Introduction

During the American Civil War, it is popularly believed that the United States troops (or ‘the Union’) wore dark blue uniforms whereas the secessionist Confederate States troops wore grey. This was generally the case. However, since both armies included assorted militias from the member states, numerous exceptions existed in which Southern troops wore blue and Northern units wore grey. Added to the smoke, noise, and confusion of battle, the inability of the respective commanders to unambiguously determine the affiliations and loyalties of approaching troops could lead to disastrous consequences. One example of the chaos resulting from this lack of certainty about the intentions of uniformed participants was seen in the Battle of Wilson's Creek in Missouri. The Union army, holding territory soon to be named Bloody Hill, mistakenly identified approaching blue-coated Confederate troops as friendly Union reinforcements, allowing them to march within a few yards before the Confederates opened fire and the Union troops were routed.

In the case of neovascular age-related macular degeneration (AMD), there is ambiguity surrounding the role of macrophages in the disease process, with conflicting evidence regarding whether they are helpful or harmful. An improved understanding of whether these cells promote macular disease or are beneficial will be essential for scientists and clinicians interested in understanding and treating AMD. In this review, we will discuss evidence that macrophages participate in exudative and non-exudative AMD, some of the relevant clinical studies, data from post-mortem human tissue, and interventional experiments using animal models. We will also suggest approaches to further clarify the roles of macrophages in AMD.

Age-related macular degeneration

Age-related macular degeneration is the leading cause of irreversible blindness in the Western world.1, 2, 3 In the United States alone, nearly two million individuals are afflicted with severe, end-stage AMD, and are legally blind. Seven million more have early stage AMD, or age-related maculopathy, and are at high risk for developing advanced AMD.4 As the population ages, the negative impact of AMD on society will increase.4, 5

Defining ‘macrophages’

For the purposes of this review, ‘macrophage’ will be defined as haematopoietic cells that are derived from haemocytoblasts in bone marrow and that travel through the systemic circulation as monocytes (Figure 1). These cells characteristically express the cluster differentiation (CD) antigens that include CD14, CD11b, CD18, and CD68, and perform roles in phagocytosis, scavenging debris, antigen presentation, and secretion of cytokines. Numerous classes of macrophages are recognized, with diversity that includes multinucleated osteoclasts in bone and alveolar macrophages that reside at the lung-air interface. In the eye, circulating monocytic cells exit the vasculature through endothelial cells in the choroid or retina during normal development, ageing, and disease. In the choroid, the distribution of resident cells is best appreciated by choroidal whole mounts that show these cells to be fairly evenly spaced in dense networks.6 Choroidal cells of monocytic origin include circulating monocytes, resident macrophages, and dendritic cells. Upon exiting the vasculature, monocytes can differentiate into either macrophages or dendritic cells depending on microenvironmental cues (Figure 1).

Figure 1
figure 1

Simplified model of leucocyte differentiation. Bone marrow haemocytoblasts (HB), which may already be specialized in some cases, give rise to progenitor cells including lymphoblasts (LB), progranulocytes (PG), and monoblasts (MB). These cells then differentiate into lymphocytes (LC) or arguably lymphoid dendritic cells (LDC), different classes of granulocytes (GC), and monocytes (MC), respectively. Circulating monocytes leave the vasculature and differentiate into myeloid dendritic cells (MDC) and macrophages (MΦ) based on microenvironmental cues.

Full size image

Association of inflammation and macrophages with choroidal neovascularization: clinical and genetic studies

Several clinical studies have been published that provide information about the possible role of inflammatory cells in choroidal neovascularization (CNV). In an early phenotype association study, Blumenkranz et al7 identified elevated white blood cell counts as one of the few systemic indicators correlating with neovascular AMD. Similarly, a recent report on the Blue Mountain cohort in Australia also found increased white blood cell counts associated with early AMD, drusen, and possibly with geographic atrophy, although there was not a significant increase in patients with CNV.8 Moreover, there is evidence of increased activation of circulating monocytes in patients with CNV.9 In addition to studies of circulating cells, biochemical studies from CNV patients also have shown levels of C-reactive protein and the cytokine interleukin (IL)-6 to be elevated in the serum in association with progression to CNV.10, 11 These experiments are intriguing as they point to systemic inflammatory factors, in addition to local ocular events, in the pathogenesis of AMD.

Anti-inflammatory treatments that affect macrophages may also be beneficial to AMD patients. In several studies of CNV patients, intravitreal injection of the anti-inflammatory steroid triamcinolone has been shown to be beneficial,12 either alone or in conjunction with additional treatment modalities (eg, photodynamic therapy).13 Administration of triamcinolone by a synthetic implant was also found to be protective against experimental CNV in a rat model.14 It should be noted however that triamcinolone likely affects multiple intercellular pathways in addition to inflammation, and the modulation of these pathways (such as apoptosis) may be as important or more important than the impact of this steroid on inflammation.

In the last few years, there has been tremendous progress in identifying genetic risk factors for AMD (recently reviewed15, 16, 17). Notably, several AMD-associated risk alleles are located in genes whose products participate in inflammation, either directly (CX3CR1, MHC alleles)18, 19 or indirectly (complement factor H, complement factor 2, complement factor B).20, 21, 22, 23 Although the biochemical details of how each of these alleles affects protein function remain to be determined, these genetic associations contribute to the notion that inflammation has a role in the pathogenesis of AMD.

Evidence of inflammation and macrophage participation in CNV: histopathologic studies

There is indirect evidence that macrophages may participate in both early AMD and in advanced, exudative AMD. Macular drusen represent important indicators of early AMD, and it is widely appreciated that the number, size, and confluency of drusen in the macula is a major risk factor for atrophic and/or neovascular changes in the ageing macula (eg,24, 25, 26). Studies examining the composition of drusen revealed that they are largely comprised of numerous mediators of inflammation.27, 28, 29

Macrophages have been shown to be localized to degraded areas of Bruch's membrane in eyes with advanced AMD30, 31 and are found in increased numbers within the choroid of eyes with AMD.32 Human CNV membranes have been shown to contain leucocytes of the macrophage lineage.33 These cells are concentrated near pathologic vascularization as found by transmission electron microscopy34 and CD68 immunohistochemistry.35 Moreover, macrophages within CNVM have been shown to express pro-angiogenic cytokines, including vascular endothelial growth factor (VEGF), indicating that macrophages might directly promote aberrant endothelial cell growth in CNV.36

Figure 2 depicts immunohistochemical labelling of CD45+ leucocytes in an unaffected eye (top) and in an eye with a CNV membrane (bottom).

Figure 2
figure 2

Leucocyte localization in human eyes. Leucocytes localized in the normal human choroid of an 84-year-old donor (top) and in an eye from an 80-year-old donor with choroidal neovascularization (bottom). Sections were labelled with antibodies directed against leucocyte common antigen, a protein expressed in all classes of leucocytes (CD45; green labelling) and with the lectin Ulex europeaus agglutinin-I to visualize the vasculature (red labelling).85 Nuclei are counterstained blue with DAPI. Yellow-orange fluorescence is due to RPE lipofuscin. Choroidal leucocytes are indicated by arrowheads, leucocytes within the CNVM are indicated by arrows, and the CNVM is indicated by the bracket. GCL, ganglion cell layer; INL, inner nuclear layer; ONL, outer nuclear layer; CHO, choroid. Scale bar=100 μm.

Full size image

How might macrophages promote macular degeneration?

One of the more surprising findings made in studies of drusen composition is the presence of MHC class II antigens in these deposits.19, 27, 37 These transmembrane proteins (that are typically expressed by antigen-presenting cells) are located in core-like structures and are often dispersed throughout drusen.19, 38 Although the possibility that these cells are removing drusen cannot be ruled out (see below), the dispersion of these antigens throughout the deposits may suggest that cells of monocytic origin are involved in the pathogenesis of drusen formation.37 Second, in order for subretinal neovascularization to occur, it is necessary for endothelial cells to migrate through defects in Bruch's membrane. In the experimental CNV mouse model (discussed below), these defects are created by laser injury; in AMD, the precipitating event is not known. The dissolution of Bruch's membrane, which is largely comprised of elastin and collagen, likely relies on matrix metalloproteinases. Macrophages and neutrophils play significant roles in extracellular matrix turnover and express a number of matrix metalloproteinases that could reasonably lead to Bruch's membrane breakdown. Third, leucocyte transmigration through the endothelium can in itself promote endothelial cell injury and death in other tissues.39 Such a mechanism could place the macular, as compared to peripheral, choriocapillaris at greater risk for injury as the macular choriocapillaris tends to exhibit higher expression of intercellular adhesion molecule-140 and might therefore subject the macular endothelium to increased monocyte transmigration. Hypoperfusion of the choriocapillaris resulting from transmigration-induced endothelial injury is a plausible mechanism for exudative conversion in AMD, especially in eyes with high risk of developing CNV in which perfusion has been shown to be decreased.41 Finally, macrophages can secrete growth factors, cytokines, and reactive oxygen species42 that may negatively affect the choriocapillaris and RPE.

It is further worth considering that two areas of great recent interest in AMD—the complement system and anti-VEGF therapy—are likely to modulate the function of circulating leucocytes in AMD. For example, anaphylotoxins generated during the formation of complement complexes have a number of bioactive effects on leucocytes, including increased migration and synthesis of pro-inflammatory mediators.43 It is also interesting, in light of the success of anti-VEGF therapies in the management of neovascular AMD, that VEGF itself increases expression of endothelial cell adhesion molecules and elevates recruitment of macrophages and neutrophils under some conditions, in addition to its well-appreciated direct effects on vascular growth.44, 45

How might macrophages protect against macular degeneration?

In addition to drusen, which are clinically apparent, the presence of membranous debris and basal linear deposit in Bruch's membrane represent morphological signs of AMD.46, 47, 48, 49, 50 It has been suggested that macrophages play a role in removal of drusen and other waste products, perhaps preventing the severe end-stage pathology that can occur after the accumulation of these deposits.51, 52 Monocyte chemotactic protein-1 (MCP-1/CCL2) is a chemokine that directs monocyte recruitment into inflamed tissues.53 Mice deficient for this protein, especially when also deficient for the chemokine receptor CX3CR1, were reported to exhibit drusen-like deposits, suggesting that impaired macrophage recruitment results in reduced clearance of debris from Bruch's membrane.54, 55 Ageing human Bruch's membrane accumulates advanced glycation endproducts (AGEs),56, 57 which can induce increased secretion of VEGF in cultured RPE cells.58 As macrophages can endocytose and remove AGEs,59, 60 a failure to recruit macrophages could thus lead to increased AGE exposure to the RPE and choriocapillaris, with concomitant injury and/or increased levels of proangiogenic cytokines. The removal of apoptotic cell debris by macrophages can also limit inflammation through inhibiting the production of inflammatory cytokines.61 These proposed functions are depicted in Figure 3.

Figure 3
figure 3

Potential harmful (left) and beneficial (right) roles of macrophages in the progression of neovascular AMD.

Full size image

Functional studies of macrophages in CNV and their limitations

As discussed in the previous sections, there is compelling histological evidence for a spatial association between immune cells (monocytes or macrophages) and human CNV membranes. Of course, the presence of macrophages and other immune cells at the site of a lesion does not in itself show whether these cells are responsible for inducing the lesion or are somehow mitigating its severity. Studies that rely on human donor tissue, while the most reliable in terms of observing actual human pathology, are limited in that only a single ‘snapshot’ of the disease process is surveyed (although it is hoped that sufficient numbers of static images from different stages of disease can produce a more dynamic and complete understanding). Animal models of CNV, in which genetic and environmental parameters can be altered, play a complementary role to human-tissue-based studies and can shed further light on the role of macrophages in CNV.

The most commonly employed animal model of CNV relies upon thermal laser-induced breaks in Bruch's membrane, which allow subsequent subretinal neovascularization at the site of injury. The use of thermal laser to induce CNV was initially described in primates62 and subsequently in rodents.63 This type of experiment has been of particular value in mice, in which the effects of different interventions and genetic backgrounds on CNV can be readily controlled. Although the events leading to human CNV are clearly different from those in laser-induced trauma, the subsequent steps of angiogenesis that occur in human CNV are likely to be reflected in the laser injury model.

The use of thermal laser to induce CNV has been used in a number of elegant studies in which leucocyte function was inhibited in mice. Two studies published in 2003 evaluated the effects of monocyte depletion on laser-induced CNV. Espinosa-Heidman et al64 pre-treated mice with clodronate liposomes prior to inducing CNV. Clodronate promotes apoptosis of monocytes and macrophages that phagocytose the liposomes, leading to depletion of both circulating and resident cell populations.65 CNV severity—as assessed by lesion size—was significantly decreased in mice that were injected with clodronate liposomes, compared with controls, following CNV induction. In a simultaneously published report, Sakurai et al66 also utilized the clodronate liposome method. In these studies, the volume of CNVMs was significantly reduced in animals treated with clodronate. Moreover, morphological studies were interpreted as showing macrophages (CD45+, F4/80+ cells) preceding endothelial cells into areas of developing CNV. Significantly, the levels of vascular endothelial growth factor were decreased in clodronate-treated animals, suggesting that macrophages are contributing to VEGF synthesis, as suggested previously in human CNVMs.36

Mice that are genetically deficient in macrophage recruitment or signalling show similarly decreased severity of CNV as mice that are pharmacologically depleted of macrophages. Knockout mice lacking Ccr2, the gene encoding the receptor for MCP-1, have severely impaired ability to recruit monocytes to areas of inflammation. When Bruch's membrane of these animals is ruptured with thermal laser, they were found to produce much smaller neovascular membranes than wild-type mice.67 Mice deficient for the endothelial surface molecule intercellular adhesion molecule-1 or one of its binding partners, CD18, also demonstrated impaired neovascularization following laser treatment.68 Thus, both pharmacologic and genetic impairment of monocytes appear protective against CNV in several studies.

In addition to monocytes/macrophages, the potential role of polymorphonuclear leucocytes has also been explored, with evidence for a modest contribution of these cells to experimental CNV,69, 70 although the significance of the observed effects on CNV size has differed between studies.

Taken together, there is considerable evidence for macrophage participation in the pathogenesis of CNV in the mouse model and, perhaps, human eyes with AMD.

The proposal that these cells exacerbate CNV has recently been challenged, however, by a multifaceted study in which Apte et al71 provided evidence that macrophages inhibit the development of experimental CNV. These investigators also employed the laser-induced model of CNV, in this case with mice deficient for IL-10, to determine the effects of macrophage suppression on subsequent CNVM formation. IL-10 is a cytokine that is generally considered anti-inflammatory with a role in inhibiting the function of macrophages. One would hypothesize that, based on the proposed role of macrophages in promoting CNV, the severity of CNV would increase in the absence of IL-10. The opposite result was actually observed—with IL-10 knocked out, or its effective concentrations reduced by the addition of exogenous anti-IL-10 antibody, mice developed less severe CNV. This effect was correlated with the extent of macrophage (or CD11b and F4/80+ cells) migration into the lesion site. To determine if the converse was true, transgenic mice overexpressing IL-10 were generated, and these animals exhibited increased susceptibility to CNV in comparison to the wild-type littermates. In addition, Apte et al found that monocytic cells injected into the vitreous, although not reflective of AMD pathophysiology, also prevented CNVMs from developing as significantly as in eyes injected with other cell types, through a mechanism likely to involve CD95 signalling. Overall, this study provided significant evidence from both positive and negative approaches that macrophages may mitigate the severity of CNV in the experimental injury model, challenging several previous studies.

In attempting to reconcile these disparate findings, and better define the role of macrophages in CNV, a few considerations are outlined below.

What are the effects of macrophage depletion in the absence of laser?

Each of these studies utilized mice with genetic or pharmacologic interventions to affect macrophage number or function, followed by treatment with thermal laser and evaluation of CNV severity. Surprisingly, the possible effects of these treatments on non-lasered eyes have not been well established. It is possible, as suggested, that clodronate liposome treatment or other modalities may lead to death or injury of the endothelium, in addition to depletion of circulating monocytes.71 This is an important consideration; clearly a modality that kills or inhibits the proliferation of the choroidal endothelium would be expected to reduce the severity of CNVMs, and its simultaneous effects on macrophages would be interesting but irrelevant. Whereas animals exposed to clodronate liposomes in other studies are viable,65 and this delivery does not appear to cause widespread vascular dropout and tissue necrosis, it is feasible that the RPE and choriocapillaris are especially sensitive to this drug. This question might be addressed by clodronate treatment, followed by reconstitution of monocytes into the circulation; if untreated monocytes are unable to restore a severe CNV phenotype, this result would call into question whether clodronate's most important biological effect in this experiment is through monocyte depletion.

Similarly, with respect to IL-10, it will be important to establish whether this cytokine affects the RPE-Bruch's membrane-choriocapillaris complex in vivo. Whereas Apte et al show that overexpression of IL-10 does not have a severe effect on the overall structure of the retina, detailed, high resolution quantitative studies will be necessary to explore whether there are subtle changes at the level of the choriocapillaris and/or Bruch's membrane in IL-10-deficient mice, IL-10 transgenic mice, and clodronate-liposome-treated mice. Careful immunohistochemical and ultrastructural studies of the choroidal vasculature, Bruch's membrane and RPE in animals with each of these treatment regimens and genotypes, such as those used previously to evaluate the effects of cigarette smoke on the RPE-choroid complex,72 are especially warranted.

Although IL-10 is generally considered an anti-inflammatory cytokine, it appears to have pleiotropic effects, with either immunosuppressive or immunostimulatory functions, depending on its microenvironment, the type of stimulus, and the model system employed.73 One effect observed in vitro is endothelial cell activation and upregulation of the endothelial cell adhesion molecule E-selectin,74 and some human patients receiving IL-10 showed an increase in plasma levels of some pro-inflammatory cytokines.75 There may also be relevant extracellular matrix changes associated with loss of IL-10; this cytokine downregulates collagen I synthesis in vitro76 and may therefore affect the kinetics of scar formation, as noted in a recent editorial.77 Moreover, collagen I is a component of Bruch's membrane,78 and changes in the ECM composition of Bruch's membrane may lead to altered propensity to form neovascular membranes. Therefore, while at face value each of these studies approaches the same question with regard to macrophage depletion, the biological impact of IL-10 administration, IL-10 deficiency, and clodronate treatment may be complex.

What are the effects of macrophage depletion in laser-independent models of CNV?

The laser-induced CNV model is highly tractable, with a predictable time course and high degree of control by the investigator; however, the injury created in this model (in which a normal, healthy Bruch's membrane is suddenly disrupted) differs substantially from the pathology in human CNV, which takes years to develop. Transgenic mice overexpressing VEGF exhibit intrachoroidal neovascularization, in a similar pattern to that described in human diabetic choroid,79 but have not been consistently found to develop subretinal neovascularization.80 Recently, alternative mouse models of sub-RPE deposits and spontaneous CNV have been described. Mice transgenic for the human apolipoprotein E epsilon 4 allele, when placed on a high-fat diet, developed sub-RPE lesions as well as choroidal (and retinal) neovascularization. In addition, mice deficient for superoxide dismutase-181 were also found to have focal drusen-like deposits, although these were less than the height of an RPE cell, and diffuse basal deposits. Spontaneous CNV was observed in these animals as well. As noted above, mice deficient for the genes encoding both MCP-1 and CX3CR1 also develop sub-RPE deposits and also develop spontaneous CNV.55 These models, and those that will be developed from genetic progress in understanding AMD, may offer new possibilities to evaluate the role of macrophages in CNV. Specifically, the effects of modifying the number and function of different classes of macrophages may be explored in what may be more medically relevant models of neovascularization.

Do different macrophage classes have different effects on CNV?

There has been considerable progress in identifying and characterizing phenotypically different classes of monocytes and macrophages in human82 and mouse,83 (recently reviewed84). These distinct classes of monocytes appear to have different functional properties and varying propensities to differentiate into macrophages and dendritic cells in response to different microenvironmental conditions. The murine studies to date have described macrophage populations in experimental CNVMs as F4/80+, CD11b+, and/or CD45+ cells. These markers appear to label all classes of mouse monocytes and macrophages, and do not discriminate between different subsets. Hence, it is possible that the increased transmigration of macrophages in IL-10-deficient mice may represent a different and more benign population of cells than those seen in other genotypes and associated with more severe CNV. Further studies using additional markers, such as CD11c and CD62L84 may further refine whether the ‘macrophages’ in each of these studies are truly the same cells.

Conclusions

In summary, despite several excellent studies to date, there is a lack of consensus on whether macrophages act to reduce CNV, exacerbate CNV, have mixed effects depending on genotype and environmental milieu, or are incidental bystanders in the process. Subtle, but important, differences in the experimental approaches taken by different groups make comparison of these experiments difficult. Understanding the role of circulating inflammatory cells in the development and progression of AMD is of clear importance regarding additional therapeutic approaches for this devastating disease. Future studies will require determining the effects of clodronate liposomes and of IL-10, CD18, and ICAM-1 deficiencies alone on the biology of the RPE-Bruch's membrane-choriocapillaris complex. If possible, separating the anti-inflammatory from the anti-angiogenic activities of IL-10 will be helpful, especially if these are found to be mediated by different receptors. In addition, it will be important to demonstrate that a severe CNV phenotype can be ‘rescued’ in mice deficient for CD18 or treated with clodronate liposomes by transfusion of healthy monocytes (of various subpopulations) into treated animals.

On the basis of preponderance of data, we propose that it is most likely that macrophages contribute to the transition to neovascular AMD. It is especially compelling that observations from human eyes in which macrophages have consistently been observed at the scene of the crime (within CNVMs) and holding a smoking revolver (or at least expressing VEGF). This suggests that inhibition of their activity would be helpful for patients.

However, there are alternative explanations for the findings in human eyes, and studies in mouse models suggest that the picture may be complicated. A more thorough understanding of the role of leucocytes in AMD will be of significant benefit. One can easily envision a situation in which some classes of leucocytes promote CNV. In this case, it would be beneficial to interfere with their migration, signalling, VEGF secretion, and other functions in an eye with high risk of developing CNVM (for example, in an eye with soft macular drusen, a high-risk genotype, and a CNVM in the contralateral eye). However, if some classes of leucocytes in the better eye are slowing the course of CNV, then inhibiting these cells would be counterproductive and harmful. Further refining the harmful and potential helpful roles of these cells in CNV will be beneficial in guiding the development of new therapies and in applying existing therapies that act on leucocytes and inflammation.