Proc. Natl. Acad. Sci. USA Vol. 77, No. 3, pp. 1442-1446, March 1980 Biochemistry Initiation of protein synthesis in bacteria at a translational of mammalian cDNA: Effects of the preceding nucleotide sequence (genetic expression//plasmid/DNA cloning/ binding) ANNIE C. Y. CHANG*tt, HENRY A. ERLICHt, ROBERT P. GUNSALUS§, JACK H. NUNBERG§, RANDAL J. KAUFMAN1, ROBERT T. SCHIMKE§, AND STANLEY N. COHEN*tll Departments of *Genetics, tMedicine, §Biological Sciences, tMedical Microbiology, and IPharmacology, Stanford University, Stanford, California 94305 Contributed by Stanley N. Cohen, December 31, 1979

ABSTRACT Plasmids containing a mouse cDNA sequence Thus, these bacterial plasmids were a potentially useful source encoding the enzyme dihydrofolate reductase (DHFR; te- of information about the effects of DNA sequence variation in trahydrofolate dehydrogenase; 5,6,7,8-tetrahydrofolate:NADP+ the translational control region on the efficiency of hetero- oxidoreductase, EC 1.5.1.3) have been used to study the effi- ciency of initiation of protein synthesis at an ATG (AUG) specific expression. The results of studies on these effects translational start codon indigenous to the eukaryotic cDNA. are reported here. Differences in DHFR production assayed phenotypically, en- zymatically, and immunologically were correlated with the MATERIALS AND METHODS primary structure of the DNA segment that precedes the Bacterial Strains and Plasmids. The construction, isolation, translational start codon. Our results indicate that initiation of and endonuclease mapping of chimeric plasmids that contain a structurally discrete and biologically functional eukaryotic encoding for mouse DHFR (pDHFR protein can occur in bacteria on a fused mRNA molecule, and double-stranded cDNA that the efficiency of expression is strongly affected by: (i) the plasmids) have been described (6). strain extent of homology of the translational control region with the X2282, a thy + derivative of X1776 (7), was used as the bacterial 3'-OH end of 16S ribosomal RNA, and (ii) the distance between host under P2 containment conditions, as specified in the NIH the protein start codon and the ribosome-binding sequence on Guidelines for Recombinant DNA Research of Dec. 22, 1978. the mRNA. Experimental Procedures. Restriction endonucleases were purchased from either Bethesda Research Labs (Rockville, MD) Introduction of double-stranded complementary DNA (cDNA) or New England BioLabs and used according to the suppliers' or chemically synthesized DNA into an endonuclease cleavage recommendations. The 5' end of endonuclease-generated DNA site in a bacterial gene has enabled the synthesis of hybrid fragments was labeled with ['y-32P]ATP by using polynucleotide that react with antibodies against their eukaryotic kinase (8) (specific activity 3000 Ci/mmol; 1 Ci = 3.7 X 1010 component (1-5). Although these hybrid proteins have not been becquerels). DNA nucleotide sequences were determined as shown to be biologically functional, biological activity has been described by Maxam and Gilbert (8); chemically cleaved DNA found for a discrete (i.e., nonhybrid) protein encoded in bacteria fragments were run on 0.35-mm gels containing 8% (9) or 20% by a mouse cDNA sequence for the enzyme dihydrofolate re- (8) acrylamide. ductase (DHFR; tetrahydrofolate dehydrogenase; 5,6,7,8-te- Enzyme assays for activity of mouse DHFR in extracts of E. trahydrofolate:NADP+ oxidoreductase, EC 1.5.1.3) (6); the high coli cells were described previously (6). Filter affinity transfer level of resistance to the antimetabolic drug trimethoprim (Tp) assays (6, 10) and in vitro synthesis of proteins were carried out specified by the mouse enzyme allowed isolation of bacterial as described (11). The [asS]-labeled cells that phenotypically expressed the eukaryotic genetic se- products from in vitro or in vivo synthesis were immunopre- quence. These bacteria were found to synthesize a protein that cipitated (12) with rabbit anti-mouse DHFR or rabbit anti-E. has the enzymatic properties, immunological reactivity, and coli f-lactamase serum (the gift of R. B. Sykes of Squibb) and molecular size of the mouse DHFR. Moreover, the cDNA Staphylococcus aureus protein A. The direct chemical transfer segment in such clones was in a different translational reading of proteins from gels to a reactive diazo filter paper has been frame from the bacterial f3-lactamase gene into which it had described by Renart et al. (13). been inserted, suggesting that the DHFR was not part of a fused protein. Together, these findings implied that initiation of RESULTS translation was occurring at the translational start codon nor- Relationship of Structure at Vector-cDNA Junction to mally used for the mouse DHFR. pDHFR Expression. Each of the plasmids studied was con- Different DHFR cDNA-containing clones showed greatly structed by of poly(dC)-"tailed" DHFR cDNA at the different levels of Tp resistance. Such differences appeared to poly(dG)-tailed PstI cleavage site of the pBR322 plasmid vector reflect control at the translational level, because all of the (6), and all except pDHFR 23 showed phenotypic expression plasmid constructs used the same 3-lactamase gene promoter of Tp resistance. Sequence analysis of the nucleotides in the and had the same site of cDNA insertion. However, the length vicinity of the vector-cDNA junction was carried out for each of the DNA segment on the 5' side of the structural gene ap- plasmid by 5' end-labeling of Hpa II and Taq I endonuclease- peared to be different in the various plasmids, and gel analysis generated DNA termini according to the scheme shown in Fig. of endonuclease-cleaved plasmid DNA showed differences in 1. This experimental plan enabled determination of the se- the sequence preceding the DHFR translational start codon. quence of both DNA strands within the region of interest.

The publication costs of this article were defrayed in part by page Abbreviations: DHFR, dihydrofolate reductase; Tp, trimethoprim; charge payment. This article must therefore be hereby marked "ad- bp, base pair(s); RBS, ribosome-binding sequence; S-D, Shine-Dal- vertisement" in accordance with 18 U. S. C. §1734 solely to indicate garno. this fact. II To whom reprint requests should be addressed. 1442 Downloaded by guest on September 26, 2021 Biochemistry: Chang et al. Proc. Natl. Acad. Sci. USA 77 (1980) 1443

3- Lactamase DH F R resistance to Tp concentrations of 500 gAg/ml or greater; clones A Pstla Taqi Hael l Hpall that expressed lower levels of resistance could not be distin- guished enzymatically, but showed different minimal inhibi- HaelIl Hpall Hinfl the 37561 3658 161 1 tory concentrations of Tp during growth. The sequence of pDHFR 7, 12, 27, 29 DNA strand equivalent to the mRNA is shown. The primary 13, 26, 28 sequencing data for some of the more interesting plasmids are 26 shown in Fig. 2. 28 As shown in Table 1 (Column IV), the A of the translational DHFR ,- Lactamase start codon is located 11 to 14 bases after the center of the 5- B Hpal HaelII Taql Pstlb base-pair (bp) sequence homologous to the C-C-U-C-C of 16S rRNA in all clones that express resistance to Tp, except for HpalI HaellI pDHFR 26, in which the ribosome-binding sequence (RBS) 1361 113548 l3489 (15-17) is located 24 nucleotides before the ATG. In the pDHFR 25 pDHFR 7 plasmid, the sequence formed by a Pst I cleavage FIG. 1. Scheme used to determine the nucleotide sequence ofthe site and the poly(dG) additions show a 5-nucleotide-long seg- DHFR cDNA insert at the vector-cDNA junctions corresponding to 3' OH end of 16S ri- the 5' end of the mRNA pDHFR plasmids. The boxed numbers at the ment of homology with a sequence at the Hae III, Hpa II, and Pst I sites indicate the nucleotide positions in bosomal RNA (Shine-Dalgarno, or S-D sequence) (15, 16). In the pBR322 plasmid as determined by Sutcliffe (14); these sites were the other plasmids, the Pst I site is further from the translational used to orient the structural gene coding for mouse DHFR. The arrows start codon, and a segment homologous with the S-D sequence above the map show the direction of translation of mouse DHFR in is formed entirely by nucleotides within the cDNA or by a the plasmid with respect to the gene coding for f3-lactamase. The as- combination of sequences from the cDNA and the homopoly- terisks indicate the 5'-labeled end of the DNA fragments selected for sequence analysis, and arrows indicate the direction of sequencing. meric "linker." (A) Scheme used for sequencing pDHFR plasmids 7, 12, 13, 26, 27, Comparison of the sequence at the vector-cDNA junction 28, and 29. (B) Scheme used for sequencing the corresponding seg- for the pDHFR 12, 13, and 28 plasmids indicates that minimal ment of pDHFR 25, which contains a cDNA insert in orientation alterations in nucleotide sequence or positional changes in the opposite to that of the f3-lactamase gene of the vector. putative ribosomal binding region can have marked effects on expression. For example, the pDHFR 12 and 28 plasmids contain 28 and 27 base pairs, respectively, between the terminal Table 1 shows the primary structure of the DNA segment nucleotide of the vector DNA and the translational start codon between the Pst I cleavage site and the ATG (AUG) transla- for DHFR, and in both cases the translational tional start codon in each of the plasmids studied, and it indi- of DHFR is different from that of the f3-lactamase. In pDHFR cates other parameters used to characterize these plasmids. 12, to an A (coincidentally generating an Alu I Expression of DHFR was assayed enzymatically in vitro and cleavage site) has occurred at position -13, which is ordinarily phenotypically. DHFR enzymatic activity is detectable in occupied by a C in DHFR mRNA (as illustrated by the se- sonicates of E. coli cells carrying pDHFR plasmids that express quences of pDHFR 23 and 26). Because the mutated locus is

Table 1. Nucleotide sequence in the region of the vector-cDNA junction corresponding to the 5' end of the mRNA, and other properties of pDHFR chimeric plasmids V IX IV DHRF Length of III bp from A of specific VI VII VIII W-lactamase Orienta- ATG to center activity, MIC Relative Reading extension, II tion of of 5-bp units/mg of Tp, DHFR frame of no. of amino Plasmid Nucleotide sequence cDNA sequence' protein ug/ml activity DHFR acids

-1.1 _I pDHFR 7 TCCACGIGG(;GGG(;GGGCA ATG(.TrA1A 12 2 >1000 1 +1 8 * 0* 0

pDHFR 12 TGCAGIGC.GGGGGG(;c(;GGGGGGGCTG(CcATCATG(;TT A 10 4.2 > 1000 2.1 + 1 13

pDHFR 13 TG;('AGGGGGGGGGGGGGGGCC('ATCATGGTT A 11 2 >1000 1 +1 11

pDHFR 26 TGCAGIG CGTGA A( GCTG(;TAGGAPTTATCC('CCGC^TGC('ATCATGGTT A 20 ND 75 0.075 +2 22 *- --*0 * -- - 3 pDHFR 27 TGC'AGI(;GGGGGGGGGGAT(;GTT A 12 - 250 0.250 +1 8 * *0 0 * pDHFR 28 TA;(-AGIG(;GGGGGGGGGGGG(1(;GCcF(;(CAT AT(;GTT A 11 ND 150 0.150 +2 >32

pDHFR 29 TGC'AGIGGGG(;(;GGGG;(G;TCATGGrT A 5or 15 500 0.500 +1 pDHFR 2:3 TGCAGjGGGGG- (;GGGGGG;AGGATTrTATCCCCGCTGC('ATCATGGTT B 20 ND <2.5 <0.0025 - NA

pDHFR 25 TGCAGIGGGGGGGGGCTGCCATCATGGTT B 11 ND 25 0.025 - NA PSI I 16S rRNA :1 AUL'CCLCCACUAGG-5

The black dots below the nucleotides indicate homology with the nucleotide sequence at the 3' OH terminus of 16S rRNA as shown in the bottom line in the table. Positive numbers start from the first base pair following the translational start codon for mouse DHFR. Negative numbers

start from the nucleotide in the 5' direction from 1. Nucleotides to the left of the vertical line at the Pst I site are within the f3-lactamase gene (14); the translational reading frame of this gene is indicated by the to the left of the Pst I site. Enzyme activity and the minimal inhibitory concentration (MIC) of Tp for each clone were determined as described (6). ND, not detectable. Column VII was derived from columns

IV and V, assigning a value of 1 to the level of activity shown by pDHFR 7. The information shown in columns III, IV, VIII, and IX was obtained from the DNA sequence data shown in the table and the reported (6) sequence of pDHFR 7. NA, not applicable. Column VIII gives reading frame with respect to the /3-lactamase gene. * Column IV gives the distance from the A of the translational start codon to the center of a sequence showing the best complementarity to CCUCC of 16S rRNA and the greatest overall complementarity to the rRNA sequence. Downloaded by guest on September 26, 2021 1444 Biochemistry: Chang et al. Proc. Natl. Acad. Sci. USA 77 (1980)

pDHFR 12 pDHFR 13 pDHFR 25 pC)HFFR 26 G-C FIG. 2. Autoradiograms showing Taq I-Hae III Taq 1*-I~-Ip A+ Hpa ll*-Hae IlIl Hpa Il*-Hinf I /GA T the DNA sequence at the vector- G C+T C G A>G A+C C+T C G A A+C G A+G C (c+--T C+T C A+I _~ _w 5A @ eT-AcDNA junction of plasmids con- |G/ taining cDNA encoding mouse w "W /=~.A T DHFR. DNAsequenceanalysiswas .- _ G /A 1G carried out as described by Maxam ._<= _c |A__ ? gT G and Gilbert (8). The fragment ana- C A !AT '-C ; * -A lyzed is indicated for each plasmid; - are rn A-T the 5'-labeled ends of fragments -A -A=CA- A C _ -A sTshownon by asterisks. Theh horizontal G G-ACT arrow in pDHFR 12 shows the site of -G C G am .-C-G-C 111 _ GS~-C mutation of the ribosomal binding -10 -A-A-T Ebb G m AG ~~Csequence of the cDNA. The vertical A-T .i~~~~~~~~i - arrows indicate the direction of -G- C -G-C reading of the mouse DHFR and -TA -G 4 -C - -C f3-lactamase genetic sequences. For -G _T _0 - T-A pDHFR 12 and pDHFR 13 the 13AG _M -A-T -C strand shown corresponds to the _A 4f_G-C DHFR _G -T-A mRNA for both the mouse -A - in 26 the -C -A T | T-A and f3-lactamase; pDHFR _-C-G -A-T sequence of the opposite strand is -CG w., A _ft Ht-T -C-G C t -C-G shown. For pDHFR 25 the strand moms -C -A -T -C-G - C-G analyzed corresponds to the mRNA *-A -A sequence for the f,-lactamase, but -G -A- T MAW_ - A - T the strand complementary to the (1)~~~DeFtutrCsq~~~~ - 5 mRNA for DHFR (i.e., the DHFR low, - A .-T cDNA insert) is in an orientation -A-T opposite to that of the f,-lactamase (11) 5' nontranslated cDNA (11l) Homopolymeric adc itions gene. contiguous with the homopolymeric segment that links the putative RBS. Although substantial homology exists between DHFR cDNA to the vector, it is not clear whether the mutation the putative RBS and the S-D sequence, and the presence of occurred by a direct.C to A change in the sequence of the insert a termination codon prior to an AUG reportedly favors reini- or whether it was a G to A change involving the terminal G of tiation of translation on polycistronic or intercistronic mRNA the homopolymeric addition; in the resulting plasmid DNA molecules (22), the level of DHFR expression from the pDHFR sequence, 9 base pairs within a 12-nucleotide segment are ho- 26 plasmid is relatively low. When an RBS the same distance mologous with the S-D sequence; in pDHFR 28, the same ex- (i.e., 24 bp) from the start codon is coupled to an apparently tent of homology extends over 13 base pairs. weak promoter (as.in the pDHFR 23 plasmid, which initiates A single nucleotide substitution within the RBS of pDHFR into the DHFR cDNA insert from the distal end 12, lengthening of the vector-cDNA junction region by one base of the f3-lactamase gene), no phenotypic expression of DHFR pair, or a combination of both factors results in a 14-fold in- is detectable at all. In contrast, resistance to Tp is observed for crease in expression for pDHFR 12 relative to pDHFR 28. The the pDHFR 25 plasmid, which contains a DHFR cDNA insert pDHFR 12 plasmid was obtained by selection for high level in the same orientation as pDHFR 23 (Table 1), but which has expression of Tp resistance (6), and this may have led to the a distance of only 14 bp between the putative RBS and trans- isolation of bacteria having a mutated translational control lational start codon. These findings indicate that small differ- region. An analogous effect of a single G to A mutational change ences in the distance between the RBS and the translational start within the RBS of gene 0.3 of bacteriophage T7 has been re- codon can have important effects on translation. ported (18). However, thermodynamic calculations suggest that Immunological Analysis of Proteins Encoded by Cloned the interaction of G-G-A-G-C with the S-D sequence would Mouse cDNA. Filter affinity transfer analysis (6, 10) (Fig. 3A) not be substantially different from the interaction of G-G-G- indicates that all of the Tp-resistant bacterial clones tested G-G with the same segment (19, 20). (including one having its cDNA insert in orientation B, pDHFR In pDHFR 25, the primary sequencing data (Fig. 2) for the 25) synthesize a that reacts immunologically with translational regulatory region verify directly our earlier con- antibody to mouse DHFR and has the same electrophoretic clusion from endonuclease mapping (6) that the cDNA segment mobility as the mouse enzyme (Mr 21,500) (26). No such pep- is oriented in a direction opposite from the direction of reading tide was made by bacteria containing either the pBR322 vector of the f3-lactamase gene into which it is inserted. Because this alone or the nonexpressing pDHFR 23 plasmid. plasmid encodes resistance to Tp (see Table 1), transcription Filter affinity transfer analysis of extracts of bacteria con- of DHFR cDNA must be proceeding from or beyond the distal taining the pDHFR 12, 25, 26, and 28 plasmids was also at- end of the f-lactamase gene on the DNA strand opposite to the tempted with antibody to bacterial f3-lactamase. Although the one used for 3-lactamase expression. The detection of such #-lactamase encoded by the pBR322 vector was identified by transcription suggests that the translationally competent DHFR filter affinity transfer assay at the appropriate gel position in cDNA inserts may be useful for the identification of bacterial such experiments, no peptide containing fl-lactamase antigenic gene promoters and for the assay of promoter activity. sites was detected in the DHFR-expressing clones. Such pep- The insert of the pDHFR 26 plasmid is the longest of those tides were identified, however, by immunological analysis of analyzed here; the sequence has been published elsewhere in proteins transferred from the gels to reactive diazo-cellulose its entirety (21). The cDNA segment of the plasmid contains filters (13) (Fig. 3B). Also seen in the analysis were adventitious a TGA triplet in the same translational reading frame as the peptide bands that reacted with the ,B-lactamase antiserum. bacterial f3-lactamase; this presumably leads to termination of These were present in all of the extracts tested and were pre- the peptide initiated at the translational start codon of the sumed to result from contaminating antibodies directed against f3-lactamase gene. Nucleotides homologous with the S-D se- non-,B-lactamase proteins. In extracts of cells containing quence follow the termination codon and are themselves fol- pDHFR 12 or the nonexpressing pDHFR 21 plasmid (6), the lowed by a translational start codon that starts 24 bp from the size of the peptides reacting specifically with 0-lactamase Downloaded by guest on September 26, 2021 Biochemistry: Chang et al. Proc. Nati. Acad. Sci. USA 77 (1980) 1445

A_ Mr 66,000

* 45,000

DHFR- p ' , -34,700 1-~~~~~~.I :

a b C d e f g h

B

1-. 4i1, ,'DOW, --*--2 *"* --o-3 -14,300

a b c d e

a b c d FIG. 4. Gel analysis of immunoprecipitated peptides synthesized FIG. 3. (A) Filter affinity transfer analysis (10) of bacterial cell in vitro with plasmid DNA templates. Protein synthesis was carried 'extracts containing pDHFR chimeric plasmids. Sonicated extracts out in vitro as described (11). Reaction mixtures were incubated for 60 min in the presence of the trypsin inhibitor phenylmethylsul- (20 of E. coli cultures or mammalian DHFR were subjected to ,l) fonyl-fluoride and [35S]methionine (58 mCi/ml) and electrophoresis at constant current for 3 hr in an 11.25% Na- (5Ogg/ml), were then precipitated with rabbit antibody (12). DodSO4/polyacrylamide gel. Peptides were transferred from the gel anti-f3-lactamase Pre- were centrifugation, to strips of a dry cellulose filter that had been covalently coupled to cipitates collected by resuspended in a solution of 10% (vol/vol) glycerol/0.7 M 2-mercaptoethanol/2.37% Na- anti-DHFR F(ab')2 fragments. Lane a, pBR322; lane b, pDHFR 7; DodSOQ0.0625 M Tris*HCl (pH and subjected to electrophoresis lane c, pDHFR 12; lane d, pDHFR 25; lane e, pDHFR 23; lane f, 6.$), for 8 hr on 12.75% used in mammalian DHFR; lane g, pDHFR 26; lane h, pDHFR 28; and lane polyacrylamide gel. Plasmid different reaction mixtures were: lane a, pBR322; lane b, pDHFR 12; lane c, i, mouse DHFR. (B) Identification of fl-lactamase and mouse DHFR antigen in bacterial extracts by direct transfer of proteins from gels pDHFR 25; lane d, pDHFR 26; lane e, pDHFR 28. Arrow 1 indicates to diazo-filters after electrophoresis. Sonicated extracts of E. coli the position of the wild-type /3-lactamase. The other arrows show the cultures containing pBR322 (lane a), pDHFR 12 (lanes b and d), and positions of truncated fl-lactamase peptides encoded -by pDHER plasmids. Other adventitious bands that precipitated with pDHFR 21 (lane c) were electrophoresed in a 12.75% NaDodSO4/ ,were which possibly represent polyacrylamide gel. The gel was thien washed for 1 hr in Na-K phos- anti-/3-lactamase antibody, and fragments of partially degraded /3-lactamase peptides (23), are seen. phate-buffered saline. (0.15 M, pH 7.4) and placed on a moist filter paper, and a reactive diazo-filter (13) was laid on the gel. After transfer for 10 hr, the filter was inactivated by a 6-hr incubation at room which electrophoreied protein bands had been coupled (Fig. temperature in 1 M /1% -bovine serum albumin and then 3B). Autoradiographic examination of the filter after incubation washed with 1% serum albumin. After a 3-hr incubation with rabbit serum anti-/3-lactamase serum (diluted 1:1000 in 1% serum albumin), the with anti-f-lactamase and 125I-labeled protein A showed filter was analyzed by autoradiography. Subsequently, the same filter a single immunoreactive band representing a truncated was incubated with rabbit anti-mouse DHFR serum (1:1000 in 1% /l-lactamase peptide (Mr 20,000). Subsequent treatment of the serum albumin) and l25I-labeled protein A. Lane d displays the au- same filter with anti-DHFR serum showed a new immuno- toradiogram of the same filter strip shown in lane b (pDHFR 12), after reactive DHFR band (Mr 21,500) above the truncated fl-lac- the additional treatment with anti-DHFR serum and 1251-labeled tamase peptide. The absence of determinants on protein A. Arrow 1, wild type 13-lactamase; arrow 2, pDHFR 12-en- -fl-lactamase the.peptide that reacted with anti-DHFR serum strongly sup- coded peptide containing f3-lactamase antigenic sites; arrow 3, pDHFR 12-encoded peptide reactive only with anti-DHFR anti- ports 'the interpretation that the biologically active mouse body. DHFR produced in these bacterial clones is not being made as part of a fusion product. antibody was smaller than the pBR322 vector-encoded f3-lac- In analogous experiments, the purified DNA of several tamase peptide and was different for the two pDHFR plasmids. pDHFR plasmids was added to an in vitro protein-synthesizing Furthermore, the amount of the immunoreactive peptide was system containing the trypsin inhibitor phenylmethylsulfonyl reduced in these extracts, and no f3-lactamase at all was found fluoride (11), and the resulting [a5S]methionine-labeled peptides in cells containing some of the pDHFR plasmids, suggesting were analyzed by gel analysis after immunoprecipitation with that insertion of the DHFR cDNA sequence has led to prema- antibody to f-lactamase (Fig. 4). In each case, except for ture termination and subsequent degradation (23) of the pDHFR 25 (which is in orientation B), the length of the peptide f3-lactamase protein. Such translational termination could be band shown by the arrow corresponds to the distance from the caused by TGA triplets within the cDNA insert that are in the NH2 terminal of the f3-lactamase protein to a nucleotide triplet same reading frame as the f3-lactamase peptide (but not in the within the DHFR sequence that could terminate translation correct reading frame for the DHFR) in these clones. The cal- in the reading frame of the f3-lactamase. The truncated f3-lac- culated molecular weight of the peptides made by both the tamase peptide encoded by pDHFR 12 DNA in vitro (lane b) pDHFR 12 and pDHFR 21 plasmids was in good agreement was the same size seen in vivo (Fig. 3B, lanes b and d). No im- with the length predicted from the position of the termination munoreactivity with antibody to mouse DHFR was seen for any codon in the DNA sequence in the cDNA insert. /3-Lactamase of the bands that contained f3-lactamase antigenic sites, as ex- and DHFR antigenic determinants were assigned to individual pected from our DNA sequence analysis showing that the gel bands by sequential immunological probing of a filter to segment of the hybrid peptide encoded by the 3-lactamase gene Downloaded by guest on September 26, 2021 1446 Biochemistry: Chang et al. Proc. Natl. Acad. Sci. USA 77 (1980) is out-of-frame with the DHFR. However, a Mr 21,500 peptide 3-lactamase gene in the direction of the NH2 terminus of the that reacted only with antibody to mouse DHFR was made in B-lactamase, nor was any such peptide detectable in gels. vitro (unpublished data) as in vivo. In the plasmids we have studied, the optimal distance be- We conclude that two separate peptides are being synthe- tween the beginning of the RBS and the A nucleotide of the sized on the same messenger RNA in vitro: (i) a fused peptide translational start codon appears to be 12-14 bp. A decrease in consisting of the NH2-terminal segment of 3-lactamase and a the homology with the S-D sequence, or positional changes that nonsense segment resulting from out-of-frame read-through bring the homologous sequence into a less optimal position, translation into the DHFR cDNA sequence and (ii) an accu- reduces translational efficiency. rately translated DHFR protein that is initiated at a translational The structural configuration described here and previously start codon indigenous to the eukaryotic DNA segment. In some (6) seems to be generally applicable to the production of other instances a termination signal for the f3-lactamase peptide oc- biologically functional eukaryotic proteins in bacteria. Very curs within a few amino acids of the initiation site of the func- recently, this approach has been used for the synthesis in E. coli tional DHFR protein (+1 frame); however, in other cases (+2 of discrete peptides having the size and immunological char- frame, e.g., pDHFR 28), out-of-frame translation can proceed acteristics of human growth hormone (24) and simian virus 40 32 amino acids into the DHFR gene sequence before encoun- t antigen (25). tering a termination triplet in the f3-lactamase reading frame. This work was supported by grants from the National Institutes of Health, the American Cancer Society, and the National Science DISCUSSION Foundation (S.N.C. and R.T.S). H.A.E. is a Fellow of the Cancer Re- search Institute, and he acknowledges the hospitality and support of Several kinds of data show that the appropriately sized, im- H. 0. McDevitt. R.P.G. is supported by a postdoctoral training grant munoreactive, and biologically functional DHFR molecules from the National Institutes of Health. synthesized in bacteria that phenotypically express Tp resis- 1. Itakura, K., Hirose, T., Crea, R., Riggs, A. D., Heyneker, H. L., tance result from initiation of the DHFR peptide at the AUG Bolivar, F. & Boyer, H. W. (1977) Science 198, 1056-1063. translational start codon normally used for expression of the 2. Villa-Komaroff, L., Efstratiadis, A., Broome, S., Lomedico, P., eukaryotic DNA sequence. All of the expressing clones we ex- Tizard, R., Naber, S. P., Chick, W. L. & Gilbert, W. (1978) Proc. amined contained DHFR cDNA in a different translational Natl. Acad. Sci. USA 75,3727-3731. reading frame from the bacterial gene into which it had been 3. Goeddel, D. L., Kleid, D. G., Bolivar, F., Heyneker, H. L., Yan- sura, D. G., Crea, R., Hirose, T., Kraszewski, A., Itakura, K. & inserted; this contrasts with the reading frame concordance Riggs, A. D. (1979) Proc. Natl. Acad. Sci. USA 76, 106-110. observed when eukaryotic peptides synthesized in bacteria are 4. Puijalon-Mercereau, O., Royal, A., Cami, B., Garapin, A., Krust, part of hybrid proteins. The occurrence of different transla- A., Gannon, F. & Kourilsky, P. (1978) Nature (London) 275, tional reading frames for the f-lactamase and DHFR in all of 505-509. the phenotypically expressing clones we examined suggests that 5. Martial, J. A., Hallewell, R. A., Baxter, J. D. & Goodman, M. M. a shift of reading frame may be required for initiation of a new (1979) Science 205, 602-607. 6. Chang, A. C. Y., Nunberg, J. H., Kaufman, R. J., Erlich, H. A., peptide on the polycistronic 3-lactamase mRNA, and that Schimke, R. T. & Cohen, S. N. (1978) Nature (London) 275, perhaps a DHFR peptide fused in-frame to the 3-lactamase 617-624. segment is not biologically active. Presumably fusion peptides 7. Curtiss, R., III, Inoue, M., Pereira, D. A., Alexander, L. & Rock, would nevertheless react immunologically, as do the truncated L. (1977) in Molecular Cloningof Recombinant DNA, eds. Scott, ,B-lactamase peptides made in DHFR-expressing clones (Figs. W. A. & Werner, R. J. (Academic, New York), pp. 99-111. that function 8. Maxam, A. M. & Gilbert, W. (1977) Proc. Natl. Acad. Sci. USA 3B and 4). Such considerations suggest biological 74,560-564. cannot always be inferred from the immunological reactivity 9. Sanger, F. & Coulson, A. R. (1978) FEBS Lett. 87, 107-110. of eukaryotic peptides made in bacteria. 10. Erlich, H. A., Levinson, J. R., Cohen, S. N. & McDevitt, H. 0. Although pDHFR plasmids direct the synthesis of immu- (1979) J. Biol. Chem. 254, 12240-12247. nologically active 3-lactamase and DHFR peptides in vivo and 11. Gunsalus, R. P., Zurawski, G. & Yanofsky, C. (1979) J. Bacteriol. in vitro, no single protein that contains antigenic sites for both 140, 106-113. ,B-lactamase and DHFR was observed. Moreover, the 12. Kessler, S. W. (1975) J. Immunol. 115, 1617-1624. only 13. Renart, J., Reiser, J. & Stark, A. R. (1979) Proc. Natl. Acad. Sci. peptides that reacted with antibody to DHFR were the same USA 76,3116-3120. size as the mouse DHFR protein, whereas peptides that reacted 14. Sutcliffe,J. G. (1978) Proc. Natl. Acad. Sci. USA 75,3737-3741. with the 3-lactamase antibody showed different lengths con- 15. Shine, J. & Dalgarno, L. (1974) Proc. Natl. Acad. Sci. USA 71, sistent with out-of-frame read-through of the /-lactamase 1342-1346. peptide into the DHFR gene segment. Taken together, these 16. Shine, J. & Dalgarno, L. (1975) Nature (London) 254, 34-38. observations make it exceedingly unlikely that frame slippage 17. Steitz, J. A. (1978) in Biological Regulation and Development, of the ed. Goldberger, T. (Plenum, New York), Vol. 1, pp. 349-399. of translational reading can account for the production 18. Dunn, J. J., Buzash-Pollert, E. & Studier, F. W. (1978) Proc. Natl. biologically active DHFR peptide that we have observed. Acad. Sci. USA 75, 2741-2745. If the DHFR protein made in these bacterial cells were a 19. Tinoco, I., Jr., Uhlenbeck, O. C. & Levine, M. D. (1971) Nature cleavage product processed from a fused 3-lactamase-DHFR (London) 230,362-367. fused protein, the amount of DHFR synthesized would be 20. Tinoco, I., Jr., Borer, P. N., Dengler, B., Levine, M. D., Uhlen- similar for all clones that contain a cDNA insert in orientation beck, 0. C., Crothers, D. M. & Gralla, J. (1974) Nature (London) would occur in each 246, 40-41. A, because transcription and translation 21. Nunberg, J. H., Kaufman, R. J., Chang, A. C. Y., Cohen, S. N. & case from the same 3-lactamase regulatory signals. Our data Schimke, R. T. (1980) Cell 19, 154-166. indicate that different amounts of DHFR are made in the 22. Platt, T., Weber, K., Ganem, D. & Miller, J. H. (1972) Proc. Natl. various clones and that DHFR synthesis is influenced by se- Acad. Sci. USA 69, 897-901. quence differences in the mRNA segment immediately pre- 23. Manley, J. L. (1978) J. Mol. Biol. 125, 449-466. ceding the translational start codon. Production of a biologically 24. Goeddel, D. V., Heynecker, H. L., Hozumi, T., Aventzen, R., their Itakura, K., Yansura, D. G., Ross, M. J., Miozzari, G., Crea, R. & and immunologically active DHFR by plasmids having Seeburg, P. H. (1979) Nature (London) 281, 544-548. cDNA insert in orientation B (Table 1 and Fig. 3) further sup- 25. Roberts, T. M., Bikel, I., Yocum, R. R., Livingston, D. M. & ports the conclusion that the DHFR is not part of a fused pep- Ptashne, M. (1979) Proc. Natl. Acad. Sci. USA 76,5596-5600. tide; no protein is known to be made from the distal end of the 26. Stone, D., Phillips, A. W. (1977) FEBS Lett. 74, 85-87. Downloaded by guest on September 26, 2021