Transcription factor allosteric regulation through substrate coordination to zinc

Abstract

The development of new synthetic biology circuits for biotechnology and medicine requires deeper mechanistic insight into allosteric transcription factors (aTFs). Here we studied the aTF UxuR, a homodimer of two domains connected by a highly flexible linker region. To explore how ligand binding to UxuR affects protein dynamics we performed molecular dynamics simulations in the free protein, the aTF bound to the inducer D-fructuronate or the structural isomer D-glucuronate. We then validated our results by constructing a sensor plasmid for D-fructuronate in Escherichia coli and performed site-directed mutagenesis. Our results show that zinc coordination is necessary for UxuR function since mutation to alanines prevents expression de-repression by D-fructuronate. Analyzing the different complexes, we found that the disordered linker regions allow the N-terminal domains to display fast and large movements. When the inducer is bound, UxuR can sample an open conformation with a more pronounced negative charge at the surface of the N-terminal DNA binding domains. In opposition, in the free and D-glucuronate bond forms the protein samples closed conformations, with a more positive character at the surface of the DNA binding regions. These molecular insights provide a new basis to harness these systems for biological systems engineering.

INTRODUCTION

Gammaproteobacteria have a remarkable capability, in stress conditions, to form complexes with regulatory proteins and different molecules, thus, acquiring structural and functional changes, to switch between metabolic pathways to accomplish an overarching purpose (e.g. growth) (1,2). The ability to control regulatory proteins (also named transcription factors: TFs) has been extensively studied, becoming one of the challenges in modern molecular biology. Several TFs are sensitive to small metabolites, ions or drugs, which can modify their binding to DNA, thereby regulating gene expression (2). One mechanism associated with these TFs, enhancing the response to environmental signals, is allosteric regulation (1). TFs that are inducible by small molecules are useful synthetic biology tools as sensors and switches (3).

Bacterial allosteric TFs (aTFs) encompass several families of proteins (1,3). The gluconate operon repressor in Bacillus subtilis (GntR) superfamily (Pfam-PF0032) (4) is a large group of TFs present in several bacterial groups (5) and they regulate fundamental biological processes, such as motility (6), development (7), antibiotic production and resistance (8,9), plasmid transfer (10) and virulence (11). Despite the fact that most of the GntR TFs available in the PDB database do not contain metals, some studies strongly suggest that the majority of fatty acid metabolism regulator proteins (FadR) (Pfam-PF07840—the largest family of GntR superfamily) (4) are metal-dependent (1,12–15), However, the underlying allosteric mechanism is poorly understood. It is not clear whether these TFs are metal-sensing, if the metal plays a structural role, or if metals are required for binding of other effector molecules (12–15).

Escherichia coli and other Gammaproteobacteria can grow using hexuronic acids as a carbon source (16). This metabolic pathway, named the Ashwell pathway, is controlled by the Uxu operon transcriptional regulator (UxuR TF) (17). UxuR TF represses the expression of its gene (uxuR) and controls the expression of operons that encode the main proteins of the Ashwell pathway (18). Experimental studies indicate that UxuR TF binds to fructuronate (19), the product of the reaction immediately downstream of glucuronate, catalyzed by D-glucuronate/D-galacturonate isomerase in the catabolic pathway. It was shown that fructuronate can influence the interaction of UxuR TF with DNA (19). However, the conformational changes that lead to this transcriptional regulation are poorly understood. Previous modeling efforts for UxuR TF assumed that the structure was a monomer or did not consider the hypothesis of metal binding (20,21). Consequently, until now, the ligand-binding site and the residues involved remain unknown, as well as the associated conformational changes.

In this work, we determined and experimentally verified, by site-directed mutagenesis, the ligand-binding site of UxuR TF (homodimer with Zn(II) ion). We furthermore studied the dynamical behavior of UxuR TF, by comparing the unbound structure to structures bound to D-fructuronate and D-glucuronate, independently. Here we have analysed the dynamics and found large NTD domain transitions at the nanosecond range.

MATERIALS AND METHODS

Modeling

We resorted to the Rosetta software package for protein structure prediction through the Robetta server (22). A GntR TF from Streptococcus agalactiae (23) (PDB code: 6AZ6, 1.91 Å resolution), with 28.9% of identity was used for homology modeling; Regions lacking enough homology, such as the peptide linkers, were modeled using the Rosetta de novo structure prediction method.

The zinc metal coordination was modeled using a semi-bonded model approach (cationic dummy atom: CaDA method) according to Pang et al. (24,25) The natural ligand, α-D-fructuronate (DFU; PubChem Identifier: CID 46878576) and β-D-glucuronate (GLU; PubChem Identifier: CID 11877136) were obtained from the PubChem Data Base. Both ligands were prepared using the UCSF Chimera program (26) and the parameters were retrieved from the GLYCAM database (27). The atomic partial charges were calculated employing the (RESP) (28) method from the HF/6–31G(d,p) single-point energy calculations.

Molecular docking

The molecular docking was performed with AutoDock4.2 suite of programs (29) with the Lamarckian Genetic Algorithm (LGA) (30). A grid box (40 × 40 × 40 Å) was centered on the Zn(II) ion, at both chains (chain A: x = 47.741; y = -2.039; z = 21.566 and chain B: x = 46.137; y = 38.828; z = 23.538), for DFU as well as for GLU. A total of 100 LGA runs were carried out for each of the ligand–protein complexes. The population was 300, the maximum number of generations was 27 000 and the maximum number of energy evaluations was 2 500 000. All rotatable bonds were kept flexible.

Molecular dynamics

To study the structural changes between the free protein and complexes with the ligands the proteins were subjected to molecular dynamics (MD) simulations (31–33). MD simulations were performed using the MD software package Amber18 (34) with the parm99SB (35) and GLYCAM06 (27) force fields. The structures were placed in an octahedral box of water molecules (10.0 Å between the surface of the protein to the box). Counter ions were added to make the entire system neutral. Each system was subjected to two initial energy minimizations and 500 ps of equilibration in an NVT ensemble, using Langevin dynamics (36) with small restraints on the protein (10.0 kcal/mol) to heat the systems from 0 K to 310.15 K (optimal temperature for the bacterial organism). Production simulations were carried out at 310.15 K in the NPT ensemble, using Langevin dynamics with a collision frequency of 1 ps^–1. Constant pressure periodic boundary conditions were imposed with an average pressure of 1 ATM. Isotropic position scaling was used to maintain pressure with a relaxation time of 2 ps. The SHAKE algorithm (37) was applied to all bonds involving hydrogen atoms. The SHAKE feature constrains the vibrational stretching of hydrogen bond lengths and fixes the bond distance to the equilibrium value. The Particle Mesh Ewald method (38) was used to calculate electrostatic interactions with a cutoff distance of 10.0 Å. The total time of the simulations was 2.1 μs (seven replicas with different initial velocities for each situation), the integration time was set to 2 fs.

We performed extensive MD simulations for three situations, intending to better understand the structural conformation change in our aTF: UxuR without any ligand, UxuR with DFU in both chains, and UxuR with GLU in both chains.

Principal components analysis and free energy landscape

To identify the overall patterns of motions in the three situations, we used principal components analysis (PCA) which was carried out using the CPPTRAJ program (39), comprising 7554 atoms in the free situation and 7598 atoms in the situations with the ligands bound. The trajectories of corresponding atoms were extracted and analyzed for 100 ns using 5 000 000 frames in each replica; the first frame was set as the reference to remove global translation/rotation. The Cartesian covariance matrix of the involved residues was calculated and then diagonalized to provide a series of eigenvectors representing different modes of conformational change and their corresponding eigenvalues. The Bio3d software (40), a package of R software (41), was used to quantify the PCA results. To identify the dominant conformational states, we calculated the free energy landscape (FEL) (42) concerning PC1 and PC2.

The study of conformational transition pathways was performed by the elastic network-driven Brownian Dynamics Importance Sampling (eBDIMS) method, which is based on the Essential Dynamics-refined Elastic Network Model force field (43). Here we used the eBDIMS server with the default parameters.

Electrostatic potential surfaces

The electrostatic potential molecular surface was calculated using the Adaptive Poisson-Boltzmann Solver (APBS) software (44). The required partial charges to molecule preparation were performed by PDB2PQR software (45). The electrostatic maps were calculated using the default parameters and the results were display using the potential at the solvent accessible surface for coloring. All procedures were done using the APBS Electrostatics Plugin into PyMOL2.0 (46).

Quantum mechanics cluster calculations

Cluster models were constructed from the last structures of the simulations in each situation, and geometry optimized with the exchange-correlation functional B3LYP and the basis set 6–31G(d) (47). The free protein has a total of 151 atoms, whereas the complex with DFU has 173 atoms and GLU has 186 atoms. The atoms of the protein residues cleaved bonds were kept frozen during the calculations.

Microbial strains

Escherichia coli strains used in this study were MG1655 with uxuB knockout (MBR) and MG1655(DE3) with gudD, uxaC and uxuR knockouts (M2BR) (Supplementary Table S1). Knockout strains were made from parent strains MG1655 and MG1655(DE3) with gudD and uxaC knockouts (MG1655(DE3)ΔgudDΔuxuR), respectively, using the λ Red knockout protocol (48) to sequentially delete uxuB and uxuR from the parent strains. Removing uxuB prevents the consumption of DFU and removing uxuR prevents interference from endogenous UxuR. λ-Red-mediated recombination was done using pKD46 with primers uxuB_FWD/uxuB_REV and uxuR_FWD/uxuR_REV for uxuB and uxuR knockouts, respectively (Supplementary Table S2). These primers were used to amplify the corresponding knockout cassette (flanked by FLP recognition target sites) from the Keio collection strains (49). MG1655 and MG1655(DE3)ΔgudDΔuxuR harboring pKD46 were transformed sequentially, first with the uxuB cassette (forming intermediate strains MG1655 with uxuB knockout (MB) and MG1655 with gudD, uxaC and uxuB knockouts (M2B), intermediate strains and then with the uxuR cassette. Following transformation, the kan selection cassette was cured of successful deletion mutants using FLP recombinase expressed from pCP20, generating strains MBR and M2BR.

Sensor plasmid construction

All primers used in this study are listed in Supplementary Table S2. All plasmids used in this study are listed in Supplementary Table S3. Also, a diagram depicting the biosensor plasmid is available in the Supporting Information (Supplementary Figure S1).

The biosensor plasmid was generated from plasmid pSNR1. Briefly, pSNR1 is a modified form of pHHD01K (50) that contains a p15A origin of replication, an mRFP gene from pTrc-RFP (located between EcoRI and BamHI cut sites located in pHHD01K) and a gfp(mut3b) gene from the Registry of Standard Biological Parts, part BBa_E0040 (http://www.partsregistry.org). mRFP is under control of the pTet promoter with RBS sequence TCACACAGGAAAG and an rrnB terminator B0010. Gfp(mut3b) is under control of a cpk promoter (see Supplementary Table S4 for sequence), with RBS sequence TATAGGGAG, and a B0014 terminator. The TetR cassette from pHHD01K was modified with a B1002 terminator. All terminators and the mRFP RBS sequence (Bba_0032) are from the Registry of Standard Biological Parts (http://www.partsregistry.org).

A negative control plasmid containing no gfp, pSNR1MCS, was generated from pSNR1 by replacing the gfp gene with a multiple cloning site. The multiple cloning site was generated by an oligo anneal of primers pSNR1 MCS FWD and pSNR1 MCS REV. The annealed oligo and pSNR1 were both digested with SacI and NotI and ligated together to form pSNR1MCS.

To incorporate a hybrid promoter for gfp that contained a UxuR binding site, the full-length hybrid promoter sequence was generated by polymerase chain reaction (PCR) amplification using primers SacI_Prom_FWD and AflII_Prom_REV that contained overlapping regions. These primers annealed to each other for amplification (with no additional template DNA required) to form the desired insert and digestion sites. The final promoter sequence generated was: TTTACAaattggtataccaatttTATTATgctagcAcgtgcaatttttaaaattaaaggcgttacccaactatagggag. −35 and −10 sequences and transcription start site are capitalized, with the binding site of UxuR located between the −35 and −10 sequences (Aaattggtataccaattt). −35 and −10 sites are from promoter J23101 from the Registry of Standard Biological Parts (http://www.partsregistry.org) and the UxuR binding site is from the promoter of the uxuR gene from E. coli (20). This PCR product and the pSNR1 backbone were both digested with SacI and AflII and ligated together. This formed the preliminary vector, pSNR1U.

To enable higher levels of gfp translation, a modified RBS was incorporated into pSNR1U using circular polymerase extension cloning (CPEC) (51). This new hybrid promoter region, Upro, (tttacaaattggtataccaattttattatgctagcacgtgcaatttttaaaattaaaggcgttacccaacAGAGG AGAaatactag) contained the same promoter, J23101, but a new, stronger RBS (capitalized). pSNR1U was amplified using primers FWD_Plasmid_CPEC_pSNR1U-Upro and REV_Plasmid_CPEC_pSNR1U-Upro, and the insert was generated by PCR with primers FWD_CPEC_pSNR1U-Upro_Insert and REV_CPEC_pSNR1U-Upro_Insert that contained overlapping regions for CPEC. As before, the insert primers self-annealed, allowing amplification without the need for template DNA. The backbone and insert were annealed and amplified by CPEC to generate the final positive control plasmid, pSNR1-Upro. This resulted in a positive control plasmid that contained no uxuR and expressed GFP constitutively in a uxuR knockout strain.

To incorporate the E. coli uxuR gene and create pSNR1-UxuR-Upro, primers BsaI_UxuR_FWD and BsaI_UxuR_REV were used to PCR amplify uxuR from E. coli genomic DNA. The PCR product was digested with BsaI (leaving Eco-RI and SacI sticky ends) and the pSNR1-Upro backbone was digested with EcoRI and SacI. The amplified gene and digested backbone were ligated to form pSNR1-UxuR-Upro.

Mutant ConstructionSite-directed mutagenesis was performed using the Q5 Site-Directed Mutagenesis (SDM) kit (New England BioLabs, Ipswich, MA, USA). Briefly, the pSNR1-UxuR-Upro plasmid was amplified with primers MutA1-A4_F and MutA1-A4_R (Supplementary Table S3). The primers A1-A4 amplified the entire vector and contained the desired mutagenized bases for mutations H152A, H201A, H223A and D148A, respectively. The amplified PCR product was digested with Dpn1 enzyme to remove template DNA at 37°C for 45 min and gel purified. The purified product was ligated using the SDM kit (per manufacturer instructions) and the ligation product used to transform DH5α.

Colonies were sequenced to verify the presence of the desired mutation using primers pTet_Forward and pHHD01K_Term_Rev. All restriction enzymes, ligases, and polymerases were purchased from New England Biolabs (Ipswich, MA, USA).

Fermentation and GFP measurement

Strain MBR was transformed with the positive control of the biosensor (pSRN1-Upro), the negative control of the biosensor (pSNR1MCS), the original, unmutated biosensor (pSNR1-UxuR-Upro) and the mutated versions of the biosensor (Supplementary Table S3). Overnight cultures of transformants were grown overnight in LB with 50 μg/ml kanamycin at 37°C. Fermentation cultures were inoculated from overnight cultures in LB and 50 μg/ml kanamycin at a dilution of 1:100 vol/vol. Fermentation was carried out for 24 h in the BioLector (m2p labs, Baesweiler, Germany). Cultures were grown in FlowerPlates (m2p labs) containing 1 ml medium per well, at 37°C, 1200 rpm, 80% relative humidity. The wells contained LB or LB medium supplemented with 3 mM GLU; 50 μg/ml kanamycin was added to all cultures. Medium in which the biosensor plasmid and mutant variants were cultured was additionally supplemented with 50 ng/ml aTc for induction of uxuR from the pTet promoter. Biomass (backscattered light at 620 nm) and GFP (488 nm excitation/520 nm emission) measurements were taken by the BioLector every ∼15 min. GFP units as measured by the BioLector were normalized by biomass units from the BioLector for data analysis.

Statistics

MATLAB statistics software was used to perform a right-tailed t-test assuming unequal variances to compare the normalized GFP signals. To compare the fold change in GFP of the unmutated biosensor to the fold change in GFP of the mutated biosensor a left-tailed t-test assuming unequal variances was used.

RESULTS AND DISCUSSION

Substrate binding to UxuR

The proteins of the GntR superfamily contain a remarkably conserved DNA-binding winged helix-turn-helix (wHTH) domain at the N-terminal region (NTD) and an effector-binding domain (EBD) at the C-terminal region (CTD) used to define GntR subfamilies. Upon binding of an effector molecule, a conformational change occurs, influencing the DNA-binding properties of the TFs, resulting in repression or activation of gene transcription (52). The nature of this mechanism is not described for any member of the GntR subfamily.

Here, we have built a new model for the aTF UxuR (Supplementary Figure S2 and S3). We have taken all the sequence information about the GntR superfamily into account. Our results show that UxuR is a dimer, which is typical for the FadR subfamily (53). The NTD of our model is formed by three α helices and three β strands. The CTD region is composed of six α helices, thus the aTF UxuR can be included in the FCD family (54). The wHTH and EBD domains are connected through a linker. Usually, this linker, composed of flexible residues, vary in secondary structure and length (from ∼ 5 to 25 amino acids) (55,56). UxuR linker model is formed by 21 amino acids, the last one being a proline residue which has been identified as a more frequent terminal linker residue, because of its rigidity that prevents unfavorable contacts between domains that could be as an unwanted result of the high flexibility of the linkers (55). The connecting between domains through a linker has an important consideration in protein architecture and function, which are both flexible and allosterically regulated (55,56). In the aTF UxuR model, we predicted some residues implied in the allosteric mechanism. The most common mode of coordination of a Zn(II) ion is via four atoms (57). The residues that dominate zinc-binding sites are well established: histidine, aspartate, glutamate and cysteine (58,59). Sequence analysis showed that most FadR TFs, namely the FadR C-terminal domain (FCD) family (Pfam-PF07729(4)), contain three conserved histidine residues and one aspartate residue, supporting the probability of metal dependence (1,13). For the UxuR model, we suggest that the conserved residues of the zinc-binding motif are composed of three histidine residues (H152, H201 and H223) and one aspartate (D148) (Figure 1). The cationic dummy atom (CaDA) method used to describe Zn(II) ion in aTF UxuR model, provides a reasonable description for Zn(II) ion's first coordination sphere (60). Our results show that during the MD simulations, the Zn(II) ion keeps its position. However, to further assess Zn(II) ion coordination we performed Quantum Mechanic (QM) cluster model calculations with the last structures of the simulations (Figure 1).

Figure 1.

Quantum Mechanic cluster model structures from the UxuR MD simulations, showing the (A) free form, (B) DFU and the (C) GLU carboxylate oxygen atoms binding to the Zn(II) ion (sphere colored orange).

Based on the information of the closest homologs (15) and our molecular docking studies, we have identified a pocket in the CTD close to the Zn(II) ion as the most probable region for substrate binding and coordinates to the substrates. We were able to identify interactions between the ligands and the protein after performing independent MD simulations for the three situations: i) free protein, ii) bound to DFU and iii) bound to GLU. Observing the free form situation (Figure 1A), the Zn(II) ion binds to the three histidine residues (H152, H201 and H223) and D148. When the UxuR is bound to their inducer (Figure 1B) and GLU (Figure 1C), the H152 side-chain rearrange its position allowing the binding of ligands to the Zn(II) ion and maintain the coordination via four atoms. Our results are in good agreement with recent literature, the average distance of metal-ligand for nitrogen atoms is 2.12 ± 0.19 Å and for oxygen atoms is 2.31 ± 0.54 Å (57). To verify our in silico results, we performed site-directed mutagenesis of the histidines and the aspartate residues. As expected, the single mutations of the histidine and the aspartate residues negatively interfere with ligand binding to UxuR (more detailed results are discussed below). These results are in unanimity with our computational predictions and reveal the important role of these four residues in allosteric regulation with the ligand. Also, the structural analysis revealed that the residues were in the correct orientation to coordinate to Zn(II) ion.

Substrate binding affects the dynamical behavior of the protein

Having determined the binding site for the substrates, we then studied the dynamic behavior of the protein for the three situations. When the ligands are bound, local changes generate an allosteric response, leading to a conformational change of UxuR. At this moment, the stimulus resulting from the perturbation caused by the ligand binding it is propagated by the linkers, from the EBD to the NTD. The flexibility of linkers allows the propagation of a signal that results in different conformational changes. The after-effect is the fast reorientation of one domain (CTD) to another (NTD). The barriers implicate in these conformational changes are quite low. Thus, molecular movements can be observed and analyzed during nanosecond simulations (56,61–63). In our study, the Root-Mean-Square Deviations (RMSDs) and Root-Mean-Square Fluctuations (RMSFs) by residue, are provided for each simulation in the SI. As it is possible to observe in Supplementary Figure S4, the values are large for all structures, showing that the structures are quite flexible and that significant domain movement occurs during our simulations. To dissect the nature of such movements, we performed PCA to extract the dominant modes in the UxuR motion from the MD simulations. By calculating the eigenvectors from the covariance matrix of a simulation we have identified the dominant motions. Briefly, the Principal Component (PC) is a vector describing the relative contribution of each selected atom's x, y and z component to the correlated mode of motion corresponding to that component (64). Here, we show for different approaches the differences between the motions in the free protein and the protein bound with either ligand. On the 2D plane projections (Figure 2 and Supplementary Figures S5–S7), each simulation frame onto a given PC is the dot product of the alpha carbon (Cα) coordinates of that frame with the PC vector. In general, for the free form situation and the one bound to DFU, the three largest PCs (PC1, PC2 and PC3) account for more than 70% of the motions (Supplementary Figures S5 and S6). However, the GLU bound situation seems to be characterized by a more complex movement, since the three first PCs were not enough to identify the dominant motions (Supplementary Figure S7). At the free protein, the two NTDs (which bind the DNA) are close together (in what we will call from now on a closed conformation) and vibrate around this position in a parallel orientation and opposite directions (Figures 2A and B; Supplementary Figure S8). When DFU binds to UxuR at the CTD region, the movements of the NTD regions are counterclockwise or clockwise twists, with a more pronounced movement of one chain (Figures 2C and D; Supplementary Figure S9). On the other hand, when GLU binds to UxuR, the movements of the NTD regions are closure twists (Figures 2E and F; Supplementary Figure S10). This movement with the GLU ligand seems to allow for the approximation of the NTD regions and the binding to the DNA. Based on results from the first two PCs, we constructed and analysed the FEL of the systems understudy (free form, bound to DFU and GLU) at 300 K (Figure 3). Our results show different minima denoting the stable states (darkest colored) and the associated higher-energy transient states (light colors) (Figure 3A, C and E). The free form FEL is significantly different from the bound systems, with just one minimum during the simulation time (Figure 3A and B). The bound-states display a lower-energy minimum and other smaller minima (Figure 3C–F). The DFU form has the largest global minimum and this is in a different position when compared to the other situations. Our results are in accordance with previous data that points to dynamic ensembles of a swiftly conformational switch, which can be described by relatively flat energy landscapes (65). For example, the ZitR TF, which contains a linker connecting two domains, displays conformational changes that amount to free energy barriers around 3 kcal/mol (63). In our work, the conformational changes observed through the simulations for the three systems are also characterized by low free energy barriers of about 3.2 kcal/mol.

Figure 2.

Right: Bio3D plot of PCA results for trajectory frames colored from blue to white to red in order of time. Left: The structures are shown as a backbone trace at the front view. The colors summarize the direction of motion, from blue to white to red. (A) and (B) UxuR free form; (C) and (D) UxuR bound to DFU; (E) and (F) UxuR bound to GLU.

Figure 3.

Two-dimensional FEL (in kcal/mol) of PC1 and PC2 of the free form protein (A and B) and protein ligand complex (C and D: DFU; E and F: GLU) at 300 K. Deeper color areas in the maps indicate lower energy. The global minima structures are shown as a backbone trace at the front view: red—free from, blue—bound to DFU and green—bound to GLU.

Comparing the global minimum structures from the PCA and FEL analysis (Supplementary Figure S11), we can observe more identical conformations between the free form and the one bound to GLU (Supplementary Figure S11a), than with the DFU form (Supplementary Figure S11b). These results are in good agreement with the dominant motions observed during the simulations. Contrary to what happens when the DFU is present, binding of the GLU still favors the bound conformation, which may explain the ability of UxuR to remain bound to DNA, as it happens in the free form.

For more detailed quantitative results of the motions in aTF UxuR model, we used the eBDIMS server (66). The eBDIMS selects the major normal modes (NMs) of the start structure for projection, i.e. the two NMs better overlapped with the modeled transition, representing more than 80% of the movements involved. Analyzing the DFU and GLU situations (Supplementary Figure S12), we observe a similar molecular movement denoting the allosteric conformational alteration. On the other hand, the free situation is distinct, which is in good agreement with the data resulting from the PCA and FEL. These results support the difference of motions between the three situations that indicated whether or not the aTF UxuR is bound to the DNA.

Based on our results during all simulations (Figures 2, 3 and Supplementary Figure S5–S12), we can establish a movement pattern for each situation, and we hypothesize that the closing movement of the chains allows the protein to bind to the DNA. This is in accordance with the flexibility of the linkers. The binding of the ligands causes a perturbation through the linkers that leads to a conformational change and reorientation of protein inter-domains region (56). Here, this allosteric propagation force seems to have different behavior whether DFU or GLU are bound to the protein.

To better understand how the allosteric regulation compromises the conformational behavior of the UxuR, we analyze the alteration of the residues side-chain on the NTD. The electrostatic surface calculations for UxuR in free and bound forms show differences at the NTD region that can further explain the binding to the DNA (Figure 4). In the free protein, the two NTD regions remain close, leading to a continuous stretch of positively charged residues in the region that binds the DNA (Figure 4A). In the structure with DFU, the two N-termini are farther apart and the negatively charged interface is more visible. This change in the protein surface charge can adversely impact binding to the DNA (which is also negatively charged) (Figure 4B). In the structure with GLU, the NTD regions are still connected, but instead of being in a parallel orientation, they are diagonally oriented (Figure 4C). It should be noted that these changes are dynamic; the NTD in the free protein remain to oscillate between closed conformations, whereas in the protein with GLU bound we observe a switch between open and closed conformations. If we consider only the crystal structure of FADR (PDB code: 1H9T) with bound DNA, we can observe that the two NTDs are in a closed conformation binding side-by-side to the major groove and different minor-groove regions (Supplementary Figure S13). Similarly, our free-protein and the protein–GLU complex can adopt closed conformations (albeit different ones), while the protein–DFU complex can sample an open-conformation. Furthermore, the movement promoted by DFU binding seems to be incompatible with closed conformations. More information about the electrostatic surface of UxuR can be found in the Supplementary Figures S14–S16.

Figure 4.

Electrostatic surface of UxuR: (A and B)—free form; (C and D)—UxuR bound to DFU; (E and F)—UxuR bound to GLU. The surface structures are shown at the front (A, C and E) and top views (B, D and F). The blue color indicates regions of positive potential whereas red shows negative potential values. The black arrows indicate the more probable DNA-binding sites.

Validation of the fructuronic acid biosensor

To allow for experimental observation of the importance of the aspartate and histidine residues upon UxuR function, a TF based biosensor that responds to DFU was successfully constructed and validated. This biosensor contained a regulator, uxuR and a fluorescent reporter, gfp. The positive and negative control set upper and lower bounds, respectively, on the expected normalized GFP signal (Figure 5A).

Figure 5.

(A) Normalized GFP for each sample at 24h. * indicates a normalized GFP signal in GLU supplemented media that is significantly (P < 0.05) higher than normalized GFP signal in unsupplemented media). (B) Fold change in normalized GFP in the presence of GLU. ** indicates a fold change in GFP that is significantly (P < 0.05) lower than observed in the biosensor. Error bars represent ±1 SD from the mean of three replicate cultures.

The UxuR operator site was located within a hybrid promoter upstream of the reporter gene. Upon expression of UxuR from the biosensor plasmid, pSNR1-UxuR-Upro, gfp transcription was repressed. This repression was alleviated upon allosteric binding by the ligand, DFU and resulting GFP levels correlated to the amount of ligand present. The ability of DFU to allosterically bind to the UxuR binding site and relieve transcriptional repression has been shown previously (19).

DFU is not commercially available; however, its structural isomer, GLU, is. DFU is formed through the isomerization of GLU via uronate isomerase (UxaC) (67). GLU can serve as a carbon source for E. coli (68) but cannot be metabolized in the absence of UxaC (69). UxaC can thus serve to convert GLU to DFU (which is then further metabolized through D-mannonate reductase, UxuB). We exploited this catabolic pathway to produce DFU in situ from GLU. Experiments were conducted in strain MBR, with knockouts of uxuB and genomic uxuR.

The expected biosensor behavior was observed as the GFP signal was found to be significantly higher in the presence of DFU (Figure 5B). The biosensor was not functional in a uxaC knockout strain (M2BR) confirming that the biosensor responds to DFU and not GLU (Supplementary Figure S17).

Further, the response of the biosensor to DFU (via 3 mM GLU) showed a clear switching dynamic with a slight lag (∼2 h) in biosensor response that is possibly due to the isomerization of GFU to DFU (Supplementary Figure S18). The dynamics of this response are similar to the responses shown in the successful application of sensor-regulation for pathway control (70,71). This indicates a potential for further application of the UxuR biosensor as a regulatory element.

This experiment was run with both positive (pSNR1-Upro) and negative (pSNR1MCS) control plasmids. The positive control plasmid had the same hybrid promoter and GFP gene as the sensor, but had no uxuR present on the plasmid, allowing for constitutive expression of GFP. The negative control contained the same uxuR gene as the biosensor but was lacking the gfp gene.

Mutations in the zinc coordination sphere

Mutations H152A, H201A, H223A and D148A were designed to disrupt the coordination of zinc-binding. Thus, testing the response of these mutants with the biosensor should reveal whether ligand binding is also disrupted. The uxuR mutants were incorporated in the biosensor plasmid, and the response to DFU was evaluated in MBR strains containing biosensor variants. In the mutant variants, no de-repression was observed upon the addition of GLU to the media (Figure 5B). The fold change in GFP response observed by the unmutated biosensor is significantly higher than all UxuR mutants suggesting that only the original biosensor retained the ability to respond to DFU while the mutants did not. Of note is the observation that there was a slight increase in the basal (leaky) biosensor signal from mutant H223A. This suggests that this particular mutation may have also impacted the affinity of UxuR for its cognate binding site within the promoter.

CONCLUSION

The GntR superfamily, broadly distributed throughout the bacteria and known to regulate essential biological processes, gives an appealing opening in the development of inducible and controllable expression systems. The presence of metals in the GntR superfamily was reported for the first time in the repressor LldR from Corynebacterium glutamicum (PDB code: 2DI3) involved in L-lactate and fructose/glucose utilization (15). Recently, Kotowska et al. described a Zn(II) ion binding a protein named HypR and suggested that zinc is a co-factor involved in the binding of an organic compound. Yet, the mechanism is still not clear (12). Zheng et al. solved the crystal structure of Thermotoga maritima TM0493—UxaR (PDB code: 3FMS), a regulator with a Zn(II) ion binding the effector domain involved in the regulation of hexuronic acid metabolism (13,72,73). The physiological role of the Zn(II) ion and its interaction was not described. Blancato et al. showed that in CitO, a regulator of citrate metabolism from Enterococcus faecalis, the coordination of a metal ion is crucial for binding of the citrate molecule and effective DNA binding (14). As we previously referred, it has been unclear if these GntR TFs are metal sensing or if the metal plays a structural role or is required for binding of other effector molecules (12,13,15). Thus, an understanding of the molecular mechanisms of regulation is required.

In this work, we have built a new homology model for the TF UxuR and performed docking and MD simulations. Our results show that this protein contains the conserved motif of three histidine residues and aspartate in a correct position to coordinate one Zn(II) ion. Furthermore, we showed that the substrate binds at the CTD and coordinates to the Zn(II) ion. We identified different fast motions of the protein that are dependent on the presence and identity of the bound substrate (i.e. there is no bound substrate, D-glucuronate or D-fructuronate). The different motions led to changes at the N-terminal positively charged surface that binds to the DNA. In both the unbound TF and the TF bound to GLU, the protein samples closed conformations that promote DNA binding; however, DFU binding to Zn(II) ion drives a change in UxuR domain movements, allowing the protein to sample an open conformation. We validated our predictions by showing that the mutations predicted to disrupt the zinc-binding motif led to an inability of the mutants to respond to fructuronic acid.

Here we propose and experimentally confirm, for the first time that the Zn(II) ion fills a role in the UxuR TF allosteric mechanism by modulating DNA binding. We further explain how this can be achieved by showing large differences in the NTDs between the different complexes.

SUPPLEMENTARY DATA

Supplementary Data are available at NARGAB Online.

FUNDING

This work was financed by Portuguese national funds via FCT - Fundação para a Ciência e a Tecnologia, under Projects MIT-Portugal seed project 6937814, UIDB/04539/2020, Grants IF/01272/2015, SFRH/BD/144303/2019 and SFRH/BD/150697/2020, and through the COMPETE 2020 - Operational Programme for Competitiveness and Internationalisation.

Conflict of interest statement. None declared.