The Ghost Map

by Steven Johnson

  The answer to that question is a measure of how far we have come since the Broad Street epidemic in our understanding both of the pathways that disease takes and the underlying genetic code that instructs bacteria and viruses. But it is also a measure of continuity: how the very same issues that Snow and Whitehead confronted on the streets of London have returned to haunt us, this time on the scale of the globe and not the city. The specific threats are different now, and in some ways they are more perilous, and the tools at our disposal are far more advanced than Snow’s statistical acumen and shoe-leather detective work. But confronting these threats requires the same kind of thinking and engagement that Snow and Whitehead so brilliantly applied to the Broad Street outbreak.

  In all the speech-making, posturing, and sober analysis about avian flu that has swept the globe in the past decade, one utterly amazing fact stands out: as far as we know, the virus that has caused such international panic does not exist yet. To be sure, H5N1 is a viciously lethal virus, with fatality rates in humans approaching 75 percent. But in its current incarnation, it is incapable of starting a pandemic, because it lacks the ability to pass directly from human to human. It can spread like wildfire through a population of chickens or ducks, and the birds can in turn infect humans. But there the chain of infection ends: so long as the overwhelming majority of humans on the planet are not in direct contact with live poultry, H5N1 is incapable of causing a global outbreak.

  So why are health officials in London and Washington and Rome worried about poultry workers in Thailand? Why, indeed, are these officials worried about avian flu in the first place? Because microbial life has an uncanny knack for mutation and innovation. All the world needs is for a single strain of H5N1 to somehow mutate into a form that is transmissible between humans, and that virus could unleash a pandemic that could easily rival the 1918 influenza pandemic, which killed as many as 100 million people worldwide.

  That new capability might come from some random mutation in the H5N1 DNA. For the H5N1, it would be like winning a genetic lottery where the odds were a trillion-to-one against you, but in a world with untold trillions of H5N1 viruses floating around, it’s not impossible to imagine. But the more likely scenario is that H5N1 will borrow the relevant genetic code directly from another organism, in a process known as transgenic shift. Recall that DNA transmission among single-celled bacteria and viruses is far more promiscuous than the controlled, vertical descent of all multicellular life. A virus can swap genes with other viruses willingly. Imagine a brunette waking up one morning with a shock of red hair, after working side by side with a redheaded colleague for a year. One day the genes for red hair just happened to jump across the cubicle and express themselves in a new body. It sounds preposterous because we’re so used to the way DNA works among the eukaryotes, but it would be an ordinary event in the microcosmos of bacterial and viral life.

  Most conventional flu viruses already possess the genetic information that allows them to pass directly from human to human. Because H5N1 is so closely related to the conventional flu virus, it would be a relatively simple matter for it to swipe a few lines of pertinent code and immediately enjoy its new capacity for human-to-human transmission. Certainly it would be easier than randomly stumbling on the correct sequence via mutation.

  And so this is why the whole world has suddenly taken an interest in whether Thai poultry workers get their flu shots: because the world wants to ensure that H5N1 stays as far away as possible from ordinary flu viruses. If the two viruses did encounter each other inside a human host, a far more ominous strain of H5N1 might emerge. It could be as infectious as the influenza bug that swept the globe in 1918, but several times more lethal. And it would find itself inhabiting a planet that was massively more interconnected and densely settled than it was in 1918.

  To appreciate how deadly transgenic shift can be, you need only look at the Broad Street epidemic. In 1996, two scientists at Harvard, John Mekalanos and Matthew K. Waldor, made an astonishing discovery about the roots of Vibrio cholerae’s killer instinct. There are two key components to the bacteria’s assault on a human body: the TCP pilus that allows it to replicate with such exponential fury in the small intestine, and the cholera toxin that actually triggers the rapid dehydration of the host. Mekalanos and Waldor discovered that the gene for cholera toxin is actually supplied by an outside source: a virus called CTX phage. Without the genes contributed by that virus, V. cholerae literally doesn’t know how to be a pathogen. It learns to be a killer by borrowing genetic information from an entirely different species. The trade between the phage and the bacterium is a classic example of coevolutionary development, two organisms cooperating at the genetic level in order to further both of their reproductive interests: the CTX phage multiplies inside the V. cholerae, and in return the virus offers up a gift that allows the bacteria to greatly increase the odds of finding another host to infect. As unlikely as it sounds, V. cholerae is not a born killer. It needs the CTX phage to switch over to the dark side.

  So we have good reason to fear genetic commingling between H5NI and the ordinary human flu virus. But we should also be comforted by how far we have advanced in our ability to anticipate these cross-species transmissions. When John Snow identified the waterborne nature of cholera in the middle of the nineteenth century, he was using the tools of science and statistics to find a way around the fundamental perceptual limits of space: the creature he was seeking was literally too small to see. So he had to detect it indirectly: in patterns of lives and deaths that played out in the streets and houses of a bustling metropolitan center. Today we have conquered that spatial dimension: we can visually inspect the kingdom of bacteria at will; we can even zoom all the way down to the molecular strands of DNA, even glimpse the atomic connections that bind them together. So now we confront another fundamental perceptual limit—not of space, but of time. We use the same methodological tools that Snow used, only now we’re using them to track a virus we can’t see because it doesn’t exist yet. Those flu vaccinations in Thailand are a preemptive strike against a possible future. No one knows when H5N1 will learn to pass directly from human to human, and it remains at least a theoretical possibility that it will never develop that trait. But planning for its emergence makes sense, because if such a strain does appear and starts spreading around the globe, there won’t be the equivalent of a pump handle to remove.

  This is why we’re vaccinating poultry workers in Thailand, why the news of some errant bird migration in Turkey can cause shudders in Los Angeles. This is why the pattern recognition and local knowledge and disease mapping that helped make Broad Street understandable have never been more essential. This is why a continued commitment to public-health institutions remains one of the most vital roles of states and international bodies. If H5N1 does manage to swap just the right piece of DNA from a type A flu virus, we could well see a runaway epidemic that would burn through some of the world’s largest cities at a staggering rate, thanks both to the extreme densities of our cities and the global connectivity of jet travel. Millions could die in a matter of months. Some experts think a pandemic on the order of 1918 is a near inevitability. Would a hundred million dead—the great majority of them big-city dwellers—be enough to derail the urbanization of the planet? It’s unlikely, as long as new pandemics didn’t start rolling in every flu season, like hurricanes. But think of the lingering trauma that 9/11 inflicted on every New Yorker—wondering if it was still safe to stay in the city. Almost everyone opted to stay, of course, and the city’s population has—wonderfully—continued to swell, thanks largely to immigration from the developing world.

  But imagine if 500,000 New Yorkers had died of the flu in September 2001, instead of 2,500 in a collapsing skyscraper. Just the deaths alone would give the year the ignominious status of the single most dramatic drop in population in the city’s history, and no doubt the deaths would be exceeded by all the migrations to the relative safety of the countryside. My wife and I are passionately co
mmitted to the idea of raising our kids in an urban environment, but if 500,000 New Yorkers were killed in the space of a few months, I know we’d find another home. We’d do it with great regret, and with the hope that, when things settled down a few years later we’d move back. But we would move, all the same.

  IT IS CONCEIVABLE, THEN, THAT A LIVING ORGANISM—whether the product of evolution or genetic engineering—could threaten our great transformation into a city-planet. But there is reason for hope. We have a window of a few decades where DNA-based microbes will retain the capability of unleashing a cascading epidemic that kills a significant portion of humanity. But at a certain point—perhaps ten years from now, perhaps fifty—the window may well close, and the threat may subside, just as other, more specific, biological threats have subsided in the past: polio, smallpox, chicken pox.

  If this scenario comes to pass, the pandemic threat will ultimately be defeated by a different kind of map—not maps of lives and deaths on a city street, or bird flu outbreaks, but maps of nucleotides wrapped in a double helix. Our ability to analyze the genetic composition of any life-form has made astonishing progress over the past ten years, but in many ways we are at the very beginning of the genomic revolution. We have already seen amazing advances in our understanding of the way genes build organisms, but the application of that understanding—particularly in the realm of medicine—is only starting to bear fruit. A decade or two from now, we may possess tools that will allow us to both analyze the genetic composition of a newly discovered bacterium and, using computer modeling, build an effective vaccine or antiviral drug in a matter of days. At that point, the primary issue will be production and delivery of the drugs. We’ll know how to make a cure for any rogue virus that shows up; the question will be whether we can produce enough supplies of the cure to stop the path of the disease. That might well require a new kind of urban infrastructure, a twenty-first-century equivalent of Bazalgette’s sewers: production plants located in every metropolitan center, ready to churn out millions of vaccines if an epidemic appears. It will take the creation of public-health institutions in the developing world—institutions that simply do not exist yet—along with a renewed commitment to public health in the industrialized world, particularly the United States. But we’ll have the tools at our disposal to deal with the emerging threats, if we’re smart enough to deploy them.

  The twentieth-century approach to battling viruses has largely operated at the same temporal scale as microbial evolution itself. It has been a classic Darwinian arms race. We take a sample of last year’s most prolific flu virus and use it as the basis for a vaccine that we then spread through the immune system of the general public; and the viruses evolve new ways around those vaccines, and so we come up with new vaccines that we hope will deal with the new bugs. But the genomic revolution means that our defense mechanisms are now starting to operate at a much faster clip than evolution. We’re no longer limited to jury-rigging vaccines out of last year’s model. We’re able to project forward, anticipate future variations, and, increasingly, address the specific threat posed by the most active virus on the ground. Our understanding of the building blocks of life is advancing at nearly exponential rates—thanks in part to the exponential advance in computation power we call Moore’s Law. But the building blocks themselves are not getting more complex. Type A influenza possesses only eight genes. Thanks to the transgenic shift of microbial life, those eight genes are capable of an astonishing amount of variation; but those possibilities are ultimately finite, and they will be no match for the modeling prowess of circa-2025 technology. Right now we’re in an arms race with the microbes, because, effectively, we’re operating on the same scale that they are. The viruses are both our enemy and our arms manufacturer. But as we enter an age of rapid molecular analysis and prototyping, the whole approach changes. The complexity of our understanding of microbial diseases is already advancing much faster than the complexity of the microbes themselves. Sooner or later, the microbes won’t be able to compete.

  But perhaps the arms race will not purely be a figure of speech. The flu virus on its own might not be able to grow complex enough to challenge the technology of genomic science, but what if the technology of genomic science were used to “weaponize” a virus? Genetic engineering may ultimately win out over evolution, but isn’t it a different matter if the viruses are themselves the product of genetic engineering? Wouldn’t the ominous trends of asymmetric warfare—increasingly advanced technology in the hands of smaller and smaller groups—be even more ominous where biological weapons are concerned? If suicide bombers with homemade explosives can effectively hold the American military hostage, imagine what they could do with a weaponized virus.

  The crucial difference, though, is that there are vaccines for biological weapons, while there are no vaccines for explosives. Any DNA-based agent can effectively be neutralized after its release, by any number of different mechanisms: early detection and mapping, quarantine, rapid vaccination, antiviral drugs. But you can’t neutralize an explosive once it has been detonated. So suicide bombers are probably destined to be a part of human civilization for as long as there are political or religious ideologies that encourage people to blow themselves up in crowded places. DNA-based weapons do not have the same future, however, because for every terrorist trying to engineer a biological weapon there are a thousand researchers working on a cure. It’s entirely likely, of course, that we will see the release of an infectious agent engineered in a rogue lab somewhere, and it’s at least conceivable that the attack could unleash a pandemic that could kill thousands or millions—particularly if such an attack took place in the next decade or so, before our defensive tools have matured. But there’s good reason to believe that defensive tools will ultimately win out in this domain as well, because they will be built on a meta-understanding of genetics itself, and because the resources put into their development will dramatically outnumber the resources devoted to developing weapons—assuming, that is, that the world’s nation-states continue the ban on the creation of biological weapons. Biological terrorism may well be in our future, and it could turn out to be one of the most hideous chapters in the history of human warfare. But in the long run, it shouldn’t threaten our transformation into a city-planet, if we continue to encourage scientific research into defensive vaccines and other treatment, and remain vigilant in our opposition to state-sponsored biological weapons research.

  Here, too, the legacy of Snow’s map is essential to the battle. The peculiar menace of a biological attack is that we may not know it is under way until weeks after the infectious agent is first released. The greatest risk of a deliberately planned urban epidemic is not that we won’t have a vaccine, it’s that we won’t recognize the outbreak until it’s too late for the vaccine to stop the spread of disease. Combating this new reality will take a twenty-first-century version of John Snow’s map: making visible patterns in the daily flow of lives and deaths that constitute the metabolism of a city, the rising and falling fortunes of the sick and the healthy. We’ll have exceptional tools at our disposal to defend ourselves against a biological attack, but we’ll have to be able to see the attack first, before we can apply those defensive measures. Before we can mobilize all the technology that would have bewildered Snow—the genomic sequencers and antiviral mass-production facilities—we’ll use a technology that Snow would have recognized instantly. We’ll use a map. Only, this map won’t be hand-illustrated from data collected via door-to-door surveys. It will draw on the elaborate network of sensors sniffing the air for potential threats in urban centers, or hospital first-responders reporting unusual symptoms in their patients, or public water facilities scanning for signs of contamination. Almost two centuries after William Farr first hit upon the idea of amassing weekly statistics on the mortality of the British population, the technique he pioneered has advanced to a level of precision and scope that would have astonished him. The Victorians could barely see microbial life-forms swimming in a
petri dish in front of them. Today, a suspicious molecule floats by a sensor in Las Vegas, and within hours the authorities at the CDC in Atlanta are on the case.

  There is less reason for optimism where nuclear weapons are concerned. A technique that effectively neutralizes the threat posed by influenza viruses could come from any number of active lines of research: from our understanding of the virus itself, from our understanding of the human immune system, even our understanding of how the respiratory system works. There are thousands of scientists and billions of dollars spent every year exploring new ways to fight lethal epidemic diseases. But no one is working on a way to neutralize a nuclear explosion, presumably for the entirely rational reason that it is impossible to neutralize a nuclear explosion. We have made some advances in detection—all nuclear devices give off a radioactive signal that can be tracked by sensors—but detection is hardly a fail-safe option. (If we were relying purely on our ability to detect emerging viruses, the long-term future for epidemic disease would look equally grim.) There is some promising research into medicines that would block the effects of radiation poisoning, which could well save millions of lives in the event of a metropolitan detonation, but millions more would still perish from the initial explosion itself.

  If you look solely at the danger side of the equation, both epidemic disease and nuclear explosions seem to present a mounting threat in the coming decades: thanks to urban density and global jet travel, it’s probably easier now for a rogue virus to spread around the globe, while the breakup of the Soviet Union and the increase in technological expertise has made it easier to both acquire radioactive materials and build the bomb itself. (As I write, the world is wrestling with the implications of Iran’s renewed commitment to a nuclear program.) But if you look at the opposing side of the equation—our ability to neutralize the threat—the story is very different. Our ability to render a virus harmless is growing at exponential rates, while our ability to undo the damage caused by the detonation of a nuclear device is, literally, nonexistent, with no sign that it will ever be technically possible.