|Home | About | Journals | Submit | Contact Us | Français|
α/β-Hydrolases are important enzymes for biocatalysis, but their stability often limits their application. As a model α/β-hydrolase, we investigated a plant esterase, salicylic acid binding protein 2 (SABP2). SABP2 shows typical stability to urea (unfolding free energy 6.9±1.5 kcal/mol) and to heat inactivation (T1/215 min 49.2±0.5 °C). Denaturation in urea occurs in two steps, but heat inactivation occurs in a single step. The first unfolding step in urea eliminates catalytic activity. Surprisingly, we found that the first unfolding likely corresponds to the unfolding of the larger catalytic domain. Replacing selected amino acid residues with proline stabilized SABP2. Proline restricts the flexibility of the unfolded protein, thereby shifting the equilibrium toward the folded conformation. Seven locations for proline substitution were chosen either by amino acid sequence alignment with a more stable homolog or by targeting flexible regions in SABP2. Introducing proline in the catalytic domain stabilized SABP2 to the first unfolding in urea for three of five cases: L46P (+0.2 M urea), S70P (+0.1) and E215P (+0.9). Introducing proline in the cap domain did not (two of two cases), supporting the assignment that the first unfolding corresponds to the catalytic domain. Proline substitutions in both domains stabilized SABP2 to heat inactivation: L46P (ΔT1/215 min = +6.4 °C), S70P (+5.4), S115P (+1.8), S141P (+4.9), and E215P (+4.2). Combining substitutions did not further increase the stability to urea denaturation, but dramatically increased resistance to heat inactivation: L46P-S70P ΔT1/215 min = +25.7 °C. This straightforward proline substitution approach may also stabilize other α/β-hydrolases.
Most proteins are only marginally stable; that is, just stable enough for their natural function.1,2 This marginal stability is not an inherent limitation of proteins since most proteins can be stabilized by single amino acid substitutions without compromising activity.3 The reason for the marginal stability is that evolution cannot select for more than the minimum stability. Further, if a hyperstable protein arises randomly, further random amino acid substitutions are likely to decrease the stability of a protein until the stability drops to the minimum needed. The result is proteins that are just stable enough for their function.4
Most applications require stable enzymes that tolerate harsh reaction conditions or maintain activity for an extended time under moderate conditions.5-7 Stable enzymes also tolerate numerous amino acid substitutions so are better starting points for protein engineering.8-11 Because of the marginal stability of most proteins, most proteins for biocatalysis need to stabilized. In this work we use an esterase from tobacco, salicylic acid binding protein 2, as a model α/β-hydrolase fold protein. This protein superfamily contains many biocatalysts including esterases, lipases, hydroxynitrile lyases and dehalogenases.12-14 General methods to stabilize this class of enzymes would be broadly useful for biocatalysis. We analyze the unfolding of this model esterase and test proline substitution as a stabilization strategy.
Two common methods to measure protein stability are denaturant-induced unfolding and heat inactivation. Stable proteins adopt folded conformations in solution, but are in equilibrium with a tiny fraction of unfolded protein conformations, Figure 1A. Adding denaturant, such as urea, disrupts hydrogen bonding between amide groups and shifts the equilibrium toward the unfolded form.15-17 By measuring the equilibrium amount of unfolded protein in different concentrations of urea, one can extrapolate to the free energy of unfolding without urea.18 Free energy values are particularly useful for rational, structure-based protein engineering since they relate directly to molecular properties like the contributions of flexibility to entropy.
Measuring the heat inactivation temperature of a protein also measures stability, Figure 1B. Heat shifts the equilibrium between folded and unfolded proteins by increasing the entropy contribution.19 The unfolded state created by heating may differ from the unfolded state created by urea denaturation because the destabilization mechanisms differ. In addition, heat denaturation usually irreversibly misfolds proteins, which prevents measuring the free energy of unfolding. Measuring stability by heat inactivation is easier than urea denaturation and more closely matches the goal of many applications – for the enzymes to remain active for long times at elevated temperatures.20-21
The α/β-hydrolase fold superfamily contains important enzymes for industrial applications.22-24 For example, lipases, esterases and hydroxynitrile lyases make enantiopure intermediates for pharmaceutical synthesis and dehalogenases degrade pollutants. Stabilization of α/β-hydrolase fold enzymes is an important goal because it can extend the applications of these enzymes.
Most measurements of stability of α/β-hydrolases use heat-induced unfolding or deactivation. For example, lipase B from Candida antarctica unfolds at 46.4 to 54.8 °C,21,25 Table 1. In a few cases, researchers measured the stability of α/β-hydrolases by urea-induced denaturation. For example, cutinase has an unfolding free energy of 7.7±0.6 kcal/mol. The unfolding occurred in one step because cutinase contains only one domain.26 For many α/β-hydrolases, the unfolding is more complex and the free energy cannot be extracted from urea denaturation curves. Most α/β-hydrolases contain two domains: a cap, or lid, domain and a catalytic domain. If these domains unfold independently via a partly unfolded intermediate, then this more complex unfolding prevents extracting free energies from the urea denaturation curves. For two special cases, esterases from thermophiles, the urea induced unfolding was a single cooperative transition and the free energies of unfolding could be derived from the data.27,28
Salicylic acid binding protein 2 from tobacco (SABP2) is a typical α/β-hydrolase.33 This esterase catalyzes hydrolysis of methyl salicylate to salicylic acid as a plant defense signal. Like most α/β-hydrolases, SABP2 consists of two domains, Figure 2. The larger catalytic domain (residues 1-111, 185-262; 189 aa, 72% of the total in SABP2) and the smaller cap domain (residues 112-184; 73 aa, 28% of the total). The catalytic domain forms the α/β-hydrolase fold with a central core of β-sheets surrounded by α-helices. It contains the catalytic triad, the oxyanion hole for catalysis, and part of the substrate-binding site. The smaller cap (or lid) domain sits on top of the catalytic domain and forms the rest of the substrate-binding site. This cap domain contains a three-stranded antiparallel β-sheets and three helices. Each domain contains approximately equal numbers of polar and nonpolar amino acid residues. The catalytic domain contains 48% polar and 52% nonpolar, while the cap domain contains 51% polar and 49% nonpolar.
Substitution of amino acid residues with proline often increases protein stability.34-36 The locations are chosen so the proline substitution does not affect the folded form. Proline restricts the flexibility of the unfolded protein as compared to other amino acids, thus lowering the entropy gain upon unfolding by ~4 cal/deg mol relative to alanine,34 which corresponds to ~1.2 kcal/mol at 300 °K, Figure 1C. This change shifts the equilibrium toward the folded form, thereby stabilizing the protein.
In this work, we examine the unfolding of a typical α/β-hydrolase, salicylic acid binding protein 2 from tobacco (SABP2). We measured the stability of this esterase by both denaturant-induced unfolding and heat-induced inactivation. Urea induced denaturation involved a partly unfolded intermediate and thus, two unfolding transitions. Fitting the data to a three-state-model yielded free energies for each unfolding step. Introducing proline substitutions in the catalytic domain stabilized this protein to urea denaturation, while substitutions in either domain stabilized the protein to heat inactivation. The largest increase was a 25.7 °C increase in the heat inactivation temperature. A similar approach may also stabilize other α/β-hydrolase fold enzymes.
Reagents were from Sigma Aldrich unless otherwise noted. House deionized water was further purified with a Milli-Q-water-system (Millipore, Billerica, MA, USA, 18 MWΩ purity). Integrated DNA Technologies (Coralville, IA) made the primers for the polymerase chain reaction and ACGT (Wheeling, IL) sequenced the DNA using BigDye terminator version 3.1. Pfu DNA polymerase was bought from Agilent Technologies (La Jolla, CA), Dpn I, DNA marker and SDS-PAGE protein marker were from Invitrogen Life Technologies (Grand Island, NY). Isopropyl-β-D-thiogalactopyranoside (IPTG) was from Gold Biotechnology (St. Louis, MO) and ampicillin was bought from Roche Diagnostics. Escherichia coli DH5α and BL21 competent cells were prepared by calcium chloride treatment.37 The protein concentrations were determined by the Bradford dye-binding assay at 595 nm using the Bio-Rad reagent and bovine serum albumin as the standard.38 Protein gels were run on sodium dodecyl sulfate polyacrylamide gradient gels (NuPage 4-12% Bis-Tris gel from Invitrogen) using BenchMark protein ladder (5 μL/lane) as the standard. DNA gels were run using 0.8% ultrapure agarose with 1×TAE buffer and 1 kb DNA ladder as a standard. Unless otherwise noted, E.coli strains for protein expression were grown in modified TB media (12 g/L bacto-tryptone, 24 g/L bacto-yeast extract, 9.86 g/L glucose with 2.3 g KH2PO4 + 12.5 g K2HPO4) containing 100 μg/mL ampicillin, and incubated in unbaffled Erlenmeyer flasks at 37 °C shaking at 250 rpm. The amino acid sequences of MeHNL39 and SABP233 were aligned using ClustalX 2.1.40
The site-directed mutagenesis of salicylic acid protein 2 (SABP2) was carried out using the Stratagene QuikChange™ method where mutagenic primers, Table S1, were used to amplify the entire plasmid using the polymerase chain reaction.41 The reaction mixture contained DNA template (wild-type SABP2 gene with a C-terminal six-histidine tag in the pET21a(+) expression vector (pET21a(+)-SABP2), 1 ng), forward and reverse primers (100 μM, 1 μL of each), 10× PCR buffer (5 μL), dNTPs (0.9 μL 10 mM), Pfu DNA polymerase (1 μL, 2.5 U) and auto-claved water to a final volume of 50 μL. The thermocycling program for plasmid amplification was 95 °C for 1 min, 25 cycles of 95 °C for 30 s, 55 °C for 30 s and 72 °C for 7 min, with a final extension at 72 °C for 10 min using an OpenPCR thermocycler (Chai Biotechnologies, Santa Clara, CA). After confirming that the amplification succeeded by agarose gel electrophoresis, the methylated parental DNA strand was digested by addition of DpnI (5 U) and incubation at 37 °C for 3 h. An aliquot of the reaction mixture (10 μL) was transformed into E. coli DH5α competent cells by heat shock. The transformed cells were spread on a lysogeny broth (LB) agar plate containing 100 μg/mL ampicillin and incubated at 37 °C overnight. Several resulting colonies were picked, grown overnight in liquid LB-ampicillin media (3 mL) and plasmid DNA was isolated using the QiaPrep MiniPrep kit. Sequencing of the entire SABP2 gene confirmed that it contained the desired mutations, but no others.
E. coli BL21 (DE3) containing pET21a(+)-SABP2 or the corresponding mutated plasmid were grown in modified TB media at 37 °C until the OD600 reached 1.4-1.6, and then cooled to 17 °C. Protein expression was induced by adding IPTG to a final concentration of 1 mM, the culture was shaken at 17 °C for 24 h and the cells were harvested by centrifugation at 5000×g for 15 min at 4 °C. Supernatants were discarded and cells were resuspended in buffer A (50 mM sodium phosphate buffer, 300 mM NaCl, 20 mM imidazole, pH 8.2). The cells were lysed at 4 °C in an ice-water bath by ultrasound (Sonifier, output magnitude 40%, duty time 3 s, interval time 6 s, 8 min). The cell extract was centrifuged at 12,000×g at 4 °C for 30 min and the supernatant used for protein purification. The recombinant proteins contained a C-terminal hexahistidine tags and were purified from the cell lysate by Ni-affinity chromatography (Ni-NTA, Qiagen) according to the manufacturer's instructions. Proteins were eluted in buffer B (50 mM sodium phosphate, 300 mM NaCl, 125 mM imidazole, pH 8.2), then the buffer was exchanged to BES (N,N-bis(2-hydroxyethyl)-2-aminoethanesulfonic acid, 5 mM, pH 7.2) using 10,000 MWCO Amicon Ultra spin concentrators (Millipore, Billerica, MA, USA)
Hydrolysis of p-nitrophenyl acetate (pNPAc) was measured at pH 7.2 and 24 °C, and was corrected for spontaneous hydrolysis. The assay was performed in a 96-well microtiter plate with reaction volume of 100 μL (light path length = 0.29 cm) containing pNPAc (0.3 mM), 6.67 vol% acetonitrile, BES (4.2 mM, pH 7.2) and about 1 μg of enzyme. The release of p-nitrophenoxide (ε 404 nm = 16,600 M−1 cm−1) was measured spectrophotometrically (SpectraMax Plus 384 plate reader, Molecular Devices, Sunnyvale, CA). Enzyme-catalyzed and spontaneous reaction rates were measured in triplicate and reported as the mean and standard deviation. One unit of activity corresponds to one μmol of p-nitrophenoxide released per minute. Steady state kinetic constants were determined with the same method by varying the p-nitrophenyl acetate concentrations from 0 to 6 mM (6.67 vol% acetonitrile). Data was fit to the Michaelis-Menten equation by a non-linear fit using Origin 7.5 (OriginLab, Northampton, MA, USA).
The unfolding of enzyme in increasing concentrations of urea (0-6.2 M) or guanidinium chloride (GdmCl) (0-4.5 M) was monitored by the decrease in intrinsic tryptophan fluorescence using SpectraMax GEMINI XS plus-384 plate reader (Molecular Devices, Sunnyvale, CA) with an excitation wavelength of 278 nm and an emission wavelength of 329 nm. Each well in the 96-well microtiter plate contained enzyme in BES buffer (20 μL, 5 mM buffer pH 7.2) with final protein concentration of 0.1-0.3 mg/mL, (180-×) μL of assay buffer (5 mM BES buffer pH 7.2), and × μL of urea stock solution (8 M). Solutions were equilibrated at 4 °C for 16 or 24 h, then warmed to room temperature for 15-min before measuring the fluorescence. The relative concentrations of the folded or native conformation, N, and unfolded or denatured state, U, were determined by comparing the fluorescence at low urea concentrations (completely folded) and high urea concentrations (completely unfolded). The equilibrium ratio of folded and unfolded forms at a particular denaturant concentration yields the free energy of protein unfolding, ΔG, at that denaturant concentration, equation 1,
where R is the gas constant and T is the temperature. A plot of this free energy versus denaturant concentration yields a straight line of the form in equation 2,
To study whether the loss of fluorescence correlated with a loss of esterase activity, we measured the esterase activity remaining after denaturation. Enzyme was incubated with denaturant for 16 h and an aliquot (10 μL, ~1 μg enzyme) was removed and immediately added to the esterase activity assay solution (90 μL). The increase in absorbance at 404 nm monitored as described above.
To measure the reversibility of unfolding, the enzyme was incubated with denaturant and then the denaturant was removed by dialysis (10 kDa MW cut off, 3.5 L BES buffer (5 mM, pH 7.2) for total of 24 h, while replacing with fresh buffer three times. The esterase activity of dialyzed enzyme was compared to the activity before incubation with denaturant.
To confirm the formation of an intermediate in the unfolding of SABP2, we measured the fluorescence intensity at different urea concentrations at two wavelengths: 320 nm (Y320) and 365 nm (Y365). If the folded protein (N) unfolds without a stable intermediate to the unfolded form (U), then the changes at the two wavelengths are proportional; a plot of Y320 versus Y365 is a straight line according to equation 3 (derivation and explanation by Bushmarina and colleagues43).
Y320(N) represents the fluorescence intensity at 320 nm due to the folded form and Y365(U) represents the fluorescence intensity at 365 nm due to the unfolded form. Several groups have used this two-wavelength approach, first described by Burstein, to detect intermediates in protein folding, although they typically use a simple two line fit to the data.43–46 This two-line method assumes that there is no overlap between transitions and no points where a significant portion of protein is in more than two states. The nonlinear solution method used here allows for overlap between transitions.
Extending equation 2 to a three-state model yielding equations 4 and 5 where ΔGni is the free energy of native to intermediate transition, and ΔGiu the free energy for the transition of intermediate to unfolded.
The free energy of each transition in pure water is ΔGniuw and ΔGiuuw and Kni is the ratio of protein in the native and intermediates states, while Kiu is the ratio of protein in the intermediate and unfolded states. The m values for the two transitions (mni and miu) are the susceptibility of native and intermediate states to the denaturant such as urea. These equations predict the fluorescence (Y) at a particular wavelength and concentration of denaturant from the parameters of fluorescence of each state (Yi, Yn, and Yu), equation 6.
Minimization of the squared residuals between measured and calculated fluorescence using the nonlinear “Solver” Add-in with Microsoft Excel yielded values for ΔGniH2O, ΔGiuH2O, and Yi for both wavelengths. To reduce the number of variables and prevent over fitting of data, m values of each domain of SABP2 were calculated using solvent accessible surface area (ProtSA47) using the relationship described by Myers and colleagues.48 The calculated m value for the catalytic domain was 1938 kcal/mol/M and that for the cap domain was 897. Yn and Yu were the average of the three highest or lowest fluorescence readings. The 95% confidence intervals of ΔGniH2O, and ΔGiuH2O values were calculated as described by Kemmer and Keller.49 We used the “FINV” function Microsoft Excel to find the range of squared residuals within the 0.95 confidence range given the degrees of freedom. ΔGniH2O, ΔGiuH2O were then varied and held constant while other variables were recalculated using Solver to determine the range of ΔGniH2O, ΔGiuH2O values that produced squared residuals within the calculated confidence range.
The kinetic stability was determined by measuring the residual enzymatic activity at room temperature after 15-min incubation (BES buffer, 5 mM, pH 7.2) at elevated temperatures. The temperature at which 50% of the enzymatic activity is lost after 15 min incubation, T1/215 min, was determined by fitting the loss of activity at different temperatures to a four-parameter Boltzmann sigmoidal function (y = a+(b-a)/(1+e(x-x0)/k) using Origin 7.5 (Origin Software Inc., Northampton, MA, USA). Half-life at 60 °C was calculated from the activity loss after 15 minute incubation with the function 15*ln(2)/ln(unheated/heated). This temperature is low enough to retain some activity for the wild type enzyme, but high enough to see some loss of activity for the stabilized variants.
The heat denaturation of SABP2 was also measured by the loss of circular dichroism (CD) signal (JASCO J-815 spectropolarimeter with Peltier-type temperature controller (PTC-348WI)) at 224 nm upon heating the sample from 20 to 90 °C at 2 °C min−1. T1/2CD is the temperature at which 50% of the signal is lost. CD measurements were made at a protein concentration of 0.2-0.3 mg/mL in BES buffer (5 mM, pH 7.2) with a 1 mm path length cell. The wavelength of 224 nm is a local minimum in the CD spectrum of SABP2 that showed the largest changes upon heating. Signals due to both α-helix and β-sheet structures typically appear in this region. The data were normalized to the lowest signal observed at 224 nm.
The FoldX algorithm (version 3.0 beta5.1, available from http://foldx.crg.es/) modelled protein stability for various amino acid replacements using empirically weighted energy terms.50 The X-ray structure of SABP2 (pdb:1Y7I33) was preprocessed using “RepairPDB” command to minimize the energy of the structure, followed by the “BuildModel” command to evaluate the effect of various substitutions.
The molecular dynamics (MD) simulation was performed using the software NAMD51 using 64 Pentium IV processors running Ubuntu Linux starting from the x-ray crystal structure of SABP2 (pdb:1Y7I,33 chain A). First, the structure was prepared for the simulation by adding hydrogen atoms and replacing the water molecules in the structure by a water sphere with a radius of 10 Å and centered at the protein's center of mass. The geometry of this model was optimized by 1,000 steps of energy minimization. Next, the molecular dynamics simulation modeled 2 ns of enzyme motion at 310 K and 1.013 bar, which required ~48 h of calculation time. The simulation results were visualized using the VMD molecular graphics program.52 The root mean square deviation (RMSD) and root mean square fluctuation (RMSF) values for backbone atoms were calculated and compared to the experimental RMSF value, which can be derived from the crystallo-graphic B-factors (B, temperature factor) according to equation 7.53,54
Incubation of wild-type SABP2 for 24 h at 4 °C in solutions of increasing urea concentration caused a decrease the fluorescence at 329 nm, Figure 3A. This decrease is consistent with protein unfolding to expose the indole ring of tryptophan to solvent, which quenches the fluorescence. SABP2 contains four tryptophan residues: Trp20 is completely buried within the catalytic domain, Trp131 is mostly buried within the cap-catalytic domain interface, Trp18 buried within the cap-catalytic domain interface as well as contacting the cap domain from the other monomer, and Trp220 is partially solvent exposed within the catalytic domain, Figure 2 above. Changes in the environment of these tryptophans decrease the fluorescence, but assigning this decrease to specific changes in structure is impossible because there are four tryptophans each with multiple possible conformations and overlapping electronic states. The midpoint of the decrease in fluorescence, which corresponds to equal amounts of folded and unfolded protein, occurred at 2.3 M urea. The slight increase in fluorescence at 0.5-1.0 M urea suggests a small conformational change, but the details of this change are unknown.
The denaturation in urea also decreased the hydrolytic activity (hydrolysis of pnitrophenyl acetate), but this decrease was reversible. When the hydrolytic activity was measured immediately after removing an aliquot from the urea solution, the midpoint of the decrease in enzymatic activity occurred at 2.4 M urea, Figure 3B, similar to the midpoint of the fluorescence decrease. Unfolding in urea required hours (data not shown), so the refolding is similarly slow since the ratio of the forward and reverse rates corresponds to the equilibrium constant. The ~5 min required for the assay is not enough time to refold significant amounts of protein. However, when the hydrolytic activity was measured after removing the urea by dialysis over 24 h, most (78%) of the original esterase activity was recovered, Table S2. This recovery indicates that the loss of hydrolytic activity is reversible by refolding of the protein. The incomplete recovery may be due to misfolding, aggregation or the formation of non-native disulfide links during the refolding. SABP2 contains five cysteine residues, none of them as disulfide links.
SABP2 also unfolded in solutions of guanidinium chloride (GdmCl), but at lower concentrations as compared to urea, Figure S1. The midpoint of unfolding occurred at 0.80 M GdmCl according to both the decrease in tryptophan fluorescence and the decrease in hydrolytic activity. While both urea and GdmCl disrupt hydrogen bonding, GdmCl also disrupts electrostatic interactions and shows a stronger unfolding effect toward most proteins.33
The changes in maximum fluorescence wavelength show a different pattern from the changes in fluorescence intensity described above. Buried tryptophan residues typically fluoresce maximally at 330 nm, while solvent exposed tryptophans fluoresce maximally at 350 nm.33a Thus, one expects the wavelength of the fluorescence maximum to shift from 330 to 350 nm as the protein unfolds. Indeed, this shift occurs for SABP2, but the midpoint for this shift occurs not at 2.1-2.4 M urea where the fluorescence intensity at 329 nm decreases and the hydrolytic activity is lost, but between 4 and 5 M urea, Figure 4A. Unfolding in GdmCl shows similar behavior, Figure S2. The midpoint for the decrease in fluorescence intensity at 329 nm occurred at 0.8 M GdmCl, but the midpoint for the shift in fluorescence wavelength occurred between 1.5 and 2.0 M GdmCl. A closer look at the unfolding curves in Figure 2A shows a second, smaller change in fluorescence intensity between 4 and 5 M urea and in Figure S1A a second change between 1.5 and 2.0 M GdmCl. The existence of two transitions indicates that unfolding of SABP2 involves at least one intermediate state.
Another way to show that unfolding involves an intermediate is to monitor the fluorescence changes at two wavelengths. If the native protein unfolds directly to the denatured form, then the fluorescence changes at all wavelengths should be linear. If they are not linear, then unfolding involves at least one intermediate. A plot of fluorescence intensity at 320 nm versus the fluorescence intensity at 365 nm was not linear indicating at least one intermediate state, Figure 4B. A plot of unfolding in GdmCl also showed similar, but less pronounced, nonlinear behavior, Figure S2B.
If protein unfolding occurs in one step (two states), the denaturant-induced shifts in the equilibrium constant can be extrapolated to 0 M denaturant to yield the unfolding free energy of the protein in water, ΔGuw, see equation 2 above. Although we know that unfolding of SABP2 involves an intermediate, we nevertheless fit the data to the one-step model, which yielded an unrealistically low value of 2.4 kcal/mol, Figure S3. Unfolding free energies of proteins similar in size to SABP2 typically range from 5-15 kcal/mol.34–36 This low value is typical for much smaller proteins of 50-70 amino acids. Using a one-step model in cases involving intermediates underestimates the true unfolding free energy.37
Fitting the data to a two-step unfolding model (three states) yielded a free energy of unfolding of SABP2 in water of 6.9±1.5 kcal/mol. The two-step unfolding model includes the relative amounts of folded, unfolded and intermediate states at various urea concentrations (equations 4 & 5) and extrapolates these values to 0 M urea. The slope of the lines is the constant m, which indicates the susceptibility of the protein to urea-induced unfolding. To avoid overfitting of the data, values of m were calculated from the surface area of the protein domains. The two-step model fit the fluorescence data more closely than the one-step model. The sum of squared residuals was 65% lower for the two-step model, Figure S3. This two-step model yielded a free energy of unfolding of 3.5±0.5 kcal/mol for the first transition (catalytic domain) and a similar 3.4±1.0 kcal/mol for the second transition (cap domain), Table 1. The sum of the two values, 6.9±1.5 kcal/mol is the free energy of unfolding of SABP2 in water. This value is similar to cutinase, a similarly sized enzyme from a mesophile, but lower than that for esterases from thermophiles, Table 1 above.
The m-values in the two-step unfolding model assumed that the first unfolding corresponds to the catalytic domain and the second unfolding corresponds to the cap domain. Reversing this assumption resulted in a poorer fit, with a sum of squared residuals 110% higher. This assumption is also consistent with substitutions in the catalytic domain causing changes in the first transition, see below. The larger domain unfolds first because it is more sensitive to urea.
Unfolding of SABP2 in GdmCl did not yield an estimate for the unfolding free energy in water. The unfolding also involves an intermediate, but the two steps are not sufficiently separated to measure the fluorescence properties of the intermediate. Fitting to the three-state model was not possible without this data.
Heating a sample of wild-type SABP2 at 1 °C/min while monitoring the circular dichroism (CD) signal at 224 nm showed a loss of signal as the protein unfolded. The midpoint for the change (T1/2CD) occurred at 49.0 ± 0.3 °C, Figure 5A. Similarly, incubation of wild-type SABP2 at elevated temperature for 15 min followed by cooling to room temperature decreased the catalytic activity, Figure 5B. This experiment measures the reversibility of the heat-induced unfolding. The midpoint for the decrease (T1/215 min) was the same to within experimental error as for the CD measurement: 49.2 ± 0.5 °C. Thus, the heat-unfolded protein does not refold quickly upon cooling. The half life at 60 °C was 3.5 min. Both transitions appear smooth with no suggestion of an intermediate state. The protein remains mostly soluble even above the transition temperature (Figure S4), although soluble inactive form may partly aggregate.
We chose locations for the proline substitutions in two different ways. The first was amino acid sequence comparison with a more stable homolog. Locations were the homolog has a proline residue should also accommodate a proline residue in the corresponding location of SABP2. The homolog chosen was hydroxynitrile lyase from Manihot esculenta (MeHNL). Both SABP2 and MeHNL are plant enzymes, they share 41% identical amino acid residues and both structures show an α/β-hydrolase fold where their main chains overlay with an rmsd of 0.75 Å, Figure S5. MeHNL is much more stable than SABP2.31,32 Its half-life at 50 °C (2.7 h) is more than ten times longer than SABP2. The temperature of heat induced unfolding as measured by circular dichroism spectroscopy (T1/2CD), 69.3 °C, is 20 °C higher than for SABP2. Alignment of the structures and amino acid sequences of SABP2 and MeHNL identified six locations where MeHNL contains a proline residue, but SABP2 does not: L46, S70, S115, K190, E215 and V227, Figure S5. All six of these substitutions were tested. Although the esterases from thermophiles in Table 1 are more stable than MeHNL, they share only 14-16% amino acid identity with SABP2, so the structures match less well to SABP2.
The second way to choose locations for proline substitutions was to add them in flexible regions within SABP2 because these regions should be able to accommodate the main chain conformations required by proline. One disadvantage of this approach is that introducing the proline in a flexible region will reduce its flexibility. Reducing the flexibility of the folded form decreases its entropy and stability, but we expected that this decrease would be smaller than the effect on the unfolded form. Other researchers have stabilized α/β-hydrolases by targeting random substitutions to the most flexible regions of proteins.60, 61 Our approach also targets the flexible regions, but only considers proline residues as replacements.
The B-factors from the x-ray crystal structure of SABP2, determined at 100 K, shows low average flexibility, Figures 6 and S6. Flexible regions include the N- and C-termini, a region in the cap domain (140-145) and two regions in the catalytic domain (66-73, and 180-203). A molecular dynamics simulation at 310 K, showed a higher average flexibility and a similar pattern. Regions 66-73 and 140-145 remain flexible, but the 180-203 region shifts to 160-180 and a region 30-60 also appears flexible. We targeted the proline substitutions to the three flexible regions identified by B-factors. The sequence comparison above already suggested proline substitutions in two of the regions: S70 and K190.
Among the six amino acid residues in the third flexible region (140-145), residues 142 and 145 are already proline. FoldX, sofware for modeling protein stability, predicted that a proline substitution at G140, E143 or E144 would destabilize the protein, but a substitution at S141 would stabilize the protein, Table S3. We chose the S141P substitution. The total of predicted locations for substitution is seven: six from sequence alignment and three from targeting flexible regions, but two predictions are common to both methods.
Site-directed mutagenesis created seven variants, each with a different single proline substitution. After expression of the proteins in E. coli and purification by Ni-affinity chromatography, SDS-PAGE analysis showed pure proteins.
The steady state kinetic constants for the single proline substitution variants were within a factor of two of those for wild-type SABP2, Table 2. The average kcat/KM values were 0.7±0.2 mM−1s−1 as compared to 0.81 mM−1s−1 for wild type SABP2. The largest changes for the kcat/KM value were a decrease by a factor of two to 0.41 mM−1s−1 for K190P and an increase by a factor of 1.2 to 0.97 mM−1s−1 for S141P. The largest changes in kcat value were a decrease by a factor of two for S115P and an increase by a factor of 1.5 for E215P. The largest changes in KM value were a decrease by a factor of 1.4 for S141P and an increase by a factor of 2 for V227P.
The increase in the concentration of urea needed to unfold half of the protein (ΔC1/2urea) indicated an increase in stability, Table 2. Due to the complexities of the unfolding, we did not convert these values to free energies. Variant E215P showed the largest increase in stability, ΔC1/2urea = +0.9 M, while variant K190P showed the large decrease in stability, ΔC1/2urea = −1.5 M. Both of these substitutions were in the catalytic domain suggesting that this first unfolding transition unfolds the catalytic domain. The changes for the other five variants were smaller, ranging from ΔC1/2urea = +0.4 M to −0.4 M. The average absolute effect on urea tolerance for mutations in the catalytic domain was larger (0.62 ±0.52 M urea) than for mutations in the cap domain (0.15 ±0.21).
The proline substitutions also increased the heat inactivation temperatures, T1/215 min, but the best variant differed from that identified by denaturation in urea. The largest increase in heat inactivation temperature was +6.4 °C for L46P, and the second best, +5.4 °C was S70P. Both these substitutions are in the catalytic domain. Similarly, the half-lives at 60 °C more than doubled for L46P, S70P, as well as for S115P and S141P. The worst variant was K190P, with a temperature decrease of 7.0 °C and half-life of under 2 minutes at 60°C. Variant K190P was also the worst one as measured by urea denaturation. This location was predicted by both targeting flexible regions and the sequence alignment.
Substitutions in the catalytic domain cause the largest increases and decreases in the first unfolding transition, thus identifying this transition as the unfolding of the catalytic domain, Figure 2 above. The two substitutions in cap domain had little effect on urea-induced unfolding. ΔC1/2urea was unchanged for S141P and decreased by 0.3 M for S115P indicating a slight destabilization. Two of the substitutions that targeted flexible regions (S70P and S141P) had little effect on the equilibrium in urea (ΔC1/2urea = +0.1 and 0.0 M, respectively), possibly because they reduce flexibility in both the folded and unfolded conformations. These two substitutions stabilized SABP2 to heat inactivation (ΔT1/215 min = +5.4 and +4.9 °C, and half-lives of 12.0 and 9.4 min at 60 °C, respectively), possibly by reducing aggregation.
Combining stabilizing proline substitutions did not further increase stability as measured by urea denaturation, but did increase stability dramatically as measured by heat inactivation, Table 2 and Figure S7. The most stable multiple substitution variant as measured by urea unfolding was L46P-S70P-E215P, but it was no more stable than E215P. The most stable multiple substitution variant as measured by heat inactivation was L46P-S70P with a T1/215 min of 74.9±1.7 °C, which is a dramatic 25.7 °C higher than wild type and a half life at 60 °C more than six times longer than wild type. Both substitutions are in the catalytic domain. Molecular dynamics simulation of SABP2-L46P-S70P folded conformation showed less flexibility in region 20-70 than wild type, Figure 4B above.
Unfolding of most α/β-hydrolases involves intermediates, but in a few case unfolding occurs in a single step (cutinase, esterases from thermophiles). The unfolding of SABP2 in urea was unusual because the intermediate can be clearly distinguished by fluorescence. The first unfolding transition affects fluorescence at 365 nm, while both the first and second transitions affect the fluorescence at 320 nm. Fitting these changes to a three state model yields a free energy of unfolding of 3.4 kcal/mol for the first transition and 3.5 kcal/mol for the second. The total, 6.9 kcal/mol, is typical for an α/β-hydrolase from a mesophile.
The two unfolding steps likely correspond to the separate unfolding of the two domains; surprisingly we found that the catalytic domain is probably the first to unfold. Multi domain proteins often unfold via intermediates where one domain is mostly unfolded while the other domain is still predominantly folded.40 While other possibilities cannot be ruled out, our data is consistent with one domain of SABP2 unfolding first, followed by the other domain. The loss of catalytic activity during the first unfolding is consistent with either domain unfolding since both domains contribute to the substrate binding site and disrupting of either would be expected to disrupt catalysis. To fit the fluorescence changes upon unfolding best required m values predicted by the surface areas of the two domains with the catalytic domain corresponding to the first transition, and the cap domain corresponding to the second transition. Substitutions within the catalytic domain have a large effect on the first transition, while substitutions within the cap domain have little effect. These results are also consistent with the first transition corresponding to the unfolding of the catalytic domain.
Heat induced inactivation showed no evidence for an intermediate. The unfolding by heat disrupts solvation and favors flexible unfolded forms, while unfolding with denaturant disrupts hydrogen bonding. It is possible, even likely, that urea-induced denaturation and heat-induced denaturation generate different unfolded proteins, so that changes to increase stability may affect the two measures differently.
Locations for proline substitution were chosen based on an amino acid sequence alignment between SABP2 and MeHNL. This approach showed a good success rate to stabilize against urea denaturation: (3/6 = 50%): substitutions L46P, S70P, and E215P. The success rate was slightly better in stabilizing against heat inactivation (4/6 = 67%): substitutions L46P, S70P, S115P, and E215P, Table 3. The three substitutions in the catalytic domain stabilized against both conditions consistent with notion that it unfoldes in both two conditions, while the substitution in cap domain on stabilized against heat denaturation constituent with the notion that the cap domain does not unfold first in urea.
Adding proline to flexible regions had a similar success rate. This approach suggested three substitutions: S70P, S141P and K190P, where S70P and K190P were also predicted by the sequence alignment. For urea denaturation, only S70P stabilized SABP2 (33% success), while for heat inactivation, both S70P and S141P stabilized SABP2 (67% success), Table 3. This approach required more effort - an x-ray crystal structure to identify the flexible regions and some computation to predict which location in these regions would be best.
Analysis of the secondary structure can also suggest locations for a proline,36,64 but we did not use this approach because it ignores the surrounding structure. Proline often occurs at the i+1 position of type I β-turns, at the i position in type II β-turns, as the N-terminal residue of α helices, and sometimes in other turn positions if the Ψ angle is near −60°. Analysis of the SABP2 structure using program PROMOTIF through PDB-sum (http://www.ebi.ac.uk/pdbsum/) identified 39 locations predicted to fit a proline residue.39 Five of these already contain a proline and three more were excluded because they are within 4 Å of the bound product, leaving 31 possible locations, Table S4. Two locations predicted above – S115 and E215 – are among these 31. S115 is at the i+1 position of a type I β turn and E215 is at the N-cap of an α helix. Both of these substitutions stabilized the protein. The other five predicted locations are not among the 31. Two of these destabilized the protein, but three stabilized it, corresponding to an overall success of 4/7= 57%. For example, residue V227 does not fit because it is in the i position of a type VIII β turn and location 141 is at the i position of a type I β turn. Location S70 may be a special case. Although it does not fit in SABP2 because it is at the i position of type I β turn, in MeHNL this location is a type II β turn, so substituting a proline may switch the conformation of the turn.
Our success with the structure approach (57%) is similar to that found by others. Using structural analysis, Prajapati and colleagues65 found 4 of 7, 0 of 1, and 2 of 4 proline substitutions (50% overall) increased thermal stability in the proteins leucine– isoleucine–valine binding protein, ribose binding protein, and thioredoxin, respectively. A meta analysis of all reported proline substitutions in the Protherm database66 found 33 proline substitutions in 10 proteins; sixteen of these (48%) increased thermal stability. Thus, the success of the structure approach matches the average of all approaches.
Homolog comparison was more successful for SABP2 (67%) and others also reported high success rates. For example, Watanabe and colleagues36 found 7 of 9 pro-line substitutions (78%) copied from a thermostable homolog increased stability of oligo-1,6-glucosidases. The improved success rate when copying from homologs suggests that structure analysis predicts the fit of proline replacements poorly, likely because it considers only secondary structure while ignoring the surrounding regions. Homolog comparison may be more successful because it adds information from evolutionary history. Destabilizing proline substitutions will be selected against; for that reason identifying proline locations in homologs may be more reliable.
Combining the substitutions yield a variant (L46P-S70P) that was dramatically stabilized to heat inactivation, but no more stable to urea denaturation. The urea stability of the double and triple mutants tested was always less than the sum of the single mutants, and usually less than the single constituent mutants, while the thermal stability was close to additive, and always greater than the individual mutations. Two possible hypotheses for this behavior are 1) thermal stability is measured at much higher temperature than chemical denaturation and 2) hydrogen bonds to the denaturant, not flexibility of the unfolded form, may drive unfolding due to chemical denaturation. Proline substitutions change the entropic advantage of unfolding and have little effect on hydrogen bonds to the denaturant. The stabilizing effect of proline substitution should be higher at higher temperatures because the entropy contribution to free energy increases a higher temperature. If hydrogen bond changes drive the unfolding in urea and guanidinium, then proline substitutions are not expected to stabilize the proteins since proline substitution does not affect hydrogen bonding. Others also reported that proline substitution improved thermal stability more than resistance to chemical denaturation.65,67
Despite the complexities of unfolding in multi-domain proteins like α/β hydrolases, the proline substitution approach is a reliable way to stabilize SABP2, especially to heat, and may be useful for other α/β hydrolases. Choosing locations by aligning amino acid sequences with a stable homolog is the easiest approach. Choosing substitutions in flexible regions was also successful, but relied on an x-ray structure. Stabilization to urea denaturation required substitutions in the catalytic domain, while stabilization to heat inactivation could be in either domain, but was most successful in the catalytic domain.
We thank the Minnesota Supercomputing Institute for use of computers and software.
Funding information: US National Science Foundation (CBET-0932762, CHE-1152804), US National Institutes of Health (1R01GM102205-01 and 5T32 GM08347), China National Natural Science Foundation (31470793; fellowship to JH)
Four tables and seven figures containing additional data. This material is available free of charge online at http://pubs.acs.org.