Biomembrane Phospholipid–Oxide Surface Interactions: Crystal Chemical and Thermodynamic Basis

Journal of Colloid and Interface Science 252, 309–319 (2002) doi:10.1006/jcis.2002.8448

Biomembrane Phospholipid–Oxide Surface Interactions: Crystal Chemical and Thermodynamic Basis Nita Sahai1 Department of Geology and Geophysics, 1215 W. Dayton Street, University of Wisconsin, Madison, WI 53706 Received January 4, 2002; accepted April 30, 2002

Quartz has the least favored surface among many oxides for bacterial attachment and for lipid bilayer or micelle interactions. Tetrahedrally coordinated crystalline silica polymorphs are membranolytic toward liposomes, lysosomes, erythrocytes, and macrophages. Amorphous silica, the octahedral silica polymorph, (stishovite), and oxides such as Al2 O3 , Fe2 O3 , and TiO2 are less cytotoxic. Existing theories for membrane rupture that invoke interactions between oxide surfaces and cell membrane phospholipids (PLs) do not adequately explain these differences in membranolytic potential of the oxides. The author presents a crystal chemical, thermodynamic model for the initial interaction of oxide surfaces with the quaternary ammonium component of the PL’s polar head group. The model includes solvation energy changes and electrostatic forces during adsorption, represented by the dielectric constant of the solid and the charge-to-radius ratio of the adsorbing solute. The nature of oxide–solute interactions compared with oxide–water, solute–water, and water–water interactions determines the membranolytic activity of the oxide, where the solute is TMA+ , the quaternary ammonium moeity. Significant membrane rupture, as on quartz, requires unfavorable adsorption entropy (Sads,TMA+ < 0) to maximize disruption of normal membrane structure and requires favorable Gibbs free energy of exchange between TMA+ and the ambient Na+ ions (Gexc,TMA+ /Na+ = Gads,TMA+ − Gads,Na+ < 0) to maximize the extent of membrane affected. For amorphous silica, Sads,TMA+ > 0, so disruption of structure is limited, even though Gexc,TMA+ /Na+ is <0. Stishovite and other oxides have Sads,TMA+ < 0, but now Gexc,TMA+ /Na+ is >0 at the acidic to circumneutral pHs of cellular and subcellular organelle fluids. The model predicts the correct sequence of membranolytic ability: quartz ≥ amorphous SiO2 > Al2 O3 > Fe2 O3 > TiO2 . The model thus explains the relatively poor adhesion of bacterial cells to quartz and the lack of quartz as a biomineral. It is proposed that one function of extracellular polymeric substances exuded by bacteria is to render mineral surfaces more hydrophilic. C 2002 Elsevier Science (USA) Key Words: quartz; silica; stishovite; bacterial adhesion; hemolysis; silicosis; solvation; adsorption; entropy.

INTRODUCTION

Interactions of mineral surfaces with cell membranes and extracellular polymeric substances (EPS) ultimately occur at 1

Fax: (608) 262-0693. E-mail: [email protected]. 309

the molecular level, with membrane phospholipids (PLs) and proteins. Many geomicrobiological, industrial, environmental health and biomedical processes involve such interactions. Examples include bacterial adhesion to mineral substrates, biofouling of filters in wastewater treatment plants, oil recovery from petroleum reservoirs, biobeneficiation of ores, biocompatibility of ceramic-based prosthetic devices, and the ability of inhaled dust particles in the lungs to rupture cell membranes and cause fibrogenesis. Bacterial adhesion is weaker and attachment densities are smaller on quartz than on other oxides such as corundum (Al2 O3 ), hematite (Fe2 O3 ), and rutile (α-TiO2 ), amorphous silica gels, and amorphous silica glasses (1–4). The amount of EPS exuded by certain strains of bacteria is also substrate dependent (5). Quartz and other tetrahedrally coordinated crystalline silica polymorphs are membranolytic and cytotoxic, whereas amorphous tetrahedral silica (glasses and gels), the octahedral silica polymorph (stishovite), corundum, hematite, rutile, and anatase (β-TiO2 ) are less harmful (6–9). The approximate sequence for the membranolytic effect of oxides is tridymite? ≥ cristobalite > quartz > coesite ? ≥ amorphous silica > corundum > stishovite > rutile (6–9). Clearly, the solid substrates with which cells interact affect the nature of the interactions, and both the chemical composition and the crystal structure of the solid are significant. A model remains to be constructed to account for the differences among the oxides and among the different silica phases with regard to their interactions with cell membranes. Oxide–PL interactions fundamentally depend on electrostatic (including charge and dipole), polarization, and H-bonding forces (10). The attachment of PLs to oxide surfaces involves several steps where electrostatic attraction is an early step followed by other forces. The electrostatic step can be examined using the solvation and electrostatic (SE) model (11–19). The SE model accounts for solvation energy changes during adsorption and for electrostatic forces driving the adsorption reaction. The other forces have been given different names such as van der Waals, electric double-layer, solvation, Lewis acid base, and “steric” and “hydrophobic” force, and they have been modeled in various ways such as the DLVO model, the extended DLVO model, and the hydrophobic interfacial model (20–22). These all are excellent approaches to modeling PL– and bacterial–mineral surface interactions, although the 0021-9797/02 $35.00

 C 2002 Elsevier Science (USA)

All rights reserved.

310

NITA SAHAI

specific nature of the solid is usually given minimal attention. The SE model was developed originally for oxide surface–small ion interactions and cannot pretend to represent whole PL–oxide surface interactions. The motivation is not to present a competing model. Rather, the idea is to take a first stab at representing the specificity of the oxide in interactions with PLs. The premise is that even though electrostatic attraction occurs predominantly during the early stage of attachment, the consequences for membrane structure are carried through to later stages of attachment. A large number of oxides have been characterized for their membranolytic and fibrogenic potential with a view to explaining the silicotic effect of inhaled oxide dust particles that are retained in the respiratory system of the body. Therefore, I use the available data on membranolytic potential of oxides to understand mineral–PL interactions in general. BACKGROUND INFORMATION

Phospholipids consist of a long hydrophobic hydrocarbon chain attached to a hydrophilic polar head group. The polar head group contains phosphoryl and, often, secondary or quaternary amine moities. A popular theory for the membranolytic action of certain oxides is that the surface interacts with portions of membrane proteins or with the phosphate or amine moeities of the PLs’ polar head group. The interactions supposedly result in conformational changes of the PL bilayer, ultimately leading to disruption of the membrane. In the case of quartz, the interaction is supposed to involve electrostatic forces between the positively charged quaternary ammonium moeity of the PL’s head group and the negatively charged surface sites (23–25). A variant of this idea invokes H-bonding interactions between charge-neutral silanol surface sites (>SiOH) and the negatively charged –PO− 4 moeity of the PL head group or negatively charged membrane proteins (26–28). A brief review of the silicosis literature relevant to the following discussion is provided in Appendix A (29–47). Alternative mechanisms that do not rely on surface interactions have been proposed. The most important alternative theory involves the toxic action of free radicals and is discussed in Appendix B (48–61). In the current article, the SE model for monovalent ion adsorption at oxide surfaces is extended to interactions with components of cell membranes. The model is used to examine the H-bonding and electrostatic-bonding theories of membranolysis. The tetramethyl ammonium ion is used as a proxy for the PL head group. Although this appears to be a long extrapolation, the use of model compounds to probe oxide surface–PL interactions is a well-established approach in the literature (e.g., 23, 24). Gas-phase adsorption studies of water, NH3 , pyridine, and alcohols have been used as proxies to study the H-bonding ability of amorphous silica, quartz, and cristobalite surfaces (62– 68). Moreover, adsorption isotherms and calorimetric and FTIR studies of actual whole PL–oxide surfaces suggest that the earliest interactions do occur through the polar head group (69–71).

This makes sense considering that oxide surfaces usually possess a net electrical charge in aqueous solution. The analysis presented below helps to explain why adsorption of the quaternary ammonium ion is more effective at disrupting PL membranes on the quartz surface as compared with other oxides. The wider implications of the model for other processes, such as bacterial adhesion to minerals, are discussed at the end of the article. MODEL REVIEW

The termination of a solid structure results in the formation of a mineral surface. The atoms at the surface have unsatisfied coordination, resulting in charged surface sites. The coordination environment of these surface sites is completed by adsorption of ions from the surrounding solution. Adsorption of monovalent cations (I+ ) at negatively charged surface sites (>SO− ) on the kth solid may be represented as >SO− + I+ = >SO− − I+

K ads,i,k

[1]

where the symbol “>” implies that the surface atoms are attached to the underlying bulk mineral and K ads,i,k is the associated equilibrium constant. A similar equation can be written for monovalent anion (I− ) adsorption at positively charged surface + − sites (>SOH+ 2 ) to form a surface complex, >SOH2 –I . It is assumed that monovalent ions retain their primary solvation sheath even when adsorbed. So the sorbed ion is separated by at least one monolayer of water from the surface, and adsorption occurs mainly by ionic or electrostatic forces. The structure of the electric double layer at the mineral–water interface is described by the triple layer model (72, 73). The total Gibbs free energy of adsorption (G 0ads,i,k ) may be expressed as the sum of two contributions: (i) G 0solv,i,k , the change in free energy of solvation involved when the dissolved ion has moved from bulk solution to the surface, and (ii) G 0nonsolv,i,k , the free energy change due to effects not associated with solvation (12, 13, 16). The free energy of adsorption may then be written as G 0ads,i,k = G 0solv,i,k + G 0nonsolv,i,k

[2]

The solvation term is calculated using Born solvation theory, and the nonsolvation term is simply represented by a function that depends only on ion-intrinsic properties. Equation [2] can now be written as (16, 74) 0 G 0ads,i,k = G 0solv,i,k + G ii, i, k

[3]

According to Born theory, the solvation term depends on the static dielectric constant of the solid, the interfacial region and bulk water (εk , εint , εw ), and on the interfacial Born solvation

311

PHOSPHOLIPID–OXIDE SURFACE INTERACTIONS

coefficient (i , kcal mol−1 ) given by (16) 

zi i = 41.50 Re,i

 − 5.72

[4]

z/Re,i = 1/Re,i , where z i is the charge on the ion and Re,i is the effective electrostatic radius of the sorbed ion defined as Re,i = rx,i +

[5]

˚ for cations and 0.0 for anions (16). The

is equal to 1.4492 A ion-intrinsic term should depend on a basic property of the ion, such as the charge-to-radius ratio of the cation (z/re,i ), where re,i is the effective electrostatic radius of the ion in solution (75) and is given by re,i = r x,i + γ

[6]

˚ for cations and 0.0 for anions (76). γ is a constant equal to 0.9 A Combining these terms, and using h 1 to h 4 to denote functions as developed in the original work (16), Eq. [3] becomes  G 0ads,i,k = h 1

1 εk

 = 



1 εk

       1 1 z + h2 + h3 + h4 εint εw re,i [7]





0 + G ii,i

[8]



0 The term G ii,i can be identified with the quantity in curly brackets in Eq. [7] and depends on z/re,i according to (16)

0 G ii,i

   1 + 1.20 = −2.303RT −3.824 re,i 

[9]

where R is the ideal gas constant = 1.98 × 10−3 kcal mol−1 K−1 and T is the absolute temperature (K ). The enthalpy of the reac0 tion (Eq. 1), Hads ,i,k , can be obtained by taking the temperature derivative of Eq. [9] and making some simplifying assumptions (18). The resulting equation is  0 Hads ,i,k = i T

    1 ∂εw 1 ∂εk − + G 0ads,i,k εw2 ∂ T εk2 ∂ T

[10]

0 The corresponding entropy of adsorption, Sads ,i,k , is calculated from

0 0 G 0ads,i,k = Hads ,i,k − T Sads,i,k

[11]

Adsorption of I+ and I− occurs at negatively and positively charged surface sites, >SO− and >SOH+ 2 . These charged sites are themselves formed by desorption and adsorption of protons at neutral surface sites, >SOH. The Gibbs free energy constants of the corresponding reactions can be predicted from the crystal chemistry of the oxide as given by Pauling’s rules (77) combined with Born solvation theory (13). Thus, surface protonation equilibrium constants (KH+ ,1,k and KH+ ,2,k ) depend on 1/εk and on s/r ; s is the Pauling bond strength of the metal in the oxide defined as valence divided by coordination number, and r is the metal–oxygen bond length in the oxide. MODEL RESULTS

The Gibbs free energy of adsorption, along with the enthalpic and entropic contributions, is shown in Fig. 1 (modified from Ref. 18). The solvation and ion-intrinsic contributions to the Gibbs free energy of adsorption are calculated elsewhere (Fig. 1 of Ref. 18). The SE model predicts negative values of G 0ads,i,k for adsorption on all oxides (Fig. 1a). The predicted affinity sequence for all of the oxides considered here, except quartz and + amorphous silica, is Li+ > Na+ > K+ > Rb+ > NH+ 4 > Cs > + + + TMA (TMA ≡ tetramethylammonium, N(CH3 )4 ); the sequence is reversed for quartz and amorphous silica. The adsorption reaction is exothermic for all oxides and all ions examined. The reaction is predominantly driven by enthalpy (Fig. 1b). The entropic contribution is negative for all oxides except amorphous silica (Fig. 1c). The adsorption free energies and affinity sequences cited above refer only to standard state conditions (G 0ads,i,k ), not to the solution conditions in the hemolysis experiments and in the cytoplasm and lysosome of the macrophage. Most hemolysis experiments were performed at pH 7.3 and 0.1 M NaCl or KCl saline to simulate body fluids such as blood where the dominant ions are Na+ and Cl− or cytoplasmic fluid where K+ and negatively charged proteins dominate (78). The pH of the lysosomal solution is 4.8, and H+ and Cl− are the major ions. The extent of adsorption under the relevant conditions (G ads,i,k ) is modeled below, where G ads,i,k depends on activity (a j ) and stoichiometric coefficient (v j ) of all species j in the reacv tion according to G ads,i,k = G 0ads,i,k + RT ln ajj . The calculations were done using the surface speciation computer program GEOSURF (17). The calculated proportion of electrically neutral >SOH sites is shown in Fig. 2. These values provide a measure of the population of neutral surface sites available for H-bonding between the PL head group and the oxide surface. This population will be compared with the population of negatively charged surface sites to estimate how much electrostatic bonding is possible. For a 0.1-M TMACl solution at pH 5, the percentage of neutral >SOH sites decreases as amorphous silica ≥ quartz > rutile > corundum, and at pH 7 the order is rutile ≈ quartz > corundum ≈ amorphous silica (Fig. 2a). At neither pH does the series match the observed hemolytic

312

NITA SAHAI

∆ G0 a d s , i , k [kcalmol- 1]

-1

-2

-3

-4

(a)

-5

∆ H0 i, k [kcalmol- 1]

-1

-2

-3

-4

∆ S 0 ads, i ,k [calmol- 1K- 1]

-5

(b)

0.25

rx, i 0.00 -0.25 TMA + -0.50

K+

-0.75 -1.00

Cs + + NH + 4 Rb

Na +

(c)

Li +

0.5

1

1.5

2 o r x, M + [ A ]

2.5

3

FIG. 1. Calculated thermodynamic quantitites at standard state for the reaction (Eq. 1), plotted as a function of adsorbate radius (a) Gibbs free energy of adsorption, (b) enthalpy, and (c) entropy. Values were calculated using Eqs. [3] to [11]. Dielectric constants of oxides decrease monotonically from rutile to amorphous SiO2 . εstishovite ∼ 11.5 (estimated, see text). Some representative +, MnO2 , pyrolusite; , oxides are indicated by larger symbols and labels.  +, SnO2 , cassiterite; , Fe2 O3 , hematite; Al2 O3 , corundum; , α-TiO2 , rutile;   , α-SiO2 , quartz; ✄, oct. SiO2 , , Al(OH)3 , gibbsite; , β-TiO2 , anatase; × + , am. SiO2 , amorphous sillica. stishovite; ×

FIG. 2. Percentage of (a) uncomplexed neutral surface sites >SOH and (b) negatively charged surface sites bonded to TMA+ , as a function of pH in 0.1 M TMACl solution. Values of model parameters were selected as follows. For all oxides, capacitances of 1 Fm−2 and 0.2 Fm−2 were selected arbitrarily for the two interfacial regions of the electric double layer. Oxide-dependent surface protonation constants and electrolyte ion adsorption constants were taken from our previous work (15, 16, 18). The amount of solid in suspension is set constant at 10 gl−1 , and the BET surface area is assumed to be constant at 20 m2 g−1 . This ensured the same total surface area and total solid mass available in all cases. Site densities were variable (Table 1 of Ref. 16; 11.4 sites nm−2 for quartz from Ref. 79), so that the total concentration of surface sites available per liter is variable and depends on the solid.

PHOSPHOLIPID–OXIDE SURFACE INTERACTIONS

ability of oxides. This result presents a problem for the Hbonding theory of membranolysis, as discussed in the next section. The calculated surface site population for adsorption of TMA+ at negatively charged sites is shown in Fig. 2b. At pH 5, the order obtained for TMA+ -bound sites decreases as rutile > quartz >amorphous silica > corundum, which is inconsistent with the hemolysis sequence. The ordering is different at pH 7—amorphous silica > quartz > rutile > corundum (Fig. 2b), nearly consistent with hemolytic ability except that amorphous silica and quartz are out of sequence. This inconsistency can be explained by comparing the manner in which the hemolysis experiments and our calculations were conducted. In the experimental system, the oxides are present in a saline solution, so the negatively charged surface sites were associated with Na+ ions. When the oxides are brought in contact with the cells, the TMA+ moeity of the phospholipid head group has to displace the Na+ ions at the oxide surface. Therefore, a more appropriate comparison with experimental results would be to calculate the proportion of surface sites where TMA+ has exchanged with or replaced Na+ (Fig. 3). Thus, G exc,TMA+ /Na+ distinguishes between silica phases, on the one hand, and other oxides, on the other. The corresponding thermodynamic quantity of relevance is G exc,TMA+ /Na+ = G ads,TMA+ − G ads,Na+ . The greatest proportion of TMA+ -replaced sites obtained at pH 5 is consistent with the observed cytotoxicity sequence quartz ≈ amorphous SiO2 > Al2 O3 > TiO2 (Fig. 3). The sequence at pH 7 is different in that amorphous SiO2 is predicted

% (>SO- -TMA+ - >SO- -Na+ ) sites

4

am. SiO 2

3 2 1

qtz

0

cor

-1 -2 -3 -4

rut -5 4

5

6

7

8

9

10

pH FIG. 3. Percentage differences between TMA+ -bound surface sites and Na+ -bound surface sites in 0.1 M TMACl and 0.1 M NaCl solution, plotted as a function of pH. Values of model parameters are the same as in Fig. 2.

313

to exchange more Na+ with TMA+ than is quartz. Also, the difference between amorphous silica and quartz is greater at pH 7 than at pH 5. The prediction that G exc,TMA+ /Na+ is more negative for amorphous silica than for quartz at pH 7 is still not consistent with the observation that lytic activity of the two phases is of comparable magnitude. Clearly, it is not sufficient to consider only TMA+ /Na+ exchange. We need to explain this inconsistency and distinguish between amorphous SiO2 and quartz. The inconsistency can be resolved because G exc,TMA+ /Na+ is not the only factor controlling lysis. The entropic contribution to the free energy must also be examined because entropy tells us something about the conformation of the PL at the oxide surface. An unfavorable change in entropy of the head group may cause conformational change of the PL and membrane rupture. We already know that the reaction is being driven by enthalpy, so TMA+ /Na+ exchange is going to occur in any case regardless of the value of Sexc,TMA+ /Na+ = Sads,TMA+ − Sads,Na+ . In other words, Sexc,TMA+ /Na+ is not the relevant quantity. Because we want to find out what conformational change will occur for the TMA+ moeity at the oxide surface, it is Sads,TMA+ that we want to consider next. Lacking an easy way to calculate 0 Sads,TMA+ , I make do with Sads ,TMA+ to provide a conceptual understanding. The reaction is entropically unfavorable for all oxides except amorphous silica (Fig. 1c). In this sense, quartz is distinct from amorphous silica and is more similar to the other oxides. 0 A positive value of Sads,TMA+ (i.e., Sads ,TMA+ for the current purposes) for amorphous silica suggests that the PL head group conformation will change less and disrupt the membrane less than for oxides with Sads,TMA+ < 0. Conversely, the negative adsorption entropy on quartz suggests that PL head group conformation is changed appreciably with greater disruption to membrane structure. Thus, the entropic effect distinguishes amorphous SiO2 from quartz. The extent of TMA+ adsorption and the effect of entropy balance each other for amorphous silica so that the net membranolytic ability observed is similar to that for quartz. Negative changes in TMA+ adsorption entropy are predicted for adsorption on other oxides such as TiO2 , Fe2 O3 , and Al2 O3 , suggesting some degree of membrane disruption. But these oxides cannot cause substantial damage because extent of Na+ replacement by TMA+ is small or negative at the low solution pH within the lysosome and at the circum-neutral pH of hemolysis experiments where these oxides have positively charged surfaces. Thus, both factors need to be fulfilled so as to disrupt the membrane sufficiently to cause lysis, as happens with quartz and other crystalline polymorphs of silica containing fourfold coordinated silicon. Note that the extent of adsorption predicted is small (Figs. 2 and 3), but the important point is not so much the absolute value of the adsorption energy or of the adsorption entropy. The relative values for the various oxides permit a conceptual understanding and allows us to predict how an oxide such as stishovite would behave.

314

NITA SAHAI

DISCUSSION

Electrostatic-Bonding Mechanism According to the above discussion, G exc,TMA+ /Na+ and Sads,TMA+ are both important in determining the membranolytic potential of oxide surfaces. The observed sequence of lytic potential can be rationalized in terms of the relative covalent versus ionic, soft versus hard, and more versus less polarizable nature of the solvated oxide surface as compared with the adsorbing solute. These properties of the oxides have been inferred previously through an interpretation of the dielectric constant of the solid (Fig. 4). The dielectric constant can be related to the

FIG. 4. Relation between extent of charge transfer (N ) in a metal (M)– oxide (O) bond and (a) bond ionicity, (b) dielectric constant of oxide. N is a measure of chemical polarizability, and depends on chemical hardness (η) and electronegativity (χ) (80). Ionicity also depends on χ . Adapted from Ref. (19).

electric polarizability and molar volume of the solid via the Clausius–Mossotti equation (81). Oxides that have smaller values of the dielectric constant are interpreted as more chemically polarizable, more chemically “soft,” more covalent, and less hydrophilic as compared with oxides with larger values of the dielectric constant that are interpreted as less polarizable, chemically harder, and more hydrophilic (19). Water is a polar, relatively hard solvent. Interactions between water and soft oxide surfaces, on the one hand, will scale differently from water–water and water–hard ion interactions, on the other. When a hydrated solute approaches the SiO2 surface, the intrinsically hardest solute (Li+ ) would have the most unfavorable interaction (least negative value of adsorption energy change [Fig. 1a]). The intrinsically softest ion (TMA+ ) would have the most favorable interaction (most negative value of adsorption energy change [Fig. 1a]). Conversely, solids with large dielectric constants that are hard and less polarizable will interact with water similar to water–water and water–hard ion interactions. Quartz has a dielectric constant value between amorphous SiO2 and the other oxides. According to the above interpretation of dielectric constant, quartz may be considered more hydrophilic than amorphous silica but less so than the other oxides. This inference is consistent with microcalorimetry, adsorption isotherm, and FTIR experiments of water vapor, NH3 , pyridine, and alcohols on silica and cristobalite surfaces that show that the type and number of hydrophobic versus hydrophilic surface sites vary for the different silica polymorphs. The amorphous silica surface contains less hydrophilic isolated silanol sites than does cristobalite, which contains more hydrophilic vicinal and geminal silanol sites (62–68). The experiments were conducted on cristobalite, but it is reasonable to expect that quartz would behave similarly. The model parameters that determine the extent of adsorption of TMA+ are the surface protonation constants and the binding constants of TMA+ and Na+ for each oxide. The protonation constants, in turn, are controlled by the bonding and structure of the oxide represented in the model by s/r and 1/εk . The ion adsorption constants depend on z/re,i and z/Re,i of the ion and on 1/εk of the oxide. The s/r and z/r terms are measures of the “intrinsic” affinity of H+ and TMA+ for the surface, respectively (18, 19). The 1/εk and z/Re,i terms are related to solvation energy changes in the adsorption reaction. In addition, εk provides a means for representing the polarizability of the oxides, which in turn can be related to their relative hydrophilicity. The model requires that the extent of TMA+ adsorption relative to Na+ be large (G exc,TMA+ /Na+ < 0) and that the entropic contribution be unfavorable ( Sads,TMA+ < 0) for appreciable membranolysis to occur (Fig. 5). Although quartz is more hydrophilic than amorphous SiO2 , the entropic contribution for TMA+ adsorption is negative for quartz but not for amorphous SiO2 . The two conditions are fulfilled only at the quartz surface. One way to check the above reasoning is to predict the membranolytic potential of stishovite. Octahedral silicon in stishovite has a Pauling bond strength of 0.67 (similar to anatase) versus

PHOSPHOLIPID–OXIDE SURFACE INTERACTIONS

FIG. 5.

315

Conceptual model for initial electrostatic interaction between PL head group and oxide surface.

a Pauling bond strength of 1.0 in tetrahedral silica polymorphs. ˚ as comAlso, stishovite has longer Si–O bonds of ∼1.7 to 1.8 A pared with tetrahedral silica polymorphs, which have Si–O bond ˚ (82). Stishovite thus has a smaller value lengths equal to 1.6 A of s/r than does quartz and should have a positively charged surface at low pH as do Al2 O3 or TiO2 . Estimating the molecular polarizability of stishovite from the Si–O bond length (Fig. 5 of Ref. 81), and knowing the molar volume of stishovite (83), the Clausius–Mossotti equation can be used to estimate the average static dielectric constant of stishovite as ∼7.0 to 11.5. This can be compared with the dielectric constant values of ∼3 to 4 for amorphous silica and quartz, and ∼10 and 121 for corundum and rutile, respectively. The value of 1/εk and the corresponding contribution of solvation energy change to total free energy of adsorption decreases as amorphous SiO2 > quartz > corundum > stishovite > anatase > hematite > rutile. Combining the surface charge and solvation behavior suggests that the membranolytic activity of stishovite should resemble Al2 O3 or TiO2 rather than the other SiO2 polymorphs. With the exception of amorphous SiO2 , the predicted and observed hemolytic sequence is consistent with increasing value of the dielectric constant. Because εk depends on both polarizability and molar volume, the model is also consistent with the observation that hemolytic ability

is inversely related to molar volume and density of the oxide (43, 44). H-Bonding Mechanism for Membranolysis The H-bonding theory does not explain the hemolytic potential of organosilane modified quartz surfaces. Surfaces modified by the primary amine –NH2 group, which is capable of H-bonding, were as inert as the methyl –CH3 modified hydrophobic surface that cannot H-bond. Also, surfaces modified by the positively charged quaternary ammonium –N(CH3 )+ 3 group, which cannot H-bond, were more hemolytic (62). According to the H-bonding theory, the –PO− 4 moeity of the PL head group H-bonds to the neutral silanol (>SiOH) sites on the quartz surface. As seen above, neutral surface sites constitute the majority (>70%) of total surface sites on nearly any oxide surface over a wide range of pHs from 4 to 10 (Fig. 2a). If H-bonding to >SiOH, >AlOH, >FeOH, and >TiOH neutral surface sites is responsible for lysis, then all of the oxides should be approximately equally harmful, or at least the sequence of decreasing percentage neutral surface sites should follow the hemolytic sequence. We have seen above that neither of these expectations is satisfied by the observed hemolytic sequence. Finally, the H-bonding theory does not explain the differences

316

NITA SAHAI

among the biological activities of amorphous silica, crystalline silica polymorphs with tetrahedral Si, and stishovite with octahedral Si, all of which have silanol surface sites. The last objection may need to be reconsidered in light of recent studies that showed that the cristobalite surface has more hydrophilic vicinal and geminal sites than does amorphous silica (62–68). Whether, and how, this would affect H-bonding of phospholipids to the oxide surfaces has not been explicitly stated by the authors, although hydrophilic vicinal sites would allow the surface to act as both H-bond acceptor and H-bond donor, consistent with theoretical calculations (84, 85). Also supporting H-bonding theory, or rather contradicting the electrostatic theory, one experimental 1 H NMR study of PL conformation at silica surfaces showed a lack of interactions between amine groups and the silica surface (86). But this is only one study, and more investigations of this type are required. Perhaps the best that can be done at this time is to suggest that both electrostatic interactions and H-bonding interactions play a role (45, 54). Such an interpretation would be consistent with the most recent results of actual PL (rather than model compounds) interacting with silica that show that the initial driving force is electrostatic followed by H-bonding and hydrophobic effects (68–70). Recall the interpretation of εk in terms of soft, covalent, and more polarizable oxides versus hard, ionic, and less polarizable oxides as well as in terms of the nature of water–water versus water– solute and oxide–solute interactions. Such a characterization of the different silica polymorphs also suggests smaller conformational changes at the silica surface in the final hydrophobically driven stage of PL adsorption. Implications of the Model for Bacterial Adhesion to Minerals Oxides can be arranged in order of decreasing bacterial adhesion as hematite > corundum > quartz (1–4). This is exactly the same sequence as increasing membranolytic potential. The stronger and more prolific attachment of some bacterial strains to oxides such as Fe2 O3 as compared with quartz may be related, in part, to the fact that the bacteria can extract and use the iron for metabolic processes. At least one additional factor may be that quartz is relatively deleterious to cell membrane structure. Extending this idea one step further, bacteria may have evolved to produce extracellular polymeric substances so as to make the mineral surfaces more hydrophilic. This would be an important evolutionary step, especially if the earliest bacteria used minerals for respiration and nutrition. According to the model, other oxides also have unfavorable entropic interactions with the PL head group, even though they are more hydrophilic than quartz and amorphous silica. So, EPS should be exuded on the surfaces of many oxide (and possibly silicate) minerals. Quartz is more harmful, however, because of the greater extent of adsorption. Larger amounts of EPS production should then be expected on quartz, all other factors being equal. Consistent with this hypothesis, the nature of the substrate and of the bacterial surfaces does, in fact, affect the amount of EPS produced (5, 87, 88). The idea that surfaces become more hydrophilic by

bacterial attachment underlies the biobeneficiation of ores during mineral separation by flotation. To my knowledge, quartz is not precipitated as a biomineral by any organism, whereas biogenic amorphous silica is produced voluminously, even though quartz is thermodynamically the more stable phase. Quartz precipitation at environmental temperatures is, of course, kinetically inhibited. But if this were the only reason, then organisms could have evolved enzymes to accelerate quartz precipitation, as in the case of biologically precipitated calcite and hydroxyapatite. The deleterious effect of quartz on lipid membranes may explain the lack of quartz biomineralization. CONCLUSIONS

In short, significant membrane rupture requires lots of PL head group adsorption (G exc,TMA+ /Na+ < 0), and it should be bad for the structure of the PL membrane (Sads,TMA+ < 0). The nature of oxide–solute interactions, as compared with oxide–water, solute–water, and water–water interactions, determines the reactivity of the oxide surface toward biological membranes, where the solute is the PL head group. Stated differently, the extent of adsorption of the quaternary ammonium moiety of the PL head group, as well as the entropy change involved in the reaction, must be considered. The extent and entropy of adsorption depend on the charge-to-radius ratio of the adsorbing ion, the dielectric constant of the oxide, and the Pauling bond strength-to-bond length ratio of the oxide. The latter two parameters depend on crystal structure and crystal chemistry. APPENDIX A

Brief Review of Relevant Silicosis Literature Inhaled dust particles of crystalline tetrahedrally coordinated silica (quartz, tridymite, cristobalite, and coesite) of size ∼5 to 30 µ are cytotoxic toward the erythrocyte (6, 7), the alveolar macrophage (8), and the lysosome and liposome (9). As seen in miners and foundry workers, prolonged exposure can result in alveolar fibrogenesis, respiratory disorders, silicosis, and even lung cancer (29–40). The cytotoxic effect of oxide dusts toward the macrophage (a type of white blood cell) has been studied in model experiments using erythrocytes (red blood cells). Hemolysis refers to the breakdown of the erythrocyte cell membrane with the liberation of measurable amounts of hemoglobin. A commonly accepted sequence for the hemolytic and membranolytic effect of oxides is tridymite? ≥ cristobalite > quartz > coesite? ≥ amorphous silica > corundum > stishovite > rutile (6–9). Corresponding fibrogenic effects were induced in vivo in animal models by inhalation or by intratracheal injection of crystalline, amorphous, and vitreous silica dusts (29–36). Amorphous silica is membranolytic but nonfibrogenic; the effects of fused silica glass or glass fibers in producing silicosis are less certain (41). Some studies show that in high doses and with certain aspect ratios, glass fibers are also fibrogenic (42). Intriguingly,

317

PHOSPHOLIPID–OXIDE SURFACE INTERACTIONS

stishovite, the silica polymorph with octahedrally coordinated Si, and other oxides such as corundum and anatase are biologically much less active or are benign (7, 29, 30, 33–36). Despite more than four decades of research, none of the proposed hypotheses in the literature provides a consistent explanation for the different biological activity of crystalline tetrahedral polymorphs of silica, crystalline octahedral silica, amorphous silica, and other metal oxides. I believe that the main reason is the singular lack of consideration (with a few exceptions, e.g., Refs. 43 and 44) of the influence of crystal structure and interfacial solvation on membranolysis. The intent of this article is to obtain a better understanding of the mechanisms involved through a crystal chemical and thermodynamic perspective. Excellent reviews of the silicosis literature already exist (40, 45). A brief summary of the fate of an inhaled particle relevant to the current article is given here. When a foreign body, such as a dust particle or bacterium, enters the airways and the lungs, it is either removed by the mucociliary escalator or deposited in the respiratory tract and lungs (46). In the latter case, the particle is immediately coated by lung surfactant PLs that mask any harmful effects of the particle. The particle is then phagocytozed by the macrophage; the protective coating is removed within the acidic (pH ∼ 4.8) environment of the phagolysosome. Once the protective PL coating is stripped off, the ingested particle is in this acidic environment and has the chance to interact with the cell membrane PLs. Normally, the particle is digested by lytic and oxidative enzymes and by other factors such as cytokines and growth factor. In the case of quartz and by other harmful particles, however, the reexposed surface interacts with the PLs of the lysosomal membrane and destroys the lysosomal membrane internally. So, the harmful particle, along with acidic and oxidizing medium and other factors, is released from the cell. Thus, quartz and other harmful particles are not digested by the macrophage, even though the macrophage attempts to digest them. The particle is then ingested by another macrophage, and the cycle continues so that a single particle can inflict significant damage. The released enzymes and other factors attack the pulmonary tissue with production of fibrous proteins (e.g., collagen) that subsequently encapsulate the inhaled particle producing the silicotic nodule, block airways in the lung, and inhibit respiration. It is worthwhile to consider the molecular composition of cellular membranes and of the membranes of subcellular organelles such as the lysosome. Phospholipids such as phosphatidylcholine, phosphatidylethanolamine, phosphatidyl serine, and sphingomyelin, as well as some proteins, constitute the membranes of the macrophage and the erythrocyte. The quaternary ammonium or, generally, some sort of ammonium moiety constitutes a part of the head group for most of these PLs. The composition of the phagosomal membrane becomes very similar to that of the lysosomal membrane when the macrophage is infected with nonpathogenic bacteria (47), and this may also be true when the foreign body is an oxide dust particle. Thus, interaction of the oxide surface with quaternary ammonium is likely.

APPENDIX B

Alternative Theories for Silicosis The relation between hemolysis and fibrogenesis is not accepted by all authors and has been debated in the literature (48–50). However, if the extent of hemolysis per unit surface area of the oxide particles is considered instead of per unit mass of oxide, then it appears that hemolysis is a direct indicator of the oxide surface’s affinity for some components of the cell membrane (51). A different hypothesis for explaining the fibrogenicity of quartz calls for the toxic action of reactive Si· and SiO· radicals. The radicals are produced by hemolytic cleavage of Si–O–Si bonds at the surface of quartz by drilling and blasting activities during mining (52–55). A related theory invokes the toxic action of reactive oxygen species such as the superoxide anion O− 2 , the hydroxyl radical OH·, and hydrogen peroxide H2 O2 , produced in the “respiratory burst” after ingestion of a foreign particle into the lungs, as being responsible for fibrogenicity (45, 56, 57). Experimental proxies that try to measure the effects of foreign body inhalation and of oxidation include (i) a measurable increase in the number of neutrophils, a type of white blood cell; (ii) increased levels of lipid peroxidation (58, 59) and DNA damage (60), which are postulated to indicate the action of reactive oxygen species; (iii) increased hydroxyproline content, which reflects fibrous collagen production; and (iv) increased levels of lactatodehydrogenase, which is a lysosomal enzyme that is leaked when the membrane is lysed (reviewed in Ref. 45). The extent of inflammation caused by inhaled foreign particles is variable and depends on the amount of clearance versus retention and macrophage activation. At least those foreign particles that are retained should stimulate the oxidative burst and cause inflammation and fibrogenesis. But we know that oxides such as TiO2 and Al2 O3 stimulate a much smaller effect (or none at all) than do quartz and bacteria. Thus, although the evidence for the involvement of oxidizing species is strong, the respiratory burst theory does not fully explain why all retained foreign particles do not elicit a fibrogenic response. On balance, a comprehensive examination of the proxy methods used to estimate cytotoxicity and fibrogenicity indicates that membranolysis and cytotoxicity can explain macrophage activation and inflammation. However, these are probably only the early steps in a complex process that includes radical oxidation reactions, which may or may not lead to fibrogenicity and silicosis (45, 49, 51, 53, 61). ACKNOWLEDGMENTS I am grateful to Dr. John Wiessner at the Zablocki VA Medical Center and the Medical School of the University of Wisconsin–Milwaukee for helpful e-mail discussions of the silicosis literature. Financial support was provided by Faculty Start-Up Grants from the Department of Geology and Geophysics and the Graduate School at the University of Wisconsin–Madison.

318

NITA SAHAI

REFERENCES 1. Mills, A. L., Herman, J. S., Hornberger, G. M., and Todd, H. DeJesus, Appl. Environ. Microbiol. 60, 3300 (1994). 2. Deo, N., and Natarajan, K. A., Min. Metall. Proc. 14, 32 (1997). 3. Deo, N., Natarjan, K. A., and Somasundaran, P., Int. J. Mineral. Proc. 62, 27 (2001). 4. Yee, N., Fein, J., and Daughney, C. J., Geochim. Cosmochim. Acta 64, 609 (2000). 5. Eginton, P. J., Gibson, H., Holah, J., Handley, P. S., and Gilbert, P., Colloids Surfaces B Biointerfaces 5, 153 (1995). 6. Harley, J. D., and Margolis, J., Nature 189, 1010 (1961). 7. Stalder, K., and St¨ober, W., Nature 207, 874 (1965). 8. Parazzi, E., Secchi, G. C., Pernis, B., and Vigliani, E., Arch. Environ. Health 17, 850 (1968). 9. Weissman, G., and Rita, G. A., Nature New Biol. 240, 167 (1972). 10. Israelachvili, J., “Intermolecular and Surface Forces,” 2nd ed., Academic Press, London (1992). 11. Sverjensky, D. A., Nature 364, 776 (1993). 12. Sverjensky, D. A., Geochim. Cosmochim. Acta 58, 3123 (1994). 13. Sverjensky, D. A., and Sahai, N., Geochim. Cosmochim. Acta 60, 3773 (1996). 14. Sverjensky, D. A., and Sahai, N., Geochim. Cosmochim. Acta 62, 3703 (1998). 15. Sahai, N., and Sverjensky, D. A., Geochim. Cosmochim. Acta 61, 2801 (1997). 16. Sahai, N., and Sverjensky, D. A., Geochim. Cosmochim. Acta 61, 2827 (1997). 17. Sahai, N., and Sverjensky, D. A., Computers Geosciences 24, 853 (1998). 18. Sahai, N., Geochim. Cosmochim. Acta 64, 3629 (2000). 19. Sahai, N., Environ. Sci. Technol. 36, 445 (2002). 20. Lyklema, J., Colloids. Surfaces 10, 33 (1984). 21. van Oss, C. J., Colloids Surfaces B Biointerfaces 5, 91 (1995). 22. Hermansson, M., Colloids Surfaces B Biointerfaces 14, 105 (1999). 23. Depasse, J., J. Colloid Interface Sci. 60, 414 (1977). 24. Depasse, J., and Warlus, J., J. Colloid Interface Sci. 56, 618 (1976). 25. Langer, A. M., and Nolan, R. P., in “In Vitro Effects of Mineral Dusts” (D. C. Beck and J. Bignon, Eds.), pp. 9–24. Springer-Verlag, Berlin, 1986. 26. Allison, A. C., Harington, J. S., and Birbeck, M., J. Exp. Med. 124, 141 (1966). 27. Nash, T., Allison, A. C., and Harington, J. S., Nature 210, 259 (1966). 28. Summerton, J., Hoenig, S., Butler, C., II, and Chvapil, M., Exp. Mol. Pathol. 26, 113 (1977). 29. King, E. J., Mohanty, G. P., Harington, C. V., and Nagelschmidt, G., Brit. J. Indust. Med. 10, 9 (1953). 30. King, E. J., Mohanty, G. P., Harington, C. V., and Nagelschmidt, G., Am. Med. Assoc. Arch. Indust. Health 7, 455 (1953). 31. Zaidi, S. H., King, E. J., Harington, C. V., and Nagelschmidt, G., Am. Med. Assoc. Arch. Indust. Health 13, 112 (1956). 32. Englebrecht, F. M., Yoganathan, M., King, E. J., and Nagelschmidt, G., Am. Med. Assoc. Arch. Ind. Health 17, 287 (1958). 33. Brieger, H., and Gross, P., Arch. Envir. Health 13, 38 (1966). 34. Brieger, H., and Gross, P., Arch. Envir. Health 14, 299 (1967a). 35. Brieger, H., and Gross, P., Arch. Envir. Health 15, 751 (1967b). 36. St¨ober, W., and Brieger, H., Arch. Envir. Health 16, 706 (1968). 37. Valiante, D. J., and Rosenman, K. D., JAMA 262, 3003 (1989). 38. Goldsmith, D. F., in “Silica: Physical Behavior, Geochemistry, and Materials Applications” (P. J. Heaney, C. T. Prewitt, and G. V. Gibbs, Eds.), Chap. 16. Mineralogical Society of America, Washington, DC, 1994. 39. Wagner, G. R., Lancet 349, 1311 (1997). 40. Fubini, B., and Are´an, C. O., Chem. Soc. Rev. 28, 373 (1999). 41. Last, J. A., and Reiser, K. M., in “Silicon Biochemistry” (Ciba Foundation Symposiun Series 121), pp. 180–193. Wiley, Chichester, UK, 1986. 42. Goldstein, B., Rendall, R. E. G., and Webster, I., Environ. Res. 32, 344 (1983).

43. Wiessner, J. H., Henderson, J. D., Jr., Sohnle, P. G., Mandel, N. S., and Mandel, G. S., Am. Rev. Resp. Dis. 138, 445 (1990). 44. Mandel, G., and Mandel, N., in “Silica and Silica-Induced Lung Diseases” (V. Castranova, V. Vallyathan, and W. E. Wallace, Eds.), pp. 63–78. CRC Press, Boca Raton, FL, 1996. 45. Fubini, B., in “The Surface Properties of Silica” (A. P. Legrand, Ed.), Chap. 5. Wiley, Chichester, UK, 1998. 46. Lehnert, B. E., in “Health Effects of Mineral Dusts” (G. D. Guthrie and B. T. Mossman, Eds.), Reviews in Mineralogy, vol. 28, Chap. 14. Mineralogical Society of America, Washington, DC, 1993. 47. Lang, T., de Chastellier, C. Ryter, A., and Thilo, L., Eur. J. Cell Biol. 46, 39 (1988). 48. Nolan, R. P., Langer, A. M., Harington, J. S., Oster, G., and Selikoff, I. J., Environ. Res. 26, 503 (1981). 49. Wallace, W. E., Keane, M. J., Mike, P. S., Hill, C. A., Vallyathan, V., and Regad, E. D., J. Toxic Environ. Health 37, 391 (1992). 50. Hemenway, D. R., Absher, M. P., Fubini, B., and Bolis, V., Arch. Environ. Health 48, 343 (1993). 51. Kozin, F., Millstein, B., Mandel, G., and Mandel, N., J. Colloid. Interface Sci. 88, 326 (1982). 52. Dalal, N. S., Shi, X., and Vallyathan, V., in “Effects of Mineral Dusts on Cells” (B. T. Mossman and R. O. B´egin, Eds.), Vol. H30, pp. 265–272. NATO, Springer-Verlag, Berlin, 1989. 53. Shi, X., Dalal, N. S., Hu, X. N., and Vallyathan, V., J. Toxicol. Environ. Health 27, 435 (1989). 54. Shi, X., Mao, Y., Daniel, L., Saffiotti, U., Dalal, N. S., and Vallyathan, V., Environ. Health Persp. 102, 149 (1994). 55. Fubini, B., Bolis, V., Giamello, E., Pugliese, L., and Volant, M., in “Effects of Mineral Dusts on Cells” (B. T. Mossman and R. O. B´egin, Eds.), Vol. H30, pp. 205–214. NATO, Springer-Verlag, Berlin, 1989. 56. Vallyathan, V., Environ. Health Persp. 102, 111 (1994). 57. Vallyathan, V., Environ. Health Persp. Suppl. 106, 1151 (1998). 58. Kilroe-Smith, T. A., Environ. Res. 7, 110 (1974). 59. Gabor, S., Anca, Z., and Zugravu, E., Arch. Environ. Health 30, 499 (1975). 60. Daniel, N. L., Mao, Y., Wang, T. C. L., Markey, S. P., Shi, X., and Saffiotti, V., Environ. Res. 71, 60 (1995). 61. Driscoll, K. E., in “Health Effects of Mineral Dusts” (G. D. Guthrie and B. T. Mossman, Eds.), Reviews in Mineralogy, Vol. 28, Chap. 16. Mineralogical Society of America, Washington, DC, 1993. 62. Wiessner, J. H., Mandel, N. S., Sohnle, P. G., Hasegawa, A., and Mandel, G. S., Am. Rev. Resp. Dis. 141, 111 (1990). 63. Bolis, V., Fubini, B., Marchese, L., Martra, G., and Costa, D., J. Chem. Soc. Faraday Trans. 87, 497 (1991). 64. Bolis, V., Cavenago, A., and Fubini, B., Langmuir 13, 895 (1997). 65. Fubini, B., Bolis, V., Cavenago, A., Garrone, E., and Ugliengo, P., Langmuir 9, 2712 (1993). 66. Fubini, B., Zanetti, G., Altila, S., Tiozzo, R., Lison, D., and Saffiotti, U., Chem. Res. Toxicol. 12, 737 (1999). 67. Cauvel, A., Brunel, D., and Di Renzo, F., Langmuir 13, 2773 (1997). 68. Zoungrana, T., Berrada, A., Douillard, J.-M., and Partyka, S., Langmuir 11, 1760 (1995). 69. Rapuano, R., and Carmona-Ribeiro, A. M., J. Colloid. Interface Sci. 193, 104 (1997). 70. Chorro, M., Chorro, C., Dolladille, O., Partyka, S., and Zana, R., J. Colloid. Interface Sci. 210, 134 (1999). 71. Zajac, J., Colloids Surfaces A 167, 3 (2000). 72. Yates, D. E., Levine, S., and Healy, T. W., J. Chem. Soc. Faraday Trans. 170, 1807 (1974). 73. Davis, J. A., James, R. O., and Leckie, J. O., J. Colloid Interface Sci. 63, 480 (1978). 74. James, R. O., and Healy, T. W., J. Colloid Interface Sci. 40, 65 (1972). 75. Helgeson, H. C., Kirkham, D. H., and Flowers, G. C., Am. J. Sci. 281, 1241 (1981). 76. Shock, E. L., and Helgeson, H. C., Geochim. Cosmochim. Acta 52, 2009 (1988).

PHOSPHOLIPID–OXIDE SURFACE INTERACTIONS 77. Pauling, L., “The Nature of the Chemical Bond,” 3rd ed., Cornell University Press, Ithaca, NY, 1960. 78. Lodish, H., Baltimore, D., Berk, A., Zipursky, S. L., Matsudaira, P., and Darnell, J., “Molecular Cell Biology,” 3rd ed., Scientific American Books, New York, 1995. 79. Yates, D. E., Greiser, F., Cooper, R., and Healy, T. W., Austral. J. Chem. 30, 1655 (1977). 80. Pearson, R. G., Inorg. Chem. 27, 734 (1988). 81. Shannon, R. D., J. Appl. Phys. 73, 348 (1993). 82. Gibbs, G. V., Downs, J. W., and Boisen, M. B., Jr., in “Silica: Physical Behavior, Geochemistry, and Materials Applications” (P. J. Heaney, C. T. Prewitt, and G. V. Gibbs, Eds.), Reviews in Mineralogy, Vol. 29, p. 341. Mineralogical Society of America, Washington, DC, 1994.

319

83. Cohen, R. E., in “Silica: Physical Behavior, Geochemistry, and Materials Applications” (P. J. Heaney, C. T. Prewitt, and G. V. Gibbs, Eds.), Reviews in Mineralogy, Vol. 29, p. 386. Mineralogical Society of America, Washington, DC, 1994. 84. Hobza, P., Sauer, J., Morgeneyer, C., Hurych, J., and Zahradnik, R., J. Phys. Chem. 85, 4061 (1981). 85. Hobza, P., Mulder, F., and Sandorfy, C., J. Am. Chem. Soc. 103, 1360 (1981). 86. Chunbo, Y., Daqing, Z., Aizhuo, L., and Jiazuan, N., J. Colloid. Interface Sci. 172, 536 (1995). 87. van Loosdrecht, M. C. M., Lyklema, J., Norde, W., Schraa, G., and Zehnder, A. J. B., Appl. Environ. Microbiol. 53, 1893 (1987). 88. van Loosdrecht, M. C. M., Lyklema, J., Norde, W., and Zehnder, A. J. B., Microbiol. Rev. 54, 75 (1990).