Pale Rider: The Spanish Flu of 1918 and How It Changed the World

by Laura Spinney

  The main sign of flu infection in mice is loss of appetite and weight. Two days after Tumpey’s team infected the mice with the revived virus, they had lost 13 per cent of their body weight. Four days after infection, they had nearly 40,000 times as many viral particles in their lungs as mice infected with a seasonal strain. And six days after infection, they were all dead, while the control mice were still standing. Mice aren’t humans, nevertheless the contrast was dramatic.

  When a virus invades the human body, the body’s immune system is spurred into action. Within minutes, immune cells start secreting a substance called interferon that blocks the synthesis of new protein, so arresting the production of new viruses. But after millennia of co-evolution with humans, flu has evolved its own means of blocking interferon. It does so by concealing the evidence that it has hijacked the cell’s reproductive machinery, so that interferon can’t shut it down. Taubenberger’s team found that the 1918 virus was exceptionally good at this, giving it a head start when it came to replication.

  Interferon is the body’s first line of defence, a generalised rapid response to invasion that is deployed while the immune system musters a rebuff that is more tailored to the invader in question. If interferon works, the invasion is halted and the individual barely feels unwell. If it fails, it means the virus has been able to replicate, and the body’s second line of defence is mobilised. Antibodies and immune cells converge on the site of infection. The immune cells release chemicals called cytokines that, among other things, increase blood flow to affected tissues so that more immune cells can reach them. They also kill other host cells, if necessary, to stop the infection spreading. The result is redness, heat, swelling and pain–collectively known as inflammation.

  Inflammation on a massive scale is what the world’s pathologists saw in 1918–those red, engorged lungs that were hard to the touch and seeped a watery, bloody fluid. Rereading their reports, immunologists from the 1940s on thought that those pathologists had witnessed the effects of a ‘cytokine storm’, an overzealous, second-line immune response that ultimately caused more damage than the virus it was intended to destroy. This is what Taubenberger and his colleagues saw in animals infected with the resurrected virus too. Whereas a benign, seasonal virus produced a transient cytokine response and localised, superficial damage to the lung, the 1918 variety produced a strong, prolonged cytokine response and damage that was severe and deep. It extended past the bronchi–the main respiratory pathways into the lungs–right down into the air sacs or alveoli that make up their very substance.

  All the viruses that Taubenberger’s group had sequenced so far came from individuals who had died in the autumn of 1918–during the most deadly wave of the pandemic.5 But the AFIP repository also contained tissue from spring-wave victims. In 2011–by which time he had moved to the Laboratory of Infectious Diseases at the National Institutes of Health (NIH) in Bethesda, Maryland–Taubenberger published a comparison of the gene sequences encoding the H antigen from the two waves. From this it became clear that the virus had undergone a small but critical change between the spring and the autumn, such that the H antigen was now less well adapted to birds, and better adapted to humans. Three-quarters of the spring-wave cases had a bird-adapted H, while three-quarters of the autumn cases had a human-adapted one.

  Since the vast majority of those who caught Spanish flu recovered, focusing on the few who died risks distorting the picture. However, the NIH team has also studied medical records that were kept in US military camps between 1917 and 1919, that logged both lethal cases and cases where the patient recovered. They show that, while the overall number of cases of influenza dwindled between April and August 1918–that is, between the spring and the autumn waves–the proportion of them that were complicated by pneumonia rose steadily over the same period. The lesions that the flu virus creates in the lining of the respiratory tract can become infected by bacteria, resulting in pneumonia. In Taubenberger’s view, the worse the underlying flu, the more likely it is to invite in opportunistic bacteria. Hence he regards the 1918 label of pneumonia as a marker flagging up the presence of the highly virulent, pandemic virus. If he’s right, then over the summer of 1918, that virus acquired the capacity to spread easily between humans.6

  Bringing all the evidence together, Taubenberger now believes that the virus emerged through a background of seasonal flu sometime in the winter of 1917–18, and was already circulating at low levels the following spring. Whether it came directly from a bird, or passed via a pig, he can’t yet say. In the summer of 1918, it mutated, becoming highly contagious between humans. This new, more virulent form spread through the viral population that summer, and in the autumn the disease erupted. By then, the seasonal background had receded, and there was nothing to dilute the ‘pure’ pandemic variety.

  What caused the virus to mutate that summer is not clear, but as we’ve seen, flu doesn’t need much prompting to change, and conditions were arguably conducive to such an event. Large parts of the world were in the grip of famine, and there is some evidence that nutritional deficiencies in the host can drive genetic changes in the flu virus, causing it to become more virulent (while simultaneously impairing the host’s immune response).7 If we accept that the second wave emerged on or close to the Western Front, then that front was awash with chemicals, some of which, mustard gas in particular, were mutagenic–meaning they were capable of inducing genetic changes in living organisms, including viruses. And those same gases had compromised the lungs of many of the young men gathered there, rendering them ripe for invasion.

  Evolutionary biologist Paul Ewald has even argued that the ratcheting up in virulence of the flu virus that summer was a direct response to conditions on the Western Front.8 It is often said that the optimal strategy for an agent of infectious disease that is transmitted directly from host to host is to moderate its virulence, so that an infected host remains alive for long enough to spread the disease far and wide. But if the pool of hosts is not very mobile–its movement being limited by being packed into trenches, say–and if those hosts are dropping dead from other causes, then there is less evolutionary pressure on the virus to moderate its virulence. In those conditions, Ewald says, there is no advantage to it keeping its host alive. Of course, the virus has no strategy in the conscious sense of the word. Rather, highly virulent strains come to dominate the viral population through natural selection, because they are the most likely to survive and reproduce.

  The human immune system takes several years to mature, and in old age it loses its potency. This is the explanation that is usually given for flu’s characteristic U-shaped death curve. But in 1918 adults in the prime of life also died in large numbers. Some have suggested that it was precisely because their immune systems were so robust that they were vulnerable, since it was in them that the cytokine storm was most aggressive. There is a problem with that explanation, however. As far as we know, the immune system is just as robust in a fifteen-year-old as it is in a twenty-eight-year-old, yet in 1918, fifteen-year-olds were down there in the first trough of the W: though they got ill in large numbers, relatively few of them died. And something else needs explaining: the W was not symmetrical. The right-hand upstroke was attenuated, meaning that the aged were in general more protected than usual. They were actually less likely to die in the 1918 pandemic than they had been in seasonal flu outbreaks throughout the previous decade.

  The answers to these puzzles may lie in the different age cohorts’ previous exposure to flu. There is a school of thought that holds that the immune system’s most effective response to flu is to the first version of the virus it ever encounters. All subsequent exposures elicit variations on that response that are never a perfect match for the new strain. There are hints, based on tests of the antibodies present in blood taken from people who were alive in the first half of the twentieth century, and stored ever since, that the flu subtype responsible for the Russian flu of the 1890s was H3N8. If so, then those who were aged between twen
ty and forty in 1918, for whom the Russian flu was probably their first exposure to influenza, were primed to deal with a very different subtype to the Spanish flu, and consequently produced an inadequate immune response in 1918. By the same logic (though there are as yet no serological data to support this hypothesis), the very old may have been afforded some protection in 1918, by virtue of having been exposed to a flu subtype containing either H1 or N1, that circulated in humans around 1830.

  What about the question of where the Spanish flu came from? We would like to know the answer to this, because it might help us to identify the conditions that give rise to a so-called ‘spillover’ event–when a virus ‘jumps’ the species barrier–and reduce the chances, as far as possible, of it happening again. In order to choose between the three current theories, or indeed, to identify a geographical origin that no one has yet proposed, scientists would need to compare the sequence of the virus that caused the Spanish flu with those of viruses that caused earlier outbreaks of respiratory disease in those places. They can’t do that yet, because the oldest human flu sequences on record belong to the Spanish flu itself. Given that, to date, they have found that virus almost everywhere they have looked for it–eventually, and with the help of intrepid flu hunters like Johan Hultin–it is possible that viable samples will still come to light that will enable them to make those comparisons. That would be the holy grail for Jeffery Taubenberger. In the meantime, however, they haven’t been idle. Other researchers have been using a new technique to make educated guesses about which of the proposed origins is most likely.

  The technique in question is based on the concept of a ‘molecular clock’. Every living organism must copy its genetic material in order to reproduce, but as we’ve seen, the mechanism by which it does so is not perfect, and flu’s copying mechanism is particularly error-prone. Some errors shape the virus–we call them, cumulatively, drift–but the majority are ‘silent’, meaning they have no effect on its structure or function. In any given host, these silent errors build up at a constant rate, which means that by counting the genetic differences between two related viruses, you can obtain a measure of the time that has elapsed since they split from a common ancestor. This is a molecular clock: it has nothing in common with a real clock, except that it counts time.

  Flu infects many animals–not just humans, birds and pigs, but also dogs, horses, bats, whales and seals. At the University of Arizona, evolutionary biologist Michael Worobey has compared all the available sequences of flu viruses that are currently circulating or have circulated in different hosts over the last century, and used them to build a family tree of influenza. The virus accumulates errors at different rates in different hosts, but because he knows that, and has calculated those rates, he can make retrospective predictions about when various historical strains were born, and with what parentage. In 2014, Worobey reported that seven of the eight genes in the 1918 virus closely resembled flu genes found in birds in the western hemisphere–in North America, to be precise.9

  Does that lay to rest all the fevered speculation about the origins of the Spanish flu? Did it begin in Kansas after all? Worobey’s work is suggestive, without being definitive. Molecular clocks are, in general, not as reliable as comparing actual sequences. Nevertheless, they have been right before. In 1963, flu broke out in horses in racing stables in Miami, eventually spreading to horses throughout the United States. Worobey found that the horse flu strain was related to one then circulating in birds in South America, corroborating contemporary veterinarians’ reports that the flu had probably reached Miami with some thoroughbreds that had been flown in from Argentina.

  Questions remain, not least over that troublesome eighth gene–the one that encodes the H1 antigen–which seems to tell a different story. The flu family tree indicates that it may have been circulating in humans for a decade or more prior to 1918, at which point it recombined with seven bird flu genes in a shift event that produced the Spanish flu. If that is what happened, it could explain that troubling cohort of the five-to-fifteen-year-olds, who got sick in droves but didn’t die, since they would have been exposed to the H1 antigen as babies, and been forearmed against it. That scenario raises questions of its own, however, not least why the emergence of that antigen in humans didn’t trigger a pandemic earlier. While scientists continue to scratch their heads over that problem, the molecular clocks have a few more insights to offer, and these may be the most troubling yet.

  The current consensus is that, for hundreds of thousands if not millions of years, wild birds have harboured a kind of primordial soup of flu viruses, some of which occasionally infect humans. The assumption is that–as with HIV, which came from monkeys inhabiting African forests–we unwittingly disturbed a pre-existing reservoir, allowing the virus to move into humans. But things may have happened differently, and we may be a much more central player in the flu ecosystem than we think.

  While teasing out the flu family tree, Worobey noticed that much of the branching in bird flu lineages was relatively recent, meaning that birds are a young reservoir in evolutionary terms. As recently as biblical times, or 212 BC, when flu ravaged the armies of Rome and Syracuse in Sicily, humans might have been more likely to catch flu from horses–another animal with which people have lived closely since the farming revolution. At some point in the last 2,000 years, birds took over as the more important reservoir. It looks as if the bird flu lineage that contributed most of its genes to the 1918 human strain became established in North America around the same time as an epidemic of horse flu that broke out in Toronto in 1872 and spread throughout that continent (newspapers described the almost deserted streets of Washington DC, and a backlog of freight at rail terminals in Philadelphia, as sick mules and horses were taken out of circulation). Worobey can’t yet tell if that flu passed from horses to birds, or vice versa, but one could speculate that the switch occurred as a result of horses being replaced by mechanised modes of transport, and the expansion of poultry farming in the late nineteenth and early twentieth centuries. The switch had occurred by 1918, but its legacy was that horses–like humans–were now vulnerable to infection by bird flu. In fact, army vets in several warring nations–as well as in the neutral but conflict-ready Netherlands–reported an epidemic of horse flu in cavalry stables, that coincided with the human one.10

  The suggestion is that we humans actively drew animal reservoirs of flu into our midst–and even created new ones–through our domestication of wild animals. If so, then the greatest threat to our health, in terms of the next flu pandemic, may not be wild birds. It may be much closer to home. Ducks are not the only avian incubators of flu, but as Claude Hannoun and others discovered in the 1970s, they are particularly effective ones. Archaeological evidence suggests that they were first domesticated in southern China around 4,000 years ago. Today there are an estimated 1 billion domestic ducks in the world, meaning they probably already outnumber wild ducks, and there is no ecological barrier between the two. The Chinese, for example, herd ducks through rice paddies to eat insects and other pests, and there they mingle with wild birds. For at least 150 years–that is, since before the Spanish flu–flu genes have been as likely to flow from domesticated birds to wild birds as in the opposite direction. Thanks to our animal-husbandry practices, in other words, we now pump flu genes into nature. The 1918 flu virus may have jumped to humans from a wild bird (either directly or via a pig), but it is just as likely to have come from one raised in a farmyard.

  Our right to blame the other is looking distinctly shaky. If the molecular clocks are right, humans contributed to their own misery both in 1918 and since. There were two further flu pandemics in the twentieth century: the 1957 ‘Asian’ flu, which claimed 2 million lives, and the 1968 Hong Kong flu, which killed perhaps twice that. They were caused by subtypes H2N2 and H3N2 respectively, but both inherited the lion’s share of their internal genes from the 1918 flu, causing Taubenberger and his colleague, epidemiologist David Morens, to dub the Spanish flu ‘t
he mother of all pandemics’.11 In the 1930s, the British and American teams who demonstrated that flu was caused by a virus surprised their peers by claiming that humans may have passed the Spanish flu virus to pigs, and not vice versa. Comparisons of human and pig flu sequences have since confirmed their suspicions, and in 2009 the H1N1 subtype that had been circulating in pigs since 1918 erupted again in humans, in modified form, triggering the first flu pandemic of the twenty-first century. It was dubbed ‘swine flu’, for obvious reasons, though in a longer timeframe it was humans who gave it to humans. Swine were mere intermediaries.

  15

  The human factor

  There was still one very large, unexplained puzzle. Granted, twentysomethings were vulnerable, but why were some twentysomethings more vulnerable than others? Why did the impact of the Spanish flu vary over space as well as time, such that in a given age group, more Kenyans died than Scots, more Indonesians than Dutch? In a future pandemic, would you die? Would your sister living on another continent? Which of your children would be more likely to survive? If we knew who was vulnerable, we could take steps to protect them.

  To understand what it was that caused some people to succumb and others to get off lightly, we have to follow the numbers. In 1918, people were struck by the terrifying randomness with which the flu seemed to choose its victims. It was only when scientists started comparing morbidity and mortality rates that they began to discern certain patterns. This led them to conclude that humans themselves had shaped the pandemic–through their unequal positions in society, the places where they built their homes, their diet, their rituals, even their DNA.