. 5). 

Figure 5. 

 

Hypothesis of RPE-derived chemokine effects. (A) Sketch showing upregulated cytokines and their receptors, illustrating the redundancy. (B) Sketch showing our hypothesis of the in vivo effects of cytokines on RPE cells. The RPE cells respond to free, circulating cytokines in the choriocapillaries by polarized secretion of chemokines. Apical secretion may influence resident inflammatory cells (microglia expressing CCR1, CCR2, CCR5, CXCR1, CXCR2, and/or CX3CR1), while basolateral secretion may influence circulating leukocytes (including monocytes, T-cells, NK cells, granulocytes, and dendritic cells).

We believe that RPE secretion of chemokines could have implications for the initial chemotaxis of T-cells into the retina in uveitis, an inflammatory state characterized by leukocyte infiltration. Though the importance of chemokines in the early stages of the disease has been shown, ⁴¹ the source has—to our knowledge—not been explored. The involvement of peripheral leukocytes is less established in AMD. However, a recent study showed that T-cell–genes CD3E and CD28 were upregulated in RPE/choroid samples from AMD eyes. ⁴²  

The most important chemokines in microglial chemotaxis are CCL2, CCL5, and CX3CL1. ⁴³ While CCL2 and CCL5 are activating, ⁶ CX3CL1 has an anti-inflammatory effect, ⁴⁴ and normally mediates monocyte migration to noninflamed tissues and differentiation into resident macrophages. ⁴⁵ This is supported by the observation that the more inflammatory (M1) macrophages mainly express CCR2, while the regulatory (M2) macrophages express only CX3CR1. ^14,46,47 There is some evidence that M2 macrophages are protective for a healthy aging retina by contributing to homeostasis, while M1 macrophages can increase inflammation and tissue damage. ^44,46,48 Interestingly, AMD patients have higher levels of CCL2 in the vitreous and aqueous humor, ^11,49,50 and several studies have found activated microglia in AMD retinas. ^51,52  

One conundrum is the high apical secretion of CXCL9, CXCL10, and CXCL11, which bind the CXCR3 receptor that is mainly present on T and NK cells. ⁵³ While apical secretion would create a chemokine gradient over the RPE, peripheral lymphocytes would not be able to migrate toward it unless the BRB is breached. ⁵⁴ To our knowledge, CXCR3 has not been observed on retinal cells. However, several studies on brain microglia identify CXCR3 as being expressed on neuroprotective microglia. ^55,56 In a mouse model of CNS inflammation, Cxcr3^−/− mice exhibited a more severe phenotype than wild-type mice, with highly activated microglia. ⁵⁷ A recent study showed that mice expressing IL12 under the control of the glial fibrillary acidic protein (GFAP-IL12, expressed in astrocytes and retinal Müller cells) had T-cell infiltration in the brain, but very little ocular disease. However, Cxcr3-deficient GFAP-IL12 mice developed severe ocular inflammation and atrophy, demonstrating a protective effect of CXCR3 in the eye. ⁵⁸ It would be interesting to study whether CXCR3 is expressed on retinal microglia, and if it alters the function of these microglia in the same manner as seen in the CNS. If so, apical RPE secretion of CXCL9, CXCL10, and CXCL11 in response to inflammation could have important effects. 

Conclusions

In conclusion, we have shown that RPE cells respond to inflammatory cytokines by upregulating expression and secretion of chemokines. Strategies aiming at regulating the inflammatory assault on RPE cells, or their response to this assault, might prove useful in preventing development of eye diseases with an inflammatory component, such as AMD or uveitis. However, since some chemokines may induce protective effects, a simple downregulation of the inflammation may prove to have deleterious side effects. 

Supplementary Materials

pdf S1 

Acknowledgments

The authors thank Mette Pries, Mohammed-Samir Belmaati, and Jin Pia Jensen for their technical assistance and Claus H. Nielsen at Institute for Inflammation Research, Rigshospitalet, for the Infliximab. 

Footnotes

 Supported by Velux Fonden and Værn om Synet.

Footnotes

 Disclosure: H.B. Juel, None; C. Faber, None; M.S. Udsen, None; L. Folkersen, None; M.H. Nissen, None