The transition state structure for binding between TAZ1 of CBP and the disordered Hif-1α CAD

Intrinsically disordered proteins (IDPs) are common in eukaryotes. However, relatively few experimental studies have addressed the nature of the rate-limiting transition state for the coupled binding and folding reactions involving IDPs. By using site-directed mutagenesis in combination with kinetics measurements we have here characterized the transition state for binding between the globular TAZ1 domain of CREB binding protein and the intrinsically disordered C-terminal activation domain of Hif-1α (Hif-1α CAD). A total of 17 Hif-1α CAD point-mutations were generated and a Φ-value binding analysis was carried out. We found that native hydrophobic binding interactions are not formed at the transition state. We also investigated the effect the biologically important Hif-1α CAD Asn-803 hydroxylation has on the binding kinetics, and found that the whole destabilization effect due the hydroxylation is within the dissociation rate constant. Thus, the rate-limiting transition state is “disordered-like”, with native hydrophobic binding contacts being formed cooperatively after the rate-limiting barrier, which is clearly shown by linear free energy relationships. The same behavior was observed in a previously characterized TAZ1/IDP interaction, which may suggest common features for the rate-limiting transition state for TAZ1/IDP interactions.


Results
The aim of the present study was to investigate the structure of the rate-limiting transition state for binding between the globular TAZ1 domain of CBP and the disordered Hif-1α CAD in terms of native binding interactions. The CD spectrum of Hif-1α CAD displays the typical characteristics for a disordered protein (Fig. 1B), which agrees well with the appearance of the 15 N-heteronuclear single quantum coherence NMR spectrum, which displayed a low degree of peak dispersion 8 . TAZ1 is a well-defined globular protein domain as previously described 13 . We have here generated 17 Hif-1α CAD point-mutant variants, and they are well distributed over the whole protein sequence. These hydrophobic deletion mutations were made at the binding interface. In addition, the biologically important Asn-803 hydroxylation was generated in this work in order to analyze its effect on the binding kinetics.
Binding kinetics of TAZ1/Hif-1α CAD. The association kinetics were measured by the stopped-flow technique. The concentration of Hif-1α CAD was varied and the fluorescence change of Trp-418 in TAZ1 was monitored. The association kinetics contained two phases (Fig. 2); a fast phase for which the observed rate constant, k obs , increased linearly with the concentration of Hif-1α CAD, and a slow phase for which the k obs value was not dependent on the concentration, centering at around 3-4 s −1 . The apparent association rate constant, k on app , was obtained from the linear dependence of Hif-1α CAD concentration of k obs for the fast phase, and was found to be equal to 8.8 × 10 6 M −1 s −1 . The apparent dissociation rate constant, k off app , was obtained from displacement experiments and determined to be 0.030 s −1 , resulting in a K d = k off app /k on app = 3.4 nM. Biphasic kinetics was also observed in a previous study where the binding between TAZ1 and another IDP, namely TAD-STAT2, was characterized by protein engineering and stopped-flow kinetics 12 . The k obs value for the slow phase for TAZ1/TAD-STAT2 is the same as the one obtained for TAZ1/Hif-1α CAD, and as was also the case for TAZ1/ TAD-STAT2, this value is very similar among the Hif-1α CAD mutant variants (Supplementary Table S1), which implies that this phase could be a result of a conformational change in TAZ1 that occurs before binding. Impact of mutations in Hif-1α CAD to the binding of TAZ1. The binding kinetics for the Hif-1α CAD mutant variants were measured using stopped-flow fluorimetry ( Fig. 3 and Supplementary Fig. S1). Point mutations that were made at the N-terminal part of Hif-1α CAD (region 778-787), did not affect the binding affinity appreciably, whereas a significant response was observed for point mutations that were made at region 792-825, with binding affinities being reduced by as much as 300-fold (Table 1). The mutations that destabilized the complex the most (>2 kcal mol −1 ) were L792A, L795A, L813A, and L818A. The Asn-803 hydroxylation lowered the binding affinity by 50-fold. The L818 position is also part of an LLXXL recognition motif in helix 3, where X may represent any amino acid. These motifs are frequently observed in protein-protein associations that play a role in transcriptional regulation 14,15 . Φ-Value binding analysis and structure of the rate-limiting transition state. A Φ-value binding analysis 16 was carried out in order to determine to what extent native binding contacts are present at the rate-limiting transition state for binding between TAZ1 and Hif-1α CAD. Φ-values are calculated as the ratio between the free energy change for the rate-limiting transition state for binding, ΔΔG TS , and at equilibrium, ΔΔG Eq :

TS Eq
The K d values that were used in this analysis were determined as k off /k on . We compared the K d obtained from the stopped-flow method with that determined by ITC for the Hif-1α CAD L813A mutation ( Supplementary  Fig. S2), which lowered the binding affinity to such an extent that an accurate determination of K d by ITC was possible, and the results show that they are in excellent agreement with each other (K d by the stopped-flow technique: 0.137 ± 0.009 μM, and K d by ITC: 0.144 ± 0.012 μM). A Φ-binding value of zero means that the residue that has been mutated does not form native contacts at the transition state for binding, while a Φ-binding value of 1 means that the change in K d is the same as the change in k on , thus the native interaction is completely present at the rate-limiting transition state for binding. Thus, the Φ-binding value is an index with values between 0 and 1 being interpreted as partially formed native binding contacts at the rate-limiting transition state for binding. A Φ-value was determined if the magnitude of ΔΔG Eq was larger or the same as 0.3 kcal mol −1 . Two mutations (I806V and V825A) had a magnitude value of ΔΔG Eq that was equal to 0.3 kcal mol −1 , whereas the rest of the calculated Φ-values had an associated ΔΔG Eq that was larger than 0.8 kcal mol −1 (Table 1). Φ-values for the I806V and V825A mutations were calculated due to the high precision in the obtained rate constants.
All of the determined Φ-binding values are low (≤0.34) which demonstrates that native hydrophobic binding interactions have not been created yet at the rate-limiting transition state for binding between TAZ1 and Hif-1α CAD (Table 1 and Fig. 4), suggesting that the rate-limiting transition state is disordered-like with low presence of native contacts in the binding interface. The whole effect of the hydroxylation of Asn-803 in Hif-1α CAD is within the dissociation rate constant (Table 1), which in turn results in a Φ-value that is low. This demonstrates that while the association kinetics for this hydroxylation remains about the same as the wild-type, the introduction of an OH-group significantly disturbs the interaction between TAZ1 and Hif-1α CAD-OH, resulting in a large increase in k off . Linear free energy relationships 17,18 (LFER) (Fig. 5) clearly demonstrate that the destabilizing effect upon mutation is dominated by changes in the dissociation rate constant. As shown in Fig. 5, the linearity and a slope that is close to 1 for log k off app vs log K d , means that native non-polar binding interactions are cooperatively created after the rate-liming barrier, which was also observed to be the case for the binding of TAZ1 The concentration dependence of k obs for the slow (inset) and fast phase. The k obs for the fast phase is linearly dependent on Hif-1α CAD concentration while the k obs for the slow phase is concentration independent, with a value of 3-4 s −1 for all Hif-1α CAD variants. k on app was obtained by fitting the concentration dependence of k obs for the fast phase to the general equation for association of two molecules 28 .
to TAD-STAT2 12 , suggesting that TAZ1/IDP interactions may share similar transition state features. However, further studies on other TAZ1/IDP binding reactions need to be performed to see if this is the case.

Discussion
The use of protein engineering in combination with kinetics measurements have during the last couple of years started to give us a better understanding on the binding mechanisms and short-lived intermediates involving binding of intrinsically disordered proteins. However, it remains unclear regarding the generality of certain mechanisms, for instance if the binding proceeds through a rate-limiting barrier that resembles the ground state structure in terms of native contact formation or not 18,19 . Such studies are also useful for benchmarking MD simulations. The analysis of the binding kinetics for TAZ1/Hif-1α CAD mutant variants reveal that native hydrophobic binding contact formations at the rate-limiting transition state are absent, since all of the calculated Φ-values are low. This was also shown to be the case in a previous study where a similar analysis was performed for the interaction between TAZ1 and the disordered TAD-STAT2 12 . Even though the topology of both Hif-1α CAD 8 and TAD-STAT2 20 , which wrap around TAZ1, is rather different compared to the single α-helix conformation that represents the bound IDP in other binding reactions 17,21,22 , the end-result of these studies are, with the exception of a few studies 22 , quite similar, i.e. the rate-limiting transition state for binding is rather unstructured, in agreement with MD simulations studies 23,24 . For instance, low Φ-binding values were consistently obtained for the interaction between PDZ domain and peptide targets 17 , and in S-protein/S-peptide interactions 25 . Similar results were also reported by Rogers et al. 21 , which showed that the transition state in the interaction between PUMA (an IDP that adopts a single α-helix when bound) and the globular MCL-1 is rather unstructured, as was also the case for the interaction between the molten-globule-like NCBD and the disordered activation domain of ACTR 26 . Although, there are a few examples where high Φ-binding values have been determined for IDP binding interactions 22 , generally it seems that the rate-limiting transition state is lacking to a large extent native hydrophobic binding contacts, forming late on the reaction pathway. Furthermore, the linearity in the LFER plot with a slope of 0.93 for log k off app vs log K d clearly show that native hydrophobic interactions in TAZ1/Hif-1α CAD are created cooperatively after crossing the rate-limiting transition state, which was also observed for TAZ1/TAD-STAT2 12  and other systems as well 17,21,26 , suggesting that these features might be common in IDP binding reactions. Studies on other IDP associations need however to be performed in order to find out if there is a general trend.
It has usually been observed that certain positions contribute more to the binding energetics than other interacting residues in protein-ligand associations. A common way of defining these so-called hot-spots, is that a mutation to Ala reduces the binding affinity by more than 2 kcal mol −1 27 . An intriguing result from the present study is that the largest mutational destabilizations of the TAZ1/Hif-1α CAD interaction are only observed to take place for substitutions at the 792-825 region in Hif-1α CAD, thus none of the mutations at the N-terminal part 778-787 resulted in significant changes (Table 1). Positions L792, L795, L813, and L818 were identified as hot-spots, which also includes a position (L818) within an LLXXL recognition motif, which are known to be important in many protein-protein associations 14 . Thus, as reported in Table 1, the very low presence of hydrophobic native binding interactions at the transition state involve all of these hot-spots as well.
In conclusion, we have shown that the rate-limiting transition state structure does not contain native hydrophobic binding interactions, but form cooperatively after crossing the rate-limiting barrier for binding between TAZ1 and Hif-1α CAD.

Materials and Methods
Protein expression and purification. Human TAZ1 (residues 340−439) was expressed and purified as described previously 12 . Hif-1α CAD (residues 776-826) was cloned into a pRSET vector, directly downstream of a His 6 -lipoyl domain−thrombin site tag. FIH-1 was inserted into the pET28a(+) vector, with a N-terminal His 6thrombin tag preceding the FIH-1 sequence. Variants of the Hif-1α CAD were obtained using the quick exchange approach with pfu Ultra and DpnI digestion. Plasmids were propagated using XL1 blue cells (Agilent) and the sequences were verified by DNA sequencing (Eurofins). Hif-1α CAD and FIH-1 were transformed into E. coli BL21 pLysS cells (Thermo Fisher Scientific). The cells were cultivated in 2xTY medium at 37 °C to an OD 600 of 0.6-0.7, followed by induction with 0.8 mM isopropyl β-D thiogalactopyranoside. The cells were grown further for 16-18 hours at 18 °C for Hif-1α CAD and 15 °C for FIH-1. Cell lysis was performed by sonication, followed by centrifugation in order to remove the cell debris, after which the supernatant was passed through a 0.2 μm filter and subsequently added to a prequilibrated nickel affinity column. The column was washed with buffer (20 mM Tris-HCl (pH = 8.0), 1 mM TCEP, 500 mM NaCl, and 5 mM imidazole) after which protein elution was done with 20 mM Tris (pH = 7.9), 1 mM TCEP, 250 mM imidazole, 500 mM NaCl. The protein solution was dialyzed overnight at 4 °C against 20 mM Tris (pH = 8.0), 1 mM TCEP, and 100 mM NaCl. Thrombin was added to the protein solution, and left to cleave for 6-8 hours at room temperature. The cleaved protein was loaded onto the sepharose nickel column to separate the fusion-tag and other impurities from the cleaved protein, which was present in the flow through. FIH-1 was dialyzed into a buffer of 20 mM Tris (pH = 7.5), 150 mM NaCl and 0.5 mM DTT overnight and then further used for hydroxylation of Hif-1α CAD. As a final purification step for Hif-1α CAD, reversed phase (RP) chromatography was used with a Resource RP column (GE Healthcare) where water/acetonitrile solvents with 0.1% (v/v) trifluoroacetic acid were used. The identity of all Hif-1α CAD variants were verified with matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry.

FIH-1 mediated hydroxylation.
For the hydroxylation of Hif-1α CAD, the purification of Hif-1α CAD was halted after the second sepharose nickel column step and was instead dialyzed into a buffer containing 20 mM Tris (pH = 7.2), 1 mM DTT and 150 mM NaCl. Thereafter, FIH-1 was added in a 1:10 (FIH-1/Hif-1α CAD) molar ratio containing 20 mM Tris (pH = 7.4), 150 mM NaCl, 1 mM DTT, 4 mM ascorbic acid, 1.5 mM FeSO 4 , 1 mM MgCl 2 , 40 µM 2-oxoglutarate and 1 mM PMSF. The hydroxylation was conducted at 37 °C for four hours. Thereafter the purification of Hif-1α CAD was continued as described above with reversed phase chromatography. The hydroxylation was confirmed with MALDI-TOF mass spectrometry. stopped-flow spectrometer (Applied Photophysics, Leatherhead, U.K.) at 293 K. TAZ1 contains a single tryptophan which was utilized as the fluorescence probe. Excitation was done at 295 nm and a 320 nm long pass-filter was used when monitoring the fluorescence change. Samples were prepared in 20 mM HEPES (pH = 6.9), 1 mM TCEP, and 190 mM NaCl. For measurement of the association rate constants the TAZ1 concentration was kept constant at 0.5-0.9 µM and the concentration of Hif-1α CAD and its variants were varied between 1 µM and 15 µM. The binding traces were biphasic for both the wild type and all of the variants of Hif-1α CAD, with a slow phase and a fast phase. The observed rate constant (k obs ) for the slow phase was concentration independent and similar for both the Hif-1α CAD and its variants (Supplementary Table S1). The k obs for the fast phase showed a linear Hif-1α CAD concentration dependence in the range of concentrations at which data were collected. The concentrations used here are mostly at pseudo-first order conditions. By using these k obs values, the apparent association rate constant (k on app ) was determined by fitting the data to the general equation for the reversible association of two molecules 28 . The dissociation rate constant (k off app ) was determined by displacement experiments in which a TAZ1 variant with the tryptophan being replaced by a tyrosine (TAZ1 W418Y ) was utilized 12 . A complex solution of TAZ1/Hif-1α CAD variant (0.75 µM/0.5 µM) was mixed with an excess of TAZ1 W418Y (varied between 10 µM and 50 µM), in which TAZ1 W418Y displaces TAZ1, resulting in traces that were single-exponential. At excess concentrations of TAZ1 W418Y the k obs is equal to k off app . The dissociation constant, K d , was then determined as K d = k off app /k on app . Circular dichroism (CD) spectroscopy. CD spectrum in the far-uv region was recorded for Hif-1α CAD using a Chirascan spectrometer using a 1 mm cuvette and at 298 K. The protein concentration was 10 μM and in a buffer containing 5 mM HEPES (pH = 6.9), 1 mM TCEP, and 50 mM NaCl.
Isothermal titration calorimetry (ITC). ITC experiments were taken using an iTC200 (Malvern Instruments) calorimeter, at 293 K. The proteins were dialyzed against 20 mM HEPES (pH = 6.9), 1 mM TCEP, and 190 mM NaCl prior to ITC measurements. A 15 µM TAZ1 solution was loaded onto the calorimeter cell, while Hif-1α CAD (L813A) was in the syringe (149 µM). The titration series started with a 1.3 µL injection, followed by 19, 1.8 µL injections. Fitting of the binding isotherm was carried out using a one-to-one model.