APPENDIX D— 14 
ing of recombinant DNAs derived from 
I the pathogenic organisms in Classes 3, 4, 
I and 5 of “Classification of Etiologic 
i Agents on the Basis of Hazard” (5), or 
' oncogenic viruses classified by NCI as 
! moderate risk (6) , or cells known to be 
I infected with such agents, regardless of 
I the host-vector system used, (ii) Delib- 
erate formation of recombinant DNAs 
containing genes for the biosynthesis of 
potent toxins (e.g., botulinum or diph- 
theria toxins: venoms from insects, 
snakes, etc.) . (iii) Deliberate creation 
from plant pathogens of recombinant 
DNAs that are likely to increase viru- 
lence and host range, (iv) Deliberate re- 
lease Into the environment of any orga- 
nism containing a recombinant DNA 
molecule, (v) Transfer of a drug resist- 
ance trait to microorganisms that are not 
known to acquire it naturally if such ac- 
quisition could compromise the use of a 
drug to control disease agents in human 
or veterinary medicine or agriculture. 
In addition, at this time large-scale 
experiments (e.g., more than 10 liters of 
culture) with recombinant DNAs known 
to make harmful products are not to be 
carried out. We differentiate between 
small- and large-scale experiments with 
such DNAs because the probability of es- 
cape from containment barriers nor- 
mally increases with increasing scale. 
However, specific experiments in this cat- 
egory that are of direct societal benefit 
may be excepted from this rule if spe- 
cial biological containment precautions 
and equipment designed for large-scale 
operations are used, and provided that 
these experiments are expressly approved 
by the Recombinant DNA Molecule Pro- 
gram Advisory Committee of NIH. 
B. Containment guidelines tor permis- 
sible experiments. It is anticipated that 
most recombinant DNA experiments in- 
itiated before these guidelines are next 
reviewed (l.e., within the year) will em- 
ploy E. coli K-12 host-vector systems. 
These are also the systems for which we 
have the most experience and knowledge 
regarding the effectiveness of the con- 
tainment provided by existing hosts and 
vectors necessary for the construction of 
more effective biological barriers. 
For these reasons, E. Coli K-12 appears 
to be the system of choice at this time, 
although we have carefully considered 
arguments that many of the potential 
dangers are compounded by using an or- 
ganism as Intimately connected with a 
man as is F. Coli. Thus, while proceeding 
cautiously with E. Coli, serious efforts 
should be made toward developing alter- 
nate host-vector systems; this subject is 
discussed in considerable detail in Appen- 
dix A. 
We therefore consider DNA recom- 
binants in E. coli K-12 before proceeding 
to other host-vector systems. 
1. Biological containment criteria us- 
ing E. coli K-12 host-vectors — EKl host- 
vectors, These are host-vector systems 
that can be estimated to already provide 
a moderate level of contaiiunent, and 
include most of the presently available 
systems. The host is always E. coli K-12, 
and the vectors Include nonconjugative 
plasmids te.g., pSClOl, ColEl or deriva- 
tives thereof (19-26)1 and variants of 
bacteriophage X (27-29) . 
The E. coli K-12 nonconjugative plas- 
mid system is taken as an example to il- 
lustrate the approximate level of co;i- 
tainment referred to here. The available 
data from experiments involving the 
feeding of bacteria to humans and calves 
(30-32) indicate that E. coli K-12 did not 
usually colonize the normal bowel, and 
exhibited little, if any, multiplication 
while passing through the alimentary 
tract even after feeding high doses (i.e., 
10® to 10'“ bacteria per human or calf) . 
However, general extrapolation of these 
results may not be warranted because 
the implantation of bacteria into the in- 
testinal tract depends on a number of 
parameters, such as the nature of the in- 
testinal fiora present in a given individual 
and the physiological state of the inoc- 
ulum. Moreover, since viable E. coli K-12 
can be found in the feces after humans 
are fed 10’ bacteria in broth (30) or 
3x10* bacteria protected by suspension 
in milk (31), transductional and conju- 
gational transfer of the plasmid vectors 
from E. coli lC-12 to resident bacteria in 
the fecal matter before and after excre- 
tion must also be considered. 
The nonconjugative plasmid vectors 
cannot promote their own transfers, but 
require the presence of a conjugative 
plasmid for mobilization and transfer to 
other bacteria. When present in the same 
cell with derepressed conjugative plas- 
mids such as P or RldrdlB, the non- 
conjugative ColEl, ColEl-frp and pSClOl 
plasmids are transferred to suitable re- 
cipient strains under ideal laboratory 
conditions at frequencies of about 0.5, 
10'* to 10 ', and 10"* per donor cell, re- 
spectively. These frequencies are reduced 
by another factor of 10“ to 10* if the con- 
jugai^e plasmid employed is repressed 
with respect to expression of donor fer- 
tility. 
The experimental transfer system 
which most closely resembles noncon- 
jugative plasmid transfer in nature is a 
triparental mating. In such matings, the 
bacterial cell possessing the nonconjuga- 
tive plasmid must first acquire a con- 
jugative plasmid from another cell be- 
fore it can transfer the nonconjugative 
plasmid to a secondary recipient. With 
ColEl, the frequencies of transfer are 
10'“ and 10"* to 10"® when using conjuga- 
tive plasmid donors possessing dere- 
pressed and repressed plasmids, respec- 
tively. Mobilization of ColEl-frp and 
pSClOl under similar laboratory condi- 
tions is so low as to be usually undetect- 
able (33). Since most conjugative plas- 
mids in nature are repressed for expres- 
sion of donor fertility, the frequency at 
which nonconjugative plasmids are 
mobilized and transferred by this se- 
quence of events in vivo is difiicult to 
estimate. However, in calves fed on an 
antibiotic-supplemented diet, it has been 
estimated that such triparental noncon- 
jugative R plasmid transfer occurs at 
frequencies of no more than 10'“ to 
10'“ per 24 hoiu-s per calf (32) . in terms 
of considering other means for plasmid 
transmission In nature. It should be 
noted that transduction does operate in 
vivo for Staphylococcus aureus (34) and 
probably for E. coli as well. However, no 
data are available to indicate the fre- 
quencies of plasmid transfer in vivo by 
either transduction or transformation. 
These observations indicate the low 
probabilities for possible dissemination 
of such plasmid vectors by accidental 
ingestion, which would probably involve 
only a few himdred or thousand bacteria 
provided that at least the standard prac- 
tices (Section II-A above) are followed, 
particularly the avoidance of mouth 
pipetting. The possibility of colonization 
and hence of transfer are increas'e^d, 
however, if the normal fiora in the bowel 
is disrup1;ed by, for example, antibiotic 
therapy (35) . For this reason, persons 
receiving such therapy must not work 
with DNA recombinants formed with 
any E. coli K-12 host-vector system dur- 
ing the therapy period and for seven 
days thereafter; similarly, persons who 
have achlorhydria or who have had sur- 
gical removal of part of the stomach or 
bowel should avoid such work, as should 
those who require large doses of ant- 
acids. 
The observations on the fate of E. coli 
K-12 in the human alimentary tract are 
also relevant to the containment of re- 
combinant DNA formed with bacterio- 
phage X variants. Bacteriophage can es- 
cape from the laboratory either as ma- 
ture infectious phage particles or in bac- 
terial host ce^ in which the phage 
genome is carried as a plasmid or pro- 
phage. The fate of E. coli K-12 host cells 
carrying the phage genome as a plasmid 
or prophage is similar to that for plas- 
mid-containing host cells as discussed 
above. The survival of the x phage 
genome when released as infectious par- 
ticles depends on their stability in na- 
ture, their infectivity and on the prob- 
ability of subsequent encounters with 
naturally occurring X-sensitive E. coli 
strains. Although the probability of sur- 
vival of X and its infection of resident 
intenstinal E. coli in animals and hu- 
mans has not been measured, it is esti- 
mated to be small given the high sensi- 
tivity of X to the low pH of the stomach, 
the Insusceptibility to x Infection of 
smooth E. coli cells (the type that nor- 
mally resides in the gut), the infre- 
quency of naturally occurring x-sensitive 
E. coli (36) and the failure to detect in- 
fective X particles in human feces after 
ingestion of up to 10“ X particles (37). 
Moreover, x particles are very sensitive 
to desiccation. 
Establishment of X as a stable lysogen 
is a frequent event (10" to 10“’) for the 
aW int* cl* phage so that this mode of 
escape would be the preponderant lab- 
oratory hazard; however, most EKl x 
vectors currently in use lack the att and 
int fimctions (27-29) thus reducing the 
probability of lysogenizatlon to about 10 ‘ 
to 10"* (38-40). The frequency for the 
conversion of x to a plasmid state for 
persistence and replication is also only 
about 10“* (41). Moreover, the routine 
treatment of phage lysates with chloro- 
form (42) should eliminate all srurvlving 
bacteria Including lysogens and X plasmid 
carriers. Lysogenizatlon could also occur 
