skip to main content

Title: Learning the language of viral evolution and escape

The ability for viruses to mutate and evade the human immune system and cause infection, called viral escape, remains an obstacle to antiviral and vaccine development. Understanding the complex rules that govern escape could inform therapeutic design. We modeled viral escape with machine learning algorithms originally developed for human natural language. We identified escape mutations as those that preserve viral infectivity but cause a virus to look different to the immune system, akin to word changes that preserve a sentence’s grammaticality but change its meaning. With this approach, language models of influenza hemagglutinin, HIV-1 envelope glycoprotein (HIV Env), and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) Spike viral proteins can accurately predict structural escape patterns using sequence data alone. Our study represents a promising conceptual bridge between natural language and viral evolution.

 ;  ;  ;  
Publication Date:
Journal Name:
Page Range or eLocation-ID:
p. 284-288
American Association for the Advancement of Science (AAAS)
Sponsoring Org:
National Science Foundation
More Like this
  1. Influenza viruses, commonly called flu, can evade our immune system and develop resistance to treatments by changing frequently. Specifically, mutations in their genome cause influenza proteins to change in ways that can help the virus evade our defences. However, these mutations come at a cost and can prevent the viral proteins from forming functional and stable three-dimensional shapes – a process known as protein folding – thereby hampering the virus’ ability to replicate. In human cells, proteins called chaperones can help our other proteins fold properly. Influenza viruses do not have their own chaperones and, instead, hijack those of their host. Host chaperones are therefore crucial to the virus’ ability to replicate. However, until now, it was not known if host chaperones can influence how these viruses evolve. Here, Phillips et al. used mammalian cells to study how host chaperones affect an evolving influenza population. First, cells were engineered to either have normal chaperone levels, elevated chaperone levels, or inactive chaperones. Next, the H3N2 influenza strain was grown in these different conditions for nearly 200 generations and sequenced to determine how the virus evolved in each distinctive host chaperone environment. Phillips et al. discovered that host chaperones affect the ratemore »at which mutations accumulate in the influenza population, and also the types of mutations in the influenza genome. For instance, when a chaperone called Hsp90 was inactivated, mutations became prevalent in the viral population more slowly than in cells with normal or elevated chaperone levels. Moreover, some specific mutations fared better in cells with high chaperone levels, whilst others worked better in cells with inactivated chaperones. These results suggest that influenza evolution is affected by host chaperone levels in complex and important ways. Moreover, whether chaperones will promote or hinder the effects of any single mutation is difficult to predict ahead of time. This discovery is significant, as the chaperones available to influenza can vary in different tissues, organisms and infectious conditions, and may therefore influence the virus' ability to change and evolve in a context-specific manner. The findings are likely to extend to other viruses such as HIV and Ebola, which also hijack host chaperones for the same purpose. More work is now needed to systematically quantify these effects so that we can better predict how specific chaperones will affect the ability of viruses to adapt, especially in pathologically relevant conditions like fever or viral host-switching. In the future, such insights could help shape the design of treatments to which viruses do not evolve resistance.« less
  2. Abstract

    Drugs of abuse, such as opiates, have been widely associated with enhancing HIV replication, accelerating disease progression and diminishing host-immune responses, thereby making it harder to effectively manage HIV infection. It is thus important to study the effects of drugs of abuse on HIV-infection and immune responses. Here, we develop mathematical models that incorporate the effects of morphine-altered antibody responses on HIV/SIV dynamics. Based on fitting the model to experimental data from simian immunodeficiency virus (SIV) infections in control and morphine-addicted macaques, we found that two of the most significant effects of virus specific antibodies are neutralizing viral particles and enhancing viral clearance. Using our model, we quantified how morphine alters virus-specific antibody responses, and how this alteration affects the key components of virus dynamics such as infection rate, virus clearance, viral load, CD4+T cell count, and CD4+T cell loss in SIV-infected macaques under conditioning with morphine. We found that in a subpopulation of SIV-infected morphine addicted macaques, the presence of drugs of abuse may cause significantly diminished antibody responses, resulting in more severe infection with increased SIV infectivity, a decreased viral clearance rate, increased viral load, and higher CD4+T cell loss.

  3. An effective vaccine that can protect against HIV infection does not exist. A major reason why a vaccine is not available is the high mutability of the virus, which enables it to evolve mutations that can evade human immune responses. This challenge is exacerbated by the ability of the virus to evolve compensatory mutations that can partially restore the fitness cost of immune-evading mutations. Based on the fitness landscapes of HIV proteins that account for the effects of coupled mutations, we designed a single long peptide immunogen comprising parts of the HIV proteome wherein mutations are likely to be deleterious regardless of the sequence of the rest of the viral protein. This immunogen was then stably expressed in adenovirus vectors that are currently in clinical development. Macaques immunized with these vaccine constructs exhibited T-cell responses that were comparable in magnitude to animals immunized with adenovirus vectors with whole HIV protein inserts. Moreover, the T-cell responses in immunized macaques strongly targeted regions contained in our immunogen. These results suggest that further studies aimed toward using our vaccine construct for HIV prophylaxis and cure are warranted.

  4. The bovine immune system is known for its unusual traits relating to immunoglobulin and antiviral responses. Peptidylarginine deiminases (PADs) are phylogenetically conserved enzymes that cause post-translational deimination, contributing to protein moonlighting in health and disease. PADs also regulate extracellular vesicle (EV) release, forming a critical part of cellular communication. As PAD-mediated mechanisms in bovine immunology and physiology remain to be investigated, this study profiled deimination signatures in serum and serum-EVs in Bos taurus. Bos EVs were poly-dispersed in a 70–500 nm size range and showed differences in deiminated protein cargo, compared with whole sera. Key immune, metabolic and gene regulatory proteins were identified to be post-translationally deiminated with some overlapping hits in sera and EVs (e.g., immunoglobulins), while some were unique to either serum or serum-EVs (e.g., histones). Protein–protein interaction network analysis of deiminated proteins revealed KEGG pathways common for serum and serum-EVs, including complement and coagulation cascades, viral infection (enveloped viruses), viral myocarditis, bacterial and parasitic infections, autoimmune disease, immunodeficiency intestinal IgA production, B-cell receptor signalling, natural killer cell mediated cytotoxicity, platelet activation and hematopoiesis, alongside metabolic pathways including ferroptosis, vitamin digestion and absorption, cholesterol metabolism and mineral absorption. KEGG pathways specific to EVs related to HIF-1 signalling, oestrogenmore »signalling and biosynthesis of amino acids. KEGG pathways specific for serum only, related to Epstein–Barr virus infection, transcription mis-regulation in cancer, bladder cancer, Rap1 signalling pathway, calcium signalling pathway and ECM-receptor interaction. This indicates differences in physiological and pathological pathways for deiminated proteins in serum-EVs, compared with serum. Our findings may shed light on pathways underlying a number of pathological and anti-pathogenic (viral, bacterial, parasitic) pathways, with putative translatable value to human pathologies, zoonotic diseases and development of therapies for infections, including anti-viral therapies.« less
  5. Abstract

    HIV is difficult to eradicate due to the persistence of a long-lived reservoir of latently infected cells. Previous studies have shown that natural killer cells are important to inhibiting HIV infection, but it is unclear whether the administration of natural killer cells can reduce rebound viremia when anti-retroviral therapy is discontinued. Here we show the administration of allogeneic human peripheral blood natural killer cells delays viral rebound following interruption of anti-retroviral therapy in humanized mice infected with HIV-1. Utilizing genetically barcoded virus technology, we show these natural killer cells efficiently reduced viral clones rebounding from latency. Moreover, a kick and kill strategy comprised of the protein kinase C modulator and latency reversing agent SUW133 and allogeneic human peripheral blood natural killer cells during anti-retroviral therapy eliminated the viral reservoir in a subset of mice. Therefore, combinations utilizing latency reversal agents with targeted cellular killing agents may be an effective approach to eradicating the viral reservoir.