Molecular confinement influences protein structure and enhances thermal protein stability

Abstract

The sol-gel method of encapsulating proteins in a silica matrix was investigated as a potential experimental system for testing the effects of molecular confinement on the structure and stability of proteins. We demonstrate that silica entrapment (1) is fully compatible with structure analysis by circular dichroism, (2) allows conformational studies in contact with solvents that would otherwise promote aggregation in solution, and (3) generally enhances thermal protein stability. Lysozyme, α-lactalbumin, and metmyoglobin retained native-like solution structures following sol-gel encapsulation, but apomyoglobin was found to be largely unfolded within the silica matrix under control buffer conditions. The secondary structure of encapsulated apomyoglobin was unaltered by changes in pH and ionic strength of KCl. Intriguingly, the addition of other neutral salts resulted in an increase in the α-helical content of encapsulated apomyoglobin in accordance with the Hofmeister ion series. We hypothesize that protein conformation is influenced directly by the properties of confined water in the pores of the silica. Further work is needed to differentiate the steric effects of the silica matrix from the solvent effects of confined water on protein structure and to determine the extent to which this experimental system mimics the effects of crowding and confinement on the function of macromolecules in vivo.

Traditionally, biochemists have studied the properties of purified macromolecules in dilute solutions due to convenience and aggregation problems that often occur on further concentration. The interior of a living cell, however, is a highly concentrated mixture of molecules where proteins and nucleic acids occupy a significant fraction of the total cell volume, e.g. 30%–40% in the cytoplasm of Escherichia coli (Zimmerman and Trach 1991). Thus, one might question the completeness of any results obtained by the reductionist approach where biomolecules are characterized in dilute solution. Can nonspecific interactions between neighboring molecules influence the vital reactions and events that define a living cell?

Statistical-thermodynamic models predict that excluded volume effects will have measurable consequences on the structure, function, and interaction of individual macromolecules under physiological conditions (Minton 1981, 1992, 1995, Zimmerman and Minton 1993). Recently, the modeling approach has been applied to the stability of proteins against denaturation by heat and by urea in crowded environments (Minton 2000). One pertinent prediction of these models is that proteins will tend to favor compact globular conformations as the extent of crowding or uniform confinement is increased. Two important corollaries may be inferred from this prediction: (1) The initial folding of a nascent polypeptide in vivo is promoted by the crowded environment in which it is synthesized; and (2) the stability of a mature protein, i.e., the ability to maintain a native globular state, is much greater in vivo than the value measured in dilute solution.

A reliable experimental method for assessing the magnitude and importance of excluded volume effects on the structure of macromolecules has been difficult to achieve. In this paper, the sol-gel method is exploited to study proteins encapsulated within the pores of a silica glass. Sol-gel glasses are prepared from metal alkoxide precursors such as tetramethoxysilane that are hydrolyzed by acid or base catalysis followed by polycondensation to form networks of the metal oxide (Dave et al. 1994). Proper buffering during the gelation step allows one to encapsulate proteins that retain their native spectroscopic properties (Ellerby et al. 1992). New sol-gel samples may be aged in contact with aqueous buffer, or the solvent may be allowed to slowly evaporate forming a denser xerogel. Wet-aged gels shrink to 80%–90% of their original volume, whereas xerogels may occupy only 10%–15% of their original volume. The size of the protein-occupied pores in these gels is believed to be the same order of magnitude as the diameter of the protein, and it has been suggested that the protein itself may dictate the pore size during the gelation process (Lan et al. 1999). Proteins are not bound covalently to the silica matrix, but the matrix substantially impedes the rotational freedom of the protein (Gottfried et al. 1999). Solvent and smaller solute molecules can diffuse into the interconnected pores of the silica matrix, but the larger protein molecules are prevented from escaping into the surrounding buffer. The optically transparent silica glass is fully compatible with most analytical techniques applied to macromolecules in dilute solutions.

We initiated this work based on the hypothesis that sol-gel encapsulation could mimic the effects of confinement on the structure and stability of proteins and could thereby provide a suitable model system for the effects of crowding and confinement in a living cell. By using silica rather than a soluble macromolecule as the "crowding agent," it is possible to perturb the structure of the test protein without simultaneously affecting the structural integrity of the species that constitutes the crowded environment. Furthermore, intermolecular protein aggregation is prohibited in this system because the glass matrix shields the individual protein molecules from each other. We demonstrate both of these attributes using circular dichroism (CD) to compare the secondary and tertiary structures of model proteins in solution and in wet-aged sol-gel glasses.

Results

Comparison of protein structures in solution and after sol-gel encapsulation

A method for preparing thin sol-gel wafers (1 mm thick) was developed to minimize the total sample absorbance and equilibration times prior to analysis by circular dichroism (see Materials and Methods). CD signals in the far-UV region originate from secondary structure contributions, whereas signals in the near-UV region arise from asymmetry in the tertiary structure surrounding aromatic residues (Schmid 1990). The structures of hen egg-white lysozyme, bovine α-lactalbumin, horse heart metmyoglobin, and apomyoglobin were analyzed in buffered solutions and analyzed in silica matrices after equilibrating with the same solutions. In the cases of lysozyme, α-lactalbumin, and metmyoglobin, the solution spectra and the sol-gel spectra are similar (Fig. 1 ▶A–C). These results indicate that the native solution structure of each of these proteins is relatively unaltered by sol-gel encapsulation.

Fig. 1.

Secondary and tertiary structures of proteins before and after sol-gel encapsulation. Far-UV CD (left) and near-UV CD (right) are shown for the following proteins: (A) lysozyme; (B) α-lactalbumin; (C) metmyoglobin; (D) apomyoglobin. Solid lines denote solution spectra (a), and broken lines denote sol-gel samples (b). All spectra were taken at 20°C in or in contact with solutions of 10 mM potassium phosphate, 50 mM KCl at pH 7.0. For α-lactalbumin samples, phosphate buffer was replaced with 10 mM Tris to accommodate addition of 5 mM CaCl₂.

For metmyoglobin, in which iron (III) is bound to the heme group, the far-UV CD spectra in each environment did not always match as closely as shown in Figure 1C ▶. Occasionally, the secondary structure of sol-gel encapsulated samples appeared to be slightly denatured with a less negative mean residue ellipticity than the solution structure. Analysis of one of these denatured sol-gel samples by UV-vis absorption revealed a broadened Soret band of decreased molar absorptivity, as reported previously by another investigator (Edmiston et al. 1994). The variability in structure of encapsulated metmyoglobin is likely attributable to partial loss or displacement of the heme group during the initial gelation and aging process.

In the case of apomyoglobin, from which the heme group has been extracted before encapsulation, the denaturing effect of sol-gel entrapment on secondary structure is pronounced and reproducible from sample to sample. Typically, the α-helical content of apomyoglobin is reduced from 60% in solution to 20%–25% on encapsulation (Fig. 1D ▶). Because of the solvent accessibility of the tryptophan residues in apomyoglobin, it is not possible to detect any changes in tertiary structure by near-UV CD.

Sensitivity of α-lactalbumin structure to calcium and DTT

Bovine α-lactalbumin is a small protein containing 123 amino acid residues and four intramolecular disulfide bonds. This protein is known to bind a single calcium ion that stabilizes its tertiary structure in the presence of disulfide reducing agents (Kuwajima et al. 1990; Ewbank and Creighton 1993). We examined the structure of α-lactalbumin on disulfide reduction in solution and in sol-gel samples to determine whether calcium binding would induce a conformational change in both environments (Fig. 2 ▶). In these experiments, the apoprotein, having no bound calcium, was obtained by addition of 5 mM EDTA to the equilibration buffer, and the partially reduced three-disulfide form was obtained by addition of 6.0 mM dithiothreitol (DTT). Consistent with previous reports, the native structure of α-lactalbumin is lost in solutions containing 6.0 mM DTT and no calcium (Fig. 2A ▶, cf. spectra b and d).

Fig. 2.

The structure of sol-gel entrapped α-lactalbumin is more strongly dependent on calcium binding and disulfide reduction than the protein in solution. Far-UV and near-UV CD of α-lactalbumin in solution (A) and encapsulated in sol-gel (B). Equilibration buffers contain 50 mM KCl, 10 mM Tris at pH 7.0, and the following additives: (a) 5.0 mM EDTA; (b) 5.0 mM EDTA and 6.0 mM DTT; (c) 5.0 mM CaCl₂; (d) 5.0 mM CaCl₂ and 6.0 mM DTT. Thicker lines denote samples containing calcium. Because of the high total absorbance in the far-UV region of sol-gel samples in contact with buffers containing EDTA or DTT, the buffer in the cuvette was diluted 5-fold with 10 mM Tris during recording of the CD spectrum.

When confined to the pores of a sol-gel, the structure of α-lactalbumin shows a stronger dependence on the presence of calcium and DTT than observed in solution. Removal of calcium from α-lactalbumin after sol-gel encapsulation results in loss of tertiary structure without the addition of reducing agents (Fig. 2B ▶, cf. spectra a and c). Furthermore, calcium was unable to rescue the native structure of sol-gel encapsulated α-lactalbumin when DTT was simultaneously present in the equilibration buffer (Fig. 2B ▶, cf. spectra c and d). The stabilizing effect of calcium on secondary structure in the sol-gel sample (in the absence of DTT) was readily reversible by changing the equilibration buffer from one that contained EDTA to one that contained excess calcium and vice versa. The loss of structure on partial disulfide reduction, however, was not reversible by simply removing DTT from the external buffer.

Effects of pH and ionic strength on encapsulated apomyoglobin

As described above, sol-gel encapsulated apomyoglobin adopts a different conformation than that observed in dilute solution. Because horse heart myoglobin is a slightly basic protein of 153 amino acid residues containing 19 lysines (pI = 7.36), it seemed possible that the unfolded state of encapsulated apomyoglobin at neutral pH was due to an electrostatic interaction between the positively charged protein and the negatively charged silica oxide groups found intermittently throughout the silica matrix. To test this hypothesis, the structure of encapsulated apomyoglobin was examined as a function of pH and ionic strength. Contrary to our expectations, the ellipticity of sol-gel samples exhibited almost no dependence on proton concentration in the range of pH 4 to pH 10 (Fig. 3A ▶). Furthermore, there was no change in the structure of encapsulated apomyoglobin when the salt concentration was increased from 50 mM to 2000 mM KCl (Fig. 3B ▶). Electrostatic interactions should have been reduced at pH values above the pI of the protein and by ionic shielding at high salt concentrations. One can rule out the possibility that the encapsulated protein was unable to equilibrate with the surrounding buffer as addition of 10 mM HCl or 6.0 M guanidinium ion resulted in further unfolding of the protein to nearly the same ellipticities found in solution (see Fig. 3B ▶, spectra c and d). The lack of effect on secondary structure on changing the pH and KCl concentration indicates that electrostatic interactions are not responsible for the unfolded conformation of encapsulated apomyoglobin.

Fig. 3.

The non-native state of sol-gel encapsulated apomyoglobin is relatively independent of pH or KCl concentration but responds to other salts in accordance with the Hofmeister effect. (A) Sol-gel samples containing apomyoglobin were equilibrated with 50 mM KCl in the following buffers: (a) 10 mM potassium acetate at pH 4.0; (b) 10 mM potassium phosphate at pH 6.0; (c) 10 mM potassium phosphate at pH 7.0; (d) 10 mM Tris at pH 8.0; (e) 10 mM glycine at pH 10.0; (f) control, in solution. (B) Sol-gel samples containing apomyoglobin were equilibrated with the following solutions: (a) control buffer, 10 mM phosphate, 50 mM KCl at pH 7.0 (solid line); (b) 10 mM phosphate, 2000 mM KCl at pH 7.0 (solid circles); (c) 10 mM HCl; (d) 10 mM phosphate, 6.0 M guanidine-HCl at pH 7.0; (f) control, in solution. (C) Sol-gel samples containing apomyoglobin were equilibrated in 10 mM phosphate at pH 7.0, plus the following salts: (a) 1.0 M LiCl; (b) sol-gel control, 50 mM KCl (thicker line); (c) 1.0 M (NH₄)SO₄; (d) 1.0 M N(CH₃)₄Cl; (e) 1.0 M KH₂PO₄; (f) control, in solution. Inset: Spectra of soluble apomyoglobin in the same 1.0 M salt solutions plus 10 mM phosphate at pH 7.0, or 10 mM sodium formate at pH 3.8. The dashed line in each panel (curve f) is a reference spectrum for the native state of apomyoglobin in solution at pH 7.0.

Hofmeister effects on encapsulated apomyoglobin

Although KCl concentration had no effect on the secondary structure of encapsulated apomyoglobin, both K⁺ and Cl⁻ ions fall in the middle of the Hofmeister series near the division between kosmotropic (strongly water binding) ions and chaotropic (weakly water binding) ions (Collins and Washabaugh 1985; Collins 1997). Thus, we decided to examine the effects of other salts from the Hofmeister series to see whether their presence would influence the structure of encapsulated apomyoglobin. When sol-gel samples were equilibrated with various 1.0 M salt solutions, a dramatic ion-specific increase in the α-helical content of encapsulated apomyoglobin was observed (Fig. 3C ▶). The rank order of effectiveness for inducing secondary structure was as follows: KH₂PO₄ > N(CH₃)₄Cl > (NH₄)₂SO₄ > KCl > LiCl. Notably, these same 1.0 M salt solutions had no significant effect on the native conformation of the soluble protein at pH 7.0 or on the molten globule-like conformation of the soluble protein at pH 3.8 (Fig. 3C ▶, inset). In the presence of 1.0 M phosphate, the α-helical content of encapsulated apomyoglobin approached that of the native protein in solution. The native-like helicity of encapsulated metmyoglobin was not affected by addition of 1.0 M phosphate (data not shown). These results suggest that water structure and protein hydration may be important determinants of apomyoglobin conformation within the pores of the silica matrix.

Alcohol effects on encapsulated protein

A useful feature of the sol-gel system for studying protein folding is that intermolecular protein aggregation is prohibited. This advantage is demonstrated by the addition of water-miscible alcohols that induce aggregation in solution. The secondary structure of apomyoglobin in solution is modestly affected by 14.5% by volume ethanol (2.5 M) or 1.0% hexafluoroisopropanol (HFIP) (Fig. 4A ▶). When the volume fraction of HFIP is increased to 5.0%, apomyoglobin aggregates in solution, and the ellipticity is greatly diminished (Fig. 4A ▶, spectrum d). In striking contrast, both ethanol and HFIP enhance the helical content of sol-gel encapsulated apomyoglobin (Fig. 4B ▶). In 5.0% HFIP for example, the ellipticity of the sol-gel sample approaches that of the native state in solution. At a volume fraction of 10% HFIP, the helical content is further increased to 81%, approximating the value of the native holoprotein in solution.

Fig. 4.

Sol-gel encapsulated proteins can be studied in solvent conditions that are prohibited in solution due to aggregation phenomena. Spectra are shown for the following proteins and conditions: (A) apomyoglobin in solution; (B) apomyoglobin in sol-gel; (C) α-lactalbumin in solution, far-UV CD; (D) α-lactalbumin in sol-gel, far-UV CD; (E) α-lactalbumin in solution, near-UV CD; (F) α-lactalbumin in sol-gel, near-UV CD. All samples contain 10 mM buffer, 50 mM KCl, plus the following: (a) control, no addition; (b) 14.5% ethanol by volume; (c) 1.0% HFIP by volume; (d) 5.0% HFIP; (e) 10% HFIP. The α-lactalbumin samples also contain 5.0 mM CaCl₂.

Similar solvent effects were observed for α-lactalbumin. In solution, 14.5% ethanol and 1.0% HFIP have no effect on the secondary or tertiary structure of α-lactalbumin (Fig. 4 ▶C,E). In the presence of 5.0% HFIP, the far-UV CD spectrum indicated a major change in secondary structure, and the near-UV signal was lost (Fig. 4 ▶, d spectra). In 10% HFIP, precipitation of α-lactalbumin in solution was visible to the eye. In contrast, α-helix formation was enhanced by high concentrations of HFIP in the sol-gel samples (Fig. 4 ▶D,F). At a volume fraction of 10% HFIP, the α-helical content of sol-gel encapsulated α-lactalbumin doubled from ∼27% to 54% with a concomitant loss in tertiary structure (compare spectra a and e, Fig. 4 ▶).

Thermal stability of encapsulated proteins

We examined the thermal transitions of lysozyme, α-lactalbumin, and apomyoglobin in solution and in the silica matrix by monitoring changes in ellipticity at a characteristic wavelength (Fig. 5 ▶). In the case of lysozyme, thermal denaturation was not reversible in solution or within the silica matrix (Fig. 5A ▶, see large symbols at 15°C). However, the encapsulated protein appeared to unfold to a lesser extent and less cooperatively than the protein in solution as measured by the change in secondary structure at 222 nm or by the change in tertiary structure at 260 nm (Fig. 5A ▶, inset).

Fig. 5.

The thermal stabilities of sol-gel encapsulated proteins are enhanced. Thermal denaturation was monitored in (a) solution, filled triangles, and in (b) sol-gel samples, open circles. Larger symbols on the left side of each panel denote the endpoint of the corresponding sample after heating and cooling back to 15°C. (A) Lysozyme transitions at 222 nm and 260 nm (inset). Buffers contain 10 mM potassium phosphate, 50 mM KCl, and 0.1 mM EDTA at pH 7.0. (B) α-Lactalbumin transitions at 222 nm and 272 nm (inset). Buffers contain 10 mM Tris, 50 mM KCl, 5.0 mM CaCl₂ at pH 7.0. (C). Apomyoglobin transitions at 222 nm under the following conditions: (a) in solution with 10 mM potassium phosphate, 50 mM KCl at pH 7.0; (b) in sol-gel with 10 mM potassium phosphate, 50 mM KCl at pH 7.0; and (c) in sol-gel with 1.0 M potassium phosphate at pH 7.0 (open squares).

The thermal stability of sol-gel encapsulated α-lactalbumin was also compared to its stability in dilute solution (Fig. 5B ▶). In this case, both samples started from the same value of ellipticity at 222 nm, but the protein in solution unfolded to a much higher degree when heated to 95°C. Similar unfolding transitions were observed by monitoring ellipticity in the near-UV region at 272 nm, although α-lactalbumin appeared to unfold more cooperatively in solution at this wavelength (Fig. 5B ▶, inset). When the temperature was slowly reduced from 95°C to the starting temperature of 15°C, the ellipticity also returned to its starting value at 222 nm and at 272 nm. Thus, the thermal denaturation of sol-gel encapsulated α-lactalbumin was completely reversible whether following changes in secondary structure or tertiary structure.

In Figure 5C ▶, we confirm previous reports that apomyoglobin unfolds irreversibly on heating in solution due to aggregation at neutral pH (curve a). In contrast, the mean residue ellipticity of sol-gel encapsulated apomyoglobin is relatively unchanged at pH 7.0 up to a temperature of 95°C (Fig. 5C ▶, curve b).

However, because encapsulated apomyoglobin is already highly unfolded at room temperature in control buffer, we also examined the thermal behavior of the entrapped protein after equilibration with 1.0 M phosphate buffer to induce a more native-like ellipticity. The conformation of encapsulated apomyoglobin in 1.0 M phosphate was also remarkably stable to increasing temperature as measured by the modest change in secondary structure at 222 nm (Fig. 5C ▶, curve c). The endpoint ellipticity at 95°C in 1.0 M phosphate remained well below the endpoint attained in solution. The relatively flat unfolding transition for each of the encapsulated apomyoglobin samples was completely reversible on cooling the temperature down to 15°C.

Discussion

CD analysis of sol-gel samples

We investigate the effects of molecular confinement using circular dichroism to examine changes in the structure and stability of sol-gel encapsulated proteins.

To our knowledge, cytochrome c and carbonic anhydrase are the only other proteins to be examined by far-UV CD following sol-gel encapsulation (Shen and Kostic 1997; Badjic and Kostic 1999). Using plastic disposable cassettes, normally employed for casting polyacrylamide gels, we were able to make uniformly thin sol-gel sheets of low optical absorbance as required for CD analysis. Comparison of structures of four different model proteins in solution and in the silica matrix revealed that not all proteins retain their native solution structure after encapsulation. Further experiments revealed that encapsulated proteins behave differently in general, on changes in buffer conditions and increases in temperature. At this time, we can envision only three phenomena that may account for the altered structural properties of sol-gel encapsulated proteins: (1) steric effects from molecular confinement; (2) adsorption to the silica matrix; and (3) the unusual physical properties of confined water. Each of these phenomena is critically evaluated in the discussion below.

Studies with apomyoglobin

The α-helical content of apomyoglobin was diminished greatly from 60% in solution to 20%–25% on sol-gel encapsulation (Fig. 1D ▶). The low fraction of helical secondary structure found in encapsulated apomyoglobin is less than that of the salt-stabilized molten globule state found in acidic solutions, for which the α-helical content is ∼35% (Goto and Fink 1990). We initially believed that this denaturing effect was attributable to an electrostatic interaction between the positively charged protein and the negatively charged silica matrix; ∼15% of the Si-O-[Si] linkages remain as Si-O⁻ groups due to incomplete cross-linking during the gelation and aging steps of the sol-gel process (Reetz et al. 1996). Also, it is known that metmyoglobin adsorbs weakly to ultrafine silica particles (Kondo and Mihara 1996). However, when we examined sol-gel encapsulated apomyoglobin in buffers at pH values above the pI of the protein, there was no apparent change in secondary structure (Fig. 3A ▶). If the protein was denatured due to direct contact with the silica matrix, this adsorption effect should have been relieved at pH 8 as seen with ultrafine silica particles (Kondo and Mihara 1996), and the protein should have regained all or part of its native structure. Furthermore, a 40-fold increase in KCl concentration had no effect on the unfolded state of encapsulated apomyoglobin (Fig. 3B ▶). Electrostatic interactions should be reduced at high salt concentrations due to ionic shielding.

We conclude from these data that the unfolded state of encapsulated apomyoglobin is not due to an electrostatic interaction between the silica matrix and the protein.

When salts other than KCl were included in the equilibration buffer, a dramatic ion-dependent increase in the α-helical content of encapsulated apomyoglobin was observed (Fig. 3C ▶). The salt solutions influenced apomyoglobin conformation in accordance with the Hofmeister series; the order of anion effectiveness was PO₄²⁻ > SO₄²⁻ > Cl⁻, and the order of cation effectiveness was N(CH₃)₄⁺ ≫L: K⁺ > Li⁺. These same ions had a negligible effect on the structure of the soluble protein under native conditions at pH 7 or under partially denatured conditions at pH 3.8 (Fig. 3C ▶, inset). Phosphate was shown previously to enhance the thermal stability of ribonuclease, and tetramethylammonium ion was shown to stabilize the helical structure of collagen (von Hippel and Wong 1964).

Although the mechanistic details of the Hofmeister effect are still under investigation, it is generally believed that Hofmeister ions influence protein structure indirectly through changes in the hydrogen-bonding properties of water (Baldwin 1996; Collins 1997; Wiggins 1997). The physical properties of water are strongly linked to its hydrogen-bonding character, and it is known that the network of hydrogen bonds is interrupted at a solid interface. Modified water structure at silica interfaces has been confirmed by dielectric relaxation (Ishida and Makino 1999), vibrational sum frequency spectroscopy (Du et al. 1994), and neutron diffraction (Teixeira et al. 1997; Soper et al. 1998). We hypothesize that the unusual water structure in the pores of the silica matrix diminishes the hydrophobic effect, a dominant force in protein folding. Addition of phosphate or tetramethylammonium ions perturbs the unfavorable water structure and reestablishes the hydrophobic effect, allowing apomyoglobin to adopt a more native conformation.

It would be interesting to know whether the tertiary structure of encapsulated apomyoglobin is also restored on addition of phosphate and tetramethylammonium ions. This cannot be determined by circular dichroism, however, because no characteristic signal exists for apomyoglobin in the near-UV region. With this in mind, we attempted to titrate hemin into sol-gel encapsulated apomyoglobin to reconstitute the holoprotein and obtain a CD spectrum in the near-UV region.

Unfortunately, hemin was observed to adsorb or precipitate onto the outer surface of the sol-gel sample, and no change in structure was detected by CD. Apparently, the chemical and physical properties of hemin hinder its diffusion into the pores of the silica matrix and prevent access to the apoprotein.

Another possible explanation for the non-native state of apomyoglobin following sol-gel encapsulation is that the protein occupies an irregularly shaped pore that is not conducive to the native structure. Using hard particle partition theory to calculate the effects of molecular confinement on protein isomerization equilibria, it has been concluded that protein conformation becomes dependent on the shape of its enclosure as the radius of the enclosure approaches the radius of the protein (Minton 1992). Unfolded proteins may become trapped in a free energy trough that does not exist in the absence of the enclosure. Although pore geometry may influence the conformation of encapsulated apomyoglobin, the Hofmeister ion effect discussed above suggests that water structure in the pores of the silica matrix is a major contributing factor to the unfolded conformation of the encapsulated protein.

In addition to salt solutions, we examined the effect of water-miscible alcohols on the structure of sol-gel encapsulated apomyoglobin. Trifluoroethanol (TFE) and hexafluoroisopropanol (HFIP) are two of the most potent alcohols for inducing helical transitions in polypeptides, and the order of alcohol effectiveness has been established as HFIP > TFE > isopropanol > ethanol > methanol (Hirota et al. 1997). Similar to the Hofmeister ions, these alcohols may affect protein structure indirectly through their ability to disrupt the hydrogen-bonding properties of the solvent. Addition of 5% HFIP to sol-gel encapsulated apomyoglobin increased the α-helical content to a value close to that of the native structure (Fig. 4 ▶A,B). In solution 5% HFIP caused apomyoglobin to aggregate, highlighting one of the key advantages of studying protein folding in the sol-gel system where aggregation is prevented by the silica matrix. Addition of 10% HFIP to the encapsulated apoprotein resulted in a further increase in helical content with a CD spectrum that resembles the holoprotein in solution (Fig. 4B ▶). A similar superhelical structure has been observed for apomyoglobin in a solution containing 80% TFE (Shiraki et al. 1995).

Other investigators have examined the properties of sol-gel encapsulated apomyoglobin by fluorescence spectroscopy (Edmiston et al. 1994). In the fluorescence study, changes in the wavelength of maximum emission led to the conclusion that apomyoglobin is partially denatured on encapsulation and that the protein could be further unfolded by guanidine-HCl. The CD data reported here are in complete agreement with the conclusions obtained by this complementary approach.

Studies with α-lactalbumin

Entrapment of α-lactalbumin in the pores of the silica matrix yielded a native-like structure in the presence of calcium (Fig. 1B ▶). However, changing the buffer conditions by addition of EDTA, DTT, or HFIP resulted in conformational changes that were substantially different from those observed in dilute solutions. Removal of bound calcium by chelation with EDTA caused a major loss in secondary and tertiary structure of the encapsulated protein in the absence of reducing agents (Fig. 2 ▶). Furthermore, in contrast to the effect observed in solution, calcium binding was unable to protect sol-gel encapsulated α-lactalbumin from unfolding on reduction of one of its four disulfide bonds. These observations can be attributed to either the misfit pore hypothesis or the unfavorable water structure hypothesis. Removal of calcium or reduction of a disulfide bond leads to a less stable protein conformation that may be more susceptible to the external influences of solvent and/or pore shape.

Similar to apomyoglobin, the addition of 10% HFIP resulted in a non-native superhelical structure in the pores of the matrix, whereas the protein aggregated under the same conditions in dilute solution (Fig. 4 ▶C–F). The presence of 40% TFE in solution has been shown to yield a CD spectrum similar to that observed for 10% HFIP in the sol-gel matrix (Gast et al. 1999). In both our experiment and the TFE experiment, there was a corresponding loss in tertiary structure as the helicity increased above the native value.

Thermal stability of sol-gel encapsulated proteins

Sol-gel entrapment has been demonstrated to be a means of stabilizing enzyme activity for bioreactor and biosensor applications. For example, the half-life in enzymatic activity at 70°C is greatly enhanced for sol-gel encapsulated alkaline phosphatase (Braun et al. 1990) and acid phosphatase (Shtelzer et al. 1992). There have been only a few reports that look directly at conformational changes in sol-gel encapsulated proteins on heating. Two studies monitored changes in intrinsic tryptophan fluorescence (Zheng and Brennan 1998; Zheng et al. 1998), and a third study monitored changes in absorbance of a covalently bound heme group (Lan et al. 1999). These earlier investigations did not compare the spectroscopic signal in solution to the signal in the sol-gel environment on a molar basis. Thus, it is unclear whether the thermal unfolding transitions started from the same protein conformation at the initial temperature under each condition.

Using the mean residue ellipticity to follow changes in structure, we examined the thermal denaturation of lysozyme, α-lactalbumin, and apomyoglobin on a molar basis (Fig. 5 ▶). In all three cases, the silica-entrapped protein appeared to be more stable than the protein in solution. For lysozyme, the unfolding transition was not reversible in solution or within the silica matrix, but the extent of unfolding was much greater in solution. The thermal denaturation curves for metmyoglobin are very similar to that observed for lysozyme; encapsulated metmyoglobin unfolds to a lesser degree than the protein in solution, but the thermal transition is not fully reversible under either condition (data not shown).

The results obtained with α-lactalbumin may provide the most compelling evidence that molecular confinement enhances protein stability. Reversible folding transitions and enhanced thermal stabilities were observed for sol-gel encapsulated α-lactalbumin by monitoring changes in both secondary structure and tertiary structure (Fig. 5B ▶). Because the silica-entrapped protein never approached a fully denatured endpoint at 95°C, it is not possible to calculate T_m, the midpoint temperature of the unfolding transition. If one assumes the thermal transition curve for encapsulated α-lactalbumin retains the same shape as the curve in solution and simply shifts to higher temperatures, then the increase in thermal stability can be estimated to be between 25°C and 32°C; the endpoint of ellipticity in the sol-gel glass at 95°C is equal to the value attained in solution at 63°C (222 nm) or 70°C (272 nm).

In the case of apomyoglobin, direct comparison of the thermal stability in solution to the stability observed in the sol-gel glass is complicated. It was demonstrated over three decades ago that myoglobin irreversibly aggregates in solution during thermal denaturation studies between pH 5.5 and 9.0 (Acampora and Hermans 1967). For this reason, most previous folding studies with apomyoglobin in solution have been carried out below pH 5.5 or in the presence of chemical denaturants like urea and guanidine-HCl. We were initially excited at the prospect of being the first to be able to measure the thermal stability of apomyoglobin in physiological buffer conditions using the sol-gel encapsulation system to avoid aggregation. However, apomyoglobin was already unfolded to a large extent at pH 7.0 in the sol-gel environment. We found that this non-native state did not change appreciably on heating to 95°C (Fig. 5C ▶). When encapsulated apomyoglobin was heated in the presence of 1.0 M phosphate to induce a more helical starting value at 222 nm, the protein appeared to be extremely stable with only a slight change in ellipticity during a temperature ramp of 80°C. A similar, flat profile has been demonstrated for the thermal transition of encapsulated cytochrome c (Lan et al. 1999).

Recently, an elegant CD study was reported on the thermal stability of sol-gel encapsulated carbonic anhydrase II (Badjic and Kostic 1999). In contrast to the present work, it was concluded that carbonic anhydrase is destabilized in the silica matrix relative to dilute solution. Differences in experimental methods and thermodynamic reversibility may account for this apparent discrepancy. First of all, the unfolding of encapsulated carbonic anhydrase was irreversible, precluding the use of the observed T_m as a meaningful parameter for comparison of stability (Pace et al. 1990). Secondly, to ensure thermal equilibration of the 8 × 8 mm thick sol-gel samples in the carbonic anhydrase study, the unfolding transition was measured at 3°C step intervals with about 30 min of equilibration time and 20 min of CD analysis time at each step. Thus, the denaturation of encapsulated carbonic anhydrase took ∼15 h to reach the endpoint temperature of 74°C, whereas the present study took only 40 min to reach an endpoint of 95°C. Perhaps a faster ramp in temperature using thinner sol-gel samples would result in a reversible unfolding transition for carbonic anhydrase. It should be noted that thermally denatured carbonic anhydrase is also difficult to refold in solution, unless the enzyme is assisted by a chemical or molecular chaperone (Rozema and Gellman 1996; Kundu and Guptasarma 1999).

Comparison to theory and relevance in vivo

Statistical-thermodynamic models, developed primarily by A.P. Minton and coworkers, predict that macromolecules will favor globular conformations as the extent of macromolecular crowding is increased. Recently these models were elaborated to estimate the magnitude of crowding effects on the thermal denaturation of proteins. This application predicts an increase in T_m at physiological solute concentrations in the range of 5–20°C (Minton 2000). Only one paper in literature has been cited as experimental support for this prediction. A 5°C increase in stability was attributed to crowding effects by detecting a subtle change in intrinsic fluorescence during thermal denaturation of G-actin in the presence or absence of 100 mg/ml polyethylene glycol, PEG-6000 (Tellam et al. 1983).

The present work appears to provide additional support for the models of molecular crowding as they pertain to protein stability. Although confinement within a silica matrix and the use of concentrated mixtures of cosolutes are two very different means of forming a crowded environment, the predicted effects are qualitatively similar in each environment for a given test protein (Minton 1992).

Direct comparison of the model predictions to the stabilities observed in the sol-gel system is difficult because the average sizes and shapes of the protein-occupied pores in the silica matrix are unknown and not easily obtained by standard techniques. Using our sol-gel entrapment system, we observed a 25–30°C increase in the thermal stability of α-lactalbumin. The stability of encapsulated apomyoglobin in the presence of 1.0 M phosphate relative to the stability of the soluble protein was even more remarkable than α-lactalbumin. However, we are hesitant to draw any conclusions from the apomyoglobin results because the tertiary structure of this protein cannot be confirmed by CD analysis; the α-helical content of encapsulated apomyoglobin in the presence of 1.0 M phosphate approximates that of the solution structure, but the protein may still be entrapped in a very non-native three-dimensional state.

Recently, the statistical-thermodynamic models of crowding have been invoked to argue that newly synthesized proteins are more susceptible to aggregation events in vivo than observed in vitro (Ellis 1997). The effective concentration of an aggregation-prone macromolecule may be increased by two to three orders of magnitude in the cell reflecting its change in thermodynamic activity on crowding.

Presumably, this is why molecular chaperone proteins are required for efficient folding of nascent polypeptides in the cellular environment. Although a recent paper examining lysozyme refolding in crowded media seems to support this notion (van den Berg et al. 1999), it should not be overlooked that the effective concentrations of the mature properly folded proteins are increased by the same factor in the cell. Thus, a newly synthesized protein should be under the positive folding influence of crowding due to nonspecific interactions with mature proteins and protein complexes (including the ribosome itself) for a significant time period before it encounters another unfolded protein and forms a species that leads off the productive folding pathway. In support of this hypothesis, the same investigators have also reported that the "fast track" of lysozyme refolding is accelerated in crowded media for those molecules that do not aggregate (van den Berg et al. 2000). Lysozyme may be predisposed to aggregation over folding in this experimental system because the starting conformation is a fully denatured and disulfide-reduced state that would not likely exist in vivo. In the cell, protein folding and disulfide formation may be initiated during cotranslational translocation into the ER (Chen et al. 1995; Hardesty et al. 1999).

In summary, we have observed many differences between the structural properties of proteins in solution and their properties in the silica matrix. The general enhancement in thermal stability of encapsulated proteins lends support to the theory-based models of macromolecular crowding and molecular confinement.

However, one protein, apomyoglobin, was found to be highly unfolded on encapsulation even though excluded volume effects should favor the native globular state. We attribute this phenomenon to water structure in the pores of the silica matrix. Because of the large interfacial area that exists in a typical sol-gel glass, a major fraction of the confined water is expected to be perturbed relative to bulk water. The present models of molecular crowding do not account for solvent effects, but water structure is likely to be an important variable in vivo and for crowding experiments in vitro. Further investigations are underway to differentiate the steric effects of confinement from the solvent effects of confined water on the structure of sol-gel encapsulated proteins.

Materials and methods

Reagents

All chemicals were purchased from Sigma-Aldrich (St. Louis, Missouri). Tetramethoxysilane (tetramethylorthosilicate, TMOS) and hexafluoroisopropanol were obtained as Aldrich products. Horse heart myoglobin, hen egg-white lysozyme, bovine α-lactalbumin, and all other reagents were obtained as Sigma products. Apomyoglobin was prepared by 2-butanone extraction of metmyoglobin followed by dialysis against 5 mM potassium acetate and 1 mM DTT at pH 5.0 (Hapner et al. 1968). The structural integrity of apomyoglobin was tested before each use by its ability to rebind hemin in solution, as assayed by UV-vis absorption at 410 nm.

Sol-gel preparation

Silica sol was made using a standard protocol (Ellerby et al. 1992). Typically, 7.61 g TMOS, 1.69 ml water, and 0.11 ml of 0.04 N HCl were mixed and sonicated in plastic centrifuge tubes in contact with an ice-water bath for 20 min. A 3.5 ml volume of the sonicated sol was mixed with 5.25 ml of protein in 2 mM potassium phosphate buffer at pH 7.0. Final protein concentrations after dilution were 0.40–0.70 mg/ml for far-UV CD analysis and 2.5–3.0 mg/ml for near-UV CD analysis. In the case of α-lactalbumin, phosphate was replaced with 2 mM Tris and 0.1 mM CaCl₂ at pH 7.0. For metmyoglobin, best results were obtained when the phosphate buffer was replaced with 10 mM potassium acetate at pH 5.0. Immediately after mixing the silica sol and aqueous protein, the solution was poured between the plates of a plastic cassette (Novex/Invitrogen, Carlsbad, California). Cassettes of 1.0 mm thickness were used for far-UV CD analysis, and cassettes of 1.5 mm thickness were used for near-UV CD. Gelation occurred within 15–30 min. After one day of aging at 4°C, the sol-gel samples were layered with a few milliliters of 2.0 mM buffer and covered with parafilm to prevent drying. Samples were allowed to age for a minimum of two weeks before removing the plastic cassette and storing the silica sheets in distilled water. Wet-aged samples were found to shrink to ∼90% of their original size. To prepare samples for experiments, the silica sheet was cut with a razor blade into rectangular shapes, ∼7.5 mm × 20 mm, for analysis in a quartz cuvette of 2 mm pathlength. Each silica sample was equilibrated with 3.0 ml of the desired buffer solution at room temperature for ≥12 h and immersed in the same buffer during recording of the spectra. Unless stated otherwise in the legends, lysozyme and myoglobin samples were equilibrated in 10 mM potassium phosphate, 50 mM KCl at pH 7.0, and α-lactalbumin samples were equilibrated in 10 mM Tris, 50 mM KCl at pH 7.0.

Circular dichroism

CD spectra were recorded with a J-715 spectropolarimeter (Jasco Inc., Easton, MD). Temperature was controlled with a peltier type thermoelectric cell holder and maintained at 20°C, except for thermal denaturation studies. Stock solution concentrations of proteins were normalized to the following mean residue ellipticity values in solution at 222 nm: metmyoglobin, −25.0 × 10³ deg•cm²•dmol⁻¹; apomyoglobin, −19.0 × 10³ deg•cm²•dmol⁻¹; lysozyme and α-lactalbumin, –10.0 × 10³ deg•cm²•dmol⁻¹. The concentrations of sol-gel encapsulated proteins were calculated based on the final dilution of the stock solutions after addition of buffer and sol. It was assumed that the aged silica samples shrink to the same extent in all dimensions. Thus, the increase in protein concentration due to shrinkage is balanced by the decrease in pathlength through the sample when converting the spectroscopic data to mean residue ellipticity. For some analyses in the far-UV region, spectra were truncated before 200 nm due to high total absorbance. Spectra in the near-UV region were averaged after four accumulations. The percentage of α-helical secondary structure was estimated using the self-consistent method (Sreerama and Woody 1993).

Thermal stability experiments were performed between 15°C and 95°C with a constant heating ramp of 2°C per min. On reaching the endpoint of 95°C, the temperature transition was immediately reversed to cool the sample down at a constant rate of 2°C or 4°C per min. For thermal studies with proteins in solution, a cuvette of 1.0 cm pathlength was employed with stirring.