At the end of September, a piece of news from Cornell University made the rounds: a human gene, BC200, had been found in the genome of the molluscum contagiosum virus. Molluscum contagiosum is a poxvirus that causes harmless skin warts and, as far as we know, can infect only skin cells. The story would have looked routine if this gene had not turned out to possess two properties at once: it works in neurons and yet remains “jumping”—able to cut itself out and insert itself into new locations in the genome.
Jumping genes, also known as transposons, are pieces of DNA that copy themselves and shove those copies across chromosomes, breaking normal genes that happen to be in their way. Barbara McClintock discovered them in the 1940s; the named professorship now held by Cedric Feschotte, one of the new paper’s authors, bears her name. Usually, a transposon adapted for useful cellular work loses its mobility and becomes an ordinary gene.
“Evolution somehow failed to disentangle these two things,” Feschotte said of BC200.
What caught me was not the theft itself. What caught me was that the traffic runs both ways: we do not merely lose genes to viruses, we also assemble our own working components out of viral and transposon junk. Before getting into what exactly we took, I searched the archive for the words “transposon,” “jumping gene,” “reverse transcriptase”: the subject had never come up even once. That does not happen often, and there is usually a story beneath it that is not sitting out in the open.
The story was found in 2010, almost by accident. Cheng Sun, then a postdoc in the joint lab of Feschotte and Ellen Pritham, was writing a program to search viral genomes for human transposons when he came across two BC200 insertions in molluscum contagiosum. They saw the importance immediately, but did not write the paper. The project came back to life a decade and a half later, when Sun and Feschotte happened to meet at a conference: by then, Sun’s graduate student Pu Gao had dug up the old files. “I can’t believe we weren’t scooped,” Feschotte commented.
The timeline came out neat. BC200 itself was discovered in the late 1980s as one of the most abundant noncoding RNAs in human neurons; it exists only in humans and closely related primates, while its precursor was co-opted from an ancient transposon in the common ancestor of primates about 40 million years ago. It entered the viral genome twice in human history, around 100 thousand years ago. And the virus infects only skin—so the jump most likely happened in skin cells.
Little is known about BC200’s function: it is thought to regulate the translation of neuronal messenger RNAs. But more is known about its behavior. It is found in small quantities in sperm and egg cells, which means its insertions can be inherited. It is abnormally expressed in certain tumors and elevated in the brains of people with Alzheimer’s disease. And it remains the only known human element that both serves the cell and retains the ability to jump.
There are precedents for this kind of “theft,” though they are rare. In the late 1980s, in a moth cell culture, a transposon jumped into a baculovirus—the first clear case of a mobile element escaping from a host genome into a virus. In 2007, a group from Japan found a transposon—one from a snake, at that—inside a rodent poxvirus. What makes the new paper different is what jumped: not an anonymous repeat, but a functioning human gene.
Our adaptive immune system rests on a stunt: the enzyme RAG1/RAG2 cuts several hundred genetic segments and stitches them together in random combinations, so that a finite set of pieces yields millions of antibody and receptor variants. The reaction is called V(D)J recombination, and without it there are no lymphocytes, no vaccines.
In 2005, researchers figured out where this enzyme came from. The core of RAG1—roughly 600 amino acids—turned out to be a close relative of a Transib-family transposase, a protein whose job is precisely to insert a mobile element into DNA. More than that: the sequences by which RAG recognizes cut sites originated from the terminal inverted repeats of an ancient Transib, and they share the same DDE catalytic triad. The entire adaptive immune system of jawed vertebrates—that is, ours—grew out of a transposon about 500 million years ago.
They acquired the tool. Then they had to disarm it.
Eleven years later, in amphioxus, a primitive marine chordate, researchers found ProtoRAG—a living, functioning transposon built exactly as predicted: it cuts itself out, inserts itself, and repairs the damage behind it. The researchers called it a molecular living fossil. Then, in 2019, ProtoRAG’s structure was captured with cryo-electron microscopy, and two modifications were counted that distinguish our version from the transposon: an arginine at position 848 in RAG1 and an acidic region in RAG2. Together, they suppress spontaneous jumping by more than 1000-fold. Domestication turned out to be a two-step operation: switch on the cutting ability and switch off the urge to scatter across the genome.
The immune systems of lampreys and hagfish show that this was not the only possible solution: they have lymphocytes and immune memory, but their VLR receptors are assembled from leucine-rich repeats through gene conversion involving cytosine deaminases. A different construction set, the same result.
A retrovirus can write a DNA copy of itself into its host’s genome. If that copy lands in a germ cell, it passes into the offspring like an ordinary gene, and such fossilized proviruses are called endogenous retroviruses. Up to 10% of mammalian genomes consists of such remnants.
The HERV-W family was brought into the genome of the ancestor of catarrhine primates by an infection 25–40 million years ago. Of the whole family, one locus—ERVWE1 on chromosome 7q21.2—retained a full-length envelope gene, and retained it under selection rather than by accident. Its protein, syncytin, fuses trophoblast cells into a single syncytial layer at the boundary between mother and fetus—without this layer, there is no placenta. Knocking out both mouse syncytins, A and B, halts placentation and kills the embryo.
This gene’s switch is assembled from the same material. Expression is controlled by a bipartite element: a retroviral promoter in the 5' LTR plus a 436-base-pair region containing a trophoblast-specific enhancer from an LTR of the MaLR family. Both parts of this switch were once foreign DNA, just from different families.
And now for the expiration date: these captures do not last forever. Syncytin-like genes have been picked up independently and repeatedly—in primates, muroids, leporids, carnivores, caviids, and sheep—over a span of roughly 10 to 85 million years ago, sometimes in pairs at once. And the envV gene, which entered the primate lineage more than 45 million years ago, still works in Old World monkeys but has lost its ability to fuse cells in humans. ERVWE1 broke even earlier in the same branch: baboons have a stop codon at position 181, macaques a frameshift at position 498. “Essential” genes acquired by chance have an expiration date.
It is worth remembering the first mistake in this story, too. The first candidate for the role of trophoblast fusion was another retrovirus, ERV-3, and the theory collapsed for a simple reason: people homozygous for an inactivating mutation were found—and without this protein, they were alive and well. The conclusion was withdrawn, the hypothesis rewritten.
Arc is a gene without which long-term memory does not form. In 2018, it emerged that the Arc protein behaves like a virus: it assembles into virus-like capsids—the protein shells viruses use to package their genomes—packs RNA inside them, leaves the neuron in a vesicle, and transfers its own messenger RNA to a neighboring cell, where it is translated depending on the cell’s activity. By origin, it is a former structural Gag gene from Ty3/gypsy-line retrotransposons—the same lineage that gave rise to retroviruses.
In fruit flies, the homolog dArc1 does the same thing at the neuromuscular synapse, except that it comes from another retrotransposon lineage. The same trick—moving information between cells via a capsid—was invented twice by evolution, and both times at the expense of mobile elements.
MER41 is a family of endogenous retroviruses found only in primates. In 2010, a strange effect showed up: roughly 5% of the signal in STAT1 binding data fell on these repeats. The cell’s interferon response, the main line of defense against viral infection, was partly being “pressed” through viral sequences.
Six years later, this was confirmed directly and expanded: different lineages of endogenous retroviruses had independently scattered across mammalian genomes regions that switch on interferon-response genes. When some of these elements were deleted from human cells, neighboring genes stopped switching on properly, including activation of the AIM2 inflammasome—one of the sensors of foreign DNA. Fighting viruses, we pull viral switches.
And the trail runs beyond immunity. In 2019, a group of researchers showed that the promoters of genes associated with intellectual disability are enriched specifically for MER41 LTRs—this stood out among more than 130 HERV families examined—and those same regions contain binding sites for the immune transcription factors STAT1, NFKB1, and STAT3. No such enrichment was found for autism or schizophrenia genes. No ready-made moral follows from this, but the outline is there: the same regulatory platforms operate in both immunity and brain development.
Now for the proportions, because this is where it is easy to throw the reader off with a big number. According to estimates from 2001, about 45% of human DNA comes from transposons, and roughly 8% consists of sequences resembling infectious retroviruses. The simplification “half of you is viruses” is wrong twice over: retroviruses make up only part of the mobile elements, and most copies have had time to accumulate mutations. Only three full proviruses with intact gag, pol, and env genes were found in the genome, and one of them has a mutation in the reverse transcriptase domain.
The youngest family, HERV-K(HML-2), has been active for roughly the past 5 million years and is implicated in germ cell tumors. But the 2001 review is cautious to the point of boredom about links between HERVs and disease: direct evidence is scarce, and high repeat expression in cancer and autoimmune conditions proves nothing by itself—dozens of families switch on at once in such cases, and finding the culprit is difficult.
Insertion rates were measured in two ways, and the results diverged. A direct count across 33 three-generation pedigrees (599 people, 437 births) found one new Alu insertion per 40 births, L1 per 63, and SVA also per 63, versus a previous estimate of 1 in 900; Alu showed a statistically significant paternal bias. A broad study of developmental disorders (9738 families) found roughly one new insertion per 12–14 births across all classes combined. Phylogenetic estimates, on which the old figures rested, turned out to be about twice as high as the direct measurements—an ordinary story for estimates obtained by different methods.
The price of the lease looks like this. Each of us carries roughly 1200 polymorphic mobile insertions in our genome, three quarters of them Alu. About one insertion per genome lands directly in a protein-coding sequence, and that accounts for about 0.6% of all protein-truncating variants in a personal genome. Some 130 pathogenic insertions have been described; among 9738 families with severe developmental disorders, nine new ones were found, four of them landing in genes already known from such diagnoses. Every 2434th case of developmental disorder in this sample is explained by such an insertion. Alu takes up about 11% of the genome and more than a million copies, and accounts for around 0.3% of genetic diseases—including through major rearrangements when repeats fuse.
A very recent example of what this looks like in the clinic: familial Marfan syndrome caused by an Alu insertion in the FBN1 gene. The insertion was missed in the first sequencing run—it was found only by the next generation of methods. In this family, the diagnosis spent more than ten years running up against an element that the previous technology simply could not see.
And BC200 in this picture is not a sensation but a refinement of the proportions. The gene works in human neurons and continues to jump; the virus has carried two copies of it for a hundred thousand years and may be using them in its own interests. That will have to be tested in separate work: it would be interesting to know whether BC200 jumps in tumor cells, and whether that is connected to its elevated level in the brains of people with Alzheimer’s.