Mitochondrial protein expression in donor retina: current findings

A protein detected in an organelle-enriched retinal pigment epithelium fraction does not represent the proteome of the intact retina. A signal localized to a retinal ganglion cell axon does not describe the same biological system as a signal measured in the soma.

This distinction is operational, not semantic. Current donor-eye proteomics can quantify thousands of proteins and identify consistent mitochondrial pathway shifts in early age-related macular degeneration. It cannot yet eliminate the pre-analytical variables that determine whether a mitochondrial signal reflects disease biology, tissue degradation, or a change in the sampled compartment.

The most substantial recent dataset quantified 5,941 proteins in organelle-enriched RPE fractions from 45 donors with early AMD and 32 age-matched healthy controls using the UHR-IonStar mass spectrometry platform. The result is a high-density molecular map. It is also a map with defined boundaries: RPE-focused, fractionation-dependent, and derived from post-mortem human tissue.

Quantitative proteomic landscapes of the human RPE

The RPE is a high-value compartment for mitochondrial analysis because it sits at the interface between photoreceptor maintenance, lipid handling, oxidative metabolism, and outer-retinal homeostasis. Its mitochondrial burden is substantial. Its molecular output is correspondingly sensitive to disruption in electron transport, ATP production, protein translation, and oxidative stress control.

The 2023 quantitative proteomic analysis of donor RPE provides a useful scale reference. The study did not measure a narrow panel of candidate proteins. It quantified 5,941 proteins across organelle-enriched RPE fractions. That level of coverage permits pathway-level interpretation rather than isolated marker tracking.

The central finding was mitochondrial dysregulation in early AMD. The observed changes involved several connected functional systems:

• Mitochondrial electron transport chain subunits, which participate in respiratory energy generation.
• ATP synthase complexes, which support oxidative phosphorylation and ATP production.
• Mitochondrial translation machinery, required for the synthesis and maintenance of mitochondrially encoded respiratory components.
• Lipid homeostasis pathways, relevant to the RPE’s handling of photoreceptor-derived material and membrane turnover.
• Oxidative stress response proteins, which regulate the consequences of elevated respiratory and metabolic load.

This pattern is more informative than a claim that one mitochondrial protein is increased or decreased. Mitochondrial dysfunction in retinal degeneration is generally distributed across linked processes. A change in ATP synthase abundance without corresponding changes in translation, lipid handling, or oxidative stress pathways would support a narrower interpretation. The donor RPE data instead point to a coordinated disturbance in mitochondrial and metabolic infrastructure.

The cohort structure is also important. The comparison included 45 early AMD donors and 32 healthy controls. The disease signal was therefore not restricted to advanced anatomical destruction. Mitochondrial pathway dysregulation was detectable in early AMD donor tissue. That shifts the analytical use of proteomics upstream. The data are relevant to molecular staging, not only to end-stage pathology description.

The useful unit of interpretation is not an isolated mitochondrial protein. It is the relationship between respiratory machinery, translation capacity, lipid processing, and oxidative stress response within a defined donor-tissue fraction.

Why protein count is not equivalent to biological resolution

A large quantified-protein total improves discovery capacity. It does not automatically provide cell-type resolution. An organelle-enriched RPE preparation can contain a highly informative mitochondrial signal while still averaging across multiple cellular states and subcellular structures.

This creates several analytical constraints.

First, enrichment changes the denominator. The measured abundance of a mitochondrial protein is interpreted within the isolated fraction, not against the entire intact retina. Second, enrichment does not produce perfect purity. Residual material from other organelles or cellular structures can influence the final profile. Third, donor tissue is heterogeneous by design. Age, disease stage, retinal region, tissue integrity, and post-mortem interval can alter the molecular output before mass spectrometry begins.

The correct formulation is therefore specific: mitochondrial protein expression was profiled in organelle-enriched donor RPE fractions. It is not equivalent to a whole-retina mitochondrial proteome. It is also not equivalent to a single-cell proteomic map of RPE mitochondria.

For procurement and biobanking workflows, this distinction affects metadata requirements. A proteomic dataset without a clear record of sampled region, fractionation method, donor phenotype, grading system, and tissue handling has limited interoperability. The raw protein list may be large. The usable biological context may still be incomplete.

Mitochondrial subproteome shifts across Minnesota Grading stages

Earlier mitochondrial subproteome work used donor eyes classified across Minnesota Grading System stages 1–4. This grading framework allows molecular measurements to be aligned with progressive AMD-associated tissue changes rather than treated as a binary healthy-versus-diseased comparison.

The analysis identified statistically significant density changes in eight protein spots across the MGS stages. The affected proteins included:

• Alpha, beta, and delta subunits of ATP synthase.
• Subunit VIb of cytochrome c oxidase.
• Mitofilin.
• Mitochondrial Hsp70.
• Mitochondrial translation factor Tu.

The list covers several mitochondrial operating layers. ATP synthase subunits relate to energy conversion. Cytochrome c oxidase is part of the respiratory chain. Mitofilin is associated with mitochondrial inner-membrane organization. Mitochondrial Hsp70 supports protein handling and folding. Translation factor Tu participates in mitochondrial protein synthesis.

The significance is architectural. The changes are not confined to one pathway label or one functional class. They extend across energy production, inner-membrane organization, protein quality control, and mitochondrial translation. That pattern is compatible with progressive stress on the mitochondrial system as donor eyes move through grading stages.

It would be incorrect to treat these eight protein spots as a complete representation of mitochondrial dysfunction. The earlier platform and spot-based design provide targeted differential information, not exhaustive proteome coverage. The later UHR-IonStar study expands the measurable protein landscape substantially. The two approaches are complementary. One emphasizes stage-associated subproteome shifts. The other provides broader quantitative coverage in early AMD.

A practical comparison is useful:

What stage-associated shifts can and cannot establish

MGS-linked changes establish association between mitochondrial protein patterns and donor-eye grading stage. They do not, by themselves, establish causal sequence. The data cannot determine whether a specific protein shift initiates retinal degeneration, compensates for it, or results from accumulated cellular stress.

They also cannot be generalized across every retinal cell population. RPE mitochondria, photoreceptor mitochondria, Müller glial mitochondria, and RGC mitochondria operate under different structural and metabolic conditions. Their proteomes should not be merged into a single undifferentiated mitochondrial signature.

This is where multiomics integration becomes necessary. Proteomic evidence can identify altered protein abundance and pathway organization. Transcriptomic data can test whether the protein pattern is accompanied by altered gene expression. Genomic data can identify inherited or somatic risk architecture. Epigenetic profiling can add regulatory context. None of these layers replaces the others.

For donor-eye research, the practical objective is not to maximize the number of assays per specimen. It is to preserve enough tissue and metadata to connect assays across the same anatomical and disease coordinates. A high-throughput proteome generated from poorly annotated tissue has lower translational value than a smaller dataset with reliable sampling provenance.

Spatial compartmentalization of retinal mitochondrial proteins

Mitochondrial proteomics becomes more precise when location is retained. Retinal ganglion cells provide a clear example. Their somatodendritic and axonal compartments have different transport requirements, structural roles, and energetic loads. A single cell-level average can conceal this division.

Spatial proteomic validation demonstrated compartmentalization of mitochondrial proteins in RGCs. Mitochondrial superoxide dismutase, or SOD2, was enriched in the somatodendritic compartment. The alpha subunit of the mitochondrial trifunctional protein, HADHA, was enriched in the axonal compartment.

This is not a minor localization detail. SOD2 is associated with mitochondrial antioxidant defense. HADHA participates in fatty-acid oxidation through the mitochondrial trifunctional enzyme complex. Their different distributions indicate that mitochondrial protein expression is organized around local functional demand.

The implication for optic neuropathy bioinformatics is direct. A bulk RGC proteome may identify mitochondrial dysregulation but fail to distinguish whether the primary signal is concentrated in the soma, dendrites, or axon. That distinction can affect interpretation of transport failure, oxidative stress, and compartment-specific degeneration.

The same logic applies to donor-eye tissue more broadly:

1. Define the anatomical region before molecular extraction. Macula, peripheral retina, optic nerve head, and choroid are not interchangeable sources.

2. Define the cellular or subcellular fraction. RPE, retina, RGCs, mitochondria, and organelle-enriched material generate different denominators.

3. Preserve spatial information where the research question depends on localization. Bulk extraction is efficient but destructive to compartment identity.

4. Link proteomic results to disease grade and donor metadata. A spatial signal without phenotype context remains difficult to interpret.

5. Report the measured compartment in every biological conclusion. This prevents a local observation from being promoted into a whole-retina claim.

Spatial context is part of the measurement. Remove it too early, and the proteome retains abundance but loses mechanism.

Relevance to single-cell and spatial multiomics

Human ocular single-cell RNA-seq has expanded the ability to identify cell-type-specific transcriptional states. Proteomics remains more difficult at equivalent resolution because protein extraction, peptide identification, and quantitative depth impose stricter material requirements.

The current donor-eye landscape therefore has an asymmetry. Transcriptomic atlases can provide broad cell-type maps. Proteomic studies can provide direct information on protein abundance and functional machinery, but often at the level of tissue fractions, organelle-enriched preparations, or selected compartments.

For mitochondrial studies, this asymmetry matters. Mitochondrial function is not fully represented by transcript abundance. Protein complex assembly, post-translational regulation, turnover, and subcellular localization can diverge from RNA measurements. Conversely, a proteomic measurement without cell identity can obscure which population produces the signal.

The most useful future datasets will combine:

• Cell-type-resolved transcriptomics.
• Quantitative proteomics with explicit fractionation metadata.
• Spatial localization of mitochondrial proteins.
• Donor-level clinical and grading information.
• Genomic data relevant to AMD, glaucoma, and inherited retinal disease.
• Controlled records of tissue ischemia, post-mortem interval, storage, and processing latency.

This is a data-integration problem as much as an assay problem. The limiting step may not be mass spectrometer capacity. It may be the absence of consistent sample identifiers, anatomical coordinates, and pre-analytical timestamps across platforms.

Metabolic pathway dysregulation in early AMD

The early AMD proteomic signal is best interpreted as a metabolic systems disturbance. Mitochondrial translation, ATP metabolism, lipid homeostasis, and oxidative stress response were all significantly dysregulated in donor RPE samples.

These pathways form a connected operating network.

Mitochondrial translation supports production of proteins required for respiratory function. Respiratory function supports ATP generation. ATP availability constrains active transport, maintenance, and biosynthetic work. Lipid handling affects membrane turnover and the processing of photoreceptor-derived material. Oxidative stress response limits damage generated by respiratory activity and other cellular processes.

A disturbance in one component can increase pressure on the others. Reduced respiratory efficiency can increase compensatory demand on energy pathways. Impaired lipid processing can alter membrane and organelle stress. Deficient protein quality control can reduce the stability of respiratory complexes. The proteomic data do not prove a single linear cascade, but they do support a network-level model rather than a one-marker disease mechanism.

The analysis also provides a useful correction to late-stage bias. Mitochondrial protein dysregulation is not restricted to advanced AMD. It is detectable in early donor tissue. That does not make mitochondrial proteins standalone diagnostic biomarkers. It does make them relevant to the molecular characterization of early disease and to the design of translational experiments focused on intervention timing.

Interpreting oxidative phosphorylation markers

Retinal mitochondrial studies often prioritize oxidative phosphorylation markers because they are biologically intuitive and technically tractable. ATP synthase subunits and cytochrome c oxidase components are valuable measurements. They should not be interpreted in isolation.

A change in one oxidative phosphorylation marker may reflect:

• Altered mitochondrial abundance.
• A shift in respiratory complex composition.
• Selective protein degradation.
• Changes in mitochondrial translation.
• Tissue or fractionation bias.
• Post-mortem degradation.
• A compensatory response rather than primary failure.

The correct analytical sequence is to compare the marker with related proteins and pathways. If ATP synthase subunits change alongside mitochondrial translation factors, electron transport components, and oxidative stress proteins, the interpretation gains coherence. If only one protein changes in a low-quality or poorly annotated specimen, the signal requires more caution.

This is why broad quantitative proteomics is useful. It increases the probability that an observed mitochondrial change can be assessed against its neighboring molecular systems. It also increases the metadata burden. More proteins do not reduce the need for provenance. They increase it.

Post-mortem mitochondrial protein stability remains a limiting variable

Human donor eyes are irreplaceable research material. They are also time-sensitive biospecimens. After death, oxygen supply stops, temperature changes, enzymatic activity continues for a period, and tissue handling introduces additional delay. Mitochondrial proteins may not degrade at the same rate. Subcellular fractions may respond differently to ischemia and storage.

The exact post-mortem interval cutoff at which mitochondrial protein stability becomes unreliable across donor retinal subfractions is not established. That uncertainty should be visible in study design and reporting. It should not be replaced with a universal threshold that the current evidence does not support.

Post-mortem mitochondrial protein stability is therefore a kinetic variable, not a binary quality label. Relevant factors include:

• Time from death to enucleation.
• Time from enucleation to dissection.
• Tissue temperature during transport.
• Freezing method and time to cryopreservation.
• Anatomical region sampled.
• Degree of tissue disruption.
• Fractionation workflow.
• Storage duration and freeze-thaw exposure.
• Donor disease state and age.
• Protease and phosphatase control during processing.

The operational consequence is clear. Procurement latency must be recorded as a sequence of timestamps, not compressed into a vague statement that tissue was processed rapidly. A research database should distinguish donor death time, recovery time, laboratory receipt, dissection start, fractionation start, and final storage. Without that chain, downstream studies cannot model degradation kinetics or compare specimens reliably.

Quality control for donor-eye mitochondrial datasets

Quality control should operate at three levels.

Specimen level. Confirm donor identity, diagnosis, grading information, anatomical source, recovery conditions, and pre-analytical timeline.

Fraction level. Record the enrichment method, expected purity, yield, storage conditions, and any deviation from the standard workflow.

Assay level. Track instrument platform, batch, peptide identification metrics, missingness, normalization, and technical replicates where available.

The objective is not to remove every variable. That is not realistic in human donor tissue. The objective is to make the variables measurable and traceable.

A useful dataset should allow an analyst to answer four questions without reconstructing the study from scattered files:

1. What tissue and compartment were measured?

2. What was the donor disease and grading context?

3. How much time elapsed at each handling stage?

4. Which part of the observed signal is biological, technical, or uncertain?

If those answers are unavailable, pathway enrichment can still be performed. Causal interpretation becomes weaker.

What the current evidence supports

The present evidence supports several conclusions.

First, mitochondrial protein expression in donor retinal tissue is measurably altered in early AMD-associated RPE samples. The signal is broad enough to involve mitochondrial translation, ATP metabolism, lipid homeostasis, and oxidative stress response.

Second, quantitative donor-eye proteomics can achieve substantial depth. The UHR-IonStar analysis quantified 5,941 proteins in organelle-enriched RPE fractions from 45 early AMD donors and 32 healthy controls.

Third, mitochondrial changes can track progressive grading context. Earlier MGS 1–4 profiling identified significant shifts in eight protein spots that included ATP synthase subunits, cytochrome c oxidase subunit VIb, mitofilin, mtHsp70, and mitochondrial translation factor Tu.

Fourth, mitochondrial protein distribution is compartment-specific. In RGCs, SOD2 was enriched in the somatodendritic compartment, while HADHA was enriched in the axonal compartment.

The evidence does not support several stronger claims. It does not establish a universal post-mortem interval cutoff for reliable mitochondrial quantification. It does not provide complete single-cell proteomic maps of human photoreceptors and bipolar cells. It does not show that isolated mitochondrial proteins represent whole-retina levels. It does not demonstrate that the observed changes are exclusive to late-stage AMD; in fact, the available cohort includes early disease.

The field is moving from marker lists toward integrated donor-eye molecular systems. That transition depends on better alignment between procurement records, tissue geography, fractionation metadata, proteomic depth, and transcriptomic identity.

The strict assessment is therefore straightforward: mitochondrial proteomics has already demonstrated a reproducible and biologically coherent signal in human donor RPE, including early AMD. Its principal constraint is no longer whether mitochondrial proteins can be detected. The constraint is whether each detected signal can be assigned to the correct donor, compartment, disease stage, and post-mortem timeline. Without that assignment, the dataset remains descriptive. With it, donor-eye proteomics becomes a usable infrastructure for retinal functional genomics and translational disease modeling.