Molecular Effects of Anionic Surfactants on LysozymePrecipitation and Crystallization
O. D. Velev,† Y. H. Pan,‡ E. W. Kaler, and A. M. Lenhoff*
Department of Chemical Engineering, University of Delaware, Newark, Delaware 19716
Received April 23, 2004; Revised Manuscript Received July 15, 2004
ABSTRACT: Surfactants are used as additives in some protein separations and crystallization procedures, butthe mechanism of their action is still poorly understood. By measuring the osmotic second virial coefficients oflysozyme solutions by static light scattering, we show that small amounts of anionic surfactants of varying molecularweight always make the protein-protein interactions in solution more attractive and that both the charge of theheadgroup and the length of the hydrophobic tail mediate the interactions. The same surfactants also modify lysozymecrystallization and promote formation of twinned phases with gross morphologies different from those seen in theabsence of surfactant, some of which display remarkably structured patterns on a micrometer scale. The surfactanteffects are, however, often only kinetic, as the phases obtained initially recrystallize slowly into large stable crystals.These crystals are of excellent X-ray diffraction quality and resolution (up to 1.4 Å). Their symmetry, unit celldimensions, crystal contacts, and protein backbone conformation are the same as those commonly observed forlysozyme, but sometimes occur at atypical pH values. The data suggest new techniques for modification of proteincrystallization.
Introduction
Increasing the efficiency with which proteins can becrystallized and improving the quality of the crystalsobtained are problems of major importance for struc-tural biology and for industrial separations. A largevariety of chemical agents are commonly added duringprotein crystallization to effect crystal growth.1 Suchadditives often include amphiphilic surfactant mol-ecules, with hydrophilic groups that can either benonionic (e.g., beta-octyl glucoside and polyoxyethylene-based surfactants) or ionic (e.g., decanoic and dodecanoicacid, sodium dodecyl sulfate, and cetyltrimethylammo-nium bromide). Surfactants may be present duringprotein crystallization even if they are not purposelyadded because amphiphilic molecules such as detergentsand lipids may persist as residual impurities afterprotein purification.
The effects of surfactants on protein crystallizationare, however, still poorly characterized and understood.A few studies have demonstrated that the addition ofnonionic surfactants can improve the quality of proteincrystals.1-5 However, little insight is available about themolecular mechanisms by which the surfactants modifythe crystallization process. In an earlier experimentalstudy,6 we characterized by quantitative fluorescencemicroscopy the infusion of lysozyme crystals with pyrene-based fluorescent surfactants. The originally nonfluo-rescent protein crystals slowly become fluorescent dueto the uptake of the surfactant into the crystal lattice,demonstrating that surfactants are adsorbed and ac-cumulate in the protein crystals. However, it has notbeen determined whether the surfactant affects thedynamics of nucleation and/or growth, or the protein
packing parameters of the crystals. Other questions atthe molecular level include whether the surfactantmolecules are bound in specific locations and orienta-tions within the crystal structure, or whether they aremerely amorphously adsorbed.
The effect of surfactants on crystallization is basedon the interaction of the surfactant with the protein.Protein-surfactant interactions in solution have beenstudied and characterized in detail previously in severalsystems utilizing ionic surfactants. Surfactant moleculesbind to proteins in solution, with the energy of thatinteraction found to be on the order of 10 kT in twoindependent studies.7,8 Coprecipitation can be expectedwhen the surfactant/protein molar ratio becomes ap-proximately equal to the net charge of the protein ofsign opposite to that of the surfactant.9-11 Largerquantities of strong detergents may solubilize theprotein molecules and cause their denaturation byunfolding of the protein.12,13
This paper reports characterization of hen egglysozyme interactions and crystallization in the presenceof small amounts of amphiphilic molecules with ananionic headgroup, with the goal of understanding howthe type and size of the surfactant hydrophobic andhydrophilic groups affect protein crystallization. Theexperiments were carried out with two homologousseries of common amphiphiles, aliphatic sulfates andcarboxylic acids of hydrocarbon chain length varyingfrom C6 to C12. The lower molecular weight amphiphilesused are generally considered to be “mild” and do notcause protein precipitation from solution (notably, mostof these molecules do not exhibit detergency and theterm “surfactants” is used here in the broader definitionof amphiphilic molecules that combine hydrophilic andhydrophobic groups). The surfactant/protein molar ra-tios range from less than 1:1 to 10:1. Due to the lowconcentrations used, formation of free surfactant mi-celles is not expected in any of these systems, and most
* Author for correspondence. Phone: 302-831-8989. E-mail:[email protected].
† Present address: Department of Chemical Engineering, NorthCarolina State University, Raleigh, NC 27695-7905.
‡ Present address: Department of Chemistry and Biochemistry,University of Delaware, Newark, DE 19716.
of the interactions involved are between pairs of inde-pendent molecules.
The effects of surfactants on the protein-proteininteractions in solution were estimated via the osmoticsecond virial coefficient measured by static light scat-tering (SLS). This parameter captures well the repul-sive-attractive balance of the interactions between theprotein molecules and can be used as a predictor forprotein crystallization.14-18 The effects of the am-phiphiles on the kinetics and morphology of crystalgrowth were followed, and the crystals obtained werecharacterized by X-ray diffraction to evaluate theirquality and to resolve whether the surfactants becomeincorporated in the crystal lattice in an ordered fashion.
Experimental Section
Materials. Lysozyme was obtained from Sigma (6× crystal-lized, L-6876). The solutions for SLS were prepared withdeionized water from a Millipore Milli-Q system. All solutionscontained 0.3 M NaCl, 10 mg/L NaN3 (both from Sigma) and3 × 10-4 M citrate buffer (citric acid from Fisher Scientific).The pH of the protein solutions was adjusted precisely to theexperimental value, 6.5 or 10, by adding small amounts of 0.1M HCl or NaOH while continuously stirring and measuringthe pH. The pH of the batch electrolyte solutions used fordiluting the protein samples was also adjusted in the sameway and sustained by the small amount of buffer present.Sodium octanoate and sodium dodecyl sulfate were obtainedfrom Sigma and all other surfactants from Aldrich. Specialcare was taken to minimize the presence of dust to obtainreliable SLS data, and to minimize the presence of externalnuclei in the crystallization. All sample preparation procedureswere carried out in a clean airflow bench (NuAir from LaminarFlow Products). The electrolyte solutions were filtered through20 nm cutoff inorganic Anotop filters (Whatman). The proteinsolutions were prepared less than 2 h before the SLS measure-ments and were filtered through 100 nm Anotop filters andimmediately diluted and sealed in ampules and vials.
Methods. The SLS data were obtained using a BrookhavenBI9000 light-scattering apparatus, equipped with a laser ofwavelength 488 nm. All measurements were performed at ascattering angle of 90° at 25 ( 0.1 °C. The absolute Rayleighratios of the samples were calculated by calibration with purebenzene before each experimental measurement. The detailsof the procedures used for data collection and processing aredescribed in detail in ref 17.
Lysozyme crystals were grown by a batch method in whichthe protein solutions were left for up to 12 months at 4 °C inplastic vials and glass ampules sealed with Teflon tape. Theplastic vials were 1.5 mL polypropylene microcentrifuge tubesfrom Sigma, handled on the clean bench. The 2 mL glassampules from Wheaton (NJ) were treated first with detergent,followed by Nochromix oxidizing solution, washed with deion-ized water, and dried in a clean environment before use. Themorphology of the growing crystals was examined periodicallyby microscopy through the walls without opening the contain-ers, to prevent contamination. The microscopy observationswere carried out in regular and polarized transmitted lighton a Nikon Optiphot-2. Water evaporation from the glass vialsamounting to up to 50 vol % was observed over periods ofseveral months.
At the end of the experimental cycle, diffraction data werecollected on some of the crystals at room temperature and-180 °C. The room temperature data were collected, forcrystals with typical dimensions of 0.35-0.5 mm, on a RigakuRU-200 (RAXIS-II) diffractometer with 0.3 µm collimation. Therotating anode was operated at 50 kV and 80 mA and a 1.5degree oscillation was used for each orientation. The low-temperature data were collected on a Rigaku RU-300 (RAXIS-IV) with a double focusing mirror system with 0.3 µmcollimation. The rotating anode was operated at 50 kV and
100 mA and one degree oscillation was used for each orienta-tion. Most of the crystals selected for data collection weresmaller than 0.3 mm to minimize mosaicity, and they werefrozen in a cryoprotectant solution made up of 30% glycerolin mother liquor.
Structural refinements of the data sets were carried outusing the programs XPLOR 3.1 and CNS 0.9a. All reflectionswere used without amplitude or sigma cutoff. Refinementswere carried out in several stages, starting from the lowerresolution range and progressing to the high-resolution limit.A typical refinement cycle at higher resolution included energyminimization by the conjugate gradient method, and re-strained individual b-factor refinement followed by simulatedannealing at 3000 K. Visual inspection of molecular modelswas carried out on a SGI graphic workstation using Chain19
and O.20 Water molecules were added to the model at a laterstage of refinement using both automatic water placementprograms in XPLOR and by manual inspection. Waters withtemperature factors greater than 50 Å2 were excluded fromthe model.
Results
Static Light Scattering. The main purpose of theSLS measurements was to compare the effect on thevirial coefficient of surfactants with two different head-groups and tails ranging from C8 to C12, all at asurfactant/protein molar ratio of 1:1. The experimentswere performed at two different pH values, 6.5 and 10.As shown earlier,14-18 the virial coefficients of lysozymesolutions at both these pH values are negative over arange of ionic strengths and are within the “crystalliza-tion slot” associated with conditions conducive to crys-tallization.14,15 Our results, presented graphically inFigure 1, show that the surfactant-containing solutionsalways have lower virial coefficients than the pureprotein solutions, indicating a shift to more attractiveintermolecular potentials. At pH 10, for example, thealkyl sulfates gradually decrease the virial coefficientfrom -5.5 to -7.2 × 10-4 mol mL/g2 and then causeprecipitation, as expected for highly attractive interac-tions below the crystallization slot.14,15 As sodium do-decyl sulfate is a strong detergent and causes immediateprecipitation even at a molar ratio of 1:1, the data inthe last column were collected at an SDS/lysozyme ratio
Figure 1. A summary of the effect of small quantities ofsurfactants with increasing hydrocarbon tail length on thelysozyme virial coefficients in solution measured by SLS. Allsolutions are at a surfactant/protein molar ratio of 1:1, exceptfor the last column (SDS). The error bar shown for one systemis characteristic of all virial coefficient measurements reportedhere.
of 1:10, where no precipitation was observed. Dodecanoicacid has a very low solubility at pH 6.5, so no data couldbe collected under these conditions.
The uniform decrease in the value of the virialcoefficients can be attributed to the binding of thesurfactant molecules and their consequent effects onprotein-protein interactions. As a rule, the proteinsolutions at pH 10, where the protein has a smallerinitial charge, have lower virial coefficients than thoseat pH 6.5, and since binding of the negatively chargedsurfactant would also offset the positive charge on the
protein, the relative electrostatic contribution to thechange in B22 is significant. Additional effects clearlyplay a role as well, though. First, comparison of thesurfactants from the two homologous series, sulfatesand carboxylic acids, shows that the systems withsulfates always have lower virial coefficients and startprecipitating first. Second, the B22 values decreasemonotonically with increasing hydrophobic tail lengthof the surfactants, until precipitation begins (Figure 1).
The effect of surfactant concentration can be exploredfor the two “mildest” surfactants from the homologousseries, which allow higher concentrations to be attained(Figure 2). As expected, in all systems the addition ofsurfactant leads to more attractive interactions, withespecially strong effects below surfactant/protein molarratios of 1:1 (sulfate) or 1:2 (carboxylate). The minimaat intermediate ratios are larger than the characteristicerrors in B22 (see error bar in Figure 1) and thusprobably reflect real interactions. However, the originof this effect is still unclear. It may arise, for instance,from formation of specific structures of two proteinmolecules bound by one or two intermediate surfactantmolecules; this effect could be lost when more am-phiphilic molecules become adsorbed on the protein andrandomize the structure.
All the data consistently demonstrate that, for thesecases, protein-protein interactions in solution can bemanipulated toward stronger attraction by addingselectively chosen surfactants. As shown below, thischange in the interaction energy leads to significantchanges in the crystallization pattern of lysozyme.
Crystal Morphology and Its Evolution withTime. The slightly negative virial coefficients measuredfor the lysozyme solutions without additives are withinthe crystallization slot for proteins, and indeed inexperiments without surfactants, crystals were observedafter an initial nucleation period of a few weeks to afew months.17 The crystals grown at pH 6.5 were of theusual tetragonal symmetry (Figure 3a), but at pH 10the crystal symmetry changed to orthorhombic (Figure3b), which correlates with the significantly lower virialcoefficients measured at that pH.17 No recrystallizationor differences between different vials of a given samplewere observed.
In contrast, when surfactants were added to thesamples a rich variety of crystal morphologies were
Figure 2. Effect of the surfactant/protein molar ratio on thelysozyme virial coefficients measured for the case of (a)octanoic acid and (b) octyl sulfate. The virial coefficients arealways in the more attractive region, but there is a specificallystrong effect at certain low molecular ratios.
Figure 3. Microphotographs of lysozyme crystals without surfactant: (a) tetragonal crystals grown at pH 6.5; (b) orthorhombiccrystals at pH 10. Scale bars ) 200 µm.
observed. These formed relatively quickly and exhibitedsubsequent metamorphoses and recrystallizations. Thefirst solid phase seen in many of the samples was a fineand visibly amorphous precipitate. Although no pre-cipitation was observed during the initial hours in thecourse of the light-scattering experiments at 25 °C,small amounts of precipitate typically formed after theprotein solutions were refrigerated overnight. Virtuallyno precipitate was observed in the samples at pH 6.5,except for a single case with the highest amount (10:1)of octanoic acid. Some degree of precipitation wasobserved in most of the samples at pH 10, with thelargest amounts in the solutions with sulfates.
A variety of crystals with different morphologiesatypical for solutions without surfactants were observedin the subsequent few weeks. In most cases, these couldbe recognized as morphologically twinned crystals con-sisting of ingrown clusters of different orientation. Thesamples at pH 6.5 usually grew rhomboidal-shapedcrystals such as the ones of tetragonal symmetrynormally seen at this pH (Figure 3a), but in thesamples with the highest concentration of octyl sulfate(3:1) the rhomboidal crystals were accompanied by aspecific “starlike” twinned modification shown in Figure4.
A much larger variety of crystalline morphologiesgrew from the protein solutions originally precipitatedat pH 10. Within two weeks, the precipitate recrystal-
lized into twinned crystals, the morphology of whichdepended on the chemical nature of the surfactant.Starlike twinned crystals similar in appearance to thosein Figure 4 were exclusively observed in solutions ofsulfates. Samples with higher molecular weight alkanoicacids demonstrated another specific “signature” in thegrowth of twinned phases: ball-like formations of large,clearly distinguishable, presumably tetragonal crystals(Figure 5a). These crystals then often grew in onespecific direction, forming pillared stacks of crystals inthe form of cylinders (Figure 6).
The common morphology observed with the shorter-chain octanoic and decanoic acids was large crystal-likefaceted chunks, which usually did not display anyspecific gross shape or symmetry (Figure 5b). Theprotein phase in the chunks was, however, crystalline,as proven by the rotation of polarized light. A commonfeature of many of these chunks was the presence oflines of different optical density when viewed along onedirection of observation, suggesting that the crystalsrepresent a specific twinned morphology. These stripessuggest that the chunks are probably twinned in theform of stacks of crystalline layers and may be a productof a step bunching growth mechanism. In some samplesof lower protein concentration, the layered chunks grewas remarkably symmetric crystals with fascinatingperiodically layered structures on a micrometer scale(Figure 7).
Figure 4. Crystals with starlike morphology grown in the presence of alkyl sulfates: (a) bright field illumination of crystalsgrown at pH 6.5 in the presence of octyl sulfate; (b) microphotograph of the same sample in crossed polarizers shows that thestructures, albeit twinned, are crystalline. Scale bars ) 200 µm.
Figure 5. Two modifications of the twinned crystals obtained in the presence of fatty acids at pH 10. (a) Single small crystalsand ball-like twinned formations; (b) shapeless layered chunk from crystalline protein obtained with octanoic acid. Scale bar (a)) 200 µm, (b) ) 100 µm.
Additional transformations were observed after thesolutions were left unperturbed for periods ranging froma few weeks to a few months. A summary of the generalpath followed in the recrystallizing solutions is pre-sented in Figure 8; it should be emphasized, however,that the course of recrystallization in any given sampleappeared to follow a random path, with several differenttwinned phases frequently coexisting with the separatedcrystals. The starlike plates formed with sulfates usu-ally recrystallized slowly into large chunky crystalssimilar to those observed with carboxylates (Figure 5b).More importantly, all of the samples evolved within 4-8months into rhomboid- or needle-shaped crystals ap-pearing identical to those seen under the same condi-tions in the absence of surfactants (Figure 3). Theultimate formation of either rhomboid or needlelikecrystals at the final stage was invariant; although bothcrystal morphologies were observed, the plastic vialsappeared to favor the needlelike modifications. In manysamples, these highly symmetric crystals grew to sizesof a few millimeters.
X-ray Characterization. Crystallographic data werecollected from 14 crystals from different batch experi-ments, selected to sample a wide range of surfactanttypes, protein concentrations, and solution pH. Thesolution conditions represented are listed in Table 1,along with data collection and refinement statistics forthese crystals. All the data sets were collected to aresolution of 1.4-2.0 Å. The rhomboidal and the layeredchunky crystals were tetragonal, belonging to space
Figure 6. Bright-field (a) and dark-field (b) images of twinned crystals with “pillared” appearance obtained at pH 10 in thepresence of dodecanoic acid. Scale bars ) 200 µm.
Figure 7. Highly symmetric structures slowly grown in the presence of octanoic acid at pH 10. (a) Large layered crystal; (b)close-up of the structure showing the regular striped pattern. Scale bar (a) ) 200 µm, (b) ) 50 µm.
Figure 8. Schematic of the general scheme of the temporalevolution of the morphology of the lysozyme crystals obtainedin the presence of small amounts of anionic surfactants.
group P43212, while the needlelike crystals were ortho-rhombic with a space group of P212121. The structureswere solved by molecular replacement using the atomiccoordinates of hen egg lysozyme files 6LYT and 193Lfrom the Protein Data Bank as the starting models forthe tetragonal and orthorhombic form, respectively. Allstructural refinements of the data sets were carried toR values of about 20% with good geometry.
The refined molecular structure of the lysozyme inall crystals was very similar to that of the initial PDB
models (193L or 6LYT); the RMS differences in all atomsand in the main chain atoms were no greater than 1.6and 1.2 Å, respectively. When compared with thestructures crystallized without surfactant, the RMSdistances in all atoms were no greater than 1.4 Å inP43212 symmetry and 0.9 Å in P212121 symmetry.Difference Fourier maps, calculated using the structurefactors obtained with and without the presence of thesurfactant, i.e., Fo⟨with surfactant⟩ - Fo⟨without sur-factant⟩, revealed no distinctive features that can be
Table 1. Solution Conditions and Crystal Collection and Refinement Statistics
Orthorhombic, Space Group P212121, Z ) 4surfactanta none C8 C8 C10Ssurfactant/protein 1:1 10:1 1:1pH 10 10 10 10
Data Collectiona (Å) 30.5b 30.8 30.8 30.4b
30.4b
b (Å) 57.9b 58.8 59.1 57.7b
57.7b
c (Å) 68.1b 68.2 68.3 67.5b
67.7b
diffraction limitc (Å) 1.60b 2.00 2.00 1.40b
1.40b
measured reflections 110955b 46043 42638 471765b
155404b
unique reflections 16379b 8388 8435 20153b
22219b
Rmerged (%) 3.9b 6.9 10.3 3.5b
3.4b
overall completeness (%) 99.9b 95.6 96.2 92.7b
91.8b
completeness (%, last shell) 99.8b 73.0 82.8 79.1b
37.1b
R value (last shell) 0.187b 0.144 0.288 0.269b
0.274b
Refinementresolution limit (Å) 1.6 2.0 1.5 1.5R value 0.196 0.198 0.206 0.178Rfree value 0.246 0.256 0.266 0.267reflections used 15700 7855 19106 18211waters 168 82 168 150
Tetragonal, Space Group P43212, Z ) 8surfactanta none C8 C8 C8S C10 C10S C10Ssurfactant/protein 1:1 10:1 1:1 1:1 1:1 1:1pH 6.5 6.5 6.5 6.5 10 6.5 10
Data Collectiona ) b (Å) 78.1b 78.9 78.9 77.0 77.3b 78.9 78.8
77.8b 77.2b
c (Å) 37.2b 38.2 38.1 38.5 37.3b 38.0 38.037.1b 37.7b
attributed to ordered amphiphilic or solvent molecules.In summary, the lysozyme structures, as well as thewater shell surrounding the protein, are very similarwithin each symmetry group for crystals obtained withor without surfactant, and no ordered surfactant mol-ecules were found in any of the refined structures.
There are, however, two notable features of the dataregarding the effect of the surfactant. The first is thefairly high-resolution achieved with many of the sur-factant-containing crystals. The resolution limit of 1.4Å is better than the limit of 1.6-1.7 Å achieved in theabsence of surfactant in this and in a previous study17
under the same procedures of batch crystallization.These high resolution data are comparable to the bestresults available in the PDB (most of which wereobtained on crystals grown in space). Thus the presenceof surfactant does not adversely impact the proteincrystal quality measurably, and may actually enhanceit, similarly to crystallization data obtained by oth-ers,1,4,5 although more carefully controlled experimentswould be needed to confirm this. The second effect,which can be specifically attributed to the presence ofthe additives, is the perturbation of many of the samplesat pH 10 to the tetragonal crystal form, contrary toprevious experiments without surfactant, where onlyorthorhombic crystals formed at this pH.17 This ran-domly observed change, despite the absence of surfac-tant effects on the respective crystal structures, suggeststhat the two crystal forms are very similar in freeenergy, and could be inferred to be a direct result ofsurfactant interference with crystal nucleation andgrowth during the spontaneous recrystallization pro-cess.
Discussion
Due to the difference in the molecular weight of theprotein and the surfactants used, the physical amountof surfactant in the solutions for the SLS and crystal-lization experiments is small, typically ranging from ca.0.01 to 0.1 wt %; this may approach the level ofamphiphilic impurities in the protein or in the crystal-lization media. The light scattering and crystallizationdata demonstrate that even this relatively small amountof surfactant can substantially change protein-proteininteractions and may be responsible for changes in theprotein crystallization pattern. Thus, the presence ofsurfactants or lipids in the crystallization media maybe responsible in part for the crystal twinning andirregular growth sometimes observed in protein crystal-lization. However, our results also demonstrate bene-ficial effects of the surfactants, which can be used tocontrol the crystallization process and enhance crystalquality.
The virial coefficient data show that the anionicsurfactants effectively shift the protein-protein interac-tions toward greater attraction. As the amphiphiles areoppositely charged to the net charge on the protein, onelikely mechanism for the increased attraction is elec-trostatic, i.e., surfactant binding decreases the charge-charge repulsion among the protein molecules. Thedifference in the strength of action of the sulfates andcarboxylates exemplifies the importance of the type ofcharged group. Similar effects of tighter electrostaticbinding of protein to sulfates compared to carboxylates
have been observed previously in protein adsorption onchromatographic media with different surface groups21
and have been explained on the basis of hydrationeffects.22
The data also show, however, that the hydrophobicprotein-surfactant interactions are also of primaryimportance in mediating the protein-protein interac-tions. One possible mechanism for this is hydrophobicanchoring of the surfactants on the protein, allowingbetter neutralization of the surface charges. An alterna-tive explanation is that an amphiphile can bridgeadjacent protein molecules by electrostatic binding toone of those molecules and hydrophobic adsorption onthe other one. This could dampen energetically unfavor-able conformations, such as when a strongly chargedhydrophilic group on the surface of one of the proteinsapproaches a hydrophobic patch on the surface of theother one. Alternatively, hydrophobic interactions mayoccur between amphiphiles electrostatically anchoredon different protein molecules.
The additional attractive component induced viasurfactant perturbation of the protein-protein interac-tions is the obvious reason for the protein precipitationobserved and the wide variety of morphologies seenupon subsequent recrystallization. Our earlier data onthe energy of ionic surfactant adsorption in proteincrystals gave an estimate of 10-12 kT per surfactantmolecule;6 this value is similar to literature values foradsorption of alkyl sulfates on proteins in solution.7,8
Such adsorption is strong enough that it could wellinterfere with protein crystal nucleation and growth. Anintriguing aspect of the crystallization abnormalitiesobserved here is that the morphologies of the twinnedphases depend specifically on the surfactant type.Understanding the way in which this surfactant “sig-nature” develops can provide interesting leads to explainand inhibit twinning in practical crystallization pro-cesses.
The long-term observations of the metamorphoses ofthe protein phase reveal that the twinning effectsobserved are kinetic in origin and ultimately transient.The most favorable crystalline symmetry and basic celldimensions, which the crystals ultimately adopt, cor-respond to those obtained without surfactant. The onlyeffect of the surfactant at these long times is seen inthe perturbation of the crystal form registered at pH10, i.e., the random formation of both tetragonal andorthorhombic crystals. The fact that no ordered surfac-tant molecules are found in the crystal structures showsthat the surfactant executes its kinetic effects via subtleinterference with the pattern of the protein-proteininteractions.
A simple model can qualitatively explain the behaviorobserved in both light scattering and crystallizationexperiments. It has been well recognized23 that thediscrete protein-protein interaction profile along acertain configuration coordinate (e.g., the angular ori-entation of the molecules) exhibits a series of minimaand maxima corresponding to more attractive andrepulsive configurations, respectively (Figure 9). Someof the deepest global minima may correspond to con-figurations found in the most extensive crystal contacts.As surfactant adsorption on the protein may dampensome of the unfavorable configurations, it could in
general lead to more energetically favorable protein-protein interactions (Figure 9), which is witnessed inthe measured lower virial coefficient.
On the other hand, it is very unlikely that adsorbedsurfactant molecules will disrupt the crystal contacts,which involve highly complementary configurations ofpairs of oppositely charged groups and/or hydrophobicpatches. The energy gained upon formation of some ofthese contacts in the crystallization of lysozyme hasbeen estimated as >20 kT.24 The adsorption of asurfactant molecule within these contacts will not beenergetically favorable, as it will lead to an energy gainof only 10-12 kT as opposed to the loss of >20 kT. Thus,the long-term crystalline structure will not be influencedby the surfactant adsorption, although the discretecrystallization and recrystallization routes by which itwill be reached can be modified by the additives. Thesmall changes in the energy vs configuration patterncaused by surfactant adsorption are the possible reasonfor formation of various twinned phases observed at pH10. This, we believe, is the mechanism of the “mild”surfactant action studied in this paper. On the otherhand, “hard” surfactants, such as larger quantities ofsodium dodecyl sulfate and other strong detergents,
adsorb strongly and form precipitates with long-termkinetic stability by profoundly modifying the protein-protein interaction pattern and possibly the proteinstructure and eventually forming hemi-micelles (Figure9, bottom panel).
We cannot conclude at this stage that the surfactantsper se enhance the growth of high-quality proteincrystals. Instead, one of the most likely reasons for theformation of the large and high-resolution crystalsappears to be the ability of the surfactants to promoterecrystallization of the phases formed. The surfactantcauses the rapid appearance of precipitates and low-quality twinned crystals. This decreases the proteinconcentration in solution, so the next generation ofcrystals will subsequently grow slowly under conditionsof low and constant supersaturation, which is knownto be the best route for the formation of highly orderedcrystals.23 This hypothesis again points to the kineticnature of the surfactant effects in protein crystallization.
Understanding the mechanism of the surfactant-protein interactions on the molecular scale may help toimprove protein separation processes of practical im-portance. Characterizing the correlation between sur-factant chemistry and its relation to a specific type
Figure 9. Schematics of the concept of how surfactant binding changes the energy of protein interactions, leading to the formationof “mutated” crystals or precipitates.
of crystal twinning may assist in identifying undesirablecontaminants during protein crystallization. On theother hand, surfactants may help in processes whererapid crystallization is required, or when time permitswaiting until high quality crystals grow via recrystal-lization. Use of surfactants in protein separations mayexpand well beyond crystallization. Protein mixturescan be controllably separated by selective precipitationwith surfactants (manuscript in preparation), and thehigh symmetry and the spontaneous organization of thetwinned structures into micrometer-sized layers (Figure7) can potentially also be of use in new materialssynthesis.
Conclusions
The surfactants studied strongly affect the patternof protein interactions and crystallization. The virialcoefficient data show that they make the proteins more“sticky”, leading to additional attraction and a propen-sity to precipitate. The surfactants mediate the attrac-tive interactions between the proteins via both electro-static and hydrophobic contributions to the interactionenergy.
The results reported here may be relevant to thewidely observed and generally undesirable crystal twin-ning. Twinning may originate from adsorption on theprotein of surfactant molecules that disrupt growth ofthe lattice. An important conclusion of this study is thatthe effect of the surfactant on the protein crystallizationprocess is kinetic and not thermodynamic. The initialprecipitation and twinning are highly disruptive andundesirable effects of the surfactant, but the crystalsultimately grown are essentially indistinguishable intheir lattice parameters and molecular conformationfrom ones obtained without additives. Moreover, crystalsgrown in the presence of surfactant were large anddiffracted to high resolution. Thus, while the uncon-trolled presence of detergents and amphiphiles inprotein crystallization solutions should be avoided, theaddition of some “mild” anionic surfactants may aidcrystallization without impairing, and possibly evenimproving, crystal quality. Not surprisingly, some sur-factants have already been successfully used as addi-tives under empirically determined experimental con-ditions.
Acknowledgment. This study was supported bygrants from NASA (NAG8-1346) and NSF (BES-9510420 and BES-0078844).
Hosseini, M. W.; Lehn, J. M.; Giege, R. J. Cryst. Growth1999, 196, 365-376.
(2) McPherson, A.; Koszelak, S.; Axelrod, H.; Day, J.; Robinson,L.; McGrath, M.; Williams, R.; Cascio, D. J. Biol. Chem.1986, 261, 1969-1975.
(3) McPherson, A.; Koszelak, S.; Axelrod, H.; Day, J.; Robinson,L.; McGrath, M.; Williams, R.; Cascio, D. J. Cryst. Growth1986, 76, 547-553.
(4) Mustafa, A. O.; Derrick, J. P.; Tiddy, G. J. T.; Ford, R. C.Acta Crystallogr. 1998, D54, 154-158.
(5) Guan, R. J.; Wang, M.; Liu, X. Q.; Wang, D. C. J. Cryst.Growth 2001, 231, 273-279.
(6) Velev, O. D.; Kaler, E. W.; Lenhoff, A. M.; J. Phys. Chem.B. 2000, 104, 9267-9275.
(7) Jones, M. N.; Manley, P., J. Chem. Soc., Faraday Trans., I1979, 75, 1736-1744.
(8) Jones, M. N.; Manley, P., J. Chem. Soc., Faraday Trans., I1980, 76, 654-664.
(9) Fukushima, K., Murata, Y., Nishikido, N., Sugihara, G.;Tanaka M. Bull. Chem. Soc. Jpn. 1981, 54, 3122-3127.
(10) Moren A. K.; Khan A. Langmuir 1995, 11, 3636-3643.(11) Moren A. K.; Khan A. Langmuir 1998, 14, 6818-6826.(12) Turro, N. J.; Lei, X.-G.; Ananthapadmanabhan, K. P.;
Aronson, M. Langmuir 1995, 11, 2525-2533.(13) Valstar, A.; Almgren, M.; Brown, W.; Vasilescu, M. Lang-
muir 2000, 16, 922-927.(14) George, A.; Wilson, W. W. Acta Crystallogr. 1994, D50, 361-
Wilson, W. W. Methods Enzymol. 1997, 276, 100-110.(16) Muschol, M.; Rosenberger. F. J. Chem. Phys. 1995, 107,
1953-1962.(17) Velev, O. D.; Kaler, E. W.; Lenhoff, A. M. Biophys. J. 1998,
75, 2682-2697.(18) Neal, B. L.; Asthagiri, D.; Velev, O. D.; Lenhoff, A. M.; Kaler,
E. W. J. Cryst. Growth 1998, 196, 377-387.(19) Sack, J. S. J. Mol. Graphics 1988 6, 224-225.(20) Jones, T. A.; Zou, J. Y.; Cowan, S. W.; Kjeldgaard, M. Acta
Crystallogr. 1991, A47, 110-119.(21) DePhillips, P.; Lenhoff, A. M. J. Chromatogr. A 2001, 933,
57-72.(22) Asthagiri, D.; Schure, M. R.; Lenhoff, A. M. J. Phys. Chem.
B 2000, 104, 8753-8761.(23) Durbin, S. D.; Feher. G. Annu. Rev. Phys. Chem. 1996, 47,
171-204.(24) Asthagiri, D.; Neal, B. L.; Lenhoff, A. M. Biophys. Chem.
