Introduction

Inorganic anion coordination by proteins is of paramount significance for recognition, transformation, and translocation processes in biology1,2,3. From a supramolecular perspective, this is simply driven by patterned cooperative, non-covalent interactions (e.g., arginine residues and backbone amides) combined with the hydrophobic effect as mimicked by synthetic systems2,4. These host design elements are often matched to the intrinsic physical properties of a guest anion, including the anion size, shape, charge, basicity, and dehydration enthalpy4,5,6,7. On this basis, proteins can exhibit an exquisite degree of anion selectivity. Historically, this has been exemplified by comparing the bacterial periplasmic proteins for phosphate and sulfate. The phosphate-binding protein (PBP) from Escherichia coli cannot recognize sulfate, whereas the sulfate-binding protein (SBP) from Salmonella typhimurium cannot recognize phosphate8,9. On the other hand, proteins can be promiscuous, resulting in the recognition of more than one anion; examples include carbonic anhydrase enzymes and chloride channels10,11. While such phenomena are observed, the underlying rules by which “selective” anion coordination is achieved by proteins remains underexplored.

To this end, we have established a program to understand and exploit promiscuity in anion-sensitive fluorescent proteins12,13,14,15,16. More recently, we broadened our scope to unravel the molecular drivers of promiscuity versus specificity in anion-binding proteins for fundamental research and translational applications17. Here, we have chosen to work with NreA, a unique nitrate sensory protein found in the food-grade bacterium Staphylococcus carnosus (ScNreA) (Fig. 1A)18,19. The first observation pointing to promiscuity is from early-stage phasing with non-native iodide to determine the ScNreA X-ray crystal structure (Fig. 1B)20. ScNreA adopts a GAF_2 domain fold with a binding cleft defined by dipole moments oriented along the α3 and α6 helices20. Specifically, three main chain amides connecting residues L67, A68, and I97, and one sidechain amide from residue W45 coordinate to nitrate and iodide. Nearby hydrophobic residues—L61, G66, Y95, P96, and I97—shield the binding pocket from bulk water20.

Fig. 1: X-ray crystal structure of NreA from Staphylococcus carnosus (ScNreA).
figure 1

A Nitrate (PDB: 4IUK) and B iodide (PDB ID: 4IUH) structures. The residues within 4 Å of nitrate and 5 Å of iodide are shown. All residues are labeled with their single letter amino acid code and position number. The colors of the residues correspond to the secondary structure as follows: alpha helix (turquoise), beta sheet (magenta), and flexible loop (pale salmon).

Full size image

The second observation of promiscuity stems from the biological function of NreA. ScNreA is a key component of the nitrate regulatory system NreABC21. This controls nitrate-induced anaerobic respiration to generate ammonia via nitrite22. Under aerobic conditions and in the absence of nitrate, ScNreA forms a complex with the oxygen sensor kinase NreB23. In turn, this prevents phosphorylation of the response regulator NreC and transcription of the nitrate reductase NarG21. However, under anaerobic conditions in the presence of nitrate, ScNreA dissociates from NreB, relieving the downstream repression23. Similar effects are observed with non-native iodide but not with native nitrite, controlling the bacterium’s metabolic state20. Whether this level of selectivity is solely at the level of ScNreA is unknown.

Building from these observations, ScNreA is ideal for our goal and affords four advantages. One, it is a soluble protein that can be readily expressed recombinantly and purified on a preparative scale from Escherichia coli20. Two, ScNreA is amenable to analysis of anion binding thermodynamic properties. Indeed, Niemann and colleagues previously measured the nitrate binding affinity of wild-type and mutated NreA with isothermal titration calorimetry (ITC)20. While mutations at W45 and Y95 can be tolerated, the nitrate binding affinity and/or biological function are attenuated. Three, ScNreA has only two tryptophan residues—W45 and W112 on a flexible loop facing bulk water20. Given this, the intrinsic fluorescence signal of W45 could be used to capture and quantitatively analyze binding kinetics. Four, the X-ray crystal structures are available which enables studies using molecular dynamics (MD) simulations. To date, no other biophysical insights for ScNreA have been reported. Here, we integrate experimental and in silico methods to dissect the interactions of ScNreA with the native nitrate and nitrite and non-native iodide.

Results

Thermodynamic characterization

We first optimized a method to express and purify ScNreA with a C-terminal polyhistidine-tag from E. coli (Figs. S1 and S2). For all experiments, the monomeric form was evaluated at pH 7.8 in 50 mM HEPES buffer with 50 mM NaCl. To quantitatively determine the thermodynamic parameters of nitrate, nitrite, and iodide recognition, ITC was employed at 25, 20, 15, and 10 °C (Figs. S3–S9, Tables S1–S9 and 1). Overall, the three anions interact with ScNreA with similar thermodynamic governing principles. All binding reactions are exothermic and driven by large enthalpic changes (ΔH) with minimal entropic gains (TΔS), except at 25 °C with nitrate. Moreover, the Gibbs free energy change (ΔG) remains relatively constant at each temperature for a given anion. As such, clear differences emerge in the relative binding affinities (Kd). This trend can be ranked from strongest to weakest Kd as follows: nitrate > iodide > nitrite. Of note, our measured binding affinity for nitrate (Kd ≈ 4 µM) is approximately (ca.) 6-fold stronger than that previously reported (Kd ≈ 22 µM) at 25 °C20. Since the previous work was carried out in buffer with 200 mM NaCl, compared to 50 mM NaCl used here, we confirmed that differences in ionic strength could account for the observed variation (Fig. S10 and Table S10). In addition, we evaluated the effect of substituting sodium with potassium as the counterion for each anion. While the absolute values of the thermodynamic parameters differ, the overall trend for the Kd is maintained (Fig. S11 and Table S11). Indeed, similar ionic strength and counterion effects have been reported in other systems24,25. Although these data are intriguing, further deconvolution of these conditions lies beyond the scope of the present study and will be the subject of a future investigation.

Table 1 aThermodynamic parameters of anion binding to ScNreA from 10–25 °C
Full size table

The ΔG for nitrate binding to ScNreA is ca. −31.2 kJ/mol (Table 1). In line with the concept of enthalpy-entropy compensation, the enthalpic term increases and the entropic term decreases as the temperature increases26. The dependence of ΔH on temperature can be interpreted in terms of a negative heat capacity change upon nitrate binding (ΔCp ≈ −308.2 J/mol K, Table S7)27. Such a value can be linked to protein motions that result in shielding of nonpolar surfaces from bulk water due to increased hydrophobic interactions upon nitrate recognition. Contributions to the total entropy change (ΔStotal) can be dissected using a previously established model28. At 25 °C, while water release (ΔSsolv) is favorable with nitrate, overcompensation by configurational changes (ΔSconf) and restricted motions (ΔSr/t) results in an unfavorable ΔStotal of −4.6 J/mol K (Tables S8 and S9).

Next, the ΔG for iodide binding to ScNreA is ca. −24.9 kJ/mol (Table 1). This is favorable but translates to a Kd that is weaker than nitrate (Table 1). Interestingly, unlike nitrate, the enthalpy-entropy change is inversely correlated with increasing temperature. This trend grants a positive ΔCp of ca. 242.3 J/mol K (Table S7). To our knowledge, this is a rare observation, particularly for an anion binding protein29. A positive ΔCp can be attributed to shielding of polar surfaces, exposing nonpolar surfaces to bulk water30. The unique hydration pattern of iodide could also be a contributing factor. Water molecules in the first hydration shell can have stronger interactions with water molecules in the second hydration shell compared to iodide itself29,31,32. In line with this, the ΔSsolv at 25 °C is unfavorable, likely stemming from dehydration of iodide, but the ΔSconf contribution at 25 °C dominates, resulting in a favorable ΔStotal of ca. 11.1 J/mol K (Tables S8 and S9).

Lastly, consistent with the weakest Kd, the ΔG for nitrite binding to ScNreA is ca. −22.7 kJ/mol (Table 1). As observed with iodide, ΔH decreases and TΔS increases to compensate as temperature increases but over a narrower range. The absolute value for ΔCp is ca. 33.1 ± 11.2 J/mol K, but the analysis of the uncertainty indicates that ΔCp fluctuates around zero (Table S7). While this suggests that there is little to no change in shielding of the protein surface area, it does not necessarily indicate an absence of a conformational change33. Looking to the ΔStotal of ca. 16.1 J/mol K, the ΔSsolv contribution is negligible whereas the ΔSconf is favorable (Tables S8 and S9).

Kinetic characterization

Building from the thermodynamic insights, we next probed if and how characteristics of association and dissociation kinetic properties could contribute to anion recognition by ScNreA. The intrinsic fluorescence signal of W45 at λem = 305 nm was monitored using time-resolved stopped-flow fluorescence emission spectroscopy for the binding of ScNreA with nitrate, iodide, and nitrite34. To maximize the signal-coverage acquisition for rapid association and dissociation events, all measurements were carried out at 10 °C. Under pseudo-first order conditions, a rapid, concentration-dependent fluorescence quenching is observed for each anion (Figs. S12–S19 and Table 2).

Table 2 aKinetic parameters of anion binding to ScNreA at 10 °C
Full size table

Each kinetic trace was fitted to a single exponential model to determine the observed apparent association rate constants (kobs)35. From the linear dependency of kobs as a function of anion concentration, supporting simple 1:1 association events, linear regression fitting was used to extrapolate the absolute association rate constant (kon) from the slope and the absolute dissociation rate constant (koff) from the y-intercept35. Tracking with the Kd, the kon can be ranked from fastest to slowest as follows: nitrate > iodide > nitrite; whereas the trend for koff is reversed: nitrate < iodide < nitrite (Table 2). These data indicate a greater contribution of the anion dissociation rates to the observed differences in Kd.

On the extreme ends, the koff for nitrate is ca. 7-fold slower than nitrite. To validate the magnitude of the absolute dissociation rate constant (koff) determined from complex-formation kinetics, apparent dissociation rates (koff-observed) were directly determined from dissociation-by-dilution stopped-flow experiments. Due to the very rapid dissociation kinetics for nitrite and iodide, this was only possible for the nitrate-ScNreA complex (Figs. S20 and S21). The magnitudes of the determined koff-observed for nitrate (Table S12) are consistent with the determined absolute nitrate dissociation rate constant (Tables 2 and S13).

In silico characterization

Finally, we employed an in silico approach to uncover the molecular basis for the experimental differences observed between each anion36. MD simulations were performed at 20 °C with a reconstructed, full-length ScNreA in the unbound or apo state along with the nitrate, iodide, and nitrite bound states (Tables S14–S16)37,38,39. The protonation states of the amino acids were set to correspond to pH 7, and the overall charge of each system was neutralized with sodium ions. Throughout the trajectories, distances were measured between the coordinating residue (W45) and backbone amides (connecting G66 and L67; L67 and A68; P96 and I97) (Figs. 1, 2 and S22–S25). In the absence of any constraints, all three anions can leave and re-enter the binding pocket through the same pathway, albeit with varying frequencies and residence times. No alternative pathways were observed. Notably, the nitrite bound form is only observed in ca. 2% of the frames acquired in a 1 µs MD simulation. Six additional MD simulations (200 ns each) of the ScNreA-nitrite system confirm the stochastic nature of this event (Figs. S26–S29). We correlate this with the weaker binding affinity determined for nitrite. Upon closer examination of the protein surface, there is a tendency for the anions to interact with electropositive regions that are composed primarily of arginine and lysine (Fig. S30). While this observation is unsurprising, analysis of anion occupancy near the binding pocket reveals a point of entry, which is most apparent with nitrate and iodide (Figs. 3 and S31 and Supplementary Movies 1–3). Thus, the following observations are based on these two anions.

Fig. 2: Molecular dynamics (MD) simulations capture the anion binding plasticity of ScNreA.
figure 2

Distances between the alpha carbon (Cα) of residue W45 to the center of mass of A nitrate, B iodide, and C nitrite are shown as a function of time from a representative trajectory (Fig. S22).

Full size image
Fig. 3: MD simulations reveal an anion entry pathway from bulk water into the ScNreA binding pocket.
figure 3

Representative snapshots of the apo, intermediate, and bound phases for nitrate and iodide are shown from left to right. Possible interactions between ScNreA and the bound anions are shown as dashed lines. Residues within 4 Å of nitrate and 5 Å of iodide are shown as sticks with their single letter amino and position number. Note: transient water molecules are not included in this schematic but are described in the Supplementary Information (Fig. S32).

Full size image

In the apo form, the electropositive sidechains of R62 and K65 face outward into the bulk water, serving as attractive anchors (Fig. 3, apo panel). Upon anion (re-)entry, the side chain of R62 can rotate inward, and along with the backbone amide connecting L61 and R62, it can directly engage with the anion (Fig. 3, phase 1 panel; Supplementary Movie 2). Following this, the sidechain of W45 and the backbone amide connecting P96 and I97 form new interactions (Fig. 3, phase 2 panel; Supplementary Movies 1, 3), drawing the anion further into the binding pocket. In the fully bound state, these backbone amides directly coordinate to the anion (Fig. 3, bound panel).

While surrounding nonpolar residues (e.g., L61, Y95, V98) are needed to promote and maintain the hydrophobic effect, water molecules can still penetrate the binding pocket. They can transiently coordinate, and even bridge nearby residues (e.g., W45) or backbone amides (e.g., P96 and I97) to the anion (Supplementary Movies 1, 3 and Fig. S32). Even though these observations can be generalized for both nitrate and iodide, the ScNreA coordination sphere is dynamic and unique for each anion. This is supported by the frequency and number (ca. 3 to 4) of interactions during the trajectories (Figs. S33–S35). On the extreme ends, W45 prefers nitrate over iodide, whereas the backbone amide in V98 prefers iodide over nitrate. We do note that charge-balancing sodium ions in the system do not interact cooperatively with the anions but instead associate transiently with electronegative surface residues (Fig. S36).

To understand how ScNreA could register anion binding at the residue level, we next used a root mean square fluctuation (RMSF) analysis on the Cα atoms40,41. The most dynamic residues (ΔRMSF ≥ 1 Å or ≤−1 Å) are found in flexible loops (Fig. S37). Of these, residues spanning 84 to 94 can form a disordered loop or an alpha helix during the simulations, but only the latter is apparent in the original crystal structures. With respect to apo ScNreA, the RMSF of residues 59–63 and 84–88 with nitrate show a large increase (up to 4 Å), whereas the RMSF of residues 84–94 with iodide shows a large decrease (up to 4 Å). Further analysis with dynamic cross-correlation (DCC) shows changes in correlated motions upon anion binding42. While no obvious differences are observed for nitrate, a clear reduction is observed with iodide (Figs. S38–S40).

These changes in protein motion are further reinforced by the solvent-accessible surface area (SASA) calculations43. Little to no differences are observed in the absence or presence of anion (SASA ≈ 9894 Å2) with the exception that ScNreA can adopt an additional, less solvent-exposed state with iodide (SASA ≈ 9107 Å2, Fig. S41). Finally, principal component analysis (PCA) of the Cα atoms reveals the global conformations for each ScNreA state (Figs. S42–S44)44. Apo ScNreA samples a wider, spatially distinct conformational space relative to the anion bound states, which are distinct themselves. Additional clustering analysis of the results from PCA for each state is in line with the RMSF analysis, indicating the most dynamic regions in the anion bound systems (Fig. S45).

Discussion

Here, we have investigated how the soluble sensor NreA from Staphylococcus carnosus can recognize not only the native nitrate and nitrite but also non-native iodide. Our calorimetric analyses reveal that anion complexation with ScNreA is an exothermic process with enthalpy as the key driving force. When comparing the binding affinities for all three anions, a clear preference emerges. ScNreA favors nitrate up to ca. 19-fold and 51-fold over iodide and nitrite, respectively. Moreover, anion binding is not rate limiting. In fact, it is a fast process that occurs on the millisecond timescale. While differences can be measured in the on-rate, these are significantly amplified by the off-rate. The ScNreA-nitrate complex dissociates ca. 4-fold and 7-fold slower than iodide and nitrite, respectively, suggesting that dissociation kinetics governs anion selectivity.

This promiscuous anion recognition is coupled to conformational plasticity of ScNreA. Experimentally, the change in heat capacity reflects differential global protein motions, leading to shielding and exposure of nonpolar surfaces to water for nitrate and iodide, respectively, or no difference for nitrite. Since the binding site is conserved, it is plausible that each anion complex could exist in a conformational equilibrium between more than one state to give rise to the calculated ΔCp values33. Our MD simulations with ScNreA not only provide evidence for such conjectures from the experimental data but also capture new elements beyond the X-ray crystal structures.

The repeated entry and exit events observed in the MD simulations reveal gating of anion entry through electropositive sidechains and final anion coordination to amide backbones and the key tryptophan at position 45. Throughout the trajectories, water molecules are dynamic and can penetrate the binding pocket, mediating transient interactions with the bound anion. We speculate that this solvation can also act as a competitive process, facilitating anion release into bulk water. Overall, local motions across flexible loops are responsive to anion binding. These changes are distinct for each anion and occur to varying degrees but remain globally subtle. Perhaps this is unsurprising given that ScNreA is a compact sensory protein with a GAF domain fold designed to bind a small inorganic anion. Together the ITC and MD simulations suggest that the exothermic nature of anion binding arises from enthalpically favorable rearrangements in hydrogen bonding networks and electrostatic interactions. These structural changes, coupled with the desolvation of both the anion and the binding pocket, likely contribute to the observed binding energetics.

Across our observations, we also considered how the physical differences, including size, shape, basicity, and dehydration enthalpy, between each anion might contribute45,46. However, when comparing all three anions, no clear correlation can be drawn between these properties and the experimental data. Expanded experimental testing beyond nitrate, nitrite, and iodide along with the development of a comprehensive theoretical model could help fill this gap, stimulating future investigations into the underlying basis of promiscuity for ScNreA and related systems.

To conclude, we demonstrate that ScNreA can recognize more than one kind of anion when taken out of its biological context, but it does so with a preference for nitrate to meet its physiological function as a nitrate sensor. This finding may reflect a broader supramolecular feature of nitrate, and potentially, anion binding proteins in general. Promiscuity arises not only from the composition of the binding pocket, but also from overall protein fold, conformational dynamics, and native biological context, where protein–protein interactions and natural anion availability may also play a role. The discovery and systematic investigation of naturally occurring anion binding proteins could help provide further evidence for the broader paradigms suggested by our findings, complementing insights from synthetic supramolecular systems, and informing protein design for future applications in biosensing and synthetic biology.

Methods

Protein expression and purification

See Supplementary Information and Figs. S1 and S2.

ITC and analysis

See Supplementary Information and Figs. S3–S11 and Tables S1–S10.

Stopped-flow fluorescence spectroscopy and analysis

See Supplementary Information and Figs. S12–S21 and Tables S11–S13.

MD simulations and analysis

See Supplementary Information and Figs. S22–S45 and Tables S14–S16.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.