Review of the Pericyte during Angiogenesis and its Role in Cancer and Diabetic Retinopathy

Introduction

Pericytes were once known as “Rouget” cells, after the initial description by Eberth 2(1871) and Rouget (1873, 1879) of a perivascular cell adjacent to capillaries. In 1923, Zimmerman introduced the term “pericyte” (Allt and lawrenson, 2001), but more recently, the terms “mural cell”, “vascular smooth muscle cell” (vSMC), and pericyte have been used interchangeably to describe a cell that forms intimate contacts to support microvessels (Figures 1 and 2). The relationship between endothelium and pericyte is for each to provide the other with growth factors and cell contacts that promote mutual proliferation and survival.

The support that pericytes provides vascular endothelium during pathological angiogenesis has attracted enormous attention (reviewed in von Tell et al., 2006 and Yamagishi and Imaizumi, 2005). It is now becoming clear that pericytes are directly involved in the pathogenesis of vascular driven diseases, such as, for example, diabetic retinopathy and neoplasia. Modulation of their function, therefore, is likely to ameliorate these diseases. Evidence generated mainly in pre-clinical models, supports this notion and indicates that targeting them may provide additional benefits to those conferred by simply targeting endothelium alone.

This review focuses on the biology and functions of pericytes, and describes their role in diabetic retinopathy and neoplasia. Their role in neoplasia is of especial interest, due to the number of therapeutic molecules that are now progressing through pre-clinical development and are specifically designed to target their function during tumor angiogenesis. Much of the work generated from these studies supports the paradigm that targeting both endothelial cells and vascular support cells (pericytes/smooth muscle) provide synergistic effects (Bergers et al., 2003; Erber et al., 2004), and may welcome in a new era of molecularly targeted drugs.

Morphology and Distribution

Classically, mature pericytes form intimate, umbrella-like contacts with the endothelial cells in capillaries, pre-capillary arterioles, collecting venules and postcapillary venules. Endothelial cells and pericytes both synthesise and share a common basement membrane (Mandarino et al., 1993). Pericytes communicate with endothelial cells directly through peg-and-socket junctions that extend through a discontinuous basement membrane (Cuevas et al., 1984). These junctions consist of membrane evaginations that are rich in tight and gap junctions. Pericyte-endothelial adhesion, recognition and signalling are also maintained by N-cadherin adhesion junctions (Gerhardt et al., 2000) and fibronectin-rich dense plaques (Courtoy and Boyles, 1983).

It is now apparent that these umbrella-like processes extend to more than one capillary (Ando et al., 1999) and are more extensive on venous capillaries and post-capillary venules (reviewed by Hirschii and D’Amore, 1996). Pericyte coverage per se is also more extensive on post-capillary venules than capillaries, with overall pericyte coverage (in the rat) varying between 11% in cardiac muscle to 41% in the retina (Sims, 1991). Pericytes are also more numerous in the distal regions of limbs (Sims et al., 1994), predominate over endothelial junctions in the ocular choroid, lung and skin (Sims and westfall, 1983) and are polarised over endothelial surfaces not involved in gas exchange (reviewed in Gerhardt and Betsholtz, 2003). The bias of pericytes towards the distal venular compartment strongly hints at their role in vivo, which is to support the microcirculation against hydrostatic pressure.

Phenotypic Markers

Pericytes express a number of markers of differentiation such as alpha smooth muscle actin (αSMA), desmin, chondroitin sulfate proteoglycan marker NG-2 (or high molecular weight melanoma antigen), PDGFRβ (platelet-derived growth factor receptorβ), aminopeptidases A and N and RGS5 (regulator of G-protein signalling) (Bondjers et al., 2003; Armulik et al., 2005). The regulation of marker expression is still largely unknown, although it is becoming clear that the switch governing αSMA and NG2/desmin expression appears to be dependent upon the release of TGFβ by endothelial cells—in its presence pericytes express αSMA, and in its absence they express NG2 and desmin (Song et al., 2005).

The lack of reliable markers has hindered pericyte identification. No single marker is able to identify all pericytes. In addition, these rather nebulous cells are able to express different markers depending upon context, species and local anatomy—for example—αSMA is perhaps the best characterised and most frequently used pericyte marker. It is expressed by most pericytes and vSMC. However, it is now apparent that both myofibroblasts (Darby et al., 1990) and cultured fibroblasts (Schurch et al., 1997) can be induced to express αSMA and synthesise collagen (Roberts et al., 1986) under the influence of TGF-β1 (transforming growth factor β1) (Desmouliere et al., 1993; Hales et al., 1994). Furthermore, microvascular pericytes derived from the mouse embryo brain and the human brain, do not express αSMA, whereas pericytes derived from the chicken embryo brain do express αSMA (Gerhardt et al., 2000). Finally, αSMA and high molecular weight melanoma antigen expression can be induced by culturing human brain pericytes in media containing serum, and further upregulated by the addition of TGFβ to the culture medium (Verbeek et al., 1994).

Similarly, NG2 expression has been identified as another putative pericyte marker (Ozerdem et al., 2001) that has subsequently been found to be of more limited value, since its expression is restricted to arteriolar and capillary perivascular cells (vSMC and pericytes) and is absent on venular pericytes (Murfee et al., 2005).

Biological Plasticity

Pericytes are multipotent cells, retaining the capacity to differentiate into vascular smooth muscle cells (Nehls and Drenkhahn, 1993) and a variety of other mesenchymal cell types. The relationship between vascular smooth muscle and pericytes was inferred by electron microscopic observations of a transitional pericyte/vSMC-like cell within terminal arterioles and venules (Meyrick and Reid, 1979; Diaz-Flores et al., 1991). More recently, it has been established that they can differentiate into vSMC in vitro (Nicosia and Villaschi, 1995; Hirschi et al., 1998) and that this in vitro phenomenon, is supported in vivo, by the generalized deficiency of both pericytes and vascular smooth muscle in PDGFRβ and PDGF-B (platelet derived growth factor-B) null mutants (Lindahl et al., 1997; Hellstrom et al., 1999).

Under the appropriate conditions, pericytes can also differentiate into myofibroblasts/fibroblasts (Ronnov-Jessen et al., 1995; Sundberg et al., 1996). The reverse transformation of vSMC (Nicosia and Villaschi, 1995) and fibroblasts (Sundberg et al., 1996) into pericytes is also possible. The myofibroblast shares many features in common with the pericyte, including the expression of αSMA and desmin (Morikawa et al., 2002). They are however, not a typical component of normal untraumatized tissues (reviewed by Eyden, 2005). They are present in pathological conditions, such as healing wounds and tumors. Here they appear as spindle shaped, stromal cells that exist free within the interstitium, lacking close contacts with blood vessels. Recent experiments have shown that stromal cells within tumors, and normal rat skin, can modulate interstitial fluid pressure through the PDGFRβ (platelet-derived growth factor receptor beta) (Rodt et al., 1996; Pietras et al., 2001). Thus treatment of tumors through the PDGFRβ that target pericytes, are also likely to target stromal cells, and thus reduce interstitial fluid pressure and improve drug delivery.

Myofibroblasts may, however, represent just one of several possible lines of pericyte differentiation. Pericytes are also thought to give rise to a number of other mesenchymal cell types, all of which are found within bones or bone marrow, such as adipocytes (Richardson et al., 1982), osteoblasts (Diaz-Flores et al., 1992; Schor et al., 1995) chondrocytes and phagocytes (reviewed in Hirschi and D’Amore, 1996 and Sims, 2000). Human mesenchymal stem cells are known to have similar properties (Muguruma et al., 2006) suggesting that pericyte plasticity may be due to their relatively undifferentiated status. This aspect of pericyte biology however remains relatively unexplored.

Derivation

Pericytes are thought to be derived from a Flk-1⁺/VEGFR-2⁺ (foetal liver kinase-1/vascular endothelial growth factor receptor-2) mesenchymal cell precursor that has the capacity to differentiate into either vSMC or endothelial cells under the influence of platelet-derived growth factor (PDGF) BB or vascular endothelial growth factor (VEGF) respectively (Yamashita et al., 2000). The phenotypic switch is thought to be regulated by the transcription factor Tal1 (Scl) (Ema et al., 2003) that acts to inhibit vSMC differentiation while promoting endothelial differentiation. In the adult, it is unknown whether endothelial cells and pericytes retain a common progenitor cell.

Postnatally, pericytes are predominantly derived, in situ, from local mesenchymal cells. Recent reports however have described an alternative source of “periendothelial vascular mural cells” (pericytes). Experiments in mice have revealed that cells expressing NG2 or the haemopoietic markers, CD11b and CD45, can be successfully recruited from the adult bone marrow to furnish capillaries with pericytes (Rajantie et al., 2004). A similar experiment using the RIP-Tag2 model of pancreatic cancer, demonstrated recruitment of Sca-1⁺ (stem cell antigen) bone marrow derived cells (Song et al., 2005). Sca-1⁺ cells constituted approximately 23% of all PDGFRβ⁺ pericytes within the pancreatic tumours (this latter Sca-1⁺/PDGFRβ⁺ subset formed approximately 50% of all Sca-1⁺ cells). Reconstitution of RIP-Tag2 mice with GFP (green fluorescent protein) expressing bone marrow, revealed that 16.6% of pericytes expressed PDGFRβ only, 14% expressed NG2 and PDGFRβ, and 40% expressed NG2 alone. GFP/NG2⁺ cells formed a subset of Sca-1⁺ cells, but unlike previous observations, did not express CD45 or CD11b (Song et al., 2005).

Pericyte Recruitment

Genetic ablation of PDGF-B or PDGFRβ in mice, results in perinatal lethality due to microaneurysm formation and widespread vascular leakage resulting from an extensive mural cell deficit (Leveen et al., 1994; Soriano, 1994; Lindahl et al., 1997). PDGF/PDGFRβ appears to be critical for the differentiation of local mesenchymal cells, and for the proliferation and possibly migration of preexisting vSMC. A recent reconstitution experiment, in which mice received a GFP (green fluorescent protein) expressing bone marrow transplant, showed that a proportion of PDGFRβ⁺ pericyte precursor cells can be recruited from the adult bone marrow (Song et al., 2005) presumably also under the influence of PDGF-B ligand.

PDGF-B is normally secreted by the endothelial cells in immature sprouting capillaries and proliferating arteries (Hellstrom et al., 1999). Locally, it acts as a paracrine factor to correctly integrate PDGFRβ⁺ precursor cells into vessel walls (Abramsson et al., 2003; Rhodin and Fujita, 1989; Hirschi et al., 1998). Unprocessed PDGF-B is secreted with a C-terminal retention motif that anchors the protein to the surface of the cell via heparin sulphate proteoglycans in the pericellular space (LaRochelle et al., 1991). This binding motif serves to effectively limit the range of influence of the protein (Eming et al., 1999), binds PDGF-B to extracellular matrix proteins to generate the proper extracellular concentration gradient, and ensure correct orientation and tight adhesion of pericytes. A similar motif in the VEGF-A protein has been shown to guide correct endothelial sprouting and branching morphogenesis (Ruhrberg et al., 2002).

Absence of the PDGF-B retention motif allows more widespread diffusion of the protein and depolarisation of the extracellular matrix gradient. PDGF-B ret/ret mice, which lack this retention motif, develop abnormally dilated tumor microvessels with defective integration of pericytes into the vascular wall. Tumor pericytes partially dissociate from the endothelial surface, and a very high proportion develop cytoplasmic processes extending more than 10 μm away from the vessel wall (Lindblom et al., 2003). Tumors from these mice also tend to be more haemorrhagic at the time of isolation (Abramsson et al., 2003) indicating vascular fragility.

Role of Factors During Vessel Maturation

Maturation of the microvasculature is under the control of a number of factors, in addition to PDGF-B, that regulate pericyte differentiation and maturation.

Angiopoietins

The Ang-Tie receptor system consists of a family of angiopoietins (Ang-1, Ang-2, Ang-3 and Ang-4) and the endothelial tyrosine kinases, Tie- 1 and Tie-2 (Thurston, 2003; Huang and Bao, 2004). Ang 1-4 act as ligands for the Tie-2 receptor—no known ligand has been found for Tie-1. The Tie receptors are expressed mainly on endothelium, although Tie-2 is now known to be expressed on 3 distinct cell populations involved in neovascularization: endothelial cells, proangiogenic bone marrow-derived monocyte/macrophages, and mesenchymal cells destined to become pericytes (De Palma et al., 2005).

Ang-1 and Ang-2 act in a context dependent manner to stimulate angiogenesis and vascular maturation. Ang-1 appears to be expressed mainly by perivascular and mural cells (Sundberg et al., 2002) functioning as a paracrine signal to promote vessel stabilisation. This is achieved by inducing pericyte attachment and reducing vascular permeability (Asahara et al., 1998; Thurston et al., 1999; Hawighorst et al., 2002).

Ang-2 antagonises Ang-1 function by inducing loosening of the attachment between endothelial cells and pericytes/vSMCs (Maisonpierre et al., 1997). In the absence of VEGF, Ang-2 induces regression of blood vessels (Sato et al., 1995; Dumont et al., 1994).

In the presence of VEGF, Ang-2 facilitates vessel sprouting (the formation of blind ending capillary tubes) (Asahara et al., 1998) and intussusception (the folding of slender tissue pillars or posts into the vessel lumen) (Patan et al., 1992). Since Ang-2 and Ang-1 both bind to Tie-2 with equal affinity (Masionpierre et al., 1997), increasing the ratio of Ang-1:Ang-2 favors vascular stability. On the other hand, high levels of Ang-2 and VEGF, normally found at the leading front of angiogenesis, favors vessel destabilization and endothelial cell migration (Yancopoulos et al., 2000; Vajkoczy et al., 2002).

The most compelling evidence for the requirement of the Tie receptors for angiogenesis and pericyte stability however, has been derived from genetic ablation experiments. Tie-1 and Tie-2 receptor null mice, like other pro-angiogenic null mice, such as VEGF/VEGFR deficient mice, are not viable and die in utero, between E9.5–10.5 (Tie-2) and E13.5- E18.5 (Tie-1) depending on strain (Dumont et al., 1994; Sato et al., 1995; Puri et al., 1995). Both Tie-1 and Tie-2 receptor knockout mice develop vascular anomalies. Tie-2 mutant mice develop abnormal vascularisation and trabeculation of the heart (Dumont et al., 1994; Sato et al., 1995) with defective vessel remodelling resulting in abnormally dilated vessels that lack mural cell support (pericytes and vSMCs) (Patan, 2004). The phenotype of Tie-2 deficient mice is similar to that of Ang-1^−/− deficient mice, and Ang-2 overexpressing mice (Masionpierre et al., 1997) supporting the theory that Ang-2 functions as the natural antagonist to Ang-1 effects on the Tie-2 receptor (Masionpierre et al., 1997) and that Ang-1/Tie-2 is necessary for mural cell recruitment (Folkman and D’Amore, 1996). A second report suggests that at least in the rat aorta, mural cell precursors express Tie-2, respond chemotactically, and can be induced to express MMP-2 (matrix metalloproteinase-2) (Iularo et al., 2003).

Gap Junctions and Transforming Growth Factor β

The mechanism of TGFβ activation, necessary for induction of the vSMC phenotype is not completely understood, but during mural cell recruitment, gap junction communication appears to be critical for this process. Gap junctions are clustered channels arranged in hexagonal arrays (Revel and Kamovsky, 1967) that allow the diffusion of secondary messengers, ions and metabolites through intercellular gaps (Gilula et al., 1972). They are composed of 6 connexin (Cx) proteins, each forming a hemichannel that docks to form a functional conduit between two adjacent cells (Willecke et al., 2002).

Both Cx43^−/− and Cx45^−/−deficient mice are unable to recruit mural cells necessary for vessel wall stabilization. Cx45 deficient mice show low levels of TGFβ1 in the epithelial layer of the yolk sac and die midgestation (E9.5–10.5) due to abnormal vessel formation resulting from arrested arterial growth and a failure to form a normal tunica media (Kruger et al., 2000). Similarly, Cx43^−/−deficient mouse embryo cultures are unable to activate latent TGFβ, since endothelial induced pericyte differentiation can be rescued by exogenous TGFβ1 (Hirchi et al., 2003).

TGFβ, angiopoietin and PDGF-B expression, and their effects on vascular proliferation/stabilisation are intimately linked via complex feedback autoregulatory loops that are now beginning to be elucidated. These regulatory loops promote vessel destabilization, necessary for sprouting angiogenesis at early time-points, but have the opposite effect at later time-points, promoting vessel maturation. Vessel stabilisation is achieved because endothelial-derived PDGF-B normally upregulates vSMC release of TGFβ and Ang-1, via the MAPK/MEK and PI3Kinase/PKC pathways respectively (Nishishita and Lin, 2004). In addition, PDGF-B down-regulates Ang-2 release (Phelps et al., 2005). Ang-1 and TGFβ together, however, form a negative feedback loop, since in concert, they profoundly inhibit endothelial derived PDGF-B release (Nishishita and Lin, 2004). In this in vitro model, at early time-points, PDGF-B downregulates Ang-1 expression (within 2 hours of stimulation), promoting vessel destabilisation and angiogenesis. At later time-points, PDGF stimulation increases the Ang-1:Ang-2 ratio and TGFβ expression, promoting vessel stabilization and pericyte differentiation (Nishishita and Lin, 2004).

Notch 3

Arterial vSMC differentiation is further controlled by Notch3 signalling. The Notch signalling pathway is an evolutionarily conserved intercellular signalling mechanism consisting of a family of transmembrane receptors that are involved in the development of most vertebrate organs (Artavanis-Tsakonas et al., 1999) as well as specifying endothelial arterial-venous identity (Lawson et al., 2001; Fischer et al., 2004). Notch3 is thought to regulate arterial-venous differentiation by repressing venous cell fate within developing arteries (Lawson et al., 2001). Notch3^−/−deficient mice show defects in postnatal vSMC maturation and arterial specification, but retain normal endothelial specification (Domenga et al., 2004).

Other factors have been implicated in pericyte recruitment and vascular morphogenesis, such as, endothelial differentiation gene (Edg)-1 (Liu et al., 2000), N-cadherin (Tillet et al., 2005), heparin binding-epidermal growth factor-like growth factor (HB-EGF) (Iivanainen et al., 2003) and endothelial nitric oxide synthase (eNOS) (Yu et al., 2005). Their roles and function however, are less well characterised.

Endothelial Differentiation Gene (Edg)-1

Sphingosine-1-phosphate (S1P), a platelet derived lipid mediator, is the extracellular ligand for Edg-1, Edg-3, Edg-5, Edg-6, and Edg-8. Edg-1^−/−deficient mice die at E12.5–14.5 due to massive embryonic haemorrhage resulting from a defect in mural cell recruitment to vessel walls (Liu et al., 2000). The basis of this phenotype appears to be due to the dependence of PDGFRβ induced migration on (S1P)/EDG-1 intracellular signalling (Rosenfeldt et al., 2001). (S1P), receptor activation is also thought to be necessary for endothelial trafficking and strengthening of N-cadherin dependent peg-and-socket adhesion junctions between endothelial and mural cells necessary for vascular stabilisation (Paik et al., 2004).

Epidermal Growth Factor Receptor

The epidermal growth factor receptors (EGFR) form a family of four tyrosine kinase receptors consisting of ErbB1 (ErbB1 or HER1) (Ullrich et al., 1984), ErbB2 (c-Neu, HER2) (Yamamoto et al., 1986) ErbB3 (HER3) (Kraus et al., 1989) and ErbB4 (Her4) (Junttila et al., 2001). Their natural ligands consist of a dozen growth factors consisting of epidermal growth factor, transforming growth factorα, and amphiregulin, betacellulin, heparin binding-EGF-like growth factor (HB-EGF), epiregulin, tomoregulin, epigen and the neuregulins/heregulins (reviewed in Normanno et al., 2005 and Singh and Harris, 2005). Endothelial derived HB-EGF is thought to regulate recruitment of perivascular mural cells by binding to ErbB1 and ErbB2 present on these cells, and that Ang-1 further promotes this process by upregulating HB-EGF expression (Iivanainen et al., 2003).

eNOS

Recently, Yu and colleges (2005) have demonstrated that genetic loss of eNOS impairs VEGF and ischaemia-initiated blood flow recovery via impaired arteriogenesis, angiogenesis and/or mobilisation and recruitment of pericytes. Genetic loss of eNOS had no effect on endothelial progenitor recruitment.

Pericyte Functions

Pericytes promote vessel stability through physical and chemical signalling with endothelium. The 2 cell types act as a functional and physical unit through the establishment of cell-cell heterotypic contacts, and synthesis and secretion of growth factors that promote their mutual survival.

As described previously, PDGF-B/PDGFRβ deficient mice develop vessels that lack mural cell support. Vessels from these mice show numerous abnormalities including endothelial hyperplasia, increased capillary diameter, abnormal endothelial shape and ultrastructure, altered cellular distribution of certain junctional proteins, and morphological signs of increased transendothelial permeability (Hellstrom et al., 2001). Tumor blood vessels are similarly abnormal, presumably due in part to their abnormal pericyte coat (Morikawa et al., 2002).

The contractile nature of pericytes has been examined both in vitro and in vivo. The ability of pericytes to contract has been demonstrated in culture by the addition of several vasoactive substances such as histamine, angiotensin II, bradykinin, and seratonin (Murphy and Wagner, 1994; Speyer et al., 1999). These findings correlate with the observations that mid-capillary pericytes at the blood–brain and blood–retinal barriers strongly express the contractile protein αSMA (Bandopadhyay et al., 2001) and myosin (Joyce et al., 1985). The role of pericyte contraction however is less clear. Pericyte contraction may regulate microvascular blood flow and permeability. Contraction may also help resist high systolic capillary hydrostatic pressure and microaneurysm formation since this is the cause of death of PDGF-B/PDGFRβ deficient mice (Lindahl et al., 1997).

Specialized Roles of Pericytes

Pericytes appear to have more specialised roles in certain organs. At the blood brain barrier, reports demonstrate that pericytes are capable of acting as brain phagocytes (van Deurs, 1976) and may convert to tissue macrophages (reviewed by Thomas, 1999). They have also been shown to express macrophage markers (reviewed by Thomas, 1999) and perform Fc-receptor antibody-dependent phagocytosis (Balabanov et al., 1996).

A specialized role for pericytes has also been reported in the glomerulus. Here, the generation of PDGF-B and PDGFRβ null mice, together with mice bearing PDGFRβ mutant alleles and PDGF-B ret/ret (retention motif) deficits have demonstrated the requirement of PDGF-B for the proper recruitment of PDGFRβ⁺ mesangial cells into the glomerular tuft during embryonic and postnatal development (reviewed in Betsholtz et al., 2004). These mice develop vascular abnormalities of the glomerular tuft, indicating that mesangial cell recruitment is instrumental for glomerular capillary splitting and branching (Leveen et al., 1994; Soriano, 1994). Postnatally, these mutant mice also develop glomerular pathology, such as decreased glomerular cellularity, and increased mesangial matrix accumulation (reviewed in Betsholtz et al., 2004).

Role During Angiogenesis

Pericytes play a lead role in angiogenesis. In the ovarian model of angiogenesis, they are the first angiogenic cells to invade the normally avascular Graafian follicle, and thus precede the sprouting vascular tip (Amselgruber et al., 1999) (Figure 3). Pericytes also appear to extend beyond the sprouting tip in a number of mouse tumours examined using CD31, αSMA and desmin confocal microscopy (Morikawa et al., 2002). This model suggests that pericyte derived VEGF, as well as cell-cell contacts, might act to promote endothelial survival and guide migration (Darland et al., 2003; Kale et al., 2005).

This role however is in contradiction to other studies that suggest that pericytes lag behind endothelial migration. Observations during the postnatal development of the retinal vasculature, indicate that the naked angiogenic sprout represents a plasticity window allowing endothelial cells to remodel. Subsequent investment by pericytes (Benjamin et al., 1998) and the expression of Ang-1/PDGF (Hoffmann et al., 2005) closes this window indicating vascular maturation. This view is supported by in vitro models showing that endothelial tube formation is followed by pericyte coverage, and that pericytes use the developing sprouts as migration guidance cues (Nicosia and Vallashi, 1995; Hashizume and Ushiki, 2002). Furthermore, in an in vitro gelatin sponge model, pericyte coverage was shown only to extend beyond the limits of the sprouting vascular tips in areas where endothelial regression had previously occurred (Hashizume and Ushiki, 2002).

Role of Pericytes in Diabetic Retinopathy

The retina has the highest pericyte density of all vascular beds (Sims, 2000) and appear to play a crucial role in the development of diabetic retinopathy. The earliest identified lesion in the diabetic retina is pericyte loss (reviewed in Hammes, 2005). Pericyte loss is also accompanied by vSMC loss (Vom Hagen et al., 2005). In mouse models, when pericyte loss exceeds 50% of normal, proliferative retinopathy invariably develops (Enge et al., 2002). Pericyte loss progresses over time to endothelial cell loss, resulting in the formation of acellular capillaries (empty basement membrane tubes lacking endothelial cells) with the eventual formation of microaneurysms—the first clinically detectable lesion in the eye of diabetic patients. Progressive vascular occlusion is accompanied by increased vascular permeability—due in part to increased expression of VEGF (Benjamin, 2001) leading to intravascular endothelial proliferations within hypercellular vessels, and eventually, neovascularisation into the vitreous with haemorrhage and visual loss (reviewed in Frank, 2004).

The pathogenesis is still a subject of debate, although a unifying hypothesis proposes that hyperglycaemia increases mitochondrial electron transport, and so drives the overproduction of reactive oxygen species, such as superoxide (Nishikawa et al., 2000a, 2000b). Reactive oxygen species are considered a causal link between elevated glucose, metabolic imbalance and the development of diabetic complications (Brownlee, 2001). Treatments based on reversing the underlying metabolic disturbances, such as anti-oxidant therapy, have been shown to reduce retinal haemodynamic abnormalities (Bursell et al., 1999), correct retinal metabolic abnormalities (Kowluru and Kennedy, 2001), attenuate the upregulation of retinal vasculature VEGF (Obrosova et al., 2001) and reduce the formation of acellular strands in the retina (Stitt et al., 2002; Hammes et al., 2003).

Uncorrected chronic hyperglycaemia however, results in the elimination of pericytes—either through direct toxic metabolite accumulation (Stitt et al., 1997) or, as has been proposed more recently, indirectly through up-regulation of Ang-2 expression (Hammes et al., 2004). This is consistent with the known functions of the angiopoietins since Ang-1 and PDGF expression marks the closure of the plasticity window for vessel remodelling within the retina (Hoffmann et al., 2005). Ang-2 expression antagonises the effects of Ang-1, inducing vessel destabilization, pericyte detachment, and eventual loss.

Role of Pericytes in Tumors

Tumor vessels are patently abnormal. They tend to be dilated, tortuous, and hyperpermeable with correspondingly high interstitial fluid pressure due to global environmental abnormalities in the signalling molecules controlling angiogenesis (reviewed in Baluk et al., 2005; Jain, 2005). The analysis of angiogenesis in experimental models of cancer has demonstrated that these rapidly growing tumours have a tendency to develop an immature vascular phenotype with continuous microvessel growth and remodelling. These abnormalities result from defects in both compartments of tumor vasculature—endothelial and pericytes.

The RIP-tag2 transgenic mouse has recently become a popular model to study pericyte biology. These studies have shown that pericytes express a variable profile of markers consisting of populations of cells expressing a subset of αSMA, desmin, NG2 and PDGFRβ (Morikawa et al., 2002; Song et al., 2005). Confocal microscopy has demonstrated αSMA⁺/desmin⁺ cells present both as abnormal mural cells (pericytes) attached to blood vessels, as well as interstitial cells (myofibroblasts) free within the stroma, with no obvious association with blood vessels (Morikawa et al., 2002). Tumour pericytes adopt an abnormally loose association with microvessels, tending to lift off endothelium, with cytoplasmic processes extending deep into the tumor parenchyma (Morikowa et al., 2002). 97% of vessels in the larger pancreatic Islet cell tumours have pericyte coverage, although in other tumor models, pericyte density has been shown to be quite variable. Some tumors such as human glioblastoma (10–20% microvessel coverage) (Eberhard et al., 2000) have relatively few pericytes.

In 2 transplantable tumors, T241 fibrosarcoma and KRIB osteosarcoma, recruitment of PDGFRβ⁺ expressing cells to endothelium was relatively sparse, despite endothelial synthesis of PDGF-B ligand (Abramsson et al., 2002). Co-injection of T241 tumor cells with embryonic stem cells from mice engineered to express the marker protein lacZ in pericytes demonstrated that these cells could be efficiently recruited to a periendothelial location (Abramsson et al., 2002).

The role of pericytes in tumours is unclear. Potentially, pericytes may stabilise blood vessels, inhibit endothelial proliferation, maintain capillary diameter, regulate blood flow, and provide endothelial survival signals via heterotypic contacts and soluble factors.

Studies in mice have demonstrated that up to 90% pericyte loss is compatible with life (Enge et al., 2002) indicating that both the abnormalities and heterogeneity of tumor pericyte coverage seen in tumors may still be consistent with residual microvascular function. Indeed, transplantable xenograft tumors grow remarkably quickly despite these limitations. Hence the role of pericytes in the most rapidly growing tumors, which tend to adopt very immature vascular phenotypes, is open to some speculation. Abnormalities in pericytes most likely contributes to the relatively poor vascular perfusion and high interstitial fluid pressure in tumors.

Anti-angiogenic treatment for tumours

The role of pericytes in tumours is open to some speculation. In some studies, tumour pericytes do not provide full protection from angiogenic inhibition since VEGF inhibitors alone are able to induce endothelial cell regression (Willett et al., 2004; Inai et al. 2004). However, other authors have demonstrated that targeting VEGFR-2 does not cause significant tumor vessel regression due to their pericyte coverage (Erber et al., 2004) and that targeting both endothelium and pericytes may confer additional advantages by overcoming pericyte stabilisation and survival signals (Bergers et al., 2003; Erber et al., 2004). The reasons for these differences is not entirely clear, but may be attributed in part, by the heterogeneity of pericyte populations, and the associated difficulties in their correct identification (reviewed by von Tell et al., 2006).

The response of tumors to VEGF inhibition may also be attributed to the nature of the models used. The xenograft tumor is the standard cancer model used for the preclinical assessment of new therapeutic drugs (Figure 4). These transplantable tumours tend to be rapidly proliferating and usually grow in a subcutaneous (ectopic) site in immunodeficient hosts (Kerbel, 2003). However, the results obtained in xenograft tumours often markedly over-predict the results obtained in human phase III clinical trials (Kerbel, 2003). In contrast, human tumors tend to be slower growing, evolving over a protracted period of time. In addition, naturally occurring tumors grow within the normal (orthotopic) stromal environment. It would therefore not be surprising if these differences in the growth of a cancer model versus the natural disease were reflected in their vascular phenotypes and dependence upon VEGF. This notion is consistent with the observations that in human tumours and xenograft models, pericyte production of VEGF is associated with endothelial survival (Darland et al., 2003), and loss of VEGF results in selective endothelial cell death in those vessels lacking pericytes (Benjamin et al., 1999).

This idea also supports the clinical success of the recombinant humanized anti-VEGF-A antibody (Avastin; Genetech). Phase III clinical trials in colorectal cancer patients have show that when Avastin is combined with chemotherapy, there is a significant survival benefit. Avastin is thought to prune some tumor blood vessels; however, the major clinical benefit probably comes from normalizing tumor microvessels resulting in reduced interstitial fluid pressure and improved drug delivery (Tong et al., 2004; Jain, 2005).

Similarly, inhibition of PDGF-B may also derive a major therapeutic benefit through an unexpected mechanism. PDGF-B inhibitors might induce vessel destabilization through pericyte loss (Wilkinson-Berka et al., 2004) but may also lower interstitial fluid pressure (Pietras et al., 2002) resulting in increased drug delivery (Baranowska-Kortylewicz et al., 2005). Treatment of advanced ovarian or advanced colorectal patients with CDP860, a humanised PEGylated di-Fab PDGFRβ inhibiting antibody resulted in changes predictive of increased uptake of concomitantly administered drug. These, patients also developed ascites and a proportional increase in tumor perfusion (Jayson et al., 2005)—the latter possibly due to recruitment of previously collapsed intra-tumor blood vessels (Padera et al., 2004) through the lowering of interstitial fluid pressure. A possible mechanism for this effect has been demonstrated in mouse with defects in the PDGFRβ PI3K domain. Fibroblasts from these mice show reduced contraction of collagen gels due to reduced binding of collagen through the α2β1 integrin (Heulchel et al., 1999).

Summary

The last decade has seen a remarkable increase in the amount of interest in the field of angiogenesis. This has been principally lead by research into the biology of the endothelial cell, but more recently, attention has focused on the pericyte. This cell appears to be multifunctional, conserving a degree of biological plasticity that has served to hamper unambiguous identification of the cell during routine angiogenesis experiments. However, despite this, the role of the cell during health and disease is now becoming clear. They appear to act in concert with endothelial cells during angiogenesis and quiescence. Their association is so intimate that both require the presence of each other for their mutual survival. Together, they are able to satisfy the vascular requirements of the body during growth and development.

However, it is their role during pathological angiogenesis, and their pharmacological modulation that promises a new approach to treat cancer. Anti-vascular treatment with Avastin (Genentech) has already shown clinical benefit. Combining VEGF inhibitors with other approaches, such as PDGFRβ inhibitors, is likely to provide additional benefits. This may be achieved through the additional pruning of blood vessels that results from pericyte destabilisation, but also from the synergistic effects of lowered interstitial fluid pressure on drug delivery. On this positive outlook, one might speculate that the ascites noted with the humanized PEGylated di-Fab PDGFRβ inhibiting antibody (CDP860), may well be ameliorated by the decreased vascular permeability offered by VEGF/VEGFR inhibitors. If so, then the generation of very specific PDGF-B/PDGFRβ inhibitors, combined with VEGF/VEGFR inhibitors, might prove to be a very effective way forward in drug development.