The periplasmic domains of Vibriocholerae ToxR and ToxS are forming a strong heterodimeric complex independent on the redox state of ToxR cysteines

Abstract The transmembrane protein ToxR plays a key role in the virulence expression system of Vibrio cholerae. The activity of ToxR is dependent on its periplasmic sensor domain (ToxRp) and on the inner membrane protein ToxS. Herein, we present the Nuclear Magnetic Resonance NMR solution structure of the sensory ToxRp containing an intramolecular disulfide bond. The presented structural and dynamic experiments with reduced and oxidized ToxRp propose an explanation for the increased proteolytic sensitivity of reduced ToxR. Additionally, for the first time, we could identify the formation of a strong heterodimer complex between the periplasmic domains of ToxR and ToxS in solution. NMR interaction studies reveal that binding of ToxS is not dependent on the redox state of ToxR cysteines, and formed complexes are structurally similar. By monitoring the proteolytic cleavage of ToxRp with NMR, we additionally provide a direct evidence of ToxS protective function. Taken together our results suggest that ToxR activity is regulated by its stability which is, on the one hand, dependent on the redox states of its cysteines, influencing the stability of its fold, and on the other hand, on its interaction with ToxS, which binds independent on the cysteines and acts as a protection against proteases.

. The regulatory cascade, called ToxR-regulon (Matson et al., 2007) leads to the production of the main virulence factors namely the toxin co-regulated pilus (TCP), and the cholera toxin (CT). While TCP is absolutely vital for the colonization in the small intestine , it is the CT that triggers the diarrheal symptoms of the cholera disease (Sánchez and Holmgren, 2008;Taylor et al., 1987). Under such virulence inducing conditions, the conserved toxRS operon is constitutively expressed (Kanjilal et al., 2010), thereby regulating numerous genes (Bina et al., 2003;Champion et al., 1997;Lee et al., 2000;Skorupski and Taylor, 1997;Wang et al., 2002;Welch and Bartlett, 1998). ToxRS co-activates, together with their homolog regulators TcpPH, the transcription of the main virulence regulator ToxT (Bina et al., 2018;Childers and Klose, 2007;Häse and Mekalanos, 1998;Higgins and DiRita, 1994;Krukonis et al., 2000). In contrast to TcpPH, which is encoded on the Vibrio Pathogenicity Island (VPI) (Jermyn and Boyd, 2002), ToxRS is also present in nonpathogenic isolates of V. cholerae, suggesting their involvement also in non-virulent associated processes such as adaption to environmental stress conditions (Almagro-Moreno et al., 2015a).
ToxR is built up of three domains, encompassing the periplasmic, membrane, and cytoplasmic compartments (Crawford et al., 2003;Miller et al., 1987;Osorio and Klose, 2000). Its activation is triggered by the sensing of environmental substances, like certain amino acids (Mey et al., 2012) or bile acids (Midgett et al., 2017), through its periplasmic domain. Bile acids are present in the small upper intestine tract and bind directly to the ToxR periplasmic sensory domain (Midgett et al., 2017;Midgett et al., 2020) thereby inducing a switch in the outer membrane porin expression from OmpT to OmpU (Miller and Mekalanos, 1988;Provenzano and Klose, 2000). The production of OmpU enables the bacterium to survive in the small intestine since it provides a more efficient exclusion of bile acids due to its narrow and negatively charged pore (Wibbenmeyer et al., 2002).
The cytoplasmic DNA binding domain of ToxR forms a winged helix-turn-helix motif (Martínez-Hackert and Stock, 1997) which binds to the so called "tox-boxes" as proposed dimers (Crawford et al., 1998;Goss et al., 2013;Krukonis and DiRita, 2003;Pfau and Taylor, 1996). Apparent dimerization of ToxR, as observed by overexpressed ToxR molecules, is achieved via the formation of intermolecular disulfide bonds between C236 and C293, located in the periplasmic domain of ToxR (Fengler et al., 2012;Lembke et al., 2020;Ottemann and Mekalanos, 1996). The formation of activity essential ToxR homodimers, containing intermolecular disulfide bonds, was shown to be dependent on periplasmic oxidoreductases DsbA/ DsbC (Fengler et al., 2012;Lembke et al., 2020). Unexpectedly, the monomeric form of ToxR, containing the intra-chain disulfide bridge, was found to be the predominant form in vivo and to be essential for the porin but not for toxT regulation (Fengler et al., 2012;Midgett et al., 2020).
The activity of ToxR depends on its effector ToxS (Almagro-Moreno et al., 2015c;Miller et al., 1989), consisting of a periplasmic domain anchored in the inner membrane. Mutations of toxS negatively influence ToxR-ToxR interactions in a homodimer and the regulation of ToxR dependent genes, including virulence factors (Lembke et al., 2020;Lembke et al., 2018). Deletions of toxS furthermore reveal an increased proteolysis of ToxR, which induces an entry of the bacterium into the dormant non-virulent state of the bacterium (Almagro-Moreno et al., 2015a;Almagro-Moreno et al., 2015c;Pennetzdorfer et al., 2019). In vitro experiments showed a physical interaction between the periplasmic domains of both proteins (Midgett et al., 2017). However, the composition as well as the binding strength of the complex formation are not clarified. Since the activity of ToxR depends on its dimerization state and its interaction with ToxS, the question arises how binding of ToxS influences the oligomerization state of ToxR.
Interestingly, the reduced form of ToxR is the protease sensitive form as shown in V. cholerae cultures incubated with reducing agent DTT (Fengler et al., 2012;Lembke et al., 2018). Proteolysis of ToxR is mainly targeted by periplasmic proteases DegS and DegP (Lembke et al., 2018;Pennetzdorfer et al., 2019). In which way stabilization by ToxS depends on the redox state of ToxR is not completely understood yet.
The versatile functions of ToxR propose separate control mechanisms, in order to regulate the activity of ToxR as direct activator, co-activator, or repressor. Since the periplasmic sensory domain seems to be crucial in the initiation of ToxR activity, we concentrated on solving the structure of this highly interesting domain. Herein, we present the Nuclear Magnetic Resonance (NMR) solution structure of the periplasmic sensory domain of V. cholerae ToxR forming an intramolecular disulfide bond (ToxRp-ox) ( Figure 1). Furthermore, this study provides an explanation for the increased proteolytic sensitivity of the reduced form of ToxRp (ToxRp-red) that was shown in V. cholerae (Lembke et al., 2020;Lembke et al., 2018). In addition, the obtained results offer new insights into the binding event of the virulence essential interaction of ToxR and ToxS. For the first time, we could identify the formation of a strong heterodimer between the periplasmic domains of ToxR (ToxRp) and ToxS (ToxSp) independent on the redox state of ToxRp. Our results reveal that ToxRp binds ToxSp in a 1:1 fashion with a dissociation constant of 11.6 nM.
Additionally, by monitoring the proteolytic cleavage of ToxRp-ox with NMR we provide a direct evidence of ToxS protective function.

| NMR solution structure and dynamics of V. cholerae ToxRp-ox
ToxRp-ox was excised from the chemically synthesized, E. coli codon optimized, full-length ToxR gene and expressed with an N-terminal 6x His tag, recombinantly in E. coli BL21 DE3 cells using 15 N and 13 C-labeled minimal medium. After purification by His-trap and size exclusion chromatography, the protein was investigated at a concentration of 0.5 mM in 50 mM NaPi buffer pH 6.5 100 mM NaCl in 90% The purity of a freshly made sample of ToxRp-ox was established by SDS page. The 2D 15 N-1 H HSQC spectrum (Figure 2a) shows mainly well-dispersed signals, which indicates that the protein is mostly well-folded.
The number of NH-resonances was found to be lower than the number of amino acids, pointing to the presence of flexibility on the intermediate NMR time scale (i.e., molecular motions in the time range of milliseconds, which leads to extensive broadening and thereby disappearing NMR signals). Sequential backbone and side chain assignments localized the well-dispersed signals to residues 201-270.
The signals of the C-terminal 24 residues were missing in the spectra ( Figure 2b).
The solution structure of the monomeric oxidized sensory periplasmic domain of V. cholerae namely ToxRp-ox was determined using NOEs and dihedral angles predicted based on chemical shifts by TALOS+ (Shen et al., 2009). It forms an αβ-fold in solution, consisting of an alpha helix stacked against a four stranded beta sheet (Figure 2c,e).
The beta strands form a four stranded beta sheet comprised by two hairpin motifs. Residues W229-K243 are forming an α-helix, the residues S227 and N228 are part of a beta turn type I and resemble an alpha-helical like formation. There are two cysteines (C236 and C293) in the periplasmic domain of ToxR. C236 is in the middle of the helix of F I G U R E 1 Schematic presentation of ToxR domains and its interaction with ToxS, both located at the inner membrane of V. cholerae. The two cysteines of the periplasmic domain of ToxR (ToxRp) can form an intramolecular disulphide bridge referred to as ToxRp-ox, the reduced state is named ToxRp-red. The periplasmic domain of ToxS (ToxSp) can form heterodimers with ToxRp-ox (ToxRSp-ox) and ToxRp-red We previously reported that the cysteines in ToxRp are in an oxidized form using a 2,2,2-trifluoroethyl 6-thio-β-d-glucopyranoside as a selective tag for cysteines (Fröhlich et al., 2012). This is confirmed by the Cβ chemical shift of C236. Therefore, the presented NMR structure represents the monomeric oxidized form of ToxRp (ToxRp-ox), which is the active conformation of ToxR in vivo (Fengler et al., 2012;Lembke et al., 2018;Mey et al., 2012;Ottemann and Mekalanos, 1996). This is supported by recently published data on a ToxR homolog from Vibrio vulnificus (amino acid sequence identity 55.2%) which suggested that the protein lacks a dimerization interface and therefore forms an intramolecular disulfide bond under nonreducing conditions (Midgett et al., 2020).

| The hydrophobic core of ToxRp-ox forms around the α-helix
The positioning of the helix of ToxRp-ox is stabilized by its hydrophobic network (Figure 2d). Half of the helix is buried in the interior of the protein with residues L230, I233, V237, and Y240 involved in the formation of the hydrophobic core.
L226 is located close to the N-terminus of the helix and interacts together with I233 from the helix with L209 from strand1. I233 from the helix is interacting with V217 of strand2. V237 from the helix, close to activity essential C236, builds apolar contacts to all four beta strands, and is therefore most probably highly significant for the fold. In detail, it contacts V212 from strand 1, V215 and V217 from strand 2, V253 from strand3 and L265 from strand4. Y240 from the helix forms π-π stacking forces with Y267 from strand4.
The β strands form a four stranded β sheet which is additionally stabilized by hydrophobic networks between the strands. Strand 1 forms hydrophobic contacts with L209 to V217 of strand 2, and additionally to V212 to V215 also from strand 2. V215 and V217 from strand 2 are close to V253 from strand 3. V253 interacts with L265 and Y267 from strand 4. Strand 3 is mostly stabilized by V253.

| Two conformations of ToxRp are found under reducing conditions
Under reducing conditions, ToxRp adapts two conformations accompanied by a second set of peaks in the backbone spectra ( Figure 3a).
One conformation resembles the ToxRp-ox structure closely. The second conformation shows signals in the typical random coil regions of 15 N-HSQC spectra. The amount of the two conformations of ToxRp under reducing conditions is similar, does not change over time and is found repeatedly in independent protein preparations.
Therefore, proteolysis can be ruled out. Interestingly, the C-terminal stretch that could not be assigned under oxidative conditions, is visible in the NMR spectra recorded under reducing conditions adapting an unstructured and highly flexible conformation. Noteworthy, we cannot exclude that the C-terminus has a second 'NMR-invisible' conformation moving on an intermediate time scale.
The presence of reducing agents has a significant effect on the ToxRp structure, especially on its C-terminal region containing the second cysteine. The presence of a second unstructured conformation visible in the NMR spectra under reducing conditions indicates that the reduction of the disulfide bridge destabilizes the ToxRp fold.
Indeed, previous in vivo studies could show a higher instability of reduced ToxRp (Lembke et al., 2020;Lembke et al., 2018). It was hypothesized that the activity of ToxR is controlled by its degradation and stability, which itself is dependent on the redox state of the cysteines in the periplasmic domain. A comparison of 15 N-HSQC spectra of ToxRp with and without reducing agents ( Figure 3a) shows that the structured conformation of ToxRp-red is similar to ToxRp-ox, which forms an intramolecular disulfide bond. The absence of a disulfide bond does not change the overall fold of the protein, but significantly alters its stability. This is corroborated by the Talos + secondary structure prediction, using the backbone assignments, that shows a similar pattern of secondary structure for both monomeric redox states: an alpha-helix, flanked by two beta strands on each side ( Figure S1).

| ToxRp-ox shows a well-structured fold in the N-terminal domain, with a flexible C-terminus
To gain further insight into the dynamical features of ToxRp in oxidized and reduced form, we carried out 15 N T 1 and T 2 relaxation as well as { 1 H}-15 N NOE measurements ( Figure 3b). These data of ToxRp-ox reveal variations of internal dynamics across the protein chain. The N-terminal part, which is located close to the inner membrane in vivo, as well as the C-terminal stretch, show a higher flexibility. ToxRp-ox contains a well-structured core flanked by these dynamic regions. The flexibility of the protein increases closer to the C-terminus. The C-terminal beta strand shows a higher flexibility than the N-terminal strands, which indicates that the following C-terminal stretch might very well be more dynamic. The overall ro-

| The periplasmic domains of V. cholerae ToxR and ToxS forming a strong salt-dependent heterodimer (ToxRSp) independent on the redox state of ToxR cysteines
Despite its essential function, ToxS remains one of the least characterized proteins in the ToxR regulon (Miller et al., 1989). We were able to express the periplasmic domain of ToxS (ToxSp), whose gene was chemically synthesized in E. coli codon optimized form, and record NMR spectra revealing its instable structure. In the absence of its interaction partner ToxRp, ToxSp tends to aggregate rapidly. The strong interaction could be confirmed by fluorescence anisotropy experiments by which we determined a K D of 11.6 ± 3.4 nM (Supplemental Figure 3).

| Binding of ToxSp slows down the proteolytic degradation of ToxRp-ox
ToxS is suggested to protect ToxR from proteolysis (Almagro-Moreno et al., 2015c;Pennetzdorfer et al., 2019). We monitored ToxRp-ox digestion by trypsin over time using NMR solution experiments and compared the proteolytic cleavage of ToxRp-ox alone to ToxRp-ox in complex with ToxSp. We chose two parameters for estimating the degree of digestion of ToxRp-ox. One parameter is the chemical shift of the HN side chain peak of W229 which is at the N-terminal end of the alpha helix of ToxRp-ox and shows a significant change in the chemical shift when it is not structured. As a second parameter, we analyze the presence of additional peaks in the middle region of the spectrum, where usually unstructured residues appear. After approximately 1.3 hr, we already observe additional signals in the unstructured region of the spectra, nevertheless the helix seems to be still intact (Figure 5a). The next spectrum is recorded after 2.0 hr and shows two peaks for the W229 side chain HN, meaning that both digested and intact helices are present in the sample. The helix is completely digested after 6.5 hr. The signals in the unstructured middle region of the spectra are gaining in intensity over time.
In comparison, the experiments with the ToxRSp-ox complex show a significant slower degradation of ToxRp-ox by trypsin ( Figure 5b). After 12.5 hr, unstructured HN signals appear in the middle region of the spectrum. Compared to ToxRp-ox alone, trypsin digestion of ToxRp-ox starts approximately 11 hr later when it is bound to ToxSp. The unstructured W229 HN side chain peak appears for the first time after 16.5 hr, which is 14.5 hr later compared to ToxRp-ox alone. The measurement was stopped after 29 hr.
The last recorded spectrum still shows two peaks for the W229 HN side chain, meaning that there are still intact helices present in the sample. The experiments clearly show that ToxRp-ox is more stable against trypsin proteolysis when it is protected by ToxSp. The digestion of ToxRp-ox by trypsin is slowed down significantly when ToxSp is bound.

| D ISCUSS I ON
The virulence expression system of the cholera causative bacterium V. cholerae, namely ToxR-regulon, represents a highly complex regulation system thereby enabling the bacterium to rapidly switch from  (Krukonis et al., 2000).
Its versatile functionality is connected to complex mechanisms in order to control its activity. Environmental signals like bile salts (Provenzano and Klose, 2000) as well as the inner membrane protein ToxS (Lembke et al., 2018), significantly stimulate the activity of ToxR by directly binding to its sensory periplasmic domain (Midgett et al., 2017;Midgett et al., 2020).
Here, we present the NMR solution structure of the ToxR periplasmic domain from V. cholerae revealing an αβ-fold, consisting of a four stranded beta sheet stacked against an alpha helix followed by a flexible C-terminus (Figure 1). In the absence of reducing agents, the protein domain forms an intramolecular disulfide bond between C236, in the middle of the helix, and C293 at the C-terminus, referred to as ToxRp-ox (Figures 1 and 2).

| ToxRp homologs share similarities in the hydrophobic core but adapt different C-terminal conformations
A comparison of the residues involved in the stabilization of the core structure of V. cholerae ToxRp-ox and V. vulnificus ToxRp-ox-V.v. (Midgett et al., 2020) show significant similarities in the amino acids and their positioning in the structure (Figure 6a). This could indicate that the hydrophobic core of ToxR is conserved in the Vibrio family.
Regarding the C-terminal region the structures of ToxRp-ox from V. cholerae and V. vulnificus significantly differ (Figure 6b).
ToxRp-ox-V.v. has an additional beta strand (β5) and an additional short helix (α2) in the C-terminal region (Midgett et al., 2020). The C-terminal stretch following beta strand 4 is invisible in the NMR spectra of V. cholerae ToxRp-ox, and thus, does not develop a stable NMR-observable structure. We propose that the cysteine-cysteine connection confines its motion leading to an intermediate exchange on the NMR time scale. Since the C-terminus is involved in the activity dependent formation of the disulfide bridge, it is possible that the C-terminal region differs among Vibrio species and may be essential for their individual functions. Another explanation for the differences observed in the C-termini could be the different physicochemical states under which the structures are determined. It is also possible that the C-terminal region has a low tendency to form structural elements, which under solution conditions is not strong enough to result in a stable folding. In a solid form, crystal packing forces may artificially stabilize the formation of these secondary structure elements in the C-terminus.

| A reduction of ToxRp cysteines lead to a destabilization of its structure
To address the question of the role of the redox state on the structure of ToxRp, we assigned the backbone of ToxRp under reducing conditions (ToxRp-red) and could detect that ToxRp adapts two conformations when its cysteines are reduced (Figure 3c). The first conformation is structurally similar to the monomeric oxidized form of ToxRp, the second one is highly flexible and probably unstructured.
The disruption of the intramolecular disulfide bond therefore seems to destabilize the ToxRp fold.

| ToxRp interacts with ToxSp in a 1:1 fashion resulting in a strong heterodimer
There exist still many unanswered questions regarding the mecha- In addition, NMR trypsin digestion studies could directly show a significant slowing down of the degradation of ToxRp when it is bound in the heterodimer complex ( Figure 5). ToxSp seems to protect putative protease cleavage sites of ToxRp and thereby increases its stability. We propose that the interaction between ToxR and ToxS is another mechanism to control the activity of ToxR by altering its accessibility to proteases. The ToxRS control-mechanism works independently of its cysteines but is dependent on the salt concentration.
Since ToxR has versatile functions and can act as activator, co-activator or repressor, it is likely that its activity is also regulated on different levels. We propose that ToxR activity is mainly controlled by its stability which is dependent on different factors (Figure 7). The reduction of ToxRp cysteines represent one possibility to decrease ToxR stability. This regulation is controlled by periplasmic oxidoreductases DsbA and DsbC in vivo (Fengler et al., 2012;Lembke et al., 2020;Lembke et al., 2018). The interaction with ToxS represents another possibility to increase ToxR stability by directly protecting under the applied conditions. Therefore, our data also support the theory that dimerization of ToxR, in order to induce transcription, is activated by the presence of DNA (Lembke et al., 2020;Midgett et al., 2020).

| Cloning expression purification
All constructs listed in Table 1 were generated by using standard with or without 4 mM β-ME.

| NMR experiments
All NMR spectra were recorded on a Bruker Avance III 700 MHz spectrometer equipped with a cryogenically cooled 5 mm TCI probe using z-axis gradients at 25°C. All NMR samples were prepared in 90% H 2 O/10% D 2 O. Spectra were processed with NMRPipe (Delaglio et al., 1995).
Secondary structure predictions were done using backbone assignments and TALOS+ (Shen et al., 2009).

| NMR trypsin degradation study
The trypsin digestion experiments were carried out in 50 mM sodium phosphate, 100 mM sodium chloride at pH 6.5 measured in a 3 mm tube. For the experiments, only ToxRp-ox was 15 N-labeled.
The concentration of 15 N-labeled ToxRp-ox and ToxRSp-ox was 3.14 mg/ml in a volume of 170 µl. A reference 2D 15 N HSQC was recorded first, then trypsin was added in a 1:10 000 ratio of trypsin to ToxRp-ox or ToxRSp-ox. After addition of the protease, a series of 15 N-HSQC experiments were acquired.

| Fluorescence anisotropy
For the fluorescence anisotropy measurement, a mutant of ToxSp containing a cysteine between the His6x tag and the protein was used. ToxSp H10_S11insC was labeled with the fluorescent dye 5-iodoacetamidofluorescein 5-IAF, which reacts to reduced cysteines according to the manufacturers protocol. In short: The mutant was expressed and purified according to the protocol used for ToxSp.
A 4 mM β-ME was added to all buffers to prevent oxidation of the cysteines. ToxSp H10_S11insC was then rebuffered in conjugate buffer (200 mM sodium phosphate, 200 mM sodium Chloride, 1 mM EDTA, and 2 mM β-ME pH7.5). 5-IAF was dissolved in DMSO then added in a 10 times excess over ToxSp H10_S11insC and incubated for 2.5 hr in the dark. To remove unbound tags, the sample was puri-