Darwin, Pasteur and the Andromeda Strain

Recombinant DNA research has led to extensive public discussions of two potential risks: the immediate risk of harm from some of the novel organisms produced, and the more conjectural, long-term risk that our interference with evolution will eventually create unforeseeable disasters. These discussions have been based largely on the assumption that any novel organism produced by this technique may well survive and spread. But this assumption ignores Darwin's great discovery: the dominating role of natural selection in determining what survive, multiplies, and evolves. While Darwin dealt only with the visible living world, Pasteur made essentially the same discovery for invisible organisms, though expressed in different terms: bacteria do not arise by spontaneous generation but are ubiquitous, and the kinds that grow out in any medium are the ones that are selected by that medium. An extension of these principles to infectious disease gave rise to the science of epidemiology, which may be viewed as a branch of microbial ecology concerned particularly with the distribution of pathogenic organisms.

I believe that epidemiology and evolution have been grossly underrepresented in the professional as well as in the public discussions of the problem. My credentials for correcting this defect are modest, for my research shifted away from medical microbiology a good many years ago, and since then my contact with the field has depended mostly on textbook writing. But I hope the information I offer will encourage deeper exploration of these topics by experts in future symposia.

To be sure, the most reliable basis for assessing the hazards will ultimately be provided by the accumulation of experience with recombinants. But meanwhile we should not act as though we are entering this new territory with no knowledge to guide us: we have a good deal of pertinent information from evolution and from epidemiology. To molecular biologists who have seen one deep mystery after another in other areas of biology settled by the extremely hard data that their field provides, the evolutionary considerations that I shall invoke may seem like mere hand-waving. But in this light nearly all of Darwin's arguments, based on inferences about the past and not on verifiable experiments, could be similarly dismissed. And I would remind you that Darwin's theory remains the most profound and unifying generalization in biology: it is enormously supported today by the evidence from DNA sequences for the genetic origin and the continuity of the observed variation, but it also involves ecological processes and populational kinetics that require a totally different set of concepts and approaches from those of molecular genetics.

Let us now review some of the pertinent principles from evolution and microbiology.

I. BACKGROUND

A. Microbiological and Evolutionary Principles

(1) The Meaning of Species. As evolution created the process of sexual reproduction, whose reassortment of genes provides a vastly increased supply of genetic diversity for the mill of natural selection, it also developed species: groups of organisms that reproduce only by mating with other members of the same group, and not with members of other species. The evolutionary function of these fertility barriers is clear: diversity is necessary for evolution, but since a successful organism must have a reasonably balanced set of genes the diversity resulting from unlimited combinations from the pool of genetic material in the living world would not be useful. Species barriers eliminate the production of grossly unfit, non-viable progeny.

Unlike eukaryotes, prokaryotes ordinarily reproduce by the asexual process of cell division. Their occasional gene transfers do not show a sharp species boundary: the transfer simply becomes less efficient the greater the evolutionary separation between the donor and the recipient. Prokaryotes therefore have no true species: they have an almost continuous spectrum of genetic patterns, and the borders between what we call bacterial species are arbitrary and often controversial. E. coli, for example, is the name given to a range of strains with certain common features but also with a variety of differences, and these differences determine their relative Darwinian fitness for various environments. This elementary concept was entirely missing from Cavaliere's discussion of the hazards of inserting genes in E. coli in the New York Times Sunday Magazine last August, and I would criticize the Times for publishing such a polemical and unqualified account.

(2) Bacterial Ecology. Every living species is adapted to a given range of habitats. The set of bacterial strains called E. coli thrive only in the vertebrate gut, and because these cells die out rather quickly in water the E. coli count of a pond or a well is a reliable index of its continuing fecal contamination. In the gut there is intense Darwinian competition between strains, depending on such variables as growth rate, growth requirements, ability to scavenge traces of food, adherence to the gut linings and resistance to antimicrobial factors in the host. Hence most novel strains are quickly extinguished. The mechanism of this extinction is the kind of selection by competition envisaged by Darwin for higher organisms, but with bacteria it happens in days rather than in eons.

This effect of the environment in the gut on the normal flora is readily recognized. For example, when breast feeding is replaced by solid food the character of the stool changes dramatically, as lactic acid bacteria (which produce sweet-smelling products) are replaced by E. coli and other foul organisms. Early in this century Mechnikov romantically hoped to promote longevity by supplying lactic acid bacteria, in the form of yogurt, to displace the presumably toxic foul organisms. The experiments were a dismal failure, but the commerical success is still seen.

(3) Pathogenesis. Only an incredibly small fraction of all bacterial species can cause disease. The rest play essential roles in the cycle of nature. Infectious bacteria differ from each other in several distinct respects: infectivity (i.e., the infectious does, ranging from a few cells of the tularemia bacillus to around 10-6 of the cholera vibrio); specific distribution in the body; virulence (i.e., the severity of the disease produced); and communicability from one individual host to another. These attributes depend on the coordinate activity of many genes, which are capable of independent variation. For our discussion the distinction between the ability to produce a serious disease and the ability to spread is particularly important.

(4) Types of natural selection. When an organism grows continuously in a relatively constant environment natural selection has a stabilizing effect, weeding out the variants that deviate too far in any direction from the well adapted norm. But when the environment is changed the same basic process of natural selection has a diversifying effect: the new circumstances select for the preferential survival and reproduction of variants with increased fitness for those circumstances. This Darwinian process explains a phenomenon that confused early workers: when pathogenic bacterial strains are isolated from infected hosts and then repeatedly transferred in artificial culture media they often rapidly lose virulence. We now know the mechanism by which this improved adaptation to the new environment occurs, at the expense of decreased adaptation to the old one (i.e., loss of virulence): the original strain is gradually outgrown by the progeny of are mutants that are better adapted to the new culture medium (i.e., that can grow slightly faster, or can grow slightly longer with a limited food supply).

It is clear that natural selection plays an overwhelming role in evolution, though with bacteria its role was long unrecognized: the population shifts seemed too rapid for an undirected process, and the existence of genes and mutations in bacteria was not recognized until the 1940s. But by now selection has become the foundation of bacterial ecology.

II.HAZARDS

In trying to estimate the immediate hazards from novel organisms it is useful to distinguish three possibilities: that experiments with a given kind of DNA will produce a dangerous organism; that the organism will infect a laboratory worker; and that the organism will spread.

A. Danger of Producing a Harmful Organism

I would like to concentrate on a kind of experiment that is causing great concern and is restricted to quite special facilities: the so-called "shotgun" experiment with random fragments of DNA from animal cells. Two considerations seem pertinent. First, the probability that any fragment will contain a gene for a toxic product, or the genes of a tumor virus, is exceedingly low, though not zero. Second, evolutionary considerations provide an independent approach to the question. It seems exceedingly doubtful that our novel ability to introduce mammalian DNA into bacteria in the laboratory will create a truly novel class of organisms, for evolution has already had a crack at the problem.

In particular, it is known that bacteria can take up naked DNA from solution; and, in fact, transfer of DNA between two strains of pneumococcus has been demonstrated in the animal body. Moreover, bacteria in the gut are constantly exposed to fragments of host DNA that are released as the cells lining the gut die, and bacteria growing in carcasses have a veritable feast of DNA. The efficiency of such uptake of mammalian DNA by bacteria is undoubtedly very low. However, because of the extraordinarily large scale of the exposure in nature, recombinants of this general class must have been formed innumerable times over millions of years. They have thus been tested in the crucible of natural selection, and if they had high survival value we would be recognizing short stretches of mammalian DNA in E. coli. We do not. If, on the other hand, naturally occurring recombinants are appearing and even causing disease, but are escaping our attention, we would have to ask how much our laboratories could add, since nature experiments with about 102)-1022 bacterial cells produced in the human species per day.

B. Danger of Laboratory Infection

Let us now consider the probability that an inadvertently produced harmful organism might cause a laboratory infection, and let us assume the worst case: an E.coli strain producing a potent toxin absorbable from the gut, such a botulinus toxin. Such a strain would indeed present a real danger of laboratory infection. But there are a number of reasons to expect this danger to be less than that with the pathogens that are handled every day by medical bacteriologists.

(a) The known laboratory infections (about 6000 recorded in the history of microbiology) have been largely respiratory infections, spread by droplets (mostly before safety cabinets were introduced in the 1940s). Enteric infections, however, occur through swallowing of contaminated food or other material. Even the most virulent enteric pathogens are relatively safe to handle in the laboratory with simple precautions, such as not putting food or a cigarette on the laboratory bench.

(b) Strain K12 of E. coli has become adapted to artificial media during transfer for at least 30 years in the laboratory. Recent tests in England showed that after a dose in man much larger that what one would expect from a laboratory accident, it disappeared from the stools within a few days. Its problems of survival are analogous to those of a delicate hothouse plant thrown out to compete with the weeds in a field.

(c) The addition of a block of foreign DNA to an organism will ordinarily decrease its adaptation to its natural habit and hence its probability of spreading.

This is a condensed version of a public lecture at the Science Center, January 5, 1977 given by Bernard D. Davis '34, Lehman Professor of Bacterial Physiology at the Medical School.

(d) A very large safety factor is added by the provision in the present Guidelines for biological containment. All work with mammalian DNA must be carried out only in an EK2 strain, which has a drastically impaired ability to multiply, or to transfer its plasmid, except under very special conditions provided in the laboratory.

In this connection, I would question the specification, in the Guidelines that an EK2 strain must have a survival frequency of less than 10-8 under natural conditions (interpreted by the committee as residual viability after 24 hours). Just as infection can be dramatically cured by a bacteriostatic antibiotic, such as chloramphenicol, as well as by a bactericidal one, such as penicillin, so the inability of an EK2 strain to multiply in the gut would be sufficient to ensure its rapid disappearance, even if it did not rapidly commit suicide. The important question, requiring extensive investigation, is not the rate of suicide of the EK2 strain but the chance of transfer of its plasmid to a better adapted strain, before disappearance of the EK2 host.

We thus see that with a strain known to have added the gene for a potent toxin a serious laboratory infection requires the compounding of four low probabilities. With strains from shotgun experiments we have a fifth, very low probability, already mentioned: that an apparently harmless mammalian tissue will yield a dangerous product.

The risks thus seem very much smaller than the public has been led to believe. Nevertheless, it is important to keep all the probabilities low. For example, even if a toxin-producing strain could survive only very briefly in the gut, a large enough dose might meanwhile cause disease. Hence a major benefit from the current discussion could be the requirement that those working in this area learn and use the standard techniques of medical microbiology, at least until we have acquired much more experience.

C. Danger of Spread

I now come to the most important point of all: the enormous difference between the danger of causing a laboratory infection and the further danger of unleashing an epidemic. In Camp Detrick, working for 25 years on the most communicable and virulent pathogens known, 423 laboratory infections were seen, most caused by respiratory pathogens. Yet there was not a single case of secondary spread to any person outside the laboratory. Similarly, in the Communicable Disease Center of the U.S. Public Health Service 150 laboratory infections were recorded, with one case of transmission to a relative. Elsewhere in the world there have been about two dozen laboratory-based microepidemics recorded, each involving a few outsiders.

With enteric pathogens the danger of secondary cases is even less, for with this class of agents modern sanitation provides infinitely better control than we can provide for respiratory infection: in contrast to influenza, the appearance of a case of typhoid in a home does not lead to an epidemic. Enteric epidemics appear only with inadequate personal hygiene or sanitation, and such epidemics are always small (except when sewage freely enters the water supply).

This information is clearly pertinent to the recombinants that we are discussing. For