Physics continues to be vital in the search for new antibodies and antivirals to fight the coronavirus. The powerful one-two combination of the world’s high-performance computers sifting through hundreds of thousands of existing compounds and billions of potential ones and the world’s battery of synchrotrons, neutron sources, and cryoelectron microscopes are providing new leads for neutralizing SARS-CoV-2.
But the inexorable mutations of the coronavirus threaten to block, or at least render less effective, the vaccines that are gaining widespread usage. Already, research is showing that the antibodies produced by the Pfizer–BioNTech and AstraZeneca vaccines are far less effective against the variant originating in South Africa than against earlier SARS-CoV-2 strains, says Dave Stuart, who heads the biosciences program at the UK’s Diamond Light Source and is joint head of structural biology at the University of Oxford. As of early March, researchers had confirmed that the variant had broken through vaccine protection in some cases. “That is a genuine cause for concern,” he says.
Providing a more comprehensive defense against the coronavirus will require not only vaccines that may target less mutagenic proteins but also therapeutics to treat infected individuals. “Tackling a virus requires a multipronged approach in the big-picture way of not only having an arsenal to respond with as the virus changes over time, but also recognizing that trying to kill something that’s not really alive is a very hard task,” says Marti Head of Oak Ridge National Laboratory (ORNL), who leads the molecular design project of the National Virtual Biotechnology Laboratory (NVBL). That consortium of nine national laboratories was formed last year by the Department of Energy to focus on COVID-19 research with funding from the Coronavirus Aid, Relief, and Economic Security (CARES) Act. (See “Q&A: DOE’s Chris Fall,” Physics Today online, 28 April 2020.)
“If I look globally at what’s happening with antivirals, it’s really disappointing at the moment,” says Stuart. “The most effective therapeutics have been things like steroids,” which are used only for late-stage disease. After screening 12 000 compounds that have been tested in humans, researchers using Diamond have identified two molecules that look particularly promising. They are about to undergo tests in animal models at the Rega Institute for Medical Research in Belgium. “One of the lessons is that the basic science of making the protein, doing the experiments at the synchrotron, and getting structures has been much quicker than joining that up with the virology, the cellular work, and the animal work,” Stuart says.
Other promising molecules discovered at Diamond have been made public. They provide leads for Covid Moonshot, a crowdsourced drug development effort by medicinal chemists around the world who work pro bono to optimize the compounds. Some have been synthesized by Enamine, a Ukrainian medicinal chemical company, and sent back to Diamond to crystallize and check for binding with their targets. “We haven’t had the resources to do a large-scale screening, but we found a handful of compounds that do actually bind to the target molecule. It’s a good start and an indication that the process is not just throwing out random noise,” Stuart says.
The most potent antivirals are likely to be compounds that are specifically designed for SARS-CoV-2, he says, rather than repurposed drugs such as remdesivir, which has been approved or granted emergency use authorization in the US and other countries in Europe, Asia, and elsewhere for treatment of COVID-19.
At ORNL, Head says the NVBL molecular design team was, as Physics Today went to press, very close to publishing a preprint describing a potential antiviral molecule that was found in in vitro tests to block the function of the papain-like protease (PLpro), one of two SARS-CoV-2 enzymes essential to processing the virus’s polyproteins. The team hopes to move the compound outside the lab into a small animal trial. But she cautions that it’s a long way from a compound to a drug that is stable, nontoxic, and bioavailable.
“One of my roles in the NVBL, since I came from pharma, is to occasionally break my team members’ hearts by telling them that success rates in pharma are around 5–8%, starting from a target hypothesis,” Head says. Nonetheless, she says she is “very excited” about the molecule.
The NVBL-funded design team identified three other compounds as potent inhibitors to viral infection, but it’s unclear how they work. “We’ve made a computational prediction that it is through interactions with one or more of a set of eight proteins. We need to do experiments to confirm that,” she says.
Those involved in the research acknowledge that development of new drugs will take a long time. “Particularly for small molecules [meaning most pharmaceuticals], the timelines are such that we are not going to have a major impact on the pandemic,” says Jim Brase, a Lawrence Livermore National Laboratory (LLNL) deputy associate director who cochairs an industry–lab consortium for computational drug development. “But we would like to use the effort to make sure we have tools in place for future needs so we’ll be in a better position next time around.”
X-ray and neutron sources play a “huge role” in the molecular design efforts, Head says (see Physics Today, May 2020, page 22). Structural biologists working at ORNL and Brookhaven National Laboratory have solved several structures of the coronavirus’s main protease (Mpro) and provided input to the computational design of inhibitors. And since the hydrogen atoms in proteins are essentially invisible to x rays, neutron crystallography gives details of where those atoms are located, which is important to understanding the details of bond formation. For covalent bonds, such as the one formed with PLpro and the lead antiviral candidate in NVBL’s effort, “understanding where all of the hydrogens are moving around as you have a chemical reaction that forms a bond is very important,” Head says.
A threat to vaccines
The spike is a powerful antigen and the standout vaccine target because the immune system immediately mounts a strong response to it. Of the coronavirus’s 28 proteins, the spike also is the most prone to adapt via mutation. Vaccines can be readily altered to cover variants by changing a few amino acids in their messenger RNA (mRNA) or DNA code, says Jason McLellan, associate professor at the University of Texas at Austin. But as with influenza vaccines, annual boosters may be needed to keep pace.
Using blood taken from recovered COVID-19 patients, Diamond’s Stuart and collaborators identified 19 antibodies that were potent neutralizers of the receptor-binding domain, the region on the spike that attaches to the host cell. Five of those are known as public antibodies because they are shared by most people. They mutate the most rapidly in response to an infection, and their nimbleness in reacting to the antigens in the vaccines helps to explain why the vaccines in use today are so highly effective against the initial variants, he says.
But the variants of SARS-CoV-2 that have originated in the UK, South Africa, and Manaus, Brazil, each have mutations in the receptor-binding domain. “From the biophysics of it, it appears that those mutations increase the affinity of the virus for the [cellular] receptor. That will allow the virus to enter cells more easily and might account in part for the transmissibility advantage of those mutant viruses,” Stuart says. Antibodies, which also attach to the receptor-binding domain, are crowded out.
Stuart and colleagues found antibodies resulting from the Pfizer–BioNTech and AstraZeneca vaccines were somewhat less effective against the UK and Brazilian variants. But their effect on the South African variant was significantly reduced, he says. Some of the public antibodies were “knocked out” by the South African variant but notably not all. “I think the response can be rescued.”
Researchers at Diamond evaluated the performance of monoclonal antibodies that are approved for use in COVID patients against the South African variant. They found one of the antibodies in Regeneron’s two-antibody cocktail effective. But results for Eli Lilly’s single antibody against the variant “were deeply disappointing,” Stuart says. AstraZeneca’s antibody cocktail, still in clinical trials, performed well.
As spike variants continue to crop up, researchers will have to move quickly to modify antibodies computationally, create physical versions, and validate them experimentally, Head says. The NVBL team has used computational approaches to sample a chemical space of more than 1040 possible antibody combinations and has identified two antibodies and one nanobody—a class of molecule that is one-tenth the size of an antibody—that potently prevented the spike from attaching to cell receptors (see Physics Today, September 2020, page 22).
A different kind of vaccine
A vaccine that bypasses the spike altogether is the goal of a partnership announced in January between the UK company ConserV Bioscience and LLNL. The candidate targets two of the virus’s other 27 proteins; their identities are considered proprietary. Because it will inhibit protein regions that are common to all coronaviruses, ConserV’s vaccine should work against existing and future viruses of that type, says Kimbell Duncan, the company’s CEO.
While development is still in the preclinical stage, the partnership calls for LLNL to contribute its nanolipoprotein particle technology as the delivery vehicle for ConserV’s mRNA construct. Those water-soluble molecules, which resemble high-density lipoproteins, bind to and coat the genetic material.
Animal studies are scheduled to begin in the second quarter of this year, followed by human trials later in the year, says Duncan. He says he first learned of the nanolipoprotein technology serendipitously when he spent time at LLNL as a graduate student at Georgetown University.
LLNL is primarily a nuclear weapons laboratory. Its bioscience program was initiated in the early 1960s to study the effects on biological systems of fallout and other radioactive hazards and was refocused on bioterrorism after 9/11. The lab is also providing delivery platforms for other vaccines under development, says senior staff scientist Amy Rasley.
71 billion molecules
Large-scale computational screening of libraries of small molecules has turned up some hits as possible antivirals. Exscalate4CoV, a public–private consortium led by the Italian pharmaceutical company Dompé and partially funded by the European Commission, last fall combined the high-performance computing assets of the petroleum giant Eni and the publicly funded Italian supercomputing consortium Cineca to look for activity against SARS-CoV-2 in 400 000 molecules—including all compounds tested in humans, unregulated substances with physiological benefits, and substances used in traditional Chinese medicine.
The screening found that raloxifene, a low-cost generic drug used to treat osteoporosis, had potential in treating COVID in mildly symptomatic patients, and the drug is now undergoing clinical trials in Europe for coronavirus use. A second drug already on the market for another use has been found to be a potent viral inhibitor and is now in preclinical testing, says Andrea Beccari, the project’s chief scientist. He declined to identify the drug pending experimental validation and publication.
Since last fall, Exscalate4CoV has screened 71 billion synthesizable molecules for activity against 15 active sites on 12 of the SARS-CoV-2 proteins. ORNL’s supercomputer performed a similar exercise last summer involving 1 billion molecules on a single protein in two conformations. “We believe that 71 billion compounds among 15 targets is a huge chemical space that can be effective to use for training neural networks for a specific protein or for addressing the entire viral functional proteome,” Beccari says. “The idea is to release the biggest possible profile of chemical space for this and future pandemics.”
The consortium routinely interacts with European light sources, including Italy’s Elettra Sincrotrone and FERMI. “Crystal structure information optimizes models,” Beccari says. “We have molecules that are active on the papain-like protease and are moving to crystallization with these compounds.”
In the US, funding for NVBL under last year’s coronavirus relief legislation has mostly dried up, but the stronger links on coronavirus persist among the DOE labs. Labs with large computational assets and those having light sources and other imaging capabilities have strengthened collaborations, says LLNL’s Brase. He coleads the consortium called Accelerating Therapeutics for Opportunities in Medicine (ATOM), which includes GlaxoSmithKline, the Frederick National Laboratory for Cancer Research, and the University of California, San Francisco (UCSF). Formed to search for cancer therapeutics, the partnership added infectious diseases to its mission last year, and ORNL, Brookhaven, and Argonne National Laboratory came aboard.
ATOM is contributing models it has developed to predict the safety of candidate molecules and whether they will travel through the body and be bioavailable, Brase says.
A nanobody inhaler
Researchers at UCSF are developing synthetic nanobodies against the coronavirus that they propose to be delivered via nasal spray or inhaler. Their technology, dubbed AeroNabs, is undergoing animal trials. After combing through databases containing 2 billion synthetic nanobody molecules, the scientists found three that neutralized the spike. At Lawrence Berkeley National Laboratory (LBNL), they were able to determine through cryoelectron microscopy and crystallography where two of the nanobodies bind to the spike. But they were unable to get structures of the third using either method.
The UCSF scientists turned to x-ray footprinting, a relatively new structural technique, performed on a synchrotron, that doesn’t require crystals but instead uses molecules in solution. Corie Ralston, facility director for biological nanostructures at LBNL’s Molecular Foundry, says the method infers the structure by whether a part of the biomolecule known as a residue and a water molecule are adjacent. “For instance, if in the process of folding, this residue has a water molecule next to it and then it doesn’t, we figure that’s where it’s making contact with another part of the protein. Then we can figure out how it folded.”
The structural work at LBNL provided information for researchers to optimize the potency of the nanobodies, using strategies such as linking two or three nanobodies together to attach to the spike at multiple points.
When the pandemic eases, it will be critical that governments not let up on developing broad-spectrum antivirals and therapeutics, says Duncan. “They need to extend the emergency work to focus on the next pandemic. They really need to stay on high alert.”