Computer-aided Discovery of Novel Influenza Endonuclease Inhibitors to Combat Flu Pandemic
1Short Project Description
4Question / Proposal
6Method / Testing and Redesign
8Conclusion / Report
9Bibliography, References and Acknowledgements
The emergence of new highly lethal influenza viruses such as H5N1 and H7N9 poses a grave threat to the world. My project is to discover novel influenza endonuclease inhibitors as leads for a new type of anti-flu medicine, effective against all influenza viruses including pandemic strains. By combining computer modeling and biological studies, I identified a number of novel, potent endonuclease inhibitors. I also performed comprehensive structural analysis, laying ground work for further design and optimization of the anti-flu drug candidates.
Influenza pandemics remain a constant threat to public health. If the highly lethal avian influenza strains H5N1 or H7N9 acquire human to human transmission, it could cause a disaster like the 1918 Spanish flu, which killed 40 million people. Anti-flu drugs are crucial to treat patients before a new vaccine becomes available, but flu strains resistant to current drugs are emerging. Therefore, new anti-influenza drugs are urgently needed to prepare for impending flu pandemics.
The Influenza endonuclease is well conserved and essential for viral propagation, and thus its inhibitors can potentially block any influenza virus. I used an interdisciplinary approach combining computational research with biological assays to expedite the discovery of new endonuclease inhibitors. Computer pharmacophore models were built and used for virtual screening of compound libraries. A fluorescence-based assay was set up to validate virtual screening hits as endonuclease inhibitors. In parallel, molecular dynamics simulations and solvent mapping of the endonuclease were performed to construct a comprehensive database of binding pockets and druggable hot spots. Molecular docking of the new inhibitors to the PA endonuclease active sites revealed interacting residues. Structure and activity relationship analysis led to identification of more drug leads.
As a result, I identified a number of new, potent and structurally diverse endonuclease inhibitors with great potential to be developed into new anti-influenza drugs. The structural study also provided valuable information for designing even more potent inhibitors. Therefore, these findings will help combat influenza and save lives. A patent has been filed on my discovery.
I am a junior at the Canyon Crest Academy in San Diego, California. Ever since childhood I was fascinated by mathematics and science, and started to do research in biology when I was a freshman. To me, performing research is a way to channel my knowledge and abilities into solving society's problems. My idols include Dr. Jonas Salk, whose contribution of the polio vaccine has changed the world immeasurably, and my mathematics teacher Brian Shay, whose energy, humor, and dedication has furthered my passions for science and math.
Besides my love for math and science, clubs, volunteering, and research, I enjoy playing piano and fencing epee.
As for my future, I may become a college professor, as I enjoy teaching. I have been a teaching assistant for college level classes at my school, founded a summer math contest program for middle school students and coached Science Olympiad for primary and high school students. Another possibility would be starting my own company leading multidisciplinary teams to tackle world problems.
I have participated in several science fairs and won various awards at levels varying from regional to international. These fairs were amazing experiences, and I was inspired by other extraordinary projects at each of the events. They also allowed me to communicate and learn from judges and other participants and further develop my project ideas. I feel like Google would be one of the biggest platforms for me to communicate my ideas to the world.
Given the urgency of possible new flu pandemics, can a computational approach combined with biological studies expedite discovery of new anti-flu medicine?
Influenza persistently causes many cases of illness and deaths every year, and lethal avian flu strains like H5N1 and H7N9 pose a potential pandemic threat that could kill millions of people. As vaccines take months to generate and currently available antiviral drugs are waning in effectiveness, it is urgent to develop novel anti-flu drugs to stop possible influenza pandemics.
Currently, the discovery of a new drug requires millions of dollars and takes an average of a decade. I believe that the increased power of computers and new structure studies can drastically speed up the drug discovery process. I chose influenza endonuclease as a drug target as it is a conserved nonstructural protein essential for viral propagation. Inhibitors of this enzyme can potentially block any influenza virus and reduce the chance of developing resistance. I hypothesize that an interdisciplinary approach combining computational research with biological assays can significantly expedite drug development of influenza endonuclease inhibitors.
First, virtual screening would significantly reduce the time and resources required to identify drug leads compared to the labor intensive conventional strategy which goes through measuring the activity of every compound in a library. Second, sensitive and reliable biochemical and biological assays can validate the virtual screening hits. Third, structural studies can provide a basis for further modifying the compounds to gain higher potency and other desirable characteristics of a successful drug.
Influenza viruses pose grave threats to human health. Every year, seasonal flu causes up to 40,000 deaths in the United States alone despite the existence of vaccines and current available antiviral drugs. The 1918 Spanish Flu, an extremely virulent pandemic influenza strain, killed over 40 million people globally. Currently, the most serious threat comes from the highly lethal avian influenza H5N1 and H7N9 viruses that are found in birds. Both viruses have very high mortality rates in humans and could become pandemic if they acquire the ability for efficient human-to-human transmission.
Since vaccines require several months to develop, effective antiviral drugs are extremely important to protect human population during the initial stage of a fast spreading pandemic. Current anti-flu drugs, such as adamantanes and zanamivir (Tamiflu), which target structural proteins M2 ion channel and neuraminidase, respectively, are becoming ineffective as resistance rapidly develops in recent years.
There are several advantages to choosing the nonstructural protein PA endonuclease, a subunit of viral RNA polymerase, as a drug target. First, the “cap-snatching” activity of the PA endonuclease is critical for virus propagation, and thus an endonuclease inhibitor should effectively kill influenza viruses. Second, PA endonuclease is unique in viruses without a human counterpart so specific drugs can be developed with less side effects. Third, the PA N-terminus is highly conserved and thus a drug targeting this endonuclease would likely be effective against all influenza viruses, and have reduced chance of resistance developing.
The crystal structures of the N-terminus of PA endonuclease (PAN) were first resolved in 2009, but were not good enough for drug design. More crystal structures were reported recently while my study was ongoing. Several inhibitor compounds of the influenza endonuclease were reported previously, however, none of them have been developed into drugs in clinics, possibly due to undesirable drug properties and lack of structure data to improve them. There was a similar situation during the development of flu neuraminidase inhibitors. While the initial discovered inhibitors had no clinical effect, breakthrough came after structure-based drug design became possible.
Developing a new drug is a long and costly process. Today many new computer-aided tools have been developed, which can significantly shorten the discovery phase. For example, virtual screening makes it possible to screen huge compound libraries within short periods of time. A three-dimensional pharmacophore model can potentially help identify compound leads with desired biological activity but diverse chemical scaffolds. Computer-based structural studies are also crucial for “lead optimization” to increase the potency of the drug candidates to the low nanomolar range. It is projected that computer simulation can cut 50% of drug development expenses.
In this project, I combined computational research with biological assays to expedite discovery of potent influenza endonuclease inhibitors for new anti-flu drug development.
Part I. Computational studies
Pharmacophore model and virtual screening: Pharmacophore models were generated using the Openeye ROCS program based on nine known inhibitors and the best scored model was used to screen compound libraries. The top-ranked compounds from virtual screening were checked for metal chelating groups and 237 compounds were selected for endonuclease assay validation.
Molecular dynamics (MD) simulation: MD simulation was performed in explicit solvent on the TACC Ranger supercomputer. From 100 nanosecond MD runs of H5N1 PAN, thousands of frames of possible conformations were obtained at atomic levels. RMSD clustering was performed using the GROMOS analysis software.
Computational solvent mapping: FTMap, a program searching the protein surface for areas that can bind probe molecules, was used with MD conformations or crystal structures of PAN.
Molecular docking: Docking of compounds into the active sites of PAN was carried out using the Glide module in Schrodinger software. The X-ray structures and the MD conformations of PAN were used in the analysis with the presence of two Mn2+ ions and water molecules.
Part II. Biological assays
PAN protein production: His tagged H1N1 PAN (1-220) was expressed in E. coli strain BL21 and purified by Ni-agarose affinity column to near homogeneity, confirmed by protein gel electrophoresis. Purified protein was aliquoted and stored at -800C.
Endonuclease FRET assay: A 17 base dual-labeled DNA oligonucleotide was used as the substrate. 1mM PAN, 0.5mM Oligo substrate and inhibitors at various concentrations were mixed in 100ml assay buffer (10 mM Tris pH8.0, 2.5 mM MnCl2, 100 mM NaCl and 10 mM b-mercaptoethanol) in 96-well plate or 30ml in 384-well plate. Flourescence was read at 485 nm excitation and 535 nm emission wavelength in the BioTek plate reader Flx-800 for 30 min at 1 min interval at 370C. The rate Vm was calculated by the software KC4 with the instrument. IC50 was determined by nonlinear curve fitting (four parameters) using GraphPad software.
Endonuclease gel-assay: 0.5 µg of M13mp18 single-stranded circular phage DNA was used as the substrate and incubated with 0.5 µg of PAN and various concentrations of inhibitors in 50 µl of assay buffer. The reactions were stopped after incubation 90 min at 370C by adding 50mM EDTA, and analyzed by agarose gel (1.0%) electrophoresis.
Compound toxicity assay: 1.5x104 Madin-Darby canine kidney (MDCK) cells (ATCC) were plated to each well of a 96-well plate. One day later, compounds at various concentrations were added and incubated for 48 hours, followed by staining with methylene blue dye. The plate was read at 630nm on a plate reader and CC50 (50% cytotoxic concentration) was calculated.
1. Pharmacophore-based virtual screening of compound library to identify new inhibitor candidates
I built pharmacophore models that capture 3D chemical features common to the known endonuclease inhibitors. The principal feature of the model is a set of consensus oxygen atoms capable of complexing with the two active-site metal ions.
Using this pharmacophore model, I performed virtual screening of libraries with roughly 450,000 compounds. Each compound was overlaid with the model and assigned a score. The top 237 hits were selected.
2. Validation of virtual screening hits by enzyme inhibition assays and discovery of six new PA endonuclease inhibitors
Virtual screening results were validated by PA endonuclease assays. I set up a FRET (fluorescence resonance energy transfer) assay using a dual-labeled oligonucleotide (figure 3A). Cleavage of the substrate separates the fluorescence dye from the quencher resulting in an increase of the fluorescence signal. Six new endonuclease inhibitors with IC50<50 µM were identified (table 1, figure 3B), and were further confirmed by a DNA gel-assay (figure 3C). Cytotoxity in MDCK cells was also determined for each new compound (table 1).
3. Molecular dynamics (MD) simulation of the PA endonuclease domain and computational solvent mapping (CS-Map)
I performed MD simulation to obtain structural information of the PA endonuclease active sites. MD simulation can illuminate the way proteins move in solution, and provide more conformations for compound docking than a single snapshot from crystal structures. The active site views of three most populated cluster centroids of MD ensembles of H5N1 PAN are shown in figure 4A.
Using the MD ensembles and the crystal structures of the PAN domain, I performed CS-Map, which predicts the hot spot residues at the endonuclease pocket surfaces favorably interacting with solvent probes (figure 4B). These results will be useful for a fragment-based drug design strategy called “fragment growth”, incorporating the new scaffolds of inhibitor candidates identified in this study.
4. Molecular docking experiments of new inhibitors to PAN to reveal key interactions for drug optimization
I performed docking simulation of these new inhibitors with the PA endonuclease domain. An optimal docking model of compound D03 and C09 is depicted in figure 5. D03 has two arms extending into two pockets of the enzyme and forming interactions with residues conserved in the PA of influenza A and B.
5. Structure and activity relationship (SAR) study of the analogs of the new inhibitors
Based on molecular docking analysis, 21 available analogs of the new inhibitors were selected for the endonuclease activity assays. Four additional new inhibitors with IC50 below 2 µM were identified.
Two compounds C03 and C09 with minimal cytotoxicity were confirmed to have good antiviral activity in reducing cytopathic effect in MDCK cells caused by influenza virus. Experiments with viruses were performed by a supervising scientist.
In summary, by using an interdisciplinary approach, I discovered a number of new potent influenza endonuclease inhibitors with diverse structures representing at least five unique classes. Some of them are the most potent endonuclease inhibitors discovered so far, which have great potential to be developed into new anti-influenza drugs. Also, structural studies including molecular docking of the new inhibitors to the endonuclease active sites and structure-activity relationship studies have laid the ground work for further drug optimization and rational drug design. The approach I used also saved huge amounts of time, resources, and effort.
In this study, I also demonstrated that combining virtual screening with biological assays provides a quick and efficient way to identify new endonuclease inhibitors from large compound libraries, saving time and resources. It took me only six months to identify and validate several new endonuclease inhibitors from almost half a million compounds, while the conventional screening method would take much longer time and require much more resources. Therefore, it supports my hypothesis that that computer-aided research can significantly expedite the process of drug discovery for influenza endonuclease inhibitors.
Computer modeling and virtual screening do have their limitations in that many leads do not have activity when tested experimentally. Therefore, I did not stop at the computational results but also worked in a biological “wet” lab to set up biochemical assays to validate the virtual screening hits and find the true endonuclease inhibitors with biological activities. Currently, through collaboration, the antiviral activities of identified inhibitors are being tested, and co-crystal structure analysis of new inhibitors with the endonuclease is being set-up. These studies will complement the computer-aid research and facilitate drug discovery.
Due to the significance of my findings, a patent on my discovery has been filed through UCSD and a manuscript is being prepared. I hope that my work will attract pharmaceutical companies to put in more resources to this development and bring new types of flu medicine to patients in the near future. Since this study will help us combat influenza and save lives, it potentially has a great impact on human health and society.
These results have also triggered possibilities in further development. First, I will collaborate with medicinal chemists to design and test new compounds to improve potency and other desirable drug properties. The various assays I developed will be very useful to find the most potent ones that will go into clinical trials. Second, similar strategies can be applied to identify drugs targeting other influenza proteins, such as PB1, PB2 and NP. Finding new, less toxic drugs may lead to “cocktail” therapies similar to the successful HIV treatment. I am going to promote this new concept of “combination therapy for flu” which may control influenza more effectively and avoid viral resistance. The strategy I have developed in this study can even be extended to hundreds of diseases currently plaguing humans, with the potential to speed up finding remedies for those in need.
Biswas, S.K., and Nayak, D.P. (1994). Mutational analysis of the conserved motifs of influenza A virus polymerase basic protein 1. Journal of virology 68, 1819-1826.
Capua, I., and Alexander, D. (2010). Perspectives on the global threat: the challenge of avian influenza viruses for the world's veterinary community. Avian diseases 54, 176-178.
Das, K., Aramini, J.M., Ma, L.C., Krug, R.M., and Arnold, E. (2010). Structures of influenza A proteins and insights into antiviral drug targets. Nature structural & molecular biology 17, 530-538.
Dias, A., Bouvier, D., Crepin, T., McCarthy, A.A., Hart, D.J., Baudin, F., Cusack, S., and Ruigrok, R.W. (2009). The cap-snatching endonuclease of influenza virus polymerase resides in the PA subunit. Nature 458, 914-918.
Dubois, R.M., Slavish, P.J., Baughman, B.M., Yun, M.K., Bao, J., Webby, R.J., Webb, T.R., and White, S.W. (2012). Structural and Biochemical Basis for Development of Influenza Virus Inhibitors Targeting the PA Endonuclease. PLoS pathogens 8, e1002830.
Durrant, J.D., and McCammon, J.A. (2011). Molecular dynamics simulations and drug discovery. BMC biology 9, 71.
Hay, A.J., Wolstenholme, A.J., Skehel, J.J., and Smith, M.H. (1985). The molecular basis of the specific anti-influenza action of amantadine. The EMBO journal 4, 3021-3024.
Iwai, Y., Murakami, K., Gomi, Y., Hashimoto, T., Asakawa, Y., Okuno, Y., Ishikawa, T., Hatakeyama, D., Echigo, N., and Kuzuhara, T. (2011). Anti-influenza activity of marchantins, macrocyclic bisbibenzyls contained in liverworts. PloS one 6, e19825.
Kowalinski, E., Zubieta, C., Wolkerstorfer, A., Szolar, O.H., Ruigrok, R.W., and Cusack, S. (2012). Structural Analysis of Specific Metal Chelating Inhibitor Binding to the Endonuclease Domain of Influenza pH1N1 (2009) Polymerase. PLoS pathogens 8, e1002831.
Krug, R.M., and Aramini, J.M. (2009). Emerging antiviral targets for influenza A virus. Trends in pharmacological sciences 30, 269-277.
Landon, M.R., Amaro, R.E., Baron, R., Ngan, C.H., Ozonoff, D., McCammon, J.A., and Vajda, S. (2008). Novel druggable hot spots in avian influenza neuraminidase H5N1 revealed by computational solvent mapping of a reduced and representative receptor ensemble. Chemical biology & drug design 71, 106-116.
Landon, M.R., Lancia, D.R., Jr., Yu, J., Thiel, S.C., and Vajda, S. (2007). Identification of hot spots within druggable binding regions by computational solvent mapping of proteins. Journal of medicinal chemistry 50, 1231-1240.
Medina, R.A., and Garcia-Sastre, A. (2011). Influenza A viruses: new research developments. Nature reviews Microbiology 9, 590-603.
Moss, R.B., Davey, R.T., Steigbigel, R.T., and Fang, F. (2010). Targeting pandemic influenza: a primer on influenza antivirals and drug resistance. The Journal of antimicrobial chemotherapy 65, 1086-1093.
Nakazawa, M., Kadowaki, S.E., Watanabe, I., Kadowaki, Y., Takei, M., and Fukuda, H. (2008). PA subunit of RNA polymerase as a promising target for anti-influenza virus agents. Antiviral research 78, 194-201.
Plotch, S.J., Bouloy, M., Ulmanen, I., and Krug, R.M. (1981). A unique cap(m7GpppXm)-dependent influenza virion endonuclease cleaves capped RNAs to generate the primers that initiate viral RNA transcription. Cell 23, 847-858.
Reid, A.H., Taubenberger, J.K., and Fanning, T.G. (2001). The 1918 Spanish influenza: integrating history and biology. Microbes and infection / Institut Pasteur 3, 81-87.
Ruigrok, R.W., Crepin, T., Hart, D.J., and Cusack, S. (2010). Towards an atomic resolution understanding of the influenza virus replication machinery. Current opinion in structural biology 20, 104-113.
Stiver, G. (2003). The treatment of influenza with antiviral drugs. CMAJ : Canadian Medical Association journal = journal de l'Association medicale canadienne 168, 49-56.
Tomassini, J.E., Davies, M.E., Hastings, J.C., Lingham, R., Mojena, M., Raghoobar, S.L., Singh, S.B., Tkacz, J.S., and Goetz, M.A. (1996). A novel antiviral agent which inhibits the endonuclease of influenza viruses. Antimicrobial agents and chemotherapy 40, 1189-1193.
von Itzstein, M., Wu, W.Y., Kok, G.B., Pegg, M.S., Dyason, J.C., Jin, B., Van Phan, T., Smythe, M.L., White, H.F., Oliver, S.W., et al. (1993). Rational design of potent sialidase-based inhibitors of influenza virus replication. Nature 363, 418-423.
Yuan, P., Bartlam, M., Lou, Z., Chen, S., Zhou, J., He, X., Lv, Z., Ge, R., Li, X., Deng, T., et al. (2009). Crystal structure of an avian influenza polymerase PA(N) reveals an endonuclease active site. Nature 458, 909-913.
I would like to thank my mentors Dr. Rommie Amaro and Dr. Gen-Sheng Feng for providing the opportunity for me to work at their laboratories in UCSD. I am grateful for their guidance and support, as well as the freedom they gave me to explore new ideas and new experiments in their laboratories. I would like to thank Dr. Robert Swift in Dr. Amaro's lab for introducing me to pharmacophore models and helping me with troubleshooting when I encountered problems. I would like to thank Dr. Robert Malmstrom and Dr.Victoria Feher in Dr. Amaro's lab for giving me suggestions on various technical issues. I would like to thank Ms. Nazilla Alderson in Dr. Feng's lab for her help arranging supplies and equipment for my experiments. I would like to thank members from both labs for helpful discussions. Finally, I would like to thank my family, teachers, and friends for their help and support of my project.