line with the first descriptions by Tidhar et al. [58], detachment of pericytes from capillaries is associated with transgene downregulation. Sometimes, pericytes change their expression profile during adulthood. For example, RGS-5 represents a marker that is specific for ‘activated’ pericytes [64], and therefore vascular maturation results in the loss of RGS5 [65]. Another marker, endosialin is expressed by pericytes during periods of embryonic angiogenesis and in tumor vessels. The analogy in the diabetic retina awaits further investigation.

Retinal Pericyte Function

Contractility and Regulation of Flow

Pericytes are the capillary counterparts to SMC on arterioles and arteries [57, 66]. One remarkable feature that pericytes have in common with SMC is their contractile phenotype. Pericytes contain both smooth muscle and nonsmooth muscle isoforms of actin and myosin; however, with an uneven distribution within the pericyte population [53]. The differential expression of smooth muscle actin in pericytes may reflect the continuum from SMC of arteries and arterioles to pericytes of true capillaries, and may correlate with the physical forces that pericytes are exposed to. These pericytes are immunolabeled with smooth muscle tropomyosin and cGK, suggestive of a contractile function [51, 67, 68]. Several paracrine factors were identified that regulate pericyte contractility in vitro. While α2-adrenergic agonists, cholinergic agonists, histamine, serotonin, angiotensin II and ET-1 lead to vasoconstriction, β2-adrenergic agonists, NO and atrial natriuretic peptide lead to a dilatation of the pericyte-covered capillaries [37]. As mentioned above, ET-1 binds to pericytes for vessel contraction, whereas NO promotes vessel relaxation by a cGTP-dependent mechanism. As demonstrated in in vivo studies, pericytes in brain and retinal capillaries constrict in response to increase in the extracellular Ca2+

concentration through electrical stimulation, superperfusion with ATP and noradrenalin. The contraction of pericyte can propagate from stimulated one to distant pericyte along the capillary [69]. Interestingly, pericytes located at branching points of capillaries (about 20% of total pericytes) express especially high concentrations of contractile proteins, suggesting an association with physiological shear stress and a subsequent stabilization function [21].

These data support the evidence that pericytes control capillary blood flow in response to local modulation by vasoactive mechanisms. However, whether the hyperglycemic milieu or the loss of pericytes has an impact on contractility, in particular when basement membrane components have changed under hyperglycemic conditions, needs further clarification.

Role of Pericytes in Vessel Formation and Stabilization

The most prominent function of pericytes is their role in vascular growth and vessel stabilization. Studies highlighted the importance of pericytes for vessel maturation in embryonic development, in vascular remodeling, and in guidance of sprouting angiogenesis [70–74].

ECs can initiate, but not complete vessel formation. After the primary network of vessels has formed during vasculogenesis (de novo formation of vessels from vascular precursor cells), maturation of the primitive network ensues. During this process, pericytes are recruited to the forming vasculature. Important molecular pathways involved in pericyte recruitment during embryonic vessel maturation are PDGF-B and its receptor PDGFR-β, TGF-β and its receptor and the an- giopoietin/Tie-2 system (Ang/Tie-2). While the PDGF/PDGFR system is crucial for pericyte migration and proliferation during vascular maturation, the angiopoietin/Tie-2 system is essential for subsequent vessel stabilization, and TGF-β is involved in the interaction of vascular cells with extracellular matrix (ECM) and ECM production

and further mural cell differentiation. Inhibition of pericyte recruitment to capillaries by interfering with the recruitment leads to abnormal remodeling of developing vessels, a process that is reversed by administration of endothelial survival factors [75].

The term angiogenesis describes a different process of blood vessel formation and is probably the predominant way of de novo blood vessel formation in the mature retina. In response to growth factor gradients, capillary sprouts start to evade from preexisiting vessels [76]. Recent studies demonstrated that in sprouting angiogenesis, individual ECs achieve a guiding tip cell phenotype while neighboring stalk cells proliferate and form the vascular lumen [77]. EC proliferation and migration of sprouting tip cells include the degradation and losing of ECM and is dependent on tip cell guidance [78]. To achieve a functional capillary network, pericytes are recruited to the growing vessels by a platelet-derived growth factor PDGF-B gradient which is produced by the sprouting ECs, but factors like Ang-1 also induce pericyte recruitment and the subsequent stabilization of newly formed vessels during development. Vessels are resistant to hyperoxic vasoregression when covered with pericytes. This led to the concept of the ‘window of plasticity’ determined by pericyte recruitment lagging behind endothelial sprouting [79]. By the use of complementary phase-specific pericyte markers outlined earlier, it was shown that pericytes play an active role in physiological and pathological angiogenesis, as they are frequently found near, at or even in front of the tips of endothelial sprouts [14, 27, 80]. Besides the PDGF-B/PDGFR-β system, other factors such as sphingosine-1-phos- phate-1 and the angiopoietins and are also involved in angiogenesis [81]. For the completion of vessel maturation, active TGF-β signaling via the ALK-1 and Smad5 pathway is needed [82]. When remodeling ceases, pericytes contribute to the stabilization of vessels by the production of collagen and ECM proteins, such as fibronectin,

laminin and glycosaminoglycans to the basal lamina [36, 83–86].

Once the entire vascular system has formed, the major function of pericytes is the maintenance of an intact vascular network. An important paracrine signaling pathways implicated in stabilization and survival of mature vessels is the angiopoietin/Tie-2 system; a growth factor system, known to be deregulated during diabetic retinopathy [87]. Ang-1 and Ang-2 signal via the tyrosine kinase receptor Tie-2. While Ang- 1 activates Tie-2, the natural antagonist Ang-2 blocks Ang-1-induced Tie-2 phosphorylation by competitive binding. Tie-2-mediated signaling via the Akt/PKB pathway regulates cell survival, cell migration and cell-cell interactions [88]. In the retina, pericytes are a predominant source of Ang-1 [74, 89] and ECs express the Tie-2 receptor. Pericyte-derived Ang-1 might promote vessel integrity and tightness of blood-retinal barrier by controlling EC proliferation, induction of cell-contact proteins, and making ECs refractory to shifts in oxygen tension and growth factor levels. Under certain pathological conditions the expression of Ang-1, its natural antagonist Ang-2 or Tie-2 receptor may change, resulting in altered EC-pericyte interaction and capillary regression.

Pericyte Loss in Diabetic Retinopathy

Given the complex cell-cell crosstalk outlined earlier, and the specific function of pericytes in maintaining blood-retinal barrier and quiescent vasculature, their loss in diabetic retinopathy predicts profound consequences for the integrity of the affected tissue.

The complex structure of retinal microvasculature, as an integrative part of the retina, and lack of adequate animal models slowed down progress in understanding the cellular changes in diabetic retinopathy. Important insight into the histopathological changes of diabetic retinopathy dates back to the early 1960s. Cogan et al. [19]

developed a method that allowed for the direct inspection of the affected retinal vasculature liberated from neuronal and neuroglial tissues due to differential susceptibility against trypsin digestion. As a result of their work and that of others presented later, pericyte loss was identified as one of the earliest changes in the diabetic retina. Given the complex function of pericytes in the mature vasculature, their early loss in developing retinopathy would have profound consequences for the underlying endothelium and the integrity of the blood-retina barrier. In fact, pericyte loss as a hallmark of human diabetic retinopathy is accompanied by EC degeneration leading to vasoregression, vascular leakage and formation of microaneurysms as a sign of vessel destabilization and focal EC proliferation.

Animal Models of Diabetic Retinopathy

Human original material is unavailable, and data on the degree and the time course of pericyte function mostly derive from animal models as widespread as from mice to monkeys.

Hyperglycemic Animal Models of Diabetic Retinopathy

Rats are the most commonly used models of experimental diabetic retinopathy. Permanent hyperglycemia is usually achieved by chemical induction with streptozotocin. In this model, pericyte loss of over 30% after 6 months of hyperglycemia and progressive capillary regression is reproducible, depending on the strain used [90– 92]. Rat models mimicking human type 2 diabetes show similar morphological characteristics, suggesting that pericyte loss and other changes characteristic for retinopathy are caused by hy- perglycemia-mediated mechanisms [93–95]. As in other animal models, the addition of sucrose to the diet in genetically obese fatty fa/fa rats resulted in a pronounced decrease in pericytes and signs of pericyte degeneration.

Withthewidespreaduseofgeneticallymanipulated mice, their use, in particular in combination

with chronic hyperglycemia, has become popular, although their retinopathic phenotype is considered less severe and the identification of pericytes more difficult than rats. Like in rats, diabetic retinal capillaries of mice show a loss of pericytes of 15–25% and significantly increased formation of acellular capillaries after 6 months of diabetes duration. Spontaneous diabetic mice as well asmouse models of type 2 diabetes are available [96–100]. Exceptional animal models of diabetic retinopathy are the spontaneous diabetic Chinese hamster, diabetic dogs and diabetic monkeys. They all show morphological characteristics of diabetic retinopathy, including early pericyte loss and vasoregression [101–109]. Therefore, pericyte loss represents a universal event in diabetic retinopathy, not restricted to specific conditions or species, and it is unlikely that humans are an exception from the rule. Nevertheless, the degree of pericyte loss and the severity of hyperglycemia-induced vascular changes differ in the different specimens and animal strains. Over and above, severity of hyperglycemia and hypertension has a clear impact on the degree of pericyte loss.

Genetically Modified Models Mimicking Diabetic Retinopathy

Availability of genetically modified organisms introduced a novel approach to study pathomechanisms of diabetic retinopathy. As noted earlier, angiogenic factors like the PDGF-B/PDGFR-β, the Ang/Tie-2 system and other factors are important in the recruitment of pericytes to the vasculature during retinal development. It becomes apparent that the same growth factors play key roles in the mechanisms underlying vascular changes in diabetic retinopathy. Thus, mice with modifications of these factors are suitable to understand the vascular consequences of pericyte deficiencies.

In diabetic mouse retinas, PDGF-B mRNA is decreased when compared to nondiabetic controls, suggesting a role of PDGF-B in hyperg- lycemia-induced pericyte loss. Heterozygous

PDGF-B-deficient mice showed a 28% reduction in the number of retinal pericytes compared to wild-type littermates and an increase in acellular capillaries during adulthood, implying that pericytes have survival-promoting functions for established retinal capillary. Hyperglycemia further aggravates pericyte dropout in this animal model. Retina of mice with an endothelium-re- stricted ablation of PDGF-B showed that pericyte loss up to 50% was accompanied by vasoregression in the retina, whereas pericyte deficits exceeding 50% induced retinal vasoproliferation mimicking proliferative diabetic retinopathy [52, 110]. The effect of inhibition of PDGF-B/PDGFR signaling in either a genetic or a therapeutic approach needs to be investigated.

Another growth factor system gaining increasing attention in regard to early pericyte loss in diabetic retinopathy is the angiopoietin/Tie receptor system. The interaction of Ang-1 with Tie-2 is crucial for the timely and coordinated recruitment of pericytes to the developing vascular system. To assess the pathogenetic role of the angiopoietin-Tie system in diabetic pericyte loss, the expression of Ang-1 and -2 was studied in a diabetic rat model, in which the onset of pericyte dropout is exactly known. It was found that Ang- 2 is dramatically upregulated prior to the onset of pericyte dropout. Injection of recombinant Ang-2 into the vitreous of nondiabetic rats reproduced pericyte dropout within days, and the 25% pericyte loss of diabetic C56BL6/J mice after 6 months of diabetes was abolished in a diabetic mouse with a 50% reduction of Ang-2 gene dose [87]. Furthermore, constitutive overexpression of Ang-2 in photoreceptor cells induces reduced pericyte coverage in the deep capillary layers of the retina during retinal development. Similarly to the changes in PDGF-B+/− mice, hyperglycemia aggravates pericyte loss and vascular changes in adult Ang-2 overexpressing retina over time [111].

Studies in animal models of VEGF and IGF-1/- receptor signaling revealed that VEGF may play

a significant role in pericyte recruitment, but the specific and predominant mechanism working in the eye, and in particular in the capillary network under pathological condition may be difficult to dissect [112]. In contrast, overexpression of IGF-1 in the outer nuclear layer and in photoreceptors of the retina led to almost 50% reduction in pericyte numbers [113]. Interestingly, the numbers of ECs remained unexpectedly unchanged, whereas acellular capillaries were more numerous already at the observed time point.

Mechanisms of Pericyte Loss

The underlying causes and mechanisms of early pericyte dropout in diabetic retinopathy still remain unclear. It is possible that pericyte loss is a result of passive processes, such as degeneration and apoptosis. For example, streptozotocin diabetic Wistar rats showed a 2.65-fold increase in the numbers of TUNEL-positive cells (including pericytes) in retinae after 11 months of hyperglycemia [114, 115]. Pericyte death was also present in retinal vessels of diabetic patients suggesting its relevance in clinical disease [116]. Alternatively to this hypothesis, growing evidence indicates that pericyte loss in diabetic retinopathy is the result of an active elimination via altered growth factor production by the neighboring cells in the retina.

Biochemical Mechanisms

The selective damage of pericytes by chronic hyperglycemia may be explained by altered biochemical pathways which have been implicated in the pathogenesis of microvascular damage, so that pericyte loss may be the consequence of hyperglycemic toxicity. The four biochemical pathways that have been discussed over years to be involved in the pathogenesis of diabetic complications are (a) increased activity of the polyol pathway, (b) activation of protein kinase C (PKC) isoforms by de novo synthesis of diacylglycerol

(DAG), (c) increased flux through the hexosamine pathway, and (d) supply of glycolytic intermediates for the formation of advanced glycation end products (AGEs). Furthermore, a direct toxic effect of modification of LDL has been discussed.

Polyol Pathway

The first pathway to be studied in this regard, was the aldose reductase pathway, as immunological evidence had suggested the selective presence of aldose reductase in pericytes [117]. When glucose levels in cells are low, the enzyme aldose reductase functions by detoxifying aldehydes to inactive alcohols. When glucose levels in cells rise in diabetes, the enzyme starts to reduce glucose to sorbitol, and this process consumes the cofactor NADPH. Since NADPH is an essential cofactor for the regeneration of an important intracellular antioxidant, reduced glutathione, its depletion may induce a significant impact on cellular defense against oxidative stress. According to novel findings, the expression and activity of aldose reductase are increased in bovine retinal pericytes in vitro when cultured in high glucose [118], and are accompanied by elevated intracellular sorbitol levels. Whether this leads to increased pericyte death and is amenable to pharmacological inhibition has not yet been demonstrated. Overall, in line with the notion that the absolute levels of aldose reductase in the retina may be too low to contribute significantly to retinopathy development, the majority of experimental, and one large clinical trial failed to establish this pathway as playing a major role.

Activation of Protein Kinase C Isoforms

Most PKC isoforms are activated by the lipid second messenger DAG. Intracellular hyperglycemia increases the amount of DAG in cultured microvascular cells and in the retina, and the de novo synthesis DAG subsequently promotes PKC activation. The β- and δ-isoforms of PKC are activated primarily and elevate PKC-α and PKC-ε isoforms in the retina. Hyperglycemia-induced

activation of PKC isoforms may mediate by the receptor of AGE and increased activity of polyol pathway (ROS). In retinal pericytes, the activation of PKC-β2 isoform is involved in control of VEGF expression through RNA-binding protein HuR, which stabilizes mRNA [119]. Retinal pericytes express VEGF and pigment epithelium-derived factor (PEDF). VEGF promotes retinal neovascularization but PEDF suppresses ischemia-induced retinal neovascularization due to induction of apoptosis in retinal ECs. The PKC-MAPK signaling suppression by retinal pericyte-conditioned medium prevents retinal EC proliferation [120]. Hyperglycemia induces ET-1 mRNA expression in capillary bovine retinal ECs and bovine retinal pericytes. The α-, β1- and δ-isoforms of PKC are significantly increased accompanied with ET-1 elevation in retinal pericytes [121–123]. PKC isoform activation can be inhibited by specific PKC inhibitors in the retina. The activation of PKC contributes to increased microvascular matrix protein accumulation by inducing TGF-β1, fibronectin and type IV collagen in mesangial cells of diabetic rats. The activation of NF-kB and overexpression of the fibrinolytic inhibitor plasminogen activator inhibitor (PAI-1) are also involved in hyperglycemia-induced activation of PKC in vSMCs.

Hexosamine Pathway

In diabetes, increased flux of fructose-6-phos- phate diverted from glycolysis to UDP-N- acetylglucosamine promotes modification of protein by O-linked N-acetylglucosamine (GlcNac). The rate-limiting enzyme in this pathway is glutamine:fructose-6-phosphate aminotransferase (GFAT), converting glucose to glucosamine. GFAT is linked to transcription of TGF-α, TGF-β and PAI-1 promoter mediated by transcription factor Sp1 in vSMCs [124]. GlcNac covalently modifies Sp1 and regulates PAI-1 transcription. Glucosamine itself also activates the PAI-1 promoter through the Sp1 site. This pathway has a key role in fator glucose-induced

insulin resistance. Apart from that, recent evidence suggests that GlcNac also contributes to hyperglycemia-induced upregulation of Ang-2 by Sp3 binding (see below).

AGE Formation

Chronic hyperglycemia-induced formation of reactive oxygen species and AGEs, which accumulate in pericytes in vivo [125] might be able to initiate pericyte degeneration. Methylglyoxal is the most important intracellular AGE precursor, which reacts with aminogroups of arginine in intracellular proteins to form AGEs. Interestingly, injection of exogenous AGE into nondiabetic animals resulted in a selective uptake in pericytes [126]. In principle, repeated injection of high doses of AGE-modified rat serum can induce selective pericyte loss in normal rats after 2 weeks [127]. Moreover, endogenous AGEs can form and accumulate in pericytes [125]. While the ingestion of exogenous AGEs is consistent with the propensity of phagocytosis of pericytes, the formation of endogenous AGEs in pericytes is inconsistent with the prior finding that pericytes and SMCs are able to downregulate glucose uptake to protect themselves from hyperglycemic damage [128]. It is thus speculated that pericytes take up AGEs from the circulation or from the direct vicinity, suggesting a clearing function under specific conditions. Since the tissue load with AGEs changes over time in diabetes, the role of pericytes as an AGEremoving cell compartment becomes increasingly relevant. However, the time course of AGE accumulation in pericytes (occurring over several months) is inconsistent with the time course of pericyte loss, starting after approximately 8 weeks of diabetes with a plateau after 6 months in diabetic animals.

Selective Pericyte Injury – Modification of LDL as an Example

Another piece of evidence suggests that pericytes may be injured by toxic plasma proteins which

predominantly form in the diabetic milieu. Pericytes are differentially susceptible to damage by modified LDL in vitro [129]. As in vivo correlate, the combination of elevated glucose and lipids in ApoE–/– db/db mice led to an almost doubling of pericyte ghosts after 6 months of exposure. The majority of genes upregulated in human pericytes exposed to modifications of LDL such as oxidized-glycated LDL belong to the families of signal transduction, enzymes, and lipid metabolism [130, 131]. One interesting gene regulated by exposure of pericytes to modified lipids is TIMP-3, which controls vessel stability and maturation in vitro. Exposure of cultured human retinal pericytes to glycatedoxidized LDL repressed TIMP-3 expression 2.4- fold vs. pericytes exposed to unmodified LDL. Given the close cell-cell communication based on physical contacts between ECs and pericytes, it is conceivable that the MMP-TIMP system can play a significant role in execution of vascular stabilization by pericytes, and to the respective alteration in diabetes [132]. Although a clear clinical benefit of lipid-lowering treatment for diabetic retinopathy has not been demonstrated [133, 134], the preclinical data and evidence from associative studies suggest an association between lipoprotein profiles and the severity of retinopathy at least in type 1 diabetes [135]. It must be kept in mind that lipid-lowering drugs can have an independent effect on pericyte survival. For instance, it was recently reported that statins, in particular simvastatin, a potent HMGCoA reductase inhibitor, can selectively induce pericyte apoptosis in vitro.

A Unifying Mechanism of Biochemical Dysregulation

As mentioned before, four pathways have been investigated over many years to explain how diabetes can affect diabetic vascular target cells: the polyol pathway, the hexosamine pathway, the PKC pathway, and increased production of AGEs. Recently, these seemingly independent

biochemical pathways have been linked by the findings that one single mechanisms, hypergly- cemia-induced mitochondrial overproduction of reactive oxygen species, is the underlying cause, which, mediated through the enzyme poly-ADP- ribose polymerase, blocks activities of the critical glycolytic enzyme glyceraldehyde-3-phosphate dehydrogenase [136, 137]. In normal cells, glucose is metabolized through glycolysis and the tricarbon cycle to generate electron donors for the mitochondrial respiratory chain. Here, energy (ATP) is generated in a precisely regulated way. In hyperglycemic cells, increased flux through glycolysis and the TCA cycle generates a voltage gradient of electrons surpassing a certain threshold which is then blocked inside complex III of the mitochondrial electron transport chain. From complex III, the surplus of electrons is purported to coenzyme Q together with molecular oxygen producing superoxide. The mitochondrial form of superoxide dismutase detoxifies this radical via hydrogen peroxide to water in the presence of oxygen. The link between the four biochemical pathways, and the mitochondrial overproduction of ROS was made when it became evident that an important change in hyperglycemic cells and in experimental animals was the reduced activity of the glycolytic enzyme glyceraldehyde- 3-phosphate dehydrogenase. When examining for biochemical modifications of the enzyme, it was observed that hyperglycemia-induced superoxides caused polymers of ADP-ribose to attach and reduce enzyme activity. These changes were prevented with the inhibition of superoxide generation, and with the inhibition of the nuclear enzyme poly-ADP-ribose polymerase using a specific PARP inhibitor. The latter is activated upon DNA strand brakes known to form in hyperglycemic cells. Reduced GAPDH activity induced by PARP activation activates the biochemical pathways by increasing intermediates such as di- acyl-glycerol (PKC pathway), glyceraldehyde-3- phosphate (AGE pathway), fructose-6-phosphate (hexosamine pathway), and intracellular glucose

(sorbitol pathway). Thus, hyperglycemia links to biochemical pathway abnormalities via a common denominator, mitochondrial overproduction of reactive oxygen species.

Pericyte Loss through Active Elimination

Alternatively to early pericyte loss in diabetic retinopathy being the result of hyperglycemic injury, a different concept was proposed. It is possible that pericyte loss is an active process involving migration of pericytes away from the capillaries, driven by the angiopoietin-Tie system.

Hyperglycemia-Induced Ang-2 Transcription As described above, gain of function experiments in nondiabetic animals revealed the induction of pericyte dropout in the vicinity of the Ang-2 overexpressing site. Superimposition of diabetes aggravated the most important vascular readout, i.e. the formation of acellular capillaries. Loss of function studies in the presence of diabetes yielded the prevention of pericyte dropout and the reduction in acellular capillary formation, accentuating the importance of Ang-2 in diabetic pericyte loss. Ang-2 is expressed in three cell types of the retina, i.e. the EC, the Müller cells, and the horizontal cells. In situ hybridization of diabetic retinae for Ang-2 yielded the expression particularly in Müller cells. Recently, the regulation of Ang-2 in chronic hyperglycemia has been delineated using renal microvascular cells as paradigm.

Yao et al. [138] proposed a mechanism by which AGEs enhance transcription of Ang-2, as a crucial factor in the development of diabetic retinopathy. It was found that increased glucose flux in renal microvascular ECs caused increased modification of the co-repressor mSin3A by the intracellular AGE methylglyoxal, resulting in recruitment of the enzyme O-GlcNAc transferase to an mSin3A-Sp3 complex. Subsequently, Sp3 modification by O-linked N-acetylglucosamine decreased its binding to the glucose-respon- sive GC box in the Ang-2 promoter and the activation of Ang-2 transcription [138]. The same