Structural determinants of persulfide-sensing specificity in a dithiol-based transcriptional regulator

Cysteine thiol-based transcriptional regulators orchestrate coordinated regulation of redox homeostasis and other cellular processes by “sensing” or detecting a specific redox-active molecule, which in turn activates the transcription of a specific detoxification pathway. The extent to which these sensors are truly specific in cells for a singular class of reactive small molecule stressors, e.g., reactive oxygen or sulfur species, is largely unknown. Here we report novel structural and mechanistic insights into a thiol-based transcriptional repressor SqrR, that reacts exclusively with organic and inorganic oxidized sulfur species, e.g., persulfides, to yield a unique tetrasulfide bridge that allosterically inhibits DNA operator-promoter binding. Evaluation of five crystallographic structures of SqrR in various derivatized states, coupled with the results of a mass spectrometry-based kinetic profiling strategy, suggest that persulfide selectivity is determined by structural frustration of the disulfide form. This energetic roadblock effectively decreases the reactivity toward major oxidants to kinetically favor formation of the tetrasulfide product. These findings lead to the identification of an uncharacterized repressor from the increasingly antibiotic-resistant bacterial pathogen, Acinetobacter baumannii, as a persulfide sensor, illustrating the predictive power of this work and potential applications to bacterial infectious disease.

illustrating the predictive power of this work and potential applications to bacterial infectious disease. 7 While the disulfide and the TMAD-adduct structures argue that the lack of reactivity toward reactive 1 oxygen-and disulfide based-oxidants could also be imposed by the required local unfolding as well as 2 the stability of the protonated thiolate, they do not provide a molecular rationale for formation of a 3 dominant tetrasulfide crosslinked species. To investigate this, we solved the structure of the tetrasulfide 4 species of C9S SqrR, obtained by reaction with an in situ organic persulfide. While this structure 5 ( Figure 3E, Figure 4C) reveals a covalent, tetrasulfide bond between C41 and C107 as expected from 6 the mass spectrometry, the tetrasulfide species, like the disulfide state, features collapse of the dithiol 7 pocket, completely shielding the tetrasulfide from solvent, particularly so for the two additional GSSH-8 derived sulfur atoms ( Figure 3E). This low solvent accessibility likely strongly attenuates the reactivity 9 of these bridging sulfur atoms toward strong electrophiles, e.g., iodoacetamide (Figure 1), and other 1 0 nucleophiles, e.g., LMW thiols that are generated in situ during the reaction, at least at these 1 1 concentrations. Moreover, this form is well folded with a decrease of highly frustrated contacts in the 1 2 cysteine pocket, suggesting that the local contacts within that pocket stabilize the folded state even more 1 3 so the reduced state ( Figure 4F).

4
It is interesting to note that the disulfide-and tetrasulfide-crosslinked SqrR species each show ArsR family repressors 36,37 . . The disulfide crosslink introduces high local structural frustration relative to the tetrasulfide 1 form. Protein structure in the proximity of the dithiol site for C9S SqrR in the disulfide (A), reduced (B) 2 and tetrasulfide (C) states are shown with the electron density corresponding to the cysteine sulfur 3 atoms, highly frustrated contacts (red broken lines) are shown for each structure (D-F).

4
While these structures hint at the mechanism of disulfide bond formation by strong electrophiles like 5 TMAD, they do not provide direct insights into the mechanism of tetrasulfide bond formation. To this 6 end, we conducted a series of kinetic reactivity experiments with various organic persulfide donors 7 followed by capping with excess IAM at variable incubation times ( Figure 5A). The conversion of the 8 reduced form and formation of the tetrasulfide bridge are kinetically well-modelled with an identical 9 pseudo first order rate constant ( Figure 5B). The minor di-and trisulfide species detected in these 1 0 profiles are, on the other hand, kinetically well-modeled as parallel reactions derived from the same 1 1 reduced SqrR, rather than as on-pathway intermediates, with much slower pseudo-first order reaction 1 2 rates independent of the identity of the sulfur donor (Supplemental table 3). A single end-point reactivity 1 3 assay of the disulfide-crosslinked C9S SqrR with GSSH indeed confirms that the disulfide is not on-1 4 pathway to the tetrasulfide product ( Figure S9). The rate constant of tetrasulfide formation can be  Figure S10). This persulfide intermediate is likely formed on C107, since it is the more nucleophilic 2 0 cysteine ( Figure 5D). We hypothesize that this initial persulfidation is followed by subsequent reactions reactions with the persulfide donor likely occur more rapidly on C107; in this case, modification of C41 would occur by intramolecular sulfur donation ( Figure 5D).  Our results define three key determinants of persulfide selectivity in dithiol based transcriptional 7 regulators: 1) a major energetic barrier to formation of the disulfide attributed to local structural 8 frustration that allows kinetic partitioning to the tetrasulfide product; 2) moderately basic thiols that are 9 poorer nucleophiles at physiological pH, and 3) a decrease in the reactivity of the less nucleophilic thiol incorporation of an extra sulfur atom in the major product (pentasulfide adduct). This is consistent with 1 7 a minimal structural perturbation of a somewhat larger cavity as inferred from a structure of the 1 8 homologous protein with putative RSS selectivity from Xylella fastidiosa (43% identity, Figure S11) 39 .

9
Consistent with the physical properties presented here, Ab BigR has been recently shown to regulate a  In conclusion, we demonstrate that it is indeed possible for non-metallated, dithiol-based 4 transcriptional regulators to achieve a level of specificity required for bacterial cells to trigger a unique 5 response to RSS/H 2 S toxicity rather than general redox misbalance. This is functionally important since 6 increased H 2 S biogenesis has been shown to be necessary for bacteria to counteract the effect of 7 antibiotic stress and host derived reactive oxygen and nitrogen species stressors 4,16,19,42 . We show that 8 this chemical specificity derives from a unique energetic landscape dictated by the protein matrix, which 9 dictates the spectrum of posttranslational modifications by minimizing local structural frustration, while 1 0 exploiting steric hindrance and low nucleophilicity. This work illustrates how the "minimal frustration  Overexpression plasmids encoding C9S, C9S/C41S and C9S/C107S SqrRs were constructed by PCR-1 7 based site-directed mutagenesis using pSUMO-CzrA as template 24 and verified using DNA sequencing.

8
Plasmids used for the expression of wild-type SqrR were reported previously 24 . Proteins were expressed 1 9 in E. coli BL21(DE3)/pLysS cells and purified as previously described for SqrR 24 .

0
AbBigR was subcloned into a pHis plasmid between NcoI and NedI sites, allowing downstream removal 2 1 of the N-terminal His 6 tag. AbBigR was expressed in E. coli BL21(DE3)/pLysS cells and purified using 2 2 a variation of a previously described protocol 46 . Briefly, freshly harvested cell expressing AbBigR paste and nucleic acid precipitation using 10% polyethylenimine (0.015 v/v) at pH 6.0. After stirring for 1 h at 1 4 °C, the solution was centrifuged at 8000 rpm for 15 min at 4 °C. The supernatant was precipitated by 2 the addition of (NH 4 ) 2 SO 4 to 70% saturation with stirring for 2 h. After centrifugation at 8000 rpm for 3 15 min, the precipitated protein was dissolved and dialyzed against Buffer A (25 mM MES, 75 mM 4 NaCl, 2 mM TCEP, 1 mM EDTA, pH 6.0). This solution was loaded onto a 10 mL SP (sulfopropyl) Fast 5 Flow cation exchange column equilibrated with Buffer A. The protein was then eluted using a 150 mL 6 linear gradient of 0.075-0.75 M NaCl. 7 C9S SqrR samples for backbone assignments were isotopically labeled using published procedures 47 8 with all isotopes for NMR experiments purchased from Cambridge Isotope Laboratories.

9
Selenomethionine labelled C9S SqrR was obtained using a published procedure 48 .  All proteins were crystallized at 20 °C by a hanging-drop vapor diffusion method at a concentration of 6 1 5 mg/mL in the crystallization buffer (25 mM Tris, 0.2 M NaCl, 2 mM EDTA, 5% glycerol, pH 8.0). WT

6
SqrR was crystallized in the presence of 2 mM TCEP with a mother liquor containing 2.5 M ammonium 1 7 sulfate, and 0.1 M sodium acetate pH 4.6. Reduced C9S SqrR was crystallized in the presence of 2 mM 1 8 TCEP with a mother liquor containing 1.8 M sodium citrate, pH 6.4. We could distinguish no difference ESI-MS. The protein was buffer exchanged in anerobic conditions into the crystallization buffer.

2
For cryoprotection, crystals were transferred for a few seconds into a reservoir solution supplemented 1 with 20% (v/v) glycerol and were subsequently flash-frozen in liquid nitrogen. Diffraction data were 2 collected at 100 K at the 4.2.2 beamline at the Advanced Light Source (Berkeley, CA). The data were 3 indexed, integrated, and scaled using the XDS package.

4
The structure of reduced C9S SqrR form (PDB code 6O8L) was solved by molecular replacement using 5 PHASER and the PDB code 3PQJ as search model 49 . Phase calculations were performed using Phaser in 6 PHENIX AutoMR module 50 . The C9S SqrR disulfide state (PDB code 6O8O) form was phased using  define while the carbonyl on the other N is likely in more than one conformation.

8
The conformational frustration patterns were calculated using the protein Frustratometer software 1 9 (www.frustratometer.tk/) 45,53 . This software that has been described in more detail elsewhere 45,53 and has contact is defined as 'minimally frustrated' if its native energy is at the lower end of the distribution of contact is defined as 'highly frustrated' if its native energy is at the other end of the distribution (i.e., 3 0 most of the other distances for that amino acid pairs in that position would be more favorable for folding 3 1 13 than the native ones by more than one standard deviation of that distribution). If the native energy is in 1 between these limits, the contact is defined as 'neutral'. In Figure 4 (main text), minimally frustrated and neutral contacts are omitted to focus in the increase of 3 the highly frustrated contacts in the disulfide state when compared to the reduced and tetrasulfide state.

4
All the minimally frustrated contacts can be found in Figure S7, with only the neutral contacts omitted 5 for clarity. was added and the reaction was sealed and stirred at room temperature for 1 hour. The pH was then 1 9 adjusted to 3.5 using pyridine and cysteine trisulfide was precipitated by addition of 3X volume of a 2 0 50/50 mixture of ethanol/THF. The precipitate was filtered, washed with 50/50 ethanol/THF and dried.  Alkylation step. Sample preparation was adapted from earlier reports 57,58 and optimized for C9S SqrR.

5
All experiments ware performed anaerobically in a glovebox. The purified C9S SqrR was first buffer   Data were collected and analyzed using MassLynx software (Waters).

1
The relative intensities of each species were obtained from the convoluted intensities of peaks the species with relative intensities higher than 10% in at least one time point were included in the 1 5 analysis, except for disulfide and trisulfide that were included and fitted to evaluate them as 1 6 intermediates or parallel reaction products.

7
The kinetic data were fitted with Dynafit 59 following a minimal kinetic model pseudo-order in the kinetic models were used to fit the data: where P represents the reduced protein monomer (quantified from the relative intensity of the peak at 2 8 +114 m/z IAM capped protein), P SSSS represents the tetrasulfide bridge within the protein monomer 2 9 monomer (quantified from the relative intensity of the peak +62 m/z), P SSSS represents the trisulfide 3 0 bridge within the protein monomer (quantified from the relative intensity of the peak +30 m/z), P SS 3 1 represents the disulfide bridge within the protein monomer (quantified from the relative intensity of the 3 2 peak -2 m/z), P SSSSS represents the penta sulfide bridge within the protein monomer (quantified from 1 the relative intensity of the peak +94 m/z considered only for Ab BigR), k TET is rate constant for 2 formation of the tetrasulfide, k TRI is rate constant for formation of the trisulfide, k DI is rate constant for 3 formation of the disulfide and k PENTA is rate constant for formation of the pentasulfide. where the representation is the same as in Mechanism (1), but also including P SSH as the representation These models and the obtained parameters imply that the rate limiting step is the formation of a with the SNAP donor suggests that the rate of the reaction depends also on the nature of the persulfide 1 9 donor.

0
The presence of persulfide radicals (as one would expect in the presence of hemin 38,60 ) results in a two- persulfides consistent with what has been previously reported 38 .

9
The parameters obtained for the reaction in the presence of DNA suggest that there is kinetic allosteric 3 0 inhibition, i.e., allokairy 61 for the reaction with GSSH which is as expected when one takes into account 3 1 the negative thermodynamic linkage in this system (Supplementary table 2).

2
The reaction with BigR is significantly slower which may derive from a restriction imposed by a well-1 folded N-terminal helix that is significantly more than that attributed to the N-terminal disordered region 2 ( Figure S6) in SqrR. TCEP to either the tetra sulfide form. This experiment validated that the tetrasulfide formation is 1 4 reversible and inhibitory of association with the DNA. All anisotropy-based data were fit to a simple 1 5 1:1, non-dissociable dimer binding model to estimate K a using DynaFit. 59 1 6