Proc. Natl. Acad. Sci. USA
`Vol. 81, pp. 4627-4631, August 1984
`Biochenustry
`
`Multiple joined genes prevent product degradation
`in Escherichia coli
`
`(multidomain/proinsulin/tandem copies/stable product)
`
`SHI-HSIANG SHEN
`
`Connaught Research Institute, 1755 Steeles Avenue West, Willowdale, Ontario, Canada M2R 3T4
`
`Communicated by David R. Davies, March 6, 1984
`
`A method is described that allows the expres-
`ABSTRACT
`sion of a stable human proinsulin product in Escherichia coli
`as encoded by, either a fused or an unfused gene construction.
`In the fused system, the human proinsulin coding sequence is
`joined to the 3’ side of a fragment containing the lac promoter
`and the coding sequence for a small part of the NH; terminus
`of B-galactosidase. In the unfused system, the proinsulin cod-
`ing sequence _is linked directly to a fragment containing the Tac
`promoter followed by a bacterial Shine—Dalgarno sequence. In
`both systems, the human proinsulin product is too unstable to
`be detected by NaDodS0./polyacrylamide gel electrophoresis
`or even pulse—chase analysis. However, when multiple copies
`of the proinsulin coding sequence are tandemly linked such
`that the resultant protein product contains multiple copies of
`the proinsulin domain, the stability of the product is markedly
`increased in both the fused and the unfused expression sys-
`tems. In the unfused system, three tandemly linked proinsulin
`polypeptide domains are required for stabilization, whereas
`two proinsulin domains plus the bacterial leader protein en-
`hance stability in the fused system. The polypeptide product of
`a multiple copy proinsulin gene can be cleaved into single
`proinsulin units by cyanogen bromide treatment.
`
`Recombinant DNA technology has been used extensively
`for the development of bacterial strains expressing useful eu-
`karyotic products such as human insulin (1, 2), human
`growth hormone (3), interferon (4), and viral vaccines (5-7).
`However, the large-scale production of certain eukaryotic
`products has often been limited because of their instability in
`the bacterial host (4, 7, 8). It has been suggested that many
`eukaryotic foreign peptides are recognized as abnormal pro-
`teins in Escherichia coli and consequently are degraded (9,
`10). For example, the half-life of humanproinsulin in E. coli
`has been reported to be 2 min (11). A common strategy to
`circumvent this problem has beenuto fuse the coding se-
`quence of the desired product to that of a host structural
`gene, resulting in the expression of a hybrid polypeptide.
`Such an expression system may provide a “native” portion
`added to the foreign product, thus preventing its degrada-
`tion. Recently, insulin, proinsulin, and a variety of other eu-
`karyotic proteins and viral vaccines have been produced by
`this method (1, 2, 12, 13). However, a major disadvantage of
`this approach is that the desired product constitutes only a
`small portion of the hybrid polypeptide, resulting in reduced
`yield and increased difficulties in purification. Efforts to re-
`duce the prokaryotic moiety to a small portion of the hybrid
`product also usually render the product unstable (13).
`A strategy is described here to prevent the degradation of
`human proinsulin in E. coli by amplifying the proinsulin cod-
`ing sequence in the expression plasmid in such a way that it
`results in a multidomain polypeptide. It is reported that such
`
`The publication costs of this article were defrayed in part by page charge
`payment. This article must therefore be hereby marked “advertisement”
`in accordance with 18 U.S.C. §1734 solely to indicate this fact.
`
`4627
`
`a multidomain polypeptide is stable in E. coli and can be
`quantitatively converted to its monomer proinsulin units by
`cyanogen bromide cleavage. A detailed description of the
`production of authentic human insulin by this system will be
`published elsewhere.
`
`MATERIALS AND METHODS
`
`Materials. Enzymes were purchased from New England
`Biolabs and Bethesda Research Laboratories. L-[35S]Meth-
`ionine (1210 Ci/mmol; 1 Ci = 37 GBq) and L-[35S]cysteine
`(1000 Ci/mmol) were obtained from Amersham.
`Bacterial Strains and Recombinant Plasmids. E. coli K-1_2
`strain JM103 [A(lac pro), thi, strA, supE, endA, sbcB, hsdR,
`F’traD36, proAB, lacI", ZAMI5] was used for all proinsulin
`expression experiments. Plasmid pAT/PI is a derivative of
`pBCA4 that carries the entire human proinsulin gene on a
`BamHI/EcoRI fragment (14). The original EcoRI linker that
`precedes the initiation codon was converted from C-C-G-G-
`A-A-T-T-C-C-G-G to the structure A-G-A-A-T-T“-C-T,
`to
`provide a modified proinsulin sequence that would maintain
`the desired reading frame in subsequent steps. The resulting
`modified proinsulin sequence on a BamHI/EcoRI fragment
`was inserted into plasmid pAT153 between the EcoRI and
`BamHI sites (unpublished data). Plasmid pTac bears a Tac
`promoter sequence modified from plasmid pDR540 (15) in
`the following way: pDR540 DNA was cut with BamHI and
`HindIII and digested with mung bean nuclease to remove the
`sticky ends. The isolated Tac promoter fragment was insert-
`ed into the vector plac 504/PI, which had been digested with
`EcoRI and HindIII and filled in with DNA polymerase I
`(Klenow fragment). The resulting construct was designated
`as pTac. Details concerning the construction of plac 504/PI
`will be published elsewhere.»Plasmid plac 239,
`to be de-
`scribed in detail elsewhere, carries a lac promoter and ‘part
`of the lacZ gene encoding 80 amino acids of the NH; termi-
`nus of /3-galactosidase.
`_
`Purification of the Proinsulin Product and Polyacrylamide
`Gel Electrophoresis. Bacterial cultures (3.0 ml) were grown
`in YT medium and induced at an approximate cell density of
`0.1 OD55o unit by the addition of isopropyl B-D-thiogalacto-
`side to 1 mM. After 12 hr, the cells were harvested by cen-
`trifugation. Cell pellets were either dissolved directly in the
`sample buffer and analyzed by NaDodSO4/polyacrylamide
`gel electrophoresis as described by Laemmli (16) or further
`purified as follows: cells were suspended in 1.0 ml of the
`sonication buffer as described by Bikel et al. (17) and sub-
`jected to some disruption. The sonicated material was centri-
`fuged. The pellets were analyzed by NaDodS04/polyacryl-
`amide gel electrophoresis as described above.
`Pulse-Chase Experiments. Bacterial cultures (5 .0 ml) were
`grown in minimal salts medium/0.2% glucose/0.001% thia-
`min at 37°C.
`t a cell density of 0.5 OD5(,o unit, the cultures
`were induce as before and, after 5 min, 300 p.Ci each of L-
`[35S]methionine and L-[35S]cysteine were added. Labeling
`Mylan v. Genentech
`Mylan V. Genentech
`IPR2016-00710
`IPR2016-00710
`Genentech Exhibit 2070
`
`Genentech Exhibit 2070
`
`

`
`4628
`
`Biochemistry: Shen
`
`Proc. Natl. Acad. Sci. USA 81 (I984)
`
`proceeded for 30 sec, after which incorporation of radioac-
`tivity was chased by the addition of 100 pl of 500 mM unla-
`beled 1'--methionine/L-cysteine. Aliquots (0.5 ml) of each cul-
`ture were removed at predetermined times and immediately
`centrifuged for 10 sec, and cells were sonicated as described
`above.
`Cyanogen Bromide Cleavage. The proinsulin polypeptide
`purified by sonication was dissolved in 20 ml of 70% formic
`acid and cleaved by treatment with 50 mg of cyanogen bro-
`
`mide per mg of protein for 35 hr at room temperature.‘Prod-
`ucts were analyzed by NaDodSO4/polyacrylamide gel elec-
`trophoresis.
`
`RESULTS
`
`Construction of the Multiple Pi-oinsulin Genes. The strategy
`for the construction of the multidomain proinsulin genes in
`both the fused and unfused systems is diagrammed in Fig. 1.
`
`P
`
`EcoRl
`
`am,”
`
`Oliqonucleotides
`
`Ecom
`
`SIaN
`
`:
`mmAAfTC
`‘G
`
`Pl
`
`(RI)
`Pl (Analog)
`
`
`
`FIG. 1. Construction of the multiple proinsulin genes. The multiple proinsulin gene was isolated from an agarose gel and inserted into the
`vectors fpTac and fplac 239. fpTac is an EcoRl/BamHI fragment of pTac. The construction was arranged to give an EcoRl site immediately 3'
`to the promoter. fplac 239 is an EcoRI/BamHI fragment of plac 239. The construction was first arranged to give an EcoRI site at amino acid 80
`in the IacZ gene for insertion of multiple genes and then to restore the proper reading frame of the proinsulin coding sequences by cleaving at the
`EcoRI site of the lac—proinsulin junction, digesting with mung bean nuclease, and recircularizing by blunt-end ligation.
`
`IIIII '
`
`Kinuo
`‘up
`
`III I
`
`
`
`

`
`Biochemistry: Shen
`
`Proc. Natl. Acad. Sci. USA 81 (1984)
`
`4629
`
`A and B are synthetic oligonucleotides having the following
`sequences:
`
`increased. The product of two joined coding sequences is
`visible (lane 2) after removal of some of the bacterial pro-
`
`A 5’ C-C-T -C-T- A-C-C -A-G- C-T-G -G- A-G-A -A-C- T-A-C -T-G- C-A-A-C-A -G-G- C-G-C 3’
`
`B 3’
`
`A- T-G-G-T-C- G-A-C -C- T-C-T -T-G—A-T-G -A-C- G-T- T-G-T -C-C- G-C-G-T- T-A-A 5’
`
`T
`
`(SfaNI cut site on the proinsulin gene)
`
`Oligonucleotide A was first phosphorylated and then hybrid-
`ized to oligonucleotide B. After hybridization, the two oligo-
`nucleotides form a part of the human proinsulin coding se-
`quence (14) phosphorylated on the end beginning at the
`SfaNI cut site near the 3’ end of the coding sequence and
`terminating with the last proinsulin codon AAC (asparagine)
`followed by the unphosphorylated additional sequence
`
`Arg Arg
`5’ ...AGG CGC
`
`3’
`
`3’
`
`.
`
`.
`
`. TCC GCG T-T-A-A 5’
`
`The extra sequence of the synthetic oligonucleotides can en-
`code two additional arginine residues and also provides a
`sticky end for ligation to an EcoRI sticky end. Note that such
`a ligation destroys the EcoRI site because of the C-G base
`pair (boxed in Fig. 1).
`The synthetic human proinsulin gene was removed from
`pAT/PI by EcoRI/BamHI digestion, dephosphorylated, and
`further digested with SfaNI. The asymmetrically phospho-
`rylated proinsulin gene fragment enables unidirectional liga-
`tion to the A and B oligonucleotides, resulting in the proinsu-
`lin gene analog [PI (analog) in Fig. 1]. The proinsulin gene
`analog was joined to another proinsulin gene sequence
`(EcoRI/BamHI fragment) in a controlled unidirectional liga-
`tion reaction canied out with T4 DNA ligase in the presence
`of EcoRI and BamHI. The resulting double proinsulin gene
`fragment was further ligated to additional proinsulin gene an-
`alogs to give plasmids containing up to seven joined proinsu-
`lin coding sequences.
`Joined proinsulin gene fragments containing different
`numbers of the proinsulin coding sequence were inserted
`into one or the other of two vectors, pTac and plac 239 (to be
`described in detail elsewhere). In the former case, a Tac pro-
`moter followed by a bacterial Shine-Dalgamo sequence is
`joined directly to the proinsulin coding sequence with its
`own initiator codon. The resulting plasmids give products
`containing one or more proinsulin domains. pTac/ZPI in Fig.
`1 has two such domains. In the latter case, the proinsulin
`coding sequence is preceded by the lac promoter and 80 co-
`dons from the 5’ end of the B-galactosidase gene. The result-
`ing plasmids give fused gene products (pan of B-galacto-
`sidase fused to one or more domains of proinsulin). plac
`239/2PI in Fig. 1 has two proinsulin domains fused to 80 ami-
`no acids of B-galactosidase. The various constructs, contain-
`ing one to seven copies of the proinsulin coding sequence,
`were confirmed by restriction mapping and DNA sequence
`analysis.
`Expression of Multiple Proinsulin Genes in E. coli. The mul-
`tidomain proinsulin gene constructs, fused to B-galacto-
`sidase and unfused, were expressed in E. coli strain JM103
`after induction, and the resultant products were analyzed by
`NaDodS04/polyacrylamide gel electrophoresis as shown in
`Fig. 2. In the unfused system (Fig. 2A), the product of one
`proinsulin coding sequence is too little to be visualized in a
`stained gel after NaDodSO4/polyacrylamide gel analysis
`(lane 1). However,
`the level of product is strikingly en-
`hanced as the proinsulin coding sequence copy number is
`
`teins by sonication, while the amount of product with three
`joined coding sequences is considerably increased (lane 3).
`A further increase in the number ofjoined coding sequences
`up to five does not appear to cause a significantly greater
`quantity of product (lane 5). A definite decrease is seen with
`seven joined coding sequences (data not shown).
`Similar results were obtained in the fused expression sys-
`tem in which the first proinsulin coding sequence was pre-
`ceded by 80 codons of the lacZ gene (Fig. 2B). The fusion
`product with a single proinsulin coding sequence was diffi-
`cult to visualize (lane 1) although, after partial purification
`by sonication, a light band could be detected (lane 4) by load-
`ing twice as much sample as in lane 1. Similarly to the un-
`fused system, the amount of product obtained in the fused
`system was greatly enhanced by increasing the number of
`joined proinsulin domains to two or more (lanes 2 and 3 as
`well as 5 and 6).
`Stability of Products. As described above, the appearance
`of prominent bands on NaDodSO4/polyacrylamide gel elec-
`trophoresis indicates that the multidomain proinsulin prod-
`ucts are present in markedly greater amounts than their
`monomeric counterparts. A possible cause of this result is a
`greater stability of the multidomain polypeptide. To investi-
`gate this factor, pulse—chase experiments were carried out.
`As shown in Fig. 3, the expected product from bacteria har-
`boring a single copy of the proinsulin coding sequence (plas-
`mid pTac/PI) could not be detected (lanes 1-4), even with a
`chase as short as 30 sec. The product from two joined proin-
`
`A
`
`1
`
`2
`
`3
`
`4
`
`"
`
`-
`
`5 M
`_.
`
`2
`
`"re
`-
`-
`
`no
`
`B
`
`..
`
`“
`
`°“”£_Q
`
`igfifl
`
`9 Q
`
`"'
`
`-
`
`so
`
`9
`
`123M456
`
`FIG. 2. Electrophoretic analysis of multiply expressed proinsu-
`lin polypeptides. Proinsulin polypeptide preparations were analyzed
`on NaDodSO4/15% polyacrylamide gels. (A) Partially purified prod-
`ucts of the unfused system construction. Lanes 1-5 represent prod-
`ucts equivalent to 500 pl of original culture from plasmids pTac/PI,
`pTac/2PI, pTac/3P1, pTac/4P1, and pTac/SP1, respectively. (B)
`Products of the fused construction system. Lanes 1-3 represent total
`cell protein equivalent to 150 pl of original culture bacteria contain-
`ing plasmids plac 239/PI, plac 239/2P1, and plac 239/3P1, respec-
`tively. Lanes 4-6 represent products from the same plasmids as in
`lanes 1-3 after partial purification by sonication, Lanes M: protein
`molecular weight markers: from top to bottom, 94,000, 67,000,
`30,000, 20,100 and 14,400.
`
`

`
`4630
`
`Biochemistry: Shen
`
`Proc. Natl. Acad. Sci. USA 81 (1984)
`
`12 3
`
`4
`
`5
`
`6 78 9 101112131/115161718
`
`1
`
`2
`
`WY
`
`C9. "'
`
`32‘
`
`M
`......
`
`iivvfi
`
`‘
`
`a
`
`1
`
`X
`
`FIG. 3. Pulse—chase experiment. Lanes 1-4 (pTac/PI, single
`copy) and lanes 5-8 (pTac/2P1, two copies) were chased for 0.5, 3,
`5, and 15 min, respectively; lanes 9-13 (pTac/3P1, three copies) and
`lanes 14-18 (pTac/4PI, four copies) were chased for 10, 30, 60, 120,
`and 200 min, respectively. Each lane represents an aliquot of cell
`equivalent to 150 pl of original culture. Samples were electropho-
`resed on a NaDodS04/ 15% polyacrylamide gel and exposed on Ko-
`dak X-Omat AR5 film for 2 days. The arrow indicates the location of
`the expected product of a single proinsulin coding sequence.
`
`sulin coding sequences (plasmid pTac/2P1) has a half-life of
`<5 min (lanes 5-8) while the half-lives of the products from
`three and four joined proinsulin coding sequences (plasmids
`p’l‘ac/3PI and pTac/4PI) were both >60 min (lanes 9-13 and
`14-18, respectively).
`Similar results were obtained in pulse—chase experiments
`in the fused system. The two-domain proinsulin product was
`found to have a half-life of >60 min, whereas the monomeric
`product was difficult to detect (data not shown).
`These two sets of experiments clearly show a large effect
`of the multidomain constructs on the stability of the resulting
`polypeptide. Moreover, they show that the effect can be ob-
`tained even when the resulting polypeptide is completely
`“foreign" to E. coli (in the unfused system).
`It should be noted that the partial purification step involv-
`ing sonication and centrifugation after the chase required
`more than 5 min. Protein degradation during this period may
`have been responsible for the failure to detect the monomer
`after the short chase.
`Cleavage of the Product. The multidomain proinsulin prod-
`ucts will have the following structure at the domain junc-
`trons:
`
`. .. Cys- Asn- Arg -Arg-Asn - Ser- Met-Phe ...
`.
`.
`. TCG AAC AGG CGC AAT TCT ATG l'l'l .
`.
`.
`
`<
`proinsulin gene
`
`L’.
`proinsulin gene
`
`Thus, cyanogen bromide treatment, for example, cleaves a
`pentadomain proinsulin molecule into four proinsulin analog
`moieties (each having the extra COOH-terminal pentapep-
`tide Arg-Arg-Asn-Ser-homoserine) plus one authentic proin-
`sulin moiety from the COOH-terminal domain. Fig. 4 shows
`the results of such a cyanogen bromide cleavage experiment.
`As expected, the pentadomain proinsulin polypeptide (lane
`2) was quantitatively cleaved to single proinsulin and proin-
`sulin analog units (lane 1). These monomeric products are
`reactive with guinea pig anti-porcine proinsulin serum in the
`protein blot radioimmunobinding assay (data not shown).
`The resultant proinsulin and proinsulin analog units should
`fold into the native conformation and be converted to au-
`
`FIG. 4. Cleavage of a multidomain proinsulin polypeptide with
`cyanogen bromide. The proinsulin polypeptide product of plasmid
`pTac/SP1 (five tandem copies of the proinsulin coding sequence)
`was purified by sonication and cleaved with cyanogen bromide, and
`the products were analyzed on a NaDodS04/15% polyacrylamide
`gel. Lanes: 1, cleaved product; 2, uncleaved product; M, molecular
`weight standards as in Fig. 1.
`
`thentic human insulin by digestion with trypsin and carboxy-
`peptidase B.
`
`DISCUSSION
`
`I have shown that expression of human proinsulin in E. coli
`can be greatly enhanced by directing its synthesis through a
`multidomain structure in which a single polypeptide is com-
`posed of several tandemly joined proinsulin analog units. I
`have also shown that this enhanced expression is, to a con-
`siderable extent, due to enhancement in the stability of the
`resulting polypeptide (as judged by the half-life). In the fused
`system in which the NH;-terminal proinsulin analog unit is
`fused to an 80-amino acid leader from the NH; terminus of B-
`galactosidase, only two proinsulin domains are required to
`make the product stable (half-life >1 hr). However, when no
`bacterial leader is present on the proinsulin polypeptide
`product, three proinsulin domains are necessary to render
`the product stable. This result also shows that the product
`can become stable even in the complete absence of any E.
`coli-related polypeptide sequences, provided that the tan-
`demly linked polypeptide domains reach a critical number.
`Nonetheless, E. coli-related polypeptide sequences do have
`some beneficial effects on yield over and above their effects
`on stability. Thus, the stability of the multidomain polypep-
`tides with and without the 80-amino acid [3-galactosidase
`leader was the same (>60 min half-life), but the yield of the
`fused product from plac 239/3P1 was at least three times
`greater than that of the unfused product from pTac/SP1. Pre-
`sumably, this is the result of other effects such as increased
`mRNA stability, increased transcription, or increased trans-
`lational efficiency. Increasing the number ofjoined domains
`from three to five did not further increase the stability of the
`products as judged by pulse—chase experiments in which the
`half-lives of the products from three and four proinsulin cod-
`
`

`
`Biochemistry: Shen
`
`Proc. Natl. Acad. Sci. USA 81 (1984)
`
`4631
`
`ing sequences were both found to be the same—approxi-
`mately 120 min. Why increasing the number of joined se-
`quences to seven resulted in a lower yield remains an inter-
`esting problem for study. Also, it would be interesting to
`determine whether changes in the number ofjoined proinsu-
`lin coding sequences alter transcriptional efficiencies or
`mRNA stability.
`It has been reported that disulfide bond formation does
`not occur in the cytoplasm of E. coli because of a low elec-
`trochemical potential in the cell (18). Thus, the product sta-
`bility observed with tandemly repeated proinsulin domains is
`unlikely due to increased numbers of intramolecular disul-
`fide bridges. I have observed, by electron microscopy, inclu-
`sion-like bodies in cells producing stable tandemly linked
`proinsulin polypeptide but could not find such structure in
`cells producing an unstable single-domain polypeptide prod-
`uct. Such inclusion-like bodies may result from the aggrega-
`tion of polypeptide products of limited solubility within the
`cell. A similar interpretation was proposed concerning the
`formation of inclusion-like bodies by such aggregation in
`production of the recombinant insulin polypeptide of E. coli
`(19). Consequently, I suggest that the aggregation of the tan-
`demly linked proinsulin polypeptide products sequestered
`into the insoluble inclusion-like bodies could avoid an attack
`
`from proteolytic enzymes in E. coli. A similar suggestion of
`product aggregation has been made for the stabilization of a
`degradable protein-X90 (20). The formation of protein ag-
`gregates may require not only a certain protein concentra-
`tion but also a critical length of individual polypeptide units.
`Achieving such conditions could be accomplished by either
`joining single polypeptide coding sequences together or fus-
`ing a coding sequence to a bacterial leader of sufficient
`length as discussed for the construction. The stabilization of
`somatostatin expression in E. coli was achieved by the latter
`strategy (21).
`The mechanisms by which abnormal proteins, including
`foreign polypeptides, are degraded in E. coli are not fully
`understood. The E. coli lon protease, a DNA-binding protein
`with ATP-dependent proteolytic activity, clearly plays a role
`in this process (22, 23). There are some E. coli mutants (22,
`24) that stabilize some otherwise unstable polypeptides and
`nonsense fragments. However, these mutants do not stabi-
`lize a single proinsulin polypeptide in E. coli (25).
`The method described here for stabilizing an_expressed
`polypeptide via a multidomain joined product could be appli-
`cable to the expression of other unstable products. It is an-
`ticipated that both the range of desirable peptides and the
`availability of methods for sequence-specific cleavage of
`peptide bonds will increase, thus improving the general ap-
`plicability of the multidomain polymer strategy here de-
`scribed.
`
`I am grateful to Dr. Keith Dorrington for his encouragement and
`critical reading of the manuscript, to Dr. Eric James for valuable
`discussions, and to Dr. Oliver Smithies for his valuable comments
`
`on this work. I thank Dr. Joel Haynes for oligonucleotide synthesis
`and comments on the manuscript, Dr. Robert Garvin for his com-
`ments on this work, and Dr. K. Tsai for electron microscopy obser-
`vation. I gratefully acknowledge Richard Elliott, Cheryl Davey, and
`Elizabeth Ciemniewski for their assistance and Lynne-Marie Me-
`Kay for her efforts in preparing the manuscript. This work was sup-
`ported by a Programme for Industry/Laboratory Projects contract
`from the National Research Council of Canada.
`
`8.
`
`1. Chan, S. J., Weiss, J., Konrad, M., White, T., Bahl, C., Yu,
`S.-D., Marks, D. & Steiner, D. F. (1981) Proc. Natl. Acad.
`Sci. USA 78, 5401-5405.
`2. Goeddel, D. V., Kleid, D. G., Bolivar, F., Heyneker, H. L.,
`Yansura, D. G., Crea, R., Hirose, T., Kraszewski, A., Ita-
`kura, K. & Riggs, A. D. (1979) Proc. Natl. Acad. Sci. USA 76,
`106-110.
`3. Goeddel, D. V., Heyneker, H. L., Hozumi, T., Arentzeh, R.,
`Itakura, K., Yansura, D. G., Ross, M. J., Miozzari, G., Crea,
`R. & Seeburg, P. H. (1979) Nature (London) 281, 544-548.
`4. Taniguchi, T., Guarente, L., Roberts, T. M., Kimelman, D.,
`Douhan, J ., II1, &.Ptashne, M. (1980) Proc. Natl. Acad. Sci.
`USA 77, 5230-5233.
`V
`5. Yelverton, E., Norton, S., Obijeski, J. F. & Goeddel, D. V.
`(1983) Science 219, 614-620.
`6. Malek, L. T., Soostmeyer, G., Garvin, R. T. & James, ‘E.
`(1984) in Modern Approaches to Vaccines, eds. Chanock,
`R. M. & Lerner, R. A. (Cold Spring Harbor Laboratory, Cold
`Spring Harbor, NY), pp. 203-208.
`7. Davis, A. R., Bos, T., Ueda, M., Nayak, D. P., Dowbenko,
`D. & Compans, R. W. (1983) Gene 21, 273-284.
`Itakura, K., Hirose, T., Crea, R., Riggs, A. D., Heyneker,
`H. L., Bolivar, F. & Boyer, H. W. (1977) Science 198, 1056-
`1063.
`9. Goldschmidt, R. (1970) Nature (London) 228, 1151-1154.
`10. Lin, S. & Zabin, I, (1972) J. Biol. Chem. 247, 2205-2211.
`11. Talmadge, K. & Gilbert, W. (1982) Proc. Natl. Acad. Sci. USA
`79, 1830-1833.
`_
`12. Shine, J., Fettes, 1., Lari, N. C. Y., Roberts, 1. L. & Baxter,
`J. D. (1980) Nature (London) 285, 456-461.
`.
`13. Reed, S. I. (1982) Gene 20, 255-265.
`14. Brosseau, R., Scarpulla, R., Sung, W., Hsiung, H. M., Nar-
`ang, S. A. & Wu, R. (1982) Gene 17, 279-289.
`15. Russell, D. R. & Bennett, G. N. (1982) Gene 20, 231-243'.
`16. Laemmli, U. K. (1970) Nature (London) 227, 680-685.
`17. Bikel, 1., Roberts, T. M., Bladon, M. T., Green, R., Amann,
`E. & Livingston, D. M. (1983) Proc. Natl. Acad. Sci. USA 80,
`906-910.
`18. Pollitt, S. _& Zalkin, H. (1983) J. Bacteriol. 153, 27-32.
`19. Williams, D. C., Van Frank, R. M., Muth, W. L. & Burnett,
`J. P. (1982) Science 215, 687-689.
`20. Cheng, Y.-S. E., Kwoh, D._Y., Kwoh, T. J., Soltvedt, B. C.
`& Zipser, D. (1981) Gene 14, 121-130.
`Itakura, K., Hirose, T., Crea, R. & Riggs, A. D. (1977) Sci-
`ence l98, 1056-1063.
`_
`22. Gottesman, S. & Zipser, D. (1978) J. Bacteriol. 133, 844-851.
`23. Charette, M. F., Henderson, G. W._& Markovitz-, A. (1981)
`Proc. Natl. Acad. Sci. USA 78, 4728-4732.
`24. Bukhari, A. I. & Zipser, D. (1973) Nature (London) New Biol.
`243, 238-241.
`25. Emerick, A. W., Bertolani, B. L., Ben-Bassat, A., White,
`T. J . & Konrad, M. W. (1984) Biotechnology 2, 165-168.
`
`21.