Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2019 Feb 11;20(7):922–930. doi: 10.1002/cbic.201800628

Experimental pK a Value Determination of All Ionizable Groups of a Hyperstable Protein

Heiner N Raum 1, Ulrich Weininger 1,✉
PMCID: PMC6619245  PMID: 30511779

Abstract

Electrostatic interactions significantly contribute to the stability and function of proteins. The stabilizing or destabilizing effect of local charge is reflected in the perturbation of the pK a value of an ionizable group from the intrinsic pK a value. Herein, the charge network of a hyperstable dimeric protein (ribbon–helix–helix (rhh) protein from plasmid pRN1 from Sulfolobus islandicus) is studied through experimental determination of the pK a values of all ionizable groups. Transitions were monitored by multiple NMR signals per ionizable group between pH 0 and 12.5, prior to a global analysis, which accounted for the effects of neighboring residues. It is found that for several residues involved in salt bridges (four Asp and one Lys) the pK a values are shifted in favor of the charged state. Furthermore, the pK a values of residues C40 and Y47, both located in the hydrophobic dimer interface, are shifted beyond 13.7. The necessary energy for such a shift is about two‐thirds of the total stability of the protein, which confirms the importance of the hydrophobic core to the overall stability of the rhh protein.

Keywords: electrostatic interactions, ionization potentials, NMR spectroscopy, pH titrations, proteins

Introduction

Protein stability results from several different forces, such as the hydrophobic effect1 or polar interactions, such as hydrogen bonds2 or aromatic interactions,3 each of which stabilize the folded state. Contrastingly, the loss in conformational entropy destabilizes the folded state relative to the unfolded state. Charge–charge interactions can have both stabilizing and destabilizing effects. Typically, charge–charge interactions result in a net stabilization effect4 and are responsible for the pH‐dependent stability profile of proteins with unfolding at high or low pH. Beyond protein stabilization, they play a key role in protein structure, dynamics, and function.5, 6, 7

The degree of electrostatic interaction between charged (ionizable) groups is reflected in deviations of pK a values from their intrinsic (or unperturbed) values.8 If the pK a value is shifted towards a lower (higher) pK a for negative charges or higher (lower) pK a for positive charges, then the presence of a charge is favorable (unfavorable). For a detailed quantification of the energetic contribution of a charge to protein stability, the determination of its pK a value in the unfolded state is necessary.9 However, if pK a values do not fall close (±0.3 units) to their intrinsic values, the pK a values of the native state alone can be used in a semiquantitative way.

Seven amino acid side chains contain ionizable groups between pH 0 and 14. Four (Asp, Glu, Tyr, and Cys) are negatively charged above their pK a or uncharged below. Three (His, Lys, and Arg) are positively charged below their pK a or uncharged above. Additionally, the termini of a protein are charged at pH 7. The protonation behavior (and the presence and absence of charges) of ionizable groups can be monitored in a site‐specific manner by means of NMR spectroscopy.10, 11, 12, 13 Chemical shifts of certain atoms in amino acids are strongly dependent on the degree of protonation and are consequently suitable probes for pK a value determination.14 However, medium‐range titrating events can also cause moderate changes in chemical shifts, which leads to additional transitions in titration curves and the need for a more involved analysis.10 By using an NMR spectroscopy based approach, about 540 pK a values have been determined experimentally in folded proteins as of 2009.15 Furthermore, for Asp and Glu residues, kinetic parameters were recently determined and linked to their pK a values.16

The small dimeric ribbon–helix–helix (rhh) protein17 from plasmid pRN1 of Sulfolobus islandicus displays an extremely high thermodynamic stability over a wide pH range.18 It consists of a β strand formed by the two monomers followed by two α helices per monomer. The native dimeric state can be stabilized relative to the monomeric unfolded state by increasing the protein concentration. Furthermore, no partially folded states are populated,18 which could lead to a misinterpretation of native‐state pK a values. Thus, it is an ideal target for a complete experimental pK a value determination over a wide pH range. The thermodynamic stability of the rhh protein is unusually high (85 kJ mol−1[18]), which must be a result of significantly more stabilizing than destabilizing interactions. For comparison, in the homologue Arc repressor, a R31M/E36Y/R40L triple mutant increased the stability from 4019 to 60 kJ mol−1.20 The rhh protein used here resembles the Arc repressor mutant at these positions. Due to the significantly higher stability of the rhh protein, it is extremely unlikely that it contains any interactions that significantly decrease its thermodynamic stability. Consequently, all significantly perturbed pK a values can be interpreted as positive or at least neutral for the overall stability in a first approximation. It contains 27 ionizable groups in a monomer: four Asp, four Glu, one Cys, three Tyr, two His, eight Lys, three Arg, and the N and C termini. A structure‐based analysis suggested a high hydrophobicity of the dimer interface as one of the main reasons for stability.17

In this study, we determine the full set of experimental macroscopic pK a values (pK a values that describe the collective behavior of a coupled system, in contrast to microscopic pK a values, which describe the behavior of one ionizable group) of a protein. We analyze the reflection of all (short and medium distance) pK a values in the NMR spectroscopy titration curves by using multiple NMR signals per ionizable group in a global manner. Using this approach, we were able to unambiguously decompose all broadened transitions to residues with neighboring ionizable groups. Furthermore, we address the impact of electrostatic contributions to the extreme thermodynamic stability of the rhh protein evaluated by large perturbations of experimentally determined pK a values compared with the intrinsic pK a values. Four possible positive Coulomb interactions (“salt bridges”), present in the crystal structure (PDB ID: 3FT7), could be confirmed in solution because five of the residues involved (four Asp, one Lys) displayed shifted pK a values in favor of the charged state. However, the contribution of the salt bridges to overall stability is minor because the protein remains folded in their absence (pH 0). In contrast, pK a values of C40 and Y47, located in the hydrophobic dimer interface, are shifted to and beyond 13.7. Around pH 12.5 (at which a low percentage of C40 and Y47 become titrated) the protein unfolds; this confirms the importance of the hydrophobicity of the protein core in the overall thermodynamic stability of the rhh protein.

Results

NMR spectroscopy signals of all 27 ionizable groups of the rhh protein could be monitored by chemical shift changes over a pH range from 0 to 12.5. Around pH 12.5, there is a continuous unfolding transition of the protein, that is, at pH 12.9 there are no signals remaining from the folded protein. Additionally, at that high pH, a much slower degradation process could be observed (Figure S1 in the Supporting Information). For all ionizable groups (between pH 0 and 14), multiple well‐resolved pH titration curves could be acquired and analyzed (Figures S2–S19). In many cases, other effects than the main transition (transition with the highest amplitude, strong impact on intraresidual probes) influenced the chemical shifts as well.

In principle, the chemical shift can be affected by three processes (other than the intraresidual ionizable group) in a pH‐dependent way: 1) pH‐induced global structural changes. This can be safely neglected for the studied protein. No protein folding intermediates or pH‐dependent changes in secondary structure were found in a previous study.18 Furthermore, such global structural changes usually happen in the slow NMR exchange regime. No such processes other than global unfolding at high pH were observed in the NMR spectra. 2) Impact of remote ionizable groups (up to 10 Å) on the chemical shift. These were treated as additional pK a values in the fit. We cross‐checked that indeed an ionizable group with matching pK a value was located in structural proximity, and that resulting chemical shift changes by these remote events were low, especially for chemical shifts that predominantly depended on the charged state of the intraresidual ionizable group.14 3) The impact of remote ionizable groups (up to 10 Å) on the charge of the intraresidual ionizable group. Ionizable groups with similar pK a values can influence each other and cause a broadened pH transition between the uncharged and charged states. Points 2) and 3) cannot be distinguished easily: the impact of 2) on the chemical shift is low, whereas 3) can only occur for two charges with similar pK a values (up to 1.3 units difference for the lower or upper microscopic pK a values21). Additionally, point 2) should affect multiple chemical shifts that report on the studied transition differently (because the influence of the nearby charge on chemical shifts is different), whereas 3) should affect all chemical shifts that report on the studied transition in the same way. This should result in the same kind of broadening caused by Coulomb interactions. In this case (3), if the chemical shift is not dependent on the ionization state of the other charge, the titration curves of the coupled system show mirror images (ideally mirrored vertically on the middle of both macroscopic pK a values). Furthermore, it must be taken into consideration that chemical shifts near both charges can be a combination of both transitions, resulting in a monophasic or non‐monotonic transition (Figure S20).

Global pK a value determination by multiple chemical shifts

To address the complications mentioned above, multiple chemical shifts were used per residue across the pH titration and a global macroscopic model was used to allow multiple pK a values (but no cooperativity factor, Hill coefficient) per pH titration curve.21, 22 The macroscopic pK a values generally reflect the total protonation/deprotonation state of a coupled system and in our protein all monitored transitions could be explained. We found that five main transitions were outside the studied pH range (R3, R24, C40, R41, and Y47). Here, only lower borders of the pK a values could be established. From the remaining 22 ionizable residues that could be quantitatively studied across the pH range, 6 main transitions can be described by a single pK a value alone (N terminus, H22, K32, D38, H51, and C terminus). For one residue, a second pK a value with a low chemical shift amplitude was necessary (K30). We interpret this example as a case in which a remote ionizable group (Y5) has an impact on the chemical shift (case 2) because some probes display monophasic behavior.

The pH transitions of the remaining 15 ionizable groups (Y5/K6, K12/Y16, E20/D23/E27, E31/D52, E46/D49, and K45/K53/K55/K66) are more complex. They can be described by two macroscopic pK a values, both with significant chemical shift amplitudes, whereby the major pK a value (macroscopic pK a value that belongs to the transition with the highest amplitude) is a close reflection of the charged state of the ionizable group. Alternatively, they can be described with one pK a value and a significantly reduced Hill coefficient (Table S1 and Figure S21). However, in the case of Y5/K6 and K12/Y16, the approach with two pK a values is significantly better because this is the only approach that enables all of the monitored transition curves to be explained (Figure 1). This approach produces two different “major” pK a values for Tyr and Lys, which are in the same range as that of the unperturbed pK a values for Tyr and Lys. Both Y5/K6 and K12/Y16 are solvent exposed with Y5/K6 located in the unstructured N terminus. A small shift from the unperturbed pK a value to lower (Tyr) and higher (Lys) values may be caused by mutual stabilization of the opposite charges. A possible coupling between K12 in each monomer predicted by the structure was not observed experimentally. For the remaining 11 ionizable groups, both approaches give similar results.

Figure 1.

Figure 1

Open in a new tab

Deconvolution of multiple pK a values by using multiple probes for the example of K12 and Y16. A) A section of the reported structure (PDB ID: 3FT7) with K12 and Y16 shown as sticks in red and labeled. K12 13Cδ (B, D), Y16 13Cϵ (C, E), 13Cα (F), and 1Hα (G) shifts are plotted versus pH. Fits are shown as solid lines and pK a values are shown as red dots and labeled. B), C), F), and G) are fitted in a global way with multiple probes (Figure S5) to two pK a values for Y16 (9.14±0.09) and K12 (10.83±0.10). D) and E) are fitted to one pK a value (9.67±0.05) and Hill coefficient (0.62±0.03) for Y16 and for K12 (10.47±0.08/0.57±0.03).

For residues E20/D23/E27, E31/D52, and E46/D49, we likely observe an impact of the remote charge on the intraresidual charge itself. E20, D23, and E27 are located on one side of the first helix (Figure 2 A). D23 is in close proximity to both E20 and E27. All main transitions for all three residues are broadened. Describing them with two macroscopic pK a values results in a significant amplitude for chemical shift changes of the second pK a value. This has to be interpreted not only as an effect on the chemical shift, but also on the charge. Describing them with one pK a value and a Hill coefficient results in a low Hill coefficient, which can be interpreted as a significant influence of one charge by the other. The major pK a value of D23 is significantly reduced compared with the intrinsic value (Figure 2 B, C). This can be explained structurally by a Coulomb interaction with H22. Furthermore, some transitions clearly look not only biphasic, but also like mirror images, which cannot be described by one pK a value and Hill coefficient.

Figure 2.

Figure 2

Open in a new tab

Coupling systems of the acidic residues. A) D23 is located between E20 and E27 (red) and next to H22 (blue) on the first helix; they are labeled and shown as sticks. For D23, the 13Cγ chemical shift (B) is weakly influenced by the charges of E20 and E27, but not by H22. Its impact on the chemical shift can be seen in D23 13Cβ (C). E46 and D49 are located at a close distance on the second helix, whereby E46 is also close to E46 of the other monomer chain (D); they are labeled and shown as sticks. For E46, 13Cδ and 13Cγ chemical shifts (E, F) are broadened as a result of “self‐coupling”, whereas, for D49, E46 has a weak impact on the 13Cγ chemical shift (G). Red points represent the pK a values used as global parameters by global analysis of multiple probes, green points represent known pK a values of neighboring ionizable groups, and orange points represent unknown pK a values that ensure a suitable baseline.

For E31/D52, we likely observe an impact of the remote charge on the inter‐residual charge itself. E31 is located on the first helix facing the second helix, where D52 is located. Both main transitions are broadened and show mirror images. An analysis with two macroscopic pK a values results in a significant amplitude for the second pK a value. Also, an individual fit with the Hill equation produces a similar Hill coefficient of about 0.6. The shift of the major pK a value of D52 to a lower value, relative to the unperturbed value, can be explained by a Coulomb interaction with R24.

E46/D49 are located at a close distance on the second helix (Figure 2 D). E46 is also close to E46 of the other monomer chain. Although the main transition of D49 shows only a small additional transition (see case 2), E46 appears to be broader; this indicates a charge effect across the dimer interface (Figure 2 E–G). Because both titration curves show the opposite of mirror images, coupling between E46 and D49 is improbable. Even though both curves can be described by the same two macroscopic pK a values (2.56 and 4.03), the lower microscopic pK a value of E46 and the major pK a value of D49 (2.56) are not separable by fitting procedures. For example, inspection of the main transition of E46 (both transitions have the same chemical shift difference) gives 2.39 and 3.60 as microscopic pK a values, which correspond to an electrostatic coupling of about 1.6 kcal mol−1. Because E46 and D49 are close to each other in space and chemical shifts are influenced by both transitions, a more detailed analysis was not possible. Through the use of the approach with one pK a value and Hill coefficient, a pK a value of 2.62 is obtained for the D49 transition and 2.97 for the E46 transition. The latter is equivalent to the mean value of the microscopic pK a values (3.00). In both approaches, the pK a value of D49 is significantly reduced compared with its unperturbed value. This can be explained structurally by a positive Coulomb interaction of D49 with K45.

The remaining system (K45/K53/K55/K56) also displays a weak impact of the remote charge on the intraresidual charge itself. K45 is located on the second helix, pointing towards the solvent and the unstructured C terminus, where K53, K55, and K56 are located. All three Lys residues show a mirror image to K53. An approach with two macroscopic pK a values in each case (K45/K53, K53/K55, and K53/K56) describes the titration curves very well, although coupling between K45, K55, and K56 was neglected because of their comparable titration curve shapes. However, some uncertainties remain. Firstly, the distance between charges is not known due to the unstructured nature of this region and the proximity to the solvent weakens Coulomb interactions. Secondly, both the baseline and titration curve shape in the basic pH range is poorly defined. In this case, the pK a value estimation with the Hill equation is a good alternative. The differences between the major macroscopic pK a value and the apparent pK a value from the Hill equation are quite low (from 0.02 to 0.17 for K45 and K53, respectively).

In summary, both approaches (macroscopic model and Hill coefficient) result in very similar pK a values that fall within 0.4 units (0.25 on average), except for E31 and Y16, for which the differences are 0.68 and 0.53 units. In the following sections, we only focus on pK a values derived from the macroscopic model, and consequently, we refer to a major macroscopic pK a value of an ionizable group simply as pK a value.

pH titration curves below pH 1

Nearly all transition curves show an additional transitions below pH 1. 4,4‐Dimethyl‐4‐silapentane‐1‐sulfonic acid (DSS) referencing, which was not necessary for all other pH regions, was unable to correct this effect completely because of possible pH deviation from the protein samples. We attribute this effect to a drastic change of the solvent. Water becomes significantly more protonated, and thus, many counterions (salt) are introduced. This was treated as an additional pK a value in the fit, but ultimately did not interfere with our results in any way because none of the pK a values of the protein were shifted below two.

Experimental pK a values of acidic residues

There are nine acidic residues per monomer in the rhh protein. Four Asp residues (D23, D38, D49, and D52), four Glu residues (E20, E27, E31, and E46) and the C terminus. The pK a values of all four Asp are significantly (one unit or more) shifted to lower values (Figures 3 A–D, S6, 10, 12, and 16, and Table 1) compared with the intrinsic pK a value;14 thus revealing a stabilization of the negatively charged state. This is in agreement with structural findings that they are within 4 Å of a positive charge (H22, R41, K45, and R24), indicating distinct positive Coulomb interactions, at least under low‐salt conditions. In the case of D49, the titration event of the Coulomb partner (K45) can also be seen in the titration curve (Figure 3 A). In contrast, the titration events of the Coulomb partners of D38 and D52 (R41 and R24, respectively) are not covered in the range of the titration curves. The titration event of the Coulomb partner of D23 (H22) is not seen in the titration curve. In contrast to the four Asp, all Glu except E46 (Figures 3 E–H, S6, S10, and S16, and Table 1) and the C terminus (Figures 4 A and S19, and Table 1) are within 0.8 units of their intrinsic pK a values;14 thus showing almost ideal behavior of surface‐exposed residues. Three Glu show negative Coulomb interactions with Asp (E20 and E27 to D23, E31 to D52), but have also positive charges in their close environment. R24 is about 5 Å away from E20, K32 and H51 about 7 Å away from E31, and K30 and R24 less than 10 Å away from E27. These positive Coulomb interactions enable them to energetically couple with Asp, which, in turn, causes these major pK a values near the intrinsic ones. Similarly, E46 has two positive charges (6 Å to H51 and about 9 Å to K32) in its environment, but, in this case, it couples with the E46 residue in the other rhh monomer across the dimerization interface.

Figure 3.

Figure 3

Open in a new tab

Experimental pH titration curves of acidic residues. For D23 (A), D38 (B), D49 (C), and D52 (D), 13COγ chemical shifts are plotted versus pH, whereas for E20 (E), E27 (F), E31 (G), and E46 (H) 13COδ chemical shifts are used. Fits estimated by global analysis of multiple probes (see Figures S6, S10, S12, and S16, and Table S2) are shown as solid lines and results are summarized in Table 1. All measurements were conducted at 25 °C on 1 mm protein in water (low salt, self‐buffering, <60 mm).

Table 1.

Macroscopic pK a values determined and the corresponding difference in chemical shift (Δδ) of the main transition.[a] The major pK a values of coupled residues are underlined.

Residue pK a Δδ (ppm) Residue pK a Δδ (ppm)
D23 2.41±0.04 2.69±0.13 K6 10.78±0.13 4.18±0.39
  3.58±0.05 0.05±0.24   9.66±0.11 0.78±0.45
  4.12±0.04 0.20±0.13 K12 10.83±0.10 3.01±0.16
D38 2.80±0.03 3.73±0.03   9.14±0.08 1.49±0.18
D49 2.56±0.04 2.73±0.10 K30 11.18±0.09 4.69±0.15
  4.03±0.08 0.33±0.09 K32 10.43±0.06 2.75±0.15
D52 2.88±0.03 2.45±0.06 K45 11.84±0.09 6.04±0.38
  4.59±0.04 0.95±0.07   10.38±0.11 0.94±0.29
E20 4.12±0.04 2.84±0.07 K53 10.38±0.11 4.65±0.64
  2.41±0.04 1.08±0.08   11.63±0.11[b] 1.06±0.35
E27 3.58±0.05 3.32±0.16 K55 11.77±0.10 4.35±0.30
  2.41±0.04 1.31±0.18   10.38±0.11 1.12±0.29
E31 4.59±0.04 2.44±0.08 K56 11.44±0.10 4.57±0.46
  2.88±0.03 1.63±0.08   10.38±0.11 0.70±0.57
E46 2.56±0.04 2.93±0.16 R3[a] >13.7  
  4.03±0.08 1.56±0.16 R24[a] >12.5  
C‐term 3.23±0.03 1.80±0.10 R41[a] >13.3  
Y5 9.66±0.11 2.52±0.30 H22 4.79±0.04 0.61±0.02
  10.78±0.13 0.79±0.31 H51 6.94±0.03 1.88±0.01
Y16 9.14±0.08 2.27±0.15 N‐term 7.76±0.04 3.13±0.01
  10.83±0.10 1.18±0.15      
Y47[b] >13.7        
C40[c] >14.9
Open in a new tab

[a] Reported errors are from Monte‐Carlo simulations. Intrinsic pK a values taken from the literature are 3.86 (D), 4.34 (E), 3.55 (C‐terminus), 10.34 (K), 8.23 (N‐terminus), 13.9 (R), 6.45 (H), 9.76 (Y), and 8.49 (C).14 A comparison between the macroscopic model and fit procedures with the Hill equation can be seen in Figure S21. Most important positions for the pK a value determination are listed in Table S2. All included titration curves are given in Figures S2–S19. [b] Estimations have been made by using 50 % of the reported chemical shift differences.14 [c] pK a value is a mean of the major pK a values of K45/K53/K56.

Figure 4.

Figure 4

Open in a new tab

Experimental pH titration curves of the protein termini and His. For the C terminus (A) and N terminus (B) 13Cα chemical shifts are plotted versus pH, while for H22 (C) and H51 (D) 13Cϵ1 chemical shifts are used. The difference in the amplitude (C, D) and the different direction of shifts (E, F) of the histidines arises from the different tautomeric forms. Fits estimated by global analysis of multiple probes (Figures S2, S7, S17, and S19, Table S2) are shown as solid lines and results are summarized in Table 1. All measurements were conducted at 25 °C on 1 mm protein in water (low‐salt, self‐buffering, <60 mm).

Experimental pK a values of basic residues

In contrast, there are 14 basic residues per monomer in the rhh protein. Two His (H22 and H51), eight Lys (K6, K12, K30, K32, K45, K53, K55, and K56), three Arg (R3, R24, and R41), and the N terminus. The N terminus (Figures 4 B and S2, and Table 1), H51 (Figures 4 D, F and S17, and Table 1), and five Lys (K6, K12, K30, K32, and K53; Figures 5 A–D and F, S4, S5, S9, S11, and 18, and Table 1) are within 0.5 (except K30 within 0.8) units to the intrinsic pK a values,14 and even closer to experimental pK a values in apo‐calmodulin.23 All of these groups are solvent exposed. H51 is not solvent exposed, but also does not display any close contacts to other charges. The pK a value of all three Arg residues (R3, R24, and R41; Figures 5 I–K, S3, S8, and S14, and Table 1) are not lower than the intrinsic pK a value,14, 24 although their pK a values are outside the studied pH region, and expected shift changes are low. In case of R41, it is unclear whether the start of the transition is observed or not, due to the proximity to Y47. Despite all of the uncertainties, they describe the lowest possible pK a value accurately. Therefore, pK a must to be close to their intrinsic values or shifted higher, which together indicate favoring of the charged state.

Figure 5.

Figure 5

Open in a new tab

Experimental pH titration curves of basic residues. For K6 (A), K12 (B), K30 (C), K32 (D), K45 (E), K53 (F), K55 (G), and K56 (H), 13Cδ chemical shifts are plotted versus pH, whereas, for R3 (I), R24 (J), and R41 (K), 1Hδ chemical shifts are used. Fits estimated by global analysis of multiple probes (Figures S4, S5, S9, S11, and S18, Table S2) are shown as solid lines and results are summarized in Table 1. All measurements were conducted at 25 °C on 1 mm protein in water (low‐salt, self‐buffering, <60 mm).

The pK a value of H22 (Figures 4 C, E, and S7, and Table 1) is significantly (1.7 units) shifted to lower values, which indicates that the positive charge is unfavorable. Additionally, it is found that more than 50 % are in the uncommon HNδ1 tautomer in its neutral form, which can be evaluated by an increasing 13Cδ2 shift with higher pH, at which the neutral form becomes populated. High 13Cδ2 shifts are very indicative of the HNδ1 tautomer, whereas the shifts are much lower and equal among the HNϵ2 tautomer and the protonated state.25, 26, 27, 28 Long‐range 1H,15N spectra of the His side chain confirm this finding (Figure S7 H, I). Because H22 is not exclusively in one tautomeric neutral form, and no potential hydrogen bonds stand out in the structure, it is likely that the shift in the pK a value is caused by the unfavored effect of the positively charged state and not by stabilizing interactions of one of the neutral tautomeric states. The pK a values of K45, K55, and K56 (Figures 5 E, G, and H, S18, and Table 1) are shifted more than one unit to higher values, which indicates a positive contribution to the protein stability of the positive charge. K45 forms a Coulomb interaction with D49 (within 4 Å distance). K55 and K56 are located in the unstructured C‐terminal region, so a difference from the intrinsic pK a value is unexpected. However, this could be explained by charge compensation with the negative charged C terminus (position 56).

Experimental pK a values of Tyr and Cys

There are three Tyr (Y5, Y16, and Y47) and one Cys (C40) residues per monomer in the rhh protein. They are neutral at acidic and neutral pH, but can be weak acids at high pH. Y5 and Y16 (Figures 6 A, B, S4, and S5, and Table 1) are solvent exposed and are within 0.6 units of their intrinsic pK a values14 and close to usually reported values in proteins.29 In contrast, Y47 and C40 (Figures 6 C, D, S15, and S13, and Table 1) are located in the hydrophobic domain interface and show dramatically shifted pK a values to higher values, which indicates a negative impact of the negative charges. For Y47, the transition starts at pH 12, and thus, is not completed at pH 12.5, whereas, for C40, no signs of a transition could be observed up to pH 12.5. The pK a values could be estimated to be larger than 13.7, assuming unaffected cooperativity and −50 % divergence of the usual amplitude.14 The drastic shift of these pK a values located in the hydrophobic core thus reflects the strength of the hydrophobicity of the domain interface, which is responsible for such a shift.

Figure 6.

Figure 6

Open in a new tab

Experimental pH titration curves of Tyr and Cys. For Y5 (A), Y16 (B), and Y47 (C), 13Cϵ chemical shifts are plotted versus pH, whereas for C40 (D) 13Cβ is used. Fits estimated by global analysis of multiple probes (Figures S4, S5, S13, and S15, and Table S2) are shown as solid lines and the results are summarized in Table 1. All measurements were conducted at 25 °C on 1 mm protein in water (low salt, self‐buffering, <60 mm).

Discussion

Deviations of experimental pK a values from intrinsic values

Sixteen of 27 pK a values are within one unit to the intrinsic values. In these cases, the charged state is neither especially favored or unfavored. In eight cases, the charged state is favored, namely, in D23, D38, D49, D52, and K45, all involved in four distinct positive Coulomb interactions, in E46, and in K55 and K56 in close proximity to the negatively charged C terminus. In three cases, the charged state is unfavored, namely, in H22, C40, and Y47. H22 is involved in hydrophobic contacts; C40 and Y47 are located in the hydrophobic domain interface.

We calculated pK a values from the crystal structure (PDB ID: 3FT7) using H++,30, 31 ROSIE,32, 33 and DEPTH,34 to see how much these shifts in pK a values could be reproduced from calculations (Table S3). H++ was able to capture the shift in pK a values of C40 and Y47 to higher values (>12). ROSIE failed on Y47 (9.4; and does not offer a calculation on C40), whereas DEPTH did not offer a calculation for C40 and Y47. Both H++ and ROSIE capture the shift in pK a values of D23, D38, and D49 to lower values, but do not provide a close match of pK a values (<0.5 units), nor the correct ranking. DEPTH only qualitatively reproduces D38, but fails on D23 and D49. ROSIE captures the shift in pK a value of K45 to a higher value (and the trends in other Lys), H++ predicts the pK a value of K45 accurately (but failed on K32). DEPTH did not reproduce pK a values in Lys. All programs failed to capture the shift in pK a value of H22 to a lower value. D52, K55, and K56 are not (well resolved) in the crystal structure. E46 was a complicated case with a charge effect across the dimer interface and was therefore excluded from the calculations. In summary, H++ was able to reproduce most shifts in pK a values qualitatively, but failed on H22.

Salt bridges

All four Asp (D23, D38, D49, and D52) are involved in distinct positive Coulomb interactions, which could be interpreted as salt bridges. They have a positive charge (H22, R41, K45, and R24) within 4 Å (Figures 7 and S22) and their pK a values are shifted to lower values (Figure 3 A–D and Table 1); thus experimentally demonstrating a stabilization of the negative charge and a direct contribution to stability under low‐salt conditions and low pH. For two salt bridges (D23–H22 and D49–K45), the distance is significantly shorter (about 3.5 Å); this indicates potentially stronger Coulomb interactions. This effect is also seen from further shifted pK a values of the two Asp involved (2.41 and 2.56, compared with 2.80 and 2.88, respectively); thus confirming two stronger and two weaker Coulomb interactions in solution. Three of the Coulomb pairs (D23‐H22, D38‐R41, and D49–K45) are intra‐monomer; the fourth (D52–R24) is inter‐monomer. On the side of the positively charged state, the results are not as uniform. The increased pK a value (Figure 5 E and Table 1) of K45 (salt bridge to D49) reports a stabilization of the positive charge and indicates a positive Coulomb interaction. Thus, this Coulomb interaction (D49–K45) contributes to the protein stability over a large pH range. For R24 and R41, no change in pK a values could be observed, although neither transition was well captured by experiments (Figure 5 J, K and Table 1). It is not clear if the unchanged pK a values are a result of experimental uncertainties caused by their high pK a value (out of the pH window of protein stability) or if the Coulomb interactions (D38–R41, D52–R24) simply contribute more to protein stability at low pH (at which a shift of the pK a value is observed) than that at high pH (at which there might not be a shift of the pK a value). In the fourth Coulomb interaction (D23–H22), the pK a value of H22 (Figure 4 C, E and Table 1) is significantly shifted to a lower value; thus reporting a preference of the uncharged state over the positively charged state, which is participating in the Coulomb interaction. The Coulomb interaction therefore is only favorable and possibly contributes to protein stability at low pH, although it is one of the two potentially stronger interactions (in terms of shorter distance and further shifted pK a value of D23). In the process of losing the four positive Coulomb interactions by pH titration at around pH 2.5, the protein loses significant thermodynamic stability (Figure 8), which is reflected in a decrease of melting temperature from 97 °C at pH 4 to 61.7 °C at pH 2.18 Because the protein is a two‐state folder and no changes in the cooperativity of the unfolding transitions are observed, transition midpoints are proportional to the free energy of unfolding.18 Despite being destabilized, the protein remains folded, even at pH 0.

Figure 7.

Figure 7

Open in a new tab

Ribbon representation of the reported structure (PDB ID: 3FT7) in two different orientations (A, B). Negatively charged residues involved in positive Coulomb interactions (D23, D38, D49, and D52) are shown as sticks, labeled and colored in red; positively charged residues are shown as sticks, labeled and colored in blue (R24, R41, and K45) and cyan (H22). Neutral residues with pK a values ≥14 (C40 and Y47) are shown as sticks, labeled and colored in green.

Figure 8.

Figure 8

Open in a new tab

Dependence of the thermodynamic stability, ΔG, and melting temperature, t m, on the pH value. ΔG (blue line) was calculated by using the major macroscopic pK a values and the reported intrinsic pK a values as an approximation for the unfolded state.12, 13 It should be noted that the melting temperature (red points) was determined with 5 μm ORF56 and 20 mm sodium phosphate and 20 mm sodium citrate as buffer.17 The gray line marks a temperature of 298 K and ΔG=0 kJ mol−1.

pK a values of buried Tyr and Cys are shifted to extremely high values

Cysteines and tyrosines are usually not charged at regular pH and prefer the uncharged state. In C40 and Y47, both located in the hydrophobic core/domain interface, the pK a values are shifted more than 6 and 3.5 units to higher values, which implies an additional disfavor of the charged state. The energy necessary for this must be provided from the hydrophobic domain interface.

This confirms the importance of the hydrophobicity of the domain interface for protein stability. This was concluded earlier on a structural comparison of different domain interfaces of rhh proteins with different stability.17 Here, the experimentally found shifted pK a values of C40 and Y47 correspond to an energy, which has to be provided by the domain interface, of more than 57 kJ mol−1 (RT(ln 103⋅9+ln 106⋅1)).12 This is a significant portion of 85 kJ mol−1[18] total stability per monomer. Also, the rhh protein is stable down to pH 0, but unfolds over pH 12, at which the deprotonation of Y47 also occurs (Figure 8). All other charge transitions are at least 90 % completed at this pH, with the exception of the 3 Arg. However, the pK a values are close to the intrinsic pK a value of Arg14 or shifted to higher values. This leaves Y47 (and C40) with pK a values of around 14 as potential reasons for the loss of stability. They are also drastically shifted compared with their intrinsic values.14 In fact, with pK a values of >14 for C40 and >13.7 for Y47, these values are significantly higher than those of the highest reported values for Cys (11.1) and Tyr (12.1).35 Taken together, the optimized hydrophobic domain interface/hydrophobic core is the main reason for hyper‐thermostability of the rhh protein from plasmid pRN1 of S. islandicus.18

Conclusion

We have determined (or estimated a lower value) all pK a values in a protein following transitions on multiple probes over a wide range of pH. Combining the results of multiple chemical shifts in a global analysis, we were able to decompose complex and broad transition curves. In four Asp (D23, D38, D49 and D52) and one Lys (K45), all involved in the four positive Coulomb interactions (salt bridges), we found significant shifts in the pK a values in favor of the charged state. Additionally, a stabilized charged state was detected in K55 and K56, which were in close proximity to the C terminus. In contrast, H22, C40, and Y47 disfavored the charged state. In particular, in C40 and Y47, both located in the hydrophobic core, the pK a values were shifted to 14 or beyond.

Experimental Section

Protein samples: The rhh protein (UniProtKB Q54323) was expressed in M9 minimal media, by using 1.5 g L−1 15N NH4Cl and 2 g L−1 13C6 glucose, and purified as described previously,36 with the modification of cell lysis directly by the heat step. It was concentrated to 1 mm. The pH was adjusted with NaOH or HCl directly on each sample in water (10 % D2O) and checked right before and after measurement in the NMR tube. pH values were measured at 22 °C by using an inoLab pH 720 pH meter with a Hamilton Spintrode pH electrode. Between pH 2 and 12, the total salt concentration (counterions of protein, H+, and OH−, NaN3) was below 60 mm.

NMR spectroscopy: All experiments were performed on a Bruker DRX 500 NMR spectrometer at a static magnetic field strength of 11.7 T at 23 °C. 1H13C constant‐time HSQC experiments optimized for aliphatic and aromatic regions and a Hx2CO experiment for Asp and Glu37, 38 were recorded at various pH values. Water suppression was achieved through gradient selection for the HSQC experiments and WATERGATE for the Hx2CO experiment. Assignments were taken from ref. 17 and confirmed through HNCACB, H(CCO)NH, (H)C(CO)NH and aromatic (HB)CB(COCD)HD, and TOCSY experiments. All spectra were processed with NMRPipe39 and analyzed with NMRView.40 The spectra were not referenced with DSS. External DSS referencing revealed a 1H chemical shift of δ=−0.104 ppm.

Data analysis: Only measurement points with a difference of pH value before and after measurement lower than 0.1 units, which did not show long‐term variations in the NMR spectra (e.g., degradation), were included in the analysis. pH titration curves monitored by multiple chemical shifts (Figures S2–S19 and Table S2) were fitted with nonlinear least‐squares regression analysis (Levenberg–Marquardt algorithm) to standard Hendersson–Hasselbalch equations for a variable number of pK a values in a global way by using MATLAB.. Convergence to the global minimum was ensured through visual inspection of the fitted curves and that the estimated results agreed with the protein structure (coupled ionizable groups shared common pK a values and all titration curves were explainable through ionizable groups at a short distance). Altogether, more than 140 different shifts were analyzed and, excluding the extrapolated pK a values, on average, 10 chemical shifts were fitted with a global pK a value. For neighboring residues with a pK a value in a comparable range (Y5/K6, K12/Y16, E20/D23/E27, E31/D52, E46/D49, and K45/K53/K55/K56), at which the transitions showed coupled behavior (mirror images), the pK a values were fitted simultaneously with the macroscopic model.21, 22 In the case of E20/D23/E27, the system was not fitted to three macroscopic pK a values in a direct way, but to two times two macroscopic pK a values for which D23 served as the interface. The K45/K53/K55/K56 system was treated similarly. In the case of known transitions in the chemical shift, this further information was added for the global pK a value determination (with an uncertainty of 0.3 for D, E, H, and the termini, and 0.9 for Y and K considered for error estimation). In unclear cases, local variable pK a values were added to ensure a suitable baseline. For transitions not fully covered by experimental pH values, assumptions of the chemical shift changes have been made according to ref. 14 by using −50 % of standard chemical shift differences caused by the titration. Effects on the chemical shift at low pH have been fitted to Hendersson–Hasselbalch curves by using pK a values of <0. This reflected the water hydronium transition or salt concentration effects. For error estimation, Monte‐Carlo simulations, with 1000 steps and a random variation of pH (±0.1), as suggested and discussed in ref. 41 and 1H (±0.01 ppm) and 13C (±0.02 ppm) chemical shifts were executed. The latter was determined by analyzing the chemical shifts without any transitions. Please note that possible systematic errors might not be captured by the data and error analysis, so the true error might be higher. The estimated pK a values refer to 10 % D2O.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

Click here for additional data file. (5.7MB, pdf)

Acknowledgements

We thank Kathrin Waldheim for excellent technical assistance. This research was supported by the Deutsche Forschungsgemeinschaft (WE 5587/1‐1).

H. N. Raum, U. Weininger, ChemBioChem 2019, 20, 922.

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

Click here for additional data file. (5.7MB, pdf)

Articles from Chembiochem are provided here courtesy of Wiley

RESOURCES