Nobel Lecture, December 11, 1951

The Development of Vaccines against Yellow Fever

The study of yellow fever may be divided into two periods. The first one occurred at the turn of the century when Walter Reed and his co-workers showed by the use of human volunteers that the causative agent of this disease was a filterable virus and that it was transmitted by the bite of the common urban mosquito, subsequently named Aedes aegypti. The second period began in 1928, when Stokes, Bauer, and Hudson found that the common Indian rhesus monkey was susceptible to yellow fever, thus making available an animal that could be used in the laboratory. The first strain of yellow fever established by these workers is known as the Asibi strain, named after the patient from whom it was isolated. It has been used extensively in yellow fever work and, as will be shown later, was the parent strain from which the 17D vaccine was eventually produced.

It was shortly after monkeys were found susceptible that, in searching for a less expensive and more readily available experimental animal, I found that the common white mouse was susceptible to the virus if inoculated by the intracerebral route. This method of inoculation was chosen as it was generally conceded that the common laboratory animals could not become infected if inoculated by the usual routes. The strain of virus with which this work was done was isolated by Mathis, Sellards, and Laigret in 1928 in Dakar, French West Africa, is known as the French strain, and, like the Asibi, is highly virulent for rhesus monkeys. The disease in mice was an encephalomyelitis with no involvement of the visceral organs, in contrast to that induced in man and monkey, in which the liver, kidney, and heart are involved. By serial passage in mice, the pathogenic action of the virus was altered in two respects. Firstly, the incubation period became progressively shortened until, after many passages, it became constant, or, to use the term introduced by Pasteur in his work on rabies, it became fixed. Now the incubation period in mice is used as a measure of the degree of neurotropism for these animals. Secondly, with the increase of virulence for the nervous system of mice, there was also evidence of a progressive loss of virulence for rhesus monkeys when inoculated parenterally. This loss of virulence for monkeys first suggested the possibility of the use of an attenuated active virus for the immunization of man. The finding that by mouse brain passage the virulence of yellow fever virus could be altered led me to undertake extensive studies on the variations induced in the virus by different laboratory procedures, which became my main scientific activity for several years. The results of only those experiments which are pertinent to the development of vaccines will be discussed here.

The finding of the susceptibility of mice and its attenuation for monkeys was rapidly confirmed by others. It was shown that many, if not all, strains of yellow fever are pathogenic for mice, and this animal came into widespread use in all yellow fever work. The pathogenicity of unmodified strains for mice varied greatly – ranging from the highly neurotropic Asibi virus to the relatively avirulent French strain. Both of these, as noted before, are highly pathogenic for rhesus monkeys and almost invariably produce a fatal disease when inoculated parenterally. With most strains it was shown that the mouse could be used for the quantitative estimation of virus. This proved of great value, for as strains of virus avirulent for monkeys were developed, the mouse was the only animal by which the presence and amount could be readily determined.

In my early work it was clearly shown that mice could be used for determining the presence of yellow fever antibodies in sera. Standardized tests for antibodies by the use of these animals were soon evolved by Sawyer and Lloyd, and myself. These have proved of great value in the study of the epidemiology and distribution of yellow fever and eventually in testing the efficacy of different vaccines as these became available.

In considering the possibility of using the modified mouse-adapted virus for human immunization, although it was clear that it had become markedly attenuated for monkeys when inoculated by parenteral routes, the discovery was made by Sellards, and Lloyd and Penna that the virus had acquired marked neurotropic affinities for these animals and produced a fatal encephalitis when inoculated into their brains. There was thus the possibility that the virus, if used as a vaccine, although it had lost the power of inducing a serious visceral disease, might induce an infection of the nervous system. Reasons for this possibility were clearly shown in experiments in monkeys. When these animals were inoculated subcutaneously with the mouse-adapted virus, they usually remained well, and developed a solid immunity to a subsequent inoculation of a highly virulent strain. The inoculation of the adapted virus was shown to produce a mild systemic infection with the production of specific antibodies; in fact, it acted like a vaccine. In an occasional animal, however, after the systemic infection had run its course, an encephalitis developed which was invariably fatal. This encephalitis was due to the mouse-adapted yellow fever virus.

In the development of vaccines for human beings, using my mouse-adapted virus, two paths were followed. In the first, used chiefly by French workers, virus alone was inoculated; and in the second, used by American and English workers, virus and human immune serum were inoculated simultaneously. The first immunizations of humans using mouse-adapted neurotropic virus alone were reported by Sellard and Laigret (1932). Several severe reactions were reported, and the method was modified by Laigret, who introduced the procedure of giving three inoculations at twenty-day intervals of virus which had been exposed for four days, two days, and one day, respectively, to a temperature of 20º C. That this method consisted essentially of the inoculation of three graded doses of fully virulent neurotropic virus and not of attenuated virus, as thought by Laigret, was shown by Whitman and myself. These early vaccines, using the neuroadapted virus alone, though producing a satisfactory immunity, were, nevertheless, not considered entirely safe because of the serious reactions associated in some cases with signs of involvement of the central nervous system. Further investigations by French workers, however, finally led to a safe and efficient method of vaccination, which is at present used on a very large scale in the French territories in Africa. This method introduced by Peltier and his co-workers (1939) consists of applying the mouse-adapted virus to the scarified skin. In retrospect, it seems probable that the early severe reactions were due to the use of a virus which, although it had undergone a considerable degree of modification, was nevertheless not sufficiently attenuated for safe use in man. At the present time this method of vaccination is usually combined with vaccinia virus. A mixture of both viruses is applied to the scarified skin, and the individual is thus immunized to the two agents at the same time. Many millions of people have been immunized thus without any very serious reactions having been reported.

The other early method of vaccination using the mouse-adapted virus consisted in the simultaneous inoculation of the virus and human immune serum. This method was based on the observation that an active immunity was readily induced in monkeys by the simultaneous inoculation of highly virulent yellow fever virus and immune serum. After extensive experiments in monkeys, using the mouse-adapted virus and immune serum, Sawyer, Kitchen, and Lloyd introduced this method for the immunization of persons working with yellow fever virus, thus bringing to an end the long series of laboratory infections that had taken such a heavy toll. Though the reactions to the vaccine were, in general, very mild, it should be noted that one very serious reaction with signs of involvement of the brain occurred like those reported by the French workers following the use of the virus alone. This method, however, had a very serious disadvantage in that it could not be used on an extensive scale because of the large quantities of human immune serum required.

Investigations were accordingly undertaken – on the one hand to find a substitute for the human immune serum and on the other hand to develop a more attenuated strain of virus. The first was readily achieved, and various workers succeeded in producing high-titer horse, rabbit, goat, and monkey immune sera. Experiments in monkeys by Whitman and myself showed that by the use of a hyperimmune serum the quantity used for human vaccination could be considerably reduced. Of the hyperimmune sera, those produced in monkeys by Smith and myself proved the most satisfactory and were used fairly extensively for some time. Reactions were few and mild and antibody production satisfactory. The main advantage, however, lay merely in the fact that only a few cubic centimeters were necessary for each individual vaccinated, whereas previously from 35 to 40 cc. of human serum were needed. The method was still cumbersome and entirely impractical for large-scale immunization. This had become an urgent problem as the epidemiological entity known as “jungle yellow fever” had been discovered. Classical urban yellow fever, transmitted by the common yellow fever mosquito, Aedes aegypti, can be readily controlled by antimosquito measures. Jungle yellow fever, on the other hand, occurs in the country districts, sometimes in vast epidemics, and in the absence of Aedes aegypti. The only rational method of protection of the exposed population called for large-scale vaccination.

In further attempts to attenuate yellow fever virus, the method of tissue culture was investigated. The cultivation of the French neurotropic virus was readily achieved by Haagen and myself and later by Lloyd, Ricci, and myself by the use of chick embryo tissue. Extensive experiments for several years with this virus strain in a variety of culture media did not lead to any further attenuation. In attempts to cultivate an unadapted strain of virus, great difficulty was encountered, but this was finally achieved by Lloyd, Ricci, and myself by using a fluid medium, the tissue component of which consisted of minced mouse-embryo tissue. Prolonged cultivation of the Asibi virus in this medium induced a marked change in its virulence; in fact, it became so attenuated that when inoculated parenterally into monkeys it no longer caused death. What was considered of equal significance was that the culture virus did not become more neurotropic; in fact, it was less neurotropic for experimental animals than the mouse-adapted French virus then in use for human vaccination. However, it was not considered sufficiently attenuated in both these attributes to be used for human immunization alone without the simultaneous inoculation of immune serum. This cultured virus, known as 17E, was consequently substituted by Lloyd for the French neurotropic virus. This method proved satisfactory.

In the culture experiments, attempts were made to propagate the virus in as wide a variety of tissues as possible, in the hope that in one the desired attenuation of both the viscerotropic as well as the neurotropic affinities would occur. While it was plain that the viscerotropic affinity could be readily reduced either by mouse brain passage or by prolonged tissue culture, what was particularly desired was a method of reducing the neurotropism. On the theory that since the virus becomes more neurotropic when maintained for a long time in brain tissue (as by passage in mice) it might lose this affinity if cultivated in the absence of nerve tissue, experiments were undertaken to determine the influence of nervous tissue on the prolonged culture of the virus. For this purpose, three parallel series of tissue cultures were maintained, all containing chick embryo tissue but differing in the amount of nervous tissue. In the first, the tissue component consisted of minced whole-chick embryo; in the second, chick embryo brain only was used; whereas the third series contained minimal amounts of nervous tissue consisting of minced chick embryo from which the brain and spinal cord had been cut away before mincing.

The Asibi virus with which all these culture experiments were undertaken is highly pathogenic, not only for monkeys by parenteral inoculation but also for mice by intracerebral inoculation; that is, its viscerotropic as well as its neurotropic affinities are highly developed. These properties made it suitable for culture experiments as any change induced could be readily determined. The culture experiments to date had shown that the viscerotropic affinity of the Asibi strain could be readily reduced without at the same time producing a change in the neurotropic affinity. Since strains of yellow fever virus as they occur in nature vary enormously in these two attributes, it was reasoned that if strains possessing less marked affinities were cultivated, then possibly the desired attenuation would occur. Numerous attempts to cultivate strains of yellow fever virus with low viscerotropic or neurotropic affinities at first ended in failure. However, success was eventually achieved by Smith and myself. Experiments had shown that when embryo mice in utero are inoculated with yellow fever, the virus has a predilection for the brain. By the use of mouse embryo brain as the tissue component in the cultures, seven different strains of yellow fever were readily cultivated. Only two of these, the unmodified French strain and the JSS, a strain isolated from a case of jungle yellow fever, were extensively studied. After their establishment in this medium, the possibility of cultivating them in other tissues was investigated. At first these attempts were unsuccessful, but eventually cultures were obtained in a medium containing minced whole-mouse embryo and finally in media containing chick embryo tissue prepared either from whole embryos or from embryos from which the central nervous system had been cut away before mincing. Three parallel series of cultures were maintained for several hundred subcultures of these two strains of virus, the tissue components of the cultures being, respectively, mouse embryo brain, minced whole-chick embryo, and chick embryos from which the central nervous system had been cut away before mincing. We thus had three different yellow fever strains (the Asibi, the unmodified French, and the JSS) of widely divergent tissue affinities running concurrently in several media containing varying amounts of nervous tissue.

Before these new culture experiments had progressed very far, a very marked change in pathogenicity was observed in the Asibi virus grown in the medium, the tissue component of which was chick embryo containing minimal amounts of nervous tissue. This is called the 17D strain. This attenuation consisted in a partial loss of neurotropism for mice and monkeys, as well as a marked loss of viscerotropism for monkeys. Monkeys inoculated intracerebrally developed a mild encephalitis which as a rule was non-fatal. This was the much hoped-for change, as both the strains then in use for human immunization, namely, the French neurotropic virus and the 17E variant of the Asibi virus, cultivated in mouse embryo tissue, invariably produced a fatal encephalitis when inoculated into the brains of rhesus monkeys.

The marked loss of viscerotropism of the 17D culture virus was clearly shown in experiments in monkeys. As a rule monkeys inoculated with the virus by extraneural routes developed no fever or other signs of illness and in their blood only minimal amounts of virus could be demonstrated. Such monkeys were shown to develop specific antibodies and to be solidly immune to the highly virulent Asibi strain.

After extensive experiments in monkeys by Smith and myself, the 17D strain was used for human vaccination without the simultaneous inoculation of immune serum. It was considered that the loss of both the viscerotropic as well as neurotropic affinities as demonstrated in monkeys made this the virus of choice for human vaccination. In a preliminary study we showed that reactions in man were either absent or minimal and that satisfactory antibody response was obtained. Smith, Penna, and Paoliello, in Brazil, made a more thorough and more extensive study of the reaction in man. The results were eminently satisfactory. The 17D virus has been used as a vaccine in many millions of individuals.

In the two parallel series of cultures with the Asibi virus in which whole-chick tissue and chick embryo brain only were used, no such marked attenuation occurred. Both of these viruses, after several hundred subcultures, produced fatal encephalitis in monkeys when inoculated intracerebrally and both produced a rather severe visceral infection. These results suggested that the amount of nervous tissue was the conditioning factor which produced the change, as loss of neurotropism occurred only in a medium containing minimal amounts of nervous tissue. It was considered at the time that these findings were a confirmation of the hypothesis on which the culture experiments had been planned.

Consequently, in order to obtain more information on the role of nervous tissue, I started three new series of tissue cultures. Virus taken from the culture series that had been maintained in chick brain only and virus that had been cultivated in whole-chick embryo were grown in two new series of cultures, the tissue component of which contained minimal amounts of nervous tissue; and, conversely, the 17D virus, which had been grown in a medium containing minimal amounts of nervous tissue, was transferred to a medium containing chick embryo brain only. At the time of the conclusion of the experiment, more than 200 subcultures in the new series had been made. At intervals, monkeys were inoculated intracerebrally to determine the neurotropism of the cultivated virus. The results showed that no modification had occurred. Thus the 17D strain, cultivated for more than 200 subcultures in chick embryo brain, had not become more neurotropic, and the virus at the beginning and the end of the series was essentially the same in pathogenicity. In like manner, the other two strains, which, it will be recalled, produced fatal encephalitis in monkeys at the beginning of these culture experiments, did not lose their neurotropism even though maintained for more than 200 subcultures in a medium containing minimal amounts of nervous tissue. The conclusion was obvious that the relative amounts of nervous tissue present in the media had not produced any demonstrable change.

The reason for the rapid change noticed in the 17D strain, which occurred between the 89th and 114th subcultures, was and still is completely unknown. However, these experiments indicate that once the mutant had occurred, it was relatively stable.

What was apparently a similar mutant was reported subsequently by Penna and Moussatché (1939). These workers maintained the Asibi virus in series in the developing chick embryo and noted a similar marked attenuation. In their experiments a marked loss of neurotropism occurred in spite of the fact that in the developing chick embryo the virus shows a marked affinity for the brain and multiplies readily there.

Our experiments with the Asibi virus in tissue culture may be summarized as follows. In one of the series of cultures, a sudden modification. occurred. The evidence is that this change was not due to the relative absence of nervous tissue in the medium, nor could chick embryo tissue per se be responsible as the change occurred in only one of six series of cultures containing chick embryo tissue.

The results with two other strains of yellow fever, the French and JSS, which were cultivated in media containing varying amounts of nervous tissue are briefly as follow. The French viscerotropic virus was cultured for several hundred subcultures in three different media, the tissue component of which consisted of, respectively, mouse-embryo brain, minced whole-chick embryo, and minced chick-embryo containing minimal amounts of nervous tissue. The results showed that the viscerotropism was rapidly lost, but there was no such extreme attenuation as was observed in the 17D virus. On the contrary, it appeared that an actual increase of neurotropism had taken place in all three parallel cultures, irrespective of the quantity of nervous tissue in the medium.

In the culture experiments with the JSS strain, maintained for over three hundred subcultures in the same media as were used for the French virus, the neurotropism tended to decrease in all three series. However, the least decrease was observed in the virus cultured in the medium containing minimal amounts of nervous tissue. The viscerotropism of the virus grown in the medium containing minced whole-chick tissue became so decreased that on inoculation subcutaneously it failed to produce an infection even when administered in large doses. This is the most extreme case of attenuation so far observed in any yellow fever virus. The loss of infectivity of this highly attenuated strain for monkeys inoculated subcutaneously recalls a somewhat similar, though not so marked, change that occurred in one of the 17D culture series, which became so attenuated for man that it failed in a fair proportion of cases to produce immunity.

In conclusion, we may summarize the culture experiments by stating that any yellow fever virus maintained in tissue culture will become attenuated somewhat in its viscerotropic affinities, irrespective of the tissue used. The neurotropic affinity as a rule does not change. Occasionally, however, for some unknown reason, a mutant appears with marked reduction in both neurotropism as well as viscerotropism. This mutant is comparatively stable, but it too has been observed to undergo change on two occasions. The first time, as reported by Soper and Smith, the cultured virus was found to have become so attenuated that it failed to produce immunity in a fair proportion of persons vaccinated, and the second time, noted by Fox and his co-workers, the virus had regained some neurotropism so that it actually produced encephalitis in a small proportion of persons vaccinated.

In comparing the two vaccines at present in use – viz., the French vaccine and the 17D – it may be stated that both vaccines produce an actual infection and a resulting immunity. The infection produced by the French vaccine is more severe than that produced by the 17D, as manifested both by subjective symptoms as well as by the amount of circulating virus. As a consequence of this relatively severe infection induced by the French vaccine, antibody production is more regular than after the extremely mild infection induced by the 17D vaccine. Only time will tell which of the two is to be preferred.

By the intelligent application of antimosquito measures combined with vaccination, public-health officials have now the means available to render what was once a prevalent epidemic disease to one which is now a comparatively rare infection of man.