Human Endogenous Retroviruses

How a retrovirus becomes heritable, why insertion position rather than presence carries the argument, and how the method has been tested across mammals is set out in Endogenous Retroviruses. This article applies it to the human genome.

About 8% of human DNA consists of the remains of retroviruses that infected ancestral germ cells and were inherited from then on — several times the fraction occupied by protein-coding genes.

Humans and chimpanzees carry these sequences at the same positions, and which primates share which insertions reconstructs the same family tree produced by anatomy and by unrelated genetic data.

What the human genome contains#

Human endogenous retroviruses are grouped into families by sequence similarity. The youngest and best studied is HERV-K, specifically the HML-2 subgroup, which contains the only insertions retaining nearly intact viral reading frames.

Older families are more degraded. Many survive only as solo long terminal repeats, the residue left when the two identical sequences flanking an insertion recombine and excise the interior.

Humans and chimpanzees share on the order of 200,000 insertions at corresponding positions. The number of shared insertions declines with gorillas, further with orangutans, further with gibbons, and further with Old World monkeys — in the order the rest of the evidence predicts.

Insertions used to build the primate tree#

Johnson and Coffin (1999) traced individual HERV-K insertions across primate species and reconstructed the primate phylogeny from insertion sites alone, without reference to anatomy or to other sequence data. The tree they recovered matched the standard one.

Every insertion site is an independent opportunity for that to fail. An insertion shared by humans and macaques but absent in chimpanzees would not fit any tree; none of that kind has been found.

Shared solo LTRs sharpen the point. Where humans and chimpanzees carry a solo LTR at matching coordinates, that records two events at one location — an insertion and a subsequent internal deletion — both occurring in a shared ancestor.

Insertions still segregating in living people#

The most recent human insertions have not finished spreading through the species.

Turner et al. (2001) identified full-length HERV-K proviruses, including HERV-K113 and HERV-K115, that are present in some people and absent in others — insertion polymorphisms still at intermediate frequency in human populations, with frequencies varying by geographic ancestry.

This is the process caught partway through in our own species. An insertion that entered the germ line recently is carried by some lineages and not others; given enough time it either disappears or fixes, at which point it becomes one of the shared markers used to reconstruct deeper relationships.

The same gradient is visible across primates: recent insertions polymorphic within humans, older ones fixed in humans, older still shared with chimpanzees, older still shared with all apes.

Syncytin and the human placenta#

The best-documented case of a co-opted human ERV is syncytin-1, derived from the envelope protein of an HERV-W insertion and essential to formation of the placental syncytiotrophoblast. Mi et al. (2000) established the function; a second gene, syncytin-2, was later found to be derived from a separate insertion.

This is a genuine function performed by a viral sequence, and it is not disputed.

It does not indicate design, for a reason that is only visible outside humans. Other mammal lineages independently captured syncytin genes from different, unrelated retroviruses — a separate capture in rodents, another in carnivores, another in marsupials (Dupressoir et al., 2012). Placental mammals did not receive one solution; separate lineages each retained a different viral accident.

Common objections#

"HERVs have functions, so they were designed rather than inherited"

The premise is correct in specific cases, syncytin most clearly.

Co-option is expected. A sequence resident in a genome for millions of years occasionally lands somewhere useful and is retained by selection.

Function also does not explain placement. If these were designed elements, there is no reason for their distribution to form a tree, and no reason for it to be the same tree produced by anatomy, chromosome structure, and unrelated gene sequences. Placing syncytin in placental mammals is intelligible as design; distributing 200,000 mostly-broken viral remnants in a pattern reproducing inheritance is not.

"Most of the genome is functional, so 'junk DNA' was a failed prediction"

This draws on the ENCODE project's 2012 report that around 80% of the genome is "biochemically active." Discovery Institute authors, including Jonathan Wells in The Myth of Junk DNA (2011), have argued this refutes the evolutionary account.

ENCODE's finding is real but narrower than the summary suggests. Its definition of biochemical activity included any reproducible transcription or protein binding, which occurs at low levels across much of the genome and is not the same as function. Graur et al. (2013) criticised the equation in detail.

It also does not reach the argument, which does not depend on ERVs being junk. Even if every insertion were functional, the question would remain why their positions form a nested hierarchy matching independent evidence.

"The same virus could have infected humans and chimpanzees separately"

The same virus family infecting different lineages is ordinary and does happen.

The argument rests on position rather than presence. Independent infections place insertions at different sites; what is observed is the same family at the same base pair in both genomes, roughly 200,000 times.

A small number of apparently shared sites do turn out on inspection to be independent insertions into the same hotspot, or alignment artifacts. They are documented in the literature and rare enough to be reported as notable exceptions, which indicates how consistent the general pattern is.

"This assumes evolution to prove evolution"

The core observation does not. Insertion positions are read directly from aligned genome sequences; determining that a sequence sits at the same coordinate in two species requires no assumption about how it got there.

What common descent supplies is a prediction about the pattern those positions should form. That prediction could have failed, and in the whale and hippopotamus case the method was run on a relationship that was still disputed.

What the evidence shows#

Humans and chimpanzees carry roughly 200,000 retroviral insertions at matching genomic positions, with shared internal deletions at some of those same sites, and the distribution across primates reconstructs the standard tree from insertion data alone.

The most recent human insertions are still polymorphic — present in some people and absent in others — which shows the process partway through in our own species rather than only in its finished state.

Some insertions have acquired functions, syncytin most clearly, which is expected and does not affect an argument resting on the pattern of positions.

The general treatment is in Endogenous Retroviruses. Related human evidence is in Chromosome 2 and Pseudogenes in the Human Genome.