Simplified gene synthesis: A one-step approach to ... - Semantic Scholar
Journal of Biotechnology 124 (2006) 496–503
Simplified gene synthesis: A one-step approach to PCR-based gene construction Gang Wu ∗,1 , Julie B. Wolf 1 , Ameer F. Ibrahim, Stephanie Vadasz, Muditha Gunasinghe, Stephen J. Freeland Applied Molecular Biology Laboratory, Department of Biological Sciences, University of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250, USA Received 4 November 2005; received in revised form 23 December 2005; accepted 13 January 2006
Abstract PCR-based gene synthesis conventionally requires two steps: first, all overlapping oligonucleotides are assembled by selfpriming; then an additional pair of primers is used to amplify the full-length gene product. Here we propose a simplified method of gene synthesis which combines these two steps into one. We have found that the efficiency of this one-step method, which we term “Simplified Gene Synthesis”, is affected by multiple parameters of the PCR reactions. In particular, the choice of polymerase is critical for successful one-step assembly. Other important factors include the concentration of assembly oligonucleotides and amplification primers. Moreover, we offer a general method to estimate, given a known mutation rate, how many clones should be sequenced in order to be confident of obtaining at least one correct gene product. Having determined the accuracy of gene products synthesized under optimal conditions with Simplified Gene Synthesis, we show that our estimation works well. Overall, the simplified gene synthesis provides an easier and more efficient approach to gene synthesis, providing a further step towards the future goal of generalized automation for this process. © 2006 Elsevier B.V. All rights reserved. Keywords: Gene synthesis; PCR-based; KOD polymerase; Optimal concentration; Accuracy
1. Introduction
Abbreviations: SGS, simplified gene synthesis; AO, assembly oligonucleotides; AP, amplification primer; GC, guanine and cytosine ∗ Corresponding author. Tel.: +1 410 455 2279; fax: +1 410 455 3875. E-mail address: [email protected] (G. Wu). 1 Gang Wu and Julie B. Wolf contributed equally to this work. 0168-1656/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jbiotec.2006.01.015
With recent advances in molecular biology techniques, constructing genes synthetically has become a time and cost-efficient alternative to conventional PCR. An obvious but significant advantage of synthetic gene construction is that no template DNA is required. This is particularly helpful if the organism of interest has sequence information readily available
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
Fig. 1. Schematic diagram compares traditional two-step assembly PCR (A) with our one-step SGS (B) method. Each assembly oligonucleotide was represented as an arrow. The amplification primers are denoted as dotted arrows. Solid lines stand for the two complementary strands of a full-length gene.
497
PCR reaction, calling the technique “Simplified Gene Synthesis” (SGS, Fig. 1B). To illustrate the utility of this method, we systematically explored various factors that affect the efficiency of gene assembly, including the use of various polymerases, and the concentrations of both the assembly and amplification oligonucleotides. To demonstrate the versatility of this method, we synthesized three DNA fragments containing whole or partial genes for the purpose of one-step epitope-tagging and mutagenesis (Gene A), the precise addition of a cell localization signal (Gene B), as well as codon usage optimization (Gene C).
2. Materials and methods (e.g. in Genbank) but is itself difficult to obtain or maintain. Moreover, gene synthesis facilitates extensive modification of nucleotide sequences that may be required to improve expression, localization, detection or purification in heterologous systems. For example, the technique has been used in mutagenesis (Hayashi et al., 1994), codon optimization (Graham et al., 1993), the construction of ancestral genes (Chang et al., 2002), gene delivery vectors (Schaffer and Lauffenburger, 2000) and even construction of an entire genome (Yount et al., 2000). A synthetic gene can be constructed by assembling multiple oligonucleotides via one of several methods including: ligation (Heyneker et al., 1976; Itakura et al., 1977; Ivanov et al., 1990; Jayaraman et al., 1991; Au et al., 1998; Borodina et al., 2003), the ForkI method (Hayden and Mandecki, 1988), serial cloning of oligonucleotides (Hayden and Mandecki, 1988), or PCR extension (Dillon and Rosen, 1990; Ciccarelli et al., 1991; Prodromou and Pearl, 1992; Chen et al., 1993; Graham et al., 1993; Hayashi et al., 1994; Vallejo et al., 1994; Stemmer et al., 1995; Xiong et al., 2004). Among these methods, PCR appears to be the most efficient and cost-effective (Withers-Martinez et al., 1999). Conventionally, PCR-based gene synthesis uses a two-step approach (Fig. 1A). The first step consists of the PCR assembly and extension of overlapping oligonucleotides by self-priming. The second step is amplification of the full-length product, which is achieved by the addition of two outside amplification primers in another round of PCR (e.g. Stemmer et al., 1995). In this study, we have simplified this method by combining these two steps into a single
2.1. Genes and oligonucleotide design Three genes of different lengths (209, 777, and 936 bp) were assembled in this study (sequences available at http://www.evolvingcode.net/wug1/sgs.xls). Gene synthesis product A is the Volvox carteri nitA (GenBank accession no: X64136) gene, encoding nitrate dehydrogenase. Gene synthesis was used to add an epitope tag to the amino terminal end of the expressed gene and at the same time a single-base substitution was made in the protein coding region. This synthesis product was 777 bp in length. Gene product B (209 bp) is a portion of a cDNA of a novel plant protease (Jaenicke and Waffenschmidt, 1979) that was cloned into a baculoviral expression vector. Gene synthesis was used to generate missing 5 cDNA sequence with the precise addition of a signal sequence necessary for heterologous protein secretion in insect cells. Gene C (936 bp) is the Renilla reniformis luciferase gene (GenBank accession no: M63501) with codons optimized for expression in E. coli. The design of the synthetic assembly oligonucleotides was similar to that of Stemmer’s method (Stemmer et al., 1995, Fig. 1A) in that the length of each oligonucleotide was 40 bases and the overlapping region was 18–20 bases. Thirty-eight oligonucleotides were designed for Gene A, 10 for Gene B, and 46 for Gene C. In addition, two outer amplification primers containing different restriction sites were designed for each gene to facilitate cloning (all oligonucleotide sequences are available at http://www.evolvingcode.net/wug1/sgs.xls).
498
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
2.2. SGS procedure Oligonucleotides were purchased from IDT (Coralville, IA) or Invitrogen (Carlsbad, CA) using the smallest scale available with standard desalting purification only. Each oligonucleotide was reconstituted to 100 M in 10 mM Tris–HCl (pH 8.5). A gene assembly mix was prepared by combining 5 l of each of the gene assembly oligonucleotides into a solution, which was then diluted to achieve a final concentration of each oligonucleotide in the assembly mix at 1 M. Using this mix, a serial dilution was made to generate assembly mixes with the final concentration of each oligonucleotide of 500, 250, 100 and 50 nM in Tris–HCl (pH 8.5). Five microliters of each mix were used in a 50 l reaction hence the final concentrations of each oligonucleotide in the PCR reaction mixture were 100, 50, 25, 10 and 5 nM, respectively. The amplification primers were used at a final concentration of 0.4 M except as indicated in the amplification primer optimization experiment. All PCR reactions were performed on a Perkin-Elmer 2400. The one-step PCR reaction conditions and cycling programs followed the manufacturer’s recommended conditions. Four different polymerases were tested: Taq (Fisher), Pfu Turbo (Stratagene), KOD XL (Novagen), and KOD HiFi (Novagen). For each reaction, 0.2 mM dNTPs was used (except Pfu Turbo, which was 0.4 mM); 2.5 unit of enzyme were used for Pfu Turbo and KOD XL, 1 unit for KOD HiFi, 5 unit for Taq, and 1 mM MgCl2 was added to KOD HiFi reactions. Each PCR reaction was carried out in 25 cycles of (i) denaturing at 94 ◦ C for 30 s for Taq and KOD XL, 95 ◦ C for 30 s for Pfu Turbo, and 98 ◦ C for 15 s for KOD HiFi; (ii) annealing at 52 ◦ C for 30 s for Taq and Pfu Turbo, and 52 ◦ C for 5 s and 2 s for KOD XL and KOD HiFi, respectively; (iii) extension at 72 ◦ C for 60 s for Taq and Pfu Turbo, and 72 ◦ C for 30 s and 20 s for KOD XL and KOD HiFi, respectively. 2.3. Cloning and sequencing Each of the three gene fragments was synthesized, cloned into different plasmid vectors and then the identity and integrity of each was confirmed by DNA sequence analysis. Additional optimization and accuracy assessments were conducted using Gene A.
Gene A PCR products amplified with KOD HiFi and KOD XL were purified on a PCR purification column (Qiagen) and digested with HindIII and BamHI. The 664 bp digested PCR products were cloned into pUC19 vector, transformed into E. coli DH5␣, and the DNA sequence of randomly selected clones was determined using BigDye Version 3.1 on an ABI 3100 DNA sequencer (Applied BioSystems). The mutation frequency was analyzed in SPSS (SPSS Inc., Chicago IL) with appropriate statistical tests. 2.4. Estimation of the number of clones required for at least an accurate gene product Given the mutation rate (p) per nucleotide associated with a particular DNA polymerase, the number of mutations (X) in a sequence of length n will follow a binomial distribution, i.e., X ∼ Bin(n, p). Then the probability (p ) of a sequence that has r mutations will be n! p (X = r) = (1) pr (1 − p)n−r r!(n − r)! For the special case of the probability of an errorfree sequence (i.e. X = 0), this simplifies to: n! p (X = 0) = p0 (1 − p)n−0 = (1 − p)n 0!(n − 0)! (2) Now, the number of sequences (Y) that are error free in a sample of N clones will similarly follow a binomial distribution, i.e., Y ∼ Bin(N, p ). Thus probability that a sample of size N will contain at least 1 error-free clone is given by: P(Y > 0) = 1 − P(Y = 0) where
P(Y = 0) =
N! 0!(N − 0)!
(3)
N−0
p (1 − p ) 0
N
= (1 − p )
(4)
Thus, if we want to achieve a 95% probability of obtaining at least 1 correct clone (i.e., P(Y > 0) = 0.95), then we must find N such that: N
P(Y > 0) = 1 − (1 − p ) = 0.95
(5)
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
499
Reorganizing Eq. (5), N=
log (0.05) log (1 − p )
(6)
Now Eqs. (1)–(6) assume that the mutation rate is equal for all four nucleotides. If the mutation rate varies for different bases then Eq. (2) requires modification that will be increasingly important as the gene under scrutiny increases in GC bias. The adjustment is simple: let us define the mutation rate of GC and AT as µs and µw , respectively. Thus, if the number of GC and AT in the gene is ns and nw , respectively, then the probability (p ) that a full-length gene is error-free transforms from Eq. (1) into: p (X = 0) = (1 − µs )ns × (1 − µw )nw
(7)
Then, substituting this revised formula for p into Eq. (6), we may once again calculate the sample size N of clones that will offer a 95% probability of containing at least one error free clone. At present, this mathematical model assumes a uniform distribution of mutation rate within a gene. This method can be easily adjusted for a non-uniform mutation rate at particular regions such as gap regions, but the empirical basis for any such adjustments remains a topic for future research.
3. Results and discussion 3.1. Performance of various thermostable DNA polymerases in creating synthetic genes Currently, Taq and Pfu polymerases or a mixture of both are commonly used in the two-step PCR-based gene assembly (Dillon and Rosen, 1990; Ciccarelli et al., 1991; Prodromou and Pearl, 1992; Chen et al., 1993; Graham et al., 1993; Hayashi et al., 1994; Stemmer et al., 1995; Xiong et al., 2004; Young and Dong, 2004). However, Taq polymerase lacks a 3 –5 exonuclease activity and is known to be error-prone (Tindall and Kunkel, 1988). Pfu polymerase is more accurate than Taq polymerase but has a slower elongation rate (Takagi et al., 1997). In our experiments, no obvious full-length gene product was obtained with either Taq or Pfu for Gene A with the SGS method (Fig. 2). Likewise, there were no full-length gene products observed in similar experiments with the 936 bp
Fig. 2. KOD polymerases perform better in SGS than Taq or Pfu. Four DNA polymerase (Taq, Pfu, KOD HiFi and KOD XL) were used to amplify the nitA gene product (Gene A). Five concentrations of assembly oligonucleotides were used in a standard synthesis reaction (lane A: 100 nM; lane B: 50 nM; lane C: 25 nM; lane D: 10 nM; lane E: 5 nM). Ten percent of each reaction was analyzed on a 1% agarose gel. The arrow indicates the expected gene product of 777 bp.
Gene C (data not shown). In contrast, the full-length 777 bp Gene A fragment was the primary product of the assembly/amplification reaction using KOD XL and KOD HiFi polymerases at several concentrations of oligonucleotides. Although the two KOD polymerases gave slightly higher background, this did not affect the purification and the cloning of the synthetic gene products. Therefore, the SGS appears to be very efficient with KOD polymerases. For genes up to 1 kb, the synthesis can be completed in less than 1 h in most thermal cyclers. 3.2. Critical assembly oligonucleotide concentrations in SGS Given that KOD polymerases out-perform other enzymes tested in SGS, we have systematically explored the optimal PCR conditions for this procedure. Besides the choice of DNA polymerases, the most critical parameter appeared to be the concentrations of the overlapping assembly oligonucleotides spanning the full-length gene. Traditionally, equal molar amounts of each assembly oligonucleotide were
500
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
3.3. Effect of amplification primer concentration on efficiency of gene synthesis
Fig. 3. The assembly oligonucleotide concentrations are crucial. The performance of KOD XL polymerase was further tested with two genes using different concentrations of assembly oligonucleotides (lanes A: 100 nM, lanes B: 50 nM, lanes C: 25 nM, lanes D: 10 nM, lanes E: 5 nM, and lanes F: 1 nM). Marker lanes M1 are the 100 bp ladder (Fermentas, Hanover, MD) and M2 are the 1 kb ladder (Fermentas). Arrows illustrate the predicted size gene products: Gene B is 209 bp and Gene C is 936 bp.
In our one-step SGS method, amplification primers are mixed with assembly oligonucleotides in a single PCR reaction; hence a potentially large fraction of amplification primers are consumed in the initial cycles as the gene assembles. Therefore, to determine whether excess amplification primers were necessary for optimal product yield, we used the optimal assembly oligonucleotide concentration determined in previous experiments and changed the ratio of amplification primer (AP) concentration to assembly oligonucleotide (AO) concentration in a range from 1:1 to 80:1 in the Gene A assembly. Indeed, amplification primers should be at least 10-fold more concentrated than assembly oligonucleotides to get an optimal yield (Fig. 4). 3.4. The effect of annealing temperatures on the efficiency of SGS
combined as the gene assembly mix (e.g. WithersMartinez et al., 1999). However, the final concentration of each oligonucleotide in the mix depends on the total number of oligonucleotides used for the assembly, usually a function of the gene length. Therefore, to determine the effects of oligonucleotide concentrations on gene assembly efficiency of each of our three fragment assemblies, we standardized the initial concentration of each oligonucleotide to 1 M. Using serial dilutions made from this assembly mix, we found the optimal concentration of each assembly oligonucleotide to be 10–25 nM in the final PCR reaction (Fig. 2 and Fig. 3). For the gene fragments larger than 700 bp, little if any full-length product was observed for concentrations outside of this range (Fig. 3, Gene C). However, shorter assembly fragments were less sensitive to oligonucleotide concentrations although optimum yields were seen for those reactions with oligonucleotide concentrations within this range (Fig. 3, Gene B). Interestingly, we have also observed the critical effects of assembly oligonucleotide concentration on the efficiency of gene synthesis in the traditional two-step assembly PCR (data not shown), suggesting that this might be a general rule. To our knowledge, this is the first time that the existence of an optimal concentration of assembly oligonucleotides in PCR-based gene synthesis has been observed.
A unique feature of gene assembly is that multiple oligonucleotides are self-priming during the annealing stages. Accordingly, previous investigators focused on oligonucleotide design with the length of each oligonucleotide and/or the length of the complementary region of oligonucleotide pairs varying in length depending on GC content, Tm or other factors deemed essential for optimal assembly (Hoover and Lubkowski, 2002). In fact, a computer program called “DNAWorks” has been written for this purpose (Hoover and Lubkowski, 2002). However, in designing the assembly oligonucleotides for this study, no attempt was made to optimize this aspect of the experiment. In general, the gene sequence needed was divided into 40mers with 20 bases of overlap for oligonucleotides on the complementary strand. Then, a compromise
Fig. 4. The ratios of amplification primers to assembly oligonucleotides are critical for successful gene synthesis. The concentrations of each assembly oligonucleotides (AO) for the 777 bp Gene A product were used at 10 nM in the final PCR reaction while the concentration of amplification primers (AP) was varied, generating AP:AO ratios of 1:1, 5:1, 10:1, 20:1, 40:1 and 80:1.
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
Fig. 5. The annealing temperature does not have a significant effect on the efficiency of gene synthesis when other parameters had been optimized. The concentrations of AP and AO were the same as the lane 40:1 in Fig. 4. The annealing temperatures for each reaction are as below: A, 48.0 ◦ C; B, 50.3 ◦ C; C, 54.0 ◦ C; D, 56.3 ◦ C; E, 58.3 ◦ C; F, 60.9 ◦ C; G, 62.0 ◦ C.
annealing temperature of 52 ◦ C was chosen which was suitable for all assembly/amplification reactions. Furthermore, when the efficiency of amplification of Gene A was tested using a gradient of annealing temperatures, the results suggest that amplification occurred at all annealing temperatures used, from 48 to 62 ◦ C, although the efficiency appeared to decline slightly when the temperature used was higher than 56 ◦ C (Fig. 5). Taken together, these results indicated that the effect of annealing temperature on the SGS efficiency was negligible if other parameters (such as AP and AO concentrations) had been optimized. 3.5. Accuracy of gene assembly products We have also compared the accuracy of the genes constructed by KOD XL and KOD HiFi polymerases. Seventeen clones of Gene A that were generated by KOD HiFi polymerase were picked at random and the DNA sequence of the inserts was determined. In total, 30 mutations were found, with a mutation frequency of 2.7 (standard deviation: 1.6) per thousand bases. This is lower than the mutation frequency (3.5/kb) reported by the manufacturer (http://www.emdbiosciences. com/docs/docs/PROT/TB320.pdf). Interestingly, 80% of the mutations (24/30) were deletions, the vast majority of which were single-base deletions. The remaining 16.7% of mutations were single-base substitutions, and 3.3% were insertions. About three-quarters of the mutation sites involved a G/C base pair. Compared with KOD HiFi polymerase, KOD XL polymerase, a mixture of a high fidelity and an error-prone polymerase, gave a slightly higher mutation frequency of 3.7 (standard deviation: 2.1) per thousand bases for the 16 clones randomly chosen. Only 41% of these mutations were deletions (16/39) and 54% were substitutions (21/39). Again, G/C base pairs comprised
501
roughly three-quarters of the total mutations. Although the error rate for KOD XL was only slightly higher than with KOD HiFi (t = 1.592, one-tailed p-value = 0.059), none of the sixteen KOD XL clones was error-free whereas three of the KOD HiFi clones had a perfect sequence. Therefore, KOD HiFi is recommended as an ideal enzyme for our SGS method. 3.6. Estimation of minimal clone number for an accurate gene product Currently, although sequencing is a necessary step to ensure that an accurate gene product has been obtained, it is common to arbitrarily determine how many clones are to be sequenced. However, given the mutation frequency, it is straightforward to make a statistical prediction of the likely number of clones that must be sequenced in order to get at least one errorfree gene (Eqs. (1)–(7)). As might be expected for any “confidence-interval” type of calculation, the number of clones required increases exponentially as the gene length increases (Fig. 6). However, this effect is reduced for low mutation rates, as the relevant part of the exponential function approximates a straight line. We illustrate the estimation method with our experimental data below. If we assume that the mutation rate of each nucleotide is the same for both GC and AT base pairs (p = 0.0027 for KOD HiFi) and independent, then the probability of each 664 bp gene being error-free is
Fig. 6. The effects of gene length and mutation rate on the number of clones required to obtain at least one correct clone. A uniform mutation rate is assumed at every position within a gene.
502
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
(1 − 0.0027)664 = 0.166 (Eq. (2)). This is very close to what we observed in this study (3/17 = 0.176). According to Eq. (6), if we want a 0.95 probability of obtaining at least one correct gene, then the number of clones we should sequence is log(0.05)/log(1 − 0.166) = 16.5. Indeed, we have obtained three correct clones out of 17 clones sequenced for genes assembled with KOD HiFi. In contrast, the mutation rate for KOD XL is 0.0037. The probability of a 664 bp gene synthesized with KOD XL being correct is (1 − 0.0037)664 = 0.085 (Eq. (2)). Accordingly, the number of clones required to be 95% sure of obtaining at least one correct gene product is log(0.05)/log(1 − 0.085) = 33.7 (Eq. (6)). Therefore, it is not surprising that we found no correct genes in the 16 clones sequenced. In addition, this method has also proved to be valid in the case of Gene B and Gene C (data not shown). Many genes show extensive GC bias. As we found in this study, the mutation rates for GC and AT can vary considerably (µs = 0.0035 and µw = 0.0016 for KOD HiFi). Therefore it is appropriate to move to the adjusted formulae for estimating errors. For instance, the GC composition of Gene A is 55.6% (369 bp GC and 295 bp AT). If we used the modified method, then the probability of a 664 bp gene being error-free is (1 − 0.0035)369 × (1 − 0.0016)295 = 0.171 (Eq. (7)), which is closer to observed value (0.176) than that calculated with previous method. Although this is only a small improvement, the differences in the predictions of the two methods would increase for a more strongly GC-biased gene.
cleotides used need no expensive purification or preparation; and the one-step assembly and amplification procedure can be used to efficiently and inexpensively create DNA fragments up to 1 kb. Larger genes may be subsequently created in blocks of 1 kb if necessary. Therefore, the SGS provides an easy and time-efficient approach to gene synthesis and can be used to create novel molecules for a variety of applications with a minimum investment in time, effort and reagents. Moreover, this one-step gene synthesis method makes the automation of this procedure possible in the future. It is important to emphasize that our method as reported here does not seek to optimize the accuracy of the gene synthesis product (aside from demonstrating that the choice of DNA polymerase can be critical). Rather our aim here is to show that the speed and cost of gene synthesis can be significantly improved without sacrificing accuracy significantly below the levels of “traditional” gene synthesis techniques. As with all PCR-based gene synthesis methods, questions relating to accuracy of synthetic gene product are likely to be influenced by many factors, such as error frequency in the oligonucleotides, fidelity of DNA polymerase, number of PCR cycles, composition of nucleotide sequences, presence of gaps and perhaps the gene length. As such, it is a clear avenue for future research to explore the strength and importance of each of these parameters in a systematic study.
3.7. Concluding remarks
Acknowledgements
As a summary of this study, we have demonstrated that our modification of the standard gene assembly method improves the time, cost-efficiency, and design of synthetic genes. Both KOD HiFi and KOD XL polymerases are ideal for this application although the accuracy of KOD HiFi polymerase is superior if absolute sequence fidelity is required. According to our study, the concentrations of the assembly oligonucleotides and the amplification primers are critical for optimal gene assembly, and the amplification primers must be in excess. Other PCR parameters could be optimized but standard conditions were suitable for all of the genes that we tested. In particular, oligonucleotide design requires no complicated planning; the oligonu-
This study is a project of UMBC AMB Master’s degree program and funded by UMBC DRIF grant and award 0317349 from the National Science Foundation’s Division of Biological Infrastructure, program in Biological Databases and Informatics. We are grateful to Dr. Lei Nie (Dept. of Math. and Stat., UMBC) for his suggestion on the statistical analysis of the sequence accuracy data and Dr. Weiwen Zhang for his valuable comments. We also want to thank Anna Raman and Dwayne Taliaferro for their help in the early stage of the luciferase gene assembly project. Finally, we would like to thank anonymous reviewers for their valuable comments that have significantly improved this manuscript.
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
References Au, L.C., Yang, F.Y., Yang, W.J., Lo, S.H., Kao, C.F., 1998. Gene synthesis by a LCR-based approach: high-level production of leptin-L54 using synthetic gene in Escherichia coli. Biochem. Biophys. Res. Commun. 248 (1), 200–203. Borodina, T.A., Lehrach, H., Soldatov, A.V., 2003. Ligationbased synthesis of oligonucleotides with block structure. Anal. Biochem. 318 (2), 309–313. Chang, B.S., Kazmi, M.A., Sakmar, T.P., 2002. Synthetic gene technology: applications to ancestral gene reconstruction and structure-function studies of receptors. Methods Enzymol. 343, 274–294. Chen, C., Mao, J., Zhang, M., Dai, J., 1993. Combination of DNA single strand synthesis with PCR to construct mung bean trypsin inhibitor gene. Chin. J. Biotechnol. 9 (1), 63–70. Ciccarelli, R.B., Gunyuzlu, P., Huang, J., Scott, C., Oakes, F.T., 1991. Construction of synthetic genes using PCR after automated DNA synthesis of their entire top and bottom strands. Nucleic Acids Res. 19 (21), 6007–6013. Dillon, P.J., Rosen, C.A., 1990. A rapid method for the construction of synthetic genes using the polymerase chain reaction. Biotechniques 9 (3), 298–300. Graham, R.W., Atkinson, T., Kilburn, D.G., Miller Jr., R.C., Warren, R.A., 1993. Rational design and PCR-based synthesis of an artificial Schizophyllum commune xylanase gene. Nucleic Acids Res. 21 (21), 4923–4928. Hayashi, N., Welschof, M., Zewe, M., Braunagel, M., Dubel, S., Breitling, F., Little, M., 1994. Simultaneous mutagenesis of antibody CDR regions by overlap extension and PCR. Biotechniques 17 (2), 314–315, pp. 310 and 312. Hayden, M.A., Mandecki, W., 1988. Gene synthesis by serial cloning of oligonucleotides. DNA 7 (8), 571–577. Heyneker, H.L., Shine, J., Goodman, H.M., Boyer, H.W., Rosenberg, J., Dickerson, R.E., Narang, S.A., Itakura, K., Lin, S., Riggs, A.D., 1976. Synthetic lac operator DNA is functional in vivo. Nature 263 (5580), 748–752. Hoover, D.M., Lubkowski, J., 2002. DNAWorks: an automated method for designing oligonucleotides for PCR-based gene synthesis. Nucleic Acids Res. 30 (10), e43. Itakura, K., Hirose, T., Crea, R., Riggs, A.D., Heyneker, H.L., Bolivar, F., Boyer, H.W., 1977. Expression in Escherichia coli of a
503
chemically synthesized gene for the hormone somatostatin. Science 198 (4321), 1056–1063. Ivanov, I., Christov, C., Alexciev, K., 1990. Two methods for rapid assembly and oligomerization of synthetic genes: construction of human calcitonin-encoding sequences. Gene 95 (2), 295–299. Jaenicke, L., Waffenschmidt, S., 1979. Matrix-lysis and release of daughter spheroids in Volvox carteri—a proteolytic process. FEBS Lett. 107 (1), 250–253. Jayaraman, K., Fingar, S.A., Shah, J., Fyles, J., 1991. Polymerase chain reaction-mediated gene synthesis: synthesis of a gene coding for isozyme c of horseradish peroxidase. Proc. Natl. Acad. Sci. USA 88 (10), 4084–4088. Prodromou, C., Pearl, L.H., 1992. Recursive PCR: a novel technique for total gene synthesis. Protein Eng. 5 (8), 827–829. Schaffer, D.V., Lauffenburger, D.A., 2000. Targeted synthetic gene delivery vectors. Curr. Opin. Mol. Ther. 2 (2), 155–161. Stemmer, W.P., Crameri, A., Ha, K.D., Brennan, T.M., Heyneker, H.L., 1995. Single-step assembly of a gene and entire plasmid from large numbers of oligodeoxyribonucleotides. Gene 164 (1), 49–53. Takagi, M., Nishioka, M., Kakihara, H., Kitabayashi, M., Inoue, H., Kawakami, B., Oka, M., Imanaka, T., 1997. Characterization of DNA polymerase from Pyrococcus sp. strain KOD1 and its application to PCR. Appl. Environ. Microbiol. 63 (11), 4504–4510. Tindall, K.R., Kunkel, T.A., 1988. Fidelity of DNA synthesis by the Thermus aquaticus DNA polymerase. Biochemistry 27 (16), 6008–6013. Vallejo, A.N., Pogulis, R.J., Pease, L.R., 1994. In vitro synthesis of novel genes: mutagenesis and recombination by PCR. PCR Methods Appl. 4 (3), S123–S130. Withers-Martinez, C., Carpenter, E.P., Hackett, F., Ely, B., Sajid, M., Grainger, M., Blackman, M.J., 1999. PCR-based gene synthesis as an efficient approach for expression of the A + T-rich malaria genome. Protein Eng. 12 (12), 1113–1120. Xiong, A.S., Yao, Q.H., Peng, R.H., Li, X., Fan, H.Q., Cheng, Z.M., Li, Y., 2004. A simple, rapid, high-fidelity and costeffective PCR-based two-step DNA synthesis method for long gene sequences. Nucleic Acids Res. 32 (12), e98. Young, L., Dong, Q., 2004. Two-step total gene synthesis method. Nucleic Acids Res. 32 (7), e59. Yount, B., Curtis, K.M., Baric, R.S., 2000. Strategy for systematic assembly of large RNA and DNA genomes: transmissible gastroenteritis virus model. J. Virol. 74 (22), 10600–10611.
Simplified gene synthesis: A one-step approach to PCR-based gene construction Gang Wu ∗,1 , Julie B. Wolf 1 , Ameer F. Ibrahim, Stephanie Vadasz, Muditha Gunasinghe, Stephen J. Freeland Applied Molecular Biology Laboratory, Department of Biological Sciences, University of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250, USA Received 4 November 2005; received in revised form 23 December 2005; accepted 13 January 2006
Abstract PCR-based gene synthesis conventionally requires two steps: first, all overlapping oligonucleotides are assembled by selfpriming; then an additional pair of primers is used to amplify the full-length gene product. Here we propose a simplified method of gene synthesis which combines these two steps into one. We have found that the efficiency of this one-step method, which we term “Simplified Gene Synthesis”, is affected by multiple parameters of the PCR reactions. In particular, the choice of polymerase is critical for successful one-step assembly. Other important factors include the concentration of assembly oligonucleotides and amplification primers. Moreover, we offer a general method to estimate, given a known mutation rate, how many clones should be sequenced in order to be confident of obtaining at least one correct gene product. Having determined the accuracy of gene products synthesized under optimal conditions with Simplified Gene Synthesis, we show that our estimation works well. Overall, the simplified gene synthesis provides an easier and more efficient approach to gene synthesis, providing a further step towards the future goal of generalized automation for this process. © 2006 Elsevier B.V. All rights reserved. Keywords: Gene synthesis; PCR-based; KOD polymerase; Optimal concentration; Accuracy
1. Introduction
Abbreviations: SGS, simplified gene synthesis; AO, assembly oligonucleotides; AP, amplification primer; GC, guanine and cytosine ∗ Corresponding author. Tel.: +1 410 455 2279; fax: +1 410 455 3875. E-mail address: [email protected] (G. Wu). 1 Gang Wu and Julie B. Wolf contributed equally to this work. 0168-1656/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jbiotec.2006.01.015
With recent advances in molecular biology techniques, constructing genes synthetically has become a time and cost-efficient alternative to conventional PCR. An obvious but significant advantage of synthetic gene construction is that no template DNA is required. This is particularly helpful if the organism of interest has sequence information readily available
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
Fig. 1. Schematic diagram compares traditional two-step assembly PCR (A) with our one-step SGS (B) method. Each assembly oligonucleotide was represented as an arrow. The amplification primers are denoted as dotted arrows. Solid lines stand for the two complementary strands of a full-length gene.
497
PCR reaction, calling the technique “Simplified Gene Synthesis” (SGS, Fig. 1B). To illustrate the utility of this method, we systematically explored various factors that affect the efficiency of gene assembly, including the use of various polymerases, and the concentrations of both the assembly and amplification oligonucleotides. To demonstrate the versatility of this method, we synthesized three DNA fragments containing whole or partial genes for the purpose of one-step epitope-tagging and mutagenesis (Gene A), the precise addition of a cell localization signal (Gene B), as well as codon usage optimization (Gene C).
2. Materials and methods (e.g. in Genbank) but is itself difficult to obtain or maintain. Moreover, gene synthesis facilitates extensive modification of nucleotide sequences that may be required to improve expression, localization, detection or purification in heterologous systems. For example, the technique has been used in mutagenesis (Hayashi et al., 1994), codon optimization (Graham et al., 1993), the construction of ancestral genes (Chang et al., 2002), gene delivery vectors (Schaffer and Lauffenburger, 2000) and even construction of an entire genome (Yount et al., 2000). A synthetic gene can be constructed by assembling multiple oligonucleotides via one of several methods including: ligation (Heyneker et al., 1976; Itakura et al., 1977; Ivanov et al., 1990; Jayaraman et al., 1991; Au et al., 1998; Borodina et al., 2003), the ForkI method (Hayden and Mandecki, 1988), serial cloning of oligonucleotides (Hayden and Mandecki, 1988), or PCR extension (Dillon and Rosen, 1990; Ciccarelli et al., 1991; Prodromou and Pearl, 1992; Chen et al., 1993; Graham et al., 1993; Hayashi et al., 1994; Vallejo et al., 1994; Stemmer et al., 1995; Xiong et al., 2004). Among these methods, PCR appears to be the most efficient and cost-effective (Withers-Martinez et al., 1999). Conventionally, PCR-based gene synthesis uses a two-step approach (Fig. 1A). The first step consists of the PCR assembly and extension of overlapping oligonucleotides by self-priming. The second step is amplification of the full-length product, which is achieved by the addition of two outside amplification primers in another round of PCR (e.g. Stemmer et al., 1995). In this study, we have simplified this method by combining these two steps into a single
2.1. Genes and oligonucleotide design Three genes of different lengths (209, 777, and 936 bp) were assembled in this study (sequences available at http://www.evolvingcode.net/wug1/sgs.xls). Gene synthesis product A is the Volvox carteri nitA (GenBank accession no: X64136) gene, encoding nitrate dehydrogenase. Gene synthesis was used to add an epitope tag to the amino terminal end of the expressed gene and at the same time a single-base substitution was made in the protein coding region. This synthesis product was 777 bp in length. Gene product B (209 bp) is a portion of a cDNA of a novel plant protease (Jaenicke and Waffenschmidt, 1979) that was cloned into a baculoviral expression vector. Gene synthesis was used to generate missing 5 cDNA sequence with the precise addition of a signal sequence necessary for heterologous protein secretion in insect cells. Gene C (936 bp) is the Renilla reniformis luciferase gene (GenBank accession no: M63501) with codons optimized for expression in E. coli. The design of the synthetic assembly oligonucleotides was similar to that of Stemmer’s method (Stemmer et al., 1995, Fig. 1A) in that the length of each oligonucleotide was 40 bases and the overlapping region was 18–20 bases. Thirty-eight oligonucleotides were designed for Gene A, 10 for Gene B, and 46 for Gene C. In addition, two outer amplification primers containing different restriction sites were designed for each gene to facilitate cloning (all oligonucleotide sequences are available at http://www.evolvingcode.net/wug1/sgs.xls).
498
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
2.2. SGS procedure Oligonucleotides were purchased from IDT (Coralville, IA) or Invitrogen (Carlsbad, CA) using the smallest scale available with standard desalting purification only. Each oligonucleotide was reconstituted to 100 M in 10 mM Tris–HCl (pH 8.5). A gene assembly mix was prepared by combining 5 l of each of the gene assembly oligonucleotides into a solution, which was then diluted to achieve a final concentration of each oligonucleotide in the assembly mix at 1 M. Using this mix, a serial dilution was made to generate assembly mixes with the final concentration of each oligonucleotide of 500, 250, 100 and 50 nM in Tris–HCl (pH 8.5). Five microliters of each mix were used in a 50 l reaction hence the final concentrations of each oligonucleotide in the PCR reaction mixture were 100, 50, 25, 10 and 5 nM, respectively. The amplification primers were used at a final concentration of 0.4 M except as indicated in the amplification primer optimization experiment. All PCR reactions were performed on a Perkin-Elmer 2400. The one-step PCR reaction conditions and cycling programs followed the manufacturer’s recommended conditions. Four different polymerases were tested: Taq (Fisher), Pfu Turbo (Stratagene), KOD XL (Novagen), and KOD HiFi (Novagen). For each reaction, 0.2 mM dNTPs was used (except Pfu Turbo, which was 0.4 mM); 2.5 unit of enzyme were used for Pfu Turbo and KOD XL, 1 unit for KOD HiFi, 5 unit for Taq, and 1 mM MgCl2 was added to KOD HiFi reactions. Each PCR reaction was carried out in 25 cycles of (i) denaturing at 94 ◦ C for 30 s for Taq and KOD XL, 95 ◦ C for 30 s for Pfu Turbo, and 98 ◦ C for 15 s for KOD HiFi; (ii) annealing at 52 ◦ C for 30 s for Taq and Pfu Turbo, and 52 ◦ C for 5 s and 2 s for KOD XL and KOD HiFi, respectively; (iii) extension at 72 ◦ C for 60 s for Taq and Pfu Turbo, and 72 ◦ C for 30 s and 20 s for KOD XL and KOD HiFi, respectively. 2.3. Cloning and sequencing Each of the three gene fragments was synthesized, cloned into different plasmid vectors and then the identity and integrity of each was confirmed by DNA sequence analysis. Additional optimization and accuracy assessments were conducted using Gene A.
Gene A PCR products amplified with KOD HiFi and KOD XL were purified on a PCR purification column (Qiagen) and digested with HindIII and BamHI. The 664 bp digested PCR products were cloned into pUC19 vector, transformed into E. coli DH5␣, and the DNA sequence of randomly selected clones was determined using BigDye Version 3.1 on an ABI 3100 DNA sequencer (Applied BioSystems). The mutation frequency was analyzed in SPSS (SPSS Inc., Chicago IL) with appropriate statistical tests. 2.4. Estimation of the number of clones required for at least an accurate gene product Given the mutation rate (p) per nucleotide associated with a particular DNA polymerase, the number of mutations (X) in a sequence of length n will follow a binomial distribution, i.e., X ∼ Bin(n, p). Then the probability (p ) of a sequence that has r mutations will be n! p (X = r) = (1) pr (1 − p)n−r r!(n − r)! For the special case of the probability of an errorfree sequence (i.e. X = 0), this simplifies to: n! p (X = 0) = p0 (1 − p)n−0 = (1 − p)n 0!(n − 0)! (2) Now, the number of sequences (Y) that are error free in a sample of N clones will similarly follow a binomial distribution, i.e., Y ∼ Bin(N, p ). Thus probability that a sample of size N will contain at least 1 error-free clone is given by: P(Y > 0) = 1 − P(Y = 0) where
P(Y = 0) =
N! 0!(N − 0)!
(3)
N−0
p (1 − p ) 0
N
= (1 − p )
(4)
Thus, if we want to achieve a 95% probability of obtaining at least 1 correct clone (i.e., P(Y > 0) = 0.95), then we must find N such that: N
P(Y > 0) = 1 − (1 − p ) = 0.95
(5)
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
499
Reorganizing Eq. (5), N=
log (0.05) log (1 − p )
(6)
Now Eqs. (1)–(6) assume that the mutation rate is equal for all four nucleotides. If the mutation rate varies for different bases then Eq. (2) requires modification that will be increasingly important as the gene under scrutiny increases in GC bias. The adjustment is simple: let us define the mutation rate of GC and AT as µs and µw , respectively. Thus, if the number of GC and AT in the gene is ns and nw , respectively, then the probability (p ) that a full-length gene is error-free transforms from Eq. (1) into: p (X = 0) = (1 − µs )ns × (1 − µw )nw
(7)
Then, substituting this revised formula for p into Eq. (6), we may once again calculate the sample size N of clones that will offer a 95% probability of containing at least one error free clone. At present, this mathematical model assumes a uniform distribution of mutation rate within a gene. This method can be easily adjusted for a non-uniform mutation rate at particular regions such as gap regions, but the empirical basis for any such adjustments remains a topic for future research.
3. Results and discussion 3.1. Performance of various thermostable DNA polymerases in creating synthetic genes Currently, Taq and Pfu polymerases or a mixture of both are commonly used in the two-step PCR-based gene assembly (Dillon and Rosen, 1990; Ciccarelli et al., 1991; Prodromou and Pearl, 1992; Chen et al., 1993; Graham et al., 1993; Hayashi et al., 1994; Stemmer et al., 1995; Xiong et al., 2004; Young and Dong, 2004). However, Taq polymerase lacks a 3 –5 exonuclease activity and is known to be error-prone (Tindall and Kunkel, 1988). Pfu polymerase is more accurate than Taq polymerase but has a slower elongation rate (Takagi et al., 1997). In our experiments, no obvious full-length gene product was obtained with either Taq or Pfu for Gene A with the SGS method (Fig. 2). Likewise, there were no full-length gene products observed in similar experiments with the 936 bp
Fig. 2. KOD polymerases perform better in SGS than Taq or Pfu. Four DNA polymerase (Taq, Pfu, KOD HiFi and KOD XL) were used to amplify the nitA gene product (Gene A). Five concentrations of assembly oligonucleotides were used in a standard synthesis reaction (lane A: 100 nM; lane B: 50 nM; lane C: 25 nM; lane D: 10 nM; lane E: 5 nM). Ten percent of each reaction was analyzed on a 1% agarose gel. The arrow indicates the expected gene product of 777 bp.
Gene C (data not shown). In contrast, the full-length 777 bp Gene A fragment was the primary product of the assembly/amplification reaction using KOD XL and KOD HiFi polymerases at several concentrations of oligonucleotides. Although the two KOD polymerases gave slightly higher background, this did not affect the purification and the cloning of the synthetic gene products. Therefore, the SGS appears to be very efficient with KOD polymerases. For genes up to 1 kb, the synthesis can be completed in less than 1 h in most thermal cyclers. 3.2. Critical assembly oligonucleotide concentrations in SGS Given that KOD polymerases out-perform other enzymes tested in SGS, we have systematically explored the optimal PCR conditions for this procedure. Besides the choice of DNA polymerases, the most critical parameter appeared to be the concentrations of the overlapping assembly oligonucleotides spanning the full-length gene. Traditionally, equal molar amounts of each assembly oligonucleotide were
500
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
3.3. Effect of amplification primer concentration on efficiency of gene synthesis
Fig. 3. The assembly oligonucleotide concentrations are crucial. The performance of KOD XL polymerase was further tested with two genes using different concentrations of assembly oligonucleotides (lanes A: 100 nM, lanes B: 50 nM, lanes C: 25 nM, lanes D: 10 nM, lanes E: 5 nM, and lanes F: 1 nM). Marker lanes M1 are the 100 bp ladder (Fermentas, Hanover, MD) and M2 are the 1 kb ladder (Fermentas). Arrows illustrate the predicted size gene products: Gene B is 209 bp and Gene C is 936 bp.
In our one-step SGS method, amplification primers are mixed with assembly oligonucleotides in a single PCR reaction; hence a potentially large fraction of amplification primers are consumed in the initial cycles as the gene assembles. Therefore, to determine whether excess amplification primers were necessary for optimal product yield, we used the optimal assembly oligonucleotide concentration determined in previous experiments and changed the ratio of amplification primer (AP) concentration to assembly oligonucleotide (AO) concentration in a range from 1:1 to 80:1 in the Gene A assembly. Indeed, amplification primers should be at least 10-fold more concentrated than assembly oligonucleotides to get an optimal yield (Fig. 4). 3.4. The effect of annealing temperatures on the efficiency of SGS
combined as the gene assembly mix (e.g. WithersMartinez et al., 1999). However, the final concentration of each oligonucleotide in the mix depends on the total number of oligonucleotides used for the assembly, usually a function of the gene length. Therefore, to determine the effects of oligonucleotide concentrations on gene assembly efficiency of each of our three fragment assemblies, we standardized the initial concentration of each oligonucleotide to 1 M. Using serial dilutions made from this assembly mix, we found the optimal concentration of each assembly oligonucleotide to be 10–25 nM in the final PCR reaction (Fig. 2 and Fig. 3). For the gene fragments larger than 700 bp, little if any full-length product was observed for concentrations outside of this range (Fig. 3, Gene C). However, shorter assembly fragments were less sensitive to oligonucleotide concentrations although optimum yields were seen for those reactions with oligonucleotide concentrations within this range (Fig. 3, Gene B). Interestingly, we have also observed the critical effects of assembly oligonucleotide concentration on the efficiency of gene synthesis in the traditional two-step assembly PCR (data not shown), suggesting that this might be a general rule. To our knowledge, this is the first time that the existence of an optimal concentration of assembly oligonucleotides in PCR-based gene synthesis has been observed.
A unique feature of gene assembly is that multiple oligonucleotides are self-priming during the annealing stages. Accordingly, previous investigators focused on oligonucleotide design with the length of each oligonucleotide and/or the length of the complementary region of oligonucleotide pairs varying in length depending on GC content, Tm or other factors deemed essential for optimal assembly (Hoover and Lubkowski, 2002). In fact, a computer program called “DNAWorks” has been written for this purpose (Hoover and Lubkowski, 2002). However, in designing the assembly oligonucleotides for this study, no attempt was made to optimize this aspect of the experiment. In general, the gene sequence needed was divided into 40mers with 20 bases of overlap for oligonucleotides on the complementary strand. Then, a compromise
Fig. 4. The ratios of amplification primers to assembly oligonucleotides are critical for successful gene synthesis. The concentrations of each assembly oligonucleotides (AO) for the 777 bp Gene A product were used at 10 nM in the final PCR reaction while the concentration of amplification primers (AP) was varied, generating AP:AO ratios of 1:1, 5:1, 10:1, 20:1, 40:1 and 80:1.
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
Fig. 5. The annealing temperature does not have a significant effect on the efficiency of gene synthesis when other parameters had been optimized. The concentrations of AP and AO were the same as the lane 40:1 in Fig. 4. The annealing temperatures for each reaction are as below: A, 48.0 ◦ C; B, 50.3 ◦ C; C, 54.0 ◦ C; D, 56.3 ◦ C; E, 58.3 ◦ C; F, 60.9 ◦ C; G, 62.0 ◦ C.
annealing temperature of 52 ◦ C was chosen which was suitable for all assembly/amplification reactions. Furthermore, when the efficiency of amplification of Gene A was tested using a gradient of annealing temperatures, the results suggest that amplification occurred at all annealing temperatures used, from 48 to 62 ◦ C, although the efficiency appeared to decline slightly when the temperature used was higher than 56 ◦ C (Fig. 5). Taken together, these results indicated that the effect of annealing temperature on the SGS efficiency was negligible if other parameters (such as AP and AO concentrations) had been optimized. 3.5. Accuracy of gene assembly products We have also compared the accuracy of the genes constructed by KOD XL and KOD HiFi polymerases. Seventeen clones of Gene A that were generated by KOD HiFi polymerase were picked at random and the DNA sequence of the inserts was determined. In total, 30 mutations were found, with a mutation frequency of 2.7 (standard deviation: 1.6) per thousand bases. This is lower than the mutation frequency (3.5/kb) reported by the manufacturer (http://www.emdbiosciences. com/docs/docs/PROT/TB320.pdf). Interestingly, 80% of the mutations (24/30) were deletions, the vast majority of which were single-base deletions. The remaining 16.7% of mutations were single-base substitutions, and 3.3% were insertions. About three-quarters of the mutation sites involved a G/C base pair. Compared with KOD HiFi polymerase, KOD XL polymerase, a mixture of a high fidelity and an error-prone polymerase, gave a slightly higher mutation frequency of 3.7 (standard deviation: 2.1) per thousand bases for the 16 clones randomly chosen. Only 41% of these mutations were deletions (16/39) and 54% were substitutions (21/39). Again, G/C base pairs comprised
501
roughly three-quarters of the total mutations. Although the error rate for KOD XL was only slightly higher than with KOD HiFi (t = 1.592, one-tailed p-value = 0.059), none of the sixteen KOD XL clones was error-free whereas three of the KOD HiFi clones had a perfect sequence. Therefore, KOD HiFi is recommended as an ideal enzyme for our SGS method. 3.6. Estimation of minimal clone number for an accurate gene product Currently, although sequencing is a necessary step to ensure that an accurate gene product has been obtained, it is common to arbitrarily determine how many clones are to be sequenced. However, given the mutation frequency, it is straightforward to make a statistical prediction of the likely number of clones that must be sequenced in order to get at least one errorfree gene (Eqs. (1)–(7)). As might be expected for any “confidence-interval” type of calculation, the number of clones required increases exponentially as the gene length increases (Fig. 6). However, this effect is reduced for low mutation rates, as the relevant part of the exponential function approximates a straight line. We illustrate the estimation method with our experimental data below. If we assume that the mutation rate of each nucleotide is the same for both GC and AT base pairs (p = 0.0027 for KOD HiFi) and independent, then the probability of each 664 bp gene being error-free is
Fig. 6. The effects of gene length and mutation rate on the number of clones required to obtain at least one correct clone. A uniform mutation rate is assumed at every position within a gene.
502
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
(1 − 0.0027)664 = 0.166 (Eq. (2)). This is very close to what we observed in this study (3/17 = 0.176). According to Eq. (6), if we want a 0.95 probability of obtaining at least one correct gene, then the number of clones we should sequence is log(0.05)/log(1 − 0.166) = 16.5. Indeed, we have obtained three correct clones out of 17 clones sequenced for genes assembled with KOD HiFi. In contrast, the mutation rate for KOD XL is 0.0037. The probability of a 664 bp gene synthesized with KOD XL being correct is (1 − 0.0037)664 = 0.085 (Eq. (2)). Accordingly, the number of clones required to be 95% sure of obtaining at least one correct gene product is log(0.05)/log(1 − 0.085) = 33.7 (Eq. (6)). Therefore, it is not surprising that we found no correct genes in the 16 clones sequenced. In addition, this method has also proved to be valid in the case of Gene B and Gene C (data not shown). Many genes show extensive GC bias. As we found in this study, the mutation rates for GC and AT can vary considerably (µs = 0.0035 and µw = 0.0016 for KOD HiFi). Therefore it is appropriate to move to the adjusted formulae for estimating errors. For instance, the GC composition of Gene A is 55.6% (369 bp GC and 295 bp AT). If we used the modified method, then the probability of a 664 bp gene being error-free is (1 − 0.0035)369 × (1 − 0.0016)295 = 0.171 (Eq. (7)), which is closer to observed value (0.176) than that calculated with previous method. Although this is only a small improvement, the differences in the predictions of the two methods would increase for a more strongly GC-biased gene.
cleotides used need no expensive purification or preparation; and the one-step assembly and amplification procedure can be used to efficiently and inexpensively create DNA fragments up to 1 kb. Larger genes may be subsequently created in blocks of 1 kb if necessary. Therefore, the SGS provides an easy and time-efficient approach to gene synthesis and can be used to create novel molecules for a variety of applications with a minimum investment in time, effort and reagents. Moreover, this one-step gene synthesis method makes the automation of this procedure possible in the future. It is important to emphasize that our method as reported here does not seek to optimize the accuracy of the gene synthesis product (aside from demonstrating that the choice of DNA polymerase can be critical). Rather our aim here is to show that the speed and cost of gene synthesis can be significantly improved without sacrificing accuracy significantly below the levels of “traditional” gene synthesis techniques. As with all PCR-based gene synthesis methods, questions relating to accuracy of synthetic gene product are likely to be influenced by many factors, such as error frequency in the oligonucleotides, fidelity of DNA polymerase, number of PCR cycles, composition of nucleotide sequences, presence of gaps and perhaps the gene length. As such, it is a clear avenue for future research to explore the strength and importance of each of these parameters in a systematic study.
3.7. Concluding remarks
Acknowledgements
As a summary of this study, we have demonstrated that our modification of the standard gene assembly method improves the time, cost-efficiency, and design of synthetic genes. Both KOD HiFi and KOD XL polymerases are ideal for this application although the accuracy of KOD HiFi polymerase is superior if absolute sequence fidelity is required. According to our study, the concentrations of the assembly oligonucleotides and the amplification primers are critical for optimal gene assembly, and the amplification primers must be in excess. Other PCR parameters could be optimized but standard conditions were suitable for all of the genes that we tested. In particular, oligonucleotide design requires no complicated planning; the oligonu-
This study is a project of UMBC AMB Master’s degree program and funded by UMBC DRIF grant and award 0317349 from the National Science Foundation’s Division of Biological Infrastructure, program in Biological Databases and Informatics. We are grateful to Dr. Lei Nie (Dept. of Math. and Stat., UMBC) for his suggestion on the statistical analysis of the sequence accuracy data and Dr. Weiwen Zhang for his valuable comments. We also want to thank Anna Raman and Dwayne Taliaferro for their help in the early stage of the luciferase gene assembly project. Finally, we would like to thank anonymous reviewers for their valuable comments that have significantly improved this manuscript.
G. Wu et al. / Journal of Biotechnology 124 (2006) 496–503
References Au, L.C., Yang, F.Y., Yang, W.J., Lo, S.H., Kao, C.F., 1998. Gene synthesis by a LCR-based approach: high-level production of leptin-L54 using synthetic gene in Escherichia coli. Biochem. Biophys. Res. Commun. 248 (1), 200–203. Borodina, T.A., Lehrach, H., Soldatov, A.V., 2003. Ligationbased synthesis of oligonucleotides with block structure. Anal. Biochem. 318 (2), 309–313. Chang, B.S., Kazmi, M.A., Sakmar, T.P., 2002. Synthetic gene technology: applications to ancestral gene reconstruction and structure-function studies of receptors. Methods Enzymol. 343, 274–294. Chen, C., Mao, J., Zhang, M., Dai, J., 1993. Combination of DNA single strand synthesis with PCR to construct mung bean trypsin inhibitor gene. Chin. J. Biotechnol. 9 (1), 63–70. Ciccarelli, R.B., Gunyuzlu, P., Huang, J., Scott, C., Oakes, F.T., 1991. Construction of synthetic genes using PCR after automated DNA synthesis of their entire top and bottom strands. Nucleic Acids Res. 19 (21), 6007–6013. Dillon, P.J., Rosen, C.A., 1990. A rapid method for the construction of synthetic genes using the polymerase chain reaction. Biotechniques 9 (3), 298–300. Graham, R.W., Atkinson, T., Kilburn, D.G., Miller Jr., R.C., Warren, R.A., 1993. Rational design and PCR-based synthesis of an artificial Schizophyllum commune xylanase gene. Nucleic Acids Res. 21 (21), 4923–4928. Hayashi, N., Welschof, M., Zewe, M., Braunagel, M., Dubel, S., Breitling, F., Little, M., 1994. Simultaneous mutagenesis of antibody CDR regions by overlap extension and PCR. Biotechniques 17 (2), 314–315, pp. 310 and 312. Hayden, M.A., Mandecki, W., 1988. Gene synthesis by serial cloning of oligonucleotides. DNA 7 (8), 571–577. Heyneker, H.L., Shine, J., Goodman, H.M., Boyer, H.W., Rosenberg, J., Dickerson, R.E., Narang, S.A., Itakura, K., Lin, S., Riggs, A.D., 1976. Synthetic lac operator DNA is functional in vivo. Nature 263 (5580), 748–752. Hoover, D.M., Lubkowski, J., 2002. DNAWorks: an automated method for designing oligonucleotides for PCR-based gene synthesis. Nucleic Acids Res. 30 (10), e43. Itakura, K., Hirose, T., Crea, R., Riggs, A.D., Heyneker, H.L., Bolivar, F., Boyer, H.W., 1977. Expression in Escherichia coli of a
503
chemically synthesized gene for the hormone somatostatin. Science 198 (4321), 1056–1063. Ivanov, I., Christov, C., Alexciev, K., 1990. Two methods for rapid assembly and oligomerization of synthetic genes: construction of human calcitonin-encoding sequences. Gene 95 (2), 295–299. Jaenicke, L., Waffenschmidt, S., 1979. Matrix-lysis and release of daughter spheroids in Volvox carteri—a proteolytic process. FEBS Lett. 107 (1), 250–253. Jayaraman, K., Fingar, S.A., Shah, J., Fyles, J., 1991. Polymerase chain reaction-mediated gene synthesis: synthesis of a gene coding for isozyme c of horseradish peroxidase. Proc. Natl. Acad. Sci. USA 88 (10), 4084–4088. Prodromou, C., Pearl, L.H., 1992. Recursive PCR: a novel technique for total gene synthesis. Protein Eng. 5 (8), 827–829. Schaffer, D.V., Lauffenburger, D.A., 2000. Targeted synthetic gene delivery vectors. Curr. Opin. Mol. Ther. 2 (2), 155–161. Stemmer, W.P., Crameri, A., Ha, K.D., Brennan, T.M., Heyneker, H.L., 1995. Single-step assembly of a gene and entire plasmid from large numbers of oligodeoxyribonucleotides. Gene 164 (1), 49–53. Takagi, M., Nishioka, M., Kakihara, H., Kitabayashi, M., Inoue, H., Kawakami, B., Oka, M., Imanaka, T., 1997. Characterization of DNA polymerase from Pyrococcus sp. strain KOD1 and its application to PCR. Appl. Environ. Microbiol. 63 (11), 4504–4510. Tindall, K.R., Kunkel, T.A., 1988. Fidelity of DNA synthesis by the Thermus aquaticus DNA polymerase. Biochemistry 27 (16), 6008–6013. Vallejo, A.N., Pogulis, R.J., Pease, L.R., 1994. In vitro synthesis of novel genes: mutagenesis and recombination by PCR. PCR Methods Appl. 4 (3), S123–S130. Withers-Martinez, C., Carpenter, E.P., Hackett, F., Ely, B., Sajid, M., Grainger, M., Blackman, M.J., 1999. PCR-based gene synthesis as an efficient approach for expression of the A + T-rich malaria genome. Protein Eng. 12 (12), 1113–1120. Xiong, A.S., Yao, Q.H., Peng, R.H., Li, X., Fan, H.Q., Cheng, Z.M., Li, Y., 2004. A simple, rapid, high-fidelity and costeffective PCR-based two-step DNA synthesis method for long gene sequences. Nucleic Acids Res. 32 (12), e98. Young, L., Dong, Q., 2004. Two-step total gene synthesis method. Nucleic Acids Res. 32 (7), e59. Yount, B., Curtis, K.M., Baric, R.S., 2000. Strategy for systematic assembly of large RNA and DNA genomes: transmissible gastroenteritis virus model. J. Virol. 74 (22), 10600–10611.
