Death and destruction
When the Vikings had departed this world and journeyed to Valhalla, some of their DNA survived to tell an epic tale of genetic pillage. Tom Gilbert helped to translate it
(31/05/2003)
By Tom Gilbert
STARTING in the 8th century, the Vikings of western Scandinavia spent more than 300 years roaming, trading and pillaging their way across much of Europe. They settled in locations as far-flung as Greenland and Russia, and their reputation as great warriors helped them to secure employment as the famed Varangian Guard of the Emperors of Constantinople. There is even evidence that by the late 10th century they had reached the eastern coast of North America some 500 years before Columbus.
The archaeological and historical record tells much about Viking history, but some mysteries remain. In particular, to what extent did the Vikings interact with local populations? Did they routinely take their own womenfolk with them or did they father children with local women as they traversed or settled in new areas? It is questions like these that my colleagues and I at the University of Oxford 's Henry Wellcome Ancient Biomolecules Centre were attempting to answer when we stumbled upon an extraordinary finding.
Our analysis of 1000-year-old mitochondrial DNA from Vikings in England and elsewhere seemed to show that a third or more of them had Middle Eastern ancestry. What was going on? As it turns out, our findings have not led us to rewrite Viking history. Instead, they give exciting new insights into how DNA changes before and after death. They lend weight to the still controversial idea that mutation is not random. And they will improve our ability to track the migrations and interactions of human populations using ancient DNA.
It all began in 2001 when the BBC engaged several groups of British scientists to do original research for its television series Blood of the Vikings. The aim was to use the latest techniques developed by population geneticists to discover the extent to which Viking genes had contributed to the modern genetic make-up of the British Isles. Several groups, including one led by David Goldstein at University College London, used DNA samples from living people for their analysis. At Oxford, however, we focused on mitochondrial DNA extracted from ancient Viking skeletons.
Mitochondria, the cell 's "powerhouses", contain multiple copies of their own DNA. In humans this takes the form of circular, double-stranded molecules, each approximately 16,570 base pairs long. What makes this genetic material ideal for population studies is that it is relatively easy to track through the generations. Because all our mitochondria are descended from the mitochondria in our mother 's egg, there is no mixing with the father 's genes, as happens with nuclear DNA. Usually a child 's mitochondrial DNA (mtDNA) is identical to its mother 's, but occasionally mutations occur during the transfer between generations normally a single base change. By plotting such changes over time, you end up with a phylogenetic tree of related individuals. Of course, in practice it is highly unlikely that generations of mtDNA will be available for analysis, so the challenge is to generate ancestral trees based on the distribution of modern sequences. This is done using computer models to work out how the DNA sequences might have changed over time to give the current distributions.
With such trees as a guide, mitochondrial samples can be divided into groups called "haplotypes", each of which contains closely related DNA sequences. The distribution of these haplotypes in indigenous populations those that are relatively unaffected by the large-scale movements of recent centuries changes across the world, reflecting historical population movements (see Diagram). For example, in north-western Europe haplotypes known as H and V predominate, and others such as M are much less common, but as you move eastward towards India, the trend is towards decreasing H and V, and increasing M.
With a global picture of haplotype distributions, researchers can then start asking more specific questions about past events. So, for example, assuming that modern Scandinavians have a similar haplotype distribution to the Vikings, you can look for Viking ancestry elsewhere in the world. This is exactly what Goldstein and his team did for the Blood of the Vikings series although, instead of using mtDNA they looked at Y chromosomes. These are passed from father to son without any mixing with maternal genetic material. So they can be tracked through the generations in the same way as mtDNA, providing an insight into the genetic contribution of Viking males.
To work out how many Brits have Viking genes, Goldstein 's team took DNA samples from British volunteers and looked to see what proportion of haplotypes they shared with Scandinavians. Their results showed that Y chromosomes in England and mainland Scotland are derived from a mixture of Angles, Saxons, Danish Vikings and Ancient Britons, with inhabitants of the north and east of England possessing the most Viking genes. The highest incidence of such genes was found in York, the site of a well-known Viking settlement, whereas Cornwall, a region believed to be little affected by the invaders, had the lowest.
But there 's a potential problem with this sort of analysis if there has been any significant but undocumented population movement in Scandinavia since the Viking age, then the modern haplotype distribution may not resemble that of the Vikings. And that 's where we came in. Our job was to look at DNA from Viking remains in England, Denmark and Greenland, to check that the pattern of haplotypes matched modern samples from Scandinavia. It was a fairly routine study entailing the extraction, amplification and sequencing of ancient mtDNA, before the more tricky Y chromosome analysis. But the results were far from routine they were startling.
Analysing the mitochondrial region known as HVS-1, we found that between 30 and 40 per cent of samples taken from 30 skeletons had a group of haplotypes known as N , which are commonly found in the Middle East, but according to our colleague Vincent Macaulay, who is an expert on haplotype distributions in modern Europe, are only found in about 5 per cent of modern Scandinavians. Could this be evidence of something not revealed in the archaeological record, we wondered? Perhaps Vikings of the Constantinople guard had fathered children with local women and then travelled back north, raising the children as Viking warriors. Or maybe the Vikings ' gene pool already contained N haplotypes but they are masked in modern Scandinavians as the result of an unknown, large-scale migration into the region in more recent times.
What 's in a millennium?
It was all very exciting. But the field of ancient DNA is full of pitfalls, and anyone who works in it soon learns to be cautious. Besides, the archaeologists were adamant that the DNA analysis must be flawed in some way. Martin Biddle from the University of Oxford and Birthe Kjolbye-Biddle from the Winchester Research Unit in Oxford, who excavated the Viking site at Repton in Derbyshire, pointed out that there is no archaeological or historical evidence to support the notion that Vikings in the late 9th century returned to northern Europe from the Middle East with women and children.
What other explanations were there? We began to wonder whether our ancient DNA had been contaminated with modern genetic material. This seemed unlikely because few Middle Eastern archaeologists would have had access to the samples. Or perhaps the Viking DNA had changed in the thousand or more years since the individuals died. This seemed more plausible. After all, we know that DNA is a relatively unstable molecule that can be fragmented and modified by all sorts of insults, from ionising radiation to free radicals. During life, a host of DNA repair enzymes counter these effects, but after death there is no way to avoid the damage.
It is almost 15 years since Svante P bo, the "guru" of ancient DNA analysis, showed that bases the "letters" of the genetic code can be modified to resemble one another after death, just as happens in some point mutations during life ( Proceedings of the National Academy of Sciences, vol 86, p 1939). Today P bo heads up the Max Planck Institute for Evolutionary Anthropology in Leipzig, and recent research there and here in Oxford has taken his findings one step further. The Leipzig team, led by Michael Hofreiter, and my colleagues and I at Oxford, have shown that the bases cytosine and adenine are particularly susceptible to modification into the molecules uracil and hypoxanthine, which are analogues of the bases thymine and guanine ( Nucleic Acids Research, vol 29, p 4793; American Journal of Human Genetics, vol 72, p 48).
Although such modifications arise relatively slowly after death, the lack of repair mechanisms in dead cells means they are permanent. Provided you have a large enough sample of DNA, the changes will be masked by the vast majority of unmodified DNA copies. But ancient DNA is difficult to extract and is often only available in tiny quantities, so analyses are often based on just a few DNA molecules that have been amplified using the polymerase chain reaction (PCR). Oliva Handt and colleagues at the University of Munich have shown that when the initial sample contains fewer than 40 molecules, modifications in some can skew the results when they are amplified ( American Journal of Human Genetics, vol 59, p 368). Could this explain our findings?
To test this idea, we used a technique called molecular cloning, which allows you to examine the individual amplified DNA strands from any single sample by inserting each into a separate bacterial cell. We could then look for minor variations between the DNA sequences from each extracted sample. And sure enough, the N haplotypes were not present in all the DNA strands, showing that their presence was not a true indication of Middle Eastern ancestry, but instead a result of distortion caused by PCR amplification of a few DNA strands that had been modified after death. That, in itself, was disappointing. But it still left the intriguing observation that these post-mortem changes were clustered around a few key areas and that these coincided with stretches of DNA that have undergone serious genetic change during our evolution.
DNA sequences evolve over time through the combined effects of mutation and natural selection. Although the biochemical processes behind mutation are assumed to be random, the constraints of natural selection result in different rates of change in different parts of the genome. Coding DNA genes tends to evolve slowly because many of the mutations that occur here will impair the function of the proteins for which it codes, reducing the survival chances of an individual, and of the mutation itself. But non-coding DNA tends to show elevated mutation rates and so greater variation. The reason is that change here is not as damaging, making the DNA more likely to be passed down to subsequent generations.
More difficult to explain is the finding that within regions of high mutation there are hot spots where mutation rates are higher still. In recent years, studies of modern mtDNA sequences have convinced most people that hot spots do exist. Our analysis of HVS-1 sheds new light on the phenomenon. These same hot spots sustain damage after death and many of the modifications to DNA that occur match those that happen during life. So hot spots cannot simply be the result of selection pressures allowing greater genetic variability in these stretches of DNA there are no such pressures on the dead. What is going on?
One idea, suggested originally by Sonja Meyer and colleagues at the Max Planck Institute in Leipzig to explain the modern hot spots is that the secondary structure of DNA the configuration of the DNA strand itself, rather than the bases in it might expose some areas to biochemical modification while protecting others ( Genetics, vol 152, p 1103). Proteins or other molecules bound to the DNA sequence could do a similar job. Our finding that genetic damage after death resembles that during life supports the notion that a structural mechanism might be involved. What 's more, one of the few areas of HVS-1 that shows unusually low mutation levels has almost exactly the same configuration as a region of coding DNA and is thought to be a protein-binding site. Here, at least, it looks as though a protein bound to a stretch of DNA has protected it from damage.
Self-defence where it counts
And bizarrely, when we extended our study to look at the gene cytochrome oxidase III, we found that, as in life, damage rates here were unusually low, even a thousand years after death. There is no way DNA repair can continue after death. The only explanation is that some structural element is protecting the coding DNA from attack. The evidence is tantalising, but we still have little idea whether this structural protection comes from conformational changes that depend on the exact base sequence, protein binding, or some other mechanism such as DNA 's ability to conduct charge away from vulnerable areas ( New Scientist, 15 March, p 38) .
Nevertheless, it looks suspiciously as if the mitochondrial genome has evolved the ability to resist attack in some areas and succumb in others. I believe the same thing may be happening in nuclear DNA, although the mechanisms may be different. If genomes do focus mutations at hot spots and reduce them in areas where they are likely to threaten survival, then mutation is not random. What 's more, it looks as though genomes that take control in this way are fitter than those that bend to the vagaries of random mutation. So they have been selected over the course of evolution.
Genomes taking control of their own destiny the idea is sure to cause a stir among evolutionary theorists. But our findings also have more practical implications they can improve our ability to understand human history and the relationship between individuals. Many of the genetic models used to analyse population and evolutionary changes require predictions of how a DNA sequence mutates. We have come a long way from the early days, when we assumed that mutations were randomly distributed along the DNA sequence and that all nucleotides were equally likely to be modified. We already take into account differences in mutation rates between coding and non-coding DNA, and between the various "letters" of the genetic code. Now we will have to adapt our models to reflect the presence of mutation hot spots.
This is particularly important in the field of ancient DNA, where sequence changes at hot spots such as those in HVS-1 are often used to assign individuals to different haplotypes. A better understanding of which positions are prone to chemical change after death will allow us to make more accurate analyses of the distribution of past mitochondrial genomes, and how populations have changed to give the haplotype distributions we see in people alive today. We should also be aware that past studies similar to our Blood of the Vikings analysis may be flawed.
If this all sounds a little negative, there is a final twist to our work that promises to strengthen the position of mtDNA analysis as a tool in population genetics. In recent years there has been heated debate over whether or not mtDNA undergoes recombination the exchange of genetic material between chromosomes during the formation of eggs or sperm. It may sound arcane, but this really matters, because if there is recombination in mtDNA, then it becomes just as tricky to track down the generations as most nuclear DNA. And that shakes the foundations upon which much of population genetics rests.
In support of recombination, Erica Hagelberg from the University of Oslo has recently argued that rapidly changing areas of mtDNA are not mutational hot spots but sites where recombination occurs. This is plausible because recombination does result in stretches of DNA that are much more variable than you would expect. But then how do you explain the hot spots in ancient DNA? Recombination only occurs in living cells you won 't find it in Vikings dead for a thousand years.
From New Scientist Magazine 31 May 03
http://www.newscientist.com/hottopics/dna/article.jsp?id=23974800&sub=Ancestry