ABSTRACT
Site-saturation mutagenesis, using degenerate oligonucleotide primers, is a frequently used method in introducing various mutations in a selected target codon. Oligonucleotides that are synthesized using equimolar concentrations of nucleoside phosphoramidites (dA, dC, dG, dT) in the positions to be saturated, result in a mutant population that is biased towards the original nucleotides. We found that this bias could be eliminated by modifying the concentrations of nucleoside phosphoramidites during the oligonucleotide synthesis. We synthesized eight degenerate oligonucleotides to saturate eight different codons, and sequenced a total of 344 mutagenized codons. In six of these eight oligonucleotides, we reduced to varying extents the concentrations of those nucleotides in the target positions that would form base pairs with the template. From the data, we analyzed the effects of different base compositions in the oligonucleotides when mutagenizing different codons, the influence of the positions of mismatches, and the significance of different non-Watson-Crick base pairs. Based on these results, we suggest levels to which different phosphoramidites should be reduced when synthesizing oligonucleotides for site-saturation mutagenesis.
Site-directed saturation mutagenesis is an efficient method to study a small, well conserved region of a protein, that is assumed or known to have a functional role. In site-saturation mutagenesis, the resulting population of mutated genes is expected to consist of otherwise identical genes, but a given codon should be random. In individual clones this codon can thus code for any amino acid, and the position is said to be saturated. The site-saturation principle gives a 2-fold advantage over the more commonly used simple substitution mutagenesis procedure. First, several different amino acid substitutions are gained by the same effort. Second, if desired, additional information can be obtained concerning the nature of acceptable substitutive amino acids.
Maximal randomness of codons is particularly important in approaches where direct phenotypic selection of the viable or functional gene products will be performed after transfecting the saturated DNA pool in appropriate cells. The actual codons in the saturated position of the products have to be sequenced, which is, in most cases, much more laborious than sequencing before transfection. When substitutions are as random as possible, a minimal number of sequences have to be determined. Of course, true randomization may be even more important when studying nucleic acid sequences themselves, e.g. sequence-specific binding or viral non-coding regions.
Several techniques are available for site-saturation mutagenesis. In cassette mutagenesis, the entire target area is synthesized, and then ligated between two restriction sites. Close to equal proportions of all four nucleotides can be ensured, at a given position of the oligonucleotide, simply by using 25% of each nucleoside phosphoramidite when synthesizing that position (1). A prerequisite for cassette mutagenesis is that there are two unique restriction sites very close to the target codon, or that such sites can be generated by site-directed mutagenesis.
To our knowledge, all other methods, based either on plasmid, bacteriophage or phagemid vectors, or polymerase chain reaction, involve a step where the mutagenic synthetic oligonucleotide is annealed to the target area. In site-saturation mutagenesis studies, the mutagenic oligonucleotide has been synthesized using equimolar concentrations of bases at the positions to be randomized (2-7). A problem arises here due to the differential efficiency of annealing: oligonucleotides containing up to three successive mismatching bases anneal less efficiently than those containing less or no mismatches, and the mutated population of clones will be biased in favor of the original sequence (8). The difference in annealing efficiency is most pronounced when the mutagenic oligonucleotide is relatively short (9). Significantly longer oligonucleotides should be able to reduce the bias, since there the mismatches have a relatively slighter effect on the melting temperature (Tm) and the association and dissociation constants (kass, kdiss). Mismatches in oligonucleotides seem to have a greater effect on kdiss than kass (9).
The bias towards the original sequence can be completely avoided by first deleting the codon (loop-out) and then re-introducing a random codon in the same place (loop-in) (8,10). This strategy has the advantages of cassette mutagenesis without needing the close-by restriction sites, but it requires two rounds of mutagenesis, and two oligonucleotides for each codon to be saturated. Wild-type background can also be avoided by introducing stop codons or restriction sites in the first round of mutagenesis, and replacing those with saturated codons in the second round (11), but these methods do not avoid the bias towards the original template sequence. In this paper we describe a one-round mutagenesis strategy that avoids the bias. It is based on modified nucleoside phosphoramidite concentrations during the synthesis of the mutagenic oligonucleotide. Furthermore, we provide data and statistics that reveal important features of annealing of degenerate oligonucleotides.
Mutagenesis was performed by primer extension reaction, using the Sculptor in vitro Mutagenesis Kit (Amersham, UK) with the M13mp18 vector (later referred to as M13 vector) and Escherichia coli TG1 strain. The method is a modification of Eckstein's phosphorothioate technique (12). The kit makes use of restriction enzyme NciI in nicking the non-mutant strands for subsequent exonuclease digestion. The mutagenic oligonucleotide can, thus, not include the recognition site for this enzyme (CCGGG, CCCGG). One of our eight oligonucleotides (VII) was capable of forming both the recognition sites with the random codons CGG and CCG, and these codons were thus expected to be absent in the products. Since codons for all the amino acids were still allowed, we made no changes in the procedure. Surprisingly, one of these codons, CGG, was found among the products. This is due to the nicking efficiency being <100%. If recognition sites for NciI can not be avoided by some of the mutagenic oligonucleotides and all the codons are needed, an alternative enzyme, or a different M13-based mutagenesis strategy may be used.
The construct that we used in mutagenesis, was an M13 vector with an insert containing nt 2440-2619 (180 bp) of the coxsackievirus A9 (CAV-9) genome. The CAV-9 cDNA in a pUBS vector (pUBS/CAV-9) was kindly provided by Dr G.Stanway (University of Essex, UK) (13,14). Annealing of the mutagenic oligonucleotide to the single-stranded template DNA was performed in 140 mM MOPS [3-(N-morpholino)propanesulfonic acid buffer], pH 8.0, containing 140 mM NaCl. The pH is critical, since at a low pH some mismatches become even more stable than a perfectly matching sequence (15). To optimize the efficiency of mutagenesis, the annealing mix was first incubated for 3 min in a +80°C waterbath, then moved to a +60°C waterbath, that was then allowed to cool to +37°C during 60 min.
In order to reliably analyze the results of a saturation mutagenesis study, one needs to achieve a large number of mutants in one mutagenesis procedure. Furthermore, the efficiency of mutagenesis, measured as the fraction of analyzed codons that contain mutations, has to be high. We found the Sculptor in vitro mutagenesis kit to be very efficient in both these terms: each mutagenesis resulted in 1200-2000 M13 plaques, each representing one mutant sequence out of the 64 possible ones, and the efficiency (plaques containing mutations / total number of plaques) was in all cases between 90 and 100%. The exact efficiency could not be calculated, since no means of recognizing the background non-mutated clones was introduced in the strategy. Theoretically, 1.6% (1/64) of the clones that have successfully gone through the mutagenesis procedure, should be wild-type.
The mutagenic oligonucleotides primed the synthesis of the sense strand of the insert. Oligonucleotides (synthesized at the Institute of Biotechnology, University of Helsinki) were 32-35 nt in length, flanking 13-17 nt at both sides of the mutagenic codon (Table 1). They were synthesized with an Applied Biosystems DNA Synthesizer, model 392. To synthesize the mutagenic positions, the four nucleoside phosphoramidites (Cruachem, Scotland) were mixed in a separate vial before synthesis, to ensure the designed proportions of nucleotides in the oligonucleotide. Biased proportions might result from incomplete mixing of the phosphoramidites in the first part of the activated amidite pulse, if on-line mixing was used (5). In all cases, the sum of phosphoramidite concentrations was 100 mM. Oligonucleotides were not purified after the synthesis, in order to maintain the composition of the stock (7). Replacing an amino acid with all other amino acids could be achieved by ensuring the existence of 32 different codons, since a codon of the form N-N-C/G allows all the amino acids to be synthesized (8), as also does a codon of the form N-N-A/G/T (16). However, to obtain more general information that can be used in all combinations of nucleotides, we allowed all nucleotides also in the third position.
Table 1.
In sequencing we used the Sequenase II sequencing kit (Amersham, UK) and the M13 -40 primer (GTTTTCCCAGTCACGAC) included in the kit.
First, we used equimolar concentrations of the four nucleoside phosphoramidites at the target codon and synthesized two oligonucleotides (I, II) to saturate codons CCT and GAG. After mutagenesis, we sequenced 73 clones to find out whether the resulting codons were random. The distribution of nucleotides was found to be heavily biased towards the original sequence: 40 and 56% of the nucleotides were not changed in the mutagenesis, when using oligonucleotides I and II, respectively. Therefore, in the synthesis of further six oligonucleotides (III-VIII), we reduced to varying extents the amounts of the `conservative' nucleotides that would form base pairs with the template strand. The data of these experiments are summarized in Table 2, and analyzed and discussed in the following.
Table 2.
When oligonucleotides uncorrected for A and T concentrations (I-IV) were used, the proportions of A and T were clearly dominating in their original (conservative) positions in the resulting mutant populations. The percentages were between 35 and 58%, rather than the expected 25%. When A and T were reduced at the conservative positions to half of the concentration of all other nucleotides (oligos V-VIII), the proportions of the original nucleotides in the products were much closer to the expected level (Table 2). In five of the six cases, they were between 17 and 32%. An exception was seen with oligonucleotide V: 51% of nucleotides at position 3 of the target codon remained A. In another attempt to saturate an identical codon, the resulting proportion of A was slightly smaller (46%), even though the concentration of A was not reduced in this oligonucleotide (III). Although these two codons are not directly comparable due to different flanking sequences, we suspect that, for an unknown reason, the concentration of A was not reduced to the designed level during the synthesis of oligonucleotide V. This assumption is supported by the observation that, after mutagenesis with oligonucleotide V, positions 1 and 2 were also dominated by A: they contained 31 and 41% A in positions that were originally G and C, respectively. The corresponding percentages were 18 and 28 after mutagenesis with oligonucleotide III.
The use of oligonucleotides with no modifications in the base concentrations (I, II) yielded mutant populations with an overwhelming majority of G and C in their conservative positions (50-68%), except in one case where C at the central position was well saturated (see section `Uneven distribution of double mutations'). Mutagenesis with oligonucleotides V-VIII, in which the proportions were reduced to 10% (C) and 6% (G), yielded good results: 16-30% of nucleotides G and C were conserved (Table 2).
The success of saturation in individual cases can be evaluated by comparing the amounts of conserved nucleotides and the other 3 nt in each position of the mutated codons. These results are given in Figure 1. Oligonucleotides I-IV resulted in a clear over-representation of the original nucleotides in positions 1 and 3, while position 2 was in a better balance. With oligonucleotide V, the bias towards A in all positions resulted in slightly reduced amounts of the original nucleotides in positions 1 and 2, while in position 3 there was a marked excess of the original A. Oligonucleotides VI-VIII yielded a good overall balance, although, in these results, the proportion of the original nucleotide in position 2 was reduced to an average level of 19%.
The most straightforward method to estimate whether a set of codons is random, is to check if the number of different codons is as expected. If the codons are random, the expected value for the number of different codons can be calculated using the formula E = 64 × [1 - (63/64)n], where E is the expected number of different codons and n is the number of codons sequenced. In this analysis, the number of different codons obtained using oligonucleotides II-V was clearly below the expected number, while that of oligonucleotides I, VI, VII and VIII was close to expected (Fig. 2).
For optimal randomization of codons, there should be no difference between the three different positions in the levels of randomness. In this respect, oligonucleotides giving rise to zero or three mutations do not need to be considered. In the case of single mutations, we assume that all positions are as likely to be mutated, and the number of single mutations achieved in this study was, as expected, too small to evaluate the matter. On the contrary, in cases where there are two mutations, the two mismatching base pairs can be either next to each other or there can be one matching base pair between them. Optimal results would be achieved if the unmutated nucleotide was as likely to be in any of the three positions. Of the analyzed 344 codons, 135 contained two mutations. Among these codons, 57 of the unmutated positions were found in position 1, 24 in position 2, and 54 in position 3. The [chi]2 probability of the distribution is 6.1 × 10-4, when tested against a random distribution. In the eight individual experiments, the numbers were small (due to using a subset of data), but the results were consistent in all of them: less than one third of the unmutated nucleotides were always in the central position.
A plausible explanation for the lower tendency of the central nucleotide to be conserved, when compared to the first and third nucleotide, would be the effect of neighbouring mismatches. If there is a matching base pair between two mismatches, the matching central base pair contributes little thermodynamic stability due to lack of base pair stacking. This is also demonstrated by nearest-neighbour analysis of thermodynamic data of DNA duplex stability (9,17). Thus, an oligonucleotide with mismatches only in positions 1 and 3 of the target codon, may anneal to the template almost as weakly as does an oligonucleotide with a mismatching trinucleotide. This is not a minor problem, since in 56% (36/64) of all possible codons, positions 1 and 3 are both mutated, and out of these, 25% (9/36) should have an unmutated central nucleotide.
In addition to normal Watson-Crick base pairing, some non-optimal base pairs are known to be stronger than others (17,18). To find out whether this has any significance in site-saturation mutagenesis, we analyzed all base pairs that were formed during the mutagenesis reactions. These numbers are not distorted by modifications in base concentrations, since only optimal base pairing was taken into account in the design of our oligonucleotides. We calculated the expected number of the different base pairs as 1/3 of the nucleotides that were not matching in each position, and compared these to the actual numbers. When analyzing all the mismatching bases, the results were very close to expectations (Fig. 3), suggesting that non-Watson-Crick base pairing does not need to be considered in site-saturation mutagenesis. This result also rules out the possibility that some nucleoside phosphoramidites would have been more reactive than others in the oligonucleotide synthesis. The stability parameters for different types of mismatches were, however, obtained when studying single mismatches (17,18). We thus analyzed separately the 83 single mutations and the 24 cases with two mutations separated by one matching base pair. These data sets both show figures that are close to expectations, but having a slight bias that may reflect the different stabilities of different base pairs. Notably, there were altogether 18 G.G base pairs [second strongest (17)] and 28 A.G base pairs [third strongest (17)] in these data sets, when 14 and 20 were expected, respectively. The weakest base pair (C.C) (17) was found six times, when 11 were expected. The remaining five base pair types, including the strongest one (G.T) (17), were found in numbers close to expectations (data not shown). We conclude that while the distinction between Watson-Crick and non-Watson-Crick base pairing is crucial in annealing of degenerate oligonucleotides, the different types of non-Watson-Crick base pairing do not need to be considered.
We thank Dr Glyn Stanway for the cloned CAV-9 genome, and Dr Lars Paulin for great flexibility in oligonucleotide synthesis. This work was supported by grants from the Academy of Finland and the Sigrid Juselius Foundation.
Nucleic Acids Research
Pages
Introduction
Materials And Methods
Vectors and mutagenesis
Mutagenic oligonucleotides
Sequencing
Results And Discussion
Mutagenesis with uncorrected or specifically tailored oligonucleotides
Results scored by individual nucleotides
Randomization of codons
Uneven distribution of double mutations
Non-Watson-Crick base pairing
Comments And Conclusions
Acknowledgements
References
Oligo
Original
codonSequence
nt
X
Y
Z
I
CCT
5'-CAATTCCGCGAGCGTAXYZGCACTCACTGCAGTTG-3'
35
25% A,C,G,T
25% A,C,G,T
25% A,C,G,T
II
GAG
5'-GCACTCACTGCAGTTXYZACAGGGCACACCTCG-3'
33
25% A,C,G,T
25% A,C,G,T
25% A,C,G,T
III
GCA
5'-GTACCTGCACTCACTXYZGTTGAGACAGGGCAC-3'
33
10% G, 30% A,C,T
10% C, 30% A,G,T
25% A,C,G,T
IV
ACC
5'-GTTGAGACAGGGCACXYZTCGCAAGTTACTCCAAG-3'
35
25% A,C,G,T
10% C, 30% A,G,T
10% C, 30% A,G,T
V
GCA
5'-CCGCGAGCGTACCTXYZCTCACTGCAGTTGAG-3'
32
6% G, 31% A,C,T
10% C, 30% A,G,T
13%A, 29% C,G,T
VI
CTC
5'-GCGAGCGTACCTGCAXYZACTGCAGTTGAGACAG-3'
34
10% C, 30% A,G,T
13%T, 29% A,C,G
10% C, 30% A,G,T
VII
ACT
5'-GCGTACCTGCACTCXYZGCAGTTGAGACAGGG-3'
32
13%A, 29% C,G,T
10% C, 30% A,G,T
13%T, 29% A,C,G
VIII
GTT
5'-CCTGCACTCACTGCAXYZGAGACAGGGCACAC-3'
32
6% G, 31% A,C,T
13%T, 29% A,C,G
13%T, 29% A,C,G
REFERENCES
This page is run by Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, as part of the OUP Journals
Comments and feedback: www-admin{at}oup.co.uk
Last modification: 6 Jan 1998
Copyright© Oxford University Press, 1998.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Airaksinen, M. Roivainen, and T. Hovi Coxsackievirus A9 VP1 Mutants with Enhanced or Hindered A Particle Formation and Decreased Infectivity J. Virol., January 15, 2001; 75(2): 952 - 960. [Abstract] [Full Text]
|
|
|
|
 |
|
 A. Airaksinen, M. Roivainen, G. Stanway, and T. Hovi Site-saturation mutagenesis of the PALTAVETG motif in coxsackievirus A9 capsid protein VP1 reveals evidence of conservation of a periodic hydrophobicity profile J. Gen. Virol., August 1, 1999; 80(8): 1919 - 1927. [Abstract] [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||