*Article* **Mesenchymal Stem Cell Induced Foxp3(+) Tregs Suppress Effector T Cells and Protect against Retinal Ischemic Injury**

**Mona Agrawal <sup>1</sup> , Pratheepa Kumari Rasiah <sup>1</sup> , Amandeep Bajwa 2,3,4 , Johnson Rajasingh 4,5 and Rajashekhar Gangaraju 1,6,\***


**Abstract:** Mesenchymal stem/stromal cells (MSC) are well known for immunomodulation; however, the mechanisms involved in their benefits in the ischemic retina are unknown. This study tested the hypothesis that MSC induces upregulation of transcription factor forkhead box protein P3 (Foxp3) in T cells to elicit immune modulation, and thus, protect against retinal damage. Induced MSCs (iMSCs) were generated by differentiating the induced pluripotent stem cells (iPSC) derived from urinary epithelial cells through a noninsertional reprogramming approach. In in-vitro cultures, iMSC transferred mitochondria to immune cells via F-actin nanotubes significantly increased oxygen consumption rate (OCR) for basal respiration and ATP production, suppressed effector T cells, and promoted differentiation of CD4+CD25+ T regulatory cells (Tregs) in coculture with mouse splenocytes. In in-vivo studies, iMSCs transplanted in ischemia-reperfusion (I/R) injured eye significantly increased Foxp3+ Tregs in the retina compared to that of saline-injected I/R eyes. Furthermore, iMSC injected I/R eyes significantly decreased retinal inflammation as evidenced by reduced gene expression of *IL1β*, *VCAM1*, *LAMA5*, and *CCL2* and improved b-wave amplitudes compared to that of saline-injected I/R eyes. Our study demonstrates that iMSCs can transfer mitochondria to immune cells to suppress the effector T cell population. Additionally, our current data indicate that iMSC can enhance differentiation of T cells into Foxp3 Tregs in vitro and therapeutically improve the retina's immune function by upregulation of Tregs to decrease inflammation and reduce I/R injury-induced retinal degeneration in vivo.

**Keywords:** CD4+CD25+; retinopathy; inflammation; iPSC; mitochondria

#### **1. Introduction**

Ischemic retinopathies, including diabetic retinopathy (DR), retinopathy of prematurity (ROP), and retinal vascular occlusion (RVO), are increasing in prevalence, represent a significant economic burden, and are major causes of vision loss and blindness worldwide [1–3]. A wide variety of traditional treatment therapies, including photocoagulation and anti-VEGF therapies during the neovascularization phase, showed benefits, with no treatments currently approved that address underlying proinflammatory pathways that are known to trigger neurovascular degeneration [2]. Stem cell therapies, mainly multipotent mesenchymal stem cells (MSCs), recently gained significant attention as a potential therapy for the treatment of ischemic retinopathies [3,4]. Our previous studies utilizing MSCs

**Citation:** Agrawal, M.; Rasiah, P.K.; Bajwa, A.; Rajasingh, J.; Gangaraju, R. Mesenchymal Stem Cell Induced Foxp3(+) Tregs Suppress Effector T Cells and Protect against Retinal Ischemic Injury. *Cells* **2021**, *10*, 3006. https://doi.org/10.3390/cells10113006

Academic Editors: Maurice Ptito, Joseph Bouskila and Karl-Wilhelm Koch

Received: 14 September 2021 Accepted: 29 October 2021 Published: 4 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

derived from the stromal vascular fraction of adult human adipose tissue (adipose-derived stem cells, ASC) [5,6], and independently corroborated with bone marrow (BM-MSC) [7] and umbilical cord (UC-MSC) [8], demonstrated substantial regeneration and recovery of the damaged retina after treatment, although the exact mechanisms by which MSCs may protect against vision loss, remains unclear.

MSCs are well known for being involved in immunomodulation, in part by donating mitochondria to damaged tissue or cells [9–11]. Recent evidence also suggests that MSCs can modulate T regulatory cells (Tregs) [12,13], in particular, Tregs expressing the Forkhead box P3 (Foxp3) transcription factor, which is part of the adaptive immune system and are principal regulators of inflammation and immune homeostasis [14]. The Tregs can migrate to the diseased tissue and dampen inflammation by increasing the milieu of anti-inflammatory cytokines and activating macrophages to clear the debris and restore the damaged tissue [15–17]. Since CD4+ T cells may mediate retinal ganglion cell (RGC) degeneration and loss of retinal function after injury [18], it may be possible to reprogram the CD4+ T cells at the site of injury to acquire a Treg phenotype. This might aid in the rescue of retinal damage and, as a result, decrease cell loss and enhance immunotolerance [19]. Therefore, in the present study, we hypothesized that intravitreal injection of MSCs in ischemia-reperfusion (I/R) injured retina reprogram CD4+ T cells to Tregs, dampen inflammation, and improve visual function. Interestingly, our data visualized that MSC actively transferred mitochondria to immune cells to suppress effector T cells and promoted differentiation of CD4+CD25+ Tregs in coculture with mouse splenocytes. Intravitreal injection of MSCs in the I/R eye significantly increased Tregs in the retina, decreased retinal inflammation, and improved visual function compared to saline-injected I/R eyes. These findings indicate that harnessing the immunosuppressive capacity of MSCs is a potential therapy for the treatment of ischemic retinal diseases.

#### **2. Materials and Methods**

#### *2.1. Cell Isolation and Culture*

iMSCs were prepared from human urinary tubular epithelial cells (UEs) through the generation of iPSCs via reprogramming with a cocktail of Oct-4, Sox-2, Klf-4, c-Myc, and Lin-28 mRNAs and subsequently differentiated into MSC as described by us previously [20,21]. Briefly, iMSCs from healthy human urine tubular epithelial cells (UEs) expressing epithelial markers CK19 and ZO1 were used to generate iPSCs via reprogramming with a cocktail of Oct-4, Sox-2, Klf-4, c-Myc and Lin-28 mRNAs [20]. To induce MSC differentiation, the UE-iPSCs were cultured under conditions conducive for mesenchymal differentiation in Mesencult ACF plus medium for 18–21 days and were characterized for MSC markers. iMSC were positive for CD105, CD90, and CD73 while negative for CD31 and CD45 (Supplemental Figure S1). To validate the mechanism of action of iMSC, ASC was used as a well-known control in in-vitro studies. ASCs used in the current study were obtained from Lonza (Cat#PT-5006), cultured in EGM-2MV media, and used between p2 and p7 in all experiments as previously described [22]. All studies involving human ASC and iMSC were approved for research as per the University of Tennessee Institutional Biosafety and as an exempt study by the Institutional Review Board. Human monocytes, THP-1 cells were purchased from ATCC (Cat#TIB-202) and cultured in RPMI 1640 complete medium as a suspension culture. In addition, primary mouse splenocytes were prepared as described previously [23]. Briefly, 6–8 weeks old wild type (C57BL/6) mice were euthanized, the spleen was collected and washed in PBS. The tissue was crushed on a 70-micron cell strainer with 5 mL syringe plunger, rinsed with RPMI 1640 media; cells were pelleted by centrifugation. Subsequently, RBCs were lysed using ACK lysis buffer, inactivated the reaction using 10% FBS. Finally, splenocytes were pelleted and resuspended in RPMI 1640 complete media and filtered through 40-micron cell strainer. Trypan blue negative splenocytes were counted as live cells and used for experiments. All cell cultures were maintained at 37 ◦C and 5% CO<sup>2</sup> in a humidified atmosphere.

#### *2.2. Coculture and Microscopy*

iMSC or ASC were plated at 1 × 10<sup>6</sup> cells/cm<sup>2</sup> in a 60 mm dish and we let them adhere. Following this, 1 × 10<sup>6</sup> cells were stained with fluorescent MitoTracker Red CMXRos (100 nm, Life Technologies, Grand Island, NE, USA) for 45 min at 37 ◦C. Labeled cells were washed 2× with PBS, trypsinized, and seeded into a 6-well plate containing 10 mm coverslips at 1 × 10<sup>5</sup> cells per well. On the second day, 1 × 10<sup>6</sup> THP-1 cells labeled with CellTracker fluorescent probe green (0.5 µM, Life Technologies) were cocultured with iMSC or ASC in a 1:10 ratio for 24 h. After 24 h of coculture, cells were fixed with 4% paraformaldehyde (PFA) and stained with DAPI nuclear stain, mounted using ProLongTM diamond antifade mountant (Life Technologies). Similarly, 1 × 10<sup>6</sup> iMSC or ASC were stained with fluorescent MitoTracker Green (100 nm, Life Technologies) to assess the mitochondrial transfer through F-actin nanotubes. Following this, 1 × 10<sup>5</sup> cells were seeded into a 6 well plate containing 10 mm coverslips. On the second day, 1 × 10<sup>6</sup> THP-1 cells labeled with CellTracker CMAC blue (0.5 µM, Life Technologies) were cocultured with iMSC or ASC in 1:10 ratio for 24 h and treated with 350 nm of cytochalasin B (CytoB, Sigma-Aldrich, Inc., St. Louis, MO, USA). Tubular microstructure tunneling nanotubes were assessed by costaining for F-actin using Phalloidin- Tetramethylrhodamine B isothiocyanate (Sigma) in the presence or absence of CytoB. The mitochondrial transfer was assessed from images captured with an EVOS fluorescence microscope (Life Technologies) or captured with a laser scanning confocal microscope (Zeiss LSM 710, Carl Zeiss Microscopy, LLC, White Plains, NY, USA). For quantification of mitochondrial transfer, THP-1 positive for CMAC blue in the coculture experiments were identified with a region of interest (ROI) and the pixel intensities of MitoTracker Green were computed for each ROI using ImageJ. At least 20 cells/image were considered, and the data were expressed as Mean fluorescent intensity (MFI) values/cell.

#### *2.3. Seahorse Flux BioanalyzerV*

THP-1 cells were pretreated for 2 h with 500 nM Rotenone (acts as a potent inhibitor of complex I of the mitochondrial respiratory chain). Following this, cells were washed (2×) and cocultured with iMSC or cultured as a monoculture. Cocultured THP-1 cells were transferred to a Seahorse 24-well tissue culture plates and oxygen consumption rate (OCR) was measured, and parameters were calculated as previously described [24]. Briefly, prior to the assay, the media was changed to unbuffered DMEM (Gibco #12800-017, pH 7.4, 37 ◦C), and cells were equilibrated for 1 h at 37 ◦C. After measuring basal respiratory rate, Oligomycin (Sigma; 1 µM; uncouples ATP-coupled respiration by inhibiting ATP synthase), FCCP (Sigma; 1 µM; carbonyl cyanide 4-(trifluoromethoxy)-phenylhydrazone (FCCP), mitochondrial uncoupling agent; uncouples mitochondrial respiration from ATP to determine maximal respiratory rate), and electron transport chain (complex I and III) inhibitors, rotenone (Sigma; 0.5 µM) and antimycin A (Sigma; 0.5 µM; to eliminate all mitochondrial respiration) were injected sequentially during the assay. Basal mitochondrial respiration and ATP-linked respiration were determined in whole cells.

#### *2.4. Coculture and Flow Cytometry*

About 1 × 10<sup>5</sup> MitoTracker green-labeled iMSC or ASC seeded in 12-well cell culture plate for 24 h. The next day, 1 × 10<sup>6</sup> freshly prepared splenocytes (1:10 ratio as described above) were added to the top of the stem cell monolayer in complete RPMI 1640 media. After 24 h of coculture, non-adherent and loosely bound splenocytes were collected, washed 2× with PBS. Single-cell suspension of splenocytes was stained in FACS buffer (2% FBS in PBS), incubated with the Fc-block anti-CD16/CD32 (2.4G2) antibody followed by a panel of fluorochrome-coupled antibodies (Table 1). While viability dye eFluor 506 (eBioscience, San Diego, CA, USA) excluded dead cells, a MitoTracker dye uptake assessed mitochondrial transfer by flow cytometry. Stem cells were pretreated for 2 h with 500 nM Rotenone to study the functional relevance of mitochondria transfer and cocultured with primary mouse splenocytes.


**Table 1.** List of fluorophores-coupled FC antibody.

To assess Treg differentiation, freshly prepared splenocytes were activated in the presence of coated anti-CD3 (BD, 10 µg/mL) and soluble anti-CD28 antibody (BD, 1 µg/mL) and recombinant IL-2 (PeproTech, Cranbury, NJ, USA, 1 ng/mL) for 24 h. Activated splenocytes were cocultured on the monolayer of iMSC or ASC in differentiation medium [RPMI 1640 medium, 10% FBS, 1 ng/mL IL-2 (PeproTech) and 5 ng/mL TGF-β (R&D)] for five days with media change at day 3. After 5 days of coculture, nonadherent and loosely bound splenocytes were collected, washed 2× with PBS. Single-cell suspension of splenocytes was then analyzed by flow cytometry as described above using specific surface markers (Table 1). For Foxp3 intracellular staining, cells were fixed and permeabilized using Foxp3/Transcription Factor Staining Buffer (eBioscience). Following this, cells were incubated with anti-Foxp3 (eBioscience) antibody for 50 min, washed and resuspended in the FACS buffer, kept at 4 ◦C, protected from light till acquisition. Data were acquired using a Bio-Rad ZE5 cell analyzer and analyzed by FlowJo software v10.8 (Flowo, Ashland, Wilmington, DE, USA).

#### *2.5. RNA Isolation and Quantitative RT–PCR*

Total RNA was isolated from cells using Nucleospin miRNA isolation kit (Macherey-Nagel Inc, Allentown, PA, USA), following the manufacturer's protocol. RNA quality and integrity were measured by absorbance at 260/280 and 260/230 nm using NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). cDNA was prepared in a reverse transcription reaction using 500 ng of RNA and a high-capacity cDNA reverse transcription kit (Thermo Fisher Scientific) following the manufacturer's instructions. About 100 nanograms of template RNA and 5 nM of each forward and reverse human-specific primers related to nanotube formation (*ACTN1*, *NEXEN*, *CAP2*, and *18S*) (Table 2) were used in SYBR green-based qPCR. Total RNA was isolated from Sham, I/R injured, and I/R injured with iMSC treated mice retina at day 7 postinjury, and gene expression was quantified using TaqMan probe-based gene-specific mouse primers (Table 3) and accompanying Master Mix (Applied Biosystems, Foster City, USA) using QuantStudio 3 (Applied Biosystems) Real-Time PCR system. Data were expressed as relative gene expression or fold mRNA expression using the 2−∆∆Ct method and normalized to *18S* housekeeping gene.

**Table 2.** List of Primer Sequences Used for SYBR green-based qPCR.



**Table 3.** List of Taqman assay IDs for qPCR.

#### *2.6. Mice, Retinal IR Injury, Electroretinography, and Immunohistology*

Animal studies were approved by the Institutional Animal Care and Use Committee, University of Tennessee Health Sciences Center (UTHSC), Memphis (IACUC ID: 20-0152, Approved 16 June 2020) following the guidelines as per the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research. C57BL/6 wild-type (B6) mice, between 12- and 16-weeks-old, were used for the study. Animals were housed under a 12-h light/dark cycle and kept under pathogenfree conditions with unlimited food and water supply. Mice were anesthetized with a mixture of 70–90 mg/kg of ketamine and 0.04–0.08 mg/kg of dexdomitor (Orion Pharma Animal Health, FI-02101 Espoo, Finland). Retinal I/R injury was induced unilaterally in the right eye. The pupil was dilated with 1% tropicamide (Akorn, Inc., Lake Forest, CA, USA), and 0.5% proparacaine hydrochloride (Alcon Laboratories, Inc., Fort Worth, TX, USA) was applied topically onto the cornea. The eye's anterior chamber was cannulated under microscopic guidance with a 32 1/2 inch-gauge needle connected to a silicone infusion line providing a balanced salt solution (Baxter, Deerfield, IL, USA), avoiding injury to the corneal endothelium, iris, and lens. Retinal ischemia was induced by raising intraocular pressure of cannulated eyes to 70 mm Hg (as measured by iCare Tonovet) for 60 min by elevating the saline reservoir. Whitening of the fundus was observed to ensure the induction of retinal ischemia. After 24 h of injury, about 1000 iMSCs/2 µL saline were intravitreally injected into the IR injured eye. Following this, on day 7, ERG (Celeris Rodent Electrophysiology system, Diagnosys LLC, Lowell, MA, USA) was recorded as described in previous publications [25]. Briefly, animals were dark-adapted overnight and anesthetized with ketamine (50 mg/kg) and dexmedetomidine (0.25 mg/kg) cocktail. Pupil dilation was achieved with 1% tropicamide. The electrodes were positioned on the surface of both the corneas. Light pulses were delivered at a frequency at 0.01, 0.1, and 1 cd-s/m2, and the responses were recorded simultaneously from both eyes. All the offline analyses were done with Diagnosys software to calculate b-wave amplitudes. At least three-five responses to light stimuli were averaged to determine the b-wave amplitude. Following ERG, mice were euthanized, enucleated the eye, fixed in paraformaldehyde; the retinal cup was isolated. Retinas were permeabilized and blocked before incubation with the primary antibody. Retinas were immuno-stained with anti-Foxp3 antibody (1:200, Cell signaling) for 48 h followed by incubation with goat antimouse IgG Alexa Fluor 546 secondary antibody. To distinguish vasculature, retinas were incubated with Alexa Fluor 488-Isolectin B (Invitrogen, Carlsbad, USA) and flat-mounted on a glass slide using ProLongTM diamond antifade mountant (Life Technologies). Imaging of retinal flat mounts was examined under a laser scanning confocal microscope (Zeiss LSM 710). The number of Foxp3 positive cells were counted in a blinded fashion from each image from all groups, and results were expressed as Foxp3 positive cells per square millimeter of the retina.

#### *2.7. Statistical Analysis*

Results are expressed as mean ± SEM for all experiments. One-way ANOVA followed by post hoc *t*-tests with the Bonferroni correction was used for multiple group comparisons using GraphPad Prism software 6.0. Comparisons of ERG data between the

groups were performed using the Student's *t*-test. Values of *p* < 0.05 were considered statistically significant.

#### **3. Results**

#### *3.1. Mitochondrial Transfer from MSCs to Immune Cells*

To explore whether MSC can transfer their mitochondria to the immune cells, donor cells (iMSC or ASC) labeled with mitochondria-specific fluorescent probes (CMXRos red) were cocultured with CellTracker (green) labeled THP-1 cells in 1:10 ratio for 24 h (Figure 1A). Fluorescence imaging of cocultures revealed both iMSC and ASC can successfully transfer their mitochondria to recipient (THP-1) cells, as shown by colocalization of red fluorescent mitochondria in green labeled THP-1 cells (yellow, Figure 1B). To study the functional relevance of mitochondria transfer, rotenone challenged MSC were cocultured with primary mouse splenocytes at a 1:10 ratio. Mitochondria-specific fluorescent probe (MitoTracker green) labeled iMSC or ASC were treated with rotenone (500 nm, 2 h) and without, cocultured with primary mouse splenocytes, and analyzed by flow cytometry after 24 h of coculture (Figure 1C). Flow cytometry analysis demonstrated mitochondria transfer from stem cells to splenocytes (Figure 1D). While splenocytes cocultured with native iMSC had 43.4 ± 0.38 percent of MitoTracker green positive splenocytes, rotenone treated cocultures showed only 30 ± 0.81 percent positive cells (Figure 1D; *p* < 0.001). Similarly, native and rotenone treated ASC when cocultured with splenocytes demonstrated 30.4 ± 0.3 and 18.67 ± 1.3 (Figure 1E; *p* < 0.001), percent of MitoTracker green positive splenocytes, respectively. Incubation of primary mouse splenocytes with culture medium obtained by MitoTracker-labeled ASCs, but without ASCs, failed to demonstrate any fluorescence signal in splenocytes (Supplemental Figure S2), ruling out the possibility of passive transfer of the mitochondrial stain to splenocytes due to MitoTracker probe leak. To determine if the changes in mitochondrial content also altered mitochondrial function, bioenergetic analysis was undertaken. THP-1 cells pre-incubated with rotenone significantly blunted oxygen consumption compared to untreated cells as expected (Figure 1F, blue to red line). Upon coculture with iMSC, those cells exposed to rotenone demonstrated near-normal basal respiration. When ATP production was quantified, rotenone treated THP-1 cells showed a reduced ATP production though the data did not reach statistical significance while those cells cocultured with iMSC demonstrated significantly greater ATP production, which is not significantly different from untreated THP-1 cells (Figure 1F; *p* < 0.05).

#### *3.2. MSC Efficiently Transfer Mitochondria in a Dose-Dependent Manner to CD4+ and CD8+ T Cells*

To further explore the mitochondrial transfer to specific cell populations in splenocytes, mitochondria-specific fluorescent probes (MitoTracker green) labeled iMSC or ASC were cocultured with primary mouse splenocytes at different ratios and analyzed by flow cytometry. After 24 h of coculture, nonadherent splenocytes analyzed demonstrated mitochondria transfer from stem cells into all studied lymphocyte subsets, mainly directed to B220+ B lymphocytes (86.8%), T helper CD4+ (50.68%) rather than T cytotoxic CD8+ (10.29%) lymphocytes (Figure 2A and Supplemental Figure S3; *p* < 0.001). Interestingly, donor iMSCs increased mitochondria transfer to T cells (CD4+ and CD8+) with splenocytes at increasing ratios (1:100, 1:25, and 1:10), with an average number of mitochondria of 31.2, 37, and 49.4% respectively in CD4+ T cells and with an average number of mitochondria of 4.6, 5.1 and 9.9% respectively for in CD8+ T cells (Figure 2B; *p* < 0.001). Similarly, when donor ASCs were cocultured with recipient splenocytes at increasing ratios (1:100, 1:25, and 1:10), an average of 31.7, 59.2, and 79.5% cells demonstrated mitochondria in CD4+ T cells and with an average of 5.7, 16.9, and 30.3% for mitochondria in CD8+ T cells, respectively (Figure 2B; *p* < 0.001). The level of mitochondria transferred to B220+ B cells from iMSCs or ASCs cocultured with primary mouse splenocytes ranged from 78–85% with minimal change noted with increasing coculture ratio (Figure 2B; *p* < 0.001). Altogether, these results

indicate that both iMSCs and ASCs can transfer their mitochondria efficiently into primary mouse splenocytes.

**Figure 1.** Both iMSC and ASC effectively transfer mitochondria to immune cells. (**A**) Schematic representation of stem cells and THP-1 coculture setup and fluorescence microscopy analysis. (**B**) MitoTracker CMXRos labeled iMSC (Red) or (**C**) ASC (Red) independently cocultured with cell tracker green-labeled THP-1 cells (Green) for 24 h, stained with nuclear dye DAPI (Blue), and images captured under fluorescence microscope shows increased donor-derived mitochondria in THP-1 cells (yellow; arrows). (**D**) Schematic representation of stem cells and THP-1 coculture setup and flow cytometry analysis. (**E**) Overlays and histograms showing flow cytometry analysis of mitochondria transfer from iMSC and (**F**) ASC to mouse splenocytes cocultured at 1:10 ratio. MitoTracker green-labeled iMSC and ASC pretreated with rotenone (RT) (500 nm, 2 h) and without rotenone treatment cocultured with splenocytes for 24 h. Mitochondria transfer to splenocytes is expressed as percent MitoTracker positive green splenocytes. Flow cytometry data were analyzed using Flowjo (v10.8) software and represented as mean ± SEM from triplicates of same experiment and statistical analysis by one-way ANOVA with Bonferroni correction (\* *p* < 0.01, \*\*\* *p* < 0.0001, ns-not significant). (**G**) Schematic representation of stem cells and THP-1 coculture setup and OCR analysis. iMSC and RT treated THP-1 coculture was set up at 1:1 ratio for 24 h. From coculture, THP-1 were harvested and seeded onto a Cell-Tak coated Seahorse XF-24e V7 PS cell culture microplate, and oxygen consumption rate (OCR) was determined, followed by quantification of basal OCR and ATP production. Data shown as mean ± SEM from a single experiment repeated independently with similar results (\* *p* < 0.01, \*\*\* *p* < 0.0001 one-way ANOVA).

**Figure 2.** Dose-dependent transfer of mitochondria from iMSC or ASC to mouse T and B cells *in vitro*. (**A**) Representative flow cytometry dot plots showing increased MitoTracker positive CD4+ T, CD8+ T, and B220+ cells in splenocyte-iMSC coculture compared to that of monoculture. (**B**) Quantification of percent mitochondria transfer to CD4+ T, CD8+ T, and B220+ B cells increased with decreased ratio with both iMSC (upper panel) and ASC (lower panel). Data shown as mean ± SEM from a single experiment repeated independently with similar results (\*\*\* *p* < 0.0001 one-way ANOVA).

#### *3.3. Tunneling Nanotubes Mediate Mitochondrial Transfer from MSC to THP-1 Cells*

After confirming the mitochondrial transfer from iMSCs or ASCs to immune cells next, we investigated whether F-actin-positive tubular microstructure known as tunneling nanotubes (TNTs) are involved in intercellular mitochondrial transfer from donor to recipient cells. To this end, MitoTracker green-labeled iMSCs or ASCs were cocultured with CMAC blue labeled THP-1 cells with and without CytoB (Figure 3A–D). As expected, fluorescence imaging revealed the transfer of mitochondria (green) from iMSC or ASC into THP-1 cells (blue) via F-actin positive TNT's (red; Figure 3E,G,J). Interestingly, those cells that were pre-incubated with CytoB demonstrated a substantial reduction in mitochondrial transfer to THP-1 cells (Figure 3F,H,I). To further confirm TNT-mediated mitochondrial transfer, cocultures in the presence or absence of CytoB were assessed for the gene expression of CAP2, NEXN, and ACTN1 that are known to associate with F-actin synthesis. Whereas the expression of CAP2, NEXN, and ACTN1 increased by 1.5–45-fold (*p* < 0.001) in cocultures as compared to monocultures, CytoB treatment significantly decreased gene expression of all three markers (Figure 3K; *p* < 0.01). Taken together, the data suggest that mitochondrial transfer from iMSCs or ASCs to immune cells occurs via F-actin-positive tubular tunneling nanotubes.

#### *3.4. MSC Suppresses T Cell Population*

Different T cell populations of splenocytes interact closely; their ratio at a given time results from a balance between their mutual effects. Mitochondrial transfer to recipient cells can increase cell metabolism, resulting in cell division or cell differentiation. To better understand the impact of mitochondrial transfer to immune cells, we first assessed cell viability in mouse splenocyte coculture with either iMSC or ASCs. While the viability of splenocytes in monoculture is 48.7%, the viability in cocultures with iMSC and ASC increased to 65.7 and 64.3%, respectively (Figure 4B; *p* < 0.001). Similarly, the frequency of B220+ B lymphocyte cells in splenocytes cocultures with iMSC and ASC also significantly increased to 64.3 (*p* < 0.001) and 58.5%, respectively (Figure 4B; *p* < 0.01). Next, to better understand the effects of mitochondria transfer from stem cells to immune cells on their subpopulation, we assessed non-adherent splenocytes from cocultures with monoclonal antibodies for CD4, CD8, and B220 and compared them to monoculture in the presence or absence of rotenone. While the helper CD4+ T cells of splenocytes in monoculture is 21%, the coculture levels with iMSC and ASC significantly reduced to 15.4 and 15.9%, respectively (Figure 4B; *p* < 0.001). Similarly, 17% cytotoxic CD8+ T cells in splenocytes monoculture reduced to 12.9 and 12.8% in iMSC and ASC coculture with splenocytes, respectively (Figure 4B; *p* < 0.001). Interestingly, rotenone challenged cocultures demonstrated a small but significant increase in percent immune cells in both iMSC and ASC as compared to cells without rotenone. Altogether, these results indicate that both iMSCs and ASCs suppress the effector T cell population upon transferring their mitochondria to immune cells.

**Figure 3.** Tunneling nanotubes mediate mitochondrial transfer from iMSC to THP-1 cells. (**A**) Schematic representation of experimental design of iMSC or ASC coculture with THP-1 cells in presence and absence of cytochalasin-B. (**B**–**H**) Representative fluorescence microscopy images showing MitoTracker green positive iMSC (**B**), CMAC blue positive THP-1 cells (**C**), F-Actin red positive ASC (**D**), ASC and THP-1 coculture in absence (**E**) and in presence (**F**) of cytochalasin-B and, iMSC and THP-1 coculture in absence (**G**) and presence (**H**) of cytochalasin-B. (**I**) Mitochondria transfer from iMSC or ASC to THP-1 cells reduced with cytochalasin-B. Mean fluorescent intensity (MFI) values calculated using ImageJ. Data shown as mean ± SEM from a single experiment (\*\*\* *p* < 0.001 one-way ANOVA). (**J**) Representative Confocal images of intercellular mitochondrial transfer between iMSC and THP-1 via F-Actin positive nanotubes. Marked area magnified to show MitoTracker positive mitochondria in F-actin positive nanotube. (**K**) qRT-PCR analysis of genes related to nanotubes formation (CAP2, NEXN, and ACTN1) increased significantly during mitochondria transfer from donor (iMSC) to recipient (THP-1) cells. On other hand, cells exposed to cytochalasin-B significantly reduced CAP2, NEXN, and ACTN1 expression. Data shown as mean ± SEM from a single experiment repeated independently with similar results (\*\* *p* < 0.001, \*\*\* *p* < 0.0001 one-way ANOVA).

**Figure 4.** Both iMSC and ASC suppress effector T cell population. (**A**) Representative flow cytometry dot plots showing decreased CD4+ T cells in iMSC-splenocytes and ASC-splenocytes coculture compared to monoculture. (**B**) Quantification data represented as bar graphs after flow cytometry analysis of live cells, CD4+ T cells, CD8+ T cells, and B220+ cells in mouse splenocytes monocultures and cocultures with iMSC, ASC, as well as iMSC, and ASC pretreated with rotenone (500 nm, 2 h) at 1:10 (iMSC/ASC and splenocytes) ratio. Data shown as mean ± SEM from a single experiment repeated independently with similar results (\* *p* < 0.01, \*\* *p* < 0.001, \*\*\* *p* < 0.0001 one-way ANOVA).

#### *3.5. MSC Differentiates T Cells into Tregs and Suppresses CD69 Expression*

The suppression of immune cell activation is one of the manifestations of MSCmediated immunomodulation [12,26]. To assess whether mitochondrial transfer from stem cells to primary mouse splenocytes impacts T-cell differentiation, activated splenocytes were cocultured with iMSCs or ASCs and subsequently assessed for the expression of T regulatory cells (Figure 5A and Supplemental Figure S4). The dot plots from flow cytometry analysis clearly showed double-positive CD25+Foxp3+ cells in the upper right quadrant that were gated on CD4+ T cells (Figure 5B). While the level of CD25+Foxp3+ cells in monoculture is 2.7%, the levels in cocultures with iMSC and ASC increased to 7.1 and 7.4%, respectively (Figure 5B,C; *p* < 0.001). To further confirm the immunosuppression capability of iMSCs and ASCs, the expression of CD69, a potent immune activation marker, was evaluated. While the level of CD69+ cells in monoculture is 18.43%, the levels in cocultures with iMSC and ASC significantly decreased to 7.2 and 6.5%, respectively (Figure 5D; *p* < 0.001). Taken together, our data suggests that both iMSC and ASC manifest their immunomodulation via increased Treg population with a significant reduction in CD69+ cells in splenocyte coculture.

#### *3.6. iMSC Significantly Increases Regulatory T Cells in the Retina of I/R Injured Mice*

MSC are well known to protect against retinal I/R damage [27–30]. To better understand if iMSC also can protect against retinal damage, we tested the intravitreal injection of iMSCs in the retinal I/R injury model in-vivo (Figure 6A). After 7 days, the retinal function, assessed by Electroretinogram (ERG), demonstrated improved b-wave amplitudes in I/R mice receiving iMSC as compared to saline-injected I/R eyes (at 1cd.s.m2 139 ± 48 v/s 48 ± 10 µvolt, *p* = 0.05) (Figure 6B). To further correlate the improved visual function observed with iMSC in I/R mice, gene expression analysis of proinflammatory markers was performed. Figure 6D shows normalized data of individual genes in all 3 groups of mice. I/R mice receiving saline had a significantly increased abundance of gene transcripts involved in microglial activation (IL1β) [31], endothelial activation (VCAM1, CCL2) [31,32], and T-cell regulation (LAMA5) [33] compared to sham mice. Interestingly, I/R mice receiving iMSC significantly ameliorated the increased gene expression (Figure 6C). The normalized fold change in expression of IL1β (I/R, 27.08 ± 8.69 vs. I/R+iMSC, 8.69 ± 1.54, *p* < 0.01), CCL2 (I/R, 13.63 ± 2.7 vs. I/R+iMSC, 3.52 ± 0.96, *p* < 0.0001), LAMA5 (I/R, 2.4 ± 0.37 vs. I/R+iMSC, 1.09 ± 0.09, *p* < 0.001) and VCAM1 (I/R, 7.4 ± 1.48 vs. I/R+iMSC, 2.2 ± 0.96, *p* < 0.01).

**Figure 5.** Both iMSC and ASC increase differentiation of T cells into Tregs and suppress CD69 expression. (**A**) Timeline and experimental details of iMSC/ASC coculture with mouse splenocytes (**B**) Representative flow cytometry dot plots showing increased CD25+Foxp3+ Tregs in iMSC/ASC cocultures with activated splenocytes. (**C**) Quantification of a percent increase in CD25+Foxp3+ cells with (**D**) a reduction in CD69+ cells in cocultures. Data shown as mean ± SEM from a single experiment repeated independently with similar results (\*\*\* *p* < 0.0001 one-way ANOVA).

β Subsequently, to correlate the iMSC ability to provide immunomodulation through increased Tregs, retinal flat mounts analyzed for Tregs in the retina by confocal microscopy revealed positive immunostaining (red) only in I/R and I/R+iMSC groups (Figure 6D). Neither the uninjured contralateral eye nor the iMSC injected into the uninjured eye was positive for Foxp3 expression (Supplemental Figure S5), suggesting the specificity of immunostaining and the correlation of Foxp3 upregulation to I/R injury. While the number of positive Foxp3 cells in the Sham group was 3.6 ± 1.2, the I/R injury retina demonstrated a significant upregulation with 57 ± 8.7 Tregs per mm2 area (*p* < 0.001). Interestingly, those I/R injury animals that received iMSC demonstrated a further significant increase in Tregs to 112 ± 11 cells per mm2 area (*p* < 0.0001; Figure 6E).

β

**Figure 6.** iMSC improves b-wave amplitudes, reduces inflammation correlated with increased regulatory T cells in retina of I/R injured mice. (**A**) Schematic representation of in-vivo experimental timeline and analyses. (**B**) B-wave amplitudes as measured by ERG show an expected decrease in I/R with a significant improvement in iMSC group. Data are shown as mean ± SEM, *n* = 7–13/group. *t*-test. (**C**) An increase in proinflammatory markers in I/R decreased with iMSC. Data are shown as mean ± SEM, one-way ANOVA. (**D**) Immunofluorescence images of Sham, I/R injured, and I/R injured with iMSC stained with isolectin B4 (green) and anti-Foxp3 (red) at day 7 post-injury. White arrows indicate Foxp3 cells. (**E**) Quantification of Foxp3 cells significantly increased in I/R injured animals with iMSC compared to both Sham and I/R. Data shown as mean ± SEM, one-way ANOVA with Bonferroni correction (\* *p* < 0.01, \*\* *p* < 0.001, \*\*\* *p* < 0.0001, ns = not significant).

#### **4. Discussion**

Our study is the first to demonstrate that iMSC reprograms mouse CD4-T cells into Foxp3 regulatory Tregs to the best of our knowledge. Additionally, we show that intravitreal delivery of iMSCs in a retinal ischemia-reperfusion (I/R) injury model display therapeutic potential and suggests the mechanism of action may involve recruitment of Foxp3Tregs. Our in-vitro culture data show that iMSCs transferred mitochondria to immune T cells via F-actin nanotubes, suppressed effector T cells, and promoted differentiation CD4+CD25+Foxp3+ Tregs in coculture with mouse splenocytes on par with ASC. Importantly, we also demonstrated that the increased recruitment of Foxp3+Tregs in the retina correlated with dampened retinal inflammation and improved b-wave amplitudes in the I/R injury model. Our findings are in keeping with published data by others that have demonstrated Tregs act as part of the adaptive immune system, and thereby (A) serve as important regulators of inflammation and play a critical role in immune homeostasis [14]; (B) are inversely correlated with retinal ischemia [34]; and (C) are well known for immunomodulation, with recent evidence suggesting that MSCs can enhance Foxp3+Treg differentiation and stability from activated T cells in part through mitochondrial transfer [12,13,35].

Recent evidence suggests that MSCs can modulate Tregs [12], particularly Tregs expressing the Foxp3 transcription factor, part of the adaptive immune system and principal regulators of inflammation and immune homeostasis [14]. Although Tregs primarily originate in the thymus (tTregs), during tissue damage, the CD4+ (TH) cells at the site of injury can also be reprogrammed and acquire the Treg phenotype, also known as peripheral Tregs (pTregs) and regulate immunotolerance [19]. Vice-versa tTregs acquire CD4+ (TH) cell phenotype and become defective to perform its immunomodulatory function [35]. Thus, the ability of Tregs to migrate to the damaged tissue and dampen inflammation is an attractive strategy to curtail the ongoing inflammation [15–17]. Following this, we show increased Foxp3 positive cells in the retina after retinal damage that further increased with intravitreal injection of iMSC in the ischemic retina. This initial increase in Tregs under acute ischemia compared to sham injury aligns with a previous observation in the oxygeninduced retinopathy mouse model [28]; however, Treg numbers are likely insufficient to repair the retinal damage. On the other hand, those animals that received iMSC probably reached the optimal levels of Foxp3 cells to dampen the retinal damage. One limitation of our initial study is that it is unclear if the increased Foxp3 Tregs are reprogrammed from the local CD4+ T cells that were shown to be upregulated and causally linked to retinal damage [19] or were recruited from elsewhere. Another limitation is that how the increased Foxp3 Tregs regulate retinal tissue inflammation is not explored. Future studies beyond the scope of this study need further analysis on how iMSC induces Tregs and, thus, regulates retinal tissue damage.

Understanding the molecular basis of Treg generation and its stability is an active area of research, with several agents proposed to upregulate and stabilize Foxp3 expression [36,37]. Notably, some agents have even been shown to be effective under hypoxic conditions [38]. MSCs grown as "feeder cells" with Tregs significantly increased Treg cell number, suppressive function, and ex vivo expansion, primarily through mitochondrial transfer from adjacent MSCs, coupled with the promotion of Foxp3 cotranscriptional proteins [13]. Thus, a primary mechanism of action of MSCs may involve an increase in Treg cell number and/or function via Foxp3 stability. To this end, we show increased differentiation of Foxp3 cells from a mixed population of mouse splenocytes cocultured with iMSC, a feature also observed with ASC cocultured with whole human peripheral blood mononuclear cells [12]. One limitation of our current study is that it is unclear if the reprogramming of Foxp3 Tregs primarily occurs through mitochondrial transfer from adjacent iMSCs or other unknown mechanism(s). To this end, genome-wide patterns of DNA methylation of Foxp3 locus were implicated with FOXP3 gene expression, which determines the tTreg generation, pTreg generation, Treg stability, and tTreg and pTreg proliferation [39]. One possibility is that iMSC affects the epigenetic stability of the FOXP3 gene in Tregs, thus might increase its stability. Future studies need to explore these hypotheses.

MSCs help in the intercellular exchange of mitochondria to restore and regenerate the damaged tissue [9], specifically in various ocular cells [10]. Therefore, iMSC by mitochondrial transfer to Tregs or immune cells may modulate them towards more stable and functional pTregs phenotype in-vivo and aid in the suppressive ability to protect from I/R-injury. To this end, we show that iMSCs can transfer mitochondria to CD4+ T-cells and suppress the CD4+ cell population. Interestingly, both these activities are dependent on mitochondria transfer as evidenced by either blocking mitochondrial function via rotenone or the use of Cytochalasin-B, a known agent that blocks actin polymerization that caused a significant reduction in tunneling nanotubes and a significant reduction in mitochondria transfer to recipient cells. In support of our study, a previous study showed such tunnel nanotube-mediated transfer of mitochondria occurs in MSC [13]. The predominant mechanism of cytochalasin B is the inhibition of actin filament polymerization by binding to the end of growing filaments. Since F-actin filaments are dynamic in nature and their formation between donor and recipient cells is influenced by the activity of many actin-binding genes/proteins, we studied the transcriptional regulation of F-actin genes as also shown previously by Jiang et al. [10]. Based on our gene expression data, it is

conceivable that cytochalasin B might indirectly affect transcriptional regulation of the F-actin-related genes and thus influences the mitochondrial transfer. Future studies are required to decipher the causal link of these actin-related genes to mitochondria transfer in our studies as mitochondria from MSCs are transferred to other cells via their exosomes [40] or gap junctions [41,42] or simply by non-mitochondrial paracrine factors [43]. Finally, one alternate hypothesis could be that a direct transfer of mitochondria, a phenomenon known as mitoception, shown for bone-marrow MSC, [44] should be explored for iMSC.

iMSC obtained in the current study were obtained by reprogramming human urinederived epithelial cells with mRNA reprogramming, the fastest and most reliable reprogramming method to date [45]. Furthermore, our study demonstrates an iPSC line that has unlimited proliferation potential [21] and has the ease of obtaining without any surgical interventions will likely address the current challenges of cell therapies in the ischemic retina [46]. Furthermore, since iPSC lines can also be obtained from diseased or aged individuals [47], our studies will likely benefit future personalized medicine studies. In conclusion, our study demonstrates that iPSC-derived MSCs can transfer mitochondria to T cells to enhance differentiation into Foxp3 Tregs. Additionally, our current data indicate that MSC can improve the retina's immune function by upregulation of Tregs to decrease inflammation and reduce I/R injury-induced retinal degeneration.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/cells10113006/s1, Figure S1: Gating strategy for iMSC characterization and representative plots. Overlay plots showing MSC characteristics positive and negative surface markers percent expression on iMSC.; Figure S2: Flow cytometry representative plots showing splenocytes incubated with media ob-tained from mitoTrackergreen stained ASC were negative for MitoTrackerGreen. ASC were stained with mitoTrackerGreen and cultured for 2 h. The cell supernatant was then incubated with splenocytes for 24 h, washed and assessed for mitoTrackerGreen.; Figure S3: Gatingstrategy for flow cytometry analysis and representative plots for transfer of mitochondria from iMSC to mouse T and B cells in vitro.; Figure S4: Representative gating strategy and flow cytometry dot plots showing CD25+Foxp3+ Tregs in iMSC/ASC co-cultures with activated splenocytes.; Figure S5: Retinal flat mounts analyzed for Tregs in the retin the iMSC injected contralateral eye (left) and un-injured cina by confocal microscopy revealed no posi-tive immunostaining ontralateral eye (right) for Foxp3 expression. IsolectinB4 staining was used to identify blood vessels. The data shown is representative of n = 3–4 eyes/group.

**Author Contributions:** Conceived and designed the experiments: M.A., P.K.R., A.B. and R.G.; performed the experiments: M.A., P.K.R. and A.B; analyzed the data: M.A., P.K.R., A.B. and R.G.; contributed reagents/materials/analysis tools: A.B., J.R. and R.G; wrote the paper: M.A., P.K.R., A.B., J.R. and R.G.; reviewed and final approval: all authors. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was funded by grants from the Department of Defense (W81XWH-16-1-0761), National Eye Institute (EY023427), and unrestricted funds from Research to Prevent Blindness to R.G., NIH(R01DK117183) to A.Band NIH (HL141345) to J.R. The funders played no role in the conduct of the study, collection of data, management of the study, analysis of data, interpretation of data, or preparation of the manuscript.

**Institutional Review Board Statement:** UTHSC Institutional Animal Care Committee, approval # 20-0152.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data are available from the authors upon request.

**Acknowledgments:** Authors wish to acknowledge Vinoth Sigamani, MTech, for technical assistance, Sheeja Rajasingh, MSc, Mphil for helpful discussions. Authors wish to acknowledge Vaishnavi Sowmya Narayn, BA, MSc. LIS, for editorial assistance.

**Conflicts of Interest:** R.G. is cofounder and holds equity in Cell Care Therapeutics Inc., whose interest is in using adipose-derived stromal cells in visual disorders. None of the other authors declare any financial conflicts.

#### **References**


### *Article* **Anti-VEGF Drugs Influence Epigenetic Regulation and AMD-Specific Molecular Markers in ARPE-19 Cells**

**Mohamed A. Hamid 1,2,† , M. Tarek Moustafa 1,2,† , Sonali Nashine <sup>1</sup> , Rodrigo Donato Costa 1,3 , Kevin Schneider <sup>1</sup> , Shari R. Atilano <sup>1</sup> , Baruch D. Kuppermann 1,4 and M. Cristina Kenney 1,5,\***


**Abstract:** Our study assesses the effects of anti-VEGF (Vascular Endothelial Growth Factor) drugs and Trichostatin A (TSA), an inhibitor of histone deacetylase (HDAC) activity, on cultured ARPE-19 (Adult Retinal Pigment Epithelial-19) cells that are immortalized human retinal pigment epithelial cells. ARPE-19 cells were treated with the following anti-VEGF drugs: aflibercept, ranibizumab, or bevacizumab at 1× and 2× concentrations of the clinical intravitreal dose (12.5 µL/mL and 25 µL/mL, respectively) and analyzed for transcription profiles of genes associated with the pathogenesis age-related macular degeneration (AMD). HDAC activity was measured using the Fluorometric Histone Deacetylase assay. TSA downregulated *HIF-1α* and *IL-1β* genes, and upregulated *BCL2L13, CASPASE-9,* and *IL-18* genes. TSA alone or bevacizumab plus TSA showed a significant reduction of HDAC activity compared to untreated ARPE-19 cells. Bevacizumab alone did not significantly alter HDAC activity, but increased gene expression of *SOD2, BCL2L13, CASPASE-3,* and *IL-18* and caused downregulation of *HIF-1α* and *IL-18*. Combination of bevacizumab plus TSA increased gene expression of *SOD2, HIF-1α, GPX3A, BCL2L13,* and *CASPASE-3*, and reduced *CASPASE-9* and *IL-β*. In conclusion, we demonstrated that anti-VEGF drugs can: (1) alter expression of genes involved in oxidative stress (*GPX3A* and *SOD2*), inflammation (*IL-18* and *IL-1β*) and apoptosis (*BCL2L13, CASPASE-3,* and *CASPASE-9*), and (2) TSA-induced deacetylation altered transcription for angiogenesis (*HIF-1α*), apoptosis, and inflammation genes.

**Keywords:** AMD; age-related macular degeneration; trichostatin A (TSA); HDAC; histone deacetylase; vascular endothelial growth factor (VEGF)

#### **1. Introduction**

Pathological angiogenesis, which subsequently leads to choroidal neovascularization, subretinal fibrosis, and exudative hemorrhage, is an underlying cause of the severe, latestage, wet form of AMD (Age-related Macular Degeneration) [1].

Wet or neovascular AMD, which accounts for 10–20% of cases, is the less common of the two types of AMD. However, 90% of AMD-associated irreversible vision loss is attributed to wet AMD [2]. Dry AMD is characterized by degeneration of Retinal Pigment Epithelial (RPE) cells and accounts for 80% of AMD cases.

VEGF (Vascular Endothelial Growth Factor) is a signaling growth factor for vascular endothelial cells and a critical angiogenic factor that stimulates ocular neovascularization. Therefore, the most widely used wet AMD treatment targets the pro-angiogenic activity of

**Citation:** Hamid, M.A.; Moustafa, M.T.; Nashine, S.; Costa, R.D.; Schneider, K.; Atilano, S.R.; Kuppermann, B.D.; Kenney, M.C. Anti-VEGF Drugs Influence Epigenetic Regulation and AMD-Specific Molecular Markers in ARPE-19 Cells. *Cells* **2021**, *10*, 878. https://doi.org/10.3390/cells10040878

Academic Editor: Maurice Ptito

Received: 2 February 2021 Accepted: 6 April 2021 Published: 12 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

VEGF to inhibit ocular neovascularization [3]. Administration of intravitreal injections of anti-VEGF drugs, such as ranibizumab (Lucentis), bevacizumab (Avastin), and aflibercept (Eylea), to wet AMD patients is successfully and routinely being used as a wet AMD therapy worldwide [4,5]. Despite the widespread use of anti-VEGF drugs, 10–15% of patients fail to respond to rigorous treatment protocols in clinical trial settings [6–10]. This variability in AMD patients' response to therapy has been attributed to several clinical, behavioral, and genetic factors [11]. Pharmacogenetic studies have identified VEGFA, VEGFR2 (VEGF Receptor 2), CFH (Complement Factor H), and ARMS2 (Age-Related Maculopathy Susceptibility 2) as potential biomarkers for response to anti-VEGF drugs [12]. Many investigators, including the CATT (Comparison of Age-related Macular Degeneration Treatments) and IVAN (Inhibition of VEGF in Age-related choroidal Neovascularisation) research groups, did not find any significant association between genes polymorphism and visual or anatomic responses to treatment [13–17]. This inconsistency in findings by pharmacogenetic studies could be explained in part by possible gene–gene or gene– environmental interactions [18,19].

Genetic and environmental factors contribute to the development and progression of AMD. Genome-wide association studies (GWAS) have identified 52 genetic variants distributed across 34 loci associated with AMD [20]. Furthermore, epigenetic modifications, which include DNA methylation, histone acetylation/deacetylation, non-coding RNA-mediated gene silencing, and chromatin remodeling [21] have been implicated in the pathogenesis of AMD by selective transcription of genes involved in angiogenesis, inflammation, and oxidative stress pathways [22,23]. Epigenetic mechanisms result in covalent modifications in the DNA and regulate gene transcription either by activation or repression, in response to environmental stimuli and are often heritable [24]. Epigenetics can elucidate gene–environment interactions and explain why a certain genotype frequently results in different phenotypes [25]. Histone acetylation is catalyzed by histone transferases (HATs) and acts to destabilize nucleosomes and unwrap DNA to make it accessible to transcription factors. Conversely, histone deacetylation, carried out by histone deacetylases (HDACs), stabilizes nucleosomes and represses DNA transcription [26]. Histone acetylation is known to regulate the expression of 2–10% of genes. Other non-histone proteins, particularly transcription factors, are also regulated by acetylation/deacetylation. This could explain the fact that gene expression is not always silenced by deacetylation [27].

AMD being a leading cause of blindness in the United States and the third major cause of visual impairment worldwide, [28] poses a major health risk to the elderly population, and AMD risk is projected to increase by 54% in the United States in the next five year [29]. Therefore, we speculate that delving into the mechanisms of action of the currently used anti-VEGF drugs might contribute to the design of more effective therapeutic strategies for wet AMD. To this end, the current in vitro study was designed to examine the effects of anti-VEGF drugs on epigenetic regulation in immortalized human ARPE-19 cell lines. The ARPE-19 cell line used in this study was originally developed from the retinal pigment epithelium (RPE) of a human donor eye and resembles the phenotype and properties characteristic of aged native human RPE cells, lack of pigmentation, weak tight junctions, reduced expression of all-trans retinol, Pigment-Epithelium-Derived Factor (PEDF), and RPE markers, and hypersensitivity to VEGF activity, thereby making the ARPE-19 cell line an ideal in vitro AMD model [30]. However, it should be emphasized that the ARPE-19 cell line loses some of the aging RPE characteristics, especially with increasing passages, such as morphology, retinoid metabolism, and VEGF secretion. Furthermore, it should be mentioned that depigmentation in the RPE of AMD eyes is different from that of ARPE-19 cells. Melanosome density in the RPE decreases significantly with normal aging and more evidently in AMD, but melanin is not completely lost. More importantly, RPE melanin in AMD loses its antioxidant properties [31].

RPE cells in vivo form the outer blood-brain barrier and support photoreceptor cells and enable phototransduction. The outer blood-brain barrier is formed of a continuous layer of tight junctions that enable transepithelial transport and phototransduction.

Pigment-Epithelium-Derived Factor (PEDF), which maintains barrier integrity and is endogenously secreted by RPE in vivo in large amounts, shows only basal levels in aged eyes. The in vitro ARPE-19 cell line is a rapidly growing immortalized human cell line derived from primary RPE cells from the globes of a 19-year-old male donor. ARPE-cell lines express RPE cell-specific markers CRALBP and RPE65 and form a viable cuboidal to columnar epithelium monolayer in culture media. Our recent study characterized and confirmed the expression of the following RPE-specific markers in the ARPE-19 cell line used in our lab: Cellular retinaldehyde binding protein-1 (CRALBP), Bestrophin1 (BEST1), and Keratin-18 (KRT18) [32]. Although ARPE-19 cells retain the characteristic features of RPE, including defined cell borders and pigmentation, they require considerable time for differentiation and are unable to completely differentiate into RPE-like layers found in vivo. ARPE-19 cells show partial polarization as some of the cells in the monolayer resemble the morphology of differentiated RPE cells such as apical microvilli, polarized distribution of organelles, basolateral infoldings, and junctional complexes on the apical plasma membrane. ARPE-19 cells exhibit low transepithelial resistance (TER) that reaches a maximum value of 50–100 Ω cm<sup>2</sup> after 28 days of culture in low-serum media in laminin-coated Transwell-COL filters. It is speculated that the low TER might be due to heterogeneity of the cell line since some of the cells show polarization [33].

The present study demonstrated that treatment of ARPE-19 cells with anti-VEGF drugs altered the total HDAC protein activity and the gene expression levels of apoptotic, inflammatory, and oxidative stress-related genes. Moreover, addition of Trichostatin-A (TSA), an HDAC inhibitor, along with an anti-VEGF drug modulated the gene expression of *VEGF*, apoptotic and inflammatory markers. These results suggest that epigenetics modulation in ARPE-19 cells is strongly influenced by anti-VEGF drug treatment.

#### **2. Materials and Methods**

#### *2.1. ARPE-19 Cell Culture*

Human RPE cells (ARPE-19 cells, ATCC, Manassas, VA, USA) were cultured till confluent in 175 cm<sup>2</sup> -flasks containing DMEM/F-12 culture medium (Dulbecco's Modification of Eagle's Medium, Mediatech, Inc., Manassas, VA, USA), 10% dialyzed fetal bovine serum, antibiotics (penicillin G 100 U/mL, streptomycin sulfate 0.1 mg/mL, gentamicin 10 µg/mL, amphotericin B 2.5 µg/mL) and 17.5 mM glucose. All ARPE-19 cells were at passage 10 and cultured side-by-side under identical standard conditions of 37 ◦C in 5% CO<sup>2</sup> and 95% relative humidity, in order to avoid any potential technical variability.

ARPE-19 cells are a spontaneously arising RPE cell line derived by Amy Aotaki-Keen from the normal eyes of a 19-year-old male who died from head trauma in a motor vehicle accident in 1986. The ARPE-19 cell line, established using the cuboidal basal cell layer cultured in specific culture media, expresses the RPE-specific markers cellular retinaldehyde binding protein and RPE-65.

In our study we utilized ARPE-19 cells at passage 10 in all our experiments to ensure the cells retained acceptable fidelity.

#### *2.2. Drug Treatment of ARPE-19 Cell Cultures*

ARPE-19 cells were plated in triplicate for 24 h in 6-well plates at a density of 500,000 cells per well, culture media were removed and replaced with the same media containing anti-VEGF drugs: aflibercept, ranibizumab or bevacizumab at 1× and 2× concentrations of the clinical intravitreal dose (12.5 µL/mL and 25 µL/mL, respectively). The clinical dose was calculated by assuming that the amount of each drug clinically used in intravitreal injections distributes equally throughout the 4 mL human vitreous. Untreated cells were used as control.

In order to further explore the potential relationship between anti-angiogenic treatment and the acetylation status of the target genes expression, a subset of ARPE-19 cells was treated with trichostatin A (TSA), an inhibitor of histone deacetylase (HDAC) activity, at 0.3 µM concentration, or a combination of 1X bevacizumab plus 0.3 µM TSA.

The control and anti-VEGF treated cultures were incubated for an additional 24 h, then RNA was isolated to be used for Quantitative Real-Time PCR (qRT-PCR) analyses.

Proteins extracted from cultures of untreated ARPE-19 cells, as well as cells treated with 1X bevacizumab, 0.3 µM TSA, and a combination of both drugs were analyzed for HDAC activity as described below.

#### *2.3. RNA Extraction, Amplification of cDNA, and Quantitative Real-Time PCR (qRT-PCR) Analysis*

ARPE-19 cells were pelleted for RNA isolation using a PureLink RNA Mini Kit (Ambion, Thermo Fisher Scientific, Waltham, MA, USA). For qRT-PCR analyses, 100 ng of individual RNA samples were reverse transcribed into complementary DNA (cDNA) using SuperScript VILO Master Mix (Thermo Fisher Scientific, Waltham, MA, USA).

We investigated transcription profiles of downstream genes known to play a role in AMD pathogenesis. RNA samples were isolated from cells that were (a) untreated; (b) treated with 1× or 2× concentrations of the three anti-VEGF drugs; (c) treated with 0.3µM TSA alone; and (d) treated with 1× bevacizumab plus TSA (bevacizumab/TSA). qRT-PCR was performed using primers for downstream pathway genes, including angiogenesis (*VEGF-A* and *HIF-1α*), apoptosis (*BCL2L13, CASPASE-3,* and *CASPASE-9*), inflammation (*IL-18* and *IL-1β*) and oxidative stress (*GPX3A* and *SOD2*) (Table 1). The qRT-PCR was performed on individual samples using QuantiFast SYBR Green PCR Kit (Qiagen, Inc., Germantown, MD, USA) on StepOne Plus Real-Time PCR system (Thermo Fisher Scientific, Waltham, MA, USA). For the various target genes, housekeeping genes that had comparable amplification efficiencies were chosen in order to maximize the accuracy of our ∆∆CT values. The housekeeper genes were either hypoxanthine phosphorbosyltransferase 1 (HPRT1) or hydroxymethylbilane synthase (HMBS). Untreated samples were used as control. ∆∆Cts (differences in cycle thresholds) were obtained and folds calculated using the formula 2∆∆Ct .

#### *2.4. Protein Extraction and HDAC Activity Assay*

ARPE-19 cell samples were lysed using RIPA buffer (Cat. #89900, Life Technologies Corp., Calsbad, CA, USA), supernatants were transferred to a new microfuge tube and protein concentrations were measured using the Bio-Rad DC Protein Assay system (Bio-Rad Laboratories, Richmond, CA, USA) according to the manufacturer's instructions. Protein samples were kept in a −80 ◦C freezer until time of use in the HDAC activity assay.

HDAC activity in the protein samples was measured using the Fluorometric Histone Deacetylase Assay Kit (Sigma–Aldrich, St. Louis, MO, USA) according to the manufacturer's protocol. Briefly, protein samples, assay buffer and HDAC substrate solution were added to the wells of a 96-well plate. Each well contained 20 µL of the protein sample, 30 µL of assay buffer and 50 µL of HDAC substrate solution. The plate was incubated at 30 ◦C for 30 min, then 10 µL of Developer solution added to each well. The plate was incubated 10 min at room temperature, and fluorescence measured with a microplate spectrofluorometer (Gemini XPS, Molecular Devices, Sunnyvale, CA, USA) using an excitation wavelength of 350 nm and an emission wavelength of 440 nm. Samples were plated in triplicate and Hela cell lysate used as a positive control for HDAC activity. Experiments were performed on three replicates i.e., in triplicate. The plate in which the cells were plated was read three times and the fluorescence intensity was averaged. The entire experiment, i.e., treatment and reading the fluorescence was repeated three times to ensure reproducibility. Protein samples from untreated ARPE-19 cells were used as control and were normalized to 100%.

#### *2.5. Statistical Analyses*

Statistical analyses of the data were performed by unpaired t-test using GraphPad Prism, Version 5 (GraphPad Software, Inc., La Jolla, CA, USA). *p* < 0.05 was considered statistically significant. Untreated samples (controls) were normalized to a value of 100% for comparison to treated samples.


**Table 1.** Gene symbols, names, gene bank accession numbers, and functions.

**<sup>a</sup>** Official gene symbol by HUGO (Human Genome Organization) Gene Nomenclature Committee (HGNC). **<sup>b</sup>** Official gene name by HUGO Gene Nomenclature Committee (HGNC). **<sup>c</sup>** Gene Accession Bank Number from the primers used (Qiagen, Valencia, CA). **<sup>d</sup>** Gene function modified from PubMed gene.

#### **3. Results**

*3.1. Measurement of HDAC Activity in ARPE-19 Cells before and after Treatment with Anti-VEGF and TSA*

Treatment of ARPE-19 cells with 0.3 µM TSA resulted in significant reduction of HDAC activity (*p* = 0.0003), as did the combination of bevacizumab 1× plus TSA (*p* < 0.0001) (Figure 1). Although treatment with bevacizumab 1X alone did not significantly alter HDAC activity (*p* = 0.15), its addition to TSA significantly potentiated its inhibition of HDAC activity (*p* = 0.0001).

**Figure 1.** Histone deacetylase (HDAC) activity as determined by relative fluorescence (%) in untreated and treated ARPE-19 cultures. \*\*\* *p* < 0.001; Bars with no asterisk represent nonsignificant difference. UNT: Untreated; B: Bevacizumab 1× conc.; T: Trichostatin A 0.3 µM; B+T: Bevacizumab 1× conc. + Trichostatin A 0.3 µM. HeLa: Hela cell lysate positive control for HDAC activity. Error bars represent 'Mean ± SEM'.

#### *3.2. Effect of Anti-VEGF Treatment and HDAC Inhibition on the Expression Profiles of Downstream Genes*

Aflibercept 1×-treated ARPE-19 cells showed upregulation of *VEGF-A* (1.12-fold, *p* = 0.67), *HIF-1α* (1.03-fold, *p* = 0.63), *GPX3A* (1.13-fold, *p* = 0.49), *BCL2L13* (1.26-fold, *p* = 0.05), *CASPASE-3* (1.21-fold, *p* = 0.04), *CASPASE-9* (1.32-fold, *p* = 0.003) and *IL-1β* (1.38-fold, *p* = 0.22); and downregulation of *SOD2* (0.76-fold, *p* = 0.14) and *IL-18* (0.95-fold, *p* = 0.051), compared to untreated cells (Table 2, Figure 2).

Treatment with aflibercept 2× downregulated the expression of *VEGF-A* (0.98-fold, *p* = 0.94), *HIF-1α* (0.67-fold, *p* = 0.002), *SOD2* (0.52-fold, *p* = 0.005), *GPX3A* (0.96-fold, *p* = 0.87), and *IL-18* (0.78-fold, *p* = 0.002), and led to upregulation of *BCL2L13* (1.21-fold, *p* < 0.0001), *CASPASE-3* (1.36-fold, *p* = 0.006), *CASPASE-9* (1.19-fold, *p* = 0.004) and *IL-1β* (1.14-fold, *p* = 0.55) (Table 2, Figure 2).

ARPE-19 cells treated with ranibizumab 1× showed downregulation of *VEGF-A* (0.75 fold, *p* = 0.40), *HIF-1α* (0.85-fold, *p* = 0.05), *GPX3A* (0.96-fold, *p* = 0.89), and *IL-1β* (0.84-fold, *p* = 0.43) and upregulation of *SOD2* (1.22-fold, *p* = 0.16), *BCL2L13* (1.21-fold, *p* > 0.0001), *CASPASE-3* (1.73-fold, *p* = 0.004), *CASPASE-9* (1.61-fold, *p* < 0.0001) and *IL-18* (1.67-fold, *p* < 0.0001) (Table 2, Figure 2).


**Table 2.** Expression folds for downstream genes in untreated and anti-vascular endothelial growth factor (anti-VEGF)-treated ARPE-19 cultures \*.

\* Fold change was calculated using the formula: 2∆∆CT. Untreated samples had a value of 1.

Ranibizumab 2×-treated cells showed downregulation of *VEGF-A* (0.72-fold, *p* = 0.20), *HIF-1α* (0.76-fold, *p* = 0.008) and *IL-1β* (0.54-fold, *p* = 0.04), and upregulation of *SOD2* (1.28-fold, *p* = 0.09), *GPX3A* (1.09-fold, *p* = 0.60), *BCL2L13* (1.4-fold, *p* < 0.0001), *CASPASE-3* (1.26-fold, *p* = 0.04), *CASPASE-9* (2.82-fold, *p* < 0.0001) and *IL-18* (1.32-fold, *p* = 0.002) (Table 2, Figure 2).

Treatment with bevacizumab 1× decreased the expression of *VEGF-A* (0.95-fold, *p* = 0.80), *HIF-1α* (0.85-fold, *p* = 0.26) and *IL-1β* (0.81-fold, *p* = 0.42), and increased the expression of *SOD2* (1.8-fold, *p* = 0.009), *GPX3A* (1.28-fold, *p* = 0.23), *BCL2L13* (1.8-fold, *p* < 0.0001), *CASPASE-3* (1.5-fold, *p* = 0.003), *CASPASE-9* (1.04-fold, *p* = 0.52) and *IL-18* (1.33-fold, *p* = 0.0003) (Table 2, Figure 2).

Bevacizumab 2× resulted in downregulation of *VEGF-A* (0.82-fold, *p* = 0.41), *HIF-1α* (0.63-fold, *p* = 0.001), *GPX3A* (0.91-fold, *p* = 0.63), *IL-18* (0.86-fold, *p* = 0.02), and *IL-1β* (0.68-fold, *p* = 0.12); and upregulation of *SOD2* (1.54-fold, *p* = 0.06), *BCL2L13* (1.07-fold, *p* = 0.16), *CASPASE-3* (1.04-fold, *p* = 0.39), and *CASPASE-9* (2.16-fold, *p* < 0.0001) (Table 2, Figure 2).

TSA treated (0.3 µM) ARPE-19 cultures resulted in downregulation of *VEGF-A* (0.75 fold, *p* = 0.24), *HIF-1α* (0.66-fold, *p* = 0.001) and *IL-1β* (0.2-fold, *p* = 0.001), and caused upregulation of *SOD2* (1.69-fold, *p* = 0.11), *GPX3A* (2.11-fold, *p* = 0.06), *BCL2L13* (1.13-fold, *p* = 0.0003), *CASPASE-3* (1.14-fold, *p* = 0.07), *CASPASE-9* (1.49-fold, *p* = 0.003) and *IL-18* (1.69-fold, *p* < 0.0001) compared to untreated cells (Table 3, Figure 3).

The combination of bevacizumab 1× plus TSA resulted in upregulation of *VEGF-A* (1.42-fold, *p* = 0.25), *HIF-1α* (1.60-fold, *p* = 0.006) *SOD2* (1.97-fold, *p* = 0.005), *GPX3A* (4.03 fold, *p* = 0.03), *BCL2L13* (1.28-fold, *p* = 0.0007) and *CASPASE-3* (1.2-fold, *p* = 0.05), and decreased expression of *CASPASE-9* (0.59-fold, *p* = 0.09), *IL-18* (0.89-fold, *p* = 0.09) and *IL-1β* (0.37-fold, *p* = 0.008) compared to untreated cells (Table 3, Figure 3).

When comparing cells treated with both drugs to cells treated with TSA alone, the addition of bevacizumab 1× significantly reversed the effect of TSA on the expression of *VEGF-A* (*p* = 0.02), *HIF-1α* (*p* = 0.0003), *CASPASE-9* (*p* < 0.0001), and *IL-18* (*p* = 0.0003), reduced its effect on an *IL-1β* (*p* = 0.0006), and potentiated its effect on *BCL2L13* (*p* = 0.008) (Figure 3).

**Figure 2.** Quantitative Real-Time PCR (qPCR) data showing Delta Cts for downstream genes in untreated and anti-VEGF treated ARPE-19 cultures. \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001; Bars with no asterisk represent nonsignificant difference. UNT: Untreated; A1×: Aflibercept 1× conc.; A2×: Aflibercept 2× conc.; R1×: Ranibizumab 1× conc.; R2×: Ranibizumab 2× conc.; B1×: Bevacizumab 1× conc.; B2×: Bevacizumab 2× conc. Error bars represent 'Mean ± SEM'.



**Table 3.** *Cont.*


\* Fold change was calculated using the formula: 2∆∆CT. Untreated samples had a value of 1.

**Figure 3.** Quantitative Real-Time PCR (qPCR) data showing Delta Cts for downstream genes after treatment with Trichostatin A and Bevacizumab. \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001; Bars with no asterisk represent nonsignificant difference. UNT: Untreated; TSA: Trichostatin A 0.3 µM; B1×+TSA: Bevacizumab 1× conc. + Trichostatin A 0.3 µM. Error bars represent 'Mean ± SEM'.

#### **4. Discussion**

In this study, we demonstrated differential HDAC total protein activity in response to anti-VEGF drug treatment in ARPE-19 cells. Treatment of ARPE-19 cells with anti-VEGF drugs alone in varying concentrations as well as additive effects of TSA and anti-VEGF drug significantly modulated the gene expression profiles of apoptotic, inflammatory, and oxidative stress markers. Therefore, our results suggest that addition of anti-VEGF drugs to cultured ARPE-19 cells strongly influence the regulation of epigenetic markers and downstream molecular markers.

Environmental stimuli are thought to induce epigenetic changes that accumulate in the cell with increasing age, as evidenced by age being the major risk factor for AMD, as well as the discordance of disease phenotype in identical twins that possess similar risk profiles for AMD [34]. Hypermethylation of genes encoding for reactive oxygen species (ROS) scavengers, namely *GSTM1* and *GSTM5*, was demonstrated in AMD RPE and choroid, downregulating their expression and rendering the cells more susceptible to oxidative damage. Additionally, several micro RNAs (miRNAs) have been implicated in AMD pathogenesis through their contribution to aberrant angiogenesis, inflammation, and apoptosis in response to oxidative stress both in vitro and in vivo [22]. Hypomethylation of the promoter region of interleukin 17 receptor C (*IL17RC*), a pro-inflammatory gene, promotes its expression and increases inflammation in AMD patients [35]. *Oliver* et al. conducted the first genome-wide epigenetic study in AMD and found hypomethylation at *HTRA/ARMS2* locus, which is a major susceptibility locus for AMD. They also observed hypermethylation of the *PRSS50* locus which had not been previously associated with AMD [23].

Over the last decade, intravitreal injection of anti-angiogenic drugs has been the mainstay of therapy for neovascular AMD as well as for macular edema associated with diabetic retinopathy and retinal vein occlusion [36]. The three drugs commonly used in clinical practice are bevacizumab, ranibizumab and aflibercept. Bevacizumab is a fulllength monoclonal antibody that has a molecular weight of 149 kilodalton (kDa) and binds all isoforms of VEGF-A rendering them inactive. Ranibizumab is an antigen-binding Fab fragment of the same parent antibody as bevacizumab. It lacks the Fc domain and has a molecular weight of 48 kDa. Aflibercept, a 96.9 kDa recombinant fusion protein, contains immunoglobulin fragments from both VEGF receptors: VEGFR1 and 2, combined with an Fc antibody fragment. It acts as a decoy receptor that not only binds VEGF-A, but also, unlike the former two drugs, can bind VEGF-B and placental-like growth factor (PlGF) [37–39]. The different molecular weights and structures of the three drugs may have an effect on their ocular and systemic pharmacokinetics and pharmacodynamics [40]. All three drugs have been shown to accumulate intracellularly. Uptake of bevacizumab and aflibercept is mediated through their Fc portion, but ranibizumab is likely internalized through a VEGF/VEGFR2-mediated mechanism [30].

RPE dysfunction is central to the pathogenesis of AMD, and RPE cells are the main source of VEGF in the retina [41]. Many investigators have tested the safety of anti-VEGF agents on RPE cells both in vitro and in animal models. Our group previously demonstrated a good safety profile of the three anti-angiogenic drugs on ARPE-19 cell cultures at the clinical dose, but mild cytotoxic effects were found at higher doses [42,43]. Other studies have shown similar results [44,45].

A study by Dinc et al. in which APRE-19 cells were subjected to H2O2-induced oxidative stress and levels of miRNA expression were evaluated, suggested an epigenetic role for anti-VEGF drugs [46]. In that study, several miRNAs were dysregulated in response to oxidative stress compared to untreated samples. Preincubation of cells with any of the three anti-VEGF drugs before H2O<sup>2</sup> treatment significantly altered the miRNA dysregulation induced by H2O2. The authors suggested that anti-VEGF drugs could protect RPE cells from oxidative stress through their effect on miRNAs.

We tried to determine whether the changes in *HDAC* expression would affect the activity of HDAC enzymes in ARPE-19 cells using an assay kit designed to measure the collective activity of all HDACs. Bevacizumab treatment alone failed to alter HDAC activity to a significant extent compared to untreated cultures. Our inability to observe a significant net effect on HDAC enzyme activity may be because bevacizumab had opposite effects on the expression of both *HDAC1* and *HDAC6* genes, so the overall HDAC activity after treatment was neutral. As expected, Trichostatin A (TSA) significantly inhibited HDAC activity. TSA is known to inhibit both class I and II HDACs. Interestingly, the combination of TSA plus bevacizumab increased significantly the frank inhibitory effect on HDAC activity, but the mechanisms for additive effects are not clear and need to be further investigated.

Next, the expression levels for genes known to regulate important pathogenetic pathways for AMD after treatment with anti-VEGF drugs and TSA were measured.

#### *4.1. Angiogenesis Genes*

HIF-1α is the oxygen-sensitive subunit of HIF-1, a transcription factor upregulated in cells in response to hypoxia. Activation of HIF-1α results in overexpression of downstream genes, including pro-angiogenic genes, mainly VEGF-A [47]. Both class I and II HDACs are upregulated by hypoxia and induce angiogenesis, particularly HDAC1 [48,49]. HIF-1α is a non-histone protein target for acetylation. It is acetylated under normoxic conditions, which decreases its stability by facilitating its proteasomal degradation. Under hypoxic conditions, HIF-1α becomes deacetylated by HDACs, which makes it more stable and prolongs its half-life [50].

We found that all three anti-VEGF drugs downregulated the expression of *HIF1-α* gene at 2×, but not 1×, concentration. This suggests that these drugs might protect cells against oxidative damage. Treatment with TSA had the same inhibitory effect. This observation comes in agreement with previous studies demonstrating that TSA could downregulate both gene and protein expression of HIF-1α in vitro and in animal models [51,52]. Other HDAC inhibitors demonstrated a protective effect against oxidative damage in vitro and animal models. These include valproic acid (VPA), which is another broad-spectrum HDACi, and tubastatin A (TST); a specific inhibitor of HDAC6 [53,54]. The inhibitory effect displayed by anti-VEGF drugs and TSA in our study was reversed by the combination of bevacizumab and TSA, which significantly upregulated *HIF1-α* expression.

VEGF-A is the main pro-angiogenic factor implicated in the pathogenesis of wet AMD [55]. The expression levels of VEGF, as well as tissue response to its secretion, are modulated by both genetic and epigenetic factors. Epigenetic regulation of angiogenesis has been extensively studied in cancer cells. Both HDAC7 and SIRT1 promote angiogenesis during development and disease by inducing pro-angiogenic factors and suppressing anti-angiogenic ones [56–58]. HDAC6 has also been associated with angiogenesis due to its ability to bind several cytoskeletal proteins in the cytoplasm and stimulate vascular proliferation and sprouting. The anti-angiogenic properties of HDAC inhibitors are the basis for their use as anti-cancer agents and extend beyond gene silencing to directly acetylating angiogenic factors in the cytoplasm. Chan and colleagues demonstrated that TSA could downregulate *VEGF-A* by epigenetically silencing its expression in the presence of CoCl2, a hypoxia-inducing agent, in human RPE cells in vitro. Furthermore, they showed that TSA could attenuate a laser induced CNV in a mouse model. However, we found that TSA did not alter *VEGF-A* gene expression in ARPE-19 cells. The ability of TSA to silence gene expression might be more pronounced under conditions of hypoxic stress that induce aberrant upregulation of VEGF-A, which was not the case in our experiment.

Two studies explored the effect of either ranibizumab or bevacizumab on *VEGF-A* gene expression in primary RPE cells. In one experiment, bevacizumab had no effect on the baseline expression levels of *VEGF-A* [59]. Similarly, in another study ranibizumab treatment did not significantly alter *VEGF-A* mRNA overexpression induced by white light illumination of RPE cells [60]. In ARPE-19 cells, both aflibercept and ranibizumab induced *VEGF-A* mRNA expression after 24 h of treatment. The authors postulated that the cells upregulated the transcription of *VEGF-A* to compensate for blocking the VEGF-A protein in the culture media [43]. Similarly, another study demonstrated that ranibizumab and bevacizumab induced overexpression of *VEGF-A* inARPE-19 cells that were subjected to oxidative stress by preincubation in a hypoxic chamber [61]. Another study showed a similar compensatory upregulation in response to bevacizumab treatment in murine RPE cells in vivo, but not in ARPE-19 cells in vitro. These investigators suggested that the compensatory upregulation of *VEGF-A* expression in complex in vivo systems might not be captured by the simpler APRE-19 cell model. It could also be that the RPE of healthy young animals was able to compensate for VEGF-A neutralization, unlike ARPE-19 cells, which carry more similarities to aged RPE [62]. Other studies showed that anti-VEGF drugs could bring down *VEGF-A* gene expression to control levels in human retinal pericytes [63] and in ARPE-19 cells [19].

*VEGF-A* expression levels were not affected by any of the used drugs in our experiment, although its expression was significantly lower in cells treated with a combination of bevacizumab plus TSA compared to treatment with TSA alone. It seems that the effect of anti-angiogenic drugs on *VEGF-A* gene expression varies according to cell type, the nature of biologic stress that the cells are exposed to as well as the in vivo versus in vitro environment.

#### *4.2. Oxidative Stress Genes*

Oxidative stress occurs when ROS accumulation overwhelms the capacity of the cell to detoxify them. Neural and RPE cells in the retina have a high metabolic demand and are most prone to oxidative damage with aging. Aging is associated with differential gene expression and chromatin reorganization mediated by epigenetic mechanisms, ultimately leading to impaired ability of the cells to adapt to environmental stress [64].

The main antioxidant in the retina is the superoxide dismutase (SOD) family [65]. *SOD2* gene encodes for a mitochondrial enzyme, which deactivates superoxide free radicals. Polymorphisms of *SOD2* have been associated with exudative AMD [66]. SOD2 expression depends on acetylation. Being a mitochondrial enzyme, SOD2 is directly deacetylated by SIRT3, a mitochondrial Class III HDAC, resulting in its activation [67]. Forkhead box O3a (FoxO3a) activates the promoter of *SOD2* gene inducing its expression. TSA was shown to increase the acetylation of *FoxO3a* promoter region, and upregulating its expression, as well as its target protein, SOD2, expression in vitro [68]. Treatment with bevacizumab 1× in our study significantly upregulated *SOD2* expression, as did combined treatment with bevacizumab and TSA. Conversely, aflibercept significantly reduced *SOD2* expression.

Glutathione peroxidase (GPX) is another antioxidant enzyme found in the RPE and photoreceptors that protects the retina from oxidative damage. GPX expression is upregulated in AMD patients, most likely due to oxidative stress [69]. VPA induced the expression of *SOD2* and *GPX* genes in ARPE-19 cells in normal conditions and maintained their expression hypoxic conditions. A similar effect was demonstrated in rat retina [70]. In our study, only the combination of bevacizumab plus TSA was able to significantly increase *GPX3A* expression.

#### *4.3. Inflammation Genes*

Inflammation has been recently recognized as a key player in the pathogenesis of AMD. Both drusen components and intracellular lipofuscin can incite inflammasome activation and the release of the pro-inflammatory cytokines IL-18 and IL-1β in retinal tissues [71,72]. It is thought that epigenetic mechanisms are involved in initiating the immune response by altering the gene expression of immune cells to allow for cytokine production and chemotaxis [24]. HDAC6 inhibition by TST suppressed mRNA expression of *IL-1β* in an inflammation model of mammary epithelial cells in vitro [73]. We found that TSA treatment had the same effect on *IL-1β* gene expression in ARPE-19 cells. Cultures treated with ranibizumab 2× and the TSA+bevacizumab combination also displayed downregulation of *IL-1β*.

Although inflammation significantly contributes to tissue damage in AMD, a certain degree of inflammation might be needed to protect against neovascularization. IL-18 has exhibited anti-angiogenic properties in tumors and post-ischemic injury and is being investigated as a potential anti-angiogenic therapy in wet AMD [74]. Shen et al. demonstrated reciprocal suppression between VEGF and IL-18. They were able to detect increased levels of IL-18 in the aqueous of patients receiving intravitreal ranibizumab injections for macular edema secondary to retinal vein occlusion. Furthermore, they found that intravitreal injection of anti-VEGF antibody in a mouse model of ischemic retinopathy upregulated mRNA expression of *IL-18*. IL-18 was able to block VEGF-induced vascular leakage and neovascularization in mice. Thus, each agent can suppress both the production and function of the other [75]. This observation could explain our findings that *IL-18* was significantly upregulated by both concentrations of ranibizumab and bevacizumab. TSA also upregulated *IL-18* expression, possibly owing to its anti-angiogenic properties. Aflibercept 2×, however, suppressed its expression. Aflibercept has a different molecular structure than both ranibizumab and bevacizumab and could have triggered another signaling mechanism that reduced *IL-18* transcription.

#### *4.4. Apoptosis Genes*

RPE cell loss characteristic of advanced AMD is thought to represent cell death by apoptosis. The BCL-2 family regulates the intrinsic mitochondrial pathway of apoptosis. Accumulation of oxidized low-density lipoproteins (LDL) in drusen and basal linear deposits, both hallmark lesions of AMD, results in upregulation of BAX, a pro-apoptotic BCL-2 family member, and downregulation of BCL-2, an anti-apoptotic member. The increased BAX/BCL-2 ratio tips the balance in favor of apoptosis [76]. Activation of pro-apoptotic BCL-2 members results in opening of mitochondrial membrane pores with subsequent release of pro-apoptotic factors, such as cytochrome c, into the cytoplasm. Cytochrome c can recruit and activate caspase-9, an initiator caspase, which activates effector caspases, such as caspase-3, that eventually cause degradation of genomic DNA and cell death [77].

Anti-VEGF drug treatment of ARPE-19 cells resulted in upregulation of the 4 proapoptotic genes in this study. This effect was seen with the three drugs tested at both 1× and 2× concentrations. We previously demonstrated some degree of reduced mitochondrial membrane potential (MMP), which is an early sign of apoptosis, in ARPE-19 cells after 24 h of treatment with higher-than-clinical concentrations of ranibizumab and aflibercept. Only bevacizumab decreased MMP at 1× concentration [42]. Another study found that bevacizumab significantly increased apoptosis in an ARPE-19 cell model of oxidative stress and as stress levels increased, the dose of bevacizumab capable of inducing apoptosis decreased. The authors postulated that bevacizumab blocked the protective effects of VEGF under high oxidative stress conditions and downregulated *BCL-2* gene expression [78]. Another study showed that ranibizumab could enhance the anti-proliferation effects of oxidative stress on ARPE-19 cells [42].

These results warrant further in vivo investigations since the net effect on retinal cells in vivo is subject to a complex interplay of many protective and detrimental factors. Antiangiogenic therapy has demonstrated protective effects in our in vitro study, as evidenced by suppressing oxidative stress and inflammatory cytokine gene transcription. However, further research is warranted as concerns have been raised about the development of geographic atrophy in 98% of wet AMD patients receiving chronic anti-VEGF injections over prolonged periods of time [79].

#### **5. Conclusions**

In conclusion, we demonstrated that anti-VEGF drugs can (1) alter expression profiles for genes involved in oxidative stress, inflammation, and apoptosis pathways and (2) modulate intracellular signal transduction in addition to blocking VEGF-A. This could have implications in management of resistance or nonresponse to anti-VEGF therapy in

some AMD patients. The phenomenon of individual variation in response to anti-VEGF treatment has also been observed in different cancers treated with bevacizumab [80–82]. Genetic variations may render vascular tissue more responsive or resistant to drug effects. Epigenetic mechanisms may render tissues less sensitive to the anti-VEGF treatment and influence pharmacogenetic interactions, as evidenced by miRNA regulation of enzymes involves in drug uptake and metabolism [83,84]. The fact that these drugs could influence the epigenome might guide precision medicine in the future by obtaining an "epigenetic" profile for wet AMD patients to predict resistance and direct the choice of therapy.

Another avenue we believe is worth exploring is the possibility of adding HDAC inhibitors to the therapeutic armamentarium of AMD. Epigenetic drugs have shown a great promise in immunomodulation, neuroprotection, and angiogenesis suppression [85]. To date, 6 epigenetic drugs have been approved by the FDA for cancer therapy [51]. A variety of epigenetic drugs, including DNMT and HDAC inhibitors are currently under investigation as potential therapeutic agents in AMD, owing to their ability to reverse inflammation and angiogenesis [55,86]. Specific HDAC inhibitors might be preferable to pan-inhibitors, such as TSA, as the latter can cause undesirable alterations in gene expression by inducing histone hyperacetylation [53].

Further studies including in vivo tests are required. Other retinal cell types are involved in the evolution of AMD pathogenesis, and the impact of the studied drugs on these cells needs to be explored as well. Human pluripotent stem cell (hPSC)-derived retinal organoids could provide an alternative platform to study drug interactions and intracellular signaling mechanisms that more closely approximates the retinal environment in vivo [87].

**Author Contributions:** Conceptualization: M.C.K. and B.D.K.; methodology: M.A.H., M.T.M., S.N., R.D.C., K.S. and S.R.A.; investigation: M.A.H., M.T.M., S.N., R.D.C., K.S. and S.R.A.; validation: M.A.H., M.T.M. and S.N.; data curation: M.A.H., M.T.M., S.N. and R.D.C.; formal analysis: M.A.H. and M.T.M.; resources: M.C.K. and B.D.K.; writing—original draft preparation: M.A.H., M.T.M. and S.N.; writing—review and editing: S.N., M.C.K. and B.D.K.; supervision: B.D.K. and M.C.K.; project administration: B.D.K. and M.C.K.; funding acquisition: B.D.K. and M.C.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Discovery Eye Foundation, Polly and Michael Smith Foundation, Iris and B. Gerald Cantor Foundation, Beckman Initiative for Macular Research, Max Factor Family Foundation, and Guenther Foundation. This study was also supported in part by an unrestricted grant from Research to Prevent Blindness, Inc. The authors acknowledge the support of the Institute for Clinical and Translational Science grant number: ICTS (ULI TR001414/TR/NCATS) at University of California, Irvine.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All data are presented within the manuscript.

**Acknowledgments:** NEI grant: NEI R01 EY027363 (MCK).

**Conflicts of Interest:** B.D.K. is consultant to Alcon, Alimera, Allegro, Allergan, Genentech, Glaukos, GSK, Neurotech, Novagali, Novartis, Ophthotech, Pfizer, Regeneron, Santen, SecondSight, Teva, ThromboGenics. The rest of the authors declare that they have no competing interests. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**

