The more we learn about emergent variants of SARS-CoV-2, the virus that causes Covid-19, the more obvious it becomes that we underestimated their potential impact on our lives. Between growing evidence of increased transmissibility, immune evasion, and possibly even lethality, I can now say with certainty what I could only speculate before. Right now these new variants are a serious cause for concern, and they could continue to be for many years to come.

The question, then, is no longer if SARS-CoV-2 will vary, but how—and in this respect it bears resemblance to another notoriously harmful virus, influenza. One theory, based on years of intensive research done by evolutionary biologist Jesse Bloom and his collaborators, is that these viruses can evolve within and across multiple scales of time and space. Though evolution, at least in the popular imagination, is generally thought to be a large-scale endeavor occurring at the population level, the common prerogative of viruses to alter themselves endlessly is as present and feasible in an individual body as it is across their host species. Per this logic, the variations that arise over the course of someone’s infection can be the same as those seen in the pandemic at large, achieving a striking degree of spatiotemporal coherence.

How can scientists possibly observe such a phenomenon? Though influenza and SARS-CoV-2 can mutate no matter the health or background of their host, immunocompromised individuals tend to develop longer-lasting infections that are easier to study. For them, what a healthy immune system can neutralize in a matter of days can take weeks or even months, giving the virus more breathing room to experiment with the very slight alterations—sometimes as miniscule as a single amino acid—that might improve its fitness. When researchers cross-referenced these changes with those in variants either already dominant or on the verge of it, they found a remarkable amount of overlap—hence the term “parallel mutations,” a phrase that captures the fact that the same mutations developed independently of one another.

One such study, published in 2017, examined four immunosuppressed cancer patients who developed very persistent and severe cases of the flu from 2006 to 2007 (Figure 1). For several months researchers obtained weekly samples of influenza virus that they sequenced, analyzed for mutations, and compared to one another. Notably, besides having a common foe in cancer and influenza, the patients were quite different — diverging in age, gender, pre-existing health conditions, and duration of infection. Patient W, for instance, was a woman in the 25 to 44 age group who was being treated for Hodgkin’s Disease, a type of lymphoma, prior to catching the flu and died 80 days after her first sample. Meanwhile Patient Y, a man in the 45 to 65 age group with leukemia, tested positive for cytomegalovirus and cold-causing coronavirus over the course of his infection but eventually recovered.

Figure 1. Overview of patient influenza infections and treatments. Periods of oseltamivir (Tamiflu) treatment are shown in orange.

Figure 1. Overview of patient influenza infections and treatments. Periods of oseltamivir (Tamiflu) … [+]“PARALLEL EVOLUTION OF INFLUENZA ACROSS MULTIPLE SPATIOTEMPORAL SCALES” HTTPS://ELIFESCIENCES.ORG/ARTICLES/26875Two major findings came of this study. First, the researchers were able to identify several mutations that developed in two or more patients in parallel (Figure 2). While these mutations appeared across nine different sites in the influenza genome, five of those sites were located on hemagglutinin, a surface protein critical to the virus for both binding and fusion purposes (Figure 3). Second, variation in four of these five hemagglutinin sites was also prevalent in influenza strains circulating globally, with two mutations in particular—V223I and N225D—reaching high frequency in the decade after the study participants were first infected.

Figure 2. (A) Number of nonsynonymous (orange) and synonymous (green) variants in each influenza gene. (B) Frequencies over time for all HA mutations in patient W. Each subplot represents a site in HA and is labeled by codon number. Ancestral identities are colored in gray and mutant ones in orange. (C) Maximum frequencies reached by all nonsynonymous (orange) and synonymous (green) mutations in each patient.

Figure 2. (A) Number of nonsynonymous (orange) and synonymous (green) variants in each influenza … [+]SOURCE: “PARALLEL EVOLUTION OF INFLUENZA ACROSS MULTIPLE SPATIOTEMPORAL SCALES” HTTPS://ELIFESCIENCES.ORG/ARTICLES/26875

Figure 3. Sites of (A) nonsynonymous and (B) synonymous within-host mutations are shown on an HA crystal structure. Sites of (C) nonsynonymous and (D) synonymous within-host mutation are shown on an NA crystal structure.

Figure 3. Sites of (A) nonsynonymous and (B) synonymous within-host mutations are shown on an HA … [+]SOURCE: “PARALLEL EVOLUTION OF INFLUENZA ACROSS MULTIPLE SPATIOTEMPORAL SCALES” HTTPS://ELIFESCIENCES.ORG/ARTICLES/26875

Today we see the same parallels arising between immunocompromised Covid-19 patients and the new SARS-CoV-2 variants. In May 2020, a man living in London—I’ll refer to him as the London patient—was hospitalized and diagnosed with Covid-19. He recovered and went home, only to be readmitted the following month when his symptoms came back with a vengeance. This time he was administered, at distinct yet overlapping intervals, a plurality of treatments, including the steroid dexamethasone; two rounds of remdesivir, an experimental drug therapy; and convalescent plasma, a highly potent preparation of anti-SARS-CoV-2 antibodies, from not one, not two, but three different patients. Like half of the influenza study participants, he was being treated for lymphoma prior to his diagnosis.

Figure 4. Clinical timeline of events with longitudinal respiratory sample CT (cycle time) values.

Figure 4. Clinical timeline of events with longitudinal respiratory sample CT (cycle time) values. SOURCE: “NEUTRALISING ANTIBODIES IN SPIKE MEDIATED SARS-COV-2 ADAPTATION” HTTPS://WWW.MEDRXIV.ORG/CONTENT/10.1101/2020.12.05.20241927V3

Over the course of the London patient’s second hospitalization, which lasted more than 100 days, researchers extracted, sequenced, and analyzed more than 20 different viral samples, a process they describe in a research paper currently undergoing peer review (Figure 4). They were able to observe the evolution of the virus in real time, watching as some amino acid changes came about in competition with others, then became fixed. One example is the H69-70 deletion in the N-terminal domain of the spike—a surface protein comparable to hemagglutinin in function and significance—which went on to become a defining feature of the B.1.1.7, or UK, variant.

Figure 5. Variants detected in the London patient from days 1-81.

Figure 5. Variants detected in the London patient from days 1-81.SOURCE: “NEUTRALISING ANTIBODIES IN SPIKE MEDIATED SARS-COV-2 ADAPTATION” HTTPS://WWW.MEDRXIV.ORG/CONTENT/10.1101/2020.12.05.20241927V3

The London patient is not the only evidence of parallel mutations in SARS-CoV-2. Another example is a 45-year-old, immunocompromised man in Boston who had Covid-19 for nearly five months. A case study of his infection, published mid-November in the New England Journal of Medicine, revealed that the E484K mutation, the very same to appear in the B.1.351 variant originating in South Africa and the P.1 variant circulating in Brazil, was detected in his viral samples. So was N501Y, yet another substitution in the spike protein most prominent for its likely role in increasing the transmissibility of the UK variant. A research paper published in Science early this month added a third precedent to the list, a cancer patient in Pittsburgh who had Covid-19 for at least 74 days before passing away. They, too, were immunocompromised—and they, too, had SARS-CoV-2 replicating inside them that sported mutations found in the UK and South Africa variants.

Recall that researchers began following the case of the London patient when he was readmitted to the hospital in June of last year. The world didn’t catch wind of the UK and South Africa variants until a full six months later—meaning that the surface protein mutations researchers identified in this single patient anticipated those to come globally. Sound familiar? That’s because the same sentence could be used to describe the outcomes of the 2017 influenza study. The similarities don’t end there, either. The variation mapping technique that first established the E484K mutation could reduce the neutralizing power of convalescent sera up to tenfold was also used on hemagglutinin back in 2019. That data pointed to a related but more general conclusion. Anti-flu antibodies recognize a portion of the surface protein so tiny, a single amino acid change is all it takes for the virus to evade the defensive attacks mounted by convalescent sera.

Is it the case that mutations don’t just arise in parallel within immunocompromised hosts and across global populations, but that the former originate the latter? We can’t be sure. What is almost certain, however, is that the evolutionary pathways traversed by SARS-CoV-2 are too close to influenza for comfort. We know that nonlethal human coronaviruses occur seasonally like the flu, each time in a new guise. Partially immunized populations, it seems, buffer these viruses much in the same way that a weakened immune system bolstered by sera does— the flipside of introducing new measures against contagion or disease being an increase in pressure on the virus to surmount those same defenses. We may think we have a grasp on the direction SARS-CoV-2 is headed in—predicting it will become less lethal or capable of causing serious disease, akin to a common cold—but allowing optimism to cloud our judgment only sets us up to be blindsided by more unpleasant surprises further down the line. At this point this virus has proven itself so flexible and wily, to think it might cease to change and conveniently fade into the background isn’t just optimistic, but naive.

Influenza has caused five pandemics in the past 100 years and continues to kill tens of thousands of people annually—and it isn’t nearly as lethal as Covid-19. If we don’t strengthen our efforts to contain this disease considerably, taking a line of attack aimed at not just eliminating the virus, but averting its evolution into more dangerous forms, we can expect it to mirror influenza in this respect, too—continuing year after year to add to the toll on human life that is already too great to bear. Analysis of persistent infections in immunocompromised patients gives us valuable insights we can use to improve our drugs, diagnostics, and vaccines, which is why my next article in this series will take a closer look at the patients from Boston and Pittsburgh.

Random variation is an essential component of all living things. It drives diversity, and it is why there are so many different species. Viruses are no exception. Most viruses are experts at changing genomes to adapt to their environment. We now have evidence that the virus that causes Covid, SARS-CoV-2, not only changes, but changes in ways that are significant. This is the eleventh part of a series of articles on how the virus changes and what that means for humanity. Read the rest: part onepart twopart threepart fourpart fivepart sixpart sevenpart eightpart nine, and part ten.