Developing Vaccines For Fast-Evolving Viruses: We Need The Help Of Computers!

Some pathogens give our immune system quite a hard time, such as pathogens like Influenza, where each season evolves new flu variants that reinfect, even if we had the flu before. Designing vaccines for such fast-evolving virus is a hassle, and there is only a very limited time between the prediction of the next coming strain and the time the vaccine should be ready. In a time where anti-vacs arguments gain strength, further pressure is added to only release flawless vaccines.

But why is it so difficult? The flu, HIV, hepatitis C virus, and other nasty drifting pathogens such as Plasmodium share one common feature: they mutate fast, and the new evolving variant is always quite different from the previous one. These viruses even have “hypervariable” motifs on their surface, which do not carry a specific function, but rather tend a trap to the immune system. In that way, the body develops very strong antibodies to these variable motifs that will mutate very fast and are not important for the virus replication, instead of targeting the functional and conserved parts of the virus. This paradox is called “immunodominance.”

For whatever reason, when you suffer an infection or when a vaccine is developed, some epitopes will mount a good antibody response while some other ones will fail to be detected. The main question is how to counteract the immunodominance of the wrong motifs and how to help our immune system to target the virus in its weak points, especially in the case of susceptible peoples like elderly persons whose immune system respond weakly to vaccination. The development of antibodies by the immune system is quite a complex and polished system that ensures we survive everyday attacks, and we need to dive into this complexity in order to develop successful vaccines.

Antibodies are proteins (immunoglobins, Ig) produced by B cells. These cells are quite fascinating because, during their development, each of them reorganizes their immunoglobulin gene locus in the chromosome and, ultimately, each expresses a different antibody with a different structure. This is quite helpful when you don’t know yet the structure of the next coming pathogens! But even better, when a B cell encounters a structure that matches its antibody, it will get activated and proliferate, and even specifically produce mutations on its antibody gene (called “somatic hypermutation”). This will generate thousands of new B cells with slightly different antibodies structures, some of them will bind the attacking pathogen with a much higher affinity. This process is called “affinity maturation” and ensures that the immune system produces efficient neutralizing antibodies that will help to clear the infection.

How is this process regulated? Cells that produce mutated random antibodies can turn very harmful and support autoimmunity. Interestingly, these mutating B cells are contained in a small 3D environment called “Germinal Center,” where they are challenged and compete to bind pathogen fragments. B cells that fail to bind with a good affinity die while B cells of higher affinity will proliferate more. More precisely, this is a tightly regulated process where cells oscillate between two zones. In one zone they proliferate and mutate for a few divisions. In the other zone, they capture the pathogen fragments depending on their affinity and present it to another immune cell type, T cells, that will only give support to the surrounding B cell that engulfed the most antigen. Overall, the selective environment inside Germinal Centers pushed the development of very specific antibodies to the target pathogen that is less dangerous for our body.

How can this help vaccines? The success of a vaccine relies on producing good B cells with high-affinity antibodies. But it is exactly inside those Germinal Centres that the immunodominance of certain epitopes occurs. Germinal centers have been studied for more than 30 years and powerful imaging techniques like 2-photon microscopy allow to track the movement of T and B cells inside them (Allen 2007 Immunity, Victora 2010 Cell, Tas 2016 Science). Further, many processes have been quantified (motility, the rate of division, the rate of recycling between the zones), that allowed to develop mathematical models of germinal centers. They basically simulate what happens to virtual “founder B cells” with a low affinity, by making them move, proliferate, be selected, and die as it would in real life (for instance Meyer-Hermann 2012 Cell Reports). These models are constantly improved by matching new experimental measurements and can right now fairly explain the immune response towards one single target epitope, and can also successfully predict what happens when the system gets perturbed.

However, the main challenge relies on understanding how Germinal Centres behave when they face multiple epitopes, typically from those nasty shifting viruses, and why the bad variable epitopes are winning the fight of immunodominance. Several mathematical models are trying to simulate the response to multiple epitopes and to predict how to design a successful, ‘broadly neutralizing vaccine’ that would protect against all strain variants of the flu or HIV for instance. These models go hand in hand with new experimental attempts to develop such vaccines and are reviewed by (Robert 2018 curr op biotech).

Interestingly, depending on how you represent the structure of the pathogen epitopes in the model, the best vaccination strategy is not the same. Several models suggest making “successive” vaccines, where you first target one viral variant, then vaccine for a second one, then a third one, and ultimately broad neutralization should be achieved. Alternative models suggest giving the maximum amount of different strains inside one vaccine in order to support B cells that are cross-reactive too different strains in the fight with immunodominant B cells that are not cross-reactive.

So, this is very complex! What is sure is that those mathematical models contain the paradox of immunodominance and the complexity of broad neutralization. They will allow us to go deeper into simulations and understand which factors on the immune or viral side, especially the design of viral structures in the models, make the difference. They will help design and perform experiments to determine why some epitopes are immunodominant and to design more efficient vaccines by combining the good epitopes at the right time. It’s a matter of time and effort, but it will come!

Further, such functional mathematical models are the first entry door to individualized medicine. From the immune profiling of a person with a weak response to vaccination, for instance, which previous strains of the flu this person already recognizes, it will be possible to predict an individualized and optimized vaccine. Viruses, be prepared — computers are joining the fight!

These findings are described in the article entitled Induction of broadly neutralizing antibodies in Germinal Centre simulations, recently published in the journal Current Opinion in Biotechnology. This work was conducted by Philippe A Robert, Andrea LJ Marschall, and Michael Meyer-Hermann from the Helmholtz Centre for Infection Research and Technische Universität Braunschweig.