Stabilization of halophilic malate dehydrogenase

Stabilization of halophilic malate dehydrogenase

J. Mol. Biol. (1989) 208, 491-500 Stabilization of Halophilic Giuseppe Zaccait, Malate Dehydrogenase Fabrice Cendrin (C.N.R.S., U.R.A. 1333) Ins...

1MB Sizes 0 Downloads 61 Views

J. Mol. Biol. (1989) 208, 491-500


of Halophilic Giuseppe


Malate Dehydrogenase Fabrice


(C.N.R.S., U.R.A. 1333) Institut Laue Lange,vin 156X, 38042 Grenoble Cedex, France



Nina Borochov

and Henryk


Polymer Research Department Weizmann Institute of Science, Rehovot 76100, Israel (Received 29 September 1988, and in revised form 14 March


Malate dehydrogenase from the extreme halophile, Halobacterium marismortui, is stable only in highly concent’rated solutions of certain salts. Previous work has established that its physiological environment is saturated in KCl; it remains soluble is saturated NaCl or KC1 solutions; also it’ unfolds in solutions containing less than 2.5 M-NaCl or -KCl, salt concentrations which are still relatively high. New data show that the structure of this enzyme can be stabilized in a range of high concentrations of Mg2+ or other “salting-in” ions, also with exceptional prot,ein-solvent interactions. “Salting-in” ions, contrary to stabilizing protein structure, usually favour unfolding. These, and most other results concerning the structure, stability and solvent interactions of the protein cannot be understood in terms of the usual effects of salts on prot#ein structure. In this paper, a novel stabilization model is proposed for halophilic malate dehydrogenase that can account for all observations so far. The model results from experiments on the protein in salt solutions chosen for their different effects on protein stability (potassium phosphate, a strongly “salting-out” agent, and MgCl,, which is “salting-in”), and previously published data from NnCl and KC1 solutions (mildly “salting-out”). Enzymic activity and stability measurements were combined with neutron scattering, ultracentrifugation and quasi-elastic lightscattering experiments. The analysis showed that the structure of the protein in solution as well as tjhe dominant stabilization mechanisms were different’ in different salt solutions in which this enzyme is active. Thus, in molar concentrations of phosphate ions, st’abilization and hydration are similar to those of non-halophilic soluble proteins, in which the hydrophobic effect, dominates. In high concentrations of KCl, NaCl or MgCl,, on the other hand, solution particles are formed in which the protein dimer interacts with large numbers of salt and water molecules (the mass of solvent molecules involved depends on the nature of thr salt but it is approximately equivalent to the protein mass). It is proposed that, under these conditions, the hydrophobicity of the protein core is t’oo weak to stabilize t’he folded structure and the main stabilization mechanism is the formation of co-operative hydrate bonds between the protein and hydrated salt ions. Model predictions are in agreement’ with all experimental results, such as the different numbers of solvent molecules found in thta solution particles formed with different salts, the loss of the exceptional solvent interactions concomitant with unfolding at non-physiological salt concentrations, and the different temperature denaturation curves observed for different salt solutions. The stabilization model is discussedin terms of a structural model proposed previously from solution smallangle scattering data, in which the prot,ein has a core similar to non-halophilic malate dehydrogenase and loops extending int,o the solvent where the exceptional hydration int’eractions t,ake place.


Cl. Zaccui et al. 1. Introduction

Halophilic bacteria live only in extremely concentrated salt environments such as ponds for the evaporation of sea-water or salt lakes such as the Dead Sea, and their cytoplasm is saturated in KCl. Halophilic enzymes are soluble and function in such extreme solvent conditions. Furthermore, many halophilic proteins are unstable and unfold if the salt concentration of their solvent is lowered (e.g. below 2.5 M-KC1 or -NaCl for halophilic malate dehydrogenase, still relatively high salt concentrations if compared to non-halophilic conditions) (Eisenberg & Wachtel, 1987). Although the aim of our work is to understand the molecular mechanisms of adaptation to extreme environments, such studies may also lead to a better general understanding of solvent effects on protein folding. Solvent effects play a crucial role in the folding of polypeptide chains to form active protein structures, and arguably the most important of these is the hydrophobic interaction (Kauzmann, 1959). The stabilization or destabilization of protein structures by high solvent salt concentrations (above molar) is probably due to a modification of the hydrophobic interaction by the presence of the salt ions in the solvent water. Certain salts stabilize protein structures and others destabilize them (von Hippel & Kchleich, 1969). This is the basis of rnany methods of protein fractionation and crystallizatron. Thus, for example, phosphate and sulphate ions are “salting-out” agents that stabilize folded protein st,ructures at high salt concentratIions, and at even higher concentrat,ions cause aggregation and precipitate proteins. Lithium, magnesium, bromide and t*hiocyanate are “salting-in” ions that break up protein aggregates and favour t,he unfolding of single proteins. Sodium, potassium and chloride ions appear to be fairly neutral or mildly salting-out agents. The hydrophobic interaction remains a topic of current interest. It has been shown, in two different studies, that stability differences in mutant proteins are directly related to the hydrophobicity of single substit,uted groups, and that the stabilization is proportional to the reduction of surface area accessible to the solvent upon folding (Kellis et al., 1988; Matsumara et al., 1988). Measurements of protein hydration and salt binding in different solvents have shown that, in “salting-out” conditions. salt ions are excluded from a hydration shell surrounding the protein, thus thermodynamically favouring a minimum surface between solvent and protein; whereas, in salting-in conditions, there is “binding” between the solute molecules and the protein, favouring unfolding (Arakawa & Timasheff, 1984a,b). A consequence of the hydrophobic interaction is that both the enthalpy and ent,ropy changes upon protein unfolding are strongly temperature dependent. For non-halophilic globular proteins, this leads to a stabilization curve (free energy change upon unfolding as a function of t,emperature) that has a maximum close to or below room temperature (Privalov. 1979). A model based

on the solution of liquid hydrocarbons in water (Baldwin, 1986) can account for some of the common quantitative features in the thermodynamic data from the folding-unfolding reactions of non-halophilic globular proteins (Privalov. 1979). The stabilization of halophilic proteins in solvents containing high salt concentrations has been discussed in terms of apparent peculiarities in their composition, i.e. higher proportions of acidic and borderline hydrophobic amino acid residues when compared to their non-halophilic homologues (Lanyi, 1974; Rao & Argos, 1981). The negative charges would compete with the salt for water of hydration, and the relatively low hydrophobicity of its composition would allow the protein to remain in solution even at high salt concentrations. Observations that halophilic proteins are stabilized by salting-out solutes reinforced the grounds for these considerations. Such simple stabilization models are incomplete, however, because they cannot account for the extensive data on the activity, stability, solvent interactions and solution structure of malate dehydrogenase (hMDHt) from Halobacterium marismortui that have been published since these models were proposed (Mevarech et al., 1977; Mevarech & Neumann, 1977; Pundak & Kisenberg, 1981; Pundak et al., 1981: Zaccai et al., 1986a,b: Calmettes et al., 1987). hMDH is significantly more stable in NaC- t,han in KCl-containing solvents (Pundak & Kisenberg, 1981), yet the salting-out effects of NaCl and KC1 are mild and similar (von Hippel & Schleich, 1969). The protein can be stabilized by high concentrations of LX1 (F. Cendrin, unpublished results) or MgCl, (see Results). Mg2+ and Li+ are salting-in ions that usually favour protein unfolding. The solut#ion structures of the enzyme are very similar in NaCI and K(JI solutions and quite different from t’hose of non-halophilic proteins under similar conditions. Neutron-scattering and ultracentrifugation experiments have shown that, in either multi-molar NaCl or multi-molar K(:t solvents, hMDH forms a particle in solution made up of a protein dimer t,hat associates 4099 motes of water and 500 rnoles of salt’ per mote of protein (Pundak et al., 1981: Zaccai et al.. 1986a,b: Calmettes et 01.. 1987). This is 985 g of water and about 0.35 g of NaCl per gram of protein. ;More common. corresponding values for non-halophilic globular proteins are 0.2 to 0.4 g of water per gram of protein and negligible amounts of salt, binding. The hydration values of non-halophilic proteins in non-denaturing conditions can usually be predicted from their composition, independently of tertiary or quaternary structure (Saenger, 1987). This is not the case for hMDH. Its exceptional hydration and salt binding properties are related to its quaternary structure. At low salt concentrations, t)he enzyme dimer dissociates. it loses it,s associated water and salt molecules. and its hydration properties become t Abbreviations used:hMDH, from Halobacterium mnrismortui:

malate dehydrogenase c.d.. circular tfjrhroism,


of Halophilic

like those of non-halophilic proteins (Pundak & Eisenberg, 1981; Zaccai et al., 19863). It was deduced from its small-angle scattering curve that the hMDH part’icle is globular with a larger surface/volume ratio than mitochondrial (non-halophilic) malate dehydrogenase, consistent with protein domains extending out int,o the solvent. The associated water and salt are located at the sameradii as these domains (Zaccai et al.. 1986a; Calmettes et al., 1987). Neutron-scattering, hydrodynamic and activity measurements on hMDH in MgCl, and potassium phosphate-containing solvents are presented in this paper. When interpreted together with the previous data, they provide a new insight into the mechanisms of stabilization of this halophilic protein. Initially, phosphate was chosen because of its pronounced salting-out’ properties and MgCl, was chosen in order to explore the effects of a divalent salting-in cation. Mg2+ is one of the dominant cations in the Dead Sea, with a concentration approaching 2 M (Nissenbaum, 1975). The Mg’+ concentration in the cytoplasm of halobacteria, however, is not known. The main conclusion of the present work is that hMDH is stabilized by different mechanisms in different salt solvents in which it is active. A novel structural model is proposed for its stabilization in its physiological (high concentration of KCl) environment’. In this model, the protein is stabilized by a hydration network in a quaternary structure made up of the protein dimer, water molecules and salt ions. In molar concentrations of potassium phosphate, however, the stabilizat,ion model is predominantly

of t,he salting-out1

type, and the physical

properties of the protein and its solvent interactions are similar to t,hose of non-halophilic globular proteins. Temperature dependence experiments showed different stability behaviour in KC1 and in potassium phosphate solvents in agreement with the proposed stabilization mechanisms.

(a) Knz2/mr was prepared as described (Mevarech 1977: Pundak & Eisenherg, 1981). (b)




Enzymic activity was measured in the following way. Samples (12 ~1) of a 67 fig hMDH/ml solution in 10 mM-


sodium phosphate. 4 M-NaCl (pH 7.0) were added t,o 3 ml of substrate solution (@lo mM-NADH. 0.25 mm-oxaloace-

tate and the indicated KCI, potassium phosphate or MgCI, concentration (pH 7.0)). The enzymic activity was measured at 37 “C by following the oxidation of NA DH by optical absorption measurements at 340 nm, (d)



The stability of the enzyme in a given solvent was calculated by measuring enzymic activity in standard conditions after incubation for 24 h in that solvent. Samples (2 ~1) of 1.7 mg hMDH/ml in 10 rnM-sodium phosphate, 4 M-E&l (pH 7.0). were added to 58 ~1 of 10 mM-sodium cacodylate and salt solut,ion (pH 7.0), to give the appropriate final KCl, potassium phosphate or MgCl, concentration. After 24 h of incubation at 19°C in this solvent, 8 ~1 of solution was added to 2 ml of substrate solution (10 mn-sodium cacodylate. 4 M-IL'aCI, @12 mM-NADH, 0.25 mw-oxaloacetate (pH 7.0)) and the enzymic activity measured as usual. (e)


centrifugation light-scattering



Analytical centrifugation and quasi-elastic light scattering experiments were performed as described for studies in NaCl and KC1 solvents (Pundak & Eisenberg. 1981; Zaccai et al., 1986a; Calmettes et al.; 1987). The sedimentation coefficient determined in a given solvent by the Beckman model E ultracentrifuge is: where (aplan,) is the mass density increment of the solution (at constant chemical potential of salt and water) per mole concentration of protein n2 (mol/ml), ‘VA is the Avogadro constant (Calmettes et al., 1987). and S is the friction coefficient. The mass concentration c2 (g/ml) is given by: The diffusion light-scattering

coefficient, is:

II, determined

~~20,sol= MT/f. The values of (dp/dn,) for each condition from t,he s20,so,/Dzo,so, ratio:

by quasi-elast’ic

were calculated

= (~~20,soI/~~20.~~l)RT%. (f)

In MgCI, solvents, the bufYer was 10 m&i-sodium cacodylate (pH 7.0). Control experiments with NaCl and KCI in cacodylate buffer gave the same results as for the phosphate buffer in which the enzyme had been studied so far. The potassium phosphate solvent was a mixture of KH,PO, and K,HPO, to give pH 7.0. In all cases, the pH was adjusted to 7.0 after the salt was brought to the appropriate concentration. (c) Activity



2. Materials and Methods

Enzyme Neumann,



All neutron data were collected at 20°C on the D 11 small-angle camera at the Institut Laue Langevin. Grenoble? as described (Calmettes et al.. 1987). Sample mass concentrations were determined by taking an absorbance va,lue of @8 cm2/mg at 280 nm (Mevarech et al., 1977); they were &8 mg hMDH/ml in 1 M-MgCl, 10 IIIMsodium cacodylate (pH 7.0), 57 mg hMDH/ml in @6 M-MgCl, and 42 mg hMDH/ml in 0.47 M-MgCl,, in the same buffer. The mass concentration of protein in 1.5 Mpotassium phosphate (pH 7.0) was 63 mg hMDH/ml. The scattered neutron intensity was interpreted in t,he range of validity of the Guinier expansion: Z(Q) = I(O) exp ( -RiQ213). where & = (47~ sin 0)/n where 28 is the scattering angle and ,I the wavelength of radiation, R, is the radius of gyration of contrast in the particle and Z(0) is the forward scattered intensity. Z(0) can be expressed in terms of the

sc&attering density increment at constant cshrmical potential of water and salt (Kisenberg. 1981: (‘almettes rt nl.. 1987):

Table 1 7’h,e corlstants

water As done previously (Zaccai et (A., 1986a), the scattered intensity was put on an absolute scale by normalizing by the incoherent scat,tering of 1.00 mm of water (Jacrot & Zaccai, 1981), and the neutron-scattering density increment (8p,,,/&,) was calculated from I(0) and the mass concentration, c2, with a value of M, = 87,000 g/mol for the molar mass of the protein. (g) Interpretation

of the density increments

The mass or scattering density increment of a solution due to a macro-molecular solute can be expressed in terms of t.he mass of the macro-molecule, its volume, and solvent interaction parameters (Eisenberg, 1981). In the special case, when each density increment varies linearly with the corresponding solvent density, a model particle can be proposed for the macro-molecule in solution. This particle has an invariant volume and associates or “binds” constant amounts of solvent water and salt (Tardieu et nZ.. 1981: Zaccai et al.. 1986a). Then, the density increments can be written: (dp/ih,) (ap,J&)

= M, +N,M, = M,b, +il’,k’,b,

+ N3M3 -p’M, +ilr,M,b-p”,M,

V,,,. V,,,.

where &fi are the molar ma.sses of water (i = 1), protein (i = 2) and salt (; = 3). NL are the numbers of moles of water or salt associated to one mole of protein in the particle, b, are the neutrorl-scattering lengths per gram of the different components. p” and pi are the density and scat,tering length densky of the solvent. respect.ively. and A4, r;,, is the volume of 1 mol of the particle:

I’rotein MgC’l,

required density

to interpret increments

M, = 18 g/m01 M, = 87,000 g/m01 ‘If, = 9.53 g/m01

the scattering

6, = -562 x IO9 cm/g 6, = 148 x 10’ (-m/g 6, = 15.4 x 109 cm/g

I’ot.assium phosphate (400/b KH,PO,, 600/, K,HPO, for pH 7.0) M, = 158~8 g/mol 6, = 11 x 109 cm/g

of potassium phosphate concentration in the solvent is very similar to those already observed for NaCl or KC1 (Fig. l(a)). It has a maximum at about O-5 M and decreases to about 50% of that value at the highest salt concentrations. The enzyme is stabilized by potassium phosphate. The stability curve for the enzyme in potassium phosphate solvents is shown in Figure 1(b). It has a shape similar to those for NaCII or KCI, with the minimum concentration, at which the protein is stable, shifted to a lower value in the case of potassium phosphate. The time dependence of the diffusion coefficient of’ hMDH, measured by quasi-elastic light-scattering in potassium phosphate solvents, showed that the solutions were unstable helow cwnwntrations of

wherca Ci is the partial specific volume of component i. The slopes of (dp,‘&,) WJ~SWSp* altd of (d&Y+) WTSUS pi should be identkal and qua1 to M, I’,,,. On the other hand. the intercept of each of these lines with thr zt‘ro tlensjtu axis depends on the relative drlkties or scattering tlrnsltlrs of thr different c-omponrnts of the particalt,. These are:

for mass and neutron-scattering length density. respectively. The comglementaritJ; of t,he neutron and ultrawntrifugation experiments arises from the fact that the 2 tlquat,ions above are independent, (in water solvent. for example: b, is negative. whereas N,M, is of course positive) and they can be solved for ,Irl and N3 (Zaccai et al., 1986a; Calmet,tes et al.. 1987). The constants required for the calculation were obtained from the chemical compositions and molar masses of the different solution components: they are given in Table 1.

Figure 1. (a) Enzymir activity of hMDH, in different

3. Results (a) Activity

and stability of hMI)H in h.iyh concentrations of pvotassiu,m phosphate

The enzyme is significantly more active in high concent’rations of potassium phosphate than in KCI solvents. The activity profile of hMDH as a function

concentrations of potassium phosphate and K(‘1. (1)) Rtahility measurements in different. concentrations of potassium phosphate and KU. The enzymir activity relative to its tnaximum value, as mrasured in standard buffer. after incubat,ion for 24 h in the different salt, wnwntrations given on t,he .x-axis. For both (a) and (b), the mrasurements were performed as described in ,VIaterials and Met,hods.


of Halophilic

about 1-OM in that salt. On the other hand, aggregation or crystallization occurred at a concentration of 2,5 M-potassium phosphate between the third and the fifth day of incubation (Hare1 et al., 1988). (b) Activity

and stability

of hMDH

in MgCY,

High concentrations of MgCl, inhibit hMDH activity. The activity of hMDH is shown in Figure 2(a) as a function of MgCl, concentration in the solvent. This salt reduces enzyme activity at all concentrations when added to the enzyme in 3 M-KCl, and beyond concentrations of about 10 mM when it is the only salt added to the buffer. Kinetic measurements have shown MgCl, to behave as a competitive inhibitor of the enzyme with respect to oxaloacetate (results not shown). The stability behaviour (or recovery of activity after incubation) of hMDH in MgCl, solvents is very different from that in the other salt solvents studied so far. Over a period of more than one day, the enzyme appears to be less stable in any concentration of MgCl, than in high concentrations of either KCl, NaCl or potassium phosphate. The time dependence of sedimentation and diffusion coefficients in MgCl, solvents showed a structural instability of the enzyme in these conditions that became


no KC1






apparent after about 24 hours. The structural instability paralleled the non-recovery of activity discussed above. Physical measurements in MgCl, solvents, therefore. were always performed on freshly prepared solutions, except when the time course of an effect was of interest. The enzyme can be stabilized by MgCl,, however, for the 24 hour incubation period used as the stability test. The stability of the enzyme as a function of MgCl, concentration in the incubat,ion solvent is plotted in Figure 2(b). The bell-shaped stability curve obtained after 24 hours of incubation suggeststhat the active structure of the protein can be maintained only in a certain range of MgCl, concentrations. This observation is very different from those in all the other salt solvents studied, for which hMDH &ability always increases beyond a minimum salt concentration (Pundak k Eisenberg, 1981, and Fig. 1). The circular dichroism (c.d.) spectra of the protein measured in high and low concentrations of NaCl showed considerable loss of secondary structure at low salt (Mevarech & Neumann, 1977). The c.d. spectra in 063 mMgC1, (not shown) are similar to those in 4 tif-SaCI. This is consistent with a folded structure in these conditions. The c2.d.spectra after incubation in 2.3 M-MgCl, (also not shown), on the other hand. are similar to those of the protein in low NaCl concentrations. The non-recover>- of activity at the higher MgCl, concentrations. therefore. is correlated with loss of struc>ture.

(a) (r) Sedimentation, diffusion n Putron-senttering

and smnll-nngk rPsu1t.s

Small-angle neutron-scattering, analytical rentrifugation and quasi-elastic: light-scattering rxperiments were performed on hMDH in MgCI, and potassium tions in 04



phosphate. the solvent.

for different salt The scatjtrrinp


and mass density increments were derived from the data and interpretrd in terms of equivalent pazticles, in which are associat)edprotein, water and salt molecules, as described in Materials and Methods. Sedimentation data could not be obt,ained in high czoncrntrations of potassium phosphate. because of the high densit)y of the solvent. Also, only the condition with 1.5 M-potassium phosphate c~~uld he measured by nrutron-sc,attering, because of instahIitS at lower cboncentrations and aggregation iLt



Figure 2. (a) Enzymic activity of hMDH, in different cLoncentra.tions of M&J,, in the presence and absence of 3 M-KU. (b) Stability measurements in different concentrations of MgCI,. The enzymic activity relative to its maximum value, as measured in standard buffer, after incubation for 24 h in the different salt, concentrations given on the .r-axis. For both (a) and (b). the measurements were performed as described in Materials and Methods.

higher potassium phosphate concentrations. It was not possible, therefore. t,o derive -VI and N,, separately. A value of 140( kO.05) x lOl5 cemjmolwas calculated for (dp,/&,) from I(0) in I.5 M-potassium phosphat,e (Fig. 3). Assuming E2= 0.75 ml/g (Zaccai et (II.. 1986a.b). this value of (ap,/&,) is not consistent with the particle of exceptional salt and water binding



for hMDH

in NaCl


KC1 solvents ((‘almettes et al., 1987). On the other hand, it is consistent with N, = 2000( + 1000) mol of wat#er per mol of prot)ein and -V3 z 0. This corresponds

to a more “nsual”


(as observed



‘l’hr ratio of associated salt to associated water in the part,icle corresponds to a 7 molal solution for IVaCI or KU, and to a 1.6 molal solution for MgCI,.

po(g/ml) I.0





I.2 I



(d) The very low resolution structures of the solution particles in MgCl,, potassium phosphate, KC1 and NaCl














Figure 3. Mass (top and right-hand y-axis) and neutron-scattering (bottom and left-hand y-axis) density increments for hMDH as functions of the respective solvent densities. Data for NaCl and KC1 are from Clamettes et al. (1987).

non-halophilic proteins in salting-out conditions) of approximately @4 g water/g protein. The density and scattering density increments derived for MgCl, are shown in Figure 3, together with previous data for NaCl and KC1 (from Calmettes et al., 1987), as functions of the corresponding solvent density. Tt is seen that for each salt, in the concentration range studied, hMDH can be described in terms of an equivalent particle. The same particle describes the NaCl and KC1 data, but the particle in MgCl, is clearly different. The compositions and volumes of these particles calculated from (ap/&,) and (ap,/%,), by using the values in Table 1, are shown in Table 2. The solution particle in MgCl,, therefore, also has exceptional water and salt binding properties, but it is different from the particle in NaCl or KC1 solvents. The enzyme forms a well-defined particle in MgCl, concentrations between about 0.3 M and about I.3 M, which binds the same amount’ of water molecules as the equivalent particle in NaCl or KCI, but only about, a quarter as many salt molecules.

The diffusion coefficient of the enzyme extrapolated to zero time and corrected for viscosity is found to be 5.70( [email protected]) x lop7 cm’/s in 0.5 M to 2.5 M-potassium phosphate. The same value is found in 0.5 M to 1.0 M-MgCl,. This value is slightly higher than the value of 5.0 x lop7 cm2/s found in NaCl solvents (Pundak et al., 1981). These diffusion coefficients correspond to a hydrodynamic radius of about 40 a (1 ,& = 0.1 nm). The small-angle neutron-scattering curves of the active particles in MgCl, and potassium phosphat,e are similar to those observed in NaCl or KC1 (Zaccai et al., 1986a,b; Calmettes et al., 1987). They correspond to those of a globular particle with a radius of gyration of 31( + 1) A and a relatively large specific surface. This agrees with the structural model proposed for the particle in solution with NaCl or KCI, having a compact protein core and domains extending outward into the solvent (Zaccai et al., 1 YSSa,b). The corrected diffusion coefficient values from quasi-elastic light-scattering, when combined with the total volumes (given in Table l), are also consistent with a globular rather than an elongated shape for the particles in NaCl, or MgCl, solvents. It is interesting to note that the hydrodynamic radius of 40 w is fully consistent with a “spherical” particle of radius of gyration 31 8. (e) The temperature dependence of the stability hMDH is different in KC1 and in potassium phosphate solvents


Stability measurements as a function of incubation temperature are shown in Figure 4. As usual, the measurements reflect the proportion of active protein after incubation in the stated conditions. The stability of the protein in KC1 (and NaCl) increases with decreasing temperature down to

Table 2 Composition hMDH Protein Particle Particle Protein “Bound” “Bound” Molality


of the hMDH

solution particles, calculated mass density increments

in (M2 g/mol) specific volume ( V,,, cm3/g) volume (A3) (M, V,,/NA) specific volume (2)* cm3/g) water N, (mol water/m01 protein) salt N3 (mol salt/m01 protein) of “bound” salt (N3 x lOOO)/(N, x MI)

t The sum of the protein and hydration volumes. $ Since mass density increments could not be determined order to interpret the neutron scattering density increment.

KC1 (1-4


87:OOO 1.79 259,000 076fO.01 4080 + 400 520 f 60 7.2 M

for potassium


porn M&l,


(0.5-1.0 87,000 1.71 247,000 @74+001 4000+400 120+30 1% M


a value






87,000 166,000~ o-75$ 2000f 1000 Negligible

of G2 = 075








O-9 I-0


of Halophilic


I.4 I.6
















Figure 4. Enzymic activity of hMDH in standard buffer after incubation for 24 h in a given salt concentration various temperatures.


details are given in Materials

- 10°C. For all potassium phosphate concentrations, however, it reaches a limiting value at a certain temperature depending on salt concentration (a higher temperature for higher concentrations).

4. Discussion (a) The structural

model proposed


A structural model for hMDH in NaCl solutions was proposed from small-angle neutron- and X-rayscattering data (Zaccai et al., 1986a). The present small-angle scattering and diffusion coefficient data show that t’he very low resolution structure of the hMDH particle in MgCl, or potassium phosphate solutions is very similar to that in NaCl. In the model, it is globular with a core and loops extending outwards to give a larger surface to volume ratio than in a non-halophilic malate dehydrogenase that was studied also (Calmettes et al., 1987). Small-angle solution scattering experiments with contrast variation can show the existence of hydration shells around macro-molecules (Lehmann & Zaccai, 1984; Zaccai & Xian, 1988). In the case of hMDH in concentrated NaCl solutions, X-ray and neutron contrast variation experiments showed that the solvent interactions occurred at the same radii as the loops and that the protein was not surrounded by a salt-free hydration shell (Reich et al.. 1982; Zaccai et aZ., 1986a). The slightly lower value found for the volume, and slightly higher value of diffusion coefficient, found for hMDH in MgCl, and potassium phosphate solutions are consistent with the lower values of “bound” solvent molecules in these conditions. (b) The stability

results are related unfolding

to the rate of

The stability of the enzyme in a given set of conditions was estimated, after 24 hours incubation in these conditions, by measuring its residual activity in a standard buffer. Physical measurements


and Methods.

(circular dichroism, neutron-scattering and quasielastic light-scattering) showed that, loss of activity is correlated with protein unfolding. Since the activity measurement in the standard buffer took only a few minutes, the stability measurement was, essentially, a measurement of the reversible or irreversible rate of unfolding of the protein in the given conditions. As was written by Lanyi (1974): “What such experimental data reveal are the properties of the forces which act as barriers to the disruption of the structures”. The data presented in this paper show that the dominant stabilizing forces for hMDH are different in the different salt solvents. (c) The stabilization of hMDH in potassium phosphate is similar to that of a non-halophilic protein

The activity. stability and structural data from hMDH in potassium phosphate solvents are compatible with what is expected from a non-halophilic protein in similar salting-out conditions. In the structural model proposed by Zaccai et al. (1986a), the water and salt association sites were located in protein loops at the periphery of the particle. The neutron-scattering curves of the protein in NaCl and potassium phosphate are very similar. It is likely, therefore, that the structure in potassium phosphate also has these loops, but they do not participate in the exceptional hydration interactions observed in NaCl or KCl. The results of the t’emperature dependence experiment’s (Fig. 4) are also in agreement with the suggestion that the stabilization mechanismsof hMDH in molar concentrations of potassium phosphate are similar to those of non-halophilic proteins. Unlike the stability of the enzyme in KCl, which rises when the temperature is lowered, the stability in potassium phosphate reaches a limiting value. The stability of non-halophilic proteins usually reaches a limiting value at temperatures close to or below room temperature (Privalov, 1979). This is becausethe enthalpic terms that favour the folded conformation at, iower

U. %accai


temperatures are counterbalanced by the entropy terms (due to the hydrophobic effect) that favour unfolding as the temperat,ure is lowered.

et, al

in potossium


(d) The stability data and solution structure in, MgCl, solvents suggest a stabilization model for hMDH The protein is inactive in high concentrations of MgCI, (Fig. 2). Nevertheless, the hMDH structure can be stabilized in a given concentration range of this salt (Fig. 2). Low concentrations of MgZf (IO-’ M) have been shown to act as a block to the renaturation of the fl-galactosidase tetramer from Escherichia coli by promoting incorrect ionic bonding between unfolded subunits (Ullman & Monod, 1969). This mechanism, however, is unlikely to occur in halophilic organisms exposed to considerably higher concentrations of monovalent and divalent salts. Furthermore, in the particular caseof hMDH, it could not account for the observation that MgCI, is stabilizing only when its concentration lies in the range between -0.3 M and - 1.3 M. The structure of the hMDH particle in solution, in the range of MgCl, concentrations in which it is stable, shows that the protein binds as many water molecules as in KCl or NaCl solvents, but’ significantly fewer salt molecules (Table 2). The observation that the upper limit of the MgCI, concentration range in which hMDH is stable corresponds approximately to the sah/water ratio associated with the protein in the solution particle (taking into account’ the fact that. due in part, to their divalent nature, Mg ions can co-ordinate more water molecules than either K+ or Na+ (Enderby et al., 1987)) suggeststhe following model for t’hr stabilization of the protein. (e) A protein-hydrated salt stabilization proposed for the hMDH solution particle NaC2 and MgCl, solvents

,model is in KCI.

In the stabilization model proposed. the halophilic protein, in NaCl, KCI or MgCl, solutions, forms a complex with hydrated salt cations and anions (Fig. 5). The weak hydrophobicity of the protein in NaCI, KCI and especially in MgCI, (which is a salting-in salt) solvents maintains its solubilit,y even in high salt concentrations but it) is not sufficient to stabilize the active structure. The protein also requires hydration interactions to be stable. Because it cannot compete with the highly concentrated salt solvent for water purely through its composition, it has evolved a quaternary structure that’ can compete, by co-ordinating hydrated salt ions at higher local concentrations than in the surrounding solution. The enrichment’ in acidic amino acids observed when the compositions of many halophilic proteins are compared to those of non-halophilic homologous proteins (Eisenberg & Wachtel,



a possible


Carboxyl groups in the outer protein loops could cooperatively nucleate and co-ordinate a network of

in KC1

Figure 5. Schematic representation of hMDH solution structures. The active structures have 2 parts: a catalytically active core, conceivably similar to that in nonhalophilic MDH, and protruding loops, required for stabilization in KCl, XaC1 or MgCI, solvents. In potassium phosphate, the protein dimer is stabilized by the hydrophobicity of the core and the protruding loops are disordered. Tn KCI (or NaCI). the protein is stabilized by the interaction of the loops in a protein-water-salt hydration network. In MgCI,, a similar structure exists with the same amount of water molecules co-ordinated by fewer salt ions. In low salt, the protein is unfolded and it,s hydrat’ion is like t,hat of non-halophilic proteins.

water and salt ions. Hydrate bonding in such a complex would be more structured than in the solvent because of the anchoring of the carboxyl groups in the protein. The resulting lower entropy. however, could be compensated for by an energy term favouring hydrate bonds in the complex over bonding in the solvent because hydrogen bonds to carboxyl groups are usually shorter (of higher binding energy) (Vinogradov & Linnell, 1971). Through the binding of hydrated salt’ ions. therefore, water molecules would be associated with the protein structure with different local salt, concentrations depending on the hydrate interactions of the particular salt. These local concentrations are observed to be 7.2 molal in NaCl and KCl and 1% modal in MgCl, (Table 2). These values are fu11y compatible with what is known about the hydration of these ions. There are data showing that Mg2+ can co-ordinate many more water molecules than either


qf Halophilic

or K+ (Enderby et al., 1987). It should be noted that the salt binding experiments measure the contribution of both the salt ions, and that Cl- in solution also has a well-defined hydration shell (Enderby et al., 1987). In terms of the structural model (Zaccai et al., 1986a), the hydration interactions would take place mainly in the loops, and the hydrophobic interaction would t,ake place mainly in the protein core. Because of their transient nature, hydration networks can be observed only in crystal structures, under special conditions, when the occupancy of well-ordered sites is high. At present, there are no examples in protein structures of networks similar to the one suggested for hMDH. Different hydration networks have been proposed to exist in the high resolution structure of vitamin B12 (Savage, 1984), and co-operative hydration networks involving ions and bonding to macro-molecular groups have been “seen” and suggested to be important for the stabilization of certain DNA structural forms (Saenger, 1987). Na+

(f) The proposed all experimental

stabilization model can account for observations on the structure and stability of hMDH

The model accounts for the exceptionally large amounts of water and salt associated with the protein in certain conditions (Pundak et al., 1981; Zaccai et al., 1986a; Calmettes et al., 1987). Unlike the case of most non-halophilic proteins for which hydration can be predicted from amino acid composition (Saenger, 1987), hydration and salt binding in the hMDH model are closely associated to tertiary or quaternary structure. The observed loss of bound water and salt upon unfolding (Pundak & Eisenberg, 1981; Zaccai et al., 19866), therefore, is as would be expected. The model also accounts for the stability of the structure in NaCl and in KC1 (despite their being weak “salting-out” agents), and for the fact that the protein is significantly more stable in NaCl than in KC1 (Pundak & Eisenberg, 1981). Hydration interactions are stronger for Na+ than for K+ (Enderby et al., 1987). The instability of the hMDH structure below a threshold salt concentration or above a maximum salt concentration (for MgCl,) can now be understood in terms of competition between protein and solvent for water and salt ions. At low salt, the chemical potential of the salt in the solvent is too low for the protein-water-salt complex to be formed. At high salt, as the solvent salt concentration approaches that in the complex, it is the chemical potential of solvent water that becomestoo low. The hydration water would tend to “flow” out from the complex into the solvent, thus destabilizing the protein structure. This is especially striking in MgCl, (which is soluble beyond 1.6 molal, the local salt concentration in the complex) where the protein is unstable beyond about 1.3 M (Fig. 2). lt is possible, however, that the destabilization in high MgCl, is also due to a salting-in effect of this salt on the




protein core, which counterbalances the salt hydrate stabilization. Since the protein would have evolved its-quaternary structure to bind hydrated K+ and Cl- in its physiological environment, the inherent instability of the structure in MgCl, might be due to a non-optimum geometry for the binding of hydrated Mg2+ as well as to the salting-in effects of this ion. In high concentrations of KCl, conditions similar to its physiological environment, the enzyme is protected from high salt destabilization because KC1 is not soluble beyond about 4 M, and its concentration in the complex is 7 molal KCl. Tn potassium phosphate, hMDH is active and it is stabilized by a classical salting-out mechanism very different from the stabilization mechanism in KCl. Given an appropriate environment, therefore, it can also behave like a non-halophilic protein. According to the model, the hydrophobicity of the protein core in high concentrations of salting-out potassium phosphate is sufficient to stabilize the active enzyme. The protein has the hydration expected from its composition. The outer domains in its structure participate in “normal” hydration interactions according to their amino acid composition, and are not involved in any specific protein hydrated-salt network. These loops are probably “disordered”, therefore (i.e. their conformations are different in different particles). (g) Th,e model qualitatively predicts the temperature dependence of the stability

The temperature dependence of the stability of hMDH in different solvents can be predicted from the model. The stability of non-halophilic globular proteins reaches a limiting value as the temperature is lowered (Privalov, 1979). Similar behaviour is expected for hMDH in potassium phosphate. On the other hand, according to the model, the hydrophobic effect does not contribute appreciably to the stabilit’y of hMDH in high concentrations of KCl, NaCl or MgCl,. The breaking of the chemically co-ordinated complex would lead to a positive enthalpy change, even at low temperatures. The stabilit)y, therefore, is expected to continue increasing as the temperature is lowered, in these salt conditions. The observed temperature dependence of the stability in the different salt conditions (Fig. 4) is in agreement with the above predictions. The stability behaviour of hMDH in potassium phosphate is similar to that, of non-halophilic globular proteins, and the continuing increase in stability with decreasing temperature, observed down to - 10°C in KC1 and NaCl is like the behaviour of a nonhalophilic protein at high temperatures, where stability is dominated by a positive enthalpy difference between the unfolded and folded states (Privalov, 1979). (h) How can the model be tested further? Crystals of hMDH have been obtained in potassium phosphate (Hare1 et al., 1988) but. in this salt,

t.he outer loops of the protein are not expected t’o participate in the stabilization interactlions with water and salt molecules. Different crystallization conditions are being pursued. conditions in which the hydration network is likely to be maintained. Studies of the cloned hMDH gene are now in The amino acid progress in our laboratories. sequence of the protein will be compared with that of other malate dehydrogenases, and explored for motifs that may give rise to potential water and salt binding loops.

We are very grateful to Bernard Jacrot and Mogens Lehmann for many discussions and a critical reading of the manuscript, and to the editor and referees for helpful comments that, led to a strengthening of the arguments in the paper. References Arakawa, T. & Timasheff, 5912-5923. Arakawa, T. & Timasheff,

S. N. (1984a). Biochemistry,


S. Npu’.(1984b). Biochemistry,




R. L. (1986). Proc.







Calmettes, Chem.

Eisenberg, Eisenberg,


H. & Zaccai, G. (1987). Biophys.

Matsumara, M.. Becktel, W. J. & Mathews, 13. W. (1988). Nature (London,), 334, 406&410. Mevarech, M. & Neumann E. (1977). Biochemistry, 16, 3786~-3791. Mevarrch, M., Eisenberg, H. & Neumann, E. (1977). Biochemistry, 16, 3781-3785. Nissenbaum. 4. (1975). Microb. Ecol. 2, 139-161. Privalov. P. T,. (1979). Advan Protein Chem. 33, 167-241. Pundak, S. & Eisenberg, H. (1981). Eur. J. Biochem. 118. 463.-470. Pundak. S., Aloni, H. & Eisenberg, H. (1981). Eur. J. Biochem. 118, 471-477. Rao, M. .J. K. & Argos. P. (1981). Biochemistry, 20, 65366543. Reich, M. H.. Kam, Z. & Eisenberg. H. (1982). Biochemistry, 21, 5189-5195. Saenger, W. (1987). Annu. Rev. Biophys. Biophys. Ohem. 16. 93-l 14. Savage, H. (1984). In Water Science Reviews (Franks, F., ed.), vol. 2. pp. 67-149, Cambridge University Press, Cambiidgr. Tardieu, A.. Vachette. P., Gulik, A. & Le Mairr, M. Ullman,

H. (1981). Quart. H. & Wachtel,





Vinogradov, Bonding,

York. von Hippel,




J. (1969). Biochem.



35, 35.

S. N. & Linnell, R. H. (1971). Hydrogen Van Nostrand Reinhold Company. New

T. (1969). In Structure and (Timasheff, S. N. & Fasman, G. D., eds), pp. 417-574. Marcel Dekker Inc., New York. Zaccai, 0. & Xian, S. Y. (1988). Biochemistry. 27. 13161320. Zaccai. G., Wachtel, E. & Eisenberg, H. (1986a). J. Mol. Biol. 190, 97-106. Zaccai, G., Bunick, G. J. & Eisenberg, H. (1986b). J. Mol. Biol. 192, 155-157. Stability

784-786. Edited


A. & Monod,






P., Eisenberg,

Biophys. 14. 141-172. E. J. (1987) Annu. Rev. Biophys. Biophys. Chem. 16, 69-92. Enderby, J. E., Cummings, S., Herdman, G. J., Neilson, G. W., Salmon, P. S. & Skipper, N. (1987). J. Phys. Chem. 91, 5851-5858. Harel, M., Shoham, M., Frolow, F., Eisenberg, H.. Mevarech, M., Yonath, A. & Sussman, J. L. (1988). J. Mol. Biol. 200, 609-610. Jacrot, B. & Zaccai, G. (1981). Biopolymers, 20, 24132426. Kauzmann, W. (1959). Advan. Protein Chem. 14, I-63. Kellis, J. T., Jr, Nyberg, K., Sali, D. & Fersht, A. R. (1988).

Larlyi. ,J. K. (1984). Bacterial. Rev. 38, 272-290. Lrhmann. MM. S. & Zaccai. G. (1984). Biochemistry.

V. Luzzati

P. H. & Schleich, of