For the best experience on the Abcam website please upgrade to a modern browser such as Google Chrome
To ensure that our antibodies are specific and provide reproducible results, we have recently introduced knockout (KO) validation at scale. Using human KO* cell lines, we are developing a growing range of KO-validated antibodies, including many recombinant monoclonal antibodies.
*We validate antibodies using haploid KO cell lines.
Dr Alejandra Solache joined Abcam in 2013 as Head of Reagents, Product Development and Manufacturing globally. She is responsible for managing the output of the Abcam Cambridge, Hangzhou and Bristol laboratories, specifically new product development, core reagents and R&D. She also plays a key role in developing our innovation strategy.
Prior to joining Abcam she held various positions at EMD-Millipore, latterly as R&D Director, leading the Antibody and Assay Development teams. She gained expertise in immunology, cell signaling and cell biology through postdoctoral fellowships at UCSF and the Trudeau Institute.
Hi everyone, and thank you for being here with us. I am really glad to be able to share with you one of our most exciting initiatives in Abcam, which is the incorporation of knockout technology in our antibody validation. This effort is quite important for us, and we believe also for the scientific community, since it contributes to increasing industry standards for antibody validation; and ultimately to increase the reproducibility of research. The aim of this presentation today is to give you an overview of some of the tools available to address the specificity, and the reproducibility issues of research reagents. In particular, we will discuss Abcam's initiatives to tackle them, and, importantly, we also will discuss how can we all collectively contribute to solve these issues. Why are we all concerned about these, and why are we in Abcam paying so much attention to this? As you probably are all aware, there is an issue with reproducibility in biomedical research. In 2012, a study led by Dr Begley at Amgen, was only able to reproduce six out of fifty-three cancer studies, and this highlighted a big issue.
Some of the red flags that he found out where for this is a producible research, included the selective presentation of positive data, the lack of blinded studies or controls, and particularly lack of reagent validation. Lack of reagent validation has been already highlighted in multiple reports recently in the literature and publications; and this is no news for anyone. Something that is making things worse is the fact that there is poor reporting in the literature, when people report how they did their experiments and what reagents they have used. These studies don't have enough information containing catalogue number, vendors, and all the information related to the reagents that have been used in the research; which makes it very difficult for other researchers to follow-up and try to reproduce that research. Although the reproducibility and reagent validation have always been on the agenda for science, we actually, we may wonder why these issues are arising more now, and one important factor is that we have wider access to advanced technologies.
There are many new tools in life sciences, such as mass spectrometry, high throughput arrays. We have now the ability to knockout and knockdown the genes and alter, and we could alter protein expression in cells, or animals. This is becoming - these techniques are becoming a lot more accessible to everyone, and therefore have created opportunities to better assess reagent specificity and reliability; and allow us to be more precise than ever before. So it's not surprising that the widespread access to these new technologies and improved validation methods, has uncovered issues with the most common reagents, in particular, with antibodies. These issues have been reported as well by several papers, and opinion pieces in Nature and Science, and many other scientific journals; and they have been implicating antibodies as an important factor contributing to the wider problem of reproducibility in research. This is highlighting a problem that has to be tackled collectively, and, in fact, there are several organizations and initiatives such as the Global Biological Standards Institute, who are advocating for better antibody standards.
As part of our commitment to improve quality standards in Abcam, we recognize that it is important that reagent suppliers do more to tackle these issues. This means offering support to different industry initiatives, improving manufacturing processes and adopting more robust validation technologies as are developed. Specificity and reproducibility need to be addressed, and they need to be addressed not just by antibody producers, but also by, it should be done by a wider effort in the scientific community. Researchers have a responsibility, and also scientific journal editors have also a role to play here. This is something that we want to address today in this webinar, and in today's presentation we will discuss how antibody types and their manufacturing can affect consistency, including the benefits of using recombinant monoclonals. We will review some of the most advanced methods used in antibody validation, and we will also describe our new initiative of antibody validation by knockout models. Finally, we will discuss the steps that both researchers and suppliers can take to promote reproducible research through the use of high-quality antibody reagents, and transparent reporting.
Starting with what can antibody suppliers and companies that produce reagents do, and the most important step that antibody producers can take to ensure consistency is to adopt high standards in manufacturing practices; and make sure that they have robust validation technologies. As we highlighted already, the performance of an antibody can be closely linked to how they are manufactured. Antibodies are available, as you know, in two forms: polyclonal and monoclonal, and both types of antibodies have their benefits and limitations, depending on their intended application. Polyclonal antibodies; the process of generating these antibodies, they generate a large variety of antibodies to different epitopes on one particular antigen, which can be advantageous for some applications, but also this can be a significant source of background and cross-activity. The complexity of polyclonals is further complicated by their finite nature, and once a batch has been depleted you need to initiate an immunization; and, therefore, the antibody performance would vary significantly between batch-to-batch. A polyclonal may be comparatively inexpensive in the short-term; maintaining consistency can become quite labor-intensive.
So now I'm moving on to monoclonals. Monoclonal antibodies recognize a single epitope, and they display higher specificity as monoclonals are manufactured from hybridoma lines. If properly maintained they should not have batch-to-batch variability. However, in rare cases, may have stability issues and lose epitope recognition. Once a monoclonal has been identified, it can continue to be produced by culturing through a hybridoma cell line; and, alternatively, the antibodies may be produced as recombinant antibodies. Progenitor recombinant antibodies, the gene is encoding the antibody binding region. The heavy and light chains of the antibodies can be cloned into a vector that will permit the expression in a more robust host cell line, which could be CHO cell line or a HEK 293 cell line, for example. From here the resulting recombinant antibody can be isolated and purified just as if they had been produced through a hybridoma. But once the antibodies sequence, recombinant antibodies are not subject to loss of activity, which could be caused by a sequence drift or other issues that could be affecting standard hybridomas when they are poorly maintained.
Once an antibody has been cloned the recombinant antibody production is scalable, and since it is a full in vitro system you are able to produce consistently antibodies that are identical from batch-to-batch. Furthermore, the inherent nature of the recombinant antibodies makes it easier to improve both antibody specificity and sensitivity, if it is required. So along with the benefits of recombinant antibodies, recombinant rabbit monoclonals, or RabMAb antibodies offer additional advantages, since they are generated in rabbit they can be developed to a diversity of epitopes. This is due to the ability of the rabbit's immune system to see multiple very small differences, due to a larger B-cell repertoire. They tend to be higher, have higher sensitivity than mouse monoclonals and the reactivity across the species can be better because they can actually work, for example, in rodent models. In Abcam, we have about 90 per cent of our rabbit monoclonals are made recombinantly, and therefore they will also have all the advantages that we have just described for recombinant antibodies.
Recombinant antibodies may require a high degree of technical expertise, and can be comparatively expensive to produce, or in the past where actually comparatively expensive to produce to a standard hybridoma. However, technology in this area is developing rapidly and the costs and timelines of production are actually coming down. The advent of Phage Display and other display technologies have allowed monoclonal antibody discovery processes to take place with minimal, or no animal involvement. Since no immunization steps are required, comprehensive Phage antibody libraries permit the targeting of antigens, which could potentially be toxic, or may have low antigenicity which will be difficult to generate through standard hybridoma technologies. We are extremely excited with that addition of the AxioMx technology to Abcam, because it provides a unique fast and efficient platform to produce recombinant monoclonal antibodies, which complement Abcam's existing production capabilities.
We have now discussed a lot to do with manufacturing practices, and the different types of antibodies and their pros and cons. Selecting a polyclonal or a standard monoclonal, or a recombinant monoclonal for your own research, will have an impact on your experimental results. But it is important to be aware that each of those may have benefits and limitations, and some of them may be suited for certain applications. However, regardless of the type of antibody or the type of production, all antibodies require a stringent validation and this is what we will be discussing now. Antibody validation has been evolving since antibodies were first used as reagents. Western blot, immunoprecipitations, immunohistochemistry, immunofluorescence and many other techniques have been available for over 30 years. All of them have advantages and disadvantages, and depending on the type of antibody that you want to obtain, or the research that you are carrying out, you will try to use an antibody that has been validated in a particular application.
More recently, more advanced techniques like knockout, knockdown or peptide array, or mass spectrometry have been more widely available, and are becoming increasingly popular. However, there is no single solution to antibody validation, and some methods are exceedingly powerful, but come with major research considerations that some labs may actually struggle to meet. You need to find the most efficient and appropriate solution for your own projects, for your own experiments, and researchers need to weigh up the cost and time restraints; and, in some cases, even the level of technical expertise or involvement of some of these techniques. So let's start with mass spectrometry. Mass spectrometry is typically used for lowering IP, and where antibodies have links to beads just to pull a target of interest from a complex solution. The recovered proteins are then digested, and the protein fragments are analyzed. This technique is very amenable to high throughput formats. However, it may come with some issues like low sensitivity, do actually sometimes, could pull down off-target proteins that have been complexed with your target. The antibody needs to be able to immunoprecipitate the target, and also it requires access to a mass spectrometer.
Moving on to protein arrays. Protein arrays are quite high throughput, and they can compare the antibody, the binding of an antibody to a target peptide and also compared to the binding to off-target peptides. This is a very good technique when validating antibodies to specific protein modifications; we particularly use these technologies to validate our antibodies to different histone modifications, and also to phosphorylated proteins. It allows screening of large numbers of expressed proteins as well, for example, when you are not using peptides. It is very general, and it can provide a lot of information about specificity. The only issue is that it's an in vitro assay and that it doesn't really have the cellular content in here. There are many models of knockdown and knockout technologies. In the case of the knockdown methods these, in general, use a small interfering RNA and this is used to reduce or deplete mRNA of a protein of interest. Therefore, you can compare the specificity of your antibody to your target when this protein is being downregulated in a particular cell line.
This is a technology that has been widely used, and several companies have actually used in these technologies, and it is quite good. However, sometimes the effects could be transient: you are sometimes not able to eliminate the protein of interest completely. Actually, the experiments are quite difficult to optimize, often requiring several siRNA sequences. Now moving on to the knockout methods. The knockout model allows scientists to understand the function of a particular gene in a whole animal, or cellular models that lack the protein of interest. In the context of antibody screening or antibody characterization, knockout models provide an excellent standard for antibody validation, as they represent a true negative control. As such, they are often viewed as the gold standard for antibody validation. In recent years, the use of CRISPR-Cas 9 genome editing has been able to achieve this knockout in a larger scale, and it's actually becoming quite an important technique widely used for many people.
Knockout cell line controls won't always be appropriate though, and as essential genes cannot be knocked out, anyone that is studying a gene that is vital to cell survival, for example, simply will not have the option to create a negative control using knockout technology. In the case of animal knockout models, obviously, they will require access to animal facilities and expertise. In general, these types of technologies can be quite time-consuming and expensive to generate. Unfortunately, there isn't a one size fits all approach to antibody validation, as suppliers need to make use of the most appropriate available technique when it comes to validating their antibodies. Whether it is mass spec, or peptide arrays, knockout, knockdowns, either of them can be quite powerful, ideally, using combinations with other technologies. What is very important is that transparency of the validation methodologies is there, so that researchers can be confident of the products and that they have sufficient quality, and enough validation for them to be able to trust that particular antibody.
In Abcam, we have been using several knockout models for quite a while already to validate antibodies, and we did these through acquiring some clones from collaborating with researchers who have some types of knockout models, or purchasing when they were available, or even developing our own. However, the scale that we could reach doing it this way was extremely small, and one of the great advancements in science has been the discovery of gene editing by certain molecules. As we discussed, CRISPR-Cas 9 has quickly become greatly applied and a well-trusted technology for precise and efficient genome editing. This technology has presented us with a great opportunity to validate the specificity of our antibodies, at an unprecedented scale.
How are we doing this validation? How are we undergoing this testing? I'm wanting to say that this knockout validation is being done on top of other applications that we have been using already for the particular antibodies. We then have selected several antibodies to a particular target. We take several endogenous - several cell lines where the protein of interest is endogenously expressed. We also test wild-type and knockout HAP1 cell lines. We would test either by western blot or by ICC, and some of the expected results that we see are, we have three different types of results. First of all, the ideal scenario is when the target with the antibody is specific for the target of interest. In this particular case, as you can see on the first western blot, we have an antibody that recognizes the PKC Alpha protein in the wild-type HAP1 cell line. In two cell lines that we have used as positive cell lines, but the band of interest disappears in the knockout cell line. This is great, this is what we want to see, so what we will do with this data is to upload it into a website, and we will add our seal of knockout validated antibody.
The second case is the case where the antibody of interest already we knew that had some cross-activity, but we wanted to identify if the bands had the correct molecular weight expected for that particular protein, in this case RIP, actually will go away to highlight if this particular antibody recognizes a RIP or not in the context of this knockout. What we see here is that the band of interest disappears in the knockout, and, however, the other band that is off-target binding stays there. These antibodies can still be useful, particularly when there are not as many antibodies to the target, and we believe that it's better to have an antibody for which knockout validation has been done. At least you know what you are getting, and that having something else that may be completely recognizing a totally different protein. The worst case scenario, obviously, is when you have a protein that recognizes something else completely different, and that just thought was the protein, your protein of interest, because the protein of interest has the same molecule and weight. However, what we find here is that the band of interest doesn't go away in the knockout cell line.
This is great, because now with this technology we are able to detect that this antibody was not specific, and we don't need to keep it in the catalogue. So we've removed the product from the catalogue and we notified the people that have purchased that particular antibody, and we also offered, ideally, one or two other antibodies that have been validated in knockout of the same target. In the case of the antibodies that don't work in western blot, there are many antibodies that are fantastic antibodies, but not necessarily work in western blot. We are doing the same testing using immunocytochemistry in this particular case, and the antibody, again, is Ki-67 was tested in the wild-type HAP cell line, and in the Ki-67 knockout HAP1. What we can see here is that in the fourth panel at the bottom, we see that the signal disappears in the knockout cell line. This is what we are doing with all of our - all the HAP1 knockout cell lines, but we are also doing a lot of validation still through our collaborations, and this contributes greatly to our internal efforts. We would like to encourage people to continue doing this, and if you have any knockouts available and want to share your data, we would love to hear from you.
In this particular case, this is data for an antibody against PD-L1, and our collaborators have sent us some beautiful data to show that the antibody detects PD-L1 only on the parental cell line, but not in the PD-L1 knockout cell line. We have discussed now all of the efforts that we are doing on knockout validation, and so far we have around 400 primary antibodies against several different targets in many areas of research. We are doing this, because the research of our customers is very important to us, and we want to take every measure to ensure that our antibodies are reproducible and robust as possible. So another thing that we believe that us and other antibody providers should be doing, is being very transparent and offering the ability for customers to provide feedback. We have been doing this for many, many years, since Abcam actually started. We have this Abreview program that allows customers to freely give positive and negative feedback, and to share their protocols. So we believe this is a really good thing for you as a researcher, because it can help you choose the best antibody. You don't necessarily just need to trust the data that the company is providing, but you can actually also see data that your peers and colleagues are providing.
The last thing is to provide enough data and information to help reproducible research. Obviously, companies need to provide detailed protocols for the application tested, and highlight applications where the particular antibody works, and, ideally, where it doesn't work, as well as any other application that can be relevant to that particular product. We have talked a lot about what companies can do, and what suppliers should be doing, but while defining the initial quality of the antibody that task that falls to the supplier, researchers at the bench have also a vital part to play in promoting reproducibility. It is important to understand that even a highly validated antibody can yield unreliable results, if the experimental protocols and controls are not in place. So, obviously, you all know that. There are other things that you as a researcher could be doing in order to facilitate reproducibility of research, and one of them is to check that the applications and protocols that you want to test the particular product is what the researcher, sorry, the supplier has recommended.
Publish with detailed antibody information or reagent information using vendor, catalogue number, and all of the information that will be useful for somebody else to be able to reproduce your research. Also share your validation data and give feedback. It's not only sites like us who have - give you the ability to share your data, but there are multiple sites out there: Antibodypedia and other sites that actually let you share data, which is very important. The last - another component of the community who has to be engaged in these concerns about reproducibility, are the publishing companies. Journal editors need to be supporting researchers by asking for detailed information, such as catalogue number, vendor and all of the information required from a particular reagent so that, again, the research can be reproduced as closely as possible. Ideally, the company should be providing information about where a particular reagent has been also cited in the literature.
Moving forward, improving the reproducibility of research across life sciences, there needs to be a partnership between all parties involved, and the suppliers, the researchers and publishers all need to do their part. As we discussed, suppliers need to go to even greater lengths to ensure that their antibodies are sufficiently validated, that there is enough testing data, and this is the first step and the most vital, as it allows researchers to review their own research, and when they are confident that the antibody has been properly validated. But also, as we have discussed, there is responsibility of people like you in the lab to take care that your experiments are robust, and that you publish with enough data, and for the scientific journals to enforce complete reagent information as well. We believe that this partnership is not only possible, but is necessary and we believe that this is actually already happening, and there are, as we say, many consortiums and a lot of different forums where people are getting together to discuss how we can improve this.
In summary today, I want to - we have discussed what antibody producers need to adopt. That antibody producers need to adopt high standards to manufacture and validate their antibodies with a combination of characterization methods. I have highlighted our commitment to continue support, the support of scientific researchers by validating hundreds of antibodies per year using knockout technologies. Also by introducing more recombinant antibodies, and also by evolving our validation strategies as new tools become available. This improvement needs to be a partnership between the antibody producers, the researchers and the journal publishers. We hope this presentation encouraged you - all of you involved in life sciences to take what steps they can to contribute towards the generation of consistent and high quality research. Thank you for your attention, and I would really like to hear any feedback or anything else that you think we should be doing as a community to help with reproducibility. Thank you.
Will you be knockout testing all of your antibodies?
Ideally, we would like to test all of our antibodies, however, we have a lot of antibodies and we will try to test as many as we can. Obviously, as discussed already, some of these - some proteins have not actually expressed in HAP, in this cell line. Also, some proteins aren't necessary for the survival of the cell, so it won't be possible to test them with this methodology, but we will try to use a different methodology for those cases.
Thank you for that. As a follow-up, we have a similar question. Will you be knockout testing all of your antibodies to a specific target?
Well, it is the same thing, depending on the target we will be testing the antibodies if we have the knockout cell line available, or if there is a knockout mouse available, so it will really depend on the target.
On a similar note, again, how are you choosing the targets to test in knockout cell lines?
We are selecting this target based on (1) availability of the knockout cell line, and (2) based how widely used it is for researchers, and then it will be this target to be validated in the research community.
Thank you for that. So moving on a little bit, how can a scientist avoid using the wrong reagent with knockout technology?
So knockout technology is only one part of the quality and the validation, and Abcam, we validate antibodies using many complementary techniques, as we've discussed, like western blot, ICC, ITC, and knockout and knockdown can be also considered.
Thank you. Regarding the experiments used for the knockout and what is published on the datasheets, what is the number of repeat experiments - western blot or immunocytochemistry - that you do for validating the antibodies prior to including this information on the datasheets?
We will do many experiments and probably in excess of five different experiments, to include data in the datasheet.
Wonderful! So when you use these knockout cell lines for antibody validation, do we generally have any negative results, and can you share some of these negative results? How are these negative results interpreted?
Yes, absolutely, and I did show one of those negative results and I can go back and show it to you in here. Just a second. We have one of the examples that we showed was a negative result, so this is in the data, is in number three that's off target actually an antibody that was negative in, or, basically, actually show a band in the knockout where we would expect not to have the band of interest. So, in those cases, we removed that particular antibody from the website.
Okay, that's good to know. Regarding the rabbit monoclonal antibodies, we see that some are not validated for flow cytometry, is there any reason for this?
Yeah, definitely. I think - well, the nature of the antibodies means that not all of the antibodies are going to work for one particular application, or another one. We often have antibodies that work in ICC and flow, and we use ICC for knockout validation, but not necessarily, or not all antibodies necessarily will work in flow cytometry. It really is quite variable.
So regarding different applications, would you perform batch testing of antibodies on knockout samples?
We will provide batch testing for only polyclonal antibodies in knockout samples. Our monoclonal antibody, due to the nature of the lack of batch-to-batch variability, we believe that when we have validated the actual clone, we won't need to be validating that clone again every time we make a new batch; but we will be validating that particular monoclonal in the other applications that we always do.
Thank you for that. Similarly, are you only validating products made at Abcam, or are you also validating outsourced products?
We are validating not only Abcam products, we are validating the antibodies that we consider that are necessary or useful, that will be useful for the customer. Or we are actually validating a lot of Abcam internal antibodies, because we have the actual clones, and we are trying to incorporate the production and the standard validation of the clones; but we are actually going back to validate some non-Abcam products as well.
So it sounds like validating for both Abcam, as well as outsourced products, which is wonderful! So when performing these validations with knockdowns, generally how many cell lines are used? This question is regards to the consideration that the pattern can be different from cell line to cell line, depending on your target?
Yes, so we are using one cell line, knockout cell line, however, we are also using different other cell lines, which express the protein endogenously. So we - the knockout is your extra, an extra piece of information, and when we have had some inconclusive information we have tried to get access to a different knockout cell type, or a different knockout model as well.
So it sounds like there's a lot of research going into the different reagents that are being used. On that note, how do you check that the protein is really knocked out in the cell lines that are being used?
Quality control checks are carried out. We also test multiple antibodies to the same target, so if in a particular case we had doubts that the cell line is not expressing the protein of interest, we will be checking with multiple different antibodies.
This actually is a good segway into a follow-up question that just came in. Some antibodies detect a protein band of different size than are expected, do you think that these antibodies are specific to the target protein, or is there potentially another explanation for this?
That depends on many things, for example, things like glycosylation and other post-translational modifications may alter the size of the protein. Actually, it could also be, depending on this cell type and any type of modifications that could happen endogenously.
So it sounds like it's a case-by-case situation to be looked at?
Regarding the phage display, there's actually a two-part question. So what is the size and diversity of the phage display library, and what is displayed in the phage display library?
The phage display library has in excess of ten to the nine clones, and the single chain they are displayed, but then these are converted to IgGs for product market.
Fantastic! We have one final follow-up question. So during the talk you mentioned that use a leukemic cell line for this knockout validation, do you feel that this gives suitable coverage of targets? For example, what about in the neuroscience area?
Well, this is a leukemic cell line, but it doesn't mean that it will then express protein. So if the particular proteins, regardless of being neuroscience or not, this cell line could actually express the same proteins. So if the protein is expressed in these HAP1 cell lines, it will be possible to use it. So these cell lines are transformed, like many other transformed cell lines express many other targets.