When chromosomes divide, they look like molecular Siamese twins: two stretches of DNA joined along a portion named the centromere.

Centromeres ensure that when the new two DNA molecules separate, new cells present the right DNA amount. The origin of the centromeres is unknown.

A new controversial theory states that crucial innovations started out at the ends of ancient chromosomes. The DNA sequence of the centromeres varies quite a bit from organism to organism, thus the centromeres could have evolved many times.

But all centromeres do the same thing: they maintain the new chromosomes together at the beginning, then help bind protein strands that pull chromosomes apart as the cell splits in two. These proteins are the same in all organisms, pointing to a single origin.

10 years ago, Alfredo Villasante, a molecular biologist at the Autonomous University of Madrid, Spain, found an amazing trait on the centromeres of fruit flies: some centromere DNA stretches fitted DNA in telomeres, the caps of chromosomes, that often shorten with each cell division and are involved in the aging process. Protein strands sometimes bind to telomeres to split chromosomes, that's why the Spanish team came with the new hypothesis.

Chromosomes were originally circular, like in bacteria, but sometimes the circles broke open and mobile elements of the round DNA molecule passed to the ragged ends, capping them and impeding circles formation.

"Over time, these mobile elements accumulated and became the telomeres," said Villasante.

When these sequences grew bigger, points inside the capped ends turned into what would be the later centromere. The DNA in these points produced RNA, which bound with certain proteins that attracted the protein strands that split off the chromosomes.

"At first, a separate set of strands attached to another, more ancient anchor point in the genome as well. With two sets of strands yanking it at different places, the chromosome tended to break up. Some of the fragments reunited, putting the protocentromeres more in the middle of the chromosomes--just as they tend to be in modern chromosomes," said Villasante.

"The authors have made a good case, within the limitations of the data," said John Kelly, a molecular biologist at the London School of Hygiene & Tropical Medicine.

Others are skeptical.

"This is highly improbable, new centromeres are evolving all the time, and not from telomeres. There is no reason to suppose any connection between centromeres and telomeres in function, in time of action, or in sequence," said Steven Henikoff, a biochemist at Fred Hutchinson Cancer Research Center in Seattle, Washington.