virology blog

About viruses and viral disease

Vincent Racaniello

Molnupiravir, a SARS-CoV-2 antiviral drug, is mutagenic in cells

by 2 Comments

Molnupiravir might be the first highly effective antiviral drug given emergency use authorization for treatment of COVID-19. Should we be concerned about the results of a recent study which show that the drug is mutagenic in cells?

Molnupiravir is an orally available pro-drug of the nucleoside analog N4-hydroxycytidine (NHC). The latter is a nucleoside analogue which is incorporated into RNA by the viral RNA-dependent RNA polymerase (pictured above). Once incorporated into RNA, NHC is recognized as either C or U by the RNA polymerase. As a consequence, many mutations are introduced into the viral genome, causing lethal mutagenesis and inhibition of infectivity. NHC has been previously shown to have broad-spectrum anti-RNA virus activity and blocks transmission of influenza virus in a guinea pig model of infection. It has been shown to block SARS-CoV-2 transmission in ferrets and results of a phase 2/3 clinical trial look promising, leading to a request for emergency use authorization.

N4-hydroxycytidine could be metabolized by the host to produce the 2’-deoxyribonucleotide form, which could be incorporated into cellular DNA and lead to mutagenesis. To test this hypothesis, a mutagenesis assay was used in Chinese hamster ovary cells (CHO-K1). These cells have one copy of the gene encoding the enzyme hypoxanthine phosphoribosyltransferase (HPRT), which makes the cells sensitive to the base analog 6-thioguanine (6-TG). If NHC were mutagenic, changes in the HPRT gene would allow cells to survive in the presence of 6-TG.

Cells were exposed to NHC for 32 days and assayed for sensitivity to 6-TG. The drug conferred 6-TG resistance in a dose-dependent manner. Two other antivirals that are base analogs, ribavirin and favipiravir, displayed either no or modest mutagenic activity in this assay. Sequence analysis of HPRT mRNA revealed the presence of base changes.

Molnupiravir is a far more active coronavirus antiviral than favipiravir and ribavirin, yet NHC has the distinct ability of causing mutations in cell DNA. The concern is that such mutations could lead to cancer or birth defects in a developing fetus. Whether or not Molnupiravir might cause cancer in humans is not known. However Merck, the developer of Molnupiravir, is required to carry out a series of gene toxicity studies before phase I testing of the compound in humans. Included is the Ames test, which uses bacteria to assess mutagenic activity of a compound. Bacteria do have the enzyme which can convert NHC to the DNA form. The results of these safety studies will not be published until after the drug receives EUA, but presumably nothing was observed that would preclude clinical trials.

Consequently until we have further information about preclinical studies on NHC, we should be cautious in our interpretation of the results of mutagenesis assays in CHO cells.

Gain of function to build therapeutically useful viral vectors

by 3 Comments

Another excellent example of gain of function research is modification of a viral vector to make it more useful for human gene therapy.

Adenovirus associated virus (AAV) is the most commonly used vector for a variety of gene therapy applications, including gene replacement and gene editing. These small viruses, which comprise a single-stranded DNA genome surrounded by an icosahedral protein shell (illustrated), have a number of features that make them useful for gene therapy, including easy manipulation and growth in large quantities, and induction of persistent gene expression for long periods of time. An example is the drug Luxturna which is an AAV vector containing a retinal pigment gene that is used to treat some forms of blindness.

Naturally occurring AAV vectors have some limitations, including the propensity to become sequestered in the liver after systemic injection. Consequently reaching other organs with AAV vectors requires injection of large amounts of recombinant virus, which may be accompanied by toxicity.

A number of approaches have therefore been developed to modify AAV vectors so they preferentially infect other tissues. A number of muscle diseases would benefit from gene therapy and therefore AAV vectors have been developed to target that tissue. In one approach that I described before, ancestral AAV viruses were recovered and shown to efficiently infect muscle cell types.

A more recent approach involved modifying the AAV genome by inserting random stretches of 7 amino acids into the viral capsid protein, infecting mice with large libraries of these variants, and identifying those that best infect muscle. In addition, muscle-specific promoter sequences were also incorporated in the viral genome. After multiple rounds of infection of mice, recovery of viruses from muscle and reinfection, capsids were identified that more efficiently infect muscle tissue and less efficiently infect liver, in multiple mouse strains and in nonhuman primates. The new AAV capsids can rescue two different types of muscle disease in mice: Duchenne muscular dystrophy and X-linked myotubular myopathy after intramuscular inoculation.

The ability of these new AAV vectors to preferentially infect muscle cells depends on the presence of three added amino acids in the viral capsid, RGD. This three amino acid motif is known to bind cell surface proteins called integrins. Indeed, the increased efficiency of these engineered vectors depends on the presence of integrins on target cells.

These findings not only have produced better AAV vectors for targeting muscle tissue, but establish a strategy to engineer viruses to preferentially infect any tissue. The new vectors are the product of gain of function research: the original AAV has been given new properties. This work constitutes another example of the broad potential benefits of gain of function research in virology.

SARS-CoV-2 related viruses from bats in Laos

by 7 Comments

Various SARS-CoV-2 like viruses have been isolated from bats in China, Thailand, and Japan, but none have a spike protein that can bind ACE2 and allow entry into human cells. Sampling of bats in Laos has now revealed the presence of such viruses.

The genome of a virus called RaTG13, from Rhinolophus affinis bats in China, is the closest to SARS-CoV-2 (although infectious RaTG13 has never been isolated). However the spike receptor binding domain (RBD) encoded in this genome has low sequence similarity with that of SARS-CoV-2 and its affinity for ACE2 is very limited. RaTG13 is clearly not the proximal ancestor of SARS-CoV-2.

Sampling of 645 bats from limestone caves in Northern Laos (see map) yielded three Sarbecovirus genomes, called BANAL-52, -103, and -236, with high sequence similarity to SARS-CoV-2 and RaTG13. These three viral genomes were obtained from three different Rhinolophus species. Of the 17 amino acids that interact with the receptor binding domain of ACE2, 16 are conserved between SARS-CoV-2 and BANAL-52 or -103, and 15/17 conserved with BANAL-236. In contrast, only 11/17 RBD amino acids are conserved among SARS-CoV-2 and RaTG13. In other words, the RBD encoded in these BANAL genomes are closer to SARS-CoV-2 than that of any other known bat virus.

Binding affinity assays done with purified proteins revealed that the BANAL-52/103 and -236 spikes bind ACE2 with affinities in the low nanomolar range, comparable to reported values for the SARS-CoV-2 spike.

Lentiviral particles with the spike protein from BANAL-236 were able to bind and enter human cells producing ACE2. Entry of this virus was blocked by human sera containing antibodies to SARS-CoV-2 but not by control sera.

Infectious BANAL-236 virus was recovered from a bat fecal swab after inoculation of Vero cells, providing a rare virus isolate for a bat SARS-CoV-2 like virus. The virus will be useful for studying the biology of infection.

Recombination analysis demonstrated that the SARS-CoV-2 genome is a mosaic of at least five different genomes, including BANAL-52, -103, and -236, and the previously published RmYN02, RpYN06, and RaTG13, the latter all discovered in China. The implication of these observations is clear: SARS-CoV-2 likely arose from recombination of viruses circulating in different species of Rhinolophus bats in the limestone caves of South China and Southeast Asia. Because the RBD of the BANAL isolates can mediate binding to and entry into cells that produce ACE2, no host to host passage need be hypothesized to explain increased RBD affinity in an intermediate host before spillover into humans.

The spike proteins encoded in the BANAL isolates do not have the furin cleavage site found in the protein from SARS-CoV-2. Such sites might be present in viruses within these bat communities, but were missed due to insufficient sampling. Alternatively, it is possible that selection for the furin cleavage site occurred after spillover into humans or another intermediate host.

These findings further emphasize and demonstrate that SARS-CoV-2 came from Nature, not from a lab.

A zero time point is essential

by

The recent release of grant materials from EcoHealth Alliance pertaining to their research with the Wuhan Institute of Virology has been used to demonstrate that gain of function research was funded by the NIH. This conclusion is correct, although not all the experiments done would fall into this category.

One of the figures from the released grant materials shows the results of two experiments using recombinant SARS-like coronaviruses that were constructed in the laboratory. As I described previously, after the SARS-CoV pandemic of 2003, wildlife sampling efforts in China revealed many SARS-like coronaviruses in bats. To assess the potential of these viruses for infecting humans, their spike protein encoding genes were substituted into the SARS-like CoV WIV1.

The reproduction of these recombinant viruses were assessed in two experimental systems. ACE2 transgenic mice were inoculated with the recombinant viruses to assess their pathogenic potential, by measuring weight loss. The results (figure below, left panel) show that only one recombinant virus, carrying the spike gene from SHC014, caused more rapid weight loss than did WIV-1. Hence, SCH014 has a gain of function. However the recombinant virus 4231 caused the same rate of weight loss as WIV-1 and therefore does not have a gain of function. Recombinant virus WIV-16 caused no weight loss – its curve resembles that of uninfected mice – and therefore this virus has a loss of function.

The second experiment is not, in my opinion, proof that any of these recombinant viruses have gained a new function. The experiment was to infect ACE2 transgenic mice with WIV1 and the three recombinant viruses, and then remove lung tissue at different times after infection and determine viral RNA loads. The first problem with this experiment is that RT-PCR was used to measure viral RNA loads, which does not directly measure viral infectivity. Therefore it cannot be concluded that any of the viruses reproduced to higher levels than WIV1. More problematic is that there is no time zero – the measurement of lung tissues begins at day 2. Consequently, one cannot reliably compare viral RNA loads of any of the recombinant viruses with WIV1. When doing kinetic experiments, a time zero is ALWAYS needed, so you can see how much virus was actually instilled in the mice. It gives you a baseline from which to compare subsequent time points. Even though the authors state the amount of virus added to the mice, they do not know how much was actually added. Operator error, binding to tissues, and a host of other factors can interfere early in the infection. For example, it is possible that the viral RNA loads on day 0 were the same as in subsequent time points, which showed little change over the course of the experiment. One cannot conclude from this experiment anything about the viral RNA loads without a time zero. No viruses gained any function in this experiment.

Unfortunately, many expert virologists do not have a problem with the absence of time zero. When I spoke to a reporter about these results, and told her my interpretation, she told me other ‘experts’ did not agree with me. You will find many examples of published time courses without a time zero – they are worthless.

Why the difference between virology experts? I was trained to always have a time zero; the others were not. They continue to get away with it so they believe it is correct. It is not proper research practice and no amount of arguing will make it right.

Gain of function research explained

by

ferret

The term ‘gain of function’ is perhaps one of the most misunderstood in the scientific lexicon. I would like to explain what the phrase means from the perspective of a scientist who has done gain of function research for the past 40 years.

Gain of function (GoF) research gives an organism a new property or enhances an existing one. The organism can be a virus, bacterium, fungus, rodent, bird, fish or anything that can be experimentally manipulated (technically whales and elephants could be included in the definition but it would be very difficult to to GoF research on them).

Many have the impression that GoF research involves making an organism more deadly – for example, increasing the capacity of a virus to cause disease. That impression is incorrect. Certainly GoF research might lead to a more dangerous organism, but most of the time that is not the goal.

There are two broad approaches to GoF research. I’ll illustrate them with viruses but the principles could apply to any organism. In one approach, a virus is passaged in a host until a virus with different properties is obtained. An example is the adaptation of a strain of poliovirus – the type 2 Lansing strain – to replicate in mice and paralyze them. The Lansing strain does not infect mice, but an investigator passed the virus 99 times from mouse to mouse and ended up with a new strain that could now paralyze the mice. The new version of the virus had a new property – it could now infect mice. This experiment was GoF research.

Another way to do GoF research is to use recombinant DNA technology to engineer changes in the genome of the organism. In experiments done in my laboratory, we took a small piece of the genome of the mouse-adapted Lansing strain of poliovirus – coding for just eight amino acids – and spliced it into the genome of another poliovirus that is unable to infect mice. The recombinant virus from this experiment had a new property – the ability to infect mice. This experiment would also be classified as GoF.

An illustration of how GoF can be done on mice is our creation of transgenic mice susceptible to poliovirus. Mice cannot be infected with the virus because they do not produce the cell receptor for poliovirus. We introduced the human poliovirus receptor gene into the mouse germline, leading to mice that produce poliovirus receptor. After infection, these poliovirus receptor transgenic mice can be infected with poliovirus and develop paralysis. The mice have a new property – susceptibility to poliovirus infection. The mice are a product of a GoF experiment.

GoF research may have a myriad of useful outcomes. Do you want to make a different tasting beer? Modify an enzyme in the yeast used for fermentation. But in the past 30 years GoF research has received a bad name. The catalyst was a series of experiments on highly pathogenic avian H5N1 influenza viruses. These viruses rarely infect humans and do not transmit well among people. In experiments to understand what limited transmission, the virus was genetically modified and passaged among ferrets. The result was a virus that could transmit among ferrets by respiratory droplets. These GoF experiments were met with criticism, entirely unwarranted as the passaged viruses had lost their virulence for ferrets! Nevertheless since then a dark cloud has unjustifiably hung over all GoF research.

GoF research has been in the press again recently as a consequence of the COVID-19 pandemic. After the SARS-CoV pandemic of 2003, wildlife sampling efforts in China revealed many SARS-like coronaviruses in bats. To assess the potential of these viruses for infecting humans, their spike protein encoding genes were substituted into the SARS-like CoV WIV1. These recombinant viruses reproduced in human airway cells – no different from WIV1 – but at least one caused more severe disease in mice. Consequently these are GoF experiments. Some have suggested that such GoF work gave rise to SARS-CoV-2 in a lab, but this notion is impossible, as none of these viruses are close enough to be a precursor of the current pandemic virus.

The production of recombinant coronaviruses to assess pandemic potential was carried out in several laboratories, all funded by the NIH. Recently Dr. Anthony Fauci told Congress that the NIH did not fund GoF coronavirus research. The press has suggested that he lied, but the truth is that his definition of GoF research is that it only involves passaged of organisms in animals. This interpretation is not correct but being wrong does not mean you are lying.

I want readers to understand that the goals of GoF research are laudable, and only a small subset has the potential to harm humans. Consequently these experiments are highly regulated and carried out under high levels of biological containment. GoF is not a dirty word.

More coronaviruses in rodents

by

This week President Biden received the report he commissioned on the origin of SARS-CoV-2. As I predicted three months ago, nothing was learned, because to find the origin of the virus it is necessary to conduct extensive wildlife surveys. Consequently this week I would rather write about a new study that identified coronaviruses in rodents.

Bats receive a great deal of attention because they harbor multiple coronaviruses (as well as many other viruses). Rodents receive far less attention, undeservedly so, because they constitute 40% of mammalian species (compared with 20% for bats), are known to host zoonotic viruses, and have frequent contact with humans. The human coronaviruses OC43 and HKU1, ubiquitous agents of the common cold, are thought to have originated from rodents.

To address gaps in our knowledge about coronaviruses that infect rodents, samples were obtained from 297 individuals animals representing eight species in three municipalities in the south of China. Viral RNA was extracted from oral and anal swabs and screened for coronavirus (CoV) RNA by RT-PCR using universal primers. Thirty-five of 297 rodents were found to harbor CoV RNA, most of them from the brown rat, Rattus norvegicus (pictured). Most interesting was that 34 of the 35 RNA positive rodents were captured in hotels and passenger stations. Just one positive animal came from a bamboo rat farm where animals are raised for food.

Unfortunately the authors were not able to recover infectious virus in cells in culture; however only primate cells, not rodent cells were used, possibly explaining this failure.

Whole genome sequencing was done for three distinct CoVs obtained from brown rats. The genome of RCoV-GCCDC3 has 96% similarity to the China Rattus CoV HKU24, while RCoV-GCCDC5 is 97% identical to the Luchang Rn rat CoV (LRNV), a recombinant virus identified in Zhejiang Province. RCoV-GCCDC4 has low similarity to known CoVs and appears to be a novel isolate.

Phylogenetic analysis reveals that RCoV-GCCDC3 and 4 belongs to the Betacoronavirus genus lineage A, while RCoV-GCCDC5 belongs to the Alphacoronavirus genus. Betacoronavirus lineage lineage A also contains the human CoVs OC43 and HKU1.

The predicted spike protein of RCoV-GCCDC4 was found to contain a polybasic cleavage site, RARR, between the S1 and S2 regions. Cleavage of such sites by furin proteases may enhance virus entry into cells. SARS-CoV-2 is the only known member of Betacoronavirus lineage B that possesses a polybasic cleavage site. RCoV-GCCDC4 also appears to be a recombinant virus as it harbors sequences similar to both MHV-1 (mouse hepatitis virus, identified in mice in 1949) and LRNV. The CoVs are known to undergo extensive recombination and this finding further emphasizes its potential to create novel viruses.

Perhaps the most important finding from this survey is that many brown rats were positive for CoV RNA. This species inhabits residential and indoor areas and has frequent contact with humans. It will be important to determine if the spike protein identified in this study can bind human cells and allow virus entry. Also essential is to determine if virus can be recovered from the full length genome sequences and if so whether the virus can reproduce in human cells. Such experiments address the pandemic potential of rodent CoV.

These findings also underscore the need to conduct more extensive virus surveillance at the human-livestock-wildlife interface to identify potential zoonotic pathogens.

Are COVID-19 vaccine boosters needed?

by

The US Centers for Disease Control have concluded that a third dose of COVID-19 vaccine will be needed for protection against disease, but the science says otherwise.

A review of the immune responses to infection will help explain why vaccine boosters are not needed. The graph below shows the relative concentrations of antibody and T cells (Y axis) as a function of time after first viral infection. Antibody levels and numbers of activated T cells initially rise, then decline after the primary viral infection is cleared. Reinfections at later times (years later) are marked by a rapid immune response because of the reanimation of memory cells. The host is infected, but virus is cleared rapidly and severe disease is prevented.

The same principles illustrated in this graph apply to the immune response after vaccination. After the first and second vaccine doses, antibodies and T cells are produced over the course of weeks and months and slowly decline to low but not non-zero levels. If a vaccinated individual is reinfected later, the memory B and T cells proliferate and virus-specific antibodies and T cells are produced. The memory response takes several days and therefore infection is not prevented. However virus reproduction in the immunized host is substantially reduced and severe disease does not occur. In addition, due to reduced viral replication in the face of a memory immune response, transmission is markedly reduced or eliminated. The impairment of transmission is the basis of herd immunity.

The COVID-19 vaccines were tested in phase III trials for their ability to prevent COVID-19 – a disease. They were not tested for their ability to prevent infection. Most human viral vaccines do not prevent infection, but they do prevent disease.

After multiple COVID-19 vaccines received emergency use authorization, studies in several countries were done to determine if immunity prevented infection. When the studies were done shortly after vaccination, antibody and T cell levels were high, and infection was prevented. But as the immune response waned (see figure above), vaccine efficiency against infection declined. This observation was incorrectly interpreted by many to indicate that the vaccines were failing. Largely ignored was the fact that vaccinated individuals were still protected from developing serious COVID-19 disease and death.

Recently public health officials in the US made the following statement:

“The available data make very clear that protection against SARS-CoV-2 infection begins to decrease over time following the initial doses of vaccination, and in association with the dominance of the Delta variant, we are starting to see evidence of reduced protection against mild and moderate disease. Based on our latest assessment, the current protection against severe disease, hospitalization, and death could diminish in the months ahead, especially among those who are at higher risk or were vaccinated during the earlier phases of the vaccination rollout. For that reason, we conclude that a booster shot will be needed to maximize vaccine-induced protection and prolong its durability.”

The decline of protection against infection is no cause for concern; as noted above, such a response occurs for most human viral vaccines. So that alone cannot be used to justify a third vaccine dose.

I am unaware of evidence that ‘we are starting to see evidence of reduced protection against mild and moderate disease’. Furthermore, that ‘current protection against severe disease, hospitalization, and death COULD diminish in the months ahead’ (capitals mine) cannot be used as a reason that a booster will be needed.

This week CDC released the results of two studies which demonstrate that vaccine protection against hospitalization remains strong. In one study, COVID-19 cases and hospitalizations among adults in New York were studied from May to July 2021. The results show that vaccine effectiveness against hospitalization in this state was stable: 91.9% in May to 95.3% in July. Despite increasing numbers of cases caused by the delta variant of SARS-CoV-2, hospitalization of vaccinated individuals did not increase. As would be expected as antibody and T cell levels decline, vaccine effectiveness against infection, determined by diagnostic tests, decreased from 91.7% in May to 79.8% in July.

A second study examined effectiveness of Pfizer-BioNTech and Moderna mRNA vaccines from March to July 2021 against hospitalization for COVID-19 in 18 US states. No decline in vaccine effectiveness against COVID-19 hospitalization was observed over 24 weeks: vaccine effectiveness was 86% 2–12 weeks after vaccination and 84% at 13–24 weeks.

These two studies provide no support for the need of a third vaccine dose in the US. If a third dose were given, the result would be an increase in virus-specific antibodies and T cells which might prevent infection for several months. But as antibody and T cell levels decline, which is inevitable, we would be in the same situation we are in now – infections occur but hospitalization and death is prevented. Would we then enter an endless cycle of giving boosters every 8 months? That is not a smart way to end this pandemic. Rather, measures to ensure that 80% of the population is fully vaccinated must be deployed.

The call for vaccine boosters in the US is another example of how some public health measures deployed during this pandemic are not based on science.

Note added after further thought: Sera from individuals who have been infected and subsequently vaccinated have broad neutralization activity against all known variants of concern. Whether boosting with existing vaccines could achieve similar breadth and potency is unknown but should be determined before giving millions of individuals a third vaccine dose.

Estimate of infectiousness during COVID-19

by

Understanding the transmission of SARS-CoV-2 is complicated by the large numbers of presymptomatic, asymptomatic, and mildly symptomatic (PAMS) patients. The reproductive number, R0, is a measure of population-level dynamics, but it cannot provide information on infectiousness of different groups such as PAMS subjects; when peak infectiousness occurs; and the effect of intrinsic properties of the virus. However information on infectiousness can be provided by a study of viral RNA load and infectivity in cell culture.

A study of 25,381 German COVID-19 cases was done to provide information on infectiousness during the course of infection. These included 6110 PAMS cases, 9159 hospitalized patients, and a series of times samples from hospitalized patients. Viral RNA loads – defined as RNA copies per swab – were determined by RT-PCR. The association between viral RNA load and probability of virus isolation in cell culture was estimated by using previously determined cell culture isolation data.

The results show that PAMS subjects have viral RNA loads (pictured) and predicted infectiousness, at the first positive test, slightly less than those of hospitalized patients. Children had similar viral RNA loads and predicted cell culture infectivity.

A small fraction of individuals’ first-positive viral RNA loads – 8%- had 10e9 RNA copies per swab. This observation is in line with previous findings that 15% of index cases are responsible for 80% of transmission, and that 8-9% of infected individuals carry 90% of the total viral load.

Time-course analysis revealed that viral RNA load peaks 1 to 3 days before onset of symptoms. From this peak, viral RNA loads decline 0.17 log10 units per day. Previous studies have shown that infectious virus is typically not recovered in cell culture beyond 10 days from symptom onset.

This study also examined viral RNA loads in 1533 patients with alpha variant infections. Compared with non-alpha infections, patients infected with the alpha variant had 10 times higher viral RNA loads and a 2.6 fold higher estimated cell culture infectivity. However, the impact of an increase in viral RNA load is, as the authors write, ‘dependent on context’ and the increase in cell culture infection probability is a ‘proxy indication of potentially higher transmissibility’. Complicating this analysis is the possibility that the correlation of viral RNA loads with cell culture infectivity might differ among variants.

The most important outcome of this study is that PAMS subjects – who in this study were tested at walk-in centers – are as infectious as hospitalized patients. As these individuals circulate in the community they can clearly trigger outbreaks of infection.

Paul and the Mosquitos

by

From the authors of Paul Has Measles and Paul Stays Home comes Paul and the Mosquitos, an illustrated book for children about mosquito-borne diseases.

In his camp, Paul and his friends discuss which is the most dangerous animal of all. They would never have imagined it would be the mosquito. Why are they dangerous and what can we do to prevent the diseases transmitted by them?

Paul and the Mosquitoes is written by Susana López, Selene Zárate, and Martha Yocupicio, with illustrations by Eva Lobatón.

A pdf of Paul and the Mosquitos can be downloaded free of charge here.

T cells will save us from COVID-19, part 3

by

In the two previous installments (one, two) of what has now become my praise of T cells, I explained that the SARS-CoV-2 protein sequences recognized by T cells do not change, likely explaining why vaccines prevent serious disease and death caused by any variant. Today I will explain that virus-specific T cells appear a week after mRNA vaccination when neutralizing antibodies are weakly detectable. Their presence is likely responsible for vaccine-mediated protection against severe disease.

Since the first COVID-19 vaccines were tested in clinical trials, much has been made of their ability to block infection. At ten days after the first vaccine dose, clear protection against severe disease is observed. But at this early time, neutralizing antibodies can barely be detected. That job is likely being done by T cells, the other arm of the adaptive immune response.

T lymphocytes come in two broad types: CD8+ and CD4+. The former, also known as cytotoxic T lymphocytes (CTLs) can recognize a virus infected cell and kill it (pictured). The latter T cells produce cytokines that are important for maturation of both CTLs and B cells, the antibody-producing cells. The importance of antibody versus T cells in controlling infection depends on the virus. For many viruses, antibody can block infection, but T cells are important for recovery. In COVID-19, antibodies rise late in the course of infection, when virus titers are already declining. This observation is in line with the resolution of infection by T cells.

People with agammaglobulinemia, who cannot make antibodies, or who have been depleted of B cells do not have a serious course of COVID-19, further supporting a role for T cells in preventing severe disease.

During the trials of mRNA vaccines it was observed that protection against severe COVID-19 arose at 10 days after the first inoculation, a time when neutralizing antibodies are barely detectable in serum. The logical candidate for this effect is the T cells.

To address this question, a group of mRNA vaccine recipients were sampled at different times after the first and second doses. CD8+ T cells in blood were examined for the ability to recognize a few spike-specific epitopes. A rapid and substantial induction of CD8+ T cells in these individuals was observed beginning at 6-8 days after the first vaccine dose. At this time, neutralizing antibodies in sera of these subjects were barely detectable. CD8+ T cell numbers are substantially increased by the second vaccine dose, which also leads to much higher levels of neutralizing antibodies. CD4+ T cells are also detected after the first vaccine dose and likely coordinate the expansion of CD8+ T cells and B cells.

mRNA vaccination also induces spike-specific memory B and T cells which circulate for at least several months. Their exact longevity remains to be determined by longer term studies.

The gradual escape of multiple variants of concern from neutralization by sera from vaccinated individuals has been met with alarm by public health officials. At least one prominent leader has suggested that the virus might even mutate so as to escape vaccine protection. An even cursory examination of the facts reveals otherwise. Despite reduced neutralization by vaccine-induced antibodies, vaccine recipients are still protected from severe disease and death by all variants. In contrast to changing B cell epitopes, T cell epitopes do not change in any variant. Furthermore, virus-specific T cells are induced early after the first mRNA vaccine dose, and confer protection against severe COVID-19 at a time when neutralizing antibodies are barely detectable. These observations indicate that T cells will save us from COVID-19.

In vaccinology, the correlate of protection refers to what is needed – antibodies or T cells – to protect from infection or disease. High levels of antibodies appear to correlate with protection against infection by SARS-CoV-2. However, when antibody levels decline, as they always do after infection or vaccination, infection can no longer be prevented. In this case, protection against severe disease and death is accomplished by T cells.