The word vaccine comes from vacca, the Latin word for cow. When 18th century English scientist Edward Jenner began inoculating people against the highly lethal smallpox virus (Variola major) by using a fairly benign related virus, cowpox virus (at the time called Variola vaccinae, or “smallpox of the cow”), he called the procedure “vaccination,” derived from vaccinae, derived from vacca. Jenner had observed that milk maids who had cowpox seemed to be protected from smallpox. But infection by one virus providing immunity against subsequent infection by another, a phenomenon known as cross-protection, doesn’t always succeed.
Most humans have been infected with one or more cold-causing coronaviruses in the past. But unlike the cowpox-smallpox scenario, prior infection by a less pathogenic coronavirus does not seem to protect against its bigger, badder cousin, SARS-CoV-2. Because there is no preexisting immunity to this virus in the human population, it falls to medical science to come up with ways to prevent people from becoming infected by SARS-CoV-2 and treat people who develop the disease it can cause, known as COVID-19.
To coordinate effort and address key challenges in responding to the pandemic, the Department of Energy established a consortium of national laboratories called the National Virtual Biotechnology Laboratory, or NVBL. One research area within the NVBL project called Molecular Design to Inform COVID-19 Medical Therapeutics is led by Los Alamos Office of Science Program Director and biochemist Srinivas Iyer.
“We are studying vaccine strategy, virus-host interactions, and small-molecule therapeutics,” says Iyer. “By integrating structural biology, computation, modeling, machine learning, and other national-laboratory capabilities, the NVBL will accelerate the development of vaccines and therapeutics against COVID-19.”
There are around 200 SARS-CoV-2 vaccines in various stages of development.
Preventing infection
In February of 2020, as SARS-CoV-2 was beginning to spread from its early enclaves to the rest of the world, Los Alamos theoretical biologist Bette Korber began to track the slowly accumulating genetic changes in the virus. Her goal was to help experimentalists identify different versions of the virus containing certain genetic changes that could impact vaccine efficacy. As a theorist, Korber studies the biology and evolution of highly pathogenic viruses, such as the Human Immunodeficiency Virus (HIV), Ebola, and Hepatitis C, and human immune responses to them. By early April, Korber saw something that, to her expert eye, looked like a pattern that was unlikely to be due to random mutation.
Coronaviruses are so named for the protuberant spike proteins that stick out from their surface, creating a halo, or corona—Latin for crown—around the particle. These spike proteins help the virus enter a human cell, so they are an obvious subject of scientific scrutiny. The spike protein for SARS-CoV-2 is formed from a string of 1,273 amino acids, and it was here that Korber saw the pattern: in the virus’s original form, the 614th amino acid is aspartic acid (denoted by the chemical symbol “D”), but an increasing number of samples from disparate geographic locations had a glycine (“G”) at that location. The replacement of D with G at location 614, or “D614G” in genomics nomenclature, began appearing in viral gene sequences in January and by June was found in nearly all new samples.
Korber and her colleagues, Will Fischer, Hyejin Yoon, James Theiler, Brian Foley, Nick Hengartner, and Werner Abfalterer, developed a bioinformatics pipeline to analyze SARS-CoV-2 gene sequence data from GISAID, the Global Initiative on Sharing All Influenza Database that was developed for influenza but is now the central coronavirus database as well. By studying GISAID coronavirus sequence data from around the world, the team looked for variants that were repetitively increasing in frequency. Their statistical analyses of global patterns showed that the D614G substitution was under positive selection; in other words, something other than mere chance was making the G-form supplant the D-form again and again.
“The viruses carrying D614G were rapidly becoming the globally dominant form, and it was important to understand why,” says Korber. “One possibility was that it was more infectious. Another possibility was that the substitution affected the human immune response to the virus.” Most vaccine efforts were based on the ancestral D-form, but as the G-form was becoming dominant, it was important to ensure that a vaccine would be effective against that form as well.
The Los Alamos team, with collaborators from Duke University, the La Jolla Institute of Immunology, and Sheffield, England, compared G-form and D-form and found some important differences. First, the G-form appears to replicate more readily in the upper respiratory tract. Second, the G-form does not appear to cause more severe disease. Third, G-form spikes were more infectious than the ancestral D-form. Then the scientists showed that the G-form is neutralized by host antibodies more, not less, as might have been expected from its increased infectivity compared to the D-form. These findings all have a structural explanation.
Laboratory structural biologist Gnana Gnanakaran, who works with Korber, wanted to pursue that structural explanation. “Why is the virus taking the G-form?” he asks. “What is the function of that single amino acid substitution?”
A team of postdocs from Gnanakaran’s group—Rachael Mansbach, Srirupa Chakraborty, and Kien Nguyen—ran molecular-dynamics simulations on the Lab’s high-performance computers, simulating every atom of the spike protein in both D-form and G-form. Each spike is a trimer—a group of three identical molecules, or protomers, that work in concert as one functional unit. In electron micrographs, these protomers usually have their terminal region, or head, folded down, but occasionally one will have its head sticking up. The “up” configuration exposes a section of the spike protein called the receptor binding domain (RBD), which interacts with host-cell receptors to allow viral entry. So, more protomers in the “up” position would allow more binding and more entry into host cells.
The team’s simulations revealed that indeed G-form spike proteins should have considerably more protomers in the up position at any given moment, about 75 percent, compared to 50 percent for D-form. Interestingly, the RBD is also a target for natural antibody-based neutralization, so the “more up” hypothesis also provides an explanation for the finding of increased neutralization.
It’s quite likely that there will be another crossover event. We’ve seen three in the last 18 years.
The location of amino acid position 614, however, is not very near the RBD; it lies about halfway down the spike protein. How can a change at that distal location affect the molecule’s likelihood to take the “up” shape? The computer models suggest that the D614G substitution acts by rearranging hydrogen bonds within and between protomers, which relieves strain caused by the “up” position. In D-form, one protomer in the “up” position creates asymmetric interactions further down the molecule, forcing neighboring protomers into the “down” position. But in the G-form, the asymmetry is relaxed, resulting in a higher proportion of spike proteins with at least one protomer “up.” Taken together, this amounts to the G-form being more infectious and more transmissible than the D-form. The theorists’ findings have been confirmed experimentally by collaborators.
The D614G change is just one amino acid substitution; viruses like SARS-CoV-2 undergo this kind of substitution frequently throughout their structural proteins. Some substitutions are consequential, like D614G, but many aren’t. Understanding these types of evolutionary mechanisms is crucial for vaccine design.
AUTHOR’S NOTE: In the time since this article was written, variations of SARS-CoV-2 with additional amino-acid substitutions have appeared throughout the world. This is not unexpected, but it does demand the attention of scientists like Korber who track viral evolution, and the vigilance of vaccine designers.
At the time of this writing, there are around 200 SARS-CoV-2 vaccine candidates in various stages of development, and preliminary results show great promise. In addition to providing structural modeling and viral evolution expertise to other vaccine designers, Korber and her team have two vaccine candidates of their own in the works. One is based on the spike protein and attempts to capture the natural diversity of the virus in key antibody targeting sites, so as to maintain efficacy as the virus evolves. The other will operate through a non-antibody immune mechanism whereby cells of the immune system track down and kill virus-infected cells. It’s unclear how crucial this pathway is for SARS-CoV-2 immunity, but it’s likely to be important because it can help resolve infection by the other highly pathogenic coronaviruses, Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV) and Middle East Respiratory Syndrome Coronavirus (MERS-CoV).
SARS-CoV and MERS-CoV emerged earlier this century, in 2002 and 2012 respectively. Three highly pathogenic coronaviruses emerging in two decades highlights the need for a broad coronavirus vaccine—one vaccine that protects against many related viruses. Korber, her Los Alamos colleagues, and the broader NVBL consortium are keeping this in their sights so that the work they do for the current pandemic can help protect humanity during the next one.
Treating disease
In addition to designing vaccines to prevent infection, Korber and colleagues are also helping design therapeutics for treating COVID-19. One method of treating an infection is to administer exogenous antibodies—that is, antibodies not made by the patient’s own body. These could be convalescent antibodies, which come from people who have survived the infection, or they could be artificially synthesized. Convalescent antibodies seem to be an effective treatment, but not one that is likely to be broadly available, so synthetic antibodies need to be developed.
“In May, when we started this work, we couldn’t get many SARS-CoV-2-specific antibodies,” says computational immunologist Kshitij Wagh. “So we used antibodies from the first SARS, SARS-CoV, as a starting point and began computationally designing variants of these that might be effective against SARS-CoV-2.”
Unfortunately, SARS-CoV-specific antibodies don’t protect against SARS-CoV-2—survivors of SARS are not immune to COVID-19. But Wagh and Korber understand the critical biophysical interactions between antibody and virus that make such antibodies effective against one but not the other. They are using this knowledge to design antibody variants that can improve interactions with SARS-CoV-2.
The antibody work that Wagh and Korber are doing for the coronavirus pandemic is underpinned by the many years of work they have done, and are still doing, for the other pandemic burning across the globe—HIV.
“HIV has been and remains a pandemic. It’s slow burning, but it’s still raging,” says Wagh. He and Korber built their antibody-modeling expertise through their pursuit of therapies for HIV. They believe, as do many in the HIV field, that the key to COVID-19 antibody therapy is to make it a cocktail—a mix of at least two or three different antibodies. Naturally arising mutations in the virus could allow it to escape neutralization by one antibody, but it would be exponentially less likely to evolve two or three escape mechanisms at once. Artificially synthesized antibodies have the advantage that they can be tailored and tweaked to alter functionality, and one such tweak on the Los Alamos team’s radar is to make an effective cocktail against other coronaviruses, not just SARS-CoV-2.
“It’s quite likely that there will be another crossover event,” emphasizes Wagh. “We’ve seen three in the last 18 years.”
Antibodies can be very particular in the targets they recognize. If they are specific to “up” configured spike proteins, they would preferentially bind to virus particles with “more up.” So Gnanakaran, in collaboration with colleagues at Duke University, is figuring out how to stabilize a spike protein in the “up” configuration, so that it might be used as an immunogen—a molecule against which antibodies are made. He is also studying the interaction between the spike protein’s RBD and the human cell receptor, angiotensin converting enzyme 2 (ACE2), to see exactly where molecular recognition occurs.
ACE2 is important to the COVID-19 picture in more ways than one. It’s the main molecule that SARS-CoV-2 uses to infect a human cell, but it has a normal job too. ACE2 is a key player in a complex blood-pressure and electrolyte-regulation pathway.
Sofiya Micheva-Viteva is a microbiologist at Los Alamos who is studying what happens when the virus, by binding to ACE2, prevents ACE2 from acting in its normal capacity. Ordinarily, when blood pressure dips, ACE2 helps bring it back up by producing vasoconstrictors and other cell-protective molecules. However, when ACE2 is bound by the virus, it can’t do its job, and inflammation results, which can be pathogenic on its own and, in the case of SARS-CoV-2 infection, might exacerbate the symptoms of COVID-19.
Rather than live virus, Micheva-Viteva uses virus-like particles (VLPs), which are essentially empty virus particles; they have the same external proteins, including the spike protein, but there is no genome inside so they are incapable of replication. VLPs can bind to live cells in a mock infection, so scientists can safely study what happens within those cells during infection.
In particular, Micheva-Viteva is looking at a preexisting but still unlicensed drug. The drug is a synthetic version of one of the molecules that ACE2 is responsible for activating, and it has anti-inflammatory effects. If SARS-CoV-2 has bound to ACE2, then the anti-inflammatory molecule can’t be activated, so inflammation and oxidative stress will rise. But the artificial version, the drug molecule, might be able to act in its place to restore regulation and minimize downstream damage. Any virus that binds ACE2 will interfere with this pathway, but this non-virus-specific therapeutic could be an effective way to restore function.
This isn’t an anti-viral approach; it’s a pro-host approach—a way to improve the outcome of infection.
Rather than an anti-viral approach, Micheva-Viteva is pursuing a pro-host approach, a way to improve the patient’s ability to handle infection. Though it wouldn’t prevent infection, the therapy may dampen the severity of the disease enough to make it non-life threatening and keep the patient out of intensive care.
The receptor ACE2 doesn’t act alone to let SARS-CoV-2 into a cell. ACE2 is how the virus knows it’s in the right place, but to get in requires several other host-cell molecules. One of these is called “transmembrane protease, serine 2,” or TMPRSS2. This molecule is an enzyme found on the cell surface, whose normal function is not entirely known. During SARS-CoV-2 infection, the enzyme acts on a particular piece of the virus’s spike protein, breaking the amino acid chain at that spot. This is part of a process called priming, which increases the spike protein’s structural flexibility and allows the membranes of the virus and cell to fuse.
Los Alamos biophysicist Julian Chen is looking closely at the role of TMPRSS2, with an eye on disrupting its function.
“In order for a coronavirus to infect a cell, there are a lot of different steps,” says Chen. “If we have a molecule that can interfere with any given step, that might be a drug that will work. There are many different points at which the process might be stopped.”
To design a good TMPRSS2 inhibitor, Chen and colleagues need to start with an accurate structure of the TMPRSS2 catalytic domain, that is, the part of the protein that acts on the spike protein.
While an experimental structure for TMPRSS2 is not currently available, computational methods can produce a highly accurate model for TMPRSS2 by using available structures of similar proteins. Through computational modeling, Chen and collaborators have begun exploring what kinds of molecules would theoretically make a good inhibitor. And the winning molecule doesn’t necessarily have to be something that prevents TMPRSS2 from cutting a protein; it could be something that TMPRSS2 cuts instead of the spike protein.
“We’ve made a short list of general candidates with certain desired chemical properties that are being tested theoretically,” Chen explains. “Candidates that meet theoretical criteria will then be tested experimentally—we’ll synthesize them and see how well they really work.”
Chen was always interested in viruses and was in high school when HIV came to prominence. He has worked on HIV therapeutics and points out that HIV therapy was transformed by the advent of drug cocktails designed to act on several different viral targets at once. This notion of hitting multiple targets is one of the major lessons learned from the past 30 years of HIV research and drug design. Now medical science is applying that concept to other viral infections, like SARS-CoV-2, to design multi-target treatment regimens. And TMPRSS2 is involved in cell entry not just for SARS-CoV-2 but also for SARS-CoV, MERS-CoV, and some influenza viruses, so drugs to inhibit its action could be included in a variety of therapeutic cocktails.
Making molecules
Whether trying to prevent infection or treating disease, computational molecular design goes hand in hand with synthesis of the actual molecules. Here too, Los Alamos scientists are leading the charge.
Los Alamos biochemist Ryszard Michalczyk oversees several molecular-synthesis projects. One of these is looking at drugs already approved by the Food and Drug Administration to see if any might be effective against COVID-19.
“We started with approved drugs because if one of them works, then the path forward is largely paved already,” explains Michalczyk. “In collaboration with other national labs, we are screening thousands of compounds, using simulation to see how well they would work, then ranking them according to expected efficacy. Any molecules deemed worthy of further experimentation are then either purchased, when possible, or synthesized at Los Alamos.”
“The NVBL molecular-design project is huge,” adds Iyer. “Some of the molecules being looked at can be purchased commercially, but many cannot. Los Alamos has always had a strong organic-synthesis capability, and we are doing nearly all of the synthesis for the NVBL.”
As molecular structures come in from computational designs, Laboratory chemists look at them from a synthesizability standpoint, to determine the best strategy, then begin making them. Sometimes the path is straightforward, having been previously published; other times there is no recipe, and the chemists have to invent one. Once the molecules are made and purified, they are sent to collaborating labs for testing, the results of which will inform subsequent iterations and improvements.
Many targets for therapeutics are enzymes, like TMPRSS2. In order to test whether the enzymes’ actions have been affected, the scientists need substrates—the molecules that enzymes bind to and modify in some way—so they need to synthesize those too.
Los Alamos chemist Jurgen Schmidt is involved in synthesizing small molecule peptides to use as substrates to test the activity of various enzymes. He’s also involved in several other molecular-synthesis projects, including non-enzyme drug candidates.
“We’ve got over 300 promising candidates so far, from computational predictions being done by us and other national labs,” Schmidt says. “When we get a hit—a molecule that looks like it will do what we want it to—we have to optimize its structure, affinity, and selectivity, then we have to make enough of it, up to gram quantities, to do toxicity and side-effect testing.”
Schmidt and collaborators are also developing unique suicide inhibitors for various viral enzyme targets. “Suicide inhibition” is a common method used in medicinal chemistry. It involves giving an enzyme a substrate that it can’t get rid of, thus preventing any further action by that enzyme molecule. The trick is to make a substrate analog that preserves the affinity between substrate and enzyme, while simultaneously adding chemical groups that will cause the substrate and enzyme to bind irreversibly to one another.
In yet another approach, Schmidt is also synthesizing peptides to directly help prevent and treat infection. He synthesized several peptides that mimic the region of the spike protein’s RBD that is the most antigenic—the region most recognized by host antibodies. The synthetic peptides were then mixed with SARS-CoV-2-positive human serum, which recognized and reacted with the synthetic peptides. This result is preliminary, but it suggests small synthetic peptides may represent a viable vaccine strategy, or a first-response treatment measure.
Looking ahead
Scientists across the globe are working around the clock to develop vaccines and drugs to end the COVID-19 pandemic. Never before have so many minds been seated at the same table, working on the same problem.
Los Alamos brings to that table not just brilliant minds but established capabilities and world-class facilities. The Laboratory excels at computational molecular design, rapid chemical synthesis, and on-demand manufacturing of custom targets, and it is now leveraging these resources to help solve the global crisis.
The first world-changing vaccine came when Edward Jenner followed a hunch. The next one will be no less world-changing, but will be much more elegantly designed. LDRD