When computation science and microbiology join forces to fight a lethal bacteria

Objective: Enhancing immunotherapeutic strategies against group A streptococcus

Streptococcus pyogenes, also known as group A streptococcus (GAS), is a bacterium responsible for several diseases (from throat infections to scarlet fever), which in some cases can be fatal, with an overall mortality rate of 10%. “Infections caused by GAS are increasing globally, including in France, and some countries are now talking about a silent epidemic”, says Hamed Khakzad, a researcher at the Inria Centre at University of Lorraine“We are observing resistance to antibiotics and the stakes surrounding the development of novel treatments are therefore high.” 

Research conducted by Hamed in the analysis of protein-protein interactions has just made a significant leap forward in this field, thanks to a collaboration with the Department of Clinical Sciences at Lund University in Sweden. “Our joint effort began in 2018, during my PhD at University of Zurich in Switzerland”, the researcher recalls. “I was developing machine learning-based algorithms capable of identifying interactions between thousands of different proteins. As it happened, Swedish scientists were trying to analyse interactions between group A streptococcus proteins and human plasma proteins.”

One of the GAS surface proteins, called the M1 protein, was of particular interest to the researchers because it plays a key role in the pathogen’s ability to evade the human immune system. Hamed Khakzad and his colleagues succeeded in identifying some of its interactions with human antibodies. “Our idea was to use this information to make the human immune system more efficient”, the scientist continues.

The power of the multidisciplinary approach

In 2021, the Swedish team brought further expertise to the project, that of Yasaman Karami, a specialist in molecular dynamics (MD) simulations at atomic scale. “Antibodies are shaped like an inverted Y, with two arms, the Fab regions, which bind to the pathogen, and a trunk, the Fc region, which enables the recruitment of cells resulting in phagocytosis, or pathogen destruction”, explains the scientist. “We knew the M1 protein of GAS was capable of obstructing the antibody’s Fc region and thus allow it to escape the immune system... but we needed to understand how this happened at atomic scale and, above all, exploit this knowledge to try to prevent it.” 

Yasaman Karami and Hamed Khakzad are now pursuing this research at the Inria Centre at University of Lorraine, which they both joined in January 2023. By combining machine learning, atomistic MD simulations and integrative structural biology, they are making one key discovery after another. First of all, the two scientists identified the exact pair of amino acids that are responsible for the interaction between the Fc region of the antibody and the M1 protein. The Swedish microbiologists then successfully extracted antibodies from a patient who had recovered from a group A streptococcus infection. With the help of machine learning algorithms, Hamed Khakzad helped them figure out that one of these antibodies binds the M1 protein with its both Fab regions; animal validations have shown this to be the most effective way of fighting infection.

Consequently, they were then able to model interactions between the different categories of antibodies (IgG1, IgG2, IgG3, IgG4) and protein M1... with surprising results: “IgG3 antibodies have less affinity but a higher phagocytosis rate”, explains Yasaman Karami. “So we ventured a guess that while it is important for Fab regions to bind the protein M1, the antibody’s flexibility must also be pivotal.” Indeed, IgG3 antibodies have a longer hinge region than other categories and are therefore more ‘flexible’. 

A more efficient hybrid antibody

Once again, atomistic simulations validated this intuition and, for the Swedish scientists, it was a crucial discovery: “It establishes a link between the structural flexibility and the functional performance of antibody-mediated immune responses”, explains Pontus Nordenfelt, head of the Lund University team. “It helped us understand why IgG3, despite its lower binding capacity, is more effective in facilitating the immune response; a concept that could not have been demonstrated by experimental analysis alone.”

On the basis of this new knowledge, the team therefore engineered an antibody combining the best of each category: the strong binding affinity of IgG1 and the flexibility of IgG3. MD simulations were used to assess its efficiency. “I carried out three simulations, of one microsecond each, for an IgG1, an IgG3 and our hybrid IgGh”, Yasaman Karami explains. “These are very intensive calculations and even with computing systems as powerful as Genci and Prace, some of them took 46 days!” 

But it was worth the wait: according to the simulations, the hybrid IgG is indeed the most flexible and therefore, logically, the most effective... as the in vivo tests confirmed. “This is a major breakthrough because it paves the way for an immunotherapy against group A streptococcus bacteria”, says Hamed Khakzad with enthusiasm. 

During the research period, the team published around ten articles in prestigious journals. The article on this new hybrid antibody which appeared in Nature Communications received particular attention.

Immunotherapy to fight bacteria but also viruses

It has to be said, this publication had an added bonus: “One of the reviewers asked us if we could prove the pertinence of engineering such antibodies to fight other pathogens”, points out Hamed Khakzad. “We therefore performed a test, designing a new hybrid antibody, this time from the IgG1 and IgG3 antibodies produced against the SARS-CoV-2 virus that causes Covid-19. And we proved that, in mice, our hybrid tackled the virus better than the natural antibodies.” 

There is, therefore, a rare potential for immunotherapy based on such antibodies, since this could be used against viruses and bacteria alike. Another project is currently on-going addressing this point, but the details remain confidential as the results will lead to the registration of a patent. “In short, this new study aims to prove that our hybrid antibody design strategy could be extended further to other pathogens”, announces Hamed Khakzad. And we’re willing to bet that this will change things for good when it comes to fighting against streptococcus A and many other causes of illness...