Thermodynamic Analysis of Point Mutations Inhibiting High-Temperature Reversible Oligomerization of PDZ3

Abstract

Differential scanning calorimetry (DSC) indicated that PDZ3 undergoes a peculiar thermal denaturation, exhibiting two endothermic peaks because of the formation of reversible oligomers at high temperature (N↔I₆↔D). This contrasts sharply with the standard two-state denaturation model observed for small, globular proteins. We performed an alanine scanning analysis by individually mutating three hydrophobic residues at the crystallographic oligomeric interface (Phe340, Leu342, and Ile389) and one away from the interface (Leu349, as a control). DSC analysis indicated that PDZ3-F340A and PDZ3-L342A exhibited a single endothermic peak. Furthermore, PDZ3-L342A underwent a perfect two-state denaturation, as evidenced by the single endothermic peak and confirmed by detailed DSC analysis, including global fitting of data measured at different protein concentrations. Reversible oligomerization (RO) at high temperatures by small globular proteins is a rare event. Furthermore, our present study showing that a point mutation, L342A, designed based on the crystal structure inhibited RO is surprising because RO occurs at a high-temperature. Future studies will determine how and why mutations designed using crystal structures determined at ambient temperatures influence the formation of RO at high temperatures, and whether high-temperature ROs are related to the propensity of proteins to aggregate or precipitate at lower temperatures, which would provide a novel and unique way of controlling protein solubility and aggregation.

Significance

Despite being a small globular protein, which normally undergoes a two-state unfolding, the thermal denaturation of PSD95-PDZ3, monitored by differential scanning calorimetry, exhibited two endothermic peaks. The second peak resulted from a reversible oligomerization (RO) at high temperatures, which is, on its own, a rare phenomenon. In this study, we show that the substitution of a single hydrophobic residue to an alanine at the interface of the crystallographic tetramers inhibited high-temperature RO, resulting in a single endothermic peak. Future studies are required to determine why mutations designed using crystal structures determined at ambient temperatures can inhibit high-temperature RO and whether the ROs are a precursor of irreversible aggregation. If so, these observations will provide an entirely new basis for creating aggregation-resistant proteins.

Introduction

Thermal denaturation of small proteins is highly cooperative and reversible as demonstrated by a sharp sigmoidal curve observed in early spectroscopic measurements (1, 2, 3, 4, 5). Furthermore, differential scanning microcalorimetry (DSC), which is a prime method for directly determining thermodynamic parameters, shows a single sharp endothermic peak corresponding to thermal transition (6, 7, 8, 9). Thus, a two-state thermal transition between the thermodynamic native (N) and the denatured (D) state is considered as a biophysical hallmark for a natively folded protein (5,6).

The sharp transition is thought to originate from the densely packed, hydrophobic interior of a protein, as exemplified in the jigsaw puzzle model (10). In small single-domain proteins, the densely packed interior forms a single hydrophobic core that would unpack in a very cooperative manner, resulting in the two-state unfolding process. In contrast, thermal denaturation of multidomain proteins can be multistepped because each domain has a hydrophobic core that can unfold independently of the rest of the protein. Another noteworthy exception to the two-state unfolding of small proteins that have a single hydrophobic core is the molten globule (MG) state. The MG state is a compact globular state with native-like secondary structures but a loosely packed protein interior (11, 12, 13, 14). The MG state was first observed as an equilibrium or thermodynamic intermediate state at nonphysiological and often extreme pHs (11, 12, 13) or when artificial mutations destabilizing the hydrophobic core were introduced into the native sequence (15). Furthermore, the equilibrium MG was associated with kinetic folding intermediates (16,17), but the relationship between the equilibrium and kinetic MG states would benefit from further analyses (18, 19, 20). Thus, under “normal” or near-physiological conditions, a natively folded, single domain protein with a single hydrophobic core will undergo a two-state thermal transition, and thus, its DSC thermogram will appear as a single endothermic peak (5,6), as mentioned above.

Recently, a few single-domain globular proteins (CheY (21), Cro repressor (22), PSD95-PDZ3 (23,24), and DEN4 ED3 (25)) exhibiting DSC thermograms with two endothermic peaks have been reported. Surprisingly, the thermal denaturation is reversible, and the proteins are monomeric both in the native and denatured states, indicating that aggregation is not the reason for the unexpected observation nor for the MG state, which usually appears only under extreme conditions. The origin of the unexpected thermal transition was the formation of a reversible oligomer at high temperature that we named the reversible oligomerization (RO) state. DSC thermograms showed two endothermic peaks, in which the first peak was associated with the unfolding of the protein and its oligomerization (N↔I_n), and the second peak reflected the dissociation of the intermediate oligomer into a monomeric denatured state (I_n↔D).

A previous study using DEN4 ED3 (PDB: 3WE1) suggested that the hydrophobic interaction at the oligomeric interface in the crystal structure of DEN4-ED3 was responsible for RO formation at high temperatures (25). DEN4 ED3 is the third domain of an envelope protein derived from dengue virus type 4, with a molecular weight of 11.4 kDa. It has an immunoglobulin fold composed of nine β-strands. The immunoglobulin fold or β-sandwich fold is one of the most common folds, and its thermal denaturation usually follows a two-state model (26), but the DSC thermogram of DEN4 ED3 exhibited two endothermic peaks and indicated the presence of a tetrameric intermediate state (N↔I₄↔D). We showed that the second peak disappeared when Val380, which is located at the crystallographic oligomer interface, was replaced with a nonhydrophobic residue (Ala, Ser, Thr, Asn, or Lys). This indicated that the formation of the high-temperature RO by DEN4 ED3 was, surprisingly, inhibited by mutating a single hydrophobic residue at the interface of its crystallographic structure, which was solved at low temperatures.

PDZ3 is the third PDZ domain of a synaptic carrier protein, PSD95, with a molecular weight of 11.0 kDa. It is made of three α-helices and eight β-strands (Fig. 1 A). DSC thermograms of PDZ3 also show two endothermic peaks, which suggested the formation of a high-temperature RO state (N↔I_n↔D) (23,24). Furthermore, earlier work on PDZ3 indicated that the ROs might be precursors of heat aggregation. Hence, strategies for inhibiting the formation of RO might facilitate the design of proteins resistant to aggregation and precipitation.

Figure 1.

Amino acid sequence (A) and ribbon model (B) of PDZ3 variants. The ribbon model was drawn using Pymol and PDB structure 3I4W. The sites of single alanine mutations are shown in red. (C) Shown is the residual conformation of the interface of PDZ3-wt (PDB: 3I4W). Two polypeptide chains are represented as wireframe models using Pymol. The backbone of PDZ3 (right gray), Phe340 (red), Leu342 (magenta), Ile389 (cyan), and Leu349 (yellow) are distinguished by different colors.

In this study, we attempted to inhibit the formation of high-temperature RO in PDZ3 by introducing a single mutation at the oligomeric interface of PSD95-PDZ3 (PDZ3) as we succeeded in doing with DEN4 ED3. To this end, we analyzed the x-ray crystal structure of PDZ3, which formed a tetramer in the unit cell (Fig. 1 B; (27)). The buried surface area (BSA) in the tetrameric crystal structure of PDZ3 (Protein Data Bank, PDB: 3I4W) suggested that Phe340/Leu342/Ile389 participated in oligomerization of protein monomers within the crystal structure. Hence, we selected four hydrophobic amino acids (Phe340, Leu342, Ile389, and Leu349) with alanine and designed four single-mutated PDZ3 variants (Fig. 1 C). Eventually, L342A yielded a mutant that denatured according to a two-state model. This finding shows that it is possible to avoid the formation of RO using single mutations that do not disturb the native state structure. It also validated our hypothesis that the RO state can be abolished by a single mutation of a hydrophobic residue at the oligomeric interface of the crystal structure.

Materials and Methods

DSSP and PDBePISA analysis

Accessible surface area (ASA) was calculated using x-ray crystallographic data of PSD95-PDZ3 (PDB: 3I4W) by DSSP (28). The four monomeric chains in each asymmetric unit cell form a crystallographic tetramer. BSA was determined by calculating the ASA of the tetrameric unit and subtracting the ASA of the monomeric structure (Tables 1 and S1).

Table 1.

ASA and BSA of Hydrophobic Residues of PDZ3-wt

	ASA(Å2)of monomer				BSA(Å2)of tetramer
Residue	A	B	C	D	A	B	C	D
Ile307	32	39	22	40	25	0	0	0
Ile314	1	1	1	3	0	0	0	0
Val315	50	52	50	53	0	0	0	3
Ile316	3	3	3	4	0	0	0	0
Leu323	13	13	12	14	0	5	0	0
Phe325	19	21	20	21	0	0	0	0
Ile327	21	23	19	23	0	2	7	0
Val328	34	35	33	33	0	33	31	0
Phe340	98∗	98∗	99∗	97∗	0	64∗	63∗	0
Ile341	41	35	41	38	0	0	0	0
Leu342	60∗	63∗	59∗	61∗	0	59∗	55∗	0
Leu349	143∗	127∗	136∗	127∗	0	0	0	0
Leu353	11	11	11	11	0	0	0	0
Leu360	23	19	20	20	20	0	0	18
Val362	1	1	1	1	0	0	0	0
Val365	74∗	75∗	75∗	76∗	0	0	0	0
Leu367	3	3	4	3	0	0	0	0
Ile377	73∗	75∗	71∗	73∗	0	12	0	0
Leu379	16	14	14	14	0	2	0	0
Ile389	50	48	49	48	46∗	0	0	45∗
Phe400	90∗	92∗	89∗	89∗	0	19	0	0

Residues with ASA > 0 Ų of at least one of the chains are listed. Hydrophobic residues (Ile, Leu, Val, and Phe) were selected using the hydropathy scale of Kyte and Doolittle (36). BSA values of the four polypeptide chains (A, B, C, and D) of PDZ3-wt (PDB: 3I4W) at the crystallographic tetramer were calculated by DSSP. Values of BSA were obtained by subtracting the ASA in the tetrameric structure from the ASA calculated from the monomer. The largest ASA and BSA are indicated with an asterisk, and selected residues are underlined.

Additionally, x-ray crystallographic data of PSD95-PDZ3 were also analyzed using PDBePISA (https://www.ebi.ac.uk/pdbe/pisa/) (29). BSA values differed between DSSP and PDBePISA because DSSP analyzes the tetramer’s interface, whereas PDBePISA analyzes the interface of dimers and calculates three parameters (ASA, BSA, and ΔⁱG). ΔⁱG (solvation free energy) calculated by PDBePISA indicates the solvation energy of the corresponding residue in kcal/M between the dissociated and the associated structures (Tables S2 and S3). Hydrogen bonds, salt bridges, and disulfide bonds are included in the calculation of ΔⁱG (−0.44, −0.15, and −4 kcal/mol per bond, respectively).

Synthesis, expression, and purification

Synthetic genes encoding PDZ3 were cloned into a pBAT4 vector, and single mutations were introduced by site-directed mutagenesis using a QuikChange protocol (Stratagene, San Diego, CA). All variants were overexpressed in Escherichia coli strain BL21 (DE3) with 1 L of lysogeny broth medium. Protein expression was induced by the addition of 0.2 mM isopropyl β-d-1-thiogalactopyranoside when the OD reached 0.6. After centrifugation, harvested cells were lysed in 50 mM Tris-HCl (pH 8.7) by ultrasonication. Next, the supernatant fraction of the cell lysate was acidified to pH 3 by adding several drops of 1 M HCl. Finally, proteins were purified by reverse-phase high performance liquid chromatography, lyophilized, and stored at −30°C until further use.

Mass spectrometry (matrix-assisted laser desorption/ionization-time of flight) measurements

The identities of the variants were confirmed by matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) with experimental errors less than 5 Da (Table S5). Mass spectrometry (MS) measurements were performed by using an Autoflex Speed MALDI-TOF mass spectrometer (Bruker, Billerica, MA). The matrix solution was prepared by dissolving 10 mg of sinapinic acid in 1 mL of a solution containing 300 μL of acetonitrile, 100 μL of 1% trifluoroacetic acid, and 600 μL of Milli-Q water (MQ). Samples for MALDI-TOF MS were prepared by mixing 1 μL of protein solution with 4 μL of the matrix solution. 1 μL of sample mixtures were spotted and air dried on the MALDI-TOF MS plate.

Differential scanning calorimetry measurements

Samples were prepared by dissolving lyophilized proteins in MQ and dialyzed for 18 h at 4°C in 50 mM potassium-phosphate buffer (pH 7.5) using a Spectra/Por 3 membrane (MWCO of 3.5 kDa) with one buffer exchange. After dialysis, the protein concentrations of samples were adjusted to 1 mg/mL, and samples were filtered with 0.20 μm membrane filters (Millex-GV; Millipore Sigma, Burlington, MA) to remove aggregates, followed by degassing of the samples. Protein concentrations and pHs of the samples were confirmed just before performing experiments.

DSC measurements were performed using a VP-DSC microcalorimeter (MicroCal, Malvern Panalytical, Malvern, UK) at a scan rate of 1.0°C/min in the temperature range of 20–100°C. Blank measurements were taken using 50 mM potassium-phosphate buffer (pH 7.5) several times before measurements. The reversibility of the thermal unfolding of the protein was checked by repeated scans of the same sample. The midpoint temperature (T_mid), the calorimetric enthalpy (ΔH_cal (T_mid)), and the van’t Hoff enthalpy (ΔH_{van’t Hoff} (T_mid)) were determined by analyzing the apparent heat capacity curves using a nonlinear, least-square fitting algorithm, DDCL3 (30,31), and assuming a linear temperature dependence of the heat capacity for the native and denatured states.

Circular dichroism measurements

Samples were prepared by dissolving lyophilized proteins in MQ, and protein concentrations of samples were adjusted to 0.5 mg/mL in 50 mM potassium-phosphate buffer (pH 7.5). Samples were filtered with 0.20-μm membrane filters (Millex-GV; MilliporeSigma) to remove aggregates. Protein concentrations and pHs of samples were confirmed just before performing the experiments.

Circular dichroism (CD) measurements were conducted using a JASCO J-820 spectropolarimeter (JASCO, Tokyo, Japan) at 0.2 mg/mL protein and 50 mM potassium-phosphate buffer (pH 7.5) using quartz cuvettes with a 2-mm optical path length. CD spectra were measured in the range of 200–260 nm. Thermal stability was measured at a protein concentration of 0.5 mg/mL in 50 mM potassium-phosphate buffer (pH 7.5), at a 1°C/min scan rate, and monitored between 20 and 90°C using the CD value at 220 nm. Melting temperature (T_m) and van’t Hoff enthalpy (ΔH_{van’t Hoff} (T_m)) were computed by means of least-squares fittings of experimental data with a two-state model using Origin 6.1 J.

Dynamic light scattering measurements

Samples were prepared by dissolving lyophilized proteins in MQ, and protein concentrations of samples were adjusted to 1 mg/mL in 50 mM potassium-phosphate buffer (pH 7.5). Samples were filtered with 0.20-μm membrane filters (Millex-GV; MilliporeSigma) to remove aggregates. Protein concentrations and pHs of samples were confirmed just before performing experiments.

Dynamic light scattering (DLS) measurements were performed by using a glass cuvette with a Zetasizer Nano (Nano S; Malvern Panalytical). Sample temperatures were increased from 25 to 60°C and then to 90°C before being cooled to 25°C to assess reversibility. The hydrodynamic radius (R_h) was calculated from size-volume graphs using the Stokes-Einstein equation.

Analytical ultracentrifugation measurements

Samples were prepared by dissolving lyophilized proteins in MQ, and protein concentrations of samples were adjusted to 1 mg/mL in 50 mM potassium-phosphate buffer (pH 7.5). Samples were filtered with 0.20-μm membrane filters (Millex-GV; MilliporeSigma) to remove aggregates. Protein concentrations and pHs of samples were confirmed just before performing experiments.

Sedimentation velocity experiments were carried out using an Optima XL-A analytical ultracentrifuge (Beckman Coulter, Brea, CA) with a four-hole An60Ti rotor at 25°C. Before centrifugation, samples were dialyzed overnight against 50 mM potassium-phosphate buffer (pH 7.5). Each sample was then transferred into a 12-mm double-sector Epon cell and centrifuged at a rotor speed of 50,000 rpm. Concentrations were monitored at 280 nm. Sedimentation velocity data were analyzed using the continuous distribution c(s) analysis module in the SEDFIT program (32). The range of sedimentation coefficients in which the main peak was present was integrated to obtain the weighted average sedimentation coefficient. The c(s) distribution was converted into c(M), a molar mass distribution. Solvent density, viscosity, and protein partial specific volumes were calculated using SEDNTERP software (33).

8-Anilino-1-naphthalenesulfonate fluorescence measurements

Samples for 8-anilino-1-naphthalenesulfonate (ANS) fluorescence measurements were prepared by dissolving lyophilized proteins in 50 mM potassium-phosphate buffer (pH 7.5) at a protein concentration of 1 mg/mL. The ANS stock solution was prepared by dissolving ANS powder (Sigma-Aldrich, St. Louis, MO) in 50 mM potassium-phosphate buffer (pH 7.5), and its concentration was determined by measuring the absorbance at 350 nm. All solutions were filtered with 0.20-μm membrane filters (Millex-GV; MilliporeSigma) to remove aggregates. The protein concentrations and pHs of the samples were confirmed just before performing experiments.

ANS fluorescence measurements were performed by using an FP-8500 (JASCO, Tokyo, Japan) fluorimeter. 200 μL of the samples were kept in quartz cuvette with 3-mm optical path length. ANS fluorescence of samples was measured by mixing the protein solution with ANS solution at a final concentration of 20 μM. The temperature of the samples was increased from 25 to 90°C. The emission spectrum was monitored with excitation at 380 nm.

Results

Two endothermic peaks in the wild-type PDZ3 thermogram

Although thermal denaturation monitored by CD exhibited a sigmoidal curve typical of a natively folded protein (Fig. 2; Table 3), the DSC thermogram of wild-type PDZ3 (PDZ3-wt) exhibited two endothermic peaks, as reported previously (23,24) (Fig. 3). The apparent T_m, defined by the maximum of the first endothermic peak, decreased when the protein concentration was increased (Fig. 3; Table S4). This dependency of the T_m on protein concentration is contrary to that observed for proteins that form oligomers in the native state (34,35) and imply the presence of an oligomeric intermediate state at high temperature. The second endothermic peak is attributed to the dissociation of the high-temperature oligomeric state into a monomeric, unfolded state.

Figure 2.

CD thermal denaturation curves of PDZ3 variants at 0.5 mg/mL and pH 7.5. Samples were heated at the scan rate of +1°C/min, and CD values were monitored at 220 nm. Black dots and gray lines represent the raw CD value and the fitting curves.

Table 3.

T_m and van’t Hoff Enthalpy (ΔH_{van’t Hoff} (T_m)) of PDZ3 Variants by Two-State Analysis in CD Measurement at 0.5 mg/mL, pH 7.5, and 1°C/min Scan Rate. These Thermodynamic Parameters Were Calculated by Fitting Thermal Denaturation Curves of PDZ3 Variants

Name	Tm (°C)	ΔHvan’t hoff (Tm) (kJ/mol)
PDZ3-wt	72.58	249.13
PDZ3-F340A	72.98	350.08
PDZ3-L342A	70.97	339.13
PDZ3-I389A	60.99	237.23
PDZ3-L349A	69.74	241.75

Figure 3.

Concentration dependence of DSC thermograms of PDZ3 variants at 0.5–1 mg/mL, pH 7.5, and 1°C/min scan rate. Black circle, gray circle, and white circle show DSC thermograms at protein concentrations of 1, 0.75, and 0.5 mg/mL, respectively.

Design of PDZ3 variants

We designed point mutations in PDZ3 intended to inhibit the formation of high-temperature reversible oligomers. Consistent with our design strategy developed using DEN4 ED3 (PDB: 3WE1), we searched for hydrophobic residues at the crystal interface and replaced them with alanine. To this end, we used the crystal structure of the PDZ3-wt, which contains a PDZ tetramer in the unit cell (Fig. 1; Tables 1 and S1). Based on our previous analysis using DEN4-ED3 (25), we assumed that high-temperature oligomers were generated by hydrophobic patches on the protein that would interact with the nearby protein. Hence, we focused on hydrophobic residues that exhibited the largest BSA, according to DSSP. BSA was determined by subtracting the ASA of the tetrameric unit from the ASA of the monomeric structure. Here, we took Ile, Leu, Val, and Phe as hydrophobic, as defined by the Kyte and Doolittle hydropathy scale (36), and we selected three hydrophobic residues, Phe340, Leu342, and Ile389, exhibiting the largest BSA (Tables 1 and S1).

The above design was confirmed using PDBePISA (Fig. S1; Table S2). Candidate residues were identified at the two largest interfaces “ID-1 (average 518.0 Ų)” and “ID-2 (average 449.6 Ų)” (Table S3) predicted by PDBePISA. According to PDBePISA, Ser371, Glu373, and Glu331 and Leu342, Ile389, and Phe340 are critical for maintaining interfaces ID1 and ID2, respectively. Phe340, Leu342, and Ile389 were the same residues selected using DSSP. Additionally, we chose, as a control, a hydrophobic residue (Leu349) with a large ASA but with a negligible BSA, meaning that this residue is not buried by the formation of the tetramer (Table S2).

Unaltered physicochemical properties at ambient temperature

The single alanine substitution on PDZ3 did not change the native structure nor its physicochemical properties at temperatures up to ∼50°C. In particular, all PDZ3 variants were monomers at 25°C, as shown by analytical ultracentrifugation (AUC) (Fig. S2; Table 2), and CD spectra of all variants were nearly identical at 25°C. (Fig. S3). Additionally, the thermal denaturation curve developed by monitoring CD at 220 nm showed a typical sigmoidal transition that could be fitted by a two-state model (N↔D) (Fig. 2; Table 3). However, PDZ3-I389A was 10°C less stable than the other variants, indicating that a single mutation could affect the thermal stability of PDZ3, as in our previous observations with bovine pancreatic trypsin inhibitor (37).

Table 2.

The Sedimentation Coefficient and Calculated Molecular Weight of PDZ3 Variants in AUC Measurements at 1 mg/mL, pH 7.5, and 25°C

Name	Weighted Average Sedimentation Coefficient (S)	Calculated molecular weight (kDa)
PDZ3-wt	1.52	10,590.1
PDZ3-F340A	1.51	10,459.3
PDZ3-L342A	1.49	10,490.9
PDZ3-I389A	1.48	10,495.6
PDZ3-L349A	1.53	9656.1

Sedimentation velocity analysis of PDZ3 variants was performed at 25°C, and experimental data were analyzed using SEDFIT (32) and SEDNTERP (33).

Inhibition of high-temperature RO by a single mutation

DSC indicated that our design for abolishing the RO state was successful. In particular, PDZ3-F340A and PDZ3-L342A at a 1 mg/mL concentration and pH 7.5 showed a single endothermic peak, and PDZ3-I389A showed a large peak overlapped with a small shoulder (Figs. 3 and S4; Table S4). The mutations changed the molar fraction of the intermediate oligomer (Figs. 4 and S5) as well as other thermodynamic properties related to the formation of the RO (Table S5). On the other hand, PDZ3-L349A, which is a surface-exposed hydrophobic residue located far from the crystal oligomeric interface, showed two endothermic peaks almost identical to PDZ3-wt.

Figure 4.

Molar fraction of PDZ3-wt (N-I₆-D model), PDZ3-F340A (N-I₄-D model), PDZ3-L342A (N-D model), PDZ3-I389A (N-I₅-D model), and PDZ3-L349A (N-I₅-D model) by DDCL3 analysis of DSC thermograms at 1 mg/mL, pH 7.5, and 1°C/min scan rate. Shown are natively folded monomer (blue), intermediate oligomer (green), and unfolded monomer (red).

The DSC results were corroborated by hydrodynamic radii measured by DLS size-volume analysis. DLS indicated that RO formation at high temperature was significantly inhibited for PDZ3-F340A and PDZ3-L342A (Fig. S6; Table 5), which showed single endothermic peaks. Their hydrodynamic radii, R_h, were almost unchanged by heating from 25 to 60°C (1.72 ± 0.11 and 1.85 ± 0.11 nm), in contrast to the R_h of PDZ3-wt, PDZ3-L349A, and PDZ3-I389A, which increased at 60°C to 2.35 ± 0.07, 4.06 ± 0.16, and 2.13 ± 0.13 nm, respectively (Table 5).

Table 5.

Hydrodynamic Radius (nm, R_h) of PDZ3 Variants by DLS Measurements at 1 mg/mL, pH 7.5, and 25–90°C

Name	25°C	60°C	90°C	25°C (after heating)
PDZ3-wt	1.78 ± 0.20	2.35 ± 0.07	1.76 ± 0.28	2.99 ± 0.11
PDZ3-F340A	1.67 ± 0.11	1.72 ± 0.11	2.05 ± 0.20	1.96 ± 0.08
PDZ3-L342A	1.79 ± 0.02	1.85 ± 0.11	1.33 ± 0.33	2.03 ± 0.06
PDZ3-I389A	1.76 ± 0.01	4.06 ± 0.16	1.72 ± 0.09	1.96 ± 0.09
PDZ3-L349A	1.73 ± 0.06	2.13 ± 0.13	2.22 ± 0.20	1.90 ± 0.02

R_h values were calculated from size-volume graphs at 25, 60, 90, and 25°C after heating. The errors are the standard deviation of three-times measurements with the same sample.

Thermodynamic parameters of PDZ3 and its variants

We performed a global fitting of the DSC thermograms measured at multiple protein concentrations to provide a detailed characterization of the thermal transition of PDZ3 variants. Both PDZ3-F340A and PDZ3-L342A showed a single endothermic peak. However, global fitting indicated that only the latter followed a perfect two-state denaturation model (N↔D); the ratio of the calorimetric and van’t Hoff enthalpy of PDZ3-L342A was essentially 1.0, indicating the absence of an intermediate state (Tables 3 and S4). Additionally, the residuals of global fittings were almost identical between the two-state (N↔D) and three-state models (N↔I_n↔D) (Table 4). Hence, the thermal transition of PDZ3-L342A was perfectly described by a two-state model (Table 4); thus, the formation of the high-temperature RO was fully inhibited.

Table 4.

Residuals of PDZ3 Variants between DSC Raw Data and Fitting Curves Calculated with DDCL3

Name	Model	Residuals (kJ mol−1 K−1)	Residuals (μW)
PDZ3-wta	N-I3-D	0.824	0.615
	N-I4-D	0.695	0.519
	N-I5-D	0.63	0.470
	N-I6-D	0.615	0.459
	N-I7-D	0.631	0.471
PDZ3-F340A	N-D	0.684	0.514
	N-I-D	0.292	0.219
	N-I2-D	0.257	0.193
	N-I3-D	0.235	0.177
	N-I4-D	0.225	0.169
	N-I5-D	0.252	0.189
PDZ3-L342A	N-D	0.339	0.248
	N-I-D	0.420	0.307
	N-I2-D	0.419	0.306
	N-I3-D	0.446	0.326
	N-I4-D	0.449	0.328
	N-I5-D	0.449	0.328
PDZ3-I389A	N-I3-D	0.349	0.261
	N-I4-D	0.287	0.215
	N-I5-D	0.267	0.200
	N-I6-D	0.274	0.205
PDZ3-L349A	N-I4-D	0.456	0.342
	N-I5-D	0.412	0.309
	N-I6-D	0.42	0.315

The oligomer number of the intermediate (I) state varied from monomer (I) to heptamer (I₇) at the thermal transition models. Optimal thermal transition models of each PDZ3 variant were determined by comparing the minimal residuals.

^aThe residuals of the global fitting suggest that PDZ3-wt forms a 5 to 7 mer in the RO state. This result does not fully coincide with the previously reported trimeric intermediate state (I₃) (23), but this minor discrepancy may be due to small differences in experimental settings and does not interfere with the final outcome of the study. Namely, a single mutation (L342A) at the crystallographic interface of PDZ3-wt abolishes the formation of the high-temperature RO state.

Although the DSC thermogram of PDZ3-F340A showed a single endothermic peak, the global fitting analysis indicated a three-state (N↔I_n↔D) denaturation. Indeed, the residuals of the global fitting drastically decreased using a three-state model, but the molar fraction of the intermediate state (I_n) was very small. Thus, F340A strongly but not fully inhibited the formation of the high-temperature RO (Figs. 4 and S5; Table S4).

PDZ3-I389A and PDZ3-L342A obviously could not be fitted using a two-state model because two endothermic peaks appeared in their DSC thermograms. The second endothermic peak of PDZ3-I389A was slightly smaller than PDZ3-wt, whereas that of PDZ3-L349A was almost identical with PDZ3-wt (Fig. S5; Table S4). Thus, I389A and L349A did not significantly inhibit the formation of high-temperature ROs.

For the purpose of discussion, we note that the apparent T_m of the wild-type determined by CD (Table 3) is higher than the T_mid measured by DSC (Table 6), but they are approximately coincident for the L342A variant. We can rationalize the above difference by assuming that the secondary structure content is conserved in the RO state because CD denaturation curve monitors the unfolding of the secondary structure content. According to this hypothesis, the T_m measured by CD should be equal to the T_mid of the sum fraction of the native and RO states determined by DSC. Note that the T_mid of the sum fraction of the native and RO states is equal to that of the fraction of the denatured state because we have a three-state denaturation system. At 0.5 mg/mL, which is the concentration at which we have data for both CD and DSC, we found that the T_mid of the PDZ3-wt is 72.8°C (calculated using the DSC fractions reported in supplemental S5; PDZ3-wt (0.5mg/mL)). The apparent T_m observed by CD was 72.6°C (Table 3), and the two values are thus in good agreement. Similarly, the midpoint and the T_m of the other mutants did also agree with an experimental error of ∼1°C, which suggests that the secondary structure content in the RO state is close to that in the native state.

Table 6.

The T_mid Calculated by DSC Measurements at 0.5 mg/mL of Protein Concentration to Confirm Whether the Secondary Structure Content Is Conserved in the RO State

Name	Tmid (N↔RO+D)	Tmid (N+RO↔D)
PDZ3-wt	68.53	72.60
PDZ3-F340A	70.16	70.71
PDZ3-L342A	69.58	69.88
PDZ3-I389A	57.33	61.05
PDZ3-L349A	66.19	70.93

T_mid (N↔RO+D) is defined as the temperature in which the sum fraction of the denatured state and the RO state is equal to the fraction of the native state. Similarly, T_mid (N+RO↔D) is defined as the temperature in which the sum fraction of the native state and the RO state is equal to the fraction of the denatured state.

To further probe the conformational nature of the RO state, we measured the ANS fluorescence (Fig. S7). As a result, the ANS fluorescence intensities of PDZ3-F340A and PDZ3-L342A (which do not form RO), at 80°C, were much lower than that of the PDZ3-wt (which forms high-temperature ROs). ANS binds to MGs but also to partially solvent-exposed hydrophobic patches or cavities of oligomerized/aggregated proteins (including RO) as we reported for the I₂ state of Che Y proteins (21). Thus, taken together with the CD results, ANS experiments suggest that the RO state is in a MG-like state (CD suggests that the secondary structure content is conserved) or at least that there are hydrophobic patches/pockets that are accessible to ANS binding, in line with our previously reported ANS measurement of PDZ3-wt (23,38).

Discussion

PSD95-PDZ3′s DSC thermogram exhibits two endothermic peaks, which is atypical of the denaturation curve observed in single-domain, monomeric, globular proteins. Such an observation is rare, but it was reported for another small, single-domain protein, DEN4 ED3, in which the second endothermic peak resulted from high-temperature RO (25). Here, our DSC analysis showed that PDZ3 undergoes thermal denaturation with an oligomeric (probably 5 to 7 mer, according to the residuals reported in Table 4), high-temperature, RO intermediate state (N↔I_n↔D; n = 5∼7). To date, reversibility is unusual when denatured proteins oligomerize because this often triggers irreversible aggregation followed by precipitation, exhibiting a strong exothermic peak during DSC measurement. In addition, high-temperature RO has rarely been reported, precisely because it occurs at high temperatures, which we speculate might make it difficult to observe.

For the purpose of discussion, let us note that it is counterintuitive that a mutation designed using the crystal structure, specifically the replacement of a hydrophobic residue at a crystallographic interface, would drastically affect the formation of unfolded oligomers at high temperature. However, this can be rationalized by assuming that the high-temperature RO occurs through oligomerization of folded proteins, either in a native state or, more likely, in an MG-like state, in which the hydrophobic core is conserved but with highly dynamic side chains (11, 12, 13, 14, 15, 16, 17, 18, 19, 20). Thus, although further studies are required to fully confirm this hypothesis, we speculate that PDZ3-wt oligomerizes in a folded state, in which the interactions observed in the crystal structure prevail (either native or MG-like). Unfolding would occur after oligomerization as confirmed by comparing CD and DSC data (Figs. 2 and 3), which increases the destabilizing effect through reversed hydrophobicity (39), becoming stronger at high temperatures (40), as we showed for oligomerization of tagged bovine pancreatic trypsin inhibitor (41). This interpretation is corroborated by our present design of PDZ3-F340A and PDZ3-L342A, in which alanine substitution of hydrophobic residues at the crystallographic oligomer interface inhibited RO formation, but not PDZ3-L349A, which was surface exposed but located far from the oligomerization site. Furthermore, a recent study revealed that the Ebola virus matrix protein VP40 also shows a DSC thermogram with two endothermic peaks. In addition, a single mutation (L117R) in the hydrophobic region of the N-terminal domain of VP40 at the dimeric crystallographic interface surprisingly suppressed the second endothermic peak (42). This observation also corroborates our present finding that the hydrophobic interaction at the interface of the natively folded oligomer actively contributes to the formation of high-temperature RO.

Finally, let us note that single mutations in recombinant proteins that drastically alter their physicochemical properties are rare because multiple interdependent factors (compensation of enthalpy and entropy (43,44), hydrogen bonds (45,46), salt bridges (47,48), and disulfide bonds (43,49)) are usually involved. Mutations often impair the original structure or function, with a few exceptions (50,51). Thus, a single alanine substitution fully inhibiting the formation of high temperature in PDZ3 is surprising and may be because interaction energies stabilizing the RO state are weak.

Conclusion

Although PSD95-PDZ3-wt is a monomeric single-domain protein, it shows a peculiar thermal denaturation curve with two endothermic peaks, as observed by DSC. Such observation is rare, but this report is the second example, along with our previous report of DEN4 ED3. Hence, RO formation might be a phenomenon common to many small globular proteins but was overlooked so far because it occurs at high temperature.

Finally, our strategy for inhibiting RO using the crystal structure of PDZ3 indicated that the substitution of a single hydrophobic residue to an alanine, L342A, of a residue located at the monomeric surface fully inhibited RO formation at high temperature. On the other hand, a control substitution replacing Leu349, an exposed hydrophobic residue far from the oligomerization site, indeed did not inhibit RO formation, indicating that RO formation must be associated with hydrophobic residues at the interface of the crystallographic oligomers. Moreover, RO formation at high temperature may be a precursor of aggregation, and our design strategy may thus have practical usefulness in designing aggregation-resistant proteins.

Author Contributions

T.S., S.-i.K., and Y.K. designed the study and wrote the manuscript. T.S., T.M., and S.-i.K. carried out the DSC analysis. T.S. and S.B. prepared the samples to perform the spectroscopic analysis. J.C.M. provided the materials and co-wrote the manuscript. S.U. carried out the AUC analysis. All authors contributed to the editing of the manuscript and approved the final version.

Editor: John Correia.