Hydrogen Ion Interactions of Horse Spleen Ferritin and Apoferritin”

The interactions of horse spleen ferritin and its derivative apoferritin with H+ ions were studied by potentiometric and spectrophotometric titration; to aid in data analysis, heats of ionization over a limited pH range and amide content were also determined. Per apoferritin subunit, all tyrosine and cysteine side chains, two of the nine lysine side chains and at least three of the six histidine side chains were found not to titrate; a preliminary but self-consistent analysis of the titration data is proposed. The titration curve of ferritin was identical with that of apoferritin in the pH range 5.5 to 3. In addition, under the conditions used, the reactivities of ferritin histidines to bromoacetate and of ferritin lysines to formaldehyde were identical with those in apoferritin. Above pH 8, a time-dependent titration of the ferritin core occurs which prevents comparison of the titration curves of the two proteins in this region. However, in the pH region 5.5 to 7.5, two extra groups per subunit titrate reversibly in ferritin relative to apoferritin. Moreover, although the isoionic points of ferritin and apoferritin are identical in water, the isoionic point of ferritin is 0.5 pH unit lower than that of apoferritin in 0.16 to 1 M KCl. The different effects of KC1 and NaCl

The interactions of horse spleen ferritin and its derivative apoferritin with H+ ions were studied by potentiometric and spectrophotometric titration; to aid in data analysis, heats of ionization over a limited pH range and amide content were also determined. Per apoferritin subunit, all tyrosine and cysteine side chains, two of the nine lysine side chains and at least three of the six histidine side chains were found not to titrate; a preliminary but self-consistent analysis of the titration data is proposed. The titration curve of ferritin was identical with that of apoferritin in the pH range 5.5 to 3. In addition, under the conditions used, the reactivities of ferritin histidines to bromoacetate and of ferritin lysines to formaldehyde were identical with those in apoferritin. Above pH 8, a time-dependent titration of the ferritin core occurs which prevents comparison of the titration curves of the two proteins in this region. However, in the pH region 5.5 to 7.5, two extra groups per subunit titrate reversibly in ferritin relative to apoferritin. Moreover, although the isoionic points of ferritin and apoferritin are identical in water, the isoionic point of ferritin is 0.5 pH unit lower than that of apoferritin in 0.16 to 1 M KCl. The different effects of KC1 and NaCl on the two proteins indicate the presence of cation binding sites in ferritin that are absent in apoferritin and possibly also the presence of anion binding sites in apoferritin that are occupied in ferritin by anions of the core. The difference between the isoionic points of the two proteins in KC1 has been interpreted to indicate the presence of approximately 2 phosphate residues per ferritin subunit which serve as cation binding sites and which are negatively charged at the isoionic point in KCl. These phosphates may also represent the additional residues that titrate in ferritin between pH 5.5 and 7.5, or may interact with positively charged residues on the inner surface of the ferritin shell, or both.
The iron storage protein ferritin consists of a spherical protein shell containing 24 subunits surrounding a micelle of ferric hydroxyphosphate (1,2). The subunits are most generally assumed to be identical with a molecular weight of 18,500 (2,3), although recent studies suggest that some variation in subunit constitution may be present (4). Relatively little is known about the primary structure or conformation of the subunits. A high LY helix content is known to be present (2,5) and a few side chain modification studies have been reported (6)(7)(8). Optical activity studies (5,9) suggest that the conformations of ferritin and its derivative iron-free protein apoferritin differ, and conformational differences between the two proteins have been invoked to account for the increased stability of ferritin relative to apoferritin (10); however, preliminary x-ray studies (2) suggest that differences between the two proteins must be small. Determination of the accessibility of individual side chains of ferritin and apoferritin to solvent is of particular interest in view of the suggestion that histidine side chains may be involved in the iron incorporation process (6). most useful methods for assessing the environment and solvent accessibility of ionizable residues on proteins is by determining their reactivity to H' ions (11,12). H+ ion reactivity can also be a sensitive indicator of conformational changes (11)(12)(13) and therefore, in this instance, might be useful in further assessing conformational differences between ferritin and apoferritin which are directly or indirectly attributable to effects of the iron core. Accordingly, we have studied the H+ ion titration properties of horse spleen ferritin and its derivative apoferritin, assessing with particular emphasis the relationship between the two proteins. In order to analyze the titration data, it was also necessary to determine the amide composition of ferritin, since this has not previously been reported. Interpretation of the H' ion titration curves has additionally been aided here by selected salt-binding and side chain modification studies. The results indicate important differences between ferritin and its derivative apoprotein in the binding of KCl, but suggest that these effects, as well as differences in the titration properties of the two proteins below pH 7.5, are probably attributable directly or indirectly to the presence of phosphate residues at the surface of the ferritin core. between the per cent of nitrogen and protein weight was determined as 5.81 mg of apoferritin/mg of nitrogen using the amino acid analysis and amide content determined in these studies. Occasional ferritin solutions were standardized by Folin analysis (18) using ferritin that had been standardized by nitrogen analysis for the standard curve. All protein concentrations were determined with a precision of 13%. A value of 18,500 was assumed for the subunit or "equivalent" molecular weight. Amino Acid Analysis-Protein samples were first precipitated with 5 N HCl, permitted to stand for 10 min, and then centrifuged at 3,000 rpm for 10 min. The precipitate was washed three times with 1 N HCl and then hydrolyzed in uacuo for 24 h in 6 N HCI at 105". An automated Durrum-500 analyzer was used for amino acid quantitation. Amino acid analyses of unmodified ferritin and apoferritin preparations were in excellent agreement with those published elsewhere (2,19).
Amide Analyses-In a typical analysis, aliquots of approximately 0.3 mg of protein each were digested in 1 M H,SO, at 100" in stoppered volumetric flasks for 3-, 4-and 5-h intervals. Samples containing no protein were put through the entire digestion procedure to serve as "blanks." The solutions were then allowed to cool to room temperature and sdjusted to 5 (21). The modified samples were freed from excess reagents either by extensive dialysis against H,O or by gel filtration on Sephadex G-25 in 0.1 M acetic acid. They were then lyophilized and the number of glycine residues incorporated determined by amino acid analysis using the unmodified protein as the co_n_trol.
Histidine residues of both apoferritin and ferritin were carboxymethylated by reaction of a 1% protein solution with 0.2 M bromoacetic acid in 1 M K,HPO,.
After reaction periods of 49 h and 7 days, the protein was precipitated with 5 N HCl and its amino acid content determined.
Sulfhydryl reactivity was determined by reaction with Ellman's reagent (5,5-dithiobis(2-nitrobenzoic acid)) using methods previously described by this laboratory (22 Each pH is the isoionic pH (PI,)' representative of the protein concentration of the fraction (11,12). A batch procedure was also used to determine the p1, of a 1% ferritin solution; successive additions of M B-l resin were made and then removed until a constant pH was achieved. The p1 of 1 to 2% ferritin in Hz0 was 4.58 * 0.02; the p1 of apoferritin of the same concentration was identical with that of ferritin within experimental error. For both proteins, isoionic pH values identical with those obtained by ion exchange were also obtained by exhaustive dialysis against H,O. H' Zen Titrations-Continuous potentiometric titrations were conducted under N, as previously described (26) with the modification that the entire titration assembly was housed in a polyethylene glove box which was pre-purged with N, to further decrease contamination by CO,. A Radiometer model pHM 64 pH meter was used and was routinely standardized at pH 7 and checked with standard buffers at pH 4 and 10 to ensure accurate response over the pH range of the titration.
Highest purity nitrogen was passed over an Ascarite bed and a solution of either 0.16 or 1 N KC1 prior to passing through the titration solution and the glove box. Generally, 2 ml of protein solution at 9 to 10 mg/ml were titrated.
Titrants were 0.1 N NaOH in 0.16 M KC1 and 0.1 N HCI in 0.06 M KC1 for titration at 0.16 ionic strength and 0.1 N NaOH in 1 M KC1 and 0.05 N HCl in 0.93 M KC1 for titration at ionic strength of 1. Titrations were always started at pH 4.5 to 5.0 with sufficient time (30 s to 3 min) allowed after each addition of acid or base for the pH to become constant, except as noted. At the completion of each titration the response of the pH meter to standard buffers was rechecked; for titrations in which results were regarded as valid, standardization drifts during titration were less than 0.02 pH unit. Spectrophotometric titrations of apoferritin at 0.16 and 1 ionic strength were performed in glycine/KCl buffer using methods previously described (13). Changes at 295 and 245 nm as a function of pH were recorded.
General Methods-All reagents were analytical grade and deionized water was used throughout. Circular dichroism studies were performed using a Cary 60 spectropolarimeter equipped with a model 6001 circular dichroism attachment.

Amide Content of Ferritin
and Apoferritin-Two methods were used to estimate this value. (see "Experimental Procedures"). First, the amide content was estimated directly from the NH, released on partial hydrolysis. Second, the carboxyl groups of the denatured protein in 7 M guanidine were modified by reaction with glycine and the excess glycine determined by amino acid analysis. Results are shown in Table I. Twenty-six non-amide side chain carboxyl groups are indicated by the partial hydrolysis data. This value is also well within experimental error of the value obtained by reaction with glycine even when potential contributions of the LAOOH to the glycine reaction are considered.

Circular
Dichroism of Ferritin and Apoferritin- Fig.  1 shows the near-ultraviolet circular dichroism spectrum of apoferritin between pH 6 and 3 and the far-ultraviolet circular dichroism spectra of ferritin and apoferritin near neutral pH; the high absorbance of ferritin in the near-ultraviolet precludes obtaining near-ultraviolet ferritin spectra. Both the near-and farultraviolet ellipticity spectra of the chemically prepared apoferritin are very similar to the spectra reported elsewhere (27) for "natural" apoferritin (that isolated directly from horse spleen), while ellipticity differences between the chemically derived apoprotein and ferritin in the far-ultraviolet are comparable in magnitude to previously observed (5) differ- ences between the two in far-ultraviolet optical rotation. We consider it unlikely that the greater far-ultraviolet ellipticity of chemically prepared apoferritin arises from the partial denaturation of apoferritin during its preparation, since (see below) acid denaturation of apoferritin leads to a diminution in apoferritin far-ultraviolet ellipticity.
It is also of interest to note that, despite the apparent similarity in CD spectra of the apoferritin prepared here and "natural" apoferritin, natural apoferritin has been reported elsewhere to have the same far-ultraviolet ellipticity spectrum as does ferritin (27); however, natural apoferritin has also been shown to differ in isoelectric focusing behavior and subunit composition from the bulk of ferritin from which the chemically prepared apoprotein is derived (28); so the validity of far-ultraviolet comparisons between natural apoferritins, ferritin, and its derivative apoprotein is difficult to assess.
The CD spectra of both ferritin and apoferritin are unaffected by increasing the pH to 11 at room temperature. Upon lowering the pH, however, conformational changes occur in apoferritin below pH 3.5. These are manifest by significant decreases in near-ultraviolet ellipticity below pH 3.5 ( Fig. 1) and by far-ultraviolet ellipticity decreases that occur below pH 3. At pH 2.0 the residue ellipticity for apoferritin is -16,090 at 222 nm, compared to values of -22,000 at neutral pH; no effects of ionic strength on acid denaturatian were noted. Changes in the far-ultraviolet ellipticity spectra of ferritin with low pH are not manifest until below pH 2.6, but by pH 2 the residue ellipticity of ferritin at 222 nm has decreased to -13,420, compared to values of -19,000 at neutral pH. Similar differences in acid lability between ferritin and apoferritin have been noted elsewhere (9); also, the earlier onset of changes in the environment of apoferritin aromatic chromophares with decreasing pH (as manifest by changes in near-ultraviolet ellipticity) than of loss of a helical content (as manifest by changes in far-ultraviolet ellipticity) is in accord with the observation (29) that, at low pH, changes in near ultraviolet absorbance precede apaferritin subunit dissociation.
Effect of Salts on the Isoionic pH of Ferritin and Apofer-&n-In water, the isoionic points of 1 to 2% preparations of ferritin and apoferritin are 4.58 + 0.02 (see "Experimental Procedures"). Both proteins are partially insoluble in the absence of salt and became soluble at very low salt concentrations. Since Hi titrations are performed in the presence of KCI, the effect of KC1 on the isoionic points of ferritin and apoferritin was determined by noting the change of pH produced by addition of increasing KCI concentrations to I.5 to 2.0% preparations of the isoionic proteins in H, O (Fig. 2); at least IO min was allowed after each addition for equilibrium to be obtained. Increasing KCI produces a large increase in the isoionic pH (PI,) of apoferritin that is particularly marked at low KCI concentrations.
This effect is too large and in the wrong direction to be attributed to effects of KC1 on activity coefficients in a pH region where carboxylic acids are the dominant buffer. For example, in agreement with theory (30), the pH of a solution of 5 x 10.' M acetic acid plus 5 x lo-" M sodium acetate decreases from 4.795 in the absence of added salt to 4.705 when sufficient KC1 is added to make the solution 0.11 M in KCI. Additionally the change in p1, is in the wrong direction to be attributed to the increased protein solubility in the presence of salt (11). Using traditional arguments (11,311, the increase in pH when KC1 is added to apoferritin is therefore tentatively attributed to binding of Cl-. By contrast with apoferritin, low concentrations of KCI produce an initial decrease in the p1, of ferritin which then gradually increases as the KCI concentration is further increased. The initial decrease in ferritin p1, is sufficiently large that it probably results only partially from salt effects on activity coefficients (see above) and is suggestive of K+ hinding. The subsequent increase in p1, at higher KC1 concentrations is attributable to Cl-binding, K+ binding apparently being stronger than Cl-binding in this case. That K+ binding to ferritin does occur is supported by the lesser effects of NaCl then of KCI on the pL of ferritin (Fig. 2). NaCl and KCI should produce similar changes in pH if their effect were only to alter activity coefficients (31). The lesser effect of NaCl than of KC1 on ferritin (which was repeatedly observed with different salt preparations) suggests that there are cation binding sites an ferritin that exhibit cation selectivity; i.e. Na+ hinds more weakly than K+. Interestingly, NaCl appears to reproducibly produce slightly smaller pH increases than does KC1 with apoferritin; while no simple explanation is available, the fact that NaCl does not generate a larger pH increase than does KC1 with apaferritin suggests that apaferritin lacks the cation binding sites found on ferritin. In sum, the data strongly suggest the presence of anion binding sites an apoferritin the strongest of which are either unavailable in ferritin or overshadowed by the presence of cation binding sites.
Titration ofApoferritin in 0.16 M KC!-Apoferritin titrations (see "Experimental Procedures") were initiated at pH 5, and the pH was adjusted continuously in small increments with either acid or base and then returned in similar fashion to the starting pH. With the exception that a small degree of irreversibility was noted on back-titration from pH 11, titrations were reversible and pH equilibrium was rapidly achieved in those studies in which the pH did not extend into the region of circular dichroism change. Where titrations extended into denaturing pH regions, the extent of irreversibility paralleled the extent of conformational change evident from CD studies. Thus, titration to and from pH 8 was rapidly reversible with no significant pH drifts noted, back-titration from pH 3 showed a small degree of irreversibility on the titration time scale and that from pH 2 showed marked irreversibility; regions of irreversibility were accompanied by drifts in pH that suggested a time-dependent return to the native structure and protein that had been exposed to denaturing conditions became markedly insoluble on back-titration to pH 5. It is relevant to note, however, that protein back-titrated to pH 5 from pH 8 or lowered in pH to 4, also showed some insolubility, despite the rapid reversibility of the titration curves and the lack of CD changes in these pH regions, suggesting that some aggregation had occurred without effect on H+ ion equilibria. Fig. 3 is a composite titration curve of all preparations of apoferritin studied. The results presented are largely of titration from pH 5 to 11 and from pH 5 to 2. Also shown are back-titrations from pH 3, pH 2, and pH 11. Not shown are the back-titration data from pH 8 (identical with the forwardtitration). The back-titration data from pH 2 must be interpreted with caution since they were obtained by very rapid titration to prevent significant conversion of denatured to native protein and equilibrium of H+ with the very heavy precipitate that is present under these conditions may not have been attained. Nonetheless, the data indicate that the protein is denatured by exposure to low pH and suggest that denatured apoferritin has more bound protons between pH 3 and 8 than does native apoferritin at equivalent pH.
Titration of apoferritin above pH 9 (Fig. 3), in contrast to titration in other pH regions, was not consistent with different preparations, but the source of this inconsistency remains unclear. As an aid in interpreting the data above pH 9, different preparations were titrated spectrophotometrically and found to give the same tyrosine ionization curve, indicating that differences in tyrosine ionization were not the cause of the scatter in the data. Tyrosine ionization data are shown in the inset of Fig. 3. Tyrosine ionization does not begin until pH 11 as evidenced by changes in absorbance at 245 and 295 nm. The pH at which tyrosine ionization begins is high even when expected (11,12) electrostatic shifts are allowed for. Above pH 11, the magnitude of the absorbance changes indicates that all 5 tyrosines titrate (complete ionization of a single tyrosine gives changes in 6 of 11,000 and 2,300 at 245 and 295 nm, respectively (32)), but the steepness of the titration curve between pH 11 and 12, and the appearance of irreversibility on back-titration from pH 13 indicate that conformational changes accompany tyrosine ionization. Collectively, the data indicate that the 5 tyrosines of apoferritin are abnormal, in agreement with the findings of Crichton and Bryce (29) that :>nly one of the 5 tyrosines can be nitrated and that the pK of the nitrated tyrosine is abnormally high.
An interesting feature of the spectrophotometric titration data is that no changes at 245 nm occur at pH values below those associated with tyrosine ionization; i.e. no changes occur at 245 nm which are not accompanied by the expected (32) 295 nm changes associated with tyrosine titration. Apoferritin has 3 half-cystine residues per subunit (2,19). The ionization of cysteine-SH groups should be accompanied by changes at 245 nm of sufficient magnitude to be observed under the conditions used here and generally occurs with a pK, of 8 to 9; i.e. the ionization of a mercaptan is accompanied by changes in absorbance at 245 nm approximately one-third that of tyrosine at the same wavelength (32). Our data therefore suggest that the half-cystines of horse spleen apoferritin are either largely not in the --SH form or not available to solvent. This is supported by our observation that no more than 0.2 --SH groups per apoferritin subunit are available for reaction with 5,5-dithiobis (2-nitrobenzoic acid), although a single -SH is reactive in both ferritin and apoferritin to bromoacetic acid (Table  II). Since, in our hands, the product of reaction with bromoacetic acid was insoluble, the results suggest to us that none of the apoferritin sulfhydryl groups are available in the native protein, but that one becomes available if the conformation is altered. These results are to be compared with those obtained elsewhere (8) which suggested that one apoferritin sulfhydryl per subunit, is available to both 5,5-dithiobis (2-nitrobenzoic acid) and to N-ethylmaleimide.
Further data on side chains potentially contributing to titration above pH 7 were obtained by form01 titration; in both ferritin and apoferritin approximately 6.7 lysines titrated at pH 9 at 25" after addition of formaldehyde (HCHO) (Table II). There are 9 lysine residues per apoferritin subunit (2,19). The results indicate either that 2.3 residues are deprotonated at pH 9 in the absence of HCHO or, more likely, that only 7 of the 9 residues are available for reaction with HCHO. Crichton and Bryce have observed that only 7 of the 9 lysines can be guanidinated (29) and analysis of the titration curve (see below) suggests that no lysines are deprotonated below pH 8.5 and that a maximum of only 0.5 lysine is deprotonated at pH 9, since a total of only 0.5 group titrate between 8.5 and 9.
In order to aid in the interpretation of potentiometric titration data between pH 5 and 8, the enthalpy of ionization was determined from the effect of temperature on titration behavior. Continuous titrations at 8" and 41" were compared with those at 25" after independently establishing the effect of temperature on pH at a single value of h, and AH was calculated as a function of h from the relationship (11): Results are shown in Fig. 4. Although AH values in the neutral pH region were somewhat lower in the 25-41" range than between 8 and 25", the results indicate a gradual increase in AH from values of 1 kcal/mol at pH 4.8 characteristic of carboxyl titration (11) to a value of about 6 kcal/mol near pH 7 (25") characteristic of imidazole titration (11); above pH 7.8, AH increases to values above 10 kcal/mol, signifying that a-NH, or t-NH2 groups, or both, are the predominant residues above this pH. Assuming normal AH values for all groups, the data suggest that the two groups titrating between the p1, (4.96) and pH 5.6 at 25" are almost exclusively carboxyl groups, since the average AH in this region is 1.2 kcal/mol. Between pH 5.6 and 7.3 at 25" the four groups that titrate appear to be a mixture of carboxyl and imidazole groups; assuming a AH of 6  to 7 kcal/mol for imidazole titration, the integrated AH in this region allows the titration of a maximum of only two normal imidazoles, even if the higher AH values obtained between 8" and 25", rather than average AH values, are used for calculatio;i. Above pH 7.3, the data allow a maximum of only one imidazole to titrate with normal AH. In sum, the data suggest that a maximum of 3 of the 6 histidines titrate between pH 4.5 and 8. This conclusion is tentatively supported by carboxymethylation studies. A maximum of 3 histidines were found to carboxymethylate in both ferritin and apoferritin, even after 1 week of reaction with bromoacetic acid (Table II).
There are several ambiguities about the structure of apoferritin which add uncertainty to any attempt at a rigorous analysis of its titration behavior. There is evidence that the o-NH, terminus is acetylated (33), but there may be two polypeptide chains per subunit (4) with one having a free (U-NH,; there therefore may be one or two cu-carboxyl groups per subunit. The number of arginine residues, 9.5, is nonintegral (2,19) which may be a reflection of subunit heterogeneity or of analytical uncertainties. In addition to the 2 unreactive lysines (see above), Crichton and Rryce report an unreactive arginine (29); it has been suggested that these residues are present in ion-pair linkage with carboxyl groups (29) but other reasons for their lack of reactivity (cf., lysine acetylation) cannot be precluded. Nonetheless, given the available facts, a self-consistent analysis of the apoferritin data using a modified Linderstrom-Lang analysis (11,12) is possible. The analysis (Table III) assumes that the 3 nonreactive histidines are masked in the unprotonated form, their protonation in the acid-denatured protein accounting for the tentative difference in & between native and denatured proteins in the pH region 5 to 8. (It can also be shown that the p1, of apoferritin cannot be accounted for if the 3 unreactive imidazoles are assumed to be charged at the p1, unless three positive charges are assumed removed from elsewhere in the molecule). Three histidines are assumed to titrate normally, although the normal titration of only 2 histidines would also be compatible with the data, given the small uncertainties in arginine and amide analyses. Two lysines and one arginine are assumed to be buried in the protonated form with 3 side chain carboxylate ions. One a-NH, and one aGOOH are assumed per subunit, but the results could accomodate two aGOOHs with little change in fit, particularly if one of the buried histidines is assumed buried in the protonated form. The results could also accomodate the absence of a titratable (Y-NH, if it is assumed that the prinicipal group titrating between pH 7.8 and 8.5 is an abnormal histidine (rather than an a-NH*) that titrates with an abnormally high AH and is unreactive to bromoacetate. The pK and electrostatic interaction factor w characteristic of each titratable class were obtained by successive approximations using the enthalpy of ionization data as a guide to histidine titration and the relationship: pK' = pK,,, -O.SSSw (~) where pK' is the apparent pK at any net charge (z) and pK,,, is the intrinsic pK for that class. (z) was calculated assuming that binding of ions other than H+ was constant over the pH region studied. No attempt was made to optimize the fit of the data, as by computer analysis, because of the tentative nature of some of the assumptions on which the analysis was based. Instead, emphasis was devoted to minimizing the number of adjustable parameters (i.e. the number of classes of titratable groups and the variability of w among different classes). Titrations of all classes of groups were then summed to give the theoretical curve, shown in Fig. 3; the parameters describing the titration curve are shown in Table III. The curve is in good agreement with the experimental data between pH 3.5 and 8.5 without postulating any significant lysine titration below pH 8.5. Above pH 8.5 the data can be accomodated using titration parameters for the 7 lysines well within the normal range, but the spread of the data does not warrant a rigorous analysis in this pH region. The analysis is internally consistent in that it predicts a net charge on the protein alone of 0 at the observed apoferritin isoionic pH in KCl. The deviation of the theoretical curve from the data below pH 3.5 is gratifying since it is in exactly this pH region that absorbance and CD data (see above) indicate that conformational changes occur. Agreement between the experimental data, titration theory, and the amino acid composition below pH 3.5 can be obtained by allowing w to gradually decrease below pH 3.5 and allowing buried carboxyls and imidazoles to re-enter H+ equilibrium, but we do not consider continuous titration data near pH 2 sufficiently accurate to warrant a detailed analysis. However, it is relevant to note that the titration analysis, in addition to postulating only 3 normal histidines, postulates two classes of titratable side chain carboxyl groups. The intrinsic pK of the prinicipal class (3.90) is slightly lower than would be predicted for a mixture of aspartyl and glutamyl side chains (34), while that of the more basic class (5.56) is decidedly high. While we do not believe that the above analysis of the carboxyl titration is unique in its fit with the data, it can readily be demonstrated that no single values for pK, and ui will fit the carboxyl region of the titration curve and that the best fits assume at least two classes of carboxyl groups whose differences in intrinsic pK exceed the known differences (34) between the intrinsic pK values of aspartate and glutamate side chains.
Values of w in Table III   zero must be known (11,12). We have assumed the isoelectric pH to he shell, but no simple model can accomodate this type of aggregation and the method of calculation does not affect the sensitivity of the analysis to changes in w over the course of the titration. Interestingly, the calculated value of w is only slightly higher than that estimated for other molecules of comparable size to apoferritin subunits (26).
Titration of Apoferritin in 1 M KC1-Titration studies of apoferritin in 1 M KC1 were carried out, in part to determine the effect of ionic strength change on titration parameters and in part for subsequent comparison with ferritin under identical conditions. Titration under these conditions was accompanied by extensive precipitation on addition of acid, although the precipitate did not appear to interfere with the rapid attainment of equilibrium or with reversibility. For this reason, data were collected only from pH 5 to 8.3 and from pH 5 to 3 and are shown in the inset in Fig. 5. Preliminary titration parameters derived from analysis of the data, using the same treatment as for the data at 0.16 ionic strength, are also included in Table III and show an expected decrease in w at the higher ionic strength with small changes in pKint; the parameters fit the data well between pH 8 and 3.5, with deviations below pH 3.5 again suggesting a change in apoferritin conformation.
It is also relevant that the 1 M KC1 apoferritin titration data, because of the sharper inflections inherent in the data, confirm that only eight groups (attributed in our analyses to four carboxyls, three imidazoles, and one a-NH,) titrate between the isoionic pH and the pH (8.3) at which lysine should begin to titrate (Fig. 5, inset).
Comparison of Ferritin and Apoferritin Behauior-Limited chemical modification studies of lysine, histidine, and cysteine were also applicable to KC1 solutions, the effect on calculated values of pK ,nt at 0.16 ionic strength would be a decrease of approximately 0.05 pH unit.
*The value of w was deliberately set at 0 to simplify curve-fitting.
revealed no significant differences between ferritin and apoferritin (Table II). The effect of pH on the solubility of ferritin was similar to its effect on apoferritin solubility. The titration of ferritin at 25" in 0.16 M KC1 is shown in Fig. 5 and compared with that of apoferritin. The titration curves of the two proteins are superimposed at the isoionic pH of apoferritin so that relationships between the two curves will be clear; additional rationale for this superposition is presented under "Discussion." Forward-titration data from pH 4.5 to 11 and from 4.5 to 2 are shown, as are back-titration for ferritin from pH 7.5, 11, and 2. Back-titration from pH 2 deviates only slightly from the forward-titration; the lesser degree of irreversibility on backtitration of ferritin than of apoferritin (Fig. 3) is in agreement with the greater stability of ferritin to low pH (see above) or suggests a more rapid reversal of acid denaturation in ferritin during the time course of the titration. Back-titration data from pH 11 show a high degree of irreversibility relative to the forward-titration curve. The irreversibility of the back-titration data from pH 11 is due to slow titration of the ferritin core; this manifests itself during forward-titrations as time-dependent drifts to lower pH that begin at pH 8,' the rate of drift 'The exact pH at which time-dependent pH drifts were observed was a function of the state of the ferritin preparation.
As preparations aged, the onset of the drift was shifted to lower pH. Thus, in good preparations, the pH was stable until 8.5 in 0.16 M KCl, while in old preparations drift was sometimes evident at pH values as low as 7.5. Preparations which manifested drift at the lower pH values were not used for titration studies. However, protein recrystallized from such aged preparations showed normal behavior (no drift until pH 8.5). The results suggest an influence of the protein on the properties of the core or on accessibility to the core. I""""""""" 1 +24i- is also evident near pH 2. As cited above, differences between the two proteins above pH 8 are clearly attributable to the ferritin core and are accompanied by drift during forwardtitration of ferritin above pH 8 and irreversibility on backtitration.
It is therefore of interest that we have observed that titration of the extra two groups in ferritin between pH 5.5 and 7.5 is reproducible, time-independent, and almost completely reversible.
Thus, with good ferritin preparations, rapid titra-'Horse spleen ferritin is heterogeneous with respect to iron content and includes a significant content (20 to 25%) of natural apoferritin (2). It is possible therefore that differences seen here between ferritin and its derivative apoprotein underestimate the real differences between the iron-loaded protein and the iron-free protein.
tion from pH 5 to 7.5 gives the same number of titratable residues as slow titration, no drifts are noted at pH 7.5 and back-titration (Fig. 5) largely follows the forward-titration curve. In an effort to identify these extra titratable residues, the AH of ionization of titratable residues in ferritin between pH 4.9 and 8.5 was determined as shown above for apoferritin. The results are shown in Fig. 4. In the pH region 5.5 to 7.5, the AH values indicate that the additional groups titrating in ferritin relative to apoferritin must have a AH of ionization of at least +5 kcal/mol, suggesting their identity as histidine side chains to which values of AH of ionization of 6 to 7 kcal/mol are typically (11) assigned. Phosphate groups from the ferritin core are alternative contenders for the role of these additional residues, since the second ionization of phosphate typically occurs near pH 6 to 8; although the AH of this second ionization is approximately +1 kcal/mol in simple molecules (II), there are compelling reasons (see "Discussion") for believing that phosphates should not be precluded on this basis.
The relationship between the titration curves of ferritin and apoferritin seen at 0.16 ionic strength extends to titration studies in 1 M KCl. In Fig. 5 (inset) the titration curves of the two proteins in 1 M KC1 at 25" are compared, again illustrating the identity in titration behavior between pH 3 and 5.5 and the titration of approximately 2 extra residues in ferritin relative to apoferritin between pH 5.5 and 7.5.

DISCUSSION
The isoionic pH of a protein (pIi) is that at which the intrinsic net charge on the protein (Le. the charge exclusive of bound ions) is zero. When a neutral salt, such as KCl, is added to an isoionic protein, the protein remains isoionic, even if ions are bound, and the intrinsic net charge on the protein remains essentially zero provided that the protein concentration is high relative to the H+ ion concentration, as is the situation here (11). The isoelectric pH of a protein (PI,) is that at which the total charge on the protein (Le. the charge including that contributed by bound ions) is zero (11). Any interpretation of the relative titration properties of ferritin and apoferritin must reconcile their similar isoionic points in the absence of salt with the divergent effects of salt on the isoionic points of each, as well as with reported identical values of p1, and electrophoretic mobility of both (see below). As it turns out, these considerations place severe restrictions on the number of possible interpretations of the data. For example, changes in isoionic pH (ApI,) accompanying addition of monovalent ions to proteins are generally interpreted as originating in a change in net protein charge and its accompanying electrostatic effect on H+ ion association constants (11,31) such that: ApI, = -0.8682& where ,?? is the charge introduced by the bound ions and w is the same electrostatic interaction factor as that derived from H+ ion titration analysis. This interpretation does not appear to be applicable in its simplest form to the different pH changes accompanying KC1 addition to isoionic ferritin and apoferritin since it would lead to the conclusion that there are large differences in total charge between the two proteins at the same pH. Thus, using the value of w = 0.08 found for carboxyl titration, data in Fig. 2, would indicate a total charge of approximately +l per subunit on ferritin at its isoionic pH (4.48) in 0.16 M KC1 and a total charge of -6 per apoferritin subunit at its isoionic point in 0.16 M KC1 (pH 4.96).4 Since 'In principle, the relationship between ApI, and the absolute number of bound ions is valid only at constant ionic strength, since changes in ionic strength lead to changes in activity coefficient. apoferritin adds only 2 protons per subunit (Fig. 3) on going from pH 4.96 to 4.48, the two proteins would therefore be predicted to differ by a charge of -5 per subunit at pH 4.48 and the isoelectric point of apoferritin would be predicted from the titration data to be 0.9 pH unit lower than that of ferritin in 0.16 M KCl. However, horse spleen ferritin and its derivative apoprotein have been shown to behave identically on free electrophoresis in several different buffers (35) and on isoelectric focusing (36), and therefore probably have similar charges in KCl. Therefore, in our interpretation of the data, a somewhat different basis for the changes in isoionic pH accompanying KC1 addition will ultimately be invoked.
A useful starting place in attempting to formulate a model that accounts for the behavior of the system is to consider the virtual identity in the titration behavior of the two proteins between pH 3 and 5.5 at the two ionic strengths studied and their apparent similarity in lysine and histidine availability to modifying reagents. Since titration behavior is sensitive to protein shape and to the number, nature, and environment of available titratable residues, these results suggest that the conformations of ferritin and apoferritin are largely similar and that differences in titration behavior of the two above pH 5.5 and in their ion-binding properties are a reflection of additional titratable residues in the ferritin core, or of a highly localized difference in conformation or environment of amino acid residues at the core-protein interface. The lower p1, value for ferritin than for apoferritin in KC1 can therefore be explained as follows. Per subunit, 2.2 protons must be added to apoferritin at its p1, in 0.16 M KC1 (pH 4.96) to reach the p1, of ferritin in 0.16 M KC1 (pH 4.48), indicating that ferritin (exclusive of bound salt) has approximately two additional negatively charged residues per subunit or two fewer positively charged residues, at pH 4.96, relative to apoferritin. 3 The similar chemical reactivities of the histidines and lysines in the two proteins suggest that the latter alternative is less probable than the former. However, any additional negative charges in ferritin probably do not originate from carboxyl groups since the titration curves of both proteins appear identical in the carboxyl region. The ferritin core contains inorganic phosphate and it is generally believed that a high fraction of the phosphate is near the core surface (2). We postulate that the two additional negative charges on ferritin at pH 4.96 arise from two such surface phosphate residues per subunit, each of which, by analogy with the known ionization behavior of phosphate (ll), would be anticipated to carry a unitary net negative charge at pH 4.96. Moreover, since the first phosphate pK, is generally near 2 in model compounds, this assumption is consistent with the observation that no additional groups in ferritin (relative to apoferritin) protonate until approximately pH 2 (Fig. 3), at which pH the data suggest that more protons are added to ferritin than to apoferritin.
Note that the assumption that the extra negative charges on ferritin arise from the core provides the rationale for our superposition of the titration curves of the two proteins at pH 4.96; i.e. differences in the absolute number of protons dissociated from isoionic ferritin (protein plus core) relative to apoferritin (protein) reflect the contribution of the core. The number of protons dissociated from the proteins alone are the same in both cases at pH 4.96.
Studies here were not conducted at constant ionic strength but conclusions as to the predicted charge differences between ferritin and apoferritin remain valid because the two proteins are being compared at the same ionic strength. leave only histidine residues as likely sites for strong iron coordination.
The above models assume that the principal differences in the titration behavior and stability of ferritin and apoferritin are not the result of differences in conformation; i.e. no differences are found between the titration behavior of the two proteins which cannot be attributed solely to their differences in core content.
Even differences in stability between the two proteins as observed here and elsewhere (9,10)