Dr. Douglas Auld is a Senior Investigator within the Department of Chemical Biology and Therapeutics at Novartis. His laboratory performs research in assay design to explore biology and enable drug discovery. He also manages a laboratory at Novartis that provides resources and expertise in high-throughput experimentation to support academic collaborations. He is the author of over 100 peer-reviewed scientific publications and is a founding editor and contributor to the Assay Guidance Manual available as an e-book on the NCBI website, which is aimed at guiding researchers in assay development and lead discovery approaches. He is also a literature editor and contributor to Assay and Drug Development Technologies.
When did you first develop an interest in science? Were there certain individuals who had a particularly strong influence on your chosen career path?
I was raised in the Boston area, which is a mecca of science and engineering. I grew up during the Apollo program, Moon-shots, and my next-door neighbor was a graduate of MIT who was designing life-support systems for the astronauts. My father is a biochemist and was a professor at Harvard Medical School. As such, there was no shortage of scientists and engineers surrounding me during my upbringing. I was fascinated by what they knew and what they could do with their knowledge. From an early age, I became interested in pursuing a career in science.
After graduating from college, I worked for 2 years as a research associate under Dr. Wolfgang Maret within the Biophysics Research Laboratory in Peter Bent Brigham Hospital. I isolated sorbitol dehydrogenase from human liver extracts and characterized this enzyme. Wolfgang was a great mentor to me, teaching some of the skills in protein chemistry and enzymology and designing experiments to answer basic research questions. My first article came from these efforts and was published in Biochemistry in 1988.
My graduate work was with Prof. Gary Pielak at UNC in the Department of Chemistry where I studied cytochrome c as a model of protein folding using two-dimensional nuclear magnetic resonance to characterize mutants of this protein. This work explored the role of weakly polar interactions in protein folding. Gary's laboratory is interested in the fundamental principles that govern protein folding, but he always told his students that they were learning much more than that. Gary taught me how to do rigorous science and to think about the proper controls so that the results could be interpreted conclusively (either positive or negative results). Later, while working at Pharmacopeia, my education in this area was applied to discover a potent inhibitor of nitric oxide synthase, which operates by inhibiting the dimerization of this enzyme in cells. I was one of Gary's first graduate students and, in fact, he had not yet moved to UNC from his position at the University of Oxford when I started to apply to work in his laboratory. Consequently, the first graduate students had to help furnish the laboratory and test new equipment needed for the work. Although this was challenging at times, working in a new laboratory gave it an exciting start-up environment.
I did my postdoctoral training at MIT with Prof. Paul Schimmel where I studied isoleucyl-tRNA synthetase (IleuRS) and the molecular recognition of RNA by tRNA synthetases. I was able to demonstrate that a single amino acid swap in a helix loop peptide of IleuRS could switch the anticodon recognition between IleuRS and the related MetRS enzyme, which operationally mimicked work where a single nucleotide in the cognate tRNAs could switch recognition between the two enzymes. Paul tried to get his postdocs to think about the essence of a scientific problem, something which had been impressed on him by one of his mentors—P.J. Flory, a Nobel Prize-winning polymer chemist who pioneered the science underlying the behavior of macromolecules. Paul taught me how to take complex experimental data and distill this down to the key points or a figure that illustrated the problem being addressed.
You have worked at various biotech and pharmaceutical companies and at the NIH. How do the scientific environments differ between these organizations? Did you need to make adjustments to move between them?
This is an interesting question and my answer to it has evolved over the last 20 years. When I was completing my postdoc, I began to consider both faculty positions in academia and industrial jobs. I talked to some people working in industry and the message was that either job could be fulfilling for a scientist, but once you made the choice, you could never switch back. About 9 years later, after working at the biotech company, Pharmacopeia, the landscape was changing. My call to NIH was not unlike that received by a few other industrial scientists who were being sought out by academic laboratories due to their expertise in lead discovery. Today there is a good deal of intermixing, with many scientists switching between public and private sector jobs as well as collaborations between the two.
I think the differences between biotech and pharmaceutical companies really come down to the people you work with. I found biotech to be enjoyable largely because I started when the company was very young. There were many motivated talented scientists, so the experience was similar to my early days of graduate school where we were all trying to build something and working together to achieve this. Such an environment can also occur within large pharmaceutical companies, and I have witnessed this when a new project or initiative is started inside a company. In my transition from NIH to Novartis, I found that my new role shared many similarities to the one I had at NIH, in that I found a very research-oriented environment and met experts in disease areas who needed guidance on lead discovery. One difference between biotech and large pharma is that in biotech, your success or failures can have a very perceptible impact on the future direction of the company. This is true in larger companies as well, but the feedback loop is longer and more complicated to navigate through. A real benefit of large pharmaceutical companies is that there are generally many more resources to draw from to drive drug development. In addition, large companies have experts in every area of drug discovery, allowing teams to obtain proper guidance and great learning opportunities.
You were one of the founding scientists at the NIH Chemical Genomics Center and played a leading role there for a number of years. What were some of the challenges and successes in those early years in the growth of drug discovery in the nonprofit sector?
Helping to bring the NIH Chemical Genomics Center (NCGC) from a whisper to a roar was an experience I will always cherish. This was an exciting time at NIH; the Molecular Libraries Initiative (MLI) was being launched as part of the NIH Road Map, with the aim of establishing a national network of chemical probe development laboratories. The NCGC started out at the National Human Genome Research Institute on the main campus of NIH located in Bethesda, MD, where we had some temporary space. The first challenge was finding laboratory space that would scale to accommodate the needs of the center as it grew. The space had to be suitable to house both the automated systems for screening and synthetic chemistry efforts. The Human Genome Science (HGS) building in Rockville, MD, had been recently vacated and a bid was made to begin renting space in this building. The HGS building contained beautiful modern laboratory space with glass walls and an award-winning atrium; however, we needed to renovate the laboratory space in three key areas. A large open room was needed to house the Kalypsys automated screening system, which was being constructed at GNF in San Diego. A significant amount of chemical fume hoods had to be added to the HGS building to support synthetic chemistry efforts. Finally, we wanted to design an open flexible laboratory space, eliminating fixed benches and replacing these with mobile work spaces, which required adding in overhead utilities. On my first day at NIH, I was put in charge of representing our scientific needs to the building committee and started working with architects on the plan. This required a significant effort and time commitment as the building committee meetings grew to over 30 people and not necessarily the same people each week. There were several times where a crisis would arise and our plans were met with adversity. Francis Collins once said to the committee that the NIH Roadmap was at the center of NIH, the MLI was at the center of the Roadmap, and that the NCGC was at the center of the MLI—so we were at the “center of the center of the center.” As such, there were many people watching us and expectations were high.
I learned an important lesson from this experience related to focus and clarity in the midst of turmoil. I remember a day early on in the building process where the foreman from Whiting-Turner construction introduced himself has “sticks-n-bricks.” He was the guy who would show up each morning and make sure that the work got done, so I realized if I wanted to know how likely something would be completed, I should talk to him. Because I came to know what he knew, I was much more confident and informed at the committee meetings. Now, when faced with a difficult situation with many voices in the air, I try to seek out the “sticks-n-bricks” of the issue to find a solution. When I performed the final walk-through of the new NCGC, it was a real relief. The laboratories were pristine and represented a world of opportunity.
My real job was choosing what projects we should focus on first at the NCGC. Being at NIH, there were plenty of investigators with ideas, which they wanted to test out, and most of the first year was spent meeting with investigators to begin to develop research plans that could be implemented at the center. There were fantastic opportunities, including the ability to work on rare and neglected diseases, which oftentimes involved new target classes and biology requiring innovative assay designs. There was also the opportunity to work with Nobel Laureates like Marshall Nirenberg who showed a large interest in the NCGC and became one of our more productive collaborators. A challenge was shepherding these expert investigators on what was possible to do at scale and how to think about what one needs to do after identifying primary hits. We realized that there was a knowledge gap between understanding how to run benchtop assays versus performing an assay in high throughput.
About this time, we were contacted by some colleagues at Eli Lilly who wanted to share a document with NIH, known as their Quantitative Biology Manual, which they had used internally for more than 10 years to educate and guide researchers in the processes and best practices in lead discovery. This became a tremendous resource to help educate investigators in the practices of assay and lead optimization. We began to expand on the existing chapters to further capture the tribal knowledge of the field inside the manual. The manual is now known as the Assay Guidance Manual (AGM) and is freely available on the NCBI Bookshelf (www.ncbi.nlm.nih.gov/books/NBK53196/). I also started a series of annual meetings called aDREAM (Development of Robust Experimental Assay Methods) aimed at encouraging researchers to share knowledge and lessons learned. Some of the topics and discussions from the aDREAM conferences would later become new chapters in the AGM. I have a strong belief that collaboration and sharing knowledge with open discussions on common problems for which the industry is currently struggling can lead to solutions, which benefit the whole group. I also think that this knowledge needs to be captured and made available so that the next generation of scientists can learn from both our successes and failures. Much of this knowledge is often lost when people leave a company or a company closes.
Later, my research on pyruvate kinase expanded to some very successful collaborations with Lewis Cantley's laboratory (who was then at Harvard Medical School) and the Structural Genomics Consortium. My laboratory identified two series of compounds, which specifically activated the M2 isoform of human pyruvate kinase (PKM2) by promoting active tetramer formation from inactive dimers. Outside of liver and red blood cells, most cells express either the PKM1 or PKM2 isoforms. The PKM1 enzyme is found in normal differentiated cells, but all proliferating cells switch expression to PKM2, including all cancer cell lines and tumors. The difference between the two isoforms lies in the allosteric regulation as PKM1 is constitutively active, while PKM2 is allosterically activated by fructose-bisphosphate. Furthermore, PKM2 is inhibited by binding to peptides containing phosphorylated tyrosine as occurs during growth factor signaling. One hypothesis is that downregulation of PKM2 due to growth factor signaling allows glycolytic carbon skeletons to be shunted toward the production of amino acids, lipids, and nucleic acids, which are required to support proliferating cells. As well, it is known that replacement of PKM2 with PKM1 reduces the Warburg effect found in cancer cells. Therefore, activators of PKM2 might restore normal cell metabolism. Our PKM2 activators were later shown to inhibit mouse xenograph tumor growth and were taken in by some biotech companies for testing. My laboratory also began to work on other glycolytic enzymes for the purpose of targeting neglected diseases. Assays were developed, which were aimed at finding inhibitors for several enzymes derived from pathogens involved with African sleeping sickness, leishmaniasis, and Chagas' disease.
Many aspects of the Molecular Libraries Probe Production Network (MLPCN), which the NCGC seeded and became a part of, were successful. The MLPCN brought industrial-scale technologies to the hands of academic scientists who were experts in an area of biology, but lacked knowledge in the area of lead discovery. This melding of minds, earlier involved with separate public and private research efforts, resulted in the production of more than 370 chemical probes, many against new targets and areas of biology. However, an intangible benefit was the education of academic scientists and postdoctoral associates in lead discovery. This occurred through both the hiring of young scientists within the MLPCN and NIH-funded grant awards, which provided opportunities for postdoctoral associates to apply for resources within the MLPCN. The NCGC has now become part of the National Center for Advancing Translational Sciences (1 of 27 Institutes and Centers at the NIH) with wide-reaching effects benefiting the improvement of human health as well as providing training to young scientists interested in biomedical research.
One of the hallmarks of high-throughput screening at the NCGC has been the quantitative HTS approach in which primary screens are conducted at multiple concentrations of test compound. Has this proved to be particularly useful with certain screening technologies or classes of target? Has this strategy been adopted in many other laboratories?
In concept, quantitative high-throughput screening (qHTS) is a simple procedure—a compound titration, but performed in an interplate manner. However, implementation of qHTS requires novel workflows in compound management and sample tracking, as well as informatics for data analysis. Once established, the ability to implement qHTS in a massively parallel manner, fit concentration–response curves (CRCs) to millions of data points, and rapidly classify CRCs can lead to truly powerful datasets.
The idea for qHTS came from discussions with Jim Inglese, who recruited me to the NCGC in 2004. Jim was a colleague of mine when I was at the combinatorial chemistry company, Pharmacopeia, located in Cranbury, NJ. He later went to Merck before taking the position as Deputy Directory of the NCGC at NIH. Pharmacopeia's screening strategy involved screening combinatorial libraries as mixtures and following up hits as single compounds at multiple equivalents of the active sublibrary. This approach yielded structure-activity-relationships (SAR) from the screen as one could view which particular R-groups were enriched around a scaffold, similar to how molecular biologists perform mutant screening. In this way, one obtained a rich dataset from the primary screen, which could be used to guide follow-up medicinal chemistry efforts. One caveat was that we screened the mixtures at one concentration, and if an active sublibrary hit strongly, many of the wells would be strongly inhibited, which could lead to very high hit rates. Jim's tenure at Merck demonstrated to him the conundrum of screening at one concentration and having to reduce the hit list to something manageable to validate compounds using full CRCs, often leading teams to question what might have been missed or to take time-consuming iterative approaches. This thinking fueled the desire at the NCGC to include concentration–response data at the primary screen stage and led to the development of the titration-based qHTS screening paradigm. Having no legacy issues, there was freedom to design both compound management and informatics pipelines to support the qHTS method. As such, we developed a process to create a titrated archive of the NIH library, which ultimately increased to ∼300K compounds.
While at the NCGC, I had to decide on an assay to first test out the qHTS method. A well-validated robust assay was the ideal, but I did not want to screen popular target classes such as protein kinases or G-protein coupled receptors (GPCRs) because the MLI was aimed at working on something different than what pharmaceutical companies were already doing. I also wanted an allosterically regulated enzyme so that one could potentially obtain both activators and inhibitors, providing a robust dataset to test out the quality and usefulness of high-throughput CRC determination. I decided that an assay that gave a strong signal increase for product formation would be best. There are some very sensitive and robust assays for measuring ATP using firefly luciferase and these had been widely used for protein kinases, except in that case, one measured substrate depletion, which is not very sensitive (the ADP-Glo assay from Promega was not available when qHTS started). However, using an enzyme that produced ATP, we could use the same technology to provide a robust luminescent signal for product formation. Pyruvate kinase seemed like a good choice as it is a well-characterized allosterically regulated enzyme that yielded ATP as a product, and I could buy boatloads of a bacterial enzyme (Bacillus stearothermophilus) from Sigma. So, the first assay NCGC screened was against pyruvate kinase and the first qHTS dataset was completed on May 31, 2005. These data were rich with both activation and inhibitory curves and the qHTS process was expanded to cover over 70K compounds as part of the first tests.
A challenge was how to rapidly fit and classify the 1,000s of CRCs obtained. There was no software available at that time that could handle thousands of curve fits and so software had to be developed in-house. We had hired a few very talented informatics scientists who were able to develop this software, which includes a curve classification scheme that allows one to rapidly sort curves based on potency, efficacy of response, and quality of the curve fit. The comprehensive view of SAR that was derived from the qHTS data allowed chemistry efforts to be launched quickly after completing the primary screen. As well, the qHTS data were very powerful for data mining purposes. The qHTS approach also gave the NCGC some competitive advantage in the MLI screening network as many collaborators appreciated the approach and wanted their assays to be screened in this manner.
At the NCGC, the qHTS process provided a flexible compound archive where one could screen at one or to up to seven different concentrations. NIH required that all primary screening data be published in PubChem, so there was a desire to make the primary screening data as robust as possible to aid future data mining activities. At the time qHTS was developed, acoustic dispensers were not available, so pin tools with fixed volumes were used for compound addition. There is significantly more flexibility in choosing compound screening concentration today with acoustic dispensing and this technology could be used to create multiple copies of a library at different concentrations.
When I arrived at Novartis in 2010, I started to suggest ways in which qHTS could enable the testing of certain focused libraries. I was very happy to find out that the software constructed at Novartis for handling the analysis of screening data already contained the ability to analyze interplate dilution data, so much of the data analysis pipeline was already in place. At Novartis, qHTS has been implemented for certain focused libraries, including a so-called mechanism of action library, which is a chemical genomic library of well-annotated compounds. qHTS has also been used at Novartis during the pilot stage of assay optimization to determine which screening concentration gives the optimal balance between true positives, false positives, and false negatives—all of which are readily obtainable from one qHTS experiment. This information can then inform on which compound concentration to choose when screening a larger set of compounds.
Titration-based screening is very useful when screening cell-based assays, particularly those employing primary cells or iPSC-derived cell types as bell-shaped behavior due to toxicity at higher compound concentrations is common. In addition, measurement of disease-relevant biological responses can often necessitate detection of very small changes in assay signal (e.g., approximately twofold), which is easier to measure reliably using a qHTS approach. Titration-based screening should always be employed when screening chemical genomic libraries of highly annotated and bioactive compounds due to the high amount of activity in these libraries.
You have published extensively on artifacts in luciferase reporter screens and other assay types and how test compound data can be misinterpreted as a result. Have you noticed a reduction in the number of misleading reports of compound activity as a result of your work and that of others? Are there examples of particularly subtle compound effects that you have encountered?
The publications on assay artifacts started with our desire at the NCGC to characterize the PubChem library for potential artifacts and bad actors in this library; all the screening results were publically available, so we wanted to provide counterscreen data to help with interpretation. Therefore, the NCGC performed fluorescent profiling studies for a subset of the PubChem library and fully characterized the library for inhibitors of firefly luciferase (FLuc), the most commonly used enzyme employed when constructing a reporter gene assay (RGA). The work on FLuc led to the discovery that certain inhibitors of the enzyme can act to stabilize the enzyme in cells, leading to a response that mimics gene activation. Our improved understanding of RGA results identified some misguided lead development efforts, prevented the following of false leads due to this artifact, and led to more optimal RGA designs. The drive to find alternatives to FLuc, resulting in the development of NanoLuc (Promega) and TurboLuc (ThermoFisher), was fueled by these findings and publications. At Novartis, I was able to further extend these studies to other classes of luciferases, including the newly introduced NanoLuc and TurboLuc enzymes. My laboratory at Novartis has also profiled the entire Novartis compound file against the FLuc enzyme to use as a counterscreen database for assays employing this reporter. Internally at Novartis, this information has proven to be valuable both when choosing a reporter system to construct an RGA and when interpreting the results of RGAs. More sophisticated RGAs are now employed, such as the coincident reporter system recently introduced by Jim Inglese's laboratory that incorporates two different luciferases with different mechanisms and inhibitor profiles, thereby providing built-in orthogonal readouts. In the last few years, I have observed laboratories moving away from FLuc and employing these improved reporter systems. The use of these improved luciferases with lower susceptibility to inhibition and more optimal assay designs should improve interpretation of the assay results. In the literature, there are some groups who have referenced this work and have taken these results into consideration when interpreting the results. I recently published a chapter on luciferase artifacts in the AGM so that others can learn from these findings.
Are there other exciting future directions in your own research or in the drug discovery field generally that you would like to highlight?
When I started in this field, large compound collections were being constructed and screened against a range of biological targets and pathways. An HTS typically covered over 1 million samples and required significant resources to support these efforts. Over time that biology was used to characterize the compound libraries and eventually understand which chemotypes were bioactive against various targets. Today, I find it intriguing how this information is being employed to create annotated chemical genomic libraries where now the compound collection is used to understand and even drive biology. For example, the production of cardiomyocytes from iPS cells can now be achieved using small molecules that modulate the Wnt pathway. Other protocols to derive neurons from iPS cells involve inhibitors of growth factor receptors, which have been developed over the past 20 years. The MLPCN efforts helped to identify several hundred new chemical probes, which can be considered when studying biology. Evaluation of the probes' utility is aided by efforts such as Chemical Probes.org (www.chemicalprobes.org), which aim to review, capture, and organize information on chemical probes. However, there is much more to do in this area and a large portion of the genome remains unexplored. This includes well-known target classes such as GPCRs and protein kinases, which have prompted NIH to start the Illuminating the Druggable Genome (IDG) initiative. The IDG program provides funds to explore new targets in the four major drug target classes—GPCRs, protein kinases, ion channels, and nuclear receptors. As well, efforts targeting entirely new macromolecules such as RNA are beginning to show up in the literature. Finally, this information has resulted in reorganizing compound libraries so these can be screened in various types of focused sets. In the last few years, it has been relatively rare that a program at Novartis will screen a full collection in an HTS of >1 million samples. Instead, focused screening is used to test out ideas and understand which assays might be predictive of downstream activity. The results of focused library testing then determine if screening a larger collection of compounds is warranted.
I am also excited about the birth of innovation centers within large pharmaceutical companies. At Novartis, I manage a laboratory that has served as a means for internal investigators to rapidly test out ideas, which need access to automated technologies. Many internal programs at Novartis have been supported by this laboratory over the past 7 years. Recently, at Novartis, this type of resource has been opened up to academic collaborators using an in-kind agreement format. Currently, there are collaborations with several local colleges and universities, as well as several universities from outside of Massachusetts. Typically a postdoc or graduate student from the collaborator's laboratory will come and work at Novartis to help perform the work. The libraries (up to 50K compounds) contain compounds that are in the public domain, so the data can be freely shared. A chemical genomic library to understand the pathways that regulate an area of biology is often employed in these collaborations. The goal is to find tool compounds for new target classes or novel biology that are of mutual interest. Such collaborations can be a key factor when attempting to gain traction in new technologies or areas of biology that show promise in supporting drug development.
—Interview by Andrew D. Napper, Editor-in-Chief