Diabetic retinopathy (DR) is the leading cause of blindness in working-age adults. Early stage DR involves inflammation, vascular leakage, apoptosis of vascular cells and neurodegeneration. In this study, we hypothesized that cells derived from the stromal fraction of adipose tissue (ASC) could therapeutically rescue early stage DR features. Streptozotocin (STZ) induced diabetic athymic nude rats received single intravitreal injection of human ASC into one eye and saline into the other eye. Two months post onset of diabetes, administration of ASC significantly improved “b” wave amplitude (as measured by electroretinogram) within 1–3 weeks of injection compared to saline treated diabetic eyes. Subsequently, retinal histopathological evaluation revealed a significant decrease in vascular leakage and apoptotic cells around the retinal vessels in the diabetic eyes that received ASC compared to the eyes that received saline injection. In addition, molecular analyses have shown down-regulation in inflammatory gene expression in diabetic retina that received ASC compared to eyes that received saline. Interestingly, ASC were found to be localized near retinal vessels at higher densities than seen in age matched non-diabetic retina that received ASC. In vitro, ASC displayed sustained proliferation and decreased apoptosis under hyperglycemic stress. In addition, ASC in co-culture with retinal endothelial cells enhance endothelial survival and collaborate to form vascular networks. Taken together, our findings suggest that ASC are able to rescue the neural retina from hyperglycemia-induced degeneration, resulting in importantly improved visual function. Our pre-clinical studies support the translational development of adipose stem cell-based therapy for DR to address both retinal capillary and neurodegeneration.


Diabetic retinopathy (DR) is the most common vascular complication in patients with long-standing diabetes, and is the leading cause of blindness in working-age adults. The estimated prevalence in the USA is 5.4% (~7.7 million) [1]. Future projections suggest that DR will become a larger public health problem, with an increase in the obese population as well as an increase in the prevalence of diabetes [2]. In the early stages of DR (non-proliferative DR; NPDR), clinically significant microvascular changes have been reported which presumably develop concomitant with pericyte loss, basement membrane thickening, and endothelial dysfunction involving loss of barrier integrity [3]. These changes cause the leakage of exudative material into the macula resulting in visual loss. It is becoming increasingly clear that neuronal cells of the retina also are affected by diabetes, resulting in visual dysfunction and even degeneration of some neuronal cells [4], [5]. Diabetes causes metabolic and physiologic abnormalities in the retina, and these changes suggest a role for inflammation in the development of DR. Using pharmacologic inhibitors or genetically modified animals, development of at least the early stages of DR, especially occlusion and degeneration of retinal capillaries has been documented[6], [7]. More recently, proinflamatory proteins released from neutrophils and monocytes have been shown to be critical in this process [8]. The development of such degenerate capillaries is preceded by the loss of pericytes (specialized perivascular cells), which are believed to provide a nourishing, anti-inflammatory and anti-angiogenic environment for endothelial cells [9]. Furthermore, the formation of the blood-retinal barrier (BRB) is dependent on the interaction of the vascular endothelial cells with both glial cells and pericytes [10]. Therefore loss of pericytes has been suggested to play a key role in the development of DR [11].

Current strategies for the therapeutic management of DR include symptomatic treatments such as laser photocoagulation, intravitreal triamcinolone and intravitreal injection of VEGF neutralizing agents (e.g., Ranibizumab), but these therapies achieve only limited success [12]. Furthermore, there are no treatments for reversing the severe ischemic changes that occur with more advanced disease and/or macular ischemia, the primary cause of reduced visual function/blindness. The Diabetes Control and Complications Trial, one of the largest clinical trials conducted between 1983–1993, suggested that insulin treatment can delay the onset and progression of diabetic complications in about 54% of the patients [13]. However, once poor glycemia occurs, subsequent normoglycemia will not prevent progression towards DR, suggesting that hyperglycemia-induced chronic cellular changes are difficult to reverse [14].

In recent years, the concept of repairing terminally differentiated organs with a cell-based therapy has evolved. The cells used in these approaches are diverse and include tissue-specific endogenous stem cells, endothelial progenitor cells, and bone-marrow derived mesenchymal stem cells. Among cell-based approaches intended to address DR, intravitreal injection of endothelial [15]_ENREF_11 and myeloid progenitor cells [16] have been shown to prevent vascular regression and protect neurons in genetic mouse models of retinal degeneration. Yet to date, the identity and cell surface markers expressed by these cells remain incompletely defined [17]. Further, most of these studies have been performed in non-diabetic models [eg. oxygen induced retinopathy (OIR) and retinal ischemia-reperfusion injury] or genetic models of retinal degeneration that are not diabetic models. Consequently, we and others turned attention to adipose stem cells (ASC), which have functional and phenotypic overlap with pericytes lining microvessels in adipose tissues [18], [19], and which also form robust functional vascular networks in vivo by cooperation of ASC with cord blood endothelial cells [20]. Mendel et al recently reported that indeed ASC-derived cells can integrate with retinal vasculature, adapting both pericyte morphology and marker expression, and provide functional vascular protection in multiple murine models of retinal vasculopathy [21]. Although this is a novel observation of direct intravitreal injection compared to the previously described intravenous injection of ASC [22], the use of OIR mice and the Akimba mouse model of DR do not represent true long-term hyperglycemia induced DR models [23]. This prompted us to use the more robust Streptozotocin—induced DR model to test the perivascular integration of ASC to rescue the capillary damage.

Apart from their role as perivascular cells, ASC are also known to produce a variety of angiogenic and antiapoptotic factors [24]. We and others have shown that ASC act in a paracrine manner, as well as by direct physical interaction with endothelial cells, to modulate angiogenesis [25]_ENREF_1, reduce skeletal muscle ischemia and tissue loss [24]_ENREF_3, limit myocardial infarction [26]_ENREF_4, promote skin repair [27], and provide neuroprotective function [28]. These seminal studies have led to the concept that ASC may not only rescue the retina from diabetic capillary damage, but also from neurodegeneration by suppressing inflammation and apoptosis. To test this hypothesis here, we employed a rat model of streptozotocin (STZ)-induced chronic diabetes model to address whether ASC ameliorates not only the structural abnormalities of early DR, but also improves neuronal activity (as assessed by electroretinogram) to enhance visual function. Additionally, we addressed mechanisms by which ASC withstand hyperglycemic stress in vitro, and thus may protect retinal endothelial cells in vivo. This is an important question, as transplanted cells typically undergo significant cell death, which could hamper the potential benefit of cell transplantation, specifically considering hyperglycemic environment observed in the diabetic vitreous [29].

Research Design & Methods
Isolation and characterization of human ASCs

Studies involving human adipose tissue sample collection were approved by Indiana University School of Medicine Institutional Review Board. The adipose specimens were obtained from elective surgical procedures and are deemed normal medical waste products resulting from these procedures. Therefore, collection of de-identified specimens was exempted from informed consent requirements. Human subcutaneous adipose tissue samples obtained from lipoaspiration procedure were processed to isolate ASC as described previously [25]. In brief, the fat tissue was digested in collagenase type I solution (Worthington Biochemical, Lakewood, NJ) under agitation for 1 hour at 37°C and centrifuged at 300 g for 8 minutes to separate the stromal cell fraction (pellet) from adipocytes. The pellet was re-suspended in DMEM/F12 containing 10% FBS (Hyclone, Thermofisher.com) filtered through 250 µm Nitex filters (Sefar America Inc, Depew, NY) and centrifuged at 300 g for 8 minutes. The cell pellet was treated with red cell lysis buffer (154 mmol/L NH4Cl, 10 mmol/L KHCO3, 0.1 mmol/L EDTA) for 10 minutes. The final pellet was suspended in EBM-2/5% FBS or in EGM2-MV (Cambrex, East Rutherford, NJ).

ASC cultured for 2 days on culture plastic were harvested with 2 mmol/L EDTA/PBS and routinely checked for both pericyte and mesenchymal cell surface markers including CD10+/CD13+/CD31−/CD34−/CD44+/CD45−/CD73​+/CD90+/CD105+[30].

ASC were labeled with lentiviral GFP per standard procedures. Briefly, ASC were plated at 10000 cells/cm2 in EGM-2-MV and three hours later the media on ASC was exchanged with fresh media supplemented with CSCGW-EGFP lentiviral stock solution (pCSCGW-EGFP construct provided by Dr Ken Cornetta, Indiana University Vector Production Facility) and 8 µg/ml of polybrene (Sigma-Aldrich, St Louis, MO). The next day the media on the cells was replaced with fresh culture media, cells were expanded, and sorted for the ones that demonstrated strong expression of fluorescent proteins using FACS Aria Sorter (BD biosciences, Franklin Lakes, NJ). Intact ASC were used during sorting procedure to establish cutoff for autofluorescence. As an additional method, ASC were also labeled with DiI per manufacturers protocol (Vybrant® DiI Cell-Labeling Solution, Life technologies, Carlsbad, CA).

Diabetic rat model and intravitreal injections

All animal studies were approved by IACUC, Indiana University School of Medicine as per the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Athymic nude rats (Hsd:RH-Foxn1rnu) were obtained from Harlan Laboratories (Indianapolis, IN) at 5–6 weeks of age. Animals were maintained in a specific pathogen-free environment in positive pressure rooms with a standard 12 hour day/12 hour night cycle. All animals were fed a normal pellet chow (Harlan Teklad, Madison, WI). After acclimatization, diabetes was induced by a single intra-peritoneal injection of a freshly prepared solution of STZ in citrate buffer (pH 4.5) at 55 mg/kg of body weight. Diabetes was confirmed within two weeks with blood glucose (Bayer Contour test strips, Bayer HealthCare LLC, Pittsburgh, PA) levels higher than 350 mg/dL on two consecutive days. Intraperitoneal glucose tolerance tests (GTT) and morphometric assessment of pancreatic islet β-cell mass were performed as previously described [31], [32]. Body weight was measured regularly. Insulin was given (0–3 times per week, 0–2 units SC of NPH insulin, Eli Lilly, Indianapolis) to achieve weight maintenance without preventing hyperglycemia (>250 but <450 mg/dL) and glucosuria. Thus, diabetic rats were insulin deficient but not grossly catabolic. Two months after diabetes onset, rats were anesthetized with isoflurane and intravitreal injections (50,000 to 250,000 of GFP-labeled ASC in 2 µL saline, typically right eye) were performed with a 30-gauge microsyringe (Hamilton, Reno, NV), using a temporal approach, 2 mm posterior and parallel to the limbus. The left eye received an equal volume of saline and served as control.
Quantitation of retinal pathology

Two months post diabetes induction, rats were euthanized and enucleated eyes were preserved in 10% neutral buffered formalin. The retinal vasculature was isolated from formalin-fixed eyes using the trypsin digest technique as described by us previously [33]. After drying the purified vessel network onto a glass slide, the preparations were stained with hematoxylin and periodic acid Schiff. Acellular capillaries were quantitated in 4–7 field areas in the mid-retina (200× magnification) in a masked manner. Acellular capillaries were identified as capillary-sized vessel tubes having no nuclei anywhere along their length, and were reported per square millimeter of retinal area. Pericyte ghosts were estimated from the prevalence of protruding bumps in the capillary basement membranes from which pericytes had disappeared [33]. At least 1,000 capillary cells (endothelial cells and pericytes) in 5 field areas in the mid-retina (400× magnification) in a masked manner were examined. Ghosts on any already acellular vessel were excluded.

Albumin Extravasation Assay

Vascular permeability in athymic nude rats was assessed by the albumin extravasation assay method as published previously [34]. Briefly, athymic nude rats after two months post diabetes induction or age matched normal rats were anesthetized and received tail vein injection of FITC-BSA (100 mg/Kg body weight, Sigma-Aldrich) One hour after injection rats were euthanized, perfused with 4% paraformaldehyde, eyes were enucleated and embedded in Tissue-Tek CRYO-OCT Compound (Thermo Fisher scientific, Inc). Frozen sections (5 µm) were cut throughout the retina to obtain 7 sections per animal with 30 µm apart. Using an epifluorescence microscope, extravasation of FITC-BSA from retinal vessels was captured and quantified using Image J software (NIH.gov). The fluorescence values were then normalized to the plasma level of FITC determined by fluorometer (Molecular Devices, Sunnyvale, CA).
Realtime RT-qPCR analysis of mRNA expression

Seven days after ASC or saline injection, rats were euthanized; enucleated eyes and retinas were collected free of vitreous and flash frozen for mRNA analysis. Each individual retina’s were processed for total RNA using NucleoSpin RNA II Kit (Clontech, Mountain View, CA). About 50 ng of RNA was mixed with SYBR green mix containing rat gene specific primers (see Table S1) as per the manufacturer’s instructions (iScript One-Step RT-PCR Kit with SYBR, Bio-Rad, Hercules, CA). Samples were analyzed on Applied Biosystems StepOne™ Real-Time PCR (Applied Biosystems, Life technologies, Carlsbad, CA) in a total reaction volume of 20 µL. The thermal cycling program consisted of an initial 10 minutes cDNA synthesis at 50°C followed by 15 minutes Thermo-start activation. The amplification included denaturation at 95°C, followed by 40 cycles of denaturation at 95°C for 15 seconds, annealing at 58°C for 30 seconds, and melting curve analysis. To compare the levels of rat DR gene transcripts between the diabetic saline treated v/s diabetic ASC treated, we used comparative CP method for relative quantification as described previously [35] and expressed compared to saline treated non-diabetic rats. The amount of target gene transcript, normalized to the elongation factor alpha (EF1α)endogenous housekeeping gene transcript and relative to the calibrator, was computed by 2−ΔΔCP, where ΔΔCP = ΔCP (unknown target gene) − ΔCP (calibrator), and ΔCP of target or calibrator is the CP of the target gene subtracted from the CP of the housekeeping gene.

TUNEL assay for apoptosis

Frozen rat retinal sections were assessed for DNA strand breaks by TUNEL assay as per the manufacturer’s instructions (ApopTag Plus In Situ Apoptosis Fluorescein Detection Kit, EMD Millipore, Billerica, MA). In some cases where FITC-BSA or GFP cells were injected, a modified HRP detection system with anti-digoxigenin HRP (1:500; Roche, Indianapolis, IN) was adopted. In each experiment, adjacent sections incubated without TdT served as control. The total number of TUNEL-positive cells in diabetic rats was normalized to total nuclear cells using MetaMorph analysis (Molecular Devices, Sunnyvale, CA) and shown as a percentage of non-diabetic animals that received saline injection.
Immunohistochemical analysis

Frozen sections of retina (5–10 µm) were fixed in 2% paraformaldehyde for 10 minutes. For blocking non-specific background staining tissues were exposed to serum free Protein Block (Dako, Carpinteria, CA) for 45 minutes followed by incubation for overnight at 4°C with the anti-alpha smooth muscle actin antibody (αSMA, Clone 1A4, mouse IgG, AbCam, Cambridge, MA; 1:200), anti-histone human IgG (Clone: AE-4; monoclonal mouse antibody, AbCam; 1:400) and anti-Von Willebrand factor (vWF, Rabbit Polyclonal Ab, AbCam, 1:200) antibodies. This was followed by washing and incubation with secondary antibody (Alexa Fluor® 647 goat anti-mouse IgG and Alexa Fluor® 546 rabbit anti-rabbit IgG, Life Technologies at 1:1000). Retinal tissues without exposure to the primary antibody were used as controls for immunostaining. Stained tissues were counterstained with DAPI and mounted using fluorescent mounting media (Sigma) and visualized using a Nikon Eclipse 80i upright digital microscope (Nikon Instruments Inc., Melville, NY).
Retinal wholemounts and confocal microscopy

Rats were euthanized by CO2 inhalation, and the eyes were enucleated and fixed with 2% paraformaldehyde. Retinas were dissected and set in 24-well cell culture plates. After washing, retinal tissues were blocked with serum free Protein block (Dako) and permeabilized with 0.3% Triton X-100 in PBS at room temperature for 1 to 2hrs. Samples were then incubated overnight in the dark at 4°C with different combinations of antibodies. Vasculature was labeled with either Alexa Fluor® 488 conjugated Isolectin GS-IB4 from Griffonia simplicifolia (Life Technologies) or rabbit polyclonal collagen IV antibody (Abcam). ASC were identified by GFP or labeled with human IgG antibody (Life Technologies). Pericytes were labeled with mouse monoclonal alpha smooth muscle actin antibody [1A4]. After washing, tissues were incubated with Alexa Fluor® 647 donkey anti-mouse IgG and Alexa Fluor® 546 conjugated goat anti-rabbit IgG antibodies for 3–4 hrs in the dark at 4°C. Retinal tissues were then counterstained with nuclear DAPI and were flat mounted with vitreous side up (Fluorescent mounting media, Sigma) on clean glass slides. Retinal flat mounts were examined under a confocal scanning laser microscope (Olympus FV1000-MPE, Center Valley, PA) configured to eliminate autofluorescence and spectral overlap allowing precise discrimination between the fluorochromes imaged. Z-stacks of confocal images of retinal wholemounts were reconstructed and analyzed (Olympus Fluoview 3.0 software).

To assess retinal function in DR model, we performed dark adapted image guided flash focal-electroretinogram (ERG, Micron III, Phoenix Research Labs, Pleasanton, CA) intensity response series in anesthetized rats before and after intravitreal injections of ASC as described previously with slight modifications [36]. Briefly, rats after 2 hr of dark adaptation anesthetized with ketamine and xylazine cocktail. Pupils were dilated with 1% tropicamide and ERG was performed on each rat sequentially beginning with the weakest excitation gradually increasing intensity. At least two regions (nasal and temporal) of the retina were targeted using the deep red real-time retinal image from Micron III camera as a guide. The reference needle was placed between two eyes and the grounding probe was inserted into the base of the tail. Corneal electrode attached to the Micron III camera lens was used to record ERG data. Roughly about 250 µm diameter of the retina was targeted with a 6-ms pulse to obtain twenty traces of readings at different light intensities beginning at 1.1 cd-second/m2 that doubled in intensity until reaching 1.5×1026 cd-second/m2. Using Labscribe2 version 2.34 (iWorx Systems Inc, Dover, NH) the amplitude (implicit time) of the b-wave was measured from the trough to the peak of the first visible b-wave.
Co-culture of retinal endothelial cells and ASC

For vascular network formation (VNF) assay, Human Retinal Microvascular Endothelial Cells (HREC; ACBRI 181, Cell Systems Corporation, Kirkland, WA) and ASC were co-cultured according to the protocol published previously [37]. Mixture of 10,000 of HREC and 60,000 of ASC (per cm2) were re-suspended and cultured in EBM-2/5% FBS medium for 6 days with media exchange at day 3. At the end of the experiment, vascular networks were visualized by staining the cultures with biotinylated Ulex Europaeus Agglutinin I (Vector labs, Burlingame, CA) and anti-αSMA IgG as previously described [37]. On an average of 9 images of each well were captured with 4× objective and processed for total tube length by Angiogenesis Tube formation assay module of MetaMorph software (Moleculardevices.com). Human cord blood derived endothelial cells served as a positive control cell type for HREC as described by us previously [37].

In contact independent co-cultures, HREC (60,000 cells) were plated on the lower surface of 24-well Transwell plates, and ASC (60,000) were plated in Transwell 0.4-µm-pore inserts. After adherence of cells to the designated surfaces, inserts with ASC were moved into the wells containing HREC. Co-cultures were cultivated under EBM-2/0.1% FBS for 3 days with medium change daily with varying doses of glucose (5.5 mM, 25 mM and 55 mM) or mannitol as an osmolality control. At the end of day 3, cells from the bottom were analyzed for cell viability/apoptosis.
Assessment of Cell viability

In vitro ASC viability upon exposure to high glucose were assessed by proliferation and apoptosis assays. Proliferation was assessed by Cell Proliferation Assay kit based on the cleavage of the tetrazolium salt WST-1 to formazan by cellular mitochondrial dehydrogenases (EMD Millipore). Briefly, about 10000 ASC were plated in a well of a 96-well flat bottom plates and left for 2 hours for attachment. Following this, the medium was removed and 200 µl of the varying doses of glucose & mannitol were distributed into each well and incubated for upto 72hrs at 37°C, 5% CO2 incubator. After incubation, 10 µL of the WST-1 dye working solution were added to the wells and plates were incubated for further 4 hours. The absorbance (A) values of each well were read at 460 nm using an automatic microplate reader (Flexstation, Molecular devices). The percentage viability was calculated using the background-corrected absorbance as follows: % viability = [(A of experimental well)/A of control well] *100.

As an additional method for cell viability, apoptotic levels in ASC were measured after exposure to increased doses of glucose and mannitol as described above. After 24 hours, cells were fixed in a glyoxal based formalin-free fixative (Prefer Ready-to-use, Anatech Ltd, Battle Creek, MI) for 20 minutes and washed with PBS. Subsequently, cells were immunostained with active caspase-3 antibody (affinity-purified rabbit polyclonal antibody, Promega, Madison, WI; 1:500) and detected with an Alexa Fluor 546 conjugated rabbit antibody. The total fluorescent intensity in a given well was computed as a ratio of total nuclear DAPI positive cells and assessed percent apoptotic rate in cells treated with glucose compared to normal glucose control. As a positive control to induce apoptosis, staurosporine (Cell Signaling, Danvers, MA) at 1 µM was used in both the experiments while normal human dermal fibroblast (HDF) cells were used a negative cell type for ASC which also originates from stroma of mesynchyme.
Statistical Analysis

Data analysis from in vivo are expressed as mean ± SEM of a group of n≥6–8. In vitro cell culture experimental data is shown as mean ± SD of triplicate measurements and repeated independently three additional times. Statistical significance was determined by Student’s t test or one way ANOVA using GraphPad Prism software (La Jolla, CA). A probability value p<0.05 was considered statistically significant.
Development of diabetes in athymic nude rats is accompanied by features of early stage retinopathy

Athymic nude rats (6 weeks old) treated with 55 mg/kg of STZ developed sustained hyperglycemia when compared to non-diabetic control rats. While blood glucose levels in non-diabetic animals were 97±15 mg/dL, diabetic animals had significantly elevated levels (413±55 mg/dL, p<0.01) as early as two days to a week after STZ, and remained >250 mg/dL until after two months of STZ injections (Fig 1a). Approximately two months post diabetes induction and prior to ASC injections, GTT was performed to document glucose intolerance and results were compared to age-matched non-diabetic rats. As expected STZ-treated (Fig 1b, squares) rats developed impaired glucose tolerance compared to non-diabetic controls (Fig 1b, round). Fasted STZ rats given 1 g/kg body weight of D-Glucose demonstrated elevated blood glucose levels of >250 mg/dL throughout the period compared to non-diabetic rats which cleared the glucose load within the two hours. Consistent with this, pancreatic islet β cell area in diabetic rats was significantly decreased (p<0.05) compared to non-diabetic rats (Figure S1 in File S1). Diabetic rats in this study were hyperglycemic and failed to gain weight at a normal rate compared to age matched non-diabetic controls (Fig 1c). Taken together, it is suggestive that these athymic nude rats develop diabetes when induced with STZ. Interestingly, intravitreal injection of ASC had no effect on elevated random blood glucose level after 3 weeks post transplantation, but demonstrated a slight natural expected increase in body weights in these rats (Figure S2 in File S1).