Ischemic Vascular Damage Can Be Repaired by Healthy, but Not Diabetic, Endothelial Progenitor Cells

Circulating CD34⁺ cells from pooled healthy human donors were purchased from Cambrex (East Rutherford, NJ). Cells from diabetic donors were isolated from patients visiting clinics at Shands Teaching Hospital at the University of Florida in accordance with an approved protocol by the institutional review board. Sixty milliliters of blood were aseptically collected into heparinized glass vacuum tubes and reacted with magnetic bead–conjugated anti-CD34 and separated according to manufacturer's directions (Stem Cell Technologies, Vancouver, BC, Canada). Enriched cells were then washed three times with PBS. Immediately before administering cells to the animals, the cells were fluorescently labeled with the membrane dye PKH-67, according to manufacturer's instructions (Sigma-Aldrich, St. Louis, MO).

All animal studies (Table 1) were approved by the institutional animal care and use committee at the University of Florida, and studies were conducted in accordance with the principles described in the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. Normal female C57BL/6J mice were purchased from Jackson Laboratories (Bar Harbor, ME). Obese male diabetic BBZDR/Wor rats and age- and sex-matched lean nondiabetic control littermates were obtained from Biomedical Research Models (Worcester, MA). Timed pregnant C57BL/6J mice (Jackson Laboratories) were obtained at E11-12. At study termination, the animals were killed by overdose of ketamine and xylazine (14 and 30 mg/kg, respectively) followed by thoracotomy, at which time the eyes were removed for immunonohistochemical processing (described below).

A previously establish model of OIR was used (16). Briefly, mouse pups (n = 8) at postnatal day (P) 7 were placed in 75% O₂ along with their nursing dam for 5 days (P12), after which they were returned to normal air. In this model, central retinal vaso-obliteration occurs during hyperoxia exposure. Neovascularization then occurs upon return to normoxia and peaks at P17. Labeled CD34⁺ cells were administered by intravitreal injection at P12 (see below), and the animals were killed at P14.

Mice (n = 20) were kept under inhalation anesthesia (isoflurane vapor) during induction of ischemia. The anterior chamber of the eye was cannulated with a 30-gauge needle attached to an infusion line of sterile saline, and the eye was subjected to 2 h of hydrostatic pressure (80–90 mmHg, Tono Pen; Medtronic Solan, Jacksonville, FL) on the anterior chamber. This resulted in retinal ischemia, as confirmed by whitening of the iris and loss of the red reflex. After 2 h, the needle was withdrawn and the intraocular pressure was normalized, resulting in reperfusion injury. The contralateral eye served as a control. Seven days after the insult, at which time retinal capillary damage was appreciable (17), the animals were given isolated labeled CD34⁺ cells by either intravitreal (n = 10) or systemic (n = 10) injection. These groups were subdivided further randomly to receive either diabetic huCD34⁺ EPCs (n = 5 intravitreous; n = 5 systemic) or nondiabetic huCD34⁺ EPCs (n = 5 intravitreous; n = 5 systemic).

C57BL/6J mice were made diabetic by STZ injection (n = 12), as described (18), and were used at 6–12 months of diabetes (mean 9). Beginning at ∼6 months of diabetes, these mice developed capillary degeneration, including acellular capillaries, similar to the histological changes associated with long-term human diabetes (18). At the time of use, the animals randomly were divided into two groups: intravitreous (n = 6) or systemic (n = 6) administration of labeled huCD34⁺ cells. Each of these groups was further subdivided randomly to receive either diabetic CD34⁺ cells (n = 3 intravitreous; n = 3 systemic) or nondiabetic CD34⁺ cells (n = 3 intravitreous; n = 3 systemic).

Obese BBZDR/WOR rats (n = 12) were diabetic for at least 3 months (range 98–361 days, mean 179.3 ± 22.9 days, median 126 days). These rats developed many characteristics of the early stages of diabetic retinopathy after 3–4 months of diabetes, including capillary degeneration, pericyte loss, and blood retinal barrier breakdown (19–21). The diabetic rats and their lean control littermates (n = 12) were equally divided to receive either diabetic CD34⁺ cells (n = 6 diabetic rats; n = 6 control rats) or nondiabetic CD34⁺ cells (n = 6 diabetic rats; n = 6 control rats) by intravitreous injection in one eye.

All animals were kept under inhalation anesthesia (as described above) while administering the cells. For all models, one set of experiments was conducted with an intraocular route of administration, giving 5 × 10⁴ cells/eye to the mice or 1.5 × 10⁵ cells/eye to the rats, all in 1 μl injection volume. A sterile 5-μl Hamilton syringe attached to a 32-gauge A-bevel needle was used to inject the cells into the vitreous directly anterior to the neural retina. In a second series of experiments using the STZ-induced diabetic and I/R insult mice, the cells were injected systemically via retro-orbital sinus (2.5 × 10⁵ cells/animal in 100 μl injection volume). The neonatal mice in the OIR model were injected only with cells from healthy donors, whereas all of the adult mice as well as the rats were equally subdivided to receive cells from either healthy or diabetic donors. The OIR animals were killed 2 days after injection with CD34⁺ cells. The adult mice in both the chronic and acute vascular injury models were killed at either 1 or 2 days after receiving the cells. The diabetic rats and their lean control counterparts were immune suppressed by intramuscular injection of cyclosporine (2 mg · kg⁻¹ · day⁻¹), beginning 1 day before administering the cells and continuing for the duration of the study. The rats were killed either 2 or 7 days after receiving the labeled CD34⁺ cells.

All of the mice in the OIR model, as well as selected mice in both the I/R and STZ-induced diabetic models (as indicated in results) were perfusion fixed via left ventricular puncture at the time of euthanasia, with 5 ml rhodamine-conjugated dextran (Sigma-Aldrich) in 4% buffered paraformaldehyde (50 mg/ml) to show vessel patency. Regardless of whether the animal was perfusion fixed, after removal all eyes were perforated with a 30-gauge needle and immersion fixed in 4% (wt/vol) buffered paraformaldehyde for 45 min then washed in three changes of PBS.

Four eyes from each experimental condition in all but the OIR and BBZ rat models were dehydrated in 2.5 mol/l sucrose and then embedded in optimal cutting temperature medium for cryosectioning. At least 50 sections (10-μm thickness, every 10th section kept) were collected and postfixed in acetone for 5 min.. The sections then were immersed in freshly prepared NaBH₄ (1 mg/ml in PBS) to reduce background autofluorescence and blocked in PBS containing 2% (wt/vol) nonfat dry milk and 2% (wt/vol) BSA. The sections then were reacted (with appropriate washes between incubations) with monoclonal rat anti-human CD31 (Abcam, Cambridge, MA) and monoclonal mouse anti-endothelium (Clone PAL-E; Abcam), both diluted 1:100 in PBS containing 1% (wt/vol) nonimmune rabbit serum and 1% nonimmune goat serum, followed by rhodamine-conjugated rabbit anti-rat IgG (Abcam) and fluorescein isothiocyanate–conjugated goat anti-mouse IgG (Abcam), both diluted 1:200 in PBS. Sections incubated without primary antibody, but with secondary antibody, were used as controls. The slides then were mounted with Vectashield antifade medium (Vector Laboratories), and digital image captures were made with a Zeiss Axioplan 2 epifluorescence microscope coupled to a Spot charge-coupled device camera.

The neural retina from all of the remaining eyes was dissected and permeabilized by overnight immersion in detergent buffer (10 mmol/l HEPES, 150 mmol/l NaCl, 0.2% [vol/vol] Triton X-100, 2% BSA, pH 8) at 4°C. Eyes from animals that were not perfused then were reacted with rhodamine-conjugated R. communis agglutinin I (1:1,000 in 10 mmol/l HEPES, pH 8; Vector Laboratories) to detect vasculature; this step was not performed for eyes from animals that had been perfusion fixed with rhodamine-conjugated dextran. The whole retinas then were mounted flat with antifade medium, and digital image captures were made with a laser scanning confocal microscope (LSCM) (BioRad MRC 1024; BioRad, Temecula, CA) and with an epifluorescence microscope (Zeiss Axioplan 2; Carl Zeiss, Thornwood, NY). ImageJ software (ImageJ 1.37n; Wayne Rasband, NIH [available at http://rsb.info.nih.gov/ij/index.html]) was used for analysis of the confocal images.

Intensity correlation analysis (22) was performed on LSCM z-section captures from the I/R and STZ-induced diabetic models in order to assess the percentage of labeled huCD34⁺ cells that incorporated into retinal vasculature. Briefly, software was used to determine total red, total green, and colocalized red and green pixel areas in each corresponding spatially calibrated z-section. The percentage of colocalization then was calculated relative to total vascular area for that field. At least three random fields (each 288,519 μm²) from each retina (n = 3, STZ-induced diabetic/nondiabetic cells/intravitreous; n = 3, STZ-induced diabetic/nondiabetic cells/systemic; n = 3, STZ-induced diabetic/diabetic cells/intravitreous; n = 3, STZ-induced diabetic/diabetic cells/systemic; n = 3, I/R with nondiabetic cells; n = 3, I/R with diabetic cells) were assessed. SPSS statistical analysis software (SPSS, Chicago, IL) was used to analyze the datasets to determine mean percent colocalized area ± SE from the mean. Significance was calculated by independent samples t test assuming equal variances.

Two days after their return to normoxia, neonatal mice in the OIR model showed extensive central vessel obliteration, vasoconstriction, poor perfusion, and other pathological abnormalities, as has been extensively described previously (Fig. 1A) (16). Those mice that were given labeled huCD34⁺ cells, however, demonstrated robust, patent vessels with large numbers of EPC-derived cells colocalizing with the vascular wall of primarily large vessels (Fig. 1B and C). There was little evidence of these cells associating with the microvasculature.

In the adult mouse I/R model of acute vascular injury, the patterns of localization of transferred huEPCs are comparable with those observed in the OIR model, albeit primarily in the capillary network rather than the larger central arterioles and venules (Fig. 2). The retina of animals receiving diabetic huEPCs show little or no association of labeled donor cells (Fig. 2A); rather the labeled cells appear to aggregate on top of the existing vasculature. By contrast, labeled huCD34⁺ cells from healthy donors appear to be integrated into both small and moderately sized vessels (Fig. 2B), unlike in the OIR model, in which localization occurred almost exclusively in large vessels. Figure 2C shows an injured vessel, as evidenced by the lack of rhodamine-dextran perfusion, in which PKH67-labeled huCD34⁺ cells evidently are attached and apparently filling in the damaged region. Figure 2D provides a schematic representation of the image in 2C, wherein the nonperfused area is shown as a hatched line in order to better visualize the nature of the process.

There does not appear to be a qualitative difference in the degree to which healthy huCD34⁺ cells associate with injured vasculature despite the route of administration. Images shown are representative of results seen 48 h after administering the cells and do not differ substantially from what is observed 24 h after giving the donor cells (not shown).

The two chronic (diabetes) injury models show similar results to those described for the acute injury models (OIR and I/R). In STZ-induced diabetic mice, 48 h after administering cells either intravitreously or systemically (Fig. 3), as well as in spontaneously diabetic BBZ/Wor rats, 7 days after intravitreous delivery of huCD34⁺ cells (Fig. 4), cells of diabetic origin do not migrate to areas of injury nor do they appear to reendothelialize damaged vasculature. Instead, the diabetic huCD34⁺ cells appear to form clumps and sheets that are associated with, but are not a part of, the existing vasculature (Fig. 3B and D and Fig. 4B and D). By contrast, healthy huCD34⁺ cells both home to damaged areas and seem to integrate extensively into the vessels in both diabetes models (Figs. 3C and 4C).

Figure 5 demonstrates that the association of the healthy labeled huEPCs with damaged vessels is not perivascular but rather represents partial integration into the microvasculature. Figure 5A depicts a typical cross-section of a normal mouse retina, wherein mouse vascular endothelium can be localized by green immunofluorescence. The retinal cross section from a STZ-induced diabetic mouse that received unlabeled huCD34⁺ cells from a healthy donor in Fig. 5B shows colocalization of the huCD34⁺ cells (red immunofluorescence) with the endothelium of the superficial retinal vascular plexus.

Quantitative assessment of percent incorporation of huCD34⁺ cells into existing vascular area yielded the following results. In the I/R mouse model, normal huCD34⁺ cells comprised 31.04 ± 3.9% of the vascular area in injured eyes, whereas diabetic huCD34⁺ cells appeared in only 2.36 ± 0.21% of vascular area (P = 0.02). In the STZ-induced diabetic model, the percentages of vascular area that colocalized with donor huCD34⁺ cells were as follows: nondiabetic huCD34⁺ cells in nondiabetic animals 1.92 ± 1.37%, diabetic huCD34⁺ cells in nondiabetic animals 2.83 ± 1.58%, nondiabetic huCD34⁺ cells in diabetic animals 26.42 ± 2.30%, and diabetic huCD34⁺ cells in diabetic animals 6.11 ± 0.77%. The difference in percent colocalized vascular area between diabetic animals receiving nondiabetic huCD34⁺ cells versus diabetic huCD34⁺ cells was highly significant (P < 0.001).

The studies presented here convincingly demonstrate for the first time that circulating healthy huCD34⁺ EPC are capable of integrating into damaged vasculature. This can be seen in both the acute and chronic ocular vascular injury models used. The association of at least some of the exogenously administered cells is not perivascular but represents true integration into the vascular wall, supportive of reendothelialization of acellular vessels. The vessels into which these cells integrate are functional by virtue of their patency, as shown by their ability to be perfused with labeled dextran. The health of donor cells greatly affects the reparative ability of the CD34⁺ EPCs. Cells from diabetic donors are unable to integrate into damaged vasculature, regardless of the type of injury, whereas cells from healthy donors are able to reendothelialize injured vasculature in all models tested, including long-term, spontaneous, or chemically induced diabetes. We have confirmed repair and incorporation using a quantitative approach. In these models, huCD34⁺ cells home exclusively to areas of vascular injury, as has been reported extensively in other in vivo model systems (2,13,23–27). The current quantitation of retinal repair using the I/R model and healthy huCD34⁺ cells suggests that at least 31% of the retina was injured and repaired. In diabetic mice, 26% of the retina is reendothelialized with healthy huCD34⁺ cells. These percentages closely correlate with the extent of capillary damage reported in these models (17,18).

Defective CD34⁺ cell function has been associated with diabetes (13–15,24–31). Specifically, diabetes is associated with reduced mobilization of CD34⁺ cells from the bone marrow, reduced numbers of CD34⁺ cells in the circulation, reduced migration of these cells into areas of ischemia, reduced incorporation of these cells into capillaries, and reduced differentiation into endothelial cells (14,15,23,32,33). Blood glucose also inversely correlates with CD34⁺ cell counts, with normal blood glucose associated with higher numbers (14,15). However, diabetic CD34⁺ cell growth defects are not reversed by cultivation in normal glucose medium, suggesting that the impairment of CD34⁺ cells is not rapidly reversible by glucose correction alone (14).

There are numerous studies demonstrating that diabetic EPCs do not repair vascular injury using both type 1 and type 2 diabetic rodent models (2,15,27–29). These studies, however, have never examined the retinal vasculature. Healthy huCD34⁺ cells promoted revascularization of skin wounds (26) and accelerated blood flow restoration in type 1 diabetic mice (30). Reduced numbers of CD34⁺ cells with impaired chemotactic and proangiogenic activity exist in type 1 diabetic subjects and, when infused, result in reduced formation of collateral vessels (33). CD34⁺ cells of patients with type 2 diabetes show impaired adhesion to the basement membrane, decreased proliferation, and aberrant tubule formation (15). Murine Sca-1⁺ (equivalent to CD34⁺ in humans) hematopoietic stem cells dramatically improved vascularization of skin wounds in obese type 2 diabetic Lepr^db mice but not in congenic lean nondiabetic C57BL/6 mice (27). Moreover, when skin wounds of Lepr^db mice were treated with Lepr^db-derived Sca-1⁺ hematopoietic stem cell–enriched cells, wound vascularization was severely inhibited (27). Awad et al. (2) demonstrated that the obese type 2 diabetes syndrome induces intrinsic defects in CD34⁺ EPCs. The behavior of bone marrow cells in diabetic and nondiabetic environments may differ (24,27), and there may be negative synergism between the diabetic environment and diabetic bone marrow–derived cells.

In diabetes, vascular basement membranes (including those of the retinal capillaries) are heavily modified by advanced glycosylation end product–crosslink formation (34,35). Thus, modifications to the basement membrane, as occur in diabetes, might result in reduced EPC attachment and incorporation into sites of microvascular injury, diminishing vascular repair and facilitating the development and persistence of acellular capillaries (36). The reendothelialization of acellular capillaries involves cell attachment of CD34⁺ cells and similarly would be adversely affected by the diabetes basement membrane. However, our studies with healthy huCD34⁺ cells showing reendothelialization and vessel patency suggest that these healthy cells are able to attach to the diabetes basement membrane. This represents an area for additional study.

We previously have demonstrated that CD34⁺ cells isolated from diabetic individuals have defective migration in response to SDF-1 and VEGF (9). We subsequently have studied the migration of CD34⁺ cells isolated from patients with type 1 or type 2 diabetes in response to IGF-1 and IGF binding protein type 3 and found that the diabetic CD34⁺ cells have defective migration to all of these factors (A.A., M.B.G., unpublished observations). Our studies in >100 diabetic patients, regardless of diabetes type (type 1 or type 2), show a strong correlation between diabetes and CD34⁺ cell dysfunction (9).

These defects in CD34⁺ EPC function imparted by the diabetes milieu result in an apparent paradox: if CD34⁺ EPCs are unable to adequately repair damaged vasculature, why then can diabetes result in increased pathological retinal neovascularization? There are a number of possible explanations. First, accumulated damage by the diabetes milieu leads to basement membrane thickening, loss of pericytes and endothelial cells, and vascular thrombosis. These changes lead to ischemia and the release of proangiogenic factors, such as VEGF, fibroblast growth factor-2, and IGF-1, inducing the resident endothelium to proliferate, migrate, and form new vascular tubes. This process is known as compensatory neovascularization. This explanation does not preclude inadequate repair by dysfunctional diabetic CD34⁺ cells. On the contrary, assuming a maintenance/repair role for these cells might imply the increased compensatory activity by resident endothelium, in part, is attributed to lack of early retinal repair by dysfunctional CD34⁺ cells.

Second, the CD14⁺ monocyte, which shares a common ancestor with CD34⁺ EPCs and is activated by the diabetic state, may contribute to aberrant neovascularization. In support, CD14⁺ monocytes are progenitors for macrophages, dendritic cells, and fibrocytes and also can generate endothelial cells (37–45). CD14⁺ cells can participate in neovascularization in experimental models (42,43). CD14⁺ cells need priming to differentiate into endothelial cells and promote vascular growth (2). Diabetes is associated with inflammation, and these inflammatory mediators presumably could be activating CD14⁺ cells (2,46–49). In diabetes, CD14⁺ cells may replace dysfunctional CD34⁺ cells in vascular repair, but their response is aberrant and proangiogenic, thereby contributing to the observed preretinal neovascularization. However, this requires further investigation.

Pharmacological strategies to promote enhanced CD34⁺ cell function certainly are feasible and attainable in the near future. The well-known beneficial cardiovascular effects of statins, in part, may be attributable to their ability to mobilize and improve the function of endogenous EPCs (50). It therefore is conceivable that in the future pharmacological armamentarium of drug therapies for diabetic patients, drugs that improve EPC function will be included, thereby preserving vascular structure in the retina in diabetes.

This work was supported by the National Eye Institute (grants EY012601, EY007739 [to M.B.G.], and EY00300 [to T.S.K.]) and the Juvenile Diabetes Research Foundation Center for Gene Therapy for the Prevention of Diabetes and Its Complications at the University of Florida and the University of Miami.