Alanine and proline content modulate global sensitivity to discrete perturbations in disordered proteins

Abstract

Molecular transduction of biological signals is understood primarily in terms of the cooperative structural transitions of protein macromolecules, providing a mechanism through which discrete local structure perturbations affect global macromolecular properties. The recognition that proteins lacking tertiary stability, commonly referred to as intrinsically disordered proteins, mediate key signaling pathways suggests that protein structures without cooperative intramolecular interactions may also have the ability to couple local and global structure changes. Presented here are results from experiments that measured and tested the ability of disordered proteins to couple local changes in structure to global changes in structure. Using the intrinsically disordered N-terminal region of the p53 protein as an experimental model, a set of proline and alanine to glycine substitution variants were designed to modulate backbone conformational propensities without introducing non-native intramolecular interactions. The hydrodynamic radius (R_h) was used to monitor changes in global structure. Circular dichroism spectroscopy showed that the glycine substitutions decreased polyproline II (PP_II) propensities relative to the wild type, as expected, and fluorescence methods indicated that substitution-induced changes in R_h were not associated with folding. The experiments showed that changes in local PP_II structure cause changes in R_h that are variable and that depend on the intrinsic chain propensities of proline and alanine residues, demonstrating a mechanism for coupling local and global structure changes. Molecular simulations that model our results were used to extend the analysis to other proteins and illustrate the generality of the observed proline and alanine effects on the structures of intrinsically disordered proteins.

Introduction

Intrinsic disorder is a common structural motif observed in many eukaryotic proteins most-widely found in transcription factors and other regulatory proteins that transduce molecular signals (1–4). Proteins, or protein domains, that have been classified as intrinsically disordered exhibit wide degrees of conformational freedom and flexibility (5), an absence of correlated dynamics between residue positions (6,7), and, overall, a persistent lack of tertiary stability (8,9). In addition to signal transduction, protein phosphorylation usually occurs within intrinsically disordered regions (10) and, accordingly, many critical biological processes are mediated by intrinsically disordered proteins (IDPs), such as transcriptional regulation, chromatin remodeling, and cell cycle control (11–13).

Induced folding studies (14–16) have led to quantitative descriptions of IDP signaling activity (17–19) in terms of allostery arising from the statistical thermodynamics of disorder-to-order transitions (i.e., cooperative folding). These allostery models have been useful for explaining distinctive signaling-related phenomena, such as conditional cooperativity (20) and agonism/antagonism switching (21). A key feature of these models is the induced folding of disordered structure from binding that results in the thermodynamic coupling of distant protein regions. Binding-induced folding has been observed for many IDPs (16,20–25) and characterized in detail, in terms of structures (23), binding kinetics (24), and interaction energies (25). Global organization of IDP structure in the absence of binding or folding has also been reported (26,27). The physicochemical mechanisms that organize disordered protein structures on a global scale without cooperative folding are not yet fully understood.

The study presented here quantitatively investigates coupled structural properties in an IDP in the absence of detectable folding. The intrinsically disordered N-terminal region of the p53 protein, p53(1–93), was used as the model IDP since this protein fragment is structurally dynamic (22,28,29), exhibits regulatory activity in the context of the full-length protein (30,31), and is proline-rich with high net charge; all common IDP properties (2–4). Motivation for the present work was the observation that substitution of all proline residues for glycine decreased the hydrodynamic radius (R_h) of p53(1–93) by 5.3 Å, though only a 2 Å reduction was expected from lower polyproline II (PP_II) propensities in the substitution variant (32). The experimental data showed that changes in intrinsic backbone conformational propensities could be coupled to changes in the global structure and, specifically, that changes in local PP_II propensities could cause changes in R_h greater than the predictions of structure-based models.

To assess the relationships between global structure, as reported by R_h, and position-specific conformational propensities, the structural characteristics of a set of proline (PRO) and alanine (ALA) to glycine (GLY) substitution variants of p53(1–93) were measured. The amino acid substitutions were designed to modulate intrinsic backbone conformational propensities without introducing non-native or long-range intramolecular interactions. Circular dichroism (CD) spectroscopy was used to estimate secondary structure content in each p53(1–93) variant while dynamic light scattering (DLS) and size exclusion chromatography (SEC) techniques were used to measure R_h. Analysis of the results estimated that per-residue PP_II propensities for ALA and PRO in the wild type were 48% and 76%, respectively. R_h dependence on intrinsic PP_II propensities followed an exponential trend, consistent with simulation data reported previously (32), showing an increased dependence on the amount of PP_II structure when R_h had expanded values and a reduced dependence with compacted R_h. To test if R_h compaction caused by the PRO and ALA to GLY substitutions buried hydrophobic surface, fluorescence methods were used to measure solvent accessibility of tryptophan groups. Overall, the results showed that disordered structures can fine-tune R_h sensitivity to local structure perturbations through intrinsic backbone conformational propensities and thus regulate the linkage between local and global structure changes in the absence of cooperative folding. Extrapolation of the p53(1–93) results to other IDPs using R_h values from published reports suggested that most IDPs have similar structural characteristics in their disordered states.

Materials and Methods

Expression and purification of recombinant protein

Recombinant human p53(1–93) and staphylococcal nuclease were expressed in bacterial cells and isolated to >95% purity using the protocols described elsewhere (33,34). Purified bovine carbonic anhydrase, chicken albumin, and horse myoglobin were purchased from Sigma-Aldrich (St. Louis, MO) and further processed by ion exchange chromatography to remove residual contaminants.

Circular dicroism spectroscopy

CD spectra were recorded using a Jasco J-710 spectropolarimeter equipped with a PFD-425S peltier unit (Jasco, Easton, MD) and employed a 1-mm path-length quartz cuvette. Samples were equilibrated at each temperature for 10 min. Spectra were collected with a resolution of 0.5 nm, a scan rate of 20 nm/min, and were the average of eight scans. Reported spectra were baseline corrected for solvent and buffer contributions.

Dynamic light scattering

DLS readings used noninvasive backscatter optics and were measured using a Zetasizer Nano ZS with peltier temperature control from Malvern Instruments (Worcestershire, UK). Samples were filtered using 0.2 mm PVDF syringe-driven filters from EMD Millipore Corporation (Billerica, MA). All measurements used 1-cm path-length quartz cuvettes. Solvent viscosity was calculated using the solvent builder software program provided by Malvern, which calculates the viscosity of dilute solutions from the solution contents (e.g., percentage weight of solutes and protein) based on the program Sednterp (35).

Size exclusion chromatography

SEC experiments used Sephadex G-75 (GE Healthcare, Piscataway, NJ) equilibrated in 10 mM sodium phosphate, 100 mM sodium chloride, pH 7. Elution volumes were determined from chromatograms measured using a Bio-Rad BioLogic LP System equipped with a UV absorbance monitor (Hercules, CA). Each sample contained 0.2–0.5 mg/mL protein in 10 mM sodium phosphate, 100 mM sodium chloride, pH 7 with 0.3 mg/mL blue dextran and 0.03 mg/mL 2,4-dinitrophenyl-L-aspartate added as indicator dyes to determine the void and total column volumes, respectively.

Flourescence spectroscopy

Fluorescence measurements were performed using a Horiba Jobin-Yvon Fluorolog 3 spectrofluorometer equipped with a Wavelength Electronics Model LF1–3751 temperature controller. Spectra of 0.5 µM p53(1–93) and of the corresponding buffer were recorded at 20°C and averaged three times from 305–440 nm using excitation wavelengths of 280 and 295 nm. The step width was 1 nm, the integration time 1 s. Isothermal acrylamide quenching of the tryptophan fluorescence for 0.5 µM p53(1–93) used an excitation wavelength of 295 nm. Following addition of buffered acrylamide solution, the protein sample was equilibrated for 5 min under slight stirring and the fluorescence intensity was recorded for 20 s at 359 nm and averaged. Thermal scans with 0.5 µM p53(1–93) were performed using a scan rate of 2°C min⁻¹ from 20–75°C under slight stirring.

Results and Discussion

Computer simulated structural relationship between PP_II content and R_h

R_h measured for p53(1–93) under normal solution conditions was 32.8 Å at 25°C, approximately 11 Å larger than statistical coil estimates for a 93-residue polypeptide (32). A statistical coil is defined as a polypeptide chain with no strongly preferred conformations (36). p53(1–93) shows a local maximum in its CD spectrum at 221 nm, indicating PP_II propensities that were estimated at 12–17% by spectral deconvolution (33). To determine if 12–17% PP_II propensities could explain the relatively large R_h that was measured for p53(1–93), simulations of randomly configured polypeptide chains using a hard sphere collision model (37) were performed to quantify the relationship between chain propensities for PP_II and R_h in disordered structures (32). The results of those computer simulations were condensed to a set of simple equations that are reproduced here. The mathematical relationships outlined by these equations frame the experimental strategy of the current study and accordingly were used for data analysis.

The first equation in this set relates R_h (in Å) to the number of residues, N, using the standard power-law scaling relationship,

$$R_h=R_0·N^v,$$ (1)

where R_o was found to be 2.16. The effects of N on R_h were insensitive to amino acid sequence in the simulations. The exponential term, v, was dependent on the applied sampling bias for PP_II, S_PPII, by,

$$v(S_{PPII})=v_0-0.088·\text{ln}(1-S_{PPII}),$$ (2)

where v_o was the value of the exponential with no applied chain bias (i.e., for S_PPII = 0), found to be 0.509. Elimination of randomly generated structures containing van der Waals contact violations (i.e., removal of steric conflicts from the simulation) caused the fractional number of PP_II folded residues, f_PPII, to differ from S_PPII by a Gaussian function that can be represented as,

$$f_{PPII}=S_{PPII}-0.062·e^{(-{(S_{PPII}-0.63)}^2/2·{0.28}^2)}.$$ (3)

Equations 1–3 provide an estimate of the dependence of R_h on PP_II content in disordered proteins of any size N.

From equations 1–3, PP_II propensities of 12–17% increase R_h for a 93-residue disordered protein to 23.0–23.6 Å from the statistical coil value of 21.7 Å (see Figure 1). Since R_h for p53(1–93) was measured to be 32.8 Å, these data indicated that either the simulation results were in gross error, the R_h measurements were wrong, PP_II propensities in p53(1–93) were much higher than 17%, or molecular properties other than PP_II propensities were dominant in determining R_h for this proline-rich protein (22 of 93 residues are proline in the wild type sequence). The CD-estimated PP_II propensities were supported by NMR analysis of the proline and pre-proline positions (38), whereas size exclusion chromatography and gel electrophoresis experiments indicated that the DLS-measured R_h values were reasonable (32). Surprisingly, the decrease in R_h, relative to the wild type, measured for a p53(1–93) variant that had all PRO residues substituted for GLY (referred to as PRO⁻) supported the simulation data in terms of the predicted change in f_PPII. Figure 1 shows that an analysis of R_h measured for this PRO⁻ variant, found to be 27.5 Å at 25°C, was consistent with an 18% decrease in f_PPII relative to the wild type, in close agreement to the CD-estimated wild type PP_II propensities (32).

Figure 1. R_h dependence on f_PPII for a 93-residue disordered protein.

R_h measured for wild type and PRO⁻ p53(1–93) are indicated in the figure. The curved line was calculated using equations 1–3 with N = 93. ∂R_h/∂f_PPII for a statistical coil was estimated empirically by calculating the change in R_h and f_PPII for the incremental change in S_PPII of 0 to 0.01. ∂R_h/∂f_PPII for wild type was estimated by a similar method. First, the wild type R_h (32.8 Å) was used to calculate a corresponding S_PPII (S_PPII,wt) using equations 1 and 2. Then ∂R_h/∂f_PPII for the wild type was estimated by calculating the change in R_h and f_PPII for the incremental change in S_PPII from S_PPII,wt to S_PPII,wt + 0.01.

The simulation results thus seemed to capture some features of the relationship between intrinsic backbone conformational propensities and the global structural parameter R_h despite its simplicity. The simulation data also indicated that changes in R_h from incremental changes in f_PPII (i.e., ∂R_h/∂f_PPII) were strongly dependent on the degree of R_h compaction. This can be seen in Figure 1 by noting that ∂R_h/∂f_PPII for wild type p53(1–93) was ~4X the statistical coil value. To test if R_h dependence on backbone conformational propensities could be affected by structural compaction in IDPs, two additional p53(1–93) variants were constructed. One had all 12 alanine residues in the wild type substituted for GLY (referred to as ALA⁻). The other used the same ALA-to-GLY substitution strategy applied to the PRO⁻ variant (referred to as ALA⁻PRO⁻). ALA residues in disordered polypeptides show significant PP_II propensities, with per-residue estimates as high as 80% (39,40). Thus, ALA should contribute to R_h in a manner similar to PRO. These two additional substitution variants also tested additivity of the effects of PP_II structure on R_h in this system, by measuring the effects of ALA-to-GLY substitutions in the background of wild type and PRO⁻, and similarly by measuring the effects of PRO-to-GLY substitutions in the background of wild type and ALA⁻.

Proline and alanine contributions to PP_II propensities

CD spectroscopy was used to estimate relative PP_II propensities among the four p53(1–93) variants. The spectra measured under normal solution conditions for wild type, ALA⁻, PRO⁻, and ALA⁻PRO⁻ from 5–85°C are provided in Figure 2. Spectra measured for ALA⁻PRO⁻ were practically superimposable with the spectra measured for PRO⁻. A local maximum centered at 221 nm was observed in the variants that contained proline residues that was mostly absent in the variants that had no prolines (figure insets). A local CD maximum from 220–225 nm is often seen in proteins containing PP_II (41–43). For p53(1–93), this local maximum disappeared with increased temperatures, consistent with known heat effects on PP_II propensities (44). Overall, these data suggested that the proline-containing variants had significant PP_II content at normal temperatures, whereas the variants lacking proline had negligible PP_II content. Panel A of Figure 3 shows that the CD spectra at high temperatures differed noticeably between variants that contained proline residues (wild type and ALA⁻) in comparison to variants that had no prolines (PRO⁻ and ALA⁻PRO⁻). Thus, at high temperatures where PP_II content should be minor, structural differences were noted owing to the presence or absence of proline residues in the sequence.

Figure 2. CD spectrum of p53(1–93) from 5–85°C.

Panels show A) wild type, B) ALA⁻ C) PRO⁻, and D) ALA⁻PRO⁻. All spectra were measured in 10 mM sodium phosphate, 100 mM sodium chloride, pH 7 using 0.1–0.2 mg/mL of protein and reported in molar residue ellipticity (MRE) units of deg cm² dmol⁻¹ res⁻¹. The data symbols (filled circles) were colored according to the temperature the spectrum was measured using: 5°C (black), 15°C (gray), 25°C (purple), 35°C (blue), 45°C (blue–green), 55°C (green), 65°C (yellow), 75°C (orange), and 85°C (red). The insets show the difference in MRE between the spectra at each temperature relative to the 85°C spectrum.

Figure 3. Effects of PRO-to-GLY and ALA-to-GLY substitutions on the CD spectrum of p53(1–93).

In both panels, wild type measurements were given with circles, ALA⁻ with squares, PRO⁻ with triangles, and ALA⁻PRO⁻ with diamonds. Panel A shows the spectrum of each variant at 85°C. Panel B shows the temperature-dependence in the height of the local CD maximum that was observed at 221 nm (from insets of Figure 2).

Spectral deconvolution using the CDPro software package (45) was performed to quantify PP_II content in the four p53(1–93) variants. Noted previously (32), CDPro predicted 12–17% PP_II propensities for the wild type spectrum, depending on the algorithm (e.g., CONTIN/LL, CDSSTR) and basis set used. The results of spectral analysis performed in this manner were surprisingly insensitive to differences in the p53(1–93) spectrum owing to heat and PRO-to-GLY and ALA-to-GLY substitutions. For example, the CONTIN/LL algorithm and SP22X basis set (46) showed PP_II content that ranged from 13.1–13.8% among the four variants at all temperatures (Figure S1). Since PRO-to-GLY substitutions (47) and increased temperatures (44) are known to decrease PP_II content, and this was not detected by spectral deconvolution, only qualitative differences in PP_II propensities among the p53(1–93) variants could be determined from CD readings. Using the local CD maximum at 221 nm as a metric for PP_II content, Figure 3 panel B shows that in terms of the height of this local maximum, wild type > ALA⁻ > PRO⁻ ≳ ALA⁻PRO⁻. These CD data were thus interpreted qualitatively to indicate that at normal temperatures, wild type p53(1–93) had the highest levels of PP_II content, with ALA⁻ slightly less, and PRO⁻ and ALA⁻PRO⁻ significantly less.

Proline and alanine contibutions to R_h

To quantify the structural size of p53(1–93), its diffusion coefficient, D, was measured by DLS and converted to an apparent hydrodynamic radius (R_h) using the Stokes-Einstein equation,

$$R_h=kT/(6\pi \eta D),$$ (4)

where k is the Boltzmann constant, T is absolute temperature, and η is solvent viscosity. The results from measuring R_h for each variant over a temperature range of 5–75°C using DLS are provided in Figure 4.

Figure 4. DLS-measured R_h for p53(1–93) from 5–75°C.

R_h values were shown using circles for wild type, squares for ALA⁻, triangles for PRO⁻, and diamonds for ALA⁻PRO⁻. R_h were measured using 0.25–0.75 mg/mL samples of protein in 10 mM sodium phosphate, 100 mM sodium chloride, pH 7. The reported values are the average of at least 5 measurements performed at each temperature and error bars represent standard deviations. The stippled line labeled “coil” shows R_h calculated for a 93-residue disordered protein using equations 1 and 2 with S_PPII = 0.

The three substitution variants each showed the same broad and gradual compaction in R_h caused by increased temperatures that was noted previously for the wild type (32), indicating heat-induced structural changes that were absent cooperative effects. At low temperatures, R_h for each variant was at least 7 Å larger than R_h calculated for a 93-residue statistical coil. At 75°C, R_h for PRO⁻ and ALA⁻PRO⁻ approached statistical coil dimensions, whereas R_h for wild type and ALA⁻ remained significantly more expanded by ~5 Å. Similar to the spectroscopic observations at high temperatures using CD (Figure 3 panel A), these data indicated structural differences owing to proline residues even when proline-associated PP_II content should be small. Also, compaction in R_h from glycine substitution followed the trend that was observed in the CD-estimated changes in PP_II content. ALA⁻, which was estimated to have slightly less PP_II content than the wild type, had a slightly smaller R_h at most temperatures. PRO⁻ and ALA⁻PRO⁻, which were estimated to have significantly less PP_II content than the wild type, had significantly smaller R_h.

To test if the DLS methods measured R_h accurately, size exclusion chromatography (SEC) experiments were performed at room temperature using the p53(1–93) variants and a set of protein standards. Correlating elution volumes (V_e) measured for the protein standards to molecular weight using the distribution coefficient,

$$K_D=(V_e-V_0)/(V_t-V_0),$$ (5)

where V_t and V_o were the total and void volumes of the column, respectively, shows that the elution volume for each p53(1–93) variant was comparatively low, indicating R_h similar to folded proteins of much larger molecular weight, and R_h in a rank order of wild type > ALA⁻ > PRO⁻ ≳ ALA⁻PRO⁻ (Figure 5 panel A). K_D that were measured for the p53(1–93) variants are provided in Table I. Though K_D for these variants did not trend with molecular weight when compared to folded proteins, K_D for all tested proteins trended with R_h (Figure 5 panel B). Here, R_h was estimated as one-half the maximum distance between any two C_α atoms in the reported crystallographic structure for each protein standard (48–51) and then used to correlate linearly with K_D. Plotting DLS-measured R_h against K_D followed the same linear trend for the p53(1–93) variants, showing that SEC and DLS methods yield consistent R_h values that likewise compare favorably to R_h estimated from high resolution structures. Because of this strong agreement, K_D values measured for the p53(1–93) variants were used to estimate R_h using the linear correlation that was observed with the protein standards. R_h determined in this manner yielded almost identical R_h values as the DLS measurements (Table I).

Figure 5. SEC-measuredK_Dfor p53(1–93) at room temperature.

Panel A shows a comparison of K_D to molecular weight for wild type, ALA⁻, PRO⁻, and ALA⁻PRO⁻ (filled circles). Open circles are K_D measured for chicken albumin (44.3 kDa), bovine erythrocyte carbonic anhydrase (29.8 kDa, indicated as CA in the figure), staphylococcal nuclease (16.9 kDa), and horse myoglobin (16.95 kDa). Reported K_D were the average of at least 3 measurements for each protein with error bars representing the standard deviations. Panel B shows a comparison of K_D to R_h. Open circles are R_h estimated as one-half the maximum C_α-C_α distance in the crystallographic structures of albumin (48), carbonic anhydrase (49), nuclease (50), and myoglobin (51). The dashed line is a linear fit of R_h to K_D applied to the open circles. Filled circles show R_h measured by DLS at 25°C for wild type, ALA⁻, PRO⁻, ALA⁻PRO⁻, carbonic anhydrase, and nuclease. K_D were measured using 0.2 mg/mL samples of protein in 10 mM sodium phosphate, 100 mM sodium chloride, pH 7. DLS measurements used identical solution conditions except for 0.25–0.75 mg/mL protein concentrations.

Table I. R_h measured for p53(1–93) variants.

R_h values are given in Å. R_h,DSL refers to DLS measured values at 25°C (from Figure 4) and R_h,SEC is room temperature measured values using SEC (from Figure 5B). R_h,DSL and R_h,SEC averaged for each variant is given by R_h,avg. R_h,seq was calculated using the equation R_h = (1.24·f_PRO + 0.904) (0.00759·Q_net + 0.963)·2.49·N^0.509 determined by Marsh and Forman-Kay (52).

	Rh,DSL	KD	Rh,SEC	Rh,avg	Rh,seq
wild type	32.80 ± 1.25	0.160 ±0.008	31.95	32.4	32.2
ALA−	31.08 ± 1.28	0.205 ±0.005	29.65	30.4	32.2
PRO−	27.53 ± 1.32	0.249 ±0.005	27.36	27.4	24.3
ALAPRO−	27.71 ± 0.92	0.252 ±0.005	27.18	27.4	24.3

Forman-Kay and Marsh showed that R_h for IDPs scale with fractional proline content and net charge and can be predicted accurately from sequence (52). For p53(1–93), sequence predicted an R_h of 32.2 Å for the wild type and 24.3 Å for the variants lacking proline residues (Table I). For a 93-residue IDP with no proline residues and no net charge, the sequence-based equation predicted an R_h of 21.8 Å. These comparisons are important for a number of reasons. First, it shows that R_h measured for wild type p53(1–93) by DLS and SEC were consistent with general R_h trends observed in other IDPs – and thus the p53(1–93) values were not anomalous. Next, it demonstrates that estimates of statistical coil R_h from equations 1 and 2 (i.e., with S_PPII = 0) are reasonable since they agree with IDP R_h values absent proline-induced or net charge-induced structural expansion. For the specific case of p53(1–93), our data indicated that proline effects on R_h depend on the presence of alanine residues at the wild type ALA positions. This observation was not predicted by the Forman-Kay and Marsh equation that calculates an identical proline contribution to R_h (7.9 Å) for both the wild type and ALA⁻.

R_h values measured for the p53(1–93) variants were also used to estimate per-residue PP_II propensities for ALA and PRO. Using averaged R_h for wild type and PRO⁻, equations 1–3 predict that f_PPII changed from 0.572 to 0.393 with the PRO-to-GLY substitutions. This represents a change in fractional PP_II content of −0.179 and indicates a per-residue PP_II propensity of 76% for each PRO in wild type. An identical calculation using the averaged R_h for ALA⁻ gave a per-residue PP_II propensity of 48% for each wild type ALA. Other research groups have shown that alanine residues favor PP_II structure in disordered peptides, with propensities reported to be as low as 30% (53) and as high as 80% (39,40). Our measurements of ALA propensities for PP_II were within these (admittedly large) extremes.

Figure 1 predicted that R_h dependence on f_PPII should decrease with structural compaction. The experimental results clearly were in qualitative agreement with that prediction. The PRO-to-GLY substitutions compacted R_h by 5 Å when applied to the wild type, but caused only a 3 Å effect in ALA⁻. Similarly, the ALA-to-GLY substitutions applied to the wild type compacted R_h by 2 Å, but had a negligible effect on R_h in PRO⁻. Using the measured R_h values to compare changes in f_PPII among the p53(1–93) variants, however, demonstrated quantitative differences between the simulation model and the experimental results. The PRO-to-GLY substitutions decreased f_PPII by 0.179 when applied to the wild type and only by 0.117 when applied to ALA⁻. If the simulation and experimental results were in full quantitative agreement, these values should have matched. Similarly, f_PPII decreased by 0.062 from wild type to ALA⁻, but had no change from PRO⁻ to ALA⁻PRO⁻.

Contributions of hydrophobic burial to R_h compaction

The PRO-to-GLY and ALA-to-GLY substitutions compacted R_h for p53(1–93). Our analysis of the changes in R_h included only the effects of reduced PP_II content. Since unfolding almost universally reports increased R_h values for stable proteins (54), R_h compaction is typically associated with protein folding rather than lowered chain propensities for PP_II. To test for “folding” effects in R_h compaction of p53(1–93), experiments were performed to determine if hydrophobic surfaces were buried with the glycine substitutions. In this pursuit, intrinsic tryptophan fluorescence was measured in the presence of increasing acrylamide concentrations for wild type, ALA⁻, PRO⁻, and ALA⁻PRO⁻. Tryptophan residues are found at positions 23, 53, and 91 in p53(1–93). Acrylamide solutions quench tryptophan fluorescence by collisional contact and concomitant energy transfer, though this effect is more pronounced for exposed tryptophan groups relative to those that are structurally buried and protected from solvent (55–58).

The intrinsic fluorescence of each variant at 20°C is provided in Figure 6 panel A. The emission maximum for each was ~360 nm from excitation at 280 and 295 nm, indicating tryptophan residues in polar environments consistent with fully exposed groups (59,60). Adding acrylamide decreased p53(1–93) fluorescence proportional to the amount that was added. This is shown in panel B, where fluorescence is given in terms of F_o/F. F_o refers to fluorescence in the absence of acrylamide and thus the ratio F_o/F increases with increased quenching. For reference, acrylamide induced fluorescence quenching of equal molar amounts of N-acetyl tryptophanamide (NATA) was included. Quenching increased with higher temperatures (Figure S2), indicating a dynamic quenching mechanism (i.e., collisional frequency increases with added heat). If fluorescence quenching was caused by a static (i.e., binding) mechanism, then temperature increases would reduce the formation of bound complexes, thus reducing quenching, which was not observed for any of the p53(1–93) variants.

Figure 6. p53(1–93) fluorescence spectrum and acrylamide induced quenching of tryptophan fluorescence.

Panel A shows the fluorescence spectrum of p53(1–93) measured at 20°C using excitation wavelengths of 280 nm (filled markers) and 295 nm (open markers). Panel B provides Stern-Volmer plots of acrylamide induced quenching of tryptophan at 20°C. Both panels used circles for wild type, squares for ALA⁻, triangles for PRO⁻, and diamonds for ALA⁻PRO⁻. NATA is shown in panel B using (X). In both panels, fluorescence was measured using 0.5 µM p53(1–93) or 1.5 µM NATA buffered at pH 7 with 10 mM sodium phosphate, 100 mM sodium chloride.

The extent of acrylamide induced fluorescence quenching, and thus tryptophan solvent accessibility, was assessed using the Stern-Volmer equation and the resulting Stern-Volmer constant, K_S-V (61). Briefly, fluorescence intensity can be described as,

$$F_0/F=1+K_{S-V}·[Q],$$ (6)

where [Q] is the quencher concentration. K_S-V was 18.9 M⁻¹ for wild type, 17.1 M⁻¹ for ALA⁻, 19.0 M⁻¹ for PRO⁻, 23.3 M⁻¹ for ALA⁻PRO⁻, and 28.2 M⁻¹ for the NATA reference. Disordered proteins (56) and proteins denatured by guanidine hydrochloride with reduced disulfides (62) characteristically have K_S-V values ≥ 12 M⁻¹, while values of 1–2 M⁻¹ are typical for folded proteins with tryptophan groups that are protected from solvent (55,56,62). K_S-V measured for each p53(1–93) variant at 20°C thus was consistent with exposed tryptophan groups. Also, the substitution effects on K_S-V values indicated mostly minor changes in tryptophan solvent accessibility. In general, the trends from PRO-to-GLY and ALA-to-GLY substitutions on K_S-V were slight increases in tryptophan solvent accessibility, possibly owing to increased chain disorder from decreased PP_II content. The fluorescence quenching results did not seem to indicate significant burial of hydrophobic surface associated with R_h compaction owing to the PRO-to-GLY and ALA-to-GLY substitutions.

R_h trends among intrinsically disordered proteins

R_h that was measured for wild type p53(1–93) is similar in character to R_h measured for other IDPs, which is demonstrated in Figure 7 panel A. Here, R_h values for IDPs (26,63–80) and folded proteins (32,54,81) were plotted according to chain length. This dataset excluded IDPs with histidine tags, since histidine tags are known to compact R_h in IDPs (52). For reference, R_h for statistical coils were shown in this figure and provide a defining separation between folded and intrinsically disordered proteins. The folded proteins were all below the coil line and IDPs were all at or, more commonly, above this line.

Figure 7. R_hcomparisons to number of residuesNfor folded and intrinsically disordered proteins.

This set was limited to N > 50 and used R_h from published reports (26,32,54,63–81). In panel A, the line labeled “coil” shows R_h calculated using equations 1 and 2 with S_PPII = 0. Filled circles are IDPs and open circles are folded proteins. In panel B R_h values for IDPs were converted to RSC_PPII RSC_PPII was determined for each IDP by first calculating the statistical coil ∂R_h/∂f_PPII (which varies with N) and IDP ∂R_h/∂f_PPII using the method given in Figure 1 for wild type p53(1–93). These two ∂R_h/∂f_PPII values were then divided (IDP/coil) to give RSC_PPII. For p53(1–93), the averaged R_h (32.4 Å) was used to determine RSC_PPII rather than the DLS-measured R_h (32.8 Å). The dashed line shows the averaged RSC_PPII (2.3 ± 0.8) from the set of IDPs in panel A.

To test if ALA content contributes to R_h for IDPs in general, as it does for p53(1–93), linear regression analysis was used to correlate fractional alanine content (f_ALA) with increases in R_h relative to the coil value (R_h,rel = R_h/R_h,coil). No correlation was observed using the set of IDPs shown in Figure 7 (coefficient of determination, R² = 0.0001, Table S1 in Supporting Information). Fractional proline content (f_PRO) and net charge (Q_net) each had positive correlations with R_h,rel in this set (R² = 0.242 and R² = 0.194, respectively), in agreement with the report by Marsh and Forman-Kay (52). Since ALA effects on R_h for p53(1–93) were detected only in the variant that had the largest R_h value (i.e., wild type with high f_PRO and Q_net), multiple linear regression was used to determine if f_ALA in combination with f_PRO and Q_net could increase sequence-based correlations to R_h,rel. Regression analysis showed that R_h,rel correlated with f_PRO and Q_net with R² = 0.539. Adding f_ALA as a fitting parameter increased R² to 0.542. The increase in R² owing to f_ALA was not meaningful, however, since adding random numbers from 0 to 1 as the additional fitting parameter rather than f_ALA increased R² on average to 0.546 ± 0.008. Of note, if the set of IDPs used in regression analysis was limited to those with f_PRO above 10%, then f_ALA correlated strongly with R_h,rel (R² = 0.472). Similarly, if regression analysis was limited to IDPs with Q_net > 10, f_ALA again correlated with R_h,rel with high confidence (R² = 0.200). These data (tabulated in Table S1) suggest that ALA contributions to R_h may be minor for IDPs overall, but significant in combination with high f_PRO and/or high Q_net.

Figure 7A implies that, on average, IDPs are structurally expanded above coil estimates. Panel B in Figure 7 provides a possible selective advantage for disordered structures with expanded R_h – to regulate the coupling of local and global structure changes in IDPs. For example, the subtle effects of ALA-to-GLY substitutions on R_h were observed in wild type p53(1–93) and not in a structurally compacted variant. In Figure 7B, the reported R_h values were used to estimate ∂R_h/∂f_PPII (using equations 1–3) for each IDP and then divided by ∂R_h/∂f_PPII for a statistical coil to give a relative metric of how changes in f_PPII affect R_h (referred to as Relative Chain Sensitivity, RCS_PPII). An RCS_PPII of 1 represents the baseline effect of changes in f_PPII on R_h for a statistical coil. The IDP averaged RCS_PPII (dashed line in Figure 7B) was more than double the baseline coil value but the standard deviation was large (averaged RCS_PPII = 2.3 ± 0.8). Wide variations in RCS_PPII among IDPs could be related to protein-specific requirements for coordinating global structure changes to discrete local structure perturbations, such as phosphorylation, though experiments are needed to quantify the relationships between Q_net, f_PRO, f_ALA, and R_h. R_h,rel correlated with Q_net (R² = 0.194), as noted above, suggesting that charge effects on IDP structure may be intertwined with f_PRO and f_ALA.

To test if charge effects on R_h depend on proline content, the set of IDPs in Figure 7 was split according to f_PRO and the relationships between R_h,rel and Q_net were re-assessed. The data in Figure 8 show that the correlation between R_h,rel and Q_net decreased to negligible levels (R² = 0.028 with ∂R_h,rel/∂Q_net = 0.002) in IDPs with f_PRO > 10%. In contrast, R_h,rel correlated strongly with Q_net in IDPs with f_PRO < 10% (R² = 0.439 with ∂R_h,rel/∂Q_net = 0.008). These results predict variable charge effects, and presumably phosphorylation effects, in IDP structures dependent on proline content. A simple mechanism could account for this observation – proline promotes structural expansion in IDPs to weaken electrostatic-based interactions between charged groups. Accordingly, IDPs with high proline content, in general, should exhibit reduced coupling of changes in charge state with changes in R_h. IDPs with low proline content should demonstrate the opposite – increased coupling of changes in charge state with changes in R_h. Each IDP in Figure 7 was predicted to contain at least one phosphorylation site using the DISPHOS algorithm (Table S2 in Supporting Information), an algorithm that shows 80% accuracy in predicting protein phosphorylation sites from sequence (10). Owing to variable proline and alanine content among these IDPs, variable phosphorylation effects on R_h should be expected.

Figure 8. Comparison ofR_h,relandQ_netfor intrinsically disordered proteins.

IDPs with f_PRO > 0.10 are shown using open circles. IDPs with f_PRO < 0.10 are shown with filled circles. Lines represent linear fits of R_h,rel to Q_net for IDPs with f_PRO > 0.10 (dashed), IDPs with f_PRO < 0.10 (solid), and all IDPs (stippled).

Conclusions

Proteins that are structurally dynamic often exhibit random coil characteristics (82–84), but IDPs under normal conditions seem to show non-random structural organization (23,26,28) that is sensitive to sequence (52), charge (27,85,86), and temperature (32,87). To investigate the structural relationships between the global parameter R_h and intrinsic backbone conformational propensities in a disordered polypeptide chain, PRO-to-GLY and ALA-to-GLY substitutions were applied to the intrinsically disordered N-terminal region of the p53 protein, p53(1–93). The effects of GLY substitutions on R_h in p53(1–93) were compared to the results of molecular simulations of PP_II effects on disordered protein structures. The experimental results qualitatively matched the simulation predictions, but lacked quantitative agreement. Differences between computational and experimental data in our study may be from non-PP_II effects on disordered structure that were not included in the simulation. For example, charge-based interactions are known to affect the structural properties of IDPs (27,85,86). An analysis of IDP R_h values presented here indicated that proline content, and thus PP_II structure, and charge may act in concert to affect R_h. Specifically, R_h compaction (e.g., owing to glycine substitutions) should decrease the contributions of PP_II content to R_h and increase the effects of charge. ALA effects on R_h also seemed to be enhanced by high Q_net in IDPs.

The experimental results showed clearly that ALA content contributes to IDP R_h values, promoting structures larger than statistical coil estimates. Analysis of CD spectra indicated that ALA-to-GLY substitutions decreased PP_II content in p53(1–93), suggesting that ALA effects on R_h were primarily owing to changes in intrinsic PP_II propensities. The magnitude of R_h compaction in p53(1–93) from the ALA-to-GLY substitutions implied per-residue PP_II propensities of 48% for ALA in the wild type. Likewise, CD spectra indicated that PRO-to-GLY substitutions decreased PP_II content in p53(1–93), causing R_h compaction that was consistent with per-residue PP_II propensities of 76% for PRO. These results demonstrated that for IDPs, R_h compaction may trend with increased chain disorder, which, for p53(1–93), was caused by decreased PP_II content. This contrasts with R_h compaction associated with folded proteins (32,54), which typically follow disorder-to-order transitions (i.e., cooperative folding).

Overall, the results seemed to show that R_h sensitivity to discrete structural perturbations could vary significantly among IDPs. For example, the PRO-to-GLY substitutions caused a 5 Å compaction in R_h when applied to the wild type, but only a 3 Å compaction in R_h when applied to ALA⁻. Also, the effects of ALA-to-GLY substitutions on R_h were context specific, causing a 2 Å compaction of the wild type, but no effect on R_h for PRO⁻. In addition, correlations of R_h,rel with Q_net in IDPs showed differences owing to proline content. These results predict that IDPs, though structurally dynamic and lacking in global conformational stability, can exhibit varied global structural responses to local structural perturbations. Accordingly, the effects of phosphorylation or cofactor interactions on disordered structure may differ substantially among IDPs, but likely in a predictable and sequence dependent manner.