Structural characterization of intrinsically disordered proteins (IDPs) is mandatory for deciphering their potential unique physical and biological properties. A large number of circular dichroism (CD) studies have demonstrated that a structural change takes place in IDPs with increasing temperature, which most likely reflects formation of transient α‒helices or loss of polyproline II (PPII) content. Using three IDPs, ACTR, NHE1, and Spd1, we show that the temperature-induced structural change is common among IDPs and is accompanied by a contraction of the conformational ensemble. This phenomenon was explored at residue resolution by multidimensional NMR spectroscopy. Intrinsic chemical shift referencing allowed us to identify regions of transiently formed helices and their temperature-dependent changes in helicity. All helical regions were found to lose rather than gain helical structures with increasing temperature, and accordingly these were not responsible for the change in the CD spectra. In contrast, the nonhelical regions exhibited a general temperature-dependent structural change that was independent of long-range interactions. The temperature-dependent CD spectroscopic signature of IDPs that has been amply documented can be rationalized to represent redistribution of the statistical coil involving a general loss of PPII conformations.
Intrinsically disordered proteins (IDPs) are abundant in higher eukaryotes and are involved in many highly regulated processes such as cell cycle regulation, intracellular signaling, translation, and transcription.1–3 Many IDPs are promiscuous and can adapt to their multiple, structurally diverse interaction partners in a process of coupled folding‒and‒binding.4, 5 The promiscuity is functionally relevant, as it allows them to function as hubs in protein interaction networks6, 7 and to regulate formation of higher order complexes.8, 9
For folded proteins, the biological function is closely linked to their structure. This suggests that the functions of IDPs can be understood by studying their transiently formed structures. IDPs do not have unique folds but consist of dynamic ensembles of interconverting, intermittent structures.10, 11 The high intramolecular flexibility inherent to IDPs means that distant elements are largely structurally decoupled. Structure formation, and hence function, is thus related to segmental parts of the sequence and dominated by local features. These are often binding sites for other proteins and have been termed molecular recognition elements or molecular recognition features.12 Despite the structural decoupling between distant regions of IDPs, even highly disordered proteins have been found not to be elongated but still to form specific tertiary and even quaternary interactions.13–15 The heterogeneous nature of IDPs means that their structure is best described by an ensemble of structures containing both long range and local interactions.
High-resolution NMR studies have recently been used in attempts to improve the description of the structural properties of disordered proteins. A particularly successful approach has been the analysis of residual dipolar couplings (RDCs) through back calculations of theoretical RDCs of large ensembles of structures. In this way, Bernado et al.16 showed that the experimental RDCs from a “random coil” like peptide chain could be reproduced by extracting amino acid‒type specific (φ,ψ)‒angle pairs from high-resolution crystal structures. This showed that even in a highly disordered state, the peptide backbone mainly populates the most favored regions of the Ramachandran plot; that is, the α‒helix region, the β‒strand region, and the polyproline II (PPII) helix region. The structure of a disordered peptide chain is thus far from random, and we will use the term statistical coil instead of “random coil” throughout the remainder of this text. Subsequent studies have shown that some IDPs required a higher proportion of helical- or PPII-like conformations to fit the experimental data,17, 18 which underscores that the free energy landscape for residual structure in IDPs is best evaluated on a case-to-case basis and with residue resolution.
The understanding of protein folding has benefited tremendously from studies in which the folding equilibrium has been perturbed by systematically varying the experimental conditions. Similarly, a number of studies have explored the energy landscapes of IDPs by changing the solvent conditions. The structure formation in IDPs has been studied as a function of ionic strength, denaturants, stabilizing agents, pH, crowding agents, solvent polarity, detergents, and temperature (Ref.19 and references therein). Temperature-induced structural changes have been observed for a large number of IDPs using circular dichroism (CD) spectroscopy. Most of these studies have identified structural changes upon heating that were interpreted mostly as formation of α‒helices and to a minor extent to a loss of PPII structure.20–26 The interpretation of the changes observed by CD spectroscopy is ambiguous. This is caused by the low resolution of this technique and the fact that structural changes in different segments may have spectroscopic contributions that cancel each other's signal. Specifically, folding of α‒helices and unfolding of PPII structures give rise to a similar change in the CD spectrum.
In this article, we have reexamined the temperature-induced structural changes in IDPs using three different proteins. From a combined analysis of data from CD spectroscopy, small-angle X‒ray scattering (SAXS), NMR chemical shift analysis, and peptide mimics, the bulk of the observed change in ellipticity with temperature is suggested to be due to a redistribution of the statistical coil ensemble, where PPII‒like conformations are lost with increasing temperature. The transiently formed α‒helices, however, lose helical structures at increased temperatures.
We have investigated temperature-induced structural changes in IDPs using three proteins with different functions and different structural contents. The analyzed proteins are the activation domain of the activator for thyroid hormone and retinoid receptors (ACTR),27 the cytosolic C‒terminal distal tail of the human sodium–proton exchanger 1 (hNHE1cdt),28 and the S‒phase delayed protein, Spd1, from yeast, a protein regulating ribonucleotide reductase activity.9 These three proteins are unrelated except for each of them being intrinsically disordered, monomeric proteins without disulfide bonds.
Macroscopic structural changes observed by CD spectroscopy
Initially, far‒UV CD spectra were recorded for all three proteins at physiological pH and temperature to confirm their unfolded characteristics (Fig. 1). All spectra had a negative minimum around 200 nm and a negative shoulder around 222 nm immediately suggesting highly unfolded proteins with a minor fraction of α‒helix. Next, the temperature dependence of the CD spectra was investigated from 5 to 95°C (Fig. 1). A significant gain in negative ellipticity around 222 nm and a loss of negative ellipticity around 200 nm was observed for all three proteins. These data suggested either a heating-induced folding of the peptide chain into α‒helical structures or an unfolding of extended PPII structures or a combination of both.
Aromatic side chains are known to contribute to the far‒UV region of the CD spectra of proteins.29, 30 To examine if aromatics contributed to the far‒UV CD spectra of the present IDPs, we recorded near‒UV CD spectra for hNHE1cdt and Spd1. ACTR was omitted as it does not have any aromatic groups. As expected, the near‒UV CD spectra did not exhibit detectable signals (data not shown), which showed that the aromatic groups do not have a fixed, asymmetric conformation. As the aromatic signals in the near‒UV regions were minuscule, the aromatic contributions to the far‒UV region are likely to be negligible.
SAXS measurements of the overall dimensions of the ensemble
To investigate the effect of temperature on the overall dimensions of the proteins, we carried out SAXS measurements at two different temperatures for each of the three proteins. The low-temperature data recorded on Spd1 were not of a suitable quality for further analysis, most likely because of the low solubility of this protein at low temperatures. ACTR and hNHE1cdt had scattering curves characteristic of disordered proteins as revealed by the Kratky plot [Fig. 2(A,B,D,E)]. To quantitatively analyze the underlying conformational features, we fitted an ensemble of structures to the scattering profile using a genetic algorithm.31 The average radius of gyration (Rg) was found to change from 26.3 Å at 5°C to 23.9 Å at 45°C for ACTR and from 37.5 Å at 5°C to 35.1 Å at 45°C for hNHE1cdt [Fig. 2(C,F)]. A similar change was observed for the maximal intramolecular distance (data not shown). These data thus showed that the IDP ensembles become more compact at higher temperatures.
Microscopic structural changes observed by chemical shifts
Chemical shifts are sensitive reporters of the dihedral angles of the peptide backbone and therefore of the content of secondary structure.32–34 To evaluate the structural changes that occur with increasing temperature, the response of the chemical shifts of the backbone nuclei (C′ and Cα) to heating was evaluated for ACTR, for which assignments of the NMR signals are available.35 To get suitable “statistical coil” reference chemical shifts, the NMR signals of ACTR were assigned in 6M urea at 5°C and at 45°C. The secondary chemical shifts were calculated as δ - δisc, where δisc is the intrinsic statistical coil shifts measured in urea.36 This referencing method produces secondary chemical shifts with less noise than the traditional method based on database chemical shifts [Figs. 3(A,B) and 4(A,B)]. The positive 13C secondary chemical shifts observed demonstrated the presence of significant amounts of transiently formed helices in the regions corresponding to the three helices formed in ACTR in complex with its ligand37 [Fig. 3(A,B)]. At 5°C the helical populations were approximately 18, 6, and 3% in helices 1, 2 and 3, respectively. The helical populations changed to ∼12, 4, and 3% at 45°C, which demonstrated a general loss of helicity with increasing temperatures. A large spike is observed in the secondary chemical shifts for H1053 when comparing the intrinsic secondary shifts at two temperatures [Fig. 4(C)]. This is the only histidine in the protein and most likely represents the effect of a small, temperature‒induced pH change.
Far‒UV CD spectra of peptides derived from the sequence of ACTR
To confirm our interpretation of the NMR data, we designed three peptides derived from the sequence of ACTR. Peptide 1 (R1022–D1031) was from a region where the secondary chemical shifts did not change with temperature (Fig. 3) and which does not have known helical propensity. Peptide 2 (G1041–H1053) corresponded to the partially formed helix 1 that was shown to lose α‒helix at increased temperature. Peptide 3 (D1068–G1080) contained helix 3, which has very little helical content. The intrinsic secondary structure content of these peptides was investigated by far‒UV CD spectroscopy. At low temperature, peptide 1 had a far‒UV CD spectrum that resembled that of a statistical coil [Fig. 5(A)]. When the temperature was increased, a broad negative band appeared centered at 218 nm, whereas the negative band at ˜200 nm lost intensity. This is consistent with the suggestion that this region loses PPII conformations in the statistical coil. The far‒UV CD spectrum of peptide 2 had a negative band at 222 nm at low temperature [Fig. 5(B)], which is consistent with the observation by NMR that this region contained significant amounts of transient helix (Fig. 3). With increased temperature, the intensity of the negative band at 222 nm decreased indicating loss of α‒helix content consistent with our chemical shift analysis. The magnitude of the decrease was smaller than expected from the chemical shift analysis, which most likely is caused by a change in the statistical coil fraction of the ensemble with an opposite spectroscopic signature. The CD spectra of peptide 3 showed a similar pattern to that of peptide 1 [Fig. 5(C)], except for a small negative band around 222 nm, which suggests a small percentage of transient helix. Taken together, these data confirmed the observations from NMR chemical shift analysis on ACTR and demonstrated that the observed effects are a property of the local peptide sequence and do not require long‒range interactions.
A characteristic temperature‒dependent, spectroscopic change has been observed in numerous IDPs.20–26 This suggests that many IDPs undergo a structural change with increasing temperature. Different models have been put forward to explain this observation; however, the data do not decisively determine which explanation is correct, mainly because of the lack of atomic resolution information.
Temperature changes have previously been reported to affect the size of both denatured proteins and IDPs.40–43 The SAXS data recorded on ACTR and hNHE1cdt show that an overall contraction of the ensemble of structures occurs with increasing temperature. The contraction could be caused by a change in either the transient secondary structure or the tertiary structure. It has been hypothesized that unfolded proteins contract because of increased tertiary interactions caused by a stronger hydrophobic effect.19 A recent study, however, showed that the contraction was independent of the hydrophobicity of the protein, and it is thus not likely to be due to increased hydrophobic interactions.40 The secondary structure content of the chain also influences the overall size of the molecule. Relative to the statistical coil, the PPII conformation is more extended, and the α‒helical conformation is more compact. Both loss of PPII conformations and gain of helicity are thus able to explain the contraction of the ensemble without requiring changes in tertiary interactions.
Structural studies by CD spectroscopy or SAXS do not reveal the nature of the temperature-induced structural transition with satisfactory detail. We thus undertook a high-resolution NMR study using ACTR as a model system. ACTR is a particularly well-suited model for this analysis as its resonances have been assigned previously35 and it is known to form three helices in complex with its binding partner, the nuclear coactivator domain (NCBD) of CREB-binding protein (CBP).37, 38 Only one of these helices was found to be partially formed in the free state.35 ACTR thus has regions with transiently formed helices, regions that are known to form helices but have no significant helix content in the unbound state, and regions that behave as a statistical coil. This allowed us to test a range of putative temperature-induced structural changes.
In this study, we identified regions of transiently formed α‒helices based on secondary chemical shifts. Helix 1 was identified previously as being partially folded35; however, our study also showed partially formed helices in helix 2 and 3. This difference is due to the different methods used for determining the statistical coil chemical shifts. Determination of an ideal set of statistical coil chemical shifts is impossible as a truly random peptide chain is unlikely to exist. Even under highly denaturing conditions, elements of local structure have been identified,44 and denaturants have been suggested to stabilize the PPII conformation.45 Several sets of random coil chemical shifts have been determined based on studies of small peptides under denaturing conditions,46, 39 including attempts to quantify the effects of the nearest neighbors.47 Chemical shift perturbations from more distant neighbors, however, contribute to the noise in the secondary chemical shifts. By using the unfolded state of the protein as the reference, the intrinsic structural preferences of the peptide chain beyond the effects of the nearest neighbors are taken into account. Consequently, this allowed for detection of very small populations (< 5%) of transiently formed helices in ACTR.
In this study, a urea-denatured state of the same protein was used to determine the intrinsic statistical coil chemical shifts. If the structural ensemble of the reference state changes with temperature, this could possibly bias our interpretation of the changes observed under native conditions. Comparison of the temperature-induced change in chemical shift in the urea-unfolded protein revealed a small but significant chemical shift change dispersed throughout the peptide chain. The sign of the chemical shift change is for a large part of the chain opposite for the C′ and Cα shifts, despite the fact that these two chemical shifts change in unison for all known secondary structure elements. This suggests that the temperature-induced chemical shift change in the urea-unfolded state primarily reflects an intrinsic effect of temperature on the chemical shifts rather than a change in the ensemble of backbone conformations. Still, the intrinsic temperature effect causes a baseline shift if the secondary chemical shifts are evaluated using a single set of reference chemical shifts. Accordingly, the Cα secondary chemical shifts displayed in Figure 4 suggest a larger degree of secondary structure than the C′ secondary chemical shifts. By using the urea-denatured state as the reference, it was possible to correct for this effect, and good agreement between the C′ and Cα at both temperatures was seen. This illustrates that it is critical that the reference chemical shifts are determined under similar solvent conditions to the protein under investigation.
It has been proposed that an increase in the α‒helix content can explain the change in ellipticity observed for IDPs by CD spectroscopy with increasing temperature.19, 26, 48 On the contrary, we observed a loss of structure in all helical elements. This is in accordance with studies of the temperature dependence of helical structure in peptides, which unfold with increasing temperature.49 A loss of helix content in transiently formed helices causes a change opposite to what we observe by CD spectroscopy. This implies that not only is the temperature-induced helix formation unlikely to cause the observed spectroscopic change, the contribution from unfolding of helices in fact eliminates part of the change in CD signal, and results in an underestimation of the extent of the structural change.
When temperature-induced formation of α‒helices has been ruled out, the CD spectroscopic signature suggests that the temperature-dependent change observed for IDPs is caused mainly by changes in the PPII content. The question remains, however, whether unfolding of discrete highly populated PPII helices occurs or whether the change takes place throughout the peptide chain. Except for the helical regions, the secondary chemical shifts are nearly zero, and they do not show a large focalized temperature-induced change. PPII helices have 13C chemical shifts that are very close to the statistical coil values and are thus difficult to detect by NMR chemical shift analysis.50 In addition to being found in numerous IDPs, the temperature-induced change in the CD spectra has also been observed in chemically denatured proteins40, 51–53 and in peptide studies.53, 54 The ubiquity of the spectroscopic change suggests that the underlying process is a property of the peptide backbone and not of specific sequence elements or a property solely associated with IDPs. This temperature effect observed by CD spectroscopy thus primarily reports on a redistribution of the statistical coil, which most likely consists of a depopulation of the PPII well in the Ramachandran space to the β‒strand well.51 It is interesting to note that this redistribution would be invisible in chemical shift analyses, as the reference state for the statistical coil chemical shifts undergoes the same redistribution.
Bioinformatic studies have shown that regions that contain transiently formed secondary structure in the unfolded state are likely to be binding sites for putative partners.55, 56 Therefore, a large number of theoretical55, 57 and experimental studies14, 18, 58–61 of IDPs has focused on identifying elements of partially formed secondary structure, particularly α‒helices. In experimental studies trying to identify partially formed helices, it is desirable to work under conditions where the helical elements have the highest population. Numerous studies have suggested that helices are formed as the IDPs are heated, which would suggest that NMR detection of residual helical regions would be most sensitive at high temperatures. In contrast to this, we showed that transiently formed helices had the highest helical content at low temperatures, and that experiments aiming to identify transiently formed helices in IDPs would be most sensitive under these conditions.
The choice of a suitable reference state for the statistical coil is aggravated by the redistribution of the statistical coil. When the chemical shifts at 5 and 45°C are compared, the changes are partially obscured by the effect of temperatures on the statistical coil. The method of intrinsic chemical shift references allows the subtraction of this perturbation effect from the secondary chemical shifts. Similarly, the statistical coil is influenced by other parameters pertaining to the solvent environment. When the many studies that are focused on solvent perturbation of IDPs are extended to high-resolution NMR studies, it is important to take the redistribution of the statistical coil into consideration. Intrinsic statistical coil chemical shift referencing offers a convenient way to accomplish this.
A multitude of studies have pointed out that IDPs undergo temperature-dependent structural changes typically interpreted as formation of helical structures at higher temperatures. Residue-specific information derived from multidimensional NMR at different temperatures allowed quantification of the temperature-dependent changes in transiently formed α‒helices. All helical regions were shown to lose helicity with increasing temperature, which is in accordance with the behavior of isolated helices and unfolded globular proteins. This unequivocally demonstrates that the helices are not responsible for the spectroscopic change observed by CD spectroscopy. The spectroscopic change is independent of tertiary interactions and is caused primarily by redistribution of the statistical coil ensemble. The presence of residue-specific effects highlights that caution should be taken in the interpretation of results from bulk methods such as CD spectroscopy.
Materials and Methods
Protein and peptide preparations
The plasmid for coexpression of ACTR(G1018‒D1088) with the NCBD of CBP was a gift from Peter E. Wright (The Scripps Research Institute) and was described previously.37, 38 Isotope labeling was achieved by growing the cells in M9 medium with 4 g/L 13C6 glucose and 1.5 g/L 15N ammonium sulfate as sole carbon and nitrogen sources. The cells were resuspended in 20 mM sodium phosphate buffer pH 7.4, heated to 55°C for 5 min, and cell debris pelleted by centrifugation. This procedure simultaneously lysed the cells and inactivated most proteases.62 Proteins were precipitated from the lysate by 60% (v/v) ammonium sulfate precipitation at 4°C. The pellet was resuspended in 4.5M urea in 20 mM sodium phosphate buffer, pH 7.4, and passed through a MonoQ‒ion exchange column (1 mL) in the same buffer in which neither the NCBD nor ACTR binds. The two proteins were separated by reversed-phase HPLC using a Source 15RPC column and a linear gradient from 0 to 70% (v/v) acetonitrile in 0.1% (v/v) TFA over 20 column volumes, room temperature.
The human NHE1 C‒terminal distal tail was cloned as Met + hNHE1(V686–Q815) (hNHE1cdt) in a pET11.a vector (Novagen). Primers were designed to amplify the DNA construct (restriction sites are denoted in brackets): hNHE1cdt forward primer: 5′‒TACCATATGGTGCCAGCCC‒3′ (NdeI), reverse primer: 5′‒GTGGCTAGCTCACTGCCCCTTGGGGA AG‒3′ (NheI). The plasmid was purified using the Wizards plus SV miniPrep DNA purification system (Promega, Madison, WI) and confirmed by sequencing (MWG biotech, Ebersberg, Germany).
Recombinant hNHE1cdt was overexpressed in E. coli BL21 Codon Plus cells in 1 L M9 minimal medium. Cells were grown at 37°C, 198 rpm until OD600= 0.6. Protein expression was induced using 0.5 mM isopropyl β‒d‒1‒thiogalactopyranoside for 4 h. Cells were harvested by centrifugation at 5000g and broken by sonication. Protein was expressed as soluble protein, and the supernatant was concentrated and directly applied to a size exclusion column, Sephadex G‒75 (column volume: 155 mL, 77.5 cm × 1.6 cm), equilibrated in 50 mM Tris‒HCl, 100 mM NaCl, pH 7.4, at 4°C. Fractions containing the desired protein were pooled and dialyzed against 50 mM Tris‒HCl, pH 7.4, at 4°C and applied to a HiTrap Q FF 5 mL anion exchange column (GE Healthcare) equilibrated in 50 mM Tris‒HCl, pH 7.4, and eluted by a linear gradient of NaCl (0–1M). Purity of hNHE1cdt batches was >95% as judged by SDS‒PAGE analysis. The expression and purification of Spd1 were performed as described elsewhere.9
Peptide 1 (RPLLRNSLDD), peptide 2 (GQSDERALLDQLH), and peptide 3 (DRALGIPELVNQG) were purchased from KJ Ross‒Petersen ApS (Klampenborg, Denmark) and were all purified by RP‒HPLC. To avoid non‒native charges, all peptides were acetylated at the N‒terminus and amidated at the C‒terminus.
Far‒UV CD measurements were recorded on a Jasco J‒810 Spectropolarimeter with peltier control using 1-mm Quartz SUPRASIL cuvettes (Hellma). Far‒UV CD spectra of peptides 1, 2, and 3 were recorded at 5, 45, and 95°C, respectively. Far‒UV CD spectra of ACTR, hNHE1cdt, and Spd1 were recorded at 5, 20, 35, 50, 65, 80, and 95°C. All far‒UV CD spectra were recorded from 250 to 190 nm with a scan speed of 20 nm/min, 10 accumulations, and a response time of 2 s. Sample conditions were 13.75 μM ACTR, 8 μM hNHE1cdt, 5 μM Spd1, 62.9 μM peptide 1, 37.5 μM peptide 2, 60 μM peptide 3, all in 20 mM NaH2PO4 buffer, pH 7.0. Near‒UV spectra were recorded from 350 to 250 nm at 20°C using the same protein concentrations and buffer conditions but in a 1-cm Quartz SUPRASIL cuvette (Hellma). Equivalent spectra of buffers were recorded and subtracted from the spectra of the proteins, and the resulting spectra smoothed using an FFT filter.
Small-angle X‒ray scattering
Samples for SAXS contained 150 mM NaCl, 20 mM sodium phosphate buffer, pH 7.4. For each condition, three concentrations of ACTR, Spd1, and hNHE1cdt were measured ranging from ˜1 to 4 mg/mL. SAXS data were recorded at the EMBL X33 beamline at the DORIS storage ring, DESY, Hamburg using an exposure time of 2 min. As the data did not show any concentration-dependent effects, we used the highest concentration for further analysis. Each of the curves was used as input for the ensemble optimization method,31 which uses a genetic algorithm to select an ensemble of structures that reproduced the scattering profile from a pool of 10,000 random peptide chains. The algorithm was only allowed to run for 50 generations, which was the point where the χ2 did not improve any further. For each scattering curve, 500 ensembles of 20 structures were selected, and the distribution of radii of gyration and maximal distances were extracted.
NMR spectra were acquired on a 2 mM13C,15N‒labeled ACTR sample in 20 mM sodium phosphate buffer, pH 6.5 including 10% (v/v) D2O and 1 mM DSS. An identical sample containing 6M urea was used to determine the reference “intrinsic statistical coil” chemical shifts, δisc. All spectra were recorded on a Varian Unity Inova 750 or 800 MHz spectrometer equipped with a room temperature probe. 15N,1H HSQC spectra were acquired at 5, 15, 25, 35, and 45°C, and HNCA and HNCO63 spectra were acquired at 5 and 45°C. Chemical shift referencing was performed relative to DSS. Resonance assignments from BMRB entry: 1539835 were adapted to the slightly different experimental conditions using the HNCA spectrum. The NMR data were zerofilled, apodized, and Fourier transformed in NMRDraw64 and analyzed in CCPNMR Analysis 2.1.65
Prediction of helicity
Residue-specific predictions of helicity were calculated using AGADIR.49, 66, 67 The calculations were performed using a pH of 6.5, an ionic strength of 0.05M, and temperatures of 5 and 45°C.
The authors thank the staff at the X33 beamline at EMBL Hamburg for help with the acquisition of the SAXS data and Peter E. Wright (The Scripps Research Institute) for generously sharing the coexpression plasmid for ACTR:NCBD.