All sources of materials including preparation of pHtrII-cyt were as described previously [20], unless otherwise stated. Buffers used were phosphate-buffered saline, 0.9 mM CaCl₂, 0.5 mM MgCl₂, 2.7 mM KCl, 138 mM NaCl, 1.5 mM KH₂PO₄, 8 mM Na₂HPO₄ (PBS); 10 mM NaP_i pH 6.0 (buffer A); 10 mM Tris-HCl pH 8.0 (buffer B); 10 mM Tris-HCl pH 8.0 in 99% D₂O (buffer C); 10 mM NaP_i pH 6.0, 10% D₂O (buffer D).

CD spectra were acquired using a Jasco 810 spectropolarimeter (Jasco Inc., Easton, Md, USA) as described previously [20]. Experimental data θ _exp were recorded in a 0.2-cm quartz cell at pHtrII-cyt concentration of 0.1 mg/mL and then converted to the mean residue ellipticity according to [θ]_MRW = θ _exp/(10*l*c*n) where l is the path length of the cell in cm, c is the protein concentration in M, and n is the number of residues. [θ]_MRW was used as the input data for the CONTIN/LL secondary structure prediction program from the CDPro software package [21, 22]. The dataset including reference spectra of denaturated proteins (SDP48) was chosen as the reference one [23]. The error bars represent the standard deviation of values predicted by CONTIN/LL assuming 10% error in the determination of concentration. The values obtained from deconvolution should be considered approximate because of inherent difficulties in CD spectra deconvolution [24, 25] and the known effects of dynamics on reliability of secondary structure prediction from CD data [26]. However, we use them here to allow quantitative comparison of structure and dynamics of pHtrII-cyt under different solvent and temperature conditions.

Measurements were performed at +20°C unless stated otherwise. Solutions for NMR spectroscopic analysis contained 5 mg/mL uniformly ¹⁵N-labeled pHtrII-cyt in buffer D. ¹H-¹⁵N-HSQC, longitudinal (T ₁) and transverse (T ₂) relaxation data, and ¹H-¹⁵N heteronuclear NOE (hetNOE) data of pHtrII-cyt in the absence and presence of 4 M salt were recorded at +10°C and +20°C using a Bruker 600 MHz spectrophotometer. Standard Bruker sequences were used to record ¹H-, ¹H-¹⁵N-HSQC, and het-NOE spectra. The T ₁, T ₂ spectra were recorded using the pulse sequence schemes described previously [27]. We used a 90° pulse of 50 microseconds for ¹⁵N for all the experiments and a 90° pulse of 14 microseconds and 9.6 microseconds for the ¹H in 4 M salt and water samples, respectively. The R ₁(1/T ₁) and R ₂(1/T ₂) rates of the 4 M salt sample were measured using the relaxation delays of 40, 80, 160, 320, 480, 640 milliseconds and 30, 64, 96, 128, 192, 240 milliseconds, respectively. Further, the water relaxation rate values were obtained at relaxation delays of 32, 64, 128, 256, 384, 512 milliseconds for R ₁ and 32, 64, 128, 192, 256, 320 milliseconds for the R ₂. HetNOE experiments were obtained by collecting the spectra with and without ¹H presaturation. A proton presaturation period of 3 seconds was applied before acquiring each ¹H‐¹⁵N HSQC spectra. The data recorded were processed using the NMRPipe, NMRview, and Sparky software packages for further analysis. The R ₁ and R ₂ rates were estimated by fitting the peak intensities at different relaxation delays using the two-parameter exponential decay function in the numerical recipes package [27, 28]. The hetNOE values for each peak were estimated by finding the ratios between the peak intensities of the ¹H-¹⁵N HSQC spectra with and without proton presaturation [27].

FTIR absorbance spectra were acquired at room temperature using a Bruker Vector 22 IR-spectrometer equipped with a DTGS detector. For film preparation, 150 μg of pHtrII-cyt in buffer B were dried under N₂ gas flow and then sealed. For measurements in solution, pHtrII-cyt was concentrated to 10 mg/mL, and buffer B was exchanged to buffer C to reduce background absorbance in the amide I region. 256 scans were taken for each sample at 1 cm⁻¹ resolution. The spectrum of residual water vapor was measured with an empty chamber and then subtracted manually from the protein spectra.

Sequence-based secondary structure predictions were performed using all of the prediction programs (SOPM, SOPMA, HNN, MLRC, DPM, DSC, GOR I, GOR III, GOR IV, PHD, PREDATOR, and SIMPA96) available on the NPS@web server [29]. The predictions of coiled-coil regions were done by the COILS prediction server [30] using the recommended window of 28 residues, MTK matrix (derived from the sequences of myosins, tropomyosins, and keratins). Since coiled-coils are often fibrous, solvent-exposed structures, all residues except for the internal a and d have a high probability of being hydrophilic. Giving equal weights to all positions therefore means being biased towards hydrophilic, charge-rich sequences. To avoid highly charged false positives a 2.5-fold weighting of positions a, d was used.

Sequence-based predictions of the dynamic properties of individual proteins were obtained from the Predictors of Natural Disordered Regions (PONDR) prediction server [31]. The analysis of the disorder propensities of the MCP signal family was carried out using RONN disorder prediction software [32, 33].

pHtrII-cyt is predicted to be a helical coiled-coil based on homology to the serine receptor (Figure 1(a)), high coiled-coil structure propensity (Figure 1(b), top panel), and high helix propensity with 63% to 97% α-helix content (Figure 1(b), bottom panel), depending on the prediction algorithm used (see Section 2.5). These predictions are not consistent with the CD spectra recorded in the commonly used buffers, PBS, buffer A and buffer B, which were all indistinguishable and essentially lacked negative ellipticity at 222 nm characteristic for helix [20]. The values obtained from deconvolution of the spectra were ∼77% unstructured, ∼7% turn, ∼13% β-sheet, and ∼3% α-helix. However, control experiments with helix-inducing solvent 2,2,2-trifluoroethanol (TFE) showed a gradual rise in negative ellipticity at 222 nm with increasing amounts of TFE added (Figure 1(d)), experimentally validating the high helix propensity inferred from the pHtrII-cyt sequence.

Figure 1(c) shows FTIR spectra in the amide I region for pHtrII-cyt in buffer C (solid line) and in dry film (dashed line). The position of the maximum of the amide I band in buffer C at 1645 cm⁻¹ is characteristic of a disordered conformation. In dry film, a shift to 1654 cm⁻¹ was observed, a value that is characteristic of α-helical secondary structure [25, 34].

Addition of glycerol to PBS at concentrations of 60% (v/v) and higher greatly stabilized helix content which based on spectral deconvolution rose from ∼19% in 60% (v/v) glycerol to ∼59% in 90% (v/v) glycerol. β-sheet and turn contributed, respectively, ∼5% and ∼12% to the spectrum in 90% (v/v) glycerol (Figure 1(e)). In sucrose, at 90% (w/v) saturation, α-helix content was ∼37%, where β-sheet and turn contributed, respectively, ∼10% and ∼21%, to the spectrum (data not shown).

Ammonium sulfate, a common agent to precipitate, stabilize, or facilitate refolding of proteins, up to 30% saturation, did not significantly alter the CD spectra of pHtrII-cyt obtained in PBS alone (Figure 1(f)). Between 30% and 60%, pHtrII-cyt adopted approximately 50–56% α-helical conformation, while at above 60% saturation, helical content was reduced and that of random coil increased. Previous crosslinking and analytical gel-filtration experiments showed that the structural transition was accompanied by protein aggregation at high ammonium sulfate concentrations [20]. Since the native intracellular environment for N. pharaonis includes 4 M KCl, we also investigated the effect of monovalent and bivalent cationic solutions on pHtrII-cyt CD spectra. In solutions of up to 2 M of either LiCl, NaCl, KCl, RbCl, or CsCl in PBS and CaCl₂ and MgCl₂ in buffer B, pHtrII-cyt CD spectra exhibited predominantly unfolded features. However, at 4 M salt concentrations, helix conformation was stabilized to different degrees depending on the nature of the salt added (Figure 2(a)). A plot of hydrated ion size (ion radii taken from [35]) versus ellipticity (Figure 2(d)) shows a minimum at the position of NaCl, indicating that NaCl is best in stabilizing pHtrII-cyt structure. We note that maximum helix induction by the high salt additions required buffer D (containing phosphate and other salts) as the base buffer, suggesting that there is additional complexity in the factors contributing to maximum stability than can be explained by a single solvent component. This also reflects the complex solvent environment of the halophilic cytoplasm. For example, studies in H. marismortui indicated that although KCl was the predominant salt, there were also considerable amounts of sodium and magnesium salts present [36].

CD was used to monitor thermal stability of pHtrII-cyt. The negative ellipticity at 222 nm as a function of temperature is shown in Figures 2(b) and 2(c). In PBS, CD spectra were essentially independent of temperature, characteristic of random coil at all temperatures probed (0°C to +99°C). In the presence of 4 M salt (Figure 2(b)), helicity was lost cooperatively with rising temperature with a single transition point, observed at +41.5°C in 4 M LiCl, +37.9°C in 4 M NaCl, +36.3°C in 4 M KCl, and +32.4°C in 4 M RbCl. The melting temperature was thus linearly dependent on ionic radius of the cations, with smaller ions stabilizing helix stronger (Figure 2(e)).

Denaturation in 90% (v/v) glycerol proceeded similar to that observed in the presence of high salt concentrations, that is, cooperatively with a transition point of +38.3°C. Thermal denaturation of pHtrII-cyt in PBS + 40% saturation ammonium sulfate was fitted by a sigmoidal curve with a transition point significantly higher than those obtained under any other condition, at +66 ± 3°C. However, the shape of the curve was not symmetric indicating the presence of multiple species with different melting temperatures. In 98.5% (v/v) TFE, a transition point of +36 ± 4.5°C was obtained, but seemed to follow a linear dependency with temperature, suggesting a gradual melting with little cooperativity.

Resonances in one-dimensional ¹H spectra of pHtrII-cyt in buffer D (data not shown) were very sharp and clustered around 8.3 ppm. No resonances were observed upfield of 0.8 ppm. These spectral features are characteristic of disordered or globally dynamic proteins [37]. ¹H-¹⁵N-HSQC spectra of uniformly ¹⁵N-labeled pHtrII-cyt at +10°C and +20°C in buffer D with and without 4 M KCl are shown in Figures 3(a), 3(b), respectively. Spectra recorded at +4°C, +10°C, +25°C, +37°C, and +41°C were qualitatively similar in chemical shift dispersion but varied greatly with respect to individual chemical shifts (data not shown). A maximum of 254 of the expected 271 backbone NH resonances were resolved at +4°C in buffer D, while only 216 were resolved at +41°C in buffer D. Only 27 peak positions were independent of temperature for low salt and 44 peak positions for high salt samples. To quantify the dynamic properties of HtrII-cyt, we measured ¹⁵N relaxation parameters T ₁, T ₂, and hetNOE at two different temperatures (+10°C and +20°C) in the presence and absence of 4 M KCl. The results are listed in Table 1. Values could only be obtained for 141–148 of the 251–258 peaks observed in the four respective HSQC spectra, while 106–110 of the peaks decayed so rapidly that reliable relaxation parameters could not be determined. This indicates that the backbone resonances of these residues are most likely rigid or undergo conformational exchange on the microsecond-to-millisecond time scale. For the 141–148 peaks for which relaxation parameters were determined in buffer D at +10°C and +20°C, the T ₁, T ₂ and hetNOE values are characteristic of random coil structure as, for example, observed in unfolded lysozyme in 8 M urea [38]. In contrast, the observed T ₁, T ₂, and hetNOE values are far from those observed in a rigid helical protein. There was very little deviation from the average in all relaxation parameters at both temperatures suggesting that the subset of resonances for which relaxation parameters could be measured experiences similarly restricted motions, on the nanosecond-to-subnanosecond time scale. In the presence of 4 M KCl, hetNOE values were significantly less negative at both temperatures. T ₂ average values were overall slightly increased as compared to those observed in the absence of high salt, probably due to the difference in solvent viscosity. Importantly, however, the standard deviation in T ₂ times was much higher in the presence of salt, reflecting the appearance of resonances with lower T ₂ times (higher relaxation rates), indicating restriction in conformational space similar to what has been observed in hydrophobic clusters in unfolded lysozyme [38].

These results strongly suggest that both in the presence and absence of salt, there is significant restriction in the conformational space being explored by the ensemble of structures. The NMR results provide detailed information on the number of resonances whose dynamics are affected by changes in solvent conditions. For 141–148 backbone amide groups out of 271 pHtrII-cyt residues (52–55%), relaxation rates indicated random-coil behavior under low salt conditions. These represent the most flexible regions of pHtrII-cyt. In these regions, the presence of 4 M KCl diminishes the degree of flexibility significantly. In the remaining 45–48% of resonances, there is stronger restriction in conformational space that is present both in low and high salt conditions. These residues experience significant changes in their environment when comparing low and high salt conditions because the chemical shift values change and a broader range of values is found under high salt conditions indicating more highly structured states.

The experimental evidence presented above uniformly suggests that the pHtrII-cyt structure is highly dynamic. Since the Tar-cyt and Tsr-cyt structures are also believed to be dynamic (see Section 1), we investigated the propensity for disorder in all three proteins and in the entire MCP family using two methods: PONDR [31] and RONN [32, 33] (Figure 4). Y-axis values (probability of disorder) higher than 0.5 generally indicate disordered regions. In all three proteins PONDR predicts substantial disorder for the cytoplasmic domain, while overall the RONN probabilities for disorder were more moderate (Figure 4(a)). Using RONN, we investigated the propensity for disorder in the entire MCP family (Figure 4(d)). A mean value of 0.5 with a low standard deviation of ~0.09 was obtained. Tsr-cyt and pHtrII-cyt closely follow the mean RONN disorder values in the MCP family at the majority of sites, indicating that high flexibility but not complete lack of structure is conserved in the MCP family. Notably, some regions at cytoplasmic domain alignment position 100–110, 190–200 and at the extreme C-terminus (>275) displayed least flexibility in the MCP family (Figure 4(d)). These correspond to regions of lowest B-factors observed in the Tsr-cyt crystal structure [13].

Previous studies showed that the shape of pHtrII-cyt is an elongated rod with a radius of gyration of 54 Å [20], both as a monomer in low salt and as a dimer in 4 M KCl. These high KCl concentrations are physiological for N. pharaonis [39, 40]. This shape is consistent with the helical coiled-coil dimer of the Tsr-cyt crystal structure [13].

However, this shape is not consistent with CD data lacking evidence for helix altogether for the monomer, and indicating only 24% helix for the dimer in 4 M KCl [20]. Here, we provide evidence that despite “random coil” spectroscopic features, pHtrII-cyt can exist as a highly flexible, loosely packed but folded helical coiled-coil: (i) sequence-based structure predictions all strongly support a helical coiled-coil structure for pHtrII-cyt, and addition of TFE, a well-known helix inducing agent, is able to generate CD spectra that are characteristic for ~80% helix; (ii) both, uncharged electrolytes and salts induce negative ellipticity in CD spectra of pHtrII-cyt, but the thermal midpoint transitions occur at very low temperatures indicating thermal instability; (iii) the majority of signals in ¹H, ¹⁵N-HSQC spectra are highly temperature-sensitive, with the largest number of visible peaks observed at low temperatures and best spectral dispersion in high salt at low temperatures. Because small angle neutron scattering indicates an identical radius of gyration [20] regardless of the differences in helix content in CD spectra and spectral changes in the NMR spectra, we propose that stabilization of more folded conformations by added stabilizing agents in the case of pHtrII-cyt means a reduction in dynamics leading to a more compact but overall similar structure.

The stabilization of folded structures of proteins by additives has been extensively studied beginning with the seminal work by Hofmeister [41]. Careful experiments, thermodynamic analyses, and discussions over the years ([42–45] and references within), including those for halophilic proteins [46–49], have led to the relatively recent “unifying theory” that both, “hard” factors, referring to excluded volume effects and “soft” factors, referring to interactions between additives, water, and proteins/nucleic acids together contribute to the mechanisms underlying the effects of cosolvents on protein/nucleic acid structural stability. While it is possible to empirically estimate the overall contribution of hard versus soft effects [44], the current uncertainty of the nature of unfolded states will preclude quantitative modeling in the near future [50]. The very high density of acidic amino acids in pHtrII-cyt typical of halophilic proteins suggests that “soft” effects will be important in the case of stabilization of the helical coiled-coil in pHtrII-cyt. To illustrate this point, we generated homology models of each monomer in the dimer from the Tsr-cyt crystal structure (Figure 5(a)). FastContact analysis [51] of the model reveals that this dimeric structure is not stable because of electrostatic repulsion between the acidic residues at the dimer interface (Figure 5(c)), while each helix in the monomer is favorably packed through hydrophobic interactions (Figure 5(b)). As a result, docking of the two helical monomers using ClusPro software [52] does not reproduce the homology model, and instead, only results in physical interaction in an area in which hydrophobic interactions are possible (Figure 5(d)). Presumably, the high salt concentrations in vivo overcome the electrostatic repulsions. To test this hypothesis, we conducted systematic comparison of monovalent and bivalent cations in the presence of a constant anion (chloride). We found that the extent and stability of the ellipticity signal in CD spectra in the presence of different cations follows exactly that predicted by Collins' law of matching water affinities [53]: the maximum stabilization is with the ion (Na⁺) that has a water affinity closest to those of the acidic amino acid residues in pHtrII-cyt, and not the one with highest charge density (Ca²⁺), or smallest ionic radius (Li⁺). However, even in the presence of 4 M NaCl, pHtrII-cyt is still a dynamic ensemble of folded coiled-coil structures, rather than a single stable conformation.

The dynamic properties of pHtrII-cyt reported here for high salt concentrations physiological for N. pharaonis are in good agreement with those reported for Tar-cyt under conditions physiological for E. coli [16]. Furthermore, Tar-cyt and Tsr-cyt were found to be globally dynamic in solution [18], similar to our observations with pHtrII-cyt. Finally, disorder predictions indicate conservation of flexible properties in the entire MCP family. This evolutionary conservation of flexibility suggests a role in function. Activation of chemotaxis receptors and phototaxis transducers occurs through very subtle effects such as small helix movements during activation and only dimers bind and activate CheA [54–56]. The compaction coupled to dimerization of a preformed ensemble of coiled-coil helical-like structures proposed here could contribute to the mechanism for activation. In support of this hypothesis, the strength of the interaction between helices in the chemoreceptor cytoplasmic domain was found to modulate kinase activity [54, 55, 57].

Recently, it was estimated that as much as 50% of human proteins belong to the “intrinsically unstructured” or “natively unfolded” class [58]. These estimates are based on predictors of disorder such as PONDR that are trained in part on sequence data of proteins with CD, FTIR, or NMR spectra like the ones we measured for pHtrII-cyt as sole evidence for disorder [59]. The findings presented in this paper supporting that pHtrII-cyt is not a random coil suggest that the terms “intrinsically unstructured” and “natively unfolded” would more appropriately be termed “highly dynamic structural ensemble." The notion of residual structure is analogous to that observed in chemically denatured proteins, where it has been proposed that a true random coil may never exist [60]. The case of pHtrII-cyt is particularly intriguing because its nonspherical shape allows us to clearly distinguish between a true random coil that may or may not retain residual structure and a near-native but highly dynamic structural ensemble. This case also reminds us that local, that is, secondary structure in the classical sense and long-range tertiary structure defining overall shape are not independent of each other. Understanding the amplitude of the motions observed in this and other systems would be an important future goal towards defining this interrelationship in more detail. Investigating how much motion is needed to lose the characteristic ellipticipty in CD spectra would be a good starting point [26].

It is an unanswered question to what extent proteins experimentally observed to be dynamic in vitro are as dynamic in vivo [42, 61]. Evidence suggests that crowding inside the cell can be mimicked with solutes in vitro. For example, FlgM, an intrinsically disordered protein gains structure in the presence of high concentrations of solutes in vitro and inside living E. coli cells [62]. Low stability of intrinsically unstructured or highly dynamic protein ensembles has been shown to be related to functions such as membrane insertion or fusion events, protein interactions, and signal transduction [63].