Mitochondria generate the bulk of cellular ATP through oxidative phosphorylation (OXPHOS) — the electron transport chain (Complexes I–IV) pumps protons across the inner membrane, and Complex V (ATP synthase) uses that proton gradient to phosphorylate ADP. When this machinery fails, the cell cannot meet its energy budget, and the cells that suffer first are the ones that spend the most: post-mitotic, high-demand tissues that cannot simply divide their way out of trouble.

This single fact — energy failure in the most energy-hungry tissues — explains the otherwise bewildering clinical picture. Brain, muscle, heart, retina, cochlea, endocrine pancreas, liver, and kidney dominate the phenotype because they have the highest mitochondrial density and the lowest tolerance for ATP shortfall. It is also why mitochondrial disease is the great mimic of medicine: a patient can present to a neurologist (stroke-like episodes, seizures, ataxia), a cardiologist (cardiomyopathy, conduction block), an endocrinologist (diabetes, short stature), or an ophthalmologist (optic atrophy, ophthalmoplegia) — and the unifying thread is only visible when you step back and ask why multiple unrelated systems are failing at once.

Why the unpredictability? Two features distinguish mitochondrial disease from textbook Mendelian disease:

• It is one of the only human disorders with two genomes in play — the small maternally inherited mitochondrial genome (mtDNA) and the much larger nuclear genome (nDNA) that encodes ~99% of mitochondrial proteins. Either can be the culprit.
• For mtDNA disease, each cell carries many copies of the genome, and a mutation may sit in some copies but not others (heteroplasmy). The same variant can therefore produce no symptoms in one tissue and devastating disease in another, and can shift dramatically from one generation to the next.

Taken together, these mean disease can present at any age and affect any organ, in almost any combination. The overall prevalence is approximately 1/5,000 — making mitochondrial disorders among the most common inherited metabolic diseases. The practical takeaway: 'think mitochondria' whenever an unexplained multi-system disorder spans organs that share nothing except a high demand for ATP.

The human mitochondrial genome is a compact circular, double-stranded molecule of 16,569 bp encoding just 37 genes: 13 OXPHOS subunit proteins, 22 transfer RNAs, and 2 ribosomal RNAs. Critically, mtDNA encodes only core subunits of the respiratory chain plus the apparatus needed to translate them — nothing else. Everything required to build, run, repair, and replicate the mitochondrion (~1,500 proteins, ~99% of the total) is encoded in the nucleus, translated in the cytosol, and imported. This division of labor is the key to understanding mitochondrial genetics: a single OXPHOS complex is assembled from subunits read off both genomes at once, so a defect in either book of instructions can break the same machine.

Why is mtDNA inherited only from the mother? The egg contributes hundreds of thousands of mitochondria to the zygote; the sperm contributes a few dozen, which are actively tagged with ubiquitin and degraded after fertilization. The result is strict maternal inheritance — an affected father, no matter how high his mutant load, transmits no mtDNA to his children. This is the single most important counseling point that separates mtDNA disease from nuclear disease, and it explains the maternal-line pedigree (diabetes-and-deafness through the mother's relatives) that should immediately raise suspicion. The historic confirmation that an mtDNA point mutation alone could cause neurological disease came from Wallace et al. 1988, who linked the m.11778G>A variant to Leber hereditary optic neuropathy and established the maternal-inheritance paradigm.

The two-genome rule has a clinically useful corollary. When the inheritance pattern is Mendelian (autosomal recessive, autosomal dominant, or X-linked) rather than maternal, the defect is almost always in a nuclear gene — even when the downstream lesion is in the mitochondrion. Some nuclear genes (POLG, TWNK/Twinkle, SLC25A4) maintain mtDNA itself; when they fail, mtDNA accumulates secondary deletions or depletion, producing a mitochondrial phenotype that nonetheless segregates as an autosomal trait. So a maternal pedigree points you toward mtDNA sequencing, while a recessive or dominant pedigree — or sporadic childhood disease — points you toward the nuclear genome, where the majority of pediatric mitochondrial disease actually lives.

Unlike the nuclear genome, where you inherit exactly two copies of each gene, mtDNA exists in hundreds to thousands of copies per cell. A pathogenic variant rarely affects all of them. When mutant and wild-type genomes coexist, the cell is heteroplasmic; when every copy is identical, it is homoplasmic. Heteroplasmy is the concept that makes mitochondrial genetics genuinely different — it turns a binary 'mutated or not' question into a continuous dose of dysfunction.

The threshold effect. Because residual wild-type mtDNA continues to make functional OXPHOS complexes, a cell can tolerate a surprising amount of mutant mtDNA before anything goes wrong. Disease only emerges once mutant load exceeds a tissue-specific threshold (~60–90%), above which the surviving wild-type genomes can no longer cover the ATP demand. This is why two relatives carrying the same variant can be asymptomatic and severely affected: it is the proportion, not the mere presence, of the mutation that matters. It also explains the mosaic biopsy — within one muscle, fibers that have crossed threshold stain COX-negative while neighboring fibers below threshold look normal, because heteroplasmy is partitioned cell by cell.

The bottleneck — why offspring are unpredictable. During early oogenesis the mtDNA population in the developing germ cell is dramatically reduced to a small number of segregating units, then re-amplified back to the full complement as the oocyte matures. Sampling a small number of genomes and re-expanding them is, statistically, a random draw — so the heteroplasmy a daughter inherits can differ sharply from her mother's. A mother with 40% heteroplasmy may have children ranging from <5% to >95%. This is not a transmission risk in the Mendelian sense (all of her children inherit some mutant mtDNA); it is a severity lottery, and it is the single hardest thing to counsel.

Two practical consequences follow. First, tissue choice matters for diagnosis: blood is convenient but, for several variants (notably m.3243A>G), mutant load drifts downward in blood leukocytes over time through negative selection, so a low blood result can be falsely reassuring — urine epithelium or muscle often carries far higher, more representative heteroplasmy. Second, recurrence risk cannot be quantified precisely for a heteroplasmic mother; counseling deals in broad ranges, not percentages.

A handful of mitochondrial syndromes recur often enough to have earned names and acronyms, and recognizing the pattern is what converts a vague multi-system presentation into a targeted genetic test. But a recurring theme cuts across all of them: the same variant can produce different syndromes, and the same syndrome can arise from different variants. This 'genotype is not phenotype' uncoupling is a direct consequence of heteroplasmy and the threshold effect — the tissues that happen to cross threshold determine which clinical picture you see.

The archetype is m.3243A>G. This single transition in MT-TL1 (the gene for mitochondrial tRNA-Leucine) was first tied to MELAS by Goto et al. 1990 and accounts for ~80% of MELAS. Yet the very same variant, at lower heteroplasmy or in different tissues, instead causes maternally inherited diabetes and deafness (MIDD) or isolated sensorineural hearing loss. One mutation, a spectrum of disease — depending entirely on dose and distribution. Because it sits in a tRNA, it impairs the translation of all thirteen mtDNA-encoded proteins, which is why tRNA mutations tend to produce such broad, multi-system encephalomyopathies rather than a single-complex defect.

The syndromes also illustrate the three mechanistic classes of mtDNA lesion, each with a characteristic inheritance and presentation:

• tRNA point mutations (MELAS / m.3243A>G, MERRF / m.8344A>G) — maternally inherited, heteroplasmic, broadly multi-system because protein synthesis is globally impaired.
• Protein-coding point mutations (LHON / Complex I subunits) — often homoplasmic yet incompletely penetrant, frequently tissue-restricted (the optic nerve in LHON), with a striking male predominance that hints at additional nuclear/hormonal modifiers.
• Large single deletions (Kearns-Sayre, Pearson, CPEO) — usually sporadic, arising de novo in the maternal oocyte rather than transmitted, which is why most KSS patients have no affected relatives. Severity tracks with deletion size and tissue distribution along the KSS → CPEO spectrum.

Mapping a presentation onto the right class — maternal pedigree vs. sporadic, multi-system vs. organ-restricted, point mutation vs. deletion — is what lets you order the correct test (targeted mtDNA sequencing vs. deletion/Southern analysis) and offer meaningful prognostic and reproductive counseling.

There is no single test that confirms or excludes mitochondrial disease, and that is the central diagnostic challenge. Because the underlying problem is energy failure in specific tissues, every line of evidence — biochemical, imaging, histological, molecular — is a sample of the wrong place at the wrong time unless interpreted carefully. The clinician's job is to assemble converging clues rather than chase one definitive result.

Why biochemistry is treacherous. Elevated lactate (and an elevated lactate:pyruvate ratio) reflects backed-up glycolysis when OXPHOS cannot keep up — a logical signature of the disease. But lactate may be normal between metabolic crises, or elevated only in the affected compartment: CSF lactate can be raised while blood lactate is normal, because the brain's energy deficit is not mirrored in the periphery. A normal lactate therefore never excludes the diagnosis; it only fails to confirm it. The same logic governs heteroplasmy testing — for respiratory-chain mtDNA disorders, muscle or urine outperforms blood, because the mutant load that matters is the load in the energy-starved tissue, not the load that happens to be measurable in a convenient one.

Why counseling is uniquely hard. Two genomes and heteroplasmy combine to defeat simple risk figures:

• For nuclear mitochondrial disease, counseling is ordinary Mendelian arithmetic (most often autosomal recessive, 25% recurrence).
• For mtDNA disease, it is not. Inheritance is strictly maternal — an affected father transmits zero risk, which is reassuring and definitive — but a heteroplasmic mother cannot be given a precise recurrence figure because the bottleneck randomizes the heteroplasmy her children inherit. Reproductive options exist (preimplantation testing for some variants; mitochondrial replacement therapy, approved in the UK for severe maternal mtDNA disease, which swaps the affected egg's cytoplasm for a donor's), but the counseling must honestly convey uncertainty of severity, not a clean percentage.

Why management is mostly defensive. With few disease-modifying drugs, the practical priorities are to avoid mitochondrial toxins and metabolic stress. The cardinal rule is that valproate is absolutely contraindicated in POLG-related disease — it inhibits the same mtDNA-replication machinery POLG encodes and can precipitate fatal hepatotoxicity (Alpers syndrome), so POLG status should be considered before valproate in any unexplained epileptic encephalopathy. Other agents that further throttle OXPHOS (metformin → lactic acidosis; aminoglycosides → hearing loss in MT-RNR1 carriers) warrant caution. Beyond drug avoidance, the energy-failure model dictates the rest: prevent prolonged fasting, treat intercurrent illness aggressively with an emergency protocol, and support with cofactors (coenzyme Q10, riboflavin, thiamine, L-carnitine) and arginine/citrulline for MELAS stroke-like episodes — interventions aimed at protecting a system that has no reserve.