DURING THE FLU EPIDEMIC of 1918, according to medical lore, victims were struck down almost in midstride. Four women in a bridge group played cards together until 11 o'clock in the evening. By the next morning three of them were dead. One man got on a streetcar feeling well enough to go to work, rode six blocks and died. During the single month of October, influenza killed 196,000 people in this country_ more than twice as many as would die of AIDS during the first 10 years of that epidemic. By the end of the winter of 1918-19, two billion people around the world had come down with influenza, and between 20 million and 40 million had died.

The flu outbreak of 1918 was "the most devastating epidemic that we have ever had in history," says John R. La Montagne, chief of infectious diseases at the National Institute of Allergy and Infectious Diseases in Bethesda, Md. "And it happened in this century. No one really knows why it occurred, but there's every expectation that if it occurred once, it can occur again." The 1918 influenza pandemic killed as many people in a single year as died in the four-year Black Death, the bubonic plague that ravaged Europe from 1347 to 1351.
 

We like to believe such plunder is an ancient relic; whatever was killing people so ruthlessly in 1918 must be something we can treat by now. Modern medicine has given us an influenza vaccine, an anti-influenza drug (amantadine) and plenty of antibiotics to prevent or treat secondary bacterial infections. But in the face of a virus that kills so rapidly, all the antiviral drugs in the physician's armamentarium would be impotent. If a strain similar to the 1918 variant were to emerge today_a strain that, last time around, killed literally overnight _some experts believe that even modern medicine would be helpless to prevent many related deaths.

THE EXISTENCE OF the influenza vaccine_ not to mention that pervasive phrase "just the flu." conveying as it does a certain harmless inevitability_ may give us a sense of false security when it comes to the possibility of a pandemic outbreak of influenza. (A pandemic is an international epidemic, with disease occurring at a higher-than- expected rate on several continents at once.) In fact, the flu vaccine must be reformulated each year to keep pace with the newest variants of this fastmutating virus, and in the case of an unanticipated outbreak it would take so long to manufacture and distribute an appropriate vaccine that thousands might die waiting.

More stable viruses, like smallpox and polio, are relatively easy to control with an effective one-time vaccine. Not influenza. Because the virus mutates so frequently, the flu vaccine must be concocted anew each year, based on scientists' best guess of what surface proteins will determine the nature of the next season's outbreak. Once they forecast the probable composition of the virus, scientists choose the appropriate antigens to make their vaccine. They must know by mid-February of any year which antigens to include in the following winter's formulations, if they hope to insure production of adequate amounts of vaccine for delivery by the start of the flu season in December.

In developing a flu vaccine, researchers have managed (so far) to outsmart their foe. They have succeeded by turning the virus's own best weapon against it. Unlike almost all other emerging viruses_viruses, like the one that causes AIDS, that turn up in a new species or in a new geographic location (box, page 55) _the influenza virus has mastered and become dependent on just one terrific trick: the speed with which it evolves. The laboratory strains developed for the influenza vaccine, the ones that can stop an epidemic in its tracks, have capitalized on the virus's notorious ability to recombine. But it is this very trait that makes any vaccine against this fast-mutating virus so frustratingly short-lived.

The recipe for making the flu vaccine is simple. Take the current year's variant of the influenza virus, throw it into a stew with a strain of virus that leads to rapid proliferation (a strain known as PR-8) and let the virus do what it does best: incorporate the fast-growing strain into its own genes and start replicating it. From there, it's an easy matter to take those plentiful viruses and attenuate them for a flu vaccine.

But the scientists who must deter mine what virus will cause the next year's illness run a high chance of being wrong. Some observers have put the odds of success at no better than 50-50. Even when they are right, the vaccine lasts only as long as that year's strain. By the next flu season, the fickle virus is almost certain to be wearing a new protein coat, a disguise sufficient to slip right past whatever antibodies were formed in response to the previous year's shot.

BECAUSE MUTATION IS SUCH A crucial factor in the virus's reproductive strategy, any discussion of influenza must include a discussion of how the virus itself changes from one strain to the next. The mutations that lead to major changes in viral surface antigens _ changes that create the pandemic strains of influenza_are quite rare. Pandemic strains have occurred just three times in this century: in 1918, the year of the so-called Spanish flu; in 1957, when a new strain called the Asian flu emerged, and in 1968, when a third strain was introduced, known as the Hong Kong flu.
 

Compared with the Spanish flu pandemic, the more recent outbreaks were extraordinarily mild. When the Asian flu first emerged, the attack rate was the expected 25 percent, but the mortality rate was relatively low; about 70,000 Americans died. With the Hong Kong flu, there were just 28,000 deaths. This lower mortality rate could be traced to the fact that of the two major antigens on the virus's surface, the Asian flu and Hong Kong flu differed in only one. This meant that most people had at least partial immunity to the new 1968 virus, which might have tempered its effects.

The explanation for the influenza virus's mutability lies in the arrangement of its genes. Because its genetic material is packaged in the form of RNA_rather than the more common DNA, which converts to RNA before reproducing, allowing for an intermediate "proofreading" step_ random mutations during replication are relatively common and are passed on intact to the virus's offspring. This kind of random mutation, which occurs during gene replication, is much more common among RNA viruses than scientists once believed. John J. Holland of the University of California at San Diego recently calculated that mutation in the flu virus occurs at a million times the rate at which mutation occurs in human beings. At this rate, Holland concluded, it takes just 12 to 15 weeks to change a flu virus's genetic arrangement by a significant degree.

The new form of the virus differs in ways that may be minor, but it nevertheless renders ineffective the antibodies formed in response to an old influenza infection. It is as though the virus took off its purple coat and put on a red one of the same cut and cloth; the body can recognize the new virus as familiar, but the differences mean that its antipurple techniques will no longer quite be enough. This change is known as antigenic drift.

Thus someone who receives a flu shot in 1992 will have little immunological memory for the slightly changed influenza virus encountered in 1994. (Natural immunity, the result of an actual bout of influenza, produces a more robust antibody response, which usually lasts longer than vaccine-induced immunity, although it, too, is finite.)

In the vast majority of mutations, there is no functional effect on the influenza virus itself. Either the new mutant virus continues producing its proteins unimpaired _ and therefore responds to the same antibodies as its parent_or it dies. But sometimes, probably as rarely as I percent of the time, the new influenza virus survives, is antigenically unique and is still capable of infecting human cells. As this virus multiplies, it has the potential to become the following year's strain of flu virus.

More significant is another kind of antigenic change that also occurs as a result of the virus's genetic structure. The flu virus has what is known as a segmented genome, a weakly bound string of genes with clefts between each of eight segments. Each viral gene is responsible for the manufacture of one or two distinct viral proteins. The segments are physically connected, but only loosely so; they easily come apart and rearrange with other viral segments if different influenza viruses are nearby. The insertion of new segments from different viruses, especially from viruses of animal (rather than human) origin, leads to the process of genetic "reassortment."

The resulting "reassortants," if they involve genes that code for proteins in the virus's surface antigens, can lead to major changes in the influenza virus's configuration_a process that is known as antigenic shift. Antigenic shift goes far beyond antigenic drift. Instead of replacing a purple coat with a similar red one, it is more like putting on a white tunic, green scarf and spangly orange cloak.

For genetic reassortment to occur, a cell must be simultaneously infected with more than one influenza virus of more than one animal strain. Such co-infection does not readily occur in human beings, even though they are susceptible to swine as well as to human influenza, but it happens often in what are known as the reservoir animals. These animals, primarily ducks and other birds and swine, can be infected with influenza from any source_ human, avian or mammalian _ without getting sick.

_Inside a reservoir animal's intestines, viruses flutter about as though they are in the hands of a master card shuffler, moving places randomly and falling back in line in random new combinations. Fortunately for us, though genetic reassortment itself happens frequently, the creation of a new reassortant capable of infecting human beings is quite rare. But when such a humaninfecting reassortant does emerge _part human, part bird or pig_ the potential for a new influenza pandemic has arfived.

OF ALL THE RESHUFFLING OF gene segments that goes on in the guts of pigs or ducks, only once in every 10 to 40 years does a new virus emerge that can rage through a human population. Most reassortants are not viable, but a hybrid virus that is viable (and is capable of infecting a human cell) is so radically different from any previous influenza virus that it can infect large populations quickly and cause serious complications.

Pandemic influenza has historically originated in China_even the misnamed Spanish flu of 1918 had Asian origins_primarily because the country has so many ducks. Wild ducks are the predominant reservoir for influenza. Waterfowl are welcome in China; they prey on many of the pests that would otherwise plague rice crops. Indeed, by some estimates, China has more ducks than it has people. On Chinese farms, ducks live close to people and close to many of the other farm animals that are also influenza reservoirs; the ducks are as likely to be infected with human influenza as with avian. These co-infections provide the ideal medium for genetic reassortment. More trouble erupts on farms where ducks and chickens are raised in proximity to pigs, another common practice in China. A pig that is coinfected with influenza from different species can serve as a mixing vessel for the creation of entirely new strains.

"This is startling information," says Stephen S. Morse, an assistant professor of virology at the Rockefeller University in New York. Morse helped popularize the notion that integrated pigduck farming is responsible for the reassortment of new pandemic influenza strains by inviting one of the theory's leading proponents Robert G. Webster of the St. Jude's Children's Research Hospital in Memphis, to speak at a 1989 conference he organized on the subject of emerging viruses. "Influenza has always been described as the classic example of viral evolution at work, and scientists have long believed that new epidemics are caused by mutations in the virus," Morse says. "Although this may be true of the smaller annual or biennial epidemics we frequently experience, it is apparently not true of the influenza viruses that cause pandemics." What makes it so startling, in other words, is that it postulates that influenza, long thought to be the most dramatic example of random mutation in all of virology, is as subject to the actions of human beings as any other virus.

Ducks have other habits, besides a tendency toward co-infection, that also make them the primary source of pandemic influenza strains. They are perfectly suited to spreading the reassortants around. During their seasonal migrations, ducks can fly enormous distances, spreading their contaminated feces across large expanses of countryside. And they take in water not only by mouth but also through the cloaca, so that as they swim they are simultaneously drinking and ingesting pond water tainted with their own or other ducks' virus-filled excrement and excreting more virus into the water. "If you could eliminate all the horses, swine and ducks in the world, you could eliminate pandemic influenza," says Edwin D. Kilbourne, professor of microbiology at the Mount Sinai School of Medicine and a national authority on influenza.

The adoption of certain new farming techniques in the developing world increases the odds that another such strain will soon emerge. Pigs, ducks and chickens live side by side on many Asian farms, especially those engaged in fish farming. Widely promoted as an energy-efficient way to generate high yields of protein foods, fish farming involves feeding hen feces to pigs and fertilizing fish ponds_where ducks also swim and drink _ with fresh pig manure. Some virologists, especially in Europe, worry about the prime opportunities for genetic reassortment that this agricultural method provides. As scientists from Wales and Germany recently wrote about fish farming, the result may well be the creation of "a considerable potential human health hazard" by bringing together the two reservoirs of influenza viruses. (Eating the fish is not thought to be a flu hazard, since fish are not reservoirs for human influenza virus.)

FOR A SHORT WHILE, INfluenza researchers believed that pandemics occurred every 11 years. This theory_developed in the late 1970's when it looked as though the next pandemic, in the form of swine flu, was on its way_was based on a cycle of three large-scale outbreaks: 1946, 1957 and 1968. Only now do we realize, in retrospect, that the 1946 outbreak represented not a totally new antigenic shift, but a higher-than-expected incidence of an old influenza.

Though the 11-year truism has gripped the public imagination and captured the fancy of a few scientists as well, "it's misinterpretation of old data, generated before the origins of epidemics became clear, or before sequence analysis on the virus genome was possible," says Brian Murphy, chief of the respiratory virus laboratory at the National Institute of Allergy and Infectious Diseases.

We now know, thanks to sequence analysis, that the 1946 virus was the same strain that had been in existence since 1918_a strain called HlN1. The H stands for hemagglutinin, the N for neuraminidase; these are the two antigens of the influenza virus. It is against them that antibodies are formed during a primary infection.

Just as the 1946 virus was the same HlN1 type that caused the 1918 epidemic, so also, it turns out, was the socalled Russian flu of 1977. When the virus from that year's outbreak was examined genetically, not only did it prove to be another HlN1 variety, but it also was genetically identica' to a type of influenza virus that had last been seen in the early 1950's, more than 25 years before. "You can anticipate certain rates of change in each gene per year, even if the hemagglutinin and neuraminidase stay the same," Murphy says. "But this virus looked as if it was in a genetically frozen state."

Many virologists now believe that it was literally just that: an influenza virus from the 1950's, stored in a laboratory freezer at a research facility in China, somehow got released into the environment in 1977. The accidentally unfrozen virus has since continued to circulate and to alter its genome gradually from year to year, the way it would be expected to. Now two different influenza types coexist: one dating from the 1977 Russian flu epidemic (type HlN1) and the other from the 1968 Hong Kong flu epidemic (type H3N2). Some years one predominates, some years the other does.

Only three hemagglutinin types and three neuraminidase types seem to be capable of infecting human beings. (These are hemagglutinin types H1, H2 and H3, and neuraminidase types N1, N2 and N8.) But there are many other antigen types circulating, a total of 14 H types and nine N types; most of them are limited to avian or swine influenza. Recent research has indicated, though, that at least one antigen previously thought unable to infect humans_ hemagglutinin type 7_can infect monkey cells in the laboratory. This raises the possibility that more antigens than we realize may be capable of causing pandemic influenza. The only reason we have not seen it happen is that the right reassortant has not yet emerged.
 

While the 11-year cycle may be a myth, there are good grounds for suspecting that influenza does emerge in a cycle of some kind_ one that suggests that the next pandemic may be caused by an H1 surface antigen. The three H types that infect human beings seem to take turns, in order, as the pandemic strains change. H1 seems always to beget H2, H2 to beget H3, H3 to beget H1 again. It seems to work this way even retrospectively, when scientists try to extrapolate, from the blood of older individuals, what type of antigens were responsible for pre-1918 pandemics. In the mid-1970's, researchers analyzed the antibodies still circulating in the bloodstreams of very old people who had been exposed to influenza pandemics in the late 19th century. They found that people in their 90's still had antibodies to H2N2 _ the probable cause of a pandemic from the mid-1800's_and those in their 70's and 80's had antibodies to a different variety _ H3N2, the probable cause of the pandemic of 1890. (Another clue about
the strain responsible for earlier epidemics was the fact that in 1968, when the H3N2 Hong Kong flu emerged, people over age 78_who had been children during the 1890 pandemic_had lower mortality and morbidity rates than people who were 10 or 15 years younger.) If the H types really do cycle in this way, the next one in line to emerge is type H1.

In evaluating the mutability of the influenza virus, it seems logical, almost inevitable, to wonder why it mutates as often as it does. But another question, just as pertinent, would be this: Why do pandemic strains arise so rarely? The big antigenic shifts, after all, are not that common; 40 years passed between the Spanish flu and the 1957 Asian flu, and more than 20 years have passed since the H3N2 Hong Kong strain emerged in 1968. It is probably a lucky stroke for people that new influenza chimeras are highly vulnerable to die off when they hit the atmosphere. Otherwise, every time a new influenza recombinant emerged from the guts of ducks, we might have trouble indeed.

INFLUENZA experts thought they saw big-league trouble coming in February 1 1976, when a few cases I of severe swine flu broke out among young military recruits in Fort Dix, N.J. One of them, Pvt. David Lewis, 19, died. Of 19 patients whose specimens were sent to the Centers for Disease Control for virus typing, Lewis and four others were shown to be infected with the same HlN1 influenza virus as was responsible for the 1918 pandemic. This got many public-health officials into gear, hoping to use modern technology, namely vaccination, to head off a
disaster of the magnitude of 1918.

The large-scale effort to manufacture and distribute enough vaccine for everyone in the country showed that millions of doses of high quality vaccine could be made in a matter of three or four months. But the swineflu pandemic never materialized. In retrospect, some critics now say 40 million Americans were vaccinated for nothing. In fact, the only real illness to result from the swine flu adventure was caused by the vaccine: about one thousand people developed Guillain-Barre syndrome, a serious paralytic disease that could be traced directly to an immunological response to the inoculation. Still, Edwin Kilbourne_an early and consistent advocate of mass immunization _ says that, based on what was known at the time, the Government acted rationally and prudently. Scientists thought they had encountered the same influenza strain that caused the devastation of 1918, and they thought they had a tool available to help avert hundreds of thousands of deaths. "Better a vaccine without an epidemic," Kilbourne says today, "than an epidemic without a vaccine."

In hindsight, those involved in the decision-making realize that major variants of influenza virus in a few individuals do not necessarily signal the start of a new pandemic. Some variants prove easily transmitted; most, like the Fort Dix strain, will die out after a few passages through human hosts. "Early detection of a new virus therefore may not be adequate evidence on which to undertake mass immunization," Kilbourne wrote in a 1979 article called "Swine Flu: The Virus That Vanished." "But it is, I believe, a signal at least to produce vaccine to hold in readiness."

The swine-flu experience may have uncovered a hidden hazard of surveillance: by looking too hard, you might actually turn up something that you cannot interpret. When influenza broke out at Fort Dix, an aggressive young publichealth director, a man trained by Kilbourne, was called in to identify the responsible strain. He had learned his lessons well: even after he examined the ffrst several samples of respiratory tract washings and found nothing but the expected H3N2 virus, he continued to look at more. "If he had not got 18 or 19 throat washings, if he had stopped at 2 or 3, then on a probability basis he would have decided the epidemic was caused by H3N2 influenza, and nothing would have happened," says Kilbourne now. "No action would have been taken, because no threat would have been perceived."

Such active surveillance continues to this day. The C.D.C. recruits 160 internists around the country every year as "sentinel physicians"; they send a weekly postcard to the headquarters in Atlanta to let Government officials know how many patients nationwide are complaining of flulike symptoms. A subset of these doctors also sends sputum samples to a central Government laboratory, which types the influenza virus and phones in the results to the C.D.C. More elaborate scouting has involved annual expeditions to the countryside of China at the start of every Chinese flu season. Because influenza strains usually emerge from China, these expeditions have often paid off. In late 1987, for instance, a C.D.C. trip to China yielded samples of a new influenza strain cultured at the Shanghai Hygiene and Anti-Epidemic Center. Antigens from that strain, named A/Shanghai/ll/87, were incorporated into the American vaccine administered in 1988.
 

LONG A SUBJECT of intensive, world-wide surveillance, influenza could offer a blueprint for how other such systems can stay one step ahead of a variety of emerging viruses. But the blueprint would come with a checkered history; with influenza, surveillance has been only of limited success. The virus changes so frequently, and so unpredictably, that the clear-cut clinical benefit of surveillance_an accurate vaccine developed in a timely manner_is often out of reach. "I am an optimist by nature, but I'm not too optimistic about what surveillance systems can do if they're trying to keep track of many viruses, especially new ones," says Kilbourne. "Influenza watchers, focused on only a single disease, are aware of the lag between discovery of change and implementation of control, and know that only the probability_not the nature_of change can be predicted. Surveillance for the
unknown will be even more difficult and will demand a level of clinical and laboratory competence not widely
available in the third world or, for that matter, anywhere else."

Thoughts about surveillance are the inevitable next step when virologists start asking themselves, "What can we do to stop it?" With very few antiviral drugs at their disposal, and with only a handful of vaccines, virologists must feel the way bacteriologists did a century ago: they know what causes human suffering, but they are relatively helpless to prevent it. The next wave in biology, the desired end result of analyzing how viruses emerge, will be characterized by the scientists fighting back.

The cyclical emergence of pandemic influenza offers the opportunity to implement some of the predictive, and possibly preventive, strategies that scientists want to use to head off more exotic viral threats. Influenza has characteristics of many other emerging viruses that make it amenable to "viral traffic control": animal reservoirs that can be monitored for signs of increasing infection; a vaccine that can be tailor-made and administered far more efficiently than it currently is; physicians in far-flung communities who can be enlisted to serve as bellwethers for new viruses and close contact with public-health officials internationally _ especially, in this case, in China.

This is not simply a dress rehearsal for some bigger, more important disease; pandemic influenza will be, almost without a doubt, a major plague when it emerges, probably in the next several years. The nature of the surface antigens of the current predominant strain of influenza has not changed appreciably since 1968. If history is any guide, we can probably expect a major alteration of those antigens_one big enough to lead to a worldwide outbreak of severe flu_before the century turns.