2.09 Å resolution structure of E. coli HigBA toxin-antitoxin complex reveals an ordered DNA-binding domain and intrinsic dynamics in antitoxin

The toxin-antitoxin (TA) systems are small operon systems that are involved in important physiological processes in bacteria such as stress response and persister cell formation. E. coli HigBA complex belongs to the type II TA systems and consists of a protein toxin called HigB and a protein antitoxin called HigA. The toxin HigB is a ribosome-dependent endoribonuclease that cleaves the translating mRNAs at the ribosome A site. The antitoxin HigA directly binds the toxin HigB, rendering the HigBA complex catalytically inactive. The existing biochemical and structural studies had revealed that the HigBA complex forms a heterotetrameric assembly via dimerization of HigA antitoxin. Here, we report a high-resolution crystal structure of E. coli HigBA complex that revealed a well-ordered DNA binding domain in HigA antitoxin. Using SEC-MALS and ITC methods, we have determined the stoichiometry of complex formation between HigBA and a 33 bp DNA and report that HigBA complex as well as HigA homodimer bind to the palindromic DNA sequence with nano molar affinity. Using E. coli growth assays, we have probed the roles of key, putative active site residues in HigB. Spectroscopic methods (CD and NMR) and MD simulations study revealed intrinsic dynamic in antitoxin in HigBA complex, which may explain the large conformational changes in HigA homodimer in free and HigBA complexes observed previously. We also report a truncated, heterodimeric form of HigBA complex that revealed possible cleavage sites in HigBA complex, which can have implications for its cellular functions. a complete HigBA complex structure from E. coli K-12 strain. We also report a crystal structure of a truncated, heterodimeric HigBA complex that suggests the possible proteolytic cleavage sites in toxin HigB and antitoxin HigA, which may have implication in HigBA complex regulation in bacteria. We have probed and validated the key putative active site residues using bacterial growth assays. Using gel-mobility shift assays, SEC-MALS, and ITC methods we have probed the interaction of HigBA complex with a 33 bp palindromic DNA sequence from its promoter region. Based on these results, we have generated a model of HigBA – DNA complex that explains the plausible mechanism of transcriptional regulation of HigBA operon expression by HigA. Furthermore, using MD simulations and spectroscopic (CD and NMR) studies, we revealed the intrinsic dynamic and conformational flexibility in HigA homodimer.


Introduction
Toxin-Antitoxin (TA) systems are pairs of genes that are either located on bacterial plasmids or integrated in the bacterial genome. One of the gene of this pair codes for a toxin, which causes growth arrest by interfering with vital cellular processes and the other gene codes for a cognate antitoxin, which neutralizes the toxin's activity during normal growth conditions [1][2][3]. When the bacterial cell encounters stressful conditions, often the antitoxin is selectively degraded, allowing toxin to exhibit its activity leading to growth arrest and/or dormancy [4][5][6]. The major biological functions of TA systems are: postsegregational killing [7], abortive infection [8], and persister cell formation [9]. The TA systems are classified into Type I to Type VI on the basis of how the antitoxin inactivates the toxin and the nature of antitoxin molecule [10,11]. The toxins in all types of TA systems are protein molecules, however the antitoxin can be a protein or an RNA molecule [10].
Type II systems are the most prevalent and are well-studied TA systems. A typical Type II TA system consists of a toxin and an antitoxin, where both toxin and antitoxin are proteins [2,12]. A type II antitoxin mainly has two functions, first it binds to the toxin and inhibit its activity and; second it acts as autorepressor by binding the promoter region thereby regulating the expression of the TA operon [13,14]. The type II toxins inhibit bacterial growth by targeting essential cellular process like DNA replication [12,15], translation [16,17], etc. Majority of the type II toxins such as RelE, HigB, MqsR and MazF are endoribonucleases. The toxins can be ribosome dependent (examples RelE, HigB, YoeB) [16,18,19] or ribosome independent (examples MqsR, MazF) endoribonucleases [20,21].
The ribosome dependent endoribonucleases share a common microbial RNase T1 like fold in their structures [22] and cleave the translating mRNAs at the A-site of the ribosomes [16,18,19,22]. Some toxins display codon specific cleavage activity, for example RelE cleaves mRNA at UAG and CAG codons [16] and YafQ cleaves at AAA codon [23], while others have a broad range of specificity where they can cleave different mRNAs.
Particularly, HigBA (host inhibition of growth) family of toxin provides a good example for such a case as the HigB toxin has been shown to cleave A-rich codons [19].
The HigBA is a unique ribosome dependent type II TA system. In HigBA operon, unlike most of the type II TA systems, the HigB toxin gene precedes the antitoxin HigA gene [24].
The E. coli K12 toxin HigB is homologous to RelE family of toxins [25]. Previous structural studies on E. coli K12 HigBA system had revealed that HigBA complex is a heterotetrameric structure formed by two interacting heterodimer of HigA-HigB, resembling an overall V shape configuration [26]. The toxin has an RNase T1 like fold consisting of three N-terminal α helices (α1-α3) followed by an antiparallel β-sheet (consisting of β1-β3 Downloaded from http://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200363/894230/bcj-2020-0363.pdf by guest on 06 October 2020 strands) and a C-terminal α helix (α4). On the other hand, the antitoxin HigA is a completely helical protein. HigA structure can be segmented into three parts: the Nterminal dimerization domain (helices α1 and α2), the central part (helices α3-α6) that interacts with the toxin HigB and the C-terminal DNA binding domain (DBD) (helices α7-α9). Interestingly, the binding of the antitoxin HigA to toxin HigB does not occupy or block the active site of HigB toxin, supporting the hypothesis that HigA antitoxin inhibits toxin activity by sterically hindering its binding to ribosome [19,26].
However, there are outstanding questions that remain to be answered. For example, the role of putative active site residues in toxin HigB's function is not systematically probed.
The mechanism of release of HigB toxin from HigBA tetramer is not well studied. The structure of DNA binding domain of HigA in HigBA complex is not well defined and the structural basis of DNA binding by HigA antitoxin remained unknown.
The existing structure of E.coli K-12 HigBA lacked the electron density for key residues in both toxin HigB and antitoxin HigA, due to the low resolution (2.7 Å) of the structure [26].
Here we report a 2.09 Å crystal structure of HigBA complex from E. coli K-12 revealing the electron density for the missing residues in both toxin HigB and antitoxin HigA. Especially, in the case of HigA we observed structured DNA binding domain (DBD), consisting of four helices (α6 to α9) with a compact hydrophobic interior. Using gel-mobility shift assays, SEC-MALS, and ITC methods we have probed the interaction of HigBA complex with DNA sequence from its promoter region. Based on these results, we have generated a model of HigBA -DNA complex that explains the plausible modes of promoter DNA binding by HigA antitoxin, which has implications in understanding the transcriptional regulation of HigBA operon. Using NMR and CD spectroscopy methods in conjunction with the all atom MD simulations, we observed that the antitoxin HigA is intrinsically dynamic in solution. This may help to understand the large conformational changes that HigA has been reported to undergo from free form to bound (HigBA) forms [27,28]. Furthermore, we report a crystal structure of the truncated, heterodimeric HigBA complex that suggests the possible proteolytic cleavage sites in toxin HigB and antitoxin HigA. We have also shown that mutation of key, putative catalytic site residues in toxin HigB results in restoration of bacterial growth. Overall, these results further our understanding of HigBA TA system's assembly, activation, and regulation.

Experimental procedures Protein expression and purification
The HigA and HigB proteins were co-expressed in E. coli BL21 Rosetta (DE3) cells transformed with pETDuet-1 plasmid in which the toxin HigB was cloned in first multiple Downloaded from http://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200363/894230/bcj-2020-0363.pdf by guest on 06 October 2020 Biochemical Journal. This is an Accepted Manuscript. You are encouraged to use the Version of Record that, when published, will replace this version. The most up-to-date-version is available at https://doi.org/10.1042/BCJ20200363 cloning site (MCS) with 6X Histidine purification tag at N-terminal and the antitoxin HigA was cloned in the second MCS with no purification tag. Full-length HigA was also cloned in pET28a vector for the expression of the antitoxin alone with an N-terminal 6X-Histidine tag.
For the protein expression, bacterial cells were grown at 37 °C till the OD 600 nm reached 0.7, after which cells were induced for protein expression with 1 mM IPTG at 20 °C for 16 h.
The cells were harvested by centrifugation at 6000 rpm for 15 min. The cell pellet was resuspended in lysis buffer (20 mM Tris, 300 mM NaCl, pH 7.5) and lysed by sonication. Cell lysates were centrifuged for 1 h at 13000 rpm and supernatant was loaded on to a 5 ml Ni 2+ -NTA Sepharose affinity column (GE Healthcare) at 4 °C. The proteins were eluted using elution buffer (20 mM Tris, 300 mM NaCl, 250 mM imidazole, pH 7.5). The elution fractions were analyzed for protein of interest using SDS-PAGE. Fractions containing the proteins were pooled and concentrated and further purified by Ion exchange chromatography followed by purification and buffer exchange (20 mM  HigBA tetramer and HigBA heterodimer, respectively. The diffraction data sets were processed by iMosflm and XDSAPP software [29][30][31]. The structures were solved by Molecular Replacement method using the reported 2.7 Å resolution structure of HigBA complex (PDB 5IFG) as the search model. Coot [32] and Phenix [33] were used for iterative model building and refinement. while the protein concentration was varied, were assembled in lysis buffer (20 mM Tris, 300 mM NaCl, pH 7.5) and incubated for 2 h in ice. The samples were electrophoresed on a native 8% polyacrylamide gel with 0.5XTBE (Tris-borate with EDTA) visualized by toluidine blue DNA staining.

SEC-MALS
SEC-MALS was performed to determine the molecular weight of HigBA and DNA bound HigBA complex in solution. Shimadzu chromatography system equipped with a miniDAWN TREOS MALS detector and a WATERS 2414 refractive index (RI) detector was used for the experiment. The system was calibrated using bovine serum albumin (BSA). 100 μL of purified, centrifuged and degassed samples were used for the experiment. Biorad Enrich S70 (10/300) column was used for protein complex whereas GE S200 (10/300) column was used for the protein-DNA complex. The molecular weight was calculated using ASTRA VI software (Wyatt Technology).

Isothermal Titration Calorimetry (ITC)
ITC experiments were performed using a VP-ITC machine (MicroCal, USA) at 30 °C. The The heat of dilution of ligands in buffer were subtracted from the integrated heat data and the data was fit for one site-binding model. The analysis was done using ORIGIN-5 software provided by the vendor. All the parameters were kept floating during the data fitting.

CD spectroscopy
CD spectra of HigA were recorded on a JASCO J-715 spectropolarimeter. The experiments were performed in 1 cm path length cuvette (Hellma Analytics) at 25°C at a scanning speed of 100 nm/min with a response time of 4 s. The wavelength scan was done from 190 nm to 260 nm. 7 μM of the HigA protein sample was used for the experiment. The average of three spectral scans was taken for each sample followed by baseline correction to negate the contribution from the buffer. For the thermal melting experiment, the changes in the secondary structure of HigA were observed by recording the CD signal at 208 nm as a Downloaded from http://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200363/894230/bcj-2020-0363.pdf by guest on 06 October 2020 function of temperature. The temperature of the sample was changed at a rate of ~1 °C/min from 15 °C to 90 °C. The melting temperature (Tm) was obtained by fitting the experimental data points (CD signal versus temperature) with a sigmoidal function for two sate transition.

E. coli growth assays
For growth inhibition assay, wild type copy of HigB was PCR amplified using pET-Duet HigBA as template DNA. The PCR product was verified by sequencing and cloned using Nde I/Hind III restriction enzymes in pET28b vector. Mutants HigB constructs (HigB-Δα4, HigB-D49A/N50A, and HigB-R68A) were generated using site directed mutagenesis of wild-type HigB construct in pET28b expression vector. These plasmid constructs were transformed in BL-21 (λDE3, pLysS) and transformants were selected on LB agar plates supplemented with kanamycin and chloramphenicol. For protein expression, cultures were grown till OD600 nm of ~0.8 and induced with the addition of 1mM IPTG. For spotting assay, at time zero and 5 h post-induction, 10-fold serial dilutions were prepared and spotted on LB agar plates.

Molecular Dynamic Simulation
The crystal structure of HigBA complex (PDB 6KML) was used for all atoms molecular dynamics (MD) simulation using GROMACS (GROningen MAchine for Chemical Simulations) 5.1 package [35]. PDB structure of HigBA complex was processed to GROMACS file format using PDB2gmx module and GROMOS96 force field [36] was used during this process.
Cubic box was generated using editconf module where HigBA complex was placed at the center of the box and distance of all edges of the box were 1.0 nm from the complex. This cubic simulation box of HigBA complex was solvated with SPC/E water model. Further, this solvated system of charged HigBA complex was neutralized by adding chloride and sodium ions as counter ions. This solvated electro neutral system was minimized energetically using steepest descent minimization algorithm followed by conjugate gradient algorithm. Both algorithms included maximum 50,000 minimization steps and maximum force below 1000 kJ mol −1 nm −1 . Energetically minimized system was equilibrated for 100 ps duration at temperature 300 K first with NVT ensemble (constant Number of particles, Volume, and Temperature) and subsequently with NPT ensemble (constant number of particles, pressure, and temperature). In both NVT and NPT ensembles, Berendsen thermostat was included in temperature coupling with time constant at 0.1 ps and constraint_algorithm LINCS [37] was included in bond parameter. In NPT, additional Parrinello-Rahman barostat was included in pressure coupling at 1 bar. This equilibrated system contained all requirements to perform MD simulation for 100 ns. MD simulation generated large trajectories were further analyzed for the root means square deviation (RMSD) generated using the g_rms module and RMSD graphs were plotted using GRACE tool (https://plasma-gate.weizmann.ac.il/Grace/). HigA crystal structure (PDB 6QJ4) was obtained from RCSB Protein Data Bank [38]. Using symmetry of HigA crystal structure HigA homodimer structure was generated and subjected to MD simulation with similar protocol mentioned for HigBA complex.

Principal component analysis and preparation of free energy landscape
In order to examine significant combined motions in the MD simulation trajectory of HigBA complex, covariance matrix dependent principal component analysis (PCA) was performed [39,40]. Atomic fluctuations in MD simulation trajectory of HigBA complex were used to generate covariance matrix and subsequently it was diagonalized. The 'g_covar' and 'g_anaeig' utilities were used for construction of covariance matrix and analysis respectively. Initial ten projection eigen vectors from MD simulation trajectory of HigBA complex were generated followed by cosine content dependent analysis. 'g_analyze' utility included in GROMACS was used for the calculation of cosine contents of all generated projection eigenvectors. Here, cosine content values of the two projection eigenvectors from diagonalized covariance matrix was calculated to be below 0.2. Usually, low value cosine contents containing projection eigenvectors, which corresponds to principal component 1 (PC1) and principal component 2 (PC2) contain essential elements to generate the free energy landscape (FEL) [41]. FEL maps (2D and 3D) were generated using Mathematica 9 (Wolfram Research, Inc., Mathematica, Version 9.0) showed free energy based different conformational ensembles of proteins, including the possible native conformation (local minima). Local minima structures of the FEL basin were generated in PDB file format and further analyzed using visualization tool UCSF Chimera [42].

Model building of HigBA -DNA complex
To generate HigBA-DNA model, we used the crystal structure of HigBA complex determined in this study (PDB 6KML). 33 bp Pal-1 DNA sequence (5' ATTCATCCGTTGCCAATCTGGCAACGGATGTTA 3') was used to generate double stranded DNA (dsDNA) using 3D-DART online server [43]. We used two HigBA complexes and one dsDNA to build HigBA-DNA model using UCSF Chimera visualization software [42] Furthermore this HigBA-DNA model was subjected to energy minimization using AMBER99SB force field [44] in GROMACS 5.1 package [45]. showed that the complex eluted as a tetramer of HigBA complex with combined molecular weight of ~ 54 kDa. A 2.7 Å resolution structure of HigBA complex had previously been determined [26]. This structure had revealed the overall architecture of E. coli HigBA complex and the molecular basis of antitoxin HigA binding to toxin HigB. However, the electron densities for several key residues/regions were missing in the structure likely due to the lower resolution. For example, residues 101 to 104 were missing in toxin HigB structure.
Similarly, in antitoxin HigA regions containing residues 95 to 104 and 131 to 138 in the Cterminus that constitute the DBD, were missing [26]. Very recently, a 3.1 Å resolution structure of homologous Shigella flexneri HigBA complex was reported (PDB 5YCL) [28].
Though the structure is similar to the E. coli HigBA complex, in this structure too, the DBD of HigA had missing residues/regions [28].
Here we report crystallization of E. coli HigBA complex under different set of crystallization conditions. Under new conditions, the HigBA complex crystallized in P2 1 space group and diffracted to a higher resolution of 2.09 Å at a synchrotron X-ray source ( Table 1). The structure of the complex was solved by the molecular replacement method using PDB 5IFG [26] as the search model. The structure was refined to a final R work of 19.5% and an R free of 23.6%. Table 1   The antitoxin DBD consists of three regular alpha helices α6 to α8, which are connected via loops (L1 and L2) followed by a short left-handed π helix at the C-terminus (named α9) ( Figure 2A). Helix α6 spans residues Gly79 to Tyr91, helix α7 spans residues Lys104 to Leu111, α8 spans residues Leu119 to Phe129, and short helix α9 spans residues Pro133 to Phe136. Two loops L1 and L2 connect helices α6 to α7 and α7 to α8 respectively. The DBD is a well-folded structure with a compact hydrophobic core. Sidechains of the hydrophobic residues from the helices α6 to α9 form the compact hydrophobic interior of the DBD. These include residues Ile81, Ile84, Leu87, Met88 from α6; residues Val107, Val110, and Leu111 from α7; residues Leu125, and Phe129 from α8; and residue Phe136 from α9. Both loops L1 and L2 are well ordered in the structure. Side chains of Leu98 and Ile101 from loop L1 and Leu117 from loop L2 face the interior contributing towards the hydrophobic core of the DBD. The Tyr91 from helix α6 and Phe129 from helixα8 have a stacking interaction, further cementing the hydrophobic interior ( Figure 2A).
The HigA DBD interacts with the toxin HigB using the amphipathic helix α6. Side chains of polar residues Arg85, Asp89, and Gln90 from helix α6 are involved in hydrogen bonds and ionic interactions with the polar residues from the helices α2 and α3 of HigB toxin. Overall, the HigBA DBD is structurally homologues to the classical DBD present in phage 434 repressor and other proteins ( Figure 2B).

HigBA heterotetramer and HigA homodimer interact with its promoter DNA with high affinity
The antitoxin HigA alone or as HigBA complex, has been shown to bind to the DNA sequences in the HigBA gene promoter region [26]. Two imperfect palindromic sequences had been identified in the promoter region namely, Pal-1 and Pal-2 (Supplementary Figure   S2A) [26]. Using gel-mobility shift assays, the HigBA complex was shown to bind to both Pal-1 and Pal-2 sequences [26], however a quantitative estimation of protein-DNA complex formation was not determined. The DNA binding was attributed to the C-terminal DBD as the deletion of the DBD was shown to abolish the binding of HigBA to the DNA [26].
We initially tested binding of HigBA complex to both 33 bp Pal-1 and 33 bp Pal-2 sequences for DNA binding using gel-shift assays. Similar to the previous observations, we found that HigBA complex binds both Pal-1 and Pal-2 sequences with Pal-1 sequence giving a clean one-band shift at 1:2 DNA to protein ratio suggesting that two HigBA complexes bind to Pal-1 DNA (Supplementary Figure S2B Figure 3B and Table 2). Under same condition, interestingly binding of HigBA heteroteramer to Pal-1 DNA showed different ITC isotherm. We observed that binding of HigBA with Pal-1 DNA proceeds with a positive heat change that reaches maximum at the third injection (please note that the first inject was incomplete and not used during data integration). The positive heat change then gradually decreases at subsequent injections. This hints at two sequential binding events ( Figure 3C) [46]. However, we achieved best fitting of integrated data for a one site binding model. Analyzed data showed that HigBA tetramer binds to Pal-1 DNA in an entropically driven manner (ΔH = 3.96±0.08 kcal/mol and TΔS = 13.73±3.97 kcal/mol) with a K D of ~91 nM and a protein to DNA stoichiometry of ~2, which is in agreement with results obtained from EMSA and SEC-MALS experiments ( Figure 3C and Table 2 repressor DBDs ( Figure 2B). Based on these collective observations, we present a model where two protomers of HigA in HigBA tetramer bind to 33 bp Pal-1 DNA with helix α7 binding the major groove of DNA to generate specific HigA -DNA complex ( Figure 3D).
Considering the symmetry of DNA sequences (palindromic nature), the helix α7 is inserted in the CG bp centered major groove of the Pal-1 DNA sequence and helix α8 is closer to this

HigB
While screening the HigBA complex for crystallization under different conditions, HigBA crystallized in a dimeric form in one of the conditions. The structure of the hetero dimeric HigBA was solved using molecular replacement method and refined to a resolution of 2.3 Å ( Figure 4A and Table 1 HigBA heterotetrameric complex, which physically cannot access the A site in the ribosome and thereby it is rendered inactive [26]. Therefore, we postulate that the removal of helix α4 will result in reduced affinity of truncated HigB (HigB-α4) for the ribosome binding and have impaired RNase activity. In bacteria, TA complexes are activated during certain stress conditions such as exposure to antibiotic [52]. Generally, the intrinsically disordered antitoxin Taken together, the truncated heterodimeric HigBA structure suggests the labile regions in antitoxin HigA and toxin HigB that may have roles in activation or deactivation of HigB activity inside bacteria.

Overexpression of HigB toxin induces bacteriostasis in E. coli
As explained earlier, E. coli HigB toxin is structurally similar to the RelE family of toxins.
However, the conservation of these sequences is poor, even at the catalytic centre, which makes it difficult to identify the catalytic residues based on sequence alignment. We have were generated using site-directed mutagenesis. To ascertain the role of C-terminal helix α4 in HigB, which was found missing in truncated HigBA dimer structure, we made a deletion mutant of HigB where helix α4 (residues 91-101) was deleted (HigB-Δα4). One of the putative catalytic site residues, H88 is close to helix α4, while residue Y91 is a part of the helix α4 in toxin HigB ( Figure 5A).
We used bacterial growth assay to access the role of these mutant and WT HigB in E. coli.
For the growth assays, pET-28b plasmids harboring either a wild type or mutants of HigB  Figure 5B). In concordance with liquid culture growth inhibition assays, we observed that in comparison to parental strain, overexpression of wild type HigB protein resulted in >10,000-fold reduction in growth on solid medium ( Figure 5C).
We also observed that the growth patterns obtained in strains overexpressing mutant proteins after 5 h of induction was comparable to strain harboring vector only ( Figure 5C). As expected, no differences in growth patterns were observed in strains harboring parental, wild type or mutant proteins before IPTG induction ( Figure 5D). Therefore, the growth assays indicate that residues D49, N50 and R68 are crucial for catalytic activity. Also, the deletion of the C-terminal helix α4 affects protein's catalytic activity, showing the importance of α4 for the catalytic activity of HigB.

Molecular dynamics simulations study reveals dynamic nature of HigA antitoxin
We performed an all atom molecular dynamics (MD) simulation of HigBA tetramer for a duration of 100 ns using GROMACS software package [35]  found well stable throughout the simulation period ( Figure 6A). Therefore, these results suggest that the dynamics observed in HigBA complex is mainly due to the antitoxin HigA homodimer in the complex.
To explore the different conformations of HigBA tetrameric complex within the MD simulation trajectory, principal component analysis (PCA) based free energy landscape (FEL) was analyzed [39,40]. The PCA process included the construction of covariance matrix and its diagonalization followed by projection on the eigenvector, which led to the selection of low value cosine content containing principal components (PCs). These PCs contained centered data, which provides collective correlated motion of atoms in HigBA complex simulated trajectory that hold all necessary components of the system. Among the ten generated eigenvectors, projections on two eigenvectors were observed with relevant correlation and cosine content of these two eigenvector projections (PC1 and PC2) were analyzed to a reasonable value of below 0.2. These two PCs comprised sufficient trajectory information of HigBA complex to perform FEL to generate final two-dimensional (2D) and three-dimensional (3D) FEL maps ( Figure 6B,C). The 3D FEL contour map clearly illustrated a broad basin, which suggested multiple minimum energy HigBA complex ensembles ( Figure 6C). This likely suggested the structural transitions in HigBA complex among the discrete conformational states with low energy barrier during simulation period of 100 ns. Three representative minimum energy structures of HigBA complex from basin of FEL contour map were extracted for further analysis (represented by red, green, and blue triangles in Figure 6B). These three FEL representative structures were extracted from 100 ns MD simulation trajectory at 35 ns (in red), 71 ns (in green) and 90 ns (in blue). From each FEL representative structure, HigA homodimer and HigB were separated, overlaid and aligned with respect to the starting structure (in steel) ( Figure 6D,E). In case of toxin HigB, the three FEL representative structures retained conformations similar to the starting structure, showing that HigB toxin maintains a stable structure during the course of simulation ( Figure 6D). However, in case of antitoxin HigA, major differences in the conformations of the three FEL representative structures, with respect to starting structure, were observed ( Figure 6E).

Antitoxin HigA forms an ordered structure but it is dynamic in solution
Generally, the antitoxins across different type II TA systems contain intrinsically disordered regions. The toxin binding domains in several antitoxins are intrinsically disordered, which become ordered upon binding to the toxin. It has been hypothesized that under stress conditions, several proteases (e.g. Lon proteases) are expressed in bacterial cell that proteolytically cleave susceptible antitoxins, thereby releasing and activating the toxin [54].
Interestingly, in a recent study, TA transcript levels were shown to increase substantially in response to diverse stress conditions without liberating the toxin from the TA complex [55]. In case of HigBA system, the antitoxin has a well folded, all helical structure [38]. However, the MD simulation results presented in current study suggested intrinsic dynamic as well as large conformational sampling of HigA protomers in HigBA complex. These contrasting The full-length antitoxin HigA was over-expressed and purified. The antitoxin HigA protein expectedly purified as a homodimer. The CD spectra of the HigA homodimer gave a spectra characteristic of a helical protein ( Figure 7A). Thermal melting of HigA homodimer was followed using CD spectroscopy. This revealed that protein has an average melting temperature of ~ 57 °C, suggesting a stable structure ( Figure 7B). We made a uniformly 15 N labeled HigA homodimeric protein and recorded a 2D 15 N-1 H TROSY HSQC NMR spectra of the protein. The spectra revealed good chemical shift dispersion suggesting that protein is folded in solution. However broad and non-uniform intensities of cross peaks in the spectra strongly suggested that HigA structure likely has enhanced dynamics and undergoes a conformational exchange ( Figure 7C). The CD spectra, although influenced by the tertiary structure, mainly reports the secondary structural elements in the protein, whereas NMR chemical shifts are very sensitive to the tertiary structure of the protein.
Taken together, these results augment the observations of MD simulation suggesting the dynamic nature of antitoxin. As mentioned earlier, a crystal structure of full-length HigA homodimer was recently reported [38]. However, it is well established that several proteins that do not give a well-dispersed 2D 15 N-1 H HSQC spectra could still crystallize readily.
Conversely, several proteins with well-dispersed 2D 15 N-1 H HSQC spectra still cannot be crystallized [56,57]. Additives in crystallization buffer can aid in the process of crystallization.
We therefore argue that although the HigA homodimer forms an ordered structure (as observed in crystal structure), it maintains the intrinsic flexibility, which is likely the reason behind the poor NMR spectral quality of the protein. This is in agreement with the MD simulation based FEL analysis ( Figure 6B, C, and E) and

Comparison of different
experimental observations on solution behaviour of HigA probed using CD and NMR spectroscopy methods (Figure 7). This dynamic behaviour of HigA homodimer will likely have a role in release of toxin HigB from HigBA complex and promoter DNA binding in bacteria.

Concluding Remarks
Structures of several type II toxin-antitoxin complexes of different classes have been determined in a number of studies [58]. These structures helped in understanding the assembly, activation, and regulations of these important cellular machines, which have implications in understanding the important processes such as persister cell formation and antibiotic tolerance by bacteria [59]. HigBA is a unique type II TA system in which the toxin HigB is a ribosome dependent RNase that cleave translating mRNAs. However, the mechanism of activation of E. coli HigBA as well as the autoregulation of HigBA operon is not well understood.
A previously determined structure of E. coli HigBA TA complex had elegantly revealed the overall organization and mechanism of antitoxin HigA binding to toxin HigB [26]. This structure lacked the electron density of key residues in both toxin HigB and antitoxin HigA.
The missing region in antitoxin corresponds to the important DBD of HigA. Here we report a 2.09 Å crystal structure of HigBA complex from E. coli K-12 that revealed the electron density for the missing residues in both toxin HigB and antitoxin HigA. Especially, in antitoxin HigA we observed structured DBD, consisting of four helices (α6 to α9) with a compact hydrophobic interior, revealing a complete HigBA complex structure from E. coli K-12 strain. We also report a crystal structure of a truncated, heterodimeric HigBA complex that suggests the possible proteolytic cleavage sites in toxin HigB and antitoxin HigA, which may have implication in HigBA complex regulation in bacteria. We have probed and validated the key putative active site residues using bacterial growth assays. Using gelmobility shift assays, SEC-MALS, and ITC methods we have probed the interaction of HigBA complex with a 33 bp palindromic DNA sequence from its promoter region. Based on these results, we have generated a model of HigBA -DNA complex that explains the plausible mechanism of transcriptional regulation of HigBA operon expression by HigA.
Furthermore, using MD simulations and spectroscopic (CD and NMR) studies, we revealed the intrinsic dynamic and conformational flexibility in HigA homodimer.

Abbreviations
HigBA, host inhibition of growth; TA, toxin-antitoxin; NMR, nuclear magnetic resonance; CD, circular dichroism; MD, molecular dynamics; GROMACS, GROningen MAchine for Chemical Simulations.        transformed with empty vector and vector encoding HigB, HigB-Δα4, HigB-D49A/N50A, and HigB-R68A. The expression was induced by addition of 1 mM IPTG and bacterial growth was monitored by measuring optical density of culture at 600 nm wavelength at indicated time points. C, D) Spotting assays was performed to monitor bacterial growth on solid LB agar plates. Monitoring growth of vector only and mutant HigB containing bacterial cultures, 10fold serial dilutions of induced (5 h post induction) (C) and uninduced (at zero time) (D) cultures were prepared and spotted on LB Agar plates. The numbers represent the dilution of the spot with 1 representing lowest dilution and 6 representing highest dilution. The plates were incubated at 37 o C overnight and images were recorded. Growth of a bacterial culture that contains empty vector is monitored in both the experiments as a control. The data shown in these panels is representative of three independent experiments.    Tables   Table 1. Crystallographic data collection and structure refinement statistics