The order of PDZ3 and TrpCage in fusion chimeras determines their properties—a biophysical characterization

Abstract

Most of the structural proteins known today are composed of domains that carry their own functions while keeping their structural properties. It is supposed that such domains, when taken out of the context of the whole protein, can retain their original structure and function to a certain extent. Information on the specific functional and structural characteristics of individual domains in a new context of artificial fusion proteins may help to reveal the rules of internal and external domain communication. Moreover, this could also help explain the mechanism of such communication and address how the mutual allosteric effect plays a role in a such multi‐domain protein system. The simple model system of the two‐domain fusion protein investigated in this work consisted of a well‐folded PDZ3 domain and an artificially designed small protein domain called Tryptophan Cage (TrpCage). Two fusion proteins with swapped domain order were designed to study their structural and functional features as well as their biophysical properties. The proteins composed of PDZ3 and TrpCage, both identical in amino acid sequence but different in composition (PDZ3‐TrpCage, TrpCage‐PDZ3), were studied using circualr dichroism (CD) spectrometry, analytical ultracentrifugation, and molecular dynamic simulations. The biophysical analysis uncovered different structural and denaturation properties of both studied proteins, revealing their different unfolding pathways and dynamics.

1. INTRODUCTION

A central paradigm of structural biology is the concept of the order of protein domains. A domain can be identified and assigned to a specific functional or structural property. Most proteins consist of two or more domains from a limited group of domain families. Domains are often rearranged in various combinations through gene fusion events to evolve new protein functions, including the use of protein intra‐ and/or inter‐ allosteric modulation through the incorporation of various regulatory domains. Despite the fact that there are a large but limited number of protein domains, countless combinations of native proteins could be built using this limited number of protein domains. ¹ , ² Therefore, the modular organization of domains is a credible evolutionary concept currently being explored and proven by bioinformatics methods ³ which has resulted in databases of specific protein domains. ⁴ , ⁵ , ⁶ Our previous computational analysis ⁷ revealed different properties of a domain in its multidomain context. Based on that study, we proposed that the function of the individual domain can be dependent on the multi‐domain protein context.

The acronym PDZ domains (PDZs) was derived from the names of the first three proteins discovered that contained the PDZ fold: the postsynaptic density protein 95 (PSD‐95), the septate junction protein Discs‐large, and the tight junction protein Zonulla occludens‐1 protein (ZO‐1). ⁸ , ⁹ PDZs are widely studied protein anchor domains present in approximately 380 different proteins in the human proteome. ¹⁰ PDZs are typical examples of so‐called dynamic allostery without conformational changes, where the dynamics of the distal side chain is modulated by a ligand binding, and the origin has been attributed to entropic effects. These domains typically bind short C‐termini of various binding partners (receptors, intracellular proteins, etc.) and act as a scaffold domain mediating the formation of intracellular multi‐protein complexes. ¹¹ , ¹² , ¹³ , ¹⁴ PDZ domains possess a very broad specificity for various substrates, ¹⁵ and the specificity of the PDZ‐ligand binding pocket has been shown to be modulated via intramolecular allosteric networks. ¹⁶ , ¹⁷ , ¹⁸ Therefore, it could be expected that the dynamics of PDZ is highly dependent on the domains in its vicinity. The human PDZ3 domain from ZO‐1 serves as an anchor in cells for cytoskeletal and trans‐membrane proteins. We selected this small protein for its single domain character with known function to bind JAM‐A ligand by PDZ3 specific binding site. ¹⁹ , ²⁰ , ²¹ The interaction of PDZ3 and JAM‐A modulates the permeability of tight junctions for immune cells and ions. Structurally, the PDZ3 is composed of six β‐strands forming a strong hydrophobic core that is complemented with 3 α‐helices around the β‐strand core. The domain‐recognition site is a cavity flanked by a β‐strand on one side, and an α‐helix on the other side. ¹⁹

The Tryptophan Cage (TrpCage) mini‐protein was used as the second domain in this study. TrpCage is a 20‐residue synthetic miniprotein that exhibits a high‐folding rate (4 μs) and high‐structure stability without a native binding partner. ²² These traits make TrpCage an optimal model system for protein folding and dynamic studies. ²³ , ²⁴ , ²⁵ , ²⁶ The TrpCage structure consists of a short α‐helix, 3₁₀‐helix, and C‐terminal polyproline‐II helix, and its structure is stabilized through interactions with a central tryptophane residue (Trp6) as a key residue for protein folding. ²² , ²⁷ , ²⁸ TrpCage has been a promising simple protein model for studies of protein dynamics, and now we have used this domain for more complex study in the target of fusion domain protein. ²⁹

To describe how the selected domain behaves out of its native context when fused with another protein domain, we proposed an artificial protein system model to study this phenomenon. To maintain the simplicity, we decided to create a system joining PDZ3 and TrpCage domains in a single fusion chimera. This simple model systems were chosen to determine how the order of different domains implies multiple arrangements which can affect overall macromolecular characteristics, functions, and the denaturation properties of a single‐chain chimera. We have already published a computational study on the influence of several artificial protein domains on PDZ3 studied by molecular dynamics in a fusion protein system. ⁷ In this work, we aim to describe biophysical and denaturation properties of two chimeras with different order of composing domains. Such information can help to reveal the principles of inter‐ or intra‐domain communication, and the mechanism of mutual structure or function modulations. We hypothesize that the allosteric modulation of the PDZ function could be demonstrated by its changed ligand specificity brought by neighboring domains. The selection of a functional domain or part of a protein combined with other protein domains is an unstoppable trend for future protein technologies, with applications in medicine, biomaterials, and biotechnology.

2. METHODS

2.1. Protein domains selection and design of fusion protein systems

The first selected domain, the PDZ3, is the naturally occurring central domain of the ZO‐1 protein located in the PDZ3‐SH3‐GUK (PSG) domain tandem (UniProtKB Q07157; PDB 3TSZ) ¹⁹ (Figure 1). Well‐characterized structural and limited functional characteristics (one binding site for native binding partner JAM‐A) ¹⁹ of the PDZ3 domain allowed us to focus on the most important features (Figure 2(a)). The second artificial domain, TrpCage (PDB 1L2Y), ²³ was chosen for two reasons, (i) TrpCage as an artificial domain has been extensively studied as a simple model of protein folding, stability and dynamics, ³⁰ , ³¹ , ³² (ii) the TrpCage is not expected to be recognized by host cells and degraded by the cellular proteolytic apparatus in a planned cellular assay. TrpCage is stabilized mainly through the interaction of the Trp residue with the poly‐proline motif (Figure 2(b)). ²² , ³² Most importantly, the TrpCage contains α‐helix ‐ capable of mimicking the role of a native α‐helical linker between PDZ3 and SH3 in ZO‐1, apparently an important element of the PDZ3 substrate specificity. ³³ The choice of linkers is an important step, not only for their influence on the behavior of both fused domains, but also for the recombinant expression of the fusion proteins (unexpected properties during an expression in the particular expression system). The design of experimental linkers is a crucial step regarding specific influence on a domain dynamic. ³⁴ , ³⁵ , ³⁶ For both chimeras, we chose the flexible linker GGGGGG (Gly linker) which is mostly used to construct recombinant fusion proteins to provide a high‐conformational flexibility for the fusion protein domains to maintain their favorable interatomic distances.

FIGURE 1.

PDZ3 present in MAGUK tandem of ZO‐1. ZO‐1 protein consists of three PDZ domains (PDZ1 and PDZ2 dark pink, PDZ3 light pink), SH3 domain (red) and GUK domain (green), PDB 3TSZ. PDZ3 domain occurs as a first member of PDZ3‐SH3‐GUK supramodule. PDZ3 and SH3 domains are linked by an α‐helical linker, which enhances the binding affinity of PDZ3 to the active binding partners

FIGURE 2.

FD3A and FD4A design and biochemical characterization. (a) PDZ3 (pink) from ZO‐1 structure (PDB:3SHU). The core of the domain consists of six β‐sheet strands (βA–βE) surrounded by two α‐helices (αA–αB). (b) TrpCage (cyan) structure (PDB:1L2Y) composed of α‐helix and poly‐proline helix which keeps the structure compact. (c) Schematic representation of FD3A and FD4A fusion proteins design. PDZ3 domain (pink) was fused with α‐helical mini‐protein TrpCage (cyan) in forward (FD3A: PZ3‐TrpCage, labeled by red) and reverse (FD4A: TrpCage‐PDZ3, labeled by blue) orders. The domains were fused by a flexible linker (gray) in order to keep domain structure individuality and promote a possible interdomain communication. (d) The size exclusion chromatography (SEC) profile of FD3A (red) and FD4A (blue) indicates slightly lower compactness of FD4A compared to FD3A. (e) Dynamic light scattering (DLS) spectra of FD3A (red) and FD4A (blue) confirmed the same trend of distinct hydrodynamic radii between FD3A and FD4A. (f) Comparison of DLS estimated hydrodynamic radii (Rh), polydispersity indexes (PDI) and predicted molecular weight (M_W) of FD3A and FD4A. The measurements were performed 20x for each sample, and final values represent the mean value with SD

2.2. Protein constructs preparation

The cDNAs of the proposed fusion domains of PDZ3‐GGGGGG‐TrpCage (internally assigned code: FD3A) and TrpCage‐GGGGGG‐PDZ3 (internally assigned code: FD4A) were synthesized by the GenScript company (Piscataway, New Jersey, United States), and subcloned into a pET15b expression vector. The PDZ3 alone was prepared from a FD3A construct by adding a stop codon at the appropriate sequence position.

2.3. Protein expression and purification

The FD3A, FD4A, and PDZ3 proteins were produced as recombinant proteins in an Escherichia coli BL21(DE3) strain and purified by two‐step chromatography techniques as described previously. ⁷ The final buffer for all produced proteins had a composition of 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.4 mM KH₂PO₄, 0.02% NaN₃, 1 mM EDTA (pH 7.4). The purity of the proteins was verified by SDS‐PAGE analysis, and predicted Mw and amino acid composition was verified by mass spectrometry. M_W of FD3A and FD4A was determined as 12.8 kDa. PDZ3 individually has 10.3 kDa. Proteins were stored at −80°C.

2.4. Peptide synthesis

TrpCage peptide was synthesized by solid‐phase peptide synthesis according to the N‐α‐Fmoc protocol in our previous publication. ³⁷

2.5. Analytical size exclusion chromatography

Analytical size exclusion chromatography (ASEC) of FD3A and FD4A proteins was performed on Akta Pure FPLC system (GE Healthcare, Uppsala, Sweden) with a Superdex 75 Increase 10/300 GL column (GE Healthcare, Uppsala, Sweden). The column was equilibrated with buffer containing the following: 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.4 mM KH₂PO₄, 0.02% NaN₃, 1 mM EDTA (pH 7.4) and calibrated with Gel Filtration Standard (Bio‐Rad, CA). The concentration of both analysed proteins was 2 mg/ml. The whole experiment was carried out at a flow rate of 1 ml/min.

2.6. Analytical ultracentrifugation

Sedimentation analyses were performed using an analytical ultracentrifuge ProteomeLab XL‐I (Beckman Coulter, Indianapolis, IN). Protein samples were dialyzed against a buffer containing 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.4 mM KH₂PO₄, 0.02% NaN₃ and 1 mM EDTA, pH 7.4 before the analysis. Sedimentation velocity experiments were conducted at 50000 rpm and 20°C using double sector cells and an An50‐Ti rotor at various protein‐loading concentrations. In total, 100 absorbance scans were recorded at 280 nm at 10 min interval. Buffer density, protein partial specific volumes, particle dimensions and the values of s_20,w sedimentation coefficients corrected to standard conditions (corrected to 20°C and the density of water) were estimated in Sednterp. ³⁸ Data were analysed using Sedfit ³⁹ using the c(s) continuous sedimentation coefficient distribution model, and the figures were prepared in GUSSI. ⁴⁰

2.7. Dynamic light scattering

The dynamic light scattering (DLS) of the FD3A and FD4A proteins was measured at a concentration of 1.0 mg/ml in the phosphate buffer containing 10 mM Na₂HPO₄, 1.4 mM KH₂PO₄, 137 mM NaCl, 2.7 mM KCl, 0.02% NaN₃, pH 7.4. Before analysis, the samples were filtered using 0.1 μm Ultrafree‐MC centrifugation filters (Millipore, Burlington, MA) to completely remove dust particle. The measurements were performed at 20°C using a Zetasizer Nano ZS instrument (Malvern Instruments, Malvern, UK) equipped with an internal 633 nm He‐Ne laser and at a given angle of 173°. Proteins were measured in a 3 × 3 mm quartz cuvette with an internal volume of 40 μl (Hellma Analytics, Müllheim, Germany). The reported values of the hydrodynamic radii (Rh) and polydispersity index (PDI) were processed using the original Zetasizer 6.2 Malvern Instruments software (Malvern Instruments, Malvern, UK). The accuracy of values was estimated as the standard errors of the mean (SD). The obtained Rh values were compared to predicted Rh values for a folded globular protein. ⁴¹ For each sample, 20 measurements were recorded.

2.8. Electronic circular dichroism

Electronic circular dichroism (ECD) measurements were performed on a Jasco‐815 spectropolarimeter equipped with a Peltier type temperature control system PTC‐423S/L (Jasco Corporation, Tokyo, Japan). The spectra were recorded in far‐UV (198–260 nm) and in the near‐UV (240–350 nm) spectral region using the following experimental setup: 0.02 mm rectangular quartz cell with the standard instrument sensitivity and 1 nm bandwidth, a scanning speed of 10 nm/min, a response time of 8 s, one accumulation and 0.1 mm rectangular quartz cell, standard instrument sensitivity, 1 nm bandwidth, a scanning speed of 10 nm/min, a response time of 8 s, one accumulation for far‐UV and near UV spectral region, respectively. The samples were kept in the phosphate buffer–137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.4 mM KH₂PO₄, 0.02% NaN₃ and 1 mM EDTA (pH 7.4) at a concentration of 0.05 mg/ml. The thermal unfolding was measured in a temperature range from 5 to 75°C, with a temperature increment of 5°C. The thermal unfolding was measured as the dependence of CD at 224 nm in the temperature range from 5 to 95°C with a slope of 15°C per hour and response time 2 s, in the same cell as previous measurements. The chemical unfolding was studied by urea in a 0–6 M concentration range with a concentration increment 0.5 M. After baseline subtraction, the final data were expressed as molar ellipticities ϴ (deg·cm²·dmol⁻¹) per residue. The numerical analysis of secondary structures was performed using the CDPro software package. ⁴² , ⁴³ , ⁴⁴ The determination of protein unfolding as a function of temperature or urea concentration was expressed as the dependence of molar ellipticities at 222 nm of far UV and 282 nm in near‐UV spectral region on temperature or urea concentration, and data were treated using the program Sigmaplot 12.5 (Systat software, San Jose, CA), according to the literature. ⁴⁵

2.9. Molecular modeling

The state‐of‐the‐art AIDA and ClusPro online servers were used to generate structural models. ⁴⁶ , ⁴⁷ In the case of the AIDA tool, the domains were taken as rigid, and their structures, together with the linker bridging both domains, were used as input. The linker is considered as a flexible part of the fusion protein, and there are no conformational constraints regarding its flexibility. The ClusPro primarily targets on a mutual orientation of both domains in an interaction and does not include any parameter for the linker. Based on the ClusPro predicted distances between the C‐terminus of the first and N‐terminus of the second domain, the linker could then be modeled to connect both domains. Analysis of the PDZ3‐SH3 interface (PDB: 3TSZ) was used to help predict the mutual interaction between the TrpCage and PDZ3 by both AIDA and ClusPro tools. To summarize, one model from AIDA was chosen for the FD3A chimera, whereas in the case of FD4A, two possible variations for orientation of composed domains were selected, the “open” form from AIDA and the “compact” from ClusPro.

2.10. Molecular dynamic simulations (MDs)

The obtained models of FD3A and FD4A constructs were all prepared for MDs and simulated by the same protocol. Preparation and MDs were run using a GROMACS 5.1.2 molecular dynamics package. ⁴⁸ The PDB:2GMX module was used to prepare the system and to add hydrogens. Proteins were treated with AMBER03 force field ⁴⁹ and explicit solvent—TIP3P water model ⁵⁰ —in a dodecahedron box. The ionic strength was set to 150 mM using sodium and chloride ions. Periodic boundary conditions were applied. Preparation of the initial system was followed by a system energy minimization. The solvent molecules were relaxed around the protein using the steepest descent algorithm. The minimized system was then equilibrated for 100 ps in the NVT ensemble with a simulation time step of 2 fs to allow the temperature to reach a plateau at 300 K. The pressure was equilibrated for 100 ps in the NPT ensemble. The velocities were generated randomly using a Maxwell‐Boltzmann distribution at 300 K with 1 atm pressure. Each structure was simulated with a production run 200 ns long. The coordinates were extracted from the trajectory every 1 ps during simulations. The leap‐frog dynamics integrator was used to integrate the equation of motion. The LINCS algorithm was used to maintain the bond parameters constrained. ⁵¹ The Particle Mesh Ewald method was used to calculate long‐range electrostatic interactions. ⁵² A short‐range van der Waals cut‐off was set by a Verlet algorithm. ⁵³ Temperature and pressure were controlled by a Berendsen thermostat and Parrinello‐Rahman algorithm, respectively. ⁵⁴

3. RESULTS AND DISCUSSION

3.1. Design and preparation of the PDZ3‐TrpCage and TrpCage‐PDZ3 chimeric proteins

The FD3A (PDZ3‐TrpCage) and FD4A (TrpCage‐PDZ3) chimeras were designed as a swapped variation of both domains (forward and reverse), linked identically by the poly‐Gly (GGGGGG) linker. The forward order of the components in the proposed fusion proteins was PDZ3 at the N‐terminus, TrpCage at the C‐terminus, and a Gly linker connecting the domains (Figure 2(c)). The reverse order in the swapped arrangement was TrpCage at the N‐terminus, PDZ3 domain at the C‐terminus, again connected by the Gly linker. The constructs prepared by recombinant expression (see Section 2) were experimentally characterized by a wide range of biochemical, biophysical, and structural methods. In order to study single domain properties out of the chimera constructs, the individual PDZ3 and TrpCage domains were also prepared and used for comparison during the whole experimental characterization. From the amino acid sequences, the molecular weights (M_W) of FD3A and FD4A were expected to be 13 kDa, for the individual PDZ3 domain 10.3 kDa, and for the TrpCage 2.7 kDa. To obtain folded proteins for characterization and denaturation studies, the FD3A and FD4A chimeras were expressed in E. coli, denatured, and refolded during the purification process (for more details, see Section 2). Besides our own results, there are numerous studies of both domains on their structure and function. ¹⁹ , ²² , ³³

3.2. The biochemical and biophysical characterizations of FD3A and FD4A

The FD3A and FD4A constructs were designed to study macromolecular properties of the recombinantly expressed and purified two‐domain chimeras to determine their differences. In addition, we studied and described the molecular size, shape, and secondary/tertiary structure content of FD3A and FD4A as well as the thermostability of the FD3A, FD4A, and PDZ3 and TrpCage domains. All measurements described in this work were performed at physiological pH (7.4). The ASEC was used as the first analytical method to verify the globular shape of the proteins and/or their potential dynamic character (Figure 2(d)). The analysis revealed slightly different molecular characteristics of FD3A and FD4A, documented by the shift of the chromatogram curves of the eluted proteins.

DLS experiments provided data to determine the hydrodynamic radius (Rh) and potential PDI of the studied proteins (Figure 2(e),(f)). The PDI value clearly indicated monodisperse solutions of the FD3A and FD4A proteins, with the highest value of 0.248 still being within a normal range for monomeric protein samples. Importantly, the estimated Rh of FD3A and FD4A was close to the predicted value for a folded globular protein of the same length (2.0 nm). ⁴¹ The FD3A and FD4A manifested Rh factors for a folded state of protein that suggested their native character. ⁴¹ DLS measurements confirmed the major monodisperse population with an Rh factor around 2 nm that corresponded to the folded states of both fusion proteins. A slight difference in their size (see calculated M_w values, Figure 2(f)) was identified, implying a distinct structure‐dynamics character. To obtain a more accurate estimate of FD3A and FD4A molecular properties, we performed a sedimentation analysis in an analytical ultracentrifugation (AUC) method.

AUC measurements showed that both FD3A and FD4A are mostly monomeric in solution, based on their c(s) sedimentation coefficients distributions (Figure 3(a),(b)). A series of c(s) distributions of FD3A showed a single peak over the whole range of loading concentrations (6–120 μM), indicating the presence of a monomeric form only with a weight‐average s_20,w value of 1.44 (± 0.014) S and an estimated M_W of 12.3 kDa. The main sedimenting species of FD4A with a weight‐average s_20,w value of 1.50 (± 0.019) S and an estimated M_W of 13.1 kDa corresponds to its monomeric forms as well. However, the c(s) distribution of FDA4 obtained for samples at 30–120 μM protein concentration showed an additional peak at higher S value, most likely corresponding to the FD4A dimer (Figure 3(a), right panel), with the fitted s_20,w value of 2.49 (± 0.087) S and the estimated M_W of 28.3 kDa, matching well the values expected for a dimeric state of FD4A. Sedimentation velocity reflects both the molecular weight and shape of a given molecule. Thus, knowing its molecular weight and sedimentation coefficient, it is possible to estimate its dimensions and degree of asymmetry, too. These parameters are estimated from the ratio of the frictional coefficients f/f₀, where f denotes a friction coefficient of the observed particle, and f₀ is the friction coefficient of a sphere of the same weight as the observed particle. Both FD3A and FD4A sedimented with a very similar overall f/f₀ ratio of 1.27 (± 0.008) and 1.28 (± 0.016), respectively (Figure 3(b)). This ratio corresponds to approximate molecular dimensions (including hydration shell) of 2–3 × 5–7 nm for a prolate ellipsoid model. The Stokes hydrodynamic radius Rh derived from the AUC sedimentation data is 2.03 (± 0.020) nm for the FD3A monomer, in perfect agreement with DLS (Figure 2(f)). For FD4A, the Rh values for its monomer and dimer species, 1.95 (± 0.025) nm and 2.34 (± 0.084) nm, respectively, again match very well the average value previously observed by DLS. It is to be expected that fusion proteins, combining two different unrelated protein domains fused by a flexible peptide linker, would exist in a number of conformations in solution, differing in mutual orientation of both domains. Although the fact that the c(s) distributions of FD4A, when compared to FD3A, show broader peaks over the whole concentration range probably reflects the FD4A monomer‐dimer equilibrium (Figure 3(a)), it might also partially reflect a difference in shape and conformation of these molecules, with FD3A being a more conformationally homogenous particle. While the FD3A protein seems to have a more rigid molecular character suggesting a potential PDZ3 and TrpCage domain contact, the FD4A protein c(s) distributions indicate a more flexible or dynamic molecular character. To partially determine the dynamics and potential different unfolding characteristics of FD3A and FD4A, we further investigated their thermal and chemical denaturation by CD spectroscopy.

FIGURE 3.

FD3A and FD4A are mostly monomeric in solution. (a) Overlay of the fitted c(s) continuous size distributions of the sedimenting species for FD3A (left panel) and FD4A (right panel at concentrations ranging from 6 to 120 μM. While FD3A is a monomer in the whole concentration range tested, FD4A showed a monomer‐dimer equilibrium from 30 μM concentration. (b) Summary of the sedimentation coefficients (s_20,w), frictional ratios (f/f₀), Stokes hydrodynamic radii (Rh), and molecular weights (M_W; estimated from the fitted s_20,w and f/f₀ values) obtained from AUC data for FD3A and FD4A chimeras

The CD spectra of FD3A and FD4A showed distinct characteristic negative maxima at ~207 nm and ~ 224 nm of comparable intensity, clearly indicating the predominant α‐helical conformation of their secondary structure at room temperature (Figure 4(a)). This observation was confirmed by numerical data analysis, which showed the major part of the α‐helical structure in the chimeras (Figure S1). The CD spectroscopy characterization of the individual PDZ3 and TrpCage domains are in the agreement with previous results (Figure 4(a)). ⁷ , ⁵⁵ The CD spectra in the near‐UV spectral region of FD3A and FD4A had a similar spectral pattern with positive maxima at ~251 nm, at ~265 nm, at ~277 nm, and at ~282 nm, which is characteristic for aromatic amino acids present in the primary structure of FD3A and FD4A, respectively (Figure 4(b)). ⁵⁶ The slightly different spectral intensity of the CD spectrum in the near‐UV spectral region observed for FD3A and FD4A may indicate a small difference in tertiary structure conformation. Thus, according to the CD data, the secondary structure of the FD3A and FD4A chimeras appears to be very similar, and the only significant difference observed relates to their tertiary structures.

FIGURE 4.

Thermal unfolding of FD3A and FD4A fusion proteins. (a) Far‐UV CD spectra of FD3A, FD4A, PDZ3, and TrpCage expressed as molar ellipticity as a function of wavelength. (b) Near‐UV spectra of FD3A and FD4A expressed as molar ellipticity as a function of wavelength. (c) The thermal unfolding of FD3A (red) and FD4A (blue) fusion proteins and PDZ3 (pink) and TrpCage (cyan) domains expressed as dependence of molar ellipticity on temperature at 224 nm. (d) Summary of melting temperatures (Tm) of FD3A (red) and FD4A (blue) fusion proteins and PDZ3 (pink) and TrpCage (cyan) domains

3.3. Thermal and chemical denaturation of FD3A and FD4A revealed distinct chimera characters

To obtain information about the denaturation pathways of FD3A and FD4A, we investigated the chemical and thermal denaturation characteristics of both proteins by CD spectroscopy. Specifically, we measured a CD signal of the proteins as a dependency of a molar ellipticity on the temperature or concentration of urea denaturant. Far‐ and near‐UV CD signals as a function of temperature or the urea concentrations were measured for FD3A and FD4A chimeras and both PDZ3 and TrpCage individual domains. PDZ3 and TrpCage folding/unfolding and dynamic studies have been extensively investigated and documented in the literature. ⁵⁵ , ⁵⁷ , ⁵⁸ PDZ3 thermal and chemical denaturation characteristics were also studied by kinetic experiments using CD spectroscopy, ⁵⁹ Trp fluorescence or NMR methods. ⁶⁰ PDZ domains commonly differ in their T_m values based on the origin of the protein, the method used to validate the T_m, and the sample concentration during the experiment. The PDZ2 domain from the ZO‐2 protein showed a T_m around 40–50°C with an obvious protein concentration dependency. ⁶¹ In one of the previous studies, the T_m of PDZ3 was estimated at a surprisingly high temperature of 72.58°C, ⁶² and the explanation of such high thermostability was provided as follows: the PDZ3 partially thermally unfolded intermediates form aggregates under native conditions, and they were detected by native‐state hydrogen exchange experiments. It implies that two processes could take place during the thermal unfolding of the PDZ3 domain. Whereas we performed low‐protein concentration rate experiments in this work, we did not expect the formation of such oligomeric intermediates; however, the PDZ3 unfolding behavior could be potentially influence by the TrpCage modulation function in FD3A and FD4A. TrpCage The TrpCage thermal unfolding equilibrium was estimated to have T_m of 41.85°C by NMR, ²⁹ and this small domain has been thoroughly studied over a range of urea denaturant concentrations as well. ⁶³

We examined the FD3A and FD4A temperature and chemical denaturation profiles by CD spectroscopy. The temperature denaturation was measured from 5 to 90°C, showing T_m ~ 46°C for FD3A, which is comparable to the T_m obtained by our measurements for the individual PDZ3 domain at ~45°C (Figure 4(c),(d), S2). ⁶¹ The FD4A with T_m ~ 59°C showed a higher stability of this construct. The T_m value for TrpCage was determined as ~37°C, concurring with its previously published thermal unfolding data. The comparable denaturation characteristics were observed for urea‐induced unfolding CD studies. The unfolding curves of FD3A, FD4A, PDZ3, and TrpCage were expressed as a function of a CD_222nm signal with urea concentrations in the range of 0–6 M (Figure 5, S3). The denaturation midpoint was observed for an urea concentration of c_urea = 2.9 M for FD3A, and c_urea = 3.1 M for FD4A. Thus, in both unfolding experiments, the FD3A construct seems to be less stable then FD4A. The urea concentration of the denaturation midpoint obtained for the PDZ3 domain was c_urea = 2.7 M, which corresponds to previously published data for PDZ urea denaturation. ⁶⁴ TrpCage values could not be determine (Figure 5). The CD data on chemical denaturation of FD3A revealed a lower isosbestic point in comparison to FD4A.

FIGURE 5.

The urea‐induced unfolding of FD3A and FD4A fusion proteins, and of PDZ3 and TrpCage individual domains. Far‐UV CD spectra of (a) FD3A, (b) FD4A, (c) PDZ3 and (d) TrpCage were measured over the 0.0–6.0 M urea concentration range with increment of 0.5 M. The CD spectra were expressed as molar ellipticities at 222 nm as a function of molar concentrations of urea

Thermal and chemical denaturation was explored to reveal the distinct behavior of two proteins with swapped domain order. Based on the AUC data suggesting on FD4A dimerization due to a higher (30–120 M) protein concentration, we assume that FD4A chimera may keep the PDZ3 based oligomeric potential independently of the fused TrpCage. In such a scenario, the FD4A chimera behaves like two independent domains just connected by the linker, it means there is no interface between PDZ3 and TrpCage. Therefore, we concluded from our data that the FD3A chimera does not show any signs of oligomers formation, and it represents a classical two‐state folder. The absence of oligomers for the FD3A chimera suggests that FD3A preserves its one‐domain compact form and uses a different denaturation mechanism. It has been described that the PDZ3 oligomer intermediates are formed by interactions of PDZ β‐sheets. ⁶² Therefore, in the case of FD3A, a stable domain–domain interface between PDZ3 and TrpCage could be possibly preserved, and this could block the association of β‐sheets necessary to form the oligomeric intermediates. Therefore, our data support the conclusion that PDZ3 and TrpCage inter‐domain communication within the FD3A and FD4A chimeras has a different character.

3.4. The order of the artificially fused PDZ3 and TrpCage domains in FD3A and FD4A is significant for defining the dynamics of the respective system

The selected molecular models provided specific conformations of the fusion protein for the direct and reverse domains combinations (Figure 2(c)). The analysis of triplicated 200 ns simulation runs for both direct and reverse order chimeras confirmed the overall structural stability of the composing domain units, PDZ3 and the TrpCage (Figure 6). In parallel, the stability of both individual domains was also studied and confirmed by independent MDs (see Figure S4). The results of MDs show a stable and preserved globular structure with a clearly defined interface between the PDZ3 domain and TrpCage for a FD3A chimeric construct (Figure 6(a)). In the case of FD4A, two individual conformational states of the domains were studied due to the higher expected dynamic between the conformational states. The first studied conformational states could be characterized as “compact” (Figure 6(b)). In this model, TrpCage was oriented toward PDZ3 in a similar way as in the FD3A, occupying a hydrophobic interaction interface. Conversely, the second state of FD4A is referred to as “open,” and the analyzed trajectories show no direct interactions between PDZ3 domain and TrpCage (Figure 6(c)). The “open” state of FD4A displayed a relatively high flexibility of mutual positions for composing domains, apparently brought on by the longer linker region and by sampling different orientations of domains several times during the simulation.

FIGURE 6.

Molecular dynamics simulations of FD3A and FD4A molecular models. The RMSD analysis of selected 200 ns molecular dynamic simulations (MDs) of (a) FD3A, (b) FD4A closed state and (c) FD4A open state. Root mean square deviation (RMSD) of C‐α atoms were analysed separately for TrpCage and PDZ3 domain to confirm the structural stability of the individual domains during molecular dynamic simulations (MDs). The structural stabilities of the whole molecules were analysed as RMSD of selected C‐α atoms corresponding to regions with defined secondary structure in the initial model. In all cases, RMSD is determined by comparing desired frame of MDs runs to the relevant initial structure. In the case of bottom right plot (C) for FD4A open state, the gray line shows RMSD computed for open conformation with compact conformation as a reference structure. The initial structures of (d) FD3A, (e) FD4A closed state and (f) FD4A open state consisting of PDZ3 domain and Trp‐cage (PDZ3 domain is shown in pink, TrpCage is shown in cyan, linker region and terminal extensions are shown in gray)

The picture obtained by MDs can help us to interpret the results of experimental studies and the biophysical properties of the studied chimeras. Based on the experimentally different behavior of the both chimeras, we can conclude that the structure and stability of a chimera are first determined by the different order of domains in the resulting protein context, which is supported by results obtained by MDs. The significant difference is one tight, well‐packed, and stable 3D structure of the FD3A (Figure 6(d)) and the unstable character of the FD4A sampling two different conformational states, the “open” state of the protein and the “closed” state (Figure 6(e),(f)) resembling the FD3A.

3.5. Conclusion

The two‐domain fusion proteins with swapped domains order were designed, expressed, and characterized by biochemical and biophysical methods. The selected functional system of the PDZ3 domain was linked with an artificially designed TrpCage domain. Biophysical and computational characteristics of the two well‐folded domains synthetically linked by Gly linker suggested that the domain order is a significant element affecting their folding and structural dynamic properties. The FD3A and FD4A proteins provided different characteristics under biochemical and biophysical experiments, and different behavior upon chemical and temperature denaturation. Both denaturation experiments, based on different unfolding chemical reactions, unambiguously suggest that a different mechanism for the folding and dynamics of both studied fusion proteins is dictated by their native structure. Still, the elucidation of the folding mechanism and dynamics of both proteins are a challenge, and will be addressed in our following work. We suppose that the different behavior of the studied chimeras reflects not only the different composition, but most probably a different way of domain communications influencing their structure, dynamics and folding/unfolding processes.

The ultimate goal of our project was to show the importance of the sequential and structural context in multi‐domain proteins. The obtained results settle that the observed differences can be manifested by functional changes due to allostery modulation. While most well characterized examples of dynamic allostery involve binding with other entities like small molecules, peptides, or nucleic acids, in this work, we anticipated that more extensive modifications like domain swapping could lead to a significant dynamic allosteric response in a PDZ protein. The combination of the PDZ3 and the artificially designed TrpCage domain created a new system to study effects that are not affected by the co‐evolution of natural domains. We believe that this will open opportunities to address the extent to which this could be a functional regulatory mechanism via allostery. The interchangeability of protein domains represents an effective method, not only to study how fusion domains could be potentially engineered with an effect of potential mutual allostery influence but also to create new highly functional fusion proteins.

CONFLICT OF INTEREST

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

AUTHOR CONTRIBUTIONS

Kristyna Bousova: Conceptualization; data curation; formal analysis; investigation; methodology; project administration; validation; visualization; writing‐original draft; writing‐review & editing. Lucie Bednarova: Data curation; formal analysis; methodology; writing‐original draft. Monika Zouharova: Investigation; methodology; validation; visualization; writing‐review & editing. Veronika Vetyskova: Investigation; methodology; validation; writing‐review & editing. Klara Postulkova: Methodology; writing‐review & editing. Kateřina Hofbauerová: Data curation; formal analysis; funding acquisition; methodology; writing‐original draft. Olivia Petrvalska: Formal analysis; investigation; methodology; validation; visualization; writing‐review & editing. Ondrej Vanek: Data curation; formal analysis; methodology; visualization; writing‐original draft. Konstantinos Tripsianes: Data curation; funding acquisition; supervision. Jiri Vondrasek: Conceptualization; data curation; formal analysis; funding acquisition; investigation; resources; supervision; validation; writing‐original draft; writing‐review & editing.

Supporting information

Appendix S1: Supplementary Information

Bousova K, Bednarova L, Zouharova M, et al. The order of PDZ3 and TrpCage in fusion chimeras determines their properties—a biophysical characterization. Protein Science. 2021;30:1653–1666. 10.1002/pro.4107