Human Choroid Transcriptomics: Fresh Donor Insights

The more useful question now is where those genes are active, which cell type carries the signal, and whether the observed pattern belongs to healthy tissue, advanced disease, or the tissue-handling process itself.

That distinction matters clinically. The choroid is not a uniform vascular sheet behind the retina. It contains endothelial beds with different molecular identities, stromal populations, immune-associated cells, and regional compartments that behave differently in the macula and periphery. A transcriptomic result that averages these structures together can conceal the biology a therapy is meant to reach.

Recent single-cell, single-nucleus, and spatial RNA-sequencing studies of postmortem human donor eyes have made that architecture more visible. One version of the human RPE and choroid atlas includes single-cell RNA sequencing data from 205,925 cells across 67 donors and single-nucleus RNA sequencing data from 742,625 nuclei across 54 donors. These numbers are not interchangeable, and neither method should be treated as a complete replacement for the other. Together, however, they provide a much more practical map of choroid molecular profiling than bulk tissue measurements alone.

Scaling the choroidal atlas: what large donor datasets actually add

Large datasets are valuable in ocular transcriptomics for a reason that is easy to miss: the human choroid is structurally heterogeneous before disease is considered. A small biopsy or a pooled tissue sample may contain a different balance of capillary, arterial, venous, stromal, and pigment-associated cells from the next sample. If those differences are not resolved, a molecular signal can be mistaken for disease biology when it is partly a sampling effect.

Single-cell RNA sequencing addresses this by separating dissociated cells according to their transcriptional profiles. The method can reveal distinct cell populations and their marker genes, but it also has physical costs. Dissociation can stress cells, alter some transcripts, and lose fragile or rare populations. It is therefore not accurate to describe a scRNA-seq atlas as a direct inventory of every cell that was present in the donor tissue.

Single-nucleus RNA sequencing works from nuclei rather than intact cells. That makes it useful for frozen or otherwise difficult-to-dissociate material, which is important in postmortem donor eye research. It also changes the biological readout: nuclear RNA is not simply a cleaner version of whole-cell RNA. The profiles reflect different aspects of gene expression and may be more or less informative depending on the cell type and the question being asked.

For translational work, the method is part of the result. A proposed target that appears consistently across single-cell and single-nucleus datasets is more persuasive than a signal found in only one preparation. That does not establish causality, but it helps separate a robust cell-type association from a technique-specific observation.

The large human RPE and choroid atlas identifies 15 major cell classes across the single-cell datasets. That scale allows researchers to examine the choroid as a set of interacting compartments rather than as a single layer. It also creates a practical reference for interpreting smaller studies: a new donor sample can be compared with known cell states, regional patterns, and disease-associated profiles rather than analyzed in isolation.

A choroidal transcriptome is only clinically useful when the tissue location, cell identity, disease stage, and preservation history remain visible in the result.

There is a logistical consequence. Donor eye procurement is not just a matter of obtaining tissue and sending it for sequencing. The value of the sample depends on how precisely its anatomical origin is recorded, how quickly it is processed, whether the relevant region is preserved, and whether the workflow supports the intended assay. A macular RPE-choroid sample, a peripheral choroidal punch, and a pooled posterior segment should not be treated as equivalent inputs.

Endothelial heterogeneity: the choriocapillaris is not interchangeable with larger vessels

The choroidal vascular tree contains distinct endothelial environments. Human donor eye transcriptomics has identified molecular differences among choriocapillaris, arterial, and venous endothelial cells. This is more than a classification exercise. A therapeutic strategy aimed at the outer retina or RPE must eventually interact with the vascular compartment that supports that region, and different endothelial populations may respond differently to inflammatory, angiogenic, or complement-related signals.

One particularly useful finding is the high and specific expression of the regulator of cell cycle gene, RGCC, in choriocapillaris endothelial cells. In the reported donor tissue analyses, RGCC was associated specifically with the choriocapillaris rather than being presented as a uniform marker across all choroidal endothelial beds. The same work linked its response to complement activation.

That combination makes RGCC interesting for human choroid gene expression profiling donor eyes, but it should be interpreted carefully. A choriocapillaris-associated marker is not automatically a disease biomarker, a drug target, or a predictor of treatment response. Its value depends on whether the signal is reproducible across donors, preserved across disease states, and connected to a measurable functional change.

For translational teams, the practical questions are straightforward:

• Is the RGCC signal detected in the relevant endothelial population, or has it been diluted by mixed RPE-choroid tissue?
• Does the signal remain visible in both single-cell and single-nucleus preparations?
• Is the donor cohort characterized well enough to distinguish healthy tissue from complement-activated disease tissue?
• Does the finding correspond to a pathway that can be measured in protein, imaging, or functional assays?
• Can the tissue workflow preserve the choriocapillaris region without excessive contamination from adjacent compartments?

These questions determine whether a marker can travel from an atlas into a therapeutic program. In a sequencing paper, cell-type specificity is an important result. In a clinical development program, it is the beginning of the validation process.

Why cell-type resolution changes therapeutic interpretation

Bulk RNA sequencing remains useful because it provides a broad measurement of tissue-level expression. It can identify regional differences and generate a stable overview across donor samples. But it averages signals from multiple populations. A gene may appear increased because endothelial cells express more of it, because the proportion of endothelial cells has changed, or because another population has been lost from the specimen.

Single-cell and single-nucleus methods help distinguish those possibilities, although neither eliminates them completely. If a gene is enriched in one cell class, the next step is to establish whether the enrichment reflects a stable identity marker, an activated state, a disease response, or a consequence of tissue processing.

The distinction becomes especially important for complement-associated biology. Complement activity may be relevant to retinal and choroidal disease, but the presence of a complement-responsive transcriptional signature in choriocapillaris endothelial cells does not by itself establish when or why that pathway becomes pathogenic. Postmortem tissue provides a molecular snapshot, not a continuous disease movie.

Spatial dynamics in macular neovascularization

Single-cell data can identify which populations carry a signal. Spatial RNA sequencing adds another layer: where the signal is located within the tissue.

In areas of macular neovascularization, endothelial cells show increased expression related to Rho family GTPase signaling and integrin signaling. VEGF and TGFB1 were identified as potential upstream regulators of these patterns. The importance of the finding lies in the anatomical context. The same pathway signal in a broad posterior segment sample would be less informative than a signal localized to endothelial cells within a neovascular lesion.

Rho family GTPase signaling is relevant to cell shape, movement, adhesion, and cytoskeletal organization. Integrin signaling also relates to how cells attach to and communicate with their surrounding matrix. In a neovascular environment, these pathways provide a plausible molecular framework for endothelial remodeling and altered interaction with the extracellular matrix. That does not mean every observed signal represents a single causal chain. It means the spatially localized transcriptional pattern is consistent with the physical behavior expected from actively remodeling vascular tissue.

The presence of VEGF and TGFB1 as potential upstream regulators adds therapeutic context without resolving it. Anti-VEGF treatment has already established the importance of VEGF biology in macular neovascular disease, but a spatial transcriptomic map may show that vascular behavior is also shaped by adhesion, matrix interaction, and tissue-state signals. A therapy that changes vessel permeability may not fully address the mechanisms supporting lesion organization or persistence.

This is where donor tissue findings need to be handled with discipline. Advanced macular neovascularization is not a molecular stand-in for every stage of age-related macular degeneration. The reported changes should not be generalized automatically to early dry AMD or to unaffected aging tissue. Disease stage is not a footnote in ocular multiomics; it determines what the tissue is allowed to tell us.

From spatial signal to experimental design

A useful donor eye study should preserve the distinction between lesion, adjacent tissue, and anatomically matched control region wherever possible. Otherwise, the strongest signal may be detectable only as a mixed average.

A practical comparison might include:

The table is not a hierarchy in which one specimen is always superior. It is a reminder that the assay must match the question. If the question concerns endothelial state within a lesion, a pooled posterior segment is poorly suited to answer it. If the question concerns broad regional differences across the choroid, bulk sequencing may provide a useful first pass before higher-resolution profiling.

Spatial methods also carry their own constraints. Tissue architecture must remain sufficiently intact, and the usable signal depends on section quality, RNA preservation, and anatomical annotation. A spatial map with uncertain lesion boundaries can create a false sense of precision. The image may be high resolution while the underlying tissue identity remains ambiguous.

Macula versus periphery: regional molecular gradients are part of normal biology

Bulk RNA sequencing across human donor eyes has shown clear regional molecular distinctions between macular and peripheral RPE-choroid. Endothelium-associated genes are upregulated in the macula compared with peripheral regions.

That result is clinically relevant because the macula is often treated as the primary reference point for vision-threatening disease, yet peripheral tissue is frequently used as a comparator. A peripheral sample may be easier to obtain or may appear less affected, but it is not biologically neutral. Regional differences exist before disease is layered onto the tissue.

For human choroid transcriptomics, this means that the phrase control tissue requires anatomical precision. A peripheral sample can help identify a regional gradient. It cannot automatically serve as an equivalent control for the macula. If a gene is more abundant in the macula under healthy conditions, its increase in macular disease cannot be interpreted without accounting for that baseline pattern.

The same issue applies to donor cohort design. Age, disease status, postmortem interval, tissue preservation, and sampling location can all influence the observed transcriptome. The exact impact of postmortem interval variation on cell-type-specific mRNA stability across all choroidal donor cohorts remains unresolved. That uncertainty should not invalidate postmortem research, but it should remain visible in the interpretation.

A strong study therefore documents more than a diagnosis label. It needs a tissue map. At minimum, researchers should know:

• whether the sample is macular, perimacular, or peripheral;
• whether RPE and choroid were analyzed together or separated;
• whether the tissue was processed for intact cells, nuclei, or bulk RNA;
• how disease-associated regions were identified;
• whether lesion and non-lesion areas were kept distinct;
• how donor-level variables may affect RNA quality and cell recovery.

The goal is not to eliminate every source of variation. That is unrealistic with human donor material. The goal is to identify which sources of variation are biological and which are introduced by the workflow.

Regional comparison is not a technical nuisance. In the human choroid, anatomy is part of the biology being measured.

What donor eye logistics determine before sequencing begins

The molecular assay receives most of the attention, but the study is often won or lost earlier. Human ocular tissue procurement has to preserve both the specimen and its context. A vial of RNA without reliable anatomical and donor metadata may still produce a sequence file, but the file will carry less translational value.

For choroidal studies, the workflow has several physical realities:

1. The intended compartment must be defined before collection.

RPE, choroid, retina, and sclera are adjacent but not interchangeable. Small differences in dissection can change the cellular composition of the final sample.

2. The macular region requires deliberate handling.

A sample described only as posterior pole tissue may combine biologically distinct areas. If macular neovascularization is the subject, lesion boundaries and neighboring tissue should be documented as precisely as the specimen allows.

3. Preservation should match the assay.

Intact-cell sequencing, single-nucleus sequencing, spatial RNA sequencing, and bulk RNA sequencing do not have identical tissue requirements. A preservation method that is acceptable for one may compromise another.

4. Cold-chain and processing time remain part of the data.

Postmortem interval can affect RNA integrity and cell recovery. The field does not yet have a universal correction that makes every donor comparable, so the interval must be tracked rather than hidden.

5. Cell loss can change the apparent biology.

Dissociation may underrepresent fragile or rare populations. A missing cell class should not automatically be read as biological absence.

6. Donor-level replication matters.

Tens of thousands of cells from one donor do not equal tens of thousands of independent donors. Large cell counts improve resolution within a specimen, while donor numbers improve confidence that the pattern is reproducible across people.

That last point is central to reading the atlas figures correctly. The reported 205,925 cells and 742,625 nuclei demonstrate substantial molecular resolution, but the units of biological replication remain the donors. A cell-rich dataset can characterize subpopulations in detail while still requiring careful donor-level analysis.

Technical constraints and the next phase of postmortem transcriptomics

The next stage of human choroid molecular profiling will likely depend less on simply increasing the number of sequenced reads and more on integrating modalities without losing tissue context.

Transcriptomics can identify cell states and pathway activity. Proteomics can help determine whether a transcriptional signal is reflected at the protein level. Imaging can localize structural changes. Functional assays can test whether a candidate pathway alters endothelial behavior, barrier properties, or interaction with the RPE. None of these layers is sufficient alone.

The same principle applies to disease genetics. A transcriptomic change near a glaucoma-associated or AMD-associated locus may be biologically interesting, but association does not establish the mechanism connecting a noncoding risk variant to a particular choroidal cell state. The definitive causal links between noncoding AMD risk variants and specific gene alterations in human choroidal stromal subpopulations remain unresolved.

That uncertainty is not a weakness of the field. It is the point at which discovery research becomes translational work. The atlas tells researchers where to look. It does not remove the need for validation.

Several developments would make donor-eye datasets more clinically useful:

• Matched regional sampling, with macular and peripheral tissue collected under a common protocol.
• Parallel single-cell and single-nucleus profiling, allowing method-related differences to be separated from biological differences.
• Spatial validation of candidate markers, particularly for endothelial subtypes and lesion-associated programs.
• Protein-level confirmation, including assessment of whether transcript abundance corresponds to measurable protein changes.
• Better postmortem metadata, including processing intervals, preservation conditions, and anatomical documentation.
• Longitudinal clinical linkage where available, so donor molecular profiles can be interpreted against documented disease course rather than diagnosis alone.

There is also a need for restraint when translating advanced lesions into treatment hypotheses. A pathway identified in macular neovascularization may suggest an intervention point, but the timing of intervention remains unknown. A signal detected after extensive remodeling may be a driver, a response, or a residual trace of an earlier process.

The clinical value lies in integration, not in the largest cell count

Human choroid transcriptomics has reached a useful level of anatomical and cellular resolution. We can distinguish endothelial compartments, identify choriocapillaris-associated expression such as RGCC, map lesion-associated signaling programs, and document meaningful molecular differences between macular and peripheral regions.

The harder work is now practical. Researchers must connect those findings to the way donor eyes are procured, dissected, preserved, and analyzed. They must also maintain the distinction between an atlas marker and a validated therapeutic target, between advanced disease and early disease, and between a large number of sequenced cells and genuine donor-level reproducibility.

That is the standard needed for translation. A useful choroidal profile should survive the journey from tissue room to sequencing platform, from sequencing platform to spatial validation, and from molecular association to a measurable functional or clinical outcome.

The promise of donor eye multiomics is therefore not that it produces a single definitive map. Its value is that it makes the map detailed enough to expose where the remaining uncertainty sits: in the endothelial subtype, the regional anatomy, the disease stage, the preservation workflow, or the missing functional experiment. That is a more grounded basis for therapeutic development—and a much safer one for the patients who may eventually depend on it.