Increasing Enzyme Stability and Activity through Hydrogen Bond-Enhanced Halogen Bonds.

The construction of more stable proteins is important in biomolecular engineering, particularly in the design of biologics-based therapeutics. We show here that replacing the tyrosine at position 18 (Y18) of T4 lysozyme with the unnatural amino acid m-chlorotyrosine ( mClY) increases both the thermal stability (increasing the melting temperature by ∼1 °C and the melting enthalpy by 3 kcal/mol) and the enzymatic activity at elevated temperatures (15% higher than that of the parent enzyme at 40 °C) of this classic enzyme. The chlorine of mClY forms a halogen bond (XB) to the carbonyl oxygen of the peptide bond at glycine 28 (G28) in a tight loop near the active site. In this case, the XB potential of the typically weak XB donor Cl is shown from quantum chemical calculations to be significantly enhanced by polarization via an intramolecular hydrogen bond (HB) from the adjacent hydroxyl substituent of the tyrosyl side chain, resulting in a distinctive synergistic HB-enhanced XB (or HeX-B for short) interaction. The larger halogens (bromine and iodine) are not well accommodated within this same loop and, consequently, do not exhibit the effects on protein stability or function associated with the HeX-B interaction. Thus, we have for the first time demonstrated that an XB can be engineered to stabilize and increase the activity of an enzyme, with the increased stabilizing potential of the HeX-B further extending the application of halogenated amino acids in the design of more stable protein therapeutics.

Over the past two decades, recombinant proteins have been developed as biotechnology medicines to treat human diseases; the first such example, insulin, entered clinical trials in 1980 to treat diabetes.[1] Today, the list of approved recombinant therapeutics has grown to include human growth hormone to treat developmental disorders,[2] interferon to control multiple sclerosis and other autoimmune diseases,[3] and erythropoietins to treat anemia associated with kidney disease or chemotherapy,[4] but the numbers remain relatively limited. One of the primary challenges to the broader adoption of therapeutic proteins is the need to increase the stability of recombinant proteins to improve on their formulation and shelf lives, while maintaining their activities or efficacies.[5]

Efforts to increase or enhance the stability of biological molecules are limited by the molecular tools provided by nature. Many approaches to stabilize proteins have been developed,[6,7] including directed evolution with the canonical amino acids,[8] applying principles learned from proteins of extremophiles,[6] or attaching proteins to a matrix material, such as surfaces or polymers.[9] While these approaches have all seen some success, it is rare that adding an electrostatic interaction would stabilize a protein by significantly more than 1 kcal/mol.[10−12] The advent of methods for incorporating noncanonical building blocks into proteins[13] has helped to overcome some limitations but continues to be constrained by the standard menu of noncovalent interactions that dictate molecular folding. We show for the first time that a more thermally stable protein [in this case, the model T4 lysozyme, or T4L (Figure A)] with increased enzymatic activity at elevated temperatures can be engineered by adding a single halogen atom to a tyrosine (Y) residue. This increased stability and activity can be attributed to a previously uncharacterized form of the halogen bond (XB) interaction in which an intramolecular hydrogen bond (HB) from the tyrosyl OH enhances the XB potential of the halogen.

Figure 1

Crystal structure of pseudo-wild-type T4 lysozyme (WT*). (A) Overall structure of WT*, with the polypeptide backbone traced as a green ribbon. Atoms of the tyrosine 18 (Y18) residue, along with the interacting residues glutamate 11 (E11), arginine 14 (R14), and glycine 28 (G28), are shown as ball and stick models colored according to atom type (green for carbon, red for oxygen, and blue for nitrogen). Two water molecules that bridge the hydroxyl substituent of Y18 and the carbonyl oxygens of E11 and G28 are shown (interactions shown as dashes). The carbonyl oxygen of G28 forms a standard HB with the backbone amino nitrogen of R14 (shown as a dash, with an O···N distance of 3.1 Å). (B) Details (as a wall-eye stereoview) of the interacting water molecules that bridge from Y18 to E11, G28, and R14 (labeled W1–W6).

XBs are analogous to HBs[14] and, although the physicochemical foundation remains debated,[15−17] their electrostatic nature is readily modeled by the σ-hole theory[18] (Figure ). In this model, one lobe of a halogen’s p orbital becomes depopulated when its valence electron is subsumed by the molecular σ-orbital of a covalent bond. The result is an electropositive σ-hole, which serves as the XB donor. However, the electronegative annulus around the waist makes the halogen amphoteric,[19,20] able to serve simultaneously as an HB acceptor perpendicular to the σ-hole.[21,22]

Figure 2

Amphoteric property of halogen substituents. The anisotropic charge distribution, as predicted by the σ-hole model,[18] allows the halogen to accept an HB from a hydroxyl and donate an XB to a carbonyl oxygen.

In biology, XBs define the binding specificity and affinity of halogenated enzyme inhibitors[23] and have been shown to affect the folding of nucleic acids,[24−26] making them important tools for both medicinal chemistry[27−30] and biomolecular engineering.[20,31] To study the effects of XBs on protein folding, we had previously replaced the OH of a tyrosine (Y18) with an XB-forming halogen[32] and showed that the interaction will partially rescue the stability of T4L, relative to a site that cannot form an XB, but not back to that of the parent (WT*) enzyme. The Y18 hydroxyl is neighboring but not sufficiently close to form HBs with carbonyl oxygens of glutamate 11 (E11) and glycine 28 (G28). Instead, these groups are bridged by water molecules from Y18 to E11 and to G28 (W2 and W1, respectively, in Figure B), which were found to be critical to both the stability and the function of T4L.[32] When the OH of Y18 is removed (as in a Y18F construct), W1 shifts in position to effectively fill the void to partially rescue the stability of the protein. Replacing the OH of Y18 with a methyl group (as in a Y18MeF construct) results in the loss of W2 and a large associated decrease in protein stability. There are now >300 studies of T4L point mutations that demonstrate the power of this system to study structure–energy–function relationships in a well-controlled protein,[33] and it is the rare instance that an engineered mutation, including those with halogenated amino acids,[32] results in significant stabilization (≥1° increase in the melting temperature, TM) of the enzyme.

In this study, a series of T4L constructs were designed in which Y18 is halogenated at the meta position (Y18-T4L, where X = Cl, Br, or I), leaving the critical OH substituent intact. The halogens are placed to potentially form an XB to the carbonyl oxygen of G28 located in a tight loop region near the enzyme active site and thus enhance T4L stability. The G28 oxygen forms a standard (3.1 Å) HB to the backbone amino nitrogen of arginine 14 [R14 (Figure A)]. However, a small void space was identified within this loop region that could potentially accommodate a halogen to form an XB to this G28 oxygen in a geometry that is perpendicular to the O···H–N HB. The theory that XBs are orthogonal interactions to this type of HB[34] would suggest that the halogen interaction should not disrupt the protein interaction. We had previously shown that the energy of an O···H–N HB is not perturbed with the addition of an XB, and vice versa. Our prediction, therefore, was that a Y18-T4L could potentially form an XB to G28, which would augment the existing HBs and, thus, increase the stability of the overall protein.

Ohtake et al.[35] had previously shown that incorporating seven Y residues (X = Cl or Br) increased glutathione S-transferase stability by ∼5.5 kcal/mol, not through XBs, but by filling void spaces in the protein core and enhancing the HBs already present through standard inductive effects. Unnatural amino acids as ligands have previously been shown to form stabilizing XBs to protein backbones,[36] as many other halogenated ligands do.[27,37−39] We demonstrate here that introducing a single Y into a protein structure can by itself increase the thermal stability of T4L by 3 kcal/mol and that this increased stability can be attributed to an HB-enhanced XB (abbreviated here as an HeX-B to simplify discussion). This distinctive interaction was revealed from quantum mechanical (QM) analyses of the crystal structures of Y18-T4Ls and their relationships to protein stability and enzymatic activities.

Materials and Methods

Site-Directed Mutagenesis and Protein Expression

All T4 lysozyme (T4L) constructs started with the gene of pseudo-wild-type (WT*) protein, the T4L double mutant C54T/C97A,[40] with the DNA sequence encoding a six-His tag appended at the C-terminus to facilitate protein purification. The Y18-T4L constructs (Y18, Y18, and Y18) had the codon for Y18 replaced with an AMBER (TAG) codon.[41] The modified DNA sequences were inserted into the pBAD vector for DNA amplification in DH5α Escherichia coli.[32]

The expression vector for the WT* construct was transformed into BL21 (DE3) pLysS E. coli. Transformed cells were grown in 2× YT medium with the appropriate antibiotics (ampicillin and chloramphenicol) and incubated at 37 °C while being shaken at 250 rpm until an OD600 of 0.4–0.6 was reached. The cells were induced with the addition of arabinose directly to the cultures to a final concentration of 0.2% (w/v) and allowed to grow for an additional 3 h. Subsequently, the cells were harvested by centrifugation at 2200 relative centrifugal force (RCF). Thereafter, the supernatant was decanted, and the bacterial pellet was stored at −80 °C.

Expression vectors for the Y18-T4L constructs were co-transformed into BL21ai E. coli with the pDule2-Mb-ClTyrRS-C6 plasmid that contains the orthogonal Mb tRNACUA and 3-halo-Tyr amino acyl-tRNA synthetase. After being rescued, the transformed cells were stored at −80 °C. Starter cultures of NIM medium[42,43] containing appropriate antibiotics (ampicillin and spectinomycin) were inoculated with these cell stocks and allowed to grow at 37 °C for 12 h while being shaken at 250 rpm. Then, 5 mL of the starter cultures was used to, in a 2 L culture flask, inoculate 500 mL of AIM[42,43] medium containing the appropriate antibiotics (ampicillin and spectinomycin), but lacking arabinose. After inoculation, the cultures grew at 37 °C while being shaken at 250 rpm. When an OD600 of ∼1.0 was reached, the noncanonical amino acid (3-chloro-l-tyrosine, 3-bromo-l-tyrosine, or 3-iodo-l-tyrosine) was added to the cultures to obtain a final concentration of 1.0 mM. The 3-halo-l-tyrosines were supplied from Ark Pharm, Inc. The cultures continued to grow at 37 °C while being shaken at 250 rpm. When an OD600 of 3.0–4.0 was reached, the cultures were induced with a final concentration of 0.2% (w/v) arabinose. After induction, the bacterial growth was continued for 3 h at 37 °C while the sample was being shaken at a reduced speed of 100 rpm. Finally, after expression for 3 h, the cells were harvested by centrifugation at 4000 RCF, the supernatants were decanted, and the bacterial pellets were stored at −80 °C.

Protein Purification

The frozen bacterial pellets were suspended in 35–45 mL of a 9:1 buffer A/buffer B mixture [buffer A consisting of 40 mM potassium phosphate (pH 7.4), 500 mM sodium chloride, and 0.02% (w/v) sodium azide and buffer B consisting of 40 mM potassium phosphate (pH 7.4), 500 mM sodium chloride, 500 mM imidazole, and 0.02% (w/v) sodium azide] and thawed in a 37 °C water bath for 15 min. Subsequently, the cells were lysed by sonication on ice for 3 × 30 s using a Branson Sonifier 450 sonicator (duty cycle of 70%, output control of 7). After cell lysis, the homogeneous suspension was centrifuged in a Beckman model J2-21 centrifuge equipped with a JA-20 rotor at 16000 rpm and 4 °C for 30 min. The supernatant was decanted and filtered twice, first through a 0.45 μm pore syringe filter and thereafter through a 0.22 μm pore filter. The filtered cell lysate was loaded, applying 10% buffer B, onto a 5 mL HisTrap HP column on an ÄKTA start FPLC system. Nonbound protein was washed out with 15% buffer B over 5 column volumes. The His-tagged T4L construct was eluted with a gradient of 20 to 100% buffer B over 13 column volumes. Selected fractions were combined and concentrated to 1 mL in an Amicon Ultra-15 10K (Millipore) centrifugal device [10000 molecular weight cutoff (MWCO)] in an Eppendorf 5810 R centrifuge at 4000 rpm and 4 °C. The concentrated protein solution was then loaded onto a gravity-fed Sephadex G-50 fine column equilibrated in buffer specific for either crystallization or differential scanning calorimetry (DSC) [crystallization buffer consisting of 100 mM sodium phosphate (pH 7.0), 500 sodium chloride, and 0.02% (w/v) sodium azide and DSC buffer consisting of 20 mM glycine-HCl (pH 3.5), 80 mM sodium chloride, and 1 mM EDTA]. After gel filtration, the selected fractions were combined and used for crystallization or DSC experiments.

Protein Crystallization

After gel filtration purification using the crystallization buffer, described above, the combined and selected fractions were concentrated to 13–20 mg/mL using an Amicon Ultra-15 10K (Millipore) centrifugal device (10000 MWCO) in an Eppendorf 5810 R centrifuge at 4000 rpm and 4 °C. Crystals of the Y18-T4L constructs were grown at 18 °C using the hanging drop vapor diffusion method with a 2:3 to 7:3 ratio of protein to precipitant solution [precipitant solutions consisting of 2.0–2.4 M potassium phosphate (pH 6.5–7.4), 50 mM 2-hydroxyethyldisulfide, and 50 mM 2-mercaptoethanol] with a final protein concentration of 8–10 mg/mL in a 3.5–4.0 μL total drop volume, similar to the process previously described.[32] Diffraction quality crystals grew after 2–5 days for the Y18 and Y18-T4L constructs and after ∼2 weeks for the Y18-T4L construct. The crystals were harvested using a cryo-loop, flash-frozen, and stored in liquid nitrogen until X-ray data were collected.

X-ray Data Collection and Structure Determination

X-ray diffraction data were collected on crystals held under a cryogenic nitrogen stream (100 K) on the Advanced Light Source (ALS) Beamline 4.2.2 at Berkeley National Laboratory (1.00 Å, Research Detectors Inc. complementary metal-oxide-semiconductor 8 M detector). Diffraction data from the ALS beamline were reduced using XDS[44] and the CCP4 suite.[45] X-ray data were phased by molecular replacement, applying the atomic coordinates of WT* [Protein Data Bank (PDB) entry 1L63(12)] as the starting model, yielding initial models with Rwork values that ranged from 29.6 to 39.5% and Rfree values that ranged from 30.9 to 41.2%. Subsequent refinement of the structure using the PHENIX suite of crystallographic software[46] resulted in final structures with Rwork values that ranged from 17.7 to 19.9% and Rfree values that ranged from 20.8 to 23.2% (Table S1).

Differential Scanning Calorimetry (DSC)

After gel filtration purification [DSC buffer (pH 3.5)], as described above, the combined fractions of the pure T4L construct were diluted to a concentration of 0.3 mg/mL with DSC buffer. Aliquots of 900 μL were prepared and stored at −80 °C. A low pH was used to help promote reversible folding.[47] Melting curves were collected on a TA Instruments Nano DSC model 602001 instrument under constant pressure (3.0 atm) with all samples matched against identical buffer in the reference cell. Samples were equilibrated for 600 s, followed by melting data collection through heating cycles from 10 to 90 °C at a scan rate of 0.75 °C/min. The reversibility was confirmed for all constructs by performing a cooling scan from 90 to 10 °C at a scan rate of 0.5 °C/min and a subsequent heating cycle. A minimum of 10 replicate experiments were conducted for each T4L construct. Melting data were analyzed, and thermodynamic parameters, including the specific heat capacities (ΔC), were determined using NanoAnalyze Data Analysis, version 3.6.0, from TA Instruments. The melting temperatures (TM) and enthalpies (ΔHM) were extracted using the TwoStateScaled model for fitting the experimental data. The ΔHfit/ΔHcal ratios were all in the range of 0.97–1.01; the Aw values were in the range of 0.99–1.05, and the standard deviation of the fits was <1.6 for all experiments.

Quantum Mechanical (QM) Calculations

QM energies and electrostatic potential maps (ESPs) were calculated using Gaussian 09 revision E.01[48] with the Møller–Plesset second order (MP2) in a cyclohexane solvent (D = 2 relative to a vacuum). This low-dielectric solvent model is appropriate for calculations on systems that involve explicit solvent and short distances between interacting atoms, as is the case in this study, and reflects the low dielectric expected for a protein interior.[49] Basis set superposition errors (BSSEs)[50] were determined from a separate counterpoise gas phase calculation and directly summed into the calculated solvent phase energy. Polarizable basis sets, including dispersion, were applied (aug-cc-PVTZ[51] for Y18 and Y18 and extended to Y18 with aug-cc-PVTZ-PP from EMSL basis set exchange). The strategy for QM calculations applied here had previously been validated against experimental XB geometries and energies in model DNA junction systems.[52−54] The atomic coordinates of the interacting residues (Y18 and G28) were taken from the refined structures of each construct. Residue 18 was reduced to 2-halophenol, and residue 28 was reduced to N-methylacetamide (NMA) to decrease computational time. Hydrogen atom positions were geometry-optimized with a semiempirical AM1 calculation. The torsional angle, δ, of the hydroxyl hydrogen is manually rotated to determine its contribution to the HeX-B.

Turbidity Assay

The activities of the T4L constructs were monitored through a standard cell clearing assay.[55,56]Microccocus lysodeikitcus bacteria were grown in 50 mL of 2× YT medium overnight at 37 °C while being shaken at 250 rpm. Then, the culture was centrifuged in an Eppendorf 5810 R centrifuge at 4000 rpm at 4 °C for 15 min. The supernatant was decanted, and the cell pellet diluted in a 1:1 mixture of 50 mM monobasic and 50 mM dibasic sodium phosphate solutions until an OD450 of 1.0 was reached. Bacterial samples of 1.0 mL were prepared and stored at −80 °C. After the samples had thawed, the purified and concentrated T4L construct in crystallization buffer was added to the bacterial sample to reach a final concentration of 0.1 mg/mL. The absorbance change over time was measured at room temperature (23 °C) and 40 °C. Three or four replicates of each construct were run for each temperature.

Results

Structures of Y18-T4L Constructs

The crystal structures of the Y18-T4Ls (Y18, Y18, and Y18), determined from 1.35 Å to 1.65 Å resolution (Table S1 and Figure S1), are all isomorphous with the parent WT* (Figure ), allowing any observed structural perturbations to be analyzed relative to each other, and relative to their effects on the constructs’ stabilities and activities. The Y18 side chain of each construct sits in a pocket formed by a pair of antiparallel loops and is rotated to place the halogen inside this pocket (i-rotamer) pointed toward G28 or outside the pocket (o-rotamer) exposed to the solvent (Figure ). The i-rotamer propensity increases as the halogen becomes smaller (Table ). The structure of the loop is invariant among all constructs, with the O···H−N HB distance (from G28 to R14) varying by <0.1 Å relative to WT* (Tables S2–S5). This loop region in the crystal structure is, therefore, very rigid, sterically constraining the size of the halogen that is accommodated and consequently its potential to participate in an XB.

Figure 3

Crystal structures of Y18-T4L constructs. (A) Overall structure of the chlorinated Y18-T4L (magenta backbone trace and carbon atoms) overlaid on WT*-T4L (blue trace and carbon atoms). The Y18 side chain sits with Cl (green) either inside the loop toward the G28 oxygen (i-rotamer) or outside this loop (o-rotamer). The o-rotamer overlays nearly exactly with WT*, while the Y18 side chain of the i-rotamer is slightly shifted away from G28. The 3.11 Å Cl···O distance from Y18 to G28 is 95% of the sum of the van der Waals radii (∑RvdW), while the angle of approach (O···Cl–C) is a shallow 150°. (B) Overall structure of brominated Y18-T4L (cyan) overlaid on WT*-T4L (blue). As with the chlorinated construct, Br (brown) is in either the i- or o-rotamer. The o-rotamer, again, overlays nearly exactly with WT*, while the Y18 side chain of the i-rotamer is more shifted significantly away from G28. The 2.88 Å Br···O distance in Y18 is 85% of the ∑RvdW, and the angle of approach (O···Br–C) is also shallow at 151°. (C) Overall structure of iodinated Y18-T4L (orange) overlaid on WT*-T4L (blue). Unlike the Cl and Br constructs, this construct places the iodine (purple atom) only in the o-rotamer position and, therefore, does not make a short interaction with G28.

Table 1

Crystallographic Analysis of Ratios of i-Rotamer to o-Rotamer, XB Geometries (where observed), and Solvent Accessible Surfaces (SASs) in the Y18-T4L Constructs (X = Cl, Br, or I)

construct	i-rotamer:o-rotamer	RX···O (%∑RvdW)a	θ1b	SAS (% max)c
mClY18-T4L	54:46	3.11 Å (95%)	150°	17.3 Å2 (13.8%)
mBrY18-T4L	22:78	2.88 Å (85%)	151°	28.9 Å2 (21.8%)
mIY18-T4L	0:100	not applicable	not applicable	17.1 Å2 (11.9%)

%∑RvdW is the percent of the sum of the standard van der Waals radii of the halogen to oxygen (X···O) interacting pair.

θ1 is the angle of approach of the oxygen acceptor to the halogen, ∠(C–X···O).

Solvent accessible surfaces for halogen atoms were calculated using PyMol.[87] The percent relative to maximum exposure (% max) was calculated relative to the exposure of each halogen in an isolated Y amino acid residue (SAS for Cl of Y = 124.8 Å2, SAS for Br of Y = 132.7 Å2, and SAS for I of Y = 143.6 Å2).

The halogens of the i-rotamers of Y18-T4L and Y18-T4L are seen to form short-range interactions with the carbonyl oxygen of G28 (Figure A,B). The Cl···O distance in Y18-T4L is ∼95% of the sum of the standard van der Waals radii (∑RvdW) of the interacting atoms, near the optimum distance for biological XBs,[20] while the Br···O distance in Y18-T4L is much shorter at ∼85%. The angles of approach of the oxygen acceptor to the halogen (θ1 = 150° for O···Cl–C, and θ1 = 151° for O···Br–C) are shallow relative to the ideal linear approach (θ1 = 180°); however, these geometries are well within the range of XB interactions observed in biological systems[20] and, as will be discussed later, are accommodated by additional polarization of the halogens in this particular system. In addition, the approach angles of halogen to the acceptor HB (X···O···N) are 80.6° for X = Cl and 82.2° for X = Br, which are consistent with the XBs being an orthogonal interaction (geometrically perpendicular and energetically independent) to the HB.[34] Thus, we are confident in classifying these interactions as XBs.

The small displacement of the Y18-T4L aromatic side chain in the i-rotamer from the o-rotamer position is likely an attempt to pull the halogen into a more linear XB geometry. The larger displacement of the Y18-T4L side chain away from G28, however, suggests destabilizing steric effects in the i-rotamer even as the halogen forms a short XB interaction. This balance between an XB attractive and steric repulsive force (and potentially bonding forces from distortion of the side chain) would account for the lower i-rotamer:o-rotamer ratio of the bromo construct. The fact that the larger iodine of Y18-T4L is entirely in the o-rotamer supports this interpretation. Given that none of the constructs are entirely in the i-rotamer position, the question is whether the XB interactions are actually stabilizing. We will address this question later by comparing the melting temperature (TM) and enthalpy (ΔHM) in solution of each Y18-T4L construct to those of the parent WT* enzyme.

Solvent Structure

The constellation of waters around E11, Y18, and G28 seen in the WT* structure (Figure B) remains mostly intact in the halogenated constructs, except to accommodate the halogens in their i- or o-rotamers (Figure , Figures S2 and S3, and Tables S2–S5). For Y18-T4L, the waters that bridge Y18 to E11 (W2–W6) are seen in positions similar to those in WT*, with the exceptions of W1 and W3. In the chlorinated construct, the position of W1, which is particularly important in stabilizing the T4L protein,[32] is filled by two partially occupied water molecules, each very close (within 1.8 Å) to the other. In addition, one of these waters sits unusually close to the OH of the Y18 side chain of the o-rotamer. We therefore interpreted this water as being a single molecule occupying two mutually exclusive positions: one assigned to the o-rotamer (W1-o, sitting in nearly the same position as W1 in WT*) and the other to the i-rotamer (W1-i, repositioned to sit in the aromatic plane) of the Y18 residue. Although not as important as W1 in terms of defining protein stability, W3 also shows two partially occupied positions, one that forms an HB to the Cl of the o-rotamer (assigned as W3-o) and one that does not (assigned as W3-i).

Figure 4

Stereoimage of the chlorinated Y18-T4L construct with its waters of interaction. The top panels show the rotamer with the chlorine (emerald green) sitting inside [Y18-T4L(i), carbon atoms colored yellow] the loop. Waters are labeled W1–W6, with those that are in positions nearly identical to those of WT* colored and labeled in blue and those in positions unique to the i-rotamer colored and labeled in yellow (along with the carbons of the Y18 side chain). The bottom panels show the rotamer with the bromine sitting outside [Y18-T4L(o), carbons colored cyan] the loop. Waters are labeled W1–W6, with those that are in positions nearly identical to those of WT* colored and labeled in blue and those in positions unique to the o-rotamer colored and labeled in cyan.

The waters around Y18-T4L (Figure S2 and Table S4) show patterns similar to those in the Y18-T4L structure, with certain solvent positions (including W1) occupied by molecules that are associated with the i-rotamer or the o-rotamer. The solvents in Y18-T4L (Figure S3 and Table S5) are similar to WT*, except that W1 is entirely missing, a consequence of the Y18 side chain being pushed closer to the carbonyl oxygen of G28, which has either completely displaced this solvent molecule or made it less specific in its positions (thereby making it unobservable in the electron density map).

Thermal Stabilities of Y18-T4L Constructs

Differential scanning calorimetry (DSC) was used to determine how the conformational features seen in each of the crystal structures affect the stability of the protein. This protein system allows precise determination of melting temperatures and enthalpies and, thus, allows us to accurately assign thermodynamic values to molecular interactions associated with specific structural modifications.[57]

The DSC-measured TM of Y18-T4L (Table ), with the iodine entirely in the exposed o-rotamer position, is ∼0.5 °C lower than that of WT*, which is consistent with our previous observation[32] that a protein is destabilized when a hydrophobic methyl or halogen substituent[58] is added to a solvent-exposed position and with previous studies on the effects of hydrophobic side chains on T4L stability.[59] This hydrophobic effect is reflected in the increased ΔC value. The Y18-T4L construct has TM and ΔHM values that are very similar to that of WT*, indicating that the stabilizing XB in the i-rotamer nearly exactly counterbalances the destabilizing effects of steric repulsion of this buried placement and the exposure in the o-rotamer. The most interesting case is that of the Y18-T4L construct, which shows an ∼1 °C increase in TM and a 2.7 kcal/mol increase in ΔHM versus those of WT*. Together, the results show that the increased stability of the protein, as measured by the TM and ΔHM, is dependent on the ability of the halogen to form an XB interaction in the i-rotamer (Figure ). Thus, for the first time, a more thermally stable protein was engineered by introducing a halogenated, in this case chlorinated, unnatural amino acid.

Table 2

DSC-Measured Melting Temperatures (TM), Enthalpies (ΔHM), and Heat Capacities of the WT* and Y18-T4L Constructs (X = Cl, Br, or I)

construct	TM (°C)	ΔHM (kcal/mol)	ΔSM (cal mol–1 K–1)a	ΔCp (kcal mol–1 K–1)
WT*	57.30 ± 0.01	120.2 ± 0.5	363.8 ± 1.5	2.6 ± 0.2
mClY18-T4L	58.28 ± 0.01	122.9 ± 0.4	370.7 ± 1.2	2.9 ± 0.3
mBrY18-T4L	57.36 ± 0.02	119.2 ± 0.4	360.6 ± 1.1	3.3 ± 0.2
mIY18-T4L	56.78 ± 0.01	115.5 ± 0.6	350.1 ± 1.9	2.8 ± 0.1

ΔSM is the melting entropy calculated as ΔHM/TM.[88]

Figure 5

Differences in melting temperatures [ΔTM (■)] and in melting enthalpies [ΔΔHM (○)] for Y18-T4L (X = Cl, Br, or I) vs WT* constructs of T4 lysozyme. Standard deviations of the measured values are shown as error bars.

The entropy of melting (ΔSM) can be calculated from the experimental ΔHM and TM values for each construct (Table ). The resulting ΔSM for Y18-T4L is ∼7 cal mol–1 K–1 higher than that of WT*, suggesting that the XB makes the protein more conformationally rigid. The alternative interpretation would be that ΔSM is defined by changes in the solvent structure, particularly because the halogens of the Y residues are hydrophobic.[58] The expectation is that if the halogen is already exposed to solvent, as is the case for the o-rotamer of Y18-T4L and Y18-T4L constructs, the entropic change upon melting would be smaller than if the halogen were more buried, as in the i-rotamer of Y18-T4L. To determine whether solvent effects are the primary determinant of ΔSM, we calculated the solvent accessible surfaces [SASs (Table )] of the halogens in the i-rotamer (when present) and o-rotamer conformations. The halogens in the i-rotamer of Y18-T4L and Y18-T4L are fully buried, as reflected in SASs of 0 Å2. The o-rotamers of all the halogenated constructs showed some degree of exposure to solvent, with the Br of Y18-T4L being most exposed and the I of Y18-T4L being the least, particularly in terms of the percentage relative to the exposure of an isolated halogenated tyrosine. Unlike our earlier study,[32] the effect of each halogen in an exposed versus buried site is internally controlled here. This observation is contrary to what was expected, but careful analyses of the structures show that the side chain of Arg14 is pulled within HB distance (3.1 Å) of the I, thus burying a significant portion of the halogen surface that is otherwise solvent-exposed in the other constructs. A comparison of the SAS and associated solvent free energies shows that they are not correlated to the ΔSM values. The ΔSMs, however, are well correlated with the occupancies of the i-rotamers (R = 99.3%), indicating that the interactions of the halogens with the loop are the primary determinants of the entropic effects on the protein structure. The DSC melting energies converted to ΔG° of stability at 40 °C follow exactly the trend for the TMs, showing that the XB renders Y18-T4L overall more stable than WT* at high temperatures.

Extrapolation of the thermodynamic DSC thermodynamic parameters to the standard temperature (25 °C) indicates that the overall stabilities, as reflected in ΔG°, for all of the halogenated constructs are lower than that of WT* at this lower temperature (Table ). The resulting standard energies follow the general trend for the melting parameters previously reported for T4L,[33,60] except for those of Y18-T4L (ΔH° and ΔS°), which are calculated as being significantly lower than those of the other constructs. This singular outlier can be ascribed to the anomalously high ΔC of Y18-T4L, which affects extrapolation of its DSC energies to room temperature (RT). This higher DSC-measured ΔC is indicative of a more hydrophobic system; ΔC values have been shown to be well correlated with SASs.[33,61−65] The experimental ΔC values listed in Table are indeed well correlated with the SAS values in Table , as the percent of the maximum exposure of the hydrophobic surface at Y18 (Figure S4), indicating that they reflect features of the crystal structures. Our interpretation is that the high ΔC of Y18-T4L may not be applicable to a RT calculation, because of temperature effects on the i-rotamer:o-rotamer ratio of the Y18 side chain. Our hypothesis is that near the TM, relaxation of the protein permits the Br-substituted Tyr side chain to be better accommodated in the pocket and to form an XB, pushing a larger proportion to convert from the o- to i-rotamer. This allows more exposure of nonpolar surfaces upon melting, resulting in the observed ΔC[65,66] near the TM for Y18-T4L being higher than we would expect at RT. Indeed, extrapolation of the Y18-T4L DSC data to RT using any of the other ΔC values in Table results in energies that are comparable to those of the other T4L constructs in this study. Similar effects on ΔC would not be expected for either the Cl or I construct, because the side chain of Y18-T4L is not seen to form an XB, while the Cl of Y18-T4L forms a stable XB. The XB favoring the buried i-rotamer of Y18-T4L, however, is constricted by the steric repulsion that favors the exposed o-rotamer. In this way, the brominated construct exposes more hydrophobic surface upon thermal unfolding, which led to the anomalously high DSC-measured ΔC value and the unexpectedly low ΔH° and ΔS° values from extrapolation to RT.

Table 3

Standard Enthalpies (ΔH°), Entropies (ΔS°), and Gibbs Free Energies (ΔG°) of Stabilization Extrapolated to 25 and 40 °C from DSC Melting Energies (Table ) of the WT* and Y18-T4L Constructs (X = Cl, Br, or I)

	25 °C			40 °C
construct	ΔH° (kcal/mol)	ΔS° (cal mol–1 K–1)	ΔG° (kcal/mol)	ΔH° (kcal/mol)	ΔS° (cal mol–1 K–1)	ΔG° (kcal/mol)
WT*	–34.8	–91.7	–7.43	–74.5	–222	–5.07
mClY18-T4L	–27.6	–67.6	–7.38	–70.5	–208	–5.31
mBrY18-T4L	–12.5	–20.8	–6.27	–61.9	–183	–4.73
mIY18-T4L	–26.9	–67.8	–6.71	–68.7	–205	–4.66

Enzymatic Activities of Y18-T4L Constructs

The thermal stabilities of T4L and its various mutants are correlated with the level of enzymatic activity.[57] Thus, we applied the standard bacterial clearing assay to monitor the effects of halogenation on the activity and, in addition, provide additional support for their observed effects on the thermal stability of the enzyme. At RT (23 °C), the activities of the Y18-T4L constructs were all lower than that of WT* (Figure ) and, with the exception of that of Y18-T4L (Figure S5), are consistent with the ΔG° values calculated from the DSC melting energies (Table ). At an elevated temperature (40 °C), the activity of the iodinated construct was not significantly changed but that of the brominated and chlorinated enzymes increased relative to that of WT*, with Y18-T4L becoming 15% greater than the native enzyme. The temperature at which Y18-T4L would become more stable than WT* is predicted from extrapolation of the DSC values to be ∼35 °C, which is also where we would expect the chlorinated enzyme to become more active. These general trends in activity at low and high temperatures follow and, therefore, confirm the DSC melting results (with the singular exception of Y18-T4L extrapolated to 25 °C, as discussed above) and serve to bridge the melting properties measured at high temperatures and the structural features of crystals grown at low temperatures. Thus, we have shown for the first time that an XB can be specifically engineered not only to increase the thermal stability of a protein but also to increase its activity at elevated temperatures.

Figure 6

Enzymatic activities of Y18-T4L constructs, as a percent of WT* activity. The enzymatic activity of each halogenated construct (Cl in green diamonds, Br in red squares, and I in purple circles) is compared to that of WT* (defined as 100% and indicated by the dashed line) at 23 and 40 °C.

Structure–Stability/Activity Relationships: The Hydrogen-Bond-Enhanced Halogen Bond

The thermal stabilities of each construct, as reflected in the DSC-measured melting temperatures (TM) and enthalpies (ΔHM) (Table ), are correlated with the percent i-rotamer (Figure ). The interactions (attributed here to XBs) within the loop convey stability to the Y18-T4L constructs, while exposure of the halogen to a solvent (similar to that previously seen with halogenated and methylated T4L analogues[32]) destabilizes the protein. These DSC values, however, may underestimate the stabilizing potential of the XB, particularly in the Y18-T4L construct, which places only 54% of the Cl in the i-rotamer position. If the Cl of Y18-T4L were entirely in the XB position, the ΔHM would be predicted to be 5.4 kcal/mol higher than that of WT*. In addition, the TMs are well correlated with their ΔHMs (a 1 kcal/mol increase in ΔHM results in an ∼0.2° increase in TM).

Even at ∼3 kcal/mol, the increase in ΔHM measured for Y18-T4L is significantly larger than that previously determined for Cl (0.5 kcal/mol) in a model DNA junction but comparable to those of Br and I (1.6–4.6 kcal/mol).[24,25,53] We attribute this remarkably stronger Cl effect to an XB that is enhanced by an intramolecular HB from the hydroxyl substituent to the negative annulus of the halogen. The electrostatic potential (ESP) surface (Figure ) calculated for 2-halophenol (a model for the Y18 side chain) shows that the σ-holes of halogen substituents become enhanced as the OH rotates from an angle δ of 180° (trans-OH) to an angle δ of 0° (cis-OH, pointing toward and within H-bonding distance of the halogen). This enhancement is interpreted as resulting from polarization of the electron density by the HB toward the p orbitals and away from the σ-hole, which renders the ESP for Cl comparable to that of the Br in bromobenzene.[18,23]

Figure 7

Effect of OH rotations on the electrostatic potential (ESP) of 2-halophenol. The QM-calculated ESPs are mapped onto the atomic surfaces of 2-halophenol, where the halogen is Cl, Br, or I. The map shows ESP from 10 to −10 kcal/mol of interaction energy to a positive point charge, reflecting a surface charge that ranges from positive (blue) to negative (red) on the halogen surface. The ESP is calculated as the OH is rotated from an angle δ of 180° (non-HB trans-OH orientation) to an angle δ of 0° (HB cis-OH orientation), in 45° increments. The neutral-charge angle (where the ESP ≈ 0) extends from an average of ∼163° (Cl, Br, and I) at an angle δ of 180° to ∼135° at an angle δ of 0° (compared to an average of 157° for halobenzenes).

The effect of the enhanced σ-hole on the XB donor ability of the Cl can be appreciated by comparing the QM-calculated XB energies (EMP2) for complexes of N-methylacetamide (NMA) with either chlorobenzene or 2-chlorophenol (a model for the XB complex between G28 and Y18), positioned according to the Y18-T4L crystal structure (Figure ). For chlorobenzene, the Cl···O XB energy is only slightly favorable (EMP2 = −0.3 kcal/mol), as expected for the inherently weak XB potential of Cl. The Cl is an even weaker XB donor (EX-MP2 = 0.06 kcal/mol) in the 2-chlorophenol complex with a trans-OH, reflecting the electron-donating property of the OH substituent, as suggested by calculations on Y-substituted insulin.[67] The cis-OH, however, forms an HB with the Cl, resulting in an enhanced XB (EMP2 = −1.4 kcal/mol relative to the trans-OH). This EMP2 is nearly identical to that of an iodine XB (−1.8 kcal/mol) that was previously shown to rescue protein stability.[32] Thus, the HB intensifies the σ-hole and extends the allowed angles of approach by the acceptor to the halogen, both of which enhance the XB potential of Cl, resulting in the HeX-B interaction. As the OH rotates from the trans-direction to the HeX-B cis-direction (Figure ), the angle at which the calculated ESP switches from being a positive σ-hole to a negative annulus (a neutral-charge angle) is increased by 22–28° for the halogen substituent (from ∼160° to < 135° for bromophenol, for example), allowing the relatively shallow 150° approach of the oxygen acceptor seen in the crystal structures to be stabilizing Cl and Br XB interactions. The HB to the halogen itself contributes significantly (−1.8 kcal/mol) to the interaction and, together with the HeX-B, accounts for the ∼3 kcal/mol enhancement of the DSC ΔHM for Y18-T4L versus that of WT*.

Figure 8

Quantum mechanics energies calculated at the MP2 level (EMP2) for complexes of N-methylacetamide (NMA) with chlorobenzene (left) or 2-chlorophenol (right). Atomic positions of the components were defined by the coordinates of the crystal structure of the m-chlorotyrosyl side chain at position 18 and the peptide bond at G28 in the chlorinated T4 lysozyme construct (Y18-T4L). The intrinsic halogen bonding potential of a Cl substituent (reflected in the EMP2 of the NMA···chlorobenzene complex) is reduced by 0.4 kJ/mol with the introduction of the o-hydroxyl substituent, an electron-donating group, when this OH is rotated with the hydrogen away from the Cl (δ = 180°). Rotation of the OH so that it can form a HB to the Cl enhances the EMP2 of the NMA···2-chlorophenol complex (δ = 0°). This more favorable EMP2 can be parsed out to the energy of the HB itself but also to the enhanced energy of the XB.

Why are these HeX-Bs not seen in all of the Y18-T4L constructs? For iodine, the answer is simply that this large halogen does not fit into the pocket of the rigid loop. The i-rotamer:o-rotamer ratio of the Br in Y18-T4L reflects an ∼0.8 kcal/mol difference between a buried and exposed halogen. Although the Br···O geometry suggests a relatively strong XB, the large displacement of the side chain of the i-rotamer would indicate that this very short distance interaction is largely offset by an unfavorable steric clash. This balance between the opposing forces would account for the apparent discrepancy between the ΔG° and activity at RT (Figure S5).

The answer to why only ∼50% of Y18-T4L forms the HeX-B interaction comes again from considering the OH group, which has no sense for whether an XB is present. The Y18 hydroxyl is bridged by HBs to the carbonyl oxygen of E11 through a water [W1-i (Figures and 9)],[32] which can form an HB to either the Cl (of the i-rotamer) or W1-o (the o-rotamer). The -OH···O=C distance between the Y18 hydroxyl and E11 carbonyl varies by <0.2 Å in Y18-T4L (3.92 and 4.10 Å for the i- and o-rotamer, respectively, compared to 4.07 Å for WT*); consequently, the direct effects of the Cl on this interaction are expected to be minimal. If the OH of Y18 is a donor to W1, it cannot simultaneously form an HB to the halogen and, thus, cannot enhance the XB capability; the approximate 1:1 i-rotamer:o-rotamer ratio suggests no preference for either Cl or W1. The higher energy of interaction of W1 with the Y18 hydroxyl is a result of it being closer when this HB donor is not oriented toward the halogen. The QM energy calculated for a ternary complex of 2-chlorophenol, NMA, and W1 in their crystal structure conformations showed a <0.1 kcal/mol difference in EMP2 between the i- and o-rotamers, which would account for the near equal distribution among the rotamers (Figure ). Furthermore, Abraham et al.[68] showed that 2-halophenols adopt a trans-OH in water but a cis-OH with a weak intramolecular HB between the hydroxyl and halogen in organic solvents. The pocket where Y18 sits is partially solvent exposed, resulting in the hydroxyl having only a slight preference as an HB donor to the Cl over W1.

Figure 9

Quantum mechanics energies (EMP2) calculated for the ternary complex of the i-rotamer (blue boxes) and o-rotamer (red boxes) forms of the Y18-T4L construct. The energies calculated for the ternary complexes can be dissected out to the major contributing interactions: the HeX-B, NMA···W1, and W1···chlorophenol interactions for the i-rotamer (top left) and the NMA···W1 and W1···chlorophenol interactions for the o-rotamer (top right). Atomic coordinates for each model complex were taken from the crystal structure of the Y18-T4L construct, with the water (W1) for each form assigned accordingly to the appropriate rotamer form (see Figure ). Each W1 species was assigned to its appropriate rotamer based on their HB distances and potential clash with the Cl atom (W1-o would have clashed with the Cl in the i-rotamer form, but that assigned as W1-i did not show a Cl collision). Positions of hydrogens for the hydroxyl group of the chlorophenol model and the interstitial waters were optimized to yield the lowest-energy structures for each complex. The differences in their respective EMP2 values predict an [i-rotamer]:[o-rotamer] ratio of 49:51, which is nearly identical to the 54:46 ratio determined from the crystal structure.

Discussion

In this study, we address the question of whether addition of an XB to augment a critical HB, by introducing an unnatural amino acid into the structure, would result in a more stable protein. The chlorinated Y18-T4L construct demonstrates the potential application of XBs in increasing the stability and associated activity of an enzyme at elevated temperatures. We had previously shown that halogenation of proteins generally has the effect of destabilizing protein structure, if the halogen is exposed to solvent and, therefore, incapable of forming an XB.[32] This effect is recapitulated here, where the stabilities of the Y18-T4Ls are dependent on burying each halogen in a protein pocket (as reflected in the o- vs i-rotamers). The earlier study also showed that this destabilizing effect can be partially rescued if the halogen, particularly iodine with its very large σ-hole, can form an XB.[32] Again, we see this same effect with the current study; however, in this case, it is the chlorinated construct that forms the more stabilizing interaction, and the interaction is sufficiently strong not only to rescue the stability but also to increase it above that of WT*.

Although the ∼3 kcal/mol stabilization of T4L may seem to be small, it should be noted that proteins are stabilized by the concerted contributions of multiple low-energy, noncovalent interactions. Indeed, introducing a 3 kcal/mol interaction would convert metastable peptides and proteins to be fully stable structures, which would affect their functions.[69] Similarly, adding a 3 kcal/mol interaction to a ligand to its interaction with a protein target would reduce its dissociation constant by >2 orders of magnitude, which in turn could make such a ligand more attractive as a potential drug candidate.[30]

The stabilizing potential of an XB with a Cl donor had previously been determined to be very small (∼0.5 kcal/mol) in a DNA system.[25] A Cl-XB had also been estimated from calculations to contribute as much as 1.5 kcal/mol to stabilizing the β-hairpin conformation of a cyclic peptide,[70] one of the first demonstrations that an XB can potentially stabilize a protein-type conformation. Finally, it had been shown that addition of halogens that fill only void spaces contributes <0.8 kcal/mol per halogen atom.[35] Thus, the ∼3 kcal/mol stabilization seen here with the addition of a single chlorine atom is surprising, leading to the question of why the Cl-XB has such as strong effect even compared to the previous I-XB in this same T4L protein system.[32]

We attribute the improved ability of the Cl to serve as an XB donor to an HB-enhanced XB. The HeX-B represents a new and potentially powerful variation on the stand-alone XB, expanding the standard menu of noncovalent interactions that dictate molecular folding.[71−73] Because XBs and HBs share a common set of acceptors, their relationships can be complex.[74] The interaction described here, however, differs significantly from the orthogonal HB/XB interaction described previously in that the HB of the HeX-B is to the XB donor (as opposed to the acceptor) and enhances, rather than being energetically independent of, the stabilizing potential of the interaction. Upon addition of an HB donor (including OH, SH, or NH2) next to a halogen, it is now possible to enhance its XB potential through a synergistic relationship, beyond tuning through standard inductive effects. This enhancement is expected to be even more dramatic (∼3–4 kcal/mol) for anionic oxygen XB acceptors compared to the neutral carbonyl acceptor in our study (Tables and 5), extending the range of stabilization potentials to as much as 6.7 kcal/mol for an iodine HeX-B (compared to the 6 kcal/mol for very strong HBs in proteins[75]) when placed in an unconstrained biomolecular environment.

Table 4

Quantum Mechanical Energies (kilocalories per mole) for the Neutral Oxygen Halogen-Bond (XB) Acceptor of N-Methylacetamide Interacting with a Halobenzene or 2-Halophenol XB Donora

XB Donor: Halobenzene
Cl	Br	I
–0.75	–1.54	–2.61

The halogen to oxygen distances (rX···O) for the halobenzene and 2-halophenol XB donors were set to 92% of the sum of the respective van der Waals radii[89] of the halogen and the XB acceptor (rCl···O = 3.01 Å, rBr···O = 3.10 Å, and rI···O = 3.22 Å). The angle of approach of the acceptor oxygen to the X–C bond (θ1 angle) was set at the optimum value of 180° for all calculations. For the 2-halophenol donor, the energies were calculated with the OH substituent rotated to align the hydrogen toward the halogen (δ = 0°), away from the halogen (δ = 180°), or perpendicular to the X···O bond (δ = 90°). The differences in energies between δ values of 0° and 180° are fairly independent of rX···O.

Table 5

Quantum Mechanical Energies (kilocalories per mole) for the Anionic Oxygen Halogen-Bond (XB) Acceptor of Hypophosphite Interacting with a Halobenzene or 2-Halophenol XB Donora

XB Donor: Halobenzene
Cl	Br	I
+0.03	–1.66	–4.29

The halogen to oxygen distances (rX···O) for the halobenzene and 2-halophenol donors were set to those seen in the crystal structures of the Cl2J (rX···O = 2.88 Å), Br2J (rX···O = 2.87 Å), and I2J (rX···O = 3.01 Å) DNA constructs.[25] The angle of approach of the acceptor oxygen to the X–C bond (θ1 angle) was set at the optimum value of 180° for all calculations. For the 2-halophenol donor, the energies were calculated with the OH substituent rotated to align the hydrogen toward the halogen (δ = 0°), away from the halogen (δ = 180°), or perpendicular to the X···O bond (δ = 90°).

More than 700 T4L variants have been characterized to date, very few of which result in more stable proteins, leading to the conclusion that it is very difficult to rationally engineer an enzyme with increased thermal stability and/or activity.[76] It has previously been shown that unnatural amino acids can be applied to engineer more stable proteins, primarily by introducing hydrophobic interactions[77] or unusual disulfide linkages,[78,79] but introducing new electrostatic interactions, including HBs, to stabilize proteins is less common.[80−86] The HeX-B from the unnatural Tyr is distinct from the multitude of interactions engineered into T4L in that both stability and function can be enhanced noncovalently through addition of a single halogen atom.

There are many applications in which halogens can be placed adjacent to HB donor substituents, such that an HeX-B can be formed. In this particular study, the HB donor is an OH group, which can rotate to either interact with the halogen or not. An NH2 in this position, however, would always result in an HB enhancing the XB donor capabilities of the halogen. For small molecules used in the design of protein inhibitors, or for crystal engineering or supramolecular assemblies, the types of HB donors that can facilitate formation of the HeX-B grow beyond the simple OH, NH2, and SH substituents, including potentially cationic HB donors (which would be expected to have an even stronger polarization effect). Thus, with the HeX-B, we have now expanded the energetic and geometric properties of XBs, making this noncanonical interaction an even more powerful tool for molecular design in biomolecular engineering, medicinal chemistry, and material science.