The finding, published in the journal Nature, suggests that new drugs that mimic the mutation's effect could do the same, researchers said.

"Less than 1 percent of the population has these alleles," or DNA variants, said Dr. William Thies, chief medical and scientific officer of the Alzheimer's Association, a research and advocacy group. "So there's no impetus to go out and get tested. But many companies are working on compounds that target the mechanism they found and so might be encouraged by these results."

Alzheimer's disease is defined by the presence of sticky plaques, composed of a peptide called amyloid beta. Amyloid beta comes from a larger protein called amyloid precursor protein, or APP. Over the last two decades, scientists have identified some two dozen mutations in the APP gene that cause early-onset Alzheimer's, which strikes before age 65.

A novel mutation in the APP gene does the opposite, scientists led by Dr. Kari Stefansson of privately held deCODE Genetics in Reykjavik found when they sequenced the genomes of 1,795 Icelanders.

Those 85 or older who have the beneficial mutation are 81 percent less likely to develop the neurodegenerative disease than others in that age group, the team found.

The variant also makes developing Alzheimer's four times less likely across all age groups.

"It confers extraordinarily strong protection," said Stefansson. "We found only five Alzheimer's cases (out of thousands of people) with it."

Because the beneficial mutation is so rare, however, its value is likely to be in pointing the way to drugs that prevent or treat Alzheimer's rather than encouraging testing to see which patients have the variant. The disease strikes an estimated 5.4 million Americans, including 13 percent of those 65 and older and 45 percent of those 85 and older.

STOPPING PLAQUE FORMATION

The mutation affects how APP, the protein, is fragmented. After a brain cell makes this protein, enzymes - designated alpha, beta and gamma - cleave it into fragments. Alpha-secretase snips APP in such a way as to prevent the production of amyloid beta and thus of plaques. Instead, a fragment that promotes the growth and survival of neurons is produced.

But beta-secretase, also called BACE, is a more deadly cleaver. Together with gamma-secretase, it snips APP so as to produce amyloid beta. Floating out of the neuron, it sticks to other amyloid beta fragments, forming plaques.

The beneficial mutation interferes with BACE's ability to cleave APP. That reduces the formation of amyloid beta by 40 to 50 percent, scientists at Roche Holding AG's Genentech unit, who co-authored the new study, found.

"This gives you a proof of concept that if you inhibit BACE it will protect against Alzheimer's," said deCODE Genetics' Stefansson. "Big pharma has been working on inhibitors for beta-secretase for 15 to 20 years, and this offers greater confidence" that those efforts will pan out.

Several companies are developing compounds that inhibit BACE. In April, Merck presented data from an early-stage human trial evaluating the safety of its MK-8931, reporting that it was generally well-tolerated and reduced amyloid beta levels as much as 92 percent.

Vitae Pharmaceuticals and Boehringer Ingelheim are jointly developing a BACE inhibitor whose progress a spokesperson describes as "robust," with an announcement expected this year. Eli Lilly and Co's BACE inhibitor recently advanced into mid-stage human trials.

But targeting enzymes that make amyloid beta has not worked so far. A drug from Lilly that inhibited gamma secretase failed in clinical trials in 2010 and made the disease worse.

"Hitting gamma secretase with a sledgehammer made people sick," said Rudolph Tanzi, a veteran Alzheimer's researcher at Harvard Medical School. One reason is that gamma secretase has many crucial functions in the body that have nothing to do with amyloid beta.

Targeting BACE could backfire for the same reason. The enzyme is involved in numerous physiological functions, said Tanzi. They include cleaving proteins to make myelin - the fatty sheath that allows neurons to carry electrical signals - and sodium channels - proteins that let ions in and out of brain cells.

Drugs that inhibit BACE too broadly might therefore cause serious side effects, said Tanzi. "That's why although this paper is going to encourage development of these compounds, I'm not optimistic."

The challenge is to develop compounds that interfere with the APP-cleaving enzymes only during that function, which Tanzi and others are pursuing.

The new study also found that people with the beneficial APP mutation are less likely to suffer other forms of age-related cognitive decline. The very elderly, aged 80 to 100, with the mutation scored better on a standard test of cognitive function than their peers without the mutation.

That suggests that normal cognitive decline also might be worsened by amyloid beta. If so, a breakthrough in developing an Alzheimer's drug that works might help other patients with memory loss, slowed mental processing and other cognitive impairments that accompany aging.

(Reporting by Sharon Begley; Editing by Michele Gershberg, Vicki Allen and Dan Grebler)

• A Permanent Economic Emergency