The Solubility of Amino Acids, Diglycine, and Triglycine in Aqueous Guanidine Hydrochloride Solutions

The solubilities of amino acids, diglycine, and triglycine have been measured in water and 1 M to 6 M guanidine hydrochloride solutions. Free energies of transfer of amino acid side chains and backbone peptide units from water to guanidine hydrochloride solutions have been calculated from these data. The results show similarity between the patterns of solubilizing effects of guanidine hydrochloride and urea, although guanidine hydrochloride is 2 to 3 times more effective than urea at the same concentrations.

show similarity between the patterns of solubilizing effects of guanidine hydrochloride and urea, although guanidine hydrochloride is 2 to 3 times more effective than urea at the same concentrations.
In previous communications the effects of aqueous urea (1,2) and aqueous ethylene glycol solutions (3) on protein denaturation were discussed on the basis of solubility studies of amino acids and their derivatives.
Urea is a fairly strong protein denaturant and ethylene glycol is known to affect protein conformation very little, and this difference in their action on proteins was explained in terms of the free energies of transfer of the constituent parts of protein molecules, calculated from the experimental solubilities.
The procedure was based on the principle of additivity of free energies of transfer; that is, the free energy of transfer of a whole protein molecule, in an unfolded state, where all parts are accessible to solvent, can be expressed as the sum of free energy contributions from various kinds of groups constituting the protein molecule. This assumption is also applied to small model compounds, so that an appropriate value of the contribution to free energy of transfer can be assigned to each particular type of group, with the further assumption that this contribution is independent of the rest of the molecule.
These basic ideas have been discussed in full detail elsewhere (2,4).
The present communication describes a similar investigation of the effects of GuHCll solutions, which are among the most powerful of the commonly used protein denaturants, and are known to convert most protein molecules to a randomly coiled state (4). EXPERIMENTAL PROCEDURE All amino acids used are L isomers, whenever optical isomers exist.
The amino acids and peptides used were purchased from * This work was supported by research grants from the National Science Foundation and from the National Institutes of Health, United States Public Health Service.
Mann except for tyrosine which was purchased from Aldrich. Most of them were used without further purification. Diglytine was recrystallized four times from 657, ethanol, and triglycine five times.
The final products in each case were shown to be chromatographically pure. Solubility Measurements--Weighed amounts of sample and solvent were placed in glass tubes with ground glass caps and Teflon sleeves.
Normally five such tubes were prepared, such that two would contain undersaturated and three saturated solutions.
The space above the mixture was flushed with nitrogen gas to minimize possible oxidation. The tubes were shaken in a water bath of 25.1" f 0.05" for 24 hours.
Except as indicated below this was sufficient time for establishment of equilibrium.
The supernatants were separated from the saturating body by suction through fritted glass filters with long legs while the tubes were still half-immersed in the bath. In most cases the supernatants were assayed by titration with 0.5 M KOH under streaming nitrogen, with a Radiometer titrator TTTlc and Titrigraph SBR-2c. Blank titrations to the same pH were subtracted.
The resultant titration curves could usually be fitted with reasonable pK values, and the amount of titrant used provided a measure of the amount of amino acid.
For a few cases, the assays were carried out by form01 titration with the same apparatus. The results were in good agreement with those by direct titrations when a solution was assayed by both methods.
For purely aqueous solutions, dry weight determinat'ion was used as described previously (2). Agreement between the results obtained by the dry weight method and by titration was very good when both methods were used for the same solution.
Solutions containing tyrosine were analyzed by spectrophotometry at 275 rnp with a Beckman DU spectrophotometer.
Some difficulty was encountered in the measurement of the Except for tryptophan and tyrosine which showed some variability in absolute solubility in water as noted below, all solubilities in water agree with the previously published values (2,7,11). The discrepancies in solubility do not necessarily indicate the use of impure samples but are presumably due to different crystalline modifications. The problem stemming from multicrystalline forms will be discussed in mole detail in a forthcoming publication. a This is not a value based on the solubility but an assumed value which falls on the extension of a smooth curve of F/t plotted against GuHCl concentration. * A different value of 1.38 g/100 g of water was reported previously with a different sample (2,3). This value is close to 1.36 g/100 g of water obtained by Hade (7).
c A different value of 0.0451 g/100 g of water was obtained previously with a different sample (2,3). This value is close to 0.0453 g/100 g of water reported by Dalton and Schmidt (8), while the present value is similar to 0.0479 g/100 g of water obtained by Dunn,Ross,and Read (9 It should be noted in this connection that the difference between the high and low solubilities is actually quite small, and that free energies of transfer would have been altered by only 30 cal per mole if the alternate values were used.
In the case of diglycine the formation of a new crystalline phase could be observed directly.
Precipitation of new crystals occurred after separation of the solution from the solid phase in some of the experiments.
In other experiments, the phase transition was apparently complete, or nearly so, before removal of solution for analysis.
Little or no precipitation occurred, but the diglycine content was always substantially lower than anticipated.
None of the results obtained with diglycine in 5 or 6 M GuHCl could be taken as representing equilibrium with the original solid phase. These results were therefore discarded, and free energies of transfer of diglycine to 6 M GuHCl were estimated on the basis of a smooth extrapolation of the results obtained at lower GuHCl concentrations.
Bello, Haas, and Bello (6) have described a loose crystalline inclusion compound between dimethylacetamide and GuHCl. It is possible that the second crystalline phase observed by us is a complex of this kind. Solubility of Amino Acids in Aqueous Guanidine Hydrochloride Vol. 245,No. 7  0 This is not a value based on the solubility but an assumed value for diglycine which falls on the extension of a smooth curve of AF', plotted against GuHCl concentration.

RESULTS
The solubilities determined in this study are listed in Table I. The free energies of transfer of the solutes from water to aqueous GuHCl solutions have been calculated directly from these results.
The method was communicated previously (2), and a brief explanation will suffice. At saturation, the following condition prevails.
pk + RT In Ni,, + RT In yiSw = pf,(t + RT In Ni,e + RT In 7i.e (1) where &w and $,a are the standard chemical potentials assigned to solute i in water and guanidine hydrochloride solution of a particular concentration, respectively; Ni,, and Ni,e are the solubilities in mole fraction units: and yi,w and ~i,e are the activity coefficients referred to P!,~ and pi,e, respectively. We have defined the free energy of transfer of solute i (AFt) as &e -$+,, so that AFI = RT In Ni,,/Ni,e + RT In Y&YJ,Q By this definition, the effect of solvent on the chemical potential of the solute at infinite dilution is contained entirely in the term AFt.
The term RT In y represents solely the effect of solutesolute self-interaction on the chemical potential, and, accordingly, we call y the self-interaction coefficient hereafter. The second term on the right side of Equation 2 represents effectively a subtraction of the effects of self-interaction so that the expression on the right side may become, as intended, a measure of the free energy of transfer at infinite dilution.
Unfortunately, we have not been able to assign reasonable values to y for GuHCl solutions, and are forced to ignore the self-interaction term in the calculation of the free energy of transfer.
The value of AFt thus obtained is necessarily approximate and is called the apparent free energy of transfer (AF'J. The values listed in Table II   acid can be considered as the sum of independent contributions from the side chain and from the +H3N-CH-COO-moiety of the molecule.
A,f't is evaluated as the difference between AF', for a givenamino acid andAF't for glycine. Strictly speaking, the quantity thus represents the effect of substituting a given side chain for a hydrogen atom.
However, the value of Ayt for a backbone peptide unit, as given in Table IV, is calculated as the contribution of a glycyl residue, which contains an extra hydrogen atom, so that the sum of the peptide unit and the side chain will correctly represent the effect of inserting an entire amino acid residue into a parent molecule.
Again the results of Tables III and IV are labeled wit'h a prime to indicate that they are subject to correction for the influence of selfinteraction.
( * Sum of Af't for a peptide unit and a methylene group. The latter values have been calculated as one-third the difference between AF', values for leucine and alanine. c Equal to Af't for a peptide unit (Table IV).
energies of transfer. Based on the principle of additivity, it has been attempted to assign a unique value for the free energy contribution to each particular kind of group, regardless of whether the group is present on an unfolded protein molecule or on a smaller molecule.
The idea was originally proposed by Cohn and Edsall (11) for any series of molecules containing hydrocarbon moieties. They showed that for any homologous series, differing only in the number of CH2 groups (nCH2), relative solubilities in 100% ethanol and in water could be expressed in terms of the relation (cf. Equation 2) AF't = KI WHZ + Kz (3) where K, and KS are constants. Essentially the same equation was shown by Wetlaufer et al. (10) to hold true for the free energies of transfer of hydrocarbons from water to aqueous urea and GuHCl solutions. Schrier and Schrier (12) have attempted to extend this concept by describing free energies of transfer of a number of compounds from water to inorganic salt solutions in terms of independent contributions from CHZ, CHZ, and amide groups.
Careful examination of the data cited above, as well as our own results, indicates, however, that the principle of additivity is not valid when applied to groups as small as CHZ, CH,, and amide groups. Equation 3, for example, applies accurately, with a unique value for K1 for a given solvent, only for CH2 groups in linear hydrocarbon chains.
In general the nature of the neighboring groups to which a particular group is attached has some influence on the Af't or Aft value that one would estimate from solubility data.
We have therefore preferred to calculate group contributions for larger groups, as large as is practical, considering the model compounds chosen and the intent to apply the results to calculation of the free energies of unfolded protein molecules.
Results have accordingly been presented in terms of contributions from entire amino acid side chains and the peptide backbone unit. Effects of neighboring groups should be much smaller when the chosen moiety is large enough so that points of attachment to neighboring groups represent only a small part of the whole unit under consideration.
Even with the choice of relatively large groups there are obvious inconsistencies in the data presented in this and previous papers.
The most striking example is provided by the backbone peptide unit (with extra hydrogen atom), which may be thought of as representing the moiety CHZCONH, CONHCH2, or NHCH&O.
Values of Aft or Af't for this grouping have shown poor agreement with each other, especially when our results with diglycine and triglycine (2,3) are compared with the results of Robinson and Jencks (13), who used a tetraglycine derivative.  Vol. 245,No. 7 concentrations in other systems. It is clearly seen that the peptide unit contribution as obtained between glycine and diglycine, which contains only one peptide unit, is larger in magnitude than that obtained between diglycine and triglycine, which contains two peptide units, in alcohols and dioxane. It may be interpreted as an indication that a single peptide unit is fully capable of interaction with these solvents, and the lower value of Afft for the triglycine-diglycine pair can then be readily explained in several different ways which need not be discussed here.
The important aspect of Table VI for the present paper is that the trend in urea and GuHCl solutions is in the opposite direction from that observed with alcohols or dioxane: the introduction of a second peptide unit causes a considerably larger effect.
This may be taken as support for the mechanism of double hydrogen bonding proposed by Robinson and Jencks (13) (14). These authors also suggest the possibility for multiple hydrogen bonding between partially charged vicinal imino groups of these compounds of the urea-guanidine series and neighboring carbonyl groups on a peptide chain (15).
It is possible, of course, t'hat the chloride ion has to be invoked to explain the foregoing, since both anion and cation contribute to the denaturing action of an electrolyte and to related interactions with model compounds.
However, possible contribut'ion from chloride ion is not likely to be significant, as judged from the studies of the effects of neutral salts by von Hippel and Schleich (16). Another evidence for the probably insignificant role of chloride ion would be the well known similarity between GuHCl and urea in their solubilizing behavior. The side chains of asparagine and glutamine have also shown considerable anomaly in the free energies of transfer in all solvents studied.
The asparagine side chain has shown a more negative free energy contribution in several solvents, including GuHCl solutions, than the glutamine side chain, whereas the additional CH2 group on the latter is expected to make its contribution more negative. Table VII shows, however, that the free energies of transfer of glutamine are in fact essentially as expected in GuHCl and glycol solutions.
The experimental values for Ajftand Ayt are compared with calculated values based on the sum independent contributions from the amide group, -CH2--CO-NH2, assumed to give an identical contribution to a single backbone peptide unit and from a methylene group.4 The agreement is seen to be excellent for GuHCl and glycol solutions, although it is less good for urea solutions.
In any event, it would seem that asparagine is much more anomalous in its behavior than glutamine.
No explanabion can be given at this time.
4 In this case it is not inappropriate to consider a methylene group as an independent entity because it is joined on both sides to hydrocarbon groups.
Comparison between Guanidine Hydrochloride and Urea-Qualitatively, GuHCl and urea evidently have similar effects on all kinds of groups, except that GuHCl is about 2 to 3 times as effective at a given denaturant concentration, and their action is distinct from that of other denaturants, such as alcohols and inorganic salts.
(For example, Aft values for the peptide backbone unit and the glutamine side chain are positive for transfer to alcohols. Aft values are positive for transfer of hydrophobic groups to inorganic salt solutions.) Quantitative examination of the data shows, however, that urea and GuHCl do not have exactly parallel effects, as can be seen, for example, in the comparisons made in Table VIII. Urea solutions seem to be, in a relative sense, favored for solubilizing aromatic side chains, particularly phenylalanyl and tyrosyl groups, while GuHCl solutions are especially powerful in solubilizing peptide units and histidine and glutamine side chains. Blocked Amino Acids-We have shown in our previous studies that Aft values based on the solubilities of N-carbobenzoxy or Nbenzoyl amino acids agree well with those based on the free amino acids for ethylene glycol (3), but the agreement is poor in urea solutions (2). Similar data have been obtained in GuHCl solutions, and have shown fair agreement for peptide units, but very poor agreement for aromatic side chains.
The latter result suggests that the major discrepancies between the blocked amino acids and the free amino acids may reflect interaction between the bulky aromatic ring in the blocking group and the amino acid side chain.
We shall present the results for the blocked amino acids in several solvents in a separate paper in the near future.