Mitochondrial DNA variants in donor retinal tissue analysis

In age-related macular degeneration, retinal samples show substantially more mitochondrial rearrangements and deletions than blood samples from the same donors. In donor retinal pigment epithelium, mitochondrial genome damage has been estimated at approximately eight times the level observed in nuclear genes.

That difference changes the procurement and analysis problem. A retinal specimen is not molecularly interchangeable with a blood specimen, and a macular punch is not equivalent to peripheral retina. The mitochondrial genome is exposed to a tissue-specific history of oxidative stress, energetic demand, cellular composition, disease progression, and postmortem degradation. Any donor eye mitochondrial genome sequencing workflow that compresses these variables into a single variant list will lose the primary structure of the signal.

The mitochondrial genome is a compact but high-density readout

Human mitochondrial DNA is a circular genome of approximately 16,569 base pairs. It encodes 13 protein subunits of the electron transport chain, 22 transfer RNAs, and two ribosomal RNAs. The genome is small relative to nuclear DNA, but its analytical behavior is more complex.

The relevant unit is not simply the presence or absence of a variant. It is the interaction between:

• the variant’s genomic position and functional class;
• the proportion of altered mitochondrial genomes within a tissue or cell population;
• the anatomical origin of the specimen;
• the donor’s age and disease status;
• the number and integrity of mitochondrial genomes recovered;
• the interval between death, tissue recovery, dissection, stabilization, and extraction.

This is the operational definition of retinal mtDNA heteroplasmy analysis. A tissue may contain a mixture of wild-type and altered mitochondrial genomes. The mixture can differ between retina and RPE, between macular and peripheral regions, and between individual cellular compartments. The available donor studies support regional and tissue-level differences, but they do not establish uniform heteroplasmy across all retinal areas or cell subpopulations.

The distinction matters because sequencing depth does not repair poor sampling design. A deeply sequenced specimen can provide a more sensitive view of the molecules that survived extraction. It cannot reconstruct a cell population that was never isolated, nor can it correct for differential degradation across anatomical regions.

In ocular mitochondrial analysis, sequencing depth improves detection. It does not eliminate pre-analytical bias.

The retina is also a demanding tissue system. Photoreceptors and RPE depend on continuous energy production and tightly regulated metabolic exchange. Mitochondrial lesions can therefore be interpreted as both a possible contributor to dysfunction and a record of cumulative tissue stress. The data do not justify reducing AMD to a single mitochondrial mechanism. They do justify treating mitochondrial genome integrity as a core variable in donor eye molecular profiling.

Retinal tissue and blood produce different mitochondrial signals

The strongest comparative result in the available evidence is not a single nucleotide variant. It is the difference between tissue compartments.

In AMD donor material, retinal mtDNA showed a higher rate of large rearrangements and deletions than blood mtDNA from the same individuals. The reported values were 14.33 ± 1.96 rearrangements or deletions per retina sample compared with 5.2 ± 0.80 in blood. The comparison is useful because paired donor material reduces some sources of between-person variation. The blood sample provides a systemic reference. The retina provides the local disease-associated molecular environment.

The two materials should not be treated as redundant biospecimens.

The numerical contrast should be read as a tissue-routing problem. If a study uses blood to infer the mitochondrial condition of the retina, it may miss lesions that are enriched in ocular tissue. If it uses bulk retina without recording the sampled region and cellular composition, it may detect a genuine signal but misassign its anatomical source.

This is where procurement latency enters the workflow. Mitochondrial DNA is physically robust compared with many RNA species, but intact long-range molecules and interpretable rearrangement profiles still depend on specimen handling. A short interval between death and stabilization does not guarantee a valid result. It creates a better starting condition. The downstream chain remains relevant: dissection precision, storage temperature, freeze-thaw exposure, extraction method, DNA fragmentation, library construction, and sequencing depth all affect the final call set.

The correct metadata model therefore needs more than donor diagnosis. At minimum, an ocular biobank record should distinguish:

1. donor disease classification and age-matched control status;

2. anatomical origin, including macular or peripheral sampling;

3. retina, RPE, or combined tissue;

4. postmortem interval and procurement timestamps;

5. stabilization and storage conditions;

6. extraction and library preparation method;

7. sequencing depth and quality thresholds;

8. whether the result represents bulk tissue or an enriched cellular fraction.

Without this structure, a mitochondrial variant database becomes a catalog of calls detached from the conditions that produced them.

Damage is not equivalent to pathogenicity

Large deletions, rearrangements, and non-synonymous substitutions are not interchangeable categories. They represent different analytical events and require different interpretation thresholds.

The evidence from AMD donor tissue indicates significantly elevated mtDNA lesion rates in diseased retina compared with age-matched normal controls. It also identifies non-synonymous variants in coding regions including NADH dehydrogenase subunits 1, 2, 4, and 5, cytochrome b, cytochrome c oxidase subunit 1, and ATP synthase subunit 6. These genes encode components of the respiratory chain. Their presence in an AMD-associated sample is biologically relevant, but detection alone does not establish that a variant initiated disease or drove progression.

A practical analysis should separate at least four levels of interpretation:

• Molecular detection: the sequence alteration is present above the method’s reporting threshold.
• Tissue distribution: the alteration is observed in retina, RPE, macula, periphery, or a defined cellular fraction.
• Quantitative burden: the proportion of altered molecules is estimated, where the assay supports that calculation.
• Functional attribution: the evidence supports a direct effect on mitochondrial activity or disease biology.

The first level is a sequencing output. The fourth requires substantially more evidence. Confusing them is a common failure mode in multiomics reporting.

Large deletions are particularly sensitive to assay design. Short-read data can identify breakpoint-supporting evidence, but detection depends on coverage distribution, breakpoint structure, library complexity, and the analytical pipeline. Long-extension PCR has been used in donor retinal studies to characterize full-length mitochondrial molecules and rearrangements. Deep sequencing has also been used to examine low-frequency variants and heteroplasmy. These methods answer related but non-identical questions.

A donor eye mitochondrial genome sequencing project should therefore report the method as part of the result, not as an implementation detail hidden in a supplementary file. A variant call without assay context has limited portability across biobanks.

The role of sequencing depth

Deep sequencing is necessary when the research question concerns low-frequency heteroplasmy. Donor retinal studies have used a minimum coverage of approximately 2,000× for deep-sequencing heteroplasmy analyses. That level of coverage increases the probability of observing minor alleles, but it does not make every low-frequency call reliable.

The error floor remains method-dependent. Polymerase errors, library artifacts, index contamination, mapping ambiguity, damaged templates, and amplification bias can all produce apparent minor variants. A robust pipeline needs technical controls and explicit thresholds for:

• minimum base and mapping quality;
• strand balance;
• local coverage uniformity;
• duplicate behavior;
• allele fraction;
• reproducibility across libraries or technical replicates;
• confirmation of structural breakpoints where applicable.

This is not an argument against deep sequencing. It is an argument for treating depth as one variable in a quality system rather than as a universal validation step.

Haplogroups provide context, not a standalone risk score

Mitochondrial haplogroups are inherited clusters of mitochondrial variants. They can act as population and lineage markers, and they may modify susceptibility to disease through interactions with mitochondrial function, nuclear background, environment, and aging.

In donor AMD studies, haplogroups J, T, and U have been associated with increased risk of age-related macular degeneration, while haplogroup H has shown a protective or lower-risk association. These observations are relevant for donor retinal tissue analysis, but they should not be converted into a deterministic classification of individual tissue.

There are several reasons.

First, a haplogroup is a background pattern, not a single pathogenic mutation. Its interpretation depends on the nuclear genome and the biological context of the donor. Second, association does not define mechanism. The available evidence does not establish exactly why haplogroup H may offer relative protection compared with J or T. Third, a haplogroup call does not describe the full burden of somatic mutations, deletions, or heteroplasmy within a particular retinal specimen.

A database should preserve these categories separately:

This separation is essential for bioinformatics of ocular mitochondrial variants. A system that stores only a final haplogroup and a list of variants cannot reconstruct whether an observed signal is inherited, somatic, region-specific, or introduced by technical artifact.

The problem becomes more pronounced when comparing macular and peripheral tissue. Existing evidence indicates that deep sequencing has been applied to donor macular and peripheral regions, but the available facts do not support treating heteroplasmy levels as uniform across those regions. The database should preserve regional identity rather than collapse both samples into a single donor-level record.

Non-synonymous variants map onto respiratory-chain vulnerability

The mitochondrial coding regions identified in AMD samples are not random from a functional perspective. Variants were detected in genes encoding components of NADH dehydrogenase, cytochrome b, cytochrome c oxidase, and ATP synthase. These proteins participate in oxidative phosphorylation and electron transport.

The analytical temptation is to translate the presence of a non-synonymous substitution directly into mitochondrial dysfunction. That shortcut is not defensible without additional evidence. A substitution may alter protein sequence, but its impact depends on conservation, allele fraction, tissue distribution, biochemical context, and interaction with the remaining mitochondrial and nuclear-encoded components of the respiratory chain.

For donor retinal multiomics, the useful approach is a layered association model:

1. Sequence layer: identify the variant and its allele fraction.

2. Structural layer: determine whether it co-occurs with deletions or rearrangements.

3. Expression layer: measure relevant mitochondrial and nuclear-encoded transcripts where RNA quality permits.

4. Protein layer: examine corresponding respiratory-chain proteins and proteomic signatures.

5. Phenotype layer: relate the molecular profile to donor diagnosis, retinal region, and tissue morphology.

The model is constrained by specimen quality. Transcriptomic yield usually declines faster than genomic DNA integrity after delayed recovery or suboptimal stabilization. A donor eye can therefore produce a usable mtDNA profile but an unusable RNA profile. That asymmetry should be recorded rather than hidden behind a combined multiomics label.

The same logic applies to mitochondrial copy number. Mitochondrial DNA copy number in human donor retina can be a valuable quantitative variable, but it requires a defined normalization strategy. Bulk tissue copy number may reflect both mitochondrial abundance per cell and shifts in cellular composition. It may also vary with extraction efficiency and nuclear DNA measurement. Without a stated reference method and tissue context, a copy-number value is not portable between studies.

A well-designed record would distinguish:

• absolute or relative copy-number measurement;
• nuclear reference loci;
• input mass and extraction yield;
• bulk versus enriched tissue;
• anatomical location;
• sequencing or quantitative assay platform;
• quality-control thresholds.

The database should not imply that every missing field is a minor documentation gap. In multiomics, missing metadata changes the interpretation space.

The primary unit of evidence is not the variant. It is the variant plus tissue origin, molecular burden, assay context, and procurement history.

m.3243A>G demonstrates why clinical interpretation must remain specific

The m.3243A>G variant in the MT-TL1 gene is a clinically recognized mitochondrial alteration. MT-TL1 encodes a mitochondrial transfer RNA for leucine, and the variant causes mitochondrial retinopathies and ocular manifestations including pattern macular dystrophy and ophthalmoplegia.

This variant should be handled as a defined pathogenic example, not as a general explanation for AMD-associated mitochondrial damage. Its presence belongs to a different interpretive category from the broader set of somatic deletions, haplogroup associations, and non-synonymous variants observed in donor AMD tissue.

The distinction prevents two opposite errors. The first is undercalling a clinically meaningful mitochondrial variant as just another entry in a high-dimensional variant table. The second is overextending a known mitochondrial retinopathy variant into a claim that it explains the mitochondrial lesion profile of age-related macular degeneration.

A tissue database can manage this distinction by assigning separate annotation fields:

• clinical or disease association;
• inheritance context;
• variant class;
• heteroplasmy measurement;
• tissue and region;
• evidence level;
• functional validation status;
• relationship to the donor’s recorded ocular diagnosis.

This is a data-modeling issue with direct scientific consequences. If clinical variants and exploratory somatic variants share one undifferentiated annotation field, downstream users may treat them as equivalent evidence. They are not equivalent.

Procurement metadata determines the usable resolution of the result

The molecular pipeline begins before extraction. For human ocular tissue, procurement latency and dissection strategy set an upper limit on analytical resolution. A retina sample acquired without a precise anatomical map cannot later be reconstructed as macular or peripheral. A combined retina-RPE specimen cannot reliably answer a cell-compartment question. A delayed sample may support a DNA-level assay while failing at the transcriptomic layer.

The operational sequence is straightforward:

1. Define the tissue question before dissection.

A project focused on RPE mitochondrial damage needs a different collection strategy from one focused on bulk retinal haplogroup profiling.

2. Record timestamps at each transfer point.

Donor death, recovery, receipt, dissection, stabilization, freezing, extraction, and library preparation should not be compressed into one postmortem interval field.

3. Preserve anatomical identity.

Macula, peripheral retina, RPE, and combined tissue should remain separate records or explicitly linked aliquots.

4. Track aliquot history.

Freeze-thaw cycles, extraction batch, and remaining material affect reproducibility and future validation.

5. Attach assay-specific quality metrics.

A high-quality DNA result does not certify RNA integrity, proteomic suitability, or single-cell compatibility.

6. Store negative information.

Failed libraries, insufficient yield, ambiguous region, and unresolved structural calls are part of the dataset. Removing them creates survivorship bias in the biobank.

This workflow is less dramatic than a new sequencing platform, but it determines whether the resulting data can be compared across donors and studies. The central bottleneck is often not read generation. It is the loss of provenance before the first read is produced.

What the current evidence supports

The evidence supports a clear but bounded assessment.

Mitochondrial genomes in AMD donor retinal tissue show a higher burden of large rearrangements and deletions than paired blood samples. In donor RPE, mitochondrial genome damage is estimated to be approximately eight times higher than damage in nuclear genes. Deep sequencing has identified non-synonymous variants in several respiratory-chain genes, and studies have used coverage of at least approximately 2,000× to examine low-frequency heteroplasmy. Haplogroups J, T, and U are associated with increased AMD risk, while haplogroup H shows a lower-risk or protective association. The m.3243A>G variant remains a defined cause of mitochondrial retinopathies with ocular manifestations.

The evidence does not support several stronger claims. It does not establish that all retinal cell types share the same heteroplasmy profile. It does not identify the exact mechanism behind the relative effect of haplogroup H. It does not make common LHON mutations primary causes of AMD in donor retinal tissue. It does not show that blood mtDNA can substitute for ocular sampling.

For ophthalmic multiomics, the practical conclusion is strict. Mitochondrial DNA variants in donor retinal tissue are interpretable only when sequence data remain attached to tissue compartment, anatomical region, procurement history, and assay performance. The most useful biobank is not the one with the largest variant count. It is the one that preserves the chain of evidence from donor eye recovery to molecular call.