Diabetic retinopathy is associated with bone marrow neuropathy and a depressed peripheral clock

BBZDR/Wor rats with 4 mo of diabetes were used and compared with their age-matched nondiabetic BBDR littermates. The BBZDR/Wor rat was specifically developed as a model of type 2 diabetes that develops diabetic complications. Male obese BBZDR/Wor rats spontaneously develop diabetes that mimics human type 2 diabetes, and this typically occurs at 74 d of age. They demonstrate hyperglycemia and dyslipidemia, which are typical of type 2 diabetes, with increased levels of cholesterol, triglycerides, and nonesterified fatty acids (Tirabassi et al., 2004). The BBZDR/Wor rat develops retinopathy, neuropathy, nephropathy, and macrovascular complications typical of human type 2 diabetes and is emerging as the most applicable model of type 2 diabetic complications (Tirabassi et al., 2004). BBZDR/Wor animals demonstrate increased vascular endothelial growth factor (VEGF) production and increased VEGF receptor expression (Ellis et al., 2000), as well as increased NADPH oxidase activity (Ellis et al., 1998) and hydrogen peroxide production (Ellis et al., 2000) in the retinal vasculature as early as 2 wk after diabetes onset. BBZDR/Wor rats also progress to early preproliferative diabetic retinopathy with the development of acellular capillaries, pericyte ghosts, and capillary basement membrane thickening after 4 mo of diabetes (Tirabassi et al., 2004).

To explore the role of the sympathetic nervous system in the release of EPCs in diabetes, we first undertook an analysis of bone marrow histology and determined 24-h circulating NE levels. Histological examination of the bone marrow in rats with 4 mo of diabetes revealed a marked reduction in the number of nerve endings as determined by neurofilament 200 (NF200) staining and increased bone marrow lipid deposition compared with age-matched controls (Fig. 1, A–F). Tyrosine hydroxylase (TH) staining showed positive nerve processes running along blood vessels in the bone marrow of controls (Fig. 1, C and C′) but was very rarely found in diabetic animals (Fig. 1, D and D′), demonstrating a defect in sympathetic innervation. As shown in Fig. 1, 4 mo of diabetes resulted in the appearance of increased numbers of acellular capillaries within the retina (Fig. 1, J and K) compared with age-matched controls (Fig. 1, I and K). We observed that denervation of the bone marrow preceded the development of this retinal pathology because at 2 mo of diabetes, the rats did not have any retinal histopathological changes (Fig. 2 C) but already showed early signs of denervation (Fig. 2, A and B).

In these studies, we characterized the progenitor population based on surface expression of thy-1 and CD133 (Peichev et al., 2000), endothelial nitric oxide synthase expression (Loomans et al., 2006), and the ability of these cells to incorporate into capillary tubes in vitro (Schatteman et al., 2007; Fig. 3). To determine whether denervation of bone marrow affected circadian EPC release in diabetes, peripheral blood EPCs were enumerated by flow cytometry analysis every 2 h in diabetic rats and compared with age-matched controls. Control rats showed a dramatic peak of circulating EPCs between ZT (Zeitgeber time) 1 and ZT-5 (Fig. 4 A). In contrast, diabetic animals exhibited a marked reduction in the amplitude and a broader peak in the number of circulating EPCs between ZT-21 and ZT-7. Moreover, the controls demonstrated a dramatic drop to baseline EPC numbers immediately after the ZT-3 peak. In contrast, diabetic rats lost this abrupt drop. This finding was confirmed in the type 1 diabetic animal model (Fig. S1 A).

We interpreted the persistent elevation of EPCs until ZT-7 in diabetics as reduced release and also diminished clearance of these cells from the blood. This reduced clearance was likely the result of a diminished ability of EPCs to migrate into tissues from the circulation and is in keeping with the profound migratory defect of these cells observed in diabetes (Segal et al., 2006; Fig. S3).

To further explore this possibility, we generated chimeric mice that underwent transplantation with bone marrow HSCs from gfp homozygous transgenic mice and then were stably engrafted for 4 mo. Retinal flat mounts were prepared from animals euthanized at either ZT-5 (peak of rest phase in mice) or ZT-13 (trough of EPC release in circulation). In agreement with thy-1 enumeration data in healthy rats (Fig. 4 A), many more gfp⁺ cells were present in the retinal circulation at the peak release for mice, ZT-5, as compared with ZT-13 (Fig. 4, B–D).

We thus postulated that in diabetes, the EPCs were trapped within the diabetic bone marrow as a result of a lack of sympathetic innervation and reduced NE release. To test this, we examined the bone marrow in the same rats that underwent 24-h blood sampling for EPC enumeration. These diabetic rats with decreased peripheral blood EPCs showed a marked elevation in the number of bone marrow EPCs compared with controls (Fig. 5 A). In addition, diabetic bone marrow EPCs showed a reduction in colony formation ability (Fig. 5, B–D) and proliferation (Fig. 5 F). Both blood and bone marrow EPCs had a reduced migration to VEGF (Fig. 5 E) compared with age-matched controls.

We reasoned that circulating NE levels in these diabetic animals with peripheral neuropathy and retinopathy may be altered and then measured NE levels in the blood of these rats. Control rats had two distinct peaks of plasma NE levels, one at ZT-1 and one at ZT-13 (Fig. S1 B). Interestingly, in diabetic animals, the first NE peak at ZT-1 was considerably higher and broader than in controls, showing a compensatory adrenal response, whereas the second peak at ZT-13 was absent, supporting the concept of sympathetic dysfunction in these rats.

These experiments demonstrate that type 2 diabetes is associated with a reduction in circadian release of EPCs from the bone marrow. Because circadian rhythm is under tight control of not only the central clock, the SCN, but also peripheral clocks, we next examined whether clock gene expression in the SCN, retina, bone marrow, and peripheral blood EPCs was altered in diabetes. At peak release of EPCs, animals were sacrificed and their SCN brain regions, retinas, bone marrow, and peripheral blood harvested for messenger RNA (mRNA) extraction. Clock genes (Clock, Bmal1, Per-1, Per-2, CRY-1, CRY-2, ERB, and RORA) were quantified by real-time PCR in these tissues and cells. Rats with diabetes demonstrated a significant decrease (P < 0.05) in expression of these clock genes in the retina and in thy-1⁺ cells isolated from both blood and bone marrow (Fig. 6, A–C). Protein studies revealed reduction in Bmal expression in diabetic rats as compared with controls (Fig. 6 E). This dramatic reduction in both the positive and negative arms of circadian gene regulators would lead to dampened oscillations and lower and broader peaks, which is the pattern we observed in the circulating EPCs in diabetic rats compared with age-matched controls. Healthy rats, in contrast, demonstrated a higher narrower peak than the diabetic rats. Although expression of clock genes was reduced in the SCN regions of diabetic rats, the difference from controls was not significant (Fig. 6 D). This lack of change suggests that diabetes is associated with dysfunction of the peripheral rather than the central clock. The mRNA and protein expression of clock genes in control and diabetic human retina (Fig. S2) is in agreement with the alterations of clock gene expression found in the diabetic rat model.

The potential to form new blood vessels and repair damaged vessels depends on the EPCs' ability to both leave the bone marrow and to migrate toward the site of angiogenesis. In this study, we observed that type 2 diabetic rats demonstrate the hallmark feature of diabetic retinopathy, acellular capillaries within the retina, at the precise time when there is denervation of the bone marrow and a reduction in peripheral clock gene expression. The resultant acellular capillaries seen at 4 mo of diabetes appear to be caused, in part, by the loss of proper EPC reparative function and by failure of circadian EPC release secondary to diabetes-associated denervation of the bone marrow. In the presence of innervation, as observed in control rats, there was normal retinal histology and a robust circadian release of EPC from the bone marrow. We report that by 4 mo of diabetes in our model of type 2 diabetes, there is inherent sympathetic denervation that can alter circadian release of EPC from the bone marrow. Our data supports that the loss of circadian EPC release results in inadequate levels of cells in the circulation at the optimal time of repair (rest phase) and that this defect, in addition to defects of EPC function (reduced proliferation and migration), contributes to the development of diabetic retinopathy.

Circadian dysfunction itself can have profound effects on the diabetic state. Prior work has shown that mice with a specific clock gene mutation demonstrate impaired endothelial function, including decreased tube formation and faster senescence and impaired EPC-mediated repair (Wang et al., 2008). Other clock genes have been shown to directly regulate glucose homeostasis, and their inactivation suppresses diurnal variation in glucose and triglycerides along with reduced gluconeogenesis (Kohsaka et al., 2007). Our novel findings would indicate that restoration of sympathetic innervation to the diabetic bone marrow should result in physiological EPC egress from the bone marrow and improved retinal vascular repair (Thomas et al., 2008). Moreover, strategies to normalize peripheral clock gene expression may be a novel way to both treat and prevent the development of diabetic macro- and microvascular complications by circumventing sympathetic nervous system dysfunction.

The protocols for the rat studies were approved by the Institutional Animal Care and Use Committee at Michigan State University, The University of Florida, and Biomedical Research Models, Inc. Multiple blood draws were performed by direct catheter placement into the femoral artery and vein. Animals were anesthetized with isoflurane (2%). The surgical approach for the femoral arterial catheters was through the ventral aspect of the leg. The catheters were tunneled subcutaneously to the nape of the neck and secured with sutures for later access while the animals were conscious. Blood draws were initiated after complete recovery of the animal, as determined by healing of the incision sites, usually 3–4 d after surgery. Blood draws (100 µl each) were performed every 2 h for 24 h using an Automatic Blood Sampler (DiLab).

Rats were euthanized by exsanguination after anesthesia. Rat hind limb bones were dislocated and the skin removed. Bones were flushed with ice cold PBS and the cells pelleted. The cells were then treated with ammonium chloride (STEMCELL Technologies Inc.) to remove any contaminating red blood cells and resuspended in PBS (Mediatech, Inc.) containing 2% FBS and 1 mM EDTA. A custom rat negative selection kit (STEMCELL Technologies Inc.) was used to deplete CD4, CD5, CD8a, and OX-43 cells of the rat. The cells were then further enriched for CD90 (thy-1)-positive cells using a positive selection kit (STEMCELL Technologies Inc.) tagged with a CD90/thy-1 antibody (Abcam).

A flow cytometer (LSR II; BD) equipped with FACSDiva software (BD) was used for thy-1⁺CD3⁻CD4⁻CD8⁻ cell enumeration. PerCP-Cy5.5– (eBioscience), FITC-, RPE-, and Alexa Fluor 647–conjugated mouse anti–rat antibodies (AbD Serotec) were used to stain thy-1, CD3, CD4, and CD8 surface markers, respectively, in concentrations according to the manufacturer's instructions. PerCP-Cy5.5, FITC, and RPE were excited at 488 nm and their respective fluorescent emission was detected at 695, 530, and 576 nm. Alexa Fluor 647 was excited at 633 nm and its emission detected at 660 nm. Isotype-negative samples for all antibodies were used to determine the background fluorescence for each fluorochrome. Single-stained samples were used to set color compensation between fluorochromes. A gate was drawn around the white blood cell population in a forward versus side scatter histogram to exclude red blood cells debris.

A competitive ELISA kit (ALPCO Diagnostics) was used to determine NE levels in 10 µl of rat plasma according to the manufacturer's instructions. In summary, NE was first extracted with a cis-diol–specific affinity gel and then acylated and converted enzymatically. After transfer on antiserum-covered microtiter plates, derivatized samples competed with solid phase–bound analytes for a fixed number of binding sites. After free analytes were removed by washing, the antibody bound to the solid phase was detected by peroxidase conjugation and read at 450 nm. Absorbance reading of the standards (linear) was plotted against known standard concentrations (logarithmic) and a standard curve was drawn to quantify unknown samples.

Decalcified rat humerus bones were embedded in paraffin, sectioned on a rotary microtome at 4 µm, and then dried and deparaffinized. Enzymatic epitope retrieval was performed. After pretreatments, standard Avidin–Biotin complex staining steps were performed at room temperature on an Autostainer (Dako). After blocking nonspecific protein with nonimmune goat serum (Vector Laboratories), sections were incubated with an Avidin (Vector Laboratories)/Biotin (Sigma-Aldrich) blocking system. Polyclonal rabbit anti-NF200 (Sigma-Aldrich) at 1:100 or polyclonal rabbit anti-TH (Millipore) at 1:100 in normal antibody diluent (ScyTek Laboratories) was used for staining. Biotinylated secondary goat anti–rabbit IgG (Vector Laboratories) in normal antibody diluent diluted 1:400, followed by R.T.U. Elite Peroxidase reagent (Vector Laboratories), was applied for visualization. The slides were developed using Nova red (Vector Laboratories), counterstained using Gill (Lerner) 2 Hematoxylin (Thermo Fisher Scientific), dehydrated, cleared, and mounted using Permount (Thermo Fisher Scientific).

The sensory retina was isolated by gently separating it from pigment epithelium and choroid using a #00 paint brush under a dissecting microscope, prepared using trypsin digestion, and stained with hematoxylin and periodic acid–Schiff (Mizutani et al., 1996). Acellular capillaries were quantified in a masked manner by three independent examiners.

10–12-wk-old female C57BL6/J mice were lethally irradiated and then their bone marrow was reconstituted with lineage-negative Sca1⁺ c-kit⁺ HSCs from donor bone marrow of transgenic mice homozygous for gfp, according to methods published previously (Grant et al., 2002). At selected times (ZT-5 And ZT-13) mice were euthanized and blood was collected for enumeration of thy-1⁺gfp⁺ circulating cells. The eyes were enucleated and preserved, retinas dissected, and mounted flat and image captures made using a fluorescent microscope (Eclipse TE200; Nikon) equipped with a digital camera (SPOT0.60XHRD06-NIK; Diagnostic Instruments, Inc.).

Rat retinas and bone marrow or blood EPCs were homogenized in Trizol reagent (Invitrogen), and RNA was isolated according to the manufacturer's instructions. First-strand complementary DNA was synthesized from isolated RNA using SuperScript II reverse transcription (Invitrogen). Prepared complementary DNA was mixed with 2× SYBR Green PCR Master Mix (Applied Biosystems) and various sets of gene-specific (Table S1) forward and reverse primers and subjected to real-time PCR quantification using the ABI PRISM 7700 Sequence Detection System (Applied Biosystems). All reactions were performed in triplicate. The relative amounts of mRNAs were calculated using the comparative threshold cycle method. Cyclophilin was used as a control, and all results were normalized to the abundance of cyclophilin.

30,000 EPCs were plated per well using a 24-well plate and then treated with cytokine mix (Stem Cell Technologies) for 12 h. Thymidine was diluted 1:10 in PBS and added to the wells directly. The cells were then placed in the CO₂ incubator for 2 h. Then cells were removed, spun at 5,000 rpm for 5 min, and washed with ice cold 10% TCA (×2). 0.2 M NaOH was added to the cells and the plates were shaken gently for 10 min at room temperature. 100 µl of cell homogenate was then added to 4 ml Cytoscint (Thermo Fisher Scientific). CPM was counted using thymidine settings.

Fig. S1 depicts decreased number of circulating EPCs in a type 1 diabetes model, increased plasma levels of NE in a type 2 diabetes model, and decreased number of circulating CD3-positive cells in a type 2 model as compared with controls. In Fig. S2, we show that clock genes demonstrate altered mRNA and protein levels in human diabetic retina. Fig. S3 illustrates that normal, but not diabetic, human EPC may participate in ocular vascular reendothelialization in diabetic mice. Table S1 shows real-time RT-PCR primers.

The authors thank Carrie Northcott, J.R. Haywood, Hannah Garver, Gregory Fink, and Sachin Kandlikar for training and assistance with femoral artery catheterization surgery and Tara Galvin and Jared Rice for providing the DiLab automated blood sampler and helping with the initial setup and data collection. Technical support was provided by Amy S. Porter, Kathleen A. Joseph, Ricky A. Rosebury, Svetlana Bozack, and Lily Yan from the Michigan State University and Sugata Hazra from the University of Florida.

This work was supported by grants from the National Institutes of Health (EY007739 and EY012601 to M.B. Grant and EY016077 to J.V. Busik), Juvenile Diabetes Research Foundation (4-2000-847 to M.B. Grant and 2-20005-97 to J.V. Busik), American Heart Association (M.B. Grant), and Michigan Agricultural Experimental Station (MICL02163 to J.V. Busik).

The authors have no conflicting financial interests.