Protein Science
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print November 6, 2006, 10.1110/ps.062370506
Protein Science (2006), 15:2682-2696. Published by Cold Spring Harbor Laboratory Press. Copyright © 2006 The Protein Society
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
ps.062370506v1
15/12/2682    most recent
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Berger, B. W.
Right arrow Articles by Kaler, E. W.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Berger, B. W.
Right arrow Articles by Kaler, E. W.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us   Add to Digg   Add to Reddit   Add to Technorati  
What's this?

Effects of additives on surfactant phase behavior relevant to bacteriorhodopsin crystallization

Bryan W. Berger1, Colleen M. Gendron2, Abraham M. Lenhoff, and Eric W. Kaler

Center for Molecular and Engineering Thermodynamics, Department of Chemical Engineering, University of Delaware, Newark, Delaware, USA

(RECEIVED May 30, 2006; FINAL REVISION September 11, 2006; ACCEPTED September 12, 2006)


   Abstract
TOP
 Abstract
Introduction
 Results and Discussion
 Conclusions
 Materials and methods
 Acknowledgments
 References
 
The interactions leading to crystallization of the integral membrane protein bacteriorhodopsin solubilized in n-octyl-beta-D-glucoside were investigated. Osmotic second virial coefficients (B22) were measured by self-interaction chromatography using a wide range of additives and precipitants, including polyethylene glycol (PEG) and heptane-1,2,3-triol (HT). In all cases, attractive protein–detergent complex (PDC) interactions were observed near the surfactant cloud point temperature, and there is a correlation between the surfactant cloud point temperatures and PDC B22 values. Light scattering, isothermal titration calorimetry, and tensiometry reveal that although the underlying reasons for the patterns of interaction may be different for various combinations of precipitants and additives, surfactant phase behavior plays an important role in promoting crystallization. In most cases, solution conditions that led to crystallization fell within a similar range of slightly negative B22 values, suggesting that weakly attractive interactions are important as they are for soluble proteins. However, the sensitivity of the cloud point temperatures and resultant coexistence curves varied significantly as a function of precipitant type, which suggests that different types of forces are involved in driving phase separation depending on the precipitant used.

Keywords: protein–detergent complex; membrane protein crystallization; cloud point temperature; osmotic second virial coefficient; self-interaction chromatography


   Introduction
TOP
 Abstract
 Introduction
Results and Discussion
 Conclusions
 Materials and methods
 Acknowledgments
 References
 
Membrane proteins represent one of the most significant challenges in modern structural biology. Genome sequence analysis indicates that 20%–40% of the open-reading frames of most organisms encode membrane proteins, yet they represent <1% of the available structures in the Protein Data Bank (PDB) (Wallin and von Heijne 1998; Wiener 2004). Although several structural genomics initiatives have recently begun to focus on membrane proteins—particularly G-protein coupled receptors (GPCRs)—difficulties in expression, purification, and crystallization remain (Loll 2003; Lundstrom 2004). In particular, identifying a suitable detergent to solubilize a given membrane protein, thereby forming a protein–detergent complex (PDC), and combining this with the appropriate additives and precipitants necessary for crystal formation while retaining PDC stability adds several variables that must be considered beyond those for soluble proteins (Rosenbusch 1990; Garavito et al. 1996). For example, including 24 common detergents and additives as variables in a typical crystallization screen using conventional 24-well formats can result in the need to sample >3000 potential solution conditions, which makes such efforts largely impractical for all but highly automated processes where substantial protein is often available (McPherson 2004). Given the additional complexities of membrane protein stability and surfactant phase behavior once in a PDC, such screens often provide little guidance as to how such underlying factors influence crystallization (Loll 2003). Therefore, developing rational approaches to membrane protein crystallization, particularly by understanding how additives and surfactants influence PDC crystal formation, can significantly improve existing methods.

Much effort has been focused on the role of surfactant phase behavior in membrane protein crystallization (Rosenbusch 1990; Garavito et al. 1996; Loll et al. 2001; Wiener 2001). In particular, many nonionic surfactants exhibit a lower critical solution temperature (LCST) or "cloud point" at elevated temperatures above their critical micelle concentration (CMC). Above the LCST, the sample becomes turbid and over time forms two equilibrium phases, one rich and the other lean in surfactant (Weckstrom and Zulauf 1985; Garavito et al. 1986). There are two explanations for this transition. In one view, spherical micelles may grow into cylindrical micelles, and increased micellar interactions lead to phase separation (Zulauf and Rosenbusch 1983; Glatter et al. 2000). An alternative explanation involves the formation of a saturated micellar network (Zilman and Safran 2002, 2003).

Experimentally, such phase transitions have been observed using n-octyl-beta-D-glucopyranoside (C8betaG1), decyl-N,N-dimethylamine oxide (DDAO), or octyl-polyoxyethylene (octyl-POE) for the crystallization of outer membrane protein F from Escherichia coli. Extended networks of surfactant–surfactant contacts were observed throughout the protein crystal lattice (Pebay-Peyroula et al. 1995). A similar network of interconnected surfactant regions was also observed in crystals of the reaction center from Rhodobacter sphaeroides using C8betaG1 (Roth et al. 1991). However, in the case of bacteriorhodopsin (BR), crystallization using C8betaG1 was complicated by the fact that significant precipitation was observed near the phase boundary (Michel and Oesterhelt 1980). This led to the "small amphiphile" concept, in which a cosolute such as heptane-1,2,3-triol (HT) or L-pipecolinic acid is added to a PDC solution to suppress surfactant phase separation and presumably to modulate surfactant structure about the PDC, thereby facilitating crystal formation and growth (Michel 1983). Experimentally, adding either HT or L-pipecolinic acid raises the cloud point temperature of the free micelles, whereas the effect on the PDC remains unclear (Schertler et al. 1991). Ultimately, a combination of lower surfactant concentration and mixture of such cosolutes proved effective in improving BR crystal nucleation and growth from a PDC (Schertler et al. 1993). Formation of membrane protein crystals has been observed near a cloud point in other cases as well, inspiring the development of screens based on surfactant phase behavior (Scarborough 1994; Song and Gouaux 1997; Loll et al. 2001; Wiener and Snook 2001). However, the influence of additives and precipitants on both surfactant phase behavior and membrane protein crystal formation is still an area of active research.

An approach that has been successful in identifying crystallization conditions for soluble proteins is to characterize protein interactions in terms of the osmotic second virial coefficient (B22), which is defined by (McQuarrie 2000):



Formula 1

The indices i, j, and k refer to individual solutes in a multicomponent mixture; {rho} is the total solute number density, x is the mole fraction, B is the osmotic virial coefficient, {Pi} is the osmotic pressure, R is the universal gas constant, and T is temperature. For a binary mixture, Equation 1 simplifies to the more familiar (George et al. 1997; Neal et al. 1999):



Formula 2

cp is the protein concentration (in mass units), and MW is the protein molecular weight. As such, B22 is a first-order correction to ideal solution behavior that, in this case, accounts for interactions between pairs of protein molecules in solution. Positive B22 values indicate an increase in the osmotic pressure of the solution, associated with repulsive interactions, whereas negative B22 values indicate the opposite.

In a pioneering study, George and Wilson (1994) observed that solution conditions leading to crystallization for nine soluble proteins gave B22 values within the range of –0.8 to –8.0 mol mL g–2, which correspond to weakly attractive protein–protein interactions. Based on these results, they proposed this range of B22 values as the "crystallization slot," thereby providing a quantitative correlation from which to identify potential crystallization conditions. This has led to further studies of crystallization patterns and interactions for a variety of model proteins aimed at understanding their relationship to B22 (Rosenbaum and Zukoski 1996; George et al. 1997; Velev et al. 1998; Neal et al. 1999; Tardieu et al. 2002; Tessier et al. 2003).

For surfactant-solubilized integral membrane proteins, the solution is comprised of not only PDCs but also surfactant micelles and monomers, all of which are influenced by precipitants and additives. Therefore, it is necessary to consider a multicomponent virial expansion for a PDC solution (Hill and Chen 1973; Tessier et al. 2004):



Formula 3

In this case, subscript 2 refers to the PDC, and subscript 3 to the surfactant micelle. Thus, B22 reflects PDC–PDC interactions similar to protein–protein interactions described in Equation 2, B23 reflects PDC-micelle interactions, and B33 reflects micelle–micelle interactions. Although much effort has been focused on measuring PDC–PDC interactions (B22), the results below illustrate, at least qualitatively, the importance of micelle–PDC (B23) and micelle–micelle (B23) interactions as well. Both substantially influence the patterns of PDC–PDC interactions and hence B22.

The crystallization slot concept has been extended to PDC interactions and crystallization as well. In particular, characterization of PDC interactions for outer membrane protein F from E. coli using mixtures of octyl-POE and either C8betaG1 or n-octyl-2-hydroxyethylsulfoxide led to the observation that favorable crystallization conditions gave rise to PDC B22 values in a range of –0.5 to –2.0 mol mL g–2, which is similar to the crystallization slot for soluble proteins (Hitscherich et al. 2000). Furthermore, similar B22 values for micellar solutions at the same conditions near the cloud point were obtained, suggesting that not only are the surfactant portions of the PDC playing a significant role in promoting crystal formation, but that the surfactant cloud point might also be used as a guide to identify weakly attractive PDC interactions favorable for crystallization (Hitscherich et al. 2000; Loll et al. 2001).

Use of the crystallization slot concept to guide protein crystallization has not been widely adopted due to experimental limitations, however. Conventional techniques for measuring B22, such as static light scattering, neutron scattering, analytical ultracentrifugation, or membrane osmometry, often require milligram quantities of protein per measurement as well as significant amounts of time and specialized equipment (Vilker et al. 1981; Velev et al. 1998; Behlke and Ristau 2003; Tessier and Lenhoff 2003; Wilson 2003). This is a particular problem for membrane proteins, since their low natural abundance and toxicity during overexpression in various hosts often limit yields to <1 mg of functional membrane protein per liter of culture (Tate 2001). Self-interaction chromatography (SIC), however, can address this difficulty by significantly reducing the amount of time and protein necessary per measurement (Tessier et al. 2002). By immobilizing the protein of interest onto chromatographic particles, packing them into a column, and then injecting a pulse of the same protein into the mobile phase, one can interpret the changes in retention in terms of B22 (Tessier et al. 2002). Furthermore, SIC can allow separation of the various components in solution, thereby allowing the interactions between them to be determined unambiguously (Tessier et al. 2004). This is important in membrane protein crystallization, where solutions contain a mixture of PDCs and free micelles, often of similar size, rather than PDCs alone, each of which may have considerably different patterns of interaction. With use of a HPLC system and an autosampler, this process can be automated and significantly reduce the amount of time and protein necessary per measurement.

Here we present osmotic second virial coefficients, measured by SIC, for the integral membrane protein bacteriorhodopsin solubilized in n-octyl-beta-D-glucoside. A previous report (Berger et al. 2005b) focused on the role that salts play in influencing BR-C8betaG1 PDC interactions, particularly in relation to the corresponding C8betaG1 phase behavior. Salts such as ammonium sulfate and sodium malonate, both of which have been shown to be effective in promoting crystallization of soluble proteins, have a dramatic effect on micelle structure at the high concentrations necessary for crystallization, and these effects may actually be detrimental to inducing attractive PDC interactions (McPherson 2001). However, by tuning the surfactant phase behavior by lowering the C8betaG1 concentration, one can overcome the adverse effects brought about by high salt concentrations and affect crystallization at lower salt and surfactant concentrations near the cloud point. Given that many membrane proteins crystallize in the presence of alkanediol additives or polyethylene glycol, however, we sought to examine their effects on surfactant phase behavior and micelle and PDC interactions in order to develop effective approaches for their use in membrane protein crystallization as well. The results illustrate some surprising patterns of PDC and micelle interactions relevant to crystallization that can be explained in the context of the corresponding C8betaG1 phase behavior.


   Results and Discussion
TOP
 Abstract
 Introduction
 Results and Discussion
Conclusions
 Materials and methods
 Acknowledgments
 References
 
Effects of heptane-1,2,3-triol on phase behavior and interactions using salts
Figure 1 shows the effect of HT on PDC B22 values. In the absence of HT, addition of sodium malonate up to concentrations above 1 M makes the interactions increasingly repulsive and thus is not conducive to crystallization. Addition of HT in a range of 10–80 mM causes a substantial reduction in apparent PDC repulsion at low to moderate ionic strengths for sodium malonate, with the apparent strong repulsion eliminated above 40 mM HT in all cases (Fig. 1). Similar results were obtained using sodium formate and ammonium sulfate in conjunction with HT (data not shown; Berger 2005; Berger et al. 2005b).


Figure 1
View larger version (31K):
[in this window]
[in a new window]

  		Figure 1. Comparison of PDC interactions in the presence of heptane-1,2,3-triol: ({blacktriangleup}) 0 mM HT, ({circ}) 40 mM HT, (•) 80 mM HT. BR concentration was 1 mg/mL, and C8betaG1 concentration 40 mM in all cases. All interaction measurements were made at 20°C.

 
This trend is consistent with quasi-elastic light scattering (QLS) results for the 80 mM HT–40 mM C8betaG1 mixture (Fig. 2), which show the apparent micelle radius of the mixture to remain ~5 nm over the entire range of sodium malonate concentrations used. In contrast, the apparent micellar radius of 40 mM C8betaG1 grows to >40 nm at 1 M sodium malonate, reflecting both an increase in micelle size as well as attractive surfactant interactions. CMC measurements for C8betaG1 and HT as well as mixtures thereof also indicate that at high salt concentrations, the CMC of the mixture is substantially higher than that of C8betaG1 alone (data not shown; Berger 2005). This reflects the much higher apparent critical aggregation concentration or solubility limit of ~1 M for HT relative to the CMC of C8betaG1, as discussed in detail below. The SIC and DLS results suggest that by reducing the apparent micelle size at high ionic strength, PDC repulsion can be reduced. Furthermore, adding HT also expands the range of conditions for which attractive PDC B22 values are found that fall within the crystallization slot (Fig. 1), with a range of 1.1–1.4 M sodium malonate giving rise to crystal formation. Examples of solution conditions that led to PDC crystal formation and growth using sodium malonate are given in Table 1. It is important to note that in the current study, the surfactant concentration is fixed at 40 mM C8betaG1, whereas in previous studies the same salt concentrations were used, but at varying surfactant concentration (Berger et al. 2005b).


Figure 2
View larger version (23K):
[in this window]
[in a new window]

  		Figure 2. Change in apparent micelle size with heptane-1,2,3-triol: ({blacktriangleup}) 0 mM HT, (•) 80 mM HT. The C8betaG1 concentration was 40 mM; and temperature, fixed at 20°C.

 


View this table:
[in this window]
[in a new window]

  		Table 1. Summary of virial coefficients for sodium malonate in conjunction with polyhydric alcohol additives and associated patterns of phase behavior

 
Comparison of the B22 results and cloud point measurements, however, shows that attractive interactions within the crystallization slot occur at solution conditions near the 40 mM C8betaG1 cloud point temperature of 20°C, rather than at the cloud point of the 80 mM HT–40 mM C8betaG1 mixture (Fig. 3). Surprisingly, no corresponding cloud point temperature of the HT–C8betaG1 mixture is observed under any conditions where HT is present; not until nearly 1.5 M sodium malonate is reached does the cloud point temperature for the C8betaG1-HT mixture appear near 90°C. At first, this may seem to contradict previous observations that attractive PDC interactions leading to crystallization occur near a surfactant liquid–liquid phase boundary. However, when the PDC is included at a concentration of 0.5 mg/mL BR (Fig. 3), the resultant cloud point temperatures fall within 20°C of those for C8betaG1 at a given sodium malonate concentration, rather than those for the HT-C8betaG1 mixture.


Figure 3
View larger version (25K):
[in this window]
[in a new window]

  		Figure 3. Differences in C8betaG1 cloud point temperature with HT and BR: ({blacktriangleup}) 0 mM HT, (•) 80 mM HT, ({bigtriangleup}) 80 mM HT with 0.5 mg/mL BR. The C8betaG1 concentration was 40 mM.

 
BR is known to have high thermal stability in its native membrane environment to temperatures above 100°C, and although this is significantly reduced once the BR is incorporated into a PDC, BR remains stable over the time necessary for such cloud point measurements to be made below 50°C (Shen et al. 1993; Heyes and El-Sayed 2002, 2003). This was confirmed by monitoring the absorbance at 550 nm and 380 nm using UV-visible spectroscopy (data not shown; Berger 2005). Inactivated BR due to loss of covalently bound retinal has an absorbance maximum near 390 nm, so elevated temperatures that caused rapid inactivation were excluded from further analysis. Furthermore, the rate at which BR aggregates increases with protein concentration as well as temperature at high salt concentrations; therefore, cloud point temperature measurements up to 50°C were limited to concentrations below 1 mg/mL BR. However, the difference in cloud point temperatures in the presence of the PDC suggests that HT affects micelles and PDCs differently.

Effects of alkanediols on phase behavior and interactions using salts
Alkanediols such as hexane-1,2-diol (1,2-HD) and hexane-1,6-diol (1,6-HD) are also common additives in membrane protein crystallization; in particular, use of 1,6-HD led to the successful crystallization of lactose permease from E. coli (Abramson et al. 2003). As with HT, increasing concentrations of 1,6-HD or 1,2-HD reduce the apparent PDC repulsion at low to moderate salt concentrations (Fig. 4). The reduced PDC repulsion coincides with a decrease in the apparent micelle size and insensitivity to salt concentration as well for mixtures of both 40 mM 1,2-HD and 40 mM 1,6-HD with 40 mM C8betaG1 (data not shown; Berger 2005). Similarly to HT, the hexanediols act to reduce micelle size, allowing attractive B22 values for BR to occur under a wider range of solution conditions, including a larger window within the crystallization slot (Fig. 4). However, the range of salt concentrations that leads to B22 values in the crystallization slot differs for 1,2-HD and 1,6-HD. Specifically, 1,2-HD leads to attractive PDC interactions at 1 M sodium malonate, which is significantly lower than the 1.6 M sodium malonate necessary for 1,6-HD, and the values for 1,2-HD remain within the slot over a significantly wider range of salt concentrations.


Figure 4
View larger version (32K):
[in this window]
[in a new window]

  		Figure 4. Changes in PDC interactions with hexanediols: ({blacktriangleup}) 0 mM hexanediol, ({diamondsuit}) 40 mM hexane-1,2-diol, ({blacksquare}) 40 mM hexane-1,6-diol. BR concentration was 1 mg/mL, and C8betaG1 concentration 40 mM in all cases. All interaction measurements were made at 20°C.

 
Interestingly, the specific structure of the hexanediol used dictates the magnitude of the cloud point temperature (Fig. 5). Whereas adding 40 mM 1,6-HD raises the cloud point temperature nearly 30°C relative to 40 mM C8betaG1 at a given salt concentration, adding 40 mM 1,2-HD actually lowers it by nearly 30°C. Furthermore, the shifts in cloud point temperatures using the hexanediols are reflected in the onset of attractive PDC interactions (Fig. 4); as mentioned before, crystallization slot values for 1,2-HD–C8betaG1 mixtures are shifted toward lower salt concentrations than for 1,6-HD–C8betaG1 mixtures, and 1,2-HD lowers the cloud point temperature of the mixture. In the case of 1,6-HD, the range of salt concentrations leading to attractive PDC B22 values is similar to that near the 40 mM C8betaG1 cloud point temperature rather than that of the 40 mM 1,6-HD–40 mM C8betaG1 mixture; this pattern is similar to that observed for HT. However, when comparing cloud point temperatures for the hexanediol mixtures including the PDC, one sees significant differences in the phase behavior compared to the surfactant mixture alone only in the case of 1,6-HD. As was observed with HT, the 1,6-HD-C8betaG1 mixture including 0.5 mg/mL BR has a cloud point temperature much closer to that of C8betaG1 alone, again suggesting that hexane-1,6-diol affects micelles and PDCs differently. In the case of 1,2-HD, however, the cloud point temperatures of the surfactant mixture with and without 1 mg/mL BR are within 10°C of one another. Therefore, unlike HT or 1,6-HD, 1,2-HD has similar effects on both micelles and PDCs.


Figure 5
View larger version (30K):
[in this window]
[in a new window]

  		Figure 5. Differences in C8betaG1 cloud point temperature with hexanediols and BR: ({blacktriangleup}) 0 mM hexanediol, ({blacksquare}) 40 mM 1,6-HD, ({square}) 40 mM 1,6-HD with 0.5 mg/mL BR, ({diamondsuit}) 40 mM 1,2-HD, ({diamond}) 40 mM 1,2-HD with 0.5 mg/mL BR. The C8betaG1 concentration was 40 mM.

 
A similar pattern is observed for other alkanediols in terms of the relationship between alkanediol structure, PDC interactions, and surfactant phase behavior. The decrease in cloud point temperature for the mixture with C8betaG1 varies directly with the alkane chain length of the 1,2-diol used, whereas the increase in cloud point temperature for the alkane-1,n-diols generally increases with chain length, although the sensitivity is much lower than for alkane-1,2-diols (data not shown; Berger 2005). This pattern is also followed in terms of attractive PDC interactions, where, with increasing alkane-1,2-diol chain length, the crystallization slot B22 values occur at progressively lower salt concentrations (Fig. 6). However, the longer chain heptane-1,7-diol and octane-1,8-diol both caused liquid–liquid phase separation and aggregation of BR at high salt concentrations, and those studies were therefore not pursued (see below).


Figure 6
View larger version (31K):
[in this window]
[in a new window]

  		Figure 6. Patterns of interaction for PDCs with 40 mM 1,2-alkanediols and sodium malonate: ({blacktriangleup}) hexane-1,2-diol, ({blacksquare}) heptane-1,2-diol, ({diamondsuit}) octane-1,2-diol. The C8betaG1 concentration was 40 mM. All interaction measurements were made at 20°C.

 
A potential explanation for the differences in phase behavior with structure of the additives is differences in their amphiphilicity. Tensiometry, vapor phase osmometry, and isothermal titration calorimetry experiments indicate that 1,2-HD is a weak amphiphile with a CMC of 0.81 M, whereas HT displays only weak surface activity near 1 M, and 1,6-HD shows no clear surface activity or aggregation phenomena (Hajji et al. 1989; Coudert et al. 1993; Hajji 2003). Detailed investigations using small-angle neutron scattering (SANS) for a broad range of alkanediols and alkanetriols also found that a transition from weakly clustered, disordered aggregates to well-defined micelle shapes occurred with HT, whereas longer-chain alkanediols such as 1,2-HD and alkanetriols led to formation of elongated, micelle-like structures (D'Arrigo et al. 2000). Thus, 1,2-HD, which displays an amphiphilic character similar to that of nonionic surfactants, is likely partitioning into the surfactant portion of the PDC as well as into the micelles, thereby affecting both in a way similar to that described previously by the small amphiphile concept (Michel 1983). In contrast, 1,6-HD and HT act indirectly through diminishing micelle interactions and size. In this sense, the latter two additives function as a competitive substrate for free surfactant in solution, thereby shifting the distribution in favor of HT–C8betaG1 mixed micelles and sequestering surfactant that interacts with the PDC, especially at the high concentrations at which such additives are typically used for membrane protein crystallization.

Effects of polyethylene glycol on phase behavior and interactions
Polyethylene glycol (PEG) is perhaps the most common precipitant used in protein crystallization; use of PEG accounts for production of >75% of all protein crystals for which the structures have been solved to high resolution (McPherson 1999). This number includes membrane proteins, where PEG is often used in conjunction with a salt such as sodium chloride or ammonium sulfate (Michel 1990). As a result, PEG has been the subject of several studies involving protein phase behavior and interactions leading to crystallization (Kulkarni et al. 2000; Loll et al. 2002; Tardieu et al. 2002; Vivares et al. 2005). Unlike salts such as ammonium sulfate or sodium malonate, however, PEG influences both the upper critical solution temperature (UCST) and the lower critical solution temperature (LCST) in mixtures with C8betaG1 (Wormuth 1991; Loll et al. 2001; Santonicola and Kaler 2005).

Use of PEG 3350 or PEG 8000 leads to an increase in attractive interactions near the surfactant phase boundary. Comparison of the C8betaG1 phase diagrams (Fig. 7) with the PDC interaction measurements (Fig. 8) shows that the pattern of increasing PDC attraction is similar to that of the salts in that the onset of attraction occurs once surfactant cloud point temperatures fall within an experimentally observable range. A summary of solution conditions that led to crystal formation using PEG is given in Table 2, where it is clear that there is good agreement between solution conditions where crystal formation was observed, and PDC B22 values near the crystallization slot. Likewise, the range of PEG concentrations that lead to B22 values within the crystallization slot is at a similar separation in terms of temperature from that at the cloud point. Interestingly, the apparent PDC repulsion observed with salts at low to moderate concentration (Fig. 1) is absent with PEG. This gives a much wider range of solution conditions (12–16 wt% PEG 3350) that fall within the crystallization slot without the use of additives such as alkanediols.


Figure 7
View larger version (23K):
[in this window]
[in a new window]

  		Figure 7. Differences in C8betaG1 cloud point temperature with PEG: ({diamondsuit}) PEG 3350, ({blacktriangleup}) PEG 8000. The C8betaG1 concentration was 40 mM.

 


Figure 8
View larger version (28K):
[in this window]
[in a new window]

  		Figure 8. Changes in PDC interactions in the presence of PEG: ({diamondsuit}) PEG 3350, ({blacktriangleup}) PEG 8000. BR concentration was 1 mg/mL, and C8betaG1 concentration 40 mM in all cases. All interaction measurements were made at 20°C.

 


View this table:
[in this window]
[in a new window]

  		Table 2. Summary of virial coefficients for PEG in conjunction with salts and associated patterns of phase behavior

 
Addition of salt also has a significant effect on the liquid–liquid phase boundary of C8betaG1–PEG mixtures as well as on PDC interactions. Addition of sodium chloride or ammonium sulfate, both of which are commonly used salts in conjunction with PEG 3350 for protein crystallization, at a concentration of 0.5 M raises the cloud point temperature of the PEG 3350–C8betaG1 mixtures by ~10°C (Fig. 9). Interestingly, both salts have a similar effect on the cloud point temperature in a range of 10–16 wt% PEG, whereas above 16 wt%, a significant increase in the cloud point temperature is seen for NaCl relative to ammonium sulfate. Raising the salt concentration to 1 M does not significantly increase the cloud point temperature beyond that at 0.5 M (data not shown); most of the sensitivity of the phase boundary to salt occurs in the concentration range 0–0.5 M (Berger 2005). In both cases, however, the decrease in cloud point temperatures relative to experimental conditions at room temperature is reflected in the increase in attractive PDC interactions with lower PEG concentrations (Fig. 10). In other words, as the cloud point temperature of the surfactant solution approaches experimental conditions, attractive micelle and PDC interactions increase.


Figure 9
View larger version (29K):
[in this window]
[in a new window]

  		Figure 9. Differences in C8betaG1 cloud point temperature with PEG 3350-salt mixtures: ({diamondsuit}) 0 M NaCl, (•) 0.25 M NaCl, ({blacktriangleup}) 0.5 M NaCl, ({square}) 0.5 M ammonium sulfate. The C8betaG1 concentration was 40 mM.

 


Figure 10
View larger version (34K):
[in this window]
[in a new window]

  		Figure 10. Changes in PDC interactions with PEG 3350-salt mixtures: ({diamondsuit}) 0 M NaCl, (•) 0.25 M NaCl, ({blacktriangleup}) 0.5 M NaCl, ({square}) 0.5 M ammonium sulfate. The BR concentration was 40 mM, and C8betaG1 concentration was 40 mM. All interaction measurements were made at 20°C.

 
This pattern of interaction is similar to that when salts are used as precipitants in the presence of heptane-1,2,3-triol or the hexanediols. In those cases, diminishing attractive surfactant interactions led to a decrease in micelle size and polydispersity, allowing more attractive PDC interactions to occur. This conclusion was reached in previous studies with C8betaG1–C8E4 mixtures that used PEG 2000 to induce liquid–liquid phase separation (Hitscherich et al. 2001; Loll et al. 2002). Using a combination of light, neutron, and X-ray scattering, it was argued that the apparent increase in micelle size observed near the phase boundary is related mainly to attractive micelle interactions, which decrease the effective diffusivity measured and increase apparent hydrodynamic size, rather than changes in micelle size. Other scattering experiments have also pointed to attractive micelle interactions being dominant over changes in micelle size near a liquid–liquid phase boundary for both polyoxyethylene surfactants such as C8E4 as well as alkyl polyglucosides such as C9betaG1. In these cases, though, it was not possible to separate effects based on size and interactions (Zulauf et al. 1985; Ericsson et al. 2004). However, recent SANS studies on polyoxyethylene surfactant micelle size and attractive interactions that specifically focused on examining both effects separately indicate that micelles first undergo structural changes far from the cloud point temperature, and then as the phase boundary is approached, attractive micelle interactions occur (Glatter et al. 2000).

One interesting observation from this study is that only higher MW PEGs, such as 3350 and 8000, lead to phase separation (Fig. 7), whereas PEG 400 and 800 do not (Berger 2005). This may be a consequence of polymer–polymer interactions, which become significant in the semidilute region above a critical overlap concentration (c*) and also depend on chain length (Larson 1998). The relevant concentration range for phase separation of C8betaG1 is in the semidilute regime, where overlap between polymers leads to formation of extended, network structures. SANS studies on PEG–C8betaG1 solutions have interpreted the appearance of apparent micelle aggregates as being due to crossover into this regime; it is also noteworthy that the onset of attractive B22 values for BR PDCs in the current study for both PEG 3350 and PEG 8000 occurs near their respective c* concentrations, and no such PDC attraction is seen for smaller MW PEG. Such a mechanism involving PEG overlap into the semidilute region has also been suggested to explain differences in micelle shape and attractive interaction leading to phase separation of C8betaG1-octyl-POE mixtures using PEG 2000 (Hitscherich et al. 2001).

Protein crystallization
Based on the above SIC results, 30 solution conditions involving alkanediols, alkanetriols, and PEG were chosen that gave PDC B22 values within the crystallization slot, and these samples were subjected to batch and vapor diffusion crystallization using a hanging drop technique. Sample images for representative cases using different combinations of salts, alkanediols, and PEG are given in Figure 11.


Figure 11
View larger version (158K):
[in this window]
[in a new window]

  		Figure 11. Crystallization patterns for BR–C8betaG1 PDC using alkanediol additives. Corresponding osmotic second virial coefficients for each panel are shown in Table 1. Images were taken using a Zeiss Axioskop 2 microscope. The approximate size of each panel is 500 x 500 µm. Panels a–c were obtained using pH 3.5, 10 mM sodium phosphate, 1.25–1.5 M sodium malonate, 100–400 mM 1,6-hexanediol, 40 mM C8betaG1, 15–20 mg/mL BR, and 10°C, whereas d was obtained using 100–400 mM 1,8-octanediol rather than 1,6-hexanediol. (a) Liquid–liquid phase separation (~24 h), (b) apparent crystal formation (1–2 d), (c) appearance of well-faceted crystals, (d) liquid–liquid phase separation and aggregation using 1,8-octanediol.

 
The first set of crystallization experiments was focused on examining the patterns of phase behavior in the presence of alkanediols in order to understand how they may influence PDC crystal formation and growth. In particular, the following conditions were chosen due to the reproducible nature of crystals obtained: pH 3.5, 25 mM sodium phosphate, 1.25–1.5 M sodium malonate, 100–400 mM hexane-1,6-diol, 40 mM C8betaG1, 15–20 mg/mL BR, and 10°C. The higher hexane-1,6-diol concentrations used in crystallization versus in the measurements of PDC B22 values (Fig. 4) and cloud point temperature (Fig. 5) were necessary to induce phase separation and crystal formation before significant denaturation of BR occurred. Overall, the entire process of crystal formation and growth took ~7–10 d, during which time various phases could be observed (Fig. 11). Within 24 h, liquid–liquid phase separation occurred, as evidenced by the appearance of large, clear droplets (Fig. 11a). These droplets appeared in clusters and were often large enough to form extensive contacts that spanned regions of the hanging drop. During this time, BR could be found at the interface between the droplets and the surrounding solution, rather than dispersed within the droplets. After ~1–2 d, the droplets gave way to a dense, textured phase, in which highly concentrated regions of BR could be observed (Fig. 11b). These concentrated BR regions ultimately grew into well-faceted crystals, while the surrounding, textured phase receded (Fig. 11c). These crystals diffracted to 4 Å. Longer alkyl chains, such as heptane-1,7-diol or octane-1,8-diol, led to liquid–liquid phase separation, where BR selectively partitioned into the droplet rather than clustering at the interface (Fig. 11d).

Liquid–liquid phase separation was also observed for crystallization using PEG (Fig. 12a). From these droplets, needle-like clusters began to form, eventually growing into long, rod-like crystals over a period of 7–10 d (Fig. 12b,c). Interestingly, the two phases persisted over the course of several weeks as the concentrated BR in solution was taken up into the crystals. These crystals also remained in one of the two liquid phases. Addition of salts and alkanediol or alkanetriol additives had no marked impact on the liquid–liquid phase separation or crystal formation. BR crystals grown using PEG diffracted weakly with a resolution of 8 Å.


Figure 12
View larger version (51K):
[in this window]
[in a new window]

  		Figure 12. Crystallization patterns for BR–C8betaG1 PDC using PEG 3350. Corresponding osmotic second virial coefficients for each panel are shown in Table 1. Images were taken using a Zeiss Axioskop 2 microscope. The approximate size of each panel is 200 x 200 µm. Conditions were pH 3.5, 10 mM sodium phosphate, 40 mM C8betaG1, 15–20 mg/mL BR, and 10°C. (a) Liquid–liquid phase separation (~24 h); (b,c) apparent crystal formation (1–2 d).

 
Correlation between the cloud point and B22, and its significance for membrane protein crystallization
This study points to the importance of surfactant interactions in promoting PDC crystallization, as reflected in the appearance of a C8betaG1 cloud point in the vicinity of crystallization slot solution conditions independent of the additives or precipitants used. This was observed previously in a few specific cases for outer membrane protein F, where attractive B22 values occurred at high PEG concentrations approaching the octyl-POE cloud curve (Hitscherich et al. 2001). As mentioned previously, attractive B22 values for the corresponding micelle solutions followed a similar trend to that of the PDCs, suggesting that the surfactant may play a constructive role in PDC crystallization. In order to draw comparisons in the present case, 40 measured osmotic second virial coefficient values for the PDC were chosen over a wide range of solution conditions for which a corresponding C8betaG1 cloud point temperature was measured. A similar analysis for BR has been made previously, in terms of the dimensionless quantities (Berger et al. 2005b):



Formula 4



Formula 5

The temperature, T, at which the interaction measurements and crystallization trials were made was 20°C in all cases.

The results (Fig. 13) show that a strong correlation exists between the measured B22 value of the BR–C8betaG1 PDCs and the corresponding cloud point temperature of C8betaG1. This clearly demonstrates the significance of C8betaG1 in the overall process of PDC crystallization and also illustrates the advantage that self-interaction chromatography provides in generating a wide range of virial coefficient values for membrane proteins. Furthermore, in the context of the crystallization slot, slightly negative values of B22 occurred at conditions that lead to crystallization in a corresponding range of ~20°–30°C below the C8betaG1 cloud point temperature for the salt solutions, whereas for PEG solutions this offset was only ~1°–5°C. In particular, for PEG-induced crystallization at 20°C, though the difference between the PEG concentrations within the crystallization slot (7–9 wt%) and at the cloud point temperature (16%) may appear large on an absolute scale (Figs. 7, 8), the various conditions collapse onto a single line on a reduced scale (Fig. 13), though the range of interaction is much longer than in the case of salts. The observed scatter at lower Tr may be a result of changes in the amount of surfactant bound to the PDC near the cloud point temperature, which in turn affects the diameter and molecular weight used in estimation of the hard-sphere contribution to B22. Likewise, the scatter is much more pronounced for salt solutions than for PEG, which might be due to the enhanced sensitivity of C8betaG1 cloud point temperatures to salts (Figs. 3, 5) than to PEG (Figs. 6, 8). In all cases, however, solution conditions that led to crystallization tended to fall within a range similar to that of the crystallization slot for soluble proteins, suggesting that similar weakly attractive forces are predominant under these conditions.


Figure 13
View larger version (23K):
[in this window]
[in a new window]

  		Figure 13. Correlation between B22 and cloud point temperature for the entire set of additives and precipitants used: ({blacktriangleup}) PEG as precipitant, ({bigtriangleup}) salts as precipitants.

 
The above correlation provides additional evidence that conditions leading to protein crystallization are often correlated with the liquid–liquid phase boundary of a protein solution. However, the mechanisms of crystal formation and growth, particularly in relation to liquid–liquid phase separation, appear different when a salt or PEG is used as the precipitant. In the case of salts driving phase separation, liquid–liquid phase separation occurs largely independent of the presence of additives, but the additives seem to play a role in the formation of the observed liquid droplets. Specifically, HT, 1,2-HD, and 1,6-HD all lead to clustering of BR at the droplet interface, from which crystal formation and growth apparently occur, whereas longer-chain alkanediols such as 1,7-heptanediol lead to a disperse liquid phase or partitioning of BR into the liquid droplets. Enhanced crystal formation and growth for BR from a PDC at an interface has been observed in several cases using benzamidine (Schertler et al. 1993). Likewise, there is speculation that the presence of a surfactant bilayer lamellar phase is a precursor from which BR crystals form and stack as in lipidic cubic phase crystallization (Nollert et al. 2001). The current work provides further evidence that such crystal formation from an interface occurs, although the dependence on additive type suggests that the major effect promoting crystal formation is the ability to modulate solubility of the BR in solution.

In many ways, this pattern of phase behavior resembles microemulsion formation for alkyl polyglucosides, where, in order to create a stable emulsion using hydrophobic oils, a cosolute such as 1,6-HD is often added to improve the surfactant solubility in the oil phase (Ryan and Kaler 2001). It is interesting to speculate whether such effects are taking place here as well.

In the case of PEG, however, crystal formation and growth occur after liquid–liquid phase separation within the droplet, even in the presence of alkanediol and alkanetriol additives, which is in direct contrast to the salt-induced crystal formation and growth at the droplet interface. This pattern of crystal formation and growth using PEG is very similar to that observed for outer membrane protein F from E. coli and B800–850 complex (Michel 1990). Although the reasons for the different mechanisms of crystal formation and growth are unclear, the results of this study suggest that PEG perturbs micelle structure much less than salts, and therefore BR solubility in a PDC is likely enhanced using PEG versus salts as precipitants, such that the integrity of the surfactant belt surrounding the transmembrane region is conserved.

Studies of lysozyme have found that the solid–liquid or crystallization temperature is directly proportional to the liquid–liquid temperature, largely independent of pH, salt type and concentration (Broide et al. 1996). Static light scattering studies on lysozyme also found that B22 values that led to crystallization fell within a narrow range of conditions at a given solubility (Rosenbaum et al. 1999). However, B22 values at the metastable liquid–liquid phase boundary varied as a function of solution conditions. These observations are all consistent with the current study, in which the magnitude and sign of B22 were similar in all cases for crystallization conditions. However, in the presence of salts, the B22 versus cloud point temperature values all fell within one band, whereas for PEG the results were shifted to another line at higher cloud point temperatures. In other words, the same range of B22 typically led to crystallization, while the corresponding cloud point temperatures varied. This may be due to the fact that the critical temperature is changing as a function of solution conditions, reflecting the different underlying interactions driving phase separation.


   Conclusions
TOP
 Abstract
 Introduction
 Results and Discussion
 Conclusions
Materials and methods
 Acknowledgments
 References
 
These results confirm that the osmotic second virial coefficient B22 is an effective measure of PDC interactions relevant to crystallization. In particular, for the bacteriorhodopsin and n-octyl-beta-D-glucoside (C8betaG1) protein–detergent complex (PDC), the onset of attractive interactions occurs near the surfactant cloud point temperature largely independent of the precipitants or additives used.

Additives such as HT and 1,6-HD, which raise the cloud point temperature of surfactant solutions, are likely affecting the free micelles in solution rather than PDCs to promote crystallization. This may be a result of the fact that HT and 1,6-HD are weakly surface-active, with no apparent critical micelle concentration (CMC), behaving more as cosolutes rather than cosurfactants to provide an alternate hydrophobic surface with which the surfactant monomers can interact. On the other hand, hexane-1,2-diol, which lowers the cloud point temperature and causes a shift to lower salt concentrations for the onset of attractive PDC interactions, has a clear CMC of 0.81 M. Therefore, it may be that the surface activity of 1,2-HD allows it to partition into the surfactant portion of the PDC, thereby influencing its interactions.

Comparing B22 values at the various solution conditions for which a surfactant cloud point temperature was observed reveals a strong correlation. This clearly points to the constructive role that the surfactant plays in promoting PDC crystallization and is similar to the type of ordering observed for colloidal particles near a liquid–liquid phase boundary in structured surfactant solutions. Although the same range of weakly attractive B22 interactions tends to promote crystallization for both salts and PEG, the corresponding cloud point temperatures are different, reflecting the different underlying forces involved in phase separation. It will be interesting to see whether additional membrane proteins and surfactants display similar behavior.


   Materials and methods
TOP
 Abstract
 Introduction
 Results and Discussion
 Conclusions
 Materials and methods
Acknowledgments
 References
 
Materials
1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) (22980), N-hydroxysuccinimide (NHS) (24500), Tris-(2-carboxyethyl)-phosphine hydrochloride (TCEP) (20460) and the Micro BCA Protein Assay Kit (components 23231, 23232, and 23234) were obtained from Pierce. Bacteriological peptone used for growth of Halobacterium halobium was from Oxoid (L37). 2-(N-morpholino)ethanesulfonic acid (MES) (M8250), sodium malonate (M4795), sodium formate (F4166), sodium acetate (241245), ammonium sulfate (A4915), heptane-1,2,3-triol (H6011), and additional cell culture reagents (molecular biology grade) were from Sigma. Hexane-1,2-diol (AC20013), hexane-1,6-diol (AC16035), heptane-1,2-diol (AC12035), heptane-1,7-diol (AC12035), and octane-1,2-diol (AC20014) were from Fisher Scientific. n-octyl-beta-D-glucoside (C8betaG1) (O311) was from Anatrace. AF-Amino-650M particles (08002) were from Tosoh Biosep. Borosilicate glass microcolumns (3 x 50 mm) (993301) were from P. J. Cobert and Associates.

Expression and purification of BR
Purple membrane (PM) was expressed and purified from Vac- Rub- Halobacterium halobium strain ET1001, which constitutively overexpresses BR, as described previously (Oesterhelt 1982; DasSarma and Fleischmann 1995). For solubilization, purified PM was diluted to 0.5 mg/mL in 10 mM MES buffer (pH 6.5), and 100 mM C8betaG1 was added. The solution was allowed to stand for 36 h in the dark at 20°C, after which residual PM was removed by centrifugation for 1 h at 50,000 rpm using a Beckman SW55-Ti rotor. Inactivated BR due to loss of bound retinal has an absorbance maximum near 380 nm; inactivation due to long-term storage could be limited, however, by keeping purified BR PDCs in 40 mM C8betaG1 solution containing 10% (v/v) glycerol at 4°C in the dark (Schertler et al. 1991). Solutions stored in 10% (v/v) glycerol showed no evidence of inactivation over a period of 6 mo.

Self-interaction chromatography
Preparation of immobilized BR particles for self-interaction chromatography, column packing, and characterization were essentially as described previously (Tessier et al. 2002; Berger et al. 2005b). Typical immobilization densities were 10–15 mg of BR per milliliter of settled particles, corresponding to ~20% surface coverage. All running buffers contained C8betaG1 at concentrations above the CMC to minimize BR aggregation. For experiments using additives such as alkanediols or alkanetriols, these were added at the appropriate concentration to the C8betaG1 running buffer and allowed to mix for at least 1 h prior to use. Injection concentrations of 0.5–1.5 mg/mL BR were found to be best for SIC measurements, and all samples were centrifuged at 90,000 rpm for 1 h prior to use to remove aggregates.

Sample runs and determination of retention factors are essentially as described previously (Berger et al. 2005b). To ensure that the measurements were not influenced by increases in viscosity, particularly at conditions near the C8betaG1 cloud point, viscosities of C8betaG1 mixtures with alkanediols using salts or PEG as precipitants were measured using an Ubbelohde capillary viscometer (Canon Instrument Co.) according to the manufacturer's instructions. All viscosity measurements were made at 20° ± 0.1°C as set by a temperature-controlled water bath and repeated five times to ensure reproducibility. No significant change in viscosity was observed until conditions near the C8betaG1 liquid–liquid phase boundary were reached; these conditions were excluded from further analysis due to peak asymmetry and appearance of significant aggregates as discussed previously (Berger et al. 2005b).

B22 was calculated according to the relation (Tessier et al. 2002):



Formula 6

{rho}s refers to the immobilization density, or the amount of protein immobilized per unit accessible surface area of the pore; {varphi} is the phase ratio, which describes the total accessible surface area of the pore per unit volume mobile phase; and k' is the chromatographic retention, which is measured for each solution condition of interest. Methods for determining each of these quantities for SIC have been discussed previously (Tessier et al. 2002, 2003; Berger et al. 2005a). Note that in this case, B22 refers to PDC–PDC interactions, whereas micelle–micelle and micelle–PDC interactions are considered separately (Equation 4). A discussion of how SIC may be extended to multicomponent solutions in order to measure each contribution separately is also available (Tessier et al. 2004). The variability in B22 from multiple independent measurements was <5% in all cases.

Due to the varying nature of the surfactant microstructure as a function of solution conditions, estimates of PDC radii were made using quasi-elastic light scattering (QLS; see below) in order to calculate the excluded volume contribution to B22 (B 22 HS ). The apparent radii were found to remain relatively constant at ~5 nm until solution conditions near the cloud point of the surfactant were reached, where the apparent PDC size would converge to clusters with an apparent average radius of ~200 nm. QLS results illustrating the changes in apparent micelle size are given in Results and Discussion, so further details are deferred to the discussion (see Fig. 2). For calculation of the excluded volume contribution to B22, the average apparent PDC radius at a given solution condition was used to calculate the equivalent spherical volume of the PDC, which was in good agreement with the volume of a BR–C8betaG1 PDC determined by NMR (Gottschalk et al. 2001).

Cloud point temperature measurements
Cloud point temperature measurements for mixtures involving C8betaG1 were determined by visual inspection as described previously (Berger et al. 2005b); the cloud point temperature is defined as the lowest temperature at which the solution spontaneously becomes turbid. This method has been successfully applied to determining the phase behavior of a wide variety of nonionic and zwitterionic surfactants (Weckstrom and Zulauf 1985; Blankschtein et al. 1986; Balzer 1996; Liu et al. 1996; Koehler and Kaler 1997). The resultant phase was determined from its ability to refract polarized light as well as by direct observation using an Olympus BHT microscope equipped with a temperature-controlled flow cell.

Quasi-elastic light scattering
Quasi-elastic light scattering measurements were made using a Brookhaven instrument with a BI200SM goniometer and BI9000AT digital correlator. The light source was a Lexel 488 nm Ar laser operated at 100 mW. The hydrodynamic radius was obtained from the diffusion coefficient using the Stokes-Einstein relation (McQuarrie 2000). Solutions were allowed to incubate for 12 h at a given temperature prior to analysis. Measurements were repeated five times to ensure reproducibility. It is important to note that the increase in micelle size near the cloud point temperature cannot be attributed entirely to micellar growth, but likely reflects both micellar growth and increasingly attractive surfactant interactions (Glatter et al. 2000). Therefore, results are reported in terms of apparent radius to reflect both of these effects.

CMC determination
The surface tensions of C8betaG1 mixtures containing alkanediols, salts, or PEG at 20°C were measured according to the Wilhelmy plate method using a Kruss tensiometer as described previously (Berger et al. 2005b). The CMC was determined from the breakpoint in the surface tension versus log CMC curve at a given solution condition.

CMC values were also measured independently using a Microcal isothermal titration calorimeter. In a typical experiment, 10 µL samples of concentrated C8betaG1 solution were injected into a 1.5-mL sample chamber containing sample buffer. The injections were chosen such that the final C8betaG1 concentration spanned a range of 0.1–100 mM in order to include pre-CMC and post-CMC concentrations; C8betaG1 has a CMC in water of 28 mM. By comparing the change in heat released with the change in surfactant concentration, the CMC can be determined from the maximum in the heat profile. The heats of dilution of buffer injections were negligible.

Crystallization
Crystallization of BR was performed using hanging-drop and microbatch methods. In all cases, crystals were grown in the dark at 4°C. For hanging drop vapor diffusion experiments, 3 µL of a 20 mg/mL PDC solution at a given pH and surfactant concentration was mixed by pipetting with 1 µL of well solution and 1 µL of additive solution at the same pH on a clean, siliconized coverslip (Hampton Research), and then set up against 500 µL of well solution. For hanging drop batch experiments, 3 µL of a 20 mg/mL PDC solution at a given pH, surfactant, and additive concentration was added to 2 µL of a concentrated precipitant solution at the same pH and mixed by pipetting. The drop was set up against a reservoir containing the same precipitant and additive concentration using 24-well crystallization trays (Hampton Research). Crystals appeared within 1–2 wk as determined by optical microscopy, whereas precipitation and surfactant phase separation could often be observed within 1–2 d. Pictures at various stages of crystal formation and growth were collected using a Zeiss Axioskop 2 microscope at 1–10 x magnification with a Zeiss Axiocam digital camera and Axioview (version 3.1) image processing software.

In a related set of experiments aimed at examining interfacial effects at the liquid–liquid interface on crystal formation, BR PDC solutions at conditions that led to significant crystal formation and growth were prepared and subjected to ultracentrifugation at 45,000 rpm for 1 h. Afterward, the samples were removed immediately; in cases where liquid–liquid phase separation occurred, the surfactant-rich (upper) phase containing all of the BR was removed using a pipette and placed on a clean, siliconized coverslip. The supernatant was examined by optical microscopy using a Zeiss Axioskop 2 Microscope at 1–10x magnification with Zeiss Axiocam digital camera and Axioview (version 3.1) image processing software.

X-ray diffraction and data collection
X-ray diffraction data for BR PDC crystals were collected using a Rigaku RU300 rotating anode generator with a RAXIS IV image plate area detector. Samples were removed from the droplet using a 0.3-mm loop and flash-frozen in liquid nitrogen. The programs DENZO and SCALEPAK were used for indexing, data processing, and scaling (Otwinowski and Minor 1997).


   Footnotes
 
1 Present addresses: Department of Biochemistry and Biophysics, School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA; Back

2 Pfizer Global Research and Development, Groton, CT 06340, USA. Back

Reprint requests to: Bryan Berger, Department of Biochemistry and Biophysics, School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA; e-mail: bergerbw{at}mail.med.upenn.edu; fax: (215) 573-7039; or Eric Kaler, Department of Chemical Engineering, University of Delaware, 102 DuPont Hall, 150 Academy Street, Newark, DE 19716, USA; e-mail: kaler@udel.edu; fax: (302) 831-6751.

Abbreviations: C8betaG1, n-octyl-beta-D-glucoside; BR, bacteriorhodopsin; HT, heptane-1,2,3-triol; 1,6-HD, hexane-1,6-diol; 1,2-HD, hexane-1,2-diol; PEG, polyethylene glycol; SIC, self-interaction chromatography.

Article published online ahead of print. Article and publication date are at http://www.proteinscience.org/cgi/doi/10.1110/ps.062370506.


   Acknowledgments
TOP
 Abstract
 Introduction
 Results and Discussion
 Conclusions
 Materials and methods
 Acknowledgments
References
 
This publication was made possible by NIH grant P20 RR-15588 from the COBRE Program of the National Center for Research Resources and NASA grant NAG8-1830 from the Microgravity Research Program. B.W.B. gratefully acknowledges support through a NIH Chemistry–Biology Interface Training Grant T32 GM-08550 and NSF IGERT Graduate Fellowship DGE-0221651. We are especially grateful to Prof. George Turner (Seton Hall University) for providing H. halobium strain ET1001, Chad Blamey and Prof. Brian Bahnson for assistance with X-ray diffraction and analysis of BR crystals, as well as Gabriella Santonicola, Dr. Kirk Czymmek (Delaware Biotechnology Institute), and Dr. Tzvetana Lazarova (Universitat Autónoma de Barcelona).


   References
TOP
 Abstract
 Introduction
 Results and Discussion