2.1 Cloning, expression, and characterization of wild-type afCR gene
Creatinase is used as a key enzyme for enzymatic measurement of creatinine which catalyzes the hydrolysis of creatine to sarcosine and urea. It has been found in various kinds of bacteria such as Flavobacterium,Pseudomonas, Arthrobacter and Bacillus [19–23]. In this work, we analyzed the creatinase family sequences and found a creatinase gene from Alcaligenes faecalis in the NCBI database. Its amino acid sequence comparison between creatinase from Pseudomonas, Flavobacterium, and Arthrobacter creatinase showed about 60% homology. We amplified afCR gene and cloned it into pANY1 expression vector. The recombinant DNA was expressed in E.coli BL21 (DE3), then purified by affinity chromatography and its activity was determined. We observed that the molecular mass of afCR was estimated to be 45 kDa. The results of afCR enzymatic characterization showed maximum activity of 14 U/mg (activity was determined by using the 1 mg/ml enzyme at 37 °C) and a lower KM value (23.6 mM), providing a new source of creatinase for the enzymatic determination of creatinine.
2.2 Effects of temperature, pH, and metal ions on wild-type afCR activity
Reaction temperature and pH have significant impacts on enzyme stability and activity, as well as on the dynamic equilibrium of a reaction. In order to provide an initial characterization of the optimal temperature for wild-type afCR (WT), we first examined its activity across temperatures ranging 25–55 °C. We found that WT activity gradually increased from 30 °C to 37 °C but was rapidly inactivated at higher temperatures, thus indicating that 37 °C was the optimum temperature for afCR reactions (Fig. 2A). After evaluating the optimal temperature, we examined WT activity over a range of reaction pH values (4.5–10.0) to determine at which values it functioned most efficiently (Fig. 2B). The results revealed a marked change in activity over this range, steadily increasing up to pH 8.0, after which the rate of afCR relative activity declined, with more than 50% of maximum activity observed at a broad pH plateau between 7.5 and 9.0. These results thus demonstrated that pH 8.0 and 37 °C were optimal conditions for afCR reactions. Interactions between metal ions and amino acid residues are essential for the stability of some proteins, however metals are also known to function as powerful inhibitors of the enzymatic activity [24, 25]. In order to investigate the effects of metal ions on WT afCR enzyme stability, the WT was individually incubated with equimolar concentrations of Co2+, Fe2+, Fe3+, Mn2+, Zn2+, Ca2+, Mg2+, Cu2+, and Na+. The WT exhibited differences in sensitivity to the metal ions (Fig. 2C). Specifically, Mn2+ and Mg2+ ions enhanced WT afCR activity; in the presence of EDTA, Fe3+, Fe2+, Ca2+ WT retained > 80% of its initial activity; however, in contrast, Cu2+, Co2+, Zn2+ and Na+ ions inhibited WT activity to varying degrees. In particular, the presence of Cu2+ at 1 mM led to complete inhibition of WT afCR activity.
2.3 Non-biased phylogenetic consensus method reveals 21 target consensus residues for mutagenesis
In order to employ a non-biased phylogenetic consensus method for identifying specific residues that will most effectively improve thermostability through mutation, we used wild-type afCR as a query for a blastP search of proteins in the NCBI database. This search yielded 45 CR sequences with a sequence similarity greater than 50% (Fig. S3) compared with wild-type afCR. We then aligned these sequences and deleted duplicates, as well as sequences that were too long or too short. A final set of 24 CR homolog sequences were selected for phylogenetic tree construction. In order to reduce the branch bias that may be introduced by overrepresentation in the database, we developed a non-biased phylogenetic consensus method based on phylogenetic relationships. To this end, we first constructed a neighbor-joining phylogenetic tree and determined the branch lengths for each sequence, representing the evolutionary distance between CR homologs as a ratio of the number of different residues to the minimum length of the sequence (Fig.S3).
Thus, the phylogenetic tree serves as a model for CR divergence caused by evolutionary pressures, with branch lengths used to calculate the weight of each sequence. The use of branch weights thus provided statistical independence among the different protein sequences, enabling the identification of conserved target residues by optimizing the occurrence frequency of amino acid residues (Table 1). We used a proprietary consensus wi.py script to introduce branch weights to calculating the consensus sequence (Fig.S1). Compared with the wild-type query sequence, 59 of the 404 amino acid positions were selected as candidate consensus residues if they appeared at a given position with > 40% frequency among aligned sequences. This consensus cut-off can be adjusted according to the screening method of consensus mutants and the accuracy of other criteria.
Table 1
Consensus design information and experimental characterization results of the single-site mutants.
Mutation
|
Secondary structurea
|
Distance to
act. site (Å)
|
Frequencyb
(%)
|
55 °C t1/2 (min)
|
Fold
Improvement
|
Relative
Activity (%)
|
M0
|
--
|
--
|
--
|
11.6
|
1
|
100
|
L6P*
|
Loop
|
12.2
|
58.97
|
19
|
1.64
|
94.28 ± 9.54
|
D17V*
|
Loop
|
11.2
|
43.04
|
150
|
12.9
|
105.00 ± 5.71
|
P20T
|
Loop
|
11.2
|
61.00
|
8.7
|
0.75
|
95.24 ± 4.77
|
V33L
|
α-Helix
|
15.6
|
59.58
|
5.45
|
0.47
|
114.29 ± 11.32
|
C52N
|
α-Helix
|
9.4
|
91.21
|
7.85
|
0.68
|
110.47 ± 6.67
|
G58D*
|
β-Turn
|
11.9
|
61.00
|
17
|
1.47
|
83.80 ± 10.41
|
W59F
|
β-Turn
|
8
|
85.88
|
3
|
0.26
|
120.95 ± 6.77
|
D73T
|
β-Turn
|
20.8
|
68.85
|
8.1
|
0.70
|
106.67 ± 17.41
|
F108Y*
|
α-Helix
|
7.8
|
71.10
|
16
|
1.37
|
105.00 ± 5.71
|
Y109F*
|
α-Helix
|
8.3
|
69.85
|
13
|
1.12
|
122.86 ± 1.91
|
L162A
|
β-Turn
|
13.0
|
66.42
|
10.4
|
0.89
|
108.57 ± 4.35
|
T117P*
|
β-Turn
|
21.0
|
44.84
|
12
|
1.03
|
100.92 ± 15.57
|
Q165I*
|
α-Helix
|
9.8
|
60.03
|
20.7
|
1.78
|
88.57 ± 9.62
|
K166A
|
α-Helix
|
10.2
|
51.51
|
11.3
|
0.97
|
103.81 ± 7.56
|
T199S*
|
α-Helix
|
10.2
|
42.81
|
14
|
1.2
|
117.14 ± 8.60
|
T251C*
|
β-Sheet
|
6.3
|
78.10
|
23
|
1.98
|
107.62 ± 2.09
|
E349V*
|
Loop
|
14.3
|
96.20
|
12
|
1.03
|
114.29 ± 13.29
|
K351E*
|
Loop
|
15.7
|
83.23
|
13.7
|
1.18
|
142.86 ± 4.85
|
V362I
|
β-Sheet
|
6.8
|
48.15
|
10.9
|
0.94
|
99.05 ± 15.23
|
V340L
|
Loop
|
6.2
|
66.40
|
3.8
|
0.33
|
80.00 ± 15.57
|
C331S
|
Loop
|
6.2
|
83.72
|
6
|
0.52
|
121.90 ± 1.90
|
a Location of the residue according to the homology structure of the afCR. b The frequency of the amino residue occurrence as calculated from the sequence alignment of the afCR. * Thermostable variant.
|
In order to further narrow the pool of candidate residues for mutagenesis, we then considered the impacts on protein structure associated with each position of the 59 residues by applying the following criteria: 1) the substitution must be farther than 6 Å away from the active site to avoid the disruption of catalysis; and 2) we excluded amino acids with side chains that directly formed hydrogen bonds or salt bridges with other residues to avoid decreasing protein structure stability. Based on these screening criteria, we selected 21 mutations for further protein engineering by mutagenesis, including E349V, W59F, C331S, K351E, T251C, F108Y, Y109F, D73T, L162A, V340L, P20T, G58D, Q165I, L6P, V362I, T117P, D17V, T199S, V33L, C52N, and K166A (See Table 1 for residue positions and distance from the active site).
2.4 Construction and characterization of the afCR mutants
Based on the report of a I304L/F395V double mutant variant of creatinase from Erwinia with lower KM than its wild type enzyme [26], we first examined the effects of introducing two non-synonymous mutations (I304L/F395V) into WT afCR. We then examined the KM value for the double mutant (15.3 mM) and found that it was lower than that of the WT afCR (23.6 mM), while retaining similar Kcat value. Accordingly, the I304L/F395V afCR variant (afCR-M0) was subsequently used as a parent template to construct more variants. To this end, 21 consensus variants were constructed based on the candidate residues identified above and expressed in E.coli. Eleven of the 21 variants showed improved thermostability compared to afCR-M0 at 55 °C with 80% or higher activity compared to afCR-M0 (Table 1). Thus, the success rate of 50% or higher among mutant variants suggested that the non-biased-phylogenetic consensus design method is an effective approach for improvement of thermostability of afCR. The best variant D17V (afCR-M1) exhibited the highest half-life at 55 °C (150 min), which was 12.9-fold higher than that of afCR-M0 (Fig. 2D). The thermostability of the other positive mutants was increased by 1-2-fold compared with that of afCR-M0.
2.5 Combination of beneficial mutations and thermostability of combinatorial variants
In order to further improve the thermostability of afCR-M1, we then gradually introduced additional positive mutations (i.e., L6P, G58D, Q165I, F108Y, Y109F, T117P, T199S, T251C, E349V, and K351E) (see Table 2). We found that thermostability improved with subsequent mutations and that these variants retained 80% or higher activity compared to afCR-M1. Among the double mutants, D17V/T199S (afCR-M2-4) showed the highest thermostability, with a half-life of 210 min at 57 °C (~ 105-fold higher than that of afCR-M0) (Table 2). We then used it as a template to individually introduce other mutations (L6P, T251C, F108Y, Y109F, K351E). All triple mutant variants exhibited significant improvements in thermostability, with the highest increase found in mutant M3-4 (D17V/T199S/L6P), which had a half-life of 1258 min at 57 °C (~ 629-fold higher than afCR-M0) (Table 2). We then generated variants of M3-4 by introduction of T251C, F108Y, Y109F, and K351E, respectively. This third round of mutagenesis produced variant M4-2 (D17V/T199S/L6P/T251C), which exhibited the highest stability of all variants up to this point, i.e., a half-life of 3371 min at 57 °C (~ 1685-fold higher than afCR-M0). All the above triple and quadruple mutants retained similar catalytic activity to that of afCR-M0 in addition to showing improved thermostability. During the combination process, we also discarded some mutation sites that led to decreased activity or failed to significantly improve thermostability, such as D17V/G58D, D17V/Q165I, D17V/T117P, and D17V/E349V.
Table 2
Experimental characterization results of the afCR multi-site mutants.
Enzyme
|
Mutation
|
57 °Ct1/2 (min)
|
Fold
Improvement
|
Relative
Activity (%)
|
M0
|
I304L/F395V
|
2
|
1
|
100
|
M1
|
M0 + D17V
|
40
|
20
|
105 ± 7.71
|
M2-1
|
M1 + L6P
|
142
|
71
|
104.76 ± 18.10
|
M2-2
|
M1 + T251C
|
71
|
35.5
|
118.09 ± 3.25
|
M2-3
|
M1 + K351E
|
101
|
50.5
|
161.90 ± 1.90
|
M2-4
|
M1 + T199S
|
210
|
105
|
111.43 ± 2.31
|
M3-1
|
M2-4 + T251C
|
599
|
299.5
|
112.38 ± 2.86
|
M3-2
|
M2-4 + F108Y
|
482
|
241
|
106.67 ± 4.76
|
M3-3
|
M2-4 + K351E
|
859
|
429.5
|
157.61 ± 5.30
|
M3-4
|
M2-4 + L6P
|
1258
|
629
|
123.81 ± 1.90
|
M4-1
|
M3-4 + F108Y
|
2498
|
1249
|
134.28 ± 2.86
|
M4-2
|
M3-4 + T251C
|
3371
|
1685.5
|
110.47 ± 1.90
|
Table 3. Kinetic and thermodynamic properties of afCR mutants.
Enzyme
|
Mutation
|
T5015 (ºC)
|
Tm (ºC)
|
M0
|
I304L/F395V
|
55.5 ± 0.1
|
59.6 ± 0.40
|
M1
|
M0 + D17V
|
59.06 ± 0.3
|
62.12 ± 0.02
|
M2-4
|
M1 + T199S
|
61.35 ± 0.2
|
63.93 ± 0.30
|
M3-4
|
M2-4 + L6P
|
62.21 ± 0.12
|
64.94 ± 0.20
|
M4-1
|
M3-4 + F108Y
|
61.0 ± 0.15
|
64.35 ± 0.31
|
M4-2
|
M3-4 + T251C
|
59.7 ± 0.1
|
64.24 ± 0.23
|
Notably, we found that some single-site mutants together produced a synergistic effect on thermostability. For example, compared with the single-site mutant D17V, the addition of T199S (D17V/T199S) resulted in another 5-fold increase in thermostability, while T199S alone led to a 1.2-fold increase compared to afCR-M0 (Table 1). Further combination of D17V/T199S with L6P produced another 6-fold increase in thermostability, whereas alone it only provided a 1.64-fold increase compared to afCR-M0 (Table 1). Thus, these results demonstrated that these amino acid residues function together resulting in a synergistic improvement to thermostability. However, we also observed that some combinations of single-site mutants resulted in antagonistic effects on the thermostability. For example, compared with the L6P (half-life of 19 min at 55 °C) and G58D (half-life of 17 min at 55 °C) single mutation variants, the L6P/G58D double mutant had reduced thermostability (half-life of 6 min at 55 °C), which was even lower than that of the afCR-M0 template. Further exploration of the contributions by each of these residues to overall thermostability will help to clarify the molecular mechanisms underlying the synergistic effects of these mutations, which will provide a useful reference for CR engineering through combinatorial mutagenesis.
2.6 Kinetic and thermodynamic stability of the improved afCR combinatorial variants
To assess the kinetic stability of the variants, activity was measured across a range of temperatures (35–70 °C) to determine the temperatures at which enzyme activity was reduced by 50% after 15 min of incubation (T5015) (Fig. 4A). We found that the residual activity of M0 was only 7% after incubation at 60 °C for 15 min (Fig. 4A), whereas, M1 retained 40% activity, M2-4 and M3-4 retained 80% activity, M4-1 retained 68% activity, and M4-2 retained 50% activity. These results thus indicated that the T5015 of these variants was 3.6 to 6.7 °C higher than that of the afCR-M0 (Table 3). Notably, although M4-1 exhibited the highest half-life (T1/2), it had a lower T5015 value than that of M3-4. Since T1/2 and T5015 are both primary indicators of protein kinetic stability, this result also demonstrated that the improvements in stability among the combinatorial variants were not due to the simple additive effects of single mutations. However, to determine the contribution of each mutation to the overall effect on thermostability requires detailed exploration of the relationship between protein structure and external reaction conditions in determining protein kinetic stability [27].
To further examine how the combined mutations affected thermodynamic stability, the melting temperatures (Tm) was determined by nanoDSF. The underlying principle of this assay is that the 350/330 nm emission ratio for tryptophan fluorescence of a given protein can indicate the temperature at which the protein unfolds. The Tm values for M1, M2-4, M3-4, M4-1, and M4-2 ranged from 2 to 5.3 °C higher than that of afCR-M0. Interestingly, unlike the changes in kinetic stability, the melting temperatures of the combined mutation variants increased only slightly over that of the template protein, which therefore indicated that these mutations in afCR affected kinetic stability more than thermodynamic stability (Fig. 4B).
2.7 Molecular mechanisms underlying higher thermostability conferred by individual mutations
Given our results of higher thermostability, we next investigated new or changed mechanistic interactions associated with individual mutations that improved thermostability among the positive mutants using afCR structural homology modeling. To this end, we first checked if mutations introduced new interactions (Fig. 5 and Table S2) and found that the F108Y substitution led to the formation of a new hydrogen bond with the Y68 side chain hydroxyl group (Fig. 5A). Similarly, the Y109F mutation was observed to form new π-π interactions with residues F133 and F108, which subsequently increased the stability of the two helical structures in which the residues are located (Fig. 5B). By contrast, the K351E mutation resulted in the loss of a hydrogen bond between that residue and E349. However, this mutation resulted in new salt bridge interactions and an H-bond formation between P352 and L350 (Fig. 5C). In addition, the T251C substitution introduced a new sulfhydryl group and may create a new disulfide bond with the neighboring C175 (Fig.S4A).
We then examined changes in hydrophobic and hydrophilic substitutions among the mutant variants and found that D17V mainly improved hydrophobic packing in the protein interior and facilitated the interaction network between the neighboring residues His12, Asn13, Lys16, Trp90, and Arg91 (Fig.S4B). Furthermore, we found that Q165I and E349V conversions, mutations similar to D17V, also stabilized the enzyme through increased hydrophobicity, using similar mechanisms as that of D17V. The advantages provided by these mutations apparently contradict conventional mutation theory, in that hydrophobic to hydrophilic amino acid conversion is more likely to improve protein thermostability. For example, the G58D and T199S mutations mainly contributed to higher stability by increasing the hydrophilicity of the protein surface, as well as the inter-helix hydrophobic interactions. In addition, introduction of proline substitutions has been previously shown to be a trend in the stabilization of proteins [28]. The leucine to proline substitution in the L6P mutation (Fig.S4C) decreased the conformational entropy, localized to this amino acid position, which resulted in made the protein space structure more rigid.
In summary, this analysis showed that individual mutations incurred a wide range of alterations to afCR-M0 structure and internal and external interactions, including changes in hydrogen bonds, salt bridge interactions, π-π interactions and disulfide bonds (F108Y, Y109F, K351E, T251C), increased hydrophilic interactions at the protein surface (T199S, G58D), reduction in the conformational entropy of local unfolded proteins (L6P), and improved hydrophobic packing in the protein interior or increased the interaction network (D17V, Q165I, E349V). Moreover, the introduction of the above interactions can improve the thermostability of proteins has been verified in other studies [2, 29, 30]. Through analysis of these mutations, we have modified our current understanding of the interaction mechanisms by which different types of amino acid conversions may improve thermostability and thereby provide guidance for engineering higher thermostability in other proteins.