WO1990004026A1 - New plasmid constructions for high level production of eukaryotic proteins - Google Patents

Abstract

Novel plasmid constructs for high level production of eukaryotic proteins in their natural form are described.

Description

NEW PLASMID CONSTRUCTIONS FOR HIGH LEVEL PRODUCTION OF EUKARYOTIC PROTEINS

Technical Field

The present invention is related generally to the construction of plasmid vectors. More particularly, the present invention is related to the construction of novel plasmids for high level production of eukaryotic proteins in their natural form in a suitable expression vector. Background of the Invention

Plasmid systems for the expression of proteins of single subunit structure have been known. Typical of such vectors are the ones described by Amann et al, 1983, Gene, 25:167-178 and Rosenberg et al, 1983, Methods in Enzymology, 101:123-138, respectively. Bacterial expression of eukaryotic proteins is a tool of ever- increasing importance in biochemistry and molecular biology. However, the majority of the recombinant eukaryotic proteins that have been expressed in bacteria are produced as fusion proteins and not in their native conformation. Despite advances in plasmid engineering, a plasmid system for high level expression of proteins with quarternary structure (that is, proteins formed by more than one subunit) in their natural form in a bacterial expression vector has not heretofore been described.

SUMMARY OF THE INVENTION It is, therefore, an object of the present invention to construct novel plasmids for high level expression of quarternary proteins in their native form, in a suitable expression vector.

It is another object of the present invention to provide plasmids containing artificial operons for the expression of dimeric or multimeric proteins in their native (natural) form and to provide expression vectors of general utility.

It is a further object of the present invention to provide high efficiency cDNA expression libraries for the screening of eukaryotic genes with antibody probes. Other objects and advantages of the present invention will become evident from the following detailed description of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS These and other objects, features and many of the attendant advantages of the invention will be better understood upon a reading of the following detailed description when considered in connection with the accom¬ panying drawings wherein: Figure 1A shows a schematic construction of pLElOl and pLE102.

Figure IB shows a schematic construction of pLE103-l from pLElOl. The sequence of pLE103-l between the BamHI site and the Hindlll site is reported above the scheme of the subcloning steps. Arrows indicate the limits of the synthetic 89-base pair BamHI-Nco I DNA fragment used for this construction. Only the restric¬ tion sites relevant for plasmid construction and the cloning sites are indicated. P_trc:trc promoter; 5S: 2_. coli rrnB 5S rRNA gene; T: Tl and T2 rrnB terminators; _>10: _10 phage T7 promoter; UG: mature UG coding sequence from pUG617; SD: Shine-Dalgarno sequence.

Figure 2A shows the expression of UG in E_^. coli BL21(DE3):pLE103-l as estimated by SDS-polyacrylaraide gel electrophoresis on a 15-25% gradient gel containing 0.1% SDS. Each lane was loaded with the equivalent of 50 _//l of bacterial culture. Lane 1: purified rabbit UG (1 iq ) ; lane 2: molecular weight standards (BRL, pre- stained); lanes 3-4: 1 hour after induction time, non- induced culture (3) and induced culture (4); lanes 5-6: 2 hours after induction time, non-induced culture (5) and induced culture (6) ; lanes 7-8: 3 hours after induction time, non-induced culture (7) and induced culture (8) .

The arrows indicate the position of the expressed bands: y-lactamase (uppermost arrow) and the two bands of UG. Aliquots of cultures (1 ml) in M9CA medium were centrifuged at 12000 x g for 2 min. Cells were washed with ice-cold PBS and centrifuged as above. Pellets were lysed in 50 ul of 2X sample buffer contain¬ ing 0.2% mercaptoethanol. Samples were boiled for 5 min. and adjusted to a final volume of 100 ul with distilled water. Aliquots (10 ul) were loaded on a 10-20% gradient polyacrylamide gel, 1.5 mm thick. Silver staining was performed by means of a Biorad kit, according to the manufacturer's instructions.

Figure 2B shows the immunoblot of expressed recombinant UG. Each lane of a 15-25% polycrylamide gradient gel containing 0.1 % SDS was loaded with the equivalent of 50 /ι l of bacterial culture, and protein bands were transfered overnight to a Nitroscreen West membrane (DuPont, 0.22 //m pore size) at 4°C, with a current of 34 mA. Lane 1: molecular weight standards (BRL, prestained); lanes 2-3: 1 hour after induction time , non-induced culture (2) and induced culture (3); lanes 4-5: 2 hours after induction time, non-induced culture (4) and induced culture (5); lanes 6-7: 3 hours after induction time, non-induced culture (6) and induced culture (7); lane 8: purified rabbit UG (250 ng) . Note that both UG monomer and dimer bands appear to be stained (arrow) .

Figures 3A and 3B show the quantitation of recombinant UG in bacterial lysate supernatants as ^• detected by RIA. Each point represents the average of three determinations, each performed in duplicate.

Figure 4 demonstrates the determination of recombinant UG molecular weight by means of size-exclu- sion chromatography on Sephacryl-S200 (Pharmacia). Sam¬ ples of pure rabbit UG (140 _^*g) and of bacterial lysate supernatant (400 l from bacteria harvested 90 min after induction) were analyzed. One ml fractions were collected, diluted 1:100 and assayed for UG by RIA. The inset shows the calibration of the column with standard proteins (gel-filtration calibration kit, Pharmacia, plus horse heart myoglobin, Sigma). Abbreviations: LYS = lysozyme; MYO = myoglobin; CHT = chymotrypsinogen A; OVA = ovalbumin; BSA = bovine serum albumin. _av values were calculated as described in Pharmacia technical booklet: Gel Filtration in Theory and Practice. Values on the y axis represent concentrations of UG in the diluted sam¬ ples, without adjustments for the dilution factor. The plot line was obtained by least-square analysis, followed by linear regression. R = 0.998. The arrow indicates the K_av of recombinant and natural UG. Apparent molecu- lar weight = 17 kd.

Figure 5 shows the results of SDS-polyacryl- amide gel electrophoresis of E_. coli proteins in IPTG induced and non-induced cultures. Samples were electro- phoresed using a 15 % polyacrylamide gel containing 0.1 % SDS. Lane a: molecular weight standards (BRL, low molecular weight standards); lane b: pure rabbit UG (4 g) non-induced culture (control), 3 hours after induc¬ tion time; lane c: bacterial lysate supernatant (about 12 g protein) induced culture, 3 hours after induction; lane d: pooled UG-containing fractions after Sephacryl- S200 superfine chromatography (about 20 μ. g protein) non- induced culture, 4 hours after induction; lane e: pooled fractions after CM-Sepharose chromatography (1.8 jug pro¬ tein) induced culture, 4 hours after induction; lane f: pooled fractions after Sephadex-G50 superfine chroma¬ tography (1.6 _u g protein) purified rabbit UG, silver stain.

Note that the bacterial lysate supernatant appears to be more concentrated than the Sephacryl pool. This is probably due to its hight content in nucleic acid fragments (A2 >5). Note also the presence of a very abundant band with an apparent molecular weight of about 25,000 in lane d. This band appears only after induction with IPTG and probably corresponds to over- produced P-lactamase.

Figure 6 shows the N-terminal sequence of recombinant UG. Figure 7 shows a schematic diagram of pLDl l, below which is shown the sequence of pLDlOl between the Bam HI site and the Hind III site.

Figure 8 shows the dose response of natural and recombinant UG as inhibitors of porcine pancreatic PLA₂. Each point represents the average of three determ¬ inations, each performed in duplicate.

DETAILED DESCRIPTION OF THE INVENTION

The above and various other objects and advan- tages of the present invention are achieved by a plasmid comprising circular, double stranded DNA of a molecular length of about 4406 base pairs containing at least a gene for ampicillin-resistance, 10 T7 promoter, ribo- somal binding site and three cloning sites, Ncol, PstI and Hindlll, wherein a cDNA for protein to be expressed is inserted at Ncol, PstI or Hindlll site. Preferred plasmids are selected from the group consisting of pLElOl, pLE102 and pLE103-l. The plasmid of the present invention is distinguishable from any other plasmid by the following features. pLE 101 and 102 contain UG cod¬ ing region under the control of "trc" promoter, whereas pLE 103-1 in addition contains (i) a T7 promoter; (ii) the T7 gene 10, 5'nontranslated region, the UG coding sequence and the cloning sites Ncol, PstI and Hind III. To our knowledge, correct intracellular formation of multimeric structures containing more than one interchain disulfide bridge has not been reported so far.

The three plasmids (pLElOl, pLE102 and pLE103- 1) are able to direct expression of recombinant rabbit uteroglobin (UG), a homodimeric protein with two inter¬ chain disulfide bridges, in _E.. coli. Among these, the plasmid pLE103-l, in which the expression of recombinant UG is controlled by a bacteriophage T7 late promoter, is by far the most efficient. With pLE103-l, recombinant UG production reached about 10% of total bacterial soluble proteins. This protein accumulated in bacterial cells in dimeric form, as it is naturally found in the rabbit uterus. Recombinant UG was purified to near-homogeneity and its N-terminal amino acid sequence was confirmed to be identical to that of its natural counterpart, except for 2 Ala residues the codons of which were added during the plasmid construction. This protein was found to be as active a phospholipase A₂ inhibitor as natural UG on a molar basis. The plasmid pLE103-l may be useful to explore the structure-function relationship of rabbit uteroglobin. In addition, this plasmid may be useful in obtaining high level bacterial expression of other eukaryotic proteins with quaternary structure, as well as for other general applications requiring efficient bac¬ terial expression of cDNAs.

Blastokinin or UG is a low molecular weight secretory protein which is found in several organs of the rabbit. The synthesis and secretion of UG are regulated by different steroid hormones in different organs. This protein has several biological properties, which include immunomodulatory effects, antiiflammatory properties and an inhibitory activity on platelet aggregation. UG is thought to play an immunomodulatory/antiinflammatory role in protecting the wet epithelia of organs which communi¬ cate with the external environment. In particular, UG has been proposed to protect the rabbit embryo from maternal immunological assault during implantation. A uteroglobin-like protein has been recently detected in the human uterus, respiratory tract and the prostate. At least some of the biological effects of UG may stem from the phospholipase A₂ (PLA₂, EC 3.1.1.4) inhibitory pro- perties of this protein. Because of its PLA₂ inhibitory effect, UG can prevent liberation of arachidonic acid from membrane phospholipids, which is the first step of the arachidonate cascade, leading to the synthesis of various eicosanoids, some of which are well known edia- tors of inflammation. UG is a homodimeric protein, formed by identical subunits of 70 amino acids each, joined in antiparallel orientation by two disulfide - 1 -

bridges.

A nonapeptide in d- -helix 3 of UG which may be the active site, or at least part of an active site, for the PLA₂ inhibitory activity of UG. As a prelude to confirming this observation by site-directed mutagenesis, a high level bacterial expression of this protein was obtained. However, the structure of the protein posed a unique problem, since to our knowledge bacterial expres¬ sion of multimeric eukaryotic proteins with two inter- chain disulfide bridges in their natural form has not been reported so far. Using plasmid pLE103-l, a high level expression of recombinant UG (about 9-11% of total bacterial soluble proteins) was obtained. In this plasmid, the transcription of UG cDNA is controlled by the _il0 late promoter of bacteriophage T7. Recombinant UG in its natural dimeric form is synthesized by J3.. coli cells harboring pLE103-l, with no apparent intracellular accumulation of free subunits. The recombinant protein was purified to near-homogeneity and its N-terminal amino acid sequence was found to be identical to that of its natural counterpart, except for 2 Ala residues the codons of which were added during the plasmid construction. Recombinant UG was found to have an identical PLA inhibitory activity as that of the natural protein. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of the pre¬ sent invention, the preferred methods and materials are now described. All publications mentioned hereunder are incorporated herein by reference. Unless mentioned otherwise, the techniques employed herein are standard methodologies well known to one of ordinary skill in the art.

The term "high level" or "high efficiency," as used herein, means about a 1000 fold or more synthesis of the protein by the plasmid system of the present inven¬ tion in an J5_. coli expression vector, compared to the ATG vector pKK-233-2. The term "artificial operon," as used herein, means a group of foreign genes, controlled by a common regulatory sequence, inserted into a plasmid for the expression of a quarternary protein.

The term "quarternary protein" as used herein means a protein formed by more than one subunit, such as heterodimeric, homodimeric or multimeric proteins.

The principles and methodologies of the present invention are now described.

The genetic material of most organisms consists of DNA. In order to synthesize proteins one helix of DNA is "transcribed" into a single stranded RNA (mRNA) by an enzyme (RNA polymerase). The mRNA is then "translated" into a protein by intracellular organelles (ribosomes) which synthesize a polypeptide chain (a chain of amino acids linked to each other by a peptide bond) according to the sequence specified by the sequence of codons (groups of three nucleotides) of the mRNA. The basic mechanisms of these processes are similar in all orga¬ nisms. However, in eukaryotic cells (cells with distinct nuclear envelope, such as cells of animal and plant origin, the protozoans and unicellular fungi) the process is much slower and more complex than in prokaryotic cells (e.g. bacteria and blue-green algae) . In prokaryotes transcription and translation are simultaneous and coupled, and both processes are much faster than in eukaryotes. Moreover, growing prokaryotes in a con¬ trolled laboratory environment is usually much cheaper than maintaining eukaryotic cells in culture, because of the simpler nutrients required and a much higher rate of growth in prokaryotes.

The Gram-negative bacterium E_. coli is by far the best characterized organism from the molecular and genetic point of view, and genetic manipulations in this organism are relatively easy. For these reasons, several systems have been developed to produce high quantities of eukaryotic proteins in Escherichia coli ( L coli) both for scientific and industrial applications (see for exam¬ ple Rosenberg et al, supra; Crow et al. Gene, 38, 31-38, 1985; Amann et al, supra; Mandecki et al. Gene, 43, 131- 138, 1986). These systems utilize plasmid vectors.

A plasmid is a circular double stranded DNA molecule which exists intracellularly (mostly in bacteria but also found in eukaryotes like the yeast) and repli¬ cates independently from the host chromosome. Biologic¬ ally, a plasmid does not "belong" to the host cell, but it is rather an endosymbiotic entity. Usually plasmids encode functions which are useful to the host cell. Typically, the plasmids used in experimental procedures contain genes which confer to the host resistance to one or more class of antibiotics. By growing plasmid-harbor- ing bacteria in antibiotic-containing medium, an inves- tigator can ensure that virtually every cell in the culture contains the plasmid which confers antibiotic- resistance to the host.

Since the massive production of a foreign pro¬ tein usually kills the bacterial host, or at least slows down its growth, most protein-expressing plasmids, con¬ structed so far, contain mechanisms of regulation. These plasmids are engineered to contain a very active promoter (i.e. a DNA region with very high affinity for RNA polym- erase) and a biochemical mechanism is incorporated which keeps the promoter from being activated until an exogen¬ ous "inducing" stimulus is given. The coding sequence for the foreign protein of interest (typically a eukary¬ otic cDNA) is inserted into the plasmid in a 3'-direction from the promoter. This allows the bacteria to be grown up to an appropriate density, and then induce the produc¬ tion of the foreign protein by starting the transcription of the plasmid promoter. An mRNA is produced from the plasmid DNA in a 3'-position of the promoter, up to the next transcriptional terminator site. At the same time, the mRNA, which includes the coding sequence for the foreign protein, is translated by bacterial ribosomes into a protein.

Demonstrated herein is the construction of a highly efficient novel plasmid for bacterial expression of an anti-inflammatory protein, uteroglobin (UG) . This protein (for review see Miele et al. Endocrine Rev. 8, 474-494, 1987) has several important biological proper¬ ties which include immunomodulatory effects, anti-inflam¬ matory properties and platelet-antiaggregating activity.

UG is a secretory homodimeric protein (i.e. formed by two identical monomers). It is significant to note that the dimeric nature of UG posed a peculiar prob¬ lem for bacterial expression, since no other proteins with quarternary structure (i.e. formed by more than one subunit) had heretofore been expressed in their natural form in bacteria. The present invention is the first successful demonstration of the synthesis of a quar¬ ternary protein in its natural form by a plasmid in a bacterial system ( L coli) . The novel plasmid directing the expression of UG is designated herein as pLEl03-l. METHODS OF CONSTRUCTION OF PLASMID VECTORS (pLE103-l, pLElOl AND pLE102) AND EXPRESSION OF UG IN E. COLI

The construction of pLE103-l is described in Figs. 1A and IB. Plasmid pUG617 was a gift from Dr. David Bullock (Lincoln College, Canterbury, New Zealand). Plasmid pKK233-2 was kindly provided by Dr. J. Brosius (Columbia University, NY). jE^. coli strain JM105 was purchased from Pharmacia and strain JM109 was a gift from Dr. J. Messing (University of Minnesota). E_. coli Strain BL21(DE3) was generously provided by Dr. W. Studier (Brookhave National Laboratory, NY) . "Library- efficient" competent J3. coli strain HBlOl was purchased from BRL. All recombinant DNA manipulations were per¬ formed according to standard techniques. Restriction enzymes, T4 DNA polymerase, T4 DNA ligase and E_. coli DNA polymerase I large fragment were obtained from either Pharmacia or BRL. All other reagents were ultra pure grade from BRL. First, a 430 bp fragment containing the entire

UG coding sequence was excised from pUG617 (Chandra et al, DNA 1:19-26, 1981) by means of digestion with the restriction enzyme PstI. Digestions with restriction endonucleases were performed according to the instruc- tions of the manufacturer (New England Biolabs, Pharmacia or Bethesda Research Laboratories). pUG617 was a generous gift from Dr. D. Bullock. This fragment does not contain the coding sequence for the UG "signal" pep- tide, i.e. the N-terminal fragment, 21 amino acids long, which is eliminated from mature UG in eucaryotic cells.

The 430 bp fragment was purified by agarose gel electrophoresis (Maniatis et al. In Molecular Cloning, a Laboratory Manual, Cold Spring Harbor Laboratory, 1982) and divided into two aliquots: one was kept intact, and the other was treated with phage T4 DNA polymerase (Maniatis et al, supra) . This treatment eliminated the protruding ends left by PstI digestion, producing a 422 bp blunt-ended fragment. The intact fragment was coval- ently joined by means of phage T4 DNA ligase to Pstl- digested pK 233-2 (Amann et al. Gene 40:183-190, 1985) purchased from Pharmacia. T4 DNA ligase and 5X T4 DNA ligase buffer were purchased from Bethesda Research Laboratories. The ligation reaction contained 0.1 units of DNA ligase, 4 ul of 5X ligase buffer, 100 ng of frag- ent and 50 ng of plasmid in a total volume of 20 ul. The reaction was carried out at 12°C for 6 hrs. The ligated DNA was diluted 1:5 with TE buffer (Maniatis et al, supra) and 5 ul were used to transform j^. coli strain JM105 (Yanisch-Perron et al. Gene 33, 103-119, 1985) and several recombinant clones were obtained. Transformation and plating of bacteria were carried out as described by Maniatis et al, supra. Small-scale preparations of plasmids of these clones were performed according to Birnboim et al (Nucleic Acid Res. 7, 1513-1523, 1979), and the correct orientation of the insert DNA was checked by digestion with appropriate restriction endonucleases (e.g. Aval and Hindlll). Clones containing the UG coding region in the appropriate orientation were designated pLElOl.

For the construction of pLE102, the Pst I- digested ends of the 430 bp DNA fragment were made blunt by treatment with bacteriophage T4 DNA polymerase. Plasmid pK 233-2 was digested with Nco I and the cohesive ends were made blunt by treatment with JE_. coli DNA poly¬ merase I "large fragment". Direct ligation of blunt- ended UG cDNA fragment into Nco I-digested, blunt-ended pK 233-2 generated pLE102. The orientation of the insert was checked by digestion with Ava I and Hindlll, and the expected reconstitution of two Nco I sites at both ends of the insert was verified.

The difference between pLElOl and pLE102 is that pLElOl three codons (Met-Ala-Ala) are added to the 5' terminus of the coding sequence of UG DNA, while in pLE102 only one codon (Met) is added to the 5' ter¬ minus. Therefore, pLElOl directs the production of a recombinant UG having three additional amino acids at the N-terminus (Met-Ala-Ala) and pLE102 directs the produc¬ tion of a recombinant UG having only one additional amino acid (Met) at the N-terminus. The presence of a Met codon at the 5' end of the coding sequence is indispens¬ able for translation of the mRNA in both prokaryotic and eukaryotic cells. In this case, the Met codon is present in the Ncol site of the vector.

The plasmid p 233-2 is an expression vector, i.e., a plasmid capable of producing foreign proteins in _E_. coli, provided that the coding sequence for such pro- teins is inserted in the appropriate restriction sites in the vector. The expression of foreign proteins in p K233-2 is controlled by the artificial "trc" promoter. a regulatory region consisting of part of the E_. coli "trp" promoter and part of the _E_. coli "lac" promoter. The "trc" promoter is regulated by the lactose operon repressor protein ("lac repressor"). When plasmids carrying the "trc" promoter are introduced into an _E. coli strain which overproduces the lac repressor (lad genotype), the expression of the foreign gene is sup¬ pressed. Upon addition of a gratuitous inducer (i.e. a molecule that binds to the lac repressor and inactivates it, without being metabolized) such as isopropyl-B-thio- galactopyranoside (IPTG), the "lac" repressor is inactiv¬ ated and the foreign gene is transcribed by the E_. coli RNA polymerase. The resulting mRNA is then translated to produce the recombinant protein (e.d. UG) . pLElOl and 102, being derivatives of pKK233-2 express the two forms of recombinant UG upon induction with IPTG. The two plasmids were tested for expression using _E_. coli strains JM105 and JM109 as hosts. Upon induction with IPTG both plasmids expressed recombinant UG detectable by RIA in the bacterial lysate super¬ natant. The best results were obtained with JM 109 when bacteria were grown in M9 medium supplemented with 0.001% thiamine and expression was induced with IPTG early dur¬ ing logarithmic growth, i.e. before the optical density of the culture at 600 nm (ODg_Qg) reached 0.4. Under these conditions, pLElOl expressed about 300 ng UG/ml supernatant and pLE102 expressed about 120 ng/ml.

Expression of recombinant UG was tested essen¬ tially as described by Amann and Brosius for pLElOl and 102 and by Rosenberg et al. for pLE103-l. Expression experiments were always performed with two identical cultures, one of which was induced with isopropyl -β -thiogalactopyranoside (IPTG, Calbiochem) early during logarithmic growth. Samples from the cultures were with- drawn before induction and at various times after induc¬ tion. Aliquots of the samples were centrifuged in an Eppendorf microcentrifuge and the bacterial pellets were lysed directly in SDS-polyacrylamide gel sample buffer for electrophoretic analysis. The remainder of the samples were centrifuged at 10,000 x g for 10 min. The bacterial pellets were washed with phosphate-buffered saline (PBS, Quality Biologicals) and recentrifuged at 10,000 x g for 10 min. The pellets were resuspended in buffer L, which consists of 50 mM Tris-HCl pH 8, 5 mM EDTA containing 4% glycerol, 250 iHL phenylmethylsulfonyl- fluoride (Sigma), 0.7 _/_g/ml pepstatin A (Calbiochem) and 0.5 jUg/ml leupeptin (Calbiochem). The samples were flash-frozen in liquid N₂, thawed on ice and sonicated for 20 sec (Heat Systems-Ultrasonics sonicator, setting 4, continuous) . Bacterial lysates were centrifuged at 30,000 x g and supernatants were transfered to clean polypropylene tubes while pellets were resuspended in buffer L. Aliquots from supernatants and from resuspended pellets were assayed for UG by RIA.

Purification of recombinant UG was accomplished as follows. Eight hundred ml of M9 medium (Quality Biologicals, Gaithersburg, MD) containing 200 g/ml ampicillin and supplemented with 0.001% thiamine were inoculated with 2 ml of a saturated culture of BL21(DE3) :pLE103-l grown in the same medium. UG expres¬ sion was induced early during logarithmic growth with IPTG at a final concentration of 0.45 mM. One hundred minutes after induction, the bacteria were harvested by centrifugation at 10,000 x g for 10 min. Bacterial pellets were flash-frozen in liquid N₂ in the centrifuge bottles, and thawed on ice. The pellets were resuspended in a total of 10 ml of ice-cold buffer L, and lysed by three cycles of sonication on ice (1 min each, setting 4.5, continuous) .

Recombinant UG was purified from E_. coli lysate by a modification of the original method published by Nieto et al. for rabbit UG. Briefly, the bacterial lysate was centrifuged at 30,000 x g for 30 min. The supernatant (^9 ml) was loaded on a column of Sephacryl- S200 superfine (Pharmacia), 2.5 x 100 cm, equilibrated in buffer L. The fractions were assayed for UG by RIA, and UG-containing fractions were pooled, dialyzed against distilled H₂0 and lyophilized. The lyophilized material was resuspended in 12 ml of 25 mM ammonium acetate buffer, pH 4.2 and centrifuged at 27,000 x g. The super¬ natant was loaded on a CM-Sepharose Fast Flow column (Pharmacia), bed volume 10 ml, equilibrated with 25 mM ammonium acetate, pH 4.2. The column was washed with 3 bed volumes of the same buffer, and eluted with a linear gradient made of 100 ml of 25 mM ammonium acetate, pH 4.2 and 100 ml of 120 mM ammonium acetate, pH 6.0. UG con¬ taining fractions (as determined by RIA) were pooled and lyophilized. The lyophilized material was resuspended in 2 ml of 10 mM ammonium bicarbonate buffer, pH 8.0, and loaded onto a column of Sephadex G50 superfine (1.5 x 70 cm) equilibrated in the same buffer. UG containing frac¬ tions were pooled and lyophilized. Recombinant UG was stored in lyophilized form at -70°C with dessicant. The concentration of purified recombinant and natural UG was estimated spectrophotometrically, using an ^M value of 1800, as published by Nieto et al. Protein concentration of complex mixtures was determined according to Bradford by means of a kit from Biorad. SDS-polyacrylamide gel electrophoresis was performed according to Laemmli and silver staining was performed by means of a kit from Biorad according to the manufacturer's instructions.

The N-terminal sequence of recombinant UG was determined by an ABI 477A gas phase sequenator following standard protocols for Edman degradation and analysis of phenylthiohydantoin derivatives. The sample (0.6 mg total) was divided into two aliquots. One aliquot was reduced with dithiothreitol under denaturing conditions and the Cys residues were pyridylethylated before sequence analysis. The other aliquot was processed with¬ out pretreatment.

Phospholipase A₂ assay was performed as previ- ously described, with modifications. Briefly, the reac¬ tion mixture contained 100 mM Tris-HCl, pH 8, 100 mM NaCl, 1 mM Na-deoxycholate, 10 _/_M 2-[l-¹ C]-arachidonyl phosphatidylcholine (Amersham, 58 mCi/mmole) and 2 nM porcine pancreatic PLA₂ (Sigma) in a total volume of 50 μl . PLA was preincubated with either recombinant or natural UG at 37°C for 5 min and the enzymatic reaction was started by addition of aliquots of the preincubation mixture to the radioactive substrate. Controls were kept in which PLA₂ was preincubated with buffer only. PIA₂ reaction was run at 37°C for 30 sec. and stopped by addi¬ tion of 50 M l of chloroform/methanol 2:1, followed by 50 ^l of chloroform and 50 i l of 4 M KC1. Radioactive arachidonic acid was separated from unhydrolyzed sub- strate by thin layer chromatography on silica plates (silica gel G, prechanneled, Analtech) . The eluent was petroleum ether/diethyl ether/acetic acid 70:30:1. Iodine-stained bands comigrating with the arachidonic acid standard were scraped and counted in a Beckman LS- 9000 liquid scintillation counter.

The maximum level of expression reached with both pLElOl and 102 was about 2 ng/ml of bacterial culture, measured by radioimmunoassay with a monospecific goat anti-UG antiserum. Radioimmunoassay (RIA) for UG was performed as previously described. Immunoblots ("Western" blots) were performed according to Burnette using Nitroscreen West membranes (New England Nuclear) . Blots were stained with a goat-anti UG antibody and a rabbit anti-goat Immunogold-Silver Staining (IGSS) kit (Janssen). This maximum level was reached in host strain JM109 (Yanisch-Perron et al, supra) in Luria-Bertani (LB) broth (Maniatis et al, supra) . These results suggested that immunoreactive UG can be produced in J3_. coli, and it is not toxic to the bacterial host at the concentration reached. The presence of 1 instead of 3 residues at the N-terminus of the recombinant protein did not modify the level of expression. Moreover, the level of expression was not affected by the use of different culture media, such as M9 minimal medium, supplemented with Casamino acids (Difco) or with a mixture of 20 pure amino acids.

The recombinant proteins produced by pLElOl and 102 were soluble, being recovered almost totally in the supernatant after centriguation of the bacterial lysate at 30,000 x g. The relatively low level of expression obtained by these plasmids is most probably explained by the peculiar problem posed by the expression of a dimeric protein in E_. coli. UG is a dimeric protein, being formed by two identical subunits joined by two disulfide bridges. Without being bound to any specific theory, it is postulated that the intracellular concentration of UG reached in _E. coli depended on the following variables: i) rate of transcription of the UG gene; ii) rate of translation of the transcribed mRNA; iii) rate of degrad¬ ation of the intracellular protein; and iv) the equi¬ librium of dimerization of the intracellular UG monomers. The dimerization of UG monomers can be described as UGm + UGm —> UG, where UGm is UG monomer. The equilibrium constant of this process is K_orr ^~- [UG]/[UGm] . Therefore [UG], i.e. the intracellular molar concentration of UG, is = _Q-_JX [UGm]². In other words, the amount of UG being produced depends on the second power of the intracellular concentration of UGm. If it is assumed that the isolated UG monomer, being much more unstable in solution than dimeric UG, has a much shorter half-life in the E_. coli, the dimerization of isolated monomers becomes a rate limiting step in the expression of recombinant UG.

As has been mentioned supra, dimerization is a second-order process, depending on the squared concen¬ tration of UGm. In practice, this means that in order to obtain efficient production of UG, the intracellular level of UGm must be kept high enough to allow efficient dimerization of UGm, because a small decrease in [UGm] would cause a dramatic decrease in [UG] . This can be obtained by maximizing the rate of transcription and translation of UGm.

To accomplish this, the regulatory sequences of pLElOl were replaced with synthetic regulatory sequences identical to those of bacteriophage T7. Such promoters have been previously shown to direct high level expres¬ sion of recombinant proteins in E_. coli. Fig. lb shows the construction of pLE103-l from pLElOl. In pLE103-l the regulatory sequences originally present in p K233-2 (lac operator, trc promoter and ribosome binding site) have been substituted with the synthetic DNA fragment whose sequence is shown in Figure IB. The synthetic DNA fragment contained the rilO late promoter of bacteriophage T7, the 5' non-translated region of T7 gene 10 and the ribosome binding site from the same gene. The sequence of this synthetic regulatory region was derived from the wild-type sequence which has been used by Studier and coworkers in their "translation" vectors, with the excep¬ tion that the 2 bases preceding the initiation ATG triplet were substituted with two cytosines. This sub¬ stitution was made in order to create an Nco I site including the initiation triplet. Additionally, a BamHI site was added at the 5' end. The plasmid obtained in this way has the same cloning sites, and should have the same possible applications, of pKK233-2 and related "ATG vectors", except that expression of the recombinant pro¬ tein is controlled by a more specific and very efficient viral promoter.

The regulatory sequences in pLElOl are derived from p K233-3 and consist of the "trc" promoter followed by an artificial "ribosome binding site" (RBS). The latter sequence is thought to control the association of a bacterial mRNA to the 3OS subunit of ribosomes, thereby regulating the rate of translation of the mRNA. The regulatory sequences in pLElOl (as in pK 233-2) are con¬ tained in a 285 bp BamHI-Ncol fragment. This fragment was excised from pLElOl (Fig. IB) by means of digestion with BamHI and Ncol restriction endonucleases. The remaining portion of the plasmid was purified by electro¬ phoresis on an agarose gel and covalently joined to a completely synthetic 89 bp BamHI-Ncol DNA fragment. The latter reproduced the sequence of the 10 promoter of phage T7 (Studier et al, J. Mol. Biol. 189:113-130, 1986; Rosenberg et al. Gene 56:125-135, 1987), followed by the 5' nontranslated region, including the RBS, of phage T7 gene 10. The synthetic regulatory region ends with an initiation codon (ATG) contained in the Ncol site. The recombinant plasmid obtained in this way was designated pLE103-l, and it contains the same cloning sites as pKK233-2, but now under the control of the synthetic T7 "gene 10-like" regulatory region. Phage T7 promoters are usually not transcribed by JE_. coli RNA polymerase, but only by T7 RNA polymerase, which is highly specific for T7 promoters and does not recognize E_. coli promoters. Therefore, unless an active T7 RNA polymerase is delivered into the bacterial host, the basal level of transcription of a foreign gene under the control of a T7 promoter is negligible. However, T7 RNA polymerase initiates transcription with very high efficiency and it elongates RNA 5-times faster than j^. coli polymerase. T7 RNA polymerase produces from plasmids such as pLE103-l longer transcripts than _E_. coli RNA polymerase, which stops at the rrnB terminators present in pKK233-2 (Fig. 1). The length of the RNAs transcribed by T7 RNA polymerase has been suggested to protect them from intracellular exonucleolytic degrada- tion starting from the 3' end, thereby increasing the half-life of these RNAs in E_. coli.

This means that when an active T7 RNA poly¬ merase and a T7 promoter are concomitantly present in an _E_. coli cell, transcription from the viral promoter will successfully inhibit transcription from the host promo¬ ters, by limiting the availability of nucleotide triphos- phates for transcription of E_. coli genes. When the T7 promoter controls the expression of a foreign gene, this generally leads to a very high accumulation of expressed protein, and eventually to the death of the bacterial host. An expression vector carrying a T7 promoter, pro- vided that it is stable in E_. coli, usually leads to a high expression of proteins which are not toxic for the bacterial host.

Possible ways of delivering an active T7 RNA polymerase into the bacterial host are: (a) by infection with T7 phage; (b) by infection with a recombinant lambda phage containing the gene of T7 RNA polymerase by use of a bacterial host such as BL21(DE3). This strain of E_. coli is lysogenic for a lambda phage containing the T7 gene for RNA polymerase. In other words, the chromosome of this strain of E_. coli contains the entire genome of a lambda phage which in turn has been engineered to contain the gene for T7 RNA polymerase, cloned under the control of a "lacUV" E_. coli promoter, also artificially inserted into the lambda genome. The "lacUV" promoter, like the "lac" promoter, is regulated by a repressor protein which is inactivated by IPTG.

When BL21(DE3) is exposed to IPTG, it produces T7 RNA polymerase. If at the same time the strain con¬ tains a plasmid carrying an active T7 promoter, the T7 RNA polymerase produced transcribes any gene cloned down¬ stream, i.e. in 3' direction, with respect to the T7 promoter. This is a very convenient system to induce the expression of any nontoxic protein from plasmids carrying T7 promoters. When applicable, the use of BL21(DE3) is preferable to the other methods of induction, because infection with T7 causes competition of viral promoters with the vector promoter, and infection with a recom¬ binant lambda phage causes lysis of the bacteria. Expression of toxic proteins in BL21(DE3) needs addi- tional measures, such as the introduction in the host of a T7 lysozyme gene, whose product inhibits the basal levels of T7 RNA polymerase produced by the lysogenic lambda. This provides a mechanism to keep the expression of a toxic protein completely repressed, so that not even trace amounts of it are produced until the moment of induction. Alternatively, methods (a) and (b) vide supra, can be used to induce the production of a toxic protein in JE_^. coli.

In accordance with the present invention, plas¬ mid pLE103-l was used in BL21(DE3) to express UG, because it was found that low levels of UG are not toxic for _E_. coli as demonstrated by the data obtained with pLelOl and 102. In this strain, phage T7 RNA polymerase is produced upon induction with IPTG from a recombinant lambda phage which is integrated into the bacterial chromosome. BL21(DE3) :pLE103-l expresses recombinant UG upon induc- tion with IPTG, and the recombinant protein is readily detectable by polyacrylamide gel electrophoresis.

Fig. 2A clearly shows the time-dependent appearance in induced bacteria of a protein band of apparent molecular weight corresponding to that of mature rabbit UG monomer. The difference in molecular weight due to the presence of the expected additional three residues in the recombinant protein is not apparent under these conditions. The lower molecular weight band appearing immediately below the putative recombinant UG band (Fig. 2A) may be a product of partial degradation of recombinant UG, or an artifact caused by the formation of an intramolecular disulfide bridge in UG during SDS- polyacrylamide gel electrophoresis. The appearance of a pure UG as a "doublet" band due to such an artifact has been described by Nieto et al. The identity of the recombinant protein was confirmed by Western blot. That the new band appearing upon induction is recognized by anti-UG antibody is shown in Fig. 2B. Interestingly, both in the control sample of rabbit UG and in the recom- binant material an immunoreactive band with apparent molecular weight of about 13,000 is present (Fig. 2B) . The presence of this band is due to incomplete reduction of disulfide bonds between UG subunits under the condi¬ tions used for sample preparation and it consistently appears when concentrated samples of pure rabbit UG are subjected to SDS-polyacrylamide gel electrophoresis followed by immunoblot. This observation gave us a pre¬ liminary indication that the recombinant protein may be present in the bacteria in dimeric form.

The optimal expression of recombinant UG with pLE103-l was obtained with a high concentration of ampi- cillin in the culture medium (200 g/ml) and a very small initial inoculum (aliquots from a saturated culture were diluted 400-fold in fresh medium) . These conditions were optimal also for pLElOl and pLE102. It is well known that saturated cultures of bacteria which harbor plasmids derived from pBR322 (such as the vectors described in this paper) contain large amounts of -lactamase. There¬ fore, unless the plasmid is extremely stable in the host, it is essential to maintain a high selective pressure in order to avoid any growth of plasmid-free bacteria. Moreover, pLE103-l contains the bla gene in trascrip- tional orientation with respect to the T7 promoter.

Thus, induction with IPTG will result in transcription of a polycistronic mRNA containing the bla coding sequence, and in overexpression of β> -lactamase. In fact, T7 RNA polymerase does not recognize E_. coli trancriptional terminators, such as the Tl and T2 rrnB terminators present in pLE103-l (see Fig. IB). Under appropriate electrophoretic conditions, we have indeed observed an overexpression of a protein band of molecular weight corresponding to that of -lactamase (data not shown) .

Table 1 shows the results of an expression experiment using pLE103-l. The production of UG, as measured by radioimmunoassay in supernatants of bacterial lysates, reached 1.9 ug/ml of bacterial culture, or 7.7 ug/mg protein, i.e. about 1000-fold more than the levels obtained with pLElOl and 102, 3 hours after induction. This amount of protein synthesis is within the range of preparative scale production, corresponding to approxi¬ mately 670 ug/g bacteria. Thus, using pLE103-l a much higher level of expression of UG in BL21(DE3) was obtained compared to pLElOl or 102 in JM109. Fig. 3 shows the quantitation of recombinant

UG, as determined by RIA in supernatants from bacterial lysates. The three plasmids pLElOl, pLE102 and pLE103-l were compared under identical experimental conditions, except for the host strains, i.e. JM109 for pLElOl and pLE102, and BL21(DE3) for pLE103-l. It is clear that with pLEl03-l production of recombinant immunoreactive UG is much higher (about 50-fold more than with pLElOl and 100-fold higher than with pLE102). With pLE103-l, the highest absolute concentration of recombinant immunoreac- tive material was reached 120 min after induction (Fig. 3A) . However, when the amount of recombinant UG was expressed as jug/mg protein, the maximum level of UG was reached 90 min after induction (Fig. 3B) . More than 99% of immunoreactive UG expressed was recovered in the supernatant after centrifugation at 30,000 x g. This indicates that the recombinant protein is soluble.

The molecular weight of the recombinant protein obtained in accordance with the present invention was then determined both by polyacrylamide gel electro- phoresis under denaturing conditions and by size exclu¬ sion chromatography under non-denaturing conditions.

Fig. 4 shows the determination of molecular weight of recombinant UG by size exclusion chromatography under nondenaturing conditions. It is evident that recombinant and natural UG have an identical chromato- graphic behaviour. Under these conditions, both proteins have an apparent molecular weight of 17,000, slightly higher than the theoretical value of 15,800. This is in agreement with the results of Nieto et al. on purified rabbit UG. No immunoreactive peak indicating the presence of isolated UG subunits was observed, although UG subunits are readily recognized by our antibody in Western blots. These results seem to indicate that in lysates of induced BL21(DE3) :pLE103-l recombinant UG exists almost solely in its natural dimeric form.

Fig. 4 shows that the recombinant protein pro- duced in E_. coli has the same apparent w (17 kd) as that of natural UG purified from the rabbit uterus. This mw is slightly higher than the calculated value of 15.8 kd. Again, this is in agreement with former data on the chromatographic behavior of native dimeric UG (Nieto et al, supra) . The three additional amino acids present in recombinant UG do not affect the chromatographhic proper¬ ties of the protein enough to alter its apparent mw. No peak in a position corresponding to UG monomer was evident in the chrom togram of bacterial extract. These results indicate that: (i) abundant quantities of dimeric UG can be produced in E_. doli by pLE103-l; (ii) the yield of UG expression obtained with pLE103-l is about three orders of magnitude higher than those obtained with pLElOl and 102; (iii) virtually all the recombinant protein produced by pLE103-l is in its natural dimeric form. This indicates that dimeric pro¬ teins can be efficiently expressed in coli under the control of a T7 promoter, to a much higher efficiency than with vectors based on prior art modified E_. coli promoters. This is then the first time that a eukaryotic protein with quarternary structure has been efficiently expressed in its natural form in prokaryotes.

Gel electrophoresis (Fig. 5) showed that upon induction with IPTG (final concentration 0.4 mM) two new protein bands appear with apparent molecular weight (mw) between 5000 and 6000. This corresponds to the electro¬ phoretic behavior of the natural UG monomer under these conditions (Nieto et al. Arch. Biochem. Biophvs 180, 82- 92, 1977). It is postulated that the reducing conditions of electrophoresis destroy the interchain disulfide bridges of dimeric UG, showing only the monomer. It is known that monomeric UG gives electrophoretic artifacts. appearing as a double band (Nieto et al, supra) . This is presumably due to formation of intrachain disulfide bonds during electrophoresis. One additional band of higher apparent mw is also induced (Fig. 5). This corresponds to the apparent mw of β-lactamase (bla gene product), whose gene lies downstream of the UG gene in pLE103-l. This indicates that transcription from the T7 promoter proceeds beyond the UG gene through the bla gene. This is expected, since T7 RNA polymerase does not stop tran- scription a E_. coli transcriptional terminators, and even natural T7 terminators have a relatively low efficiency.

Despite the higher complexity of the protein mixture present in E_. coli lysates compared to rabbit uterine flushings, the very large headstart allowed us to obtain near-homogeneous recombinant UG by a modification of the original procedure (Figure 5). The chromato- graphic properties of recombinant UG in the columns used for purification were indistinguishable from those of the natural protein. The final yield of the purification was 3.2 mg recombinant UG from 800 ml of induced bacteria, as estimated by UV absorption using the published value of 1800 for the £^M. The starting material (bacterial lysate supernatant) contained a total of about 50 mg protein, and approximately 5 mg of recombinant UG (data not shown). Therefore, the final recovery of recombinant UG can be estimated as about 64%.

The N-terminal amino acid sequence of reduced and non-reduced recombinant UG is shown in Figure 6. Sequence a) represents non-reduced recombinant UG. In this sequence, Cys 3 was detectable only after in situ reduction and pyridylethylation in the sequenator cartridge. Sequence b) represents a sample which was reduced with a 100-fold molar excess of dithiotreitol in 8 M urea at 56°C for 1 hour and Cys residues were pyridylethylated prior to Edman degradation. Unnumbered residues were added to the N-terminus as a consequence of plasmid construction. Numbers indicate positions in the sequence of rabbit UG.

The recombinant UG produced by BL2(DE3) :pLE103- 1 is totally dimeric and no accumulation of free subunits could be detected. This might indicate that i) the rate of association of the subunits is very high and/or ii) free subunits are highly unstable and rapidly degraded in the bacterial host. To our knowledge, this is the first report of high level bacterial expression of a full- length dimeric eukaryotic protein with two interchain disulfide bridges in its natural quaternary structure. Our results demonstrate that a recombinant protein can form correct quaternary structures during overexpression in E_. coli even when correct formation of two interchain disulfide bridges is essential for its structure, pro- vided that the rate of intracellular accumulation of subunits and the rate of association of free subunits are high enough.

Non-covalent self-association of recombinant eukaryotic proteins in E_. coli has been described for human tumor necrosis factor and rat liver aldehyde dehydrogenase. However, in the first case most of the recombinant protein appeared in the insoluble fraction due to incorrect folding and in the second case the high efficiency of expression was suggested to be due to unique features of the 5' non-translated region of the cDNA (which contained a potential prokaryotic Shine- Dalgarno sequence) and its relationship with the lac promoter present in pUC8.

In contrast, pLE103-l, the 5'-non translated region and the Shine-Dalgarno sequence are built in the vector, so that theoretically any open reading-frame could be expressed in place of the UG cDNA. Until recently, it was generally believed that the intracellu¬ lar environment of _E_. coli is not conducive to the for a- tion of quaternary structures which require formation of interchain disulfide bridges. Correct folding and assem¬ bly of heterodimeric fragments of immunoglobulins after proteolytic processing of fusion precursors and secretion into the periplasmic space of E_. coli has been recently described. In one case the assembly involved the forma¬ tion of one interchain disulfide bridge. Our findings confirm and extend this observation further, demonstrat¬ ing that: i) quaternary structures containing more than one interchain disulfide bridge can also be properly assembled in E_. coli and ii) at least in the case of UG, correct assembly of quaternary structure can take place in the bacterial cytoplasm without the need for correct proteolytic processing of a precursor protein and trans- membrane transport of the product. It should be noted that natural UG is a secretory protein which is synthe¬ sized as a precursor that naturally undergoes transmem- brane transport and proteolytic processing.

Our data indicate that assembly of quarternary structure in E_. coli does not necessarily require the construction of a fusion precursor protein with a bacter¬ ial secretion signal sequence. Such constructions can be advantageous if secretion of correctly processed recombi¬ nant protein in the medium is achieved, but require a precise "in frame" fusion between the bacterial signal sequence and the eukaryotic coding sequence. This may require extensive manipulations on vector and/or insert DNA.

All in all, our observations seem to suggest that if a sufficiently high level of intracellular accumulation of recombinant protein(s) is obtained, the possibility of formation of multimeric structures involv- ing disulfide bridges depends essentially on the phys¬ ical-chemical factors controlling the folding of the protein(s) and the interaction between subunits. Thus, it may be possible to obtain efficient bacterial expres¬ sion of eukaryotic multimeric proteins other than UG, provided that: i) the vector/host system used insures a high efficiency of expression and intracellular accumula¬ tion of the product(s); and ii) the tertiary and quarter- nary structure of the recombinant protein(s) are thermo- dynamically stable and the kinetics of folding and assembly are not too slow. The latter conditions obviously depend on the particular protein(s) being expressed.

The results of Edman degradation experiments on recombinant UG, together with the chromatographic and electrophoretic data (see above) strongly support the hypothesis that the dimeric structure of recombinant UG is stabilized by two disulfide bridges identical to those of natural UG. In theory, it is possible that recombi¬ nant UG could form "inverted" dimerε in which the two disulfide bridges join Cys 3 and 3'; 69 and 69'. How¬ ever, this possibility is made unlikely by the fact that besides the two disulfide bridges, several other stereo- specific intermolecular contacts (Van der Waals interac¬ tions and H-bonds) , contribute to the stabilization of the UG dimer. Moreover, the identical properties of recombinant and natural UG as PLA₂ inhibitors further support the hypothesis that the two proteins are struc¬ turally identical, with the exception of the two addi¬ tional Ala residues in recombinant UG.

With the vector/host system used in this study, considerable overexpression of -^ -lactamase along with UG was observed. Simultaneous overexpression of a recombi¬ nant protein and j_> -lactamse has also been described with other vectors based on T7 promoters. Therefore, these systems are able to support overexpression of two dif¬ ferent polypeptides and could be used for the construc- tion of artificial operons to express heterodimeric proteins. The vector pLE103-l could be a useful addition to the already existing expression plasmids based on T7 promoters. In fact, after excision of the UG coding sequence from the Pst I site, pLE103-l can be converted into a general purpose expression vector, which we have denominated pLDlOl. The cloning sites Nco I, Pst I and Hindlll give to pLE103-l the same potential applications of "ATG vectors" with the advantage of the high efficiency and specificity of the T7 promoter. In particular, the Nco I site (CCATGG) is frequently present in eukaryotic translational starts. Moreover, after restriction endonuclease diges¬ tion the Nco I site can be easily "filled in" with _E_. coli DNA polymerase I large fragment. This process reconstitutes the ATG triplet, thereby allowing "blunt"- ended DNAs to be cloned in-frame directly into the "filled-in" Ncol site. In addition, the presence of the Pst I and Hindlll sites allows "forced" or "directional" cloning. Finally, the presence of the Pst I site, and the relative positions of the three cloning sites allow this plasmid to be used for the construction of cDNA libraries by several different methods.

The data obtained on the biochemical properties of recombinant UG have already yielded some valuable information on the structure/function relationship of this protein. Since the recombinant protein appears to fold correctly in E_. coli, the presence of the "leader peptide" which is physiologically present in rabbit preUG is not necessary for the folding of UG during transla¬ tion. Furthermore, the addition of two Ala residues at the N-terminus of recombinant UG does not affect its activity as a PLA₂ inhibitor.

Of course, as demonstrated for UG, similar procedure can be applied for the expression in E_. coli of any homodimeric, heterodimeric or multimeric protein by inserting the appropriate coding sequences for such pro- teins in place of UG coding sequence. It is noted that many proteins of medical importance fall in this group, including human hormones such as insulin, thyroid stimu¬ lating hormone, the gonadotropic hormones FSH and LH, human chorionic gonadotropin and the like. All these proteins are dimeric consisting of two relatively small subunits, whose genes can be assembled from totally syn¬ thetic DNA fragments and then inserted in tandem into an appropriate expression vector carrying the T7 promoter, such as pLE103-l.

As shown by the results presented herein, this vector directs the synthesis of both subunits, which then associate within the bacterial cell to produce the com¬ plete protein. The overexpression of both UG and ^*P_> - lactamase by pLEI103-l shows that T7 RNA polymerase can easily transcribe adjacent genes into polycistronic mRNAs (i.e. mRNAs containing more than one gene) and thereby direct the expression of two different polypeptide chains. This, of course, allows the construction of artificial "operons" (i.e. groups of adjacent genes con¬ trolled by a common regulatory sequence) in T7 expression vectors for the expression of dimeric or multimeric pro- teins in _E_. coli.

II Immunoglobulins (antibodies) are a class of multimeric proteins of enormous biological and medical interest that could be expressed in JS_. coli using the system of the present invention. This is accomplished by cloning cDNAs for an immunoglobulin light chain and heavy chain into the T7 vector of the present invention with a synthetic "spacer" sequence containing a prokaryotic RBS. Furthermore, genetic manipulations in plasmids could then allow the construction of mutant antibodies of altered antigen specificity, both for practical uses and for detailed studies of the molecular basis of antibody specificity.

Excision of the UG gene from pLE103-l by means of digestion with PstI, followed by ligation of the plas- mid, generated pLDlOl (Fig. 7). pLDlOl is a novel T7 expression vector with three cloning sites: Ncol, PstI and Hindlll. The complete sequence of the synthetic regulatory element containing the T7 «510 promoter and the gene 10 leader region and RBS is also shown in Fig. 7. This figure shows the main features of pLDlOl. It is a circular double stranded DNA with a molecular length of 4406 base pairs (bp) . It carries the gene for ampicil- lin-resistance (Amp^r) and an incomplete remnant of the tetracycline resistance gene (Tet^s). The _>10 T7 promoter is located between these two genes, followed by the T7 gene 10 non-translated region and RBS. This whole regu- latory region, shown in Fig. 7, is 89 bp long and it is followed by the three cloning sites Ncol, PstI and Hindlll. The nucleotide sequence of the regulatory region and the cloning sites is shown in the lower part of Fig. 7. The 5S rRNA gene (5S) and the Tl and T2 _E_. coli rRNA terminators (T) are from pKK233-2.

This recombinant plasmid pLDlOl provides the advantages of both T7-promoter vectors and of ATG vectors in a single system.

It may be pointed out that the pre-existing translation vectors using the T7 promoter have one restriction site (Ndel) into which foreign genes can be inserted. Another site (BamHI) is placed downstream, but insertion of a coding sequence in this site results in expression of a fusion protein containing 14 additional amino acids. The Ndel site contains the start codon ATG. However, use of a Ndel site presents some distinct disadvantages: (i) Ndel is a rare site on eukaryotic DNA; (ii) the restriction enzyme Ndel is a very unstable enzyme (half life of 15 min at 37°C, according to manu- facturer specifications) and does not work well unless the substrate DNA is thoroughly purified (see New England Biolabs catalog, 1987); (iii) after restriction cut, Ndel leaves a 2 bp overhang (AT) which is more difficult to ligate than 4 bp overhangs (in fact, ligation of Ndel ends relies on the formation of an AT-TA base pairing, which is held together by only 4 hydrogen bonds, and is rather unstable); (iv) when Ndel ends are made "blunt" by treatment with the "Klenow" fragment of DNA polymerase, the ATG is not reconstituted. This prevents blunt-ended fragments to be attached directly "in frame" to a "blunted" Ndel end reconstructing the start codon ATG. Finally, in some cases the presence of a relatively long extraneous peptide at the N- terminus of an expressed protein is not acceptable. This limits the applications of the downstream BamHI site to cases where the addi¬ tional 14 residues do not interfere with studied struc- tural or functional features ofthe expressed protein.

In contrast, the Ncol site present in the novel construct of the present invention allows a greater ver¬ satility of cloning, for the following reasons: (i) the Ncol cite (CCATGG) is frequently present in eudaryotic translational starts. The consensus sequence derived from 211 mRNAs is in fact CC,CCATG(G) (Kosak, Nucleic Acid Res. 12, 857-872, 1984). This means that many eudaryotic cDNAs can be directly cloned in the NCol site, producing an unfused recombinant protein; (ii) Ncol is a stable enzyme, and it also works properly on partially purified DNA< such as plasmid "minipreps;" (iii) after cut, Ncol leaves a 4 bp overhang, which allows easier ligations; (iv) when a cut Ncol site is "filled in" with Klenow fragment, the ATG is reconstituted. This allows "Blunt"-ended DNAs to be cloned in-frame directly by attaching them to a "filled-in" Ncol site. This situa¬ tion also results in production of an unfused recombinant protein, provided that the inserted "blunt-ended" frag¬ ment is in the correct reading frame (see, for example, construction of pLE102).

The presence of the PstI and Hindlll sites allows "forced" or "directional" cloning (i.e. using an insert with an Ncol or blunt end and a PstI orHindlll end, so that it can be inserted into the plasmid only in one orientation) . Cloning inserts into the PstI or Hindlll results in the insertion of only two (PstI), or three (Hindlll) alanine residues after N-terminal methionine (see for example construction of pLElOl and pLel03-l) . Finally, the presence of the PstI site, and the relative positions of the three cloning sotes om ATG vectors and in pLDlOl allow use of these plasmids for cDNA construction by different methods, such as the homo- polymeric tailing methods of Land et al (Nucleic Acid Res. 9, 2251-2266, 1981); Okayama et al (Mol. Cell. Biol. 2, 161-170, 1982); Heidecker et al (Nucleic Acid Res. 11, 4891-4906, 1983) and the double-linker method of Helfman et al (Proc. Natl. Acad. Sci. USA 80, 31-35, 1983). Moreover, constructing cDNA libraries in pLDlOl provides the benefit of high efficiency expression libraries, particularly useful for antibody screening. Fig. 8 shows does-response curves of recombi¬ nant and natural UG as PLA₂ inhibitors. Both proteins were tested in the range of concentrations which have been reported to be optimal for the PLA₂ inhibitory activity. It is evident that the two curves are essen- tially identical. This indicates that purified recombi¬ nant UG is as potent a PLA₂ inhibitor as the natural protein. The slightly lower percent inhibition obtained in the present study with UG, with respect to previously published data is most likely due to differences in the assay procedure, particularly the change in PLA₂ source and batch. With the batch of PLA currently used in our laboratory, inhibition observed with UG and other poly¬ peptide inhibitors rarely exceeds 40%. This may be due to contaminant protease(s) present in some batches of PLA2 ^{tnat ma}y cause degradation of these inhibitors.

In summary, with respect to previously avail¬ able ATG vectors, the plasmid constructs of the present invneiton provide much higher efficiency of expression; and with respect to previously available T7 promoter vectors, the plasmids of the present invention provide (i) the Ncol site; (ii) the PstI site; (iii) the Hindlll site; (iv) the feasibility of direct or directional clon¬ ing of blunt-ended cDNAs while conserving the ATG; and (v) the feasibility of direct construction of expression libraries.

Of course, it should be clear to one of ordi¬ nary skill in the art that the plasmids constructed in accordance with the present invention can be utilized for expression in any strain of E_^. coli of a suitable geno¬ type. Screening of cDNA expression libraries with anti¬ body can be performed by any standard methodology well known in the art and described in such texts as Davis et al, 1986, Basic Methods in Molecular Biology, Elsevier publication.

A deposit of pLElOl, 102 and 103-1 has been made at the American Type Culture Collection (ATCC), 12301 Parklawn Drive, Rockville, Maryland 20852, U.S.A. on October 5, 1988 under accession numbers 67815, 67816 and 67817, respectively. These deposits shall be viably maintained, replacing if they became non-viable, for a period of 30 years from the date of the deposit, or for 5 years from the last date of request for a sample of the deposits, whichever is longer, and made available to the public without restriction in accordance with the provi¬ sions of the law. The Commissioner of Patents and Trade¬ marks, upon request, shall have access to the deposits. It is understood that the examples and embodi¬ ments described herein are for illustrative purposes only and that various modifications or changes in light there¬ of will be suggessted to persons skilled in the art and are to be includes within the spirit and purview of this application and scope of the appended claims.

EXPRESSION OF RECOMBINANT UG IN E. COLI

Time after Induct ion, UG ( fg/ml)* hrs non-induced

10 * μg UG per ml of bacterial culture.

Wet weight. Approximate weight of cells was calculated from ODg_nn assuming 1 OD/ml = 8 x 10 cells/ml and weight of a single cell = 2 pg.

TABLE 1

Claims

WHAT IS CLAIMED IS:

1. A plasmid, comprising circular, double stranded DNA of a molecular length of about 4406 base pairs containing at least a gene for ampicillin-resis- tance, 10 T7 promoter, ribosomal binding site and three cloning sites, Ncol, PstI and Hindlll, wherein a cDNA for protein to be expressed is inserted at Ncol, PstI or Hindlll site.

2. The plasmid of claim 1 selected from the group consisting of pLElOl, PLE102 and PLE103-1.

3. The plasmid of claim 2 being PLElOl.

4. The plasmid of claim 2 being PLE102.

5. The plasmid of claim 2 being pLE103-l.

6. The plasmid of claim 1 inserted in a bac- terial host.

7. The plasmid of claim 6 wherein said expres¬ sion vector is L coli.

8. The plasmid of claim 1 wherein the cDNA is for a eukaryotic quarternary protein.

9. The plasmid of claim 1 wherein said quar¬ ternary protein is uteroglobin, immunoglobulin or a hor¬ mone.

10. The plasmid of claim 9 wherein said protein is uteroglobin.

11. A method for high level production of an eukaryotic protein, comprising culturing in a growth medium a bacterial host containing the plasmid of claim 1 having cDNA for a protein desired to be expressed in natural form at a high efficiency; and then recovering the desired protein from the growth medium.

12. The method of claim 11 wherein said protein is uteroglobin, immunoglobulin or a hormone.

13. The method of claim 12 wherein said protein is uteroglobin.

14. The method of claim 12 wherein said protein is an immunoglobulin.

15. The method of claim 12 wherein said protein is a hormone .

16. A method for constructing a general utility expression vector from pLE103-l, comprising digesting pLE103-l with PstI, eliminating UG insert from pLE103-l, religating 4406 base pair vector to reconstitute PstI site and then religating to produce pLDlOl.

17. A cDNA library constructed by employing pLE103-l.