APPENDIX P 
Excerpted from an article by Robin Holliday: 
Should Genetic Engineers Be Contained ? 
New Scientist 73:399-401, February l7, 1977 
Ever since Professor Paul Berg 
and his colleagues drew attention 
in the summer of 1974 to the poten- 
tial dangers of certain experiments 
involving the formation of DNA 
molecules containing segments 
from widely different species, there have been numerous 
conferences and reports on the precautions which should 
be taken. Researchers agree that the hazards are difficult 
to assess; for this reason the supposed damages have 
almost always been referred to in a very general and non- 
specific way. In this article I shall attempt to make a more 
specific assessment of the various accidents and biological 
events which would have to occur before human life was 
endangered. By doing this I hope to place in a more rational 
perspective the current debate on the wisdom of pursuing 
this research. 
The experiments we are concerned with have been 
variously labelled as “genetic manipulation”, “genetic en- 
gineering” or just “recombinant DNA”. As genetic mani- 
pulation of chromosomes and the formation of recom- 
binant DNA has been the basis of genetic experimentation 
for many decades, these terms are grossly misleading and 
have certainly lead to confusion among non-geneticists. 
Genetic engineering is more acceptable, but this is too 
general a term because it has been widely used in other 
contexts (for instance, the possibility of correcting human 
genetic defects by inserting normal DNA in place of 
mutant DNA). I prefer the term heterogenetics: the syn- 
thesis and study of replicating DNA molecules containing 
nucleotide sequences from unrelated organisms. 
It is not of course possible to consider all experiments 
of this type so I simply discuss one standard “shotgun” 
experiment, which has been referred to as often as any, 
and which is listed in the most potentially dangerous cate- 
gory in the Williams working party on genetic manipula- 
tion. In such a shotgun experiment, DNA from a mam- 
malian, perhaps human, source is fragmented with a par- 
ticular restriction enzyme (specific endonuclease). The 
number of different pieces of DNA produced is extremely 
large, of the order of half a million. These pieces are in- 
serted at random into bacterial plasmid DNA, and the 
plasmid is introduced into an Escherichia coli host. The plas- 
mid replicates autonomously and can be transmitted by 
cbnjugation to other E.coli cells, or related bacteria, which 
do not already harbour a similar plasmid. In such an 
experiment, clones of bacteria can be recognised which 
contain plasmids with inserted DNA, and since the num- 
ber of fragments of DNA is so large, almost every one of 
these clones is distinct. The investigator then has to find 
the particular clone in which he is interested (for instance, 
one containing a particular DNA sequence such as ribo- 
somal DNA). This aspect of the experimentalist’s problem 
need not concern us here. 
We start our scenario with a standard microbiological 
laboratory without special containment facilities, which is 
using standard laboratory strains of E.coli. In the scenario 
I will not initially assign particular probabilities to each 
event, but simply refer to them as Pi, Pz, P 3 . . . etc. 
Similarly, 1 call the possible deleterious consequences Ci, 
Cz, Cz . . etc At the end of the article I shall attempt to 
assess the atiua! probabilities involved, and so produce an 
overall value of risk. But one point of my article is to 
allow each reader to make his or her own judgements. 
One careless scientist or technician, whilst pipetting by 
mouth, accidentally swallows a few thousand or million 
viable bacteria (Probability=Pi. We need not assign a 
different probability to other types of careless experimenta- 
tion, for instance, the spilling of bacteria on hands or 
clothes or the contamination of food or drink). The cells 
survive and proliferate in his gut (Pz. Note that although 
E.coli is a component of the human intestinal flora, the 
laboratory strains are far removed from their original 
“wild type” ancestor). The plasmid the cells contain is 
transmitted to more natural bacteria of the flora (P3). 
This plasmid still contains mammalian DNA and it is one 
of the axioms of those who worry about these experi- 
ments that such DNA in a foreign environment may behave 
abnormally. It may contain a latent or lysogenic mammalian 
virus which is induced to proliferate, perhaps because a 
stretch of DNA which keeps it in the lysogenic state in 
the mammalian cell has been replaced by bacterial DNA. 
Let us examine the probability of this (P»): note that about 
half a million different clones of bacteria exist containing 
mammalian DNA, so if there is one latent virus per mam- 
malian genome, Pi=0- 000002. 
There could, however, be 10 or 100 such latent virus 
genomes; let us be pessimistic and assume there are 50, 
in which case P«= 0-0001. (In the context of this article 
“pessimistic” means assigning a high probability to a 
biological event). The plasmid containing this virus is 
present in many bacteria and in at least some of these it is 
induced to form virus particles. The virus is of mammalian 
origin, but it must be able to replicate its genome, pro- 
duce coat proteins and so on, in a prokaryotic environment 
(the bacterial vehicle). I assign an overall probability Ps 
for the induction, growth, maturation and release of virions, 
(Remember that so far no mammalian virus has so far 
been grown on any bacterium, although many people have 
attempted to do this). We then suppose that the virus is 
pathogenic (Ps) and that it infects the individual who is 
now its host. One consequence is that the individual dies 
(Cl); another is that he is a resistant carrier and transmits 
the virus to others by contact, or any of the usual channels 
of transmission (P 7 ). We suppose that some of these 
individuals are more susceptible than the original carrier 
(Ps) and become ill or die (C2). Presumably infected 
individuals will be diagnosed as suffering from an unknown 
but potentially lethal disease, and would be isolated in 
hospitals. However, if only a proportion of infected indi- 
viduals showed symptoms but were infective during a long 
incubation period (P»), an epidemic is ptossible (C3). 
Another version of the scenario assumes that the virus 
is oncogenic (causes cancer). It might be reinserted in 
altered form into the chromosomes of cells of the human 
host and subsequently induce a malignant transformation 
(Pio). The individual originally infected would then die of 
cancer, but probably some years later (Ci). More in- 
sidiously, the virus could be both transmissible and 
oncogenic (Pn). In this case it would be capable of pro- 
liferating and spreading to other individuals, but would 
produce no outward symptoms. Nevertheless, it inserts 
itself into chromosomal DNA and eventually causes cancer. 
In this situation an epidemic of widespread cancer (C*) 
might not occur until years after the original laboratory 
accident. It is almost certainly the possibility of conse- 
quence C4, or perhaps C3, which forms the basis of all the 
concern about “shotgun” or other related experiments. 
This article first appeared in New Scientist London, the weekly review 
of Science and Technology. 
Appendix P--1 
