Trp RNA-Binding Attenuation Protein: Modifying Symmetry and Stability of a Circular Oligomer

Background Subunit number is amongst the most important structural parameters that determine size, symmetry and geometry of a circular protein oligomer. The L-tryptophan biosynthesis regulator, TRAP, present in several Bacilli, is a good model system for investigating determinants of the oligomeric state. A short segment of C-terminal residues defines whether TRAP forms an 11-mer or 12-mer assembly. To understand which oligomeric state is more stable, we examine the stability of several wild type and mutant TRAP proteins. Methodology/Principal Findings Among the wild type B. stearothermophilus, B. halodurans and B. subtilis TRAP, we find that the former is the most stable whilst the latter is the least. Thermal stability of all TRAP is shown to increase with L-tryptophan concentration. We also find that mutant TRAP molecules that are truncated at the C-terminus - and hence induced to form 12-mers, distinct from their 11-mer wild type counterparts - have increased melting temperatures. We show that the same effect can be achieved by a point mutation S72N at a subunit interface, which leads to exclusion of C-terminal residues from the interface. Our findings are supported by dye-based scanning fluorimetry, CD spectroscopy, and by crystal structure and mass spectrometry analysis of the B. subtilis S72N TRAP. Conclusions/Significance We conclude that the oligomeric state of a circular protein can be changed by introducing a point mutation at a subunit interface. Exclusion (or deletion) of the C-terminus from the subunit interface has a major impact on properties of TRAP oligomers, making them more stable, and we argue that the cause of these changes is the altered oligomeric state. The more stable TRAP oligomers could be used in potential applications of TRAP in bionanotechnology.


Introduction
Cyclic protein oligomers pose promise as the basis for engineering molecular machines, including those that operate by rotation of one or more components about a principal axis. Their suitability, owing to their unique geometry, is evident by their frequent occurrence in key bio-mechanical and regulatory roles. Examples include circular protein oligomers found in flagellar motors [1], the ATP synthase [2] and the molecular motor of tailed bacteriophages which translocate DNA into the capsid during viral particle assembly [3][4][5]. The stoichiometry of the MSring and C-ring in flagellar motors are variable [6]. Likewise, the central component of the bacteriophage DNA-translocating motor can also exist in different oligomeric states [5]. This demonstrates a general problem associated with instability of circular oligomers and variation in their oligomeric states when they are extracted from their natural environment. Engineering stable molecular devices of suitable size, geometry, and hence subunit number is of interest for applications in molecular biology and medicine. For example, applications of viral DNA-translocating devices in the transfer of genetic information across biological membranes are attractive. Recent studies aimed at the incorporation of a portal protein into lipid bilayers look particularly promising [7].
The tryptophan RNA-binding attenuation protein, TRAP, possesses similar geometric features to other circular assemblies, being assembled of multiple subunits and containing a central tunnel. Typically composed of 11 or 12 identical subunits, with a molecular weight of 90.6 to 101.9 kDa [8,9], TRAP regulates Ltryptophan biosynthesis genes by attenuation in many Bacilli [10]. When activated by increased levels of L-tryptophan -bound to receptor sites nestled between adjacent subunits -TRAP binds single stranded RNA at the leader region of the trpEDCFBA operon transcript. This modulates terminator hairpin formation and so downregulates L-tryptophan biosynthesis [8]. In this study TRAP is employed as a model for investigating the influences of oligomeric state and stability of such assemblies. To date TRAP has already alluded to its value in bionanotechnology, showing that mutations to residues protruding into the central cavity allow gold nanoparticle binding [11]. Nanotubes consisting of stacked TRAP rings linked by disulfide bonds have also been shown to selfassemble, in response to the introduction of cysteine residues on the ring-faces orthogonal to the principal axis [12].
TRAP oligomers have previously been reported to be highly thermostable [13]. Predictably, we find that the melting temperature is highest in the case of TRAP from B. stearothermohphilus, a hyperthermophile. The presence of L-tryptophan -and hence the bound holo-TRAP state -has been associated with a greater structural rigidity and integrity [14]. Prior to the recent determination of the B. stearothermophilus apo-TRAP crystal structure [15] TRAP was crystallized in the presence of Ltryptophan [8,9]. We find that L-tryptophan binds with a K d in the 2.8 mM to 6.9 mM range, with small variation between TRAP from different species. We explore how the presence of Ltryptophan and its' binding affects thermal stability in the context of the three wild type and three mutant TRAP molecules. We also explore the influence of the oligomeric state on thermal stability by comparing wild type 11-subunit and mutant 12-subunit molecules from the same species. Finally, we show how the 11-subunit to 12subunit switch could be induced by the point mutation S72N. The analysis includes crystal structure and native mass spectrometry data on B. subtilis S72N TRAP. We find that the increase in subunit number increases the stability of the assembly.

Crystal Structure and Native Mass Spectrometry of B. subtilis S72N TRAP
We previously demonstrated the importance of the C-terminus in oligomer formation during B. subtilis TRAP assembly [9]. However, this work required deletion of the C-terminus to increase the subunit stoichiometry and we questioned if a more subtle mutation at the C-terminus could induce formation of a 12subunit assembly. Based on the previous structural work, we chose the S72N mutation because we expected that the bulky side chain at this position would result in the exclusion of N72 and following C-terminal residues from the subunit interface. The crystal structure of B. subtilis S72N TRAP, determined at 2.7 Å , reveals a circular 12-mer assembly (Figure 1), which is similar to previously reported crystal structures of TRAP. In the crystal the 12 subunit assembly is generated by combination of six TRAP subunits, present in the asymmetric unit, with the crystallographic 2-fold axis of the P2 1 2 1 2 space group ( Table 1). The final electron density maps allowed positioning of all residues, with the exception of the four N-terminal residues and the five C-terminal residues of most subunits, for which no clear electron density has been observed, indicating their flexibility. The structure of the S72N TRAP was compared against the wild type TRAP (Figure 2A), which highlighted change in the conformation of the C-terminus in the mutant protein, starting from position 69. The weighted F o -F c omit electron density maps for residues E69 and M70 ( Figure 2B) demonstrate the absence of interpretable electron density beyond position 70 in all chains, indicating that five Cterminal residues are disordered. The crystal structure compares well with our predictions on the effect of the point mutation, in inducing the removal of the C-terminus from the subunit interface through steric hindrance, thus facilitating rigid-body rotations of subunits for the accommodation of one additional subunit into the ring.
To understand if the 12-mer state of the S72N TRAP observed in the X-ray structure, predominates in solution, we analyzed this protein by native mass spectrometry (Text S1). The 12-mer state was identified as the dominant species in the mass spectra ( Figure  S1), as opposed to the wild type protein which forms 11-mers [16].

12-mer TRAP Assemblies are More Stable
Melting temperatures for six TRAP molecules, including three wild type and three mutant TRAP listed in Table 2, were determined at varied L-tryptophan concentrations (0-100 mM) using dye-based scanning fluorimetry as described in the methods. Fluorescence versus temperature melting curve data ( Figure S2, S3) was averaged between four observations and the midpoints of fitted sigmoidal curves were taken as the melting temperatures, T m . Melting temperatures (Table S1) were then plotted against Ltryptophan concentration ( Figure 3).
As the fluorescent dye could potentially bind in the tryptophanbinding pocket, we employed CD spectroscopy (Text S2) to verify that melting temperatures were not perturbed by interaction with the dye, Figure S2. We note that there is coherence between melting temperatures derived from the two respective techniques.
Increasing the L-tryptophan concentration increases the thermal stability of all the TRAP proteins except for wild type B. subtilis TRAP. B. halodurans TRAP was included as a reference, as it is naturally a 12-mer, without engineering. Throughout the Ltryptophan concentrations for which the T m was measured, the wild type B. subilis TRAP was least stable, while the B. stearothermophilus was most stable. Significantly, the mutant 12 subunit assemblies were at least as stable, but usually more stable than the 11 subunit wild-type, at all L-tryptophan concentrations measured. This stabilization was particularly marked at high Ltryptophan concentrations as the mutant 12-mer subunit TRAPs have a larger increase in T m with increasing L-tryptophan concentration than their wild type 11-mer counterparts. For example, from 0 to 100 mM L-tryptophan, B. subtilis wild type increases in stability by 0.8uC, while the T m of the K71stop and S72N 12-mer mutants increases by 23.2uC and 13.3uC respectively. The B. stearothermophilus wild type TRAP increases in stability by 7.1uC, whereas the T m of E71stop TRAP increases by 20.2uC. The thermal stability of B. subtilis wild type TRAP appeared to show little dependence on L-tryptophan concentration up to 100 mM. We observed that the stability of the wild type 12-mer B. halodurans TRAP lies midway between that of mutant 12-mers, B. subtilis K71stop and B. stearothermophilus E71stop TRAP ( Figure 3).

L-Tryptophan Binding
The increased stabilization by L-tryptophan of the 12-subunit assemblies, as compared to 11-mers, was unexpected. To investigate this further we measured the affinity of each TRAP for L-tryptophan ( Table 2). A comparison of the dissociation constants of the mutant 12-mers to their 11-mer wild type counterparts showed no obvious correlation. Although an 11-mer to 12-mer transition in B. subtilis TRAP increases the affinity of binding, this is not observed in B. stearothermophilus TRAP. This suggests stabilization by L-tryptophan is not directly related to affinity towards L-tryptophan.

Discussion
The conformation of the C-terminus has previously been shown to play a major role in determining whether the 11-mer or the 12-mer oligomeric state assembles [9], and the new structure of the S72N mutant reinforces this. The relationship between subunit number and thermal stability has for the first time been probed, revealing the 12-mer configuration is more stable than the 11-mer counterpart.

Comparison of Thermal Stability in Wild Type TRAP
We compared the thermal stability of B. stearothermophilus, B. halodurans, and B. subtilis wild type TRAP. The relative thermal stabilities of the oligomers reflect the temperature environments of the respective species' habitats, as might be expected. For example, at 50 mM L-tryptophan, the melting temperatures for the respective wild type TRAP are: B. stearothermophilus, 81.4uC; B. halodurans, 75.6uC; B. subtilis, 44.3uC. These melting temperatures reflect the respective species' hyperthermophilic, thermophilic and mesophilic origins respectively. This relationship appears to hold across all of the L-tryptophan concentration range ( Figure 3).
Within these wild type TRAP oligomers, B. stearothermophilus and B. subtilis both have 11-mer TRAP, and B. halodurans has 12-mer TRAP. It is apparent therefore that the temperature environment in which the species have evolved is the overriding factor in determining the thermal stability of the wild type TRAP, as we have an 11-mer as both the most and least stable. However, we find that one effect of inducing a change of oligomeric state from 11-mer to 12-mer by mutations is increased thermal stability, implying that species-specific structural characteristics unrelated to subunit number are not the only determinant of thermal stability.

C-terminal Truncation Induces 11-to 12-mer Transition and Increases Stability
Comparing the structure of the 12-mer B. halodurans TRAP with that of the 11-mer structures of B. subtilis TRAP and B. stearothermophilus TRAP, the importance of the C-terminus (residues 71 to 75) becomes apparent. In B. halodurans TRAP, the nature of the last four amino acids dictate that the C-terminal segment is excluded from the subunit-subunit interface, thereby generating favorable interactions at the outer surface of the oligomer. This is considered as the crucial factor in permitting the higher oligomeric state, which is related to the 11-mer state by a 2.7 o rigid-body rotation of adjacent subunits [9]. Hence what can be considered macroscopically as the result of the extraction of a thin ''wedge'' between subunits on the outer side of the ring, has the effect of decreasing the angle that a single subunit contributes towards the full 360u.
Taking our investigation a step further, we focused further studies at determining whether there is a general relationship between the oligomeric state of TRAP and its thermal stability, hence exploring factors that are not species-specific. The scenario found in the 12-mer B. halodurans TRAP, regarding the exclusion of the C-terminus, could logically be mimicked by truncating the 4 or 5 C-terminal residues, removing this segment from the subunit interface. This is observed in the previously determined structures of B. stearothermophilus E71stop TRAP and B. subtilis K71stop TRAP [9]. With respect to thermal stability accompanying the 11mer to 12-mer transition, we find that this truncation results in an increase in melting temperature, so that at 50 mM L-tryptophan the increase in melting temperature between B. stearothermophilus wild type and E71stop is 6uC, from 81.4uC to 87.4uC. The equivalent increase between B. subtilis wild type and K71stop is 18.7uC, 44.3uC to 63.0uC. This hints at an increase in thermal stability in response to incorporation of an additional subunit into the ring. However, this difference in melting temperature could also be a result of the absence of the C-terminal residues themselves. For instance, we note that both B. subtilis and B. stearothermophilus wild type TRAP contain two flexible lysine residues at their C-termini [17] that can contribute electrostatically to the overall stability of TRAP. To understand which of these two factors -C-terminus or oligomeric state -has the greater influence we compared thermal stability of wild type B. subtilis TRAP 11mer with that of the B. subtilis S72N TRAP 12-mer, which still possesses the C-terminal segment but has it excluded from the subunit interface for steric reasons.

S72N Mutation also Increases Subunit Number and Stability
Exclusion of the C-terminus from the subunit interface can be achieved by introducing steric strain at a key region towards the Cterminus that causes the segment to pivot out. In this respect, exclusion of the C-terminus in S72N is achieved in a manner more similar to how exclusion is achieved in B. halodurans TRAP. Residue 72 in particular (Ser 72 in B. subtilis TRAP) -which sits at the subunit-subunit interface in both wild type 11-mer structures studied -serves as a critical determinant in the conformation of the C-terminus. This is evident from the crystal structure presented in this study. From the structure of the S72N TRAP we see that substituting serine 72 for a bulkier amino acid, asparagine, does indeed introduce the necessary constraints to exclude the subsequent segment from the subunit interface. The thermal stability of this mutant 12-mer TRAP, S72N, is higher than that of its wild type 11-mer counterpart. For example at 50 mM L-tryptophan the melting temperature of the wild type and mutant oligomers are 44.uC and 54.5uC respectively, yielding a stabilization of 10.2uC. This is significant, albeit it is less compared to the 18.7uC stabilization we see from C-terminal truncation (i.e. B. subtilis K71stop).
We conclude that by inducing an expansion in the oligomeric state from 11 to 12 subunits the thermal stability of the oligomer is increased. This increase is predominantly a result of the oligomeric state transition, and to a much lesser extent the influence of Cterminal or other specific residues. A potential, general relationship in TRAP oligomers therefore exists, in which melting temperature depends on subunit number. Further studies should be focused on evaluation of influence of the oligomeric state of TRAP on its activity in vivo.

Binding Affinity Towards L-tryptophan
To explore whether the stabilization by L-tryptophan was directly related to the affinity of TRAP oligomers to the binding of L-tryptophan, dissociation constants were measured. Results showed no obvious correlation with thermal stability or subunit number ( Table 2). Although between B. subtilis wild type and S72N mutant TRAP, the 11-mer to 12-mer transition does coincide with a slight increase in affinity for L-tryptophan, these changes are only twofold, implying there are other more significant influences. . is average value of intensity, the sum g hkl is over all measured reflections and the sum g i is over i measurements of a reflection. b Crystallographic R = g hkl ||F obs -F calc ||/g hkl |F obs |, R free was calculated using a randomly chosen set of reflections that were excluded from the refinement. doi:10.1371/journal.pone.0044309.t001

Increasing L-tryptophan Concentration Increases Thermal Stability
For most TRAP oligomers studied we observe an increase in melting temperature in response to an increase in L-tryptophan concentration from 0 mM to 100 mM added. The trend does not appear in B. subtilis wild type TRAP however, over the Ltryptophan concentration range studied. Most concentrations in the monitored range represent many times the typical K d value,  hence L-tryptophan binding-sites in TRAP are expected to be saturated. This therefore suggests a degree of non-specific binding of L-tryptophan to TRAP as the principal cause of the stabilization, over a thermodynamic or kinetic influence involving the L-tryptophan binding sites. It is likely that non-specific binding, beyond the 1:1 stoichiometry, would hence increase with increasing concentration of L-tryptophan.

Conclusions
This study demonstrates how the oligomeric state of a circular protein can be changed by introducing a point mutation at a subunit interface. Importantly, we demonstrate that a change in subunit number changes the stability of a circular assembly in vitro.
In the case of TRAP, used in our studies, an increase in subunit number increases the stability for the oligomeric assembly. Predictably, this stability is dependent on solvent conditions, specifically L-tryptophan concentration. We conclude that a similarly subtle approach could be used during protein engineering of other ring-like structures, for stabilizing a particular oligomeric state thereby modifying dimensions, symmetry and stability.

Gene Cloning, Protein Purification, Crystallization and Data Collection
The B. subtilis S72N mutant TRAP was generated using the QuikChange kit (Stratagene, US) and a pET9a plasmid containing the wild type gene. S72N TRAP proteins were produced and purified as described previously [18]. Before crystallization, protein samples were transferred into solution containing 2 mM Tris (pH 8.5), 300 mM NaCl and purified by size-exclusion chromatography using Superdex 200 Column (GE healthcare, UK). TRAP samples were initially analysed by size-exclusion chromatography, employing a Superdex 200 10/300 column (GE healthcare, UK) for the preliminary oligomeric assembly analysis. A 200-300 mg TRAP sample was injected and eluted by a buffer containing 20 mM Tris-HCl, 30 mM NaCl, and 0.1 mM Ltryptophan (pH 8.5). TRAP samples with an elution volume less than 14.5 ml were selected for further analysis by native mass spectrometry analysis ( Figure S1) and X-ray crystallographic studies.
Crystallization was carried out at 18uC by hanging drop vapor diffusion. For crystallization, B. subtilis S72N TRAP was transferred into solution containing 10 mM triethanolamine (pH 8.0), 100 mM NaCl, 15 mM L-tryptophan and concentrated to 33.5 mg/ml. The reservoir contained 100 mM Bis-Tris propane (pH 8.5), 200 mM KSCN and 13% PEG 3350 (v/v). Protein crystals were frozen using solutions containing all the crystallization ingredients with addition of 20% glycerol (v/v). The X-ray data were collected at 120 K at I04 beamline at the Diamond Light Source. Data were processed using HKL2000 [19], ( Table 1).

Structure Determination and Refinement
All crystallographic calculations were carried out using the CCP4 program package [20]. The structure was determined by molecular replacement using MOLREP [21] with three adjacent subunits of B. stearothermophilus TRAP as a search model. Refinement was performed by REFMAC [22] and model rebuilding was done using COOT [23]. Water molecules were added automatically with the program ARP/wARP [24] and further corrected using maximum likelihood-weighted 2|F o | -|F c | and |F o | -|F c | electron density maps. Molecular contacts between adjacent monomers of TRAP were examined by CONTACT [20]. Figure 1 was generated using Bobscript [25], Figure 2A using UCSF Chimera [26], and Figure 2B using CCP4 mg [27].

Dye-based Scanning Fluorimetry
Dye-based scanning fluorimetry, or the Thermofluor method, was performed as previously described [28,29]. Protein samples were prepared by transferring to 50 mM Tris pH 8, 150 mM NaCl, 10 mM MgCl 2 , with variable L-tryptophan concentration, in 30 kDa cut-off Vivaspin concentrators, and diluting to 1 mg/ ml, monitoring concentration by the Comassie-blue Bradford method. L-tryptophan concentration was varied from 0 to 0.1 mM (final well concentration). Samples were transferred to 96-well thin-walled clear PCR plates. Each well contained 30 ml of solution, consisting of 15 ml of the 1 mg/ml protein (yielding a final concentration of 0.5 mg/ml), 1/1000 diluted SYPRO Orange (Sigma-Aldrich) and 50 mM Tris pH 8 with a specified L-tryptophan concentration. A Stratagene Mx3005P QPCR instrument was programmed to increase temperature in 1uC increments at 30 second intervals, from 25uC to 95uC with 470 nm excitation, measuring fluorescence at 570 nm.

Measurement of L-tryptophan Dissociation Constants
L-tryptophan binding to TRAP was measured using a fluorescence assay based on competition between L-tryptophan and 1-anilionaphthalene-8-sulfonic acid (ANS) as described previously [30]. ANS fluoresces weakly in solution but fluoresces strongly at 460 nm with 372 nm excitation upon binding to TRAP. L-tryptophan has a greater affinity than ANS for TRAP, and displaces ANS thus reducing the fluorescence. The change in fluorescence, DF, was monitored using a LS-50B fluorometer (Perkin Elmer). A 100 mL sample of 15 mM TRAP was prepared in 60 mM ANS and 0.5 mM sodium phosphate at pH 8.0. Readings of fluorescence intensity were taken from 0 to 2000 mM tryptophan. At each concentration of tryptophan, fluorescence intensity was collected at room temperature for one minute to allow the reaction to reach equilibrium. Fluorescence intensity was averaged for each concentration of L-tryptophan and then used to calculate the change in fluorescence (DF). The maximal change in fluorescence was set to 100% and values of DF were normalised accordingly. GraphPad Prizm 4 was used to plot % DF as a function of tryptophan concentration to obtain L-tryptophan dissociation constants and binding curves.  Figure S1 Native mass (Nanoflow electrospray) spectrum of B. subtilis S72N TRAP. m/z corresponds to the massto-charge ratio. Red hexagons correspond to 12-mer TRAP species without bound tryptophan.  (11mer). Sigmoidal curves were fitted to CD spectroscopy data (red) for comparison against fluorescence data. The melting temperatures represented by the midpoints of these curves were found to be 78.5uC and 81.6uC for (A) and (B) respectively. These compare well with the melting temperatures derived from curves fitted to the fluorescence data, 75.6uC and 81.4uC respectively (Table S1)