ABSTRACT Title of Document: PROBING ALLOSTERIC COMMUNICATION BETWEEN DISORDERED SURFACES IN A PROTEIN William J. Cressman Master of Science, 2015 Directed By: Dr. Dorothy Beckett Department of Chemistry and Biochemistry Molecular mechanisms of protein allostery are not well understood, particularly in those systems that undergo disorder to order transitions upon activation. BirA is a model allosteric protein in which corepressor, bio-5’-AMP, binding enhances dimerization by -4 kcal/mol and is coupled to disordered loop folding on both functional surfaces. BirA variants with single alanine substitutions in dimerization surface residues are investigated to further characterize communication between the two sites. Isothermal titration calorimetry measurements of corepressor binding of these BirA variants indicate only the G142A substitution perturbs the free energy of binding. The G142A crystal structure indicates a mechanism of communication from the corepressor binding to the dimerization surface involving α- helical extension of residues 143-146. Measurements of heat capacity changes associated with corepressor binding to the BirA variants support a model in which the helical extension enhances dimerization by enabling the formation of a network of intramolecular interactions on the dimerization surface. PROBING ALLOSTERIC COMMUNICATION BETWEEN DISORDERED SURFACES IN A PROTEIN By William J. Cressman Thesis submitted to the Faculty of the Graduate School of the University of Maryland, College Park, in partial fulfillment of the requirements for the degree of Master of Biochemistry 2015 Advisory Committee: Professor Dorothy Beckett, Chair Professor Douglas Julin Professor Nicole LaRonde © Copyright by William J. Cressman 2015 ii Dedication For my father, Robert Cressman Jr., whose love of science could only be surpassed by his passion for helping others. iii Acknowledgements Foremost, my advisor, Dr. Dorothy Beckett, whose efforts and guidance have made this thesis possible and made me the scientist I am today. My wife, Stephanie, for her unwavering love and support. My coworkers and friends, Chris Eginton and Jingheng Wang, who were a constant source of encouragement, suggestions and ideas. My family and friends, for their unending support and encouragement. iv Table of Contents Dedication ..................................................................................................................... ii Acknowledgements ...................................................................................................... iii Table of Contents ......................................................................................................... iv List of Tables ............................................................................................................... vi List of Figures ............................................................................................................. vii Chapter 1 : Introduction ................................................................................................ 1 Allostery in Protein Regulation ................................................................................ 1 BirA as a model allosteric system ............................................................................. 2 Disordered loops play a role in the BirA allosteric mechanism ............................... 4 Chapter 2 : Materials and Methods ............................................................................... 9 Chemicals and biochemicals ..................................................................................... 9 Isothermal titration calorimetry (ITC) measurements ............................................ 11 Direct ITC binding measurements of biotin binding parameters ........................... 13 Displacement ITC binding measurement of bio-5’-AMP binding parameters ....... 14 Total association at partial saturation (TAPS) calorimetry .................................... 15 Molar heat capacity changes of bio-5’-AMP binding ............................................ 16 Chapter 3 : Results ...................................................................................................... 17 Selection of alanine-substituted dimerization surface variants ............................... 17 Dimerization loop variants bind biotin with affinities similar to wild type BirA ... 19 Alanine substitution at residue G142 greatly perturbs bio-5’-AMP binding. ......... 22 ΔC°p,bio-5’-AMP is more negative for BirA dimerization surface variants ................. 26 Chapter 4 : Discussion ................................................................................................ 31 BirA integrity is maintained upon alanine substitution .......................................... 31 Residue G142 displays evidence for communication to the ligand binding site .... 32 Correction of enthalpy of bio-5’-AMP binding for holoBirA dimerization ........... 33 ∆C°p,bio-5’-AMP values directly correlate to BirA dimerization ................................ 38 Formation of intramolecular interactions is observed upon allosteric activation ... 44 v Chapter 5 : Conclusions and Future Directions .......................................................... 46 Chapter 6 :Appendicies ............................................................................................... 47 Appendix 1: Discrepancy between ∆HTAPS and ∆Hdisplacement of bio-5’-AMP binding .................................................................................................................... 47 Possible sources of experimental discrepancy .................................................... 47 ∆G°bio-5’-AMP measured by biotin displacement agrees with those measured by direct titrations. ................................................................................................... 51 ∆H°biotin measured directly agrees with that measured by TAPS titration. ......... 53 Appendix 2:Considerations in the Design of ITC experiments .............................. 56 Wiseman’s c-value for the design of ITC titrations ............................................ 56 Ligand concentration and injection scheme ........................................................ 59 Appendix 3: Expression and purification of BirA variants .................................... 62 Preparation and regeneration of FPLC chromatography resins .......................... 67 Bibliography ............................................................................................................... 70 vi List of Tables Table 3.1: Homodimerization energetics of chosen BirA variants: ............................ 18 Table 3.2: Thermodynamic parameters of biotin binding for BirA variants .............. 21 Table 3.3: Thermodynamic parameters of bio-5’-AMP binding. ............................... 24 Table 3.4: Heat capacity change of bio-5’-AMP binding for BirA dimerization surface variants. .......................................................................................................... 30 Table 4.1: Fraction of BirA dimer present in ITC titrations ....................................... 37 Table 4.2: Corrected thermodynamic parameters of bio-5’-AMP binding to dimerization surface variants ...................................................................................... 39 Table 4.3: Corrected heat capacity changes for Bio-5’-AMP binding to BirA variants ..................................................................................................................................... 41 Table 4.4: Intramolecular interactions formed upon BirA monomer activation ........ 45 Table 6.1. Reproduction of displacement titration measurements .............................. 50 Table 6.2 Comparison of direct and displacement bio-5’-AMP binding measurements. ..................................................................................................................................... 52 Table 6.3 Biotin binding enthalpy obtained from TAPS measurements. ................... 55 vii List of Figures Figure 1.1: BirA in the E.coli biotin regulatory system. ............................................... 3 Figure 1.2: Thermodynamic cycle illustrating coupled bio-5’-AMP binding and dimerization. ................................................................................................................. 3 Figure 1.3: Structural comparison of apo and holoBirA............................................... 5 Figure 1.4: Functional loops in the BirA dimer. ........................................................... 5 Figure 1.5: BirA∙bio-5’-AMP homodimerization free energies of dimerization surface variants. ......................................................................................................................... 8 Figure 3.1: Direct biotin titration of I280A BirA. ...................................................... 20 Figure 3.2: Comparison of thermodynamic parameters of biotin binding for BirA variants. ....................................................................................................................... 21 Figure 3.3: Biotin Displacement bio-5’-AMP titration of I280A BirA. ..................... 24 Figure 3.4: Energetic contribution to bio-5’-AMP binding for BirA variants. ........... 25 Figure 3.5: Bio-5’-AMP TAPS titration of G196A BirA ........................................... 28 Figure 3.6: Heat capacity change of bio-5’-AMP binding for G196A BirA. ............. 29 Figure 3.7: Comparison of heat capacity change measurements. ............................... 30 Figure 4.1: Structural comparison of G142A BirA to wild type BirA. ...................... 34 Figure 4.2: Corrected thermodynamic contributions to bio-5’-AMP binding. ........... 39 Figure 4.3: Effects of the correction for wild type BirA homodimerization on the heat capacity change of bio-5’-AMP binding..................................................................... 40 Figure 4.4: Comparison of corrected heat capacity change measurements. ............... 41 Figure 4.5.Functional correlations of the heat capacity change of bio-5’-AMP binding ..................................................................................................................................... 42 Figure 6.1. Discrepancy between bio-5’-AMP binding enthalpy from TAPS and displacement ITC measurements. ............................................................................... 48 Figure 6.2 Direct versus Displacement titration of W223A BirA. ............................. 52 Figure 6.3 Biotin TAPS titration of P143A BirA. ...................................................... 54 Figure 6.4 Difference of bio-5’-AMP binding enthalpy between direct and TAPS ITC measurements. ............................................................................................................. 55 Figure 6.5 Simulated influence of c-value on single-site ITC binding curve. ............ 57 Figure 6.6 Example ITC titration using excess ligand. ............................................... 60 Figure 6.7 Simulation of a direct biotin binding titration. .......................................... 61 1 Chapter 1 : Introduction Allostery in Protein Regulation Allostery is defined as the functional regulation of one site on a protein by an effector at a second site. This effector could be any small ligand, macromolecule, or post-translational modification1,2. Biologically, allostery allows modulation of protein function based on environmental conditions and is therefore critical for regulation of numerous cellular processes including metabolism3, transcription4, and signal transduction5. Historically, the MWC6 and KNF7 allosteric models were developed to describe enzyme cooperativity. These models describe homotropic cooperativity using a 2-state, conformational change from a tense, T, to a relaxed, R, state. However, the molecular details of these T and R states could only be described when crystal structures, such as those first obtained for horse hemoglobin were obtained by Max Perutz in 19708. With this data, he was able to develop a structural model for the T to R transition in hemoglobin. These structures allowed development of a molecular mechanism of the allosteric transition communication in hemoglobin in which O2 binding induces the breakage of constraining inter-subunit salt bridges. When these intermolecular salt bridges are removed, the free energy of subsequent O2 binding steps is more favorable leading to positive homotropic cooperativity. Allosteric mechanisms for proteins have now been characterized for a number of other systems. For example, Datta et. al. observed a pathway of 21 hydrogen bonds connecting the allosterically activated site to the modified functional sites in the 2 caspase-1 dimer. Only 12 of these hydrogen bonds were present in the apo protein9. Alanine mutagenesis significantly alters protease activity for four residues spanning the distance between the two sites indicating an apparent allosteric pathway. However, the secondary structure of proteins is not necessarily well defined. Both partially disordered proteins and intrinsically disorder proteins (IDPs) that have no well-defined secondary structure have been characterized10,11. Biologically, protein disorder is thought to allow for multispecificity in protein-protein interactions and is widely observed in signaling and regulatory pathways12,13. The folding of disordered regions upon effector binding allows the formation of additional intramolecular interactions associated with allosteric regulation. However, disorder to order transitions and their mechanistic roles in protein allostery are not well understood. Above 30% of all protein residues in eukaryotes are predicted to contain disordered regions14 and this knowledge gap could prove to be a significant barrier to protein structure-function prediction and drug design. Within this thesis the allosteric mechanism in one model protein, BirA, for which disordered loop folding is coupled to its allosteric activation is investigated. BirA as a model allosteric system E. coli BirA functions as both a transcriptional repressor and critical enzyme for fatty acid biosynthesis (Figure 1.1). Both functions require BirA-catalyzed synthesis of its corepressor, bio-5’-AMP, from biotin and ATP. The corepressor remains tightly bound to BirA after synthesis, enhancing homodimerization by -4 3 Figure 1.1: BirA in the E.coli biotin regulatory system. Following bio-5’-AMP synthesis, holoBirA dimerizes to form either a catalytic heterodimer (left) or a DNA-binding homodimer (right) on the same dimerization surface15. Figure 1.2: Thermodynamic cycle illustrating coupled bio-5’-AMP binding and dimerization. Binding of corepressor bio-5’-AMP enhances dimerization by the coupling free energy, -4.0±0.3kcal/mol16. 4 kcal/mol17. Following dimerization, BirA binds to the biotin operator, BioO, repressing transcription of biotin biosynthetic genes17. Bio-5’-AMP binding and subsequent homodimerization are important elements of the regulation of biotin biosynthesis in E. coli. Moreover this process serves as a model system for studying the mechanistic basis of protein allostery. Disordered loops play a role in the BirA allosteric mechanism Bio-5’AMP binding occurs at a site which includes two loops that are unfolded prior to ligand binding. Residues 116-126, the biotin binding loop (BBL), fold when biotin is bound. In addition to the BBL, residues 211-234, the adenylate binding loop (ABL), become ordered upon bio-5’-AMP binding (Figure 1.3). The ABL is critical for adenylate binding as evidenced by single alanine substitutions in this loop, which perturb bio-5’-AMP binding by up to 3.4 kcal/mol18. Upon bio-5’- AMP binding, a majority of these BirA variants also show decreased protein protection from site-specific subtilisin cleavage between residues 217 and 218 in the ABL, consistent with perturbation of loop folding18,19. Substitutions in the ABL also perturb holoBirA dimerization free energy by 0.5-1.5 kcal/mol, indicating functional communication to the dimerization surface. Together, these results are consistent with communication from the ligand binding surface to the dimerization surface by coupled corepressor binding and ABL folding. Similarly, perturbations were made in several dimerization surface loops to investigate their role in assembly of the BirA dimer. Alanine substitutions were made in four dimerization surface loops comprised of residues 140-143, 170-167, 193-197, and 280-283. These substitutions are all located in the dimerization interface 5 Figure 1.3: Structural comparison of apo and holoBirA. Apo BirA (PDB entry 1BIA) and a monomer of the holoBirA dimer (PDB entry 2EWN) are shown with the non-hydrolysable bio-5’-AMP analog, biotinol-5’-AMP (black), bound. Following ligand binding and subsequent dimerization, residues 116-124 (red), 212-222 (blue), and 194-197 (pink), and glycine 142 of the 140-143 (dark green) loops become ordered. Residues in the 170-176 (orange) and 280-283 (purple) loops are folded regardless of ligand binding. Figure 1.4: Functional loops in the BirA dimer. Due to their presence on the dimerization surface, alanine substitutions in the 140-143 (green), 194-197, (pink) and 280-283 loops are investigated for their role in allosteric activation. 6 (Figure 1.4) and the resulting proteins have variable ΔG°DIM values consistent with their roles in BirA dimerization20(Figure 1.5). Alanine substitution at residue G142 gives the largest perturbation to the dimerization free energy of 4.4±0.6 kcal/mol. In other words, G142A BirA dimerizes with energetics similar to those measured for apo BirA. This variant is also perturbed in adenylate binding by 3.0±0.4 kcal/mol relative to wild type BirA19. Consistent with abolition of coupling between corepressor binding and dimerization, the monomeric structure G142A BirA bound to biotinol-adenylate, a non-hydrolyzable bio-5’-AMP analog, resembles that of wild type apo BirA. Additionally residues 193-199 on the dimerization surface and residues 211-221 in the ABL are disordered in the G142A structure. This provides evidence for the role of coupled disorder-to-order transitions on the two functional surfaces in allosteric activation of the BirA monomer19. The data obtained for G142A BirA was the first to functionally demonstrate communication from the dimerization surface to the ligand binding surface. However, it is not known if other residues on the dimerization surface also communicate to the ligand binding surface. This paper further investigates the mechanism of communication from the dimerization to the ligand binding surface in the BirA monomer. By measuring corepressor binding of other alanine-substituted variants in the dimerization surface loops using isothermal titration calorimetry (ITC) titrations, both Gibbs free energy and enthalpy of bio-5’-AMP binding are determined. From these parameters, the entropy change of corepressor binding can be calculated so that the thermodynamic contributions to ligand binding can be discussed. Similar to G142A BirA, significant alterations in bio-5’-AMP binding free energy would 7 indicate other residues that also functionally communicate to the ligand binding surface. To further characterize the source of alterations in corepressor binding, the heat capacity change of bio-5’-AMP association was determined by linear regression of the temperature dependence of the binding enthalpy. Binding enthalpies from 10 to 25°C were measured using ITC at total association at partial saturation (TAPS) conditions21. The heat capacity change of bio-5’-AMP binding is negative for all variants and significant deviations from that of wild type BirA are observed. Following a correction for the contribution of BirA dimerization, a strong correlation between heat capacity change and the free energy of dimerization is observed for most variants. Residues for which the heat capacity change is more positive are those for which alanine mutagenesis results in the loss of intramolecular interactions on the dimerization surface. 8 Figure 1.5: BirA∙bio-5’-AMP homodimerization free energies of dimerization surface variants. Homodimerization free energies of bio-5’-AMP bound wild type (white) and single alanine variants in the 140-143 (green), 170-176 (orange), 193-197 (pink), and the 280-283 (purple) dimerization surface loops. Error bars represent the standard deviation from at least two sedimentation equilibrium measurements in 10mM Tris (pH 7.50), 200mM KCl, and 2.5mM MgCl2 at 20°C. Figure adapted from Adikaram, Beckett 201220 9 Chapter 2 : Materials and Methods Chemicals and biochemicals All chemicals and biochemicals used in this study were at least reagent grade. Isopropyl-β-D-1-thiogalactopyranoside (IPTG) was purchased from Teknova and 1M solutions were prepared in MQH20 (Millipore—Milli-Q synthesis, resistance=18.2MΩ∙cm) from dry powder immediately prior to use. A 100mM phenylmethylsulfonyl fluoride (PMSF, Sigma) solution was prepared by dissolving the dry powder in 95% ethanol. Luria Bertani (LB) broth for bacterial growth was prepared using 10g/L tryptone (Bacto), 5g/L yeast extract (Bacto), 5g/L NaCl (Mallinckrodt), 0.1% 1M NaOH (Fisher), and was used less than a week after sterilization by autoclaving at 250°F, 22 psi, for 20 minutes. Ampicillin (AMP, Fisher) was prepared as a fresh 100mg/mL solution in MQH2O immediately before addition to liquid LB media or LB-agar. Polyethylenimine (PEI) was purchased from Sigma and diluted to a 10% working solution in MQH20. This solution was adjusted to a pH of 7.5 with HCl. Because biotin has no spectral signal, it must be quantitatively prepared so that the concentration can be accurately known. D-biotin was obtained from Sigma- Aldrich and prepared in Standard Buffer (10mM Tris pH 7.5 at 20°C, 2.5mM MgCl2, 200mM KCl). Due to its poor solubility in H20, (0.22mg/mL, or 900μM at 25°C22), care was taken to prepare biotin solutions at a concentration well below this value. This ensures the weight of biotin measured accurately represents dissolved biotin and that no undissolved biotin solids are lost upon filtration. Biotin powder was weighed 10 into a small beaker, in which the stock solutions for 25mL of Standard Buffer (1M Tris Base, 1M MgCl2, 2M KCl) were added. After the addition of 80% of the total volume as MQH20 to dissolve the lyophilized powder, the pH of the solution was adjusted to 7.50 at 20°C with HCl. Following quantitative transfer to a 25mL volumetric flask, the remaining volume of MQH2O was added to bring the final volume to 25mL. The biotin solution was subsequently filtered with a 0.22µm mixed- cellulose GSWP filter (Millipore) that was pre-rinsed with MQH20. Following the filtration, the solution was aliquoted into 1mL fractions and stored at -80°C to minimize oxidation. The bio-5’-AMP was synthesized and purified in lab as previously described23,24. As an anhydride, hydrolysis of bio-5’-AMP to its constituents, biotin and adenosine monophosphate (AMP), can occur. Adenylate integrity was assessed by chromatographic separation on a cellulose F, TLC membrane (EMD) using a mobile phase of 3:2:1 (v:v:v) of amyl alcohol, formic acid, H20 respectively. Greater than 1nmol of biotin or bio-5’-AMP was visualized with a stain comprised of 1% H2SO4, 0.1% p-dimethylcinnemaldehyde solution in 95% ethanol. This solution quickly stains biotin-containing compounds bright pink. Purity was assessed by the presence of only a single bio-5’-AMP band (Rf=0.53). Expression and purification of BirA variants occurred as previously published20 except for wild type, E140A, and K283A BirA. These proteins were isolated as described in Appendix 3: Expression and Purification of BirA Variants. 11 Isothermal titration calorimetry (ITC) measurements All titrations were performed on a Microcal VP-ITC (Malvern) calorimeter with a 1.44mL cell volume. For all experiments, the sample cell contained protein and the 250μL mixing syringe contained the ligand. In each titration, 21-23 injections of ligand were added into the sample cell. Titrations were designed with two factors in mind: first the c-value ( 𝑐 = 𝐾𝐴𝑃𝑡𝑛; where KA is the association constant, Pt is the total protein concentration, and n is the stoichiometry of ligand binding), for the binding curve was within the recommended range for association constant determination of 5-500 (500 ≥ 𝑃𝑡∙𝑛 𝐾𝐷 ≥ 5)28. Second, care was taken so that both pre- saturation and post-saturation enthalpy plateaus were represented by several injections. This required the use of a larger number of lower-volume ligand injections. These factors, as well as usage of an experimental design tool29 are detailed in Appendix 2. Prior to performing a titration, all protein solutions (200-500uL) were dialyzed twice, the second dialysis overnight, in 250mL of 0.22um-filtered (GSWP) Standard Buffer (10mM Tris pH 7.5 at 20°C, 2.5mM MgCl2, 200mM KCl) at 4°C. To eliminate any precipitate, the protein solution was then filtered with a 0.45um PTFE syringe filter. The concentration of the protein solution was determined using absorbance spectroscopy (εM, 280nm=47510M-1cm-1). During this process the buffer from dialysis was degassed for at least 8 min using the Microcal ThermoVac accessory. To ensure buffer matching, this dialysate was used to dilute both protein and ligand solutions to the appropriate final concentrations (see results). Directly prior to ligand and protein loading, both solutions were degassed for 8-10 minutes in 12 the Microcal ThermoVac accessory 2-3°C above the working temperature. To further prevent interference from bubble formation/popping during the titration, each experiment was initiated with a 2μL injection to clear the syringe tip. The default stir speed and data filter time of 310 rpm and 2 seconds respectively were used for all titrations. Both the ‘Fast Equilibration’ and ‘Auto Equilibration’ options were used with the ‘High’ gain/feedback setting. Integration of the raw titration peaks was performed in the Microcal software suite in Origin 7.0; this process consisted of first computer-assisted baseline assignment followed by manual refinement. Multiple analyses of identical titration data indicated little variation in binding parameters. The heat of injection was assumed to be constant for all titrations based on previous biotin and bio-5’-AMP dilution experiments. This heat was calculated from the average value of heats associated with the final 4-5 injections of the titration, which were after protein saturation. Prior to model-fitting, this heat was subtracted from each heat of injection to obtain the heat of ligand binding. Injection heats were normalized to the moles of ligand injected, resulting in the final molar heats of injection. Each isotherm was fit using nonlinear least squares regression to a binding model in Origin 7.0 (see Direct, Displacement, and TAPS titrations below). It was from this regression that the equilibrium constant, molar enthalpy, and stoichiometry of binding was determined. The sample cell contents were removed following each titration, after which the cell was thoroughly washed with 800mL of working concentration of RBS35 detergent (1:50 RBS35:diH20) using the Microcal ThermoVac. To remove this detergent, the cell was rinsed with two or three 1L volumes of MQH2O. When not in 13 use, the sample cell was filled with MQH2O to prevent both the need for later equilibration in water and the deposition of any residue by evaporation of the cell contents. Direct ITC binding measurements of biotin binding parameters The direct binding method was used to determine the binding parameters for the BirA dimerization surface variants to biotin and for BirA variant G142A to bio-5- AMP. The design of direct binding titrations was aided by a spreadsheet-based simulation tool for a single-site binding model29. Each biotin titration consisted of twenty-one 7μL injections of 50 µM biotin into a 2.5μM BirA solution and was terminated by a single 100μL injection. This final injection sufficiently saturated BirA with biotin for a following displacement titration (see Displacement ITC). The Bio-5’-AMP direct titrations were performed with twenty 13μL injections of 20 µM bio-5’-AMP into a 2μM BirA solution. For all direct titrations, a 300 second delay between injections was sufficient for complete baseline equilibration. Integration of the resulting injection peaks yields the molar heat of injection, qi. Following baseline subtraction (see ITC measurements), the isotherms were fit to the single-site binding model: 𝑞𝑖 = n𝑀𝑡∆H°𝑉0 2 [1 + 𝑋𝑡 𝑛𝑀𝑡 + 1 𝑛𝐾𝐴𝑀𝑡 −√(1 + 𝑋𝑡 𝑛𝑀𝑡 + 𝑋𝑡 𝑛𝑀𝑡 ) 2 − 4𝑋𝑡 𝑛𝑀𝑡 ] using Origin 7.0 least-squares analysis software to determine the association constant, KA, binding stoichiometry, n, and molar enthalpy change, ∆H°, of ligand binding. 14 Displacement ITC binding measurement of bio-5’-AMP binding parameters Direct measurement of KA, bio-5’-AMP was not possible for several variants due to the corepressor’s tight interaction with BirA (KD, bio-5’-AMP<10-10M)30. To achieve an adequate c-value for KD,bio-5’-AMP measurement, a nanomolar protein concentration would be required. However, this protein concentration would provide an insufficient heat of binding to distinguish itself from the heat of dilution. Instead, bio-5’-AMP displacement of biotin from saturated BirA was measured. With this method, the apparent association constant, KA, app, is within an accessible range. This value is a function of the KA, Biotin (see direct binding) and KA, bio-5’-AMP, as well as the bulk biotin concentration, [biotin]. 𝐾𝑎𝑝𝑝 = 𝐾𝐴,𝑏𝑖𝑜5′𝐴𝑀𝑃 1 + 𝐾𝐴,𝑏𝑖𝑜𝑡𝑖𝑛[𝑏𝑖𝑜𝑡𝑖𝑛] Similarly, the enthalpy of bio-5’-AMP binding can be deconvoluted from the apparent measured binding enthalpy using the following expression in which ΔH°biotin has been previously measured by a direct titration. ∆𝐻°𝑎𝑝𝑝 = ∆𝐻°𝑏𝑖𝑜5′𝐴𝑀𝑃 − ∆𝐻°𝑏𝑖𝑜𝑡𝑖𝑛 𝐾𝐴,𝑏𝑖𝑜𝑡𝑖𝑛[𝑏𝑖𝑜𝑡𝑖𝑛] 1 + 𝐾𝐴,𝑏𝑖𝑜𝑡𝑖𝑛[𝑏𝑖𝑜𝑡𝑖𝑛] However, binding enthalpies determined by this method were not used in analysis of the titration data (see appendix 1). Instead, those values obtained from direct TAPS titrations were used. Displacement titrations were designed so that upon injection, bio-5’-AMP displaces biotin at its shared binding site on BirA. To ensure complete biotin saturation prior to titration, a 6μM final biotin concentration was used ([biotin]>KD, 15 Biotinx100), which was achieved at the end of each direct titration. To perform each displacement titration, twenty-one 12μL injections of 22μM bio-5’-AMP and 6µM biotin were titrated into the sample cell, containing 2.2µM BirA and 6µM biotin. Due to the slower kinetics of the displacement reaction, a longer, 360 second, equilibration period was required between injections. Total association at partial saturation (TAPS) calorimetry Calorimetric titrations in TAPS conditions21 were utilized to measure ∆Hbio-5’- AMP at temperatures ranging from 10°C-25°C for the purpose of calculating the molar heat capacity of corepressor binding. This method allows for more accurate determination of the binding enthalpy by decoupling it from the equilibrium association constant. TAPS conditions are achieved at high c-values (when 𝑃𝑡𝑛 𝐾𝐷 >500), at which there is complete binding upon ligand injection, for multiple injections. Because of the large dependence of pKatris on temperature, the pH of standard buffer used was adjusted using HCl to pH=7.5 at each working temperature. The faster kinetics of direct binding permitted a 300 second delay between injections. For all TAPS titrations, five 7μL injections of 20μM bio-5’-AMP are followed by a single, saturating, 180μL injection and terminated by six 10μL injections. The integrated heat from the first five injections represents the sum of the heat of binding and heat of dilution, ∆𝐻°𝑡𝑜𝑡. ∆𝐻°𝑡𝑜𝑡 = ∆𝐻°𝑏𝑖𝑜−5′−𝐴𝑀𝑃 + ∆𝐻°𝑑𝑖𝑙 The final six injections represent the molar heat of dilution, ∆𝐻𝑑𝑖𝑙. The difference of these terms yields the enthalpy of bio-5’-AMP binding: 16 ∆𝐻°𝑏𝑖𝑜−5′−𝐴𝑀𝑃 = ∆𝐻°𝑡𝑜𝑡 − ∆𝐻°𝑑𝑖𝑙 By subtraction of the average experimental heat of dilution, five measurements of the molar enthalpy, corresponding to the five pre-saturation injections, were obtained at each temperature. Molar heat capacity changes of bio-5’-AMP binding Heat capacity changes of bio-5’-AMP binding by the BirA variants were calculated using the ∆𝐻bio-5’-AMP values from the ITC-TAPS titrations above. The heat capacity change, ΔC°P, is the slope of the linear enthalpy versus temperature relationship, ∆𝐻° = 𝑞𝑃 = ∆𝐶𝑃° ∙ ∆𝑇 , as determined by linear regression. Each calculation of ∆C°p was comprised of five enthalpies of bio-5’-AMP binding at each of four temperatures, 10°C, 15°C, 20°C, 25°C. 17 Chapter 3 : Results Allosteric activation of BirA for dimerization occurs via communication of disorder-to-order transitions at the corepressor binding and dimerization surfaces. This was communication was further investigated by measuring the ligand binding properties of dimerization surface variants using isothermal titration calorimetry (ITC). Selection of alanine-substituted dimerization surface variants BirA variants with alanine substitutions in four dimerization surface loops reveal a range of effects (Figure 1.5) on the equilibrium dimerization constant20. A subset of these variants was selected for ligand binding measurements based on the following criteria: Variants in the 170-176 dimerization loop were not studied due to their proximity to the active site and known role in catalysis20. In the remaining three loops, only variants with dimerization affinities similar to or lower than that measured for wild type BirA were studied. This approach was adopted to target variants that, like G142A BirA, are more likely to be defective in communication to the ligand binding surface. Moreover, weak dimerization also minimizes contributions from the 41±3 kcal/mol homodimerization enthalpy to the ITC measurements31. Using these criteria, variants E140A, G142A, P143A, T195A, G196A, D197A, I280A, G281A, and K283A were selected. The dimerization energetics of these variants are summarized in Table 3.1. 18 Table 3.1: Homodimerization energetics of chosen BirA variants: Errors represent the standard deviation from at least two sedimentation equilibrium measurements in 10mM Tris (pH 7.50), 200mM KCl, and 2.5mM MgCl2 at 20°C.20 BirA variant K dim (M) a ∆G o dim (kcal/mol) a Wild Type 7(±3)x10 -6 -6.8±0.3 E140A 22(±5)x10 -6 -6.2±0.2 G142A 17(±52)x10 -3 -2.4±0.9 P143A 4(±1)x10 -5 -5.9±0.2 T195A 10(±4)x10 -5 -5.3±0.2 G196A 5(±4)x10 -6 -7.1±0.4 D197A 5(±3)x10 -4 -4.5±0.3 I280A 4(±3)x10 -3 -3.3±0.7 G281A 9(±4)x10 -5 -5.4±0.3 K283A 12(±7)x10 -6 -6.6±0.4 19 Dimerization loop variants bind biotin with affinities similar to wild type BirA As a control for the presence of non-allosteric effects on ligand binding, such as loss of binding site integrity, biotin binding to all variants was measured. Since this ligand does not elicit an allosteric response16, perturbations to biotin binding are unlikely to reflect effects on inter-surface communication. Thermodynamic parameters of biotin binding were obtained by direct ITC binding titrations. See materials and methods for experimental details. ITC titration data for biotin binding to I280A BirA are shown in Figure 3.1. Analysis of the data indicated excellent agreement to a single site binding model, yielding the dissociation constant, molar enthalpy, and stoichiometry of biotin binding from which the Gibbs free energy, ∆G°biotin, and entropy, -T∆S°biotin were calculated. The values of these parameters for the I280A variant (Table 3.2) are identical, within one standard deviation, to those obtained for wild type BirA. Biotin binding titrations performed on all other variants yielded data that are also well described by a single site model. Moreover, the thermodynamic parameters obtained from analysis of the titration data are, with few exceptions, equivalent to those obtained for wild type BirA (Table 3.2). For all variants, the equilibrium dissociation and Gibbs free energy of binding are identical and the same as those obtained for wt BirA. However, the thermodynamic contributions to the free energy of binding for G142A BirA differ from those obtained for wild type BirA. The molar binding enthalpy for this variant is -1.8 ±0.4kcal/mol more favorable than wild type 20 Figure 3.1: Direct biotin titration of I280A BirA. This figure illustrates data for the direct binding ITC method. Each titration was performed by injecting twenty 7uL injections of 50 µM biotin into 1.44mL of BirA. Both protein and ligand solutions were prepared in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, and 2.5mM MgCl2. The top panel is the ITC data trace following manual baseline correction. After integration and normalization to the moles of injectant, the binding isotherm, bottom panel, is obtained. The binding isotherm for I280A BirA is fit to a single site model for the biotin binding parameters. 21 Table 3.2: Thermodynamic parameters of biotin binding for BirA variants Measured at 20°C by direct ITC titration in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, 2.5mM MgCl2. Each parameter represents the average from at least two independent measurements. The uncertainty provided is the propagated 68% confidence interval from the nonlinear-least-squares regression to a single site model. Figure 3.2: Comparison of thermodynamic parameters of biotin binding for BirA variants. A) Change in the Gibbs free energy of biotin binding, ∆∆G°biotin, variant - ∆∆G°biotin, wt, shown for the nine variants. Error bars shown represent propagated 68% confidence intervals from wild type and the appropriate variant. B) The enthalpic (red) and entropic (blue) components of the biotin binding free energy for each BirA variant at 20oC. Errors bars shown are the propagated 68% confidence intervals resulting from nonlinear least squares analysis. Protein K D (M) ΔG° (kcal/mol) ΔH° (kcal/mol) –TΔS° (kcal/mol) N wild type 4.7(±0.8)x10 -8 -9.8±0.1 -20.1±0.4 10.3±0.5 0.93±0.01 E140A 4.0(±0.4)x10 -8 -9.92±0.05 -20.1±0.2 10.2±0.2 0.931±0.006 G142A 4.0(±0.4)x10 -8 -9.92±0.06 -21.9±0.2 12.0±0.3 1.085±0.009 P143A 4.5(±0.3)x10 -8 -9.85±0.04 -19.7±0.1 9.9±0.1 0.955±0.004 T195A 4.5(±0.5)x10 -8 -9.85±0.06 -19.9±0.2 10.1±0.2 0.894±0.007 G196A 5.1(±0.5)x10 -8 -9.78±0.06 -20.3±0.2 10.5±0.2 0.984±0.007 D197A 4.2(±0.4)x10 -8 -9.89±0.05 -20.1±0.2 10.2±0.5 1.260±0.008 I280A 4.6(±0.5)x10 -8 -9.84±0.06 -20.6±0.2 10.8±0.2 1.027±0.007 G281A 4.1(±0.4)x10 -8 -9.91±0.05 -20.4±0.2 10.5±0.2 0.878±0.006 K283A 3.9(±0.2)x10 -8 -9.93±0.03 -20.8±0.3 10.8±0.3 0.904±0.009 B. A. 22 BirA. Additionally, the entropic component of binding, -T∆S°biotin, is 1.7±0.6 kcal/mol less favorable. Nonetheless, the free energy of biotin binding remains equivalent to that of wild type at -9.92±0.06 kcal/mol. With one exception, the resolved stoichiometries of biotin binding range from 0.878 and 1.085 are consistent with a 1:1 binding stoichiometry. The exception, BirA D197A, yields a reproducibly higher value of 1.260±0.008. Since this value is consistent with results of stoichiometric titrations monitored by fluorescence, (unpublished data) this behavior appears to be an intrinsic property of the variant. Alanine substitution at residue G142 greatly perturbs bio-5’-AMP binding. The perturbed ABL folding of the G142A BirA variant indicates functional communication from the BirA dimerization surface to the ligand binding surface32. The alanine-substituted variants that are also defective in dimerization were analyzed to determine if this communication, measured by loss of bio-5’-AMP binding affinity, extends to other residues on the dimerization surface. Because of the high c-values that would be required, the affinity (KD,bio-5’-AMP = 5(±2)x10-11M) of bio-5’-AMP to wild type BirA precludes the use of direct binding ITC titrations. Therefore, it was necessary to perform corepressor binding measurements using the displacement titration method33. For this experiment, biotin-saturated BirA is titrated with bio-5’-AMP which displaces biotin in the course of binding. Nonlinear least squares analysis of this isotherm using a single site model yields the apparent binding parameters used to calculate (see Materials and Methods) the bio-5’-AMP binding parameters. 23 Data obtained for the I280A BirA displacement titration are shown in Figure 3.3. In contrast to biotin binding, this titration is endothermic in its initial injections. This is because the positive enthalpy of biotin release is larger in magnitude than the negative enthalpy of bio-5’-AMP binding. Consequently, the apparent molar binding enthalpy for these measurements was about +6kcal/mol, compared to -20 kcal/mol for the direct biotin binding titrations. This results in a lower signal-to-noise for the parameters obtained using the competitive binding method. Nonetheless, the data are well described by the single site model. Moreover, the thermodynamic parameters of bio-5’-AMP binding (Table 3.3) derived from the analysis indicate that I280A BirA is characterized by corepressor binding parameters identical to those determined for wild type BirA. Displacement titrations were used to measure bio-5’-AMP binding to all variants with the exception of G142A BirA. These measurements yielded free energies of bio-5’-AMP binding (Figure 3.4A) identical, within one standard deviation, to that of wild type BirA. The direct titration of G142A BirA with the corepressor indicated much weaker binding with an equilibrium dissociation constant of 8(±1) x10-9 M, approximately 160-fold greater than that measured for wt BirA. The enthalpies of ligand binding obtained by displacement titration were deemed innacurate (see appendix1) and therefore are excluded from these results.The enthalpies derived from TAPS titrations give insight into the effects observed in the variants. In contrast to the Gibbs free energies of binding, the BirA variants display variable enthalpic and entropic components to bio-5’-AMP binding, illustrated in Figure 3.4B. The loss of corepressor binding affinity for G142A BirA compared to 24 Figure 3.3: Biotin Displacement bio-5’-AMP titration of I280A BirA. This figure illustrates data for the competitive binding ITC method. Each titration was performed by injecting 20-22, 12uL injections of 22 µM bio-5’-AMP into 1.44mL of BirA. Both protein and ligand solutions were prepared in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, 2.5mM MgCl2, and 6µM biotin. The top panel is the ITC data trace following manual baseline correction. Following integration and normalization to the moles of injectant, the binding isotherm, bottom panel, is obtained. The binding isotherm for I280A BirA is fit to a single site model for the apparent binding parameters from which the bio-5’-AMP binding parameters are calculated Table 3.3: Thermodynamic parameters of bio-5’-AMP binding. Above binding parameters were measured at 20°C by displacement ITC titration in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, 2.5mM MgCl2. Each parameter represents the average from at least two independent measurements. Uncertainty provided is the propagated 68% confidence interval from the nonlinear-least-squares regression to a single site model. a Obtained from nonlinear least squares analysis of displacement titration data to a single site binding model in Origin 7.0. bEnthalpy obtained from direct TAPS titration at 20°C. c Entropy, -T∆S°, calculated using the expression ∆G°=∆H°-T∆S° Protein K D (M) a ΔG° b (kcal/mol) ΔH° c (kcal/mol) –TΔS° d (kcal/mol) n a wild type 3(±3)x10 -11 -14.1±0.5 -12.6±0.3 -1.4±0.5 1.06±0.02 E140A 7(±3)x10 -11 -13.6±0.2 -12.43±0.06 -1.2±0.2 0.90±0.01 G142 8 (±1)x10 -9 -10.85±0.09 -13.3±0.1 2.9±0.2 1.01±0.01 P143A 8(±2)x10 -11 -13.5±0.2 -12.2±0.01 -1.4±0.2 0.96±0.01 T19 A 10(±4)x10 -11 -13.4±0.2 -12.58±0.07 -0.9±0.2 0.88±0.01 G196A 2(±3)x10 -11 -14.2±0.8 -11.90±0.1 -2.4±0.8 1.09±0.03 D197A 9(±3)x10 -11 -13.5±0.2 -12.43±0.2 -1.1±0.3 1.21±0.02 I280A 9(±2)x10 -11 -13.5±0.1 -12.5±0.1 -0.9±0.2 1.06±0.01 G281A 8(±3)x10 -11 -13.6±0.2 -12.7±0.2 -0.9±0.3 0.89±0.01 K283A 9(±2)x10 -11 -13.5±0.3 -13.3±0.2 -0.2±0.2 0.89±0.01 25 Figure 3.4: Energetic contribution to bio-5’-AMP binding for BirA variants. A) Change in the Gibbs free energy of bio-5’-AMP binding, ∆∆G°bio-5’-AMP, variant-∆∆G°bio-5’-AMP, wt, for the nine variants studied. Error bars are the propagated uncertainties from both wild type and appropriate variant. B) The enthalpic (red) and entropic (blue) components of the free energy of bio-5’-AMP binding for each BirA species studied. Errors bars shown are the propagated 68% confidence intervals. 26 wild type reflects both a modestly more favorable ∆H°G142A, by -0.g237±0.3 kcal/mol , and less favorable -T∆S°G142A of 4.3±0.5 kcal/mol. In agreement with the biotin binding measurements, the stoichiometries of bio-5’-AMP binding, shown in Table 3.3, are consistently between 0.89 and 1.09, except for variant D197A ΔC°p,bio-5’-AMP is more negative for BirA dimerization surface variants To further probe the structural basis of the BirA allosteric mechanism, the heat capacity changes associated with bio-5’-AMP binding to the variants were determined from measurements of the temperature dependence of the enthalpy of ligand binding. For each protein, the enthalpy of bio-5’-AMP binding was measured between 10°C and 25°C using the ITC total association at partial saturation (TAPS) method21. TAPS titration data for BirA variant G196A are shown in Figure 3.5. Analogous to the direct biotin titration in Figure 3.1, the initial five injections, when integrated, represent the combined heats of ligand binding and ligand dilution. A large injection of bio-5’-AMP then saturates the protein so that the remaining six injections only measure the heat of ligand dilution. The difference in the heats for these injections, when normalized to the ligand concentration, yields the molar enthalpy of binding. Therefore, from a single titration, five measurements of the bio- 5’-AMP binding enthalpy were obtained. Linear regression of this data for G196A BirA yields a heat capacity change associated with bio-5’-AMP binding of -240±6 cal/mol∙K. The r2 value of the linear regression, 0.95, indicates the heat capacity change was constant over the temperature range used. 27 Linear regression of the molar bio-5’-AMP binding enthalpy versus temperature data for the remaining BirA variants (Table 3.4) indicated a constant heat capacity over the temperature range employed as judged by r2 values ranging from 0.91 to 0.98. The ∆C°p,bio-5’-AMP is more negative for all variants than the value obtained for wild type. The ∆C°p, bio-5’-AMP values are, within one standard deviation, identical within the 280-283 loop. Additionally, no significant differences in ∆C°p, bio- 5’-AMP are observed among variants E140, P143A, T195A and D196A. The tightest dimerizing variant, G196A BirA, has the most positive heat capacity but, G142A BirA, which has the following most positive ∆C°p,bio-5’-AMP, is most defective in dimerization. 28 Figure 3.5: Bio-5’-AMP TAPS titration of G196A BirA Each titration was performed by injecting first five, 7µL injections of 20 µM bio-5’-AMP into 1.44mL of BirA. To saturate the protein solution, 180µL of 20µM bio-5’-AMP were injected. After saturation, six 10µL injections of 20 µM bio-5’-AMP were performed. Both protein and ligand solutions were prepared in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, 2.5mM MgCl2. The top panel is the ITC data trace following manual baseline correction. After integration and normalization to the moles of injectant, the binding isotherm, bottom panel, is obtained. The difference between the molar heat before and after protein saturation provides the enthalpy of ligand binding. 29 Figure 3.6: Heat capacity change of bio-5’-AMP binding for G196A BirA. The heat capacity change was obtained from the slope of linear least squares regression of the enthalpy-temperature relationship, ∆𝐶𝑝° = ∆𝐻° ∆𝑇° . Either four or five measurements of ∆H°bio-5’-AMP were obtained at each temperature by a single ITC-TAPS titration at 10, 15, 20, and 25°C. 30 Table 3.4: Heat capacity change of bio-5’-AMP binding for BirA dimerization surface variants. Calculated from linear regression of the temperature dependence of ∆H°bio-5‘-AMP (Figure 3.6). Uncertainty given in ∆C°p is the 68% confidence interval from linear regression. Change in ∆C°p from wild type, ∆∆C°p, is calculated by ∆C°p,variant-∆C°p,wt. Uncertainty in ∆∆C°p is the propagated error in heat capacity change for wild type and the BirA variant. W T B ir A E 1 4 0 A G 1 4 2 A P 1 4 3 A T 1 9 5 A G 1 9 6 A D 1 9 7 A I 2 8 0 A G 2 8 1 A K 2 8 3 A -0.5 -0.4 -0.3 -0.2 -0.1 -0.0  C p ( k c a l / m o l  K ) Figure 3.7: Comparison of heat capacity change measurements. The ∆C°p of wild type (grey), upon alanine substitution in the 140-143 loop (green), the 193- 197 loop (pink) and the 280-283 loop (purple). Error bars are the standard deviation of the slope from linear regression Protein ∆C P a (kcal/mol∙K) ∆ΔC P (kcal/mol∙K) wild type -0.30±0.01 — E140A -0.37±0.03 -0.07±0.01 G142A -0.300±0.008 -0.00±0.03 P143A -0.363±0.008 -0.06±0.01 T195A -0.37±0.01 -0.07±0.01 G196A -0.240±0.006 0.06±0.01 D197A -0.364±0.009 -0.06±0.01 I280A -0.31±0.01 -0.01±0.01 G281A -0.31±0.01 -0.01±0.01 K283A -0.29±0.02 0.01±0.02 31 Chapter 4 : Discussion The loss of Gibbs free energy of both dimerization and bio-5’-AMP binding measurements in G142A BirA indicate that coupling between ligand binding and dimerization has been abolished. Of the alanine substitutions introduced on the dimerization surface, only G142A results in defective bio-5’-AMP binding indicating disruption of elements critical for site-to-site communication in the wild type protein. Consistent with the role of disorder-to-order transitions in allosteric activation of the BirA monomer, the G142A crystal structure indicates an unfolded ABL and 193-197 loop. Additionally, N-terminal extension of the 147-164 helix by four residues that is observed in wild type BirA upon corepressor binding is not observed in the G142A structure. Large heat capacity changes upon bio-5’-AMP binding for the variants indicate that although communication to the ligand binding surface is not observed for the remaining residues, perturbations to the dimerization surface are present and correlate well with the resulting effects on BirA dimerization free energy. Moreover, these results are consistent with formation of intramolecular interactions upon loop folding being critical for the modulation of dimerization free energy. BirA integrity is maintained upon alanine substitution Alanine substitutions in all dimerization surface variants elicit no observable penalty to the free energy of biotin binding. Therefore, it is reasonable to assume that there is no large conformational change or loss of protein stability resulting from mutagenesis that would functionally disturb bio-5’-AMP binding. This assumption is 32 further supported by the BirA G142A crystal structure which demonstrates retention of secondary structural elements present in apo BirA and the holo BirA monomer. Moreover, alanine substitutions in the ABL induced deficient bio-5’-AMP binding without disturbing the free energy of biotin binding18. Therefore, because alanine substitutions in these flexible loops do not indicate disruption of the protein structure, any alteration to the Gibbs free energy of bio-5’-AMP binding upon alanine substitution can be considered due to perturbation of long distance allosteric communication. Residue G142 displays evidence for communication to the ligand binding site The free energy of bio-5’-AMP binding is only perturbed by alanine substitution at residue G142. As discussed above, this result demonstrates perturbation of a long distance communication from the dimerization loop residues to the ligand binding surface. Therefore, the structural differences observed in G142A BirA represent perturbations to the features of the allosteric activation of BirA (Figure 4.1). The 193-197 loop remains folded in the G142A structure and alanine substitutions in this region do not perturb ligand binding. Therefore the folding of residues 193-197 is not a direct consequence of ABL folding. Instead, it must be influenced by other local effects, such as intramolecular interactions that form as a consequence of the 143-146 alpha helical extension. However, it is not clear how this alpha helical extension is communicated from the ligand binding surface. 33 Correction of enthalpy of bio-5’-AMP binding for holoBirA dimerization Titration data indicate that all variants except for G196A have different energetic contributions to bio-5’-AMP binding than wild type. Variant G196A is unique in that it is the only protein with a more favorable dimerization free energy than wild type BirA. Because of the linkage between corepressor binding and dimerization, dimer assembly is a complicating factor in the analysis of ITC-derived binding enthalpies. Moreover, because of the large positive enthalpy of dimerization, 41±3 kcal/mol31, low quantities of dimer formation contributes significantly to the measured heats. For this reason, despite designing the experiments to minimize this effect, a correction for the heat of dimerization was performed on the measured enthalpies of bio-5-AMP binding. The fraction of BirA dimer, 𝜒𝐷𝐼𝑀, was calculated for each variant at the total protein concentration, Ptot, at the end of each titration. This calculation was performed using the following expression and the equilibrium dimerization constants, KA,DIM, previously measured at 20°C20: 𝜒𝐷𝐼𝑀 = 1 − 𝜒𝑚𝑜𝑛𝑜𝑚𝑒𝑟 = 1 − −1 + √1 + 8𝐾𝐴,𝐷𝐼𝑀𝑃𝑡𝑜𝑡 4𝐾𝐴,𝐷𝐼𝑀𝑃𝑡𝑜𝑡 34 Figure 4.1: Structural comparison of G142A BirA to wild type BirA. Structural features of the dimerization surface are compared between wt apo (pdb entry 1BIA), wt holo (2EWN), and G142A (4WF2) BirA. Folding of the 193-197 loop (pink) is dependent on the alpha helical extension observed in residues 143-146 (dark green). It should be noted, however, that this is in context of the holoBirA dimer (left). 35 At the protein concentrations used, at least 10% dimer was present in wild type BirA and variants G196A, K283A, and E140A BirA (Figure 4.1) Therefore, the experimental contribution from BirA dimerization, qdim, to the measured heat of injection, qinj, is: 𝑞𝑖𝑛𝑗 = 𝑞𝑑𝑖𝑚 + 𝑞𝑏𝑖𝑜−5′−𝐴𝑀𝑃 where qbio-5’-AMP is the desired parameter, heat due to ligand binding. This value was calculated using the equation: 𝑞𝑑𝑖𝑚 = ∆𝐻°𝑑𝑖𝑚 ∙ [𝐵𝑖𝑟𝐴2] = ∆𝐻°𝑑𝑖𝑚 ([ℎ𝑜𝑙𝑜𝐵𝑖𝑟𝐴] − −1 + √1 + 8𝐾𝐴,𝐷𝐼𝑀[ℎ𝑜𝑙𝑜𝐵𝑖𝑟𝐴] 4𝐾𝐴,𝐷𝐼𝑀 ) using the enthlapy of BirA dimerization, ∆𝐻°𝑑𝑖𝑚, obtained for wild type BirA. The incremental change in ligand-bound protein concentration, [holoBirA], was equal to the incremental ligand concentration calculated by the Microcal VP-ITC software due to stoichiometric binding conditions. The dimerization constants for each variant, KA,dim, are known at 20°C20, and were determined at the remaining temperatures using the wild-type BirA temperature dependence31. To determine the enthalpy of binding, the measured heat due to bio-5’-AMP binding is normalized to the moles of injectant using the following equation: ∆𝐻°𝑏𝑖𝑜−5′−𝐴𝑀𝑃 = 𝑞𝑏𝑖𝑜−5′−𝐴𝑀𝑃 ∙ 𝐶𝑖𝑛𝑗 𝑉𝑖𝑛𝑗 Where Cinj is the concentration of the ligand solution and Vinj is the volume of each injection. By performing this correction, it is assumed that the enthalpy of dimerization of the BirA variants was constant and identical to that measured for wild type BirA. 36 The basis for this assumption is that measurements of the effect of replacing H2O with D2O are consistent with solvent release serving as the molecular origin of this large positive enthalpy34. Because similar isotope effects are observed on the dimerization free energies of T195A and V219A BirA, it seems reasonable to assume ∆HDIM does not differ dramatically from that of wild type despite their free energy penalties to dimerization. Correcting for dimerization reveals binding enthalpies for wild type and G196A BirA now more consistent with the other variants (Figure 4.2). The entropy of binding is revealed to be unfavorable for wild type, G196A, and K283A BirA. This enthalpy-entropy compensation phenomenon, in which large variations in thermodynamic parameters lead to little or no change in ∆G°binding has been seen in other systems in which small perturbations are made to interacting species35,36. In G142A BirA, the loss of free energy stems from a less favorable enthalpy by 2±1 kcal/mol compared to wild type and a more negative, but within error, -T∆S° by 1±1 kcal/mol. A less favorable binding enthalpy is also observed in T195A, D197A, I280A, and G281A BirA, each of which dimerizes at least 1 kcal/mol less favorably than wild type. This loss of binding enthalpy may represent the inability to 37 Table 4.1:Fraction of BirA dimer present in ITC titrations Fraction dimers above were calculated using the expression: 𝜒𝐷𝐼𝑀 = 1 − 𝜒𝑚𝑜𝑛𝑜𝑚𝑒𝑟 = 1 − −1 + √1 + 8𝐾𝐴,𝐷𝐼𝑀𝑃𝑡𝑜𝑡 4𝐾𝐴,𝐷𝐼𝑀𝑃𝑡𝑜𝑡 at 1.8uM BirA, the average concentration at the end of the TAPS measurements performed. The Dimerization parameters used were obtained from Adikaram 201220 BirA variant Fraction Dimer Wild Type 0.272 E140A 0.125 G142A 0.000 P143A 0.077 T195A 0.034 G196A 0.327 D197A 0.007 I280A 0.001 G281A 0.037 K283A 0.195 38 optimize electrostatic interactions on the dimerization surface upon allosteric activation. Although the observed changes in bio-5’-AMP binding enthalpy could also reflect structural changes on the ligand binding surface, all hydrogen bonding interactions to the ligand remain present in G142ABirA and the ligand binding affinity remains unchanged despite the enthalpic penalty in the remaining variants (Figure 4.2,Table 4.2). For this reason it appears the loss of dimerization free energy stems from disrupting intramolecular interactions on the dimerization surface of the monomer. ∆C°p,bio-5’-AMP values directly correlate to BirA dimerization All BirA variants show a more negative heat capacity change of bio-5’-AMP binding than wild type and G196A BirA. Additionally, no correlation is observed between heat capacity change and dimerization free energy of each variant (Figure 4.5A) .However, because the heat capacity changes were derived from the enthalpies of bio- 5’-AMP binding, the heat of BirA dimerization is also a consideration. Therefore, the correction was performed as above for the binding enthalpies at each temperature obtained by ITC TAPS titrations. Due to the temperature dependence of BirA dimerization31, the KD,DIM at each temperature was calculated from the variant’s dimerization free energy at 20°C. This was performed by using the Van’t Hoff equation, assuming the BirA temperature dependence of dimerization was conserved upon alanine mutagenesis. The corrected heat capacity changes ( Figure 4.3,Table 4.3) remain negative but are higher in magnitude compared to the non-corrected values. It is revealed that wild type, K283A, and G196A BirA have the most negative heat 39 Table 4.2: Corrected thermodynamic parameters of bio-5’-AMP binding to dimerization surface variants All measurements were performed in 10mM Tris (pH 7.5), 200mM KCl, 2.5mM MgCl2 at 20°C. Each value represents the average of at least two independent ITC titrations. Errors given are 68% confidence intervals propagated from the nonlinear least squares analysis. a Obtained from nonlinear least squares analysis of the titration data to a single site binding model in Origin 7.0. b Measured using direct TAPS titrations, corrected for BirA homodimerization as described in discussion c Entropy, -T∆S°, calculated from titration results using the expression ∆G°=∆H°-T∆S° W T E 1 4 0 A G 1 4 2 A P 1 4 3 A T 1 9 5 A G 1 9 6 A D 1 9 7 A I 2 8 0 A G 2 8 1 A K 2 8 3 A -16 -12 -8 -4 0 4 H -TS E n e r g y ( k c a l / m o l ) Figure 4.2: Corrected thermodynamic contributions to bio-5’-AMP binding. The enthalpic (red) and entropic (blue) components of the free energy of bio-5’AMP binding for each BirA species studied following correction for BirA homodimerization. Errors bars shown are the propagated 68% confidence intervals. Protein K D (M) a ΔG° a (kcal/mol) ΔH° b (kcal/mol) –TΔS° c (kcal/mol) n a wild type 3(±3)x10 -11 -14.1±0.5 -15±1 1±1 1.06±0.02 E140A 7(±3)x10 -11 -13.6±0.2 -13.7±0.5 0.0±0.5 0.90±0.01 G142A 8 (±1)x10 -9 -10.85±0.09 -13.2±0.1 2.3±0.2 1.010±0.01 P143A 8(±2)x10 -11 -13.5±0.2 -12.9±0.2 -0.6±0.3 0.96±0.01 T195A 1.0(±0.4)x10 -10 -13.4±0.2 -12.9±0.1 -0.5±0.2 0.88±0.01 G196A 2(±3)x10 -11 -14.2±0.8 -16±1 2±2 1.09±0.03 D197A 9(±3)x10 -11 -13.5±0.2 -12.2±0.2 -1.3±0.3 1.21±0.02 I280A 9(±2)x10 -11 -13.5±0.1 -12.6±0.1 -0.9±0.2 1.06±0.01 G281A 8(±3)x10 -11 -13.6±0.2 -13.1±0.3 -0.5±0.4 0.89±0.01 K283A 9(±2)x10 -11 -13.5±0.3 -15.3±0.7 1.8±0.2 0.89±0.01 40 Figure 4.3: Effects of the correction for wild type BirA homodimerization on the heat capacity change of bio-5’-AMP binding. Data shown are enthalpy of bio-5’-AMP binding as determined by TAPS titrations at 10-25C before (black) and after (grey) applying the correction for BirA homodimerization. 41 Table 4.3: Corrected heat capacity changes for Bio-5’-AMP binding to BirA variants Heat capacity changes of bio-5’-AMP binding to the BirA dimerization surface variants corrected for the presence of BirA dimerization. aCalculated from linear regression of the temperature dependence of the corrected ∆H°bio-5‘- AMP (Figure 3.6). Uncertainty given in ∆C°p is the 68% confidence interval from linear regression. b Change in ∆C°p from wild type, ∆∆C°p Calculated by the expression ∆C°p,variant-∆C°p,wt. Uncertainty in ∆∆C°p is the propagated error in heat capacity change for wild type and BirA variant. W T E 1 4 0 A G 1 4 2 A P 1 4 3 A T 1 9 5 A G 1 9 6 A D 1 9 7 A I 2 8 0 A G 2 8 1 A K 2 8 3 A -0.8 -0.6 -0.4 -0.2 -0.0  C  p ( k c a l / m o l ) Figure 4.4: Comparison of corrected heat capacity change measurements. The ∆C°p of bio-5’-AMP binding before correction (dark) corrected for the enthalpy of BirA homodimerization (light). Error bars are the standard deviation of the slope from linear regression. Protein ∆C P a (kcal/mol∙K) ∆ΔC P b (kcal/mol∙K) Corrected Wild Type -0.75±0.06 — E140A -0.58±0.05 0.17±0.08 G142A -0.31±0.01 0.44±0.06 P143A -0.49±0.02 0.26±0.06 T195A -0.43±0.02 0.32±0.06 G196A -0.68±0.06 0.07±0.08 D197A -0.37±0.01 0.38±0.06 I280A -0.31±0.01 0.44±0.06 G281A -0.38±0.01 0.37±0.06 K283A -0.59±0.04 0.16±0.07 42 -7 -6 -5 -4 -3 -2 -1  G  d i m ( k c a l / m o l ) -0.45 -0.40 -0.35 -0.30 -0.25 -0.20 -16 -15 -14 -13 -12 -11 -10 C p,bio-5'-AMP (kcal/mol)  G  b i o - 5 '- A M P ( k c a l / m o l ) -8 -6 -4 -2 0  G  d i m ( k c a l / m o l ) -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -15.5 -13.0 -10.5 -8.0 C p,bio-5'-AMP (kcal/mol)  G  b i o - 5 '- A M P ( k c a l / m o l ) Figure 4.5.Functional correlations of the heat capacity change of bio-5’-AMP binding The change in heat capacity is illustrated in relation to both dimerization and bio-5’-AMP binding free energies for each BirA variant. Error bars represent the 68% confidence intervals of the respective thermodynamic parameters. A) Correlation plot of the heat capacities as measured. B) Correlation plot using the corrected heat capacity values. BirA variants G142A and I280A (□) are excluded from the linear regression shown. B. A. 43 capacity changes (Figure 4.4). This correction reveals a striking positive correlation between dimerization free energy and ∆∆C°p (Figure 4.5B).Excluding variants G142A and I280A, which are not well-described by a linear function, the remaining variants assume a linear relationship with a slope of 4.5 kcal/mol of dimerization free energy per kcal/mol increase in ∆C°p. Interpretations of ΔC°p have often focused on correlations established to changes in solvent accessible surface area37. However, this interpretation has not been demonstrated to be consistent in all systems. Changes in hydrogen bonding and electrostatic interactions have also been demonstrated to give rise to negative heat capacity changes38,39. An alternative interpretation given by Alan Cooper is that negative heat capacity changes may be an expected consequence for the disorder to order transitions occurring in BirA. He demonstrates this by deriving ∆C°p using a statistical thermodynamic model in which a network of interacting subunits undergo a cooperative transition40. Following this interpretation, it would be expected that the heat capacity change would be more positive in variants that do not undergo as significant of a disorder-to-order transition. Using this interpretation, the most negative heat capacities are consistent with the full disorder-to-order transition observed in wild type BirA including G196A, E140A and K283A. The remaining BirA variants have significantly more positive heat capacity changes consistent with perturbations of loop folding. Because there are no functional effects observed to ligand binding, these perturbations to loop folding can be tied to the 193-197 disorder to order transition. The heat capacity change results also affirm the predictions32 from 44 the G142A structure that disorder-to-order loop folding is coupled to allosteric activation of BirA by demonstrating a direct correlation between dimerization free energy and heat capacity change, a parameter associated with both protein folding and disorder-order transitions. Formation of intramolecular interactions is observed upon allosteric activation Analysis of the wild type apo and holoBirA structures reveals multiple intramolecular interactions occurring upon helical extension of residues 143-146 and folding of the 193-197 loop (Table 4.4). Disruption of these intramolecular interactions formed upon allosteric activation by alanine substitution has moderate effects on the heat capacity change of bio-5’-AMP binding (Figure 4.4). Due to the correlation between ∆∆G°DIM and ∆∆C°p these variants also have more positive dimerization free energies consistent with perturbed loop folding. However, there are no obvious intramolecular interactions that exist in wild type holoBirA which would be perturbed in I280A BirA, despite its deficit to free energy of dimerization. Because of the pronounced functional effect observed for these variants and their lack of correlation with heat capacity change, the effects that are observed on the dimerization free energy are beyond perturbation to folding of the 193-197 loop. 45 Table 4.4: Intramolecular interactions formed upon BirA monomer activation Residues Interaction Atomic Distance (Å) K194-D197 Salt Bridge 3.4 K194b-D197b Hydrogen bond 3.0 K194b-A198b Hydrogen bond 3.4 P143b-A146b Hydrogen bond 2.9 R119-D282 Salt Bridge 4.1 R119-E313 Salt Bridge 3.7 R118-D176 Salt Bridge 3.0 R118-V189b Hydrogen bond 3.5 Interactions above are those observed are on the dimerization surface (top) and those at conserved R118 and R119 residues in the BBL(bottom). These were determined from comparing wild type apo BirA (1BIA) to the holo (2EWN) structure. 46 Chapter 5 : Conclusions and Future Directions The results indicate a model for allosteric activation of the BirA monomer in which bio-5’-AMP binding and folding of the ABL induce an alpha helical extension of residues 143-146. This helical extension optimizes the dimerization surface for formation of intramolecular interactions and subsequent folding of the 193-197 loop. Proper folding and arrangement of these regions occur in activation of the molecule for dimerization. However, the mechanism by which the allosteric signal is communicated to residues 143-146 is not clear from this study. Because there are no other observable structural consequences in G142A BirA, investigation of this mechanism must proceed by either garnering further structural data of alanine mutations that perturb the alpha helical transition, or by non- structural methods such as molecular dynamics (MD) simulations that investigate the changes of protein dynamics upon BirA activation. Evidence from prior H/D exchange mass spectra41 indicate a global dampening of BirA dynamics upon allosteric activation, suggesting that investigation of the role of protein motions may be required for full understanding the molecular mechanism of BirA allostery. Analysis of correlated motions and residue-residue contacts42 using molecular dynamics simulations could provide insight and, more importantly, testable hypotheses to further investigate how the α-helical extension of the 143-146 residues is occurring. 47 Chapter 6 :Appendicies Appendix 1: Discrepancy between ∆HTAPS and ∆Hdisplacement of bio-5’- AMP binding The purpose of this appendix is to address a discrepancy that is observed when measuring the enthalpy of bio-5’-AMP binding using two different ITC titration methods. The enthalpies measured by the displacement technique are 1-3 kcal/mol more negative than those measured directly by TAPS titrations. The P143A variant has the largest deviation, 3.1±.1 kcal/mol. Displacement titrations for the remaining proteins also yielded more favorable enthalpies by 1-2.5 kcal/mol except for G142A BirA, for which a displacement titration could not be performed (Figure 6.1). For this variant, no difference was observed between the direct and TAPS measurement. For this reason and due to the relative simplicity of the direct measurements, investigation into possible sources of the discrepancy was focused on the displacement method. However, a likely source of the more negative enthalpy of binding from displacement titrations was not found. Nontheless, the accuracy of the Gibbs free energies measured by displacement titrations and the enthalpy of bio-5’-AMP binding measured by TAPS titrations are strongly supported by experimental evidence. Possible sources of experimental discrepancy To attempt to discern the source of the 1-3 kcal/mol deviation observed between TAPS and displacement-obtained enthalpies of binding, differences in both experimental protocol and reagents used were examined. The enthalpies determined for I280A and K283A BirA by the displacement method were first reproduced using 48 P 1 4 3 A D 1 9 7 A G 2 8 1 A E 1 4 0 A I 2 8 0 A T 1 9 5 A W T K 2 8 3 A G 1 9 6 A G 1 4 2 A * 0 1 2 3 BirA Variant   H  b i o - 5 ' - A M P ( k c a l / m o l ) Figure 6.1. Discrepancy between bio-5’-AMP binding enthalpy from TAPS and displacement ITC measurements. Values of ∆∆H °Calculated by ∆∆H°=∆H°TAPS-∆H°Disp for each BirA variant. Protein and ligand solutions for both methods were prepared in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, 2.5mM MgCl2.Errors reported are the propagated 68% confidence interval of the individual enthalpy measurements. 49 new reagents (Table 3.1Table 6.1). The deviation was reproducible using different reagents at a later date. Because of this, it is not believed that there is any systematic error due to the reagents used or the order in which experiments were performed. To further demonstrate the ligand was intact for these titrations, the bio-5’-AMP solution was analyzed by TLC (see materials and methods). The presence of contaminating biotin in the bio-5’-AMP reagent was a concern because of its more negative enthalpy of binding. However, the presence of only a single spot at the correct retention distance, 0.53, was indicative of no contaminating biotin in the solution. Additionally, systematic experimental deviations could stem from differences in the protocol between the two methods. Displacement titrations uniquely require the inclusion of saturating biotin and this presence of excess biotin represents a difference in the chemical environment of the bio-5’-AMP binding reaction. To determine if the excess biotin used affected the enthalpy of corepressor binding, displacement titrations were performed using both 3µM and 6µM biotin for I280A BirA. A ∆∆H° between methods of 1.4±0.1 and 1.6±0.1 kcal/mol, respectively, was still observed at both concentrations. These results indicate that the presence of excess biotin, which was present in each displacement titration, does not significantly affect the measured enthalpy of bio-5’-AMP binding. 50 Table 6.1. Reproduction of displacement titration measurements I280A BirA KD,bio-5’-AMP (M) ΔG°bio-5’-AMP (kcal/mol) ΔH°bio-5’-AMP (kcal/mol) -TΔSbio-5’-AMP (kcal/mol) n Original 9(±1)x10-11 -13.5±0.1 -14.5±0.1 1.0±0.1 1.16±0.01 Reproduction 1.0(±0.2)x10-10 -13.4±0.2 -14.4±0.1 1.0±0.2 0.97±0.01 K283A BirA Original 1.1(±0.2)x10-11 -13.4±0.1 -14.8±0.1 1.4±0.1 0.87±0.1 Reproduction 1.1(±0.4)x10-11 -13.3±0.2 -15.2±0.2 1.9±0.2 0.87±0.1 Measured at 20°C by displacement ITC titration in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, 2.5mM MgCl2 . Each parameter represents the results from non-linear least squares regression of a single ITC isotherm. The uncertainty provided is the 68% confidence interval provided by linear regression. 51 ∆G°bio-5’-AMP measured by biotin displacement agrees with those measured by direct titrations. Because of the presence of this discrepancy between the measured enthalpies of bio-5’-AMP, the accuracy of the parameters obtained from both methods were in question. Measurement of the Gibbs free energy and the enthalpy of bio-5’-AMP binding by ITC is inherently coupled, however it was possible to validate the free energy values obtained by direct comparison to the results of a direct titration. To perform this experiment, a variant was carefully chosen so that its c-value would be within the measureable range for both direct and biotin displacement titrations. The BirA variant W223A allowed for these measurements, providing c-values of ~8 and ~300 for the displacement and direct titrations respectively (Figure 6.2). These measurements yielded ∆G°bio-5’-AMP of -11.7±0.2 -11.2±0.2 indicating that the Gibbs free energies as measured by displacement titrations well-represent those that cannot be determined directly(Table 6.2). In additional support of the of the Gibbs free energy values obtained by this method, the wild type ∆G°bio-5’-AMP (Table 3.3) is within 0.5 kcal/mol of that calculated from the results of previous kinetic measurements, -13.6±0.2 kcal/mol43. For this reason it is believed that the free energies obtained by biotin displacement ITC titrations are valid and therefore were not excluded from this study. 52 Figure 6.2 Direct versus Displacement titration of W223A BirA. Data is illustrated for both direct binding(A) and displacement(B) ITC methods. Protein and ligand solutions were prepared in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, and 2.5mM MgCl2. The top panel is the ITC data trace following manual baseline correction. After integration and normalization to the moles of injectant, the binding isotherm, bottom panel, is obtained. Non-linear least squares regression of each isotherm to the single-site binding model provides the ligand binding parameters. Table 6.2 Comparison of direct and displacement bio-5’-AMP binding measurements. All measurements were performed in 10mM Tris (pH 7.5), 200mM KCl, 2.5mM MgCl2 at 20°C. Each value represents the average of at least two independent ITC titrations. Errors given are 68% confidence intervals propagated from the nonlinear least squares analysis. 53 ∆H°biotin measured directly agrees with that measured by TAPS titration. Because of the discrepancy between direct and displacement ITC titrations, it must be determined which method most accurately provides the enthalpy of bio-5’- AMP binding. Ultimately, biotin displacement is an indirect measurement of both the enthalpy and free energy of bio-5’-AMP binding. The parameters determined from this competitive assay have a lower signal-to-noise and rely on accurate measurement of the enthalpy and association constant of biotin, both of which have an associated error. Due to the comparative complexity of the displacement titrations, it is believed that the binding enthalpy obtained from TAPS measurements provides a more accurate value. To demonstrate this, the enthalpies of biotin binding, which have also been characterized by full-direct titrations in this study, were reproduced for three dimerization surface variants using the TAPS method (Figure 6.3). No deviation was observed for both K283A BirA and E140A BirA. P143A BirA was also examined in this way and a small deviation, ∆∆H° = 0.5±0.3 kcal/mol, was measured between TAPS and direct titrations (Figure 6.4, Table 6.3). However, this difference is much smaller than that measured between TAPS and displacement methods, 3.1±0.1kcal/mol. These results demonstrate reproducibility and consistency of the enthalpies determined from TAPS measurements compared to those determined from direct titrations. Therefore, the molar enthalpies of bio-5’-AMP binding determined from displacement titrations were not analyzed in this study and instead only those values obtained from TAPS titrations were considered. Ultimately, these enthalpies were used to calculate the entropic components (-T∆S) and heat capacity changes of bio-5’-AMP binding to the BirA variants. 54 Figure 6.3 Biotin TAPS titration of P143A BirA. Both protein and ligand solutions were prepared in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, 2.5mM MgCl2. The top panel is the ITC data trace following manual baseline correction. After integration and normalization to the moles of injectant, the binding isotherm, bottom panel, is obtained. The difference between the molar heat before and after protein saturation provides the enthalpy of ligand binding. 55 Figure 6.4 Difference of bio-5’-AMP binding enthalpy between direct and TAPS ITC measurements. Values of ∆∆H °Calculated by ∆∆H°=∆H°TAPS-∆H°Direct for each BirA dimerization surface variant. Protein and ligand solutions for both methods were prepared in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, 2.5mM MgCl2.Errors reported are the propagated 68% confidence interval of the individual enthalpy measurements. Table 6.3 Biotin binding enthalpy obtained from TAPS measurements. BirA Variant ΔH°biotin (kcal/mol) ∆ΔH°biotin (kcal/mol) P143A -20.2±0.4 0.5±0.4 E140A -20.1±0.5 0.0±0.5 K283A -20.6±0.4 0.2±0.5 Biotin TAPS titrations performed in 10mM Tris (pH=7.50 at 20°C), 200mM KCl, 2.5mM MgCl2 A large protein concentration, 5µM, was used to assure stoichiometric biotin binding. Enthalpies reported from the average of two TAPS titratrations containing five pre- saturation injections each. Errors reported are propagated 68% confidence intervals. ∆∆H°biotin calculated using the equation ∆∆H°biotin=ΔH°TAPS-ΔH°Direct where ΔH°Direct values are those listed in Table 3.2. 56 Appendix 2:Considerations in the Design of ITC experiments All ITC titrations in this study were designed with several experimental factors in mind which are detailed below. Particularly, careful consideration was given to the Wiseman ‘c’ value44 and the concentration and size of each ligand injection. Wiseman’s c-value for the design of ITC titrations The ‘c’ parameter, where c=KA∙[P]∙n, is a function of the association constant, KA, total protein concentration, [P], and binding stoichiometry, n, for a single-site binding interaction. Of these parameters, only the protein concentration is modifiable in the design of a simple binding ITC experiment. This choice is limited on the upper and lower bounds by the quantity of protein able to be purified and the minimum protein concentration required to achieve a measureable heat of interaction upon injection (ideally ~5µcal or more, but as low as ~1µcal can be analyzed with quality data). In this study, increasing the protein concentration was also limited by the presence of a coupled homodimerization reaction. Nonetheless, the c-value was maintained within the recommended range of about 5-500, because titration data with a c-value outside this range provide a binding curve with less information about the binding process The titration curve asymptotically approaches a step-function (Figure 6.5)., which contains no information regarding the association constant of binding, at large-c-values (stoichiometric binding). However, both the experimental stoichiometry and enthalpy of binding are very well-determined under these 57 Figure 6.5 Simulated influence of c-value on single-site ITC binding curve. Heat of injection from c=0.01 to c=1x104 were simulated using the parameters: ΔH=- 6kcal/mol, KD=2x10-8M, and n=1. The protein concentration was varied in each case to obtain differing c-values. 58 conditions, which is the basis of total association at partial saturation (TAPS) measurements. Values obtained from small c-value binding curves can be misleading, because they generally do not reach saturation over the course of titration. Therefore, these measurements provide only a small portion of the full binding curve from which an accurate stoichiometry and enthalpy of binding cannot be determined. The association constant of binding is theoretically accessible at a low c-value (as long as the heats of binding upon injection are measureable), but precautions must be taken to assure full protein saturation45. C-values utilized were approximately 100 and 100,000 for the direct biotin binding and bio-5’-AMP TAPS measurements, respectively. Although the c-value is a parameter derived from the single-site binding equation, it can also be used to describe the biotin displacement titrations in this study. This is because the competition titrations performed were all using saturating concentrations of the competing ligand, which results in a pseudo-single-site binding curve for which the apparent association constant is a function of the KA of both ligands and the concentration of competing ligand as shown below: 𝐾𝑎𝑝𝑝 = 𝐾𝐴,𝑏𝑖𝑜5′𝐴𝑀𝑃 1 + 𝐾𝐴,𝑏𝑖𝑜𝑡𝑖𝑛[𝑏𝑖𝑜𝑡𝑖𝑛] Therefore the c-value of a displacement titration is calculated using the apparent association constant. Decoupling of the actual binding constant from the c-value is what permits competitive binding assays to measure association constants significantly tighter than those accessible by only direct binding. A typical c-value obtained for these displacement titrations was ~250. 59 Ligand concentration and injection scheme Another important consideration in the design of ITC titrations is determining the quantity of ligand to be injected and the number of injections over which this will occur. The quantity of ligand injected should be large enough that the protein is ultimately saturated with ligand at the end of the titration. Unless the c-value of the titration is particularly low, a final ligand/protein ratio of 2.2-2.5 is acceptable for complete saturation. This translates to a ligand concentration about 10-fold larger in the injection syringe than the protein concentration present in the sample cell. Titration with a quantity of ligand much greater than this will than this will, because the protein is saturated after only several injections, only provide a fraction of the binding curve. Determination of the stoichiometry and enthalpy of binding will then be difficult (Figure 6.6). In the determination of an injection scheme, or what quantity and size of ligand injections to use, the use of a simulation program is highly recommended. To maximize the amount of data available for nonlinear-least squares regression, a full, symmetrical binding curve, with several injections before and after protein saturation, should be used (see Figures 3.1, 3.3 for examples). For this study, the titration parameters were optimized by simulating ITC titrations using a spreadsheet provided by T. Biswas and O.V. Tsodikov29 (Figure 6.7). Injection volumes were chosen so that multiple injections (3-4) were present on the saturation and pre-saturation binding plateaus. 60 Figure 6.6 Example ITC titration using excess ligand. Too much ligand was used, resulting in early saturation of the protein in the course of this direct biotin titration. -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2-0.35 -0.30 -0.25 -0.20 -0.15 -0.10 -0.05 0.00 0.05 -10 0 10 20 30 40 50 60 70 80 90100110120130140150 Time (min) µ c a l / s e c Molar Ratio k c a l / m o l e o f i n j e c t a n t 61 Figure 6.7 Simulation of a direct biotin binding titration. This Microsoft-Excel-based tool allows simulation of single-site and pseudo single-site binding titrations. The user provides the experimental parameters (green, yellow, or grey cells) and the heat of injection is calculated for an indicated number of injections u sing the single-site binding model. 62 Appendix 3: Expression and purification of BirA variants BirA alanine variants used in this study were expressed and purified as previously described18,20, except for variants E140A and K283A which were prepared as follows. Approximately 50ng of pBtac2 BirA-(His6) plasmid25 DNA containing the appropriate variant coding sequence was transformed at 1500V into 50μL of JM109 electrocompetent cells. Immediately after electroporation (BTX-ECM399), cells were suspended in 1mL of cold SOC media and incubated at 37°C for an hour. Both 5μL and 20μL of transformed cells were spread evenly onto separate LB/AMP agar plates and subsequently grown at 37°C overnight. The pBtac2 BirA-(His6) plasmid contains the β-lactamase gene that confers ampicillin resistance. Inclusion of ampicillin (100μg/mL) in both LB agar and the liquid media allowed for selection of only transformed colonies. Following transformation, 2-3 well-separated colonies were inoculated into 5mL of LB/AMP and shaken overnight at 37°C. Prior to their usage for the large- scale preparation of BirA, the induction of each cell line was assessed by a small- scale induction trial. To perform this trial, half of each overnight culture (2.5mL) was inoculated into 20mL of sterile LB/AMP. The remaining 2.5mL of culture was stored at 4°C until needed. After inoculation, the 20mL culture was incubated at 30°C until the OD600 reached 0.6. At this point, 0.5mL aliquots of each culture were removed as a pre-induction sample. These samples were pelleted by centrifugation for 5 minutes at 10,000rpm and stored at -20°C for subsequent SDS-PAGE analysis. Because the pBtac2 BirA-(His6) plasmid is transcriptionally controlled by the tac promoter, overexpression of BirA variants was induced by addition of IPTG to 1mM final 63 concentration. After 4 hours of 200 rpm shaking at 30°C, the OD600 of each culture was recorded and a second 0.5mL aliquot was removed from each induction trial for analysis. These aliquots were similarly pelleted as above, and all cell-pellets were decanted then dissolved in a volume, 𝑂𝐷600 12 mL, of Laemmeli sample buffer26 (LSB) normalized to the OD600 of the culture at the time it was removed. This process confers a constant protein concentration to all SDS-PAGE samples so they may be qualitatively compared. The culture that demonstrated the highest induction of BirA, and of no other proteins, was chosen for large-scale protein preparation. The remaining 2.5mL of the chosen cell line was inoculated into 300mL of LB/AMP and shaken (200rpm) overnight at 37°C as a starter culture for protein expression. The 300mL overnight culture was evenly divided into two 4L Erlenmeyer flasks, each containing 2L of sterile LB/AMP media. Cells were grown at the induction temperature, 30°C, with shaking at 175rpm until the OD600 reached approximately 0.8. At this point, a 0.5mL aliquot was removed and pelleted as above for SDS-PAGE analysis. Overexpression was induced by adding IPTG to a final concentration of 1mM. After starting induction, the cultures were shaken for an additional 4 hours in identical conditions. Once induction was complete, the cell solutions were cooled on ice to 4°C. Care was taken to keep the cell solutions, cell pellet, and cell lysate at this temperature throughout the preparation steps to preserve BirA integrity. A 1L centrifugation bottle (Hitachi) was weighed prior to pelleting the cells, so that total cell mass could later be determined. The cell suspension was pelleted by centrifugation 2L at a time at 4500 rpm for 20 minutes at 4°C. This centrifugation (Sorvall RC 5B Plus) step was performed by dividing the cell 64 suspension evenly (about 500mL x4 bottles) and carefully balancing (±0.5g) the suspension between bottles placed oppositely into the F8S-4x1000y rotor. The cell pellets were then washed by suspension in excess lysis buffer (50mM NaH2PO4, 300mM NaCl, 10mM Imidazole, 5% glycerol, 100μM DTT, pH 8), and combined into a single pre-weighed centrifuge bottle. After a second centrifugation at 4500rpm for 20 minutes, the lysis buffer was carefully decanted and the mass of the remaining cell pellet weighed on an analytical balance. Cells were gently suspended by Pasteur or transfer pipet into 5mL of lysis buffer per gram of cell pellet. For lysis, the cells were transferred to the smallest-allowable, pre-chilled metal lysis cup. The sample was sonicated (Branson Sonifier 450) for 2-minute intervals in lysis buffer at sonifier output and duty cycle of 60% to minimize sample heating. The cell lysate was maintained below 10°C during lysis using a salt/ice bath. To inhibit serine protease activity, 0.01% of the total cell lysate volume of 100mM PMSF was introduced prior to each sonication interval. Lysis was considered complete when the OD600 of a cell lysate sample (diluted 1:100 in lysis buffer) was <10% of the initial value. Centrifugation at 5.4 x g for 30min at 4°C was performed to remove cell debris. Nucleic acids were precipitated from the decanted supernatant by adding PEI (pH=7.5) to a final concentration of .1% (v/v) at 4°C for at least 15 minutes. Another centrifugation step at 5.4 x g for 30min at 4°C pelleted the resulting PEI precipitate. Both cell debris and PEI pellet were suspended into lysis buffer in a volume equal to the respective supernatant. Aliquots (0.5mL) of both precipitate suspensions and both supernatants were removed for SDS-PAGE analysis. Before continuing with protein purification, both overexpression of BirA and its 65 presence in the PEI supernatant was confirmed by a 12.5% SDS-PAGE gel. The sample from each expression and purification step was diluted 1:1 into LSB, mixed well and heated at 90°C prior to loading 10uL onto a 12.5% acrylamide SDS-PAGE gel, prepared according to the Laemmli discontinuous gel system26. Purification of BirA variants from the resulting crude lysate was performed by several chromatography steps using an ÄKTA Prime (GE) liquid chromatography system. All resins were packed in a 1.5cm diameter Bio-Rad econo-column. Between 4 and 5cm (about 7mL) of resin was packed in each 15cm column. Each resin used for purification, except the HA Ultrogel, was regenerated before purification using a resin-specific protocol (see resin regeneration). All resins were equilibrated with at least 10 column volumes of resin-specific starting buffer directly before use. Because the variants have a C-terminal His6-tag, a nickel affinity column (Ni-NTA agarose, Qiagen) was the first BirA purification step. After dialysis of the PEI supernatant into fresh lysis buffer (at least 2 L, twice), the protein solution (about 150mL) was loaded at 1mL/min onto 7mL of Ni-NTA resin. BirA was eluted in 5mL fractions using an imidazole step gradient starting from 20mM, and proceeding to 50mM, 100mM and finally 250mM. This gradient was performed by eluting with 0%, 13%, 34%, and 100%, respectively, of elution buffer in Ni-NTA starting buffer (50mM NaH2PO4, 300mM NaCl, 20mM imidazole, 100μM DTT, 5% glycerol, pH=8) at 0.5mL/min. Each elution step was begun only after the A280 was constant for at least 5 minutes. The E140A variant and K283 variant eluted at 50mM and 100mM imidazole, respectively. Following SDS-PAGE (12.5%) analysis of each fraction (1:1 in LSB), those that indicated significant purification of BirA as compared to a 5μL pre-column 66 sample were pooled and exchanged twice by dialysis against 2L (about 25:1 buffer:dialysate) of SP-sepharose starting buffer. The second purification step for both E140A and K283A was chromatography on SP-sepharose (GE-Healthcare) cation exchange resin. After equilibration of 7mL of resin with SP-sepharose starting buffer (50mM Tris pH 7.5 at 4oC, 50mM KCl, 100mM DTT, 5% glycerol), the protein solution was loaded at 0.9mL/min. The BirA variants were eluted using a linear KCl gradient from 0.05M-0.8M (0-100% elution buffer) over 110mL. BirA elution was observed at about 0.3M KCl (33% elution buffer) for both variants. Additional purification of the E140A BirA variant was performed using a Q- sepharose (GE-Healthcare) anion exchange resin using the same loading, elution, and regeneration steps as the SP resin, except that the resin was stored in only 20% ethanol. BirA is eluted at a low salt concentration from the Q-sepharose column, for some mutants this occurs in the wash step. Consequently, 2mL fractions were taken during loading and starting buffer wash in addition to the elution gradient. Following two 500mL (33:1) dialysis steps of the 15mL protein solution into HA starting buffer, further purification of the E140A BirA variant was achieved using a hydroxyapatite Ultrogel (Pall) mixed-mode resin. The protein solution was loaded at 0.5mL/min onto 5mL of HA resin equilibrated in starting buffer. The protein was eluted at 0.5mL/min using a linear 0.05M-0.5M phosphate gradient, or 0-100% HA elution buffer (50mM sodium phosphate pH=6.5, 100μM DTT, 5% glycerol) in HA starting buffer over 100mL. Each 3mL fraction containing pure E140A BirA, according to 12.5% SDS-PAGE gel analysis, was pooled and dialyzed in 1L (37:1) 67 of storage buffer (200mM KCl, 50mM Tris, 100μM DTT, 5% glycerol, pH=7.5 at 4°C). All purified variants were dialyzed twice in at least 1:30 of protein:storage buffer and filtered using a 0.22um PVDF syringe filter pre-rinsed in storage buffer. The final protein solutions were mixed carefully to achieve homogeneity and aliquoted into 1mL fractions for storage at -80°C. Coomassie brilliant blue stains of samples subjected to 12.5% SDS-PAGE indicated that all BirA variants were ≥95% pure. Stoichiometric bio-5’-AMP binding titrations, monitored by fluorescence spectroscopy, indicate ≥85% ligand binding activity. BirA concentration was determined by the average of three measurements of the absorbance at 280nm using the molar extinction coefficient, 47510 M-1cm-1, calculated from the protein sequence27. Each measurement of concentration was a different BirA dilution prepared gravimetrically in storage buffer to obtain a range of A280 between 0.2 and 0.8. Preparation and regeneration of FPLC chromatography resins After use, each resin was regenerated as described below. Additionally, all resins that had been previously used were regenerated prior to protein purification. To avoid cross-contamination between protein solutions, a single resin aliquot was used to prepare only one specific BirA variant. In contrast to the other resins, the hydroxyapatite (HA) Ultrogel (PALL) was only regenerated after use since it was newly obtained. All resins were cleaned at 4°C by gravity-flow of a series of reagents. Each reagent was allowed to reach the column bed before the next was introduced. 68 Regeneration and storage of the Ni-NTA resin following purification was performed with slight modification from the manufacturer’s recommendation (QIAGEN). First 1 column volume (c.v.) of Ni-NTA elution buffer (50mM sodium phosphate pH=8, 300mM NaCl, 250mM imidazole, 100μM DTT, 5% glycerol,) followed by 2 c.v. of Ni-NTA regeneration buffer (6M guanidinium hydrochloride, 0.2M acetic acid) were flowed through the resin. It was then equilibrated with 5 c.v. of MQH2O followed by 3 c.v. of 2% sodium dodecyl sulfate (SDS) in water. Subsequently, 1 c.v. of first 25%, then 50%, and finally 75% ethanol was used. The column was then washed with 5 c.v. of 95% ethanol, then again with 1 c.v. of 75%, 50%, and finally 25% ethanol. After a 1 c.v. rinse with MQH2O, the column was equilibrated in 5 c.v. of 100mM ethylenediaminetetraacetic acid (EDTA). This step removes chelated nickel (II), so the resin should lose its blue color. Following a 5 c.v. MQH2O wash to remove any EDTA present, the nickel is restored by rinsing the resin with 2 c.v. of 100mM NiSO4 in water. A 2 c.v. regeneration buffer rinse was used afterwards to remove excess nickel (II). If significant color does not remain, new resin is required. The resin is then equilibrated in 5 c.v. lysis buffer (50mM NaH2PO4, 300mM NaCl, 10mM Imidazole, 5% glycerol, 100μM dithiothreitol (DTT), pH 8), at which point it was used for protein purification. For storage after regeneration, the resin was equilibrated in 5 c.v. of 30% ethanol at 4°C. Regeneration of the SP-sepharose column was performed by gravity-flow of 2 c.v. of 2M NaCl, followed by 10 c.v. of 1M NaOH, and finally 8 c.v. of 70% ethanol. The regenerated resin was equilibrated with 5 c.v. of SP-sepharose starting buffer (50mM Tris, 50mM KCl, 100μM DTT, 5% glycerol, pH=7.5 at 4°C) for protein 69 purification, or in 20% ethanol, 0.2M sodium acetate for storage at 4°C. The Q- sepharose column was regenerated and stored identically, except that it was stored without sodium acetate. The HA resin was regenerated by gravity flow of 4 c.v. of 500mM sodium phosphate pH=6.8, followed by 4 c.v. of 0.1M NaOH, and another 4 c.v. of 500mM sodium phosphate pH=6.8. Prior to storage, the resin was equilibrated in 5 c.v. of HA starting buffer (50mM sodium phosphate pH=6.5, 100μM DTT, 5% glycerol) containing 0.04% sodium azide to inhibit bacterial growth. 70 Bibliography 1. Nussinov, R. & Tsai, C.-J. Allostery in Disease and in Drug Discovery. Cell 153, 293–305 (2013). 2. Johnson, L. N. Glycogen phosphorylase: control by phosphorylation and allosteric effectors. FASEB J. 6, 2274–2282 (1992). 3. Macol, C. P., Tsuruta, H., Stec, B. & Kantrowitz, E. R. Direct structural evidence for a concerted allosteric transition in Escherichia coli aspartate transcarbamoylase. Nat. Struct. Mol. Biol. 8, 423–426 (2001). 4. Daly, T. J. & Matthews, K. S. Allosteric regulation of inducer and operator binding to the lactose repressor. Biochemistry 25, 5479–5484 (1986). 5. Ferreon, A. C. M., Ferreon, J. C., Wright, P. E. & Deniz, A. A. Modulation of allostery by protein intrinsic disorder. Nature 498, 390–394 (2013). 6. Monod, J., Wyman, J. & Changeux, J.-P. On the nature of allosteric transitions: A plausible model. J. Mol. Biol. 12, 88–118 (1965). 7. Koshland, D. E., Némethy, G. & Filmer, D. Comparison of Experimental Binding Data and Theoretical Models in Proteins Containing Subunits*. Biochemisty 5, 365–385 (1966). 8. Perutz, M. F. Stereochemistry of Cooperative Effects in Haemoglobin: Haem– Haem Interaction and the Problem of Allostery. Nature 228, 726–734 (1970). 9. Datta, D., Scheer, J. M., Romanowski, M. J. & Wells, J. A. An Allosteric Circuit in Caspase-1. J. Mol. Biol. 381, 1157–1167 (2008). 71 10. Zhang, J. & Matthews, C. R. Ligand Binding Is the Principal Determinant of Stability for the p21H-ras Protein. Biochemistry 37, 14881–14890 (1998). 11. Uversky, V. N. A Protein-Chameleon: Conformational Plasticity of α- Synuclein, a Disordered Protein Involved in Neurodegenerative Disorders. J. Biomol. Struct. Dyn. 21, 211–234 (2003). 12. Schmitz, M. L. et al. Structural and functional analysis of the NF-kappa B p65 C terminus. An acidic and modular transactivation domain with the potential to adopt an alpha-helical conformation. J. Biol. Chem. 269, 25613–25620 (1994). 13. Oldfield, C. J. et al. Flexible nets: disorder and induced fit in the associations of p53 and 14-3-3 with their partners. BMC Genomics 9, S1 (2008). 14. Ward, J. J., Sodhi, J. S., McGuffin, L. J., Buxton, B. F. & Jones, D. T. Prediction and Functional Analysis of Native Disorder in Proteins from the Three Kingdoms of Life. J. Mol. Biol. 337, 635–645 (2004). 15. Zhao, H. & Beckett, D. Kinetic Partitioning Between Alternative Protein– Protein Interactions Controls a Transcriptional Switch. J. Mol. Biol. 380, 223–236 (2008). 16. Streaker, E. D., Gupta, A. & Beckett, D. The Biotin Repressor:  Thermodynamic Coupling of Corepressor Binding, Protein Assembly, and Sequence- Specific DNA Binding†. Biochemistry 41, 14263–14271 (2002). 17. Streaker, E. D. & Beckett, D. Coupling of Protein Assembly and DNA Binding: Biotin Repressor Dimerization Precedes Biotin Operator Binding. J. Mol. Biol. 325, 937–948 (2003). 72 18. Naganathan, S. & Beckett, D. Nucleation of an Allosteric Response via Ligand-induced Loop Folding. J. Mol. Biol. 373, 96–111 (2007). 19. Eginton, C., Naganathan, S. & Beckett, D. Sequence–function relationships in folding upon binding. Protein Sci. 24, 200–211 (2015). 20. Adikaram, P. R. & Beckett, D. Functional Versatility of a Single Protein Surface in Two Protein:Protein Interactions. J. Mol. Biol. 419, 223–233 (2012). 21. Bains, G. & Freire, E. Calorimetric determination of cooperative interactions in high affinity binding processes. Anal. Biochem. 192, 203–206 (1991). 22. The Merck Index. (Merck & Co., Inc, 1960). 23. Abbott, J. & Beckett, D. Cooperative binding of the Escherichia coli repressor of biotin biosynthesis to the biotin operator sequence. Biochemistry 32, 9649–9656 (1993). 24. Lane, M. D., Rominger, K. L., Young, D. L. & Lynen, F. The Enzymatic Synthesis of Holotranscarboxylase from Apotranscarboxylase and (+)-Biotin II. Investigation of the Reaction Mechanism. J. Biol. Chem. 239, 2865–2871 (1964). 25. Kwon, K., Streaker, E. D., Ruparelia, S. & Beckett, D. Multiple disordered loops function in corepressor-induced dimerization of the biotin repressor. J. Mol. Biol. 304, 821–833 (2000). 26. Laemmli, U. K. Cleavage of Structural Proteins during the Assembly of the Head of Bacteriophage T4. Nature 227, 680–685 (1970). 27. Gill, S. C. & von Hippel, P. H. Calculation of protein extinction coefficients from amino acid sequence data. Anal. Biochem. 182, 319–326 (1989). 28. VP-ITC MicroCalorimeter User’s Manual. (MicroCal, LLC). 73 29. Biswas, T. & Tsodikov, O. V. An easy-to-use tool for planning and modeling a calorimetric titration. Anal. Biochem. 406, 91–93 (2010). 30. Brown, P. H. & Beckett, D. Use of Binding Enthalpy To Drive an Allosteric Transition. Biochemistry 44, 3112–3121 (2005). 31. Zhao, H., Streaker, E., Pan, W. & Beckett, D. Protein−Protein Interactions Dominate the Assembly Thermodynamics of a Transcription Repression Complex†. Biochemistry 46, 13667–13676 (2007). 32. Eginton, C., Cressman, W. J., Bachas, S., Wade, H. & Beckett, D. Allosteric Coupling via Distant Disorder-to-Order Transitions. J. Mol. Biol. 427, 1695–1704 (2015). 33. Sigurskjold, B. W. Exact Analysis of Competition Ligand Binding by Displacement Isothermal Titration Calorimetry. Anal. Biochem. 277, 260–266 (2000). 34. Eginton, C. & Beckett, D. A Large Solvent Isotope Effect on Protein Association Thermodynamics. Biochemistry 52, 6595–6600 (2013). 35. Breiten, B. et al. Water Networks Contribute to Enthalpy/Entropy Compensation in Protein–Ligand Binding. J. Am. Chem. Soc. 135, 15579–15584 (2013). 36. Ferrante, A. & Gorski, J. Enthalpy–Entropy Compensation and Cooperativity as Thermodynamic Epiphenomena of Structural Flexibility in Ligand–Receptor Interactions. J. Mol. Biol. 417, 454–467 (2012). 37. Loladze, V. V., Ermolenko, D. N. & Makhatadze, G. I. Heat capacity changes upon burial of polar and nonpolar groups in proteins. Protein Sci. Publ. Protein Soc. 10, 1343–1352 (2001). 74 38. Jung, H.-I., Bowden, S. J., Cooper, A. & Perham, R. N. Thermodynamic analysis of the binding of component enzymes in the assembly of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus. Protein Sci. 11, 1091–1100 (2002). 39. Krogsgaard, M. et al. Evidence that Structural Rearrangements and/or Flexibility during TCR Binding Can Contribute to T Cell Activation. Mol. Cell 12, 1367–1378 (2003). 40. Cooper, A., Johnson, C. M., Lakey, J. H. & Nöllmann, M. Heat does not come in different colours: entropy–enthalpy compensation, free energy windows, quantum confinement, pressure perturbation calorimetry, solvation and the multiple causes of heat capacity effects in biomolecular interactions. Biophys. Chem. 93, 215–230 (2001). 41. Laine, O., Streaker, E. D., Nabavi, M., Fenselau, C. C. & Beckett, D. Allosteric signaling in the biotin repressor occurs via local folding coupled to global dampening of protein dynamics. J. Mol. Biol. 381, 89–101 (2008). 42. Johnson, Q. R., Lindsay, R. J., Nellas, R. B., Fernandez, E. J. & Shen, T. Mapping Allostery through Computational Glycine Scanning and Correlation Analysis of Residue–Residue Contacts. Biochemistry 54, 1534–1541 (2015). 43. Xu, Y. & Beckett, D. Kinetics of Biotinyl-5’-adenylate Synthesis Catalyzed by the Escherichia coli Repressor of Biotin Biosynthesis and the Stability of the Enzyme-Product Complex. Biochemistry 33, 7354–7360 (1994). 75 44. Wiseman, T., Williston, S., Brandts, J. F. & Lin, L.-N. Rapid measurement of binding constants and heats of binding using a new titration calorimeter. Anal. Biochem. 179, 131–137 (1989). 45. Turnbull, W. B. & Daranas, A. H. On the Value of c:  Can Low Affinity Systems Be Studied by Isothermal Titration Calorimetry? J. Am. Chem. Soc. 125, 14859–14866 (2003).