Transient Inhibition of Transforming Growth Factor-β1 in Human Diabetic CD34⁺ Cells Enhances Vascular Reparative Functions

All animal studies were approved by the institutional animal care and use committee, and studies were conducted in accordance with The Guiding Principles in the Care and Use of Animals (National Institute of Health) as well as the Association of Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research.

HSCs were harvested from the bone marrow obtained from femurs and tibiae of C57BL/6-Tg (UBC-GFP) 30Scha/J mice homozygous for green fluorescent protein (GFP) antibodies. Fluorescently labeled c-kit (CD117) and Sca-1 (BD PharMingen, San Diego, CA) were used to enrich HSCs from mononuclear cell fractions using a BD cell sorter (FACS Calibur, BD Bioscience, San Jose, CA). We have previously shown that this technique yields a 95% pure hemangioblast cell/HSC population (5).

The detailed protocol for the isolation of long-term repopulating HSCs (LTR-HSCs) was previously described (22). Briefly, a dose response of LTR-HSCs (30–1,000 cells) pretreated for 1 h with anti-TGF-β1 neutralizing antibody (clone 1D11.16) or IgG1K isotype control antibody, both at 20 μg/ml (R & D Systems, Minneapolis, MN), were transplanted intravenously into lethally irradiated mice, and survival was examined over time.

Acute vascular injury was induced in mice using the ischemia reperfusion model, as detailed in an online appendix at http://diabetes.diabetesjournals.org/cgi/content/full/db10-0287/DC1. Briefly, the anterior chamber of the eye was then subjected to 2 h of increased hydrostatic pressure, resulting in retinal ischemia. This model results in the generation of acellular capillaries similar to what is observed in longstanding diabetic retinopathy but occurs after 7 days of injury (23). At this time, the animals received injections of isolated human CD34⁺ cells pretreated with either TGF-β1-PMO or control PMO into their vitreous.

TGF-β1-PMOs or control PMOs (AVI-Biopharma, Portland, OR) were synthesized, lyophilized, dissolved in sterile distilled water, and serially diluted. CD34⁺ cells were incubated for 16 h at 37°C in Iscove's modified Dulbecco's medium containing 10% FBS, 10% selected horse serum, and 40 μg/ml PMO. The detailed protocol for PMO treatment is described in an online appendix.

Lin⁻Sca-1⁺ cells (described above) were treated with anti-TGF-β1 antibodies or TGF-β1-PMO. Endogenous expression of TGF-β1 was then studied after fixation and permeabilization using phycoethrin-labeled TGF-β1 antibodies (IQ Products, Groningen, Holland) and analyzed by flow cytometry (FACS Vantage DiVa I flow cytometer, San Jose, CA). Anti-actin antibodies (BD PharMingen) served as control.

Peripheral blood samples (50 ml) were collected from type 2 diabetic patients and healthy age-matched volunteers. The study was approved by the University of Florida Institutional Review Board. All participants gave written informed consent to participate. Criteria for inclusion and exclusion of participants are provided in an online appendix, and detailed patient characteristic are described in supplementary Tables 1 and 2 (also available in an online appendix).

Mononuclear cells were isolated from peripheral blood using Lymphoprep and then subjected to magnetic selection of lin⁺ cells yielding lin⁻ cells (Stem Cell Technologies, Vancouver, BC, Canada), which were then incubated with anti-CD34-phycoethrin and anti-CD45-phycoethrin-Cy7 (BD PharMingen) and other select antibodies. Lin⁻ CD34⁺ CD45^mid cells were clearly resolved from lymphocytes and other myeloid cells using a FACS Vantage DiVa flow cytometer. CD34⁺ cells enriched by Miltenyi positive selection from human cord blood and bone marrow obtained from AllCells (Emeryville, CA) and were >95% pure as assessed by fluorescence-activated cell sorter (FACS) analysis. Viability was determined by 7AAD exclusion and was routinely >95% after washings (detailed protocol in an online appendix).

Lin⁻ CD34⁺CD45^mid cells from cord blood or peripheral blood were directly sorted into 96-well plates (10 ± 3 cells per well) containing plain medium, control PMO, or TGF-β1-PMO. Cell viability and proliferation were assessed daily using direct light microscopy (×200), both before growth factors and after their addition. Recombinant hematopoietic growth factor cocktail was added on day 5 to each well that contained human TPO (20 ng/ml), stem cell factor (50 ng/ml), IL-3 (50 ng/ml), and IL-6 (20 ng/ml) (PeproTech, NJ). Cells were scored that survived after the initial 5-day period in the absence of growth factors.

Cell migration after overnight inhibition with TGF-β1-PMOs or control PMOs was analyzed using a modified Boyden chamber assay as described previously (24).

At T = 0 and after treatment with TGF-β1-PMO for 16 h, healthy or diabetic CD34⁺ cells were incubated with CD34, CD45, and CXCR4 antibodies (BD Bisociences) to analyze CXCR4 expression using FACS (BD Bioscience).

NO production was quantified in peripheral blood CD34⁺ cells using NO-sensitive 4-amino-5-methylamino-2-difluorofluorescein (DAF-FM) fluorescence, as described previously (25).

Data are presented as mean values ± SEM. Statistical analyses were performed using GraphPad InStat 3.0 (GraphPad Software, San Diego, CA). Statistical differences in the mean were assessed using one-way ANOVA followed by Tukey Carmer post hoc test or unpaired t test.

To study the effect of TGF-β1 inhibition on survival and engraftment of LTR-HSCs, mice were lethally irradiated with 950 rads and anti-TGF-β1 antibody–treated LTR-HSCs were transplanted. As shown in Fig. 1 A, isotype antibody-treated LTR-HSC did not mediate survival of lethally irradiated mice over a range of 30–500 cells per mouse, and even at 1,000 LTR-HSC per mouse, the six-month survival was only 10%. In striking contrast, anti-TGF-β1 treatment for 1 h before transplant resulted in survival of ∼70% of mice at 60 cells per mouse. However, if 250 LTR-HSC were treated with anti-TGF-β1 antibody before transplant, essentially 100% of mice survived for >6 months post transplant.

Neutralizing antibody targets only cell surface TGF-β1. We also wanted to evaluate the pharmacological effect of both neutralization of cell surface TGF-β1 and intracellular TGF-β1 and used PMOs. FACS analysis on murine lin⁻ Sca1⁺ c-kit⁺ cells treated with either TGF-β1-PMO or TGF-β1 neutralizing antibody resulted in specific inhibition of endogenous TGF-β1 (Fig. 1 B). A functional assay that we used is survival of enriched progenitor cells in the absence of growth factors. Both murine LTR-HSC and human CD34⁺ cells exhibit a rapid death curve in the absence of stem cell growth factors. We observed that inhibition of endogenous TGF-β1 in LTR-HSC resulted in prolonged survival of the cells even in single cell per well cultures (data not shown). Human CD34⁺ cells also survive in the absence of growth factors if TGF-β1-PMO was present (supplementary Fig. 1, available in an online appendix). Unexpectedly, human CD34⁺ cells did not survive in the absence of growth factors when TGF-β1 neutralizing antibody was present (in contrast to murine LTR-HSC) (supplementary Fig. 1) but did so when cultured with TGF-β1-PMO. Human CD34⁺ cells not only proliferated in response to the added growth factors after 5 days but generated large numbers of progeny during the subsequent 7-day growth period with growth factors present. In contrast, control antibody, control PMO, and TGF-β1 neutralizing antibody did not promote survival of CD34⁺ cells in the absence of growth factors, nor did they respond to added growth factors at day 5. At this point, we chose to exclusively use TGF-β1-PMO and not TGF-β1 neutralizing antibody for subsequent studies with human cells.

The PMOs used in these studies were 18–21 oligonucleotides in length, which showed excellent cellular uptake without the aid of any transfection reagent or peptide conjugation. To further establish kinetics PMO uptake by CD34⁺ cells, fluorescein isothiocyanate (FITC)-conjugated PMO were developed (FITC-PMO kindly provided by Pat Iversen, AVI-Biopharma). FITC-PMOs were rapidly taken up by peripheral blood CD34⁺ cells in an exponential manner; at 40 μg/ml FITC-PMO, a plateau was observed after overnight ex vivo incubation, while at a concentration of 160 μg/ml, saturation occurred within 1–3 h (supplementary Fig. 2, available in an online appendix).

We pretreated murine HSCs (lin⁻c-kit⁺Sca1⁺) with TGF-β1 PMO or control PMO and tested cell homing to areas of laser injury. As shown in supplementary Fig. 3 (available in an online appendix), pretreatment with TGF-β1-PMO substantially increased the number of GFP⁺ HSCs that were recruited to the sites of laser rupture of Bruch's membrane compared with control PMO–treated cells (P < 0.05).

We next examined the reparative function of diabetic CD34⁺ cells after pretreatment with TGF-β1-PMO or control PMO in the ischemia reperfusion model, which recapitulates key aspects of the vasodegenerate phase (acellular capillaries) of diabetic retinopathy. Analysis of retinal flat mounts showed that pretreatment of both diabetic and healthy CD34⁺ cells with TGF-β1-PMO increased homing to areas of retinal injury to a greater degree than control PMO (Fig. 2,A). Diabetic CD34⁺ cells showed a 2.2-fold lower colocalization to injured retinal areas than healthy CD34⁺ cells (Fig. 2,B). Pretreatment of healthy CD34⁺ cells with TGF-β1-PMO resulted in a 2.51-fold (P < 0.01) increase in the numbers of CD34⁺ cells homing to injured vessels. On the other hand, treatment of diabetic CD34⁺ cells with TGF-β1-PMO produced a robust reparative response (P < 0.01), as seen by a 5.4-fold (P < 0.01) increase in numbers of CD34⁺ cells colocalized with vessels supporting the restored ability of these cells to home to acellular capillaries (Fig. 2 B).

Our animal studies showed that inhibition of TGF-β1 improved homing to areas of injury, the critical first step in repair. To evaluate the mechanism responsible for the improvement in CD34⁺ cell homing after transient inhibition of endogenous TGF-β1, we first tested the impact of PMO treatment on survival of human CD34⁺ cells. For these studies, we used CD34⁺ cells from sources that would be typically used for human revascularization trials, such as cord blood, bone marrow, and peripheral blood. Cord blood and bone marrow CD34⁺ cells were cultured in medium with or without PMOs for 5 days. In the absence of growth factors, 95% of human lin⁻ CD34⁺ cells died after 2–3 days, even though they were cultured in a rich medium (Iscove's modified Dulbecco's medium with 10% FBS and 10% horse serum). In contrast, TGF-β1-PMO–treated normal cord blood and bone marrow cells survived (P < 0.05) in the absence of growth factors and continued their growth over the course of an additional 7 days (Fig. 3,A). Next, peripheral blood CD34⁺ cells from diabetic subjects and age- and sex-matched control subjects were tested. As observed for cord blood and bone marrow CD34⁺ cells, the addition of growth factors elicited greater proliferation in TGF-β1-PMO–treated peripheral blood cells than in control PMO-treated cells (P < 0.05) (Fig. 3 B).

Inhibition of endogenous TGF-β1 expression enhanced migration of both diabetic (P < 0.05 vs. control PMO) and healthy cells (P < 0.01 vs. control PMO) (Fig. 4). To understand the mechanisms involved in this enhanced migratory function of CD34⁺ cells, we analyzed expression of the SDF-1 receptor, CXCR4. Enriched CD34⁺ cells were gated for the CD34⁺ CD45^mid population (Fig. 5,A), and CXCR4 expression was analyzed (Fig. 5 B). We found that CXCR4 surface expression was greater in TGF-β1-PMO–treated diabetic cells than in control PMO-treated cells (P = 0.004). A similar trend was observed in healthy cells; however, it did not achieve statistical significance.

Bioavailable NO is reduced in diabetic progenitors, and this contributes to the migratory dysfunction observed in these cells (13,26). Because inhibition of TGF-β1 in healthy and diabetic CD34⁺ cells increased migration and homing in both animal models, we asked whether this effect was coupled to enhanced NO generation. In healthy CD34⁺ cells, SDF-1 (100 nmol/l) increased NO generation by 45 ± 4% (Fig. 6). The signal transduction pathway involved in SDF-1–induced NO release in CD34⁺ cells has not been elucidated. To delineate this pathway, we tested different pharmacological inhibitors. SDF-1–induced NO release was significantly decreased when cells were pretreated with one of the following: 10 μmol/l AMD3100, a specific nonpeptide CXCR4 antagonist (3 ± 2%, P < 0.001); 100 ng/ml pertussis toxin, a G_i protein inhibitor (20 ± 9%, P < 0.01); 20 μmol/l LY294002, a phosphoinositide 3-kinase (PI3K) inhibitor (12 ± 3%, P < 0.001); or 30 μmol/l triciribin, an Akt inhibitor (12 ± 3%, P < 0.01) (Fig. 6). These results suggest that, in CD34⁺ cells, SDF-1–induced NO release involves the CXCR4/G_i-protein/PI3K/Akt pathway.

Next, we determined whether pretreatment with control or TGF-β1-PMO influenced NO release in these cells. We found that SDF-1 elicited a comparable degree of NO release in healthy, peripheral blood CD34⁺ cells that had been pretreated with either control PMO (60 ± 5%) or TGF-β1-PMO (54 ± 4%) (Fig. 7,A, B). In contrast, SDF-1–induced marked NO generation in diabetic CD34⁺ cells treated with TGF-β1-PMO (69 ± 7%, P < 0.0001) but not in cells treated with control PMO (9.2%) (Fig. 7 A, B). This ability of TGF-β1-PMO to restore NO release in diabetic CD34⁺ cells was diminished by the selective CXCR4 inhibitor, AMD3100 (28 ± 2%, P < 0.01), supporting that this effect was mediated by CXCR4 activation.

Therapeutic revascularization with endothelial progenitor cells holds promise as a treatment modality to prevent tissue damage and restore blood flow in individuals such as diabetic subjects who are not ideal candidates for standard revascularization procedures because of their small vessel disease. However, while cell therapy is needed in diabetic patients, this approach has limited utility because of endothelial progenitor cell dysfunction (27,28). Specifically, endothelial progenitors isolated from diabetic individuals demonstrate reduced proliferation, migration, and differentiation into endothelial cells. Exposure to high concentrations of glucose reduces endothelial nitric oxide synthase (eNOS) expression in these cells. HIF-1α modification by methylglyoxal, a consequence of increased reactive oxygen species formed during high glucose exposure, is likely responsible (27). BMPCs from db/db mice show reduced expression of eNOS and phospho-eNOS (29). Consistent with this finding, we have shown that human diabetic CD34⁺ cells have reduced NO bioavailability, which is associated with decreased migration that can be restored through exposure to NO donors (13).

In this study, we tested whether the reparative function of diabetic progenitors could be enhanced through inhibition of TGF-β1, a novel approach that we hypothesized could correct multiple functional defects in these cells. We found that pretreating CD34⁺ progenitors with TGF-β1-PMO enhanced proliferation and migration in vitro as well as homing in two preclinical in vivo models of injury. Endogenous TGF-β1 has been shown to be a primary factor maintaining HSCs in G0/G1. However, exogenous TGF-β1 (in plasma and tissue), SDF-1/CXCR4 signaling, and circulating growth factors could also regulate HSC cell cycle status. Although the majority of TGF-β1 is expressed as a latent, inactive form, HSCs and BMPCs express transglutaminase and other proteolytic enzymes that can activate endogenous TGF-β1 (S.H.B., unpublished data). Accordingly, active intracellular TGF-β1 has been identified in HSCs and endothelial progenitors (30). HSC quiescence, division, and daughter cell fate decisions have been shown to be mediated by the TGF-β1/hematopoietic growth factor axis (22,31).

Previously, it has been reported that inhibition of TGF-β1 induces G0-to-G1 transition (32). Here we find that it promotes cell migration and proliferation and improves homing of CD34⁺ cells to the vasculature. If intracellular or extracellular TGF-β1 is withdrawn from CD34⁺ cells in the presence of optimal hematopoietic growth factors, then G0 cells rapidly shift to G1 and the G2/M/S interface. TGF-β1 inhibition drives CD34⁺cells into G1 (33) and has been shown to increase the engraftment potential of HSCs (34). In contrast to TGF-β1 inhibition, withdrawal of hematopoietic growth factors results in CD34⁺ cell death, some of which is mediated by TGF-β1 (35).

Here we show that if TGF-β1 is inhibited intracellularly in the absence of hematopoietic growth factors, human CD34⁺ cells can survive for >1 week ex vivo. In vivo we observe an enhanced reparative response that is likely due not only to increased migration of the TGF-β1-PMO–treated human CD34⁺ cells but also to increased proliferation at the site of vascular injury. Similar to engraftment into the bone marrow of LTR-HSC after TGF-β1-PMO treatment, we observed correction of defective homing in the diabetic CD34⁺ cells treated with TGF-β1-PMO. To better understand the mechanisms of this beneficial effect, we first focused on the SDF-1/CXCR4 pathway that has been show to be central to mediating tissue repair after injury. SDF-1 is highly expressed by ischemic tissue and can also increase progenitor cell homing back to the bone marrow (36,37). SDF-1 plays an important role in vascular development (38), and ablation of the SDF-1/CXCR4 gene in mice produces defects in blood vessel formation. As shown in Fig. 1,A, anti-TGF-β–treated LTR-HSC demonstrates increased homing to the bone marrow, and as shown in Fig. 2 and supplementary Fig. 3, anti-TGF-β–treated cells migrate better to injured tissue. This can be due to increased CXCR-4 expression or increased proliferative and migratory potential. Our results revealed that the migration of human CD34⁺ cells to SDF-1 was reduced in diabetic cells and that this migratory defect was corrected by endogenous TGF-β1 inhibition, suggesting that efficient signal transduction of the SDF-1/CXCR4 pathway was restored.

To further understand the influence of TGF-β1 inhibition on this pathway, we evaluated NO release in response to SDF-1 and used different pharmacological inhibitors to delineate the involvement of CXCR-4 receptor signaling pathway in SDF-1–induced NO release. Importantly, our studies also indicate that SDF-1–induced NO release occurred via activation of CXCR-4, and this involved G_i protein, PI3K and subsequent Akt activation. Akt activates eNOS by phosphorylation at Ser¹¹⁷⁷, resulting in NO production (39,40) Our findings are consistent with earlier studies showing that SDF-1–induced chemotaxis of T lymphocytes involves G_i protein-coupled PI3K activation and Akt phosphorylation (41,42).

We found that healthy CD34⁺ cells demonstrated robust NO release in response to SDF-1 and TGF-β1 inhibition did not increase NO release further. This suggests that CXCR-4 activation required for NO release in response to SDF-1 was already maximal in the nondiabetic cells, before TGF-β1 inhibition using PMOs. In contrast, TGF-β1 inhibition in diabetic CD34⁺ cells corrected defective NO release in response to SDF-1, in part by significantly increasing the expression of CXCR-4 on the diabetic cells, which was not the case for the nondiabetic cells (Fig. 5,B). This implies that, in healthy nondiabetic CD34⁺ cells, activation of CXCR-4 triggers NO-independent signaling pathways, such as tyrosine phosphorylation of Wiskott-Aldrich syndrome protein (WASp). WASp is a critical regulator of actin cytoskeleton remodeling after CXCR-4 activation. SDF-1 exposure results in Cdc42 activation and WASp tyrosine phosphorylation as well as in WASp association with Fyn and Pyk-2 tyrosine kinases (43). Here, we observed that cells were recruited to areas of vascular injury in two distinct but complementary injury models. Systemic administration of TGF-β1-PMO–treated murine LTR-HSC's increased the homing of these cells to sites of neovascularization (supplementary Fig. 3) and intravitreal administration of TGF-β1-PMO–treated CD34⁺ cells enhanced homing to injured retinal vessels (Fig. 2). Both the laser injury model and the ischemia reperfusion model result in release of angiogenic growth factors (44). We have previously shown that laser injury alone is sufficient to induce stem cell recruitment as early as 1 week after injury (45). In this model, SDF-1 antibodies reduce the recruitment of HSCs to the choroidal neovascularization lesion, suggesting that SDF-1/CXCR-4 participates in the recruitment of cells to sites of laser injury (46). Our current study suggests that treatment of the murine HSCs and human CD34⁺cells with TGF-β1-PMO increases their homing to areas of injury by enhancing their migratory potential. This is likely due to increased surface expression of CXCR-4 mediated by TGF-β1-PMO exposure. Activation of CXCR-4 generates NO, which is critical to CD34⁺ cell migration (13,47).

Retinal and subretinal ischemia contributes to visual impairment and blindness in diseases as diverse as retinopathy of prematurity, diabetic retinopathy, and age-related macular degeneration. The ischemia reperfusion model mimics many aspects of the pathophysiology of retinal ischemia and leads to the development of acellular capillaries. We have previously shown that, in this model, healthy CD34⁺/endothelial precursors reendothelialize ischemic capillaries; however, diabetic CD34⁺/endothelial precursor cells do not (11). In the present study, inhibition of TGF-β1 enhanced the recruitment of diabetic as well as healthy CD34⁺ cells to sites of retinal injury and corrected this defect in diabetic CD34⁺ cells.

In conclusion, our studies show that transient inhibition of TGF-β1 in CD34⁺ cells ex vivo enhances repair after vascular damage. This finding may have a profound impact on disease states associated with vascular dysfunction such as ischemic heart disease and diabetic vascular complications. While an attempt is being made to replace traditional approaches for alleviating tissue ischemia (e.g., stents, angioplasty, or vessel grafts) with cell therapy, autologous cellular therapy has not been feasible in diabetic patients because of dysfunctional cells. Transient inhibition of TGF-β1 may represent a promising therapeutic strategy for restoring vascular reparative function in diabetic CD34⁺ cells and may increase the likelihood of successful cellular therapy in diabetic individuals.

This work was supported by grants to M.B.G. from the National Institutes of Health (EY-007739 and EY-012601), U01-HL-087366, the Loftus family, and the Juvenile Diabetes Research Foundation and to A.D.B. from the American Heart Association. S.H.B. received a grant from BetaStem Therapeutics Inc. No other potential conflicts of interest relevant to this article were reported.

A.D.B., E.P.G., Y.P.R.J., S.H.B., and S.C. generated and analyzed data. C.S., D.K., F.W.R., C.J.P., and P.J.H. contributed to discussion. L.K. and M.C.L. provided access to the diabetic clinic at University of Florida. S.H.B. and M.B.G. analyzed data and wrote the manuscript.

The authors thank Dr. Martin Friedlander (The Scripps Research Institute, LA Jolla, CA) and Dr. George King (Joslin Diabetes Center, Boston, MA) for their helpful discussions during the completion of these studies and the preparation of this manuscript. We also thank Dr. Patrick L. Iversen (AVI Biopharma Inc, Portland, OR) for many helpful discussions concerning the approaches used in these studies and for his ability to design PMOs that are effective.