A hypocretin molecule. Credit: Boghog2
A hypocretin molecule. Credit: Boghog2

Well, this is disappointing.

Last December, I wrote a story for Nature News about a new paper, published in Science Translational Medicine, which seemed to confirm narcolepsy as an autoimmune disease. Here’s what I wrote at the time:

Narcolepsy is mostly caused by the gradual loss of neurons that produce hypocretin, a hormone that keeps us awake. Many scientists had suspected that the immune system was responsible, but the Stanford team [led by Elizabeth Mellins and Emmanuel Mignot] has found the first direct evidence: a special group of CD4+ T cells (a type of immune cell) that targets hypocretin and is found only in people with narcolepsy.

“Up till now, the idea that narcolepsy was an autoimmune disorder was a very compelling hypothesis, but this is the first direct evidence of autoimmunity,” says Mellins. “I think these cells are a smoking gun.” The study is published today in Science Translational Medicine1.

Thomas Scammell, a neurologist at Harvard Medical School in Boston, Massachusetts, says that the results are welcome after “years of modest disappointment”, marked by many failures to find antibodies made by a person’s body against their own hypocretin. “It’s one of the biggest things to happen in the narcolepsy field for some time.”

Now, it looks like the years of modest disappointment will continue because this paper is being retracted. The notice says:

“The researchers report that they have been unable to reproduce the paper’s key findings. Specifically, they could not demonstrate a differential Enzyme-Linked ImmunoSpot response of CD4+ T cells from patients with narcolepsy compared to those from normal controls after exposure to the hypocretin peptides HCRT56–68 or HCRT87–99. Because the validity of the conclusions reported in the study cannot be confirmed, they are retracting the article.”

To explain, in the study, the team showed that two fragments of hypocretin activated a specific group of CD4+ cells in narcolepsy patients, but not in healthy people. The Enzyme-Linked ImmunoSpot, or ELISpot, was a test they used to confirm the existence of these cells… and it’s no longer giving the same results that are described in the paper.

I contacted Mellins for more details, and was referred to the institute’s Director of Media Relations, Ruthann Richter. She sent the following statement from Mignot who is currently in China.

Beginning in March, my lab could not make the ELISPot test work. After many attempts in several settings and verification of all reagents, we decided to withdraw the manuscript. We have no reason to believe that the DQ0602 in vitro binding studies of peptides reported in the paper are problematic. We have already moved forward with other interesting findings on the narcolepsy/H1N1 association.

That’s not massively illuminating about what caused the original problem but neither researcher was available for interview, and Richter added, “We don’t have any more information at this point.”

(DQ0602 is a version of a protein found on the surface of immune cells. It’s found in the majority of people with narcolepsy and only a minority of the general population. The two hypocretin fragments that the team used in their experiments were those that can stick to DQ0602.)

I contacted Gert Lammers, president of the European Narcolepsy Network, who originally told me that “the results are very important, but they need to do a replication study in a large group of patients and controls.” He said:

“This may indeed be considered as a blow to the field, although I already wrote you that the findings needed replication. Studies like this are very complex and subject to many known and in part unknown or unidentified influences, and therefore always need replication preferentially by an independent lab. The methods applied seemed to be sound and most of the findings seemed to make sense (although also inducing new questions) and therefore I had and have no reason, with the knowledge I currently have, to mistrust the data presented in the paper. However, it was remarkable that after almost a year there were no reports about successful replication of the findings in Stanford nor anywhere else. This was starting to become a hint that there could have been a problem with the experiments and discussions about this possibility were starting.”

As always, Ivan Oransky is on the case. He has more comments from Mignon and Mellins, and the comments on his post may eventually shed more light on the background to this retraction. I also wrote a pointer to the Nature piece on this blog, which I have since amended.