Nuclear Magnetic Resonance Titration Curves of
Histidine Ring Protons
III*. Ribonuclease

ALAN N. SCHECHTER, DAVID H. SACHS,

Laboratory of Chemical Biology

National Institute of Arthritis and Metabolic Diseases

STEPHEN R. HELLER, RICHARD I. SHRAGER AND JACK S. COHEN

Physical Sciences Laboratory

Division of Computer Research and Technology

National Institutes of Health

Bethesda, MD 20014, USA.

(Received 24 March 1972, and in revised form 5 July 1972)

The nuclear magnetic resonance titration curves of chemical shift versus pH of two of the histidine ring C (2) protons of ribonuclease show deviations from the curve expected for a simple proton association equilibrium. We have previously shown (Sachs, Schechter & Cohen, 1971), using model systems, that titrating groups in spatial proximity to an imidazole ring give rise to such effects and that these data can be analyzed by computer curve-fitting to give microscopic and macroscopic apparent ionization constants of the imidazole ring and the interacting groups. When applied to the curve with the greater deviation, such an analysis shows that the histidine residue giving rise to this resonance has a macroscopic apparent pK of 6.I and is interacting with an adjacent group of pK 4.6. This resonance has previously been assigned to histidine residue 119 and, on the basis of information X-ray crystallographic studies, we believe its interacting group is the carboxyl function of aspartic acid residue 121. The curve probably corresponding to histidine residue 12 has a macroscopic apparent pK of about 6.2 but displays small acid and alkaline inflections. These result from the interactions with groups of pK 5 and 8.4. The alkaline group is possibly the e-amino function of lysine residue 41, on the basis of the know three-dimensional structure. This method of analysis may be a general one for measuring interactions of charged groups with histidine residues in proteins.

1. Introduction

High resolution proton magnetic resonance spectroscopy can provide detailed information about the environment of histidine residues within a protein molecule, because the imidazole ring C(2) protons can be resolved from the resonances of other protons as a result of their unique chemical environment (Roberts & Jardetzky, 1970). These protons can also be resolved from each other when there are several histidine residues in a protein as a result of the fact that the three-dimensional-folding of the protein produces different local environments. Further the chemical

*Paper II in this series is Shrager, Cohen, Heller, Sachs & Schechter (1972). A preliminary account of this work was presented in Cohen, Schechter, Sharager, Sachs & Heller (1971).

shift values of these resonances change continuously with the degree of protonation of the imidazole rings, so that the individual ionization constants of these residues may be determined from their p.m.r.* titration curves (Brandbury & Scheraga, 1966; Meadows, Markley, Cohen & Jardetzky, 1967).

The p.m.r. titration curves of the four imidazole C(2) proton resonances of bovine pancreatic ribonuclease have been described, and an assignment of each resonance to a corresponding histidine residue in the amino-acid sequence has been made (Meadows, Jardetzky, Epand, Ruterjans & Scheraga, 1968). Recently, several groups have noted that the titration curves of two of these imidazole C(2) proton resonances,, assigned to the active-site histidine residues 12 and 119, show asymmetries in the low pH region, and thus cannot be described by a simple proton association equilibrium (Ruterjans & Witzel, 1969; Cohen, Shrager, McNeel & Schechter, 1979; King & Bradbury, 1971). This phenomenon has been interpreted as reflecting an interaction between these two histidine residues in the active site (Ruterjans & Witzel, 1969; King & Bradbury, 1971).

We present here an alternative explanation of these findings which results from an analysis of the interaction of imidazole rings with other titrating groups, as measured by p.m.r. spectroscopy. This analysis is based on a study of the effect of the protonation of the amino and carboxyl groups of low molecular weight imidazole derivatives on the titration curves of the imidazole protons (Sachs, Schechter & Cohen, 1971), and on mathematical models to determine the microscopic apparent pK values of these compounds from these curves (Shrager, Cohen, Heller, Sachs & Schechter, 1972).

Our interpretation of the titration curves of these two histidine residues postulates the interactions of these imidazole rings with carboxyl and amino groups on the side chains of neighboring amino-acid residues in the three-dimensional conformation of the protein. It is possible that this method of analysis will provide a general means of studying in detail the environments around histidine residues in proteins and of determining the pK values of interacting charged groups.

2. Materials and Methods

Bovine pancreatic ribonuclease was purchased from Worthington Biochemical Corp. (Rutherford, NJ) as a phosphate-free powder. Concentrated solutions were dialyzed extensively against deionized water, lyophilized 3 times from 99.7% deuterium oxide (Aldrich Chemical Co., Milwaukee, WI) and then once from 100% deuterium oxide (Aldrich). Chemical analysis of these samples, after experimental studies, showed no detectable phosphate or sulfate. Solutions were made to 100 mg/ml. with 100% deuterium oxide and contained 0.1 M-NaC1 and NaOD (Merck, Canada) solutions were used for pH adjustments.

Spectra were recorded on a Varian Associates HR 220 nuclear magnetic resonance spectrometer. A Varian C1024 time-averaging computer accumulated spectra for periods between 1 to 4 hr at each pH value. All spectra were recorded at the probe temperature of 22^oC. Standardization of chemical-shift values was relative to external tetramethylsilane in carbon tetrachloride solution.

Measurements of pH were made with a Radiometer no. 26 meter (Copenhagen) with a long thin combination electrode (Instrumentation Labs., Lexington, Mass.) that fitted into the nuclear magnetic resonance tube. pH readings were made before and after each spectrum was recorded; values generally differed by less than 0.03 pH

unit. The final value was used routinely and is quoted as an uncorrected "pH meter reading." The values are not

*Abbreviation used: p.m.r., proton magnetic resonance.

strictly pD of pH measurements but are closer to the latter, probably differing by about 0.1 pH unit (Roberts, Meadows & Jardetzky, 1969; Sachs et al., 1971).

The data for chemical-shift values as a function of pH meter readings were analyzed by computer curve-fitting using the MODELAIDE program by methods we have described (Shrager, 1970) using several mathematical models developed for interacting charged groups (Shrager et al., 1972). In accordance with this published analysis, we describe the titration of two charged interacting groups, A and B, in the same molecule by the following

equilibrium, where, C₀₀ = concentration of species with both A and B unprotonated; C₁₀ = concentration with only A protonated; C₀₁ = concentration with only B protonated, and C₁₁ = concentration with both A and B protonated. Each species has a corresponding chemical shift:

Since there is fast exchange of protons the observed chemical shift at any pH is a weighted average of the chemical-shift values of all of these species,

i.e.

where C = total concentration of all species.

Substitution for the concentration terms provides an equation in terms of the microscopic pK values, pK_A0, pK_B0, etc., which may be evaluated by computer curve-fitting to the chemical-shift versus pH data. The microscopic pK values, pK_A0, pK_B0, etc. which may be evaluated by computer curve-fitting to the chemical-shift versus pH data. The macroscopic pK values, pK_A and pK_B, may then be derived (Edsall, Martin & Hollingworth, 1958). Several mathematical models for approximating these sets of equations were delineated in the previous paper (Shrager et al., 1972) and these will be referred to in this paper by the same numbering system.

3. Results

Figure 1 shows chemical shift values at 220 MHZ as a function of pH for three of the four histidine residues C(2) protons of ribonuclease. These data are from two experiments involving 24 and 25 time-averaged spectra. The fourth histidine C(2) proton resonance is clearly seen at about 8.0 p.p.m. below pH 5 and at about 7-6 p.p.m. above pH 7. Its behavior in the intermediate pH region is complex and will be discussed in a further communication.

The establishment of the continuity of lines in titration curves when there are two or more intersecting lines may cause difficulty. We are reasonably certain of the continuities indicated because of the following aids to the identification of resonances: (a) the line-widths of the several resonances differ, (b) perturbants such as phosphate ions and substrate analogs affect resonances H-2 and H-3 (differentially) but hardly affect H-1 (Meadows, Roberts & Jardetzky, 1969; and unpublished results), thereby allowing the points for H-1 to be identified distinctly in a given titration and aiding in separating H-2 from H-3, and © the results of curve-fitting of various mathematical models (see below) to alternative continuities. The continuities indicated by hand-drawn lines in Figure 1 are also in accord with those used by the other workers (Meadows et al., 1968; Ruterjans & Witzel, 1969; King & Bradbury, 1971) who have reported p.m.r. data on ribonuclease at 100 MHZ.

Inspection of the lines in Figure 1 indicates the following. The curve of resonance H-1 follows the symmetric function expected for a simple proton association equilibrium. The slope of the curve for resonance H-2 changes more gradually than curve 1 in the region between pH 4 and pH 5, and between pH 7 and 8. The curve for resonance H-3 changes much more gradually between pH 3 and 7 than would be expected for a simple proton association equilibrium.

The physical bases of the asymmetric shapes of these curves may be tested by the application of curve-fitting of the data to mathematical models for titrating groups as previously presented. The results obtained from such curve-fits are shown in Table 1. Figure 2(a) shows the data for curve H-1 from Figure 1 fitted to the simple proton association equilibrium (model 1, Shrager et al., 1972). This close fit is well within the experimental variance of individual points (approx. +0.02 p.p.m.). In all experiments, including those with various perturbants, curve H-1 varied little in position, with apparent pK of 6-7, and could always be described by the Henderson-Hasselbalch function.

The data for curve H-3 from this titration experiment are poorly represented by the equation for the simple proton association equilibrium (Fig. 2(b)). The computer generated fit has significant and consistent deviations from the experimental points over most of the range of pH values. The greatest deviation is in the region between pH 3 and 5, where the slope is more gradual than that expected for the simple proton association equilibrium. A better fit to the data is obtained by use of mathematical model 2 (Fig. 2(b), which includes interaction with another titrating group. The improved quality of fit now may be judged by inspection and by the reduced value of the sum of squares of the errors (Table 1). The microscopic apparent pK values derived from this model, as well as the calculated macroscopic apparent pK values, are also listed in Table 1. The macroscopic apparent pK for the histidine residue (pK_A) is 6-05, while for the interacting group (pK_B) is 4-52. Curve-fitting with assumed chemical shift values for the intermediate forms (8_A10) of 5 and 10% compared to the over-all chemicals shift change (8_A11- 8_A00) (model 3), does not significantly change the values of the microscopic or macroscopic apparent pK values. This finding is similar to the results for histidine methyl ester (Shrager et al., 1972).

The deviation of the data for curve H-2 form the simple proton association equilibrium is smaller than that for H-3 but significantly greater than that for H-1 (Fig. 2© ). In this case, there are deviations both between pH 4 and 6 and between pH 7 and 8, the latter being larger in magnitude. Model 1 gives an over-all apparent pK of 6-18 (Table 1). The data have been fitted to model 2 in two steps, with emphasis on either the acid or alkaline perturbations. In the first step, the pK_A for the H-2 histidine is 6-32 while its interacting group has a pK_B of 4-94. These values are sensitive, however to the chosen value of intermediate chemical shift (model 3) and vary from 6-3 to 6-4 for pK_B. This uncertainty is not surprising in view of the small magnitude of the acid perturbation. The alkaline perturbation is less sensitive to this variable and values of pK of 6-10 and pK of 8-36 are obtained from the application of model 2 or 3 in the second step. We may conclude that the histidine residue H-2 has a pK of between 6-1 and 6-3 (probably closer to the latter value) and that it is interacting with two other titrating groups which have pK values of about 8-4 and 5-0.

The data for curves H-2 and H-3 have also been fitted simultaneously to a mathematical model (model 4) for mutual interaction of the two titrating imidazole groups (Fig. 3, Table 1). Inspection of the results and comparisons of the sum of squares of errors show that the quality of the fits of this model are much poorer than those of the above methods of analysis. The introduction into the model, with mutual interactions, of possible values for the intermediate chemical-shift values (model 5) does not significantly change either the quality of the fit or the computed pK values. Another test of the physical applicability of the model for mutually interacting groups is the degree of consistency of the derived apparent pK values (6-2 and 4-8) with those from the separate fits of H-2 and H-3 (6-2 (average of the two fits) and 6-05). Thus these results are also not in accord with the concept that the group interacting with each of these histidine residues is the other histidine residue.

4. Discussion

In previous papers we have shown that detailed measurements of changes in chemical-shift values as a function of pH can be used to make inferences about the electronic and magnetic environments around imidazole protons (Sachs et al., 1971). We have developed mathematical models to describe the effects of adjacent charged groups on p.m.r. titration curves and have found that the number of parameters needed to describe these effects may be decreased when certain assumptions are made. Results of microscopic and macroscopic pK values which are in agreement with titrimetric values have been obtained in the analysis of the titration curves of these histidine compounds by these mathematical models (Shrager et al., 1972). Asymmetries in the p.m.r. titration curves were observed for several histidine model compounds and in each case the pK values of the imidazole groups, as well as of those amino and carboxyl groups interacting with them, were accurately determined by these methods. We have therefore analyzed the titration curves of ribonuclease A using the same methods.

Our computations show that the H-2 resonance has a macroscopic apparent pK between 6-1 and 6-3 and is interacting significantly with a group of apparent pK of approximately 5-0. The H-3 resonance, with a macroscopic apparent pK of 6-05, is interacting significantly with a group of apparent pK 4-5. The group with apparent pK 8-4 is most likely to be a lysine residue, although -amino or anomalous tyrosine residues cannot be excluded. The acid titrating groups are most likely to be aspartic or glutamic acid residues.* Computer curve-fitting to models assuming mutual histidine-histidine interaction indicate that such interaction is an unlikely explanation of our results.

The acid asymmetries in the titration curves of histidine resonances H-2 and H-3 of ribonuclease. A have now been described by three groups (Ruterjans & Witzel, 1969; Cohen et al., 1970; and King & Bradbury, 1971) using a variety of enzyme preparations and spectrometers. The alkaline asymmetry in H-2 is small in magnitude, but can be noted by inspection of the data published by Ruterjans & Witzel (1969) and by King & Bradbury (1971) as well as our own. These other workers interpreted the deviations at acid pH as being consistent with a mutual interaction between the two histidine residues in the active site of ribonuclease A. While this is not indicated by our analyses, it should be noted that, as with all fitting of data to mathematical models, the results can never prove the validity of a model but can only be used to decrease or increase the probability of certain interpretations relative to others. An important test is that the results be consistent with other information about the protein being studied.

Using the assignments of H-2 as histidine residue 12 and H-3 as histidine residue 119, from the work of Meadows et al., (1968), we may correlate these estimations of pK values and identification of residues with the structure determined by X-ray crystallography of ribonuclease S (Wyckoff et al., 1967, 1970). Aspartic acid residue 121 is about 5 A from the position of histidine residue 119 with a closest approach of the carboxyl oxygen of about 3-3 A. Lysine residue 41 is about 10 A from the position of histidine residue 12 with a closest approach of about 8-5 A. The separation of aspartic acid residue 121 and histidine residue 12 is also about 10 A. The two imidazole rings have a closest approach of about 7-5 A. Although these distances, as indicated in the crystal, may vary for the protein in solution, they may nevertheless be taken as an indication of the relative juxtapositions of these groups in the protein.**

Thus, the residue giving rise to resonance H-3, histidine residue 119, might well be expected to be interacting with a group of apparent pK about 4-6, aspartic acid residue 121. Histidine residue 12, resonance H-2, might also be expected to be sensitive to the protonation of lysine 41 (apparent pK about 8-4). Histidine residue 12 is also affected by an acidic group which might be aspartic acid residue 121 or 9, or glutamic acid residue 83, or some combination of these. Interactions of this type have also been deduced by Loeb & Saroff (1964) on the basis of ion-binding studies. Secondary effects, including histidine 12-histidine 119, histidine 119-lysine 41 and other interactions may also occur but our data indicate that such effects, if present, are too small in magnitude to perturb the histidine titration curves significantly.

The values of the macroscopic apparent pK values of the three histidine resonances are in reasonable agreement with those obtained in the other published studies (Table 2). The differences are, in general, in the direction that one would expect from the drawing of one Henderson-Hasselbalch curve through more complex curves or using other approximations to analyze these data.

The assignments of Meadows et al. (1968) are based on several assumptions and are not a unique interpretation of their results (King & Bradbury, 1971). However, even if two or more of the correlations of resonance and residue numbers are changed, the measured apparent pK values of the interacting groups still indicate that acidic and basic functions are interacting with al least two of the histidine residues of ribonuclease A. The interaction of histidine residue 48 with aspartic acid residue 14 and/or tyrosine residue 25 (Wyckoff et al., 1967, 1970) for example, would also be consistent with our analysis of the p.m.r. titration curves.

*The extend of this inflection is greatly increased on the addition of phosphate ion and nucleotide inhibitors. While we cannot rule out the presence of a small amount of bound phosphate in our preparations, although the accuracy of the phosphate analysis is 3 p.p.m., it is quite clear that the inflection could not be a direct result of the presence of phosphate. Thus, the phosphoric acid pK values are 2-2 and 7-2 and the interaction with a positively charged imidazole group would be expected, if anything, to lower these values. Further, an inflection is also seen in the presence of sulphate ion, which has no pK value above 1-9.

**Although we have calculated these distances based on the co-ordinates of ribonuclease S (Wyckoff et al., 1967), we have been informed that the high resolution model of ribonuclease A (Kartha, personal communication) shows a similar arrangement for residues in the active site, with Asp-121 being almost in contact with His-119. Conversely, we have found similar deviations from the simple proton association equilibrium in the p.m.r. titration curves of the histidine ring C(2) protons of ribonuclease S (Griffin, Schechter & Cohen, unpublished results).

It is important to compare these perturbations with results that have been observed by other forms of spectroscopy. At 22^oC ribonuclease is extremely stable to changes in pH and most studies by other techniques have not detected structural transitions until pH 2 or lower (Simons & Blout, 1968; Richards & Wyckoff, 1971). However, Donovan (1965) has noted small perturbations in the ultraviolet difference absorption spectra between pH 2 and 5, which he has interpreted as charge perturbation of neighboring tyrosine and phenylalanine residues by imidazole groups or the -amino group. The changes we have observed and those observed by Donovan (1965) might be interpreted implicitly as due to "conformational changes" in the protein during titration with protons.

However, by the p.m.r. technique we can now identify explicitly the specific protonation events that are reflected as electromagnetic perturbations in the molecule. The perturbations of chemical-shift values reflected in asymmetric titration curves may be considered to reflect both the changes in magnetic and electric fields which occur for a reporter resonance when a nearby group changes its state of ionization. The intermediate chemical-shift values may be considered an index of the electromagnetic environment, while the changes in microscopic pK values probably reflect more the charge difference. Examination of the computed values for these parameters indicates that both factors have significant effects. For example, the extra down field shift of the protonated form of the histidine residue giving rise to curve H-3 (Fig. 1) may be interpreted as deshielding by the neighboring carboxyl group (King & Bradbury, 1971). Experimentally, however, the changes in the microscopic apparent pK values appear to make the more significant contributions to the shapes of the individual titration curves. Phenomenologically, this means that the ease of protonation of each histidine residue is sensitive to a change in the state of ionization of the neighboring groups.

The mechanism of action of ribonuclease has recently been reviewed in detail (Richards & Wyckoff, 1971). Two possible roles for the two active site histidine residues are still widely discussed: (a) the histidine residues being on opposite sides of bound substrate so as to act in a "push-pull" way with regard to proton transfer or (b) the histidine residues being adjacent and acting in concert. The absence of evidence for a histidine-histidine interaction in our analysis would tend to support a mechanism of the former type.

We thank Dr. H. G. Wyckoff for the three-dimensional co-ordinates of ribonuclease S and Dr. G. Kartha for information about the structure of ribonuclease A. We are grateful to Mr. R. Bradley for aid with the spectrometer.

TABLE 1

Ionization constants for ribonuclease

* For definition of these models see Shrager et al. (1972).

* This represents the value of the chemical shift of the intermediate (8_A11-8_A00).

TABLE 2

Comparison of apparent pK values for three of the histidine resonances of ribonuclease A

as determined by p.m.r.

* 0-4 substracted from pH values.

* pK in both studies appears to be defined as the midpoint of the total chemical shift change, including the region of observed asymmetry.

* The large dependence of the pK of H-2 and H-3 on ionic strength is interpreted in termsneighboringvoring positively charge group, lysine residue 41.

* The "edown fieldnfield shift in H-3 below pH 4 is interpreted as indicating a neighboring carboxyl group, aspartic acid residue 121.

REFERENCES

Bradbury, J. H. & Scheraga, H. A. (1966). J. Amer. Chem. Soc. 88, 4240.

Cohen, J. S., Schechter, A. N., Shrager, R., Sachs, D. H. & Heller, S. R. (1971). Fed. Proc., Fed. Amer. Soc. Exp. Biol. 30, 1293A.

Cohen, J. S., Shrager, R. I., McNeel, M. & Schechter, A. N. (1970). Biochem, Biophy. Res. Comm. 40, 144.

Donavan, J. W. (1965). Biochemistry, 4, 823.

Edsall, J. R., Martin, R. B. & Hollingworth, B. R. (1958). Proc. Nat. Acad. Sci., Wash. 44, 505.

King, N. L. R. & Bradbury, J. H. (1971). Nature, 229, 404.

Loeb, G. I. & Saroff, H. A. (1964), Biochemistry, 3, 1819.

Meadows, D. H., Jardetzky, O., Epand, R. M., Ruterjans, H. H. & Scheraga, H. A. (1968), Proc. Nat. Acad. Sci., Wash. 60, 766.

Meadows, D. H., Markley, J., Cohen, J. S. & Jardetzky, O. (1967). Proc. Nat. Acad. Sci., Wash. 58, 1307.

Meadows, D. H., Roberts, G. C. K. & Jardetsky, O. (1969). J. Mol. Biol. 45, 491.

Richards, F. M. & Wyckoff, H. W. (1971). In The Enzymes, ed. by P. D. Boyer, 2nd edit., vol. 4, p. 647. Academic Press: New York.

Roberts, G. C. K. & Jardetzky, O. (1970). Adv. Prot. Chem. 24, 448.

Roberts, G. C. K., Meadows, D. H. & Jardetzky, O. (1969). Biochemistry, 8, 2053.

Ruterjans, H. & Witzel, H. (1969). Europ. J. Biochem. 9, 118.

Sachs, D. H., Schechter, A. N. & Cohen, J. S. (1971). J. Biol. Chem. 246, 6576.

Shrager, R. (1970). J. Ass. Comp. Mach. 17, 446.

Shrager, R., Cohen, J. S., Heller, S. R., Sachs, D. H. & Schechter, A. N. (1972). Biochemistry, 11, 541.

Simons, E. R. & Blout, E. R. (1968). J. Biol. Chem. 243, 218.

Wyckoff, H. W., Hardman, K. D., Allewell, N. M., Inagami, T., Johnson, L. N. & Richards, F. M. (1967). J. Biol. Chem. 242, 3985.

Wyckoff, H. W., Tsernoglou, D., Hanson, A. W., Knox, J. K., Lee, B. & Richards, F. M. (1970). J. Biol. Chem. 245, 305.