Choriocapillaris Loss vs RPE Breakdown in Dry AMD

The more difficult question is temporal: which compartment begins to fail, which one amplifies the injury, and whether the sequence is stable across the different pathological forms grouped under dry AMD and geographic atrophy.

The tissue does not offer a clean answer. In complete geographic atrophy, regions of profound RPE loss can contain approximately 50% less choriocapillaris vascular area, with the remaining capillaries narrowed to an extreme degree. Yet histological sections from early AMD also reveal choriocapillaris dropout outside visible RPE degeneration. These empty vascular outlines—ghost vessels—are not simply the aftermath of an established atrophic lesion. They can appear while the overlying RPE remains structurally present.

That is the contradiction at the center of the choriocapillaris loss vs RPE degeneration AMD debate: the two tissues behave less like separate suspects and more like connected compartments in a progressive pathological cascade.

The ghost vessel phenomenon: where the capillary disappears but its outline remains

The choriocapillaris is a dense, specialized capillary bed positioned immediately beneath Bruch’s membrane and the RPE. Its geometry is not incidental. Photoreceptors maintain one of the highest metabolic demands in the body, while the RPE handles phagocytosis of shed outer segments, retinoid cycling, barrier maintenance, and the transport of metabolites between the neural retina and the vascular supply below.

When choriocapillaris endothelial cells are lost, the tissue does not always collapse into an obvious void. The capillary basement membrane may remain as an acellular tube. In donor retina and choroid preparations, these remnants are recognized as ghost vessels: a vascular architecture preserved after the cellular lining has disappeared.

The distinction between a surviving capillary and a ghost vessel depends on the marker used. Ulex europaeus agglutinin-I, or UEA-I, can label the vascular structures and expose the pattern of capillary loss, while endothelial markers such as CD31 help identify whether viable endothelial cells are still present. A vessel-like tube that retains structural material but lacks its endothelial population is not a functioning microvessel. It is the anatomical residue of a previous circulation.

This matters because a histological map of choriocapillaris density can underestimate the severity of functional vascular loss if the analysis treats every preserved tube as an intact vessel. The basement membrane may remain visible after perfusion capacity has already been removed from the tissue.

A ghost vessel is not a small vessel. It is the trace left when the vascular compartment has already lost its cellular identity.

Human donor eye studies using UEA-I labeling identified significant choriocapillaris loss and ghost vessels in early AMD, as well as in geographic atrophy tissue beyond the immediate zone of RPE degeneration. That distribution weakens any model in which the choriocapillaris fails only after the RPE has disappeared. It does not prove that vascular injury is always the first event. It does show that vascular pathology can precede, accompany, or extend beyond visible epithelial atrophy.

The pathology therefore has a spatial problem as well as a temporal one. The question is not only whether the RPE or choriocapillaris fails first. It is also whether the lesion begins in one compartment and spreads into the other, or whether both are damaged by a shared local environment involving Bruch’s membrane, drusen-associated stress, complement activity, altered extracellular matrix, and impaired cellular clearance.

RPE breakdown and choriocapillaris loss are anatomically coupled

RPE cells and choriocapillaris endothelial cells maintain a reciprocal relationship. The RPE supplies trophic and regulatory signals to the vascular bed, while the choriocapillaris provides oxygen, nutrients, and metabolic clearance for the outer retina and its supporting epithelium. Disturbance in either direction changes the conditions experienced by the other.

With RPE senescence, several pathological processes converge. The cells become less efficient at phagocytosing photoreceptor outer segments, their lysosomal processing becomes less reliable, and their ability to regulate inflammatory and vascular signals changes. Drusen biogenesis adds another layer: deposits accumulate between the RPE and Bruch’s membrane, altering diffusion, mechanical properties, and the local distribution of signaling molecules.

A capillary immediately beneath such a region is not operating in a neutral environment. The endothelial cells are separated from the RPE by a barrier whose composition and permeability have already changed. At the same time, the RPE is increasingly dependent on a vascular bed that may be losing density and constricting its remaining lumens.

This is why the phrase RPE versus choriocapillaris atrophy post mortem can be misleading. Post-mortem tissue provides a high-resolution record of structure, but it captures the endpoint or an intermediate state rather than a continuous cellular movie. A section can show RPE loss over a large area and capillary dropout beneath it, but the tissue alone cannot always establish which injury occurred weeks, months, or years earlier.

The morphology does, however, preserve clues:

• Ghost vessels outside complete RPE atrophy indicate that choriocapillaris injury is not confined to terminal epithelial lesions.
• Severe capillary narrowing beneath complete RPE atrophy is consistent with a later stage in which the loss of epithelial support and vascular demand reinforce one another.
• Drusen-associated vascular attenuation suggests that microvascular change can be spatially linked to deposits before the tissue reaches confluent geographic atrophy.
• The persistence of basement membrane tubes after endothelial loss separates structural remnants from active perfusion.

The result is a pathological sequence with feedback rather than a single linear chain. RPE stress can reduce support for the choriocapillaris; vascular insufficiency can increase metabolic stress on the RPE and photoreceptors; both processes can then accelerate local degeneration.

What extramacular drusen reveal about vascular dropout

The most useful evidence for an early vascular component comes from tissue that has not yet become a fully developed geographic atrophy lesion. In human donor maculae, choriocapillaris vascular density beneath extramacular drusen was 45% lower than in adjacent areas. The comparison is important because it places the vascular deficit beside, rather than only inside, the most visibly damaged atrophic zones.

Extramacular drusen are not merely passive deposits located away from the central lesion. They mark regions where the interface between RPE, Bruch’s membrane, and choroidal circulation has undergone pathological remodeling. When vascular density is reduced directly beneath these deposits, the finding raises a specific possibility: the choriocapillaris may be responding to local extracellular and epithelial abnormalities before complete RPE disappearance.

The analysis reported a statistically significant difference, with P < 0.01, between the vascular density beneath extramacular drusen and adjacent areas. That does not convert an association into proof of causation. A druse may damage the underlying interface, or it may form in a region already affected by impaired transport and altered vascular support. Both could also be consequences of a third process.

Still, the spatial relationship is difficult to dismiss as a terminal feature of atrophy. If choriocapillaris loss occurred exclusively as a secondary response to the final collapse of the RPE, the vascular deficit would be expected to remain tightly restricted to areas where epithelial degeneration was already advanced. Donor tissue instead shows a broader field of microvascular attenuation.

The pattern is particularly revealing when ghost vessels are quantified rather than described qualitatively. A linear relationship has been reported between ghost vessel count and choriocapillaris vascular density loss, with an r² of 0.55 and P < 0.001. In practical terms, more ghost vessels correspond to a greater reduction in the remaining vascular network. The relationship is not perfect—biological systems rarely grant that convenience—but it supports the interpretation that ghost vessels are part of a progressive vascular attrition process rather than isolated histological curiosities.

A comparison of the two pathological models

The table does not resolve the argument. It shows why the argument persists. Each model explains a portion of the tissue record, while neither explains all of it without borrowing mechanisms from the other.

The RPE-debridement model: a controlled sequence of secondary vascular decay

Experimental work in pigs provides a different kind of evidence. In these models, physical removal of RPE cells from the subretinal surface led to rapid choriocapillaris loss within two months. Photoreceptor loss followed progressively, accompanied by remodeling of the choroid toward a pachychoroid-like configuration.

This sequence is important because the initial insult is experimentally defined. The RPE is removed first; the vascular compartment then deteriorates. It demonstrates that RPE integrity is not merely correlated with choriocapillaris survival. The presence of RPE cells can be necessary for maintaining the local vascular phenotype.

The model also clarifies what it cannot prove. An experimentally induced RPE debridement injury is not equivalent to the heterogeneous sequence of events in human dry AMD. It does not reproduce the full history of drusen accumulation, age-related changes in Bruch’s membrane, complement-associated stress, genetic susceptibility, or the spatial gradients found in donor maculae. It establishes biological plausibility for one direction of causation, not universal priority.

The chronology nevertheless forms a coherent cascade:

1. RPE removal eliminates epithelial support. Signals that help maintain the choriocapillaris are abruptly withdrawn.

2. The choriocapillaris loses vascular integrity. Capillary density falls, and remaining vessels may constrict or regress.

3. Photoreceptors become progressively deprived. Their metabolic demand remains high while support from both the RPE and choroidal circulation deteriorates.

4. The choroid remodels. The tissue does not simply become smaller; its architecture changes in response to altered cellular and vascular conditions.

This sequence resembles the later stages of geographic atrophy, where extensive RPE loss coexists with severe choriocapillaris attenuation. Human donor tissue showed a reduction of up to 50% in choriocapillaris vascular area within regions of complete RPE atrophy, together with extreme constriction of viable capillaries. The experimental model offers a mechanism for how the vascular bed might deteriorate after epithelial collapse.

But the early human lesions remain the obstruction to a purely RPE-first interpretation. Ghost vessels beneath or beyond regions without complete RPE loss indicate that the choriocapillaris can be injured before the epithelium has vanished. The most defensible reading is not that one model defeats the other. It is that RPE loss can produce a powerful secondary vascular response, while earlier vascular abnormalities may already be developing in parallel.

CFH Tyr402His and the question of endothelial vulnerability

Genetic variation adds another layer to the tissue sequence. Eyes homozygous for the CFH Tyr402His risk allele showed significantly higher densities of ghost vessels in the choriocapillaris than eyes without the risk genotype.

CFH encodes complement factor H, a regulator of the alternative complement pathway. In the context of macular degeneration, the Tyr402His variant has long been associated with altered disease susceptibility, but donor eye histology gives the association a more localized anatomical form: the choriocapillaris itself may carry a greater burden of acellular vascular remnants in genetically susceptible tissue.

The finding does not identify the precise molecular trigger for endothelial cell loss. It does not show whether complement activation acts directly on endothelial cells, changes the RPE–choriocapillaris interface, modifies Bruch’s membrane, or amplifies an inflammatory environment in which several cell types become vulnerable at once. The pathway from genotype to ghost vessel remains incomplete.

It does, however, complicate any account of dry AMD that treats vascular dropout as an automatic consequence of RPE disappearance. If ghost vessel density is higher in tissue carrying two copies of the CFH Tyr402His risk allele, the vascular compartment may possess its own genetically conditioned susceptibility. The choriocapillaris is not simply waiting passively for the RPE to fail.

Microglial activation may also become relevant at this interface, although the donor findings described here do not establish a complete inflammatory mechanism. Microglia, RPE cells, endothelial cells, pericytes, and complement components occupy overlapping pathological territory. Their interactions could determine whether a local vascular defect remains limited, spreads beneath drusen, or becomes integrated into a larger atrophic field.

The distinction between association and mechanism is essential. Higher ghost vessel density in risk-genotype eyes is a clue about vulnerability. It is not a direct demonstration of endothelial apoptosis caused by one specific complement event.

The geography of geographic atrophy changes the argument

A complete GA lesion is the easiest place to observe the final convergence of degeneration. The RPE is absent or severely depleted, photoreceptors are lost, and the choriocapillaris is markedly attenuated. If analysis stops there, the tissue can appear to support a straightforward RPE-first model: remove the RPE, then watch the vascular bed disappear.

The margins and extra-lesional tissue are more informative. They contain partial damage, preserved cell populations, and the spatial traces of processes that have not yet reached their terminal state. Choriocapillaris dropout outside the zone of RPE degeneration suggests that the vascular field may be affected before the lesion becomes clinically or histologically complete.

This is where donor retina studies provide something imaging alone cannot. In vivo imaging can show reduced flow or altered vascular architecture, but post-mortem histology can distinguish viable vessels from basement membrane remnants and can place the vascular deficit beside drusen, RPE loss, and photoreceptor degeneration at cellular resolution.

The interpretation still depends on sampling. Human donor eyes are not longitudinal specimens from the same individual. A donor eye provides a snapshot, and the apparent sequence must be inferred from patterns across specimens with different degrees of pathology. That limitation does not make the evidence weak; it defines the type of question the evidence can answer.

Histology is strongest when asking:

• Where are viable endothelial cells absent?
• Where do ghost vessels persist?
• How does vascular density change beneath drusen and adjacent tissue?
• Is choriocapillaris loss confined to complete RPE atrophy?
• How do vascular and epithelial abnormalities overlap spatially?

It is less decisive when asking the exact order of molecular events within one individual lesion. For that, spatial transcriptomics, carefully matched donor tissue, and models that preserve the RPE–Bruch’s membrane–choriocapillaris interface will be necessary.

Toward a model of mutual dependence rather than a single primary insult

The most coherent interpretation of the current donor evidence is a reciprocal degeneration model. It begins with a local disturbance that may differ between lesions: RPE stress, altered Bruch’s membrane, drusen-associated signaling, complement susceptibility, endothelial vulnerability, or a combination of these factors. The initial disturbance then changes the environment of the neighboring compartment.

An RPE under chronic stress may provide weaker trophic support and alter the signals that regulate the choriocapillaris. A compromised choriocapillaris may reduce metabolic support and clearance, increasing the burden on the RPE and photoreceptors. Once both compartments are involved, the lesion no longer has a single direction. Each damaged layer becomes a source of stress for the other.

This model accounts for the major findings without forcing them into a false hierarchy:

• Complete RPE atrophy can be accompanied by approximately 50% loss of choriocapillaris vascular area.
• RPE removal in experimental tissue can lead to rapid vascular loss within two months.
• Ghost vessels can appear in early AMD before complete RPE degeneration.
• Choriocapillaris density beneath extramacular drusen can be 45% lower than in adjacent regions.
• CFH Tyr402His homozygosity is associated with higher ghost vessel density.
• Ghost vessel burden correlates with the degree of vascular density loss.

None of these observations alone identifies a universal first lesion. Together, they describe a system in which the RPE and choriocapillaris are mutually dependent, but not necessarily equally vulnerable at every stage or in every phenotype.

That distinction is more than semantic. If one subgroup of dry AMD begins with endothelial dysfunction while another begins with RPE senescence, a single treatment strategy aimed at a presumed primary tissue could miss the relevant initiating mechanism. The pathology may require stratification by lesion geography, drusen context, complement genotype, vascular phenotype, and the relative preservation of RPE and choriocapillaris.

The unresolved cellular question

The remaining mystery is not whether the choriocapillaris can disappear before complete RPE atrophy. Human donor histology shows that it can. Nor is it whether RPE loss can drive secondary capillary decay. Experimental RPE-debridement models show that sequence with unusual clarity.

The unresolved question is what turns an early vascular defect into a self-sustaining lesion. Which signal causes endothelial cells to die or retract, leaving behind ghost vessels? Does the initiating event involve complement-mediated injury, altered RPE trophic support, mechanical obstruction at the drusen–Bruch’s membrane interface, impaired pericyte function, or a metabolic failure that is shared by both compartments? And why do some areas with vascular attenuation progress to geographic atrophy while others remain structurally incomplete?

The tissue has recorded the aftermath, but the trigger remains partly concealed in the interval between a viable capillary and its acellular outline.

For now, the choriocapillaris loss vs RPE degeneration AMD debate should not be framed as a contest with one winner. The donor eye points to a sequence with branching paths: RPE breakdown can remove the support required for capillary survival, while choriocapillaris loss can emerge in advance of overt epithelial collapse and deepen the stress imposed on the outer retina. Geographic atrophy is therefore not exclusively an RPE disease or exclusively a vascular disease. It is the visible endpoint of an interdependent cellular failure whose first decisive event may vary from lesion to lesion—and whose earliest molecular signature has yet to be fully identified.