An Alternate Method for Synthesis of Double-stranded DNA Segments*

of interferon I-mediated synthesis of oligonucleotide substrates having short stretches of comple- mentary sequence at their 3’ termini. the presence of DNA polymerase I and the four deoxyribonucleoside triphosphates, those primer-templates are converted to full length double-stranded DNAs. The economy in chemical synthesis using this approach is substantial with a in the amount of required as compared with the conventional approach. We describe in detail this meth- odology for the assembly of long gene segments from synthetic oligodeoxyribonucleotides.

' The abbreviations used are: IFa, human leukocyte interferon a,; b.p., base pair. several of the above-mentioned features. The total gene synthesis involved the chemical syntheses of 67 oligodeoxyribonucleotides which were assembled into the complete coding sequence by a series of template-dependent enzymatic ligations, a technique pioneered by Khorana and his co-workers (6).
In an effort to speed up and simplify the assembly of large DNA segments for cloning, we have developed an alternate approach which can reduce the amount of required chemical synthesis by more than 40%. Basically, this procedure as described here involves the chemical synthesis of large DNA fragments (up to 43 bases long) which share a short stretch of complementary sequence a t their 3' termini. When these sequences are annealed in the presence of all four deoxyribonucleoside triphosphates and DNA polymerase I (Klenow fragment), a full length double-stranded product can be synthesized. We have used this approach to assemble the COOHterminal segment of a sequence coding for IFa2.
The synthetic gene segment uses codons corresponding to the most abundant tRNAs in Escherichia coli (5), with conveniently placed restriction endonuclease sites for cloning into a bacterial plasmid vector.
The total chemical synthesis for a final product of 144-b.p. nucleotides involved only four synthetic fragments of 43, 42, 39, and 39 bases long (or a total of 163 bases). This represents a 44% reduction in the amount of chemical synthesis required by the conventional approaches (1,6). The IFm gene segment has been cloned into an E. coli plasmid vector using a threepart ligation, and the correct DNA sequence of the synthetic gene segment has been verified subsequent to its cloning.
We report in detail here the biochemical methodology used to construct and clone the synthetic IFa2 gene segment. This approach to gene synthesis should prove generally useful for the synthesis and cloning of DNAs encoding regulatory and protein sequences. The chemical synthesis of the starting oligodeoxyribonucleotides by the solid phase method will be reported elsewhere.

EXPERIMENTAL PROCEDURES
DNA Synthesis and Purification-Oligonucleotides were chemically synthesized by the phosphotriester method using a solid support. Details of the chemical synthesis will be reported elsewhere (7). Individual oligodeoxyribonucleotides were purified by gel electrophoresis using 8% polyacrylamide gels containing 7 M urea. The DNA fragments were visualized in the gel by ultraviolet shadowing using fluorescent thin layer chromatographic plates and a short wavelength UV lamp. DNAs of the desired chain length were excised from the gel, electroeluted, and further purified by benzoylated naphthoylated DEAE-cellulose (Serva) column chromatography (8).
Enzymes and Chemicals-Restriction endonucleases Eco RI and Pst I, T, polynucleotide kinase, Tq DNA ligase, and bacterial alkaline phosphatase were purchased from Bethesda Research Laboratories. The large fragment of E. coli DNA polymerase I (Klenow) was purchased from Boehringer Mannheim. Acrylamide and bisacrylamide were from Bio-Rad. The [ySPP]ATP (6000-8000 Ci/mmol) was synthesized according to published procedures (9) using commercially available 32P04 (ICN or New England Nuclear); [a-"PIdCTP was purchased from New England Nuclear.
DNA polymerase I reactions were all performed in EndoR buffer (IO Gel and Enzyme Buffers-Restriction endonuclease digestions and m~ Tris-HC1 (pH 7.9), 6.6 mM MgC12, 6 m M P-mercaptoethanol, 60 mM NaC1). Gel electrophoresis was carried out using a one-half concentrate of Tris-borate-EDTA buffer (10). Bacterial alkaline phosphatase treatment, ligations, and kinations were carried out under conditions recommended by the supplier.
DNA Sequencing-DNA sequences were determined using the method of Maxam and Gilbert (11) on restriction fragments labeled at their 3' termini with [cx-~'P]~CTP and DNA polymerase I (Klenow). Labeling was carried out in EndoR buffer at room temperature for 5 min.
Bacterial Transformations-E. coli strain MC1061 (12) was used as the transformation host, employing the procedure of Kushner (13). Antibiotics, ampicillin or tetracycline, were used in the growth media at 30 pg/ml.

RESULTS
Synthetic Design-The complete nucleotide sequences of several different human leukocyte interferon species have  Opal 0-1 Amber 1-0 Schematic of strategy for assembly and cloning of the I F N a z gene segment. The synthetic gene fragments consist of the following deoxyribooligonucleotides: A, 43-mer; B , 39-mer; C, 42mer; and D, 39-mer. Fragments A and B share 10 bases of complementary sequence at their 3' termini, while fragments C and D share nine bases of a 3' complementary sequence. The appropriate pairs of fragments were annealed, and the double-stranded products were polymerized by the addition of all four deoxyribonucleoside triphosphates and E. coli DNA polymerase I (Klenow fragment). The double-stranded products formed are 72 base pairs long in each case. Subsequent to the polymerization reaction, the double-stranded po-been determined from analyses of cDNA and genomic clones (14-17). We have utilized the amino acid coding sequences deduced for a2 (17) to chemically synthesize four oligodeoxyribonucleotides which, when assembled, comprise the coding sequence for IFa2 amino acid residues 126 through 165. In addition, restriction endonuclease recognition sites for Eco RI and Pst I were introduced into the synthetic design to facilitate cloning of the IFa2 segment into a bacterial plasmid vector. As stated under "Introduction," the codon usage takes advantage of the most abundant E. coli tRNA species ( 5 ) and, therefore, differs considerably from that found in the corresponding natural IFa2 (Table I).
Four oligodeoxyribonucleotide fragments, 43, 42, 39, and 39 bases long, were chemically synthesized by the phosphotriester method. Each pair of fragments was designed to have a short region of complementary sequence at its 3' termini ( Fig.  1). For the A and B fragments, this stretch is 10 b.p., while for the C and D fragments there is 9 b.p. of complementary sequence. When the pairs of fragments are annealed in the presence of deoxyribonucleoside triphosphates, they become a substrate for DNA polymerase I (K1enow)-mediated polymerization to give a full length duplex product ( Figs. 1 and 2).

I n Vitro
Construction of Double-stranded Gene Fragments-Prior to polymerization, each synthetic fragment was phosphorylated at its 5' terminus, using polynucleotide kinase and ATP. This was done to facilitate detection and subsequent cleavage of the double-stranded products by the appropriate restriction endonucleases. In a typical experiment, equimolar amounts of 5' terminally phosphorylated fragments A and B or C and D were mixed together, heated in a boiling water bath for 3 min, quickly chilled, and then allowed to form the desired annealing a t their 3' termini. The fragments were then incubated with all four deoxyribonucleoside triphosphates and DNA polymerase I. The products of these reactions were electrophoresed in acrylamide to resolve duplexes from the single-stranded oligonucleotides (Fig. 2). From the relative

T A G T A C G C A A G A A A G A G G G A C A G A T G A T T G G A A G T C C T C 5'
( D l -

T C T C T G C G T l C T A A A G A A l A G C~f i~~~G T G G P * t l A G A G A C G C A A G A T T T C T T A T C G I A C G T C A C C ID\
lymerization products were cleaved with either Eco RI for A and B or Pst I for C and D. The desired products of these cleavages were purified by gel electrophoresis. The appropriate fragments were eluted from the gel and ligated with the Eco RI-Pst I-cleaved plasmid pXJOOl as depicted. See text for additional details concerning screening and selection of colonies containing the cloned IFa2 gene segment. The complete nucleotide sequence of the synthetic, polymerized products is depicted below. The heavy underlining depicts the original, synthetic oligonucleotides A, B, C, and D. The underlinings overlap at the regions of 3' complementary sequences. intensities of the bands in the autoradiograph presented in Fig. 2, it can be seen that approximately 40-50% of the starting single-stranded oligonucleotides were polymerized into duplex structures. To prepare the synthetic sequences for cloning, 62 pmol of each oligonucleotide were treated as described above. Subsequent to the polymerization reaction, the duplexes formed from A plus B were cleaved with the restriction endonuclease Eco RI, while the C plus D duplexes were cleaved with Pst I. The efficiencies of these cleavages were examined by electrophoresis in a denaturing acrylamide gel (data not shown). After restriction endonuclease digestions were deemed complete, the appropriately cleaved fragments were electrophoresed in a nondenaturing acrylamide gel, and the DNAs were electroeluted from the gel slices in preparation for cloning.
Strategy for Cloning the Synthetic IFa:! Fragments-The bacterial vector used for cloning the IFa2 fragments is a derivative of pBR327 (la), designated pXJ00l (19). This vector has unique Eco RI and Pst I sites into which the synthetic fragments were inserted. Removal or replacement of the -650b.p. plasmid Eco RI-Pst I fragment destroys the plasmidcoded ampicillinase gene. The strategy used for ligation of the synthetic oligonucleotides into the plasmid vector is depicted in Fig. 1 and involves a three-part ligation of Eco RI ends, Pst I ends, and a blunt end ligation of the two synthetic fragments. Approximately 30 pmol each of the double-stranded Eco RIor Pst I-cleaved fragments were ligated with 2 pmol of Eco RI-Pst I-cleaved alkaline phosphatase-treated pXJ001. The ligation mixture was used to transform the E. coli strain MC1061 (12), and selection was made for colonies resistant to 30 p g / d tetracycline. From this mixture, there were approximately IO00 transformants. Twenty out of approximately 400 colonies tested were ampicillin-sensitive. Plasmid DNA was prepared from 10 of these ampicillin-sensitive colonies using a plasmid DNA miniscreen procedure (20). Eight of the 10 colonies tested had plasmids which, when cleaved with Eco RI and Pst I, yielded a restriction fragment of the expected length, 132 b.p. Plasmid DNA from one of these transformants was purified in large scale and the insert DNA was subjected to DNA sequence analyses (Fig. 3). The results presented in this figure demonstrate that the desired IFa2 sequence had been successfully cloned by the combined chemical-biochemical approach.

DISCUSSION
The development of more efficient methods for the chemical synthesis of oligodeoxyribonucleotides by the solid phase approach allows us to chemically synthesize DNAs of over 40 bases in length (7). A marked advantage of utilizing long DNAs for cloning is that it greatly simplifies the gene assembly process. Prior to this communication, the only approach for gene assembly was that pioneered by Khorana (6) and his coworkers, template-dependent ligations of short oligonucleotides. However, this approach has some disadvantages. The entire DNA sequence to be synthesized must be carefully screened by computer analyses to search for regions of overlapping homologies (1). These regions could hamper desired placements of restriction endonuclease sites as well as codon usage in certain regions of the gene sequence. The methodology we have utilized to assemble the IFa2 gene segment provides a rapid and convenient alternative to constructing double-stranded gene segments. Although homologous complementary sequences other than the target sequences could lead to undesired polymerization products, this is more easily controlled by adding or deleting a few nucleotides in the design of the region of 3' complementary sequence. It should be noted that DNA polymerase has been previously used to complete a double-stranded sequence, but in this case, only one strand served as primer (6).
The fidelity of the overlapping polymerization technique is demonstrated by the nucleotide sequence analyses of the cloned gene segment (Fig. 3) which agree precisely with the designed sequence (Fig. 1). The three-part ligation methodology used to clone the IFa2 gene segment also greatly simplifies the gene assembly process (Fig. 1). The frequency of clones containing the desired insert was somewhat low, approximately 4%. This could be due to the lowered probability of properly aligning all three segments of DNA for the desired joinings. Although we used a DNA miniscreen procedure to analyze the phenotypically correct clones, colony hybridiza- The plasmid was secondarily cleaved with HindIII and the appropriate fragment was purified from the gel and sequenced. The nucleotide sequence presented in A corresponds to the upper strand of the Fig. 1 sequence beginning with the Pst I site. For clarity, two different lengths of sequencing gel runs are presented. For the B sequence, the HindIII site was 3'-labeled with [a-32P] dCTP and DNA polymerase I (Klenow).
The plasmid was secondarily cleaved with Pst I. The sequence presented corresponds to the lower strand of the Fig.  1 sequence beginning with amino acid 128. The correct sequence upstream of this position has also been confirmed (data not presented). tion with radioactively labeled synthetic DNA probes could greatly facilitate the screening process (21).
One of the most important features in synthesizing a designed gene coding for a protein would be higher expression than the corresponding natural gene in a microbial environment. It has been demonstrated for several organisms that a definite bias in codon usage exists for highly expressed proteins. In E. coli, where many tRNA species have been identified, this codon bias seems to be well correlated with the relative abundance of iso-accepting tRNA species (5). We have incorporated the E. coli preferred codons into the synthetic IFa2 segment (Table I). Obviously, we will have to await the total gene synthesis for IFaz in order to draw any conclusions as to whether or not codon usage will play a significant role in the expression of this gene.