Every multicellular organism is, strictly speaking, a mosaic: each cell division carries a small probability of error, so cells inevitably accumulate private mutations. Mosaicism as a clinical concept refers to a postzygotic mutational event — occurring after fertilization — that creates a subpopulation of cells carrying an alteration distinct from the original zygote, in a proportion large enough to matter biologically.

The developmental-timing principle is the key to understanding mosaicism. Picture mutation as a single point dropped onto the branching tree of cell lineages. Every cell descended from that branch inherits the change; everything on other branches does not. A mutation in the first cleavage division is copied into roughly half of all body cells and, because the embryo has not yet committed cells to specific germ layers, can seed ectoderm, mesoderm, and endoderm alike — and the germline. A mutation occurring late, during organogenesis, is confined to whatever tissue patch that lineage builds. This is why timing dictates both the burden of mutant cells and their anatomical reach, and why two patients with the identical variant can look entirely different depending on when in development it arose.

A crucial corollary for counseling: the germline branches off from somatic lineages early in development. A mutation arising before that split can land in both soma and germ cells; one arising after it is locked into either the body or the gonad, but not both. This single fact underlies the entire distinction between somatic and germline mosaicism and the very different recurrence risks each carries.

Types of alteration: Any class of genetic change can occur postzygotically — chromosomal non-disjunction, structural rearrangement, single-nucleotide variant, copy-number variant, or epigenetic (methylation) change. The mechanism of detection and the clinical consequence differ for each, but the underlying logic of timing and lineage is shared.

Chromosomal mosaicism traces back to one of two kinds of event. In a chromosomally normal zygote, mitotic non-disjunction in an early cleavage division creates an aneuploid line: when sister chromatids fail to separate, one daughter cell gains a chromosome (trisomy) and the other loses one (monosomy), while their euploid neighbors continue normally — producing two or three coexisting cell lines from one zygote. A related mechanism, anaphase lag, loses a lagging chromosome from one daughter cell and is a common route to losing the second sex chromosome in 45,X/46,XX mosaicism. In a trisomic zygote, trisomy rescue — loss of the extra chromosome in some cells — leaves a mosaic of trisomic and disomic cells, and in about one third of rescue events the two remaining copies come from the same parent (uniparental disomy). A SNP array can often distinguish the two: a third parental haplotype points to a meiotic error followed by rescue.

Why mosaic aneuploidy is generally milder than the constitutional form is itself instructive. A full trisomy burdens every cell with a gene-dosage imbalance throughout development; in mosaicism, a reservoir of normal cells can compensate, and there is often selection against the aneuploid line — many trisomic cells proliferate poorly and are progressively diluted out. This is precisely why a constitutional trisomy that is usually lethal (trisomy 8, trisomy 9) can be compatible with life when mosaic: the normal cell line carries the organism.

The central counseling caveat is the disconnect between the tissue you sample and the tissue that matters. Because the two cell lines were distributed semi-randomly to different organs during early embryogenesis, the percentage of abnormal cells in peripheral blood need not match the percentage in brain, gonad, or skin. A reassuringly low blood mosaicism level cannot guarantee a comparably low burden in the nervous system, and a high blood level need not predict severe cognitive involvement. Phenotype therefore cannot be read off a single karyotype, and prognostic statements based on blood percentage alone are unsound.

For decades, a focal brain malformation in a child with normal blood genetics and unaffected parents was filed under "sporadic" — a diagnostic dead end. Somatic mosaicism reframed that entire category. Many of these lesions are not the absence of a genetic cause; they are a genetic cause hiding in the lesion itself, undetectable wherever you happened to look first.

The developmental logic explains the anatomy. During cortical neurogenesis, a mutation in a single neural progenitor is clonally propagated to all of its descendants — the neurons and glia that migrate out to populate a circumscribed territory of cortex. The result is a discrete malformed patch within an otherwise normal brain, with the variant present in a minority of cells (in one surgical series, allele fractions mostly 0.4–8.9% in FCD type II (up to 18.6% in one large FCD IIa) and 7.5–34% in hemimegalencephaly; Baldassari et al. 2019) and frequently absent altogether from blood and other organs, because the founding mutation arose after those lineages had already diverged. The earlier the hit, the larger and more bilateral the malformation; a very early event can produce hemimegalencephaly, while a later one yields a small focal cortical dysplasia.

The landmark demonstration came from paired brain–blood sequencing in hemimegalencephaly, where de novo somatic activating variants in the PI3K–AKT3–mTOR growth pathway (PIK3CA, AKT3, MTOR) were found in roughly 30% of resected specimens but were undetectable in the same patients' blood Lee et al. 2012. These variants drive constitutive mTOR signaling, abnormal neuronal size and migration, and intrinsic epileptogenicity — the mechanistic basis for the term "mTORopathies." The same logic extends across mosaic disease: in Sturge-Weber and McCune-Albright syndromes, a single early activating mutation (GNAQ and GNAS respectively) is distributed to whichever tissues the mutant clone happened to build, so the rash, the angioma, and the endocrine lesion trace the embryonic lineage map rather than any inherited pattern. Tuberous sclerosis and DEPDC5-related epilepsy add a second-hit twist: an inherited loss-of-function variant in a repressor of mTORC1 is present everywhere, and a focal lesion can arise where a postzygotic somatic hit removes the remaining allele — the Knudson two-hit model operating in the brain. The model is well supported for SEGAs and angiomyolipomas and for DEPDC5-related dysplasia (Ribierre et al. 2018), but second hits were found in only about a third of cortical tubers (Martin et al. 2017).

The unifying detection lesson: if the disease lives in the lesion, that is where you must sequence, and you must sequence deep — standard germline pipelines built to call ~50% heterozygous variants will silently miss a variant present in a small minority of cells.

Germline (gonadal) mosaicism is the somatic story turned inside out. Here the postzygotic mutation lands in a precursor of the germline, so it is confined to — or substantially enriched in — the gonad while sparing the body. The parent looks entirely normal, tests negative on blood DNA, and yet carries the variant in a fraction of eggs or sperm, ready to pass it on.

This is the single most important — and most counterintuitive — concept in recurrence counseling. A variant labeled "de novo" because it is absent from both parents' blood is not the same as a variant with zero recurrence risk. The phrase clinicians must internalize is that de novo describes where we failed to find the variant, not where it does not exist. The classic red flag is two affected children of clinically unaffected parents carrying the identical "de novo" variant — vanishingly unlikely as two independent events, but exactly what germline mosaicism predicts. Because a single mutant clone can populate a large share of one gonad, transmission is not a rare fluke; the average recurrence risk after a de novo variant is usually quoted at 1–2%, but that average hides very different families: in a study that deep-sequenced several tissues from both parents (and sperm from fathers), the risk fell below 0.1% for 35 of 58 families and rose for 5, reaching 5.6–12.1% in fathers with mixed mosaicism (Bernkopf et al. 2023). In multi-sibling families, 3.8% of de novo variants were mosaic in a parent's germline and 1.3% were shared by siblings (Rahbari et al. 2016).

The best-quantified example is Duchenne muscular dystrophy, where roughly one-third of cases arise de novo and maternal germline mosaicism is repeatedly documented among them — giving an empirical recurrence risk of about 4% when the mother's blood is negative, and about 9% if a later son inherits the same maternal X as the affected boy Helderman-van den Enden et al. 2009. The principle generalizes broadly:

• Autosomal dominant conditions: osteogenesis imperfecta (where some families show strikingly high recurrence), the RASopathies, and many skeletal dysplasias
• X-linked conditions: DMD/dystrophinopathies and Rett syndrome (MECP2), where nearly all de novo variants arise on the paternal X (Trappe et al. 2001) and parental germline mosaicism leaves a low but above-background risk despite negative parental testing

The practical consequence: pure gonadal mosaicism is invisible to blood-based testing, and mixed mosaicism is often present in blood below the floor of a clinical exome, so germline mosaicism must be assumed possible unless deep testing of several parental tissues (and, in fathers, sperm) has excluded it. That single shift — from "recurrence risk is zero" to "there is a small but real residual risk" — is what makes prenatal diagnosis (CVS/amniocentesis) and PGT-M legitimate options to offer after any apparently de novo dominant or X-linked diagnosis.

Detecting mosaicism is a problem of signal versus noise at low allele fractions, and it has two independent failure modes: sequencing the wrong tissue, and sequencing the right tissue too shallowly. Understanding both is what separates a confident negative from a missed diagnosis.

Why depth sets the detection floor. A variant present in a given fraction of cells contributes a proportional fraction of the sequencing reads at that position. To call a true 5% variant, you must distinguish 5 mutant reads out of 100 from the background error rate of the sequencer (~0.1–1% per base) — a margin that simply does not exist at the 50–100× depth of a routine exome, where a handful of reads is within the range of random error and gets filtered as noise. Adding depth is the remedy: at 500–2000× a 1% variant is represented by many reads, lifting it cleanly above the error floor, provided the analysis uses a somatic variant caller that does not assume the ~50% allele fraction of a germline heterozygote. This is the conceptual reason a negative standard exome cannot exclude low-level mosaicism — the variant was below the platform's detection floor, not absent.

Why tissue choice is non-negotiable. No depth of sequencing recovers a variant from a tissue the mutant clone never colonized. Because mosaic variants track embryonic lineage, the sample must be chosen to capture the affected lineage: resected epilepsy cortex for a focal cortical malformation, affected-skin biopsy for a segmental skin disorder, and — when invasive tissue is unavailable — saliva or buccal swabs, which include neural-crest- and ectoderm-derived cells and outperform blood for many mosaic neurocutaneous conditions. Sampling several tissues raises the odds of intersecting the mutant clone.

Matching the tool to the question completes the strategy: chromosomal microarray (especially SNP arrays read through B-allele frequency) detects mosaic copy-number and aneuploidy down to about 5–10% (Conlin 2012); high-depth NGS is the discovery tool for unknown low-level point mutations; and droplet digital PCR, once a specific variant is known, quantifies it down to ~0.01–0.1% to confirm and track mosaicism that NGS could only suggest.