Elucidation of Key Interactions between VirF and the virB Promoter in Shigella flexneri Using E. coli MarA- and GadX-Based Homology Models and In Vitro Analysis of the DNA-Binding Domains of VirF and MarA

ABSTRACT

Infection with Shigella, the organism responsible for the diarrheal disease shigellosis, leads to approximately 200,000 deaths globally annually. Virulence of this pathogen is primarily controlled by the DNA-binding transcriptional activator VirF. This AraC family protein activates transcription of two major virulence genes, virB and icsA, which lead to the pathogen’s ability to invade and spread within colonic epithelial cells. While several AraC proteins have been studied, few studies of VirF’s binding to its DNA promoters have been reported, and VirF’s three-dimensional structure remains unsolved. Here, we used structures of two E. coli VirF homologs, GadX and MarA-marRAB, to generate homology models of the VirF DNA-binding domain in free and DNA-bound conformations. We conducted alanine scanning mutagenesis on seven residues within MarA that make base-specific interactions with its promoter, marRAB, and the corresponding residues within VirF (identified by sequence and structural homologies). In vitro DNA-binding assays studying both wild-type and mutant MarA and VirF proteins identified residues important for binding to the marRAB and virB promoters, respectively. Comparison of the effects of these DNA-binding domain mutants validated our MarA-based homology model, allowing us to identify crucial interactions between VirF and the virB promoter. Proteins with mutations to helix 3 within both MarA(W42A, R46A) and MalE-VirF(R192A, K193A) exhibited significant reductions in DNA binding, while the effects of mutations in helix 6 varied. This suggests the shared importance of helix 3 in the binding to these promoters, while helix 6 is transcription factor specific. These results can inform further development of virulence-targeting inhibitors as an alternative to traditional antimicrobial drug design.

IMPORTANCE Globally, infection with Shigella flexneri is a leading cause of bacterial dysentery, particularly affecting children under the age of 5 years. The virulence of this pathogen makes it highly infectious, allowing it to spread easily within areas lacking proper sanitation or access to clean drinking water. VirF is a DNA-binding transcription factor that activates S. flexneri virulence once the bacteria infect the human colon. Development of drugs that target VirF’s DNA-binding activity can be an effective treatment to combat shigellosis as an alternative or addition to traditional antibiotics. Due to the lack of structural data, analysis of VirF’s DNA-binding activity is critical to the development of potent VirF inhibitors.

INTRODUCTION

Bacterial infections remain a major disease burden, particularly in regions of the world where access to adequate medication and sanitation is lacking (1, 2). Due to many factors, including lack of new antibiotic production and their improper use, antibiotic resistance continues to emerge, with many bacteria now becoming multidrug resistant (1). It has been estimated that 10 million deaths will occur every year due to infection with antimicrobial-resistant pathogens by 2050 (3); however, that may be an underestimate because we had already reached 5 million deaths in 2019 (4). One such antimicrobial-resistant pathogen the U.S. Centers for Disease Control and Prevention (CDC) has labeled a “serious threat” is the microorganism Shigella (5). Infection with Shigella species, termed shigellosis, is a bacillary dysentery often presenting bloody diarrhea, fever, and severe dehydration. Globally, infection with Shigella results in nearly 270 million cases and over 200,000 deaths per year, particularly in children under the age of 5 years (6). Of the current treatments, resistance of Shigella to ciprofloxacin or azithromycin has risen to 17% in clinically isolated strains in the United States (5). Worldwide, antibiotic resistance in Shigella is rising significantly, with “complete ciprofloxacin resistance (MIC ≥ 4 mg/L)” noted as a “serious concern” by the World Health Organization (7). Increased incidence of antibiotic resistance in Shigella, among other pathogens, and increased mortality highlight a critical need to develop novel antibiotics, vaccines, or other therapies to treat infection. One promising approach to developing improved therapeutics to treat shigellosis, and other infections, is targeting the virulence pathways by which the causative pathogens invade and propagate within infected hosts. By targeting virulence rather than cell viability, it is thought that there will be a weaker selective pressure to evolve resistance mechanisms to the drug (8). Additionally, targeting virulence pathways should have no effect on the avirulent microbiome, thereby reducing the risk of opportunistic infections—e.g., Clostridium difficile (8).

Shigella species are extremely adapted and virulent human pathogens. One of the most virulent species, Shigella flexneri, is also the most prevalent globally. Shigella flexneri is a species of Gram-negative, facultative anaerobic bacteria that invade the colonic epithelium to create their replicative niche. Unlike most bacterial pathogens, which often require millions of cells to establish an infection, S. flexneri is known to establish an infection with as few as 10 cells (9). After ingestion, S. flexneri can resist the high acidity of the stomach and then migrate to the basolateral membrane of the colon, where it penetrates the epithelial layer (10, 11). S. flexneri relies on the transcriptional activator VirF to activate its virulence pathway. VirF, an AraC family protein encoded by the virF gene present on a 230-kbp virulence plasmid (pINV), is expressed only under conditions found in the host colonic lumen (12–15). VirF binds to both the virB and icsA promoters, leading to activation of transcription of both genes (16–21). Upon activation of virB, the resulting secondary transcription factor VirB activates transcription of ipaB, ipaC, and ipaD, as well as other genes encoding structural and effector proteins leading to the expression of the type 3 secretion system (18, 19). This allows the bacteria to invade the colonic epithelium and manipulate the host immune response (e.g., escape from macrophages). VirF also activates transcription of icsA, initiating cell-to-cell spread by facilitating the polymerization of host cell actin at one pole of the bacterium (20, 21). Additionally, VirF represses transcription of the antisense RNA, RnaG (22). RnaG acts as an attenuator of virulence by downregulating translation of icsA (23).

VirF is a very promising antivirulence target. Initial studies reported that when VirF was inactivated via Tn5 insertion, there was a significant reduction in expression of four major virulence antigens (IpaB, IpaC, IpaD, and IcsA), which was restored when VirF was reintroduced (20). Sansonetti et al. (24) also reported that S. flexneri, when expressing a mutant form of IcsA, did not present colonic tissue damage in test animals as it did not possess the ability for cell-to-cell spread. The importance of IcsA in Shigella was further demonstrated by the Theriot lab, who made videos showing a dramatic drop in motility in a ΔicsA mutant (25). Compound screening has identified VirF inhibitors with 50% inhibitory concentration (IC₅₀) values of less than 100 μM in cell-based β-galactosidase reporter assays which also attenuated the virulence of S. flexneri in vivo (26–30). While these screening studies exemplify validated approaches to identifying VirF inhibitors, they failed to produce inhibitors potent enough to be tested in animal models.

AraC family proteins are widespread within bacteria and contain a highly conserved DNA-binding domain (DBD) consisting of two helix-turn-helix (HTH) motifs (16, 17). However, these proteins are often highly insoluble, making them difficult to purify and work with in vitro (31). There have been attempts to obtain native VirF to study in vitro, but the yields have been extremely low and insufficient for study (22). In contrast, much success has been had with MalE-VirF fusion proteins (18, 28, 29). Relatively few three-dimensional structures of AraC family proteins have been solved. Unfortunately, no crystal structure of VirF has been solved. This has slowed the analysis of the VirF-DNA binding interaction, thus reducing our ability to improve and develop more potent VirF inhibitors. Structures of several AraC family proteins from Escherichia coli and one from Vibrio cholerae have been solved (see Table S1 in the supplemental material). We selected two proteins, GadX and MarA, for use in homology modeling of the VirF DBD and in vitro analyses. GadX is involved in promoting acid resistance (32) and shows a 29% sequence identity to the DBD of VirF. The GadX structure (PDB no. 3MKL) does not include DNA and is likely reflective of the GadX conformation in solution.

MarA, a regulator of multiple-antibiotic resistance, has been studied extensively, and a crystal structure of it bound to its cognate DNA promoter marRAB (PDB no. 1BL0) has been solved by Rhee et al. (33), who identified amino acids in positions that very likely contribute to the protein-DNA binding interactions via hydrogen bonding or van der Waals (vDW) interactions (Table S2). In 2000, Gillette et al. (34) published their results of an in vivo alanine scan of 107 amino acids of MarA and evaluated their ability to activate transcription of five mar regulon promoters in vivo. Six of those mutants were purified, and their in vitro affinities for marRAB and micF were determined in a ³²P DNA binding assay.

Using GadX and MarA as models, we prepared two structural homology models for the VirF DBD: free and DNA bound. Here, we have conducted our own alanine scan of seven MarA residues that were identified by Rhee et al. to make base-specific interactions with marRAB (33). Two of these were studied in vitro by Gillette et al. (34); we added the other five identified by Rhee et al. and evaluated the affinities of all seven for the marRAB promoter DNA in vitro. Based on sequence alignments with MarA and the MarA-based VirF homology model, the corresponding residues in VirF were mutated to alanine and tested in vitro to both validate the MarA-based model of VirF in a DNA-bound conformation and elucidate key DNA-binding interactions between VirF and the virB promoter.

RESULTS

VirF DNA-binding domain homology modeling.

The VirF DBD was subjected to a homology search using SWISS-MODEL (35) to identify suitable structural templates. GadX (PDB no. 3MKL) had a sequence identity score of 29 and a global model quality estimate (GMQE) score of 0.76. MarA (PDB no. 1BL0) had a sequence identity score of 19 and a GMQE score of 0.70. As discussed above, these two structures were chosen as the templates for our homology modeling.

The qualities of the generated homology models (Fig. 1) were assessed by evaluating the stereochemical parameters on a Ramachandran plot using the Molecular Operating Environment (MOE) software (see Fig. S1A and B in the supplemental material). In the GadX-based homology model, there are no bond angles that form sterically disallowed confirmations. Two residues fall just outside the ideal tertiary structure regions (outlined in blue in Fig. S1A) but are still within the sterically allowed region for beta sheets (within the red outline). The MarA-based homology model is similar, but with more residues falling into the sterically allowed region of the plot for atoms that are closer together (outside the blue line but within the red in Fig. S1B). This model also contains one outlier, serine 38, which lies outside the sterically allowed region (Fig. S1B).

FIG 1.

VirF C-terminal DNA-binding domain (DBD) homology models and the structural templates MarA and GadX. (A) Crystal structure of MarA (PDB no. 1BL0 [green]) bound to the marRAB promoter DNA; (B) VirF DBD homology model (red) created using MarA as a template overlaid (RMSD = 3.57) with MarA bound to marRAB; (C) overlay from panel B with the MarA protein removed, leaving the homology model and marRAB DNA; (D) VirF DBD homology model (cyan) created using GadX (PDB no. 3MKL) as a structural template overlaid (RMSD = 0.61) with the GadX structure (purple); (E) both VirF DBD homology models, MarA based (red) and GadX based (cyan), overlaid (RMSD = 3.57) to show shift in confirmation.

When superimposing the VirF models with the templates (Fig. 1B and D), the percentage of identity is plotted above the sequence alignment and the root mean square deviation (RMSD) for each residue is color coded on the sequence (spectrum from green to red) (Fig. S1A). Several regions of the proteins have high identity and a small RMSD value, particularly the GadX-based model. In the MarA-marRAB structure, helix 3 and helix 6 are binding to the major groove on the DNA promoter (Fig. 1A and 2). Helix 6 shared a similar position in space among the two homology models and two crystal structures, whereas helix 3 showed variability in its placement, although the percentage of identity did not vary much between the two regions. In the MarA-based VirF model, helix 2 is shown as a free loop, but when overlaid with the MarA- or GadX-based model (Fig. 1B and E), it appears that this region has helical characteristics. This is likely an artifact of MarA having lower sequence identity to the VirF DBD. We decided to use the model that MOE created and chose not to impose this region into an alpha helix. The alignment of the homology model with MarA allowed us to visualize which side chains on these binding regions (helix 3 and helix 6) could potentially be contributing to VirF’s ability to bind to pvirB (Fig. 1 and 2).

FIG 2.

Spatial orientation of the seven amino acids in MarA (PDB no. 1BL0) that make base contacts with the marRAB promoter. Residues are color coded according to the alignment of the MarA and VirF DNA-binding domains indicated below. Small red spheres are water molecules.

MarA alanine mutants binding to the marRAB promoter.

DNA-binding assays were used to identify the binding activity of MarA for their cognate promoters. (The promoter sequences, fluorophores used, and method to obtain probes can be found in Table S5 and Fig. S2.) First, to evaluate the ability of MarA to bind to its DNA promoter, alanine scanning mutagenesis was utilized, creating seven MarA mutants. Three mutations were made to helix 3 in the first HTH motif (W42A, Q45A, and R46A), and four were made to helix 6 in the second HTH (Q92A, T93A, T95A, and R96A). To evaluate how these alanine mutations affected binding, an electrophoretic mobility shift assay (EMSA) was used to compare each mutant to the wild-type (WT) MarA (Fig. 3). The quantitative analyses to obtain the binding affinities for all MarA proteins for marRAB are presented in Fig. 4 and Table 1. Three of the mutants (W42A, R46A and R96A) did not exhibit saturation of the DNA at any concentration tested. When their data were plotted and fitted, none of them achieved a maximum binding extent (B_max) of >40%, indicating that these mutations dramatically impaired DNA binding. Consequently, the fitted parameters for these three mutants should be considered estimates. For R46A, the two highest concentration points were plotted but not used in the nonlinear regression. The helix 3 Q45A mutation slightly attenuated MarA’s ability to bind to the marRAB promoter, with an experimental equilibrium dissociation constant (K_D) value of 8.1 μM, relative to 4.8 μM for WT MarA. Interestingly, R46A exhibited significant binding to marRAB at concentrations of <3.5 μM (Fig. 3), but the band shift is weaker at concentrations higher than 3.5 μM (Fig. 4). This suggests that the mutant protein may be aggregating at higher concentrations. The only mutation on helix 6 that attenuated binding was R96A. This change significantly increased the experimental estimated K_D to 20 μM. The other three helix 6 mutations all slightly strengthened MarA’s affinity for marRAB, specifically Q92A, which presented a K_D of 0.23 μM. Both T93A and T95A improved binding to a lesser degree, displaying K_Ds of 2.0 μM and 2.5 μM, respectively. Circular dichroism (CD) spectra were collected for the WT and our weakest binding mutant (W42A) to determine if the W42A protein was folded properly (Fig. S3A and B). W42A presented a similar CD spectrum to the MarA WT, indicating it is likely folded properly. Since the other mutants exhibited binding in the EMSA, we surmise that all proteins were properly folded.

FIG 3.

EMSAs displaying the MarA WT and each MarA alanine mutant’s ability to recognize and bind the marRAB promoter. Each protein was tested in 2-fold serial dilutions from the starting concentration shown on the left of each gel (Table S6).

FIG 4.

Nonlinear regression analysis of MarA alanine mutants compared to MarA WT by EMSA. Binding affinities are presented in Table 1 with 95% confidence intervals presented in parentheses. Fits were constrained to B_max = 100, except for W42A, R46A, and R96A which did not reach full saturation. Every protein concentration data point was tested in duplicate, and error bars are included indicating the standard deviation. §, the two R46A data points circled were not included in the fit. †, for W42A, R46A and R96A, the fits are displayed as dashed lines and the fitted parameters (in italic in Table 1) are included as estimates only since the proteins were unable to fully saturate marRAB in the EMSAs.

TABLE 1.

Binding affinities of WT and alanine mutant MarA from the fits plotted in Fig. 4^a

Protein	Result for:
	WT	Mutant
		W42A	Q45A	R46A	Q92A	T93A	T95A	R96A
KD, μM (95% CI)	4.8 (2.9–6.5)	11 (6.6–20)	8.1 (6.1–10)	1.9b (0.40–12)	0.23 (0.14–0.33)	2.0 (1.5–2.4)	2.5 (2.0–3.0)	20 (9.2–49)
Bmax, % (95% CI)	100	28 (22–38)	100	34 (20–ND)	100 (88–ND)	100 (90–ND)	100	37 (26–62)

^aSee the Fig. 4 legend for details. For W42A, R46A, and R96A, fitted values are shown in italic and included as estimates only since the proteins were unable to fully saturate marRAB in the EMSAs. ND, not determined.

^bTwo R46A data points were not included in the fit.

Fluorescence polarization (FP) experiments were attempted as an additional method to assess the binding of each of our MarA mutants. We first tried a 5′-6-carboxyfluorescein (FAM) fluorophore on a 60-bp probe of the marRAB DNA (Table S5). Unfortunately, the data showed no trends in polarization when the protein was added at any concentration. We reasoned that the size difference between MarA and the large DNA probe was too small to adequately determine polarization if binding was occurring. To remedy this, we reduced the size of the probe from 60 bp to 32 bp while keeping the binding site intact (Fig. S2). Concerns about the fluorescence half-life led us to make three of these shorter probes, each with a different 5′ fluorophore (FAM, Cy3, or Cy5). Regrettably, the same issue was observed when using each of these new probes in the FP experiments. The polarization data showed no determinable trend while titrating in the MarA WT or MarA(W42A) or MarA(T95A) mutant (data not shown).

MalE-VirF alanine mutant binding to the virB promoter.

Seven MalE-VirF residues (I189, R192, and K193 on helix 3 and S238, Y239, I241, and R242 on helix 6) were mutated to alanine individually. The MalE-VirF numbering refers to the primary sequence of VirF from Shigella flexneri. To test how these alanine mutants bound to DNA, initially an EMSA was used (Fig. 5). The MalE-VirF I189A, Y239A, I241A, and R242A mutants all presented titratable binding shifts comparable to WT MalE-VirF. Conversely, the R192A and K193A mutants presented a lack of binding activity at the tested concentrations and the S238A mutant presented a visible reduction in DNA-binding activity. WT and active DNA-binding mutants presented similar affinities for pvirB according to their experimental K_D values and corresponding 95% confidence intervals (95% CIs) (Fig. S4). S238A MalE-VirF bound to pvirB with a similar affinity to the other mutants but visually presented weaker binding as well as a lower maximum extent of binding to the probe (B_max). Additionally, at higher concentrations for WT and mutant proteins, a significant amount of probe can be visualized in the wells (Fig. 5). This is likely due to protein aggregating at these higher concentrations (>~10 μM) as precipitate does form when the reaction mixtures are loaded into the gel. These data points are not included in the fits provided in Fig. S4. Probe binding to these aggregates is likely due to MalE-VirF on the surface of the aggregates binding specifically to pvirB since reaction mixtures containing boiled MalE-VirF at the same concentrations do not trap the probe in the well (data not shown). Overall, the trapping of probe in the well due to protein aggregation confounded the data analysis and binding fits of the EMSA results.

FIG 5.

EMSAs testing the binding of WT and MalE-VirF alanine mutants for pvirB. Each protein was tested in 2-fold serial dilutions from the starting concentration shown on the left of each gel (Table S6). For gels testing the titrations of S238A, I241A, and R242A against pvirB, the negative control (−) was cropped and placed to the right of the titrations for presentation consistency. One negative control was used for R192A and K193A titrations as they were tested together on one gel.

FP was used to quantitate the affinities of MalE-VirF for pvirB (Fig. 6 and Table 2). WT MalE-VirF presented a 2.3 μM affinity for pvirB. The MalE-VirF I189A mutant (4.2 μM) presented affinity for pvirB similar to that of the WT, with overlapping 95% confidence intervals. The I241A (6.5 μM) and R242A (10.1 μM) mutants, which showed qualitatively equivalent binding in the EMSA (Fig. 5 and Fig. S4), exhibited 3- to 4-fold reduced binding relative to the WT. The S238A mutant presented a 3-fold weaker affinity for pvirB (6.3 μM) than the WT. This is consistent with the EMSA of the S238A MalE-VirF mutant, where there was visually weaker binding than the WT and the other active DNA-binding mutants. As seen in the EMSA, no binding was observed in the FP for both the R192A and K193A mutants (Fig. S5). Given the loss of observed DNA binding for the R192A and K193A mutants, CD spectra were obtained for them and the WT to probe for protein misfolding. We also included S238A MalE-VirF in the study to observe CD spectra for a weaker-binding mutant. Overall, the CD spectra were very similar, suggesting that all proteins were folded similar to WT MalE-VirF (Fig. S3C and D). Additionally, the Hill slopes for WT and all mutants in the FP assay ranged from 0.7 to 1.5 (Fig. 6). This indicates that the proteins are not likely aggregating (Hill slope of ≫1) and are likely binding with a 1:1 ratio of protein to DNA. Currently, it is not known if MalE-VirF must dimerize to bind to DNA promoters, so it is unknown if this 1:1 binding ratio is with a monomer or preformed dimer. Interestingly, the Y239A mutant did not present optimal binding to pvirB in the FP, unlike in the EMSA (Fig. S6). Y239A MalE-VirF bound to pvirB with an affinity of 11 μM but had an extremely high Hill slope of 16, potentially indicating aggregation in solution. We tested for Y239A MalE-VirF aggregation using a turbidity assay, but no significant aggregation was observed (data not shown). Due to these inconsistencies, we did not include the Y239A mutant data in Fig. 6.

FIG 6.

Nonlinear regression analysis of WT MalE-VirF and DNA-binding alanine mutants binding to a fluorescein-labeled virB promoter in the FP assay. Sigmoidal curves were generated to calculate the binding affinities (K_D), Hill slopes, and spans (mP differences between the maximums and minimums of the fits) of the I189A, S238A, I241A, and R242A mutants and WT MalE-VirF as presented in Table 2. In Table 2, for each affinity and Hill slope, 95% confidence intervals (95% CI) are presented in parentheses. Every protein concentration data point was tested in triplicate, and error bars indicate the standard deviation.

TABLE 2.

Binding affinities of WT and MalE-VirF and DNA-binding domain alanine mutants from the fits plotted in Fig. 6^a

Protein	Result for:
	WT	Mutant
		I189A	S238A	I241A	R242A
KD, μM (95% CI)	2.3 (1.7–2.9)	4.2 (2.7–18)	6.3 (5.1–7.9)	6.5 (5.2–8.3)	10.1 (7.7–15)
Hill slope (95% CI)	1.1 (0.9–1.3)	1.1 (0.6–1.8)	1.1 (0.9–1.3)	1.2 (0.9–1.5)	0.8 (0.6–1.1)
Span, mP	38	38	33	35	34

^aSee the Fig. 6 legend for details.

Based on footprinting analysis, VirF has been shown to have two binding sites on pvirB (binding sites are underlined in Table S5) (18). Additionally, VirF belongs to a subclass of AraC family proteins that are thought to dimerize (17, 36). VirF’s DNA-binding interaction could be unlike MarA’s interaction with marRAB since there is only one identified binding site and MarA lacks a dimerization domain and is therefore monomeric (33, 34). To better elucidate the binding interaction between VirF and pvirB, we independently scrambled each binding site on the promoter and tested DNA binding in the EMSA (Fig. S7). WT MalE-VirF bound nearly identically to both the wild-type (pvirB) and “binding site 2 scrambled” (BS2 Scram) probes. However, when binding site 1 was scrambled (BS1 Scram), there was a significant reduction in binding present when >6 μM WT MalE-VirF was tested. Additionally, we were unable to visualize a second binding shift, which could indicate either two binding events, or dimerization, when MalE-VirF interacted with pvirB. These observations indicate that binding site 1 is a higher-affinity site for MalE-VirF (comparable to MarA binding marRAB) and suggest that binding to site 2 may be significantly facilitated by MalE-VirF dimerization.

MalE-VirF and MarA bind specifically to their DNA promoters.

Despite high sequence similarity in the DNA-binding domains (DBDs) of AraC family proteins, they are known to bind to their DNA promoters with high specificity (16, 17). To test the DNA-binding specificity of these proteins, we performed EMSAs testing WT and mutant MalE-VirF and MarA proteins binding to their noncognate promoters marRAB and pvirB, respectively. When WT and mutant MalE-VirF proteins were tested at 5 μM against marRAB, there were no visible binding shifts present even at the highest testable concentration, with WT MalE-VirF at 23 μM (Fig. 7A). We also verified that none of the individual mutations to MarA created any affinity for pvirB (Fig. 7B), even at the highest concentration testable (data not shown). To further probe the binding specificity, we ordered from Twist Bioscience a MarA mutant containing each of the MalE-VirF amino acids that were mutated in the alanine scan in place of the aligned MarA amino acids (W42I, Q45R, R46L, Q92S, T93Y, and T95I). The mutant MarA showed no binding affinity for pvirB or marRAB in the EMSA at 50 μM (Fig. S8).

FIG 7.

EMSA testing the binding of WT and DNA-binding mutants for (A) MalE-VirF for the marRAB promoter and (B) MarA for pvirB. All MalE-VirF mutants were tested at 5 μM; the WT, which contained the highest testable concentration of WT MalE-VirF, was tested at 23 μM. All MarA mutants were tested at the highest possible concentrations as shown in Table S6.

DISCUSSION

Homology modeling is a powerful tool when experimental determination of the three-dimensional structure of a protein (or other large biomolecule) is not feasible and structures of homologous proteins are available. Although many attempts have been made, there are no experimental structures of full-length VirF or the VirF DBD. VirF is a member of the AraC family of transcriptional regulators, which is widely distributed across Gram-negative bacteria and is typically involved in carbon metabolism, stress response, and virulence. Members of this family have a highly conserved C-terminal domain that contains two helix-turn-helix (HTH) DNA binding motifs (16, 17). This universally conserved region of the protein family allowed us to find several templates in the Protein Data Bank (PDB) (see Table S1 in the supplemental material) with high sequence coverage and strong GMQE scores to the VirF DBD (amino acids 144 to 262 in WT VirF).

The two structures selected to serve as the templates for homology modeling were GadX (PDB no. 3MKL) and MarA (PDB no. 1BL0). In the GadX structure, the protein is not bound to DNA. In the MarA structure, it is bound to the marRAB promoter (Fig. 1A). When superimposed, the MarA- and GadX-based VirF homology models have slightly different orientations of the HTH DNA-binding domains (Fig. 1E). This could potentially represent bound and unbound conformations of the VirF DBD, but further analysis is needed to be certain. By overlaying the MarA-based VirF DBD homology model with the MarA crystal structure, not only do we get an idea of how the DBDs align (Fig. 1B), but we can visualize how the VirF DBD might interact with pvirB (Fig. 1C). The function of each HTH motif in the AraC family of transcriptional regulators has been somewhat debated (16, 37, 38), but slight variations in the orientation of both helices may be needed for each protein to bind its cognate promoter.

Examination of the crystal structure of MarA in Fig. 2 reveals three amino acid side chains in helix 3 and four in helix 6 that interact with DNA. Each of these side chains extends out of the helical core and makes contacts with the marRAB promoter DNA bases directly or via an intervening water molecule (Table S2) (33, 34). By aligning the primary VirF sequence and overlaying our homology models with the MarA structure, we were able to identify VirF DBD amino acids that correspond to these seven MarA residues (Fig. 2).

To probe the interactions of MarA with marRAB, we performed alanine scanning mutagenesis in vitro on each of the DNA-interacting residues in MarA and evaluated their affinities for marRAB via electrophoretic mobility shift assays (EMSAs). The K_D for MarA WT that we determined in this study diverges significantly from what was reported by Gillette et al. (34). It has been shown that deletions to the marRAB promoter, specifically to the binding site (i.e., ΔmarO281), significantly reduce the affinity of MarA to DNA by up to 100-fold (39). Hence, we tested binding of our purified WT MarA to both our 60-bp marRAB promoter and the 143-bp promoter used by Gillette et al. (34) (Fig. S9A and B). MarA affinities for both the 143- and 60-bp probes were not significantly different under our assay conditions and did not significantly vary when determined in Gillette dialysis buffer (50 mM HEPES, 500 mM NaCl, 250 mM imidazole, 25% glycerol [pH 8.5]) (Fig. S9C to F). Upon running the EMSA according to the Gillette protocol, comparable binding affinities were again observed (Fig. S9G and H). Changes to experimental conditions and promoter length have resulted in reported K_Ds ranging from ~3 to 75 nM for the MarA-marRAB binding interaction (34, 39–43). While our K_D is higher than the highest previously reported K_D, it is consistent with our determination of the K_D for the DNA promoter used by Gillette et al. (34).

Overall, our observations are largely consistent with those from previous studies, in that alanine substitutions of R46 and R96 severely reduced MarA’s affinity for marRAB DNA. While it is obvious that R46A decreased MarA’s binding affinity (Fig. 4), attempts to determine its K_D were unsuccessful (Fig. 5). Interestingly, when we replaced W42 with alanine, despite showing little decrease in MarA activity in vivo (34), we observed a 2.3-fold decrease in binding affinity relative to the WT (Table 3). The cumulative effect of losing the three vDW contacts that W42 makes with C31, C32, and T18 of the marRAB DNA is likely responsible for the decrease in affinity (Table S2). Replacing R96 with alanine resulted in a 4.1-fold decrease in binding affinity, the largest decrease of all the alanine substitution mutants (Table 3). In general, alanine substitutions made to helix 3 residues resulted in larger decreases in binding affinity than those in helix 6. While R96A dramatically decreased binding affinity, the Q92A, T93A, and T95A mutants each displayed an increase in binding affinity for marRAB. The Q92A mutant showed a surprising increase in affinity, increasing the K_D more than 20-fold compared to that of the WT (Table 3). These findings align with observations made by Gillette et al. that the protein is more sensitive to alterations made to its N-terminal HTH (helices 2 and 3) than to alterations in the C-terminal HTH (helices 5 and 6) (34).

TABLE 3.

Comparison of the binding affinities of WT and mutant MarA and MalE-VirF proteins for their cognate promoters^a

MarA mutation	KD (μM) determined by EMSA	Relative KD	MalE-VirF mutation	KD (μM) determined by FP	Relative KD
WT	4.8	1		2.3	1
Mutant
W42A	11b	2.3	I189A	4.2	1.8
Q45A	8.1	1.7	R192A	NDA
R46A	1.9b	0.40	K193A	NDA
Q92A	0.23	0.04	S238A	6.3	2.7
T93A	2.0	0.42	Y239A	11c	4.8
T95A	2.5	0.57	I241A	6.5	2.8
R96A	20b	4.1	R242A	10.1	4.4

^aRelative K_Ds compare each mutation to its WT counterpart. NDA, no detectable activity.

^bEstimated K_D determined from Fig. 4.

^cAggregation was inferred due to very steep Hill slope in the FP experiment (Fig. S6).

We also performed alanine scanning mutagenesis in vitro on each of the corresponding residues in the MalE-VirF DBD. Overall, there were similar trends seen for both MarA and MalE-VirF mutants binding to their cognate promoters (Table 3), validating our MarA-based VirF DBD homology model. While both WT proteins bound with micromolar affinity to their cognate promoters, mutation of amino acids capable of hydrogen bonding or electrostatic interactions with their cognate promoter (the MarA R46 and R96 mutants and MalE-VirF R192, K193, and R242 mutants) showed reductions or loss of DNA-binding activity. While R192 and K193 MalE-VirF alanine-mutants led to a complete loss of binding activity, R242A in MalE-VirF exhibited a >4-fold change in binding relative to WT MalE-VirF but retained activity in both the FP assay and EMSA (Fig. 5 and 6). Similarly, R96A MarA showed a 4-fold decrease in binding relative to MarA WT (Fig. 4). Given that these residues are identical in both proteins and are likely to make electrostatic or hydrogen-bonding interactions with their DNA promoters, this similarity in affinity upon mutation is unsurprising.

Not much is known about which amino acids are crucial for VirF DNA-binding activity. Porter and Dorman (36), employed random and site-directed mutagenesis on full-length virF to test the ability of the protein to activate gene expression of an mxiC-lacZ fusion protein, which was measured by β-galactosidase activity in vivo (36). These results are comparable to ours since VirF activates transcription of virB (via binding to pvirB), which in turn positively regulates mxiC. It was determined that mutations in both the VirF N-terminal domain and the C-terminal DBD significantly affected gene transcription, but further studies were not performed to determine how these mutations affected DNA binding. They tested similar mutants with DBD mutations K193A, Y239F, and I241N that showed no activity in their β-galactosidase assay (36). Consistent with Porter and Dorman’s report for K193A VirF, we observed a lack of DNA-binding activity for K193A MalE-VirF at the protein concentrations tested in the EMSA and FP assay (Fig. 5 and Fig. S5) (36). This is not surprising since losing a positively charged amino acid in the DBD could have deleterious effects on binding simply via loss of an electrostatic interaction between the terminal charged amine of lysine and the phosphate backbone of pvirB. Alternatively, there could also be a loss of hydrogen bonding between the base pairs of the promoter and the side-chain amine of lysine after mutation. Similarly, we did not see any binding associated with R192A MalE-VirF, likely for similar reasons. As seen with MarA, mutations to residues in the N-terminal HTH resulted in reduced affinity or loss of DNA-binding activity (Table 3).

Conversely, we did see DNA-binding activity in both assays with our Y239A and I241A MalE-VirF mutants (Table 3). Porter and Dorman reported a loss of in vivo activity with the corresponding Y239F and I241N VirF mutants (36). For our Y239A MalE-VirF mutant, we unexpectedly observed binding similar to WT in our EMSA (Fig. 5). Porter and Dorman did see expression of Y239F VirF in a Western blot (36). In vivo aggregation or misfolding of Y239F VirF could cause a reduction in β-galactosidase activity, and this is consistent with our FP data since the very high Hill slope of Y239A MalE-VirF in the FP assay suggested significant aggregation (Fig. S6). Despite the suggestion of aggregation of the Y239A mutant in the FP assay, we suspect that Y239 participates via hydrophobic or vDW interactions with pvirB or potentially through a weak/suboptimal hydrogen bond. Similarly, Porter and Dorman saw a loss of activity with their I241N VirF mutant (36). However, our corresponding I241A MalE-VirF mutant presented a modest reduction in binding compared to WT MalE-VirF in the FP assay. I241A MalE-VirF incorporates a relatively conservative change to the protein, maintaining the hydrophobic nature of the amino acid, albeit while significantly reducing the steric bulk. The modest change in DNA-binding activity of I241A MalE-VirF compared to the WT suggests that I241 interacts with pvirB mostly through weak hydrophobic or vDW interactions since these interactions involve less energy than hydrogen bonding or electrostatic interactions. On the other hand, Porter and Dorman’s VirF I241N mutation introduced a nonconservative change to asparagine which altered the polarity/hydrophobicity at that residue, likely accounting for its reduced DNA-binding activity (36). Additionally, protein expression was seen in a Western blot, so it is likely that a nonconservative change at this position led to a decrease in β-galactosidase activity (36).

The other three mutants, I189A, S238A, and R242A MalE-VirF, had not been previously studied. As expected, the mutation I189A in MalE-VirF introduced a hydrophobically conservative change to the protein, and we did not observe a significant difference in DNA-binding activity. This indicates that I189 interacts with pvirB most likely through weak hydrophobic or vDW interactions. S238A MalE-VirF showed a decrease in binding affinity for pvirB in the FP assay. The loss of the serine hydroxyl led to an ~3-fold reduction of DNA-binding activity relative to the WT (Fig. 6 and Table 3). It seems likely that S238 participates via weak hydrophobic or VDW interactions, despite containing a hydroxyl which could contribute to a hydrogen-bonding interaction with pvirB. A loss of a hydrogen-bonding interaction would likely lead to a >10-fold loss in DNA-binding affinity, so this suggests that S238 might not contribute to the pvirB binding interaction via hydrogen bonding, but we cannot rule out a weak hydrogen bond. Finally, R242A MalE-VirF bound to pvirB ~4-fold weaker than the WT. Initially, it was expected that an alanine at this position would elicit a significant (>10-fold) reduction in DNA-binding activity via loss of assumed electrostatic interactions. However, the mutation of this arginine led to a very modest loss of affinity, unlike the dramatic loss of binding seen with the R192A and K193A mutants. Therefore, it seems unlikely that the R242 side chain contributes to the pvirB binding interaction via electrostatics or hydrogen bonding. However, we cannot rule out that R242 could be interacting with pvirB via a suboptimal/weak hydrogen bond or hydrophobic and vDW interactions. A summary of all these speculated interactions based on the experimental data is presented in Fig. 8. Overall, if any of these mutations, except I189A, occurred in S. flexneri, it is possible these decreases in binding activity could have a negative impact on its virulence phenotype. It has been shown that VirF needs to reach a threshold of activity (~40% relative to Shigella WT grown at 37°C) to activate the virulence pathway (44), so these mutations could be sufficient to be detrimental to the bacteria.

FIG 8.

Diagram of studied MalE-VirF amino acids and their potential binding interactions with pvirB based on the experimental data. The diagram uses the MarA-based homology model (Fig. 1C).

AraC family proteins are known to bind specifically to their cognate promoters. We confirmed that neither MarA nor MalE-VirF WT nor their alanine mutant proteins were capable of binding to the other noncognate promoter used in this study (Fig. 7). This is consistent with the fact that they recognize different DNA promoter sequences. Even with mutation of all seven residues of MarA to the corresponding VirF residues, there was no binding detected when tested with pvirB (Fig. S8). This indicates that DNA binding and recognition is more complex than simply the identity of the seven amino acids at these key positions. Both MarA and VirF bind to other DNA promoter sequences in addition to the promoters used within this study. MarA binds to specific promoter sequences termed the marbox, which control transcription of >15 genes in the mar regulon (45, 46). VirF is known to bind three distinct promoters which regulate transcription of two downstream genes (virB and icsA) and an antisense RNA, RnaG (18, 22). We have identified specific amino acids on both helices 3 and 6 that are required for MalE-VirF DNA-binding affinity and specificity for pvirB over a noncognate promoter, marRAB. We have also identified amino acids within these that show little effect on DNA binding to pvirB when mutated to alanine. It is certainly possible that the same mutations may have much more significant effects on MalE-VirF binding to the picsA and prnaG promoters. Differential DNA-binding activity of these MalE-VirF mutants for picsA and prnaG is currently being explored.

In contrast to MarA, VirF contains an N-terminal dimerization domain. This adds another layer of complexity to VirF that is absent in MarA. It has been shown that a form of VirF (VirF₂₁) lacking a portion of its N-terminal domain is capable of binding to the virF promoter (via footprinting analysis) and negatively autoregulates the expression of full-length VirF; however, it was not shown to bind to the virB promoter (47). This suggests that the N-terminal dimerization domain may influence the DNA-binding domain’s affinity for its promoters. As presented in Fig. S7, we found that MalE-VirF preferentially binds to one of the two binding sites on pvirB in the EMSA, which were first identified via footprinting analysis (18). This result made the comparison between MarA and MalE-VirF more feasible as these DNA-binding interactions are more closely comparable with both presenting a 1:1 binding interaction under the conditions of our study. Although VirF is thought to dimerize, the dimerization activity of VirF has been severely understudied and the order in which promoter binding and dimerization occur to activate virulence gene transcription is unknown. Porter and Dorman reported that when coexpressed with WT VirF and tested in a β-galactosidase assay, only two VirF mutations presented a dominant-negative effect, suggesting that the protein dimerizes (36). There have not been any reported studies that directly test the dimerization activity of VirF and its effects on DNA binding. Fortunately, AraC and other homologs such as ToxT from V. cholerae and HilD from Salmonella enterica serovar Typhimurium, have been studied (48–50). It has been shown that changes to the dimerization domain, either through effector binding (AraC) or site-directed mutagenesis (ToxT), altered the abilities of these proteins to bind to their respective DNA promoters (51, 52). Further studies are being performed to elucidate VirF dimerization and how mutations to this domain affect DNA binding.

In conclusion, we have generated models of the VirF DBD using GadX and MarA-marRAB structures. The two models of VirF, based on GadX and MarA, suggest a conformational change of the AraC proteins when they bind to DNA (Fig. 1). Based on alignments with MarA, we tested seven alanine mutants within the DBD of the MalE-VirF fusion protein. Comparisons of the effects of the alanine mutants on DNA binding validated the MarA-based VirF model and identified key interactions between VirF and one of its cognate promoters, pvirB. These DNA-binding assays, in conjunction with the observations we made of the VirF DBD through homology modeling, give a more detailed understanding of how VirF interacts with pvirB. Given the lack of structural data for this protein, these results and models will be useful in the continued efforts to analyze VirF’s ability to bind its other promoters as well as identify and improve upon current DNA-binding inhibitors for the novel treatment of shigellosis.

MATERIALS AND METHODS

Materials.

All standard buffer components were purchased from Millipore Sigma or Thermo Fisher. Specific reagents or biological products not purchased from these are noted in parentheses. DNA oligonucleotides were purchased from Integrated DNA Technologies (IDT). Equipment utilized for these experiments was purchased from various companies which are indicated in parentheses throughout this section.

Alignment and homology modeling.

Using SWISS-MODEL, the Protein Data Bank database was searched for structures that showed homology with the C-terminal DBD of VirF (35, 53–56). The two templates selected, GadX (PDB no. 3MKL, 2.15-Å resolution) and MarA (PDB no. 1BL0, 2.30-Å resolution), stood out at the top of the list, with high sequence coverage, identity/homology, and global model quality estimate (GMQE) scores. To develop the VirF DBD models, the protein sequence was entered into the MOE (Molecular Operating Environment) platform (57). Here, the sequence was used to search the PDB database to obtain the GadX and MarA structures that we wanted to use as the templates, and both structures were opened for analysis in the program. In the homology model window, the VirF DBD sequence was entered as the sequence to model and each template was selected as a primary structure template in two separate consecutive modeling runs. The geometries of the models with the highest RMSDs for each template were evaluated and selected as the representative homology model. RMSD values were calculated using MOE. The protein sequences were aligned and analyzed using MOE, giving a percentage of identity match. MarA and VirF DBDs were also aligned and evaluated using EXPASY SIM-Alignment Tool (58) (Fig. 2).

Alanine scanning mutagenesis.

Site-directed mutagenesis of the marA and malE-virF genes was performed using oligonucleotides described in Table S3 in the supplemental material. The alanine mutation was incorporated using the GCG alanine codon for each MalE-VirF mutant (I189, R192, K193, S238, Y239, I241, R242) and each MarA mutant (W42, Q45, R46, Q92, T93, T95, R96), except for the R46A mutation, which contains the GCT codon. MalE-VirF numbering refers to the primary sequence of VirF from Shigella flexneri. Two-step PCR was used on pET-15b marA and pBAD202-MALVirF with PFU Turbo DNA polymerase in a MiniAmp Plus thermal cycler (Thermo Fisher). All mutations were confirmed by DNA sequence analysis.

MarA expression and purification.

Our procedure to express and purify MarA was adapted from that of Jair et al. (59). Starter cultures (10 mL) of E. coli BL21(DE3) (Table S4) containing pET15b-marA were grown overnight in 2× TY broth (16 g Bacto tryptone, 10 g yeast extract, and 5 g NaCl per L of water) supplemented with carbenicillin at 37°C under vigorous agitation. The next day, the starter culture was used to inoculate 1 L of 2× TY broth supplemented with carbenicillin. The cells were grown to an optical density at 600 nm (OD₆₀₀) of 0.8 before expression was induced with the addition of isopropyl-β-d-1-thiogalactopyranoside (IPTG) at a final concentration of 0.4 mM, and the culture continued to shake overnight at 16°C. Cells were then harvested by centrifugation (6,000 × g, 4°C, 15 min) before being resuspended in 25 mL MarA lysis buffer (50 mM Tris-HCl, 1 mM EDTA, 1 M NaCl [pH 7.5]) supplemented with a Roche cOmplete miniprotease inhibitor cocktail tablet (Roche) and 0.1 mM phenylmethylsulfonyl fluoride (PMSF). All of the proceeding steps were performed on ice or at 4°C. Cells were lysed via sonication (8 cycles, 15-s pulse, 3-min intervals, 60% setting) utilizing an ultrasonic XL2020 sonicator (Misonix). Following sonication, the solution was pelleted by ultracentrifugation (120,000 × g, 4°C, 30 min). The supernatant was discarded, and the pellet was washed with 30 mL of MarA denature buffer 1 (50 mM Tris-HCl, 4 M urea [pH 8.5]) before the ultracentrifugation was repeated. The supernatant was discarded again, and the pellet was resuspended in 25 mL MarA denature buffer 2 (50 mM Tris-HCl, 6 M guanidinium chloride [pH 8.5]). The mixture was subjected to ultracentrifugation a third time, and the supernatant was collected. Next, 2 mL of Ni-nitrilotriacetic acid (NTA) agarose (Qiagen) was added, and the solution was rocked gently overnight at 4°C. The following day, the resin slurry was poured into an empty purification column. Once settled, the resin bed was washed stepwise with increasing concentrations of imidazole by combining MarA elution buffer (50 mM Tris-HCl, 500 mM NaCl, 1 M imidazole [pH 8.5]) and MarA wash buffer (50 mM Tris-HCl, 500 mM NaCl [pH 8.5]) with 100, 300, 500, and 750 mM imidazole. The fraction eluted with 300 mM imidazole contained MarA, was verified by SDS-PAGE, and was dialyzed with a Slide-A-Lyzer dialysis cassette (3-kDa molecular weight cutoff [MWCO]; Thermo Fisher) at 4°C overnight in MarA dialysis buffer (50 mM Tris-HCl, 250 mM NaCl, 0.1% Triton X-100, 20% glycerol [pH 8.0]). The protein concentrations were determined with a Bradford assay (Bio-Rad) using bovine serum albumin (BSA) standards. Protein stocks were flash frozen in liquid nitrogen and then stored at −80°C.

MalE-VirF expression and purification.

WT and mutant MalE-VirF proteins were expressed similarly to those previously described by Emanuele and Garcia (28). E. coli TOP10 cells (Table S4) were transformed with the MalE-VirF expression plasmid pBAD202-MALVirF. TOP10 cells harboring the expression plasmid and supplemented with kanamycin were grown, and protein expression was induced as previously described (28). The cells were resuspended in MalE-VirF binding buffer (20 mM Tris-HCl, 1 mM EDTA, 1 mM dithiothreitol [DTT], 500 mM NaCl [pH 7.4]) and supplemented with a Roche cOmplete miniprotease inhibitor cocktail tablet (Roche) and 0.1 mM PMSF. Following resuspension, all subsequent steps were on ice or kept at 4°C. The resuspended cells were lysed via sonication (6 cycles, 30-s pulse, 4-min rest, 60% of maximum pulse setting) using an ultrasonic XL2020 sonicator (Misonix). Cellular debris was separated via centrifugation (45 min, 25,000 × g, 4°C). Following centrifugation, the supernatant was loaded onto a 5-mL MBPTrap HP column (Cytiva) using an AKTA fast protein liquid chromatography (FPLC) column (GE Healthcare). The column was then washed with 15 column volumes (CVs) of MalE-VirF binding buffer. Protein was eluted from the column using MalE-VirF elution buffer (20 mM Tris-HCl, 1 mM EDTA, 1 mM DTT, 500 mM NaCl, 5% glycerol, 10 mM maltose [pH 7.4]). Eluent was collected in 1.5-mL fractions. Fractions with the highest resulting at a UV absorbance at 280 nm were collected and concentrated to approximately 600 μL using an Amicon Ultra-15 kDa MWCO centrifugal filter units and then filtered to remove any precipitate. The resulting filtered protein was loaded onto a Superdex 200 GL10/300 gel filtration column to separate MalE-VirF from truncated MalE and other protein impurities eluted from the MBPTrap HP column. Protein was eluted from the column using MalE-VirF binding buffer, and 0.5-mL fractions were collected and tested via SDS-PAGE to determine where MalE-VirF eluted. The corresponding MalE-VirF fractions were dialyzed with a Slide-A-Lyzer dialysis cassette (10-kDa MWCO; Thermo Fisher) at 4°C overnight in MalE-VirF dialysis buffer (20 mM Tris-HCl, 1 mM EDTA, 5 mM DTT, 200 mM NaCl, 40% glycerol [pH 7.4]). The concentration of each protein was tested with a Bradford assay (Bio-Rad) using BSA standards. Protein stocks were flash frozen in liquid nitrogen and then stored at −80°C.

Electrophoretic mobility shift assays.

EMSAs were performed as previously described by Emanuele and Garcia (28). Prior to preparing the gel and reactions, the Cy5-labeled virB (pvirB) and marRAB oligonucleotide promoter probes were annealed as previously described using the oligonucleotide primers found in Table S5 (28). The marRAB probe contains the MarA binding site at the center of the 60-bp DNA, whereas pvirB, also 60 bp, contains two VirF binding sites with extra flanking bases to support DNA duplex stability (18, 28, 33). Although in the crystal structure, MarA is bound to a 20-bp DNA fragment, the length of our probe was increased to support duplex stability and for consistency with previous MalE-VirF EMSAs (28, 33). A 6% native polyacrylamide gel was prepared using 30% acrylamide–bisacrylamide (29:1 ratio) solution and TBE buffer (0.25× final concentration; 22 mM Tris base, 22 mM boric acid, 0.5 mM EDTA [pH 8.5 or 9.5]). EMSAs containing MarA proteins and the marRAB promoter used 0.25× TBE gel and running buffer at pH 8.5, whereas EMSAs containing MalE-VirF proteins and pvirB were run with the same buffer at pH 9.5. MalE-VirF was unable to enter the native polyacrylamide gel when the 0.25× TBE buffer was below pH 9.5. Prior to running the reactions, the empty gel was electrophoresed for 1 h at 150 V in 0.25× TBE. For MarA EMSAs, each reaction mixture was composed of 3 μL marRAB probe (42 nM), 9 μL MarA WT or mutant protein, 1 μL salmon sperm DNA (0.7 mg/mL; Invitrogen), 0.5 μL BSA (0.07 mg/mL), and 1.5 μL Milli-Q H₂O for a final volume of 15 μL. Final concentrations of DNA probe, salmon sperm DNA, and BSA within each reaction mixture are presented in parentheses. Final buffer conditions were 40 mM Tris-HCl, 160 mM NaCl, 0.04 mM EDTA, 0.06% Triton X-100, and 12% glycerol (pH 8.0). Serial dilutions were prepared by diluting the highest possible concentration for each protein with MarA dialysis buffer. Additionally, we ordered and obtained a 143-bp 5′-Cy5-labeled marRAB promoter from IDT, which was used to compare our assay protocol and DNA promoter with those utilized by Gillette et al. (34).

For testing with MalE-VirF, reaction mixtures were prepared with 6 μL MalE-VirF WT or mutant protein, 6 μL pvirB (83 nM), 1.5 μL Milli-Q H₂O, 1 μL salmon sperm DNA (0.7 mg/mL), and 0.5 μL BSA (0.07 mg/mL). Final buffer conditions were 12 mM Tris-HCl, 100 mM NaCl, 16% glycerol, 2 mM DTT, and 0.44 mM EDTA (pH 7.4). To perform titrations of all MalE-VirF proteins against pvirB, 2-fold dilutions were prepared using MalE-VirF dialysis buffer. The MalE-VirF and MarA concentration ranges that were tested in the assay can be found in Table S6. MarA and MalE-VirF WT and mutant protein concentrations varied by purification yield. Negative-control reaction mixtures were prepared with 6 or 9 μL native gel loading dye (300 mM Tris-HCl, 50% glycerol, 0.05% bromophenol blue [pH 7]) instead of MalE-VirF or MarA proteins, respectively. All reaction mixtures were incubated at 37°C for 15 min in a water bath before addition of 6 μL of each reaction mixture to the corresponding wells on the gel. The gel was electrophoresed at 150 V for an additional 1.5 h in the dark at 4°C. Gel visualization was performed using a Molecular Dynamics Typhoon 9200 molecular imager by excitation (Ex) at 607 nm and reading the 710-nm emission (Em). Quantitative data were obtained by measuring the density of the bands on the gel using ImageJ software (60). Prism 9 software (61) (using the “one-site specific binding” equation) was used to generate binding affinities (K_D) and B_max (percentage).

Fluorescence polarization assay.

The fluorescence polarization (FP) assays were performed as described by Emanuele and Garcia (28). The assays were conducted in low-volume round-bottom 384-well plates (Corning). The fluorophore-labeled pvirB and marRAB oligonucleotide probes were annealed as previously described using the oligonucleotide primers found in Table S5 (28). The top strands of the pvirB or marRAB promoters (Table S5) contained either a 5′-fluorescein, 5′-Cy3, or 5′-Cy5 and were annealed to their corresponding, unlabeled bottom strands to prepare stocks with concentrations of 5 μM. The pvirB probe contained a 5′-fluorescein, whereas the marRAB promoter probes were prepared with all three fluorophores, individually. The DNA probes were diluted with probe buffer (50 mM Tris-HCl, 1 mM EDTA, 5 mM DTT, and 200 mM NaCl [pH 7.4]) to a final working concentration of 20 nM and supplemented with 1.4 mg/mL salmon sperm DNA (Invitrogen) and 0.14 mg/mL BSA. Next, 14-step 2-fold serial dilutions of the MarA WT or MarA mutant were prepared with MarA dialysis buffer, and then 10 μL of each dilution was added to the appropriate wells in triplicate for each concentration tested. For MalE-VirF, 2-fold dilutions were prepared using MalE-VirF dialysis buffer prior to loading of 10 μL of each dilution onto the plate. MalE-VirF and MarA concentration ranges tested in the assay can be found in Table S6. Following loading of protein into the wells, 10 μL of the 20 nM DNA probe solution was added to each well. The final concentrations of DNA probe, salmon sperm DNA, and BSA in each reaction mixture are 10 nM, 0.7 mg/mL, and 0.07 mg/mL, respectively. Negative-control reactions were also tested in triplicate with mixtures containing 10 μL MarA or MalE-VirF dialysis buffers and 10 μL of their corresponding 20 nM DNA probe solutions to determine the baseline FP for each DNA probe. Additionally, for each protein concentration tested, a blank reaction mixture containing the corresponding tested protein concentration and the DNA probe solution lacking the labeled probe was used to subtract fluorescent contributions of MalE-VirF or MarA, salmon sperm DNA, BSA, and other buffer components from the test reaction mixtures. The plate was incubated at 37°C for 2 h before the raw fluorescence was measured using a Biotek Synergy H1 plate reader after excitation at the appropriate wavelength for the corresponding pvirB and marRAB probes (fluorescein Ex/Em = 485/528, Cy3 Ex/Em = 554/568, Cy5 Ex/Em = 649/666). Fluorescence polarization (FP) was calculated from the corrected parallel (F‖) and perpendicular (F_⊥) fluorescence values using equation 1:

$$\text{FP}=(F_{\left|\right|}-F_{\perp })/(F_{\left|\right|}+F_{\perp })\times \text{\hspace{0.17em}1,}000$$ (1)

The G factor, often used in polarization calculations, was not included in our calculations as we determined it to be negligible under these conditions (data not shown).

The plots were fit by nonlinear regression to the following sigmoidal four-parameter equation using Prism 9 (62) (GraphPad Software) as shown in equation 2:

$$\text{mP}=\text{min}+(\text{max}-\text{min})/\{1+{10}^{[({\text{log\hspace{0.17em}EC}}_{\text{50}}-X)\times \text{Hill\hspace{0.17em}slope]}}\}$$ (2)

In equation 2, max and min are the maximum and minimum plateaus of the mP and X is the log of sample concentration. When unconstrained, the values of max, min, EC₅₀ (50% effective concentration), and Hill slope are fit by the regression plot. For all plots except the I189A mutant, the mP max was constrained to the observed mP max for WT MalE-VirF (e.g., 95). The mP plot for the I189A mutant is shifted ~10 mP units higher than the other plots; therefore, the mP max was fit. The span (mP range) for the I189A mutant matched those of the WT and other mutants.

Solubility via turbidity assay.

Y239A MalE-VirF was tested in duplicate at various concentrations (the same as the FP concentrations present in Table S6) using the DNA probe buffer (including BSA and salmon sperm DNA with and without pvirB present) to determine at which concentrations the protein aggregates in solution. Reactions were dispersed into a UV-transparent 96-well plate and allowed to incubate at 37°C for 2 h. OD₆₂₀ was measured to determine the point of protein aggregation, which was where the tested concentration showed a significant increase in absorbance compared to a control lacking Y239A MalE-VirF.

Circular dichroism.

Buffer exchange was performed on WT and selected MalE-VirF and MarA mutant proteins to place the samples in 10 mM NaH₂PO₄ (pH 7.5) using Amicon Ultra-0.5 centrifugal filter units (3-kDa MWCO; MilliporeSigma). The samples were centrifuged (13,000 × g, 4°C, 10 min), flowthrough was discarded, and then samples were diluted again with 10 mM NaH₂PO₄ (pH 7.5). This process was repeated for 5 to 10 rounds. The concentration of each sample was determined via Bradford assay (Bio-Rad). The resultant samples were flash frozen in liquid nitrogen and stored at −80°C prior to circular dichroism (CD) testing. For CD testing, samples were loaded into a 1-mm-path-length quartz cuvette, and CD spectra were collected from 195 to 250 nm using a JASCO J-810 spectrometer. Spectra of a sample containing only buffer was used to correct the raw data for each protein. JASCO Spectra Manager was used to visualize the spectra and export the raw data to be plotted using Prism 9 software (63).