What is the embryonic origin of cancer in newborn twins?

One twin's tumour crossed over only once

The paper asks whether identical cancers in monozygotic twins arise twice independently or arise once and then spread between the twins in utero. The two scenarios leave different genetic signatures: independent tumours should carry separate mutation sets, whereas a single tumour that spread should share its mutations across both children's lesions. Whole genome sequencing of 10 tumour samples, eight from twin A and two from twin B, found only twin A's genotype in every tumour, wherever it was sampled. That establishes the sarcoma formed in twin A and then transferred to twin B.

The transfer was not repeated. Reconstructing the tumour's subclonal structure from shared mutations and copy number alterations showed that spread within twin A likely involved multiple clones, while twin B's tumours most likely descended from a single dissemination event that happened relatively late. Assuming a linear mutation rate, the authors date the tumour's origin to the end of the first trimester and the inter-twin spread to late in the second trimester. Those dates are model-based estimates, not direct observations, and they depend on the linear-rate assumption.

Variant allele frequency measures lineage share

Most of the argument rests on variant allele frequency, or VAF: the fraction of sequencing reads at a locus that carry the mutant base. In a bulk tissue sample, VAF approximates how large a share of cells descends from the cell in which that mutation arose. A mutation approaching VAF 0.5 in twin A therefore indicates that the cell carrying it gave rise to nearly all of that twin's cells. This is why VAF, rather than the mere presence or absence of a mutation, carries the developmental information here.

The authors used this logic on 20 developmental mutations found across normal and placental tissue, sorting them into three groups: mutation A present in both twins, mutations B to H making a major contribution to twin A, and mutations I to T restricted to twin B. The VAFs summed to 0.5 across the three parallel lineages in each twin, which the authors take as evidence that they captured all early lineages contributing to the sampled tissues. That claim is bounded by which tissues were sampled.

The placenta came mostly from twin B's lineage

The lineage map is strikingly lopsided. One lineage, marked by mutations C and F, contributed almost exclusively to twin A. A second, marked by mutation A, contributed to both twins and the placenta. A third, marked by mutations J, L, Q, T, I, N and P, contributed to twin B and the placenta. Only one mutation was found in both twins, so each twin was built mainly from its own separate early lineage.

Placental sampling across six territories, 18 bulk samples plus 12 trophoblast-enriched samples collected by laser capture microdissection, showed a dominant contribution from twin B's lineages. The authors flag that they could not sample the entire organ, so this asymmetry describes the sampled placenta rather than a proven whole-organ composition. It is nonetheless the observation that drives their reinterpretation of how twinning happens.

A hundred-cell split cannot explain these labels

The classical model places the split at about day 6 after conception, when the embryo has roughly 100 cells, after trophectoderm formation and before amnion formation. The authors argue this fits their data poorly. If the zygote's descendant cells were already mixed, or the fission plane randomly oriented, cutting a 100-cell blastocyst in two would be unlikely to yield one twin built almost entirely from one lineage and the other from a different one. Yet that is essentially what the mutation labels show.

Their data instead favour an earlier split, though not at the first cleavage: mutation A appears in both twins' soma, so at least four cells must have existed at the time of separation. The authors also invoke trophoblast fusion, the idea that one twin's trophectoderm can replace the other's, to explain why the placenta is mostly twin B's lineage. They are explicit that this is an argument from a single twinning event and from surviving cells only, so other arrangements may exist.

Twin B's cells dominate twin A's spleen

Screening normal tissues for twin-defining mutations turned up twin B's mutations in the solid spleen sample from twin A, at unusually high VAFs. Applying a mixing equation to the observed and expected VAFs, the authors estimate that about 75% of cells in twin A's spleen actually came from twin B. This is a separate measurement from the tumour transfer: it concerns blood-cell mixing between the twins, with flow from twin B to twin A, and it is why the spleen sample from twin A and the skin sample from twin B were excluded from the aggregated normal-tissue VAF analysis because of cross-contamination.

A rare spread, not an inevitable one

The most consequential reading is about likelihood rather than possibility. Because twin B's tumours trace to one late dissemination event after the tumour had already diversified and metastasised within twin A, the authors conclude that transplacental tumour spread was a rare event and not a foregone outcome of sharing a circulation. The same sequencing depth that made the tumour phylogeny resolvable also produced the twinning reconstruction, which is why the paper presents the two findings together.

The priority for readers is therefore the pairing of two claims of different strength. The tumour's origin in twin A and its single transfer to twin B rest on genotype evidence found in every tumour sample. The earlier-split model of twinning rests on one case, on sampled rather than whole organs, and on surviving cells, so it should be treated as a challenge to the classical model rather than a replacement for it.

Sources used in this answer