With COVID-19 on the scene, talk of other respiratory viruses has taken a backseat. But this doesn’t change the fact that they are as present as ever, lurking in the background. Influenza is one such virus, causing an estimated 1 billion cases globally every year—of these, upwards of 500,000 end up being fatal. Although lockdowns and social distancing efforts aimed at curbing the spread of COVID-19 also helped put a dent in seasonal influenza epidemics, the flu is now returning with full force (Figure 1).
Recently there has been some success on the road to developing a much sought-after broadly neutralizing vaccine against all influenza A viruses. If successful, it may act as a precursor to a truly universal flu vaccine, one that protects against all types and subtypes of the virus. The breakthrough may also provide a blueprint for developing a COVID-19 vaccine that retains its efficacy in the face of new variants.
There are seven types of influenza, four of which can infect mammals: A, B, C, and D. Influenza A and B are the most common, driving yearly seasonal flu epidemics. Influenza C infections in humans are rare and are thought to cause relatively mild disease. Influenza D mainly infects cattle with no recorded human infections to date.
On the whole, influenza A viruses are considered the biggest threat. All of the major influenza pandemics of the last century have been caused by influenza A viruses, the most notable being the Spanish flu pandemic of 1918 which led to the death of roughly 50 million people, one-tenth of the entire global population at the time.
Influenza A viruses are also particularly adept at evading our immune system; they have a high rate of genetic mutation, resulting in a process known as antigenic variation. This is what allows the virus to reinfect us again every season, it is what makes the virus volatile and capable of causing the occasional pandemic, and it is what makes it so difficult for us to develop long-lasting vaccines. To help explain antigenic variation, it’s worth quickly looking at the structure of the virus (Figure 2).
First, all influenza viruses have single-stranded, negative-sense RNA. This is the genetic material used by the virus to replicate. In the case of influenza A, the genome is broken into eight separate segments. Surrounding and protecting these RNA strands is the viral envelope. The envelope contains two key surface proteins hemagglutinin (HA) and neuraminidase (NA)—the former helps the virus enter our cells, and the latter helps newly-formed viral particles leave the cell.
There are 18 different subtypes of the hemagglutinin surface protein, aptly named H1 through H18. These are split into two groups, organized according to their evolutionary history. Crucially, immunity against one group does not confer immunity against the other group. Any universal influenza A vaccine has to account for this fact. Neuraminidase, in turn, has 11 different subtypes—N1 through N11.
It is on the basis of these surface protein subtypes that influenza A viruses are classified. For example, the viral subtype that caused the 1918 Spanish flu had an H1 hemagglutinin protein and a N1 neuraminidase protein, making it an influenza A (H1N1) virus. The influenza virus that caused the 1968 pandemic, in contrast, had a different HA subtype, H3, and a different NA subtype as well, N2. As a result, it is classified as influenza A (H3N2). These are the two most common subtypes of influenza A, responsible for the majority of yearly infections.
Influenza A subtypes can be broken down even further into strains. This is determined by the sum of their genetic material, known as the genome. Two viruses may have the same broad hemagglutinin and neuraminidase subtypes, but have subtle differences throughout their genome. For example, the H1N1 subtype was responsible for both the 1918 and 2009 influenza pandemics, but in each case it was a different strain of the virus. (Figure 3).
So, how does all of this relate to antigenic variation and why does it make it difficult for us to develop a universal influenza A vaccine?
The HA and NA surface proteins are recognized by our immune system as foreign substances—known as antigens. Detection of antigens in our body leads to the activation of various immune responses aimed at clearing the threat. One kind of immune response involves training B cells to produce antibodies that specifically target the antigens in question. In a sense, our B cells develop a type of “immune memory” towards the virus. Doing so allows our immune system to react more quickly and effectively in case of reinfection, blocking the virus before it has the chance to get us sick.
As hinted at above, influenza A viruses are notorious for their ability to mutate. This happens when the virus makes small “mistakes” while copying its genetic information during replication. Over time, these small mutations to the genome can accumulate, changing the structure of various parts of the virus, including HA and NA. If it changes enough, the virus will be able to evade our immune system; our B cells will no longer be able to recognize the antigens, nullifying what they may have learned during a prior infection. This is a type of antigenic variation called “drift”, and it is the reason we can repeatedly become sick with the flu.
Vaccination against influenza works by exposing our immune system to killed or weakened forms of the virus. In the same way a natural infection would, this helps our immune system build up a memory of the viral antigens and teaches it to keep an eye out for them in the future. But like the antibodies produced after natural infection, those produced in response to vaccination are also rendered useless over time by antigenic drift; as before, the antigens become too different to be recognized, forcing B cells to rebuild their immune memory all over again. This is why scientists have to develop a new flu vaccine every year, so that our immune system can build up its memory based on the most recent circulating influenza antigens.
There’s also another, more drastic kind of antigenic variation called antigenic “shift”. Whereas drift happens over an extended period of time, via the accumulation of small changes, antigenic shift can lead to substantial changes very quickly. How? Usually when a cell is infected by an influenza virus, eight gene segments come in, these eight segments are copied and packaged into a newly formed viral particle, and the new particle then leaves the cell to go infect other cells. But sometimes a single cell—whether human or avian or porcine—can become infected by two different influenza A viruses at the same time. In such cases, the two viruses can end up “exchanging” RNA segments with one another. This is called reassortment, a process that leads to the creation of a third, novel virus that’s distinct from both of the parent strains. Often this happens when an influenza A virus jumps between different animal hosts. For example, the 1957 influenza pandemic was caused by a novel subtype, H2N2, composed partly of genes from the 1918 strain and partly from new genes gained while infecting birds (Figure 3).
Such sudden jumps in genetic code can lead to influenza viruses that look very different to those usually circulating, rendering the surface proteins unrecognizable to our immune system. Once again, our immune system is confronted with what is essentially an entirely new threat and has to build its immune memory from the ground up. Whereas antigenic drift is responsible for yearly seasonal flu outbreaks, antigenic shift is often responsible for new influenza pandemics. Any universal influenza A vaccine has to successfully account for both antigenic drift and antigenic shift.
The seasonal approach
In an attempt to circumvent the issue of antigenic variation, vaccine manufacturers update the flu shot each year based on the latest circulating influenza strains. In the United States, all influenza vaccines are “quadrivalent”, meaning they contain four different influenza viruses. This is done to broaden protection against the various influenza subtypes and lineages known to drive seasonal outbreaks: influenza A (H1N1), influenza A (H3N2), influenza B/Victoria, and influenza B/Yamagata. Quadrivalent vaccines will also protect against any other influenza viruses that are antigenically similar.
Unfortunately, vaccines produced in this way are nowhere near as effective as we might hope. At best, they protect 60% of people from illness, but this number can, and often does, drop much lower. For the influenza A (H3N2) subtype, vaccine effectiveness hovers around 33%. So where are we going wrong?
Missing the target: Egg-based vaccines
Selection of candidate vaccine viruses (CVVs) is only one part of the equation, growing them is another. This is no simple feat considering they need to be available in bulk, enough to make millions of vaccines. For the past 70 years, the majority of manufacturers have turned to chicken eggs in order to achieve the necessary growth (Figure 4). The candidate vaccine viruses are injected into fertilized hen’s eggs and left to incubate for a few days. During this period, the viruses are able to replicate. The fluid in the eggs is then extracted and the viruses are “killed” (inactivated). Finally, the antigen of choice—usually the hemagglutinin surface protein—is isolated from the killed viruses and purified, making it ready for use in vaccines. Even now, most flu vaccines continue to be egg-based.
In response to these issues, manufacturers have tried to develop new production methods that avoid using chicken eggs to culture candidate viruses. Cell-based and recombinant vaccines aim to resolve the issues on the production side of things. The former by skipping the need for eggs, and by extension, the threat of egg-adapted changes. The latter by skipping the need for eggs as well as cutting down the time it takes to produce the vaccines, reducing the risk of escape mutants.
However, there is a paucity of comparative research on these new production methods, making it difficult to draw any firm conclusions, but early findings suggest recombinant vaccines may be more effective than traditional egg-based and cell-based vaccines, including improved antibody production.
Possibility of mRNA influenza vaccines?
Messenger RNA (mRNA) technology could also help solve the issues associated with slow production. Unlike traditional flu vaccines, mRNA vaccines do not depend on the cultivation of viruses. Instead, they are based on the synthesis of carefully designed mRNA. The process is more akin to how a pharmaceutical drug like aspirin is produced than it is to how current flu shots are produced.
As soon as researchers have a complete sequence of a virus’ genome, they can isolate the section they want to use for vaccination—usually, the gene corresponding to the surface protein the virus uses to bind and enter our cells—and artificially synthesize it in the lab. The gene can then be inserted into a small, circular piece of DNA known as a plasmid. Plasmids are easy to replicate, allowing researchers to produce large amounts of the target viral gene at once. Finally, the plasmids are put into a reactor with a molecule that transcribes the viral RNA into mRNA. Any remaining plasmid material is broken down and the mRNA put through a series of purifiers before being packaged into the protective lipid shell. The whole process takes a matter of weeks rather than months. Or maybe even less: in 2013, a team of scientists at Novartis generated an mRNA vaccine candidate in the span of eight days.
Unlike COVID-19 outbreaks, which are usually driven by one dominant circulating variant, seasonal flu outbreaks are characterized by multiple cocirculating strains. These strains are derived from the four main influenza viruses: influenza A(H3N2), influenza A(H1N1), and two influenza B viruses. To account for differences in the hemagglutinin surface protein between strains, traditional flu shots are “quadrivalent”, meaning they contain an inactivated version of each of the four viruses. But a synthesized mRNA sequence can only make one particular protein; an effective mRNA influenza vaccine would need four different sequences to make four different HA proteins. A 2020 study led by researchers at the Icahn School of Medicine at Mount Sinai suggests this shouldn’t be an issue. The team of scientists manufactured an mRNA vaccine against an influenza subtype that encoded four different proteins at once. They have since repeated the experiment for two additional influenza subtypes, and they plan to test a combined vaccine that encodes for a total of 10 to 12 different antigens across subtypes. This is part of a larger movement towards mRNA flu shots.
What remains unclear is whether mRNA technology by itself could address a major issue plaguing current influenza vaccines: waning immunity. A 2018 meta-analysis of influenza vaccine durability found that protection can vanish as early as 90 days post vaccination. Recombinant and inactivated vaccines both suffer from this issue, even though they trigger immune responses in different ways. If the COVID-19 mRNA vaccines are anything to go by, waning immunity will likely continue to be a factor even if influenza vaccines made the switch to mRNA technology.
Nasal influenza vaccines
There has been tremendous excitement around mucosal vaccines, with many in the field pinning their hopes on long-term protection from infection as well as severe disease. I do not share the same unbridled enthusiasm that nasal vaccines will provide long-lasting protection from infection and protection from all variants for several reasons.
The first nasal flu vaccine, FluMist, was approved by the U.S. Food and Drug Administration (FDA) in 2012. By 2016 it had been pulled off the market, not to reappear until 2018. The reason? Decreased vaccine effectiveness. Combined data from 2013 to 2016 indicates that nasal spray vaccine effectiveness was roughly 26% in children between the ages of 12 and 17. This is compared to 51% effectiveness for inactivated vaccines, administered intramuscularly.
Although these low numbers were brushed off as being a result of poorly matched vaccine virus strains, there may also be other reasons the nasal spray influenza vaccine hasn’t quite lived up to expectations.
For one, IgA is typically quite short-lived; it is produced only for a brief period after exposure. Any barrier immunity that the nasal spray vaccine does offer us will likely fade as soon as IgA numbers drop. And durability is already a point of concern with traditional, inactivated flu vaccines, which depend on the longer-lived circulating IgG class of antibodies for protection from infection. So, we cannot expect it to be all that durable.
Another worry is that the nasal spray vaccine may be cleared by our immune system before it has the chance to complete its job. Adults who have previously been exposed to the flu—and that means more or less every adult—will still have at least partial mucosal immunity against the influenza viruses. The attenuated, or “weakened”, viruses used in the nasal spray vaccine might simply be getting neutralized before they can infect nasal cells, preventing our B cells from updating sIgA to match the viruses circulating that season.
On a more theoretical note, one might also question the belief that vaccine-induced mucosal immunity can provide lasting protection if natural infection, through the same path of entry, fails to do so. Many people are infected by influenza on a seasonal basis, with waning immunity and viral mutation leading to renewed vulnerability year after year. This happens even though they would have built up some degree of mucosal immunity during prior infection. If prior infection through the nose doesn’t protect us from reinfection, why would a nasal vaccine?
Vaccine platform vs. vaccine antigens
How the antigen is delivered is one side of the coin, what antigen gets chosen in the first place is the other side. Delivery platforms continue to evolve, yet progress in antigen selection has been lagging. To optimize our influenza vaccines, we need both.
Targeting conserved regions: HA stalk and M2e
Whether egg-based, cell-based, recombinant or otherwise, most of the currently available flu shots use the hemagglutinin (HA) surface protein as their antigen of choice. This makes sense; influenza viruses depend on the hemagglutinin protein to gain entry into our cells. Block the protein, and you have a chance at blocking infection. The issue with using hemagglutinin as the primary antigen is that it is extremely mutable. The structure of the protein can change to avoid previously established immunity, including B cell-derived antibodies. Thus, the seasonal flu cycle and yearly flu shots.
The universal vaccine approach aims to change things. Instead of using the head of the hemagglutinin protein as the antigen, researchers working towards an universal flu vaccine look for viral proteins, or specific protein regions, that are highly conserved across strains and across time—essentially, antigens that only very rarely mutate. If successful, universal vaccines could provide a much broader immunity, eliminating the need for new vaccines every year.
So far, there have been two promising contenders: the stalk portion of the hemagglutinin protein and, for influenza A viruses, the extracellular region of the Matrix-2 protein (M2e).
The hemagglutinin surface protein can be separated into two regions, one being the highly mutable head (HA1) and the other being the stable stem (HA2) (Figure 5). The head region contains the receptor binding domain (RBD) which the influenza viruses use to bind to our cells. Using this area as the reference antigen creates a highly specific immune response, with antibodies that can block the RBD and, if present in sufficient numbers, may prevent infection. But the specificity comes at the cost of breadth. Even a single mutation can throw off our immune system, rendering the antibodies practically ineffective.
Where the HA head domain is used to bind to our cells, the stem region is used for fusion—the process by which the virus “injects” its genetic material into the host cell cytoplasm, kicking off viral replication. Briefly, once a viral particle has bound to a host cell it is “ingested” by a section of the cell membrane that then buds off to form a membrane “bubble”. The resulting bubble is called a vesicle, and foreign substances internalized in this way are usually promptly digested and disposed of by the cell. But influenza viruses are crafty and have developed a way to use the acid that normally helps our cells digest such vesicle contents to instead free up the RNA inside the viral particle. This is where the stem portion of the hemagglutinin protein comes into play. Once the RNA has been freed up, the HA stem is used as a lever of sorts to pull the viral membrane into the cellular membrane, finally creating a “tear” that lets the genetic material spill out into the cytoplasm. Once there, the virus can begin replicating before spreading to other cells.
A vaccine that targets the HA stem and blocks it from initiating fusion would interrupt the replication process, preventing further infection and stopping the virus in its tracks. Indeed, researchers have already developed vaccines based on “headless” HA proteins as antigens, but results so far have been slightly underwhelming; although the protection the vaccines have provided has been fairly broad, it has not been very strong. Mice inoculated with vaccines based on HA stalk as an antigen still lose noticeable amounts of body weight, a common symptom of illness. This has been particularly true for influenza A viruses with group 2 HA proteins (Figure 5), which also happen to be those causing the majority of seasonal infections.
In response, attention has also been directed towards the matrix-2 (M2) protein (Figure 6). This protein spans the membrane of the influenza viral particle and functions to internalize the acid I mentioned above. Without the M2 protein, the viral RNA cannot be untethered from its membrane and replication is once again impaired. Vaccines that target the extracellular domain of M2 (M2e) have fared better than those that target the HA head alone.
More recently, efforts have been made to merge HA stem and M2e into one single molecule, to boost immunogenicity. Linking them chemically has yielded promising results, with higher protective potency than HA stem or M2e vaccines alone.
Design of the antigen: Perfecting the “headless” HA protein
Interest in the stem portion of HA has its roots in the discovery of human broadly neutralizing antibodies (bnAbs) that target the region. This discovery revitalized hopes of developing universal flu vaccines by giving researchers an entry point for the production of a universal antigen. Since the full HA protein only rarely elicits broadly neutralizing antibodies, various approaches have emerged to expose and properly present the stem domain to our immune system. One of the more successful strategies involves knocking off the HA head domain while ensuring structural stability of the leftover HA stem.
But this is easier said than done. The head of the HA protein stabilizes the prefusion protein resulting in a tightly coiled triplex of three proteins. Broadly neutralizing monoclonal antibodies only recognize this tightly coiled triplex and are lost once the fusion process commences or the head of the HA protein is removed.
To perfect the hemagglutinin portion of their antigen, Subbiah and colleagues drew upon previous research by Impagliazzo et al. and Yassine et al. Both groups managed to engineer stable, headless HA stem antigens—called “mini-HAs”—that exhibited structural and antibody binding properties similar to full-length HA. Antibodies produced in response to these mini-HAs managed to protect against a wide range of HAs in mice and nonhuman primates.
Figure 7 outlines the various stages of Impagliazzo et al.’s design, culminating in the creation of a stable, headless “mini-HA” stem that retains all the relevant neutralizing binding sites. Figure 8 outlines the design stages of Yassine et al, who achieved great minimization of the HA protein, ultimately succeeding in having the stem domain contribute 94% of the total surface area of their antigen.
Forming a stable trimer and preventing protein aggregation
Influenza hemagglutinin is naturally a trimeric molecule, meaning it is composed of three distinct hemagglutinin proteins bound together. In order to accurately mimic this three-part structure, Subbiah et al. excluded the C-terminus of the HA stem portion and instead replaced it with the foldon sequence of a virus that infects bacteria.
When added to the end of a monomer that has a propensity to dissociate, the foldon sequence stabilizes it and allows it to form a trimer (Figure 9). Failure to do so may lead the headless HA stem antigen to lose its structure, and by extension its epitope sites. If our antibodies cannot bind to the stem, they cannot learn about its structure and thus cannot take on the specific shape needed to protect us from future reinfection.
The researchers also added specific point mutations along HA1 head and HA2 stalk to prevent protein aggregation issues. All of the mutations were in regions outside of those recognized by broadly neutralizing antibodies. Each one of the mutations was in a long helical region with a tendency to self-aggregate. By changing the amino acids in these stretches, they substantially reduced non-specific aggregation and instead favored the stable form of the protein.
In addition to this, they mutated a cysteine residue and replaced it with a serine residue, again strengthening the stability of the protein and minimizing the chance of any unwanted disulfide bonds from forming.
Addition of M2e
To broaden the neutralization of their antigen and increase its potency, the researchers added in two identical segments of the M2 channel protein. The segments they chose correspond to a 23 amino acid-long sequence that encodes the epitopes for the extracellular region of the protein. This area protrudes from the viral membrane, allowing it to be bound by neutralizing antibodies.
Crucially, M2e has previously been shown to confer cross-group immunity. Influenza A viruses are split up into two groups according to phylogenetic differences in their HA protein. Usually immunity against group 1 HAs (H1, H2, H5, H6, H8, H9, H11, H12, H13, H16, H17, H18) does not confer immunity against group 2 HAs (H3, H4, H7, H10, H14, H15). This means that infection by an influenza A/(H3N2) virus, for example, generally does not confer protection against a strain of influenza A/(H1N1).
Finally, Subbiah et al. connected the various domains of the M2e-H3 construct using flexible linkers, which are small amino acids that function like bungee cords; they keep the proteins tethered together but are flexible enough not to interfere with the folding process.
The full genetic sequence of the M2e-H3 antigen can be seen in Figure 10, and the structure of the antigen—as compared to wild type hemagglutinin—can be seen in Figure 11.
Production and purification of the antigen
For ease of purification of the antigen, the researchers added a histidine tag (6xHis) to the N-terminal of the M2e-H3 stem domain. A histidine tag is made up of a DNA sequence that encodes six to nine histidine residues. These histidine residues bind to several different types of immobilized metal ions. So, once a protein has been “His-tagged” it can be easily identified and separated from other proteins by exposure to such metal ions. The histidine tag can then be removed from the protein or it can be retained, either way it does not affect the antigenicity of the protein.
Next, Subbiah et al. synthesized a codon-optimized gene encoding the M2e-H3 stem protein. This gene is then inserted into escherichia coli (E. coli) for quick and easy mass production.
Does it retain antigenicity?
After extraction and purification, Subbiah et al. began testing the M2e-H3 stalk molecule for its ability to be recognized by antibodies, known as antigenicity. They exposed the chimeric molecule to antibodies known to target the stem portion of the hemagglutinin protein as well as antibodies known to target M2e; in both cases they witnessed high reactivity. The scientists recorded similar results from antisera—blood samples from mice that had previously been infected with an influenza virus and had built up a corresponding antibody response. Importantly, the construct retained its antigenicity when exposed to influenza A viruses belonging to different HA groups. And, even after storage at 122 degrees Fahrenheit for 11 days, the antigen continued to elicit strong antibody reactivity, suggesting solid thermostability.
Having confirmed that the M2E-H3 stem protein can be recognized and bound by antibodies, Subbiah and colleagues moved on to vaccination trials. Often, vaccines based on viral proteins instead of inactivated or attenuated viruses require an adjuvant to help stimulate a stronger immune response, and by extension, better protective efficacy. For their experiments, the researchers used a particularly strong adjuvant similar to the AS01 liposome adjuvant licensed for use in herpes Zoster vaccination. This is a double-edged sword: the adjuvant ensures a quick immune response, but because of its strength, it also comes at the cost of side effects. Although the side effects are not much cause for concern—headache, fever, chills, and so on— they are uncomfortable all the same. Using a different delivery platform, like mRNA technology, may help circumvent these issues in the future.
The adjuvanted M2e-H3 stalk vaccine was administered to mice as two doses, a primary vaccination followed by a boost two weeks later. Following the first dose, there was a noticeable increase in antibodies specific to M2e and HA stem regions. Addition of the boost further increased antibody titers 10 fold. Antisera built up in response to M2e-H3 exposure managed to protect against both group 1 and group 2 influenza A viruses. Subbiah et al. suspect this might be explained by antibody-dependent cellular cytotoxicity (ADCC), a mechanism of immune defense whereby effector cells recognize antibodies stuck to antigens and actively break down the foreign threat. Their suspicions were proven right by western blotting, which showed that M2e antibodies, in particular, raised ADCC activity 20 fold.
To compare protection against an M2e only vaccine, the group of researchers designed another chimera protein combining M2e fused with the receptor binding domain (RBD) of SARS-CoV-2 spike protein. When exposed to a lethal dose of influenza A/HK/H3N2 virus, all of the mice inoculated with M2e-H3 stem survived, suffering only moderate body weight loss of around 10%. On the other hand, none of the mice inoculated with the M2e-only vaccine survived the infection (Figure 12).
Along with a robust antibody response, M2e-H3 stalk vaccination also triggered a strong T cell response. Where antibodies can bind to pathogens to prevent infection, T cells are in charge of getting rid of already-infected cells, curbing the spread of the virus. Indeed, the vaccine significantly reduced the amount of virus in the lungs of infected mice within a span of six days. Depleting T cell levels in vaccinated mice before exposing them to a lethal challenge with influenza A H7N9 produced twice as much weight loss (12%) than in mice that had not been depleted of their T cells.
The work by Subbiah et al. acts as a clear proof of concept: using M2e and HA stalk regions as antigens can stimulate broad neutralizing immunity. Crucially, this immunity should hold up to new strains, as it targets those viral regions that only very rarely mutate. Although their work focused specifically on influenza A viruses, there is no reason to think that, given enough time and research, we won’t be able to produce a truly universal influenza vaccine.
William R. Haseltine, PhD, is chair and president of the think tank ACCESS Health International, a former Harvard Medical School and School of Public Health professor and founder of the university’s cancer and HIV/AIDS research departments. He is also the founder of more than a dozen biotechnology companies, including Human Genome Sciences.