Small Extracellular Vesicle-Associated MiRNAs in Polarized Retinal Pigmented Epithelium

Abstract

Purpose

Oxidative stress in the retinal pigmented epithelium (RPE) has been implicated in age-related macular degeneration by impacting endocytic trafficking, including the formation, content, and secretion of extracellular vesicles (EVs). Using our model of polarized primary porcine RPE (pRPE) cells under chronic subtoxic oxidative stress, we tested the hypothesis that RPE miRNAs packaged into EVs are secreted in a polarized manner and contribute to maintaining RPE homeostasis.

Methods

Small EVs (sEVs) enriched for exosomes were isolated from apical and basal conditioned media from pRPE cells grown for up to four weeks with or without low concentrations of hydrogen peroxide using two sEV isolation methods, leading to eight experimental groups. The sEV miRNA expression was profiled using miRNA-Seq with Illumina MiSeq, followed by quality control and bioinformatics analysis for differential expression using the R computing environment. Expression of selected miRNAs were validated using qRT-PCR.

Results

We identified miRNA content differences carried by sEVs isolated using two ultracentrifugation-based methods. Regardless of the sEV isolation method, miR-182 and miR-183 were enriched in the cargo of apically secreted sEVs, and miR-122 in the cargo of basally secreted sEVs from RPE cells during normal homeostatic conditions. After oxidative stress, miR-183 levels were significantly decreased in the cargo of apically released sEVs from stressed RPE cells.

Conclusions

We curated RPE sEV miRNA datasets based on cell polarity and oxidative stress. Unbiased miRNA analysis identified differences based on polarity, stress, and sEV isolation methods. These findings suggest that miRNAs in sEVs may contribute to RPE homeostasis and function in a polarized manner.

The retinal pigmented epithelium (RPE) is a highly polarized monolayer that functions to maintain the outer blood-retina barrier, phagocytose photoreceptor outer segments, transport nutrients and ions, and recycle visual pigments.¹ Situated between the photoreceptors and the systemic circulation of the choroid, RPE polarity is responsible for directional secretion of proteins, lipoprotein particles, and extracellular vesicles (EVs), including exosomes.^2–4 Stressors including oxidative stress and complement activation can damage the RPE and have been implicated in in the development of age-related macular degeneration (AMD),^5,6 the leading cause of irreversible blindness among the elderly of developed countries.^7–11 In fact, dysfunction in the RPE appears to drive AMD onset.¹²

There are three main subtypes of EVs: exosomes and other small EVs (sEVs), microvesicles (or ectosomes), and apoptotic bodies.¹³ Exosomes are the smallest subpopulation of EVs (30–150 nm in diameter) containing proteins and genetic material including microRNAs (miRNAs), which make up approximately 19% of the RNA content of EVs,^14–16 from their cells of origin, enabling cells to communicate without contact and to modulate immunoregulatory and other processes. Exosomes and other sEVs are formed as intraluminal vesicles by inward budding into early endosomes, which mature into multivesicular endosomes, and are released by fusion of multivesicular endosomes with the plasma membrane into the extracellular milieu.¹⁷ Small EVs are released from most cell types under normal or pathological conditions, influencing the activity of recipient cells by carrying active signals.^18,19 More specifically, EVs house and transfer genetic material to other cells as a form of intercellular communication²⁰ and are known to be involved in pathologic processes in cancer, cardiovascular disease, Parkinson's disease, Alzheimer's disease, and ocular diseases.^2,21–23

Exosomes contain miRNAs, which have been shown to be essential for RPE function.²⁴ miRNAs are small non-coding RNA molecules²⁵ that regulate gene expression by degrading messenger RNA expression through complementary base pairing.^24,26 Several studies have shown a correlation between a dysregulation of miRNAs and ocular diseases including AMD.^27–31 In vivo and in vitro RPE oxidative stress studies have provided evidence of dysregulation of miRNAs secreted from the RPE.^24,32–35 What remains unknown is how miRNA-mediated gene expression can be detected before RPE dysfunction/morphological changes. EVs facilitate intercellular communication between cells in both homeostasis and diseased states; therefore analyzing EV-associated miRNAs may provide insight into RPE regulation.^34,35

In this study, we interrogated the role of miRNAs in sEV preparations enriched for exosomes, in normal RPE, and in RPE affected by an early-stage AMD-relevant stressor. We used our established chronic subtoxic oxidative stress model³⁶ to analyze the miRNA cargo in RPE-derived sEVs isolated from the apical and basal media at a stage that precedes overt morphological changes in the RPE monolayer. Because the polarization of the RPE is essential for barrier function and transport of nutrients and ions to the retina and choroid,³⁷ we hypothesized that signaling via sEV miRNA cargo would also be polarized. Herein, we identify differences in miRNA content within sEVs that emerge based on polarity, chronic oxidative stress, and the method of EV isolation. We also identified changes in specific miRNAs, including miR-183 in the sEV cargo, under oxidative stress conditions, suggesting that these miRNAs may serve as potential biomarkers of RPE health.

Material and Methods

Polarized Porcine RPE Cell Culture Model

Primary porcine RPE (pRPE) cell cultures were prepared as described in our previous publications.^3,36 Briefly, porcine eyes were obtained from a local North Carolina slaughterhouse and trimmed of excess tissue and anterior segments (including the vitreous and lens). pRPE cells were seeded at high density (300,000 cells/cm²) onto cell culture inserts (Transwell no. 3450; Corning Inc., Corning, NY, USA) and were maintained in complete pig RPE medium, which consisted of DMEM with glucose and sodium pyruvate (Gibco no. 11995-065; Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 1% (v/v) heat-inactivated fetal bovine serum (FBS) (no. 35-010-CV; Mediatech, Manassas, VA, USA), 100 units/mL penicillin, 100 µg/mL streptomycin, 2 mM L-glutamine (no. G6784; Sigma-Aldrich Corp., St. Louis, MO, USA), MEM non-essential amino acids (Gibco no. 11140-050; Thermo Fisher Scientific), 0.25 µg/mL Amphotericin B (Gibco no. 15290-018; Thermo Fisher Scientific), and 10 µg/mL Ciprofloxacin (no. 61-227-RF; Corning Inc., Corning, NY, USA). Porcine RPE cells isolated from 12 to 18 pig eyes per isolation session (in a single day). All RPE cells were pooled from the total number of eyes processed in one session. One pool of RPE cells was used for the four six-well transwell plates required for a single experiment of control and H₂O₂-stressed RPE cultures. Two or three separate pools were generated for the experiments in this study.

Chronic Subtoxic Oxidative Stress Treatment

We used a chronic subtoxic oxidative stress pRPE cell culture model that we recently described.³⁶ Briefly, oxidative stress was induced by adding 200 µM H₂O₂ daily to both the apical and basal compartments of fully differentiated pRPE transwell cultures. Our subtoxic oxidative stress condition mimics early stages of pRPE dysfunction without disrupting the barrier integrity (which occurs at transepithelial electrical resistance [TER] values below 100 Ω ∙ cm²;³⁸) or causing cell death within the four-week experimental treatment duration as described previously.³⁶

TER Measurements

Barrier integrity and function of the cultured pRPE cells were established by measuring the TER, which is inversely proportional to the paracellular permeability of cultured pRPE cells.^39–42 TER measurements were obtained using an epithelial volt-ohm meter (EVOM) equipped with a 24 mm EndOhm chamber (World Precision Instruments, Sarasota, FL, USA). TER values were determined from a minimum of three individual cultures and corrected for the inherent cell culture insert resistance within five minutes after removing the plates from the incubator as we have previously described.³⁶ All values represent the mean ± SEM.

Conditioned Media for sEV Isolation

The generation of EV-depleted FBS and conditioned media for sEV isolation was done as described previously.^36,43 Briefly, 20% (v/v) FBS was centrifuged in a Beckman Optima XE-90 ultracentrifuge (Beckman Coulter, Inc, Southfield, MI, USA) using an SW 28 Ti rotor at 100,000g_avg for 18 hours at 4°C. The supernatant was carefully collected without disturbing the loose pellet and sterile-filtered through a 0.22 µm PVDF filter bottle (Millipore, Burlington, MA, USA), aliquoted and frozen at −20°C until used. EV-conditioned media was generated using the complete media described above substituting 1% (v/v) heat-inactivated FBS for 2% (v/v) EV-depleted FBS.

Collections from RPE cell culture inserts of conditioned media for sEV isolation were as follows: for each treatment condition, conditioned media from two six-well cluster plates with 24 mm permeable inserts were collected every other day (every 48 hours) for four weeks. The volumes collected were 1.5 mL in the upper (apical) chamber and 2.6 mL in the lower (basal) chamber for each insert. For an average sEV preparation, 100 to 200 mL of apical and 200 to 300 mL of basal conditioned media was used as starting material. During the four-week experiments, conditioned media was stored at −80°C until further analyses were performed. Samples were then thawed and pooled according to timepoint, treatment, and polarity.

Small EV Isolation Methods

We isolated sEVs following the MISEV2018 guidelines⁴⁴ using two different protocols: differential ultracentrifugation (DUC) and cushioned iodixanol buoyant density gradient ultracentrifugation (C-DGUC), as detailed below. We extracted miRNA from sEVs in parallel with our sEV proteomic studies.³⁶ Small EVs were authenticated by immunoblotting for exosome and sEV markers (Supplementary Fig. S1), size analysis using Nanoparticle tracking analysis (ZetaView PMX-110; Particle Metrix, Inning am Ammersee, Germany), and transmission electron microscopy as shown and previously described.^3,36

SEV Isolation by DUC

Small EVs were isolated using a modification of a well-established standard DUC protocol as previously described.^36,43,45 The resulting 100K pellet was lysed with 700 µL of TriZol (miRNeasy Micro Kit, no. 217084; Qiagen, Hilden, Germany). The total RNA quality and quantity was determined using an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA) with RNA Pico chips at the Duke Microbiome Core Facility.

SEV Isolation by C-DGUC

Conditioned media was ultracentrifuged on top of a cushion of 60% OptiPrep (no. 1556; Sigma-Aldrich) as previously described.³⁶ After ultracentrifugation of the subsequent density gradient, 1 mL fractions were collected manually from the top of the self-generated gradient and weighed to determine density. A total of 12 fractions (1 mL each) were collected. Fractions with known densities for exosomes and sEVs (1.07–1.11 g/mL), were pooled and used for downstream analyses. The pooled 1 mL fractions were diluted with PBS to 12 mL and subjected to centrifugation at 100,000g_avg for 90 minutes in an SW 41 Ti rotor using a Beckman Optima XE-90 ultracentrifuge (Beckman Coulter, Inc). The resulting pellets were resuspended in 700 µL of TriZol (no. 217084; Qiagen). The total RNA quantity was measured as described above.

MiRNA Sequencing

MiRNAs were isolated from sEVs using a Qiagen miRNeasy micro kit (no. 217084; Qiagen). A range of between 1 ng to 10 ng of RNA was obtained from the sEV samples. To maintain equivalent amounts of RNA across all the samples, a total of 1 ng of RNA per reaction was used to synthesize each of the miRNA libraries from the sEVs isolated from all conditions. miRNA libraries were prepared using a QIAseq miRNA Library kit (no. 331502; Qiagen). Illumina indexes (no. 331592; Qiagen) were incorporated into the miRNA libraries for sequencing using an Illumina NGS system (Illumina, San Diego, CA, USA). The miRNA libraries were sequenced at the Duke Sequencing and Genomic Technologies Center using the Illumina MiSeq v3 chemistry with a read length of 75 bases.

We separated our four-week experiment into two different time points for control and oxidative stress conditions. The “Early” timepoint spans weeks 1 and 2 in culture from the start of the experiment and the “Late” timepoint spans weeks 3 and 4 in culture. Three replicate experiments were done for the following conditions to identify miRNAs found in the cargo of apical sEVs based on the following eight treatments, timepoints, and isolation method combinations: (1) apical DUC early control, (2) apical DUC early chronic stress (H₂O₂), (3) apical DUC late control, (4) apical DUC late H₂O₂, (5) apical C-DGUC early control, (6) apical C-DGUC early H₂O₂, (7) apical C-DGUC late control, and (8) apical C-DGUC late H₂O₂. There is an N = 2 for the following conditions to identify miRNAs from basal sEVs: (1) basal DUC early control, (2) basal DUC early H₂O₂, (3) basal DUC late control, (4) basal DUC late H₂O₂, (5) basal C-DGUC early control, (6) basal C-DGUC early H₂O₂, (7) basal C-DGUC late control, and (8) basal C-DGUC late H₂O₂. See Supplementary Figure S2 and Supplementary Figure S3 for a schematic of the experiment design.

Sequencing Quality Control and Statistical Analysis

Poor quality miRNA sequencing reads and adapters sequences were removed with Cutadapt⁴⁶ and UMI Tools,⁴⁷ respectively. Reads between 18-30 base pairs were aligned to the Sscrofa 10.2 genome assembly⁴⁸ (https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_000003025.5/), as well as mature miRNA and hairpin structure sequence references from miRbase dataset⁴⁹ with Bowtie2.⁵⁰ The alignment parameters allowed one mismatch per 20 nucleotide seed and multiple mappings. The single best scoring mapping was used for miRNA quantification with BEDTools.⁵¹ All statistical analyses were done within the R (v4.3.1) computing environment.⁵² The miRNAs with at least 10 read counts in 75% of samples were normalized using the locally weighted least squares regression method.⁵³ Differential expression analyses were performed with multiple testing correction (5% FDR) using limma,⁵⁴ and results were visualized with ggplot2.⁵⁵

MiRNA Validation

Validation of selected miRNAs were performed as previously described.²³ We aligned the nucleotide sequences of selected miRNAs against human miRNA equivalents and found that these porcine and human miRNAs are almost identical. Therefore we selected human miRNA assays from Qiagen after further confirmation with the Qiagen Technical Support team. Briefly, RNA isolated from apical and basal sEVs in RPE cell media from all conditions was synthesized into cDNA using the miRCURY LNA RT Kit (no. 339340; Qiagen) per the manufacturer's protocol. Quantitative reverse-transcription PCR (qRT-PCR) for specific miRNAs was achieved using miRCURY LNA SYBR Green PCR Kit (no. 339346; Qiagen) according to the manufacturer's protocol. PCR reactions were performed in triplicate using QuantSudio3 (Applied Biosystems, Waltham, MA, USA), and data were analyzed using the comparative C_t (ΔΔC_t) method.⁵⁶ The miRNAs used as references for qRT-PCR were Homo sapiens (human) hsa-miR-103a-3p (no. 339306; Qiagen; GeneGlobe ID:YP0020463) and hsa-miR-30c-5p (no. 339306; Qiagen; GeneGlobe YP00204783). The miRNA expression was normalized to either of the reference miRNA above, which were selected based on their consistent equal expression across different experimental groups. Reference miRNAs were selected based on the following criteria of a fold change between −1.04 and 1.04. The miR103a-3p was used as a reference for polarity comparisons, and oxidative stress comparisons for apical sEVs. MiR-30c-5p (no. 339306; ; Qiagen; GeneGlobe YP00204783) was used as reference miRNA for the method comparisons for basolateral sEV samples.

The miRNAs tested for validation by qRT-PCR were hsa-miR-182-5p (no. 339306; Qiagen; GeneGlobe ID: YP00206070), hsa-miR-183-5p (no. 339306; Qiagen; GeneGlobe ID: YP00206030), and hsa-miR-122-5p (no. 339306; Qiagen; GeneGlobe ID: YP00205664). These three miRNAs have perfect sequence matches to the porcine miRNA equivalents. The qRT-PCR quantification was tested for statistical significance using an unpaired two-tailed Student t-test. P values < 0.05 were considered statistically significant. All statistical analyses were performed using Prism Software (GraphPad Prism Version 9.5.1 [733] for PC; GraphPad, San Diego, CA, USA).

Pathway Analysis of EV MiRNAs

Bioinformatics analyses were performed using the Qiagen Ingenuity Pathway Analysis software (Ingenuity IPA – 94302991; Qiagen). The miRNAs that were significantly different based on polarity and oxidative stress were loaded into IPA Datasets, followed by the “Core Analysis for Expression” to identify enriched “Canonical Pathways.” The miRNA Target Filter in IPA detects target genes from multiple sources (e.g., TargetScan, TarBase, miRecords, and Ingenuity Knowledge Base).^57–59 For accuracy, we limited our miRNA targets to those experimentally validated or predicted at high confidence in IPA. The list of enriched pathways and networks were exported from the Ingenuity IPA analysis platform.

Results

Differences in MiRNA Yields Based on SEV Isolation Method

The initial analysis of miRNA yields based on sEV isolation method was determined using primary pRPE cells grown on cell culture inserts with and without chronic subtoxic oxidative stress using our previously described model.³⁶ Apical and basal conditioned media was collected from control and stressed pRPE cultures that were at 100% confluence and that had fully differentiated (TER ∼800 Ohm ∙ cm²) for at least two weeks before use in experiments. Small EVs were isolated from the pooled pRPE cell culture media (apical plus basal) using either the C-DGUC or the standard DUC method. Overall sEV yields for the different conditions were assessed by immunoblotting for sEV marker Syntenin 1, or by Nanoparticle Tracking analysis, see Supplementary Table S1. The miRNAs were then isolated, used to prepare miRNA libraries, and sequenced. We combined apical and basal sEVs isolated by DUC from control and stressed RPE cultures and compared the same conditions with sEVs isolated from DUC to identify differences in miRNA yields between methods.

We performed differential miRNA expression analyses within each experimental group (four conditions, listed in Methods under miRNA sequencing). Comparison of the isolation methods using apical preparations showed there was higher abundance of miRNAs using the DUC method, in which we identified 213 miRNAs (Fig. 1, Supplementary Table S2; 16 miRNAs decreased and 16 increased miRNAs using DUC vs. C-DGUC, that were changed at least twofold). To confirm that the miRNA yield differences detected between the C-DGUC and DUC methods were consistent with what is shown in Figure 1, we performed an additional comparison of isolation methods using only EVs released on the basal side, from control and stressed RPE cells (Fig. 2, Supplementary Table S3). Twelve miRNAs decreased and 20 increased miRNAs using DUC versus C-DGUC for basolateral EVs, that were changed at least twofold. There was a decrease in alignment rate percentage matched to the Sus scrofa (porcine) sequence in basal sEVs compared to apical sEVs (Supplementary Fig. S4), which is why basal sEVs were chosen for method comparison for validation of Figure 1. Comparing both methods, there is a significantly higher miRNA abundance detected when using the DUC method compared to the C-DGUC method. The DUC method had a higher alignment rate percentage matched to the Sus scrofa sequence compared to the C-DGUC method (Supplementary Fig. S4). Next, we sought to validate the differences between miRNA content in both isolation methods, with quantitative reverse transcription PCR (qRT-PCR) on selected miRNAs. Indeed, we detected a significantly higher relative abundance of miR-183 (6-fold change increase) and miR-122 (9-fold change increase) content in sEVs isolated by the DUC method (Supplementary Fig. S5), correlating with what was seen in Figures 1 and 2, Tables 1 and 2, and Supplementary Tables S2 and S3.

Figure 1.

Differential miRNA abundance in all sEVs by isolation method. Volcano plot of miRNAs in sEVs based on a comparison of the isolation method used. Apical and Basal RPE-derived sEVs under stressed and control conditions were isolated using either the DUC or C-DGUC method. The DUC isolation group consists of all of the apical plus basal sEVs from media pooled from stressed and control conditions and then isolated using the DUC method. The C-DGUC isolation group consists of all the apical and basal sEVs isolated from cell media of RPE cultures under stressed and control conditions. The baseline was normalized for the C-DGUC method, and the changes indicated refer to sEV miRNAs isolated using the DUC method. See Supplementary Table S2 for miRNA dataset.

Figure 2.

Differential miRNA abundance in Basal sEVs by isolation method. Volcano plot of comparison of miRNAs in sEVs based on isolation methods. Basolateral RPE-derived sEVs from both stressed and control conditions were pooled and isolated using the DUC or C-DGUC method under normal control and stressed conditions. The baseline was normalized for the C-DGUC method and changes are indicated for miRNAs isolated with the DUC method. See Supplementary Table S3 for table format of this miRNA dataset.

Table 1.

Differential Abundance of miRNAs in Apically Versus Basolaterally Released sEVs From the RPE During Normal Homeostasis

MiRNA Name	Fold Change	P Value	Adjusted P Value
Ssc-miR-339	−1.82	0.002	0.02
Ssc-miR-574-3p	−1.78	0.0008	0.02
Ssc-miR-328	−1.70	0.03	0.08
Ssc-miR-29a-3p	−1.56	0.001	0.02
Ssc-miR-29b-1	−1.32	0.008	0.05
Ssc-miR-339-5p	−1.25	0.005	0.04
Ssc-miR-532-5p	−1.21	0.02	0.09
Ssc-miR-218b	−1.20	0.005	0.04
Ssc-miR-128-1	−1.15	0.007	0.05
Ssc-miR-320	1.11	0.003	0.04
Ssc-miR-425-5p	1.13	0.009	0.05
Ssc-miR-7134-5p	1.17	0.002	0.03
Ssc-miR-708-5p	1.18	0.004	0.04
Ssc-miR-148a	1.22	0.008	0.05
Ssc-miR-222	1.23	0.01	0.06
Ssc-miR-181a-2	1.40	0.003	0.04
Ssc-miR-423-5p	1.53	0.003	0.04
Ssc-miR-374a-5p	1.58	0.0004	0.02
Ssc-miR-98	1.60	0.02	0.09
Ssc-miR-181a-1	1.63	0.02	0.09
Ssc-miR-340-2	1.71	0.005	0.04
Ssc-miR-181d	1.86	0.0009	0.02
Ssc-miR-363	1.89	0.004	0.04
Ssc-miR-143-3p	2.24	0.003	0.04
Ssc-miR-122	2.59	8.10E-05	0.006
Ssc-miR-122-5p	2.59	8.10E-05	0.006

miRNAs in basolateral released sEVs compared to apically released sEVs with a log² fold-change difference are listed. Apical sEVs and basolateral sEVs were isolated from control cell cultures using the DUC method. Apical sEVs from early and late control timepoints were combined and compared to basolateral sEVs from early and late control timepoints. The ranking of miRNAs is sorted from low (negative value) to high (positive value) miRNA content found in basolateral sEVs. Adjusted P value resulted from multiple testing correction.⁵⁴

Table 2.

Differences in Abundance of sEVs miRNA Based on RPE Polarity After Chronic Oxidative Stress

MiRNA Name	Fold Change	P Value	Adjusted P Value
Ssc-miR-532-5p	−2.79	0.007	0.05
Ssc-miR-339	−2.67	2.48E-05	0.003
Ssc-miR-210	−2.21	0.02	0.09
Ssc-miR-339-5p	−1.84	0.0001	0.004
Ssc-miR-2320-5p	−1.59	0.007	0.05
Ssc-miR-29a-3p	−1.45	0.003	0.03
Ssc-miR-30b-5p	−1.43	0.02	0.09
Ssc-miR-27a	−1.34	0.02	0.09
Ssc-miR-29b-1	−1.14	0.008	0.05
Ssc-miR-24-3p	−1.04	0.005	0.04
Ssc-miR-340	0.99	0.01	0.07
Ssc-miR-423-3p	1.13	0.008	0.05
Ssc-miR-148a-3p	1.17	0.02	0.10
Ssc-miR-181b	1.32	0.002	0.03
Ssc-miR-126-5p	1.59	0.004	0.04
Ssc-miR-98	1.60	0.007	0.05
Ssc-miR-4332	1.73	0.02	0.08
Ssc-miR-423	1.80	0.002	0.02
Ssc-miR-423-5p	1.83	0.003	0.03
Ssc-miR-181a-2	1.90	0.0002	0.002
Ssc-miR-34c	2.04	0.004	0.04
Ssc-miR-92b-3p	2.08	0.02	0.08
Ssc-miR-181a	2.16	5.16E-05	0.004
Ssc-miR-374a-5p	2.17	0.0004	0.008
Ssc-miR-340-2	2.52	8.78E-05	0.004
Ssc-miR-181d	2.67	0.0008	0.01
Ssc-miR-181d-5p	2.74	0.0006	0.01
Ssc-miR-184	2.91	0.0003	0.006
Ssc-miR-190a	3.33	0.008	0.05
Ssc-miR-122-5p	3.82	0.0002	0.005
Ssc-miR-142	6.73	1.98E-06	0.00042

The table shows miRNAs in basolateral released sEVs compared to apically released sEVs with a log² fold-change difference. Apically and basolateral released sEVs specifically from stressed RPE cell cultures were isolated using the standard DUC method. Apical sEVs from early and late oxidative stress timepoints were combined and compared to basolateral sEVs from early and late oxidative stress timepoints. The ranking of miRNAs is sorted from low (negative value) to high (positive value) miRNA content found in basolateral sEVs. Adjusted P value resulted from multiple testing correction.⁵⁴

Differences in MiRNA Content Based on RPE Polarization

Based on our comparison of the miRNAs from preparations using different sEV isolation methods, the DUC method resulted in higher miRNA yields. Therefore, going forward, to interrogate changes in miRNA content in sEVs based on RPE polarity, we used the DUC method for all subsequent isolations. Because the polarization of the RPE monolayer is essential for RPE barrier function and the transport of ions and nutrients to the retina and choroid,⁶⁰ and it is largely unknown whether specific miRNAs are loaded into sEVs specifically targeted for apical or basal secretion from the RPE, we identified and compared the miRNA cargo of sEVs secreted into the apical versus the basal media by highly polarized pRPE cells grown on cell culture inserts. By isolating and sequencing miRNAs from preparations of apical or basal RPE sEVs (Fig. 3), we identified specific miRNAs in the resulting miRNA datasets within sEVs that were secreted preferentially from either the apical or basolateral surface of control RPE cells (Table 1; nine Apical and 17 Basolateral miRNAs that were at least twofold changed).

Figure 3.

Differential miRNA abundance by polarity from sEVs released by control pRPE cells. Volcano plot of miRNAs in apical and basolateral (Basal) sEVs isolated from control RPE cell culture media. Small EVs were isolated using the DUC method. The baseline is set for apical sEVs, and the change shown refers to basolateral sEVs. See Supplementary Table S4 for excel of this miRNA dataset.

Select miRNAs that were significantly different based on their abundance in apical and basolateral sEVs isolated from control pRPE cultures were validated by qRT-PCR (Table 1). We identified miR-183 (1.3-fold change, P = 0.611) as highly abundant in apically released sEVs compared to basolaterally released sEVs. We identified miR-122-5p (2.6-fold change, P = 0.006) as highly abundant in basolateral sEVs compared to apically released sEVs. We pooled sEVs isolated by both methods to rule out the effect of isolation method on the sEV miRNA cargo recovered from either side of the control pRPE cultures. miR-182 (eightfold change, P = 0.006) and miR-183 (ninefold change, P = 0.0007) were more abundant in the cargo of apically released sEVs compared to basolaterally released sEVs, and this was validated by qRT-PCR (Figs. 4A, 4B). miR-122 (sixfold change, P = 0.0009) was confirmed to be more abundant in the cargo of basolateral compared to apical sEVs by qRT-PCR (Fig. 4C) as previously shown in Figure 3.

Figure 4.

The RPE secretes miRNAs via sEVs in a polarized manner during homeostasis. qRT-PCR analysis of miR-182, miR-183, and miR-122 from apical and basolateral sEVs isolated from control RPE cells. Small EVs were isolated using the DUC (100K) and C-DGUC methods, and pooled. miR-182 (A) and miR-183 (B), were present in higher levels in apical sEVs. (C) miR-122 was present at higher levels in basal sEVs. **P < 0.01; ***P < 0.001.

Oxidative Stress Affects the MiRNA Cargo in RPE sEVs

Based on our findings that the RPE secretes different miRNAs in a highly polarized manner via sEVs during homeostasis, we next interrogated whether there are differences in the miRNA cargo of sEVs released based on polarity under conditions that cause RPE dysfunction. To determine changes in miRNA content secreted via sEVs that are relevant to early stages of RPE dysfunction, we used the same chronic subtoxic oxidative stress model that we recently showed causes mild dysfunction preceding any functional or morphological changes.³⁶ We treated pRPE cell cultures for four weeks with 0.2 mM H₂O₂ daily. We separated the experiment into two timepoints: “Early,” which corresponds to pooled conditioned media collected during weeks 1 and 2 of the oxidative stress treatment; and “Late,” which corresponds to pooled conditioned media collected during weeks 3 and 4 of the oxidative stress treatment.

Interestingly, statistical analyses identified miRNAs that were preferentially released either apically or basolaterally under these RPE stressor conditions (Supplementary Fig. S6 and Table 2; 10 Apical and 21 Basolateral miRNAs that were at least twofold changed). For example, miR-184, miR-142 and miR-190a were significantly higher in the cargo of basolateral sEVs compared to apical sEVs under oxidative stress conditions (Table 2). miR-182 (twofold change, P = 0.02) and miR-183 (2.5-fold change, P = 0.02) were significantly decreased in the cargo of apically released sEVs from RPE cells exposed to late oxidative stress compared to the early treatment (Fig. 5). Based on the decreases in the miR-182 and miR-183 in apically released sEVs from stressed RPE cell cultures during the late oxidative stress timepoint compared to the early oxidative stress timepoint obtained by miRNA sequencing (Fig. 5), we performed qRT-PCR to validate these changes, as well as compare the changes in these miRNAs at the same time points in the apical sEVs isolated from control RPE cell media. There was no difference in miR-182 levels in apical sEVs comparing the early (weeks 1 and 2) abundance to the late (weeks 3 and 4) abundance of chronically stressed RPE cells (Fig. 6A). However, miR-183 (fivefold change, P = 0.006) levels were significantly decreased after four weeks of oxidative stress compared to two weeks of oxidative stress (Fig. 6B). The difference we observed in miR-183 levels between early and late treatment of oxidative stress suggest that stress accumulates over time in the RPE, which validates the statistical analysis results in Figure 5. When the same validation by qRT-PCR was performed on these two miRNAs with the control apical sEVs there were no significant changes in the abundance of either miR-182 or miR-183 during the first two weeks of oxidative stress treatment (Figs. 7A, 7B). Nor was there a significant decrease in the miR-182 levels during weeks 3 and 4 of oxidative stress treatment, although it trended downward (Fig. 7C). However, the miR-183 (threefold change, P = 0.04) was significantly decreased at the late timepoint of oxidative stress treatment relative to the late control (Fig. 7D). These results suggest that individual miRNA levels (e.g., miR-183) in sEVs may serve as indicators of RPE stress.

Figure 5.

Apical sEVs secreted by the RPE contain lower amounts of miR-182 and miR-183 after oxidative stress. The top and bottom whisker of each boxplot represents where 25% of data is located, and the horizontal line is the median. Apical sEVs released from stressed RPE cells were isolated using the DUC (100K) method. The early timepoint (yellow box) refers to cell media collected during weeks 1 and 2, and the late timepoint (blue box) refers to weeks 3 and 4 of oxidative stress treatment, respectively. As oxidative stress accumulates in the RPE over time, miR-182 and miR-183 levels decrease.

Figure 6.

Differences in sEV miR-182 and miR-183 levels at early and late stages of chronic oxidative stress in RPE. qRT-PCR was performed on apical sEVs isolated from cell culture media of chronically stressed RPE cell cultures at early and late timepoints. Weeks 1 and 2 of oxidative stress treatment are labeled as “Early H₂O₂.” Weeks 3 and 4 of oxidative stress treatment are labeled as “Late H₂O₂.” Apical sEVs were isolated using the DUC (100K) method. (A) miR-182 abundance was not significantly different in sEVs at the early compared to the late timepoint. (B) miR-183 was significantly decreased in apical sEVs at the late compared to the early timepoint in oxidatively stressed RPE cells. **P < 0.01; ns, not significant.

Figure 7.

MiR-183 levels in apical sEVs are statistically significantly decreased in RPE cultures at late stage of chronic oxidative stress. qRT-PCR was performed on apical sEVs from weeks 1 and 2 (labeled “Early”), and weeks 3 and 4 (labeled “Late”), from control and H₂0₂ stressed RPE cell cultures. Apical sEVs were isolated using the standard DUC method. Neither miR-182 (A) nor miR-183 (B) levels were significantly affected by oxidative stress during the early timepoint. At the late timepoint, miR-182 (C) was not significantly affected, but miR-183 (D) was significantly decreased in response to chronic oxidative stress. *P < 0.05; ns, not significant.

Pathway Analysis From SEV MiRNA Profiling

We performed IPA (Qiagen) to identify pathways associated with miR-182 and miR-183, because these miRNAs displayed significant changes in apical sEVs from the RPE in response to oxidative stress. We identified the circadian rhythm pathway (BRTC), as well as the melanocyte development and pigmentation signaling (binding to MITF), as targets of miR-183 (Fig. 8A) and miR-182 (Fig. 8B). IPA analyses also identified miR-183 as regulating tumor necrosis factor-α (TNF-alpha), IDH2, YAP1, FOXO1, IL1B, IL6, and PPP2CA (Fig. 8A). We also identified targets of miR-182 and miR-183 that were specific for the RPE. MITF is associated with RPE development⁶¹ and antioxidant signaling.⁶² miR-183 targets IDH2 and FOXO1 that are involved in protecting the RPE from oxidative stress.^63–65 TNF-α is known to disturb the visual cycle and RPE barrier function.^66,67 The miR-182 was also identified as regulating ITGA4, IFITI, SMAD9, CASP12, CASP10, and IGFBP1 through the Cell movement of epithelial cells pathway (Fig. 8C).

Figure 8.

Pathway analysis of miR-183 and miR-182. (A) IPA analysis identified targets of miR-183. The dashed lines show direct targets of miR-183. (B) Genes identified are targets of miR-182. (C) “Cell movement of epithelial cell line” is a pathway that was identified to be a target of miR-182.

Discussion

To date, EV studies in the RPE have focused on solely studying miRNAs from apically released EVs.^68–72 Thus there is a gap in knowledge of the specific miRNAs that are secreted from the RPE based on polarity and after chronic oxidative stress. Using highly polarized RPE cell cultures, we show that miRNAs are secreted via EVs from the RPE in a polarized manner during homeostasis, and that there is a change in EV miRNAs on oxidative stress. Previous studies have shown that miRNAs impact RPE development and differentiation,^24,73,74 and high levels of miR-23a-3p and miR-27b-3p were detected in AMD donor eyes.⁷⁵ Our study focuses on exploring the role of EVs as indicators of RPE dysfunction at a transcription and/or cargo-loading level, based on their miRNA cargo. We generated miRNA datasets from RPE-derived EVs, based on multiple parameters including the impact of different isolation methods, polarity, and chronic oxidative stress.

The EV lipid bilayer protects miRNA cargo from degradation, facilitating miRNA secretion from the cell and extracellular survival.¹³ The three major subtypes of EVs (small EVs including exosomes, microvesicles, and apoptotic bodies) all contain varying amounts and types of miRNAs and can be involved in cell signaling.^13,76–78 The majority of RPE studies that have analyzed miRNAs from EVs have used ExoQuick (PEG precipitation), which captures a heterogeneous population of large and small EVs, lipoprotein particles, and varying amounts of soluble protein contaminants, and DUC, which enriches for small EVs, as isolation methods.^68–70,79 miRNAs are loaded into EVs, but how EV miRNAs affect RPE function when taken up by neighboring cells is a question that needs to be further investigated. One of the essential principles of EV studies is to use proper isolation methods to enrich for the EVs of interest.⁸⁰ Thus we compared the miRNA content in sEVs isolated using two different isolation methods that are accepted in the EV research field as being preferred over precipitation methods, for their increased sEV enrichment.⁴⁴

Statistical analyses of the miRNA sequences identified, show a significant difference in miRNA content based on the different sEV isolation methods. There is a larger abundance of miRNAs detected when using the DUC method, which may be due to the larger heterogeneity of EVs, and possibly lipoprotein particles; compared to C-DGUC preparations that are composed of a more homogeneous subset of sEVs and exosomes.^81,82 Regardless of the isolation method, it is a challenge to separate exosomes and sEVs from other EV subtypes. There are significant differences in the composition of the preparations generated by the most commonly used EV isolation methods, making it difficult to determine which population of EVs carries specific miRNAs.⁸³ The lower miRNA yields in C-DGUC compared to DUC preparations is not surprising, as it is known that using a sucrose or iodixanol buoyant density gradient results in a decrease in EV yield while resulting in a higher enrichment of sEVs and exosomes.⁸⁴ It is well known that EV preparations differ widely in their protein composition depending on the isolation methods used.⁸⁵ However, for sEV miRNA isolation, the DUC method does not appear to co-isolate any significant amounts of contaminants compared to the C-DGUC method, that interfere with miRNA sequencing. Previous miRNA studies including a wide variety of different EV isolation methods, concluded that differences in miRNA content between isolation methods are relatively small.^86–88 Determining which method will yield the majority of EV miRNA, or miRNA cargo of a particular EV subtype, is challenging. For example, a recent cancer study, comparing the miRNA content between exosomes and microvesicles⁸⁶ found 53 unique miRNAs in the SW620 cancer cell line, and 20 of those miRNAs were shared between exosomes and microvesicles. A small number of miRNAs were unique for either exosomes or microvesicles.⁸⁶ The shared miRNAs expressed in both exosomes and microvesicles suggest that miRNAs are secreted in all EV subtypes. The mechanism(s) governing miRNA loading into a specific EV subtype remains unclear.

Exosomes and sEVs share some characteristics with larger EVs, such as somewhat overlapping size and tetraspanin components.⁸⁹ We likely removed the majority of microvesicles in both of our isolation methods by discarding the 10,000g pellet³⁶ (which is known to contain microvesicles⁸⁶), prior to the 100,000g centrifugation and subsequent C-DGUC. Another possible explanation for the higher levels of miRNA content in the 100 K pellets from the DUC method is the presence of lipoproteins, which also carry miRNAs.⁹⁰ High-density lipoproteins in particular, have been reported as carriers of miRNAs.⁹¹ High-density lipoproteins are not exclusively a cholesterol transporter but has the capacity to transfer miRNAs to cells, which can lead to altered gene expression.

We validated the differential abundance of selected miRNAs based on RPE polarity and oxidative stress by qRT-PCR of apically and basolaterally released sEVs, using both isolation methods to validate the polarity differences. We found that miR-182 and miR-183 are highly specific for the cargo of apically released sEVs regardless of the isolation method (Fig. 4). The miRNA differences in polarity were consistent between apically and basolaterally released sEVs and again not impacted by the EV isolation method. miR-182 and miR-183 are involved in photoreceptor maintenance and differentiation.^92,93 A recent study also implicated miR-182 in increasing expression of VEGFR2 and VEGF under hyperglycemic conditions in ARPE-19 cells,⁹⁴ suggesting a role in angiogenic signaling. miR-122 has been implicated in angiogenesis in the RPE,⁷³ and it is therefore not surprising that miR-122 was specifically found as cargo in basolateral sEVs because these could provide signaling to the choroidal blood vessels.

The chronic oxidative stress treatment of the RPE cells used here did not lead to significant changes in many different sEV miRNAs, which may be due in part to the fact that this a subtoxic oxidative stress.³⁶ This oxidative stress model mimics the earliest stages of RPE dysfunction, and allows the RPE to retain pigmentation, normal morphology, and optimal TER.³⁶ Our study can be contrasted to other studies that use an acute oxidative stress model whereby a greater number of miRNAs were identified having robust changes. For example, ARPE-19 cells exposed to 0.6 mM to 0.8 mM H₂O₂ for 24 hours, induced RPE cell death and 50 significantly differentially abundant miRNAs.⁹⁵ Another study used photo-oxidative damage in mice to induce retinal degeneration to recapitulate an AMD phenotype, and approximately 10 miRNAs were identified that increased upon retina damage.⁹⁶ Using a chronic oxidative stress paradigm, our unbiased sEV miRNA analysis revealed differences in apical versus basal sEVs (Table 2 and Figs. 2, 5, and 7). Some of these miRNAs have previously been shown to be involved in RPE biology, including miR-184, which affects RPE differentiation, efficient uptake of photoreceptor outer segments, and prevention of RPE dysfunction.⁹⁷ MiR-142 is considered a mediator of choroidal neovascularization and induces microglia activation.⁹⁸ MiR-190a is downregulated in RPE undergoing epithelial-to-mesenchymal transition.⁹⁹ Based on the miRNA comparisons by polarity (Fig. 3, Table 2, and Supplementary Fig. S5), there was a higher abundance of miRNAs released basolaterally than apically during normal homeostasis and upon oxidative stress. Recent studies in mouse retina have identified miR-183 and miR-182 in small EVs.^96,100 Interestingly, Slc1a1, a neuronal glutamate transporter, is a known target of miR-182 and miR-183.¹⁰¹ It remains unclear what the specific targets of miR-182 and miR-183 are in the RPE, or possibly in photoreceptors that are known to use glutamate for signal transduction to bipolar cells,¹⁰¹ thus highlighting the need for further studies.

Some limitations of the study are worth discussing. The current study used pRPE primary cultures to analyze the secreted sEV miRNA cargo under control and stressed conditions. We have recently shown that changes to protein cargo in pRPE-derived sEVs under the same oxidative stress conditions correlated with the responses detected in several human iPSC-RPE models.³⁶ Nevertheless, it would be worthwhile to repeat the current study in human RPE models in future studies. In addition, studies using other types of EV isolation methods may be able to stratify in greater detail, which subpopulations of EVs and other particles carry these miRNAs.

In conclusion, our study provides an assessment of different parameters that are important in studies of RPE-derived sEVs, such as polarity, which have been largely understudied. We have provided evidence that miRNAs are released via sEVs in a polarized manner during homeostasis and are altered upon oxidative stress. Our comparison of sEV isolation methods highlights the difference in miRNA content between methods and is important for future RPE EV studies. The miRNAs validated in this study and the pathways identified based on our miRNA profiling suggest that miRNAs may serve as biomarkers for detecting early AMD.