The Wrong Path
(July 4th, 2017) Non-functioning antibodies are a nuisance in every lab. But when two decades of research are based on them, problems are of a much bigger dimension
For almost 20 years, researchers around the world have focussed their attention on an oestrogen receptor – known as ESR2 - present in breast tissue, hoping that it could be the answer to a new treatment to improve breast cancer survival rates. However, it turns out this receptor is not even there!
The problem in this case was not intentionally fraudulent results from the original studies but reliance on antibodies that cannot detect what they were supposed to detect. After an exhaustive comparison of 12 antibodies, oncologist Cecilia Williams, based at the KTH Royal institute of Technology in Stockholm, found that only one provided reliable results.
Williams started getting suspicious about her own results when antibodies showed extensive expression but mRNA levels were low. “I have tried to understand the role of this receptor for many years and have struggled to get clear data of its expression,” says Williams. “This was especially clear in cell lines where we did not have effect nor expression of ESR2 but still antibodies would stain.”
Most researchers are aware that there is a problem with antibodies but many disagree about which ones work and which ones don’t. The worrying part is that this problem has been going on for the best part of 20 years. “There has certainly been efforts to address this before,” says Williams but no answer yet “partly because good controls have been lacking and the understanding of where the receptor was supposed to be expressed (positive control) or not (negative control) might not have been correct.” Luckily, “the more recent databases of gene expression based on RNA-seq of tissues, such as GTEx and TCGA”, adds the researcher, “have opened up new possibilities to compare expression levels between individuals and tissues in an unbiased manner, and to set its expression levels into context, which was not possible some years ago.”
So, does this mean 20 years-worth of results looking at this receptor have just gone down the drain? Well … maybe not. “I think the understanding of where ESR2 is expressed, in both normal and cancer tissues, needs to be re-evaluated. This does not necessarily mean that everything is wrong but I would argue that it all needs to be repeated with a validated antibody. We cannot trust the data that is in the literature, as the majority has been produced with unspecific antibodies,” says Williams. For the researcher, the trick is not to rely on one single method of analysis. “mRNA levels HAVE to be taken into account - if mRNA is lacking or very, very low, this should not be disregarded, it needs to be looked into and explained.”
Then there’s also the fact that ESR2 is present in other cell types and the idea of a new treatment based on this receptor may yet come true. “We found a lot of it expressed in granulosa cell tumours (a type of ovarian tumour) where it might very well be a suitable target for therapy. Also, in some melanoma and thyroid tumours, there could be opportunities,” says Williams.
Curiously, the only antibody validated by Williams’ team – PPZ0506 – was often used as a control because it was not “detecting” the oestrogen receptor with the same efficiency as the remaining antibodies. “As this antibody would not stain normal breast tissue, it would likely be regarded as not working. Also, as many antibodies had been reported to work well in publications, other researchers would rather take that as validation of their functionality and favour them, rather than try untested ones.”
This work raises interesting questions about validation of antibodies used in research. Namely, who should be responsible? Companies selling or researchers using them? The easy answer would be the vendors (after all, would you buy a car if it wasn’t working?) but the issue cannot be that black and white. “I know that it is not quite as simple as that,” says Williams. “An antibody might work well in one application, but not in others and a company selling antibodies cannot reasonably be expected to be experts in every protein that they sell antibodies for.” The answer may come in the form of new guidelines recently issued, which Williams hopes will be adopted soon.
Having said that, it cannot all fall on the shoulders of companies developing new antibodies. Researchers must play their part in the validation process and use adequate and tested controls. “This is part of a systemic problem, though: researchers generally are awarded for the number of publications they produce and are under great pressure to be ‘productive’. They are not actually rewarded for the quality or reproducibility of the data, per se, and being thorough with validations could result in much fewer publications, which could mean the end a researcher's career,” concludes Williams.