THE CLONING OF A XELF-RI<PLICA?'ING RNA AIOLECULP'

BY REUBEN 1,EVIBOHNt ANI) s. SPINGELMAN

DEPlRTMENT OF MICllOLlIOLOGY, IJNIVERSI'L'Y OF ILIdNOIH, URUAN k

Communicated May 9, 1968

The availability of purified'

                    which can mediate virtually uri-
limited synthesis of biologically active3 and infectious viral RNA, has made
possible a variety of informative investigations.  The use of a temperature-
sensitive mutant led to the demonstration that the RNA molecule is the instruc-
tive agent in the replicative reaction, hence the rigorous conclusion4 that the
in vitro synthesis of a self-replicating molecule had in fact been achieved.

Among the interesting possibilities thus generated was the feasibility of per-
forming Darwinian experiments in which the replicating molecules are exposed
to a variety of selection pressures.  Mills, Peterson, and Spiegelmad reported
experiments of this nature designed to select molecules that could replicate
faster than the original viral RNA.  During the course of the serial transfers
involved, variants of decreasing length made their appearance sequentially.
The mutant isolated after the 74th transfer replicated some 15 times faster and
contained only 550 of the 3600 residues originally present in the parental QP-
RNA.

We shall now focus attention on the use of this system to generate clones
descended from individual RNA strands.  Such experiments possess an inherent
historical interest since the successful cloning of a molecule replicating in vitro has
not, thus far, been recorded.  Nore important, however, is the fact that the
resulting clones provide the sort of uniformity required for sequence studies of
variant molecular structures.  The same experiments can also provide useful
information on molecular weights and genetic stability of distinguishable vari-
ant molecules of independent origin.

The approach used depends on a straightforward comparison of the observed
frequency distribution with that expected from Poisson statistics in a series of
repeated syntheses.  Thus, if one strand is suficient to start a synthesis, then the
proportion of tubes showing no synthesis should correspond to e-m, m being the
average number of strands inoculated per tube.  Further, if the onset and syn-
theses were adequately synchronized, one might hope to identify tubes that
received one, two, or three strands; these should appear with frequencies corre-
sponding to me-", (m2/2!)e-", and (m3/3 !)ecm, respectively.

It is evident that the contemplated experiments make rather severe demands
on the purity of the replicase preparation employed.  It must be sufficiently
free of contaminating nucleases so that there is a high probability that a single
strand will initiate and complete its replication.  Further, the content of con-
taminating RNA must be low enough so that tubes that receive no added tem-
plate molecule will not show evidence of synthesis in the time period of the experi-
ment.  Note that as little as 1 ppg of residual RNA in 10 pg of enzyme (i.e., 1 in
lo7 contamination by weight) would correspond to the presence of 1.2 x lo7
strands.  This difficulty can be obviated with the use of a variant QP-RNA that

866

VOL. SO, 10623 GENETICS: LEVISOHN AND SPIEGELMAN SS7

grows much faster than the ordinary &@-RNA molecules expected as contami-
nants; this procedure was followed in the experiments to be described.

It is the purpose of the present paper to detail thc relevant experiments.
The data obtained show that the necessary prerequisites can be met arid that it
is in fact possible to produce clones of variant-RNA molecules.

                          QS-replicase preparations were
purified from &@-infected Escherichia coli Q13 cells according to Haruiia and Spiegelman12
except that the purification on an 0-(diethylaminoethy1)cellulose (1)EAE-cellulose)
column was done twice. The enzyme preparations used were checked repeatedly for ab-
sence of "background reaction," i.e., synthesis of RNA without addition of external
template.

(b) Assay of enzyme activity: The procedure of Spiegelman et aL3 was used with slight
modifications. The standard replicase reaction mixture included, in addition to the com-
ponents used previously, 0.75 pmole ethylenedianiinetetraacetate (EDTA) and an addi-
tional 0.75 pmole magnesium chloride per 0.25 ml. Unless otherwise specified, each reac-
tion had a total volume of 0.125 ml, containing 20 pg Qp-replicase, and was incubated for
15 min at 38°C.  ~P~~-uridine 5'-triphosphate (UP) or ~P~~-guanosine 5'-triphosphate
(GTP) was synthesized and used to monitor incorporation into acid-insoluble (5T0 tri-
chloroacetic acid (TCA)) polynucleotide as described previously.6

Unless specified otherwise, results are expressed in terms of standard reaction mixtures
with an input of 106 cpm P3*-UTP or P32-GTP per 0.25 ml after subtraction of background
counts. At this value, incorporation of 4100 cpm of P3*-UTP and 4150 cpm of P32-GTP
are equivalent to the synthesis of 1 pg of variant RNA.

            As detailed previously,7 electrophoresis through 3.6% pre-
swollen, 0.9 x 4.5 cm, bis-acrylamide cross-linked polyacrylamide gels was carried out
for 2 hr at 10 ma/gel.  After the gels were frozen, 0.5-mm slices were made with a pre-
cooled microtome.  These were washed with 4y0 TCB, dissolved for 6 hr at 80°C in
30% Hz02, and counted in liquid scintillation fluid.

(d) Ultraviolet irradiation: A solution of 7 pg/ml &@-RNA was irradiated with a 15-
watt Champion germicidal ultraviolet lamp from a distance of 142 mm for 7 niiii (to be
described elsew11ere.s).

ResuLk.-(a) Selection of variaizt-2 RNA molecules: In the original serid
transfer  experiment^,^ which led to the isolation of variant-1 (V-1), the product
of each reaction was diluted 12.5-fold in the course of being used as a template for
the next tube.  To maintain the selection pressure, the period of incubation was
shortened at intervals.

To isolate a new fast-growing mutant for the present investigation, a modifi-
cation was introduced in the selection procedure.  The incubation interval at
38°C was held constant at 15 minutes and increasing selection pressure was
achieved by recurrent sharp increases in the dilution experienced by successive
transfers.                          First, it was inter-
esting to see whether a fast-growing mutant, distinguishable from V-1, would
emerge.  Further, this type of selection could increase the severity and extent of
the selection pressure.  Finally, the use of very low initiating inputs would more
closely mimic the kind of experiments required for the Poisson analysis.  Con-
sequently, the mutant finally evolved would be more suitable for experiments of
this nature.

The &@-RNA used as the initial template was subjected to ultraviolet irradi-
ation as described in Materials and Methods.  This treatment is not essential but
was included to accelerate the rate of mutant production.  Experiments with

Materials and Methods.-(a) Enzyme preparutions:

(c) Gel electrophoresis:

This variation was employed for several reasons.

868 GEMETICS: LEVISOHN AND SPIEGELMAN PROC. N. A. s.

unirradiated RNA yielded similar mutants.  The first reaction was initiated by
the addition of 0.07 pg RNA, 0.01 ml being transferred to the next tube.  This
125-fold dilution transfer was continued for five tubes, at which point a fast-
growing mutant had already arisen.  In the next seven transfers, 0.01 ml of u
1 x 106-fold dilution of each reaction was used, correspondirig to approximately
1 x loj strands.  In the final five transfers, 0.01 ml of dilutions ranging from
1 X 107-fold to 2.5 X 10'O-fold were employed.  The variant RNA that evolved
by the 17th transfer was selected for more detailed study and will be called
variant-2 (V-2).

                 It was interesting to compare the proper-
ties of V-1 and V-2, particularly in their response to low inputs of template.
Figure 1 and Table 1 show the behavior of the two variants in a 15-minute
incubation at, 38°C.  V-2 is clearly superior at initi:bt,ing levels hrlow 100 pppg
of RNA.

Figure 2 compares the growth kinetics of the two variants during the arithmetic
(A) and exponential (R) growth phases.  During the exponential period, the
doubling time of V-2 was 0.403 minute as compared with 0.456 minute for V-1.
With this difference in growth rate, V-2 would experience four more doublings
(i.e., a 16-fold increase) than V-1 in a 15-minute period of logarithmic growth.

In view of the different rates of synthesis, it was interesting to make a size
comparison.  The two variants were radioactively labeled, one with H3, the
other with                             The results shown in
Figure 3 indicate that the double-stranded intermediates ($wl peak) and the
single strands (secoiad peak)5 all fall in the same region of the gel.  The barely
detectable difference in size is not enough to account for the higher growth rate of
v-2.

(b)

Comparison of V-i aird V-2:

and coelectrophoresed in the same gel.

FIQ. 1 .--Comparison between
the synthesis of V-1 RNA and
V-2 ltNA when used as tem-
platesat very low concentrations.
The indicated concentrations
of V-1 RNA and V-2 RNA
were added as templates to 0.125
in1 of standard reaction mix-
ture. Following a 15-min in-
cubation at 38"C, the P32-
UTP incorporation into acid-in-
soluble material was determined
as described in Materials and
Methods.

TEMPLATE: pppglO.125 ml

Vor.. 60, 1968 GENETICS: LEVISOHN AND SPIE'GELAIAN

869

TABLE 1.

RNA synthesis initiated by V-I and V-2.

Initial concentration

(~~pg/0.125 lnl)

of template v- 1

((yim / 0.25 ml)

1 0
10 0
102 0
103 5,850
104 11,300

106 17 000

105 13,200

   v-2
(rpni/0.25 nil)

I, 180
6,700
10,300
16,700
18,800
20,100
22 400

Conditions are the same as in Fig. 1.

(c)

CloTiing of V-2:

           1,ike V-l,5 V-2 has a molecular weight of ahout 177,000,
corresporiding to approximately 550 residues; thus one strand is equivalent to
0.29 pppg of RNA.

A reaction was run to syrithesizc 13.2 pg of P32-labeled V-2 RNA as described
in Materials and Methods.  The product was diluted in a standard Qp-reaction
mixture to 2.9 pppg/ml, and 0.1 ml was placed in each of 82 tubes.  Thus,
each tube received, on the average, one strand of template.  All these steps
were carried out at O'C, at which no detectable synthesis occurs.  The rack of
tubes was then placed in a 38°C water bath and held there for 30 minutes.  The
reactions were stopped arid the amount of radioactive polynucleotide synthesized
in each tube was determined (see Materials and Methods).  The results are
summarized in Table 2.

If the assumptions underlying the Poisson distribution have been satisfied by
the conditioiis of the experiment, one would expect 36.8% of the S2 tubes (Le.,
30.2) to show no synthesis.  This is in excellent agreement with the 30 tubes
found.

It was interesting to see whether among the tubes exhibiting synthesis one
could distinguish those ittitiatcd by  on^ strand from those that recrivrd two or

FIG. 2.-Kinetics of RNA synthesis initiated by V-1 and V-2 RNA.

                                           Standard reaction
mixtwes containing 0.0001 pg variant RNA per 0.25 ml were incubated at 38°C.  Samples of
0.05 ml were withdrawn at the indicated times to determine the incorporation of P32-UTP into
acid-insoluble material.  The actual counts per sample were 14 times higher than the stan-
dardized counts (see Materials and Methods), which were plotted on an arithmetic scale (A) and
a semilogarithmic scale (B).

870 GENETICS: LEVISOHN AND SPIEGELMAN PROC. N. A. S.

                            FIG. 3.-Gel electrophoresis
                                   of V-l and V-2. Standard
1 reaction mixtiires of 0.125 nil
                            initiated with 0.2 pg of V-1 or
                           0.01 pg of V-2 arid labeled with

05 or 4 x 106 cpm P32-UTP, re-
5 spectively, were incubated for

   were terminated with 0.01 ml of
5 2.5% sodium dodecyl sulfate.
   Portions of the two reactions
   were mixed and electrophoresis
   through polyacrylamide gel was
   carried out as in Materials and

Methods.

c about 1.5 X 107 cpm H8-ATP


m w 20 min at 38°C. The reactions
I,

9

XI

I

more.  To achieve the desired identification, the following calculation was per-
formed.                                          The
sum of counts observed in all tubes (25,340 cpm after correction for controls
without added RNA, which was 22 cpm/tube) is a measure of the total RNA
initiated by all 82 templates.  If the sum is divided by 82, one obtains 309 cpm
as the average amount ascribable to a single template.  The actual incorporation
observed in each tube can then be divided by 309 arid the result rounded out to
the nearest integer to yield an approximation of the number of strands that initi-
ated the synthesis in that tube.  The resulting data can then be compared with
that expected, as is done in Table 2.  The average and the corresponding devi-
ation (2~) for each group are also listed.  One finds here a remarkable agree-
ment between what was found and what was expected from a Poisson distri-
bution of template strands in the 82 tubes.  The chances are clearly excellent
that tubes that exhibited incorporation close to the average expected for one
strand (309 cpm) were in fact initiated by single strands.

As a further check on the validity of t,liis sort of approximation, a modification
of the original Poisson experiment was performed, in which the average input was
varied.  Three sets of 30 tubes each were set up as described for Table 2.  Each
set was seeded with a different average number of initiating strands ranging

The total experiment involved the functioning of 82 templates.

TABLE 2.

Distribution of template strands at an average input of one template strand per
tube.

Strands Polynucleotide Among 82 Tubes Average f 2u

per tube        (r)         Expected       Found       (cpm/tube)
  0          0.368          30.2         30          17 f 76
  1          0.368          30.2         29         321 i 173
  2          0.184          15.1          19         613 f 184
  3          0.0613          5.4          3         841 i 58
  4         0.0153          1.3           1        1334

A reaction mixture in which 13.2 pg of V-2 RNA were synthesized was diluted to a final concen-
tration of 2.9 pppg/rnl RNA in a standard reaction mixture containing 7.8 X lo5 cpm P2-GTP per
0.25 rnl.                     Following a 30-min incubation at 38OC, the acid-in-
soluble P32-GTP was determined as described in Materials and Methods.  The average is for the
actual number of counts (not standardized) after subtraction of the average of nine control tubes
(22 cprn) to which no variant RNA was added.  In this experiment. 3240 cpm is equivalent to 1 fig
of variant RNA.

Each of 82 tubes received 0.1 ml.

VUL. 60, 1968 GENETICS: LEVISOHN AND SPIEGELMAN 871

TABLE 3.  Distribution of template strands at an average of one, two, or three template strands
     per tube.

(r)

Strands
per tube

0
1
2
3

>3

--___~-~ Average Input---

One Strand Two Strands

Found    Expected    Found   Expected
13       11.0        3       4.1
9       11 .0       9       8.1
6       5 .5      13      8.1
  1       1.8       2      5.4
  1       0.7       3      4.3

7

-_

Three Strands

Found    Expected
  2       1.5
  2       4.5
7       6.7
6       6.7

13 10.6

V-2 RNA was diluted to final concentrations of 2.9, 5.8, and 8.7 p,~pg/inl RNA in :I standard
reaction mixture containing 80 pg/0.25 nil (20 replicase.  Three sets of 30 tubes each were set up.
Each set received 0.1-ml portions of one of the three reaction mixtures and was then incubated for
25 inin at 38" ad treated as in T:ible 1.

from one to three.  The results :m sunimariactl in Tahlc 3.  Ag:Lin, thc zgrc('-
merit between thc obrcrved and expcctccl at each avcragc input, is excellent.

Discusszm-It is apparent from the results clcscritml that the replicasc system
has been sufficiently freed of interfering reactions to pcrmit initiation of syn-
thesis by a single strand of template.  Thus, the generation of clones of de-
scendants from an individual lWA molecule was made possible.  It is of inter-
est and no little convenience that one can actually identify, by the amount of
RNA synthesized, those tubes that were initiated by a single strand.  It is not,
however, necessary to depend 011 this property of the system.  One can start with
a dilution such that the vast majority of the tubes receive no templates.  Under
these conditions, those that do receive templates are very unlikely to receive more
than one.

We may note here another method of cloning that is, in principle, possible.
The enzyme and substrates could be incorporated into a semisolid medium
(e.g., a soft agar layer), and thc initiating strands spread on the surface.  The
resulting clones could thcn be located aid picked much as one does with bacterial
colonies.

The fact that 0.29 pppg of RXA satisfies the Poisson expcctat,ion for an average
of one strand immediately allows us to calculate the molecular weight as 174,000
daltons, which is in good agreement with that deduced from gel electrophoresis.
Needless to say, there is no guarantce that all individuals in a clone will be
identical.  They will be as similar as the mutation frequency will permit, which
is as close as one can get with self-duplicating objects.  In this connection, it
should be noted that V-1 and V-2 have retained their phenotypic difference in
growth rates over many transfers.  This suggests the possibility of isolating
and maintaining other mutant types, a possibility that has been realized in
experiments recently performed.

We have pointed out elsewhere5 the potential usefulness of developing mutant
molecules.  Cloning them increases their utility for more detailed studies such
as sequencing.  It is also it necessary requisite for a ``genetic" analysis of the
replicating molecules.

Finally, a note of caution: the synthesis of self-duplicating molecules entails
the risk of introducing them as laboratory environmental contaminants.  As
was seen in the experiments described, one molecule can take over a reaction.

This potential sourcc of cotifusion has in fact been realized several times in our
laborat,org, beginning early in 1967 soon after we synthesized our first fast-
growing mutant.  The initial indication that we were being inconvenienced by
our own creations was a sudden inability to prepare replicase exhibiting depend-
ence on added t.ernplate.  In all inst'ances, t'he difficulty was traced to contami-
nation of a conimorily employed assay reagent with a fast-growing variant.
The uriiquc phenotypic properties enabled us to identify it as one we had
synthesized and isolated. These molecules are remarkably stable under a
variet,y of conditions.  As much care as is normally employed with bacteria
and viruses must be exercised to exclude these molecules as unwanted intruders
in experiments.

  5ummal.y.-Experiments have been described that show that purified
QP-replicase can be initiat>ed t,o synt,hesize copies of a mutant &@-RNA by a
single strand of template.  The resulting clone of descendants provides a popu-
lation of individuals possessing the liind of uniformity required for sequence
studies.  Further, the fact that distinguishable clones can be isolated and
maintained makes possible the inception of an in vitro genetics of replicating
molecules.

tional Cancer Institute and the National Science Foundation.

* This investigation was supported by USPHS research grant no. CA-01094 from the Na-

t Recipient of a Damon Runyon Cancer Research Fellowship.

Pace, N. R., and S. Spiegelman, these PROCEEDINGS, 55, 1608 (1966).
Haruna, I., and S. Spiegelman, these PROCEEDINGS, 54, 579 (1965).
Spiegelman, S., I. Haruna, I. B. Holland, G. Beaudreau, and D. 11.. Mills, these PROCEED-

Pace, N. R., a,nd S. Spiegelman, Science, 153, 64 (1966).
Mills, 1). It., E. L. Peterson, and S. Spiegelman, these PROCEEDINGS, 58, 217 (1967).
Harnna, I., K. NOZII, Y. Ohtaka, and S. Spiegelman, these PROCEEDINGS, 50, 905 (1963).

INGS, 54, 919 (1965).

' Bishop, D. 11. L., J. It. Claybrook, and S. Spiegelman, J. Mol. Biol., 26, 373 (1967).
8 Levisohn, It., G. Weber, and S. Spiegelman, in preparation.

Levisohn, lt., arid S. Spiegelman, in preparation.