PEDS Advance Access originally published online on September 2, 2006
Protein Engineering Design and Selection 2006 19(11):483-489; doi:10.1093/protein/gzl034
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Directed evolution by accumulating tailored mutations: Thermostabilization of lactate oxidase with less trade-off with catalytic activity
1 Tsukuba Research Institute, Novartis Pharma K.K. Ohkubo 8, Tsukuba 300-2611, Japan 2 Department of Functional Materials Science, Saitama University Saitama 338-8570, Japan
3To whom correspondence should be addressed.E-mail: yasuhiko.shibanaka{at}novartis.com
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
We assumed that adverse effects posed by introducing multiple mutations could be decomposed into those of each of the component mutations and that the risk could be reduced by the accumulation of mutations that were finely tuned for directed improvement of a specific property. We propose here a directed evolution strategy for improving a specific property with less effect on other ones. This strategy is composed of fine-tuning of mutations and their accumulation by our original mutation-assembling method. In this study, we selected lactate oxidase (LOX) as a model enzyme, because its directed evolution had showed a trade-off between thermostability and catalytic activity. Mutation profiling at each of the sites found by error-prone PCR revealed a strong inverse relationship between the two properties. Thermostable mutations with less effect on catalytic activity were selected at each site and accumulated with ideal combinations by our method. The resultant multiple mutants exhibited 5- to 10-fold superior catalytic activity and comparable thermostability with those created by accumulating thermostable mutations, which were not tuned for catalytic activity. This result demonstrates that the accumulation of fine-tuned mutations is an advantageous approach to reduce the risk of adverse effects posed by accumulating multiple mutations.
Keywords: additivity principles/directed evolution/lactate oxidase/local fitness landscape/thermostability
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Directed evolution is a powerful method to create enzymes suited for industrial uses. We, however, often encounter a trade-off between a property to be improved and other ones, such as catalytic activity, which diminishes the value of created enzymes. There have been several reports demonstrating an improvement of a property independently of another one, e.g. thermostabilization without sacrificing catalytic activity (Giver et al., 1998
; Miyazaki et al., 2000; Wintrode et al., 2000). Although these researches successfully developed promising mutants of their target enzymes without losing other properties, they were stochastic approaches, which are highly dependent on the characteristic and potential of target enzymes. Methods applicable to a wide variety of enzymes are still not available. So far, we have theoretically demonstrated the near-additivity feature of advantageous effects of multiple mutations in local fitness landscapes (Aita et al., 2001, 2002). Based on the typical feature of local fitness landscapes, we have also demonstrated that accumulation of advantageous mutations based on the near-additivity is an approach suited for making a great leap in a protein property (Hamamatsu et al., 2005). According to this line of research, we hypothesized that adverse effects posed by introducing multiple mutations could be decomposed into those of each of its component mutations and that the risk could be reduced by the accumulation of mutations tuned for directed improvement of a specific property. Kuchner and Arnold (1997) also predicted that accumulation of mutations, which exclusively contribute to a certain property, would be one of the approaches for improving a specific property.
The lactate oxidase (LOX) from Aerococcus viridans is a member of the family of flavin mononucleotide-dependent 
-hydroxy acid oxidizing enzymes, catalyzing the oxidation of L-lactate to pyruvate with reduction of O2 to H2O2 (Maeda-Yorita et al., 1995). LOX has been widely utilized as a component of electrochemical sensor systems for monitoring lactate concentration in blood, because the blood level is important in monitoring shocks, respiratory insufficiency and heart failure: in sport medicine it is a very useful indicator for assessing the general physical condition of athletes (Mascini et al., 1985; Palleschi et al., 1990; Ito et al., 1995). The LOX-based enzyme sensors were commercially available. Prolonged stability of the sensors is, however, required for expanding its application. Several approaches have been reported to improve stability and/or sensitivity of the LOX sensors from the point of view of their composition, such as carrier polymers (Palmisano et al., 1994; Wei et al., 2003). Another way in which to achieve prolonged stability would be to increase the thermostability of the immobilized LOX, as reported by Minagawa et al. (1998). Our own study, as well as a report by Kaneko et al. (2005), however, showed that thermostabilization of LOX tended to pose deterioration of catalytic activity.
Here we present a strategy for improving a specific property with less trade-off with a critical one, such as catalytic activity, and the creation of thermostable LOX mutants with less deterioration of catalytic activity by this strategy. First, we discovered primary mutations at the sites found by error-prone PCR, which yet affected catalytic activity. Then amino acid substitutions at the sites were finely tuned by saturation mutagenesis to reduce the deteriorative effect on catalytic activity. After that, the fine-tuned mutations were accumulated by our original method, called ‘biased-mutation assembling’. This method is designed to accumulate as many advantageous mutations as possible via a recombination process similar to an overlap extension PCR method without epistatic effects, which, owing to unpredictable interference between mutations, diminishes the positive effect of each mutation when they are introduced concurrently. Finally, we succeeded in developing LOX multiple mutants, which exhibited superior catalytic activity and comparable thermostability with those with not tuned mutations.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Development of primary thermostable mutations by error-prone PCR
Complementary DNA of LOX derived from A.viridans IFO12219 was kindly provided by the Asahi Kasei Corporation, Japan (Maeda-Yorita et al., 1995). The cDNA was transferred to our expression vector and LOX was expressed under the control of the tac promoter in Escherichia coli. The LOX gene was subjected to error-prone PCR to introduce point mutations as previously described (Uchiyama et al., 2000), where experimental conditions were optimized so that only one or at most a few nucleotide substitutions were introduced (Leung et al., 1989). The amplified DNA fragments were ligated into the expression vector and the resultant vectors were used as a LOX random mutant library. The library was screened for mutants that showed improved thermostability as described below, and mutations introduced were determined as ‘primary mutations’ by DNA sequencing.
Screening for thermostable LOX mutants by an active-staining method
The JM109 strain of E.coli was transformed with a LOX mutant library and spread on a nitrocellulose membrane (Schleicher & Schuell) placed on TY agar medium containing 100 µg/ml ampicillin. Colonies forming on the membrane were printed onto a hydrophilized polyvinylidene fluoride membrane (Immobiron, Millipore). The replica membrane was incubated on TY agar medium containing 100 µg/ml ampicillin at 37°C for 16 h. Colonies forming on the membrane were lysed by soaking the membrane into 10 ml of lysis buffer consisting of 100 mM Tris–HCl (pH 8.0), 2 mM EDTA, 1 mg/ml lysozyme and 1% v/v Triton X-100, to make LOX mutants adsorbed onto the membrane. After blocking with 5% w/v skim milk in 50 mM phosphate buffer, the membrane was put into 20 mM phosphate buffer and incubated at 50 or 60°C for 30 min. To detect mutants that retained LOX activity after the heat treatment, the membrane was put into 7 ml of Konica immunostain HRP-1000 solution supplemented with 70 mM L(+)-lactic acid (Sigma) and 12.5 unit/ml horseradish peroxidase (Roche). After emerging clear blue spots with lactate-oxidation by LOX mutants, the membrane was put into distilled water to terminate the enzyme reaction. Transformants expressing LOX mutants with residual catalytic activity after the heat treatment were picked from the original nitrocellulose membrane and cultured in 1.2 ml of Luria broth medium containing 100 µg/ml ampicillin; the transformants were then examined for activity half-life at several temperatures.
LOX enzyme assay
Transformants expressing promising LOX mutants were cultured at 37°C for 16 h and then lysed in 200 µl of lysis buffer consisting of 50 mM Tris–HCl (pH 8.0), 2 mM EDTA and 1 mg/ml lysozyme; the crude lysates were then cleared by a filter plate (Corning). The oxidation activity of LOX contained in the lysate was determined by a chromogenic assay. Five microliters of the crude lysate was added to LOX assay solution consisting of 100 mM KH2PO4 (pH 6.5), 25 mM L(+)-lactic acid (SIGMA), 100 µg/ml BSA (Boehringer Mannheim), 1.25 mM 4-aminoantipyrine (WAKO), 85 mM phenol (Amresco) and 12 unit/ml horseradish peroxidase. LOX catalyzes the oxidation of two molecules of lactic acid with one molecule of dissolved oxygen to produce one molecule of hydrogen peroxide and pyruvate. This reaction can be coupled to the oxidation of 4-aminoantipyrine and phenol with two molecule of hydrogen peroxide, which is catalyzed by horseradish peroxidase. The latter reaction results in the generation of a red dye, which can be monitored by the change in absorbance at 510 nm at 30°C. Here, 1 unit of LOX activity was defined as the amount of the enzyme that generated 1 µmol of hydrogen peroxide per min. The concentration of LOX in the crude lysate was determined by enzyme-linked immunosorbent assay, using a specially prepared rabbit anti-LOX antibody, with purified LOX (Asahi Kasei) serving as a standard. No LOX activity and LOX presence were detected in a lysate of the vector control.
Evaluation of LOX activity half-life at high temperatures
Crude lysates were diluted with a buffer consisting of 50 mM Tris–HCl (pH 8.0) and 2 mM EDTA to adjust the concentration of LOX activity. Twenty microliters of the lysate was subjected to heat treatment at 62.5, 65, 68, 70 or 75°C for 0, 15, 30, 60, 120 or 180 min, respectively. The residual LOX activity after the heat treatment was determined and the LOX activity half-life (t1/2) was calculated on the assumption that the deterioration of activity obeyed the first-order kinetics. In addition, for temperatures that were so high or low that an empirical determination of the activity half-life was impossible, the t1/2 was calculated from the slope of an Arrhenius plot. As we did not find any marked decrease in LOX activity in the crude lysate below 50°C for 180 min, we concluded that the effect of proteolytic degradation was negligible during the heat treatment.
Profiling of amino acid substitutions at six primary mutation sites by saturation mutagenesis
Six primary sites identified by error-prone PCR were subjected to saturation mutagenesis (Kegler-Ebo et al., 1996; Zheng et al., 2004) in which the amino acid codes were diversified by using randomized NNK codons (where N is G, A, T, or C and K is G or T). The codons were introduced by using primers specific for each primary site and the QuickChange mutagenesis kit (Stratagene). The libraries constructed were subjected to screening for thermostable mutants by the active-staining method; the activity half-life at 65°C and the specific activity of selected mutants were then determined. Mutants showing improved activity half-life and basal catalytic activity of more than 40% of that of the wild-type were selected as candidates for each of the sites. Amino acids introduced in the candidates were determined by DNA sequencing.
Construction of a LOX multiple mutant library and identification of promising multiple mutants
Similar to the approach used in our previous study (Hamamatsu et al., 2005), we divided the LOX gene into six blocks to generate an assembling library for recombining the six mutation sites. Figure 1 shows the design of the blocks. Precursor DNA fragments corresponding to each block were amplified from the wild-type and a mutant gene, which had a mutation at either N2T, K15R, D70V, R141C, D164V or F239L. To generate the corresponding DNA fragments, the precursors for each of them were mixed in equimolar concentrations to give equal incorporation probability. The DNA fragments I to VI were then mixed in equimolar concentration and subjected to a recombination reaction by using overlap extension PCR without primers during the first five thermal cycles. After five cycles, the primers FrgI-f and FrgVI-r were added and the thermal reaction cycle was repeated 20 more times to amplify reconstructed full-length LOX genes. Full-length genes were purified and digested with EcoRI and PstI and then ligated back into the expression vector. This LOX multiple mutant library was subjected to screening for thermostability by the active-staining method and then to evaluation of activity half-life at 75°C. For the recombination of fine-tuned amino acid substitutions, unlike the previous report, several of the sites had multiple amino acid candidates. For the preparation of the corresponding DNA fragments including such sites, their precursors were amplified from mutants with each candidate mutation and then mixed in equimolar concentrations to give equal incorporation probability of mutations at each site. Next, these mixtures were mixed with their corresponding wild-type fragments in equimolar concentration. For fragments I and II, only DNA fragments derived from the wild-type gene were used because no candidate of mutation satisfied our criteria. Unlike our previous report in which libraries biased toward multiple mutants were generated, we generated a library in which all conceivable combinations of selected mutations existed equally because there was no mutation that contributed to thermostabilization independently of catalytic activity.
|
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Identification of thermostable mutations from a random mutation library and their accumulation
We screened a random mutation library generated by error-prone PCR for thermostable mutants and eventually identified six mutations for improving the thermostability of LOX, N2T, K15R, D70V, R141C, D164V and F239L. These mutations were then combined to generate a multiple mutant library as described in Materials and methods. In total, 10 000 transformants in the library were examined for thermostability by using the active-staining method and by evaluating activity half-life at 75°C. As a result, we identified 21 transformants that expressed thermostable LOX multiple mutants. Table I lists the top seven thermostable multiple mutants, together with their activity half-life at 75°C and specific activity. Although the most thermostable mutant showed 5000-fold greater thermostability at 75°C than the wild-type enzyme, its catalytic activity was only 2% that of the wild-type enzyme. Furthermore, Figure 2 showed the inverse relationship between thermostability and specific activity of the multiple mutants. This trade-off must reduce the practical value of these thermostable mutants, emphasizing the necessity of another approach to reduce the risk of the deterioration of catalytic activity. Since each of the component mutations reduced catalytic activity by 70–80% (data not shown), we assessed if decreasing the adverse effect of each of mutations would reduce the risk in multiple mutants. Incidentally, the top three multiple mutants had four or five of the six mutations. This high incorporation frequency and significant improvement of thermostability by accumulating mutations strongly supported the usefulness of our strategy based on the additivity principles, being consistent with the results in our previous reports (Aita et al., 2002; Hamamatsu et al., 2005).
|
|
Development of thermostable mutations with less effect on catalytic activity
By using a saturation mutagenesis method, we prepared small libraries comprising mutants with a complete set of the 20 standard amino acids incorporated at each of the six primary mutation sites we identified. Approximately 3000 transformants in each library were subsequently examined for thermostability by using the active-staining method and for evaluating activity half-life at 65°C. Thermostable mutants identified in these screenings were then examined for catalytic activity. As shown in Figure 3, mutants from each library showed a strong inverse relationship between thermostability and catalytic activity, although one mutant containing a mutation at R141 and one containing a mutation at D164 retained 64 and 78%, respectively, of catalytic activity of the wild-type enzyme despite their high thermostability. To achieve our intended improvement, we consequently selected mutations that contributed to thermostabilization and showed catalytic activity of more than 40% that of the wild-type enzyme. Table II lists the selected mutants together with their catalytic activities and activity half-lives at 65 and 75°C. All mutants from both N2 and K15 saturation libraries did not satisfy the criteria. Notably, 8 of 13 mutations, including R141F and D164I mentioned above, found in the mutants required at least two contiguous nucleotide mutations, indicating the usefulness of expanding sequence search by saturation mutagenesis after error-prone PCR to identify ideal mutations.
|
|
Accumulation of mutations with less effect on catalytic activity
The mutations listed in Table II were accumulated by the mutation-assembling method as described in Materials and methods. We examined 20 000 transformants for their thermostability by using the active-staining method and subsequently selected 261 transformants after evaluating their activity half-lives at 75°C. They were subjected to the evaluation of specific activity and, ultimately, 21 transformants were selected as those that expressed promising LOX multiple mutants. As shown in Figure 4, although these mutants still showed an inverse relationship between thermostability and catalytic activity, their catalytic activities were improved in comparison with those of the multiple mutants in Figure 2. Table III lists the top 17 thermostable multiple mutants, which showed 5- to 10-fold greater catalytic activity despite having comparable thermostability when comparing with those of the mutants in Table I. This result indicates that fine-tuning of component mutations directed to the improvement of an intended property can avoid the accumulation of negative effects on other properties. As shown in Table III, the top three multiple mutants had mutations at all four sites. This result also supports our strategy based on the additivity principles together with that in Table I.
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
We have theoretically demonstrated the near-additivity of advantageous effects of mutations in traversing local fitness landscapes. Given the near-additivity of mutational effects, we can simply hypothesize that the accumulation of mutations advantageous for a property with small change in other ones, such as catalytic activity, could allow to create multiple mutants not at the cost of other properties. To verify this hypothesis, we selected LOX as a model case because of strong trade-off between thermostability and catalytic activity in its directed evolution. We assessed, (i) if there could be mutations that were advantageous in thermostabilization of LOX not at the cost of catalytic activity and (ii) the accumulation of which could show an advantage for exclusive improvement of thermostability. The accumulation of mutations was conducted by our mutation-assembling method, which facilitated to accumulate as many advantageous mutations as possible without epistatic effects, which diminish the positive effect of each mutation when they are present concurrently (Hamamatsu et al., 2005
).
In this study, we identified mutation sites by an error-prone PCR method, which is useful to generate a random mutant library. This method is, however, unlikely to generate two, much less three, contiguous nucleotide changes or more than five to seven amino acid changes. In addition, the likely amino acid substitutions have the similar physicochemical properties, such as hydrophobicity and charge group. Miyazaki and Arnold (1999) demonstrated the impact of expanding search space by saturation mutagenesis. They showed that non-conservative amino acid substitutions, which require at least two contiguous nucleotide mutations, offered a potential to improve thermostability of a psychrophilic subtilisin-family protease, which was not able to be achieved by conservative ones. This method was, therefore, employed in our study to discover our intended mutations at the sites found by an error-prone PCR method. As shown in Figure 3 and Table II, LOX mutants showed a strong inverse relationship between thermostability and catalytic activity. Several mutations, such as R141F and D164I, however, showed less effect on catalytic activity despite great contribution to thermostabilization. High population of non-conservative mutations in Table II indicates the advantage of expanding search space by saturation mutagenesis again. Furthermore, this profile of single point mutations in terms of the change in several properties is useful in estimating the extent to which the improvement of a property is expected to be achieved if the deteriorative effect on the other ones is taken into consideration. For our assembling technique, this analysis allows us to set criteria for the selection of a basis set of component mutations on the basis of both required improvement and acceptable deterioration. Even though mutations are finely tuned by saturation mutagenesis, the original selection of primary mutation sites was still limited by codon degeneracy. In future studies, this issue could be addressed by another efficient method for identifying promising mutation sites, such as that reported by Murakami et al. (2002), in which mutations can be introduced into a gene without the limitations imposed by codon degeneracy.
As shown in Tables I and III, the accumulation of tuned mutations of LOX achieved 5- to 10-fold increase in catalytic activity with comparable improvement of thermostability in comparison with that of not tuned mutations. This result indicates that the accumulation of fine-tuned mutations with optimum combinations is one of the useful approaches to create multiple mutants with directed improvement. In addition, the multiple mutants showed high incorporation frequency of the selected mutations as exemplified in our previous reports (Aita et al., 2001; Hamamatsu et al., 2005). The retention of catalytic activity achieved in this study might be, however, not sufficient for practical use of the LOX mutants because it is still only 10% of the wild-type activity. One way to address this issue would be the co-accumulation of mutations that contribute to improving catalytic activity without deterioration of thermostability. Alternatively, the criteria for selecting mutation candidates could be tightened, although in this case thermostability would be probably sacrificed. Because it is not currently clear how much thermostabilization is required to prolong the life of lactate sensors, the effect of thermostabilization on the lifetime of the sensor should be evaluated with our developed mutants.
The structure of LOX has not been elucidated owing to the difficulties involved in generating crystals of the protein (Morimoto et al., 1998). Kaneko et al. (2005) have, however, reported its predicted tertiary structure on the basis of homology modeling using the structure of glycolate oxidase (EC 1.1.3.1
[EC]
) as a template. Similarly, we generated a model structure of LOX and investigated the location of mutations incorporated in the multiple mutants to look at the mechanism of their contribution to thermostabilization. LOX is thought to form a (ß/)8 barrel structure and a homo-tetramer (Morimoto et al., 1998). Strengthening the quaternary structure is one way in which to improve the thermostability. Kaneko et al. (2005) predicted that the A68 site was a crucial residue for formation of the homo-tetramer and, furthermore, that its substitution with a hydrophobic amino acid was expected to strengthen the interaction with a hydrophobic pocket in the adjoining subunit. Although substitutions at the A68 site failed to improve the thermostability of LOX in their study, we found that substitutions at the D70 site close to the A98 site resulted in great thermostabilization and favored substitution with hydrophobic amino acids. Indeed, substitution with leucine and valine showed approximately 4- to 10-fold greater thermostability than substitution with methionine, serine and lysine (data not shown), although methionine and serine were selected as candidates at the D70 site for our modification because the hydrophobic amino acids led to a greater reduction in catalytic activity. The failure at the A68 site is likely to be attributable to the accuracy of the model. In addition, this finding of substitution with non-hydrophobic amino acids, such as methionine and serine, suggests the possibility that another mechanism of thermostabilization may operate in the subunit boundary. Minagawa et al. (2002) reported that E160G mutation contributed to the thermostabilization of LOX by strengthening the interaction between the 2 and 3 helices of the (ß/)8 barrel structure via weakening electrostatic repulsion between the E160 and E130 residues. We found that the D164 site, which is spatially close to the E160 residue and the side chain of which faces the same direction on the 3 helix as that of E160 residue, contributed to LOX thermostability. In contrast to the result of Minagawa, however, substitution of the D164 residue with the hydrophobic amino acids, leucine and isoleucine, was preferable for thermostabilization. This suggested the presence of a hydrophobic environment between the 2 and 3 helices. Indeed, our model showed that the side chain of the E164 residue was spatially close to that of the F129 residue, which was likely to comprise a hydrophobic pocket with the F143 and I132 residues. The side chain of F129 was also close to the R141 residue, whose substitution with phenylalanine was favored for thermostabilization (Table II). As shown in Table III, multiple mutants in which both the R141 and D164 residues were substituted with hydrophobic amino acids showed greater thermostability than those that did not have these substitutions. This line of evidence indicates that the formation of a hydrophobic interaction between the 2 and 3 helices has a marked contribution to stabilization of the (ß/)8 barrel structure. As for the F239 residue, its side chain was close to that of the Trp142 residue, which is next to the R141 residue. This suggests that substitutions at the F239 site is likely to aid the formation of a hydrophobic interaction between the R141 and D164 residues. Indeed, substitutions at the F239 site showed an additive effect on the thermostability of mutants that had both the R141 and D164 mutations (Table III). Taken together, the profile of mutations at each site and multiple mutants have provided useful information about both the structure of LOX and strategies for its modification.
In conclusion, we hypothesized the presence of single-mutation variants around the wild-type on a sequence space, which show the exclusive improvement of a specific property, and the advantage to accumulating the mutations by our assembling method to make a leap of the property without sacrificing other ones. The result in this study indicates that the comprehensive search by error-prone PCR followed by saturation mutagenesis is one of the useful methods for discovering appropriate mutations and estimating the feasibility of exclusive improvement of a property. Furthermore, the successful development of LOX multiple mutants by the accumulation of such mutations supports the concept that adverse effects in multiple mutants could be avoidable by reducing the effect of each of the component mutations to be introduced. Our method, reported previously, had potential risk of the changes in other properties. Since our strategy is based on the typical features of local fitness landscapes, the method coupled with the fine-tuning of mutations would be applicable to a wide variety of enzymes, although further validation is required.
|
Footnotes |
|---|
Edited by Michael Hecht
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
We thank Dr Keigo Goda for the study of LOX homology modeling. This work was performed as a part of the R&D Project of the Industrial Science and Technology Frontier Program supported by New Energy and Industrial Technology Development Organization in Japan.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Aita T., Iwakura M., Husimi Y. (2001) Protein Eng. 14:633–638.
Aita T., Hamamatsu N., Nomiya Y., Uchiyama H., Shibanaka Y., Husimi Y. (2002) Biopolymers 64:95–105.[CrossRef][Web of Science][Medline]
Giver L., Gershenson A., Freskgard P.O., Arnold F.H. (1998) Proc. Natl Acad. Sci. USA 95:12809–12813.
Hamamatsu N., Aita T., Nomiya Y., Uchiyama H., Nakajima M., Husimi Y., Shibanaka Y. (2005) Protein Eng. Des. Sel. 18:265–271.
Ito N., Matsumoto T., Fujiwara H., Matsumoto Y., Kayashima S., Arai T., Kikuchi M., Karube I. (1995) Analytica Chimica Acta 312:323–328.[CrossRef]
Kaneko H., Minagawa H., Shimada J. (2005) Biotechnol. Lett. 27:1777–1784.[CrossRef][Web of Science][Medline]
Kegler-Ebo D.M., Polack G.W., DiMaio D. (1996) Methods Mol. Biol. 57:297–310.[Medline]
Kuchner O. and Arnold F.H. (1997) Trends Biotechnol. 15:523–530.[CrossRef][Web of Science][Medline]
Leung D.W., Chen E., Goeddel D.V. (1989) Technique 1:11–15.
Maeda-Yorita K., Aki K., Sagai H., Misaki H., Massey V. (1995) Biochimie 77:631–642.[Medline]
Mascini M., Fortunati S., Moscone D., Palleschi G., Massi-Benedetti M., Fabietti P. (1985) Clin. Chem. 31:451–453.
Minagawa H., Nakayama N., Matsumoto T., Ito N. (1998) Biosens. Bioelectron. 13:313–318.[CrossRef]
Minagawa H., Kaneko H., Shimada J. (2002) Nec Res. Dev. 43:251–255.
Miyazaki K. and Arnold F.H. (1999) J. Mol. Evol. 49:716–720.[CrossRef][Web of Science][Medline]
Miyazaki K., Wintrode P.L., Grayling R.A., Rubingh D.N., Arnold F.H. (2000) J. Mol. Biol. 297:1015–1026.[CrossRef][Web of Science][Medline]
Morimoto Y., Yorita K., Aki K., Misaki H., Massey V. (1998) Biochimie. 80:309–312.[Medline]
Murakami H., Hohsaka T., Sisido M. (2002) Nat. Biotech. 20:76–81.[CrossRef][Web of Science][Medline]
Palleschi G., Mascini M., Bernardi L., Zeppilli P. (1990) Med. Biol. Eng. Comput. 28:B25–B28.[CrossRef][Web of Science][Medline]
Palmisano F., Centonze D., Zambonin P.G. (1994) Biosens. Bioelectron. 9:471–479.[CrossRef][Web of Science][Medline]
Uchiyama H., Inaoka T., Ohkuma-Soyejima T., Togame H., Shibanaka Y., Yoshimoto T., Kokubo T. (2000) J. Biochem. (Tokyo) 128:441–447.
Wei X., Zhang M., Gorski W. (2003) Anal. Chem. 75:2060–2064.[Medline]
Wintrode P.L., Miyazaki K., Arnold F.H. (2000) J. Biol. Chem. 275:31635–31640.
Zheng L., Baumann U., Reymond J.L. (2004) Nucleic Acids Res. 32:e115.
Received April 12, 2006; revised July 17, 2006; accepted July 24, 2006.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||