Commercial antibodies are common in biology laboratories. Researchers use these giant Y-shaped proteins to detect specific molecules in cells, tissues, and test tubes. But sometimes the protein also detects other molecules, or – instead. When this happens, confusion can snowball.
Consider gene CR9ORF72. It is often mutated in people with neurodegenerative diseases amyotrophic lateral sclerosis and familial frontotemporal dementia. But what this actually does is difficult to do, partly because the widely changing locations of proteins in the cell provide more confusion for clarity.
Peter McPherson, a neuroscientist at McGill University in Montreal, Canada, raised suspicions that arise from multiple locations, often viewed as a trivial decision to detect proteins: the choice of antibodies. Antibodies work by binding to specific parts of a protein, depending on the size and chemical properties of the protein, but one antibody produced to bind one protein can often bind to another, and sometimes with better affinity.
It originates in MacPherson’s work. He and his team purchased 16 antibodies marketed to detect CR9ORF72. They then took a cell line that produces a high level of protein and used the genome-editing tool CRISPR – Cas9 to create a line in which CR9ORF72 Was knocked out, so the protein would not be present. They then assessed how antibodies were performed in two lines in a series of common tests and found that the antibody used in most publications (and most frequently cited) was still a protein when it was not. The best doers for each assay did not appear in the literature1.
Others have reported comparable experiences. Cancer researcher Cecilia Williams of KTH Royal Institute of Technology in Stockholm tested 13 antibodies to try conflicting data about estrogen receptor angle, a protein discovered in 1996 that is a potential anticancer target. Twelve of the antibodies, including the two most popular, gave either false positives or false negatives, or both, he and his team reported2. “Don’t take either literature or antibodies,” she warns.
Researchers often purchase antibodies according to the number of times the product is cited in the literature, but this strategy may ignore new products that have been put through more rigorous tests. They also believe that people who used antibodies did this work before they were examined, and that it would work in their own experiments, opening the door for self-destructive artifacts.
“When I look at the papers in general, I get depressed with the quality of antibody characterization,” said Simon Goodman, a science consultant at the Antibody Society, a nonprofit professional association. Goodman is based in Darmstadt, Germany, and has organized a series of educational webinars on technologies suitable for society3. “If you ask ‘How did you validate the antibody?’, The researchers would say, ‘We bought it and the manufacturer says it behaves like this.”
Often, the data that companies provide to show their antibody functions come from a cell line that has been engineered to express proteins at higher levels than under physiological conditions. Researchers have improved to test whether an antibody can detect proteins at the physiological level, the technique and tissue type they plan to use and, ideally, indicate when it disappears or when the protein The level disappears.
Researchers have found a steady slope of effort to be more careful. In 2016, the US National Institutes of Health (NIH) made the necessary start to grant applicants how they would certify antibodies and other key resources. The validation road map has been printed, summits have been held and web portals have been established – Antibodypedia, Antibody-Online, Antibody Resource, Biocomper, CiteAb and Labome, to name a few.
“CRISPR has been the true game-changer,” says Allred Edwards, who led the Structural Genomics Consortium from Toronto, Canada, a public-private partnership dedicated to doing basic science that can promote drug discovery. This is because the technique has made it easier to perform useful control experiments, as did McPherson and his team. Earlier this year, antibody vendor Abcam in Cambridge, UK, introduced a suite of knockout cell lines and preparations that researchers purchased under specific conditions in their laboratories to test their antibodies as well as positive and negative cells. Can. The company now has more than 1,600 cell lines and 2,400 cell laces available.
Edwards and McPherson helped establish a Toronto-based charity, called YCharOS (pronounced Ikaros), to put commercial antibodies to the test. They plan to use McPherson’s strategy to assess more antibodies against other targets, demonstrating performance in three general assays: immunoblot, immunopreservation, and immunofluorescence. They are also working with several antibody suppliers and pharmaceutical companies to develop standard operating protocols. Michael J. in New York City, along with some contributions from the NIH and Parkinson’s disease. The Fox Foundation is providing an initial fund of US $ 300,000 to test a suite of antibodies used in neuroscience.
Not every antibody can be tested using knockout controls, Edwards admits. About 10% of the genes are required for life, so a knockout cell line is not viable for them. In addition, an antibody that performs well in one cell line may be deficient in another. Nevertheless, these simple experiments can help identify antibodies that are not binding with their target proteins. There are other methods that researchers can try, such as mass spectrometry to pair immunopreservation, to see which proteins bind to which proteins.4. Ultimately, Edwards says, “there is an onus on the experimenter”.
James Trimmer said that this is more than correct experiments, James Trimmer, who directs the Neuro Lab at the University of California, Davis, strives to produce high-quality antibodies for neuroscience. An antibody that functions reliably when a protein (‘basic’) is inside a cell, when the protein is transformed into chemically protected tissue or appears in the cell mixture, and in sample preparation Small changes can also make a big impact. .
Researchers need to know how their own method compares with the methods used in validation experiments, and antibodies should be avoided when validation details are unavailable. “If you use them for the wrong purpose, they won’t be a good fit,” Trimmer says.
It is not uncommon for laboratories to purchase multiple antibodies and select the one that works best. But many developers license their antibodies to many distributors, who do not always disclose the origin of the antibodies. While venturing out to do antibody testing for C9ORF72, McPherson’s postdoc Carl LaFlam used CiteAb and research literature to identify more than 100 antibody products. He then sent the vendors for questioning and scanned the data sheet to find duplicates. Nevertheless, the team later found out that 2 of the 16 antibodies purchased from different companies were identical, so they unnecessarily ran a full set of experiments. “We wasted our money and our time and effort,” McPherson says.
Sometimes it is not even clear which antibodies researchers have used, especially in older studies. According to one analysis, approximately 11% of antibodies used in the papers published in 1997 are identifiable5 Led by researchers at the data-sharing platform SciCrunch at the University of California, San Diego (UCSD), and La Jolla, California; Nowadays, this figure has increased to 43%.
Anita Bandrovsky, head of SciCrunch at UCSD and a bioinformatist, is attempting to assign a unique identifier, called RRID, to each antibody and include it in publications. Researchers can locate or request an RRID on the Antibody Registry’s website, and the identifiers remain the same, regardless of whether an antibody-supplying vendor changes its inventory or goes out of business. Bandrosky states that when magazines mandate RRIDs, antibodies are much easier to find. The journal CellFor example, the author asks for RRID, and 97% of its antibodies are detectable.5. Both Nature And nature research journals encourage the use of RRIDs to track key biological resources including antibodies, cell lines, model organisms, and equipment.
RRIDs can reduce, but cannot solve, the same antibody problem being sold by multiple vendors: if the original source is explicitly disclosed, the same RRID may be assigned to all antibodies. Bandrovsky estimated that 2.5 million antibodies with RRIDs represent perhaps 700,000 unique molecules. But RRIDs do not differentiate between different batches of the same product, which can be particularly problematic for polyclonal antibodies purified from the blood of immunized animals and therefore compared to a mixture made from cultured cells. Are more.
The bottom line is this: Although an antibody-driven experiment unfolds, researchers would be skeptical. Goodman says that when experiments fail, researchers often question their technique. “Of course you blame yourself as a young scientist.” But the scientific community must be equally skeptical of antibodies that seem to work, says Edwards, and ask for evidence that they do before trusting them. “We buy antibodies, we don’t test them, and then we publish articles that send the field sideways.”