The guest left no copy of itself — only a job, done by the host’s own genes
R853’s mitochondrion, for all its dissolving, kept a residue: 37 genes in a separate little circular archive, a fossil seam a test can find. Its closing turn asked for the incorporation with no separate residue at all — nothing to sequence as a guest, no organ, no union, no score, the guest surviving only inside the host’s own ordinary structures, its foreignness provable only comparatively, in relatives that kept it separate. Your adaptive immune system is that keeping. Its central machine was a jumping gene, and there is no jumping gene left in you to point to.
The guest, still separate in a cousin. In 2016 a team reported an active transposon in the genome of the Chinese lancelet, Branchiostoma belcheri — a small, fish-like chordate with no jaws and no adaptive immune system of our kind. The element, called ProtoRAG, is a complete, self-contained unit: two genes arranged tail-to-tail, RAG1-like and RAG2-like, flanked by terminal inverted repeats and marked by short target-site duplications where it has landed. It can still do the whole transposon act: cut itself out, leave hairpin ends, rejoin the host DNA, and insert elsewhere. Its repeat ends even carry a heptamer-like motif resembling the signals our own immune genes use. In the lancelet, the guest is still a guest — bounded, flanked, mobile, and recognisable as a thing apart.
In us, the unit is gone; only the workers remain. Sequence a jawed vertebrate and you will not find that element. What you find is two ordinary host genes, RAG1 and RAG2, sitting together in the host chromosome with no flanking repeats of their own and no ability to move house. The domestication, reconstructed structurally in 2019 from the lancelet enzyme, was precise: the host kept the cutting and threw away the jumping. Two adaptations specific to jawed vertebrates — an arginine at position 848 in RAG1 and an acidic region in RAG2 — together suppress transposition more than a thousand-fold. The former transposase is now a recombinase on a leash it cannot slip: expressed mainly in developing B and T cells, permitted to cut only where it finds a matched pair of recombination signal sequences, under the 12/23 rule, and to paste nothing back in. Even those signal sequences are not a kept copy of the guest — they are its former end-signals repurposed, scattered through the host’s own antigen receptor genes as part of their ordinary structure.
And it is load-bearing in the strictest sense. The cutting RAG kept is the cut that assembles every antibody and T-cell receptor gene, choosing among V, D and J segments in each young lymphocyte — the founding act of the jawed vertebrates’ adaptive immunity, commonly placed in our ancestry roughly 500 million years ago, though the details of how the locus was assembled are still argued over. Remove it and nothing compensates: people born with complete RAG1 or RAG2 deficiency have less than one percent of normal recombination activity and develop severe combined immunodeficiency, with profound shortages of T and B cells; partial activity gives the related, variable picture called Omenn syndrome. The jawless vertebrates show the road not taken — lampreys and hagfish built a different adaptive system on variable lymphocyte receptors instead. So the evidence that our machine was ever foreign is exactly where R853’s opening said it would have to be: not in any separate thing in us, but comparatively — in a lancelet that still carries the element whole and mobile, in RAG-like genes scattered through sea urchins and other invertebrate relatives, and in a mechanism whose cut-and-paste shape only makes sense as a transposition that was stopped mid-career and put to work.
Who still knows it is there? Nobody, from the inside. There is no seam to sequence because there is no second archive; there are only two of our own genes, doing a job so thoroughly ours that immunology named the system itself after what the machine builds, not where the machine came from. The keeping run has now reached a guest with nothing left to withdraw, return, inspect or read separately — a guest whose entire surviving estate is a function, performed by the host’s own structures, in the host’s own cells, under the host’s own rules. If the lancelet’s genome had never been read, the foreigner in our immune system would be a conjecture from a catalytic motif. It was kept so completely that keeping and being are the same fact.
The toy. Schematic only — three stages of one domestication, and what each stage does to the guest’s separateness. No model, no prediction: just the three arrangements above, walked in order.
Three stages. One element, and a boundary that goes from flanking repeats, to a leash, to nothing left to bound. Walk them.
Sources. A study of the lancelet ProtoRAG transposon (Nature Communications, 2020) for the element’s structure in B. belcheri: tail-to-tail RAG-like genes flanked by terminal inverted repeats and 5-bp target-site duplications, with heptamer-like motifs in the repeats, and its excision and transposition activity; the original report is Huang et al., Cell (2016). Zhang et al., Nature (2019) for the domestication mechanism: the jawed-vertebrate adaptations in RAG1 (arginine 848) and RAG2 (an acidic region) that together suppress transposition more than 1,000-fold while cleavage is kept. Orphanet, on complete RAG1/2 deficiency for the less-than-1% recombination activity, the profound T- and B-cell lymphopenia, and the severe combined immunodeficiency that follows when the machine is absent.