Miracle fat-burning hormone doesn't exist after all

Scientists are humans, and as such, they can sometimes get carried away when they make a breakthrough discovery. Because of this premature excitement, they may lose attention to detail, over-interpret results, or cut corners to speed up that much-desired Nature publication. The discovery of irisin, or ‘exercise hormone’, is one such example. Once thought to be a promising exercise-free solution for obesity and diabetes, irisin has now been shown to be no more than a random blood protein detected by flawed reagents.

Irisin was first discovered in 2012 by Bruce Spiegelman and colleagues at Harvard Medical School (US). In a Nature article, the researchers reported that after exercise muscle cells release a fragment of a pre-hormone-like protein called FNDC5 into the bloodstream, where it travels to adipose cells to trigger the conversion of white fat into calorie-burning brown fat. They concluded that this small molecule is a “newly identified hormone”, which they named irisin, after the Greek messenger goddess Iris.

Whereas white fat stores energy, brown fat is converted to heat—that’s how hibernating animals and newborn babies stay warm. So, unless you starve or exercise a lot, your white fat will remain stubbornly lodged on your hips, while brown fat burns calories. Unfortunately for most of us though, only about 10% of our adipose tissue consists of brown fat-producing cells. And this is why the discovery of irisin was so exciting. What if we could take an irisin pill to turn our white fat into brown fat? Could we burn calories while lying comfortably on the couch eating ice cream? 

It is no surprise then that in just three years over 170 studies were published on irisin. It didn’t take that long though for someone to question the Spiegelman study. Harold Erickson from Duke University (US) first voiced his concerns about irisin in 2013, and recently he showed that the commercial antibodies most widely used to detect irisin are unspecific—instead of irisin, they detect cross-reaction blood proteins, basically unknown random proteins.

Antibodies are proteins produced by immune cells that stick to specific bits of other proteins, and they’re used by scientists to detect their proteins of interest. In the original irisin paper, Spiegelman’s team identified irisin with a polyclonal antibody produced by Abcam that should in theory attach to the tail of FNDC5. But irisin is a fragment of FNDC5 that is chopped from the other end of the protein, so this antibody couldn’t possibly detect it, Erickson argued back in 2013. Spiegelman replied to this by saying that Abcam had not correctly annotated the antibody in their catalogue.

Erickson also noticed that none of the commercially available irisin antibodies had been properly tested by the companies that made them. But despite this worrying observation, several research groups continued to use them, and what’s worse, without attempting to verify their specificity for irisin. And there was more.

A few months after Erickson published these findings, Juergen Eckel and colleagues at the German Diabetes Centre (Dusseldorf, German) found that the human FNDC5 gene has an unusual START codon (the bit of DNA that is translated into the first ‘letter’ of a protein). This weird (and rare) codon is associated with very inefficient protein production. In the case of FNDC5, only about 1% of normal FNDC5 protein levels are produced by human cells, Eckel showed. At such low amounts, it would be highly unlikely that irisin had a physiological role in humans.

Over the years contradictory data from dozens of studies that relied on dodgy reagents cast doubts on whether irisin really exists or is a miracle fat-burning hormone, but that wasn’t enough to dissuade most researchers from working on it. Could this be about to change?

In their new study, Erickson's team and colleagues from three other research groups tested four commercial irisin antibodies used in over 80 studies. They employed a technique called ‘western blotting’, which separates proteins by size. To be sure they were looking at the right thing, the researchers synthesised irisin molecules and then compared them side-by-side with the proteins detected by the commercial antibodies. They tested several tissue samples from humans and other animals, including blood serum from horses after strenuous exercise. None of the antibodies detected a protein with the predicted size for irisin, and even more worrying, they didn’t detect synthesised irisin. However, the antibodies reacted with many other proteins of the wrong size. This shows that all previously published studies based on assays using these antibodies “were reporting unknown cross-reacting proteins”, the authors claim in the study.

The question now was… does irisin exist at all?  

To answer this question, the team looked for irisin in human blood serum using a sensitive technique that detects tiny amounts of molecules without the use of antibodies, called mass spectrometry. They were able to identify a molecule corresponding to FNDC5 or irisin, which is the “first mass spectrometry identification of an irisin peptide at the correct size, and might be considered as supporting the existence of irisin in human serum”, the authors say in the study. However, the very low amounts of irisin detected “makes a physiological role for irisin very unlikely”, they add. According to Erickson and colleagues, the exercise hormone is a myth.

These new findings are bad news for irisin researchers and food lovers, but they’re very good news for science. They show than even though human nature might at times corrupt scientific discoveries (voluntarily or involuntarily), science infallibly corrects itself, and we can therefore trust the scientific process.


Reference:

Albrecht E., Bernd Thiede, Torgeir Holen, Tomoo Ohashi, Lisa Schering, Sindre Lee, Julia Brenmoehl, Selina Thomas, Christian A. Drevon & Harold P. Erickson & (2015). Irisin – a myth rather than an exercise-inducible myokine, Scientific Reports, 5 8889. DOI: http://dx.doi.org/10.1038/srep08889

An edited version of this article was published in Lab Times on the 17-04-2015. You can read it here.