Introduction

It has long been recognized that protein-water interactions play an important role in the determination and maintenance of the three-dimensional structure of proteins. Quite apart from its fundamental importance and interest, knowledge of processes occurring on hydration or dehydration of proteins is also important in biotechnological applications of proteins, such as their use as catalysts in anhydrous organic solvents, the stabilization of protein preparations for pharmaceutical use, and in food preservation. It is not surprising that protein-water interactions have been the subject of intense study and have provided very significant advances in our understanding of the involvement of water in protein stability, dynamics, and function [1-3]. Such studies can be classified into those employing protein solutions, in which the necessary variation in water activity is achieved by the use of water-cosolvent mixtures, and those employing hydrated protein powders, films, or glasses. As will become apparent in this chapter, most protein properties of interest are expressed and can be studied in the absence of a bulk water phase. Such studies also serve to emphasize the types of water of greatest importance to the determination of protein properties, namely, internal and surface water. It was once believed that drying a protein always resulted in major disruption of the native conformation to give a collection of randomly unfolded polypeptide chains—an inactive polypeptide "dust" such that studies of dry or partially hydrated proteins were of little relevance to issues of structure-function relationships. This view was challenged by the extensive systematic work of Rupley, Careri, and their coworkers on lysozyme hydration in the 1980s, which suggested that little or no conformational rearrangement occurred on dehydration [3]. Recent Fourier-transform infrared spectroscopy (FTIR) studies by Prestrelski et al. [4], however, indicate that the true situation may be more complicated than either of these extremes, the extent of dehydration-induced conformational rearrangement being quite variable and protein dependent.

Studies of protein hydration using solid samples allow the water activity to be varied over a very wide range, from the dry state to formation of the solution state, which occurs at about 0.9-1.0 h (g water/g protein). Although the point at which hydrated protein powders collapse to form true solutions represent concentrations that are orders of magnitude larger than those normally employed in physical studies of proteins, hydration levels in the range from 0 to 1.0 h prove to be highly relevant to a discussion of the role of water in protein function. Indeed, one of the most remarkable conclusions from studies of protein hydration conducted in the last 10 to 15 years is that very little water is required for the onset of functionally important protein processes. As we shall see, enzymes begin to regain their catalytic activity at hydration levels as low as 0.12-0.2 h, and most important changes in the thermodynamic and dynamic properties of proteins occur over the hydration range from 0 to 0.4 h.

The intracellular environment in which proteins function bears very little resemblance to the dilute solutions normally employed in physical studies of enzymes. The amount of intracellular water available to solvate proteins is small, about enough for two or three layers of water per protein molecule, which corresponds approximately to the hydration level at which hydrated powders collapse to form solutions. Nuclear magnetic resonance (NMR) titration of lysozyme solutions suggests that bulk water properties are only seen at hydration levels in excess of 1.4 h, which represents about 60% water by weight, within the range of water contents observed for some biological tissues [5]. Thus, some tissues and organisms may possess little, if any, bulk water.

Rupley and Careri [3] have recently provided an excellent and extensive review of protein hydration. We will not repeat such a monumental task here, but instead describe the events that occur as the protein is brought from the dry state to the fully hydrated state with particular emphasis on the role of water as a "plasti-cizer" of the protein conformation and in the reestablishment of the dynamic properties of proteins. A consideration of the results derived from hydration studies of proteins and synthetic polymers, recent work on glass transitions, and the identification of dynamically distinct structural classes in globular proteins derived from hydrogen isotope exchange results leads us to a picture of the dynamic properties of proteins in which free volume rearrangement and mobile internal water are the primary determining factors. No direct, model-independent measure of the elusive population of internal water has been made, except by x-ray crystallography, which can define only the sites with the highest water occupancies. As our understanding of the role of water in protein dynamics has developed, however, we have become increasingly convinced that the size of the population of mobile buried water and its importance to protein dynamics has been greatly underestimated.

II. PREPARATION OF SOLID STATE SAMPLES

Most work on protein hydration has involved the use of protein powders that are brought to the required hydration level by isopiestic equilibration with a solution of a salt or sulfuric acid of known water activity [6], Sample preparation is straightforward. The protein solution is usually dialyzed against water or buffer and then lyophilized. Further drying is obtained by storage in a vacuum desiccator over phosphorus pentoxide. The lowest hydration level we have obtained in this way for lysozyme is about 0.01 h, which represents about 8 moles water/mole protein. Removal of the last few water molecules is extremely difficult and may require drying under high vacuum (10~6 torr) for several days. The dry protein powder is transferred to a sealed vessel over a solution of known water activity and is allowed to equilibrate for a period of one to a few days, depending on the physical state of the powder and the final hydration level required. Uniform hydration can also be achieved by addition of water directly to the dry protein powder. The hydration level of the sample can be determined gravimetrically or by Karl Fischer titration, although problems of protein solubility in the Karl Fischer reagent, usually a solution in pyridine, and reaction of certain protein groups with the reagent can cause problems with the latter method [7,8], Correction for the small amount of water associated with the lyophilized protein is required for the most precise work, but this is easily determined from the weight loss of a sample of the lyophilized protein after drying in an oven at 105°C for about 24 hours.

Hydrated powder samples are not suitable for transmittance optical spectroscopy because of scattering, and instead it is possible to employ hydrated films of protein that are cast from concentrated protein solutions spread onto an appropriate optically transparent support (i.e., CaF2 for infrared measurements). The hydration level of the films may be controlled by drying in a sealed vessel over a solution of known water activity. It is also possible to prepare large, clear "glassy" pellets by drying concentrated protein solutions in a suitable container [8],

III. ADSORPTION OF WATER VAPOR BY PROTEINS: THE SORPTION ISOTHERM

Measurement of the uptake of water by protein at constant temperature as a function of water activity or relative vapor pressure provides the adsorption isotherm, although as Kuntz and Kauzmann [1] note, the term adsorption is not particularly appropriate for what is essentially a solvation or absorption process. The adsorption isotherm is conveniently measured using the isopiestic equilibration method described in Sec. II. The sigmoidal isotherms are of the BET type [9]. The adsorption isotherm for uptake of D2O by HEW lysozyme at 27°C obtained by Careri et al. [10] is shown in Fig. 1. At low water activities, water uptake by the protein increases rapidly. There is a distinct "knee" at a relative humidity of between 0.05 and 0.2, corresponding to a water uptake of about 0.07 h, beyond which the uptake of water occurs more slowly with increases in relative humidity. At relative humidities greater than about 0.9, water uptake increases rapidly again. We should note that the uptake of water by proteins at a particular relative humidity is independent of the surface area of the protein powder, in contrast to adsorption isotherms for inert gases such as nitrogen or argon, which do depend on the degree of subdivision or fineness of the powder samples [11].

One of the most significant features of the water sorption isotherms is the appearance of pronounced hysteresis below relative humidities of 0.8 to 0.9, so

P/P0

Figure 1 D20 sorption isotherm for lysozyme at 27°C. The curves labeled a, b, and c represent the three classes of sorption sites defined by the D'Arcy and Watt equation (Eq. (3)). From Careri et al. [10]. Reprinted by permission of John Wiley & Sons, Inc.

P/P0

Figure 1 D20 sorption isotherm for lysozyme at 27°C. The curves labeled a, b, and c represent the three classes of sorption sites defined by the D'Arcy and Watt equation (Eq. (3)). From Careri et al. [10]. Reprinted by permission of John Wiley & Sons, Inc.

that the adsorption curve lies 10-30% below the desorption curve. Adsorption isotherms for nitrogen and argon show no hysteresis [1], Luscher-Mattli and Ruegg [12] have shown that water sorption hysteresis depends on the extent of prior dehydration of the sample. For example, adsorption-desorption cycles for a-chymotrypsin between relative humidities of 0 and 0.93 exhibit a pronounced hysteresis (Fig. 2). When samples are prepared by evaporation of a solution at a relative humidity of 0.92 (corresponding to a hydration level or water uptake of 0.3 h) and then cycled between relative humidities of 0.15 and 0.93, however, no hysteresis is observed within the limits of experimental error, which were reported as 0.002-0.005 g water/g protein (Fig. 2). The lowest hydration level attained was 0.065 h. Evidently, hysteresis is a consequence of removal of the most tightly bound water molecules.

A. Conventional Sorption Isotherms

A variety of theoretical models have been employed to describe sorption isotherms and the hysteresis effects in proteins. Much of the early work has been critically reviewed by Kuntz and Kauzmann [1], who also discuss the virtues of solution theories, such as those due to Flory [13], Hailwood and Horrobin [14,15], and D'Arcy and Watt [16], versus surface models such as Brunauer, Emmett, and Teller (BET) theory [9]. A more complex but far more powerful analysis has been developed by Cerofolini and Cerofolini [17], which accounts for the heterogeneity of water-binding sites through the use of an adsorption energy distribution. The theory is also capable of describing hysteresis phenomena.

Although protein water sorption isotherms resemble the sigmoidal BET isotherm, they are found to obey the BET equation only up to a relative humidity of about 0.6. The BET equation is given by

where x is the relative vapor pressure, p is the experimental water vapor pressure, and po is the water vapor pressure of pure water at the same temperature and pressure employed in the experiment; W is the hydration level or uptake of water by the protein, and K and Wm are constants proportional to the energy of adsorption and the number of binding sites, respectively. The BET model assumes a fixed number of independent binding sites that accommodate a monolayer of adsórbate molecules (the Langmuir term on the right-hand side of Eq. (1)) and allows subsequent layers to bind more weakly.

Other isotherms derived from surface models are equally limited in their description of water adsorption to proteins. More success in fitting the entire range of adsorption data, however, is obtained with solution models such as the Hailwood and Horrobin [14,15] equation,

Figure 2 Hysteresis in sorption isotherms for a-chymotrypsin. (a) Two adsorption-desorption cycles on lyophilized a-chymotrypsin. Curves 1 represent the first cycle, curves 2 the second cycle. The relative humidities of each cycle went from 0 to 0.93 and back to 0. (b) Desorption-adsorption cycle. Relative humidities went from 0.92 to 0.15 and back to 0.92. (c) Second cycle on the sample shown in (b) over the relative humidity range from 0.92 to 0 and back to 0.92. From Luscher-Mattli and Ruegg [12]. Reprinted by permission of John Wiley & Sons, Inc.

Figure 2 Hysteresis in sorption isotherms for a-chymotrypsin. (a) Two adsorption-desorption cycles on lyophilized a-chymotrypsin. Curves 1 represent the first cycle, curves 2 the second cycle. The relative humidities of each cycle went from 0 to 0.93 and back to 0. (b) Desorption-adsorption cycle. Relative humidities went from 0.92 to 0.15 and back to 0.92. (c) Second cycle on the sample shown in (b) over the relative humidity range from 0.92 to 0 and back to 0.92. From Luscher-Mattli and Ruegg [12]. Reprinted by permission of John Wiley & Sons, Inc.

Wm 1 + Kx 1 - -yx w which assumes binding to a set of independent sites (the Langmuir term again) as well as formation of an ideal solid solution, or with the D'Arcy and Watt [16] equation,

1 + Kx 1 —yjc which adds a class of weak binding sites with affinity proportional to C to the Hail-wood and Horrobin model. Equations (1), (2), and (3) represent two-, three-, and five-parameter models, respectively, and therefore it is not surprising that they provide successively better fits to the data. Luscher-Mattli and Ruegg [12] have employed these models to fit adsorption isotherms for a number of proteins. Kuntz and Kauzmann [ 1 ] suggest that the addition of the linear term in Eq. (3) does little to improve fits to adsorption data and is of doubtful physical significance for proteins because it corresponds to such weak binding that saturation of these sites would require an unreasonably large number of water molecules per monolayer site. Careri et al. [10] employed the D'Arcy-Watt equation in the analysis of DjO sorption isotherms, however, and obtained coverages for the strong and weak binding sites that were in agreement with independent infrared measurements and gave values of about 100 water molecules per protein molecule for the coverage of the weak binding sites. This is not an unreasonably large number and was assigned to coverage of carbonyl groups along the polypeptide backbone. The fit of the D'Arcy and Watt equation to the sorption isotherm for lysozyme is shown in Fig. 1. Typical values of Wm for proteins obtained with Eqs. (1) and (2) vary between 0.05 and 0.1 g water/g protein. The corresponding values for K vary between 10 and 20. The parameter 7 in Eq. (2) has a value of 0.8-0.9 and is related to the activity of water in the solid solution. Despite their different physical derivation and interpretation, the BET and Hailwood and Horrobin equations are very similar. In fact, the only difference is the appearance of the factor 7 in the Hailwood and Horrobin equation, which thus effectively serves to adjust the degree of curvature and the size of the BET multilayer term.

Kuntz and Kauzmann [ 1 ] cite a number of reasons for preferring the solution models, among them that no phase transition is observed in the adsorption isotherm over the entire composition range, despite the change in sample appearance from solid to liquid (the lysozyme-water system forms a solution at a hydration level of about 1.0 h). Among the solution models, the Hailwood and Horrobin equation is the simplest that fits the data well over a broad range of relative humidities. The characterization of the "monolayer" sites by a single adsorption energy, however, is not very realistic given the variety of polar groups found at the protein surface. Nor is it reasonable to expect the protein conforma-

tion to be unaffected on binding water. Adsorption energies are significant with respect to the interaction energies among protein groups so, as Cerofolini and Cerofolini [17] note, "A kind of surface reconstruction takes place after adsorption." A theory for the adsorption of water by proteins that accounts both for the heterogeneity of the water-binding sites and the possibility that adsorption can alter the protein conformation has been developed by Cerofolini and Cerofolini [17] and is described below.

B. Site Heterogeneity and Conformational Perturbations

The heterogeneity of the protein surface is due to two distinct factors. First, the water-binding sites are chemically diverse and include a variety of charged groups and uncharged polar groups. Second, the protein samples a variety of conformational states, which gives rise to a distribution of local environments for each water-binding site. This heterogeneity is measured by the distribution of adsorption energies, q, through the probability density function (pdf) <¡>(<7), so that §{q)dq is the fraction of sites with adsorption energy between q and q + dq. The probability that a water-binding site with adsorption energy q is exposed is denoted by P(q | W). This function accounts for the influence of hydration on the conformational state of the protein. The hydration level at a vapor pressure p is given by

o where Q(p,q) is the local adsorption isotherm describing the uptake of water at sites with adsorption energy q at a vapor pressure p. Cerofolini and Cerofolini [17] employ a BET adsorption isotherm,

qQ is the heat of vaporization of water. To simplify subsequent calculations, Cerofolini and Cerofolini [17] assume C(q) » 1 and employ the following approximation for 6(p, q):

where py. = Co"'po exp {qJkT).T\\e expression for water uptake becomes

W(p) = (l - J p[p + PL exp(-q/kT))-i P(q | W)4>(q) dq v o

Note that p/{p + pi e\p( -q/kT)} is simply the Langmuir isotherm 0lang(/m).

It is necessary to specify the function P(q \ W). Cerofolini and Cerofolini [17] develop a mean field expression for P(q \ W). Sites with adsorption energy q are considered to be either exposed or masked. The probability that a site with adsorption energy q is exposed in the dry protein is assumed to be

where q* is the adsorption energy of a reference site with P — 0.5 and ct is a suitable parameter. Masking is suggested to occur via long-range electrostatic interactions between sites that are modified by the presence of water. Sites with greater adsorption energy (i.e., more polar) are therefore assumed to have a greater probability of being masked in the dry protein, while hydration increases exposure of water-binding sites. All sites are assumed to interact with a mean field generated by the presence of water, so that the energy difference between masked and exposed sites is a function of the hydration level. The probability that a water-binding site is exposed is given by

where a is a constant. The presence of W(p) on the right-hand side of Eq. (10) greatly complicates calculations with Eq. (8). The following simple approximation, valid when ct is very small, is therefore employed:

P(„|W)=|1 for q < q* + aW(p) {q 1 ; lO for q > q* + aW(j}) K '

for which the following expression for water uptake is obtained:

W{p) = (l - jo9*+ aW,<" p\p + pL 4>(?) dq (12)

The sorption isotherm can be calculated for any given adsorption energy pdf. For example, if we suppose that the protein has a uniform distribution of adsorption energies, defined over some range qm < q < qM ,

I/to forqm<q<qM 0 otherwise where 00 = — qm, then integration of Eq. (12) gives

An example of this sorption isotherm, with to = 4kT, p,Jpo = 0.1, q* = qm + 0.5o), and a = 0.25w is shown in Fig. 3.

Equation (8) is a far more realistic description of sorption isotherms for proteins than the conventional isotherms. The difficulty, of course, is to define the functions P(q \ W) and <f>(<7)- What is of more interest is the possibility of determining these functions from experimental sorption isotherms by numerical solution of Eq. (8). Equation (8) is a Fredholm integral equation of the first kind, the solution of which is very unstable to noise in the data. Very powerful regular-ization methods, however, are now available for the numerical solution of Fredholm integral equations that address the problems of stability to noise and solution uniqueness. Probably the best known of these is the regularized, least squares algorithm called CONTIN, developed by Provencher [18,19], but maximum entropy methods [20,21] offer an alternative approach. Given an experimental estimate of W{p) and the reasonable assumption of a Langmuir or BET model for the local sorption isotherm, these methods should yield P(q \ W)^(q) directly. Hysteresis phenomena complicate the analysis, but may also be amenable to analysis with integral transform methods. We outline such an analysis in the next section.

P/P0

Figure 3 A numerical example of the adsorption and desorption isotherms given by Eq. (14). From Cerofolini and Cerofolini [17],

C. Sorption Hysteresis

We have referred, so far, to the sorption isotherm W(p) without distinguishing adsorption isotherms (Wa(p)) from desorption isotherms (Wd(p)); that is, we have assumed complete reversibility. As noted earlier, however, water sorption isotherms for proteins show pronounced hysteresis. In their development, Cerofolini and Cerofolini [17] suppose that W(p) can be measured under equilibrium conditions starting from the completely dry protein; that is, they employ the adsorption isotherm. This is, however, probably not possible for most proteins. For a-chymotrypsin, only sorption cycles starting at high relative humidities (p/po), which do not go below a relative humidity of about 0.15 (corresponding to a hydration level of about 0.07 h), represent equilibrium conditions. Interestingly, Luscher-Mattli and Ruegg [12] found the desorption branches of all scanning cycles to be identical. In other words, it is the desorption branch rather than the adsorption branch of sorption cycles displaying hysteresis that overlap the reversible sorption cycle. Desorption isotherms may thus represent equilibrium conditions regardless of the prior treatment of the sample, at least for a-chymotrypsin and tropocollagen.

A number of explanations of hysteresis have been suggested, including capillary condensation [22] and, most recently, the possibility that nucleation sites for condensation of water on the surface is limiting during adsorption [3,23]. There is much evidence, however, to suggest that conformational changes occur in the region of low hydration, and this strongly supports the notion that sorption hysteresis is a molecular phenomenon related to conformational change in the protein molecule [12,17,24], As will be discussed later, the critical hydration level reported by Luscher-Mattli and Ruegg [12] corresponds to the hydration level where one begins to observe the development of internal protein mobility, suggesting that conformational flexibility and rates of conformational change are intimately linked to the hysteresis phenomena.

Bryan [24] has described several thermodynamic models of sorption hysteresis involving slow confrontational changes or intermolecular rearrangements, some of which are now given.

Model I: Slow Conformational Change Hysteresis. All protein molecules are assumed to have the same conformation at low relative humidities, but undergo a slow conformational change at some intermediate relative humidity. The conformational change is sufficiently slow that the new equilibrium position is not reached in the time required for the sorption experiment. The adsorption isotherm should be reversible below the vapor pressure range over which the conformational change takes place, while the desorption isotherm should be reversible above this range.

Model II: Conformational Hysteresis. The protein conformation changes continuously with addition or removal of water. At any particular vapor pressure, the protein is capable of adopting a number of conformations characterized by different hydration levels, W,(p) over the range Wa(p) < W,(p) < W<i(p). Each conformation corresponds to a local free energy minimum and is stable for an extended period of time. The sorption isotherm becomes reversible only when enough water has been added (pr) to the protein to increase the rates of conformational changes sufficiently for the protein to attain a global minimum energy conformation.

Model III: Phase-Annealing Hysteresis. Relatively rapid intramolecular conformational changes can occur, but intermolecular repositioning of protein molecules relative to one another to attain the most favorable intermolecular arrangement (free energy minimum) is limited at low hydrations. Phase annealing becomes fast, and the isotherms become reversible only when enough water (pr) is present to assist in repositioning.

Bryan [24] has described several experimental tests of these models and others involving reversibility studies around the whole hysteresis loop and long-term weight changes and vapor pressure reversal studies over part of the hysteresis loop. For example, models I and III predict long-term changes in hydration levels as the water-protein system slowly moves to its global free energy minimum. Unfortunately, no studies of long-term weight changes have been performed, and the only detailed reversibility study we are aware of is that of Luscher-Mattli and Ruegg [12].

The theory of Cerofolini and Cerofolini [17] accounts for hysteresis through the function P{q | W), which, it will be recalled, is the probability that a site with adsorption energy q is exposed. This probability depends on the conformational state of the protein, which in turn depends on the hydration level and the previous history of the sample. Cerofolini and Cerofolini generate sorption cycles resembling those observed experimentally by adopting different functions for P(q | W) in expressions for the adsorption and desorption branches; that is, P(q | Wd) is assumed to equal to one (all sites are equally exposed during desorption) but P(q | M7;,) is given by Eq. (11), although it is possible that some buried water molecules may be trapped by dehydration-induced conformational changes that prevent access of these buried water molecules to the surface.

If P(q | Wd) = 1 is assumed for the desorption isotherm in Eq. (8), numerical solution of the integral in this equation using Wa{p) gives §(q) directly. Comparison with the solution P(q \ Wd )<J>(<?) obtained from a similar analysis of Wa(p) would then give an estimate of P(q \ Wa). This function gives the probability that a site is exposed during adsorption (relative to its exposure during desorption) and thus should indicate which sites, as indexed by their adsorption energy, are influenced by hydration of the protein.

The treatment of sorption isotherms in terms of site exposure probabilities emphasizes hydration-induced protein conformational change as the source of hysteresis. Preliminary investigations of the hysteresis loop in lysozyme do appear to support the conformational hysteresis model of Bryan [24], which is consistent with the theory of Cerofolini and Cerofolini [17] to the extent that both models propose continuous changes in protein conformation with changes in hydration. Changes in hydration, however, also influence the dynamic properties of proteins, and this would appear to be the key to understanding the irreversibility of the sorption isotherms. In solution, proteins fluctuate among a number of conformational states ("mobile defect" conformers) of approximately the same free energy, through redistribution of free volume and rearrangements in hydrogen bonding and other noncovalent interactions [25]. As water is removed, the flexibility of the protein conformation is reduced. Below a hydration level of about 0.1 h, the protein conformation becomes rigid [3], Recent solid state NMR studies (see Sec. VI.A) indicate that the distribution of conformational states sampled by the protein is a function of the hydration level, the distribution of conformations being much broader for the dry protein than for the fully hydrated protein [26,27]. Fluctuations between these states, however, are greatly reduced or do not occur at low hydrations. Water acts as a "plasticizer" of the protein conformation, allowing conformational rearrangements to occur. In addition, water binding serves to order the protein conformation, enhancing the probabilities of certain conformational states and suppressing others, so that the width of the distribution of conformational states is greatly reduced.

As the protein is dehydrated and its conformational flexibility decreases, a protein molecule can become "trapped" in a local free energy minimum corresponding to some conformational state it happened to be sampling at the time the water molecules were removed. Activation barriers between this conformational state and neighboring states increase when certain binding sites lose water, and thus conformational transitions from this state to others are slower or not possible in the absence of water. As dehydration continues, one can imagine protein molecules becoming "trapped" in successive local free energy minima corresponding to different conformational states. The broad distribution of isotropic chemical shifts observed by solid state NMR for the dry protein would be one consequence of this accumulation of different conformational states on dehydration of the protein. In this way, a number of different conformational states, each characterized by a different hydration level, can be found at a given vapor pressure, which is a requirement of the conformational hysteresis model of Bryan [24]. At low hydrations, these conformations are not present in the equilibrium proportions characteristic of the global free energy minimum, and thus could give rise to the hysteresis observed in the adsorption isotherm.

1. Buried Water as a Source of Sorption Hysteresis

It should be noted that the protein conformation is coupled to the pattern of occupied water-binding sites, so that changes in one must modify the other. For example, tosylation is known to induce conformational changes in ot-chymotrypsin that have a significant effect on the hydration of the protein [28]. Conformations sampled during adsorption that differ from those characteristic of the global free energy minimum sampled during desorption will have different patterns of occupied water-binding sites; that is, the site exposure probabilities will be different during adsorption and desorption. Evidently, the sites responsible for the hysteresis behavior are those with the highest adsorption energies (the most tightly bound water molecules). These include ionizable side chains but must also include buried water sites, since it is known from hydrogen exchange studies that internal motions recover at very low hydrations, and these motions require water to enter the protein in order to plasticize the conformation. Which type of site (ionizable side chains or buried water) is responsible for sorption hysteresis? The occupancy of buried water sites is likely to be very sensitive to changes in protein conformation, and thus one would expect these sites to be the most relevant to any conformational model of hysteresis. At a given relative humidity, buried water sites that are occupied in the desorption isotherm are not occupied in the adsorption isotherm, because the conformational rearrangements necessary for access to these sites cannot occur or at least do not occur on the time scale of the sorption measurements. Some water must enter the protein interior during the early phase of adsorption because exchange of internal peptide protons can be observed at about 0.05-0.07 h in samples hydrated by isopiestic equilibration from the lyophilized protein [29]. Exchange rates reach dilute solution values at about 0.15/? in lysozyme [29], suggesting that the full population of internal water is not achieved until hydration levels reach this value. If the source of sorption hysteresis is due to the failure to recover the full population of buried water molecules as the dry protein is hydrated up to 0.15 h, some estimate of the difference in this population for the adsorption and desorption branches of the sorption cycle can be made from the difference in the hydration levels of the two sorption branches at a given relative humidity. The data for a-chymotrypsin [12] suggest a difference of about 0.02 g water/g protein, which corresponds to about 20-25 moles of water per mole of protein. The full population of internal water must be larger than this to account for the water that enters the protein during the early phase of adsorption to begin to plasticize the protein. This appears to be a very large number of water molecules. Twenty to thirty water molecules would represent about 4-6% of the water required to fully hydrate chymotrypsin. Birktoft and Blow [30] identified 13 buried water molecules in the original crystal structure, but an analysis of this structure by Nedev et al. [31] suggested an internal capacity, based on the volume of cavities accessible to the solvent, for about 50 water molecules. Sreenivasan and Axelsen [32] identified 21 conserved, buried water sites among the serine proteases and proteins with trypsinlike substrate specificity. Sixteen of these sites were common to all the serine protease crystal structures analyzed. Not all these sites were found to be occupied by water in a particular structure. Only those water sites with high occupancies, however, can be reliable identified by x-ray crystallography. In rat trypsin, 28 buried sites were found. The pattern of water occupancy is highly dynamic, and occupancies of individual sites may be quite variable. They are coupled to the rearrangements in protein-free volume and conformation. In lysozyme, all peptide NH protons can exchange with bulk solvent from the native state via a mechanism of catalyst penetration if given sufficient time. Thus, water molecules and exchange catalysts (H+ and OH") must be capable of reaching all peptide groups in the protein, albeit only very rarely in the case of the exchange-stable regions. Before discussing the plasticizing action of buried water and its effect on protein dynamics in any further detail, we turn to a description of the sorption sites and the extent of conformational change observed during hydration.

+1 0

Post a comment