Re-engineering a β-lactamase using prototype peptides from a library of local structural motifs

Abstract

B. licheniformis exo-small β-lactamase (ESBL) has a complex architecture with twelve α helices and a five-stranded beta sheet. We replaced, separately or simultaneously, three of the ESBL α helices with prototype amphiphatic helices from a catalog of secondary structure elements. Although the substitutes bear no sequence similarity to the originals and pertain to unrelated protein families, all the engineered ESBL variants were found able to fold in native like structures with in vitro and in vivo enzymic activity. The triple substituted variant resembles a primitive protein, with folding defects such as a strong tendency to oligomerization and very low stability; however it mimics a non homologous recombinant abandoning the family sequence space while preserving fold. The results test protein folding and evolution theories.

Introduction

The three-dimensional (3D) structure of proteins is determined by the amino acid sequence,1 but the underlying principles of specific folding are poorly understood. A theory of folding should explain several apparently contradictory features of the specific relationship between sequence and folding. For instance: (a) proteins tolerate a surprisingly large number of mutations2–8; (b) there are several examples of similar protein folds encoded by completely different sequences9; (c) conversely, proteins with highly similar sequence but different folds have been created10–15; (d) novel folds can be generated by nonhomologous domain or subdomain recombination16–18; (e) identical sequences of 8–11 residues can adopt different structure in different proteins17, 19, 20; (f) well documented examples of alternative native folds for the same sequence are known10, 17, 21; and (g) specific long-range tertiary interactions might not be the main determinant of protein conformation.22

Yet, protein functions and folds evolve by sequence changes, and therefore there should be a way to reconcile all the above properties in a single folding theory. To that end, evolutionary studies and experiments aimed to test under controlled conditions the relationship between sequence and structure are in great demand.

Insertions, duplications, and deletions of sequence contribute to the structural and functional diversity of proteins, and have been relatively well studied in model proteins.23 In this regard, exchange of gene segments seems to have been instrumental in protein evolution. It has been suggested that the present protein domains have evolved by assembly and/or exchange of small gene segments. This process can also be simulated in vitro. A range of methodologies is available for this purpose.16, 24

Interestingly, an attempt to exchange pieces of helical elements between two proteins by random nonhomologous recombination failed to produce active proteins, leaving open the question as to if local structure fitness is not sufficient to override defects in tertiary interactions.24

Most of the above studies aimed to discover how to evolve new folds (i.e., how to explore the fold space), or to explore sequence space within a fold (i.e., how to introduce sequence variability). Inherent to them is the idea of the existence of a sequence space associated to a protein family. Less attention was paid to the exploration of how a fold is preserved from the flow of sequences from unrelated folds that may occur by nonhomologous recombination.

Herein, we describe an exercise of protein engineering designed to test if a family fold can tolerate substitution of parts of its sequence by standard sequences from unrelated families.

ESBL (Fig. 1), a 29 kDa, cysteine-free protein, possesses two domains and a complex architecture.25, 26 The α+β domain (residues 26–60 and 216–295) is sequentially discontinuous and comprises a central, five-stranded β sheet and three superficial α helices. The α domain (residues 61–215) is a globular array of helical elements. The catalytic site is located in the interface between the two domains. ESBL is a very stable protein that under denaturing conditions populates several partially folded states.

Figure 1. The structure of ESBL. The cartoon depicts the α (left) and α+β (right) domains and the location of α helices 9, 10, and 12 (in black).

For the exercise, we selected ESBL helices 9, 10, and 12 (see Fig. 1) as annotated in the pdb entry 4blm. The first two helices are in the α domain (residues 184–196 and 201–212), run head to tail into each other, and are separated by a β turn. Helix 12 is located in the α+β domain and comprises the last 20 residues of the protein (276–291).

We will show that ESBL is highly tolerant to sequence substitution, extending, and confirming our previous discovery that no part of ESBL sequence is essential to the lactamase fold.22 We will also show that the exercise can mimic the creation of a protoenzyme that might further evolve to an efficient catalyst crossing sequence family boundaries while preserving the essential features of the fold.

Results

Prototype peptide selection

The I-sites library collects protein sequences of 3–19 residues that correlate strongly with local structural motifs and are not limited by family boundaries.27, 28 Each structural motif is associated to several sequence clusters, which are in turn associated to a paradigm sequence. In this work, the I-sites library was used as a source of archetypal sequences, to avoid subjective bias in choosing the mutations to be introduced into the lactamase α-helical segments.

The selected sequences (see Fig. 2) have the propensity to acquire the proper local structure but exhibit no homology to the lactamase family. Using standard similarity scores, a vast majority of sequences in the data bank are more similar to the corresponding ESBL sequences than the chosen archetypal peptides. Moreover, the E-values for the actual and a perfect match differ by 8–21 decimal logarithms. These figures rule out any possible evolutionary relationship between the replaced sequences and their substitutes.

Figure 2. Features of ESBL helices 9, 10, and 12. The original and substitute sequence are shown in upper case. The E value represents the number of sequences expected by chance to yield scores as those calculated for the ungapped alignment (in parenthesis are the E values for perfect matches). The engineered helices are shown in a wheel representation with ESBL and prototype residues placed in the outer and inner layer, respectively. Hydrophobic residues are highlighted by a gray circular area.

Like the target helices, the selected archetypal peptides are amphiphilic. However, the latter are not optimized for long range interactions within ESBL. A wheel representation (see Fig. 2) shows that the coincidence in the hydrophobicity profile is reasonable for α₉ but less than satisfactory for α₁₀ and α₁₂ ESBL. To alleviate this problem and properly fit into the ESBL native body, α₁₀ and α₁₂ may change side-chain rotamers and also undergo previously described readjustments such us “register shift” (also named “residue translocation”) or “loop out” (also named “bulging”).29

Protein expression and purification

Wild-type, α₉, α₁₀, α₁₂, and α_9,10,12 ESBL were expressed in E. coli with high yield (200–300 mg of protein per liter of culture). As previously reported,30 wild-type ESBL is expressed as a folded protein and can be recovered soluble from the osmotically sensitive compartment of the bacterium. In contrast, α₉, α₁₀, α₁₂, and α_9,10,12 ESBL were mostly directed to inclusion bodies and, accordingly, purified from that fraction after urea solubilization.31 Judging from the SDS-PAGE analyses, the proteins were purified to homogeneity (not shown). Mass spectrometry results for α₉, α₁₀, α₁₂, and α_9,10,12 ESBL were 29,602.4 ± 1.1, 29,408.3 ± 1.5, 29,785.1 ± 1.1, and 29,774.8 ± 1.2 Da, respectively, identical within 1 Da to the masses calculated from the corresponding sequences. Refolding was accomplished by dialysis, yielding 100% of soluble protein.

Hydrodynamic behavior

The aggregation state of the mutants was investigated by analytical size exclusion chromatography under nondenaturing conditions (not shown), and the R_s values calculated from the chromatograms are given in Table I. As previously reported, wild-type ESBL behaves in this analysis as a mono disperse spherical molecule the size of a monomer.30, 31 The chromatograms indicated that α₉ and α₁₀ ESBL populate monomers (∼80%), dimers (∼10%), and higher order aggregates (∼10%). This tendency was reversed in the case of α₁₂ ESBL (∼60% aggregates and ∼30% monomer). Reinjection of the collected monomeric fractions evidenced that the aggregation of these variants is slow, compared with the chromatographic time, and that the hydrodynamic behavior of the monomers is similar to that of wild-type ESBL (Table I). A stronger aggregation tendency was observed for the triple mutant, which populates mostly high-order aggregates. However, aggregation of α_9,10,12 ESBL was much less pronounced in the presence of 250 mM Na₂SO₄ (∼50% monomer).

Table I. General Properties of ESBL Variantsa

	RSb (Å)	Qc	λmaxd (nm)	Enzymic activitye (%)	ANS bindingf
a Except for the triple mutant, the reported values correspond to the monomeric fractions isolated from the refolding solution by size exclusion chromatography. For the triple mutant, values correspond to the unfractioned refolding solution in the presence of 250 mM Na2SO4, and the reported RS is that of the monomeric peak. The errors are the standard deviation of 2–3 determinations. b Determined by size exclusion chromatography. c Fluorescence quantum yield. Q of free tryptophan in aqueous buffer is 0.13.32 d Emission peak upon excitation at 295 nm. e Hydrolysis of benzylpenicillin, as percentage of wild-type specific activity. f A variant was considered negative for ANS binding if the titration assay (see Materials and Methods) yielded an affinity constant larger than 500 μM.
Wild type	25.3 ± 0.7	0.24 ± 0.02	334	100	No
α9 ESBL	26.9 ± 2.0	0.16 ± 0.02	333	10.2 ± 3.0	No
α10 ESBL	27.9 ± 2.0	0.31 ± 0.02	331	10.1 ± 2.5	Yes
α12 ESBL	28.0 ± 1.2	0.17 ± 0.02	331	2.2 ± 0.04	No
α9,10,12 ESBL	26.0 ± 1.0	0.15 ± 0.02	331	0.0022	Yes

Optical properties

The secondary structure of the mutants was assessed by far UV CD (see Fig. 3). α₁₀ ESBL shows a α-helix content similar to that of wild type ESBL, whereas α₉ and α₁₂ ESBL exhibit a slightly decreased content of that structure. The triple mutant spectrum showed a small blue shift of the minimum, suggesting a higher content of β structure. However, this result must be interpreted with caution because the analyzed triple mutant sample contains mostly high-order aggregates.

Figure 3. Far (Panel A) and near (Panel B) UV CD spectra. Full line, wild-type ESBL; dots, α₉ ESBL_; short dashes, α₁₀ ESBL; dots/dashes, α₁₂ ESBL; and short/long dashes, α₉,_10,12 ESBL.

The near-UV CD spectrum of wild-type ESBL is strong and structured, with distinct shoulders and minima ascribable to tryptophan and tyrosine rotatory power. Although about half as intense, these features are present in the spectra of all the single-helix mutants (see Fig. 3), suggesting that most of the tertiary structure of ESBL is preserved in them. On the contrary, little rotatory strength and structure is seen in the triple mutant spectrum, which indicates that only a small fraction of the molecules adopts a tertiary structure.

Fluorescence features of ESBL are dominated by the emission of its three tryptophan residues (at sequence positions 210, 229, and 259, with 2, 16, and 12% solvent accessible surface, respectively; Fig. 4). Like the wild-type protein, the mutants exhibit emission maxima consistent with solvent-shielded tryptophan residues (Fig. 4; Table I). However, the Q value for α₉, α₁₂ and α_9,10,12 ESBL is somewhat lower than that of wild-type ESBL, which suggests an increased conformational fluctuation for these variants. In contrast, the Q value for α₁₀ ESBL is significantly larger than that for wild-type ESBL. A preliminary analysis of tryptophan mutants of ESBL (VAR and MRE, unpublished results) suggests that Trp 229 is the ESBLs main light emitter, whereas Trp 210 and 251 emission is significantly quenched. In this scenario, the increased Q value described above for α₁₀ ESBL would be the result of the mutation of Lys 212, whose NE atom is located at 4.9 Å of to the center of the aromatic ring of Trp 251 and likely responsible for the low Q value of the latter. Work is in progress to confirm this hypothesis.

Figure 4. Fluorescence spectra. Full line, wild-type ESBL; dots, α₉ ESBL_; short dashes, α₁₀ ESBL; dots/dashes, α₁₂ ESBL; short/long dashes, α₉,_10,12 ESBL; double dots/dashes, N-acetyl tryptophanamide. All samples were dissolved in Buffer A.

Binding of ANS

In aqueous milieu, ANS is a very poor fluorescence emitter, but it fluoresces intensely when bound to hydrophobic cavities and patches that become accessible in protein partially folded states.33, 34 For α₉ and α₁₂ ESBL titrated with ANS, the fluorescence of the dye increases steadily and it was not possible to obtain saturation curves in the micromolar range of ligand concentrations (not shown), which indicates that the affinity of these variants for the dye is in the millimolar range, a value larger than that usually observed for typical molten globules. We conclude that the conformational ensemble populated by these variants contains a fraction of molten globule-like states, but either these states are in equilibrium with a much larger fraction of natively folded molecules or themselves are more native-like than typical molten globules. On the contrary, α₁₀ ESBL and the triple mutant yield saturation isotherms (not shown), from which K_D of 250 and 80 μM, respectively, were calculated. Since this is characteristic of ANS binding to molten globules, we conclude that these mutants populate in solution a large fraction of molten globule state.

Enzymic activity

The lactamase active site is made of residues from different parts of the sequence. The number and the intricacy of its geometric constraints indicate that the conformational integrity of most of the two domains is required for catalysis, and that ESBL can only be enzymically active if able to fold in the native state (see Ref.22 for a detailed analysis of the structure-activity relationship for ESBL). For this reason, probing the enzymic activity of the ESBL variants permits evaluation of the degree of conformational perturbation of the native state introduced by the helix replacement. The results shown in Table I indicate that the α₉ and α₁₀ helix replacements are the less perturbing, since 10% of the wild type activity is measured for these variants. Replacement of the C-terminal helix is somewhat more detrimental to the native conformation, since α₁₂ retains only two percent of the wild-type activity. On the other hand, the triple mutant shows only traces of activity.

A more detailed analysis of the catalytic activity showed that all of the of the above mutants exhibit Michaelis-Menten behavior allowing calculation of k_cat and K_m (Table II) and comparison with the catalytic efficiencies of the unperturbed protein and with previously characterized mutants22 lacking altogether the helices under study.

Table II. Catalyitic Parameters of ESBL Variantsa

	kcat (s^−1)	Km (M)	kcat/Km (s^−1 M^−1)	log (kcat/Km)
a Data for CP181\195, CP201\212, and Δ277\295 are from Ref.22.
Wild type	2.4 × 10^3	1.3 × 10^−4	1.8 × 10^7	7.2
α9 ESBL	3.8 × 10^2	5.6 × 10^−4	6.9 × 10^5	5.8
CP181\195	5.0 × 10^−2	8.0 × 10^−5	6.2 × 10^2	2.8
α10 ESBL	5.4× 10^2	6.1 × 10^−4	8.8 × 10^5	5.9
CP201\212	6.6 × 10^0	2.6 × 10^−4	2.5 × 10^4	4.4
α12 ESBL	4.9 × 10^0	1.8 × 10^−3	2.7 × 10^3	3.4
Δ277\295	9.5 × 10^−2	2.5 × 10^−4	3.8 × 10^2	2.6
α9,10,12 ESBL	7.7 × 10^−2	3.1 × 10^−4	2.5 × 10^2	2.4

The k_cat/K_m value for α₉ and α₁₀ ESBL places these variants in the category of efficient catalysts. By the same criterion, α₁₂ ESBL can be considered a moderately good enzyme. The triple mutant, even though a weak catalyst, accelerates the hydrolysis of the substrate by a factor of 500 compared with the uncatalyzed reaction.22 Importantly, the catalytic efficiency of α_9,10,12 ESBL (log(k_cat/K_m)>1.622); suggests that it is a genuine enzyme, which in turn implies that at least a small fraction of the molecules of this mutant is still able to fold in a native-like structure. This was further confirmed by in vivo experiments (not shown): E. coli JM109 cells expressing α_9,10,12 ESBL grow in culture plates containing 16 μg/mL ampicillin, whereas no growth is observed the for the untransformed bacterium at 1 μg/mL ampicillin).

The comparison with circularly permuted and truncated ESBL variants lacking helices 9, 10, and 12 (Table II) clearly shows that replacing the sequences of these helices with standard sequences, perturbs less the native conformation than its elimination altogether. This suggests that the grafted helices fold and make a positive contribution to the context native conformation.

Urea induced unfolding

Figure 5 shows unfolding curves obtained by monitoring fluorescence, far UV CD, and specific activity for ESBL and its mutants. The α domain of ESBL concentrates ∼75% of the protein helical structure; therefore the CD signal should be particularly sensitive to changes in that domain. Fluorescence informs on the environment of tryptophan residues, of which there are one in the α domain and two in the α+β domain. Specific activity, as discussed above, reflects the existence of a functional active site built with the contribution of the two lactamase domains and a large number of sequentially disperse side chains. The unfolding data were globally fit to Eqs. (5)–(8), which are based on a three state mechanism. However, this simplification may not describe appropriately all the states involved in the unfolding of ESBL and its variants. Indeed, the mechanism is likely to be far more complex. Thus, the analysis below, only attempts to capture gross differences between ESBL and the variants, and to compare relative stabilities, without claming the true existence of only three states.

Figure 5. Urea induced unfolding equilibrium. From top to bottom, the curves for wild type, α₉, α₁₀, α₁₂, and α_9,10,12, respectively are shown. Panels on the left show the conformational changes monitored by far UV CD (circles), fluorescence emission (squares), and enzymic activity (triangles). A three-state unfolding mechanism was assumed to fit the data for the three probes simultaneously [see Eqs. (5)–(8)], except for the triple mutant for which only CD and fluorescence data were globally fit. Right panels show the changes in the equilibrium fraction of native (dashed line), partially folded (full line), and unfolded state (dotted line) derived from the fit.

The unfolding profile of wild type ESBL is in agreement with previous reports and reveals a very stable protein (∼10 kcal mol⁻¹ for complete unfolding; Table III). At low urea concentrations, wild type ESBL seems to populate an equilibrium intermediate with full activity and very subtle differences in fluorescence and CD properties. Indeed, only the global fit of the three probes allows detecting the presence of this intermediate, as a slight lack of register for the unfolding curves in the first pre transition; and it would be perfectly acceptable to consider this intermediate as part of a native state ensemble.

Table III. Urea-Induced Unfolding Parameters

ESBL	ΔG0,NI	ΔG0,IU	S0,I,CDa	S0,I,FL	S0,I,SA
a S0,I,CD, S0,I,FL, and S0,I,SA are the percentage of intrinsic unfolding of the state I (i.e., a value of 10 indicates that the I state preserves 90% of the signal characteristic of the folded state) as measured by CD, fluorescence, and specific activity, respectively. The units of free energy are kcal mol^−1. The parameters were calculated by nonlinear least square fit of the data shown in Figure 5 as described in Materials and Methods. Errors were calculated by Monte Carlo simulation as described.35
Wild type	5.7 ± 1.9	5.0 ± 0.4	10 ± 1	17 ± 2	0 ± 1
α9	2.2 ± 0.7	1.7 ± 0.2	60 ± 6	82 ± 8	82 ± 8
α10	1.8 ± 0.6	3.1 ± 0.3	16 ± 2	10 ± 1	93 ± 9
α12	1.6 ± 0.5	0.7 ± 0.1	0 ± 1	14 ± 1	42 ± 4
α9,10,12	0.0 ± 0.6	6.5 ± 0.5	47 ± 5	94 ± 9	—

The ESBL variants exhibit cooperative transitions but their global stability is significantly lower (40–60%) than that of the wild-type protein. The stability of the native state for the single-helix mutants is low 1.6–2.2 kcal mol⁻¹, but enough to populate the native conformation in the absence of denaturant. On the contrary, the triple mutant in this condition is mostly in a partially folded state. The features of the intermediate that becomes populated at intermediate denaturing conditions differ for each mutant: (i) α₉ ESBL intermediate state is nearly completely unfolded by fluorescence and enzymic activity criteria, but it retains 40% of the initial α-helix signal; (ii) the intermediate of α₁₀ ESBL is inactive but it retains most of the fluorescence and ellipticity measured in the absence of urea; (iii) α₁₂ ESBL intermediate exhibits almost full CD signal, slightly decreased fluorescence, and 60% of the activity displayed in the absence of denaturant; and (iv) the triple mutant exhibit a second intermediate state with a 50 and 100% percent reduction in ellipticity and fluorescence, respectively.

Altogether, these results show that the single-helix mutations, although greatly destabilize and affect the mechanism of unfolding, permit the subsistence of unfolding cooperativity and of a fraction of molecules with a native-like conformation.

Temperature melting

Further insight on the unfolding equilibrium of the ESBL variants was obtained by thermal unfolding and far UV signal monitoring. Under the conditions tested, thermal unfolding is reversible for wild-type ESBL and the single helix variants, but not for the triple helix mutant (not shown). Therefore, thermodynamic parameters could be derived from the unfolding curves (see Fig. 6) assuming two-state equilibriums. The experiment was conducted at pH 6.0, 7.0, and 8.0, and a global fit of Eqs. (3) and (4) to the whole set of data was performed yielding the parameters listed in (Table IV) and in supplementary material. In this analysis, it should be kept in mind that the CD signal only reports on the secondary structure of the protein and that change in the tertiary structure may go undetected. This is important because the changes in tertiary structure may occur at lower temperatures and the thermal unfolding curves likely represent only I↔U transitions.

Figure 6. Thermal unfolding. Panel A shows the unfolding curves for the ESBL variants wild type (squares), α₉ (triangles), α_10, (diamond), and α₁₂ (circles). For clarity, arbitrary constants were added to each set of experimental data to displace the curves along the ordinate axis. Lines are the best fit of Eqs. (3) and (4) (see Materials and Methods). Only the pH 7.0 data and fit are shown. However, the experiment was performed at pH 6.0, 7.0, and 8.0 (not shown), and the fit was global to the whole set of data. The thermodynamic parameters derived from the fit are shown in Table 4 (pH 7.0) and in Supplementary Material (pH 6.0 and 8.0). Panel B shows the free energy of unfolding as a function of temperature, as calculated from the thermodynamic parameters listed in Table 4. Full, dot, dash, and dot-dash correspond to wild type, α₉, α_10, α₁₂ ESBL.

Table IV. Thermal Unfolding Parameters at pH 7.0a

ESBL	ΔCP (kcal mol^−1 K^−1)	ΔHTm (kcal mol^−1)	Tm (°C)	Tmax (°C)	ΔGTmax (kcal mol^−1)
a Unfolding transitions as a function of temperature were monitored by CD at 220 nm and the unfolding curves (Fig. 6) analyzed as described in Materials and Methods. Tm and Tmax are the temperatures of melting (ΔG = 0) and maximum stability, respectively. ΔGmax is the ΔG at Tmax. The corresponding figures for the thermal unfolding conducted at pH 6.0 and 8.0 are given in the Supplementary Information. A global fit to Eqs. (3)–(4) for pH 6.0, 7.0 and 8.0 was performed with the additional constraint given by ΔHTm = ΔHTm,pH=7.0 + ΔCP (Tm − Tm,pH=7.0). The errors were calculated by performing four fits with one quarter of equally spaced points each time and averaging the results.
Wild type	3.91 ± 0.03	139.9 ± 0.5	66.9 ± 0.1	35.0	7.1
α9	0.49 ± 0.01	6.6 ± 0.9	37.9 ± 1.6	25.2	0.2
α10	0.86 ± 0.07	32.53 ± 1.9	55.7 ± 0.6	17.6	1.9
α12	0.37 ± 0.02	0.0 ± 0.1	22.6 ± 0.1	23.0	0.0

Unfolding of wild type ESBL is characterized by a high and positive ΔH with a ΔC_p = 3.91 kcal mol⁻¹ K⁻¹. ΔC_p values can be accurately predicted from structural parameters,36 and the ΔC_p so estimated for ESBL is 4.3 ± 0.2 kcal mol⁻¹ K⁻¹ (95% confidence interval). Furthermore, the ΔC_p of ESBL unfolding determined by differential scanning calorimetry is 3.8 kcal mol⁻¹ K⁻¹ (VAR and MRE, unpublished results). Thus, the results for wild type ESBL validate the procedure to evaluate the impact of the mutations on the thermodynamic of folding.

The data for α₉, α_10, and α₁₂ ESBL show a dramatic decrease in ΔC_p for the unfolding transition, which must be taken as the consequence of a great reduction in the exposed surface difference between the involved conformational states. This may be due to the variant's unfolded states being of a highly compact nature, or to the transition being from an expanded I state. Considering the CD spectra (see Fig. 3), the ANS binding, and enzymic properties, the pretransition state of the variants at room temperature is likely to be an ensemble containing a fraction of native-like molecules in equilibrium with solvated, partially unfolded conformations. This is particularly evident for α₉ and α₁₂ ESBL, which have a temperature of melting of 38 and 23°C, respectively, with a ΔG close to zero at 25°C. It is also evident from the melting curves that α₉ and α₁₂ ESBL undergo cold unfolding at temperatures above 0°C. On the other hand, α₁₀ ESBL is significantly more stable than the other two variants, with a T_m of 56°C and a ΔG of ∼2 kcal mol⁻¹ at 25°C.

It is worth to mention that in the pH range 6.0–8.0 the stability of wild-type and α₁₂ ESBL decreases with pH; whereas the opposite happens with α₉ ESBL, and no variation is seen for α₁₀ ESBL (see supplementary information). The pH effect is large, resulting in a ΔT_m for wild type, α₉, and α₁₂ ESBLof 4°, 17°, and 11°, respectively.

Discussion

We previously demonstrated that the ESBL fold can tolerate the elimination of virtually any local conformational information implicit in specific sequence segments, reinforcing the notion that the fold is encoded redundantly and robustly along the entire chain of proteins.22 We also proposed that one efficient way to endow fold robustness against deletions and mutations is modular organization, which would effectively protect the global fold from genetic disruption and greatly simplify searching the conformational space.22

Herein, we demonstrated that a protein fold is also highly tolerant to internal segment substitution by nonhomologous sequences from unrelated folds. This kind of tolerance may blurry the frontiers in the sequence space between different protein families.

The exercise that we performed illustrates DNA swapping, a general evolutionary move thought to have been of great importance in the generation of the extant molecular diversity.37–40 The target molecule was a well-characterized and highly stable enzyme, and the transferred sequences from unrelated proteins were chosen from a prototype library of secondary structural elements. It is shown for α_9,10,12 ESBL that the exercise leads to what can be considered a primitive enzyme, with structural defects, greatly diminished stability, and weak functional activity.

Structural defects of α_9,10,12 ESBL might be partially remedied by its oligomerization, as suggested for primitive proteins that need to bury excessive surface hydrophobicity.16 The engineered enzyme, however, must retain the essential features of the parent fold because the latter is necessary to construct a functional active site.22

Further directed evolution of α_9,10,12 ESBL showed that a few point mutations, far in sequence from the engineered helices, can significantly improve the activity of the enzyme in vivo (VAR and MRE, unpublished results). Since it is so unlikely that an entirely new fold could have been generated while preserving function and in a single attempt, we can only conclude that it is possible to cross the borders between presumed basins in sequence space without changing the essence of the fold.

Although the general relevance of our results must be confirmed for other folds and functions, they are to be taken into account for the construction of a general protein folding and evolution theory.

ASA, accessible surface area; CD, circular dichroism; ESBL, B. licheniformis exo-small β-lactamase; MS, mass spectroscopy; N, U, and I, native, unfolded, and partially folded states, respectively; PAGE, polyacrylamide gel electrophoresis.

Materials and Methods

General details

Benzylpenicillin was from Sigma (St. Louis, Missouri). Protein purity was assessed by SDS-PAGE. Enzymic activity was determined at 25°C (Δε_{240 nm} = 570 M⁻¹ cm⁻¹41) in 50 mM sodium phosphate, pH 7.0 supplemented with 1.5 μM bovine serum albumin and containing 0.5 mg/mL benzylpenicillin. Least-square fit was done using the Solver module in Microsoft Excel 2000. Mass spectroscopy (MS) was performed on a VG Quatro II (VG Biotech, Altrinchan, UK) triple quadrupole instrument equipped with an electrospray ionization source. Unless otherwise indicated, the nonconsecutive residue numbering system of Ambler42 was used for the ESBL sequence. The ESBL substitute sequences are paradigm sequences obtained by searching the I-sites (invariant or initiation sites) library (http://www.bioinfo.rpi.edu/∼bystrc/Isites/index.html) using the wild-type ESBL sequence as the query.

The wild type and corresponding prototype peptides were aligned without gaps, and the sequence similarity was calculated using the blosum62 matrix.43 The resulting score, S, was used to calculated the expected random matches from the universe of known sequences as

• (1)

and

• (2)

where m is the length of the aligned segment, and n is the number of residues in the data bank. n, λ, and K are scaling parameters estimated by running a Blast search in the NCBI protein bank using the ESBL peptides as the query.

Protein expression and purification

DNA sequences encoding ESBL variants were prepared by PCR mutagenesis using Platinum Pfx DNA polymerase (Invitrogen, Corp.), appropriate primers, and pELB3 as template.30, 44 PCR products were cut with restriction enzymes and ligated into the XbaI/BamHI site of pET9a generating the expression plasmids pELBα_9,10,12, pELBα₉, pELBα₁₀, pELBα₁₂. Protein expression and purification was done as described earlier.30 The variants purified from inclusion bodies dissolved in 6.5 M urea were refolded at 4°C by dialysis against 100 mM sodium phosphate, pH 7.0 (buffer A) (α₉ and α₁₀ ESBL were refolded at 1 mg/mL, and α₁₂ and α_9,10,12 at 0.3 mg/mL).

Hydrodynamic and optical studies

Analytical size exclusion chromatography30 was carried out at 22°C using Buffer A. UV-absorption and CD spectra were acquired and processed following published procedures.45 Unless otherwise indicated the buffer for optical measurements was 25 mM sodium phosphate, 100 mM sodium fluoride, pH 7.0 (Buffer B) and the temperature was set to 20°C. Near-UV measurements were carried out with a 1.0-cm cell containing 15-μM protein. In the far UV, cell path, and protein concentration were 0.1 cm and 1.5 μM, respectively. The final spectra were smoothed using a forty-point moving window and a 4th order polynomial.46

Fluorescence measurements were made at 20°C with a K2 ISS spectrofluorometer (ISS, Champaign, IL). Protein solutions (3 μM) were prepared in Buffer A. Excitation was at 295 nm (8 nm bandwidth), and data were acquired at 1-nm intervals between 250 and 450 nm. Quantum yield (Q) was calculated as described previously.31

Equilibrium unfolding

Unfolding transitions as a function of temperature were monitored by CD at 220 nm. Protein concentration was 1.5 μM, and a 1.0-cm cell was used. Buffer B was adjusted to pH 6.0, 7.0, and 8.0. Temperature was varied from 0 to 95°C with a rate of 2°C min⁻¹, and the melting curve was sampled at 0.2-min intervals. Assuming equilibrium with only native (N) and unfolded (U) states (N↔U), data were fit to the following equations47:

• (3)

and

• (4)

where f_U and f_N are the unfolded and folded fractions, T_m is the temperature at which f_U = f_N, S is the observed CD signal, S_0,N and S_0,U are the intrinsic spectroscopic signals for the native and unfolded state, respectively, and l_N and l_U are the slopes for the assumed linear dependence of S_0,N and S_0,U with the temperature, respectively. The fit was performed simultaneously for all three pH values, with a global ΔC_P and pH specific energy and signal parameters.

Isothermal unfolding experiments were carried out incubating the ESBL variants with 0–8 M urea in Buffer A for 3 h at room temperature and then measuring CD, fluorescence, and enzymic activity. To avoid refolding during the measurement, enzymic activity was measured at 10°C in a 15-s assay as described before.30

For data analysis, a three-state unfolding mechanism with a partially folded state (I) at equilibrium with N and U was assumed. The raw optical values for 0 and 8 M urea were as expected for native and fully unfolded states, respectively. This allowed normalization of the data to unfolded fractions. The following equations48 were used in the simultaneous fit of the three normalized signals:

• (5)

• (6)

• (7)

and

• (8)

where f_N, f_I, and f_U are the fractions of native, partially folded and unfolded state, ‘Tr’ stands for NI or IU, K_Tr are equilibrium constants, D is the denaturant concentration, C_Tr is the denaturant concentration at which ΔG_Tr is zero, m_Tr is the slope of the linear dependency of ΔG_Tr on denaturant concentration, ΔG is ΔG_Tr at zero denaturant concentration, S₀ is the value of the signal for each state extrapolated to zero denaturant concentration, and l is the denaturant dependence of the intrinsic signal for each state. The fit was performed with C_i and m_i as common parameters for each variant, whereas the parameters related to the signals were specific to each particular probe. The fit was constrained by fixing m_NI+m_IU = 3.0, the predicted value for a protein the size of ESBL.36 Without this constraint, the fit converged to unreasonably high values of m and ΔG.