Antibodies are part of the body's defense system. They latch onto intruders with specific aim, marking them for destruction. This specificity makes antibodies versatile lab tools. Unless it's Tuesday. Or there's a full moon. Antibody-based assays fail on too many lab benches and often for no clear-cut reasons.

Antibody companies use rabbits to produce polyclonal antibodies. Credit: Thinkstock

Scientists interviewed by Nature Methods say that troubleshooting their experiments frequently involves returning the antibody vial to the supplier. Antibody companies tend to be untroubled about this practice. Money-back guarantees and no-questions-asked exchanges or credits for buyers are common in this competitive marketplace populated by over 350 companies.

As they address the expanding needs of researchers, antibody companies are refining how they help researchers find the quality antibody for their purposes, increasing interactions with customers and adding validation methods. Separately, scientists who have learned hard truths about antibodies offer tips on how to be a discerning antibody shopper.

Variability factor: biology

Monoclonal antibodies are produced from cell lines, whereas polyclonal antibodies are harvested from animal serum, usually from rabbits. The animals produce antibodies in reaction to an injected substance, an immunogen. Variability is almost “the nature of the business,” says Jeff Xu, who works on antibodies at EMD Millipore. Antibodies from different rabbits reacting to the same immunogen have to be compared, he says. Biological variability can explain a multimonth absence of an antibody from the company catalog when it has failed internal quality controls, says Roisin Burns of Abcam, which sells antibodies and other proteomics tools.

The animal origin causes variability in several other ways. What might seem to be the same antibody with the same order number in a catalog is not always the identical antibody. During a rabbit's lifetime, the antibody quality can change, says Mike Browning, president of the antibody company PhosphoSolutions. Variability slips into the production process, too, when companies do not test each batch or when antibodies are purified from insufficient starting material. As Chandra Mohan from EMD Millipore explains, his colleagues purify large batches and use serum when the animal is producing maximum antibody titers. But when there are months between batches, the antibodies can vary slightly, he says.

The synthetic nature of monoclonal antibodies can address variability. Abcam bought the firm Epitomics last year, which sells rabbit monoclonals produced with proprietary technology. Abcam hopes these antibodies can reduce issues of “batch-to-batch variation” seen with polyclonal antibodies, says Burns. The company wants to use these antibodies to develop applications for in vitro diagnostics, says Danielle Miller, who manages the company's operations.

Monoclonal antibodies are produced by immortalized cells cultured outside of an animal and, theoretically, produce antibodies indefinitely, says Anita Bandrowski, a researcher at the University of California in San Diego. But it seems unclear whether the antibodies “are indeed exactly the same from year to year and decade to decade.”

Variability with monoclonal antibodies happens more frequently than the scientific community first assumed, says Arizona State University researcher Josh LaBaer. Bandrowski recommends that researchers include antibody lot numbers in their papers to help track reagent “drift” and address reproducibility of scientific findings.

For their large-scale effort to characterize all human proteins, scientists in the Human Protein Atlas (HPA), a Swedish research project funded by the Knut and Alice Wallenberg Foundation, need plenty of antibodies. When they evaluated the performance of over 5,000 antibodies, both monoclonals and polyclonals from different unnamed vendors, some companies' antibodies showed a 100% success rate in accurate protein detection. Others flunked the test with a 0% success rate1.

Antibody validation can include evaluating images stained with immunohistochemical methods. Credit: Cell Signaling Technology

The researchers used two antibody-testing techniques: western blotting, in which an antibody linked to a dye or other marker detects proteins separated by molecular weight; and immunohistochemistry, in which antibodies are used to stain cells or tissues for imaging to help identify where proteins are located. The average success rate was around 50%, says Emma Lundberg, who directs part of the HPA.

Other studies have produced similar numbers. “It is like a lottery, when you come to think of it,” she says. “Although this is awful, it is the reality, and one has to deal with it.” She advises colleagues to thoroughly evaluate application and validation data and other background information before purchasing any antibody.

Variability factor: application

“Antibodies are only as good as the applications to which they can be applied,” says LaBaer. Shopping scientists will want to see how well an antibody performs in a western blot and with immunoprecipitation or immunohistochemistry methods. But companies should show western blotting data not only on purified proteins but also on a protein expressed in a cell lysate, he says. This experiment shows that the antibody binds the protein but also that “it doesn't bind anything else in that lysate.”

Lundberg has tested over 20,000 antibodies for human proteins for HPA in a variety of assays. But she and her team find it “impossible” to predict which antibodies will work and which will not. They have concluded that they must perform application-specific validation of each antibody they use, she says.

For example, she says, a well-known centrosomal protein can stain “beautifully” in immunofluorescence. But a western blot with the same antibody will show many bands, all of the wrong size. “One explanation for this can be the high local concentration of the protein to the centrosomes in cells,” she says. “This high local concentration is then disrupted when the cells are lysed for the western blot.”

Henrik Wernérus, Atlas Antibodies' chief scientific officer, explains that HPA scientists realized they could not secure the antibodies they need from vendors. They also found that some antibodies did not work for immunohistochemistry. These results led the HPA to set up its own polyclonal antibody production and quality controls. Atlas Antibodies was spun out of HPA in 2006 to commercialize the project's polyclonal antibodies and, potentially, the diagnostic biomarkers that the project identifies.

As LaBaer explains, western blotting and immunohistochemistry results do not tell scientists whether the antibody is the right choice for a so-called ChIP-seq experiment, in which chromatin immunoprecipitation is followed by second-generation sequencing. ChIP-seq experiments help researchers to study proteins bound to DNA and to track their influence on gene expression.

In western blotting, antibodies probe denatured proteins and recognize a linear epitope, which is a continguous stretch of amino acids. But for ChIP-seq, scientists need to know whether the antibody can latch onto a conformational epitope, which comprises possibly noncontiguous amino acids on a folded protein. A protein's folds can hide a linear epitope. In that case, an antibody that works in a western blot might not work for ChIP-seq because the antibody does not 'see' the epitope to which it binds, says LaBaer.

Variability factor: validation

A critical step in antibody research is validation2. Some distributors sell antibodies made only by other companies, and many do not validate the antibodies they sell. The market barriers for vendors of research antibodies are low.

Commercially sourced reagents “are often of variable quality and all should be validated internally,” says Tom Bumol, vice president of Lilly Biotechnology. His team also takes a “trust and verify” approach to the scientific literature, he says. Antibody shoppers should be cautious shoppers.

Some antibody companies expand their own offerings with antibodies from others. For example, Abcam produces its own antibodies and also has around 300 suppliers. “At Abcam we've got experience with both, making sure we're doing things properly ourselves and making sure our suppliers are doing things properly,” says Miller.

The company tests antibodies in many applications, including immunofluorescence, immunohistochemistry, immunoprecipitation and flow cytometry, and in different species. “Once we have set the bar for the antibody, we will continue to batch-test all new batches in the same way,” says Burns. “If something doesn't pass those criteria, then we won't stock it.”

An antibody that works in a western blot might not work for ChIP-seq, says Josh LaBaer. Credit: J. LaBaer/ASU

Cell Signaling Technology (CST) says it manufactures and validates all antibodies it sells. All of Atlas Antibodies' polyclonal antibodies are manufactured using the same production process, which starts with selection of protein fragment antigens chosen for their sequence uniqueness when compared to other human proteins, says Wernérus. “This process, together with an advanced three-step affinity-based antibody purification and in-depth annotation of antibody staining in tissues and cells, makes us unique,” he says. The company makes its primary application data available on the HPA web portal, which includes immunohistochemistry, immunofluorescence and western blot images.

EMD Millipore's Xu says that antibody quality can be affected at many production stages. Attempts to speed the purification process, for example, can lead to an accumulation of contaminants that cross-react in experiments. Mohan adds that the company tests antibodies in applications with recombinant protein and also with cells.

Scientists should look for antibody validation that shows good performance in relevant biological systems with appropriate controls, says Craig Thompson, vice president of development and production at CST. His company performs validation, maintenance testing and production validation to ensure that lots remain consistent.

Smaller in-house production operations can offer better characterization of the antibodies they offer, says Craig Thompson. Credit: Cell Signaling Technology

Too many companies present too little validation or quality control data, which is a warning sign, LaBaer says. Scientists need these data to determine whether an antibody might be right for their experiment. “I would look for a western blot first,” says PhosphoSolutions' Browning. “If they don't have a western blot, I'd be very worried,” he says. The omission could mean the western blot result is “dirty” with numerous bands, which could imply specificity issues. Last month, his team began including information on the company's data sheets about the protocols in which antibodies work as well as those in which they do not.

Not only are antibody companies frequently bought and sold, they also change their business approach. Some companies transition from selling licensed antibodies to producing their own antibodies. Companies also sell to one another to keep their catalogs thick and increase sales. “Suddenly, every company is making their own as well as buying it from someone else,” says Browning. Because some companies sell antibodies they have not validated, researchers need to inquire about validation data.

Immunofluorescence can help to validate an antibody's specificity. Here, tubulin (stained orange) is shown in the cell's criss-crossing network of microtubules. Credit: Abcam

The larger, more “supermarket-style vendors” can have a broader selection of antibodies, but smaller in-house production operations can offer better characterization of the antibodies they offer, says CST's Thompson. Because his team produces and validates every antibody the company sells, “we are able to offer a level of insightful service to our customers that large vendors with outsourced antibody production lack.”

Scientists face many choices when hunting for the right antibody. “The market of antibodies is a jungle, and it is very easy to get lost,” says Lundberg. A search for antibodies against P53, a tumor-suppressor protein, yields an offering of over 7,000 antibodies. Choosing is hard even when the options are fewer.

Researchers can find around 600 antibodies to the enzyme tyrosine hydroxylase. The enzyme catalyzes reactions in the dopamine synthesis pathway, which appears to play a role in Parkinson's disease. This antibody is among the bestsellers in neuroscience and is one that PhosphoSolutions makes. Even in a market with dominant players and a wide selection, Browning sees room for “boutiques” such as his company.

Catering to the fast-growing neuroscience community, Browning focuses on phosphoantibodies, which help scientists study protein function. A regular antibody is similar to “an FBI agent sitting in a car outside a house, whereas a phosphoantibody is a camera inside the house,” he says. Browning originally set out using distributors to sell antibodies he manufactured. He then began selling them himself as well, building on relationships in the neuroscience community. As scientists shop with diligence, they should bear in mind how companies may change, which affects if and how validation is performed.

Customer affairs

The global market for research antibodies reached around $1.6 billion in 2011, says life science analyst Christi Bird with the consulting company Frost & Sullivan. The hunt for quality antibodies that work is a challenge for both suppliers and end users. “In a market where quality is the biggest complaint, there will always be companies entering the market to try to fix that,” says Bird. Customers return to companies whose antibodies work for their purposes, she says. Intensified customer service is part of how companies define quality.

At EMD Millipore, Mohan and his colleague Kevin Long ensure that the technical data sheets about the company's antibodies are up-to-date, and they interact with researchers as well as the company to strategize about new antibodies. They may license antibodies developed in research labs where they have customers, but they always test an antibody before adding it to the catalog, Mohan says. When a scientist mentions an antibody works in a type of tissue not tested at EMD, “we have to authenticate that data.”

Information not authenticated in-house is added to the antibody's internal file but not to the data sheet. Even when data are published in a journal, the team first tracks whether others have replicated the findings.

Emma Lundberg says scientists need application-specific validation for their antibodies.

Data sheets and online product information are updated continuously as the scientific knowledgebase grows, says CST's Thompson. The company fosters “rich relationships” with scientists, “keeping us in tune with current research trends and needs,” he says. If an antibody does not work, Mohan says his staff will help researchers troubleshoot experiments. He might tell customers about other feedback relating to that specific antibody. Conversations can be extensive and include data review, says Long. These exchanges can address variability relating to applications and protocols, the species of animal tissue being studied, or buffers. EMD has a technical support phone line dedicated to antibodies.

“Providing a replacement isn't always the solution to a problem,” says Randy Wetzel, who directs cytometry at CST. His team delves into the details of a researcher's experimental protocol. “If necessary, our scientists will try to replicate customer experiments in our lab and share our results with them so that we can provide detailed protocols that will enable them to continue their research.”

Other companies, such as Atlas Antibodies and Abcam, also help scientists troubleshoot experiments. Abcam says it wants feedback, even negative feedback. The company has a “no-quibble” approach to product exchanges or refunds if an antibody doesn't work, says Karen Halls, who addresses customer concerns for the company. Scientists struggling with an antibody might hesitate to tell the company. “When products don't work, we want to know,” she says. Abcam includes on its data sheets customer feedback and ways to optimize protocols, which often stem from customer input.

Antibodies are purified from animal serum at Atlas Antibodies in a three-step process that also uses the antigen to pull out a specific antibody. Credit: Atlas Antibodies

EMD Millipore can apply a “technical hold” to antibodies, which halts sales temporarily, Mohan says. “If it's not working in your hands, it is our problem too.” Quality-control groups at the company retest the antibody to resolve the issue.

Asking for special data

When it comes to particular validation data that may not be part of the product information, it is worth a scientist's time to ask whether it is available, both for commercial and nonprofit ventures (Box 1).

Lundberg says a new release of the HPA slated for December will show improved validation of antibodies, including data in which the target protein is downregulated with short interfering RNA in immunofluorescence and western blotting applications. This information shows whether the antibody binds the intended target protein and also whether the observed subcellular location is correct, she says.

To show the specificity of a class of the company's monoclonal antibodies, OriGene uses high-density protein microarray chips with more than 10,000 overexpressed proteins3. The tests address issues of cross-reactivity, in which antibodies react to several proteins in a sample, says Xuan Liu, who directs marketing at OriGene.

Cross-reactivity is often a coincidental experimental finding, she says. Scientists may know to disregard a signal of a cytosolic protein when they know their protein of interest is found in the nucleus. They can tell themselves, “Don't look at a cytosolic signal, just look at a nuclear signal,” she says. But cross-reactivity can hamper many studies, including those that seek to develop potential diagnostic tests for patients.

Although OriGene's specificity-testing method is promising for the clues it provides, several scientists interviewed by Nature Methods said that neither are its results easy to interpret nor do they readily transfer to gauging performance across applications.

Another bit of information about an antibody that can help scientists relates to the immunogen used to generate the antibody. Companies differ in their willingness to share that information.

Lundberg says the sequence is helpful when devising so-called high-sensitivity sandwich assays, for which researchers need to buy two antibodies targeting different epitopes of the same protein. But for companies, revealing the exact sequence can allow a competitor replicate its commercial product. “The companies do have an obligation to protect themselves,” says Miller. For commercial reasons, Abcam does not divulge the sequence on its web site. But “wherever possible,” it will tell a scientist the part of the protein that the antibody targets or the specific immunogen sequence. Browning at PhosphoSolutions tells his customers, for example, the 50-amino-acid stretch that an antibody will target.

At EMD Millipore, western blots are performed with robots. Credit: EMD Millipore

Mohan explains that if the antibody is one generated in-house at EMD Millipore and the scientist is about to publish with a reviewer requesting the immunogen sequence, then “Yes, certainly we'll oblige.” Atlas Antibodies' Wernérus believes the antigen sequence is “very important information for the customer,” so his company puts the antigen sequence for all products in its product data sheets. CST also discloses the complete antigen sequence and other information to the scientists, says Thompson.

Minding the clock

Scientists appreciate when companies allow them to return vials of antibodies, but it is not always a practical option. It is a “time and money sink” to buy antibodies for a few hundred dollars a vial and work out the details of an assay, only to discover that a second antibody aliquot might not be available or might not work, says LaBaer. “You've lost time, you've lost money. It can be very expensive fiddling with these antibodies.”

When Lundberg buys antibodies, she consults published references to save time in making her choice. “If I can see that other researchers have used a certain antibody and published the results in peer-reviewed journals, I feel more confident that the antibody is robust, detects the correct protein, and the likelihood is higher that it will work also in my hands.” She would like companies to offer application-specific validation accompanied by primary data as well as protocol recommendations to obtain results and basically learn “how to make this antibody work in your hands.”

If a graduate student tries and tries again to reproduce a result from a paper that applied an antibody-based assay, but the student keeps failing, he or she does not tend to doubt the paper's result, says Browning. The student “doubts himself or herself.” But consideration of the many variables that can lead to failed antibody assays can be a more productive approach to troubleshooting. “You've got to look at these things or else you're going to be chasing the same antibody that doesn't work through 12 different companies.”